diff options
| author | Samuel Thibault <samuel.thibault@ens-lyon.org> | 2013-02-04 10:27:44 +0100 |
|---|---|---|
| committer | Samuel Thibault <samuel.thibault@ens-lyon.org> | 2013-02-04 10:27:44 +0100 |
| commit | ba1b3afd50913473f3036a63b4a82d7ba5c42009 (patch) | |
| tree | 9dff0ddec4bf8b927a025b4bf9882cb1731170f3 | |
| parent | bfdb3be16e5a20eebc97b3ca613d9a4da4465533 (diff) | |
| parent | 51e87d005139a435cd846ac5c224eed5042c4fa0 (diff) | |
Merge branch 'master' into master-gdb_stubs
285 files changed, 26324 insertions, 5481 deletions
diff --git a/.gitignore b/.gitignore new file mode 100644 index 00000000..4488bc2d --- /dev/null +++ b/.gitignore @@ -0,0 +1,39 @@ +*~ + +.deps/ +.dirstamp +*.o +*.a +/gnumach +/gnumach-undef +/gnumach-undef-bad +/gnumach.msgids + +# Makerules.am +*.gz +*.stripped + +# Makerules.mig.am +*.user.* +*.server.* + +autom4te.cache/ +build-aux/ +/INSTALL +/Makefile +/Makefile.in +/aclocal.m4 +/config.h +/config.h.in +/config.log +/config.status +/config.status.orig +/configure +/stamp-h1 +/version.c + +# Ignore arch symlinks +/machine +/mach/machine +/linux/src/include/asm +/linux/dev/include/asm diff --git a/Makefile.am b/Makefile.am index 04885ddb..319b7e80 100644 --- a/Makefile.am +++ b/Makefile.am @@ -1,6 +1,6 @@ # Makefile for GNU Mach. -# Copyright (C) 2006, 2007, 2008 Free Software Foundation, Inc. +# Copyright (C) 2006, 2007, 2008, 2009 Free Software Foundation, Inc. # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by the @@ -51,6 +51,9 @@ AM_CPPFLAGS += \ -I$(top_srcdir)/$(systype)/include/mach/sa \ -I$(top_srcdir)/include +AM_CFLAGS += \ + -fno-builtin-log + # Yes, this makes the eyes hurt. But perhaps someone will finally take care of # all that scruffy Mach code... Also see <http://savannah.gnu.org/task/?5726>. AM_CFLAGS += \ @@ -68,6 +71,26 @@ AM_CFLAGS += \ endif # +# Silent build support. +# + +AWK_V = $(AWK_V_$(V)) +AWK_V_ = $(AWK_V_$(AM_DEFAULT_VERBOSITY)) +AWK_V_0 = @echo " AWK $@"; + +GZIP_V = $(GZIP_V_$(V)) +GZIP_V_ = $(GZIP_V_$(AM_DEFAULT_VERBOSITY)) +GZIP_V_0 = @echo " GZIP $@"; + +NM_V = $(NM_V_$(V)) +NM_V_ = $(NM_V_$(AM_DEFAULT_VERBOSITY)) +NM_V_0 = @echo " NM $@"; + +MIGCOM_V = $(MIGCOM_V_$(V)) +MIGCOM_V_ = $(MIGCOM_V_$(AM_DEFAULT_VERBOSITY)) +MIGCOM_V_0 = @echo " MIG $@"; + +# # MIG Setup. # @@ -133,22 +156,22 @@ noinst_PROGRAMS += \ gnumach.o # This is the list of routines we decide is OK to steal from the C library. -clib_routines := memcmp memcpy memmove memset bcopy bzero \ - strchr strstr strsep strpbrk strtok \ +clib_routines := memcmp memcpy memmove \ + strchr strstr strsep strtok \ htonl htons ntohl ntohs \ udivdi3 __udivdi3 \ __rel_iplt_start __rel_iplt_end \ _START _start etext _edata end _end # actually ld magic, not libc. gnumach-undef: gnumach.$(OBJEXT) - $(NM) -u $< | sed 's/ *U *//' | sort -u > $@ + $(NM_V) $(NM) -u $< | sed 's/ *U *//' | sort -u > $@ MOSTLYCLEANFILES += gnumach-undef gnumach-undef-bad: gnumach-undef Makefile - sed '$(foreach r,$(clib_routines),/^$r$$/d;)' $< > $@ + $(AM_V_GEN) sed '$(foreach r,$(clib_routines),/^$r$$/d;)' $< > $@ MOSTLYCLEANFILES += gnumach-undef-bad clib-routines.o: gnumach-undef gnumach-undef-bad - if test -s gnumach-undef-bad; \ + $(AM_V_at) if test -s gnumach-undef-bad; \ then cat gnumach-undef-bad; exit 2; else true; fi - $(CCLD) -nostdlib -nostartfiles -r -static \ + $(AM_V_CCLD) $(CCLD) -nostdlib -nostartfiles -r -static \ -o $@ `sed 's/^/-Wl,-u,/' < $<` -x c /dev/null -lc -lgcc gnumach_LINK = $(LD) $(LINKFLAGS) $(gnumach_LINKFLAGS) -o $@ diff --git a/Makefrag.am b/Makefrag.am index cc5a01ea..42026f33 100644 --- a/Makefrag.am +++ b/Makefrag.am @@ -1,6 +1,7 @@ # Main Makefile fragment for GNU Mach. -# Copyright (C) 1997, 1999, 2004, 2006, 2007 Free Software Foundation, Inc. +# Copyright (C) 1997, 1999, 2004, 2006, 2007, 2009 Free Software +# Foundation, Inc. # Permission to use, copy, modify and distribute this software and its # documentation is hereby granted, provided that both the copyright @@ -61,6 +62,9 @@ libkernel_a_SOURCES += \ ddb/nlist.h \ ddb/stab.h \ ddb/tr.h + +# We need frame pointers for trace to work properly. +AM_CFLAGS += -fno-omit-frame-pointer endif # @@ -135,7 +139,6 @@ libkernel_a_SOURCES += \ kern/ast.h \ kern/boot_script.h \ kern/bootstrap.c \ - kern/compat_xxx_defs.h \ kern/counters.c \ kern/counters.h \ kern/cpu_number.h \ @@ -156,9 +159,9 @@ libkernel_a_SOURCES += \ kern/ipc_sched.h \ kern/ipc_tt.c \ kern/ipc_tt.h \ - kern/kalloc.c \ kern/kalloc.h \ kern/kern_types.h \ + kern/list.h \ kern/lock.c \ kern/lock.h \ kern/lock_mon.c \ @@ -166,7 +169,6 @@ libkernel_a_SOURCES += \ kern/mach_clock.h \ kern/mach_factor.c \ kern/mach_factor.h \ - kern/mach_param.h \ kern/machine.c \ kern/machine.h \ kern/macro_help.h \ @@ -180,7 +182,12 @@ libkernel_a_SOURCES += \ kern/profile.c \ kern/queue.c \ kern/queue.h \ + kern/rbtree.c \ + kern/rbtree.h \ + kern/rbtree_i.h \ kern/refcount.h \ + kern/slab.c \ + kern/slab.h \ kern/sched.h \ kern/sched_prim.c \ kern/sched_prim.h \ @@ -205,13 +212,12 @@ libkernel_a_SOURCES += \ kern/timer.h \ kern/xpr.c \ kern/xpr.h \ - kern/zalloc.c \ - kern/zalloc.h \ kern/elf-load.c \ kern/boot_script.c EXTRA_DIST += \ kern/mach.srv \ kern/mach4.srv \ + kern/gnumach.srv \ kern/mach_debug.srv \ kern/mach_host.srv @@ -233,6 +239,7 @@ libkernel_a_SOURCES += \ libkernel_a_SOURCES += \ vm/memory_object_proxy.c \ + vm/memory_object_proxy.h \ vm/memory_object.c \ vm/memory_object.h \ vm/pmap.h \ @@ -287,7 +294,6 @@ libkernel_a_SOURCES += \ device/device_types_kernel.h \ device/ds_routines.c \ device/ds_routines.h \ - device/errno.h \ device/if_ether.h \ device/if_hdr.h \ device/io_req.h \ @@ -348,6 +354,7 @@ include_mach_HEADERS = \ include/mach/exc.defs \ include/mach/mach.defs \ include/mach/mach4.defs \ + include/mach/gnumach.defs \ include/mach/mach_host.defs \ include/mach/mach_port.defs \ include/mach/mach_types.defs \ @@ -390,6 +397,7 @@ include_mach_HEADERS = \ include/mach/time_value.h \ include/mach/version.h \ include/mach/vm_attributes.h \ + include/mach/vm_cache_statistics.h \ include/mach/vm_inherit.h \ include/mach/vm_param.h \ include/mach/vm_prot.h \ @@ -406,7 +414,7 @@ include_mach_eXec_HEADERS = \ # mach-debug-headers:= $(addprefix mach_debug/, hash_info.h ipc_info.h \ # mach_debug.defs mach_debug_types.defs mach_debug_types.h \ -# pc_info.h vm_info.h zone_info.h) +# pc_info.h vm_info.h slab_info.h) # Other headers for the distribution. We don't install these, because the # GNU C library has correct versions for users to use. @@ -485,6 +493,7 @@ nodist_libkernel_a_SOURCES += \ nodist_lib_dep_tr_for_defs_a_SOURCES += \ kern/mach.server.defs.c \ kern/mach4.server.defs.c \ + kern/gnumach.server.defs.c \ kern/mach_debug.server.defs.c \ kern/mach_host.server.defs.c nodist_libkernel_a_SOURCES += \ @@ -494,6 +503,9 @@ nodist_libkernel_a_SOURCES += \ kern/mach4.server.h \ kern/mach4.server.c \ kern/mach4.server.msgids \ + kern/gnumach.server.h \ + kern/gnumach.server.c \ + kern/gnumach.server.msgids \ kern/mach_debug.server.h \ kern/mach_debug.server.c \ kern/mach_debug.server.msgids \ @@ -502,6 +514,7 @@ nodist_libkernel_a_SOURCES += \ kern/mach_host.server.msgids # kern/mach.server.defs # kern/mach4.server.defs +# kern/gnumach.server.defs # kern/mach_debug.server.defs # kern/mach_host.server.defs @@ -509,8 +522,8 @@ nodist_libkernel_a_SOURCES += \ MOSTLYCLEANFILES += \ gnumach.msgids gnumach.msgids: $(filter %.msgids,$(nodist_libkernel_a_SOURCES)) - cat $^ > $@.new - mv $@.new $@ + $(AM_V_at) cat $^ > $@.new + $(AM_V_GEN) mv $@.new $@ # `exec_' prefix, so that we don't try to build that file during when running # `make install-data', as it may fail there, but isn't needed there either. exec_msgidsdir = $(datadir)/msgids diff --git a/Makerules.am b/Makerules.am index 37d383ae..b1f17d12 100644 --- a/Makerules.am +++ b/Makerules.am @@ -1,6 +1,6 @@ # Makerules: how to do some things. -# Copyright (C) 2006, 2007 Free Software Foundation, Inc. +# Copyright (C) 2006, 2007, 2009 Free Software Foundation, Inc. # Permission to use, copy, modify and distribute this software and its # documentation is hereby granted, provided that both the copyright @@ -20,11 +20,11 @@ EXTRA_DIST += \ gensym.awk %.symc: %.sym gensym.awk - $(AWK) -f $(word 2,$^) $< > $@ + $(AWK_V) $(AWK) -f $(word 2,$^) $< > $@ %.symc.o: %.symc - $(COMPILE) -S -x c -o $@ $< + $(AM_V_CC) $(COMPILE) -S -x c -o $@ $< %.h: %.symc.o - sed < $< > $@ \ + $(AM_V_GEN) sed < $< > $@ \ -e 's/^[^*].*$$//' \ -e 's/^[*]/#define/' \ -e 's/mAgIc[^-0-9]*//' @@ -37,7 +37,7 @@ include Makerules.mig.am # %.gz: % - $(GZIP) -9 < $< > $@ + $(GZIP_V) $(GZIP) -9 < $< > $@ # # strip files. diff --git a/Makerules.mig.am b/Makerules.mig.am index b3f76da2..30609846 100644 --- a/Makerules.mig.am +++ b/Makerules.mig.am @@ -1,6 +1,6 @@ # Makerules.mig: how to do some MIG-related things. -# Copyright (C) 2006, 2007 Free Software Foundation, Inc. +# Copyright (C) 2006, 2007, 2009 Free Software Foundation, Inc. # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by the @@ -72,20 +72,20 @@ lib_dep_tr_for_defs_a_CPPFLAGS = $(AM_CPPFLAGS) \ -E %.server.defs.c: %.srv - rm -f $@ - cp -p $< $@ + $(AM_V_at) rm -f $@ + $(AM_V_GEN) cp -p $< $@ %.user.defs.c: %.cli - rm -f $@ - cp -p $< $@ + $(AM_V_at) rm -f $@ + $(AM_V_GEN) cp -p $< $@ %.server.h %.server.c %.server.msgids: lib_dep_tr_for_defs_a-%.server.defs.$(OBJEXT) - $(MIGCOM) $(MIGCOMFLAGS) $(MIGCOMSFLAGS) \ - -sheader $*.server.h -server $*.server.c \ - -list $*.server.msgids \ + $(MIGCOM_V) $(MIGCOM) $(MIGCOMFLAGS) $(MIGCOMSFLAGS) \ + -sheader $*.server.h -server $*.server.c \ + -list $*.server.msgids \ < $< %.user.h %.user.c %.user.msgids: lib_dep_tr_for_defs_a-%.user.defs.$(OBJEXT) - $(MIGCOM) $(MIGCOMFLAGS) $(MIGCOMUFLAGS) \ - -user $*.user.c -header $*.user.h \ - -list $*.user.msgids \ + $(MIGCOM_V) $(MIGCOM) $(MIGCOMFLAGS) $(MIGCOMUFLAGS) \ + -user $*.user.c -header $*.user.h \ + -list $*.user.msgids \ < $< # This is how it should be done, but this is not integrated into GNU Automake diff --git a/configfrag.ac b/configfrag.ac index 1be36cb8..57146c93 100644 --- a/configfrag.ac +++ b/configfrag.ac @@ -102,6 +102,12 @@ AC_DEFINE([STAT_TIME], [1], [STAT_TIME]) # Kernel tracing. AC_DEFINE([XPR_DEBUG], [1], [XPR_DEBUG]) + +# Slab allocator debugging facilities. +AC_DEFINE([SLAB_VERIFY], [0], [SLAB_VERIFY]) + +# Enable the CPU pool layer in the slab allocator. +AC_DEFINE([SLAB_USE_CPU_POOLS], [0], [SLAB_USE_CPU_POOLS]) # # Options. diff --git a/configure.ac b/configure.ac index 50ec6b40..3a7d3be6 100644 --- a/configure.ac +++ b/configure.ac @@ -32,6 +32,9 @@ dnl Do not clutter the main build directory. dnl We require GNU make. [-Wall -Wno-portability] ) + +m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([no])], + [AC_SUBST([AM_DEFAULT_VERBOSITY], [1])]) # # Deduce the architecture we're building for. @@ -149,7 +152,8 @@ m4_include([linux/configfrag.ac]) [ssp_possible=yes] AC_MSG_CHECKING([whether the compiler accepts `-fstack-protector']) # Is this a reliable test case? -AC_LANG_CONFTEST([[void foo (void) { volatile char a[8]; a[3]; }]]) +AC_LANG_CONFTEST( + [AC_LANG_SOURCE([[void foo (void) { volatile char a[8]; a[3]; }]])]) [# `$CC -c -o ...' might not be portable. But, oh, well... Is calling # `ac_compile' like this correct, after all? if eval "$ac_compile -S -fstack-protector -o conftest.s" 2> /dev/null; then] diff --git a/ddb/db_examine.c b/ddb/db_examine.c index 2df261f5..96c5eee1 100644 --- a/ddb/db_examine.c +++ b/ddb/db_examine.c @@ -480,7 +480,7 @@ db_xcdump(addr, size, count, task) if (!DB_CHECK_ACCESS(next_page_addr, sizeof(int), task)) bcount = next_page_addr - addr; } - db_read_bytes((char *)addr, bcount, data, task); + db_read_bytes(addr, bcount, data, task); for (i = 0; i < bcount && off != 0; i += size) { if (i % 4 == 0) db_printf(" "); diff --git a/ddb/db_macro.c b/ddb/db_macro.c index e372c236..43bb5837 100644 --- a/ddb/db_macro.c +++ b/ddb/db_macro.c @@ -165,7 +165,7 @@ db_exec_macro(name) return(0); } -int +long /* ARGSUSED */ db_arg_variable(vp, valuep, flag, ap) struct db_variable *vp; diff --git a/ddb/db_macro.h b/ddb/db_macro.h index 12ed16e4..da5626f9 100644 --- a/ddb/db_macro.h +++ b/ddb/db_macro.h @@ -32,7 +32,7 @@ extern void db_show_macro (void); extern int db_exec_macro (char *name); -extern int db_arg_variable ( +extern long db_arg_variable ( struct db_variable *vp, db_expr_t *valuep, int flag, diff --git a/ddb/db_output.c b/ddb/db_output.c index 57d6856d..3ea2caac 100644 --- a/ddb/db_output.c +++ b/ddb/db_output.c @@ -213,7 +213,7 @@ db_printf(const char *fmt, ...) db_printf_enter(); /* optional multiP serialization */ #endif va_start(listp, fmt); - _doprnt(fmt, &listp, db_id_putc, db_radix, 0); + _doprnt(fmt, listp, db_id_putc, db_radix, 0); va_end(listp); } @@ -225,7 +225,7 @@ kdbprintf(const char *fmt, ...) { va_list listp; va_start(listp, fmt); - _doprnt(fmt, &listp, db_id_putc, db_radix, 0); + _doprnt(fmt, listp, db_id_putc, db_radix, 0); va_end(listp); } diff --git a/ddb/db_output.h b/ddb/db_output.h index 3203e307..1159c6ba 100644 --- a/ddb/db_output.h +++ b/ddb/db_output.h @@ -35,6 +35,6 @@ extern void db_force_whitespace(void); extern int db_print_position(void); extern void db_end_line(void); -extern void db_printf( const char *fmt, ...); -extern void db_putchar(int c); - +extern void db_printf(const char *fmt, ...); +extern void db_putchar(int c); +extern void kdbprintf(const char *fmt, ...); diff --git a/ddb/db_sym.c b/ddb/db_sym.c index dd721384..5c5f7006 100644 --- a/ddb/db_sym.c +++ b/ddb/db_sym.c @@ -445,7 +445,7 @@ db_symbol_values(stab, sym, namep, valuep) * not accept symbols whose value is zero (and use plain hex). */ -unsigned int db_maxoff = 0x4000; +unsigned long db_maxoff = 0x4000; void db_task_printsym(off, strategy, task) diff --git a/ddb/db_task_thread.c b/ddb/db_task_thread.c index 37629425..1146223b 100644 --- a/ddb/db_task_thread.c +++ b/ddb/db_task_thread.c @@ -245,7 +245,7 @@ db_init_default_thread(void) * in the command line */ /* ARGSUSED */ -int +long db_set_default_thread(vp, valuep, flag) struct db_variable *vp; db_expr_t *valuep; @@ -270,7 +270,7 @@ db_set_default_thread(vp, valuep, flag) /* * convert $taskXXX[.YYY] type DDB variable to task or thread address */ -int +long db_get_task_thread(vp, valuep, flag, ap) struct db_variable *vp; db_expr_t *valuep; diff --git a/ddb/db_variables.h b/ddb/db_variables.h index c01a5e25..5249d18c 100644 --- a/ddb/db_variables.h +++ b/ddb/db_variables.h @@ -42,7 +42,7 @@ struct db_variable { char *name; /* Name of variable */ db_expr_t *valuep; /* pointer to value of variable */ /* function to call when reading/writing */ - int (*fcn)(struct db_variable *, db_expr_t *, int, db_var_aux_param_t); + long (*fcn)(struct db_variable *, db_expr_t *, int, db_var_aux_param_t); short min_level; /* number of minimum suffix levels */ short max_level; /* number of maximum suffix levels */ short low; /* low value of level 1 suffix */ @@ -50,7 +50,7 @@ struct db_variable { #define DB_VAR_GET 0 #define DB_VAR_SET 1 }; -#define FCN_NULL ((int (*)())0) +#define FCN_NULL ((long (*)())0) #define DB_VAR_LEVEL 3 /* maximum number of suffix level */ diff --git a/device/cons.c b/device/cons.c index e3e95ffb..f26f22c5 100644 --- a/device/cons.c +++ b/device/cons.c @@ -22,22 +22,10 @@ #include <string.h> #include <kern/debug.h> -#ifdef MACH_KERNEL #include <sys/types.h> #include <device/conf.h> #include <mach/boolean.h> #include <device/cons.h> -#else -#include <sys/param.h> -#include <sys/user.h> -#include <sys/systm.h> -#include <sys/buf.h> -#include <sys/ioctl.h> -#include <sys/tty.h> -#include <sys/file.h> -#include <sys/conf.h> -#include <hpdev/cons.h> -#endif #ifdef MACH_KMSG #include <device/io_req.h> @@ -46,9 +34,6 @@ static int cn_inited = 0; static struct consdev *cn_tab = 0; /* physical console device info */ -#ifndef MACH_KERNEL -static struct tty *constty = 0; /* virtual console output device */ -#endif /* * ROM getc/putc primitives. @@ -76,10 +61,8 @@ void cninit() { struct consdev *cp; -#ifdef MACH_KERNEL dev_ops_t cn_ops; int x; -#endif if (cn_inited) return; @@ -103,7 +86,6 @@ cninit() * Initialize as console */ (*cp->cn_init)(cp); -#ifdef MACH_KERNEL /* * Look up its dev_ops pointer in the device table and * place it in the device indirection table. @@ -111,7 +93,6 @@ cninit() if (dev_name_lookup(cp->cn_name, &cn_ops, &x) == FALSE) panic("cninit: dev_name_lookup failed"); dev_set_indirection("console", cn_ops, minor(cp->cn_dev)); -#endif #if CONSBUFSIZE > 0 /* * Now that the console is initialized, dump any chars in @@ -134,97 +115,9 @@ cninit() /* * No console device found, not a problem for BSD, fatal for Mach */ -#ifdef MACH_KERNEL panic("can't find a console device"); -#endif } -#ifndef MACH_KERNEL -cnopen(dev, flag) - dev_t dev; -{ - if (cn_tab == NULL) - return(0); - dev = cn_tab->cn_dev; - return ((*cdevsw[major(dev)].d_open)(dev, flag)); -} - -cnclose(dev, flag) - dev_t dev; -{ - if (cn_tab == NULL) - return(0); - dev = cn_tab->cn_dev; - return ((*cdevsw[major(dev)].d_close)(dev, flag)); -} - -cnread(dev, uio) - dev_t dev; - struct uio *uio; -{ - if (cn_tab == NULL) - return(0); - dev = cn_tab->cn_dev; - return ((*cdevsw[major(dev)].d_read)(dev, uio)); -} - -cnwrite(dev, uio) - dev_t dev; - struct uio *uio; -{ - if (cn_tab == NULL) - return(0); - dev = cn_tab->cn_dev; - return ((*cdevsw[major(dev)].d_write)(dev, uio)); -} - -cnioctl(dev, cmd, data, flag) - dev_t dev; - caddr_t data; -{ - if (cn_tab == NULL) - return(0); - /* - * Superuser can always use this to wrest control of console - * output from the "virtual" console. - */ - if (cmd == TIOCCONS && constty) { - if (!suser()) - return(EPERM); - constty = NULL; - return(0); - } - dev = cn_tab->cn_dev; - return ((*cdevsw[major(dev)].d_ioctl)(dev, cmd, data, flag)); -} - -cnselect(dev, rw) - dev_t dev; - int rw; -{ - if (cn_tab == NULL) - return(1); - return(ttselect(cn_tab->cn_dev, rw)); -} - -#ifndef hp300 -/* - * XXX Should go away when the new CIO MUX driver is in place - */ -#define d_control d_mmap -cncontrol(dev, cmd, data) - dev_t dev; - int cmd; - int data; -{ - if (cn_tab == NULL) - return(0); - dev = cn_tab->cn_dev; - return((*cdevsw[major(dev)].d_control)(dev, cmd, data)); -} -#undef d_control -#endif -#endif int cngetc() @@ -236,7 +129,6 @@ cngetc() return (0); } -#ifdef MACH_KERNEL int cnmaygetc() { @@ -246,7 +138,6 @@ cnmaygetc() return ((*romgetc)(0)); return (0); } -#endif void cnputc(c) diff --git a/device/cons.h b/device/cons.h index a6b04fff..6a0ae850 100644 --- a/device/cons.h +++ b/device/cons.h @@ -21,16 +21,16 @@ * Utah $Hdr: cons.h 1.10 94/12/14$ */ +#ifndef _DEVICE_CONS_H +#define _DEVICE_CONS_H #include <sys/types.h> struct consdev { -#ifdef MACH_KERNEL char *cn_name; /* name of device in dev_name_list */ -#endif - int (*cn_probe)(); /* probe hardware and fill in consdev info */ - int (*cn_init)(); /* turn on as console */ - int (*cn_getc)(); /* kernel getchar interface */ - int (*cn_putc)(); /* kernel putchar interface */ + int (*cn_probe)(struct consdev *cp); /* probe hardware and fill in consdev info */ + int (*cn_init)(struct consdev *cp); /* turn on as console */ + int (*cn_getc)(dev_t dev, int wait); /* kernel getchar interface */ + int (*cn_putc)(dev_t dev, int c); /* kernel putchar interface */ dev_t cn_dev; /* major/minor of device */ short cn_pri; /* pecking order; the higher the better */ }; @@ -57,3 +57,4 @@ extern int cngetc(void); extern int cnmaygetc(void); extern void cnputc(char); +#endif /* _DEVICE_CONS_H */ diff --git a/device/dev_lookup.c b/device/dev_lookup.c index 2391e8d7..98a2d02c 100644 --- a/device/dev_lookup.c +++ b/device/dev_lookup.c @@ -32,7 +32,7 @@ #include <mach/vm_param.h> #include <kern/queue.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <device/device_types.h> #include <device/dev_hdr.h> @@ -62,7 +62,7 @@ queue_head_t dev_number_hash_table[NDEVHASH]; decl_simple_lock_data(, dev_number_lock) -zone_t dev_hdr_zone; +struct kmem_cache dev_hdr_cache; /* * Enter device in the number lookup table. @@ -151,7 +151,7 @@ device_lookup(name) simple_unlock(&dev_number_lock); - new_device = (mach_device_t) zalloc(dev_hdr_zone); + new_device = (mach_device_t) kmem_cache_alloc(&dev_hdr_cache); simple_lock_init(&new_device->ref_lock); new_device->ref_count = 1; simple_lock_init(&new_device->lock); @@ -187,7 +187,7 @@ device_lookup(name) simple_unlock(&dev_number_lock); if (new_device != MACH_DEVICE_NULL) - zfree(dev_hdr_zone, (vm_offset_t)new_device); + kmem_cache_free(&dev_hdr_cache, (vm_offset_t)new_device); } return (device); @@ -233,7 +233,7 @@ mach_device_deallocate(device) simple_unlock(&device->ref_lock); simple_unlock(&dev_number_lock); - zfree(dev_hdr_zone, (vm_offset_t)device); + kmem_cache_free(&dev_hdr_cache, (vm_offset_t)device); } /* @@ -376,9 +376,6 @@ dev_lookup_init() simple_lock_init(&dev_port_lock); - dev_hdr_zone = zinit(sizeof(struct mach_device), 0, - sizeof(struct mach_device) * NDEVICES, - PAGE_SIZE, - FALSE, - "open device entry"); + kmem_cache_init(&dev_hdr_cache, "mach_device", + sizeof(struct mach_device), 0, NULL, NULL, NULL, 0); } diff --git a/device/dev_pager.c b/device/dev_pager.c index 224be850..e0ca2c76 100644 --- a/device/dev_pager.c +++ b/device/dev_pager.c @@ -44,8 +44,7 @@ #include <kern/debug.h> #include <kern/printf.h> #include <kern/queue.h> -#include <kern/zalloc.h> -#include <kern/kalloc.h> +#include <kern/slab.h> #include <vm/vm_page.h> #include <vm/vm_kern.h> @@ -55,6 +54,7 @@ #include <device/ds_routines.h> #include <device/dev_hdr.h> #include <device/io_req.h> +#include <device/memory_object_reply.user.h> extern vm_offset_t block_io_mmap(); /* dummy routine to allow mmap for block devices */ @@ -125,7 +125,7 @@ typedef struct dev_pager *dev_pager_t; #define DEV_PAGER_NULL ((dev_pager_t)0) -zone_t dev_pager_zone; +struct kmem_cache dev_pager_cache; void dev_pager_reference(register dev_pager_t ds) { @@ -143,7 +143,7 @@ void dev_pager_deallocate(register dev_pager_t ds) } simple_unlock(&ds->lock); - zfree(dev_pager_zone, (vm_offset_t)ds); + kmem_cache_free(&dev_pager_cache, (vm_offset_t)ds); } /* @@ -160,12 +160,12 @@ struct dev_pager_entry { typedef struct dev_pager_entry *dev_pager_entry_t; queue_head_t dev_pager_hashtable[DEV_PAGER_HASH_COUNT]; -zone_t dev_pager_hash_zone; +struct kmem_cache dev_pager_hash_cache; decl_simple_lock_data(, dev_pager_hash_lock) #define dev_pager_hash(name_port) \ - (((natural_t)(name_port) & 0xffffff) % DEV_PAGER_HASH_COUNT) + (((vm_offset_t)(name_port) & 0xffffff) % DEV_PAGER_HASH_COUNT) void dev_pager_hash_init(void) { @@ -173,13 +173,8 @@ void dev_pager_hash_init(void) register vm_size_t size; size = sizeof(struct dev_pager_entry); - dev_pager_hash_zone = zinit( - size, - 0, - size * 1000, - PAGE_SIZE, - FALSE, - "dev_pager port hash"); + kmem_cache_init(&dev_pager_hash_cache, "dev_pager_entry", size, 0, + NULL, NULL, NULL, 0); for (i = 0; i < DEV_PAGER_HASH_COUNT; i++) queue_init(&dev_pager_hashtable[i]); simple_lock_init(&dev_pager_hash_lock); @@ -191,7 +186,7 @@ void dev_pager_hash_insert( { register dev_pager_entry_t new_entry; - new_entry = (dev_pager_entry_t) zalloc(dev_pager_hash_zone); + new_entry = (dev_pager_entry_t) kmem_cache_alloc(&dev_pager_hash_cache); new_entry->name = name_port; new_entry->pager_rec = rec; @@ -219,7 +214,7 @@ void dev_pager_hash_delete(ipc_port_t name_port) } simple_unlock(&dev_pager_hash_lock); if (entry) - zfree(dev_pager_hash_zone, (vm_offset_t)entry); + kmem_cache_free(&dev_pager_hash_cache, (vm_offset_t)entry); } dev_pager_t dev_pager_hash_lookup(ipc_port_t name_port) @@ -272,7 +267,7 @@ kern_return_t device_pager_setup( return (D_SUCCESS); } - d = (dev_pager_t) zalloc(dev_pager_zone); + d = (dev_pager_t) kmem_cache_alloc(&dev_pager_cache); if (d == DEV_PAGER_NULL) return (KERN_RESOURCE_SHORTAGE); @@ -341,7 +336,7 @@ kern_return_t device_pager_data_request( #endif /* lint */ if (device_pager_debug) - printf("(device_pager)data_request: pager=%p, offset=0x%x, length=0x%x\n", + printf("(device_pager)data_request: pager=%p, offset=0x%lx, length=0x%x\n", pager, offset, length); ds = dev_pager_hash_lookup((ipc_port_t)pager); @@ -725,15 +720,11 @@ void device_pager_init(void) register vm_size_t size; /* - * Initialize zone of paging structures. + * Initialize cache of paging structures. */ size = sizeof(struct dev_pager); - dev_pager_zone = zinit(size, - 0, - (vm_size_t) size * 1000, - PAGE_SIZE, - FALSE, - "device pager structures"); + kmem_cache_init(&dev_pager_cache, "dev_pager", size, 0, + NULL, NULL, NULL, 0); /* * Initialize the name port hashing stuff. diff --git a/device/device.srv b/device/device.srv index 06aa0be3..f63813f9 100644 --- a/device/device.srv +++ b/device/device.srv @@ -24,6 +24,4 @@ #define KERNEL_SERVER 1 -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ - #include <device/device.defs> diff --git a/device/ds_routines.c b/device/ds_routines.c index 93569f28..68589dee 100644 --- a/device/ds_routines.c +++ b/device/ds_routines.c @@ -73,7 +73,7 @@ #include <kern/debug.h> #include <kern/printf.h> #include <kern/queue.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <kern/thread.h> #include <kern/task.h> #include <kern/sched_prim.h> @@ -99,6 +99,7 @@ extern struct device_emulation_ops linux_block_emulation_ops; #ifdef CONFIG_INET extern struct device_emulation_ops linux_net_emulation_ops; +extern void free_skbuffs (); #ifdef CONFIG_PCMCIA extern struct device_emulation_ops linux_pcmcia_emulation_ops; #endif @@ -129,6 +130,9 @@ static struct device_emulation_ops *emulation_list[] = &mach_device_emulation_ops, }; +static struct vm_map device_io_map_store; +vm_map_t device_io_map = &device_io_map_store; + #define NUM_EMULATION (sizeof (emulation_list) / sizeof (emulation_list[0])) io_return_t @@ -852,7 +856,7 @@ device_write_get(ior, wait) */ if (ior->io_op & IO_INBAND) { assert(ior->io_count <= sizeof (io_buf_ptr_inband_t)); - new_addr = zalloc(io_inband_zone); + new_addr = kmem_cache_alloc(&io_inband_cache); memcpy((void*)new_addr, ior->io_data, ior->io_count); ior->io_data = (io_buf_ptr_t)new_addr; ior->io_alloc_size = sizeof (io_buf_ptr_inband_t); @@ -932,7 +936,7 @@ device_write_dealloc(ior) * Inband case. */ if (ior->io_op & IO_INBAND) { - zfree(io_inband_zone, (vm_offset_t)ior->io_data); + kmem_cache_free(&io_inband_cache, (vm_offset_t)ior->io_data); return (TRUE); } @@ -1242,7 +1246,7 @@ kern_return_t device_read_alloc(ior, size) return (KERN_SUCCESS); if (ior->io_op & IO_INBAND) { - ior->io_data = (io_buf_ptr_t) zalloc(io_inband_zone); + ior->io_data = (io_buf_ptr_t) kmem_cache_alloc(&io_inband_cache); ior->io_alloc_size = sizeof(io_buf_ptr_inband_t); } else { size = round_page(size); @@ -1335,7 +1339,7 @@ boolean_t ds_read_done(ior) if (ior->io_count != 0) { if (ior->io_op & IO_INBAND) { if (ior->io_alloc_size > 0) - zfree(io_inband_zone, (vm_offset_t)ior->io_data); + kmem_cache_free(&io_inband_cache, (vm_offset_t)ior->io_data); } else { register vm_offset_t end_alloc; @@ -1445,7 +1449,7 @@ static void ds_no_senders(notification) mach_no_senders_notification_t *notification; { - printf("ds_no_senders called! device_port=0x%x count=%d\n", + printf("ds_no_senders called! device_port=0x%lx count=%d\n", notification->not_header.msgh_remote_port, notification->not_count); } @@ -1548,11 +1552,9 @@ void mach_device_init() queue_init(&io_done_list); simple_lock_init(&io_done_list_lock); - device_io_map = kmem_suballoc(kernel_map, - &device_io_min, - &device_io_max, - DEVICE_IO_MAP_SIZE, - FALSE); + kmem_submap(device_io_map, kernel_map, &device_io_min, &device_io_max, + DEVICE_IO_MAP_SIZE, FALSE); + /* * If the kernel receives many device_write requests, the * device_io_map might run out of space. To prevent @@ -1572,11 +1574,8 @@ void mach_device_init() */ device_io_map->wait_for_space = TRUE; - io_inband_zone = zinit(sizeof(io_buf_ptr_inband_t), 0, - 1000 * sizeof(io_buf_ptr_inband_t), - 10 * sizeof(io_buf_ptr_inband_t), - FALSE, - "io inband read buffers"); + kmem_cache_init(&io_inband_cache, "io_buf_ptr_inband", + sizeof(io_buf_ptr_inband_t), 0, NULL, NULL, NULL, 0); mach_device_trap_init(); } @@ -1612,7 +1611,7 @@ void iowait(ior) */ #define IOTRAP_REQSIZE 2048 -zone_t io_trap_zone; +struct kmem_cache io_trap_cache; /* * Initialization. Called from mach_device_init(). @@ -1620,24 +1619,21 @@ zone_t io_trap_zone; static void mach_device_trap_init(void) { - io_trap_zone = zinit(IOTRAP_REQSIZE, 0, - 256 * IOTRAP_REQSIZE, - 16 * IOTRAP_REQSIZE, - FALSE, - "wired device trap buffers"); + kmem_cache_init(&io_trap_cache, "io_req", IOTRAP_REQSIZE, 0, + NULL, NULL, NULL, 0); } /* * Allocate an io_req_t. - * Currently zalloc's from io_trap_zone. + * Currently allocates from io_trap_cache. * - * Could have lists of different size zones. + * Could have lists of different size caches. * Could call a device-specific routine. */ io_req_t ds_trap_req_alloc(mach_device_t device, vm_size_t data_size) { - return (io_req_t) zalloc(io_trap_zone); + return (io_req_t) kmem_cache_alloc(&io_trap_cache); } /* @@ -1653,7 +1649,7 @@ ds_trap_write_done(io_req_t ior) /* * Should look at reply port and maybe send a message. */ - zfree(io_trap_zone, (vm_offset_t) ior); + kmem_cache_free(&io_trap_cache, (vm_offset_t) ior); /* * Give up device reference from ds_write_trap. @@ -1729,7 +1725,7 @@ device_write_trap (mach_device_t device, dev_mode_t mode, */ mach_device_deallocate(device); - zfree(io_trap_zone, (vm_offset_t) ior); + kmem_cache_free(&io_trap_cache, (vm_offset_t) ior); return (result); } @@ -1820,7 +1816,7 @@ device_writev_trap (mach_device_t device, dev_mode_t mode, */ mach_device_deallocate(device); - zfree(io_trap_zone, (vm_offset_t) ior); + kmem_cache_free(&io_trap_cache, (vm_offset_t) ior); return (result); } diff --git a/device/ds_routines.h b/device/ds_routines.h index e1f6aadb..a00a12d5 100644 --- a/device/ds_routines.h +++ b/device/ds_routines.h @@ -41,7 +41,9 @@ /* * Map for device IO memory. */ -vm_map_t device_io_map; +extern vm_map_t device_io_map; + +extern queue_head_t io_done_list; kern_return_t device_read_alloc(io_req_t, vm_size_t); kern_return_t device_write_get(io_req_t, boolean_t *); diff --git a/device/errno.h b/device/errno.h deleted file mode 100644 index e65aa984..00000000 --- a/device/errno.h +++ /dev/null @@ -1,45 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1991 Carnegie Mellon University - * All Rights Reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" - * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR - * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * Author: David B. Golub, Carnegie Mellon University - * Date: 8/89 - * - * Old names for new error codes, for compatibility. - */ - -#ifndef _ERRNO_ -#define _ERRNO_ - -#include <device/device_types.h> /* the real error names */ - -#define EIO D_IO_ERROR -#define ENXIO D_NO_SUCH_DEVICE -#define EINVAL D_INVALID_SIZE /* XXX */ -#define EBUSY D_ALREADY_OPEN -#define ENOTTY D_INVALID_OPERATION -#define ENOMEM D_NO_MEMORY - -#endif /* _ERRNO_ */ diff --git a/device/io_req.h b/device/io_req.h index 162524d7..65e23e60 100644 --- a/device/io_req.h +++ b/device/io_req.h @@ -35,6 +35,7 @@ #include <mach/port.h> #include <mach/message.h> #include <mach/vm_param.h> +#include <kern/slab.h> #include <kern/kalloc.h> #include <kern/lock.h> #include <vm/vm_page.h> @@ -124,7 +125,7 @@ struct io_req { void iodone(io_req_t); /* - * Macros to allocate and free IORs - will convert to zones later. + * Macros to allocate and free IORs - will convert to caches later. */ #define io_req_alloc(ior,size) \ MACRO_BEGIN \ @@ -136,6 +137,6 @@ void iodone(io_req_t); (kfree((vm_offset_t)(ior), sizeof(struct io_req))) -zone_t io_inband_zone; /* for inband reads */ +struct kmem_cache io_inband_cache; /* for inband reads */ #endif /* _IO_REQ_ */ diff --git a/device/kmsg.h b/device/kmsg.h index adfc9453..b8c1f366 100644 --- a/device/kmsg.h +++ b/device/kmsg.h @@ -1,7 +1,6 @@ #ifndef _DEVICE_KMSG_H_ #define _DEVICE_KMSG_H_ 1 -#ifdef MACH_KERNEL #include <sys/types.h> @@ -15,6 +14,5 @@ io_return_t kmsggetstat (dev_t dev, int flavor, int *data, unsigned int *count); void kmsg_putchar (int c); -#endif /* MACH_KERNEL */ #endif /* !_DEVICE_KMSG_H_ */ diff --git a/device/net_io.c b/device/net_io.c index 1db9bca9..4ebf9964 100644 --- a/device/net_io.c +++ b/device/net_io.c @@ -61,6 +61,7 @@ #include <kern/printf.h> #include <kern/queue.h> #include <kern/sched_prim.h> +#include <kern/slab.h> #include <kern/thread.h> #include <machine/machspl.h> @@ -302,7 +303,7 @@ struct net_rcv_port { }; typedef struct net_rcv_port *net_rcv_port_t; -zone_t net_rcv_zone; /* zone of net_rcv_port structs */ +struct kmem_cache net_rcv_cache; /* cache of net_rcv_port structs */ #define NET_HASH_SIZE 256 @@ -324,7 +325,7 @@ struct net_hash_entry { }; typedef struct net_hash_entry *net_hash_entry_t; -zone_t net_hash_entry_zone; +struct kmem_cache net_hash_entry_cache; /* * This structure represents a packet filter with multiple sessions. @@ -393,7 +394,7 @@ int net_add_q_info (ipc_port_t rcv_port); int bpf_match ( net_hash_header_t hash, int n_keys, - unsigned int *keys, + unsigned long *keys, net_hash_entry_t **hash_headpp, net_hash_entry_t *entpp); @@ -1195,7 +1196,7 @@ net_set_filter(ifp, rcv_port, priority, filter, filter_count) * If there is no match instruction, we allocate * a normal packet filter structure. */ - my_infp = (net_rcv_port_t) zalloc(net_rcv_zone); + my_infp = (net_rcv_port_t) kmem_cache_alloc(&net_rcv_cache); my_infp->rcv_port = rcv_port; is_new_infp = TRUE; } else { @@ -1205,7 +1206,7 @@ net_set_filter(ifp, rcv_port, priority, filter, filter_count) * a hash table to deal with them. */ my_infp = 0; - hash_entp = (net_hash_entry_t) zalloc(net_hash_entry_zone); + hash_entp = (net_hash_entry_t) kmem_cache_alloc(&net_hash_entry_cache); is_new_infp = FALSE; } @@ -1310,7 +1311,8 @@ net_set_filter(ifp, rcv_port, priority, filter, filter_count) ipc_port_release_send(rcv_port); if (match != 0) - zfree (net_hash_entry_zone, (vm_offset_t)hash_entp); + kmem_cache_free(&net_hash_entry_cache, + (vm_offset_t)hash_entp); rval = D_NO_MEMORY; goto clean_and_return; @@ -1526,20 +1528,12 @@ net_io_init() register vm_size_t size; size = sizeof(struct net_rcv_port); - net_rcv_zone = zinit(size, - 0, - size * 1000, - PAGE_SIZE, - FALSE, - "net_rcv_port"); + kmem_cache_init(&net_rcv_cache, "net_rcv_port", size, 0, + NULL, NULL, NULL, 0); size = sizeof(struct net_hash_entry); - net_hash_entry_zone = zinit(size, - 0, - size * 100, - PAGE_SIZE, - FALSE, - "net_hash_entry"); + kmem_cache_init(&net_hash_entry_cache, "net_hash_entry", size, 0, + NULL, NULL, NULL, 0); size = ikm_plus_overhead(sizeof(struct net_rcv_msg)); net_kmsg_size = round_page(size); @@ -1644,7 +1638,7 @@ bpf_do_filter(infp, p, wirelen, header, hlen, hash_headpp, entpp) register unsigned long A, X; register int k; - long mem[BPF_MEMWORDS]; + unsigned long mem[BPF_MEMWORDS]; /* Generic pointer to either HEADER or P according to the specified offset. */ char *data = NULL; @@ -1655,10 +1649,9 @@ bpf_do_filter(infp, p, wirelen, header, hlen, hash_headpp, entpp) buflen = NET_RCV_MAX; *entpp = 0; /* default */ -#ifdef lint A = 0; X = 0; -#endif + for (; pc < pc_end; ++pc) { switch (pc->code) { @@ -2039,7 +2032,7 @@ int bpf_match (hash, n_keys, keys, hash_headpp, entpp) net_hash_header_t hash; register int n_keys; - register unsigned int *keys; + register unsigned long *keys; net_hash_entry_t **hash_headpp, *entpp; { register net_hash_entry_t head, entp; @@ -2168,7 +2161,7 @@ net_free_dead_infp (dead_infp) nextfp = (net_rcv_port_t) queue_next(&infp->input); ipc_port_release_send(infp->rcv_port); net_del_q_info(infp->rcv_qlimit); - zfree(net_rcv_zone, (vm_offset_t) infp); + kmem_cache_free(&net_rcv_cache, (vm_offset_t) infp); } } @@ -2191,7 +2184,7 @@ net_free_dead_entp (dead_entp) ipc_port_release_send(entp->rcv_port); net_del_q_info(entp->rcv_qlimit); - zfree(net_hash_entry_zone, (vm_offset_t) entp); + kmem_cache_free(&net_hash_entry_cache, (vm_offset_t) entp); } } diff --git a/doc/.gitignore b/doc/.gitignore new file mode 100644 index 00000000..829355ba --- /dev/null +++ b/doc/.gitignore @@ -0,0 +1,4 @@ +/*.info +/*.info-* +/stamp-vti +/version.texi diff --git a/i386/Makefrag.am b/i386/Makefrag.am index db58c9c6..3338e6e6 100644 --- a/i386/Makefrag.am +++ b/i386/Makefrag.am @@ -136,7 +136,6 @@ libkernel_a_SOURCES += \ i386/i386/vm_param.h \ i386/i386/vm_tuning.h \ i386/i386/xpr.h \ - i386/i386/zalloc.h \ i386/intel/pmap.c \ i386/intel/pmap.h \ i386/intel/read_fault.c \ @@ -206,9 +205,16 @@ EXTRA_DIST += \ i386/ldscript if PLATFORM_at gnumach_LINKFLAGS += \ - --defsym _START=0x100000 \ + --defsym _START=0xC0100000 \ + --defsym _START_MAP=0x100000 \ -T '$(srcdir)'/i386/ldscript endif + +AM_CFLAGS += \ + -mno-3dnow \ + -mno-mmx \ + -mno-sse \ + -mno-sse2 # # Installation. diff --git a/i386/configfrag.ac b/i386/configfrag.ac index e4ce97ef..1eaabcad 100644 --- a/i386/configfrag.ac +++ b/i386/configfrag.ac @@ -23,7 +23,9 @@ dnl USE OF THIS SOFTWARE. # Some of the i386-specific code checks for these. AC_DEFINE([__ELF__], [1], [__ELF__]) - AC_DEFINE([i386], [1], [i386]) + + # Determines the size of the CPU cache line. + AC_DEFINE([CPU_L1_SHIFT], [6], [CPU_L1_SHIFT]) [# Does the architecture provide machine-specific interfaces? mach_machine_routines=1;; diff --git a/i386/i386/.gitignore b/i386/i386/.gitignore new file mode 100644 index 00000000..4520a2a1 --- /dev/null +++ b/i386/i386/.gitignore @@ -0,0 +1 @@ +/i386asm.h diff --git a/i386/i386/db_interface.c b/i386/i386/db_interface.c index d149adc5..66cc8b59 100644 --- a/i386/i386/db_interface.c +++ b/i386/i386/db_interface.c @@ -39,6 +39,7 @@ #include <i386/setjmp.h> #include <i386/pmap.h> #include <i386/proc_reg.h> +#include <i386/locore.h> #include "gdt.h" #include "trap.h" @@ -187,10 +188,10 @@ kdb_trap( * instead of those at its call to KDB. */ struct int_regs { - int edi; - int esi; - int ebp; - int ebx; + long edi; + long esi; + long ebp; + long ebx; struct i386_interrupt_state *is; }; @@ -390,6 +391,8 @@ db_write_bytes( oldmap1 = *ptep1; *ptep1 |= INTEL_PTE_WRITE; } + if (CPU_HAS_FEATURE(CPU_FEATURE_PGE)) + set_cr4(get_cr4() & ~CR4_PGE); flush_tlb(); } @@ -404,6 +407,8 @@ db_write_bytes( *ptep1 = oldmap1; } flush_tlb(); + if (CPU_HAS_FEATURE(CPU_FEATURE_PGE)) + set_cr4(get_cr4() | CR4_PGE); } } diff --git a/i386/i386/db_interface.h b/i386/i386/db_interface.h index a81c8eec..10a02e2b 100644 --- a/i386/i386/db_interface.h +++ b/i386/i386/db_interface.h @@ -62,6 +62,7 @@ extern void db_task_name (task_t task); #define I386_DB_LEN_1 0 #define I386_DB_LEN_2 1 #define I386_DB_LEN_4 3 +#define I386_DB_LEN_8 2 /* For >= Pentium4 and Xen CPUID >= 15 only */ #define I386_DB_LOCAL 1 #define I386_DB_GLOBAL 2 diff --git a/i386/i386/db_machdep.h b/i386/i386/db_machdep.h index 95e37971..1dba2cd8 100644 --- a/i386/i386/db_machdep.h +++ b/i386/i386/db_machdep.h @@ -39,7 +39,7 @@ #include <i386/trap.h> typedef vm_offset_t db_addr_t; /* address - unsigned */ -typedef int db_expr_t; /* expression - signed */ +typedef long db_expr_t; /* expression - signed */ typedef struct i386_saved_state db_regs_t; db_regs_t ddb_regs; /* register state */ diff --git a/i386/i386/db_trace.c b/i386/i386/db_trace.c index c4019b59..4e3bea39 100644 --- a/i386/i386/db_trace.c +++ b/i386/i386/db_trace.c @@ -45,7 +45,7 @@ #include "trap.h" -int +long db_i386_reg_value( struct db_variable *vp, db_expr_t *valuep, @@ -56,22 +56,22 @@ db_i386_reg_value( * Machine register set. */ struct db_variable db_regs[] = { - { "cs", (int *)&ddb_regs.cs, db_i386_reg_value }, - { "ds", (int *)&ddb_regs.ds, db_i386_reg_value }, - { "es", (int *)&ddb_regs.es, db_i386_reg_value }, - { "fs", (int *)&ddb_regs.fs, db_i386_reg_value }, - { "gs", (int *)&ddb_regs.gs, db_i386_reg_value }, - { "ss", (int *)&ddb_regs.ss, db_i386_reg_value }, - { "eax",(int *)&ddb_regs.eax, db_i386_reg_value }, - { "ecx",(int *)&ddb_regs.ecx, db_i386_reg_value }, - { "edx",(int *)&ddb_regs.edx, db_i386_reg_value }, - { "ebx",(int *)&ddb_regs.ebx, db_i386_reg_value }, - { "esp",(int *)&ddb_regs.uesp,db_i386_reg_value }, - { "ebp",(int *)&ddb_regs.ebp, db_i386_reg_value }, - { "esi",(int *)&ddb_regs.esi, db_i386_reg_value }, - { "edi",(int *)&ddb_regs.edi, db_i386_reg_value }, - { "eip",(int *)&ddb_regs.eip, db_i386_reg_value }, - { "efl",(int *)&ddb_regs.efl, db_i386_reg_value }, + { "cs", (long *)&ddb_regs.cs, db_i386_reg_value }, + { "ds", (long *)&ddb_regs.ds, db_i386_reg_value }, + { "es", (long *)&ddb_regs.es, db_i386_reg_value }, + { "fs", (long *)&ddb_regs.fs, db_i386_reg_value }, + { "gs", (long *)&ddb_regs.gs, db_i386_reg_value }, + { "ss", (long *)&ddb_regs.ss, db_i386_reg_value }, + { "eax",(long *)&ddb_regs.eax, db_i386_reg_value }, + { "ecx",(long *)&ddb_regs.ecx, db_i386_reg_value }, + { "edx",(long *)&ddb_regs.edx, db_i386_reg_value }, + { "ebx",(long *)&ddb_regs.ebx, db_i386_reg_value }, + { "esp",(long *)&ddb_regs.uesp,db_i386_reg_value }, + { "ebp",(long *)&ddb_regs.ebp, db_i386_reg_value }, + { "esi",(long *)&ddb_regs.esi, db_i386_reg_value }, + { "edi",(long *)&ddb_regs.edi, db_i386_reg_value }, + { "eip",(long *)&ddb_regs.eip, db_i386_reg_value }, + { "efl",(long *)&ddb_regs.efl, db_i386_reg_value }, }; struct db_variable *db_eregs = db_regs + sizeof(db_regs)/sizeof(db_regs[0]); @@ -82,8 +82,8 @@ struct db_variable *db_eregs = db_regs + sizeof(db_regs)/sizeof(db_regs[0]); struct i386_frame { struct i386_frame *f_frame; - int f_retaddr; - int f_arg0; + long f_retaddr; + long f_arg0; }; #define TRAP 1 @@ -99,18 +99,18 @@ boolean_t db_trace_symbols_found = FALSE; struct i386_kregs { char *name; - int offset; + long offset; } i386_kregs[] = { - { "ebx", (int)(&((struct i386_kernel_state *)0)->k_ebx) }, - { "esp", (int)(&((struct i386_kernel_state *)0)->k_esp) }, - { "ebp", (int)(&((struct i386_kernel_state *)0)->k_ebp) }, - { "edi", (int)(&((struct i386_kernel_state *)0)->k_edi) }, - { "esi", (int)(&((struct i386_kernel_state *)0)->k_esi) }, - { "eip", (int)(&((struct i386_kernel_state *)0)->k_eip) }, + { "ebx", (long)(&((struct i386_kernel_state *)0)->k_ebx) }, + { "esp", (long)(&((struct i386_kernel_state *)0)->k_esp) }, + { "ebp", (long)(&((struct i386_kernel_state *)0)->k_ebp) }, + { "edi", (long)(&((struct i386_kernel_state *)0)->k_edi) }, + { "esi", (long)(&((struct i386_kernel_state *)0)->k_esi) }, + { "eip", (long)(&((struct i386_kernel_state *)0)->k_eip) }, { 0 }, }; -int * +long * db_lookup_i386_kreg( char *name, int *kregp) @@ -119,19 +119,19 @@ db_lookup_i386_kreg( for (kp = i386_kregs; kp->name; kp++) { if (strcmp(name, kp->name) == 0) - return (int *)((int)kregp + kp->offset); + return (long *)((long)kregp + kp->offset); } return 0; } -int +long db_i386_reg_value( struct db_variable *vp, db_expr_t *valuep, int flag, db_var_aux_param_t ap) { - int *dp = 0; + long *dp = 0; db_expr_t null_reg = 0; register thread_t thread = ap->thread; extern unsigned int_stack_high; @@ -153,15 +153,15 @@ db_i386_reg_value( } else if ((thread->state & TH_SWAPPED) == 0 && thread->kernel_stack) { dp = db_lookup_i386_kreg(vp->name, - (int *)(STACK_IKS(thread->kernel_stack))); + (long *)(STACK_IKS(thread->kernel_stack))); if (dp == 0) dp = &null_reg; } else if ((thread->state & TH_SWAPPED) && thread->swap_func != thread_exception_return) { /*.....this breaks t/t $taskN.0...*/ /* only EIP is valid */ - if (vp->valuep == (int *) &ddb_regs.eip) { - dp = (int *)(&thread->swap_func); + if (vp->valuep == (long *) &ddb_regs.eip) { + dp = (long *)(&thread->swap_func); } else { dp = &null_reg; } @@ -170,8 +170,8 @@ db_i386_reg_value( if (dp == 0) { if (thread->pcb == 0) db_error("no pcb\n"); - dp = (int *)((int)(&thread->pcb->iss) + - ((int)vp->valuep - (int)&ddb_regs)); + dp = (long *)((long)(&thread->pcb->iss) + + ((long)vp->valuep - (long)&ddb_regs)); } if (flag == DB_VAR_SET) *dp = *valuep; @@ -212,18 +212,18 @@ db_numargs( struct i386_frame *fp, task_t task) { - int *argp; - int inst; - int args; + long *argp; + long inst; + long args; extern char etext[]; - argp = (int *)db_get_task_value((int)&fp->f_retaddr, 4, FALSE, task); - if (argp < (int *)VM_MIN_KERNEL_ADDRESS || argp > (int *)etext) + argp = (long *)db_get_task_value((long)&fp->f_retaddr, sizeof(long), FALSE, task); + if (argp < (long *)VM_MIN_KERNEL_ADDRESS || argp > (long *)etext) args = db_numargs_default; - else if (!DB_CHECK_ACCESS((int)argp, 4, task)) + else if (!DB_CHECK_ACCESS((long)argp, sizeof(long), task)) args = db_numargs_default; else { - inst = db_get_task_value((int)argp, 4, FALSE, task); + inst = db_get_task_value((long)argp, sizeof(long), FALSE, task); if ((inst & 0xff) == 0x59) /* popl %ecx */ args = 1; else if ((inst & 0xffff) == 0xc483) /* addl %n, %esp */ @@ -236,16 +236,16 @@ db_numargs( struct interrupt_frame { struct i386_frame *if_frame; /* point to next frame */ - int if_retaddr; /* return address to _interrupt */ - int if_unit; /* unit number */ + long if_retaddr; /* return address to _interrupt */ + long if_unit; /* unit number */ spl_t if_spl; /* saved spl */ - int if_iretaddr; /* _return_to_{iret,iret_i} */ - int if_edx; /* old sp(iret) or saved edx(iret_i) */ - int if_ecx; /* saved ecx(iret_i) */ - int if_eax; /* saved eax(iret_i) */ - int if_eip; /* saved eip(iret_i) */ - int if_cs; /* saved cs(iret_i) */ - int if_efl; /* saved efl(iret_i) */ + long if_iretaddr; /* _return_to_{iret,iret_i} */ + long if_edx; /* old sp(iret) or saved edx(iret_i) */ + long if_ecx; /* saved ecx(iret_i) */ + long if_eax; /* saved eax(iret_i) */ + long if_eip; /* saved eip(iret_i) */ + long if_cs; /* saved cs(iret_i) */ + long if_efl; /* saved efl(iret_i) */ }; /* @@ -263,7 +263,7 @@ db_nextframe( struct i386_frame **lfp, /* in/out */ struct i386_frame **fp, /* in/out */ db_addr_t *ip, /* out */ - int frame_type, /* in */ + long frame_type, /* in */ thread_t thread) /* in */ { struct i386_saved_state *saved_regs; @@ -277,7 +277,7 @@ db_nextframe( * it is an (struct i386_saved_state *). */ saved_regs = (struct i386_saved_state *) - db_get_task_value((int)&((*fp)->f_arg0),4,FALSE,task); + db_get_task_value((long)&((*fp)->f_arg0),sizeof(long),FALSE,task); db_printf(">>>>> %s (%d) at ", trap_name(saved_regs->trapno), saved_regs->trapno); db_task_printsym(saved_regs->eip, DB_STGY_PROC, task); @@ -310,10 +310,10 @@ db_nextframe( default: miss_frame: *ip = (db_addr_t) - db_get_task_value((int)&(*fp)->f_retaddr, 4, FALSE, task); + db_get_task_value((long)&(*fp)->f_retaddr, sizeof(long), FALSE, task); *lfp = *fp; *fp = (struct i386_frame *) - db_get_task_value((int)&(*fp)->f_frame, 4, FALSE, task); + db_get_task_value((long)&(*fp)->f_frame, sizeof(long), FALSE, task); break; } } @@ -401,7 +401,7 @@ db_stack_trace_cmd( } else { frame = (struct i386_frame *)addr; th = (db_default_thread)? db_default_thread: current_thread(); - callpc = (db_addr_t)db_get_task_value((int)&frame->f_retaddr, 4, + callpc = (db_addr_t)db_get_task_value((long)&frame->f_retaddr, sizeof(long), FALSE, (th == THREAD_NULL) ? TASK_NULL : th->task); } @@ -420,13 +420,13 @@ db_i386_stack_trace( { task_t task; boolean_t kernel_only; - int *argp; - int user_frame = 0; + long *argp; + long user_frame = 0; struct i386_frame *lastframe; int frame_type; char *filename; int linenum; - extern unsigned int db_maxoff; + extern unsigned long db_maxoff; if (count == -1) count = 65535; @@ -438,7 +438,7 @@ db_i386_stack_trace( if (!db_trace_symbols_found) db_find_trace_symbols(); - if (!INKERNEL((unsigned)callpc) && !INKERNEL((unsigned)frame)) { + if (!INKERNEL((unsigned long)callpc) && !INKERNEL((unsigned long)frame)) { db_printf(">>>>> user space <<<<<\n"); user_frame++; } @@ -449,7 +449,7 @@ db_i386_stack_trace( char * name; db_expr_t offset; - if (INKERNEL((unsigned)callpc) && user_frame == 0) { + if (INKERNEL((unsigned long)callpc) && user_frame == 0) { db_addr_t call_func = 0; db_sym_t sym_tmp; @@ -474,7 +474,7 @@ db_i386_stack_trace( frame_type = 0; narg = db_numargs(frame, task); } - } else if (INKERNEL((unsigned)callpc) ^ INKERNEL((unsigned)frame)) { + } else if (INKERNEL((unsigned long)callpc) ^ INKERNEL((unsigned long)frame)) { frame_type = 0; narg = -1; } else { @@ -492,7 +492,7 @@ db_i386_stack_trace( argp = &frame->f_arg0; while (narg > 0) { - db_printf("%x", db_get_task_value((int)argp,4,FALSE,task)); + db_printf("%x", db_get_task_value((long)argp,sizeof(long),FALSE,task)); argp++; if (--narg != 0) db_printf(","); @@ -519,7 +519,7 @@ db_i386_stack_trace( break; } if (!INKERNEL(lastframe) || - (!INKERNEL((unsigned)callpc) && !INKERNEL((unsigned)frame))) + (!INKERNEL((unsigned long)callpc) && !INKERNEL((unsigned long)frame))) user_frame++; if (user_frame == 1) { db_printf(">>>>> user space <<<<<\n"); @@ -570,6 +570,7 @@ static void db_cproc_state( } /* offsets in a cproc structure */ +/* TODO: longs? */ int db_cproc_next_offset = 0 * 4; int db_cproc_incarnation_offset = 1 * 4; int db_cproc_list_offset = 2 * 4; diff --git a/i386/i386/debug_i386.c b/i386/i386/debug_i386.c index f3e18835..937d7b4b 100644 --- a/i386/i386/debug_i386.c +++ b/i386/i386/debug_i386.c @@ -30,19 +30,21 @@ void dump_ss(struct i386_saved_state *st) { printf("Dump of i386_saved_state %p:\n", st); - printf("EAX %08x EBX %08x ECX %08x EDX %08x\n", + printf("EAX %08lx EBX %08lx ECX %08lx EDX %08lx\n", st->eax, st->ebx, st->ecx, st->edx); - printf("ESI %08x EDI %08x EBP %08x ESP %08x\n", + printf("ESI %08lx EDI %08lx EBP %08lx ESP %08lx\n", st->esi, st->edi, st->ebp, st->uesp); - printf("CS %04x SS %04x DS %04x ES %04x FS %04x GS %04x\n", + printf("CS %04lx SS %04lx " + "DS %04lx ES %04lx " + "FS %04lx GS %04lx\n", st->cs & 0xffff, st->ss & 0xffff, st->ds & 0xffff, st->es & 0xffff, st->fs & 0xffff, st->gs & 0xffff); - printf("v86: DS %04x ES %04x FS %04x GS %04x\n", + printf("v86: DS %04lx ES %04lx FS %04lx GS %04lx\n", st->v86_segs.v86_ds & 0xffff, st->v86_segs.v86_es & 0xffff, st->v86_segs.v86_gs & 0xffff, st->v86_segs.v86_gs & 0xffff); - printf("EIP %08x EFLAGS %08x\n", st->eip, st->efl); - printf("trapno %d: %s, error %08x\n", + printf("EIP %08lx EFLAGS %08lx\n", st->eip, st->efl); + printf("trapno %ld: %s, error %08lx\n", st->trapno, trap_name(st->trapno), st->err); } diff --git a/i386/i386/fpu.c b/i386/i386/fpu.c index 2a4b9c09..8efe9e80 100644 --- a/i386/i386/fpu.c +++ b/i386/i386/fpu.c @@ -44,20 +44,19 @@ #include <kern/debug.h> #include <machine/machspl.h> /* spls */ -#include <kern/mach_param.h> #include <kern/printf.h> #include <kern/thread.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <i386/thread.h> #include <i386/fpu.h> #include <i386/pio.h> +#include <i386/pic.h> #include <i386/locore.h> #include "cpu_number.h" #if 0 #include <i386/ipl.h> -extern int curr_ipl; #define ASSERT_IPL(L) \ { \ if (curr_ipl != L) { \ @@ -72,7 +71,7 @@ extern int curr_ipl; extern void i386_exception(); int fp_kind = FP_387; /* 80387 present */ -zone_t ifps_zone; /* zone for FPU save area */ +struct kmem_cache ifps_cache; /* cache for FPU save area */ static unsigned long mxcsr_feature_mask = 0xffffffff; /* Always AND user-provided mxcsr with this security mask */ void fp_save(thread_t thread); @@ -110,9 +109,9 @@ init_fpu() { unsigned short status, control; -#ifdef MACH_HYP +#ifdef MACH_RING1 clear_ts(); -#else /* MACH_HYP */ +#else /* MACH_RING1 */ unsigned int native = 0; if (machine_slot[cpu_number()].cpu_type >= CPU_TYPE_I486) @@ -124,7 +123,7 @@ init_fpu() * the control and status registers. */ set_cr0((get_cr0() & ~(CR0_EM|CR0_TS)) | native); /* allow use of FPU */ -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ fninit(); status = fnstsw(); @@ -158,10 +157,10 @@ init_fpu() struct i386_xfp_save save; unsigned long mask; fp_kind = FP_387X; -#ifndef MACH_HYP +#ifndef MACH_RING1 printf("Enabling FXSR\n"); set_cr4(get_cr4() | CR4_OSFXSR); -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ fxsave(&save); mask = save.fp_mxcsr_mask; if (!mask) @@ -170,14 +169,14 @@ init_fpu() } else fp_kind = FP_387; } -#ifdef MACH_HYP +#ifdef MACH_RING1 set_ts(); -#else /* MACH_HYP */ +#else /* MACH_RING1 */ /* * Trap wait instructions. Turn off FPU for now. */ set_cr0(get_cr0() | CR0_TS | CR0_MP); -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ } else { /* @@ -193,10 +192,9 @@ init_fpu() void fpu_module_init() { - ifps_zone = zinit(sizeof(struct i386_fpsave_state), 16, - THREAD_MAX * sizeof(struct i386_fpsave_state), - THREAD_CHUNK * sizeof(struct i386_fpsave_state), - 0, "i386 fpsave state"); + kmem_cache_init(&ifps_cache, "i386_fpsave_state", + sizeof(struct i386_fpsave_state), 16, + NULL, NULL, NULL, 0); } /* @@ -221,7 +219,7 @@ ASSERT_IPL(SPL0); clear_fpu(); } #endif /* NCPUS == 1 */ - zfree(ifps_zone, (vm_offset_t) fps); + kmem_cache_free(&ifps_cache, (vm_offset_t) fps); } /* The two following functions were stolen from Linux's i387.c */ @@ -335,7 +333,7 @@ ASSERT_IPL(SPL0); simple_unlock(&pcb->lock); if (ifps != 0) { - zfree(ifps_zone, (vm_offset_t) ifps); + kmem_cache_free(&ifps_cache, (vm_offset_t) ifps); } } else { @@ -356,7 +354,7 @@ ASSERT_IPL(SPL0); if (ifps == 0) { if (new_ifps == 0) { simple_unlock(&pcb->lock); - new_ifps = (struct i386_fpsave_state *) zalloc(ifps_zone); + new_ifps = (struct i386_fpsave_state *) kmem_cache_alloc(&ifps_cache); goto Retry; } ifps = new_ifps; @@ -396,7 +394,7 @@ ASSERT_IPL(SPL0); simple_unlock(&pcb->lock); if (new_ifps != 0) - zfree(ifps_zone, (vm_offset_t) new_ifps); + kmem_cache_free(&ifps_cache, (vm_offset_t) new_ifps); } return KERN_SUCCESS; @@ -609,7 +607,7 @@ fpextovrflt() clear_fpu(); if (ifps) - zfree(ifps_zone, (vm_offset_t) ifps); + kmem_cache_free(&ifps_cache, (vm_offset_t) ifps); /* * Raise exception. @@ -686,7 +684,7 @@ fpexterrflt() /*NOTREACHED*/ } -#ifndef MACH_XEN +#ifndef MACH_RING1 /* * FPU error. Called by AST. */ @@ -743,7 +741,7 @@ ASSERT_IPL(SPL0); thread->pcb->ims.ifps->fp_save_state.fp_status); /*NOTREACHED*/ } -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ /* * Save FPU state. @@ -785,7 +783,7 @@ fp_load(thread) ASSERT_IPL(SPL0); ifps = pcb->ims.ifps; if (ifps == 0) { - ifps = (struct i386_fpsave_state *) zalloc(ifps_zone); + ifps = (struct i386_fpsave_state *) kmem_cache_alloc(&ifps_cache); memset(ifps, 0, sizeof *ifps); pcb->ims.ifps = ifps; fpinit(); @@ -836,7 +834,7 @@ fp_state_alloc() pcb_t pcb = current_thread()->pcb; struct i386_fpsave_state *ifps; - ifps = (struct i386_fpsave_state *)zalloc(ifps_zone); + ifps = (struct i386_fpsave_state *)kmem_cache_alloc(&ifps_cache); memset(ifps, 0, sizeof *ifps); pcb->ims.ifps = ifps; @@ -865,7 +863,7 @@ fp_state_alloc() * This comes in on line 5 of the slave PIC at SPL1. */ void -fpintr() +fpintr(int unit) { spl_t s; thread_t thread = current_thread(); diff --git a/i386/i386/fpu.h b/i386/i386/fpu.h index 1a1b61f6..21561875 100644 --- a/i386/i386/fpu.h +++ b/i386/i386/fpu.h @@ -124,5 +124,6 @@ extern void fpextovrflt(void); extern void fpexterrflt(void); extern void fpastintr(void); extern void init_fpu(void); +extern void fpintr(int unit); #endif /* _I386_FPU_H_ */ diff --git a/i386/i386/gdt.c b/i386/i386/gdt.c index f26a50c9..5523fea3 100644 --- a/i386/i386/gdt.c +++ b/i386/i386/gdt.c @@ -39,10 +39,10 @@ #include "seg.h" #include "gdt.h" -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS /* It is actually defined in xen_boothdr.S */ extern -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ struct real_descriptor gdt[GDTSZ]; void @@ -57,25 +57,29 @@ gdt_init() LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS, LINEAR_MAX_KERNEL_ADDRESS - (LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) - 1, ACC_PL_K|ACC_DATA_W, SZ_32); -#ifndef MACH_HYP +#ifndef MACH_PV_DESCRIPTORS fill_gdt_descriptor(LINEAR_DS, 0, 0xffffffff, ACC_PL_K|ACC_DATA_W, SZ_32); -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS unsigned long frame = kv_to_mfn(gdt); pmap_set_page_readonly(gdt); if (hyp_set_gdt(kv_to_la(&frame), GDTSZ)) panic("couldn't set gdt\n"); +#endif +#ifdef MACH_PV_PAGETABLES if (hyp_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments)) panic("couldn't set 4gb segments vm assist"); #if 0 if (hyp_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments_notify)) panic("couldn't set 4gb segments vm assist notify"); #endif -#else /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ + +#ifndef MACH_PV_DESCRIPTORS /* Load the new GDT. */ { struct pseudo_descriptor pdesc; @@ -84,7 +88,7 @@ gdt_init() pdesc.linear_base = kvtolin(&gdt); lgdt(&pdesc); } -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* Reload all the segment registers from the new GDT. We must load ds and es with 0 before loading them with KERNEL_DS @@ -101,7 +105,7 @@ gdt_init() "movw %w1,%%es\n" "movw %w1,%%ss\n" : : "i" (KERNEL_CS), "r" (KERNEL_DS), "r" (0)); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES #if VM_MIN_KERNEL_ADDRESS != LINEAR_MIN_KERNEL_ADDRESS /* things now get shifted */ #ifdef MACH_PSEUDO_PHYS @@ -109,6 +113,6 @@ gdt_init() #endif /* MACH_PSEUDO_PHYS */ la_shift += LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS; #endif -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } diff --git a/i386/i386/gdt.h b/i386/i386/gdt.h index 41ace791..d865640b 100644 --- a/i386/i386/gdt.h +++ b/i386/i386/gdt.h @@ -40,16 +40,16 @@ */ #define KERNEL_CS (0x08 | KERNEL_RING) /* kernel code */ #define KERNEL_DS (0x10 | KERNEL_RING) /* kernel data */ -#ifndef MACH_XEN +#ifndef MACH_PV_DESCRIPTORS #define KERNEL_LDT 0x18 /* master LDT */ -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ #define KERNEL_TSS 0x20 /* master TSS (uniprocessor) */ #define USER_LDT 0x28 /* place for per-thread LDT */ #define USER_TSS 0x30 /* place for per-thread TSS that holds IO bitmap */ -#ifndef MACH_HYP +#ifndef MACH_PV_DESCRIPTORS #define LINEAR_DS 0x38 /* linear mapping */ -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ /* 0x40 was USER_FPREGS, now free */ #define USER_GDT 0x48 /* user-defined GDT entries */ diff --git a/i386/i386/hardclock.c b/i386/i386/hardclock.c index a8846b95..66150332 100644 --- a/i386/i386/hardclock.c +++ b/i386/i386/hardclock.c @@ -39,13 +39,17 @@ #include <i386/ipl.h> #endif -extern void clock_interrupt(); +#ifdef LINUX_DEV +#include <linux/dev/glue/glue.h> +#endif + extern char return_to_iret[]; void -hardclock(iunit, old_ipl, ret_addr, regs) +hardclock(iunit, old_ipl, irq, ret_addr, regs) int iunit; /* 'unit' number */ int old_ipl; /* old interrupt level */ + int irq; /* irq number */ char * ret_addr; /* return address in interrupt handler */ struct i386_interrupt_state *regs; /* saved registers */ diff --git a/i386/i386/i386asm.sym b/i386/i386/i386asm.sym index b1670e8b..dd1a2edf 100644 --- a/i386/i386/i386asm.sym +++ b/i386/i386/i386asm.sym @@ -78,6 +78,7 @@ offset i386_saved_state r err offset i386_saved_state r efl R_EFLAGS offset i386_saved_state r eip offset i386_saved_state r cr2 +offset i386_saved_state r edi offset i386_interrupt_state i eip offset i386_interrupt_state i cs @@ -86,12 +87,14 @@ offset i386_interrupt_state i efl offset i386_tss tss esp0 offset i386_tss tss ss0 +offset machine_slot sub_type cpu_type + expr I386_PGBYTES NBPG expr VM_MIN_ADDRESS expr VM_MAX_ADDRESS expr VM_MIN_KERNEL_ADDRESS KERNELBASE expr KERNEL_STACK_SIZE -#if VM_MIN_KERNEL_ADDRESS == LINEAR_MIN_KERNEL_ADDRESS +#if defined MACH_PSEUDO_PHYS && (VM_MIN_KERNEL_ADDRESS == LINEAR_MIN_KERNEL_ADDRESS) expr PFN_LIST pfn_list #endif @@ -121,9 +124,9 @@ expr KERNEL_RING expr KERNEL_CS expr KERNEL_DS expr KERNEL_TSS -#ifndef MACH_XEN +#ifndef MACH_PV_DESCRIPTORS expr KERNEL_LDT -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ expr (VM_MIN_KERNEL_ADDRESS>>PDESHIFT)*sizeof(pt_entry_t) KERNELBASEPDE @@ -148,5 +151,7 @@ offset shared_info si vcpu_info[0].evtchn_upcall_pending CPU_PENDING offset shared_info si vcpu_info[0].evtchn_pending_sel CPU_PENDING_SEL offset shared_info si evtchn_pending PENDING offset shared_info si evtchn_mask EVTMASK +#ifdef MACH_PV_PAGETABLES offset shared_info si vcpu_info[0].arch.cr2 CR2 +#endif /* MACH_PV_PAGETABLES */ #endif /* MACH_XEN */ diff --git a/i386/i386/idt.c b/i386/i386/idt.c index b5e3d080..882764f4 100644 --- a/i386/i386/idt.c +++ b/i386/i386/idt.c @@ -38,10 +38,10 @@ extern struct idt_init_entry idt_inittab[]; void idt_init() { -#ifdef MACH_HYP +#ifdef MACH_PV_DESCRIPTORS if (hyp_set_trap_table(kvtolin(idt_inittab))) panic("couldn't set trap table\n"); -#else /* MACH_HYP */ +#else /* MACH_PV_DESCRIPTORS */ struct idt_init_entry *iie = idt_inittab; /* Initialize the exception vectors from the idt_inittab. */ @@ -59,6 +59,6 @@ void idt_init() pdesc.linear_base = kvtolin(&idt); lidt(&pdesc); } -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ } diff --git a/i386/i386/idt_inittab.S b/i386/i386/idt_inittab.S index 63e554bd..8e92d805 100644 --- a/i386/i386/idt_inittab.S +++ b/i386/i386/idt_inittab.S @@ -39,7 +39,7 @@ ENTRY(idt_inittab) /* * Interrupt descriptor table and code vectors for it. */ -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS #define IDT_ENTRY(n,entry,type) \ .data 2 ;\ .byte n ;\ @@ -47,14 +47,14 @@ ENTRY(idt_inittab) .word KERNEL_CS ;\ .long entry ;\ .text -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ #define IDT_ENTRY(n,entry,type) \ .data 2 ;\ .long entry ;\ .word n ;\ .word type ;\ .text -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* * No error code. Clear error code and push trap number. @@ -108,7 +108,7 @@ EXCEP_SPC(0x0b,t_segnp) EXCEP_ERR(0x0c,t_stack_fault) EXCEP_SPC(0x0d,t_gen_prot) EXCEP_SPC(0x0e,t_page_fault) -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS EXCEP_ERR(0x0f,t_trap_0f) #else EXCEPTION(0x0f,t_trap_0f) @@ -133,7 +133,7 @@ EXCEPTION(0x1f,t_trap_1f) /* Terminator */ .data 2 .long 0 -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS .long 0 -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ diff --git a/i386/i386/io_map.c b/i386/i386/io_map.c index 5b77552d..b095f224 100644 --- a/i386/i386/io_map.c +++ b/i386/i386/io_map.c @@ -49,7 +49,7 @@ io_map(phys_addr, size) */ start = kernel_virtual_start; kernel_virtual_start += round_page(size); - printf("stealing kernel virtual addresses %08x-%08x\n", start, kernel_virtual_start); + printf("stealing kernel virtual addresses %08lx-%08lx\n", start, kernel_virtual_start); } else { (void) kmem_alloc_pageable(kernel_map, &start, round_page(size)); diff --git a/i386/i386/io_perm.c b/i386/i386/io_perm.c index df25cc63..8bacb8d5 100644 --- a/i386/i386/io_perm.c +++ b/i386/i386/io_perm.c @@ -54,7 +54,8 @@ #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> -#include <kern/zalloc.h> +#include <kern/slab.h> +#include <kern/kalloc.h> #include <kern/lock.h> #include <kern/queue.h> #include <kern/thread.h> @@ -257,12 +258,12 @@ i386_io_perm_modify (task_t target_task, io_perm_t io_perm, boolean_t enable) if (!iopb) { simple_unlock (&target_task->machine.iopb_lock); - iopb = (unsigned char *) zalloc (machine_task_iopb_zone); + iopb = (unsigned char *) kmem_cache_alloc (&machine_task_iopb_cache); simple_lock (&target_task->machine.iopb_lock); if (target_task->machine.iopb) { if (iopb) - zfree (machine_task_iopb_zone, (vm_offset_t) iopb); + kmem_cache_free (&machine_task_iopb_cache, (vm_offset_t) iopb); iopb = target_task->machine.iopb; iopb_size = target_task->machine.iopb_size; } diff --git a/i386/i386/ipl.h b/i386/i386/ipl.h index 557cd8df..8f729e1d 100644 --- a/i386/i386/ipl.h +++ b/i386/i386/ipl.h @@ -73,5 +73,6 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. extern void (*ivect[])(); extern int iunit[]; extern int intpri[]; +extern spl_t curr_ipl; #endif /* __ASSEMBLER__ */ #endif /* KERNEL */ diff --git a/i386/i386/ktss.c b/i386/i386/ktss.c index 66432f3e..1b2938a0 100644 --- a/i386/i386/ktss.c +++ b/i386/i386/ktss.c @@ -45,12 +45,12 @@ ktss_init() /* XXX temporary exception stack */ static int exception_stack[1024]; -#ifdef MACH_XEN +#ifdef MACH_RING1 /* Xen won't allow us to do any I/O by default anyway, just register * exception stack */ - if (hyp_stack_switch(KERNEL_DS, (unsigned)(exception_stack+1024))) + if (hyp_stack_switch(KERNEL_DS, (unsigned long)(exception_stack+1024))) panic("couldn't register exception stack\n"); -#else /* MACH_XEN */ +#else /* MACH_RING1 */ /* Initialize the master TSS descriptor. */ fill_gdt_descriptor(KERNEL_TSS, kvtolin(&ktss), sizeof(struct task_tss) - 1, @@ -58,13 +58,13 @@ ktss_init() /* Initialize the master TSS. */ ktss.tss.ss0 = KERNEL_DS; - ktss.tss.esp0 = (unsigned)(exception_stack+1024); + ktss.tss.esp0 = (unsigned long)(exception_stack+1024); ktss.tss.io_bit_map_offset = IOPB_INVAL; /* Set the last byte in the I/O bitmap to all 1's. */ ktss.barrier = 0xff; /* Load the TSS. */ ltr(KERNEL_TSS); -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ } diff --git a/i386/i386/ldt.c b/i386/i386/ldt.c index 0ef7a8c4..43b9efb5 100644 --- a/i386/i386/ldt.c +++ b/i386/i386/ldt.c @@ -39,23 +39,25 @@ extern int syscall(); -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS /* It is actually defined in xen_boothdr.S */ extern -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ struct real_descriptor ldt[LDTSZ]; void ldt_init() { -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS +#ifdef MACH_PV_PAGETABLES pmap_set_page_readwrite(ldt); -#else /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ +#else /* MACH_PV_DESCRIPTORS */ /* Initialize the master LDT descriptor in the GDT. */ fill_gdt_descriptor(KERNEL_LDT, kvtolin(&ldt), sizeof(ldt)-1, ACC_PL_K|ACC_LDT, 0); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* Initialize the LDT descriptors. */ fill_ldt_gate(USER_SCALL, @@ -72,9 +74,9 @@ ldt_init() ACC_PL_U|ACC_DATA_W, SZ_32); /* Activate the LDT. */ -#ifdef MACH_HYP +#ifdef MACH_PV_DESCRIPTORS hyp_set_ldt(&ldt, LDTSZ); -#else /* MACH_HYP */ +#else /* MACH_PV_DESCRIPTORS */ lldt(KERNEL_LDT); -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ } diff --git a/i386/i386/locore.S b/i386/i386/locore.S index 11214e25..eeef09d7 100644 --- a/i386/i386/locore.S +++ b/i386/i386/locore.S @@ -442,12 +442,12 @@ ENTRY(t_debug) ENTRY(t_page_fault) pushl $(T_PAGE_FAULT) /* mark a page fault trap */ pusha /* save the general registers */ -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES movl %ss:hyp_shared_info+CR2,%eax -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ movl %cr2,%eax /* get the faulting address */ -#endif /* MACH_XEN */ - movl %eax,12(%esp) /* save in esp save slot */ +#endif /* MACH_PV_PAGETABLES */ + movl %eax,R_CR2-R_EDI(%esp) /* save in esp save slot */ jmp trap_push_segs /* continue fault */ /* @@ -785,6 +785,7 @@ ast_from_interrupt: * #ifndef MACH_XEN * iunit * saved SPL + * irq * #endif * return address == return_to_iret_i * saved %edx @@ -800,6 +801,7 @@ ast_from_interrupt: * #ifndef MACH_XEN * iunit * saved SPL + * irq * #endif * return address == return_to_iret * pointer to save area on old stack @@ -1299,48 +1301,51 @@ Entry(copyoutmsg) movl $USER_DS,%eax /* use user data segment for accesses */ mov %ax,%es +#if !defined(MACH_HYP) && !PAE + cmpl $3,machine_slot+SUB_TYPE_CPU_TYPE + jbe copyout_retry /* Use slow version on i386 */ +#endif /* !defined(MACH_HYP) && !PAE */ + + movl %edx,%eax /* use count */ + /*cld*/ /* count up: always this way in GCC code */ + movl %eax,%ecx /* move by longwords first */ + shrl $2,%ecx + RECOVER(copyout_fail) + rep + movsl + movl %eax,%ecx /* now move remaining bytes */ + andl $3,%ecx + RECOVER(copyout_fail) + rep + movsb /* move */ + xorl %eax,%eax /* return 0 for success */ + +copyout_ret: + mov %ss,%di /* restore ES to kernel segment */ + mov %di,%es + + popl %edi /* restore registers */ + popl %esi + ret /* and return */ + +copyout_fail: + movl $1,%eax /* return 1 for failure */ + jmp copyout_ret /* pop frame and return */ + +#if !defined(MACH_HYP) && !PAE /* * Check whether user address space is writable - * before writing to it - hardware is broken. - * XXX only have to do this on 386's. + * before writing to it - i386 hardware is broken. */ copyout_retry: -#ifdef MACH_HYP - movl cr3,%ecx /* point to page directory */ -#else /* MACH_HYP */ movl %cr3,%ecx /* point to page directory */ -#endif /* MACH_HYP */ -#if PAE - movl %edi,%eax /* get page directory pointer bits */ - shrl $(PDPSHIFT),%eax /* from user address */ - movl KERNELBASE(%ecx,%eax,PTE_SIZE),%ecx - /* get page directory pointer */ -#ifdef MACH_PSEUDO_PHYS - shrl $(PTESHIFT),%ecx - movl pfn_list,%eax - movl (%eax,%ecx,4),%ecx /* mfn_to_pfn */ - shll $(PTESHIFT),%ecx -#else /* MACH_PSEUDO_PHYS */ - andl $(PTE_PFN),%ecx /* isolate page frame address */ -#endif /* MACH_PSEUDO_PHYS */ -#endif /* PAE */ movl %edi,%eax /* get page directory bits */ shrl $(PDESHIFT),%eax /* from user address */ -#if PAE - andl $(PDEMASK),%eax -#endif /* PAE */ movl KERNELBASE(%ecx,%eax,PTE_SIZE),%ecx /* get page directory pointer */ testl $(PTE_V),%ecx /* present? */ jz 0f /* if not, fault is OK */ -#ifdef MACH_PSEUDO_PHYS - shrl $(PTESHIFT),%ecx - movl pfn_list,%eax - movl (%eax,%ecx,4),%ecx /* mfn_to_pfn */ - shll $(PTESHIFT),%ecx -#else /* MACH_PSEUDO_PHYS */ andl $(PTE_PFN),%ecx /* isolate page frame address */ -#endif /* MACH_PSEUDO_PHYS */ movl %edi,%eax /* get page table bits */ shrl $(PTESHIFT),%eax andl $(PTEMASK),%eax /* from user address */ @@ -1354,17 +1359,9 @@ copyout_retry: /* * Not writable - must fake a fault. Turn off access to the page. */ -#ifdef MACH_HYP - pushl %edx - pushl %ecx - call hyp_invalidate_pte - popl %ecx - popl %edx -#else /* MACH_HYP */ andl $(PTE_INVALID),(%ecx) /* turn off valid bit */ movl %cr3,%eax /* invalidate TLB */ movl %eax,%cr3 -#endif /* MACH_HYP */ 0: /* @@ -1395,18 +1392,8 @@ copyout_retry: subl %eax,%edx /* decrement count */ jg copyout_retry /* restart on next page if not done */ xorl %eax,%eax /* return 0 for success */ - -copyout_ret: - mov %ss,%di /* restore ES to kernel segment */ - mov %di,%es - - popl %edi /* restore registers */ - popl %esi - ret /* and return */ - -copyout_fail: - movl $1,%eax /* return 1 for failure */ - jmp copyout_ret /* pop frame and return */ + jmp copyout_ret +#endif /* !defined(MACH_HYP) && !PAE */ /* * int inst_fetch(int eip, int cs); @@ -1436,14 +1423,14 @@ _inst_fetch_fault: ENTRY(dr6) -#ifdef MACH_XEN +#ifdef MACH_RING1 pushl %ebx movl $6, %ebx call __hyp_get_debugreg popl %ebx -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %db6, %eax -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ ret /* dr<i>(address, type, len, persistence) @@ -1451,67 +1438,67 @@ ENTRY(dr6) ENTRY(dr0) movl S_ARG0, %eax movl %eax,EXT(dr_addr) -#ifdef MACH_XEN +#ifdef MACH_RING1 pushl %ebx movl $0,%ebx movl %eax,%ecx call __hyp_set_debugreg -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %eax, %db0 -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl $0, %ecx jmp 0f ENTRY(dr1) movl S_ARG0, %eax movl %eax,EXT(dr_addr)+1*4 -#ifdef MACH_XEN +#ifdef MACH_RING1 pushl %ebx movl $1,%ebx movl %eax,%ecx call __hyp_set_debugreg -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %eax, %db1 -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl $2, %ecx jmp 0f ENTRY(dr2) movl S_ARG0, %eax movl %eax,EXT(dr_addr)+2*4 -#ifdef MACH_XEN +#ifdef MACH_RING1 pushl %ebx movl $2,%ebx movl %eax,%ecx call __hyp_set_debugreg -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %eax, %db2 -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl $4, %ecx jmp 0f ENTRY(dr3) movl S_ARG0, %eax movl %eax,EXT(dr_addr)+3*4 -#ifdef MACH_XEN +#ifdef MACH_RING1 pushl %ebx movl $3,%ebx movl %eax,%ecx call __hyp_set_debugreg -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %eax, %db3 -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl $6, %ecx 0: pushl %ebp movl %esp, %ebp -#ifdef MACH_XEN +#ifdef MACH_RING1 movl $7,%ebx call __hyp_get_debugreg movl %eax, %edx -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %db7, %edx -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl %edx,EXT(dr_addr)+4*4 andl dr_msk(,%ecx,2),%edx /* clear out new entry */ movl %edx,EXT(dr_addr)+5*4 @@ -1533,14 +1520,14 @@ ENTRY(dr3) shll %cl, %eax orl %eax, %edx -#ifdef MACH_XEN +#ifdef MACH_RING1 movl $7,%ebx movl %edx, %ecx call __hyp_set_debugreg popl %ebx -#else /* MACH_XEN */ +#else /* MACH_RING1 */ movl %edx, %db7 -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ movl %edx,EXT(dr_addr)+7*4 movl %edx, %eax leave diff --git a/i386/i386/locore.h b/i386/i386/locore.h index 57014304..bfd13177 100644 --- a/i386/i386/locore.h +++ b/i386/i386/locore.h @@ -1,6 +1,5 @@ /* - * Header file for printf type functions. - * Copyright (C) 2006 Free Software Foundation. + * Copyright (C) 2006, 2011 Free Software Foundation. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -24,16 +23,39 @@ #include <kern/sched_prim.h> -extern int copyin (const void *userbuf, void *kernelbuf, size_t cn); +/* + * Fault recovery in copyin/copyout routines. + */ +struct recovery { + int fault_addr; + int recover_addr; +}; -extern int copyinmsg (const void *userbuf, void *kernelbuf, size_t cn); +extern struct recovery recover_table[]; +extern struct recovery recover_table_end[]; -extern int copyout (const void *kernelbuf, void *userbuf, size_t cn); +/* + * Recovery from Successful fault in copyout does not + * return directly - it retries the pte check, since + * the 386 ignores write protection in kernel mode. + */ +extern struct recovery retry_table[]; +extern struct recovery retry_table_end[]; -extern int copyoutmsg (const void *kernelbuf, void *userbuf, size_t cn); extern int call_continuation (continuation_t continuation); +extern int discover_x86_cpu_type (void); + +extern int copyin (const void *userbuf, void *kernelbuf, size_t cn); +extern int copyinmsg (const void *userbuf, void *kernelbuf, size_t cn); +extern int copyout (const void *kernelbuf, void *userbuf, size_t cn); +extern int copyoutmsg (const void *kernelbuf, void *userbuf, size_t cn); + +extern int inst_fetch (int eip, int cs); + +extern void cpu_shutdown (void); + extern unsigned int cpu_features[1]; #define CPU_FEATURE_FPU 0 diff --git a/i386/i386/machine_task.c b/i386/i386/machine_task.c index 35b89e02..62b22e3a 100644 --- a/i386/i386/machine_task.c +++ b/i386/i386/machine_task.c @@ -22,15 +22,14 @@ #include <kern/lock.h> #include <mach/mach_types.h> -#include <kern/zalloc.h> -#include <kern/mach_param.h> +#include <kern/slab.h> #include <machine/task.h> #include <machine/io_perm.h> -/* The zone which holds our IO permission bitmaps. */ -zone_t machine_task_iopb_zone; +/* The cache which holds our IO permission bitmaps. */ +struct kmem_cache machine_task_iopb_cache; /* Initialize the machine task module. The function is called once at @@ -38,11 +37,8 @@ zone_t machine_task_iopb_zone; void machine_task_module_init (void) { - machine_task_iopb_zone = zinit (IOPB_BYTES, 0, - TASK_MAX * IOPB_BYTES, - IOPB_BYTES, - ZONE_COLLECTABLE | ZONE_EXHAUSTIBLE, - "i386 machine task iopb"); + kmem_cache_init (&machine_task_iopb_cache, "i386_task_iopb", IOPB_BYTES, 0, + NULL, NULL, NULL, 0); } @@ -62,7 +58,8 @@ void machine_task_terminate (task_t task) { if (task->machine.iopb) - zfree (machine_task_iopb_zone, (vm_offset_t) task->machine.iopb); + kmem_cache_free (&machine_task_iopb_cache, + (vm_offset_t) task->machine.iopb); } @@ -74,7 +71,8 @@ machine_task_collect (task_t task) simple_lock (&task->machine.iopb_lock); if (task->machine.iopb_size == 0 && task->machine.iopb) { - zfree (machine_task_iopb_zone, (vm_offset_t) task->machine.iopb); + kmem_cache_free (&machine_task_iopb_cache, + (vm_offset_t) task->machine.iopb); task->machine.iopb = 0; } simple_unlock (&task->machine.iopb_lock); diff --git a/i386/i386/mp_desc.c b/i386/i386/mp_desc.c index 2fd5ec2d..95f55af2 100644 --- a/i386/i386/mp_desc.c +++ b/i386/i386/mp_desc.c @@ -150,9 +150,9 @@ mp_desc_init(mycpu) * Fix up the entries in the GDT to point to * this LDT and this TSS. */ -#ifdef MACH_HYP +#ifdef MACH_RING1 panic("TODO %s:%d\n",__FILE__,__LINE__); -#else /* MACH_HYP */ +#else /* MACH_RING1 */ fill_descriptor(&mpt->gdt[sel_idx(KERNEL_LDT)], (unsigned)&mpt->ldt, LDTSZ * sizeof(struct real_descriptor) - 1, @@ -165,7 +165,7 @@ mp_desc_init(mycpu) mpt->ktss.tss.ss0 = KERNEL_DS; mpt->ktss.tss.io_bit_map_offset = IOPB_INVAL; mpt->ktss.barrier = 0xFF; -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ return mpt; } diff --git a/i386/i386/pcb.c b/i386/i386/pcb.c index f687db14..7d632a6b 100644 --- a/i386/i386/pcb.c +++ b/i386/i386/pcb.c @@ -36,9 +36,9 @@ #include "vm_param.h" #include <kern/counters.h> #include <kern/debug.h> -#include <kern/mach_param.h> #include <kern/thread.h> #include <kern/sched_prim.h> +#include <kern/slab.h> #include <vm/vm_kern.h> #include <vm/pmap.h> @@ -65,7 +65,7 @@ extern void Thread_continue(); extern void user_ldt_free(); -zone_t pcb_zone; +struct kmem_cache pcb_cache; vm_offset_t kernel_stack[NCPUS]; /* top of active_stack */ @@ -78,7 +78,7 @@ vm_offset_t kernel_stack[NCPUS]; /* top of active_stack */ void stack_attach(thread, stack, continuation) register thread_t thread; register vm_offset_t stack; - void (*continuation)(); + void (*continuation)(thread_t); { counter(if (++c_stacks_current > c_stacks_max) c_stacks_max = c_stacks_current); @@ -93,9 +93,10 @@ void stack_attach(thread, stack, continuation) * This function will not return normally, * so we don`t have to worry about a return address. */ - STACK_IKS(stack)->k_eip = (int) Thread_continue; - STACK_IKS(stack)->k_ebx = (int) continuation; - STACK_IKS(stack)->k_esp = (int) STACK_IEL(stack); + STACK_IKS(stack)->k_eip = (long) Thread_continue; + STACK_IKS(stack)->k_ebx = (long) continuation; + STACK_IKS(stack)->k_esp = (long) STACK_IEL(stack); + STACK_IKS(stack)->k_ebp = (long) 0; /* * Point top of kernel stack to user`s registers. @@ -151,15 +152,16 @@ void switch_ktss(pcb) */ pcb_stack_top = (pcb->iss.efl & EFL_VM) - ? (int) (&pcb->iss + 1) - : (int) (&pcb->iss.v86_segs); + ? (long) (&pcb->iss + 1) + : (long) (&pcb->iss.v86_segs); -#ifdef MACH_XEN +#ifdef MACH_RING1 /* No IO mask here */ - hyp_stack_switch(KERNEL_DS, pcb_stack_top); -#else /* MACH_XEN */ + if (hyp_stack_switch(KERNEL_DS, pcb_stack_top)) + panic("stack_switch"); +#else /* MACH_RING1 */ curr_ktss(mycpu)->tss.esp0 = pcb_stack_top; -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ } { @@ -171,28 +173,28 @@ void switch_ktss(pcb) /* * Use system LDT. */ -#ifdef MACH_HYP +#ifdef MACH_PV_DESCRIPTORS hyp_set_ldt(&ldt, LDTSZ); -#else /* MACH_HYP */ +#else /* MACH_PV_DESCRIPTORS */ set_ldt(KERNEL_LDT); -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ } else { /* * Thread has its own LDT. */ -#ifdef MACH_HYP +#ifdef MACH_PV_DESCRIPTORS hyp_set_ldt(tldt->ldt, (tldt->desc.limit_low|(tldt->desc.limit_high<<16)) / sizeof(struct real_descriptor)); -#else /* MACH_HYP */ +#else /* MACH_PV_DESCRIPTORS */ *gdt_desc_p(mycpu,USER_LDT) = tldt->desc; set_ldt(USER_LDT); -#endif /* MACH_HYP */ +#endif /* MACH_PV_DESCRIPTORS */ } } -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS { int i; for (i=0; i < USER_GDT_SLOTS; i++) { @@ -204,14 +206,14 @@ void switch_ktss(pcb) } } } -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ /* Copy in the per-thread GDT slots. No reloading is necessary because just restoring the segment registers on the way back to user mode reloads the shadow registers from the in-memory GDT. */ memcpy (gdt_desc_p (mycpu, USER_GDT), pcb->ims.user_gdt, sizeof pcb->ims.user_gdt); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* * Load the floating-point context, if necessary. @@ -368,10 +370,8 @@ thread_t switch_context(old, continuation, new) void pcb_module_init() { - pcb_zone = zinit(sizeof(struct pcb), 0, - THREAD_MAX * sizeof(struct pcb), - THREAD_CHUNK * sizeof(struct pcb), - 0, "i386 pcb state"); + kmem_cache_init(&pcb_cache, "pcb", sizeof(struct pcb), 0, + NULL, NULL, NULL, 0); fpu_module_init(); } @@ -381,7 +381,7 @@ void pcb_init(thread) { register pcb_t pcb; - pcb = (pcb_t) zalloc(pcb_zone); + pcb = (pcb_t) kmem_cache_alloc(&pcb_cache); if (pcb == 0) panic("pcb_init"); @@ -421,7 +421,7 @@ void pcb_terminate(thread) fp_free(pcb->ims.ifps); if (pcb->ims.ldt != 0) user_ldt_free(pcb->ims.ldt); - zfree(pcb_zone, (vm_offset_t) pcb); + kmem_cache_free(&pcb_cache, (vm_offset_t) pcb); thread->pcb = 0; } @@ -811,7 +811,7 @@ set_user_regs(stack_base, stack_size, exec_info, arg_size) arg_addr = stack_base + stack_size - arg_size; saved_state = USER_REGS(current_thread()); - saved_state->uesp = (int)arg_addr; + saved_state->uesp = (long)arg_addr; saved_state->eip = exec_info->entry; return (arg_addr); diff --git a/i386/i386/pcb.h b/i386/i386/pcb.h index f8671a2e..21bdfd9c 100644 --- a/i386/i386/pcb.h +++ b/i386/i386/pcb.h @@ -64,7 +64,7 @@ extern void load_context (thread_t new); extern void stack_attach ( thread_t thread, vm_offset_t stack, - void (*continuation)()); + void (*continuation)(thread_t)); extern vm_offset_t stack_detach (thread_t thread); diff --git a/i386/i386/phys.c b/i386/i386/phys.c index 925593b3..ed4a309a 100644 --- a/i386/i386/phys.c +++ b/i386/i386/phys.c @@ -105,9 +105,5 @@ vm_offset_t addr; if ((pte = pmap_pte(kernel_pmap, addr)) == PT_ENTRY_NULL) return 0; - return i386_trunc_page( -#ifdef MACH_PSEUDO_PHYS - ma_to_pa -#endif /* MACH_PSEUDO_PHYS */ - (*pte)) | (addr & INTEL_OFFMASK); + return pte_to_pa(*pte) | (addr & INTEL_OFFMASK); } diff --git a/i386/i386/pic.h b/i386/i386/pic.h index 7a177d86..52f6ec16 100644 --- a/i386/i386/pic.h +++ b/i386/i386/pic.h @@ -179,6 +179,10 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #ifndef __ASSEMBLER__ extern void form_pic_mask (void); extern void picinit (void); +extern int curr_pic_mask; +extern int pic_mask[]; +extern void prtnull(int unit); +extern void intnull(int unit); #endif #endif /* _I386_PIC_H_ */ diff --git a/i386/i386/proc_reg.h b/i386/i386/proc_reg.h index 64d8c43f..f1b2c89f 100644 --- a/i386/i386/proc_reg.h +++ b/i386/i386/proc_reg.h @@ -77,54 +77,54 @@ #include <i386/ldt.h> #endif /* MACH_HYP */ -static inline unsigned +static inline unsigned long get_eflags(void) { - unsigned eflags; - asm("pushfd; popl %0" : "=r" (eflags)); + unsigned long eflags; + asm("pushfd; pop %0" : "=r" (eflags)); return eflags; } static inline void -set_eflags(unsigned eflags) +set_eflags(unsigned long eflags) { - asm volatile("pushl %0; popfd" : : "r" (eflags)); + asm volatile("push %0; popfd" : : "r" (eflags)); } #define get_esp() \ ({ \ - register unsigned int _temp__ asm("esp"); \ + register unsigned long _temp__ asm("esp"); \ _temp__; \ }) #define get_eflags() \ ({ \ - register unsigned int _temp__; \ - asm("pushf; popl %0" : "=r" (_temp__)); \ + register unsigned long _temp__; \ + asm("pushf; pop %0" : "=r" (_temp__)); \ _temp__; \ }) #define get_cr0() \ ({ \ - register unsigned int _temp__; \ + register unsigned long _temp__; \ asm volatile("mov %%cr0, %0" : "=r" (_temp__)); \ _temp__; \ }) #define set_cr0(value) \ ({ \ - register unsigned int _temp__ = (value); \ + register unsigned long _temp__ = (value); \ asm volatile("mov %0, %%cr0" : : "r" (_temp__)); \ }) #define get_cr2() \ ({ \ - register unsigned int _temp__; \ + register unsigned long _temp__; \ asm volatile("mov %%cr2, %0" : "=r" (_temp__)); \ _temp__; \ }) -#ifdef MACH_HYP +#ifdef MACH_PV_PAGETABLES extern unsigned long cr3; #define get_cr3() (cr3) #define set_cr3(value) \ @@ -133,24 +133,24 @@ extern unsigned long cr3; if (!hyp_set_cr3(value)) \ panic("set_cr3"); \ }) -#else /* MACH_HYP */ +#else /* MACH_PV_PAGETABLES */ #define get_cr3() \ ({ \ - register unsigned int _temp__; \ + register unsigned long _temp__; \ asm volatile("mov %%cr3, %0" : "=r" (_temp__)); \ _temp__; \ }) #define set_cr3(value) \ ({ \ - register unsigned int _temp__ = (value); \ - asm volatile("mov %0, %%cr3" : : "r" (_temp__)); \ + register unsigned long _temp__ = (value); \ + asm volatile("mov %0, %%cr3" : : "r" (_temp__) : "memory"); \ }) -#endif /* MACH_HYP */ +#endif /* MACH_PV_PAGETABLES */ #define flush_tlb() set_cr3(get_cr3()) -#ifndef MACH_HYP +#ifndef MACH_PV_PAGETABLES #define invlpg(addr) \ ({ \ asm volatile("invlpg (%0)" : : "r" (addr)); \ @@ -178,34 +178,34 @@ extern unsigned long cr3; : "+r" (var) : "r" (end), \ "q" (LINEAR_DS), "q" (KERNEL_DS), "i" (PAGE_SIZE)); \ }) -#endif /* MACH_HYP */ +#endif /* MACH_PV_PAGETABLES */ #define get_cr4() \ ({ \ - register unsigned int _temp__; \ + register unsigned long _temp__; \ asm volatile("mov %%cr4, %0" : "=r" (_temp__)); \ _temp__; \ }) #define set_cr4(value) \ ({ \ - register unsigned int _temp__ = (value); \ + register unsigned long _temp__ = (value); \ asm volatile("mov %0, %%cr4" : : "r" (_temp__)); \ }) -#ifdef MACH_HYP +#ifdef MACH_RING1 #define set_ts() \ hyp_fpu_taskswitch(1) #define clear_ts() \ hyp_fpu_taskswitch(0) -#else /* MACH_HYP */ +#else /* MACH_RING1 */ #define set_ts() \ set_cr0(get_cr0() | CR0_TS) #define clear_ts() \ asm volatile("clts") -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ #define get_tr() \ ({ \ diff --git a/i386/i386/seg.h b/i386/i386/seg.h index a7f65736..163a1c52 100644 --- a/i386/i386/seg.h +++ b/i386/i386/seg.h @@ -38,11 +38,11 @@ */ /* Note: the value of KERNEL_RING is handled by hand in locore.S */ -#ifdef MACH_HYP +#ifdef MACH_RING1 #define KERNEL_RING 1 -#else /* MACH_HYP */ +#else /* MACH_RING1 */ #define KERNEL_RING 0 -#endif /* MACH_HYP */ +#endif /* MACH_RING1 */ #ifndef __ASSEMBLER__ @@ -161,12 +161,12 @@ MACH_INLINE void fill_descriptor(struct real_descriptor *_desc, unsigned base, unsigned limit, unsigned char access, unsigned char sizebits) { - /* TODO: when !MACH_XEN, setting desc and just memcpy isn't simpler actually */ -#ifdef MACH_XEN + /* TODO: when !MACH_PV_DESCRIPTORS, setting desc and just memcpy isn't simpler actually */ +#ifdef MACH_PV_DESCRIPTORS struct real_descriptor __desc, *desc = &__desc; -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ struct real_descriptor *desc = _desc; -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ if (limit > 0xfffff) { limit >>= 12; @@ -179,10 +179,10 @@ fill_descriptor(struct real_descriptor *_desc, unsigned base, unsigned limit, desc->limit_high = limit >> 16; desc->granularity = sizebits; desc->base_high = base >> 24; -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS if (hyp_do_update_descriptor(kv_to_ma(_desc), *(uint64_t*)desc)) panic("couldn't update descriptor(%p to %08lx%08lx)\n", (vm_offset_t) kv_to_ma(_desc), *(((unsigned long*)desc)+1), *(unsigned long *)desc); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ } /* Fill a gate with particular values. */ diff --git a/i386/i386/spl.S b/i386/i386/spl.S index f1d4b45f..3c075092 100644 --- a/i386/i386/spl.S +++ b/i386/i386/spl.S @@ -143,7 +143,7 @@ ENTRY(spl7) SETIPL(SPL7) ENTRY(splx) - movl 4(%esp),%edx /* get ipl */ + movl S_ARG0,%edx /* get ipl */ testl %edx,%edx /* spl0? */ jz EXT(spl0) /* yes, handle specially */ cmpl EXT(curr_ipl),%edx /* same ipl as current? */ @@ -160,7 +160,7 @@ ENTRY(splx) .align TEXT_ALIGN .globl splx_cli splx_cli: - movl 4(%esp),%edx /* get ipl */ + movl S_ARG0,%edx /* get ipl */ cli /* disable interrupts */ testl %edx,%edx /* spl0? */ jnz 2f /* no, skip */ diff --git a/i386/i386/task.h b/i386/i386/task.h index ca8de044..0060ad4e 100644 --- a/i386/i386/task.h +++ b/i386/i386/task.h @@ -24,7 +24,7 @@ #define _I386_TASK_H_ #include <kern/kern_types.h> -#include <kern/zalloc.h> +#include <kern/slab.h> /* The machine specific data of a task. */ struct machine_task @@ -41,7 +41,7 @@ struct machine_task typedef struct machine_task machine_task_t; -extern zone_t machine_task_iopb_zone; +extern struct kmem_cache machine_task_iopb_cache; /* Initialize the machine task module. The function is called once at start up by task_init in kern/task.c. */ diff --git a/i386/i386/thread.h b/i386/i386/thread.h index f2ae8bf0..eddd25c2 100644 --- a/i386/i386/thread.h +++ b/i386/i386/thread.h @@ -51,31 +51,31 @@ */ struct i386_saved_state { - unsigned int gs; - unsigned int fs; - unsigned int es; - unsigned int ds; - unsigned int edi; - unsigned int esi; - unsigned int ebp; - unsigned int cr2; /* kernel esp stored by pusha - + unsigned long gs; + unsigned long fs; + unsigned long es; + unsigned long ds; + unsigned long edi; + unsigned long esi; + unsigned long ebp; + unsigned long cr2; /* kernel esp stored by pusha - we save cr2 here later */ - unsigned int ebx; - unsigned int edx; - unsigned int ecx; - unsigned int eax; - unsigned int trapno; - unsigned int err; - unsigned int eip; - unsigned int cs; - unsigned int efl; - unsigned int uesp; - unsigned int ss; + unsigned long ebx; + unsigned long edx; + unsigned long ecx; + unsigned long eax; + unsigned long trapno; + unsigned long err; + unsigned long eip; + unsigned long cs; + unsigned long efl; + unsigned long uesp; + unsigned long ss; struct v86_segs { - unsigned int v86_es; /* virtual 8086 segment registers */ - unsigned int v86_ds; - unsigned int v86_fs; - unsigned int v86_gs; + unsigned long v86_es; /* virtual 8086 segment registers */ + unsigned long v86_ds; + unsigned long v86_fs; + unsigned long v86_gs; } v86_segs; }; @@ -97,12 +97,12 @@ struct i386_exception_link { */ struct i386_kernel_state { - int k_ebx; /* kernel context */ - int k_esp; - int k_ebp; - int k_edi; - int k_esi; - int k_eip; + long k_ebx; /* kernel context */ + long k_esp; + long k_ebp; + long k_edi; + long k_esi; + long k_eip; }; /* @@ -144,16 +144,16 @@ struct v86_assist_state { */ struct i386_interrupt_state { - int gs; - int fs; - int es; - int ds; - int edx; - int ecx; - int eax; - int eip; - int cs; - int efl; + long gs; + long fs; + long es; + long ds; + long edx; + long ecx; + long eax; + long eip; + long cs; + long efl; }; /* diff --git a/i386/i386/trap.c b/i386/i386/trap.c index 22dd4914..c6aab488 100644 --- a/i386/i386/trap.c +++ b/i386/i386/trap.c @@ -33,6 +33,7 @@ #include <mach/machine/eflags.h> #include <i386/trap.h> #include <i386/fpu.h> +#include <i386/locore.h> #include <i386/model_dep.h> #include <intel/read_fault.h> #include <machine/machspl.h> /* for spl_t */ @@ -122,25 +123,6 @@ user_page_fault_continue(kr) /*NOTREACHED*/ } -/* - * Fault recovery in copyin/copyout routines. - */ -struct recovery { - int fault_addr; - int recover_addr; -}; - -extern struct recovery recover_table[]; -extern struct recovery recover_table_end[]; - -/* - * Recovery from Successful fault in copyout does not - * return directly - it retries the pte check, since - * the 386 ignores write protection in kernel mode. - */ -extern struct recovery retry_table[]; -extern struct recovery retry_table_end[]; - static char *trap_type[] = { "Divide error", @@ -235,10 +217,10 @@ dump_ss(regs); printf("now %08x\n", subcode); #endif if (trunc_page(subcode) == 0 - || (subcode >= (int)_start - && subcode < (int)etext)) { + || (subcode >= (long)_start + && subcode < (long)etext)) { printf("Kernel page fault at address 0x%x, " - "eip = 0x%x\n", + "eip = 0x%lx\n", subcode, regs->eip); goto badtrap; } @@ -295,6 +277,7 @@ dump_ss(regs); */ register struct recovery *rp; + /* Linear searching; but the list is small enough. */ for (rp = retry_table; rp < retry_table_end; rp++) { if (regs->eip == rp->fault_addr) { regs->eip = rp->recover_addr; @@ -311,6 +294,7 @@ dump_ss(regs); { register struct recovery *rp; + /* Linear searching; but the list is small enough. */ for (rp = recover_table; rp < recover_table_end; rp++) { @@ -344,7 +328,7 @@ dump_ss(regs); printf("%s trap", trap_type[type]); else printf("trap %d", type); - printf(", eip 0x%x\n", regs->eip); + printf(", eip 0x%lx\n", regs->eip); #if MACH_TTD if (kttd_enabled && kttd_trap(type, code, regs)) return; @@ -378,7 +362,7 @@ int user_trap(regs) if ((vm_offset_t)thread < phys_last_addr) { printf("user_trap: bad thread pointer 0x%p\n", thread); - printf("trap type %d, code 0x%x, va 0x%x, eip 0x%x\n", + printf("trap type %ld, code 0x%lx, va 0x%lx, eip 0x%lx\n", regs->trapno, regs->err, regs->cr2, regs->eip); asm volatile ("1: hlt; jmp 1b"); } @@ -526,7 +510,7 @@ printf("user trap %d error %d sub %08x\n", type, code, subcode); /*NOTREACHED*/ break; -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES case 15: { static unsigned count = 0; @@ -556,7 +540,7 @@ printf("user trap %d error %d sub %08x\n", type, code, subcode); return 0; #endif /* MACH_KDB */ splhigh(); - printf("user trap, type %d, code = %x\n", + printf("user trap, type %d, code = %lx\n", type, regs->err); dump_ss(regs); panic("trap"); @@ -601,7 +585,7 @@ i386_astintr() int mycpu = cpu_number(); (void) splsched(); /* block interrupts to check reasons */ -#ifndef MACH_XEN +#ifndef MACH_RING1 if (need_ast[mycpu] & AST_I386_FP) { /* * AST was for delayed floating-point exception - @@ -614,7 +598,7 @@ i386_astintr() fpastintr(); } else -#endif /* MACH_XEN */ +#endif /* MACH_RING1 */ { /* * Not an FPU trap. Handle the AST. diff --git a/i386/i386/user_ldt.c b/i386/i386/user_ldt.c index dfe6b1e6..74c10a4c 100644 --- a/i386/i386/user_ldt.c +++ b/i386/i386/user_ldt.c @@ -196,17 +196,17 @@ i386_set_ldt(thread, first_selector, desc_list, count, desc_list_inline) if (new_ldt == 0) { simple_unlock(&pcb->lock); -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS /* LDT needs to be aligned on a page */ vm_offset_t alloc = kalloc(ldt_size_needed + PAGE_SIZE + offsetof(struct user_ldt, ldt)); new_ldt = (user_ldt_t) (round_page((alloc + offsetof(struct user_ldt, ldt))) - offsetof(struct user_ldt, ldt)); new_ldt->alloc = alloc; -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ new_ldt = (user_ldt_t) kalloc(ldt_size_needed + sizeof(struct real_descriptor)); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* * Build a descriptor that describes the * LDT itself @@ -272,19 +272,21 @@ i386_set_ldt(thread, first_selector, desc_list, count, desc_list_inline) simple_unlock(&pcb->lock); if (new_ldt) -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS { int i; +#ifdef MACH_PV_PAGETABLES for (i=0; i<(new_ldt->desc.limit_low + 1)/sizeof(struct real_descriptor); i+=PAGE_SIZE/sizeof(struct real_descriptor)) pmap_set_page_readwrite(&new_ldt->ldt[i]); +#endif /* MACH_PV_PAGETABLES*/ kfree(new_ldt->alloc, new_ldt->desc.limit_low + 1 + PAGE_SIZE + offsetof(struct user_ldt, ldt)); } -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ kfree((vm_offset_t)new_ldt, new_ldt->desc.limit_low + 1 + sizeof(struct real_descriptor)); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ /* * Free the descriptor list, if it was @@ -417,17 +419,19 @@ void user_ldt_free(user_ldt) user_ldt_t user_ldt; { -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS int i; +#ifdef MACH_PV_PAGETABLES for (i=0; i<(user_ldt->desc.limit_low + 1)/sizeof(struct real_descriptor); i+=PAGE_SIZE/sizeof(struct real_descriptor)) pmap_set_page_readwrite(&user_ldt->ldt[i]); +#endif /* MACH_PV_PAGETABLES */ kfree(user_ldt->alloc, user_ldt->desc.limit_low + 1 + PAGE_SIZE + offsetof(struct user_ldt, ldt)); -#else /* MACH_XEN */ +#else /* MACH_PV_DESCRIPTORS */ kfree((vm_offset_t)user_ldt, user_ldt->desc.limit_low + 1 + sizeof(struct real_descriptor)); -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ } diff --git a/i386/i386/user_ldt.h b/i386/i386/user_ldt.h index 8d16ed87..6c6c858e 100644 --- a/i386/i386/user_ldt.h +++ b/i386/i386/user_ldt.h @@ -36,9 +36,9 @@ #include <i386/seg.h> struct user_ldt { -#ifdef MACH_XEN +#ifdef MACH_PV_DESCRIPTORS vm_offset_t alloc; /* allocation before alignment */ -#endif /* MACH_XEN */ +#endif /* MACH_PV_DESCRIPTORS */ struct real_descriptor desc; /* descriptor for self */ struct real_descriptor ldt[1]; /* descriptor table (variable) */ }; diff --git a/i386/i386/vm_param.h b/i386/i386/vm_param.h index 95df6044..ffd91d65 100644 --- a/i386/i386/vm_param.h +++ b/i386/i386/vm_param.h @@ -25,25 +25,41 @@ /* XXX use xu/vm_param.h */ #include <mach/vm_param.h> +#ifdef MACH_PV_PAGETABLES #include <xen/public/xen.h> +#endif + +/* The kernel address space is usually 1GB, usually starting at virtual address 0. */ +/* This can be changed freely to separate kernel addresses from user addresses + * for better trace support in kdb; the _START symbol has to be offset by the + * same amount. */ +#define VM_MIN_KERNEL_ADDRESS 0xC0000000UL -/* The kernel address space is 1GB, starting at virtual address 0. */ #ifdef MACH_XEN -#define VM_MIN_KERNEL_ADDRESS 0x20000000UL +/* PV kernels can be loaded directly to the target virtual address */ +#define INIT_VM_MIN_KERNEL_ADDRESS VM_MIN_KERNEL_ADDRESS #else /* MACH_XEN */ -#define VM_MIN_KERNEL_ADDRESS 0x00000000UL +/* This must remain 0 */ +#define INIT_VM_MIN_KERNEL_ADDRESS 0x00000000UL #endif /* MACH_XEN */ -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES #if PAE #define HYP_VIRT_START HYPERVISOR_VIRT_START_PAE #else /* PAE */ #define HYP_VIRT_START HYPERVISOR_VIRT_START_NONPAE #endif /* PAE */ #define VM_MAX_KERNEL_ADDRESS (HYP_VIRT_START - LINEAR_MIN_KERNEL_ADDRESS + VM_MIN_KERNEL_ADDRESS) -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ #define VM_MAX_KERNEL_ADDRESS (LINEAR_MAX_KERNEL_ADDRESS - LINEAR_MIN_KERNEL_ADDRESS + VM_MIN_KERNEL_ADDRESS) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ + +/* Reserve mapping room for kmem. */ +#ifdef MACH_XEN +#define VM_KERNEL_MAP_SIZE (224 * 1024 * 1024) +#else +#define VM_KERNEL_MAP_SIZE (192 * 1024 * 1024) +#endif /* The kernel virtual address space is actually located at high linear addresses. @@ -51,14 +67,14 @@ #define LINEAR_MIN_KERNEL_ADDRESS (VM_MAX_ADDRESS) #define LINEAR_MAX_KERNEL_ADDRESS (0xffffffffUL) -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES /* need room for mmu updates (2*8bytes) */ #define KERNEL_STACK_SIZE (4*I386_PGBYTES) #define INTSTACK_SIZE (4*I386_PGBYTES) -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ #define KERNEL_STACK_SIZE (1*I386_PGBYTES) #define INTSTACK_SIZE (1*I386_PGBYTES) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ /* interrupt stack size */ /* diff --git a/i386/i386/xen.h b/i386/i386/xen.h index 731e5a2f..5bdaf0b8 100644 --- a/i386/i386/xen.h +++ b/i386/i386/xen.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006, 2009, 2010, 2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -25,6 +25,7 @@ #include <mach/machine/vm_types.h> #include <mach/vm_param.h> #include <mach/inline.h> +#include <mach/xen.h> #include <machine/vm_param.h> #include <intel/pmap.h> #include <kern/debug.h> @@ -37,7 +38,7 @@ #define wmb() mb() MACH_INLINE unsigned long xchgl(volatile unsigned long *ptr, unsigned long x) { - __asm__ __volatile__("xchgl %0, %1" + __asm__ __volatile__("xchg %0, %1" : "=r" (x) : "m" (*(ptr)), "0" (x): "memory"); return x; @@ -151,6 +152,7 @@ MACH_INLINE type hyp_##name(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type _hypcall1(long, set_trap_table, vm_offset_t /* struct trap_info * */, traps); +#ifdef MACH_PV_PAGETABLES _hypcall4(int, mmu_update, vm_offset_t /* struct mmu_update * */, req, int, count, vm_offset_t /* int * */, success_count, domid_t, domid) MACH_INLINE int hyp_mmu_update_pte(pt_entry_t pte, pt_entry_t val) { @@ -169,6 +171,7 @@ MACH_INLINE int hyp_mmu_update_pte(pt_entry_t pte, pt_entry_t val) #define hyp_mmu_update_la(la, val) hyp_mmu_update_pte( \ (kernel_pmap->dirbase[lin2pdenum((vm_offset_t)(la))] & INTEL_PTE_PFN) \ + ptenum((vm_offset_t)(la)) * sizeof(pt_entry_t), val) +#endif _hypcall2(long, set_gdt, vm_offset_t /* unsigned long * */, frame_list, unsigned int, entries) @@ -215,6 +218,7 @@ MACH_INLINE void hyp_free_page(unsigned long pfn, void *va) /* save mfn */ unsigned long mfn = pfn_to_mfn(pfn); +#ifdef MACH_PV_PAGETABLES /* remove from mappings */ if (hyp_do_update_va_mapping(kvtolin(va), 0, UVMF_INVLPG|UVMF_ALL)) panic("couldn't clear page %d at %p\n", pfn, va); @@ -223,11 +227,13 @@ MACH_INLINE void hyp_free_page(unsigned long pfn, void *va) /* drop machine page */ mfn_list[pfn] = ~0; #endif /* MACH_PSEUDO_PHYS */ +#endif /* and free from Xen */ hyp_free_mfn(mfn); } +#ifdef MACH_PV_PAGETABLES _hypcall4(int, mmuext_op, vm_offset_t /* struct mmuext_op * */, op, int, count, vm_offset_t /* int * */, success_count, domid_t, domid); MACH_INLINE int hyp_mmuext_op_void(unsigned int cmd) { @@ -273,6 +279,7 @@ MACH_INLINE void hyp_invlpg(vm_offset_t lin) { if (n < 1) panic("couldn't invlpg\n"); } +#endif _hypcall2(long, set_timer_op, unsigned long, absolute_lo, unsigned long, absolute_hi); #define hyp_do_set_timer_op(absolute_nsec) ({ \ diff --git a/i386/i386/zalloc.h b/i386/i386/zalloc.h deleted file mode 100644 index bf7cf6b2..00000000 --- a/i386/i386/zalloc.h +++ /dev/null @@ -1,29 +0,0 @@ -/* - * Copyright (c) 1996-1994 The University of Utah and - * the Computer Systems Laboratory (CSL). All rights reserved. - * - * Permission to use, copy, modify and distribute this software is hereby - * granted provided that (1) source code retains these copyright, permission, - * and disclaimer notices, and (2) redistributions including binaries - * reproduce the notices in supporting documentation, and (3) all advertising - * materials mentioning features or use of this software display the following - * acknowledgement: ``This product includes software developed by the - * Computer Systems Laboratory at the University of Utah.'' - * - * THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF THIS SOFTWARE IN ITS "AS - * IS" CONDITION. THE UNIVERSITY OF UTAH AND CSL DISCLAIM ANY LIABILITY OF - * ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. - * - * CSL requests users of this software to return to csl-dist@cs.utah.edu any - * improvements that they make and grant CSL redistribution rights. - * - * Utah $Hdr: zalloc.h 1.4 94/12/16$ - * Author: Bryan Ford - */ - -#ifndef _I386_ZALLOC_H_ -#define _I386_ZALLOC_H_ - -#include <kern/zalloc.h> - -#endif /* _I386_ZALLOC_H_ */ diff --git a/i386/i386at/autoconf.c b/i386/i386at/autoconf.c index 7713dec6..93c71412 100644 --- a/i386/i386at/autoconf.c +++ b/i386/i386at/autoconf.c @@ -25,16 +25,7 @@ */ #include <kern/printf.h> -#ifdef MACH_KERNEL #include <mach/std_types.h> -#else /* MACH_KERNEL */ -#include <cpus.h> -#include <platforms.h> -#include <generic.h> -#include <sys/param.h> -#include <mach/machine.h> -#include <machine/cpu.h> -#endif /* MACH_KERNEL */ #include <i386/pic.h> #include <i386/ipl.h> #include <chips/busses.h> @@ -144,7 +135,7 @@ void take_dev_irq( printf("The device below will clobber IRQ %d.\n", pic); printf("You have two devices at the same IRQ.\n"); printf("This won't work. Reconfigure your hardware and try again.\n"); - printf("%s%d: port = %x, spl = %d, pic = %d.\n", + printf("%s%d: port = %lx, spl = %ld, pic = %d.\n", dev->name, dev->unit, dev->address, dev->sysdep, dev->sysdep1); while (1); diff --git a/i386/i386at/boothdr.S b/i386/i386at/boothdr.S index 45cd599f..567851e4 100644 --- a/i386/i386at/boothdr.S +++ b/i386/i386at/boothdr.S @@ -39,6 +39,19 @@ boot_hdr: #endif /* __ELF__ */ boot_entry: + /* use segmentation to offset ourself. */ + lgdt boot_gdt_descr - KERNELBASE + ljmp $8,$0f +0: + movw $0,%ax + movw %ax,%ds + movw %ax,%es + movw %ax,%fs + movw %ax,%gs + movw $16,%ax + movw %ax,%ds + movw %ax,%es + movw %ax,%ss /* Switch to our own interrupt stack. */ movl $_intstack+INTSTACK_SIZE,%esp @@ -80,3 +93,27 @@ iplt_done: .comm _intstack,INTSTACK_SIZE +.align 16 + .word 0 +boot_gdt_descr: + .word 3*8+7 + .long boot_gdt - KERNELBASE +.align 16 +boot_gdt: + /* 0 */ + .quad 0 + /* boot CS = 8 */ + .word 0xffff + .word (-KERNELBASE) & 0xffff + .byte ((-KERNELBASE) >> 16) & 0xff + .byte 0x9a + .byte 0xcf + .byte ((-KERNELBASE) >> 24) & 0xff + /* boot DS = 8 */ + .word 0xffff + .word (-KERNELBASE) & 0xffff + .byte ((-KERNELBASE) >> 16) & 0xff + .byte 0x92 + .byte 0xcf + .byte ((-KERNELBASE) >> 24) & 0xff + diff --git a/i386/i386at/com.c b/i386/i386at/com.c index 325e0991..93c3faaa 100644 --- a/i386/i386at/com.c +++ b/i386/i386at/com.c @@ -35,7 +35,7 @@ #include <kern/mach_clock.h> #include <sys/time.h> #include <device/conf.h> -#include <device/errno.h> +#include <device/device_types.h> #include <device/tty.h> #include <device/io_req.h> @@ -70,10 +70,6 @@ int comtimer_state[NCOM]; #define RCBAUD B9600 static int rcline = -1; static struct bus_device *comcndev; -int comcnprobe(struct consdev *cp); -int comcninit(struct consdev *cp); -int comcngetc(dev_t dev, int wait); -int comcnputc(dev_t dev, int c); /* XX */ extern char *kernel_cmdline; @@ -215,6 +211,8 @@ comcnprobe(struct consdev *cp) cp->cn_dev = makedev(maj, unit); cp->cn_pri = pri; + + return 0; } @@ -231,7 +229,7 @@ comattach(struct bus_device *dev) u_short addr = dev->address; take_dev_irq(dev); - printf(", port = %x, spl = %d, pic = %d. (DOS COM%d)", + printf(", port = %lx, spl = %ld, pic = %d. (DOS COM%d)", dev->address, dev->sysdep, dev->sysdep1, unit+1); /* comcarrier[unit] = addr->flags;*/ @@ -267,7 +265,7 @@ comcninit(struct consdev *cp) outb(LINE_CTL(addr), iDLAB); outb(BAUD_LSB(addr), divisorreg[RCBAUD] & 0xff); outb(BAUD_MSB(addr), divisorreg[RCBAUD] >>8); - outb(LINE_CTL(addr), i7BITS|iPEN); + outb(LINE_CTL(addr), i8BITS); outb(INTR_ENAB(addr), 0); outb(MODEM_CTL(addr), iDTR|iRTS|iOUT2); @@ -287,6 +285,7 @@ comcninit(struct consdev *cp) } } + return 0; } /* @@ -336,13 +335,13 @@ io_return_t comopen( io_return_t result; if (unit >= NCOM) - return ENXIO; /* no such device */ + return D_NO_SUCH_DEVICE; /* no such device */ if ((isai = cominfo[unit]) == 0 || isai->alive == 0) { /* * Try to probe it again */ if (!com_reprobe(unit)) - return ENXIO; + return D_NO_SUCH_DEVICE; } tp = &com_tty[unit]; @@ -448,7 +447,7 @@ comgetstat(dev, flavor, data, count) dev_t dev; int flavor; int *data; /* pointer to OUT array */ -unsigned int *count; /* out */ +natural_t *count; /* out */ { io_return_t result = D_SUCCESS; int unit = minor(dev); @@ -471,7 +470,7 @@ comsetstat(dev, flavor, data, count) dev_t dev; int flavor; int * data; -unsigned int count; +natural_t count; { io_return_t result = D_SUCCESS; int unit = minor(dev); @@ -622,7 +621,9 @@ comstart(tp) struct tty *tp; { char nch; +#if 0 int i; +#endif if (tp->t_state & (TS_TIMEOUT|TS_TTSTOP|TS_BUSY)) { comst_1++; @@ -656,7 +657,7 @@ comst_4++; #else nch = getc(&tp->t_outq); if ((nch & 0200) && ((tp->t_flags & LITOUT) == 0)) { - timeout(ttrstrt, (char *)tp, (nch & 0x7f) + 6); + timeout((timer_func_t *)ttrstrt, (char *)tp, (nch & 0x7f) + 6); tp->t_state |= TS_TIMEOUT; comst_4++; return; @@ -884,6 +885,8 @@ comcnputc(dev_t dev, int c) if (c == '\n') comcnputc(dev, '\r'); outb(addr, c); + + return 0; } int diff --git a/i386/i386at/com.h b/i386/i386at/com.h index e53e9482..49f23eec 100644 --- a/i386/i386at/com.h +++ b/i386/i386at/com.h @@ -27,6 +27,7 @@ #define _COM_H_ #include <mach/std_types.h> +#include <device/cons.h> /* * Set receive modem state from modem status register. @@ -42,4 +43,9 @@ extern void commodem_intr(int unit, int stat); extern int comgetc(int unit); +extern int comcnprobe(struct consdev *cp); +extern int comcninit(struct consdev *cp); +extern int comcngetc(dev_t dev, int wait); +extern int comcnputc(dev_t dev, int c); + #endif /* _COM_H_ */ diff --git a/i386/i386at/cons_conf.c b/i386/i386at/cons_conf.c index 8a4b37c5..cf42bb63 100644 --- a/i386/i386at/cons_conf.c +++ b/i386/i386at/cons_conf.c @@ -31,11 +31,11 @@ #include <device/cons.h> #ifdef MACH_HYP -extern int hypcnprobe(), hypcninit(), hypcngetc(), hypcnputc(); +#include <xen/console.h> #else /* MACH_HYP */ -extern int kdcnprobe(), kdcninit(), kdcngetc(), kdcnputc(); +#include "kd.h" #if NCOM > 0 -extern int comcnprobe(), comcninit(), comcngetc(), comcnputc(); +#include "com.h" #endif #endif /* MACH_HYP */ diff --git a/i386/i386at/kd.c b/i386/i386at/kd.c index e4ac8325..c9778629 100644 --- a/i386/i386at/kd.c +++ b/i386/i386at/kd.c @@ -85,6 +85,7 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #include <device/io_req.h> #include <device/buf.h> /* for struct uio (!) */ #include <vm/vm_kern.h> +#include <i386/locore.h> #include <i386/loose_ends.h> #include <i386/vm_param.h> #include <i386/machspl.h> @@ -112,10 +113,6 @@ static void charput(), charmvup(), charmvdown(), charclear(), charsetcursor(); static void kd_noopreset(); boolean_t kdcheckmagic(); -int kdcnprobe(struct consdev *cp); -int kdcninit(struct consdev *cp); -int kdcngetc(dev_t dev, int wait); -void kdcnputc(dev_t dev, int c); int do_modifier (int, Scancode, boolean_t); /* @@ -595,7 +592,7 @@ io_return_t kdgetstat(dev, flavor, data, count) dev_t dev; int flavor; int * data; /* pointer to OUT array */ - unsigned int *count; /* OUT */ + natural_t *count; /* OUT */ { io_return_t result; @@ -625,7 +622,7 @@ io_return_t kdsetstat(dev, flavor, data, count) dev_t dev; int flavor; int * data; - unsigned int count; + natural_t count; { io_return_t result; @@ -735,9 +732,8 @@ int flags; /* flags set for console */ */ /*ARGSUSED*/ void -kdintr(vec, regs) +kdintr(vec) int vec; -int regs; { struct tty *tp; unsigned char c; @@ -802,7 +798,7 @@ int regs; goto done; } else if (kd_kbd_mouse && kd_kbd_magic(scancode)) { goto done; - } else if (kdcheckmagic(scancode, ®s)) { + } else if (kdcheckmagic(scancode)) { goto done; } else if (kb_mode == KB_EVENT) { kd_enqsc(scancode); @@ -988,9 +984,8 @@ boolean_t up; * are still held down. */ boolean_t -kdcheckmagic(scancode, regs) +kdcheckmagic(scancode) Scancode scancode; -int *regs; { static int magic_state = KS_NORMAL; /* like kd_state */ boolean_t up = FALSE; @@ -2969,16 +2964,18 @@ kdcngetc(dev_t dev, int wait) return kdcnmaygetc(); } -void +int kdcnputc(dev_t dev, int c) { if (!kd_initialized) - return; + return -1; /* Note that tab is handled in kd_putc */ if (c == '\n') kd_putc('\r'); kd_putc(c); + + return 0; } /* diff --git a/i386/i386at/kd.h b/i386/i386at/kd.h index 98c8bfee..1d53538b 100644 --- a/i386/i386at/kd.h +++ b/i386/i386at/kd.h @@ -75,6 +75,7 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #include <mach/boolean.h> #include <sys/types.h> #include <sys/time.h> +#include <device/cons.h> /* @@ -729,10 +730,16 @@ extern void kd_handle_ack (void); extern int kd_kbd_magic (int); extern int kdstate2idx (int, boolean_t); extern void kd_parserest (u_char *); +extern int kdcnprobe(struct consdev *cp); +extern int kdcninit(struct consdev *cp); +extern int kdcngetc(dev_t dev, int wait); extern int kdcnmaygetc (void); +extern int kdcnputc(dev_t dev, int c); extern void kd_slmwd (void *start, int count, int value); extern void kd_slmscu (void *from, void *to, int count); extern void kd_slmscd (void *from, void *to, int count); +extern void kdintr(int vec); + #endif /* _KD_H_ */ diff --git a/i386/i386at/kd_event.c b/i386/i386at/kd_event.c index 3983a118..4d2ea008 100644 --- a/i386/i386at/kd_event.c +++ b/i386/i386at/kd_event.c @@ -59,20 +59,9 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #include <kern/printf.h> #include <string.h> -#ifdef MACH_KERNEL #include <device/ds_routines.h> -#include <device/errno.h> +#include <device/device_types.h> #include <device/io_req.h> -#else /* MACH_KERNEL */ -#include <sys/file.h> -#include <sys/errno.h> -#include <kern/thread.h> -#include <sys/user.h> -#include <sys/proc.h> -#include <sys/kernel.h> -#include <sys/ioctl.h> -#include <sys/tty.h> -#endif /* MACH_KERNEL */ #include <i386/machspl.h> #include <i386/pio.h> #include <i386at/kd.h> @@ -92,24 +81,12 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ kd_event_queue kbd_queue; /* queue of keyboard events */ -#ifdef MACH_KERNEL queue_head_t kbd_read_queue = { &kbd_read_queue, &kbd_read_queue }; -#else /* MACH_KERNEL */ -struct proc *kbd_sel = 0; /* selecting process, if any */ -short kbdpgrp = 0; /* process group leader when dev is open */ - -int kbdflag = 0; -#define KBD_COLL 1 /* select collision */ -#define KBD_ASYNC 2 /* user wants asynch notification */ -#define KBD_NBIO 4 /* user wants non-blocking I/O */ -#endif /* MACH_KERNEL */ void kbd_enqueue(); -#ifdef MACH_KERNEL io_return_t X_kdb_enter_init(); io_return_t X_kdb_exit_init(); -#endif /* MACH_KERNEL */ static boolean_t initialized = FALSE; @@ -147,14 +124,6 @@ kbdopen(dev, flags) splx(o_pri); kbdinit(); -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - if (flags & FWRITE) - return(ENODEV); - - if (kbdpgrp == 0) - kbdpgrp = u.u_procp->p_pgrp; -#endif /* MACH_KERNEL */ return(0); } @@ -173,18 +142,11 @@ kbdclose(dev, flags) spl_t s = SPLKD(); kb_mode = KB_ASCII; -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - kbdpgrp = 0; - kbdflag = 0; - kbd_sel = 0; -#endif /* MACH_KERNEL */ kdq_reset(&kbd_queue); splx(s); } -#ifdef MACH_KERNEL io_return_t kbdgetstat(dev, flavor, data, count) dev_t dev; int flavor; @@ -234,91 +196,12 @@ io_return_t kbdsetstat(dev, flavor, data, count) return (D_SUCCESS); } -#else /* MACH_KERNEL */ -/* - * kbdioctl - handling for asynch & non-blocking I/O. - */ - -/*ARGSUSED*/ -int -kbdioctl(dev, cmd, data, flag) - dev_t dev; - int cmd; - caddr_t data; - int flag; -{ - spl_t s = SPLKD(); - int err = 0; - - switch (cmd) { - case KDSKBDMODE: - kb_mode = *(int *)data; - /* XXX - what to do about unread events? */ - /* XXX - should check that "data" contains an OK value */ - break; - case KDGKBDTYPE: - *(int *)data = KB_VANILLAKB; - break; - case K_X_KDB_ENTER: - X_kdb_enter_init((struct X_kdb *) data); - break; - case K_X_KDB_EXIT: - X_kdb_exit_init( (struct X_kdb *) data); - break; - case FIONBIO: - if (*(int *)data) - kbdflag |= KBD_NBIO; - else - kbdflag &= ~KBD_NBIO; - break; - case FIOASYNC: - if (*(int *)data) - kbdflag |= KBD_ASYNC; - else - kbdflag &= ~KBD_ASYNC; - break; - default: - err = ENOTTY; - break; - } - - splx(s); - return(err); -} - - -/* - * kbdselect - */ - -/*ARGSUSED*/ -int -kbdselect(dev, rw) -{ - spl_t s = SPLKD(); - - if (!kdq_empty(&kbd_queue)) { - splx(s); - return(1); - } - - if (kbd_sel) - kbdflag |= KBD_COLL; - else - kbd_sel = (struct proc *)current_thread(); - /* eeeyuck */ - - splx(s); - return(0); -} -#endif /* MACH_KERNEL */ /* * kbdread - dequeue and return any queued events. */ -#ifdef MACH_KERNEL boolean_t kbd_read_done(); /* forward */ int @@ -391,44 +274,6 @@ boolean_t kbd_read_done(ior) return (TRUE); } -#else /* MACH_KERNEL */ -/*ARGSUSED*/ -kbdread(dev, uio) - dev_t dev; - struct uio *uio; -{ - int s = SPLKD(); - int err = 0; - kd_event *ev; - int i; - char *cp; - - if (kdq_empty(&kbd_queue)) - if (kbdflag & KBD_NBIO) { - err = EWOULDBLOCK; - goto done; - } else - while (kdq_empty(&kbd_queue)) { - splx(s); - sleep((caddr_t)&kbd_queue, TTIPRI); - s = SPLKD(); - } - - while (!kdq_empty(&kbd_queue) && uio->uio_resid >= sizeof(kd_event)) { - ev = kdq_get(&kbd_queue); - for (cp = (char *)ev, i = 0; i < sizeof(kd_event); - ++i, ++cp) { - err = ureadc(*cp, uio); - if (err) - goto done; - } - } - -done: - splx(s); - return(err); -} -#endif /* MACH_KERNEL */ /* @@ -462,22 +307,11 @@ kbd_enqueue(ev) else kdq_put(&kbd_queue, ev); -#ifdef MACH_KERNEL { register io_req_t ior; while ((ior = (io_req_t)dequeue_head(&kbd_read_queue)) != 0) iodone(ior); } -#else /* MACH_KERNEL */ - if (kbd_sel) { - selwakeup(kbd_sel, kbdflag & KBD_COLL); - kbd_sel = 0; - kbdflag &= ~KBD_COLL; - } - if (kbdflag & KBD_ASYNC) - gsignal(kbdpgrp, SIGIO); - wakeup((caddr_t)&kbd_queue); -#endif /* MACH_KERNEL */ } u_int X_kdb_enter_str[512], X_kdb_exit_str[512]; @@ -538,7 +372,6 @@ register u_int *u_ip, *endp; kdb_in_out(u_ip); } -#ifdef MACH_KERNEL io_return_t X_kdb_enter_init(data, count) u_int *data; @@ -564,28 +397,3 @@ X_kdb_exit_init(data, count) X_kdb_exit_len = count; return D_SUCCESS; } -#else /* MACH_KERNEL */ -void -X_kdb_enter_init(kp) -struct X_kdb *kp; -{ - if (kp->size > sizeof X_kdb_enter_str) - u.u_error = ENOENT; - else if(copyin(kp->ptr, X_kdb_enter_str, kp->size) == EFAULT) - u.u_error = EFAULT; - - X_kdb_enter_len = kp->size>>2; -} - -void -X_kdb_exit_init(kp) -struct X_kdb *kp; -{ - if (kp->size > sizeof X_kdb_exit_str) - u.u_error = ENOENT; - else if(copyin(kp->ptr, X_kdb_exit_str, kp->size) == EFAULT) - u.u_error = EFAULT; - - X_kdb_exit_len = kp->size>>2; -} -#endif /* MACH_KERNEL */ diff --git a/i386/i386at/kd_mouse.c b/i386/i386at/kd_mouse.c index 640209c5..6e7b68a8 100644 --- a/i386/i386at/kd_mouse.c +++ b/i386/i386at/kd_mouse.c @@ -67,21 +67,10 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #include <mach/boolean.h> #include <sys/types.h> #include <kern/printf.h> -#ifdef MACH_KERNEL #include <device/ds_routines.h> -#include <device/errno.h> +#include <device/device_types.h> #include <device/io_req.h> #include <device/subrs.h> -#else /* MACH_KERNEL */ -#include <sys/file.h> -#include <sys/errno.h> -#include <kern/thread.h> -#include <sys/user.h> -#include <sys/proc.h> -#include <sys/kernel.h> -#include <sys/ioctl.h> -#include <sys/tty.h> -#endif /* MACH_KERNEL */ #include <i386/ipl.h> #include <i386/pic.h> #include <i386/pio.h> @@ -100,20 +89,8 @@ extern struct bus_device *cominfo[]; kd_event_queue mouse_queue; /* queue of mouse events */ boolean_t mouse_in_use = FALSE; -#ifdef MACH_KERNEL queue_head_t mouse_read_queue = { &mouse_read_queue, &mouse_read_queue }; -#else /* MACH_KERNEL */ -struct proc *mouse_sel = 0; /* selecting process, if any */ -short mousepgrp = 0; /* process group leader when dev is open */ -#endif /* MACH_KERNEL */ - -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ -int mouseflag = 0; -#define MOUSE_COLL 1 /* select collision */ -#define MOUSE_ASYNC 2 /* user wants asynch notification */ -#define MOUSE_NBIO 4 /* user wants non-blocking I/O */ -#endif /* MACH_KERNEL */ + /* * The state of the 3 buttons is encoded in the low-order 3 bits (both @@ -176,20 +153,11 @@ mouseopen(dev, flags) dev_t dev; int flags; { -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - if (flags & FWRITE) - return(ENODEV); -#endif /* MACH_KERNEL */ if (mouse_in_use) - return(EBUSY); + return (D_ALREADY_OPEN); mouse_in_use = TRUE; /* locking? */ kdq_reset(&mouse_queue); lastbuttons = MOUSE_ALL_UP; -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - mousepgrp = u.u_procp->p_pgrp; -#endif /* MACH_KERNEL */ switch (mouse_type = ((minor(dev) & 0xf8) >> 3)) { case MICROSOFT_MOUSE7: @@ -294,12 +262,6 @@ mouseclose(dev, flags) kdq_reset(&mouse_queue); /* paranoia */ mouse_in_use = FALSE; -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - mousepgrp = 0; - mouseflag = 0; - mouse_sel = 0; -#endif /* MACH_KERNEL */ } /*ARGSUSED*/ @@ -335,7 +297,6 @@ kd_mouse_close(dev, mouse_pic) splx(s); } -#ifdef MACH_KERNEL io_return_t mousegetstat(dev, flavor, data, count) dev_t dev; int flavor; @@ -354,75 +315,10 @@ io_return_t mousegetstat(dev, flavor, data, count) return D_SUCCESS; } -#else /* MACH_KERNEL */ -/* - * mouseioctl - handling for asynch & non-blocking I/O. - */ - -/*ARGSUSED*/ -int -mouseioctl(dev, cmd, data, flag) - dev_t dev; - int cmd; - caddr_t data; - int flag; -{ - int s = SPLKD(); - int err = 0; - - switch (cmd) { - case FIONBIO: - if (*(int *)data) - mouseflag |= MOUSE_NBIO; - else - mouseflag &= ~MOUSE_NBIO; - break; - case FIOASYNC: - if (*(int *)data) - mouseflag |= MOUSE_ASYNC; - else - mouseflag &= ~MOUSE_ASYNC; - break; - default: - err = ENOTTY; - break; - } - - splx(s); - return(err); -} - - -/* - * mouseselect - check for pending events, etc. - */ - -/*ARGSUSED*/ -int -mouseselect(dev, rw) -{ - int s = SPLKD(); - - if (!kdq_empty(&mouse_queue)) { - splx(s); - return(1); - } - - if (mouse_sel) - mouseflag |= MOUSE_COLL; - else - mouse_sel = (struct proc *)current_thread(); - /* eeeyuck */ - - splx(s); - return(0); -} -#endif /* MACH_KERNEL */ /* * mouseread - dequeue and return any queued events. */ -#ifdef MACH_KERNEL boolean_t mouse_read_done(); /* forward */ int @@ -495,45 +391,6 @@ boolean_t mouse_read_done(ior) return (TRUE); } -#else /* MACH_KERNEL */ -/*ARGSUSED*/ -int -mouseread(dev, uio) - dev_t dev; - struct uio *uio; -{ - int s = SPLKD(); - int err = 0; - kd_event *ev; - int i; - char *cp; - - if (kdq_empty(&mouse_queue)) - if (mouseflag & MOUSE_NBIO) { - err = EWOULDBLOCK; - goto done; - } else - while (kdq_empty(&mouse_queue)) { - splx(s); - sleep((caddr_t)&mouse_queue, TTIPRI); - s = SPLKD(); - } - - while (!kdq_empty(&mouse_queue) && uio->uio_resid >= sizeof(kd_event)) { - ev = kdq_get(&mouse_queue); - for (cp = (char *)ev, i = 0; i < sizeof(kd_event); - ++i, ++cp) { - err = ureadc(*cp, uio); - if (err) - goto done; - } - } - -done: - splx(s); - return(err); -} -#endif /* MACH_KERNEL */ /* @@ -780,12 +637,8 @@ int kd_mouse_read(void) while (mousebufindex <= mouse_char_index) { mouse_char_wanted = TRUE; -#ifdef MACH_KERNEL assert_wait((event_t) &mousebuf, FALSE); thread_block((void (*)()) 0); -#else /* MACH_KERNEL */ - sleep(&mousebuf, PZERO); -#endif /* MACH_KERNEL */ } ch = mousebuf[mouse_char_index++]; @@ -955,20 +808,9 @@ mouse_enqueue(ev) else kdq_put(&mouse_queue, ev); -#ifdef MACH_KERNEL { register io_req_t ior; while ((ior = (io_req_t)dequeue_head(&mouse_read_queue)) != 0) iodone(ior); } -#else /* MACH_KERNEL */ - if (mouse_sel) { - selwakeup(mouse_sel, mouseflag & MOUSE_COLL); - mouse_sel = 0; - mouseflag &= ~MOUSE_COLL; - } - if (mouseflag & MOUSE_ASYNC) - gsignal(mousepgrp, SIGIO); - wakeup((caddr_t)&mouse_queue); -#endif /* MACH_KERNEL */ } diff --git a/i386/i386at/lpr.c b/i386/i386at/lpr.c index 57f6f857..c92795ef 100644 --- a/i386/i386at/lpr.c +++ b/i386/i386at/lpr.c @@ -30,28 +30,15 @@ #if NLPR > 0 -#ifdef MACH_KERNEL #include <mach/std_types.h> #include <sys/types.h> #include <kern/printf.h> #include <kern/mach_clock.h> #include <sys/time.h> #include <device/conf.h> -#include <device/errno.h> +#include <device/device_types.h> #include <device/tty.h> #include <device/io_req.h> -#else /* MACH_KERNEL */ -#include <sys/param.h> -#include <sys/conf.h> -#include <sys/dir.h> -#include <sys/user.h> -#include <sys/proc.h> -#include <sys/ioctl.h> -#include <sys/tty.h> -#include <sys/systm.h> -#include <sys/uio.h> -#include <sys/file.h> -#endif /* MACH_KERNEL */ #include <i386/ipl.h> #include <i386/pio.h> @@ -68,9 +55,7 @@ int lprprobe(); void lprstop(); void lprintr(), lprstart(); void lprattach(struct bus_device *); -#ifdef MACH_KERNEL int lprgetstat(), lprsetstat(); -#endif /* MACH_KERNEL */ void lprpr_addr(); struct bus_device *lprinfo[NLPR]; /* ??? */ @@ -117,7 +102,7 @@ void lprattach(struct bus_device *dev) u_short addr = (u_short) dev->address; take_dev_irq(dev); - printf(", port = %x, spl = %d, pic = %d.", + printf(", port = %lx, spl = %ld, pic = %d.", dev->address, dev->sysdep, dev->sysdep1); lprinfo[unit] = dev; @@ -130,9 +115,7 @@ int lpropen(dev, flag, ior) int dev; int flag; -#ifdef MACH_KERNEL io_req_t ior; -#endif /* MACH_KERNEL */ { int unit = minor(dev); struct bus_device *isai; @@ -140,22 +123,16 @@ struct tty *tp; u_short addr; if (unit >= NLPR || (isai = lprinfo[unit]) == 0 || isai->alive == 0) - return(ENXIO); + return (D_NO_SUCH_DEVICE); tp = &lpr_tty[unit]; -#ifndef MACH_KERNEL - if (tp->t_state & TS_XCLUDE && u.u_uid != 0) - return(EBUSY); -#endif /* MACH_KERNEL */ addr = (u_short) isai->address; tp->t_dev = dev; tp->t_addr = *(caddr_t *)&addr; tp->t_oproc = lprstart; tp->t_state |= TS_WOPEN; -#ifdef MACH_KERNEL tp->t_stop = lprstop; tp->t_getstat = lprgetstat; tp->t_setstat = lprsetstat; -#endif /* MACH_KERNEL */ if ((tp->t_state & TS_ISOPEN) == 0) ttychars(tp); outb(INTR_ENAB(addr), inb(INTR_ENAB(addr)) | 0x10); @@ -172,9 +149,6 @@ int unit = minor(dev); struct tty *tp = &lpr_tty[unit]; u_short addr = (u_short) lprinfo[unit]->address; -#ifndef MACH_KERNEL - (*linesw[tp->t_line].l_close)(tp); -#endif /* MACH_KERNEL */ ttyclose(tp); if (tp->t_state&TS_HUPCLS || (tp->t_state&TS_ISOPEN)==0) { outb(INTR_ENAB(addr), inb(INTR_ENAB(addr)) & 0x0f); @@ -182,7 +156,6 @@ u_short addr = (u_short) lprinfo[unit]->address; } } -#ifdef MACH_KERNEL int lprread(dev, ior) int dev; @@ -212,7 +185,7 @@ lprgetstat(dev, flavor, data, count) dev_t dev; int flavor; int *data; /* pointer to OUT array */ -unsigned int *count; /* out */ +natural_t *count; /* out */ { io_return_t result = D_SUCCESS; int unit = minor(dev); @@ -230,7 +203,7 @@ lprsetstat(dev, flavor, data, count) dev_t dev; int flavor; int * data; -unsigned int count; +natural_t count; { io_return_t result = D_SUCCESS; int unit = minor(dev); @@ -244,43 +217,6 @@ unsigned int count; } return (D_SUCCESS); } -#else /* MACH_KERNEL */ -int lprwrite(dev, uio) - int dev; - struct uio *uio; -{ - struct tty *tp= &lpr_tty[minor(dev)]; - - return ((*linesw[tp->t_line].l_write)(tp, uio)); -} - -int lprioctl(dev, cmd, addr, mode) - int dev; - int cmd; - caddr_t addr; - int mode; -{ - int error; - spl_t s; - int unit = minor(dev); - struct tty *tp = &lpr_tty[unit]; - - error = (*linesw[tp->t_line].l_ioctl)(tp, cmd, addr,mode); - if (error >= 0) - return(error); - error = ttioctl(tp, cmd, addr,mode); - if (error >= 0) - return (error); - s = spltty(); - switch (cmd) { - default: - splx(s); - return(ENOTTY); - } - splx(s); - return(0); -} -#endif /* MACH_KERNEL */ void lprintr(unit) int unit; @@ -317,28 +253,15 @@ struct tty *tp; } if (tp->t_outq.c_cc <= TTLOWAT(tp)) { -#ifdef MACH_KERNEL tt_write_wakeup(tp); -#else /* MACH_KERNEL */ - if (tp->t_state & TS_ASLEEP) { - tp->t_state &= ~TS_ASLEEP; - wakeup ((caddr_t)&tp->t_outq); - } - if (tp->t_wsel) { - selwakeup(tp->t_wsel, tp->t_state & TS_WCOLL); - tp->t_wsel = 0; - tp->t_state &= ~TS_WCOLL; - } -#endif /* MACH_KERNEL */ } if (tp->t_outq.c_cc == 0) { splx(s); return; } -#ifdef MACH_KERNEL nch = getc(&tp->t_outq); if ((tp->t_flags & LITOUT) == 0 && (nch & 0200)) { - timeout(ttrstrt, (char *)tp, (nch & 0x7f) + 6); + timeout((timer_func_t *)ttrstrt, (char *)tp, (nch & 0x7f) + 6); tp->t_state |= TS_TIMEOUT; return; } @@ -346,32 +269,10 @@ struct tty *tp; outb(INTR_ENAB(addr),inb(INTR_ENAB(addr)) | 0x01); outb(INTR_ENAB(addr),inb(INTR_ENAB(addr)) & 0x1e); tp->t_state |= TS_BUSY; -#else /* MACH_KERNEL */ - if (tp->t_flags & (RAW|LITOUT)) - nch = ndqb(&tp->t_outq,0); - else { - nch = ndqb(&tp->t_outq, 0200); - if (nch == 0) { - nch = getc(&tp->t_outq); - timeout(ttrstrt,(caddr_t)tp,(nch&0x7f)+6); - tp->t_state |= TS_TIMEOUT; - splx(s); - return(0); - } - } - if (nch) { - nch=getc(&tp->t_outq); - outb(DATA(addr), nch); - outb(INTR_ENAB(addr),inb(INTR_ENAB(addr)) | 0x01); - outb(INTR_ENAB(addr),inb(INTR_ENAB(addr)) & 0x1e); - tp->t_state |= TS_BUSY; - } -#endif /* MACH_KERNEL */ splx(s); return; } -#ifdef MACH_KERNEL void lprstop(tp, flags) register struct tty *tp; @@ -380,17 +281,6 @@ int flags; if ((tp->t_state & TS_BUSY) && (tp->t_state & TS_TTSTOP) == 0) tp->t_state |= TS_FLUSH; } -#else /* MACH_KERNEL */ -void lprstop(tp, flag) -struct tty *tp; -{ - int s = spltty(); - - if ((tp->t_state&TS_BUSY) && (!(tp->t_state&TS_TTSTOP))) - tp->t_state |= TS_FLUSH; - splx(s); -} -#endif /* MACH_KERNEL */ int lprpr(unit) { diff --git a/i386/i386at/model_dep.c b/i386/i386at/model_dep.c index 04b8228f..3db03d76 100644 --- a/i386/i386at/model_dep.c +++ b/i386/i386at/model_dep.c @@ -128,8 +128,6 @@ extern void setup_main(); void halt_all_cpus (boolean_t reboot) __attribute__ ((noreturn)); void halt_cpu (void) __attribute__ ((noreturn)); -int discover_x86_cpu_type (void); - void inittodr(); /* forward */ int rebootflag = 0; /* exported to kdintr */ @@ -263,6 +261,8 @@ void db_reset_cpu(void) void mem_size_init(void) { + vm_offset_t max_phys_size; + /* Physical memory on all PCs starts at physical address 0. XX make it a constant. */ phys_first_addr = 0; @@ -284,15 +284,16 @@ mem_size_init(void) phys_last_addr = phys_last_kb * 0x400; #endif /* MACH_HYP */ - printf("AT386 boot: physical memory from 0x%x to 0x%x\n", + printf("AT386 boot: physical memory from 0x%lx to 0x%lx\n", phys_first_addr, phys_last_addr); - /* Reserve 1/6 of the memory address space for virtual mappings. + /* Reserve room for virtual mappings. * Yes, this loses memory. Blame i386. */ - if (phys_last_addr > ((VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 6) * 5) { - phys_last_addr = ((VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 6) * 5; - printf("Truncating memory size to %dMiB\n", (phys_last_addr - phys_first_addr) / (1024 * 1024)); - /* TODO Xen: free lost memory */ + max_phys_size = VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS - VM_KERNEL_MAP_SIZE; + if (phys_last_addr - phys_first_addr > max_phys_size) { + phys_last_addr = phys_first_addr + max_phys_size; + printf("Truncating memory size to %luMiB\n", (phys_last_addr - phys_first_addr) / (1024 * 1024)); + /* TODO Xen: be nice, free lost memory */ } phys_first_addr = round_page(phys_first_addr); @@ -318,7 +319,7 @@ i386at_init(void) /* XXX move to intel/pmap.h */ extern pt_entry_t *kernel_page_dir; int nb_direct, i; - vm_offset_t addr; + vm_offset_t addr, delta; /* * Initialize the PIC prior to any possible call to an spl. @@ -380,29 +381,34 @@ i386at_init(void) pmap_bootstrap(); /* - * Turn paging on. * We'll have to temporarily install a direct mapping * between physical memory and low linear memory, * until we start using our new kernel segment descriptors. - * Also, set the WP bit so that on 486 or better processors - * page-level write protection works in kernel mode. */ - init_alloc_aligned(0, &addr); - nb_direct = (addr + NPTES * PAGE_SIZE - 1) / (NPTES * PAGE_SIZE); +#if INIT_VM_MIN_KERNEL_ADDRESS != LINEAR_MIN_KERNEL_ADDRESS + delta = INIT_VM_MIN_KERNEL_ADDRESS - LINEAR_MIN_KERNEL_ADDRESS; + if ((vm_offset_t)(-delta) < delta) + delta = (vm_offset_t)(-delta); + nb_direct = delta >> PDESHIFT; for (i = 0; i < nb_direct; i++) - kernel_page_dir[lin2pdenum(VM_MIN_KERNEL_ADDRESS) + i] = + kernel_page_dir[lin2pdenum(INIT_VM_MIN_KERNEL_ADDRESS) + i] = kernel_page_dir[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS) + i]; +#endif + /* We need BIOS memory mapped at 0xc0000 & co for Linux drivers */ +#ifdef LINUX_DEV +#if VM_MIN_KERNEL_ADDRESS != 0 + kernel_page_dir[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS)] = + kernel_page_dir[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS)]; +#endif +#endif -#ifdef MACH_XEN - { - int i; - for (i = 0; i < PDPNUM; i++) - pmap_set_page_readonly_init((void*) kernel_page_dir + i * INTEL_PGBYTES); +#ifdef MACH_PV_PAGETABLES + for (i = 0; i < PDPNUM; i++) + pmap_set_page_readonly_init((void*) kernel_page_dir + i * INTEL_PGBYTES); #if PAE - pmap_set_page_readonly_init(kernel_pmap->pdpbase); + pmap_set_page_readonly_init(kernel_pmap->pdpbase); #endif /* PAE */ - } -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #if PAE set_cr3((unsigned)_kvtophys(kernel_pmap->pdpbase)); #ifndef MACH_HYP @@ -414,16 +420,19 @@ i386at_init(void) set_cr3((unsigned)_kvtophys(kernel_page_dir)); #endif /* PAE */ #ifndef MACH_HYP - /* already set by Hypervisor */ + /* Turn paging on. + * Also set the WP bit so that on 486 or better processors + * page-level write protection works in kernel mode. + */ set_cr0(get_cr0() | CR0_PG | CR0_WP); set_cr0(get_cr0() & ~(CR0_CD | CR0_NW)); if (CPU_HAS_FEATURE(CPU_FEATURE_PGE)) set_cr4(get_cr4() | CR4_PGE); #endif /* MACH_HYP */ flush_instr_queue(); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES pmap_clear_bootstrap_pagetable((void *)boot_info.pt_base); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ /* Interrupt stacks are allocated in physical memory, while kernel stacks are allocated in kernel virtual memory, @@ -441,6 +450,7 @@ i386at_init(void) ldt_init(); ktss_init(); +#if INIT_VM_MIN_KERNEL_ADDRESS != LINEAR_MIN_KERNEL_ADDRESS /* Get rid of the temporary direct mapping and flush it out of the TLB. */ for (i = 0 ; i < nb_direct; i++) { #ifdef MACH_XEN @@ -451,9 +461,17 @@ i386at_init(void) #endif /* MACH_PSEUDO_PHYS */ printf("couldn't unmap frame %d\n", i); #else /* MACH_XEN */ - kernel_page_dir[lin2pdenum(VM_MIN_KERNEL_ADDRESS) + i] = 0; + kernel_page_dir[lin2pdenum(INIT_VM_MIN_KERNEL_ADDRESS) + i] = 0; #endif /* MACH_XEN */ } +#endif + /* Keep BIOS memory mapped */ +#ifdef LINUX_DEV +#if VM_MIN_KERNEL_ADDRESS != 0 + kernel_page_dir[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS)] = + kernel_page_dir[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS)]; +#endif +#endif /* Not used after boot, better give it back. */ #ifdef MACH_XEN @@ -485,7 +503,7 @@ void c_boot_entry(vm_offset_t bi) /* Before we do _anything_ else, print the hello message. If there are no initialized console devices yet, it will be stored and printed at the first opportunity. */ - printf(version); + printf("%s", version); printf("\n"); #ifdef MACH_XEN @@ -513,7 +531,7 @@ void c_boot_entry(vm_offset_t bi) strtab_size = (vm_offset_t)phystokv(boot_info.syms.a.strsize); kern_sym_end = kern_sym_start + 4 + symtab_size + strtab_size; - printf("kernel symbol table at %08x-%08x (%d,%d)\n", + printf("kernel symbol table at %08lx-%08lx (%d,%d)\n", kern_sym_start, kern_sym_end, symtab_size, strtab_size); } @@ -699,9 +717,9 @@ init_alloc_aligned(vm_size_t size, vm_offset_t *addrp) } /* Skip our own kernel code, data, and bss. */ - if ((avail_next > (vm_offset_t)start) && (addr < (vm_offset_t)end)) + if ((phystokv(avail_next) > (vm_offset_t)start) && (phystokv(addr) < (vm_offset_t)end)) { - avail_next = (vm_offset_t)end; + avail_next = _kvtophys(end); goto retry; } @@ -716,9 +734,9 @@ init_alloc_aligned(vm_size_t size, vm_offset_t *addrp) avail_next = mods_end_pa; goto retry; } - if ((avail_next > kern_sym_start) && (addr < kern_sym_end)) + if ((phystokv(avail_next) > kern_sym_start) && (phystokv(addr) < kern_sym_end)) { - avail_next = kern_sym_end; + avail_next = _kvtophys(kern_sym_end); goto retry; } if (boot_info.flags & MULTIBOOT_MODS) diff --git a/i386/i386at/pic_isa.c b/i386/i386at/pic_isa.c index ef7d84c7..e48fb507 100644 --- a/i386/i386at/pic_isa.c +++ b/i386/i386at/pic_isa.c @@ -27,11 +27,11 @@ #include <sys/types.h> #include <i386/ipl.h> #include <i386/pic.h> - +#include <i386/fpu.h> +#include <i386at/kd.h> /* These interrupts are always present */ -extern void intnull(), fpintr(), hardclock(), kdintr(); -extern void prtnull(); +extern void hardclock(); void (*ivect[NINTR])() = { /* 00 */ hardclock, /* always */ diff --git a/i386/i386at/rtc.c b/i386/i386at/rtc.c index e0a03dee..67768013 100644 --- a/i386/i386at/rtc.c +++ b/i386/i386at/rtc.c @@ -53,7 +53,6 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. #include <i386/pio.h> #include <i386at/rtc.h> -static unsigned char rtc[RTC_NREG]; static int first_rtcopen_ever = 1; void @@ -139,11 +138,7 @@ readtodc(tp) int i, days = 0; spl_t ospl; -#ifdef MACH_KERNEL ospl = splclock(); -#else /* MACH_KERNEL */ - ospl = spl5(); -#endif /* MACH_KERNEL */ if (rtcget(&rtclk)) { splx(ospl); return(-1); @@ -170,12 +165,6 @@ readtodc(tp) days += yeartoday(i); n += days * 3600 * 24; -#ifdef MACH_KERNEL -#else /* MACH_KERNEL */ - n += tz.tz_minuteswest * 60; - if (tz.tz_dsttime) - n -= 3600; -#endif /* MACH_KERNEL */ *tp = n; @@ -190,24 +179,14 @@ writetodc() int diff, i, j; spl_t ospl; -#ifdef MACH_KERNEL ospl = splclock(); -#else /* MACH_KERNEL */ - ospl = spl5(); -#endif /* MACH_KERNEL */ if (rtcget(&rtclk)) { splx(ospl); return(-1); } splx(ospl); -#ifdef MACH_KERNEL diff = 0; -#else /* MACH_KERNEL */ - diff = tz.tz_minuteswest * 60; - if (tz.tz_dsttime) - diff -= 3600; -#endif /* MACH_KERNEL */ n = (time.tv_sec - diff) % (3600 * 24); /* hrs+mins+secs */ rtclk.rtc_sec = dectohexdec(n%60); n /= 60; @@ -231,11 +210,7 @@ writetodc() rtclk.rtc_dom = dectohexdec(++n); -#ifdef MACH_KERNEL ospl = splclock(); -#else /* MACH_KERNEL */ - ospl = spl5(); -#endif /* MACH_KERNEL */ rtcput(&rtclk); splx(ospl); diff --git a/i386/include/mach/i386/exec/elf.h b/i386/include/mach/i386/exec/elf.h index 5155b3d5..cfa988d2 100644 --- a/i386/include/mach/i386/exec/elf.h +++ b/i386/include/mach/i386/exec/elf.h @@ -23,11 +23,11 @@ #ifndef _MACH_I386_EXEC_ELF_H_ #define _MACH_I386_EXEC_ELF_H_ -typedef unsigned long Elf32_Addr; +typedef unsigned int Elf32_Addr; typedef unsigned short Elf32_Half; -typedef unsigned long Elf32_Off; -typedef signed long Elf32_Sword; -typedef unsigned long Elf32_Word; +typedef unsigned int Elf32_Off; +typedef signed int Elf32_Sword; +typedef unsigned int Elf32_Word; /* Architecture identification parameters for i386. */ #define MY_EI_DATA ELFDATA2LSB diff --git a/i386/include/mach/i386/vm_param.h b/i386/include/mach/i386/vm_param.h index 6d7c5f3c..8f708f0f 100644 --- a/i386/include/mach/i386/vm_param.h +++ b/i386/include/mach/i386/vm_param.h @@ -52,8 +52,8 @@ * No rounding is used. */ -#define i386_btop(x) (((unsigned)(x)) >> I386_PGSHIFT) -#define i386_ptob(x) (((unsigned)(x)) << I386_PGSHIFT) +#define i386_btop(x) (((unsigned long)(x)) >> I386_PGSHIFT) +#define i386_ptob(x) (((unsigned long)(x)) << I386_PGSHIFT) /* * Round off or truncate to the nearest page. These will work @@ -61,9 +61,9 @@ * bytes.) */ -#define i386_round_page(x) ((((unsigned)(x)) + I386_PGBYTES - 1) & \ +#define i386_round_page(x) ((((unsigned long)(x)) + I386_PGBYTES - 1) & \ ~(I386_PGBYTES-1)) -#define i386_trunc_page(x) (((unsigned)(x)) & ~(I386_PGBYTES-1)) +#define i386_trunc_page(x) (((unsigned long)(x)) & ~(I386_PGBYTES-1)) /* User address spaces are 3GB each, starting at virtual and linear address 0. diff --git a/i386/include/mach/i386/vm_types.h b/i386/include/mach/i386/vm_types.h index d54008ef..1439940b 100644 --- a/i386/include/mach/i386/vm_types.h +++ b/i386/include/mach/i386/vm_types.h @@ -73,7 +73,7 @@ typedef unsigned int uint32; * A vm_offset_t is a type-neutral pointer, * e.g. an offset into a virtual memory space. */ -typedef natural_t vm_offset_t; +typedef unsigned long vm_offset_t; typedef vm_offset_t * vm_offset_array_t; /* @@ -88,11 +88,11 @@ typedef natural_t vm_size_t; */ typedef signed char signed8_t; typedef signed short signed16_t; -typedef signed long signed32_t; +typedef signed int signed32_t; typedef signed long long signed64_t; typedef unsigned char unsigned8_t; typedef unsigned short unsigned16_t; -typedef unsigned long unsigned32_t; +typedef unsigned int unsigned32_t; typedef unsigned long long unsigned64_t; typedef float float32_t; typedef double float64_t; diff --git a/i386/include/mach/sa/stdarg.h b/i386/include/mach/sa/stdarg.h index ba0f78a1..550fec4f 100644 --- a/i386/include/mach/sa/stdarg.h +++ b/i386/include/mach/sa/stdarg.h @@ -39,7 +39,7 @@ typedef __builtin_va_list va_list; #else -#define __va_size(type) ((sizeof(type)+3) & ~0x3) +#define __va_size(type) ((sizeof(type)+sizeof(unsigned long)-1) & ~(sizeof(unsigned long)-1)) #ifndef _VA_LIST_ #define _VA_LIST_ diff --git a/i386/intel/pmap.c b/i386/intel/pmap.c index 0b8ae903..490c1d95 100644 --- a/i386/intel/pmap.c +++ b/i386/intel/pmap.c @@ -63,7 +63,7 @@ #include <kern/debug.h> #include <kern/printf.h> #include <kern/thread.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <kern/lock.h> @@ -113,7 +113,7 @@ pv_entry_t pv_head_table; /* array of entries, one per page */ /* * pv_list entries are kept on a list that can only be accessed * with the pmap system locked (at SPLVM, not in the cpus_active set). - * The list is refilled from the pv_list_zone if it becomes empty. + * The list is refilled from the pv_list_cache if it becomes empty. */ pv_entry_t pv_free_list; /* free list at SPLVM */ decl_simple_lock_data(, pv_free_list_lock) @@ -133,7 +133,7 @@ decl_simple_lock_data(, pv_free_list_lock) simple_unlock(&pv_free_list_lock); \ } -zone_t pv_list_zone; /* zone of pv_entry structures */ +struct kmem_cache pv_list_cache; /* cache of pv_entry structures */ /* * Each entry in the pv_head_table is locked by a bit in the @@ -155,9 +155,6 @@ boolean_t pmap_initialized = FALSE; vm_offset_t kernel_virtual_start; vm_offset_t kernel_virtual_end; -/* XXX stupid fixed limit - get rid */ -vm_size_t morevm = 128 * 1024 * 1024; /* VM space for kernel map */ - /* * Index into pv_head table, its lock bits, and the modify/reference * bits starting at phys_first_addr. @@ -330,7 +327,7 @@ lock_data_t pmap_system_lock; #define MAX_TBIS_SIZE 32 /* > this -> TBIA */ /* XXX */ -#ifdef MACH_HYP +#ifdef MACH_PV_PAGETABLES #if 1 #define INVALIDATE_TLB(pmap, s, e) hyp_mmuext_op_void(MMUEXT_TLB_FLUSH_LOCAL) #else @@ -342,7 +339,7 @@ lock_data_t pmap_system_lock; hyp_mmuext_op_void(MMUEXT_TLB_FLUSH_LOCAL); \ } while(0) #endif -#else /* MACH_HYP */ +#else /* MACH_PV_PAGETABLES */ #if 0 /* It is hard to know when a TLB flush becomes less expensive than a bunch of * invlpgs. But it surely is more expensive than just one invlpg. */ @@ -361,7 +358,7 @@ lock_data_t pmap_system_lock; flush_tlb(); \ } #endif -#endif /* MACH_HYP */ +#endif /* MACH_PV_PAGETABLES */ #if NCPUS > 1 @@ -403,7 +400,7 @@ struct pmap_update_list cpu_update_list[NCPUS]; struct pmap kernel_pmap_store; pmap_t kernel_pmap; -struct zone *pmap_zone; /* zone of pmap structures */ +struct kmem_cache pmap_cache; /* cache of pmap structures */ int pmap_debug = 0; /* flag for debugging prints */ @@ -531,10 +528,10 @@ vm_offset_t pmap_map_bd(virt, start, end, prot) register pt_entry_t template; register pt_entry_t *pte; int spl; -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES int n, i = 0; struct mmu_update update[HYP_BATCH_MMU_UPDATES]; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ template = pa_to_pte(start) | INTEL_PTE_NCACHE|INTEL_PTE_WTHRU @@ -549,7 +546,7 @@ vm_offset_t pmap_map_bd(virt, start, end, prot) pte = pmap_pte(kernel_pmap, virt); if (pte == PT_ENTRY_NULL) panic("pmap_map_bd: Invalid kernel address\n"); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES update[i].ptr = kv_to_ma(pte); update[i].val = pa_to_ma(template); i++; @@ -559,20 +556,20 @@ vm_offset_t pmap_map_bd(virt, start, end, prot) panic("couldn't pmap_map_bd\n"); i = 0; } -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ WRITE_PTE(pte, template) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ pte_increment_pa(template); virt += PAGE_SIZE; start += PAGE_SIZE; } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (i > HYP_BATCH_MMU_UPDATES) panic("overflowed array in pmap_map_bd"); hyp_mmu_update(kvtolin(&update), i, kvtolin(&n), DOMID_SELF); if (n != i) panic("couldn't pmap_map_bd\n"); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ PMAP_READ_UNLOCK(pmap, spl); return(virt); } @@ -620,8 +617,7 @@ void pmap_bootstrap() * and extends to a stupid arbitrary limit beyond that. */ kernel_virtual_start = phystokv(phys_last_addr); - kernel_virtual_end = phystokv(phys_last_addr) + morevm - + (phys_last_addr - phys_first_addr); + kernel_virtual_end = phystokv(phys_last_addr) + VM_KERNEL_MAP_SIZE; if (kernel_virtual_end < kernel_virtual_start || kernel_virtual_end > VM_MAX_KERNEL_ADDRESS) @@ -653,7 +649,7 @@ void pmap_bootstrap() kernel_pmap->dirbase[i] = 0; } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES /* We don't actually deal with the CR3 register content at all */ hyp_vm_assist(VMASST_CMD_enable, VMASST_TYPE_pae_extended_cr3); /* @@ -671,25 +667,27 @@ void pmap_bootstrap() pt_entry_t *l1_map[NSUP_L1]; { pt_entry_t *base = (pt_entry_t*) boot_info.pt_base; - int i; + vm_offset_t la; int n_l1map; + for (n_l1map = 0, la = VM_MIN_KERNEL_ADDRESS; la >= VM_MIN_KERNEL_ADDRESS; la += NPTES * PAGE_SIZE) { #ifdef PAE - pt_entry_t *l2_map = (pt_entry_t*) ptetokv(base[0]); + pt_entry_t *l2_map = (pt_entry_t*) ptetokv(base[lin2pdpnum(la)]); #else /* PAE */ - pt_entry_t *l2_map = base; + pt_entry_t *l2_map = base; #endif /* PAE */ - for (n_l1map = 0, i = lin2pdenum(VM_MIN_KERNEL_ADDRESS); i < NPTES; i++) { - if (!(l2_map[i] & INTEL_PTE_VALID)) { + /* Like lin2pdenum, but works with non-contiguous boot L3 */ + l2_map += (la >> PDESHIFT) & PDEMASK; + if (!(*l2_map & INTEL_PTE_VALID)) { struct mmu_update update; int j, n; l1_map[n_l1map] = (pt_entry_t*) phystokv(pmap_grab_page()); for (j = 0; j < NPTES; j++) - l1_map[n_l1map][j] = (((pt_entry_t)pfn_to_mfn((i - lin2pdenum(VM_MIN_KERNEL_ADDRESS)) * NPTES + j)) << PAGE_SHIFT) | INTEL_PTE_VALID | INTEL_PTE_WRITE; + l1_map[n_l1map][j] = (((pt_entry_t)pfn_to_mfn(lin2pdenum(la - VM_MIN_KERNEL_ADDRESS) * NPTES + j)) << PAGE_SHIFT) | INTEL_PTE_VALID | INTEL_PTE_WRITE; pmap_set_page_readonly_init(l1_map[n_l1map]); if (!hyp_mmuext_op_mfn (MMUEXT_PIN_L1_TABLE, kv_to_mfn (l1_map[n_l1map]))) panic("couldn't pin page %p(%p)", l1_map[n_l1map], (vm_offset_t) kv_to_ma (l1_map[n_l1map])); - update.ptr = kv_to_ma(&l2_map[i]); + update.ptr = kv_to_ma(l2_map); update.val = kv_to_ma(l1_map[n_l1map]) | INTEL_PTE_VALID | INTEL_PTE_WRITE; hyp_mmu_update(kv_to_la(&update), 1, kv_to_la(&n), DOMID_SELF); if (n != 1) @@ -700,7 +698,7 @@ void pmap_bootstrap() } } } -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ /* * Allocate and set up the kernel page tables. @@ -720,7 +718,7 @@ void pmap_bootstrap() * to allocate new kernel page tables later. * XX fix this */ - for (va = phystokv(phys_first_addr); va < kernel_virtual_end; ) + for (va = phystokv(phys_first_addr); va >= phystokv(phys_first_addr) && va < kernel_virtual_end; ) { pt_entry_t *pde = kernel_page_dir + lin2pdenum(kvtolin(va)); pt_entry_t *ptable = (pt_entry_t*)phystokv(pmap_grab_page()); @@ -738,24 +736,24 @@ void pmap_bootstrap() WRITE_PTE(pte, 0); } else -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (va == (vm_offset_t) &hyp_shared_info) { *pte = boot_info.shared_info | INTEL_PTE_VALID | INTEL_PTE_WRITE; va += INTEL_PGBYTES; } else -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ { extern char _start[], etext[]; if (((va >= (vm_offset_t) _start) && (va + INTEL_PGBYTES <= (vm_offset_t)etext)) -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES || (va >= (vm_offset_t) boot_info.pt_base && (va + INTEL_PGBYTES <= (vm_offset_t) ptable + INTEL_PGBYTES)) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ ) { WRITE_PTE(pte, pa_to_pte(_kvtophys(va)) @@ -763,7 +761,7 @@ void pmap_bootstrap() } else { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES /* Keep supplementary L1 pages read-only */ int i; for (i = 0; i < NSUP_L1; i++) @@ -773,7 +771,7 @@ void pmap_bootstrap() break; } if (i == NSUP_L1) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ WRITE_PTE(pte, pa_to_pte(_kvtophys(va)) | INTEL_PTE_VALID | INTEL_PTE_WRITE | global) @@ -786,11 +784,11 @@ void pmap_bootstrap() WRITE_PTE(pte, 0); va += INTEL_PGBYTES; } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES pmap_set_page_readonly_init(ptable); if (!hyp_mmuext_op_mfn (MMUEXT_PIN_L1_TABLE, kv_to_mfn (ptable))) panic("couldn't pin page %p(%p)\n", ptable, (vm_offset_t) kv_to_ma (ptable)); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } } @@ -798,7 +796,7 @@ void pmap_bootstrap() soon after we return from here. */ } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES /* These are only required because of Xen security policies */ /* Set back a page read write */ @@ -830,26 +828,27 @@ void pmap_set_page_readonly(void *_vaddr) { } /* This needs to be called instead of pmap_set_page_readonly as long as RC3 - * still points to the bootstrap dirbase. */ + * still points to the bootstrap dirbase, to also fix the bootstrap table. */ void pmap_set_page_readonly_init(void *_vaddr) { vm_offset_t vaddr = (vm_offset_t) _vaddr; #if PAE pt_entry_t *pdpbase = (void*) boot_info.pt_base; - vm_offset_t dirbase = ptetokv(pdpbase[0]); + /* The bootstrap table does not necessarily use contiguous pages for the pde tables */ + pt_entry_t *dirbase = (void*) ptetokv(pdpbase[lin2pdpnum(vaddr)]); #else - vm_offset_t dirbase = boot_info.pt_base; + pt_entry_t *dirbase = (void*) boot_info.pt_base; #endif - struct pmap linear_pmap = { - .dirbase = (void*) dirbase, - }; + pt_entry_t *pte = &dirbase[lin2pdenum(vaddr) & PTEMASK]; /* Modify our future kernel map (can't use update_va_mapping for this)... */ - if (*pmap_pde(kernel_pmap, vaddr) & INTEL_PTE_VALID) + if (*pmap_pde(kernel_pmap, vaddr) & INTEL_PTE_VALID) { if (!hyp_mmu_update_la (kvtolin(vaddr), pa_to_pte (kv_to_ma(vaddr)) | INTEL_PTE_VALID)) panic("couldn't set hiMMU readonly for vaddr %p(%p)\n", vaddr, (vm_offset_t) kv_to_ma (vaddr)); + } /* ... and the bootstrap map. */ - if (*pmap_pde(&linear_pmap, vaddr) & INTEL_PTE_VALID) + if (*pte & INTEL_PTE_VALID) { if (hyp_do_update_va_mapping (vaddr, pa_to_pte (kv_to_ma(vaddr)) | INTEL_PTE_VALID, UVMF_NONE)) panic("couldn't set MMU readonly for vaddr %p(%p)\n", vaddr, (vm_offset_t) kv_to_ma (vaddr)); + } } void pmap_clear_bootstrap_pagetable(pt_entry_t *base) { @@ -872,7 +871,7 @@ void pmap_clear_bootstrap_pagetable(pt_entry_t *base) { #endif /* PAE */ for (i = 0; i < NPTES; i++) { pt_entry_t pde = dir[i]; - unsigned long pfn = mfn_to_pfn(atop(pde)); + unsigned long pfn = atop(pte_to_pa(pde)); void *pgt = (void*) phystokv(ptoa(pfn)); if (pde & INTEL_PTE_VALID) hyp_free_page(pfn, pgt); @@ -890,7 +889,7 @@ void pmap_clear_bootstrap_pagetable(pt_entry_t *base) { #endif /* PAE */ hyp_free_page(atop(_kvtophys(base)), base); } -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ void pmap_virtual_space(startp, endp) vm_offset_t *startp; @@ -941,13 +940,13 @@ void pmap_init() pmap_phys_attributes = (char *) addr; /* - * Create the zone of physical maps, + * Create the cache of physical maps, * and of the physical-to-virtual entries. */ s = (vm_size_t) sizeof(struct pmap); - pmap_zone = zinit(s, 0, 400*s, 4096, 0, "pmap"); /* XXX */ + kmem_cache_init(&pmap_cache, "pmap", s, 0, NULL, NULL, NULL, 0); s = (vm_size_t) sizeof(struct pv_entry); - pv_list_zone = zinit(s, 0, 10000*s, 4096, 0, "pv_list"); /* XXX */ + kmem_cache_init(&pv_list_cache, "pv_entry", s, 0, NULL, NULL, NULL, 0); #if NCPUS > 1 /* @@ -1013,7 +1012,7 @@ pmap_page_table_page_alloc() /* * We cannot allocate the pmap_object in pmap_init, - * because it is called before the zone package is up. + * because it is called before the cache package is up. * Allocate it now if it is missing. */ if (pmap_object == VM_OBJECT_NULL) @@ -1052,21 +1051,38 @@ void pmap_map_mfn(void *_addr, unsigned long mfn) { pt_entry_t *pte, *pdp; vm_offset_t ptp; pt_entry_t ma = ((pt_entry_t) mfn) << PAGE_SHIFT; + + /* Add a ptp if none exist yet for this pte */ if ((pte = pmap_pte(kernel_pmap, addr)) == PT_ENTRY_NULL) { ptp = phystokv(pmap_page_table_page_alloc()); +#ifdef MACH_PV_PAGETABLES pmap_set_page_readonly((void*) ptp); if (!hyp_mmuext_op_mfn (MMUEXT_PIN_L1_TABLE, pa_to_mfn(ptp))) panic("couldn't pin page %p(%p)\n",ptp,(vm_offset_t) kv_to_ma(ptp)); +#endif /* MACH_PV_PAGETABLES */ pdp = pmap_pde(kernel_pmap, addr); + +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pdp), pa_to_pte(kv_to_ma(ptp)) | INTEL_PTE_VALID | INTEL_PTE_USER | INTEL_PTE_WRITE)) panic("%s:%d could not set pde %p(%p) to %p(%p)\n",__FILE__,__LINE__,kvtophys((vm_offset_t)pdp),(vm_offset_t) kv_to_ma(pdp), ptp, (vm_offset_t) pa_to_ma(ptp)); +#else /* MACH_PV_PAGETABLES */ + *pdp = pa_to_pte(kvtophys(ptp)) | INTEL_PTE_VALID + | INTEL_PTE_USER + | INTEL_PTE_WRITE; +#endif /* MACH_PV_PAGETABLES */ pte = pmap_pte(kernel_pmap, addr); } + +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pte), ma | INTEL_PTE_VALID | INTEL_PTE_WRITE)) panic("%s:%d could not set pte %p(%p) to %p(%p)\n",__FILE__,__LINE__,pte,(vm_offset_t) kv_to_ma(pte), ma, ma_to_pa(ma)); +#else /* MACH_PV_PAGETABLES */ + /* Note: in this case, mfn is actually a pfn. */ + WRITE_PTE(pte, ma | INTEL_PTE_VALID | INTEL_PTE_WRITE); +#endif /* MACH_PV_PAGETABLES */ } #endif /* MACH_XEN */ @@ -1117,11 +1133,11 @@ pmap_t pmap_create(size) } /* - * Allocate a pmap struct from the pmap_zone. Then allocate - * the page descriptor table from the pd_zone. + * Allocate a pmap struct from the pmap_cache. Then allocate + * the page descriptor table. */ - p = (pmap_t) zalloc(pmap_zone); + p = (pmap_t) kmem_cache_alloc(&pmap_cache); if (p == PMAP_NULL) panic("pmap_create"); @@ -1131,13 +1147,19 @@ pmap_t pmap_create(size) panic("pmap_create"); memcpy(p->dirbase, kernel_page_dir, PDPNUM * INTEL_PGBYTES); -#ifdef MACH_XEN +#ifdef LINUX_DEV +#if VM_MIN_KERNEL_ADDRESS != 0 + /* Do not map BIOS in user tasks */ + p->dirbase[lin2pdenum(LINEAR_MIN_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS)] = 0; +#endif +#endif +#ifdef MACH_PV_PAGETABLES { int i; for (i = 0; i < PDPNUM; i++) pmap_set_page_readonly((void*) p->dirbase + i * INTEL_PGBYTES); } -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #if PAE if (kmem_alloc_wired(kernel_map, @@ -1149,9 +1171,9 @@ pmap_t pmap_create(size) for (i = 0; i < PDPNUM; i++) WRITE_PTE(&p->pdpbase[i], pa_to_pte(kvtophys((vm_offset_t) p->dirbase + i * INTEL_PGBYTES)) | INTEL_PTE_VALID); } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES pmap_set_page_readonly(p->pdpbase); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #endif /* PAE */ p->ref_count = 1; @@ -1211,32 +1233,32 @@ void pmap_destroy(p) if (m == VM_PAGE_NULL) panic("pmap_destroy: pte page not in object"); vm_page_lock_queues(); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!hyp_mmuext_op_mfn (MMUEXT_UNPIN_TABLE, pa_to_mfn(pa))) panic("pmap_destroy: couldn't unpin page %p(%p)\n", pa, (vm_offset_t) kv_to_ma(pa)); pmap_set_page_readwrite((void*) phystokv(pa)); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ vm_page_free(m); inuse_ptepages_count--; vm_page_unlock_queues(); vm_object_unlock(pmap_object); } } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES { int i; for (i = 0; i < PDPNUM; i++) pmap_set_page_readwrite((void*) p->dirbase + i * INTEL_PGBYTES); } -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ kmem_free(kernel_map, (vm_offset_t)p->dirbase, PDPNUM * INTEL_PGBYTES); #if PAE -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES pmap_set_page_readwrite(p->pdpbase); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ kmem_free(kernel_map, (vm_offset_t)p->pdpbase, INTEL_PGBYTES); #endif /* PAE */ - zfree(pmap_zone, (vm_offset_t) p); + kmem_cache_free(&pmap_cache, (vm_offset_t) p); } /* @@ -1279,10 +1301,10 @@ void pmap_remove_range(pmap, va, spte, epte) int num_removed, num_unwired; int pai; vm_offset_t pa; -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES int n, ii = 0; struct mmu_update update[HYP_BATCH_MMU_UPDATES]; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #if DEBUG_PTE_PAGE if (pmap != kernel_pmap) @@ -1311,7 +1333,7 @@ void pmap_remove_range(pmap, va, spte, epte) register int i = ptes_per_vm_page; register pt_entry_t *lpte = cpte; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES update[ii].ptr = kv_to_ma(lpte); update[ii].val = 0; ii++; @@ -1321,9 +1343,9 @@ void pmap_remove_range(pmap, va, spte, epte) panic("couldn't pmap_remove_range\n"); ii = 0; } -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *lpte = 0; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ lpte++; } while (--i > 0); continue; @@ -1344,7 +1366,7 @@ void pmap_remove_range(pmap, va, spte, epte) do { pmap_phys_attributes[pai] |= *lpte & (PHYS_MODIFIED|PHYS_REFERENCED); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES update[ii].ptr = kv_to_ma(lpte); update[ii].val = 0; ii++; @@ -1354,9 +1376,9 @@ void pmap_remove_range(pmap, va, spte, epte) panic("couldn't pmap_remove_range\n"); ii = 0; } -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *lpte = 0; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ lpte++; } while (--i > 0); } @@ -1402,13 +1424,13 @@ void pmap_remove_range(pmap, va, spte, epte) } } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (ii > HYP_BATCH_MMU_UPDATES) panic("overflowed array in pmap_remove_range"); hyp_mmu_update(kvtolin(&update), ii, kvtolin(&n), DOMID_SELF); if (n != ii) panic("couldn't pmap_remove_range\n"); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ /* * Update the counts @@ -1554,12 +1576,12 @@ void pmap_page_protect(phys, prot) do { pmap_phys_attributes[pai] |= *pte & (PHYS_MODIFIED|PHYS_REFERENCED); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pte++), 0)) panic("%s:%d could not clear pte %p\n",__FILE__,__LINE__,pte-1); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *pte++ = 0; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } while (--i > 0); } @@ -1589,12 +1611,12 @@ void pmap_page_protect(phys, prot) register int i = ptes_per_vm_page; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pte), *pte & ~INTEL_PTE_WRITE)) - panic("%s:%d could not enable write on pte %p\n",__FILE__,__LINE__,pte); -#else /* MACH_XEN */ + panic("%s:%d could not disable write on pte %p\n",__FILE__,__LINE__,pte); +#else /* MACH_PV_PAGETABLES */ *pte &= ~INTEL_PTE_WRITE; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ pte++; } while (--i > 0); @@ -1683,14 +1705,14 @@ void pmap_protect(map, s, e, prot) spte = &spte[ptenum(s)]; epte = &spte[intel_btop(l-s)]; -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES int n, i = 0; struct mmu_update update[HYP_BATCH_MMU_UPDATES]; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ while (spte < epte) { if (*spte & INTEL_PTE_VALID) { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES update[i].ptr = kv_to_ma(spte); update[i].val = *spte & ~INTEL_PTE_WRITE; i++; @@ -1700,19 +1722,19 @@ void pmap_protect(map, s, e, prot) panic("couldn't pmap_protect\n"); i = 0; } -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *spte &= ~INTEL_PTE_WRITE; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } spte++; } -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (i > HYP_BATCH_MMU_UPDATES) panic("overflowed array in pmap_protect"); hyp_mmu_update(kvtolin(&update), i, kvtolin(&n), DOMID_SELF); if (n != i) panic("couldn't pmap_protect\n"); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } s = l; pde++; @@ -1751,7 +1773,7 @@ void pmap_enter(pmap, v, pa, prot, wired) vm_offset_t old_pa; assert(pa != vm_page_fictitious_addr); -if (pmap_debug) printf("pmap(%x, %x)\n", v, pa); +if (pmap_debug) printf("pmap(%lx, %lx)\n", v, pa); if (pmap == PMAP_NULL) return; @@ -1786,7 +1808,7 @@ if (pmap_debug) printf("pmap(%x, %x)\n", v, pa); /* * Must allocate a new pvlist entry while we're unlocked; - * zalloc may cause pageout (which will lock the pmap system). + * Allocating may cause pageout (which will lock the pmap system). * If we determine we need a pvlist entry, we will unlock * and allocate one. Then we will retry, throughing away * the allocated entry later (if we no longer need it). @@ -1849,7 +1871,7 @@ Retry: /*XX pdp = &pmap->dirbase[pdenum(v) & ~(i-1)];*/ pdp = pmap_pde(pmap, v); do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES pmap_set_page_readonly((void *) ptp); if (!hyp_mmuext_op_mfn (MMUEXT_PIN_L1_TABLE, kv_to_mfn(ptp))) panic("couldn't pin page %p(%p)\n",ptp,(vm_offset_t) kv_to_ma(ptp)); @@ -1858,11 +1880,11 @@ Retry: | INTEL_PTE_USER | INTEL_PTE_WRITE)) panic("%s:%d could not set pde %p(%p,%p) to %p(%p,%p) %p\n",__FILE__,__LINE__, pdp, kvtophys((vm_offset_t)pdp), (vm_offset_t) pa_to_ma(kvtophys((vm_offset_t)pdp)), ptp, kvtophys(ptp), (vm_offset_t) pa_to_ma(kvtophys(ptp)), (vm_offset_t) pa_to_pte(kv_to_ma(ptp))); -#else /* MACH_XEN */ - *pdp = pa_to_pte(ptp) | INTEL_PTE_VALID - | INTEL_PTE_USER - | INTEL_PTE_WRITE; -#endif /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ + *pdp = pa_to_pte(kvtophys(ptp)) | INTEL_PTE_VALID + | INTEL_PTE_USER + | INTEL_PTE_WRITE; +#endif /* MACH_PV_PAGETABLES */ pdp++; ptp += INTEL_PGBYTES; } while (--i > 0); @@ -1902,12 +1924,12 @@ Retry: do { if (*pte & INTEL_PTE_MOD) template |= INTEL_PTE_MOD; -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pte), pa_to_ma(template))) panic("%s:%d could not set pte %p to %p\n",__FILE__,__LINE__,pte,template); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ WRITE_PTE(pte, template) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ pte++; pte_increment_pa(template); } while (--i > 0); @@ -1970,9 +1992,9 @@ Retry: PMAP_READ_UNLOCK(pmap, spl); /* - * Refill from zone. + * Refill from cache. */ - pv_e = (pv_entry_t) zalloc(pv_list_zone); + pv_e = (pv_entry_t) kmem_cache_alloc(&pv_list_cache); goto Retry; } } @@ -2012,12 +2034,12 @@ Retry: template |= INTEL_PTE_WIRED; i = ptes_per_vm_page; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!(hyp_mmu_update_pte(kv_to_ma(pte), pa_to_ma(template)))) panic("%s:%d could not set pte %p to %p\n",__FILE__,__LINE__,pte,template); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ WRITE_PTE(pte, template) -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ pte++; pte_increment_pa(template); } while (--i > 0); @@ -2072,12 +2094,12 @@ void pmap_change_wiring(map, v, wired) map->stats.wired_count--; i = ptes_per_vm_page; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!(hyp_mmu_update_pte(kv_to_ma(pte), *pte & ~INTEL_PTE_WIRED))) panic("%s:%d could not wire down pte %p\n",__FILE__,__LINE__,pte); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *pte &= ~INTEL_PTE_WIRED; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ pte++; } while (--i > 0); } @@ -2208,7 +2230,7 @@ void pmap_collect(p) register int i = ptes_per_vm_page; register pt_entry_t *pdep = pdp; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES unsigned long pte = *pdep; void *ptable = (void*) ptetokv(pte); if (!(hyp_mmu_update_pte(pa_to_ma(kvtophys((vm_offset_t)pdep++)), 0))) @@ -2216,9 +2238,9 @@ void pmap_collect(p) if (!hyp_mmuext_op_mfn (MMUEXT_UNPIN_TABLE, kv_to_mfn(ptable))) panic("couldn't unpin page %p(%p)\n", ptable, (vm_offset_t) pa_to_ma(kvtophys((vm_offset_t)ptable))); pmap_set_page_readwrite(ptable); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *pdep++ = 0; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } while (--i > 0); } @@ -2435,12 +2457,12 @@ phys_attribute_clear(phys, bits) { register int i = ptes_per_vm_page; do { -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!(hyp_mmu_update_pte(kv_to_ma(pte), *pte & ~bits))) panic("%s:%d could not clear bits %lx from pte %p\n",__FILE__,__LINE__,bits,pte); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *pte &= ~bits; -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } while (--i > 0); } PMAP_UPDATE_TLBS(pmap, va, va + PAGE_SIZE); @@ -2790,7 +2812,7 @@ void pmap_update_interrupt() } #endif /* NCPUS > 1 */ -#ifdef i386 +#if defined(__i386__) /* Unmap page 0 to trap NULL references. */ void pmap_unmap_page_zero () @@ -2801,12 +2823,12 @@ pmap_unmap_page_zero () if (!pte) return; assert (pte); -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES if (!hyp_mmu_update_pte(kv_to_ma(pte), 0)) printf("couldn't unmap page 0\n"); -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ *pte = 0; INVALIDATE_TLB(kernel_pmap, 0, PAGE_SIZE); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ } -#endif /* i386 */ +#endif /* __i386__ */ diff --git a/i386/intel/pmap.h b/i386/intel/pmap.h index 7ba7d2cb..93293e3d 100644 --- a/i386/intel/pmap.h +++ b/i386/intel/pmap.h @@ -37,7 +37,6 @@ #ifndef __ASSEMBLER__ -#include <kern/zalloc.h> #include <kern/lock.h> #include <mach/machine/vm_param.h> #include <mach/vm_statistics.h> @@ -49,7 +48,7 @@ * Define the generic in terms of the specific */ -#if i386 +#if defined(__i386__) #define INTEL_PGBYTES I386_PGBYTES #define INTEL_PGSHIFT I386_PGSHIFT #define intel_btop(x) i386_btop(x) @@ -59,7 +58,7 @@ #define trunc_intel_to_vm(x) trunc_i386_to_vm(x) #define round_intel_to_vm(x) round_i386_to_vm(x) #define vm_to_intel(x) vm_to_i386(x) -#endif /* i386 */ +#endif /* __i386__ */ /* * i386/i486 Page Table Entry @@ -102,6 +101,13 @@ typedef unsigned int pt_entry_t; #endif /* + * Convert linear offset to page directory pointer index + */ +#if PAE +#define lin2pdpnum(a) (((a) >> PDPSHIFT) & PDPMASK) +#endif + +/* * Convert page descriptor index to linear address */ #define pdenum2lin(a) ((vm_offset_t)(a) << PDESHIFT) @@ -126,15 +132,15 @@ typedef unsigned int pt_entry_t; #define INTEL_PTE_NCACHE 0x00000010 #define INTEL_PTE_REF 0x00000020 #define INTEL_PTE_MOD 0x00000040 -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES /* Not supported */ #define INTEL_PTE_GLOBAL 0x00000000 -#else /* MACH_XEN */ +#else /* MACH_PV_PAGETABLES */ #define INTEL_PTE_GLOBAL 0x00000100 -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #define INTEL_PTE_WIRED 0x00000200 #ifdef PAE -#define INTEL_PTE_PFN 0xfffffffffffff000ULL +#define INTEL_PTE_PFN 0x00007ffffffff000ULL #else #define INTEL_PTE_PFN 0xfffff000 #endif @@ -172,13 +178,13 @@ typedef struct pmap *pmap_t; #define PMAP_NULL ((pmap_t) 0) -#ifdef MACH_XEN +#ifdef MACH_PV_PAGETABLES extern void pmap_set_page_readwrite(void *addr); extern void pmap_set_page_readonly(void *addr); extern void pmap_set_page_readonly_init(void *addr); extern void pmap_map_mfn(void *addr, unsigned long mfn); extern void pmap_clear_bootstrap_pagetable(pt_entry_t *addr); -#endif /* MACH_XEN */ +#endif /* MACH_PV_PAGETABLES */ #if PAE #define set_pmap(pmap) set_cr3(kvtophys((vm_offset_t)(pmap)->pdpbase)) diff --git a/i386/ldscript b/i386/ldscript index f2b90fa2..ddbbf910 100644 --- a/i386/ldscript +++ b/i386/ldscript @@ -13,6 +13,7 @@ SECTIONS */ . = _START; .text : + AT (_START_MAP) { *(.text.start) *(.text .stub .text.* .gnu.linkonce.t.*) @@ -69,7 +70,8 @@ SECTIONS .rodata1 : { *(.rodata1) } .eh_frame_hdr : { *(.eh_frame_hdr) } .eh_frame : ONLY_IF_RO { KEEP (*(.eh_frame)) } - .gcc_except_table : ONLY_IF_RO { *(.gcc_except_table .gcc_except_table.*) } + .gcc_except_table : ONLY_IF_RO { *(.gcc_except_table + .gcc_except_table.*) } /* Adjust the address for the data segment. We want to adjust up to the same address within the page on the next page up. */ . = ALIGN (CONSTANT (MAXPAGESIZE)) - ((CONSTANT (MAXPAGESIZE) - .) & (CONSTANT (MAXPAGESIZE) - 1)); . = DATA_SEGMENT_ALIGN (CONSTANT (MAXPAGESIZE), CONSTANT (COMMONPAGESIZE)); diff --git a/i386/xen/Makefrag.am b/i386/xen/Makefrag.am index b15b7db1..ecb33ffc 100644 --- a/i386/xen/Makefrag.am +++ b/i386/xen/Makefrag.am @@ -28,6 +28,7 @@ libkernel_a_SOURCES += \ if PLATFORM_xen gnumach_LINKFLAGS += \ - --defsym _START=0x20000000 \ + --defsym _START=0xC0000000 \ + --defsym _START_MAP=0xC0000000 \ -T '$(srcdir)'/i386/ldscript endif diff --git a/i386/xen/xen.c b/i386/xen/xen.c index aa3c2cc8..a46ee2c6 100644 --- a/i386/xen/xen.c +++ b/i386/xen/xen.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -18,6 +18,7 @@ #include <kern/printf.h> #include <kern/debug.h> +#include <kern/mach_clock.h> #include <mach/machine/eflags.h> #include <machine/thread.h> @@ -43,7 +44,6 @@ void hyp_failsafe_c_callback(struct failsafe_callback_regs *regs) { panic("failsafe"); } -extern void clock_interrupt(); extern void return_to_iret; void hypclock_machine_intr(int old_ipl, void *ret_addr, struct i386_interrupt_state *regs, unsigned64_t delta) { diff --git a/i386/xen/xen_boothdr.S b/i386/xen/xen_boothdr.S index 1b32efb8..ac6ad25d 100644 --- a/i386/xen/xen_boothdr.S +++ b/i386/xen/xen_boothdr.S @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006, 2009, 2010, 2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -22,11 +22,11 @@ .ascii "GUEST_OS=GNU Mach" .ascii ",GUEST_VERSION=1.3" .ascii ",XEN_VER=xen-3.0" - .ascii ",VIRT_BASE=0x20000000" - .ascii ",ELF_PADDR_OFFSET=0x20000000" + .ascii ",VIRT_BASE=0xC0000000" + .ascii ",ELF_PADDR_OFFSET=0xC0000000" .ascii ",HYPERCALL_PAGE=0x2" #if PAE - .ascii ",PAE=yes" + .ascii ",PAE=yes[extended-cr3]" #else .ascii ",PAE=no" #endif @@ -36,6 +36,15 @@ #else /* MACH_PSEUDO_PHYS */ .ascii ",FEATURES=!auto_translated_physmap" #endif +#ifndef MACH_PV_PAGETABLES + .ascii "|!writable_page_tables" +#endif /* MACH_PV_PAGETABLES */ +#ifndef MACH_PV_DESCRIPTORS + .ascii "|!writable_descriptor_tables" +#endif /* MACH_PV_DESCRIPTORS */ +#ifndef MACH_RING1 + .ascii "|!supervisor_mode_kernel" +#endif /* MACH_PV_DESCRIPTORS */ .byte 0 /* Macro taken from linux/include/linux/elfnote.h */ @@ -59,7 +68,7 @@ ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, .long, start) ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, .long, hypcalls) #if PAE - ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz, "yes") + ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz, "yes[extended-cr3]") #else ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz, "no") #endif @@ -68,8 +77,17 @@ #ifdef MACH_PSEUDO_PHYS "pae_pgdir_above_4gb" #else /* MACH_PSEUDO_PHYS */ - "!auto_translated_physmap|" + "!auto_translated_physmap" #endif +#ifndef MACH_PV_PAGETABLES + "|!writable_page_tables" +#endif /* MACH_PV_PAGETABLES */ +#ifndef MACH_PV_DESCRIPTORS + "|!writable_descriptor_tables" +#endif /* MACH_PV_DESCRIPTORS */ +#ifndef MACH_RING1 + "|!supervisor_mode_kernel" +#endif /* MACH_RING1 */ ) #include <mach/machine/asm.h> diff --git a/i386/xen/xen_locore.S b/i386/xen/xen_locore.S index 51f823f2..1468ef80 100644 --- a/i386/xen/xen_locore.S +++ b/i386/xen/xen_locore.S @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by diff --git a/include/device/device.defs b/include/device/device.defs index 2bbd5563..d9234e39 100644 --- a/include/device/device.defs +++ b/include/device/device.defs @@ -32,10 +32,6 @@ * block and character device interfaces to the kernel. */ -#ifdef MACH_KERNEL -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ -#endif - subsystem #if KERNEL_SERVER KernelServer @@ -99,27 +95,9 @@ routine device_read_inband( out data : io_buf_ptr_inband_t ); -/* obsolete */ -routine xxx_device_set_status( - device : device_t; - in flavor : dev_flavor_t; - in status : dev_status_t, IsLong - ); - -/* obsolete */ -routine xxx_device_get_status( - device : device_t; - in flavor : dev_flavor_t; - out status : dev_status_t, IsLong - ); - -/* obsolete */ -routine xxx_device_set_filter( - device : device_t; - in receive_port : mach_port_send_t; - in priority : int; - in filter : filter_array_t, IsLong - ); +skip; /* old xxx_device_set_status */ +skip; /* old xxx_device_get_status */ +skip; /* old xxx_device_set_filter*/ routine device_map( device : device_t; diff --git a/include/mach/gnumach.defs b/include/mach/gnumach.defs new file mode 100644 index 00000000..73313343 --- /dev/null +++ b/include/mach/gnumach.defs @@ -0,0 +1,39 @@ +/* + * Copyright (C) 2012 Free Software Foundation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +subsystem +#if KERNEL_SERVER + KernelServer +#endif /* KERNEL_SERVER */ +#if KERNEL_USER + KernelUser +#endif /* KERNEL_USER */ + gnumach 4200; + +#include <mach/std_types.defs> +#include <mach/mach_types.defs> + +type vm_cache_statistics_data_t = struct[11] of integer_t; + +/* + * Return page cache statistics for the host on which the target task + * resides. + */ +routine vm_cache_statistics( + target_task : vm_task_t; + out vm_cache_stats : vm_cache_statistics_data_t); diff --git a/include/mach/mach.defs b/include/mach/mach.defs index 4531a220..58510805 100644 --- a/include/mach/mach.defs +++ b/include/mach/mach.defs @@ -30,10 +30,6 @@ * Matchmaker definitions file for Mach kernel interface. */ -#ifdef MACH_KERNEL -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ -#endif /* MACH_KERNEL */ - subsystem #if KERNEL_USER KernelUser @@ -388,16 +384,8 @@ skip; /* old pager_flush_request */ * boolean). The new routine is backwards compatible at the C * language interface. */ -simpleroutine xxx_memory_object_lock_request( - memory_control : memory_object_control_t; - offset : vm_offset_t; - size : vm_size_t; - should_clean : boolean_t; - should_flush : boolean_t; - lock_value : vm_prot_t; - reply_to : mach_port_t = - MACH_MSG_TYPE_MAKE_SEND_ONCE|polymorphic); +skip; /* old xxx_memory_object_lock_request */ simpleroutine memory_object_lock_request( memory_control : memory_object_control_t; @@ -409,46 +397,11 @@ simpleroutine memory_object_lock_request( reply_to : mach_port_t = MACH_MSG_TYPE_MAKE_SEND_ONCE|polymorphic); -/* obsolete */ -routine xxx_task_get_emulation_vector( - task : task_t; - out vector_start : int; - out emulation_vector: xxx_emulation_vector_t, IsLong); - -/* obsolete */ -routine xxx_task_set_emulation_vector( - task : task_t; - vector_start : int; - emulation_vector: xxx_emulation_vector_t, IsLong); - -/* - * Returns information about the host on which the - * target object resides. [This object may be - * a task, thread, or memory_object_control port.] - */ -routine xxx_host_info( - target_task : mach_port_t; - out info : machine_info_data_t); - -/* - * Returns information about a particular processor on - * the host on which the target task resides. - */ -routine xxx_slot_info( - target_task : task_t; - slot : int; - out info : machine_slot_data_t); - -/* - * Performs control operations (currently only - * turning off or on) on a particular processor on - * the host on which the target task resides. - */ -routine xxx_cpu_control( - target_task : task_t; - cpu : int; - running : boolean_t); - +skip; /* old xxx_task_get_emulation_vector */ +skip; /* old xxx_task_set_emulation_vector */ +skip; /* old xxx_host_info */ +skip; /* old xxx_slot_info */ +skip; /* old xxx_cpu_control */ skip; /* old thread_statistics */ skip; /* old task_statistics */ skip; /* old netport_init */ @@ -491,11 +444,7 @@ routine task_set_special_port( which_port : int; special_port : mach_port_t); -/* obsolete */ -routine xxx_task_info( - target_task : task_t; - flavor : int; - out task_info_out : task_info_t, IsLong); +skip; /* old xxx_task_info */ /* @@ -537,17 +486,8 @@ routine thread_resume( routine thread_abort( target_thread : thread_t); -/* obsolete */ -routine xxx_thread_get_state( - target_thread : thread_t; - flavor : int; - out old_state : thread_state_t, IsLong); - -/* obsolete */ -routine xxx_thread_set_state( - target_thread : thread_t; - flavor : int; - new_state : thread_state_t, IsLong); +skip; /* old xxx_thread_get_state */ +skip; /* old xxx_thread_set_state */ /* * Returns the current value of the selected special port @@ -567,11 +507,7 @@ routine thread_set_special_port( which_port : int; special_port : mach_port_t); -/* obsolete */ -routine xxx_thread_info( - target_thread : thread_t; - flavor : int; - out thread_info_out : thread_info_t, IsLong); +skip; /* old xxx_thread_info */ /* * Establish a user-level handler for the specified diff --git a/include/mach/mach_host.defs b/include/mach/mach_host.defs index 85ee4dc5..2644146f 100644 --- a/include/mach/mach_host.defs +++ b/include/mach/mach_host.defs @@ -34,10 +34,6 @@ * control. */ -#ifdef MACH_KERNEL -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ -#endif - subsystem #if KERNEL_SERVER KernelServer @@ -59,19 +55,8 @@ routine host_processors( host_priv : host_priv_t; out processor_list : processor_array_t); -/* obsolete */ -routine yyy_host_info( - host : host_t; - flavor : int; - out host_info_out : host_info_t, IsLong); - - -/* obsolete */ -routine yyy_processor_info( - processor : processor_t; - flavor : int; - out host : host_t; - out processor_info_out: processor_info_t, IsLong); +skip; /* old yyy_host_info */ +skip; /* old yyy_processor_info */ /* * Start processor. @@ -87,10 +72,7 @@ routine processor_start( routine processor_exit( processor : processor_t); -/* obsolete */ -routine yyy_processor_control( - processor : processor_t; - processor_cmd : processor_info_t, IsLong); +skip; /* old yyy_processor_control */ /* * Get default processor set for host. @@ -99,13 +81,7 @@ routine processor_set_default( host : host_t; out default_set : processor_set_name_t); -/* - * Get rights to default processor set for host. - * Replaced by host_processor_set_priv. - */ -routine xxx_processor_set_default_priv( - host : host_priv_t; - out default_set : processor_set_t); +skip; /* old xxx_processor_set_default_priv */ /* * Create new processor set. Returns real port for manipulations, @@ -122,12 +98,7 @@ routine processor_set_create( routine processor_set_destroy( set : processor_set_t); -/* obsolete */ -routine yyy_processor_set_info( - set_name : processor_set_name_t; - flavor : int; - out host : host_t; - out info_out : processor_set_info_t, IsLong); +skip; /* old yyy_processor_set_info */ /* * Assign processor to processor set. diff --git a/include/mach/mach_types.defs b/include/mach/mach_types.defs index 4e448b81..607d5d92 100644 --- a/include/mach/mach_types.defs +++ b/include/mach/mach_types.defs @@ -121,11 +121,11 @@ type vm_statistics_data_t = struct[13] of integer_t; type vm_machine_attribute_t = int; type vm_machine_attribute_val_t = int; -type thread_info_t = array[*:1024] of natural_t; +type thread_info_t = array[*:1024] of integer_t; type thread_basic_info_data_t = struct[11] of integer_t; type thread_sched_info_data_t = struct[7] of integer_t; -type task_info_t = array[*:1024] of natural_t; +type task_info_t = array[*:1024] of integer_t; type task_basic_info_data_t = struct[8] of integer_t; type task_events_info = struct[7] of natural_t; type task_thread_times_info_data_t = struct[4] of integer_t; @@ -174,7 +174,7 @@ type host_priv_t = mach_port_t #endif /* KERNEL_SERVER */ ; -type host_info_t = array[*:1024] of natural_t; +type host_info_t = array[*:1024] of integer_t; type host_basic_info_data_t = struct[5] of integer_t; type host_sched_info_data_t = struct[2] of integer_t; type host_load_info_data_t = struct[6] of integer_t; @@ -189,7 +189,7 @@ type processor_t = mach_port_t ; type processor_array_t = ^array[] of processor_t; -type processor_info_t = array[*:1024] of natural_t; +type processor_info_t = array[*:1024] of integer_t; type processor_basic_info_data_t = struct[5] of integer_t; @@ -215,7 +215,7 @@ type processor_set_name_t = mach_port_t type processor_set_name_array_t = ^array[] of processor_set_name_t; -type processor_set_info_t = array[*:1024] of natural_t; +type processor_set_info_t = array[*:1024] of integer_t; type processor_set_basic_info_data_t = struct[5] of integer_t; type processor_set_sched_info_data_t = struct[2] of integer_t; @@ -228,10 +228,6 @@ type time_value_t = struct[2] of integer_t; type emulation_vector_t = ^array[] of vm_offset_t; -type xxx_emulation_vector_t = array[*:1024] of vm_offset_t - ctype: emulation_vector_t; - /* XXX compatibility */ - type rpc_signature_info_t = array[*:1024] of int; #if KERNEL_SERVER diff --git a/include/mach/mach_types.h b/include/mach/mach_types.h index f6ceac3b..87684824 100644 --- a/include/mach/mach_types.h +++ b/include/mach/mach_types.h @@ -52,6 +52,7 @@ #include <mach/vm_inherit.h> #include <mach/vm_prot.h> #include <mach/vm_statistics.h> +#include <mach/vm_cache_statistics.h> #ifdef MACH_KERNEL #include <kern/task.h> /* for task_array_t */ diff --git a/include/mach/port.h b/include/mach/port.h index 6dafb2f3..53f60716 100644 --- a/include/mach/port.h +++ b/include/mach/port.h @@ -39,7 +39,7 @@ #include <mach/machine/vm_types.h> -typedef natural_t mach_port_t; +typedef vm_offset_t mach_port_t; typedef mach_port_t *mach_port_array_t; typedef int *rpc_signature_info_t; diff --git a/include/mach/syscall_sw.h b/include/mach/syscall_sw.h index af14c8de..89597e99 100644 --- a/include/mach/syscall_sw.h +++ b/include/mach/syscall_sw.h @@ -52,6 +52,7 @@ kernel_trap(mach_reply_port,-26,0) kernel_trap(mach_thread_self,-27,0) kernel_trap(mach_task_self,-28,0) kernel_trap(mach_host_self,-29,0) +kernel_trap(mach_print,-30,1) kernel_trap(swtch_pri,-59,1) kernel_trap(swtch,-60,0) diff --git a/include/mach/vm_cache_statistics.h b/include/mach/vm_cache_statistics.h new file mode 100644 index 00000000..072976af --- /dev/null +++ b/include/mach/vm_cache_statistics.h @@ -0,0 +1,41 @@ +/* + * Copyright (C) 2012 Free Software Foundation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +#ifndef _MACH_VM_CACHE_STATISTICS_H_ +#define _MACH_VM_CACHE_STATISTICS_H_ + +#include <mach/machine/vm_types.h> + +struct vm_cache_statistics { + integer_t cache_object_count; /* # of cached objects */ + integer_t cache_count; /* # of cached pages */ + integer_t active_tmp_count; /* # of active temporary pages */ + integer_t inactive_tmp_count; /* # of inactive temporary pages */ + integer_t active_perm_count; /* # of active permanent pages */ + integer_t inactive_perm_count; /* # of inactive permanent pages */ + integer_t dirty_count; /* # of dirty pages */ + integer_t laundry_count; /* # of pages being laundered */ + integer_t writeback_count; /* # of pages being written back */ + integer_t slab_count; /* # of slab allocator pages */ + integer_t slab_reclaim_count; /* # of reclaimable slab pages */ +}; + +typedef struct vm_cache_statistics *vm_cache_statistics_t; +typedef struct vm_cache_statistics vm_cache_statistics_data_t; + +#endif /* _MACH_VM_CACHE_STATISTICS_H_ */ diff --git a/include/mach/xen.h b/include/mach/xen.h index f1d9e418..6fc626f2 100644 --- a/include/mach/xen.h +++ b/include/mach/xen.h @@ -1,6 +1,6 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009, 2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -28,6 +28,7 @@ extern struct start_info boot_info; extern volatile struct shared_info hyp_shared_info; +#ifdef MACH_PV_PAGETABLES /* Memory translations */ /* pa are physical addresses, from 0 to size of memory */ @@ -77,6 +78,11 @@ extern unsigned long *mfn_list; #define kv_to_mfn(a) pa_to_mfn(_kvtophys(a)) #define kv_to_ma(a) pa_to_ma(_kvtophys(a)) +#else /* MACH_PV_PAGETABLES */ +#define mfn_to_pfn(n) (n) +#define pfn_to_mfn(n) (n) +#endif /* MACH_PV_PAGETABLES */ + #define mfn_to_kv(mfn) phystokv(ptoa(mfn_to_pfn(mfn))) #include <machine/xen.h> diff --git a/include/mach_debug/mach_debug.defs b/include/mach_debug/mach_debug.defs index 2a58dc4c..053c3fe6 100644 --- a/include/mach_debug/mach_debug.defs +++ b/include/mach_debug/mach_debug.defs @@ -42,17 +42,7 @@ skip; /* host_ipc_statistics_reset */ skip; /* host_callout_info */ skip; /* host_callout_statistics */ skip; /* host_callout_statistics_reset */ - -/* - * Returns information about the memory allocation zones. - */ -routine host_zone_info( - host : host_t; - out names : zone_name_array_t, - CountInOut, Dealloc; - out info : zone_info_array_t, - CountInOut, Dealloc); - +skip; /* host_zone_info */ skip; /* host_ipc_bucket_info */ #if !defined(MACH_IPC_DEBUG) || MACH_IPC_DEBUG @@ -228,6 +218,14 @@ routine mach_vm_object_pages( out pages : vm_page_info_array_t, CountInOut, Dealloc); +/* + * Returns information about the memory allocation caches. + */ +routine host_slab_info( + host : host_t; + out info : cache_info_array_t, + CountInOut, Dealloc); + #else /* !defined(MACH_VM_DEBUG) || MACH_VM_DEBUG */ skip; /* mach_vm_region_info */ skip; /* mach_vm_object_info */ diff --git a/include/mach_debug/mach_debug_types.defs b/include/mach_debug/mach_debug_types.defs index 9f1976fe..f60125a0 100644 --- a/include/mach_debug/mach_debug_types.defs +++ b/include/mach_debug/mach_debug_types.defs @@ -32,11 +32,8 @@ #include <mach/std_types.defs> -type zone_name_t = struct[80] of char; -type zone_name_array_t = array[] of zone_name_t; - -type zone_info_t = struct[9] of integer_t; -type zone_info_array_t = array[] of zone_info_t; +type cache_info_t = struct[19] of integer_t; +type cache_info_array_t = array[] of cache_info_t; type hash_info_bucket_t = struct[1] of natural_t; type hash_info_bucket_array_t = array[] of hash_info_bucket_t; diff --git a/include/mach_debug/mach_debug_types.h b/include/mach_debug/mach_debug_types.h index 2ba0cb1f..5d4efcde 100644 --- a/include/mach_debug/mach_debug_types.h +++ b/include/mach_debug/mach_debug_types.h @@ -32,7 +32,7 @@ #include <mach_debug/ipc_info.h> #include <mach_debug/vm_info.h> -#include <mach_debug/zone_info.h> +#include <mach_debug/slab_info.h> #include <mach_debug/hash_info.h> typedef char symtab_name_t[32]; diff --git a/include/mach_debug/zone_info.h b/include/mach_debug/slab_info.h index 1b36fe02..37dcb8c4 100644 --- a/include/mach_debug/zone_info.h +++ b/include/mach_debug/slab_info.h @@ -24,38 +24,39 @@ * the rights to redistribute these changes. */ -#ifndef _MACH_DEBUG_ZONE_INFO_H_ -#define _MACH_DEBUG_ZONE_INFO_H_ +#ifndef _MACH_DEBUG_SLAB_INFO_H_ +#define _MACH_DEBUG_SLAB_INFO_H_ -#include <mach/boolean.h> -#include <mach/machine/vm_types.h> +#include <sys/types.h> /* * Remember to update the mig type definitions * in mach_debug_types.defs when adding/removing fields. */ -#define ZONE_NAME_MAX_LEN 80 - -typedef struct zone_name { - char zn_name[ZONE_NAME_MAX_LEN]; -} zone_name_t; - -typedef zone_name_t *zone_name_array_t; - - -typedef struct zone_info { - integer_t zi_count; /* Number of elements used now */ - vm_size_t zi_cur_size; /* current memory utilization */ - vm_size_t zi_max_size; /* how large can this zone grow */ - vm_size_t zi_elem_size; /* size of an element */ - vm_size_t zi_alloc_size; /* size used for more memory */ -/*boolean_t*/integer_t zi_pageable; /* zone pageable? */ -/*boolean_t*/integer_t zi_sleepable; /* sleep if empty? */ -/*boolean_t*/integer_t zi_exhaustible; /* merely return if empty? */ -/*boolean_t*/integer_t zi_collectable; /* garbage collect elements? */ -} zone_info_t; - -typedef zone_info_t *zone_info_array_t; - -#endif /* _MACH_DEBUG_ZONE_INFO_H_ */ +#define CACHE_NAME_MAX_LEN 32 + +#define CACHE_FLAGS_NO_CPU_POOL 0x01 +#define CACHE_FLAGS_SLAB_EXTERNAL 0x02 +#define CACHE_FLAGS_NO_RECLAIM 0x04 +#define CACHE_FLAGS_VERIFY 0x08 +#define CACHE_FLAGS_DIRECT 0x10 + +typedef struct cache_info { + int flags; + size_t cpu_pool_size; + size_t obj_size; + size_t align; + size_t buf_size; + size_t slab_size; + unsigned long bufs_per_slab; + unsigned long nr_objs; + unsigned long nr_bufs; + unsigned long nr_slabs; + unsigned long nr_free_slabs; + char name[CACHE_NAME_MAX_LEN]; +} cache_info_t; + +typedef cache_info_t *cache_info_array_t; + +#endif /* _MACH_DEBUG_SLAB_INFO_H_ */ diff --git a/include/string.h b/include/string.h index 8059bedb..c77d387b 100644 --- a/include/string.h +++ b/include/string.h @@ -32,7 +32,7 @@ extern void *memcpy (void *dest, const void *src, size_t n); extern void *memmove (void *dest, const void *src, size_t n); -extern int *memcmp (const void *s1, const void *s2, size_t n); +extern int memcmp (const void *s1, const void *s2, size_t n); extern void *memset (void *s, int c, size_t n); diff --git a/include/sys/types.h b/include/sys/types.h index d79e077c..19e7b242 100644 --- a/include/sys/types.h +++ b/include/sys/types.h @@ -30,7 +30,7 @@ #ifndef _SIZE_T #define _SIZE_T -typedef natural_t size_t; +typedef unsigned long size_t; #endif #ifndef _SSIZE_T diff --git a/ipc/ipc_entry.c b/ipc/ipc_entry.c index 42e8dd8e..3a062447 100644 --- a/ipc/ipc_entry.c +++ b/ipc/ipc_entry.c @@ -41,7 +41,7 @@ #include <mach/port.h> #include <kern/assert.h> #include <kern/sched_prim.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <ipc/port.h> #include <ipc/ipc_types.h> #include <ipc/ipc_entry.h> @@ -51,7 +51,7 @@ #include <ipc/ipc_table.h> #include <ipc/ipc_object.h> -zone_t ipc_tree_entry_zone; +struct kmem_cache ipc_tree_entry_cache; /* * Routine: ipc_entry_tree_collision diff --git a/ipc/ipc_entry.h b/ipc/ipc_entry.h index a577cf0c..6afa4f68 100644 --- a/ipc/ipc_entry.h +++ b/ipc/ipc_entry.h @@ -41,7 +41,7 @@ #include <mach/mach_types.h> #include <mach/port.h> #include <mach/kern_return.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <ipc/port.h> #include <ipc/ipc_table.h> #include <ipc/ipc_types.h> @@ -129,10 +129,10 @@ typedef struct ipc_tree_entry { #define ite_request ite_entry.ie_request #define ite_next ite_entry.hash.tree -extern zone_t ipc_tree_entry_zone; +extern struct kmem_cache ipc_tree_entry_cache; -#define ite_alloc() ((ipc_tree_entry_t) zalloc(ipc_tree_entry_zone)) -#define ite_free(ite) zfree(ipc_tree_entry_zone, (vm_offset_t) (ite)) +#define ite_alloc() ((ipc_tree_entry_t) kmem_cache_alloc(&ipc_tree_entry_cache)) +#define ite_free(ite) kmem_cache_free(&ipc_tree_entry_cache, (vm_offset_t) (ite)) extern ipc_entry_t diff --git a/ipc/ipc_hash.c b/ipc/ipc_hash.c index c2412890..5eec58cb 100644 --- a/ipc/ipc_hash.c +++ b/ipc/ipc_hash.c @@ -326,7 +326,7 @@ ipc_hash_global_delete( */ #define IH_LOCAL_HASH(obj, size) \ - ((((mach_port_index_t) (obj)) >> 6) % (size)) + ((((mach_port_index_t) (vm_offset_t) (obj)) >> 6) % (size)) /* * Routine: ipc_hash_local_lookup @@ -535,13 +535,9 @@ ipc_hash_init(void) { ipc_hash_index_t i; - /* if not configured, initialize ipc_hash_global_size */ + /* initialize ipc_hash_global_size */ - if (ipc_hash_global_size == 0) { - ipc_hash_global_size = ipc_tree_entry_max >> 8; - if (ipc_hash_global_size < 32) - ipc_hash_global_size = 32; - } + ipc_hash_global_size = IPC_HASH_GLOBAL_SIZE; /* make sure it is a power of two */ diff --git a/ipc/ipc_hash.h b/ipc/ipc_hash.h index 58c56ca8..929ba77d 100644 --- a/ipc/ipc_hash.h +++ b/ipc/ipc_hash.h @@ -67,6 +67,8 @@ ipc_hash_delete(ipc_space_t space, ipc_object_t obj, * and the local primitives, for table entries. */ +#define IPC_HASH_GLOBAL_SIZE 256 + extern boolean_t ipc_hash_global_lookup(ipc_space_t space, ipc_object_t obj, mach_port_t *namep, ipc_tree_entry_t *entryp); diff --git a/ipc/ipc_init.c b/ipc/ipc_init.c index 90e6b783..ca7e7912 100644 --- a/ipc/ipc_init.c +++ b/ipc/ipc_init.c @@ -35,8 +35,8 @@ */ #include <mach/kern_return.h> -#include <kern/mach_param.h> #include <kern/ipc_host.h> +#include <kern/slab.h> #include <vm/vm_map.h> #include <vm/vm_kern.h> #include <ipc/ipc_entry.h> @@ -52,14 +52,10 @@ -vm_map_t ipc_kernel_map; +static struct vm_map ipc_kernel_map_store; +vm_map_t ipc_kernel_map = &ipc_kernel_map_store; vm_size_t ipc_kernel_map_size = 8 * 1024 * 1024; -int ipc_space_max = SPACE_MAX; -int ipc_tree_entry_max = ITE_MAX; -int ipc_port_max = PORT_MAX; -int ipc_pset_max = SET_MAX; - /* * Routine: ipc_bootstrap * Purpose: @@ -77,28 +73,17 @@ ipc_bootstrap(void) ipc_port_timestamp_lock_init(); ipc_port_timestamp_data = 0; - ipc_space_zone = zinit(sizeof(struct ipc_space), 0, - ipc_space_max * sizeof(struct ipc_space), - sizeof(struct ipc_space), - IPC_ZONE_TYPE, "ipc spaces"); - - ipc_tree_entry_zone = - zinit(sizeof(struct ipc_tree_entry), 0, - ipc_tree_entry_max * sizeof(struct ipc_tree_entry), - sizeof(struct ipc_tree_entry), - IPC_ZONE_TYPE, "ipc tree entries"); - - ipc_object_zones[IOT_PORT] = - zinit(sizeof(struct ipc_port), 0, - ipc_port_max * sizeof(struct ipc_port), - sizeof(struct ipc_port), - 0, "ipc ports"); - - ipc_object_zones[IOT_PORT_SET] = - zinit(sizeof(struct ipc_pset), 0, - ipc_pset_max * sizeof(struct ipc_pset), - sizeof(struct ipc_pset), - IPC_ZONE_TYPE, "ipc port sets"); + kmem_cache_init(&ipc_space_cache, "ipc_space", + sizeof(struct ipc_space), 0, NULL, NULL, NULL, 0); + + kmem_cache_init(&ipc_tree_entry_cache, "ipc_tree_entry", + sizeof(struct ipc_tree_entry), 0, NULL, NULL, NULL, 0); + + kmem_cache_init(&ipc_object_caches[IOT_PORT], "ipc_port", + sizeof(struct ipc_port), 0, NULL, NULL, NULL, 0); + + kmem_cache_init(&ipc_object_caches[IOT_PORT_SET], "ipc_pset", + sizeof(struct ipc_pset), 0, NULL, NULL, NULL, 0); /* create special spaces */ @@ -127,8 +112,8 @@ ipc_init() { vm_offset_t min, max; - ipc_kernel_map = kmem_suballoc(kernel_map, &min, &max, - ipc_kernel_map_size, TRUE); + kmem_submap(ipc_kernel_map, kernel_map, &min, &max, + ipc_kernel_map_size, TRUE); ipc_host_init(); } diff --git a/ipc/ipc_init.h b/ipc/ipc_init.h index b2f1dd4b..8dd64bb5 100644 --- a/ipc/ipc_init.h +++ b/ipc/ipc_init.h @@ -37,14 +37,6 @@ #ifndef _IPC_IPC_INIT_H_ #define _IPC_IPC_INIT_H_ -/* all IPC zones should be exhaustible */ -#define IPC_ZONE_TYPE ZONE_EXHAUSTIBLE - -extern int ipc_space_max; -extern int ipc_tree_entry_max; -extern int ipc_port_max; -extern int ipc_pset_max; - /* * Exported interfaces */ diff --git a/ipc/ipc_kmsg.c b/ipc/ipc_kmsg.c index a12c9476..c2689a48 100644 --- a/ipc/ipc_kmsg.c +++ b/ipc/ipc_kmsg.c @@ -1368,7 +1368,7 @@ ipc_kmsg_copyin_body(kmsg, space, map) mach_msg_type_number_t number; boolean_t is_inline, longform, dealloc, is_port; vm_offset_t data; - vm_size_t length; + unsigned64_t length; kern_return_t kr; type = (mach_msg_type_long_t *) saddr; @@ -1419,7 +1419,7 @@ ipc_kmsg_copyin_body(kmsg, space, map) /* calculate length of data in bytes, rounding up */ - length = ((number * size) + 7) >> 3; + length = (((unsigned64_t) number * size) + 7) >> 3; if (is_inline) { vm_size_t amount; @@ -2382,7 +2382,7 @@ ipc_kmsg_copyout_body(saddr, eaddr, space, map) mach_msg_type_size_t size; mach_msg_type_number_t number; boolean_t is_inline, longform, is_port; - vm_size_t length; + unsigned64_t length; vm_offset_t addr; type = (mach_msg_type_long_t *) saddr; @@ -2413,7 +2413,7 @@ ipc_kmsg_copyout_body(saddr, eaddr, space, map) /* calculate length of data in bytes, rounding up */ - length = ((number * size) + 7) >> 3; + length = (((unsigned64_t) number * size) + 7) >> 3; is_port = MACH_MSG_TYPE_PORT_ANY(name); diff --git a/ipc/ipc_marequest.c b/ipc/ipc_marequest.c index 540382af..06c53eb4 100644 --- a/ipc/ipc_marequest.c +++ b/ipc/ipc_marequest.c @@ -37,9 +37,8 @@ #include <mach/message.h> #include <mach/port.h> #include <kern/lock.h> -#include <kern/mach_param.h> #include <kern/kalloc.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <ipc/port.h> #include <ipc/ipc_init.h> #include <ipc/ipc_space.h> @@ -58,11 +57,10 @@ #endif -zone_t ipc_marequest_zone; -int ipc_marequest_max = IMAR_MAX; +struct kmem_cache ipc_marequest_cache; -#define imar_alloc() ((ipc_marequest_t) zalloc(ipc_marequest_zone)) -#define imar_free(imar) zfree(ipc_marequest_zone, (vm_offset_t) (imar)) +#define imar_alloc() ((ipc_marequest_t) kmem_cache_alloc(&ipc_marequest_cache)) +#define imar_free(imar) kmem_cache_free(&ipc_marequest_cache, (vm_offset_t) (imar)) typedef unsigned int ipc_marequest_index_t; @@ -100,13 +98,9 @@ ipc_marequest_init(void) { ipc_marequest_index_t i; - /* if not configured, initialize ipc_marequest_size */ + /* initialize ipc_marequest_size */ - if (ipc_marequest_size == 0) { - ipc_marequest_size = ipc_marequest_max >> 8; - if (ipc_marequest_size < 16) - ipc_marequest_size = 16; - } + ipc_marequest_size = IPC_MAREQUEST_SIZE; /* make sure it is a power of two */ @@ -142,11 +136,8 @@ ipc_marequest_init(void) bucket->imarb_head = IMAR_NULL; } - ipc_marequest_zone = - zinit(sizeof(struct ipc_marequest), 0, - ipc_marequest_max * sizeof(struct ipc_marequest), - sizeof(struct ipc_marequest), - IPC_ZONE_TYPE, "ipc msg-accepted requests"); + kmem_cache_init(&ipc_marequest_cache, "ipc_marequest", + sizeof(struct ipc_marequest), 0, NULL, NULL, NULL, 0); } /* @@ -439,7 +430,7 @@ ipc_marequest_info(maxp, info, count) info[i].hib_count = bucket_count; } - *maxp = ipc_marequest_max; + *maxp = (unsigned int)-1; return ipc_marequest_size; } diff --git a/ipc/ipc_marequest.h b/ipc/ipc_marequest.h index eb3746d2..4f6f7584 100644 --- a/ipc/ipc_marequest.h +++ b/ipc/ipc_marequest.h @@ -70,6 +70,7 @@ typedef struct ipc_marequest { #define IMAR_NULL ((ipc_marequest_t) 0) +#define IPC_MAREQUEST_SIZE 16 extern void ipc_marequest_init(void); diff --git a/ipc/ipc_mqueue.c b/ipc/ipc_mqueue.c index f3d5d4b0..80a34d3a 100644 --- a/ipc/ipc_mqueue.c +++ b/ipc/ipc_mqueue.c @@ -374,6 +374,8 @@ ipc_mqueue_send(kmsg, option, time_out) } } + current_task()->messages_sent++; + return MACH_MSG_SUCCESS; } @@ -684,6 +686,8 @@ ipc_mqueue_receive( ip_unlock(port); } + current_task()->messages_received++; + *kmsgp = kmsg; *seqnop = seqno; return MACH_MSG_SUCCESS; diff --git a/ipc/ipc_notify.c b/ipc/ipc_notify.c index d06346ea..25fa421b 100644 --- a/ipc/ipc_notify.c +++ b/ipc/ipc_notify.c @@ -264,7 +264,7 @@ ipc_notify_port_deleted(port, name) kmsg = ikm_alloc(sizeof *n); if (kmsg == IKM_NULL) { - printf("dropped port-deleted (0x%p, 0x%x)\n", port, name); + printf("dropped port-deleted (0x%p, 0x%lx)\n", port, name); ipc_port_release_sonce(port); return; } @@ -298,7 +298,7 @@ ipc_notify_msg_accepted(port, name) kmsg = ikm_alloc(sizeof *n); if (kmsg == IKM_NULL) { - printf("dropped msg-accepted (0x%p, 0x%x)\n", port, name); + printf("dropped msg-accepted (0x%p, 0x%lx)\n", port, name); ipc_port_release_sonce(port); return; } @@ -437,7 +437,7 @@ ipc_notify_dead_name(port, name) kmsg = ikm_alloc(sizeof *n); if (kmsg == IKM_NULL) { - printf("dropped dead-name (0x%p, 0x%x)\n", port, name); + printf("dropped dead-name (0x%p, 0x%lx)\n", port, name); ipc_port_release_sonce(port); return; } diff --git a/ipc/ipc_object.c b/ipc/ipc_object.c index a7a7ddb2..b8cae8f5 100644 --- a/ipc/ipc_object.c +++ b/ipc/ipc_object.c @@ -47,8 +47,14 @@ #include <ipc/ipc_pset.h> #include <kern/debug.h> #include <kern/printf.h> +#include <kern/slab.h> -zone_t ipc_object_zones[IOT_NUMBER]; +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ + + +struct kmem_cache ipc_object_caches[IOT_NUMBER]; diff --git a/ipc/ipc_object.h b/ipc/ipc_object.h index 2bbf8bd7..adf5bca4 100644 --- a/ipc/ipc_object.h +++ b/ipc/ipc_object.h @@ -39,7 +39,7 @@ #include <ipc/ipc_types.h> #include <kern/lock.h> #include <kern/macro_help.h> -#include <kern/zalloc.h> +#include <kern/slab.h> typedef unsigned int ipc_object_refs_t; typedef unsigned int ipc_object_bits_t; @@ -57,7 +57,7 @@ typedef struct ipc_object { #define IO_VALID(io) (((io) != IO_NULL) && ((io) != IO_DEAD)) #define IO_BITS_KOTYPE 0x0000ffff /* used by the object */ -#define IO_BITS_OTYPE 0x7fff0000 /* determines a zone */ +#define IO_BITS_OTYPE 0x7fff0000 /* determines a cache */ #define IO_BITS_ACTIVE 0x80000000U /* is object alive? */ #define io_active(io) ((int)(io)->io_bits < 0) /* hack */ @@ -75,13 +75,13 @@ typedef struct ipc_object { #define IOT_PORT_SET 1 #define IOT_NUMBER 2 /* number of types used */ -extern zone_t ipc_object_zones[IOT_NUMBER]; +extern struct kmem_cache ipc_object_caches[IOT_NUMBER]; #define io_alloc(otype) \ - ((ipc_object_t) zalloc(ipc_object_zones[(otype)])) + ((ipc_object_t) kmem_cache_alloc(&ipc_object_caches[(otype)])) #define io_free(otype, io) \ - zfree(ipc_object_zones[(otype)], (vm_offset_t) (io)) + kmem_cache_free(&ipc_object_caches[(otype)], (vm_offset_t) (io)) #define io_lock_init(io) simple_lock_init(&(io)->io_lock_data) #define io_lock(io) simple_lock(&(io)->io_lock_data) diff --git a/ipc/ipc_port.c b/ipc/ipc_port.c index ce0dbeb4..8e41c3ca 100644 --- a/ipc/ipc_port.c +++ b/ipc/ipc_port.c @@ -51,6 +51,9 @@ #include <ipc/ipc_mqueue.h> #include <ipc/ipc_notify.h> +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ decl_simple_lock_data(, ipc_port_multiple_lock_data) diff --git a/ipc/ipc_pset.c b/ipc/ipc_pset.c index 141cbdb8..e2b3c862 100644 --- a/ipc/ipc_pset.c +++ b/ipc/ipc_pset.c @@ -46,6 +46,10 @@ #include <ipc/ipc_right.h> #include <ipc/ipc_space.h> +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ + /* * Routine: ipc_pset_alloc diff --git a/ipc/ipc_space.c b/ipc/ipc_space.c index 0f50f15b..ab55e838 100644 --- a/ipc/ipc_space.c +++ b/ipc/ipc_space.c @@ -43,7 +43,7 @@ #include <mach/port.h> #include <kern/assert.h> #include <kern/sched_prim.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <ipc/port.h> #include <ipc/ipc_entry.h> #include <ipc/ipc_splay.h> @@ -55,7 +55,7 @@ -zone_t ipc_space_zone; +struct kmem_cache ipc_space_cache; ipc_space_t ipc_space_kernel; ipc_space_t ipc_space_reply; diff --git a/ipc/ipc_space.h b/ipc/ipc_space.h index d030bf76..c4683d20 100644 --- a/ipc/ipc_space.h +++ b/ipc/ipc_space.h @@ -44,6 +44,7 @@ #include <mach/mach_types.h> #include <kern/macro_help.h> #include <kern/lock.h> +#include <kern/slab.h> #include <ipc/ipc_splay.h> #include <ipc/ipc_types.h> @@ -82,10 +83,10 @@ struct ipc_space { #define IS_NULL ((ipc_space_t) 0) -extern zone_t ipc_space_zone; +extern struct kmem_cache ipc_space_cache; -#define is_alloc() ((ipc_space_t) zalloc(ipc_space_zone)) -#define is_free(is) zfree(ipc_space_zone, (vm_offset_t) (is)) +#define is_alloc() ((ipc_space_t) kmem_cache_alloc(&ipc_space_cache)) +#define is_free(is) kmem_cache_free(&ipc_space_cache, (vm_offset_t) (is)) extern struct ipc_space *ipc_space_kernel; extern struct ipc_space *ipc_space_reply; diff --git a/ipc/ipc_table.c b/ipc/ipc_table.c index e5723586..cbb6a894 100644 --- a/ipc/ipc_table.c +++ b/ipc/ipc_table.c @@ -39,6 +39,7 @@ #include <ipc/ipc_port.h> #include <ipc/ipc_entry.h> #include <kern/kalloc.h> +#include <kern/slab.h> #include <vm/vm_kern.h> /* @@ -50,13 +51,6 @@ void ipc_table_fill( unsigned int min, vm_size_t elemsize); -/* - * We borrow the kalloc map, rather than creating - * yet another submap of the kernel map. - */ - -extern vm_map_t kalloc_map; - ipc_table_size_t ipc_table_entries; unsigned int ipc_table_entries_size = 512; @@ -151,7 +145,7 @@ ipc_table_alloc( if (size < PAGE_SIZE) table = kalloc(size); else - if (kmem_alloc(kalloc_map, &table, size) != KERN_SUCCESS) + if (kmem_alloc(kmem_map, &table, size) != KERN_SUCCESS) table = 0; return table; @@ -177,7 +171,7 @@ ipc_table_realloc( { vm_offset_t new_table; - if (kmem_realloc(kalloc_map, old_table, old_size, + if (kmem_realloc(kmem_map, old_table, old_size, &new_table, new_size) != KERN_SUCCESS) new_table = 0; @@ -201,5 +195,5 @@ ipc_table_free( if (size < PAGE_SIZE) kfree(table, size); else - kmem_free(kalloc_map, table, size); + kmem_free(kmem_map, table, size); } diff --git a/ipc/ipc_thread.h b/ipc/ipc_thread.h index e8bfe4a8..fbeea46a 100644 --- a/ipc/ipc_thread.h +++ b/ipc/ipc_thread.h @@ -46,6 +46,11 @@ typedef thread_t ipc_thread_t; #define ith_lock(thread) simple_lock(&(thread)->ith_lock_data) #define ith_unlock(thread) simple_unlock(&(thread)->ith_lock_data) +/* + * Note that this isn't a queue, but rather a stack. This causes + * threads that were recently running to be reused earlier, which + * helps improve locality of reference. + */ typedef struct ipc_thread_queue { ipc_thread_t ithq_base; } *ipc_thread_queue_t; @@ -103,6 +108,7 @@ MACRO_BEGIN \ (thread)->ith_prev = _last; \ _first->ith_prev = (thread); \ _last->ith_next = (thread); \ + (queue)->ithq_base = (thread); \ } \ MACRO_END diff --git a/ipc/mach_msg.c b/ipc/mach_msg.c index 43ae918a..00ab085b 100644 --- a/ipc/mach_msg.c +++ b/ipc/mach_msg.c @@ -218,7 +218,7 @@ mach_msg_receive(msg, option, rcv_size, rcv_name, time_out, notify) if (mr != MACH_MSG_SUCCESS) { if (mr == MACH_RCV_TOO_LARGE) { mach_msg_size_t real_size = - (mach_msg_size_t) (natural_t) kmsg; + (mach_msg_size_t) (vm_offset_t) kmsg; assert(real_size > rcv_size); @@ -309,7 +309,7 @@ mach_msg_receive_continue(void) if (mr != MACH_MSG_SUCCESS) { if (mr == MACH_RCV_TOO_LARGE) { mach_msg_size_t real_size = - (mach_msg_size_t) (natural_t) kmsg; + (mach_msg_size_t) (vm_offset_t) kmsg; assert(real_size > rcv_size); diff --git a/ipc/mach_port.c b/ipc/mach_port.c index c5688c90..d0310b55 100644 --- a/ipc/mach_port.c +++ b/ipc/mach_port.c @@ -571,7 +571,7 @@ mach_port_destroy( kr = ipc_right_lookup_write(space, name, &entry); if (kr != KERN_SUCCESS) { if (name != MACH_PORT_NULL && name != MACH_PORT_DEAD && space == current_space()) { - printf("task %p destroying an invalid port %u, most probably a bug.\n", current_task(), name); + printf("task %p destroying an invalid port %lu, most probably a bug.\n", current_task(), name); if (mach_port_deallocate_debug) SoftDebugger("mach_port_deallocate"); } @@ -615,7 +615,7 @@ mach_port_deallocate( kr = ipc_right_lookup_write(space, name, &entry); if (kr != KERN_SUCCESS) { if (name != MACH_PORT_NULL && name != MACH_PORT_DEAD && space == current_space()) { - printf("task %p deallocating an invalid port %u, most probably a bug.\n", current_task(), name); + printf("task %p deallocating an invalid port %lu, most probably a bug.\n", current_task(), name); if (mach_port_deallocate_debug) SoftDebugger("mach_port_deallocate"); } @@ -30,8 +30,7 @@ #include <mach/kern_return.h> #include <mach/alert.h> -#include <kern/mach_param.h> /* XXX INCALL_... */ -#include <kern/zalloc.h> +#include <kern/slab.h> #include <kern/thread.h> #include <kern/task.h> #include <kern/debug.h> @@ -47,7 +46,7 @@ static void special_handler(ReturnHandler *rh, struct Act *act); #endif #ifndef ACT_STATIC_KLUDGE -static zone_t act_zone; +static struct kmem_cache act_cache; #else static Act *act_freelist; static Act free_acts[ACT_STATIC_KLUDGE]; @@ -68,11 +67,8 @@ void global_act_init() { #ifndef ACT_STATIC_KLUDGE - act_zone = zinit( - sizeof(struct Act), 0, - ACT_MAX * sizeof(struct Act), /* XXX */ - ACT_CHUNK * sizeof(struct Act), - 0, "activations"); + kmem_cache_init(&act_cache, "Act", sizeof(struct Act), 0, + NULL, NULL, NULL, 0); #else int i; @@ -104,7 +100,7 @@ kern_return_t act_create(task_t task, vm_offset_t user_stack, int rc; #ifndef ACT_STATIC_KLUDGE - act = (Act*)zalloc(act_zone); + act = (Act*)kmem_cache_alloc(&act_cache); if (act == 0) return(KERN_RESOURCE_SHORTAGE); #else @@ -170,9 +166,9 @@ static void act_free(Act *inc) /* Drop the task reference. */ task_deallocate(inc->task); - /* Put the act back on the act zone */ + /* Put the act back on the act cache */ #ifndef ACT_STATIC_KLUDGE - zfree(act_zone, (vm_offset_t)inc); + kmem_cache_free(&act_cache, (vm_offset_t)inc); #else /* XXX ipt_lock(act_freelist); */ inc->ipt_next = act_freelist; @@ -68,7 +68,7 @@ #define AST_PER_THREAD (AST_HALT | AST_TERMINATE | MACHINE_AST_PER_THREAD) -typedef unsigned int ast_t; +typedef unsigned long ast_t; extern volatile ast_t need_ast[NCPUS]; diff --git a/kern/boot_script.c b/kern/boot_script.c index 93491267..b2e9393b 100644 --- a/kern/boot_script.c +++ b/kern/boot_script.c @@ -17,7 +17,7 @@ struct sym int type; /* Symbol value. */ - int val; + long val; /* For function symbols; type of value returned by function. */ int ret_type; @@ -44,7 +44,7 @@ struct arg int type; /* Argument value. */ - int val; + long val; }; /* List of commands. */ @@ -67,23 +67,23 @@ static int symtab_index = 0; /* Create a task and suspend it. */ static int -create_task (struct cmd *cmd, int *val) +create_task (struct cmd *cmd, long *val) { int err = boot_script_task_create (cmd); - *val = (int) cmd->task; + *val = (long) cmd->task; return err; } /* Resume a task. */ static int -resume_task (struct cmd *cmd, int *val) +resume_task (struct cmd *cmd, long *val) { return boot_script_task_resume (cmd); } /* Resume a task when the user hits return. */ static int -prompt_resume_task (struct cmd *cmd, int *val) +prompt_resume_task (struct cmd *cmd, long *val) { return boot_script_prompt_task_resume (cmd); } @@ -91,9 +91,9 @@ prompt_resume_task (struct cmd *cmd, int *val) /* List of builtin symbols. */ static struct sym builtin_symbols[] = { - { "task-create", VAL_FUNC, (int) create_task, VAL_TASK, 0 }, - { "task-resume", VAL_FUNC, (int) resume_task, VAL_NONE, 1 }, - { "prompt-task-resume", VAL_FUNC, (int) prompt_resume_task, VAL_NONE, 1 }, + { "task-create", VAL_FUNC, (long) create_task, VAL_TASK, 0 }, + { "task-resume", VAL_FUNC, (long) resume_task, VAL_NONE, 1 }, + { "prompt-task-resume", VAL_FUNC, (long) prompt_resume_task, VAL_NONE, 1 }, }; #define NUM_BUILTIN (sizeof (builtin_symbols) / sizeof (builtin_symbols[0])) @@ -294,7 +294,8 @@ boot_script_parse_line (void *hook, char *cmdline) for (p += 2;;) { char c; - int i, val, type; + int i, type; + long val; struct sym *s; /* Parse symbol name. */ @@ -349,7 +350,7 @@ boot_script_parse_line (void *hook, char *cmdline) if (! s->run_on_exec) { (error - = ((*((int (*) (struct cmd *, int *)) s->val)) + = ((*((int (*) (struct cmd *, long *)) s->val)) (cmd, &val))); if (error) goto bad; @@ -371,7 +372,7 @@ boot_script_parse_line (void *hook, char *cmdline) else if (s->type == VAL_NONE) { type = VAL_SYM; - val = (int) s; + val = (long) s; } else { @@ -658,7 +659,7 @@ boot_script_exec () /* Create an entry for the variable NAME with TYPE and value VAL, in the symbol table. */ int -boot_script_set_variable (const char *name, int type, int val) +boot_script_set_variable (const char *name, int type, long val) { struct sym *sym = sym_enter (name); @@ -681,7 +682,7 @@ boot_script_define_function (const char *name, int ret_type, if (sym) { sym->type = VAL_FUNC; - sym->val = (int) func; + sym->val = (long) func; sym->ret_type = ret_type; sym->run_on_exec = ret_type == VAL_NONE; } diff --git a/kern/boot_script.h b/kern/boot_script.h index c436ac21..c5ad6732 100644 --- a/kern/boot_script.h +++ b/kern/boot_script.h @@ -102,7 +102,7 @@ int boot_script_exec (void); /* Create an entry in the symbol table for variable NAME, whose type is TYPE and value is VAL. Returns 0 on success, non-zero otherwise. */ -int boot_script_set_variable (const char *name, int type, int val); +int boot_script_set_variable (const char *name, int type, long val); /* Define the function NAME, which will return type RET_TYPE. */ int boot_script_define_function (const char *name, int ret_type, diff --git a/kern/bootstrap.c b/kern/bootstrap.c index 7dd7c140..4a39f26a 100644 --- a/kern/bootstrap.c +++ b/kern/bootstrap.c @@ -42,11 +42,13 @@ #include <kern/debug.h> #include <kern/host.h> #include <kern/printf.h> +#include <kern/kalloc.h> #include <kern/task.h> #include <kern/thread.h> #include <kern/lock.h> #include <vm/vm_kern.h> #include <vm/vm_user.h> +#include <vm/pmap.h> #include <device/device_port.h> #if MACH_KDB @@ -112,10 +114,10 @@ task_insert_send_right( void bootstrap_create() { int compat; + int n = 0; #ifdef MACH_XEN struct multiboot_module *bmods = ((struct multiboot_module *) boot_info.mod_start); - int n = 0; if (bmods) for (n = 0; bmods[n].mod_start; n++) { bmods[n].mod_start = kvtophys(bmods[n].mod_start + (vm_offset_t) bmods); @@ -168,19 +170,19 @@ void bootstrap_create() /* Initialize boot script variables. We leak these send rights. */ losers = boot_script_set_variable ("host-port", VAL_PORT, - (int)ipc_port_make_send(realhost.host_priv_self)); + (long)ipc_port_make_send(realhost.host_priv_self)); if (losers) panic ("cannot set boot-script variable host-port: %s", boot_script_error_string (losers)); losers = boot_script_set_variable ("device-port", VAL_PORT, - (int) ipc_port_make_send(master_device_port)); + (long) ipc_port_make_send(master_device_port)); if (losers) panic ("cannot set boot-script variable device-port: %s", boot_script_error_string (losers)); losers = boot_script_set_variable ("kernel-command-line", VAL_STR, - (int) kernel_cmdline); + (long) kernel_cmdline); if (losers) panic ("cannot set boot-script variable %s: %s", "kernel-command-line", boot_script_error_string (losers)); @@ -199,12 +201,12 @@ void bootstrap_create() get_compat_strings(flag_string, root_string); losers = boot_script_set_variable ("boot-args", VAL_STR, - (int) flag_string); + (long) flag_string); if (losers) panic ("cannot set boot-script variable %s: %s", "boot-args", boot_script_error_string (losers)); losers = boot_script_set_variable ("root-device", VAL_STR, - (int) root_string); + (long) root_string); if (losers) panic ("cannot set boot-script variable %s: %s", "root-device", boot_script_error_string (losers)); @@ -225,7 +227,7 @@ void bootstrap_create() char *var = memcpy (alloca (len), *ep, len); char *val = strchr (var, '='); *val++ = '\0'; - losers = boot_script_set_variable (var, VAL_STR, (int) val); + losers = boot_script_set_variable (var, VAL_STR, (long) val); if (losers) panic ("cannot set boot-script variable %s: %s", var, boot_script_error_string (losers)); @@ -246,7 +248,7 @@ void bootstrap_create() if (eq == 0) continue; *eq++ = '\0'; - losers = boot_script_set_variable (word, VAL_STR, (int) eq); + losers = boot_script_set_variable (word, VAL_STR, (long) eq); if (losers) panic ("cannot set boot-script variable %s: %s", word, boot_script_error_string (losers)); @@ -279,8 +281,10 @@ void bootstrap_create() panic ("ERROR in executing boot script: %s", boot_script_error_string (losers)); } - /* XXX at this point, we could free all the memory used - by the boot modules and the boot loader's descriptors and such. */ + /* XXX we could free the memory used + by the boot loader's descriptors and such. */ + for (n = 0; n < boot_info.mods_count; n++) + vm_page_create(bmods[n].mod_start, bmods[n].mod_end); } static void diff --git a/kern/compat_xxx_defs.h b/kern/compat_xxx_defs.h deleted file mode 100644 index 1878bb22..00000000 --- a/kern/compat_xxx_defs.h +++ /dev/null @@ -1,64 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1991 Carnegie Mellon University. - * Copyright (c) 1993,1994 The University of Utah and - * the Computer Systems Laboratory (CSL). - * All rights reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON, THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF - * THIS SOFTWARE IN ITS "AS IS" CONDITION, AND DISCLAIM ANY LIABILITY - * OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF - * THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * Compatibility definitions for the MiG-related changes - * to various routines. - * - * When all user code has been relinked, this file and the xxx_ - * and yyy_ routines MUST be removed! - */ - -/* from mach.defs */ - -#define xxx_task_info task_info -#ifdef MIGRATING_THREADS -#define xxx_thread_get_state act_get_state -#define xxx_thread_set_state act_set_state -#define xxx_thread_info act_info -#else -#define xxx_thread_get_state thread_get_state -#define xxx_thread_set_state thread_set_state -#define xxx_thread_info thread_info -#endif /* MIGRATING_THREADS */ - -/* from mach_host.defs */ - -#define yyy_host_info host_info -#define yyy_processor_info processor_info -#define yyy_processor_set_info processor_set_info -#define yyy_processor_control processor_control - -/* from device.defs */ - -#define ds_xxx_device_set_status ds_device_set_status -#define ds_xxx_device_get_status ds_device_get_status -#define ds_xxx_device_set_filter ds_device_set_filter - - - diff --git a/kern/debug.c b/kern/debug.c index 178d0788..8a04f1db 100644 --- a/kern/debug.c +++ b/kern/debug.c @@ -100,7 +100,7 @@ void SoftDebugger(message) gimmeabreak(); #endif -#ifdef i386 +#if defined(__i386__) asm("int3"); #endif } @@ -168,7 +168,7 @@ panic(const char *s, ...) #endif printf(": "); va_start(listp, s); - _doprnt(s, &listp, do_cnputc, 0, 0); + _doprnt(s, listp, do_cnputc, 0, 0); va_end(listp); printf("\n"); @@ -205,7 +205,7 @@ log(int level, const char *fmt, ...) level++; #endif va_start(listp, fmt); - _doprnt(fmt, &listp, do_cnputc, 0, 0); + _doprnt(fmt, listp, do_cnputc, 0, 0); va_end(listp); } diff --git a/kern/debug.h b/kern/debug.h index f4e8200d..e429bdd1 100644 --- a/kern/debug.h +++ b/kern/debug.h @@ -57,6 +57,8 @@ #endif /* NDEBUG */ +extern void log (int level, const char *fmt, ...); + extern void panic_init(void); extern void panic (const char *s, ...) __attribute__ ((noreturn)); diff --git a/kern/gnumach.srv b/kern/gnumach.srv new file mode 100644 index 00000000..38bc9090 --- /dev/null +++ b/kern/gnumach.srv @@ -0,0 +1,23 @@ +/* + * Copyright (c) 2012 Free Software Foundation. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +/* This is a server presentation file. */ + +#define KERNEL_SERVER 1 + +#include <mach/gnumach.defs> diff --git a/kern/ipc_host.c b/kern/ipc_host.c index ea8f03ad..cd1c11ab 100644 --- a/kern/ipc_host.c +++ b/kern/ipc_host.c @@ -199,10 +199,9 @@ ipc_pset_terminate( } /* - * processor_set_default, processor_set_default_priv: + * processor_set_default: * - * Return ports for manipulating default_processor set. MiG code - * differentiates between these two routines. + * Return ports for manipulating default_processor set. */ kern_return_t processor_set_default( @@ -217,19 +216,6 @@ processor_set_default( return KERN_SUCCESS; } -kern_return_t -xxx_processor_set_default_priv( - host_t host, - processor_set_t *pset) -{ - if (host == HOST_NULL) - return KERN_INVALID_ARGUMENT; - - *pset = &default_pset; - pset_reference(*pset); - return KERN_SUCCESS; -} - /* * Routine: convert_port_to_host * Purpose: diff --git a/kern/ipc_kobject.c b/kern/ipc_kobject.c index 3d8775b4..37d4eb99 100644 --- a/kern/ipc_kobject.c +++ b/kern/ipc_kobject.c @@ -46,6 +46,7 @@ #include <ipc/ipc_port.h> #include <ipc/ipc_thread.h> #include <vm/vm_object.h> +#include <vm/memory_object_proxy.h> #include <device/ds_routines.h> #if MACH_MACHINE_ROUTINES @@ -150,7 +151,8 @@ ipc_kobject_server(request) mach_host_server_routine(), device_server_routine(), device_pager_server_routine(), - mach4_server_routine(); + mach4_server_routine(), + gnumach_server_routine(); #if MACH_DEBUG extern mig_routine_t mach_debug_server_routine(); #endif @@ -169,20 +171,27 @@ ipc_kobject_server(request) || (routine = mach_debug_server_routine(&request->ikm_header)) != 0 #endif /* MACH_DEBUG */ || (routine = mach4_server_routine(&request->ikm_header)) != 0 + || (routine = gnumach_server_routine(&request->ikm_header)) != 0 #if MACH_MACHINE_ROUTINES || (routine = MACHINE_SERVER_ROUTINE(&request->ikm_header)) != 0 #endif /* MACH_MACHINE_ROUTINES */ ) { (*routine)(&request->ikm_header, &reply->ikm_header); - } - else if (!ipc_kobject_notify(&request->ikm_header,&reply->ikm_header)){ + kernel_task->messages_received++; + } else { + if (!ipc_kobject_notify(&request->ikm_header, + &reply->ikm_header)) { ((mig_reply_header_t *) &reply->ikm_header)->RetCode = MIG_BAD_ID; #if MACH_IPC_TEST printf("ipc_kobject_server: bogus kernel message, id=%d\n", request->ikm_header.msgh_id); #endif /* MACH_IPC_TEST */ + } else { + kernel_task->messages_received++; + } } + kernel_task->messages_sent++; } check_simple_locks(); @@ -318,7 +327,7 @@ ipc_kobject_destroy( default: #if MACH_ASSERT - printf("ipc_kobject_destroy: port 0x%p, kobj 0x%x, type %d\n", + printf("ipc_kobject_destroy: port 0x%p, kobj 0x%lx, type %d\n", port, port->ip_kobject, ip_kotype(port)); #endif /* MACH_ASSERT */ break; diff --git a/kern/ipc_tt.c b/kern/ipc_tt.c index de4edc65..6d32e5b0 100644 --- a/kern/ipc_tt.c +++ b/kern/ipc_tt.c @@ -36,6 +36,7 @@ #include <mach/thread_special_ports.h> #include <vm/vm_kern.h> #include <kern/debug.h> +#include <kern/kalloc.h> #include <kern/task.h> #include <kern/thread.h> #include <kern/ipc_kobject.h> diff --git a/kern/kalloc.c b/kern/kalloc.c deleted file mode 100644 index 8256305b..00000000 --- a/kern/kalloc.c +++ /dev/null @@ -1,254 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University. - * Copyright (c) 1993,1994 The University of Utah and - * the Computer Systems Laboratory (CSL). - * All rights reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON, THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF - * THIS SOFTWARE IN ITS "AS IS" CONDITION, AND DISCLAIM ANY LIABILITY - * OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF - * THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * File: kern/kalloc.c - * Author: Avadis Tevanian, Jr. - * Date: 1985 - * - * General kernel memory allocator. This allocator is designed - * to be used by the kernel to manage dynamic memory fast. - */ - -#include <mach/machine/vm_types.h> -#include <mach/vm_param.h> - -#include <kern/debug.h> -#include <kern/zalloc.h> -#include <kern/kalloc.h> -#include <vm/vm_kern.h> -#include <vm/vm_object.h> -#include <vm/vm_map.h> - - - -vm_map_t kalloc_map; -vm_size_t kalloc_map_size = 64 * 1024 * 1024; -vm_size_t kalloc_max; - -/* - * All allocations of size less than kalloc_max are rounded to the - * next highest power of 2. This allocator is built on top of - * the zone allocator. A zone is created for each potential size - * that we are willing to get in small blocks. - * - * We assume that kalloc_max is not greater than 64K; - * thus 16 is a safe array size for k_zone and k_zone_name. - */ - -int first_k_zone = -1; -struct zone *k_zone[16]; -static char *k_zone_name[16] = { - "kalloc.1", "kalloc.2", - "kalloc.4", "kalloc.8", - "kalloc.16", "kalloc.32", - "kalloc.64", "kalloc.128", - "kalloc.256", "kalloc.512", - "kalloc.1024", "kalloc.2048", - "kalloc.4096", "kalloc.8192", - "kalloc.16384", "kalloc.32768" -}; - -/* - * Max number of elements per zone. zinit rounds things up correctly - * Doing things this way permits each zone to have a different maximum size - * based on need, rather than just guessing; it also - * means its patchable in case you're wrong! - */ -unsigned long k_zone_max[16] = { - 1024, /* 1 Byte */ - 1024, /* 2 Byte */ - 1024, /* 4 Byte */ - 1024, /* 8 Byte */ - 1024, /* 16 Byte */ - 4096, /* 32 Byte */ - 4096, /* 64 Byte */ - 4096, /* 128 Byte */ - 4096, /* 256 Byte */ - 1024, /* 512 Byte */ - 1024, /* 1024 Byte */ - 1024, /* 2048 Byte */ - 1024, /* 4096 Byte */ - 4096, /* 8192 Byte */ - 64, /* 16384 Byte */ - 64, /* 32768 Byte */ -}; - -/* - * Initialize the memory allocator. This should be called only - * once on a system wide basis (i.e. first processor to get here - * does the initialization). - * - * This initializes all of the zones. - */ - -#ifndef NDEBUG -static int kalloc_init_called; -#endif - -void kalloc_init() -{ - vm_offset_t min, max; - vm_size_t size; - register int i; - - assert (! kalloc_init_called); - - kalloc_map = kmem_suballoc(kernel_map, &min, &max, - kalloc_map_size, FALSE); - - /* - * Ensure that zones up to size 8192 bytes exist. - * This is desirable because messages are allocated - * with kalloc, and messages up through size 8192 are common. - */ - - if (PAGE_SIZE < 16*1024) - kalloc_max = 16*1024; - else - kalloc_max = PAGE_SIZE; - - /* - * Allocate a zone for each size we are going to handle. - * We specify non-paged memory. - */ - for (i = 0, size = 1; size < kalloc_max; i++, size <<= 1) { - if (size < MINSIZE) { - k_zone[i] = 0; - continue; - } - if (size == MINSIZE) { - first_k_zone = i; - } - k_zone[i] = zinit(size, 0, k_zone_max[i] * size, size, - size >= PAGE_SIZE ? ZONE_COLLECTABLE : 0, - k_zone_name[i]); - } - -#ifndef NDEBUG - kalloc_init_called = 1; -#endif -} - -vm_offset_t kalloc(size) - vm_size_t size; -{ - register int zindex; - register vm_size_t allocsize; - vm_offset_t addr; - - /* compute the size of the block that we will actually allocate */ - - assert (kalloc_init_called); - - allocsize = size; - if (size < kalloc_max) { - allocsize = MINSIZE; - zindex = first_k_zone; - while (allocsize < size) { - allocsize <<= 1; - zindex++; - } - } - - /* - * If our size is still small enough, check the queue for that size - * and allocate. - */ - - if (allocsize < kalloc_max) { - addr = zalloc(k_zone[zindex]); - } else { - if (kmem_alloc_wired(kalloc_map, &addr, allocsize) - != KERN_SUCCESS) - addr = 0; - } - return(addr); -} - -vm_offset_t kget(size) - vm_size_t size; -{ - register int zindex; - register vm_size_t allocsize; - vm_offset_t addr; - - assert (kalloc_init_called); - - /* compute the size of the block that we will actually allocate */ - - allocsize = size; - if (size < kalloc_max) { - allocsize = MINSIZE; - zindex = first_k_zone; - while (allocsize < size) { - allocsize <<= 1; - zindex++; - } - } - - /* - * If our size is still small enough, check the queue for that size - * and allocate. - */ - - if (allocsize < kalloc_max) { - addr = zget(k_zone[zindex]); - } else { - /* This will never work, so we might as well panic */ - panic("kget"); - } - return(addr); -} - -void -kfree(data, size) - vm_offset_t data; - vm_size_t size; -{ - register int zindex; - register vm_size_t freesize; - - assert (kalloc_init_called); - - freesize = size; - if (size < kalloc_max) { - freesize = MINSIZE; - zindex = first_k_zone; - while (freesize < size) { - freesize <<= 1; - zindex++; - } - } - - if (freesize < kalloc_max) { - zfree(k_zone[zindex], data); - } else { - kmem_free(kalloc_map, data, freesize); - } -} diff --git a/kern/kalloc.h b/kern/kalloc.h index a80f6dbd..004e3a6b 100644 --- a/kern/kalloc.h +++ b/kern/kalloc.h @@ -28,11 +28,9 @@ #define _KERN_KALLOC_H_ #include <mach/machine/vm_types.h> - -#define MINSIZE 16 +#include <vm/vm_types.h> extern vm_offset_t kalloc (vm_size_t size); -extern vm_offset_t kget (vm_size_t size); extern void kfree (vm_offset_t data, vm_size_t size); extern void kalloc_init (void); diff --git a/kern/list.h b/kern/list.h new file mode 100644 index 00000000..03414718 --- /dev/null +++ b/kern/list.h @@ -0,0 +1,359 @@ +/* + * Copyright (c) 2009, 2010 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * + * Simple doubly-linked list. + */ + +#ifndef _KERN_LIST_H +#define _KERN_LIST_H + +#include <stddef.h> +#include <sys/types.h> + +#define structof(ptr, type, member) \ + ((type *)((char *)ptr - offsetof(type, member))) + +/* + * Structure used as both head and node. + * + * This implementation relies on using the same type for both heads and nodes. + * + * It is recommended to encode the use of struct list variables in their names, + * e.g. struct list free_list or struct list free_objects is a good hint for a + * list of free objects. A declaration like struct list free_node clearly + * indicates it is used as part of a node in the free list. + */ +struct list { + struct list *prev; + struct list *next; +}; + +/* + * Static list initializer. + */ +#define LIST_INITIALIZER(list) { &(list), &(list) } + +/* + * Initialize a list. + */ +static inline void list_init(struct list *list) +{ + list->prev = list; + list->next = list; +} + +/* + * Initialize a list node. + * + * An entry is in no list when its node members point to NULL. + */ +static inline void list_node_init(struct list *node) +{ + node->prev = NULL; + node->next = NULL; +} + +/* + * Return true if node is in no list. + */ +static inline int list_node_unlinked(const struct list *node) +{ + return node->prev == NULL; +} + +/* + * Macro that evaluates to the address of the structure containing the + * given node based on the given type and member. + */ +#define list_entry(node, type, member) structof(node, type, member) + +/* + * Return the first node of a list. + */ +static inline struct list * list_first(const struct list *list) +{ + return list->next; +} + +/* + * Return the last node of a list. + */ +static inline struct list * list_last(const struct list *list) +{ + return list->prev; +} + +/* + * Return the node next to the given node. + */ +static inline struct list * list_next(const struct list *node) +{ + return node->next; +} + +/* + * Return the node previous to the given node. + */ +static inline struct list * list_prev(const struct list *node) +{ + return node->prev; +} + +/* + * Get the first entry of a list. + */ +#define list_first_entry(list, type, member) \ + list_entry(list_first(list), type, member) + +/* + * Get the last entry of a list. + */ +#define list_last_entry(list, type, member) \ + list_entry(list_last(list), type, member) + +/* + * Return true if node is after the last or before the first node of the list. + */ +static inline int list_end(const struct list *list, const struct list *node) +{ + return list == node; +} + +/* + * Return true if list is empty. + */ +static inline int list_empty(const struct list *list) +{ + return list == list->next; +} + +/* + * Return true if list contains exactly one node. + */ +static inline int list_singular(const struct list *list) +{ + return (list != list->next) && (list->next == list->prev); +} + +/* + * Split list2 by moving its nodes up to (but not including) the given + * node into list1 (which can be in a stale state). + * + * If list2 is empty, or node is list2 or list2->next, nothing is done. + */ +static inline void list_split(struct list *list1, struct list *list2, + struct list *node) +{ + if (list_empty(list2) || (list2->next == node) || list_end(list2, node)) + return; + + list1->next = list2->next; + list1->next->prev = list1; + + list1->prev = node->prev; + node->prev->next = list1; + + list2->next = node; + node->prev = list2; +} + +/* + * Append the nodes of list2 at the end of list1. + * + * After completion, list2 is stale. + */ +static inline void list_concat(struct list *list1, const struct list *list2) +{ + struct list *last1, *first2, *last2; + + if (list_empty(list2)) + return; + + last1 = list1->prev; + first2 = list2->next; + last2 = list2->prev; + + last1->next = first2; + first2->prev = last1; + + last2->next = list1; + list1->prev = last2; +} + +/* + * Set the new head of a list. + * + * This function is an optimized version of : + * list_init(&new_list); + * list_concat(&new_list, &old_list); + * + * After completion, old_head is stale. + */ +static inline void list_set_head(struct list *new_head, + const struct list *old_head) +{ + if (list_empty(old_head)) { + list_init(new_head); + return; + } + + *new_head = *old_head; + new_head->next->prev = new_head; + new_head->prev->next = new_head; +} + +/* + * Add a node between two nodes. + */ +static inline void list_add(struct list *prev, struct list *next, + struct list *node) +{ + next->prev = node; + node->next = next; + + prev->next = node; + node->prev = prev; +} + +/* + * Insert a node at the head of a list. + */ +static inline void list_insert(struct list *list, struct list *node) +{ + list_add(list, list->next, node); +} + +/* + * Insert a node at the tail of a list. + */ +static inline void list_insert_tail(struct list *list, struct list *node) +{ + list_add(list->prev, list, node); +} + +/* + * Insert a node before another node. + */ +static inline void list_insert_before(struct list *next, struct list *node) +{ + list_add(next->prev, next, node); +} + +/* + * Insert a node after another node. + */ +static inline void list_insert_after(struct list *prev, struct list *node) +{ + list_add(prev, prev->next, node); +} + +/* + * Remove a node from a list. + * + * After completion, the node is stale. + */ +static inline void list_remove(struct list *node) +{ + node->prev->next = node->next; + node->next->prev = node->prev; +} + +/* + * Forge a loop to process all nodes of a list. + * + * The node must not be altered during the loop. + */ +#define list_for_each(list, node) \ +for (node = list_first(list); \ + !list_end(list, node); \ + node = list_next(node)) + +/* + * Forge a loop to process all nodes of a list. + */ +#define list_for_each_safe(list, node, tmp) \ +for (node = list_first(list), tmp = list_next(node); \ + !list_end(list, node); \ + node = tmp, tmp = list_next(node)) + +/* + * Version of list_for_each() that processes nodes backward. + */ +#define list_for_each_reverse(list, node) \ +for (node = list_last(list); \ + !list_end(list, node); \ + node = list_prev(node)) + +/* + * Version of list_for_each_safe() that processes nodes backward. + */ +#define list_for_each_reverse_safe(list, node, tmp) \ +for (node = list_last(list), tmp = list_prev(node); \ + !list_end(list, node); \ + node = tmp, tmp = list_prev(node)) + +/* + * Forge a loop to process all entries of a list. + * + * The entry node must not be altered during the loop. + */ +#define list_for_each_entry(list, entry, member) \ +for (entry = list_entry(list_first(list), typeof(*entry), member); \ + !list_end(list, &entry->member); \ + entry = list_entry(list_next(&entry->member), typeof(*entry), \ + member)) + +/* + * Forge a loop to process all entries of a list. + */ +#define list_for_each_entry_safe(list, entry, tmp, member) \ +for (entry = list_entry(list_first(list), typeof(*entry), member), \ + tmp = list_entry(list_next(&entry->member), typeof(*entry), \ + member); \ + !list_end(list, &entry->member); \ + entry = tmp, tmp = list_entry(list_next(&entry->member), \ + typeof(*entry), member)) + +/* + * Version of list_for_each_entry() that processes entries backward. + */ +#define list_for_each_entry_reverse(list, entry, member) \ +for (entry = list_entry(list_last(list), typeof(*entry), member); \ + !list_end(list, &entry->member); \ + entry = list_entry(list_prev(&entry->member), typeof(*entry), \ + member)) + +/* + * Version of list_for_each_entry_safe() that processes entries backward. + */ +#define list_for_each_entry_reverse_safe(list, entry, tmp, member) \ +for (entry = list_entry(list_last(list), typeof(*entry), member), \ + tmp = list_entry(list_prev(&entry->member), typeof(*entry), \ + member); \ + !list_end(list, &entry->member); \ + entry = tmp, tmp = list_entry(list_prev(&entry->member), \ + typeof(*entry), member)) + +#endif /* _KERN_LIST_H */ diff --git a/kern/lock.c b/kern/lock.c index 0c61227a..44d4448e 100644 --- a/kern/lock.c +++ b/kern/lock.c @@ -162,8 +162,8 @@ void simple_lock( info = &simple_locks_info[simple_locks_taken++]; info->l = l; /* XXX we want our return address, if possible */ -#ifdef i386 - info->ra = *((unsigned int *)&l - 1); +#if defined(__i386__) + info->ra = *((unsigned long *)&l - 1); #endif /* i386 */ } @@ -180,8 +180,8 @@ boolean_t simple_lock_try( info = &simple_locks_info[simple_locks_taken++]; info->l = l; /* XXX we want our return address, if possible */ -#ifdef i386 - info->ra = *((unsigned int *)&l - 1); +#if defined(__i386__) + info->ra = *((unsigned long *)&l - 1); #endif /* i386 */ return TRUE; @@ -622,7 +622,7 @@ void db_show_all_slocks(void) info = &simple_locks_info[i]; db_printf("%d: ", i); db_printsym(info->l, DB_STGY_ANY); -#if i386 +#if defined(__i386__) db_printf(" locked by "); db_printsym(info->ra, DB_STGY_PROC); #endif diff --git a/kern/mach.srv b/kern/mach.srv index 3ed92593..b1cec606 100644 --- a/kern/mach.srv +++ b/kern/mach.srv @@ -37,6 +37,4 @@ #define thread_get_special_port act_get_special_port #endif /* MIGRATING_THREADS */ -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ - #include <mach/mach.defs> diff --git a/kern/mach_clock.c b/kern/mach_clock.c index 04a31153..edf87f07 100644 --- a/kern/mach_clock.c +++ b/kern/mach_clock.c @@ -47,7 +47,6 @@ #include <kern/host.h> #include <kern/lock.h> #include <kern/mach_clock.h> -#include <kern/mach_param.h> #include <kern/processor.h> #include <kern/queue.h> #include <kern/sched.h> @@ -91,7 +90,7 @@ int bigadj = 1000000; /* adjust 10*tickadj if adjustment * } while (secs != mtime->check_seconds); * to read the time correctly. (On a multiprocessor this assumes * that processors see each other's writes in the correct order. - * We may have to insert fence operations.) + * We have to insert write fence operations.) FIXME */ mapped_time_value_t *mtime = 0; @@ -100,7 +99,9 @@ mapped_time_value_t *mtime = 0; MACRO_BEGIN \ if (mtime != 0) { \ mtime->check_seconds = (time)->seconds; \ + asm volatile("":::"memory"); \ mtime->microseconds = (time)->microseconds; \ + asm volatile("":::"memory"); \ mtime->seconds = (time)->seconds; \ } \ MACRO_END @@ -512,7 +513,7 @@ int timeclose() /* * Compatibility for device drivers. * New code should use set_timeout/reset_timeout and private timers. - * These code can't use a zone to allocate timers, because + * These code can't use a cache to allocate timers, because * it can be called from interrupt handlers. */ diff --git a/kern/mach_clock.h b/kern/mach_clock.h index 2009c709..4e4e8ff1 100644 --- a/kern/mach_clock.h +++ b/kern/mach_clock.h @@ -37,10 +37,12 @@ extern int hz; /* number of ticks per second */ extern int tick; /* number of usec per tick */ +typedef void timer_func_t(void *); + /* Time-out element. */ struct timer_elt { queue_chain_t chain; /* chain in order of expiration */ - void (*fcn)(); /* function to call */ + timer_func_t *fcn; /* function to call */ void * param; /* with this parameter */ unsigned long ticks; /* expiration time, in ticks */ int set; /* unset | set | allocated */ @@ -99,7 +101,7 @@ extern kern_return_t host_adjust_time( extern void mapable_time_init (void); /* For public timer elements. */ -extern void timeout(void (*fcn)(void *), void *param, int interval); -extern boolean_t untimeout(void (*fcn)(void *), void *param); +extern void timeout(timer_func_t *fcn, void *param, int interval); +extern boolean_t untimeout(timer_func_t *fcn, void *param); #endif /* _KERN_MACH_CLOCK_H_ */ diff --git a/kern/mach_factor.c b/kern/mach_factor.c index 6fbda83d..558c4a06 100644 --- a/kern/mach_factor.c +++ b/kern/mach_factor.c @@ -36,10 +36,8 @@ #include <kern/mach_clock.h> #include <kern/sched.h> #include <kern/processor.h> -#if MACH_KERNEL #include <mach/kern_return.h> #include <mach/port.h> -#endif /* MACH_KERNEL */ #include "mach_factor.h" diff --git a/kern/mach_host.srv b/kern/mach_host.srv index 30d78db2..a18ab1cf 100644 --- a/kern/mach_host.srv +++ b/kern/mach_host.srv @@ -23,8 +23,6 @@ #define KERNEL_SERVER 1 -simport <kern/compat_xxx_defs.h>; /* for obsolete routines */ - #ifdef MIGRATING_THREADS #define thread_assign act_thread_assign #define thread_assign_default act_thread_assign_default diff --git a/kern/mach_param.h b/kern/mach_param.h deleted file mode 100644 index 10376d81..00000000 --- a/kern/mach_param.h +++ /dev/null @@ -1,67 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University. - * Copyright (c) 1993,1994 The University of Utah and - * the Computer Systems Laboratory (CSL). - * All rights reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON, THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF - * THIS SOFTWARE IN ITS "AS IS" CONDITION, AND DISCLAIM ANY LIABILITY - * OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF - * THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * File: kern/mach_param.h - * Author: Avadis Tevanian, Jr., Michael Wayne Young - * Date: 1986 - * - * Mach system sizing parameters - * - */ - -#ifndef _KERN_MACH_PARAM_H_ -#define _KERN_MACH_PARAM_H_ - -#define THREAD_MAX 1024 /* Max number of threads */ -#define THREAD_CHUNK 64 /* Allocation chunk */ - -#define TASK_MAX 1024 /* Max number of tasks */ -#define TASK_CHUNK 64 /* Allocation chunk */ - -#define ACT_MAX 1024 /* Max number of acts */ -#define ACT_CHUNK 64 /* Allocation chunk */ - -#define ACTPOOL_MAX 1024 -#define ACTPOOL_CHUNK 64 - -#define PORT_MAX ((TASK_MAX * 3 + THREAD_MAX) /* kernel */ \ - + (THREAD_MAX * 2) /* user */ \ - + 40000) /* slop for objects */ - /* Number of ports, system-wide */ - -#define SET_MAX (TASK_MAX + THREAD_MAX + 200) - /* Max number of port sets */ - -#define ITE_MAX (1 << 16) /* Max number of splay tree entries */ - -#define SPACE_MAX (TASK_MAX + 5) /* Max number of IPC spaces */ - -#define IMAR_MAX (1 << 10) /* Max number of msg-accepted reqs */ - -#endif /* _KERN_MACH_PARAM_H_ */ diff --git a/kern/machine.c b/kern/machine.c index bcf394c3..c2a19b99 100644 --- a/kern/machine.c +++ b/kern/machine.c @@ -67,60 +67,6 @@ queue_head_t action_queue; /* assign/shutdown queue */ decl_simple_lock_data(,action_lock); /* - * xxx_host_info: - * - * Return the host_info structure. This routine is exported to the - * user level. - */ -kern_return_t xxx_host_info(task, info) - task_t task; - machine_info_t info; -{ -#ifdef lint - task++; -#endif /* lint */ - *info = machine_info; - return(KERN_SUCCESS); -} - -/* - * xxx_slot_info: - * - * Return the slot_info structure for the specified slot. This routine - * is exported to the user level. - */ -kern_return_t xxx_slot_info(task, slot, info) - task_t task; - int slot; - machine_slot_t info; -{ -#ifdef lint - task++; -#endif /* lint */ - if ((slot < 0) || (slot >= NCPUS)) - return(KERN_INVALID_ARGUMENT); - *info = machine_slot[slot]; - return(KERN_SUCCESS); -} - -/* - * xxx_cpu_control: - * - * Support for user control of cpus. The user indicates which cpu - * he is interested in, and whether or not that cpu should be running. - */ -kern_return_t xxx_cpu_control(task, cpu, runnable) - task_t task; - int cpu; - boolean_t runnable; -{ -#ifdef lint - task++; cpu++; runnable++; -#endif /* lint */ - return(KERN_FAILURE); -} - -/* * cpu_up: * * Flag specified cpu as up and running. Called when a processor comes diff --git a/kern/macro_help.h b/kern/macro_help.h index e13b01d9..a3d156b7 100644 --- a/kern/macro_help.h +++ b/kern/macro_help.h @@ -45,8 +45,8 @@ boolean_t ALWAYS; #define ALWAYS TRUE #endif /* lint */ -#define MACRO_BEGIN do { -#define MACRO_END } while (NEVER) +#define MACRO_BEGIN ({ +#define MACRO_END }) #define MACRO_RETURN if (ALWAYS) return diff --git a/kern/pc_sample.c b/kern/pc_sample.c index c82707b2..57002581 100644 --- a/kern/pc_sample.c +++ b/kern/pc_sample.c @@ -31,6 +31,7 @@ #include <mach/std_types.h> /* pointer_t */ #include <mach/pc_sample.h> #include <machine/trap.h> +#include <kern/kalloc.h> #include <kern/host.h> #include <kern/thread.h> #include <kern/pc_sample.h> @@ -56,7 +57,7 @@ void take_pc_sample( pc = interrupted_pc(t); cp->seqno++; sample = &((sampled_pc_t *)cp->buffer)[cp->seqno % MAX_PC_SAMPLES]; - sample->id = (natural_t)t; + sample->id = (vm_offset_t)t; sample->pc = pc; sample->sampletype = flavor; } diff --git a/kern/printf.c b/kern/printf.c index 88a527ba..a3a771d0 100644 --- a/kern/printf.c +++ b/kern/printf.c @@ -39,7 +39,10 @@ * FILE *fd; * char *format; * { + * va_list listp; + * va_start(listp, fmt); * _doprnt(format, &args, fd); + * va_end(listp); * } * * would suffice. (This example does not handle the fprintf's "return @@ -165,7 +168,7 @@ void printf_init(void) void _doprnt( register const char *fmt, - va_list *argp, + va_list argp, /* character output routine */ void (*putc)( char, vm_offset_t), int radix, /* default radix - for '%r' */ @@ -248,7 +251,7 @@ void _doprnt( } } else if (c == '*') { - length = va_arg(*argp, int); + length = va_arg(argp, int); c = *++fmt; if (length < 0) { ladjust = !ladjust; @@ -266,7 +269,7 @@ void _doprnt( } } else if (c == '*') { - prec = va_arg(*argp, int); + prec = va_arg(argp, int); c = *++fmt; } } @@ -284,8 +287,8 @@ void _doprnt( boolean_t any; register int i; - u = va_arg(*argp, unsigned long); - p = va_arg(*argp, char *); + u = va_arg(argp, unsigned long); + p = va_arg(argp, char *); base = *p++; printnum(u, base, putc, putc_arg); @@ -333,7 +336,7 @@ void _doprnt( } case 'c': - c = va_arg(*argp, int); + c = va_arg(argp, int); (*putc)(c, putc_arg); break; @@ -345,7 +348,7 @@ void _doprnt( if (prec == -1) prec = 0x7fffffff; /* MAXINT */ - p = va_arg(*argp, char *); + p = va_arg(argp, char *); if (p == (char *)0) p = ""; @@ -428,7 +431,7 @@ void _doprnt( goto print_unsigned; print_signed: - n = va_arg(*argp, long); + n = va_arg(argp, long); if (n >= 0) { u = n; sign_char = plus_sign; @@ -440,7 +443,7 @@ void _doprnt( goto print_num; print_unsigned: - u = va_arg(*argp, unsigned long); + u = va_arg(argp, unsigned long); goto print_num; print_num: @@ -514,7 +517,7 @@ void _doprnt( int vprintf(const char *fmt, va_list listp) { - _doprnt(fmt, &listp, (void (*)( char, vm_offset_t)) cnputc, 16, 0); + _doprnt(fmt, listp, (void (*)( char, vm_offset_t)) cnputc, 16, 0); return 0; } @@ -550,7 +553,7 @@ void iprintf(const char *fmt, ...) } } va_start(listp, fmt); - _doprnt(fmt, &listp, (void (*)( char, vm_offset_t)) cnputc, 16, 0); + _doprnt(fmt, listp, (void (*)( char, vm_offset_t)) cnputc, 16, 0); va_end(listp); } @@ -577,7 +580,7 @@ sprintf(char *buf, const char *fmt, ...) char *start = buf; va_start(listp, fmt); - _doprnt(fmt, &listp, sputc, 16, (vm_offset_t)&buf); + _doprnt(fmt, listp, sputc, 16, (vm_offset_t)&buf); va_end(listp); *buf = 0; @@ -601,12 +604,12 @@ snputc(char c, vm_offset_t arg) } int -vsnprintf(char *buf, int size, const char *fmt, va_list args) +vsnprintf(char *buf, size_t size, const char *fmt, va_list args) { struct vsnprintf_cookie cookie = { .buf = buf, .index = 0, .max_len = size }; - _doprnt (fmt, &args, snputc, 16, (vm_offset_t)&cookie); + _doprnt (fmt, args, snputc, 16, (vm_offset_t)&cookie); cookie.buf[cookie.index] = '\0'; return cookie.index; diff --git a/kern/printf.h b/kern/printf.h index c5effe5d..8b4e7606 100644 --- a/kern/printf.h +++ b/kern/printf.h @@ -30,7 +30,7 @@ extern void printf_init (void); extern void _doprnt (const char *fmt, - va_list *argp, + va_list argp, void (*putc)(char, vm_offset_t), int radix, vm_offset_t putc_arg); @@ -40,6 +40,7 @@ extern void printnum (unsigned long u, int base, vm_offset_t putc_arg); extern int sprintf (char *buf, const char *fmt, ...); +extern int vsnprintf (char *buf, size_t size, const char *fmt, va_list args); extern int printf (const char *fmt, ...); diff --git a/kern/priority.c b/kern/priority.c index feddd8ee..17541b8b 100644 --- a/kern/priority.c +++ b/kern/priority.c @@ -39,7 +39,6 @@ #include <mach/machine.h> #include <kern/host.h> #include <kern/mach_clock.h> -#include <kern/mach_param.h> #include <kern/sched.h> #include <kern/sched_prim.h> #include <kern/thread.h> diff --git a/kern/processor.c b/kern/processor.c index 718ff3ad..19868609 100644 --- a/kern/processor.c +++ b/kern/processor.c @@ -35,6 +35,7 @@ #include <mach/vm_param.h> #include <kern/cpu_number.h> #include <kern/debug.h> +#include <kern/kalloc.h> #include <kern/lock.h> #include <kern/host.h> #include <kern/ipc_tt.h> @@ -46,8 +47,8 @@ #include <ipc/ipc_port.h> #if MACH_HOST -#include <kern/zalloc.h> -zone_t pset_zone; +#include <kern/slab.h> +struct kmem_cache pset_cache; #endif /* MACH_HOST */ @@ -112,10 +113,10 @@ void pset_sys_init(void) register processor_t processor; /* - * Allocate the zone for processor sets. + * Allocate the cache for processor sets. */ - pset_zone = zinit(sizeof(struct processor_set), 0, 128*PAGE_SIZE, - PAGE_SIZE, 0, "processor sets"); + kmem_cache_init(&pset_cache, "processor_set", + sizeof(struct processor_set), 0, NULL, NULL, NULL, 0); /* * Give each processor a control port. @@ -394,7 +395,7 @@ void pset_deallocate( /* * That's it, free data structure. */ - zfree(pset_zone, (vm_offset_t)pset); + kmem_cache_free(&pset_cache, (vm_offset_t)pset); #endif /* MACH_HOST */ } @@ -538,7 +539,7 @@ processor_set_create( if (host == HOST_NULL) return KERN_INVALID_ARGUMENT; - pset = (processor_set_t) zalloc(pset_zone); + pset = (processor_set_t) kmem_cache_alloc(&pset_cache); pset_init(pset); pset_reference(pset); /* for new_set out argument */ pset_reference(pset); /* for new_name out argument */ diff --git a/kern/rbtree.c b/kern/rbtree.c new file mode 100644 index 00000000..0f5eb9ae --- /dev/null +++ b/kern/rbtree.c @@ -0,0 +1,483 @@ +/* + * Copyright (c) 2010, 2012 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include <kern/assert.h> +#include <kern/rbtree.h> +#include <kern/rbtree_i.h> +#include <sys/types.h> + +#define unlikely(expr) __builtin_expect(!!(expr), 0) + +/* + * Return the index of a node in the children array of its parent. + * + * The parent parameter must not be null, and must be the parent of the + * given node. + */ +static inline int rbtree_index(const struct rbtree_node *node, + const struct rbtree_node *parent) +{ + assert(parent != NULL); + assert((node == NULL) || (rbtree_parent(node) == parent)); + + if (parent->children[RBTREE_LEFT] == node) + return RBTREE_LEFT; + + assert(parent->children[RBTREE_RIGHT] == node); + + return RBTREE_RIGHT; +} + +/* + * Return the color of a node. + */ +static inline int rbtree_color(const struct rbtree_node *node) +{ + return node->parent & RBTREE_COLOR_MASK; +} + +/* + * Return true if the node is red. + */ +static inline int rbtree_is_red(const struct rbtree_node *node) +{ + return rbtree_color(node) == RBTREE_COLOR_RED; +} + +/* + * Return true if the node is black. + */ +static inline int rbtree_is_black(const struct rbtree_node *node) +{ + return rbtree_color(node) == RBTREE_COLOR_BLACK; +} + +/* + * Set the parent of a node, retaining its current color. + */ +static inline void rbtree_set_parent(struct rbtree_node *node, + struct rbtree_node *parent) +{ + assert(rbtree_check_alignment(node)); + assert(rbtree_check_alignment(parent)); + + node->parent = (unsigned long)parent | (node->parent & RBTREE_COLOR_MASK); +} + +/* + * Set the color of a node, retaining its current parent. + */ +static inline void rbtree_set_color(struct rbtree_node *node, int color) +{ + assert((color & ~RBTREE_COLOR_MASK) == 0); + node->parent = (node->parent & RBTREE_PARENT_MASK) | color; +} + +/* + * Set the color of a node to red, retaining its current parent. + */ +static inline void rbtree_set_red(struct rbtree_node *node) +{ + rbtree_set_color(node, RBTREE_COLOR_RED); +} + +/* + * Set the color of a node to black, retaining its current parent. + */ +static inline void rbtree_set_black(struct rbtree_node *node) +{ + rbtree_set_color(node, RBTREE_COLOR_BLACK); +} + +/* + * Perform a tree rotation, rooted at the given node. + * + * The direction parameter defines the rotation direction and is either + * RBTREE_LEFT or RBTREE_RIGHT. + */ +static void rbtree_rotate(struct rbtree *tree, struct rbtree_node *node, + int direction) +{ + struct rbtree_node *parent, *rnode; + int left, right; + + left = direction; + right = 1 - left; + parent = rbtree_parent(node); + rnode = node->children[right]; + + node->children[right] = rnode->children[left]; + + if (rnode->children[left] != NULL) + rbtree_set_parent(rnode->children[left], node); + + rnode->children[left] = node; + rbtree_set_parent(rnode, parent); + + if (unlikely(parent == NULL)) + tree->root = rnode; + else + parent->children[rbtree_index(node, parent)] = rnode; + + rbtree_set_parent(node, rnode); +} + +void rbtree_insert_rebalance(struct rbtree *tree, struct rbtree_node *parent, + int index, struct rbtree_node *node) +{ + struct rbtree_node *grand_parent, *uncle, *tmp; + int left, right; + + assert(rbtree_check_alignment(parent)); + assert(rbtree_check_alignment(node)); + + node->parent = (unsigned long)parent | RBTREE_COLOR_RED; + node->children[RBTREE_LEFT] = NULL; + node->children[RBTREE_RIGHT] = NULL; + + if (unlikely(parent == NULL)) + tree->root = node; + else + parent->children[index] = node; + + for (;;) { + if (parent == NULL) { + rbtree_set_black(node); + break; + } + + if (rbtree_is_black(parent)) + break; + + grand_parent = rbtree_parent(parent); + assert(grand_parent != NULL); + + left = rbtree_index(parent, grand_parent); + right = 1 - left; + + uncle = grand_parent->children[right]; + + /* + * Uncle is red. Flip colors and repeat at grand parent. + */ + if ((uncle != NULL) && rbtree_is_red(uncle)) { + rbtree_set_black(uncle); + rbtree_set_black(parent); + rbtree_set_red(grand_parent); + node = grand_parent; + parent = rbtree_parent(node); + continue; + } + + /* + * Node is the right child of its parent. Rotate left at parent. + */ + if (parent->children[right] == node) { + rbtree_rotate(tree, parent, left); + tmp = node; + node = parent; + parent = tmp; + } + + /* + * Node is the left child of its parent. Handle colors, rotate right + * at grand parent, and leave. + */ + rbtree_set_black(parent); + rbtree_set_red(grand_parent); + rbtree_rotate(tree, grand_parent, right); + break; + } + + assert(rbtree_is_black(tree->root)); +} + +void rbtree_remove(struct rbtree *tree, struct rbtree_node *node) +{ + struct rbtree_node *child, *parent, *brother; + int color, left, right; + + if (node->children[RBTREE_LEFT] == NULL) + child = node->children[RBTREE_RIGHT]; + else if (node->children[RBTREE_RIGHT] == NULL) + child = node->children[RBTREE_LEFT]; + else { + struct rbtree_node *successor; + + /* + * Two-children case: replace the node with its successor. + */ + + successor = node->children[RBTREE_RIGHT]; + + while (successor->children[RBTREE_LEFT] != NULL) + successor = successor->children[RBTREE_LEFT]; + + color = rbtree_color(successor); + child = successor->children[RBTREE_RIGHT]; + parent = rbtree_parent(node); + + if (unlikely(parent == NULL)) + tree->root = successor; + else + parent->children[rbtree_index(node, parent)] = successor; + + parent = rbtree_parent(successor); + + /* + * Set parent directly to keep the original color. + */ + successor->parent = node->parent; + successor->children[RBTREE_LEFT] = node->children[RBTREE_LEFT]; + rbtree_set_parent(successor->children[RBTREE_LEFT], successor); + + if (node == parent) + parent = successor; + else { + successor->children[RBTREE_RIGHT] = node->children[RBTREE_RIGHT]; + rbtree_set_parent(successor->children[RBTREE_RIGHT], successor); + parent->children[RBTREE_LEFT] = child; + + if (child != NULL) + rbtree_set_parent(child, parent); + } + + goto update_color; + } + + /* + * Node has at most one child. + */ + + color = rbtree_color(node); + parent = rbtree_parent(node); + + if (child != NULL) + rbtree_set_parent(child, parent); + + if (unlikely(parent == NULL)) + tree->root = child; + else + parent->children[rbtree_index(node, parent)] = child; + + /* + * The node has been removed, update the colors. The child pointer can + * be null, in which case it is considered a black leaf. + */ +update_color: + if (color == RBTREE_COLOR_RED) + return; + + for (;;) { + if ((child != NULL) && rbtree_is_red(child)) { + rbtree_set_black(child); + break; + } + + if (parent == NULL) + break; + + left = rbtree_index(child, parent); + right = 1 - left; + + brother = parent->children[right]; + + /* + * Brother is red. Recolor and rotate left at parent so that brother + * becomes black. + */ + if (rbtree_is_red(brother)) { + rbtree_set_black(brother); + rbtree_set_red(parent); + rbtree_rotate(tree, parent, left); + brother = parent->children[right]; + } + + /* + * Brother has no red child. Recolor and repeat at parent. + */ + if (((brother->children[RBTREE_LEFT] == NULL) + || rbtree_is_black(brother->children[RBTREE_LEFT])) + && ((brother->children[RBTREE_RIGHT] == NULL) + || rbtree_is_black(brother->children[RBTREE_RIGHT]))) { + rbtree_set_red(brother); + child = parent; + parent = rbtree_parent(child); + continue; + } + + /* + * Brother's right child is black. Recolor and rotate right at brother. + */ + if ((brother->children[right] == NULL) + || rbtree_is_black(brother->children[right])) { + rbtree_set_black(brother->children[left]); + rbtree_set_red(brother); + rbtree_rotate(tree, brother, right); + brother = parent->children[right]; + } + + /* + * Brother's left child is black. Exchange parent and brother colors + * (we already know brother is black), set brother's right child black, + * rotate left at parent and leave. + */ + rbtree_set_color(brother, rbtree_color(parent)); + rbtree_set_black(parent); + rbtree_set_black(brother->children[right]); + rbtree_rotate(tree, parent, left); + break; + } + + assert((tree->root == NULL) || rbtree_is_black(tree->root)); +} + +struct rbtree_node * rbtree_nearest(struct rbtree_node *parent, int index, + int direction) +{ + assert(rbtree_check_index(direction)); + + if (parent == NULL) + return NULL; + + assert(rbtree_check_index(index)); + + if (index != direction) + return parent; + + return rbtree_walk(parent, direction); +} + +struct rbtree_node * rbtree_firstlast(const struct rbtree *tree, int direction) +{ + struct rbtree_node *prev, *cur; + + assert(rbtree_check_index(direction)); + + prev = NULL; + + for (cur = tree->root; cur != NULL; cur = cur->children[direction]) + prev = cur; + + return prev; +} + +struct rbtree_node * rbtree_walk(struct rbtree_node *node, int direction) +{ + int left, right; + + assert(rbtree_check_index(direction)); + + left = direction; + right = 1 - left; + + if (node == NULL) + return NULL; + + if (node->children[left] != NULL) { + node = node->children[left]; + + while (node->children[right] != NULL) + node = node->children[right]; + } else { + struct rbtree_node *parent; + int index; + + for (;;) { + parent = rbtree_parent(node); + + if (parent == NULL) + return NULL; + + index = rbtree_index(node, parent); + node = parent; + + if (index == right) + break; + } + } + + return node; +} + +/* + * Return the left-most deepest child node of the given node. + */ +static struct rbtree_node * rbtree_find_deepest(struct rbtree_node *node) +{ + struct rbtree_node *parent; + + assert(node != NULL); + + for (;;) { + parent = node; + node = node->children[RBTREE_LEFT]; + + if (node == NULL) { + node = parent->children[RBTREE_RIGHT]; + + if (node == NULL) + return parent; + } + } +} + +struct rbtree_node * rbtree_postwalk_deepest(const struct rbtree *tree) +{ + struct rbtree_node *node; + + node = tree->root; + + if (node == NULL) + return NULL; + + return rbtree_find_deepest(node); +} + +struct rbtree_node * rbtree_postwalk_unlink(struct rbtree_node *node) +{ + struct rbtree_node *parent; + int index; + + if (node == NULL) + return NULL; + + assert(node->children[RBTREE_LEFT] == NULL); + assert(node->children[RBTREE_RIGHT] == NULL); + + parent = rbtree_parent(node); + + if (parent == NULL) + return NULL; + + index = rbtree_index(node, parent); + parent->children[index] = NULL; + node = parent->children[RBTREE_RIGHT]; + + if (node == NULL) + return parent; + + return rbtree_find_deepest(node); +} diff --git a/kern/rbtree.h b/kern/rbtree.h new file mode 100644 index 00000000..5a65d1ef --- /dev/null +++ b/kern/rbtree.h @@ -0,0 +1,310 @@ +/* + * Copyright (c) 2010, 2011 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * + * Red-black tree. + */ + +#ifndef _KERN_RBTREE_H +#define _KERN_RBTREE_H + +#include <stddef.h> +#include <kern/assert.h> +#include <kern/macro_help.h> +#include <kern/rbtree.h> +#include <sys/types.h> + +#define structof(ptr, type, member) \ + ((type *)((char *)ptr - offsetof(type, member))) + +/* + * Indexes of the left and right nodes in the children array of a node. + */ +#define RBTREE_LEFT 0 +#define RBTREE_RIGHT 1 + +/* + * Red-black node. + */ +struct rbtree_node; + +/* + * Red-black tree. + */ +struct rbtree; + +/* + * Static tree initializer. + */ +#define RBTREE_INITIALIZER { NULL } + +#include "rbtree_i.h" + +/* + * Initialize a tree. + */ +static inline void rbtree_init(struct rbtree *tree) +{ + tree->root = NULL; +} + +/* + * Initialize a node. + * + * A node is in no tree when its parent points to itself. + */ +static inline void rbtree_node_init(struct rbtree_node *node) +{ + assert(rbtree_check_alignment(node)); + + node->parent = (unsigned long)node | RBTREE_COLOR_RED; + node->children[RBTREE_LEFT] = NULL; + node->children[RBTREE_RIGHT] = NULL; +} + +/* + * Return true if node is in no tree. + */ +static inline int rbtree_node_unlinked(const struct rbtree_node *node) +{ + return rbtree_parent(node) == node; +} + +/* + * Macro that evaluates to the address of the structure containing the + * given node based on the given type and member. + */ +#define rbtree_entry(node, type, member) structof(node, type, member) + +/* + * Return true if tree is empty. + */ +static inline int rbtree_empty(const struct rbtree *tree) +{ + return tree->root == NULL; +} + +/* + * Look up a node in a tree. + * + * Note that implementing the lookup algorithm as a macro gives two benefits: + * First, it avoids the overhead of a callback function. Next, the type of the + * cmp_fn parameter isn't rigid. The only guarantee offered by this + * implementation is that the key parameter is the first parameter given to + * cmp_fn. This way, users can pass only the value they need for comparison + * instead of e.g. allocating a full structure on the stack. + * + * See rbtree_insert(). + */ +#define rbtree_lookup(tree, key, cmp_fn) \ +MACRO_BEGIN \ + struct rbtree_node *___cur; \ + int ___diff; \ + \ + ___cur = (tree)->root; \ + \ + while (___cur != NULL) { \ + ___diff = cmp_fn(key, ___cur); \ + \ + if (___diff == 0) \ + break; \ + \ + ___cur = ___cur->children[rbtree_d2i(___diff)]; \ + } \ + \ + ___cur; \ +MACRO_END + +/* + * Look up a node or one of its nearest nodes in a tree. + * + * This macro essentially acts as rbtree_lookup() but if no entry matched + * the key, an additional step is performed to obtain the next or previous + * node, depending on the direction (left or right). + * + * The constraints that apply to the key parameter are the same as for + * rbtree_lookup(). + */ +#define rbtree_lookup_nearest(tree, key, cmp_fn, dir) \ +MACRO_BEGIN \ + struct rbtree_node *___cur, *___prev; \ + int ___diff, ___index; \ + \ + ___prev = NULL; \ + ___index = -1; \ + ___cur = (tree)->root; \ + \ + while (___cur != NULL) { \ + ___diff = cmp_fn(key, ___cur); \ + \ + if (___diff == 0) \ + break; \ + \ + ___prev = ___cur; \ + ___index = rbtree_d2i(___diff); \ + ___cur = ___cur->children[___index]; \ + } \ + \ + if (___cur == NULL) \ + ___cur = rbtree_nearest(___prev, ___index, dir); \ + \ + ___cur; \ +MACRO_END + +/* + * Insert a node in a tree. + * + * This macro performs a standard lookup to obtain the insertion point of + * the given node in the tree (it is assumed that the inserted node never + * compares equal to any other entry in the tree) and links the node. It + * then It then checks red-black rules violations, and rebalances the tree + * if necessary. + * + * Unlike rbtree_lookup(), the cmp_fn parameter must compare two complete + * entries, so it is suggested to use two different comparison inline + * functions, such as myobj_cmp_lookup() and myobj_cmp_insert(). There is no + * guarantee about the order of the nodes given to the comparison function. + * + * See rbtree_lookup(). + */ +#define rbtree_insert(tree, node, cmp_fn) \ +MACRO_BEGIN \ + struct rbtree_node *___cur, *___prev; \ + int ___diff, ___index; \ + \ + ___prev = NULL; \ + ___index = -1; \ + ___cur = (tree)->root; \ + \ + while (___cur != NULL) { \ + ___diff = cmp_fn(node, ___cur); \ + assert(___diff != 0); \ + ___prev = ___cur; \ + ___index = rbtree_d2i(___diff); \ + ___cur = ___cur->children[___index]; \ + } \ + \ + rbtree_insert_rebalance(tree, ___prev, ___index, node); \ +MACRO_END + +/* + * Look up a node/slot pair in a tree. + * + * This macro essentially acts as rbtree_lookup() but in addition to a node, + * it also returns a slot, which identifies an insertion point in the tree. + * If the returned node is null, the slot can be used by rbtree_insert_slot() + * to insert without the overhead of an additional lookup. The slot is a + * simple unsigned long integer. + * + * The constraints that apply to the key parameter are the same as for + * rbtree_lookup(). + */ +#define rbtree_lookup_slot(tree, key, cmp_fn, slot) \ +MACRO_BEGIN \ + struct rbtree_node *___cur, *___prev; \ + int ___diff, ___index; \ + \ + ___prev = NULL; \ + ___index = 0; \ + ___cur = (tree)->root; \ + \ + while (___cur != NULL) { \ + ___diff = cmp_fn(key, ___cur); \ + \ + if (___diff == 0) \ + break; \ + \ + ___prev = ___cur; \ + ___index = rbtree_d2i(___diff); \ + ___cur = ___cur->children[___index]; \ + } \ + \ + (slot) = rbtree_slot(___prev, ___index); \ + ___cur; \ +MACRO_END + +/* + * Insert a node at an insertion point in a tree. + * + * This macro essentially acts as rbtree_insert() except that it doesn't + * obtain the insertion point with a standard lookup. The insertion point + * is obtained by calling rbtree_lookup_slot(). In addition, the new node + * must not compare equal to an existing node in the tree (i.e. the slot + * must denote a null node). + */ +static inline void +rbtree_insert_slot(struct rbtree *tree, unsigned long slot, + struct rbtree_node *node) +{ + struct rbtree_node *parent; + int index; + + parent = rbtree_slot_parent(slot); + index = rbtree_slot_index(slot); + rbtree_insert_rebalance(tree, parent, index, node); +} + +/* + * Remove a node from a tree. + * + * After completion, the node is stale. + */ +void rbtree_remove(struct rbtree *tree, struct rbtree_node *node); + +/* + * Return the first node of a tree. + */ +#define rbtree_first(tree) rbtree_firstlast(tree, RBTREE_LEFT) + +/* + * Return the last node of a tree. + */ +#define rbtree_last(tree) rbtree_firstlast(tree, RBTREE_RIGHT) + +/* + * Return the node previous to the given node. + */ +#define rbtree_prev(node) rbtree_walk(node, RBTREE_LEFT) + +/* + * Return the node next to the given node. + */ +#define rbtree_next(node) rbtree_walk(node, RBTREE_RIGHT) + +/* + * Forge a loop to process all nodes of a tree, removing them when visited. + * + * This macro can only be used to destroy a tree, so that the resources used + * by the entries can be released by the user. It basically removes all nodes + * without doing any color checking. + * + * After completion, all nodes and the tree root member are stale. + */ +#define rbtree_for_each_remove(tree, node, tmp) \ +for (node = rbtree_postwalk_deepest(tree), \ + tmp = rbtree_postwalk_unlink(node); \ + node != NULL; \ + node = tmp, tmp = rbtree_postwalk_unlink(node)) \ + +#endif /* _KERN_RBTREE_H */ diff --git a/kern/rbtree_i.h b/kern/rbtree_i.h new file mode 100644 index 00000000..69dfb9d0 --- /dev/null +++ b/kern/rbtree_i.h @@ -0,0 +1,186 @@ +/* + * Copyright (c) 2010, 2011 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef _KERN_RBTREE_I_H +#define _KERN_RBTREE_I_H + +#include <kern/assert.h> + +/* + * Red-black node structure. + * + * To reduce the number of branches and the instruction cache footprint, + * the left and right child pointers are stored in an array, and the symmetry + * of most tree operations is exploited by using left/right variables when + * referring to children. + * + * In addition, this implementation assumes that all nodes are 4-byte aligned, + * so that the least significant bit of the parent member can be used to store + * the color of the node. This is true for all modern 32 and 64 bits + * architectures, as long as the nodes aren't embedded in structures with + * special alignment constraints such as member packing. + */ +struct rbtree_node { + unsigned long parent; + struct rbtree_node *children[2]; +}; + +/* + * Red-black tree structure. + */ +struct rbtree { + struct rbtree_node *root; +}; + +/* + * Masks applied on the parent member of a node to obtain either the + * color or the parent address. + */ +#define RBTREE_COLOR_MASK 0x1UL +#define RBTREE_PARENT_MASK (~0x3UL) + +/* + * Node colors. + */ +#define RBTREE_COLOR_RED 0 +#define RBTREE_COLOR_BLACK 1 + +/* + * Masks applied on slots to obtain either the child index or the parent + * address. + */ +#define RBTREE_SLOT_INDEX_MASK 0x1UL +#define RBTREE_SLOT_PARENT_MASK (~RBTREE_SLOT_INDEX_MASK) + +/* + * Return true if the given pointer is suitably aligned. + */ +static inline int rbtree_check_alignment(const struct rbtree_node *node) +{ + return ((unsigned long)node & (~RBTREE_PARENT_MASK)) == 0; +} + +/* + * Return true if the given index is a valid child index. + */ +static inline int rbtree_check_index(int index) +{ + return index == (index & 1); +} + +/* + * Convert the result of a comparison into an index in the children array + * (0 or 1). + * + * This function is mostly used when looking up a node. + */ +static inline int rbtree_d2i(int diff) +{ + return !(diff <= 0); +} + +/* + * Return the parent of a node. + */ +static inline struct rbtree_node * rbtree_parent(const struct rbtree_node *node) +{ + return (struct rbtree_node *)(node->parent & RBTREE_PARENT_MASK); +} + +/* + * Translate an insertion point into a slot. + */ +static inline unsigned long rbtree_slot(struct rbtree_node *parent, int index) +{ + assert(rbtree_check_alignment(parent)); + assert(rbtree_check_index(index)); + return (unsigned long)parent | index; +} + +/* + * Extract the parent address from a slot. + */ +static inline struct rbtree_node * rbtree_slot_parent(unsigned long slot) +{ + return (struct rbtree_node *)(slot & RBTREE_SLOT_PARENT_MASK); +} + +/* + * Extract the index from a slot. + */ +static inline int rbtree_slot_index(unsigned long slot) +{ + return slot & RBTREE_SLOT_INDEX_MASK; +} + +/* + * Insert a node in a tree, rebalancing it if necessary. + * + * The index parameter is the index in the children array of the parent where + * the new node is to be inserted. It is ignored if the parent is null. + * + * This function is intended to be used by the rbtree_insert() macro only. + */ +void rbtree_insert_rebalance(struct rbtree *tree, struct rbtree_node *parent, + int index, struct rbtree_node *node); + +/* + * Return the previous or next node relative to a location in a tree. + * + * The parent and index parameters define the location, which can be empty. + * The direction parameter is either RBTREE_LEFT (to obtain the previous + * node) or RBTREE_RIGHT (to obtain the next one). + */ +struct rbtree_node * rbtree_nearest(struct rbtree_node *parent, int index, + int direction); + +/* + * Return the first or last node of a tree. + * + * The direction parameter is either RBTREE_LEFT (to obtain the first node) + * or RBTREE_RIGHT (to obtain the last one). + */ +struct rbtree_node * rbtree_firstlast(const struct rbtree *tree, int direction); + +/* + * Return the node next to, or previous to the given node. + * + * The direction parameter is either RBTREE_LEFT (to obtain the previous node) + * or RBTREE_RIGHT (to obtain the next one). + */ +struct rbtree_node * rbtree_walk(struct rbtree_node *node, int direction); + +/* + * Return the left-most deepest node of a tree, which is the starting point of + * the postorder traversal performed by rbtree_for_each_remove(). + */ +struct rbtree_node * rbtree_postwalk_deepest(const struct rbtree *tree); + +/* + * Unlink a node from its tree and return the next (right) node in postorder. + */ +struct rbtree_node * rbtree_postwalk_unlink(struct rbtree_node *node); + +#endif /* _KERN_RBTREE_I_H */ diff --git a/kern/sched_prim.c b/kern/sched_prim.c index ff942aee..ba2dc8ab 100644 --- a/kern/sched_prim.c +++ b/kern/sched_prim.c @@ -139,14 +139,14 @@ void thread_check(thread_t, run_queue_t); * The wait event hash table declarations are as follows: */ -#define NUMQUEUES 59 +#define NUMQUEUES 1031 queue_head_t wait_queue[NUMQUEUES]; decl_simple_lock_data(, wait_lock[NUMQUEUES]) /* NOTE: we want a small positive integer out of this */ #define wait_hash(event) \ - ((((int)(event) < 0) ? ~(int)(event) : (int)(event)) % NUMQUEUES) + ((((long)(event) < 0) ? ~(long)(event) : (long)(event)) % NUMQUEUES) void wait_queue_init(void) { @@ -637,6 +637,7 @@ boolean_t thread_invoke( thread_lock(new_thread); new_thread->state &= ~TH_UNINT; thread_unlock(new_thread); + thread_wakeup(&new_thread->state); if (continuation != (void (*)()) 0) { (void) spl0(); @@ -658,6 +659,7 @@ boolean_t thread_invoke( new_thread->state &= ~(TH_SWAPPED | TH_UNINT); thread_unlock(new_thread); + thread_wakeup(&new_thread->state); #if NCPUS > 1 new_thread->last_processor = current_processor(); @@ -787,6 +789,7 @@ boolean_t thread_invoke( new_thread->state &= ~(TH_SWAPPED | TH_UNINT); thread_unlock(new_thread); + thread_wakeup(&new_thread->state); /* * Thread is now interruptible. diff --git a/kern/server_loop.ch b/kern/server_loop.ch index 1aa7edbc..409e013d 100644 --- a/kern/server_loop.ch +++ b/kern/server_loop.ch @@ -39,6 +39,7 @@ */ #include <kern/debug.h> +#include <kern/kalloc.h> #include <mach/port.h> #include <mach/message.h> #include <vm/vm_kern.h> /* for kernel_map */ diff --git a/kern/slab.c b/kern/slab.c new file mode 100644 index 00000000..99d5bca3 --- /dev/null +++ b/kern/slab.c @@ -0,0 +1,1615 @@ +/* + * Copyright (c) 2011 Free Software Foundation. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +/* + * Copyright (c) 2010, 2011 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * + * Object caching and general purpose memory allocator. + * + * This allocator is based on the paper "The Slab Allocator: An Object-Caching + * Kernel Memory Allocator" by Jeff Bonwick. + * + * It allows the allocation of objects (i.e. fixed-size typed buffers) from + * caches and is efficient in both space and time. This implementation follows + * many of the indications from the paper mentioned. The most notable + * differences are outlined below. + * + * The per-cache self-scaling hash table for buffer-to-bufctl conversion, + * described in 3.2.3 "Slab Layout for Large Objects", has been replaced by + * a red-black tree storing slabs, sorted by address. The use of a + * self-balancing tree for buffer-to-slab conversions provides a few advantages + * over a hash table. Unlike a hash table, a BST provides a "lookup nearest" + * operation, so obtaining the slab data (whether it is embedded in the slab or + * off slab) from a buffer address simply consists of a "lookup nearest towards + * 0" tree search. Storing slabs instead of buffers also considerably reduces + * the number of elements to retain. Finally, a self-balancing tree is a true + * self-scaling data structure, whereas a hash table requires periodic + * maintenance and complete resizing, which is expensive. The only drawback is + * that releasing a buffer to the slab layer takes logarithmic time instead of + * constant time. But as the data set size is kept reasonable (because slabs + * are stored instead of buffers) and because the CPU pool layer services most + * requests, avoiding many accesses to the slab layer, it is considered an + * acceptable tradeoff. + * + * This implementation uses per-cpu pools of objects, which service most + * allocation requests. These pools act as caches (but are named differently + * to avoid confusion with CPU caches) that reduce contention on multiprocessor + * systems. When a pool is empty and cannot provide an object, it is filled by + * transferring multiple objects from the slab layer. The symmetric case is + * handled likewise. + */ + +#include <string.h> +#include <kern/assert.h> +#include <kern/mach_clock.h> +#include <kern/printf.h> +#include <kern/slab.h> +#include <kern/kalloc.h> +#include <kern/cpu_number.h> +#include <mach/vm_param.h> +#include <mach/machine/vm_types.h> +#include <vm/vm_kern.h> +#include <vm/vm_types.h> +#include <sys/types.h> + +#ifdef MACH_DEBUG +#include <mach_debug/slab_info.h> +#endif + +/* + * Utility macros. + */ +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#define P2ALIGNED(x, a) (((x) & ((a) - 1)) == 0) +#define ISP2(x) P2ALIGNED(x, x) +#define P2ALIGN(x, a) ((x) & -(a)) +#define P2ROUND(x, a) (-(-(x) & -(a))) +#define P2END(x, a) (-(~(x) & -(a))) +#define likely(expr) __builtin_expect(!!(expr), 1) +#define unlikely(expr) __builtin_expect(!!(expr), 0) + +/* + * Minimum required alignment. + */ +#define KMEM_ALIGN_MIN 8 + +/* + * Minimum number of buffers per slab. + * + * This value is ignored when the slab size exceeds a threshold. + */ +#define KMEM_MIN_BUFS_PER_SLAB 8 + +/* + * Special slab size beyond which the minimum number of buffers per slab is + * ignored when computing the slab size of a cache. + */ +#define KMEM_SLAB_SIZE_THRESHOLD (8 * PAGE_SIZE) + +/* + * Special buffer size under which slab data is unconditionnally allocated + * from its associated slab. + */ +#define KMEM_BUF_SIZE_THRESHOLD (PAGE_SIZE / 8) + +/* + * Time (in ticks) between two garbage collection operations. + */ +#define KMEM_GC_INTERVAL (5 * hz) + +/* + * The transfer size of a CPU pool is computed by dividing the pool size by + * this value. + */ +#define KMEM_CPU_POOL_TRANSFER_RATIO 2 + +/* + * Redzone guard word. + */ +#ifdef __LP64__ +#if _HOST_BIG_ENDIAN +#define KMEM_REDZONE_WORD 0xfeedfacefeedfaceUL +#else /* _HOST_BIG_ENDIAN */ +#define KMEM_REDZONE_WORD 0xcefaedfecefaedfeUL +#endif /* _HOST_BIG_ENDIAN */ +#else /* __LP64__ */ +#if _HOST_BIG_ENDIAN +#define KMEM_REDZONE_WORD 0xfeedfaceUL +#else /* _HOST_BIG_ENDIAN */ +#define KMEM_REDZONE_WORD 0xcefaedfeUL +#endif /* _HOST_BIG_ENDIAN */ +#endif /* __LP64__ */ + +/* + * Redzone byte for padding. + */ +#define KMEM_REDZONE_BYTE 0xbb + +/* + * Size of the VM submap from which default backend functions allocate. + */ +#define KMEM_MAP_SIZE (128 * 1024 * 1024) + +/* + * Shift for the first kalloc cache size. + */ +#define KALLOC_FIRST_SHIFT 5 + +/* + * Number of caches backing general purpose allocations. + */ +#define KALLOC_NR_CACHES 13 + +/* + * Values the buftag state member can take. + */ +#ifdef __LP64__ +#if _HOST_BIG_ENDIAN +#define KMEM_BUFTAG_ALLOC 0xa110c8eda110c8edUL +#define KMEM_BUFTAG_FREE 0xf4eeb10cf4eeb10cUL +#else /* _HOST_BIG_ENDIAN */ +#define KMEM_BUFTAG_ALLOC 0xedc810a1edc810a1UL +#define KMEM_BUFTAG_FREE 0x0cb1eef40cb1eef4UL +#endif /* _HOST_BIG_ENDIAN */ +#else /* __LP64__ */ +#if _HOST_BIG_ENDIAN +#define KMEM_BUFTAG_ALLOC 0xa110c8edUL +#define KMEM_BUFTAG_FREE 0xf4eeb10cUL +#else /* _HOST_BIG_ENDIAN */ +#define KMEM_BUFTAG_ALLOC 0xedc810a1UL +#define KMEM_BUFTAG_FREE 0x0cb1eef4UL +#endif /* _HOST_BIG_ENDIAN */ +#endif /* __LP64__ */ + +/* + * Free and uninitialized patterns. + * + * These values are unconditionnally 64-bit wide since buffers are at least + * 8-byte aligned. + */ +#if _HOST_BIG_ENDIAN +#define KMEM_FREE_PATTERN 0xdeadbeefdeadbeefULL +#define KMEM_UNINIT_PATTERN 0xbaddcafebaddcafeULL +#else /* _HOST_BIG_ENDIAN */ +#define KMEM_FREE_PATTERN 0xefbeaddeefbeaddeULL +#define KMEM_UNINIT_PATTERN 0xfecaddbafecaddbaULL +#endif /* _HOST_BIG_ENDIAN */ + +/* + * Cache flags. + * + * The flags don't change once set and can be tested without locking. + */ +#define KMEM_CF_NO_CPU_POOL 0x01 /* CPU pool layer disabled */ +#define KMEM_CF_SLAB_EXTERNAL 0x02 /* Slab data is off slab */ +#define KMEM_CF_NO_RECLAIM 0x04 /* Slabs are not reclaimable */ +#define KMEM_CF_VERIFY 0x08 /* Debugging facilities enabled */ +#define KMEM_CF_DIRECT 0x10 /* No buf-to-slab tree lookup */ + +/* + * Options for kmem_cache_alloc_verify(). + */ +#define KMEM_AV_NOCONSTRUCT 0 +#define KMEM_AV_CONSTRUCT 1 + +/* + * Error codes for kmem_cache_error(). + */ +#define KMEM_ERR_INVALID 0 /* Invalid address being freed */ +#define KMEM_ERR_DOUBLEFREE 1 /* Freeing already free address */ +#define KMEM_ERR_BUFTAG 2 /* Invalid buftag content */ +#define KMEM_ERR_MODIFIED 3 /* Buffer modified while free */ +#define KMEM_ERR_REDZONE 4 /* Redzone violation */ + +#if SLAB_USE_CPU_POOLS +/* + * Available CPU pool types. + * + * For each entry, the CPU pool size applies from the entry buf_size + * (excluded) up to (and including) the buf_size of the preceding entry. + * + * See struct kmem_cpu_pool_type for a description of the values. + */ +static struct kmem_cpu_pool_type kmem_cpu_pool_types[] = { + { 32768, 1, 0, NULL }, + { 4096, 8, CPU_L1_SIZE, NULL }, + { 256, 64, CPU_L1_SIZE, NULL }, + { 0, 128, CPU_L1_SIZE, NULL } +}; + +/* + * Caches where CPU pool arrays are allocated from. + */ +static struct kmem_cache kmem_cpu_array_caches[ARRAY_SIZE(kmem_cpu_pool_types)]; +#endif /* SLAB_USE_CPU_POOLS */ + +/* + * Cache for off slab data. + */ +static struct kmem_cache kmem_slab_cache; + +/* + * General purpose caches array. + */ +static struct kmem_cache kalloc_caches[KALLOC_NR_CACHES]; + +/* + * List of all caches managed by the allocator. + */ +static struct list kmem_cache_list; +static unsigned int kmem_nr_caches; +static simple_lock_data_t __attribute__((used)) kmem_cache_list_lock; + +/* + * VM submap for slab caches. + */ +static struct vm_map kmem_map_store; +vm_map_t kmem_map = &kmem_map_store; + +/* + * Time of the last memory reclaim, in clock ticks. + */ +static unsigned long kmem_gc_last_tick; + +#define kmem_error(format, ...) \ + printf("mem: error: %s(): " format "\n", __func__, \ + ## __VA_ARGS__) + +#define kmem_warn(format, ...) \ + printf("mem: warning: %s(): " format "\n", __func__, \ + ## __VA_ARGS__) + +#define kmem_print(format, ...) \ + printf(format "\n", ## __VA_ARGS__) + +static void kmem_cache_error(struct kmem_cache *cache, void *buf, int error, + void *arg); +static void * kmem_cache_alloc_from_slab(struct kmem_cache *cache); +static void kmem_cache_free_to_slab(struct kmem_cache *cache, void *buf); + +static void * kmem_buf_verify_bytes(void *buf, void *pattern, size_t size) +{ + char *ptr, *pattern_ptr, *end; + + end = buf + size; + + for (ptr = buf, pattern_ptr = pattern; ptr < end; ptr++, pattern_ptr++) + if (*ptr != *pattern_ptr) + return ptr; + + return NULL; +} + +static void * kmem_buf_verify(void *buf, uint64_t pattern, vm_size_t size) +{ + uint64_t *ptr, *end; + + assert(P2ALIGNED((unsigned long)buf, sizeof(uint64_t))); + assert(P2ALIGNED(size, sizeof(uint64_t))); + + end = buf + size; + + for (ptr = buf; ptr < end; ptr++) + if (*ptr != pattern) + return kmem_buf_verify_bytes(ptr, &pattern, sizeof(pattern)); + + return NULL; +} + +static void kmem_buf_fill(void *buf, uint64_t pattern, size_t size) +{ + uint64_t *ptr, *end; + + assert(P2ALIGNED((unsigned long)buf, sizeof(uint64_t))); + assert(P2ALIGNED(size, sizeof(uint64_t))); + + end = buf + size; + + for (ptr = buf; ptr < end; ptr++) + *ptr = pattern; +} + +static void * kmem_buf_verify_fill(void *buf, uint64_t old, uint64_t new, + size_t size) +{ + uint64_t *ptr, *end; + + assert(P2ALIGNED((unsigned long)buf, sizeof(uint64_t))); + assert(P2ALIGNED(size, sizeof(uint64_t))); + + end = buf + size; + + for (ptr = buf; ptr < end; ptr++) { + if (*ptr != old) + return kmem_buf_verify_bytes(ptr, &old, sizeof(old)); + + *ptr = new; + } + + return NULL; +} + +static inline union kmem_bufctl * +kmem_buf_to_bufctl(void *buf, struct kmem_cache *cache) +{ + return (union kmem_bufctl *)(buf + cache->bufctl_dist); +} + +static inline struct kmem_buftag * +kmem_buf_to_buftag(void *buf, struct kmem_cache *cache) +{ + return (struct kmem_buftag *)(buf + cache->buftag_dist); +} + +static inline void * kmem_bufctl_to_buf(union kmem_bufctl *bufctl, + struct kmem_cache *cache) +{ + return (void *)bufctl - cache->bufctl_dist; +} + +static vm_offset_t kmem_pagealloc(vm_size_t size) +{ + vm_offset_t addr; + kern_return_t kr; + + kr = kmem_alloc_wired(kmem_map, &addr, size); + + if (kr != KERN_SUCCESS) + return 0; + + return addr; +} + +static void kmem_pagefree(vm_offset_t ptr, vm_size_t size) +{ + kmem_free(kmem_map, ptr, size); +} + +static void kmem_slab_create_verify(struct kmem_slab *slab, + struct kmem_cache *cache) +{ + struct kmem_buftag *buftag; + size_t buf_size; + unsigned long buffers; + void *buf; + + buf_size = cache->buf_size; + buf = slab->addr; + buftag = kmem_buf_to_buftag(buf, cache); + + for (buffers = cache->bufs_per_slab; buffers != 0; buffers--) { + kmem_buf_fill(buf, KMEM_FREE_PATTERN, cache->bufctl_dist); + buftag->state = KMEM_BUFTAG_FREE; + buf += buf_size; + buftag = kmem_buf_to_buftag(buf, cache); + } +} + +/* + * Create an empty slab for a cache. + * + * The caller must drop all locks before calling this function. + */ +static struct kmem_slab * kmem_slab_create(struct kmem_cache *cache, + size_t color) +{ + struct kmem_slab *slab; + union kmem_bufctl *bufctl; + size_t buf_size; + unsigned long buffers; + void *slab_buf; + + if (cache->slab_alloc_fn == NULL) + slab_buf = (void *)kmem_pagealloc(cache->slab_size); + else + slab_buf = (void *)cache->slab_alloc_fn(cache->slab_size); + + if (slab_buf == NULL) + return NULL; + + if (cache->flags & KMEM_CF_SLAB_EXTERNAL) { + assert(!(cache->flags & KMEM_CF_NO_RECLAIM)); + slab = (struct kmem_slab *)kmem_cache_alloc(&kmem_slab_cache); + + if (slab == NULL) { + if (cache->slab_free_fn == NULL) + kmem_pagefree((vm_offset_t)slab_buf, cache->slab_size); + else + cache->slab_free_fn((vm_offset_t)slab_buf, cache->slab_size); + + return NULL; + } + } else { + slab = (struct kmem_slab *)(slab_buf + cache->slab_size) - 1; + } + + list_node_init(&slab->list_node); + rbtree_node_init(&slab->tree_node); + slab->nr_refs = 0; + slab->first_free = NULL; + slab->addr = slab_buf + color; + + buf_size = cache->buf_size; + bufctl = kmem_buf_to_bufctl(slab->addr, cache); + + for (buffers = cache->bufs_per_slab; buffers != 0; buffers--) { + bufctl->next = slab->first_free; + slab->first_free = bufctl; + bufctl = (union kmem_bufctl *)((void *)bufctl + buf_size); + } + + if (cache->flags & KMEM_CF_VERIFY) + kmem_slab_create_verify(slab, cache); + + return slab; +} + +static void kmem_slab_destroy_verify(struct kmem_slab *slab, + struct kmem_cache *cache) +{ + struct kmem_buftag *buftag; + size_t buf_size; + unsigned long buffers; + void *buf, *addr; + + buf_size = cache->buf_size; + buf = slab->addr; + buftag = kmem_buf_to_buftag(buf, cache); + + for (buffers = cache->bufs_per_slab; buffers != 0; buffers--) { + if (buftag->state != KMEM_BUFTAG_FREE) + kmem_cache_error(cache, buf, KMEM_ERR_BUFTAG, buftag); + + addr = kmem_buf_verify(buf, KMEM_FREE_PATTERN, cache->bufctl_dist); + + if (addr != NULL) + kmem_cache_error(cache, buf, KMEM_ERR_MODIFIED, addr); + + buf += buf_size; + buftag = kmem_buf_to_buftag(buf, cache); + } +} + +/* + * Destroy a slab. + * + * The caller must drop all locks before calling this function. + */ +static void kmem_slab_destroy(struct kmem_slab *slab, struct kmem_cache *cache) +{ + vm_offset_t slab_buf; + + assert(slab->nr_refs == 0); + assert(slab->first_free != NULL); + assert(!(cache->flags & KMEM_CF_NO_RECLAIM)); + + if (cache->flags & KMEM_CF_VERIFY) + kmem_slab_destroy_verify(slab, cache); + + slab_buf = (vm_offset_t)P2ALIGN((unsigned long)slab->addr, PAGE_SIZE); + + if (cache->slab_free_fn == NULL) + kmem_pagefree(slab_buf, cache->slab_size); + else + cache->slab_free_fn(slab_buf, cache->slab_size); + + if (cache->flags & KMEM_CF_SLAB_EXTERNAL) + kmem_cache_free(&kmem_slab_cache, (vm_offset_t)slab); +} + +static inline int kmem_slab_use_tree(int flags) +{ + return !(flags & KMEM_CF_DIRECT) || (flags & KMEM_CF_VERIFY); +} + +static inline int kmem_slab_cmp_lookup(const void *addr, + const struct rbtree_node *node) +{ + struct kmem_slab *slab; + + slab = rbtree_entry(node, struct kmem_slab, tree_node); + + if (addr == slab->addr) + return 0; + else if (addr < slab->addr) + return -1; + else + return 1; +} + +static inline int kmem_slab_cmp_insert(const struct rbtree_node *a, + const struct rbtree_node *b) +{ + struct kmem_slab *slab; + + slab = rbtree_entry(a, struct kmem_slab, tree_node); + return kmem_slab_cmp_lookup(slab->addr, b); +} + +#if SLAB_USE_CPU_POOLS +static void kmem_cpu_pool_init(struct kmem_cpu_pool *cpu_pool, + struct kmem_cache *cache) +{ + simple_lock_init(&cpu_pool->lock); + cpu_pool->flags = cache->flags; + cpu_pool->size = 0; + cpu_pool->transfer_size = 0; + cpu_pool->nr_objs = 0; + cpu_pool->array = NULL; +} + +/* + * Return a CPU pool. + * + * This function will generally return the pool matching the CPU running the + * calling thread. Because of context switches and thread migration, the + * caller might be running on another processor after this function returns. + * Although not optimal, this should rarely happen, and it doesn't affect the + * allocator operations in any other way, as CPU pools are always valid, and + * their access is serialized by a lock. + */ +static inline struct kmem_cpu_pool * kmem_cpu_pool_get(struct kmem_cache *cache) +{ + return &cache->cpu_pools[cpu_number()]; +} + +static inline void kmem_cpu_pool_build(struct kmem_cpu_pool *cpu_pool, + struct kmem_cache *cache, void **array) +{ + cpu_pool->size = cache->cpu_pool_type->array_size; + cpu_pool->transfer_size = (cpu_pool->size + + KMEM_CPU_POOL_TRANSFER_RATIO - 1) + / KMEM_CPU_POOL_TRANSFER_RATIO; + cpu_pool->array = array; +} + +static inline void * kmem_cpu_pool_pop(struct kmem_cpu_pool *cpu_pool) +{ + cpu_pool->nr_objs--; + return cpu_pool->array[cpu_pool->nr_objs]; +} + +static inline void kmem_cpu_pool_push(struct kmem_cpu_pool *cpu_pool, void *obj) +{ + cpu_pool->array[cpu_pool->nr_objs] = obj; + cpu_pool->nr_objs++; +} + +static int kmem_cpu_pool_fill(struct kmem_cpu_pool *cpu_pool, + struct kmem_cache *cache) +{ + void *obj; + int i; + + simple_lock(&cache->lock); + + for (i = 0; i < cpu_pool->transfer_size; i++) { + obj = kmem_cache_alloc_from_slab(cache); + + if (obj == NULL) + break; + + kmem_cpu_pool_push(cpu_pool, obj); + } + + simple_unlock(&cache->lock); + + return i; +} + +static void kmem_cpu_pool_drain(struct kmem_cpu_pool *cpu_pool, + struct kmem_cache *cache) +{ + void *obj; + int i; + + simple_lock(&cache->lock); + + for (i = cpu_pool->transfer_size; i > 0; i--) { + obj = kmem_cpu_pool_pop(cpu_pool); + kmem_cache_free_to_slab(cache, obj); + } + + simple_unlock(&cache->lock); +} +#endif /* SLAB_USE_CPU_POOLS */ + +static void kmem_cache_error(struct kmem_cache *cache, void *buf, int error, + void *arg) +{ + struct kmem_buftag *buftag; + + kmem_error("cache: %s, buffer: %p", cache->name, (void *)buf); + + switch(error) { + case KMEM_ERR_INVALID: + kmem_error("freeing invalid address"); + break; + case KMEM_ERR_DOUBLEFREE: + kmem_error("attempting to free the same address twice"); + break; + case KMEM_ERR_BUFTAG: + buftag = arg; + kmem_error("invalid buftag content, buftag state: %p", + (void *)buftag->state); + break; + case KMEM_ERR_MODIFIED: + kmem_error("free buffer modified, fault address: %p, " + "offset in buffer: %td", arg, arg - buf); + break; + case KMEM_ERR_REDZONE: + kmem_error("write beyond end of buffer, fault address: %p, " + "offset in buffer: %td", arg, arg - buf); + break; + default: + kmem_error("unknown error"); + } + + /* + * Never reached. + */ +} + +/* + * Compute an appropriate slab size for the given cache. + * + * Once the slab size is known, this function sets the related properties + * (buffers per slab and maximum color). It can also set the KMEM_CF_DIRECT + * and/or KMEM_CF_SLAB_EXTERNAL flags depending on the resulting layout. + */ +static void kmem_cache_compute_sizes(struct kmem_cache *cache, int flags) +{ + size_t i, buffers, buf_size, slab_size, free_slab_size, optimal_size; + size_t waste, waste_min; + int embed, optimal_embed = optimal_embed; + + buf_size = cache->buf_size; + + if (buf_size < KMEM_BUF_SIZE_THRESHOLD) + flags |= KMEM_CACHE_NOOFFSLAB; + + i = 0; + waste_min = (size_t)-1; + + do { + i++; + slab_size = P2ROUND(i * buf_size, PAGE_SIZE); + free_slab_size = slab_size; + + if (flags & KMEM_CACHE_NOOFFSLAB) + free_slab_size -= sizeof(struct kmem_slab); + + buffers = free_slab_size / buf_size; + waste = free_slab_size % buf_size; + + if (buffers > i) + i = buffers; + + if (flags & KMEM_CACHE_NOOFFSLAB) + embed = 1; + else if (sizeof(struct kmem_slab) <= waste) { + embed = 1; + waste -= sizeof(struct kmem_slab); + } else { + embed = 0; + } + + if (waste <= waste_min) { + waste_min = waste; + optimal_size = slab_size; + optimal_embed = embed; + } + } while ((buffers < KMEM_MIN_BUFS_PER_SLAB) + && (slab_size < KMEM_SLAB_SIZE_THRESHOLD)); + + assert(!(flags & KMEM_CACHE_NOOFFSLAB) || optimal_embed); + + cache->slab_size = optimal_size; + slab_size = cache->slab_size - (optimal_embed + ? sizeof(struct kmem_slab) + : 0); + cache->bufs_per_slab = slab_size / buf_size; + cache->color_max = slab_size % buf_size; + + if (cache->color_max >= PAGE_SIZE) + cache->color_max = PAGE_SIZE - 1; + + if (optimal_embed) { + if (cache->slab_size == PAGE_SIZE) + cache->flags |= KMEM_CF_DIRECT; + } else { + cache->flags |= KMEM_CF_SLAB_EXTERNAL; + } +} + +void kmem_cache_init(struct kmem_cache *cache, const char *name, + size_t obj_size, size_t align, kmem_cache_ctor_t ctor, + kmem_slab_alloc_fn_t slab_alloc_fn, + kmem_slab_free_fn_t slab_free_fn, int flags) +{ +#if SLAB_USE_CPU_POOLS + struct kmem_cpu_pool_type *cpu_pool_type; + size_t i; +#endif /* SLAB_USE_CPU_POOLS */ + size_t buf_size; + +#if SLAB_VERIFY + cache->flags = KMEM_CF_VERIFY; +#else /* SLAB_VERIFY */ + cache->flags = 0; +#endif /* SLAB_VERIFY */ + + if (flags & KMEM_CACHE_NOCPUPOOL) + cache->flags |= KMEM_CF_NO_CPU_POOL; + + if (flags & KMEM_CACHE_NORECLAIM) { + assert(slab_free_fn == NULL); + flags |= KMEM_CACHE_NOOFFSLAB; + cache->flags |= KMEM_CF_NO_RECLAIM; + } + + if (flags & KMEM_CACHE_VERIFY) + cache->flags |= KMEM_CF_VERIFY; + + if (align < KMEM_ALIGN_MIN) + align = KMEM_ALIGN_MIN; + + assert(obj_size > 0); + assert(ISP2(align)); + assert(align < PAGE_SIZE); + + buf_size = P2ROUND(obj_size, align); + + simple_lock_init(&cache->lock); + list_node_init(&cache->node); + list_init(&cache->partial_slabs); + list_init(&cache->free_slabs); + rbtree_init(&cache->active_slabs); + cache->obj_size = obj_size; + cache->align = align; + cache->buf_size = buf_size; + cache->bufctl_dist = buf_size - sizeof(union kmem_bufctl); + cache->color = 0; + cache->nr_objs = 0; + cache->nr_bufs = 0; + cache->nr_slabs = 0; + cache->nr_free_slabs = 0; + cache->ctor = ctor; + cache->slab_alloc_fn = slab_alloc_fn; + cache->slab_free_fn = slab_free_fn; + strncpy(cache->name, name, sizeof(cache->name)); + cache->name[sizeof(cache->name) - 1] = '\0'; + cache->buftag_dist = 0; + cache->redzone_pad = 0; + + if (cache->flags & KMEM_CF_VERIFY) { + cache->bufctl_dist = buf_size; + cache->buftag_dist = cache->bufctl_dist + sizeof(union kmem_bufctl); + cache->redzone_pad = cache->bufctl_dist - cache->obj_size; + buf_size += sizeof(union kmem_bufctl) + sizeof(struct kmem_buftag); + buf_size = P2ROUND(buf_size, align); + cache->buf_size = buf_size; + } + + kmem_cache_compute_sizes(cache, flags); + +#if SLAB_USE_CPU_POOLS + for (cpu_pool_type = kmem_cpu_pool_types; + buf_size <= cpu_pool_type->buf_size; + cpu_pool_type++); + + cache->cpu_pool_type = cpu_pool_type; + + for (i = 0; i < ARRAY_SIZE(cache->cpu_pools); i++) + kmem_cpu_pool_init(&cache->cpu_pools[i], cache); +#endif /* SLAB_USE_CPU_POOLS */ + + simple_lock(&kmem_cache_list_lock); + list_insert_tail(&kmem_cache_list, &cache->node); + kmem_nr_caches++; + simple_unlock(&kmem_cache_list_lock); +} + +static inline int kmem_cache_empty(struct kmem_cache *cache) +{ + return cache->nr_objs == cache->nr_bufs; +} + +static int kmem_cache_grow(struct kmem_cache *cache) +{ + struct kmem_slab *slab; + size_t color; + int empty; + + simple_lock(&cache->lock); + + if (!kmem_cache_empty(cache)) { + simple_unlock(&cache->lock); + return 1; + } + + color = cache->color; + cache->color += cache->align; + + if (cache->color > cache->color_max) + cache->color = 0; + + simple_unlock(&cache->lock); + + slab = kmem_slab_create(cache, color); + + simple_lock(&cache->lock); + + if (slab != NULL) { + list_insert_tail(&cache->free_slabs, &slab->list_node); + cache->nr_bufs += cache->bufs_per_slab; + cache->nr_slabs++; + cache->nr_free_slabs++; + } + + /* + * Even if our slab creation failed, another thread might have succeeded + * in growing the cache. + */ + empty = kmem_cache_empty(cache); + + simple_unlock(&cache->lock); + + return !empty; +} + +static void kmem_cache_reap(struct kmem_cache *cache) +{ + struct kmem_slab *slab; + struct list dead_slabs; + + if (cache->flags & KMEM_CF_NO_RECLAIM) + return; + + list_init(&dead_slabs); + + simple_lock(&cache->lock); + + while (!list_empty(&cache->free_slabs)) { + slab = list_first_entry(&cache->free_slabs, struct kmem_slab, + list_node); + list_remove(&slab->list_node); + list_insert(&dead_slabs, &slab->list_node); + cache->nr_bufs -= cache->bufs_per_slab; + cache->nr_slabs--; + cache->nr_free_slabs--; + } + + simple_unlock(&cache->lock); + + while (!list_empty(&dead_slabs)) { + slab = list_first_entry(&dead_slabs, struct kmem_slab, list_node); + list_remove(&slab->list_node); + kmem_slab_destroy(slab, cache); + } +} + +/* + * Allocate a raw (unconstructed) buffer from the slab layer of a cache. + * + * The cache must be locked before calling this function. + */ +static void * kmem_cache_alloc_from_slab(struct kmem_cache *cache) +{ + struct kmem_slab *slab; + union kmem_bufctl *bufctl; + + if (!list_empty(&cache->partial_slabs)) + slab = list_first_entry(&cache->partial_slabs, struct kmem_slab, + list_node); + else if (!list_empty(&cache->free_slabs)) + slab = list_first_entry(&cache->free_slabs, struct kmem_slab, + list_node); + else + return NULL; + + bufctl = slab->first_free; + assert(bufctl != NULL); + slab->first_free = bufctl->next; + slab->nr_refs++; + cache->nr_objs++; + + /* + * The slab has become complete. + */ + if (slab->nr_refs == cache->bufs_per_slab) { + list_remove(&slab->list_node); + + if (slab->nr_refs == 1) + cache->nr_free_slabs--; + } else if (slab->nr_refs == 1) { + /* + * The slab has become partial. + */ + list_remove(&slab->list_node); + list_insert_tail(&cache->partial_slabs, &slab->list_node); + cache->nr_free_slabs--; + } else if (!list_singular(&cache->partial_slabs)) { + struct list *node; + struct kmem_slab *tmp; + + /* + * The slab remains partial. If there are more than one partial slabs, + * maintain the list sorted. + */ + + assert(slab->nr_refs > 1); + + for (node = list_prev(&slab->list_node); + !list_end(&cache->partial_slabs, node); + node = list_prev(node)) { + tmp = list_entry(node, struct kmem_slab, list_node); + + if (tmp->nr_refs >= slab->nr_refs) + break; + } + + /* + * If the direct neighbor was found, the list is already sorted. + * If no slab was found, the slab is inserted at the head of the list. + */ + if (node != list_prev(&slab->list_node)) { + list_remove(&slab->list_node); + list_insert_after(node, &slab->list_node); + } + } + + if ((slab->nr_refs == 1) && kmem_slab_use_tree(cache->flags)) + rbtree_insert(&cache->active_slabs, &slab->tree_node, + kmem_slab_cmp_insert); + + return kmem_bufctl_to_buf(bufctl, cache); +} + +/* + * Release a buffer to the slab layer of a cache. + * + * The cache must be locked before calling this function. + */ +static void kmem_cache_free_to_slab(struct kmem_cache *cache, void *buf) +{ + struct kmem_slab *slab; + union kmem_bufctl *bufctl; + + if (cache->flags & KMEM_CF_DIRECT) { + assert(cache->slab_size == PAGE_SIZE); + slab = (struct kmem_slab *)P2END((unsigned long)buf, cache->slab_size) + - 1; + } else { + struct rbtree_node *node; + + node = rbtree_lookup_nearest(&cache->active_slabs, buf, + kmem_slab_cmp_lookup, RBTREE_LEFT); + assert(node != NULL); + slab = rbtree_entry(node, struct kmem_slab, tree_node); + assert((unsigned long)buf < (P2ALIGN((unsigned long)slab->addr + + cache->slab_size, PAGE_SIZE))); + } + + assert(slab->nr_refs >= 1); + assert(slab->nr_refs <= cache->bufs_per_slab); + bufctl = kmem_buf_to_bufctl(buf, cache); + bufctl->next = slab->first_free; + slab->first_free = bufctl; + slab->nr_refs--; + cache->nr_objs--; + + /* + * The slab has become free. + */ + if (slab->nr_refs == 0) { + if (kmem_slab_use_tree(cache->flags)) + rbtree_remove(&cache->active_slabs, &slab->tree_node); + + /* + * The slab was partial. + */ + if (cache->bufs_per_slab > 1) + list_remove(&slab->list_node); + + list_insert_tail(&cache->free_slabs, &slab->list_node); + cache->nr_free_slabs++; + } else if (slab->nr_refs == (cache->bufs_per_slab - 1)) { + /* + * The slab has become partial. + */ + list_insert(&cache->partial_slabs, &slab->list_node); + } else if (!list_singular(&cache->partial_slabs)) { + struct list *node; + struct kmem_slab *tmp; + + /* + * The slab remains partial. If there are more than one partial slabs, + * maintain the list sorted. + */ + + assert(slab->nr_refs > 0); + + for (node = list_next(&slab->list_node); + !list_end(&cache->partial_slabs, node); + node = list_next(node)) { + tmp = list_entry(node, struct kmem_slab, list_node); + + if (tmp->nr_refs <= slab->nr_refs) + break; + } + + /* + * If the direct neighbor was found, the list is already sorted. + * If no slab was found, the slab is inserted at the tail of the list. + */ + if (node != list_next(&slab->list_node)) { + list_remove(&slab->list_node); + list_insert_before(node, &slab->list_node); + } + } +} + +static void kmem_cache_alloc_verify(struct kmem_cache *cache, void *buf, + int construct) +{ + struct kmem_buftag *buftag; + union kmem_bufctl *bufctl; + void *addr; + + buftag = kmem_buf_to_buftag(buf, cache); + + if (buftag->state != KMEM_BUFTAG_FREE) + kmem_cache_error(cache, buf, KMEM_ERR_BUFTAG, buftag); + + addr = kmem_buf_verify_fill(buf, KMEM_FREE_PATTERN, KMEM_UNINIT_PATTERN, + cache->bufctl_dist); + + if (addr != NULL) + kmem_cache_error(cache, buf, KMEM_ERR_MODIFIED, addr); + + addr = buf + cache->obj_size; + memset(addr, KMEM_REDZONE_BYTE, cache->redzone_pad); + + bufctl = kmem_buf_to_bufctl(buf, cache); + bufctl->redzone = KMEM_REDZONE_WORD; + buftag->state = KMEM_BUFTAG_ALLOC; + + if (construct && (cache->ctor != NULL)) + cache->ctor(buf); +} + +vm_offset_t kmem_cache_alloc(struct kmem_cache *cache) +{ + int filled; + void *buf; + +#if SLAB_USE_CPU_POOLS + struct kmem_cpu_pool *cpu_pool; + + cpu_pool = kmem_cpu_pool_get(cache); + + if (cpu_pool->flags & KMEM_CF_NO_CPU_POOL) + goto slab_alloc; + + simple_lock(&cpu_pool->lock); + +fast_alloc: + if (likely(cpu_pool->nr_objs > 0)) { + buf = kmem_cpu_pool_pop(cpu_pool); + simple_unlock(&cpu_pool->lock); + + if (cpu_pool->flags & KMEM_CF_VERIFY) + kmem_cache_alloc_verify(cache, buf, KMEM_AV_CONSTRUCT); + + return (vm_offset_t)buf; + } + + if (cpu_pool->array != NULL) { + filled = kmem_cpu_pool_fill(cpu_pool, cache); + + if (!filled) { + simple_unlock(&cpu_pool->lock); + + filled = kmem_cache_grow(cache); + + if (!filled) + return 0; + + simple_lock(&cpu_pool->lock); + } + + goto fast_alloc; + } + + simple_unlock(&cpu_pool->lock); +#endif /* SLAB_USE_CPU_POOLS */ + +slab_alloc: + simple_lock(&cache->lock); + buf = kmem_cache_alloc_from_slab(cache); + simple_unlock(&cache->lock); + + if (buf == NULL) { + filled = kmem_cache_grow(cache); + + if (!filled) + return 0; + + goto slab_alloc; + } + + if (cache->flags & KMEM_CF_VERIFY) + kmem_cache_alloc_verify(cache, buf, KMEM_AV_NOCONSTRUCT); + + if (cache->ctor != NULL) + cache->ctor(buf); + + return (vm_offset_t)buf; +} + +static void kmem_cache_free_verify(struct kmem_cache *cache, void *buf) +{ + struct rbtree_node *node; + struct kmem_buftag *buftag; + struct kmem_slab *slab; + union kmem_bufctl *bufctl; + unsigned char *redzone_byte; + unsigned long slabend; + + simple_lock(&cache->lock); + node = rbtree_lookup_nearest(&cache->active_slabs, buf, + kmem_slab_cmp_lookup, RBTREE_LEFT); + simple_unlock(&cache->lock); + + if (node == NULL) + kmem_cache_error(cache, buf, KMEM_ERR_INVALID, NULL); + + slab = rbtree_entry(node, struct kmem_slab, tree_node); + slabend = P2ALIGN((unsigned long)slab->addr + cache->slab_size, PAGE_SIZE); + + if ((unsigned long)buf >= slabend) + kmem_cache_error(cache, buf, KMEM_ERR_INVALID, NULL); + + if ((((unsigned long)buf - (unsigned long)slab->addr) % cache->buf_size) + != 0) + kmem_cache_error(cache, buf, KMEM_ERR_INVALID, NULL); + + /* + * As the buffer address is valid, accessing its buftag is safe. + */ + buftag = kmem_buf_to_buftag(buf, cache); + + if (buftag->state != KMEM_BUFTAG_ALLOC) { + if (buftag->state == KMEM_BUFTAG_FREE) + kmem_cache_error(cache, buf, KMEM_ERR_DOUBLEFREE, NULL); + else + kmem_cache_error(cache, buf, KMEM_ERR_BUFTAG, buftag); + } + + redzone_byte = buf + cache->obj_size; + bufctl = kmem_buf_to_bufctl(buf, cache); + + while (redzone_byte < (unsigned char *)bufctl) { + if (*redzone_byte != KMEM_REDZONE_BYTE) + kmem_cache_error(cache, buf, KMEM_ERR_REDZONE, redzone_byte); + + redzone_byte++; + } + + if (bufctl->redzone != KMEM_REDZONE_WORD) { + unsigned long word; + + word = KMEM_REDZONE_WORD; + redzone_byte = kmem_buf_verify_bytes(&bufctl->redzone, &word, + sizeof(bufctl->redzone)); + kmem_cache_error(cache, buf, KMEM_ERR_REDZONE, redzone_byte); + } + + kmem_buf_fill(buf, KMEM_FREE_PATTERN, cache->bufctl_dist); + buftag->state = KMEM_BUFTAG_FREE; +} + +void kmem_cache_free(struct kmem_cache *cache, vm_offset_t obj) +{ +#if SLAB_USE_CPU_POOLS + struct kmem_cpu_pool *cpu_pool; + void **array; + + cpu_pool = kmem_cpu_pool_get(cache); + + if (cpu_pool->flags & KMEM_CF_VERIFY) { +#else /* SLAB_USE_CPU_POOLS */ + if (cache->flags & KMEM_CF_VERIFY) { +#endif /* SLAB_USE_CPU_POOLS */ + kmem_cache_free_verify(cache, (void *)obj); + } + +#if SLAB_USE_CPU_POOLS + if (cpu_pool->flags & KMEM_CF_NO_CPU_POOL) + goto slab_free; + + simple_lock(&cpu_pool->lock); + +fast_free: + if (likely(cpu_pool->nr_objs < cpu_pool->size)) { + kmem_cpu_pool_push(cpu_pool, (void *)obj); + simple_unlock(&cpu_pool->lock); + return; + } + + if (cpu_pool->array != NULL) { + kmem_cpu_pool_drain(cpu_pool, cache); + goto fast_free; + } + + simple_unlock(&cpu_pool->lock); + + array = (void *)kmem_cache_alloc(cache->cpu_pool_type->array_cache); + + if (array != NULL) { + simple_lock(&cpu_pool->lock); + + /* + * Another thread may have built the CPU pool while the lock was + * dropped. + */ + if (cpu_pool->array != NULL) { + simple_unlock(&cpu_pool->lock); + kmem_cache_free(cache->cpu_pool_type->array_cache, + (vm_offset_t)array); + goto fast_free; + } + + kmem_cpu_pool_build(cpu_pool, cache, array); + goto fast_free; + } + +slab_free: +#endif /* SLAB_USE_CPU_POOLS */ + + kmem_cache_free_to_slab(cache, (void *)obj); +} + +void slab_collect(void) +{ + struct kmem_cache *cache; + + if (elapsed_ticks <= (kmem_gc_last_tick + KMEM_GC_INTERVAL)) + return; + + kmem_gc_last_tick = elapsed_ticks; + + simple_lock(&kmem_cache_list_lock); + + list_for_each_entry(&kmem_cache_list, cache, node) + kmem_cache_reap(cache); + + simple_unlock(&kmem_cache_list_lock); +} + +void slab_bootstrap(void) +{ + /* Make sure a bufctl can always be stored in a buffer */ + assert(sizeof(union kmem_bufctl) <= KMEM_ALIGN_MIN); + + list_init(&kmem_cache_list); + simple_lock_init(&kmem_cache_list_lock); +} + +void slab_init(void) +{ + vm_offset_t min, max; + +#if SLAB_USE_CPU_POOLS + struct kmem_cpu_pool_type *cpu_pool_type; + char name[KMEM_CACHE_NAME_SIZE]; + size_t i, size; +#endif /* SLAB_USE_CPU_POOLS */ + + kmem_submap(kmem_map, kernel_map, &min, &max, KMEM_MAP_SIZE, FALSE); + +#if SLAB_USE_CPU_POOLS + for (i = 0; i < ARRAY_SIZE(kmem_cpu_pool_types); i++) { + cpu_pool_type = &kmem_cpu_pool_types[i]; + cpu_pool_type->array_cache = &kmem_cpu_array_caches[i]; + sprintf(name, "kmem_cpu_array_%d", cpu_pool_type->array_size); + size = sizeof(void *) * cpu_pool_type->array_size; + kmem_cache_init(cpu_pool_type->array_cache, name, size, + cpu_pool_type->array_align, NULL, NULL, NULL, 0); + } +#endif /* SLAB_USE_CPU_POOLS */ + + /* + * Prevent off slab data for the slab cache to avoid infinite recursion. + */ + kmem_cache_init(&kmem_slab_cache, "kmem_slab", sizeof(struct kmem_slab), + 0, NULL, NULL, NULL, KMEM_CACHE_NOOFFSLAB); +} + +static vm_offset_t kalloc_pagealloc(vm_size_t size) +{ + vm_offset_t addr; + kern_return_t kr; + + kr = kmem_alloc_wired(kmem_map, &addr, size); + + if (kr != KERN_SUCCESS) + return 0; + + return addr; +} + +static void kalloc_pagefree(vm_offset_t ptr, vm_size_t size) +{ + kmem_free(kmem_map, ptr, size); +} + +void kalloc_init(void) +{ + char name[KMEM_CACHE_NAME_SIZE]; + size_t i, size; + vm_offset_t min, max; + + size = 1 << KALLOC_FIRST_SHIFT; + + for (i = 0; i < ARRAY_SIZE(kalloc_caches); i++) { + sprintf(name, "kalloc_%lu", size); + kmem_cache_init(&kalloc_caches[i], name, size, 0, NULL, + kalloc_pagealloc, kalloc_pagefree, 0); + size <<= 1; + } +} + +/* + * Return the kalloc cache index matching the given allocation size, which + * must be strictly greater than 0. + */ +static inline size_t kalloc_get_index(unsigned long size) +{ + assert(size != 0); + + size = (size - 1) >> KALLOC_FIRST_SHIFT; + + if (size == 0) + return 0; + else + return (sizeof(long) * 8) - __builtin_clzl(size); +} + +static void kalloc_verify(struct kmem_cache *cache, void *buf, size_t size) +{ + size_t redzone_size; + void *redzone; + + assert(size <= cache->obj_size); + + redzone = buf + size; + redzone_size = cache->obj_size - size; + memset(redzone, KMEM_REDZONE_BYTE, redzone_size); +} + +vm_offset_t kalloc(vm_size_t size) +{ + size_t index; + void *buf; + + if (size == 0) + return 0; + + index = kalloc_get_index(size); + + if (index < ARRAY_SIZE(kalloc_caches)) { + struct kmem_cache *cache; + + cache = &kalloc_caches[index]; + buf = (void *)kmem_cache_alloc(cache); + + if ((buf != 0) && (cache->flags & KMEM_CF_VERIFY)) + kalloc_verify(cache, buf, size); + } else + buf = (void *)kalloc_pagealloc(size); + + return (vm_offset_t)buf; +} + +static void kfree_verify(struct kmem_cache *cache, void *buf, size_t size) +{ + unsigned char *redzone_byte, *redzone_end; + + assert(size <= cache->obj_size); + + redzone_byte = buf + size; + redzone_end = buf + cache->obj_size; + + while (redzone_byte < redzone_end) { + if (*redzone_byte != KMEM_REDZONE_BYTE) + kmem_cache_error(cache, buf, KMEM_ERR_REDZONE, redzone_byte); + + redzone_byte++; + } +} + +void kfree(vm_offset_t data, vm_size_t size) +{ + size_t index; + + if ((data == 0) || (size == 0)) + return; + + index = kalloc_get_index(size); + + if (index < ARRAY_SIZE(kalloc_caches)) { + struct kmem_cache *cache; + + cache = &kalloc_caches[index]; + + if (cache->flags & KMEM_CF_VERIFY) + kfree_verify(cache, (void *)data, size); + + kmem_cache_free(cache, data); + } else { + kalloc_pagefree(data, size); + } +} + +void slab_info(void) +{ + struct kmem_cache *cache; + vm_size_t mem_usage, mem_reclaimable; + + printf("cache obj slab bufs objs bufs " + " total reclaimable\n" + "name size size /slab usage count " + " memory memory\n"); + + simple_lock(&kmem_cache_list_lock); + + list_for_each_entry(&kmem_cache_list, cache, node) { + simple_lock(&cache->lock); + + mem_usage = (cache->nr_slabs * cache->slab_size) >> 10; + mem_reclaimable = (cache->nr_free_slabs * cache->slab_size) >> 10; + + printf("%-19s %6lu %3luk %4lu %6lu %6lu %7luk %10luk\n", + cache->name, cache->obj_size, cache->slab_size >> 10, + cache->bufs_per_slab, cache->nr_objs, cache->nr_bufs, + mem_usage, mem_reclaimable); + + simple_unlock(&cache->lock); + } + + simple_unlock(&kmem_cache_list_lock); +} + +#if MACH_DEBUG +kern_return_t host_slab_info(host_t host, cache_info_array_t *infop, + unsigned int *infoCntp) +{ + struct kmem_cache *cache; + cache_info_t *info; + unsigned int i, nr_caches; + vm_size_t info_size = info_size; + kern_return_t kr; + + if (host == HOST_NULL) + return KERN_INVALID_HOST; + + /* + * Assume the cache list is unaltered once the kernel is ready. + */ + + simple_lock(&kmem_cache_list_lock); + nr_caches = kmem_nr_caches; + simple_unlock(&kmem_cache_list_lock); + + if (nr_caches <= *infoCntp) + info = *infop; + else { + vm_offset_t info_addr; + + info_size = round_page(nr_caches * sizeof(*info)); + kr = kmem_alloc_pageable(ipc_kernel_map, &info_addr, info_size); + + if (kr != KERN_SUCCESS) + return kr; + + info = (cache_info_t *)info_addr; + } + + if (info == NULL) + return KERN_RESOURCE_SHORTAGE; + + i = 0; + + list_for_each_entry(&kmem_cache_list, cache, node) { + simple_lock(&cache_lock); + info[i].flags = ((cache->flags & KMEM_CF_NO_CPU_POOL) + ? CACHE_FLAGS_NO_CPU_POOL : 0) + | ((cache->flags & KMEM_CF_SLAB_EXTERNAL) + ? CACHE_FLAGS_SLAB_EXTERNAL : 0) + | ((cache->flags & KMEM_CF_NO_RECLAIM) + ? CACHE_FLAGS_NO_RECLAIM : 0) + | ((cache->flags & KMEM_CF_VERIFY) + ? CACHE_FLAGS_VERIFY : 0) + | ((cache->flags & KMEM_CF_DIRECT) + ? CACHE_FLAGS_DIRECT : 0); +#if SLAB_USE_CPU_POOLS + info[i].cpu_pool_size = cache->cpu_pool_type->array_size; +#else /* SLAB_USE_CPU_POOLS */ + info[i].cpu_pool_size = 0; +#endif /* SLAB_USE_CPU_POOLS */ + info[i].obj_size = cache->obj_size; + info[i].align = cache->align; + info[i].buf_size = cache->buf_size; + info[i].slab_size = cache->slab_size; + info[i].bufs_per_slab = cache->bufs_per_slab; + info[i].nr_objs = cache->nr_objs; + info[i].nr_bufs = cache->nr_bufs; + info[i].nr_slabs = cache->nr_slabs; + info[i].nr_free_slabs = cache->nr_free_slabs; + strncpy(info[i].name, cache->name, sizeof(info[i].name)); + info[i].name[sizeof(info[i].name) - 1] = '\0'; + simple_unlock(&cache->lock); + + i++; + } + + if (info != *infop) { + vm_map_copy_t copy; + vm_size_t used; + + used = nr_caches * sizeof(*info); + + if (used != info_size) + memset((char *)info + used, 0, info_size - used); + + kr = vm_map_copyin(ipc_kernel_map, (vm_offset_t)info, used, TRUE, + ©); + + assert(kr == KERN_SUCCESS); + *infop = (cache_info_t *)copy; + } + + *infoCntp = nr_caches; + + return KERN_SUCCESS; +} +#endif /* MACH_DEBUG */ diff --git a/kern/slab.h b/kern/slab.h new file mode 100644 index 00000000..47bef218 --- /dev/null +++ b/kern/slab.h @@ -0,0 +1,254 @@ +/* + * Copyright (c) 2011 Free Software Foundation. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +/* + * Copyright (c) 2010, 2011 Richard Braun. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * + * Object caching memory allocator. + */ + +#ifndef _KERN_SLAB_H +#define _KERN_SLAB_H + +#include <kern/lock.h> +#include <kern/list.h> +#include <kern/rbtree.h> +#include <mach/machine/vm_types.h> +#include <sys/types.h> +#include <vm/vm_types.h> + +#if SLAB_USE_CPU_POOLS +/* + * L1 cache line size. + */ +#define CPU_L1_SIZE (1 << CPU_L1_SHIFT) + +/* + * Per-processor cache of pre-constructed objects. + * + * The flags member is a read-only CPU-local copy of the parent cache flags. + */ +struct kmem_cpu_pool { + simple_lock_data_t lock; + int flags; + int size; + int transfer_size; + int nr_objs; + void **array; +} __attribute__((aligned(CPU_L1_SIZE))); + +/* + * When a cache is created, its CPU pool type is determined from the buffer + * size. For small buffer sizes, many objects can be cached in a CPU pool. + * Conversely, for large buffer sizes, this would incur much overhead, so only + * a few objects are stored in a CPU pool. + */ +struct kmem_cpu_pool_type { + size_t buf_size; + int array_size; + size_t array_align; + struct kmem_cache *array_cache; +}; +#endif /* SLAB_USE_CPU_POOLS */ + +/* + * Buffer descriptor. + * + * For normal caches (i.e. without SLAB_CF_VERIFY), bufctls are located at the + * end of (but inside) each buffer. If SLAB_CF_VERIFY is set, bufctls are + * located after each buffer. + * + * When an object is allocated to a client, its bufctl isn't used. This memory + * is instead used for redzoning if cache debugging is in effect. + */ +union kmem_bufctl { + union kmem_bufctl *next; + unsigned long redzone; +}; + +/* + * Buffer tag. + * + * This structure is only used for SLAB_CF_VERIFY caches. It is located after + * the bufctl and includes information about the state of the buffer it + * describes (allocated or not). It should be thought of as a debugging + * extension of the bufctl. + */ +struct kmem_buftag { + unsigned long state; +}; + +/* + * Page-aligned collection of unconstructed buffers. + */ +struct kmem_slab { + struct list list_node; + struct rbtree_node tree_node; + unsigned long nr_refs; + union kmem_bufctl *first_free; + void *addr; +}; + +/* + * Type for constructor functions. + * + * The pre-constructed state of an object is supposed to include only + * elements such as e.g. linked lists, locks, reference counters. Therefore + * constructors are expected to 1) never fail and 2) not need any + * user-provided data. The first constraint implies that object construction + * never performs dynamic resource allocation, which also means there is no + * need for destructors. + */ +typedef void (*kmem_cache_ctor_t)(void *obj); + +/* + * Types for slab allocation/free functions. + * + * All addresses and sizes must be page-aligned. + */ +typedef vm_offset_t (*kmem_slab_alloc_fn_t)(vm_size_t); +typedef void (*kmem_slab_free_fn_t)(vm_offset_t, vm_size_t); + +/* + * Cache name buffer size. + */ +#define KMEM_CACHE_NAME_SIZE 32 + +/* + * Cache of objects. + * + * Locking order : cpu_pool -> cache. CPU pools locking is ordered by CPU ID. + * + * The partial slabs list is sorted by slab references. Slabs with a high + * number of references are placed first on the list to reduce fragmentation. + * Sorting occurs at insertion/removal of buffers in a slab. As the list + * is maintained sorted, and the number of references only changes by one, + * this is a very cheap operation in the average case and the worst (linear) + * case is very unlikely. + */ +struct kmem_cache { +#if SLAB_USE_CPU_POOLS + /* CPU pool layer */ + struct kmem_cpu_pool cpu_pools[NCPUS]; + struct kmem_cpu_pool_type *cpu_pool_type; +#endif /* SLAB_USE_CPU_POOLS */ + + /* Slab layer */ + simple_lock_data_t lock; + struct list node; /* Cache list linkage */ + struct list partial_slabs; + struct list free_slabs; + struct rbtree active_slabs; + int flags; + size_t obj_size; /* User-provided size */ + size_t align; + size_t buf_size; /* Aligned object size */ + size_t bufctl_dist; /* Distance from buffer to bufctl */ + size_t slab_size; + size_t color; + size_t color_max; + unsigned long bufs_per_slab; + unsigned long nr_objs; /* Number of allocated objects */ + unsigned long nr_bufs; /* Total number of buffers */ + unsigned long nr_slabs; + unsigned long nr_free_slabs; + kmem_cache_ctor_t ctor; + kmem_slab_alloc_fn_t slab_alloc_fn; + kmem_slab_free_fn_t slab_free_fn; + char name[KMEM_CACHE_NAME_SIZE]; + size_t buftag_dist; /* Distance from buffer to buftag */ + size_t redzone_pad; /* Bytes from end of object to redzone word */ +}; + +/* + * Mach-style declarations for struct kmem_cache. + */ +typedef struct kmem_cache *kmem_cache_t; +#define KMEM_CACHE_NULL ((kmem_cache_t) 0) + +/* + * VM submap for slab allocations. + */ +extern vm_map_t kmem_map; + +/* + * Cache initialization flags. + */ +#define KMEM_CACHE_NOCPUPOOL 0x1 /* Don't use the per-cpu pools */ +#define KMEM_CACHE_NOOFFSLAB 0x2 /* Don't allocate external slab data */ +#define KMEM_CACHE_NORECLAIM 0x4 /* Never give slabs back to their source, + implies KMEM_CACHE_NOOFFSLAB */ +#define KMEM_CACHE_VERIFY 0x8 /* Use debugging facilities */ + +/* + * Initialize a cache. + */ +void kmem_cache_init(struct kmem_cache *cache, const char *name, + size_t obj_size, size_t align, kmem_cache_ctor_t ctor, + kmem_slab_alloc_fn_t slab_alloc_fn, + kmem_slab_free_fn_t slab_free_fn, int flags); + +/* + * Allocate an object from a cache. + */ +vm_offset_t kmem_cache_alloc(struct kmem_cache *cache); + +/* + * Release an object to its cache. + */ +void kmem_cache_free(struct kmem_cache *cache, vm_offset_t obj); + +/* + * Initialize the memory allocator module. + */ +void slab_bootstrap(void); +void slab_init(void); + +/* + * Release free slabs to the VM system. + */ +void slab_collect(void); + +/* + * Display a summary of all kernel caches. + */ +void slab_info(void); + +#endif /* _KERN_SLAB_H */ diff --git a/kern/startup.c b/kern/startup.c index 75a05b0e..0c04ccf8 100644 --- a/kern/startup.c +++ b/kern/startup.c @@ -48,7 +48,6 @@ #include <kern/timer.h> #include <kern/xpr.h> #include <kern/time_stamp.h> -#include <kern/zalloc.h> #include <vm/vm_kern.h> #include <vm/vm_map.h> #include <vm/vm_object.h> @@ -146,10 +145,10 @@ void setup_main() timestamp_init(); - mapable_time_init(); - machine_init(); + mapable_time_init(); + machine_info.max_cpus = NCPUS; machine_info.memory_size = phys_last_addr - phys_first_addr; /* XXX mem_size */ machine_info.avail_cpus = 0; diff --git a/kern/strings.c b/kern/strings.c index 80e410e2..3676f98e 100644 --- a/kern/strings.c +++ b/kern/strings.c @@ -172,3 +172,22 @@ strlen( return string - 1 - ret; } + +/* + * Abstract: + * memset writes value "c" in the "n" bytes starting at address "s". + * The return value is a pointer to the "s" string. + */ + +void * +memset( + void *_s, int c, size_t n) +{ + char *s = _s; + int i; + + for (i = 0; i < n ; i++) + s[i] = c; + + return _s; +} diff --git a/kern/syscall_emulation.c b/kern/syscall_emulation.c index 06781d5b..c1c3096c 100644 --- a/kern/syscall_emulation.c +++ b/kern/syscall_emulation.c @@ -336,24 +336,6 @@ task_set_emulation_vector(task, vector_start, emulation_vector, } /* - * Compatibility entry. Vector is passed inline. - */ -kern_return_t -xxx_task_set_emulation_vector(task, vector_start, emulation_vector, - emulation_vector_count) - task_t task; - int vector_start; - emulation_vector_t emulation_vector; - unsigned int emulation_vector_count; -{ - return task_set_emulation_vector_internal( - task, - vector_start, - emulation_vector, - emulation_vector_count); -} - -/* * task_get_emulation_vector: [Server Entry] * * Get the list of emulated system calls for this task. @@ -460,53 +442,6 @@ task_get_emulation_vector(task, vector_start, emulation_vector, } /* - * xxx_task_get_emulation: [Server Entry] - * get the list of emulated system calls for this task. - * Compatibility code: return list in-line. - */ -kern_return_t -xxx_task_get_emulation_vector(task, vector_start, emulation_vector, - emulation_vector_count) - task_t task; - int *vector_start; - emulation_vector_t emulation_vector; /* pointer to OUT array */ - unsigned int *emulation_vector_count; /*IN/OUT*/ -{ - register eml_dispatch_t eml; - - if (task == TASK_NULL) - return( EML_BAD_TASK ); - - task_lock(task); - - eml = task->eml_dispatch; - if (eml == EML_DISPATCH_NULL) { - task_unlock(task); - *vector_start = 0; - *emulation_vector_count = 0; - return( KERN_SUCCESS ); - } - - simple_lock(&eml->lock); - - if (*emulation_vector_count < eml->disp_count) { - simple_unlock(&eml->lock); - task_unlock(task); - return( EML_BAD_CNT ); - } - - *vector_start = eml->disp_min; - *emulation_vector_count = eml->disp_count; - memcpy(emulation_vector, eml->disp_vector, - *emulation_vector_count * sizeof(vm_offset_t)); - simple_unlock(&eml->lock); - - task_unlock(task); - - return( KERN_SUCCESS ); -} - -/* * task_set_emulation: [Server Entry] * set up for user space emulation of syscalls within this task. */ diff --git a/kern/syscall_subr.c b/kern/syscall_subr.c index 395b9b8f..ae2d7d73 100644 --- a/kern/syscall_subr.c +++ b/kern/syscall_subr.c @@ -34,6 +34,7 @@ #include <kern/counters.h> #include <kern/ipc_kobject.h> #include <kern/mach_clock.h> +#include <kern/printf.h> #include <kern/processor.h> #include <kern/sched.h> #include <kern/sched_prim.h> @@ -380,3 +381,18 @@ register thread_t thread; (void) splx(s); return(KERN_SUCCESS); } + +/* + * mach_print + * + * Display a null-terminated character string on the Mach console. + * This system call is meant as a debugging tool useful to circumvent + * messaging altogether. + */ +#ifdef MACH_KDB +void +mach_print(const char *s) +{ + printf("%s", s); +} +#endif /* MACH_KDB */ diff --git a/kern/syscall_subr.h b/kern/syscall_subr.h index 2b8bcd36..a2e39205 100644 --- a/kern/syscall_subr.h +++ b/kern/syscall_subr.h @@ -36,5 +36,6 @@ extern int swtch_pri(int); extern int thread_switch(mach_port_t, int, mach_msg_timeout_t); extern void thread_depress_timeout(thread_t); extern kern_return_t thread_depress_abort(thread_t); +extern void mach_print(const char *); #endif /* _KERN_SYSCALL_SUBR_H_ */ diff --git a/kern/syscall_sw.c b/kern/syscall_sw.c index b2e20e66..607d843e 100644 --- a/kern/syscall_sw.c +++ b/kern/syscall_sw.c @@ -122,8 +122,12 @@ mach_trap_t mach_trap_table[] = { MACH_TRAP(mach_thread_self, 0), /* 27 */ MACH_TRAP(mach_task_self, 0), /* 28 */ MACH_TRAP(mach_host_self, 0), /* 29 */ +#ifdef MACH_KDB + MACH_TRAP_STACK(mach_print, 1), /* 30 */ +#else /* MACH_KDB */ + MACH_TRAP_STACK(kern_invalid, 0), /* 30 */ +#endif /* MACH_KDB */ - MACH_TRAP(kern_invalid, 0), /* 30 */ MACH_TRAP(kern_invalid, 0), /* 31 */ MACH_TRAP(kern_invalid, 0), /* 32 */ MACH_TRAP(kern_invalid, 0), /* 33 emul: task_by_pid */ diff --git a/kern/task.c b/kern/task.c index 88da16e8..114dd319 100644 --- a/kern/task.c +++ b/kern/task.c @@ -40,10 +40,9 @@ #include <ipc/ipc_space.h> #include <ipc/ipc_types.h> #include <kern/debug.h> -#include <kern/mach_param.h> #include <kern/task.h> #include <kern/thread.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <kern/kalloc.h> #include <kern/processor.h> #include <kern/sched_prim.h> /* for thread_wakeup */ @@ -52,7 +51,7 @@ #include <machine/machspl.h> /* for splsched */ task_t kernel_task = TASK_NULL; -zone_t task_zone; +struct kmem_cache task_cache; extern void eml_init(void); extern void eml_task_reference(task_t, task_t); @@ -60,11 +59,8 @@ extern void eml_task_deallocate(task_t); void task_init(void) { - task_zone = zinit( - sizeof(struct task), 0, - TASK_MAX * sizeof(struct task), - TASK_CHUNK * sizeof(struct task), - 0, "tasks"); + kmem_cache_init(&task_cache, "task", sizeof(struct task), 0, + NULL, NULL, NULL, 0); eml_init(); machine_task_module_init (); @@ -77,38 +73,6 @@ void task_init(void) (void) task_create(TASK_NULL, FALSE, &kernel_task); } -/* - * Create a task running in the kernel address space. It may - * have its own map of size mem_size (if 0, it uses the kernel map), - * and may have ipc privileges. - */ -task_t kernel_task_create( - task_t parent_task, - vm_size_t map_size) -{ - task_t new_task; - vm_offset_t min, max; - - /* - * Create the task. - */ - (void) task_create(parent_task, FALSE, &new_task); - - /* - * Task_create creates the task with a user-space map. - * Remove the map and replace it with the kernel map - * or a submap of the kernel map. - */ - vm_map_deallocate(new_task->map); - if (map_size == 0) - new_task->map = kernel_map; - else - new_task->map = kmem_suballoc(kernel_map, &min, &max, - map_size, FALSE); - - return new_task; -} - kern_return_t task_create( task_t parent_task, boolean_t inherit_memory, @@ -120,7 +84,7 @@ kern_return_t task_create( int i; #endif - new_task = (task_t) zalloc(task_zone); + new_task = (task_t) kmem_cache_alloc(&task_cache); if (new_task == TASK_NULL) { panic("task_create: no memory for task structure"); } @@ -144,6 +108,13 @@ kern_return_t task_create( new_task->active = TRUE; new_task->user_stop_count = 0; new_task->thread_count = 0; + new_task->faults = 0; + new_task->zero_fills = 0; + new_task->reactivations = 0; + new_task->pageins = 0; + new_task->cow_faults = 0; + new_task->messages_sent = 0; + new_task->messages_received = 0; eml_task_reference(new_task, parent_task); @@ -235,7 +206,7 @@ void task_deallocate( pset_deallocate(pset); vm_map_deallocate(task->map); is_release(task->itk_space); - zfree(task_zone, (vm_offset_t) task); + kmem_cache_free(&task_cache, (vm_offset_t) task); } void task_reference( @@ -783,6 +754,30 @@ kern_return_t task_info( break; } + case TASK_EVENTS_INFO: + { + register task_events_info_t event_info; + + if (*task_info_count < TASK_EVENTS_INFO_COUNT) { + return KERN_INVALID_ARGUMENT; + } + + event_info = (task_events_info_t) task_info_out; + + task_lock(&task); + event_info->faults = task->faults; + event_info->zero_fills = task->zero_fills; + event_info->reactivations = task->reactivations; + event_info->pageins = task->pageins; + event_info->cow_faults = task->cow_faults; + event_info->messages_sent = task->messages_sent; + event_info->messages_received = task->messages_received; + task_unlock(&task); + + *task_info_count = TASK_EVENTS_INFO_COUNT; + break; + } + case TASK_THREAD_TIMES_INFO: { register task_thread_times_info_t times_info; diff --git a/kern/task.h b/kern/task.h index 9d902435..9bfea571 100644 --- a/kern/task.h +++ b/kern/task.h @@ -102,6 +102,15 @@ struct task { /* Hardware specific data. */ machine_task_t machine; + + /* Statistics */ + natural_t faults; /* page faults counter */ + natural_t zero_fills; /* zero fill pages counter */ + natural_t reactivations; /* reactivated pages counter */ + natural_t pageins; /* actual pageins couter */ + natural_t cow_faults; /* copy-on-write faults counter */ + natural_t messages_sent; /* messages sent counter */ + natural_t messages_received; /* messages received counter */ }; #define task_lock(task) simple_lock(&(task)->lock) @@ -162,8 +171,6 @@ extern kern_return_t task_hold(task_t); extern kern_return_t task_dowait(task_t, boolean_t); extern kern_return_t task_release(task_t); -extern task_t kernel_task_create(task_t, vm_size_t); - extern task_t kernel_task; #endif /* _KERN_TASK_H_ */ diff --git a/kern/thread.c b/kern/thread.c index 1548e143..87be9231 100644 --- a/kern/thread.c +++ b/kern/thread.c @@ -45,7 +45,6 @@ #include <kern/eventcount.h> #include <kern/ipc_mig.h> #include <kern/ipc_tt.h> -#include <kern/mach_param.h> #include <kern/processor.h> #include <kern/queue.h> #include <kern/sched.h> @@ -54,7 +53,8 @@ #include <kern/thread.h> #include <kern/thread_swap.h> #include <kern/host.h> -#include <kern/zalloc.h> +#include <kern/kalloc.h> +#include <kern/slab.h> #include <kern/mach_clock.h> #include <vm/vm_kern.h> #include <ipc/ipc_kmsg.h> @@ -67,7 +67,7 @@ thread_t active_threads[NCPUS]; vm_offset_t active_stacks[NCPUS]; -struct zone *thread_zone; +struct kmem_cache thread_cache; queue_head_t reaper_queue; decl_simple_lock_data(, reaper_lock) @@ -208,7 +208,7 @@ void stack_alloc( * addresses of a stack given an address in the middle. */ - if (kmem_alloc_aligned(kernel_map, &stack, KERNEL_STACK_SIZE) + if (kmem_alloc_aligned(kmem_map, &stack, KERNEL_STACK_SIZE) != KERN_SUCCESS) panic("stack_alloc"); @@ -268,7 +268,7 @@ void stack_collect(void) #if MACH_DEBUG stack_finalize(stack); #endif /* MACH_DEBUG */ - kmem_free(kernel_map, stack, KERNEL_STACK_SIZE); + kmem_free(kmem_map, stack, KERNEL_STACK_SIZE); s = splsched(); stack_lock(); @@ -300,11 +300,8 @@ void stack_privilege( void thread_init(void) { - thread_zone = zinit( - sizeof(struct thread), 0, - THREAD_MAX * sizeof(struct thread), - THREAD_CHUNK * sizeof(struct thread), - 0, "threads"); + kmem_cache_init(&thread_cache, "thread", sizeof(struct thread), 0, + NULL, NULL, NULL, 0); /* * Fill in a template thread for fast initialization. @@ -414,7 +411,7 @@ kern_return_t thread_create( * Allocate a thread and initialize static fields */ - new_thread = (thread_t) zalloc(thread_zone); + new_thread = (thread_t) kmem_cache_alloc(&thread_cache); if (new_thread == THREAD_NULL) return KERN_RESOURCE_SHORTAGE; @@ -710,7 +707,7 @@ void thread_deallocate( evc_notify_abort(thread); pcb_terminate(thread); - zfree(thread_zone, (vm_offset_t) thread); + kmem_cache_free(&thread_cache, (vm_offset_t) thread); } void thread_reference( @@ -1329,6 +1326,13 @@ kern_return_t thread_suspend( hold = FALSE; spl = splsched(); thread_lock(thread); + /* Wait for thread to get interruptible */ + while (thread->state & TH_UNINT) { + assert_wait(&thread->state, TRUE); + thread_unlock(thread); + thread_block(NULL); + thread_lock(thread); + } if (thread->user_stop_count++ == 0) { hold = TRUE; thread->suspend_count++; @@ -2395,7 +2399,7 @@ kern_return_t host_stack_usage( vm_size_t *maxusagep, vm_offset_t *maxstackp) { - unsigned int total; + natural_t total; vm_size_t maxusage; if (host == HOST_NULL) diff --git a/kern/zalloc.c b/kern/zalloc.c deleted file mode 100644 index a95c7f6b..00000000 --- a/kern/zalloc.c +++ /dev/null @@ -1,1007 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1993-1987 Carnegie Mellon University. - * Copyright (c) 1993,1994 The University of Utah and - * the Computer Systems Laboratory (CSL). - * All rights reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON, THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF - * THIS SOFTWARE IN ITS "AS IS" CONDITION, AND DISCLAIM ANY LIABILITY - * OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF - * THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * File: kern/zalloc.c - * Author: Avadis Tevanian, Jr. - * - * Zone-based memory allocator. A zone is a collection of fixed size - * data blocks for which quick allocation/deallocation is possible. - */ - -#include <string.h> - -#include <kern/debug.h> -#include <kern/macro_help.h> -#include <kern/printf.h> -#include <kern/mach_clock.h> -#include <kern/sched.h> -#include <kern/zalloc.h> -#include <mach/vm_param.h> -#include <vm/vm_kern.h> -#include <machine/machspl.h> - -#if MACH_DEBUG -#include <mach/kern_return.h> -#include <mach/machine/vm_types.h> -#include <mach_debug/zone_info.h> -#include <kern/host.h> -#include <vm/vm_map.h> -#include <vm/vm_user.h> -#include <vm/vm_kern.h> -#endif - -#define ADD_TO_ZONE(zone, element) \ -MACRO_BEGIN \ - *((vm_offset_t *)(element)) = (zone)->free_elements; \ - (zone)->free_elements = (vm_offset_t) (element); \ - zone_count_down(zone); \ -MACRO_END - -#define REMOVE_FROM_ZONE(zone, ret, type) \ -MACRO_BEGIN \ - (ret) = (type) (zone)->free_elements; \ - if ((ret) != (type) 0) { \ - zone_count_up(zone); \ - (zone)->free_elements = *((vm_offset_t *)(ret)); \ - } \ -MACRO_END - -#define ALIGN_SIZE_UP(size, align) \ -((size) = (((size) + ((align) - 1)) & ~((align) - 1))) - -/* - * Support for garbage collection of unused zone pages: - */ - -struct zone_page_table_entry { - struct zone_page_table_entry *next; - short in_free_list; - short alloc_count; -}; - -extern struct zone_page_table_entry * zone_page_table; -extern vm_offset_t zone_map_min_address; - -#define lock_zone_page_table() simple_lock(&zone_page_table_lock) -#define unlock_zone_page_table() simple_unlock(&zone_page_table_lock) - -#define zone_page(addr) \ - (&(zone_page_table[(atop(((vm_offset_t)addr) - zone_map_min_address))])) - - -extern void zone_page_alloc(vm_offset_t, vm_size_t); -extern void zone_page_dealloc(vm_offset_t, vm_size_t); -extern void zone_page_in_use(vm_offset_t, vm_size_t); -extern void zone_page_free(vm_offset_t, vm_size_t); - -zone_t zone_zone; /* this is the zone containing other zones */ - -boolean_t zone_ignore_overflow = TRUE; - -vm_map_t zone_map = VM_MAP_NULL; -vm_size_t zone_map_size = 64 * 1024 * 1024; - -/* - * The VM system gives us an initial chunk of memory. - * It has to be big enough to allocate the zone_zone - * and some initial kernel data structures, like kernel maps. - * It is advantageous to make it bigger than really necessary, - * because this memory is more efficient than normal kernel - * virtual memory. (It doesn't have vm_page structures backing it - * and it may have other machine-dependent advantages.) - * So for best performance, zdata_size should approximate - * the amount of memory you expect the zone system to consume. - */ - -vm_offset_t zdata; -vm_size_t zdata_size = 420 * 1024; - -#define zone_lock(zone) \ -MACRO_BEGIN \ - if (zone->type & ZONE_PAGEABLE) { \ - lock_write(&zone->complex_lock); \ - } else { \ - simple_lock(&zone->lock); \ - } \ -MACRO_END - -#define zone_unlock(zone) \ -MACRO_BEGIN \ - if (zone->type & ZONE_PAGEABLE) { \ - lock_done(&zone->complex_lock); \ - } else { \ - simple_unlock(&zone->lock); \ - } \ -MACRO_END - -#define zone_lock_init(zone) \ -MACRO_BEGIN \ - if (zone->type & ZONE_PAGEABLE) { \ - lock_init(&zone->complex_lock, TRUE); \ - } else { \ - simple_lock_init(&zone->lock); \ - } \ -MACRO_END - -static vm_offset_t zget_space(vm_offset_t size, vm_size_t align); - -decl_simple_lock_data(,zget_space_lock) -vm_offset_t zalloc_next_space; -vm_offset_t zalloc_end_of_space; -vm_size_t zalloc_wasted_space; - -/* - * Garbage collection map information - */ -decl_simple_lock_data(,zone_page_table_lock) -struct zone_page_table_entry * zone_page_table; -vm_offset_t zone_map_min_address; -vm_offset_t zone_map_max_address; -int zone_pages; - -extern void zone_page_init(vm_offset_t, vm_size_t, int); - -#define ZONE_PAGE_USED 0 -#define ZONE_PAGE_UNUSED -1 - - -/* - * Protects first_zone, last_zone, num_zones, - * and the next_zone field of zones. - */ -decl_simple_lock_data(,all_zones_lock) -zone_t first_zone; -zone_t *last_zone; -int num_zones; - -/* - * zinit initializes a new zone. The zone data structures themselves - * are stored in a zone, which is initially a static structure that - * is initialized by zone_init. - */ -zone_t zinit(size, align, max, alloc, memtype, name) - vm_size_t size; /* the size of an element */ - vm_size_t align; /* alignment of elements */ - vm_size_t max; /* maximum memory to use */ - vm_size_t alloc; /* allocation size */ - unsigned int memtype; /* flags specifying type of memory */ - char *name; /* a name for the zone */ -{ - register zone_t z; - - if (zone_zone == ZONE_NULL) - z = (zone_t) zget_space(sizeof(struct zone), 0); - else - z = (zone_t) zalloc(zone_zone); - if (z == ZONE_NULL) - panic("zinit"); - if (alloc == 0) - alloc = PAGE_SIZE; - - if (size == 0) - size = sizeof(z->free_elements); - /* - * Round off all the parameters appropriately. - */ - - if ((max = round_page(max)) < (alloc = round_page(alloc))) - max = alloc; - - if (align > 0) { - if (PAGE_SIZE % align || align % sizeof(z->free_elements)) - panic("zinit"); - ALIGN_SIZE_UP(size, align); - } - - z->free_elements = 0; - z->cur_size = 0; - z->max_size = max; - z->elem_size = ((size-1) + sizeof(z->free_elements)) - - ((size-1) % sizeof(z->free_elements)); - z->align = align; - - z->alloc_size = alloc; - z->type = memtype; - z->zone_name = name; -#ifdef ZONE_COUNT - z->count = 0; -#endif - z->doing_alloc = FALSE; - zone_lock_init(z); - - /* - * Add the zone to the all-zones list. - */ - - z->next_zone = ZONE_NULL; - simple_lock(&all_zones_lock); - *last_zone = z; - last_zone = &z->next_zone; - num_zones++; - simple_unlock(&all_zones_lock); - - return(z); -} - -/* - * Cram the given memory into the specified zone. - */ -void zcram(zone_t zone, vm_offset_t newmem, vm_size_t size) -{ - register vm_size_t elem_size; - - if (newmem == (vm_offset_t) 0) { - panic("zcram - memory at zero"); - } - elem_size = zone->elem_size; - - zone_lock(zone); - while (size >= elem_size) { - ADD_TO_ZONE(zone, newmem); - zone_page_alloc(newmem, elem_size); - zone_count_up(zone); /* compensate for ADD_TO_ZONE */ - size -= elem_size; - newmem += elem_size; - zone->cur_size += elem_size; - } - zone_unlock(zone); -} - -/* - * Contiguous space allocator for non-paged zones. Allocates "size" amount - * of memory from zone_map. - */ - -static vm_offset_t zget_space(vm_offset_t size, vm_size_t align) -{ - vm_offset_t new_space = 0; - vm_offset_t result; - vm_size_t space_to_add = 0; /*'=0' to quiet gcc warnings */ - - simple_lock(&zget_space_lock); - if (align > 0) { - assert(align < PAGE_SIZE); - ALIGN_SIZE_UP(zalloc_next_space, align); - } - - while ((zalloc_next_space + size) > zalloc_end_of_space) { - /* - * Add at least one page to allocation area. - */ - - space_to_add = round_page(size); - - if (new_space == 0) { - /* - * Memory cannot be wired down while holding - * any locks that the pageout daemon might - * need to free up pages. [Making the zget_space - * lock a complex lock does not help in this - * regard.] - * - * Unlock and allocate memory. Because several - * threads might try to do this at once, don't - * use the memory before checking for available - * space again. - */ - - simple_unlock(&zget_space_lock); - - if (kmem_alloc_wired(zone_map, - &new_space, space_to_add) - != KERN_SUCCESS) - return(0); - zone_page_init(new_space, space_to_add, - ZONE_PAGE_USED); - simple_lock(&zget_space_lock); - if (align > 0) - ALIGN_SIZE_UP(zalloc_next_space, align); - continue; - } - - - /* - * Memory was allocated in a previous iteration. - * - * Check whether the new region is contiguous - * with the old one. - */ - - if (new_space != zalloc_end_of_space) { - /* - * Throw away the remainder of the - * old space, and start a new one. - */ - zalloc_wasted_space += - zalloc_end_of_space - zalloc_next_space; - zalloc_next_space = new_space; - } - - zalloc_end_of_space = new_space + space_to_add; - - new_space = 0; - } - result = zalloc_next_space; - zalloc_next_space += size; - simple_unlock(&zget_space_lock); - - if (new_space != 0) - kmem_free(zone_map, new_space, space_to_add); - - return(result); -} - - -/* - * Initialize the "zone of zones" which uses fixed memory allocated - * earlier in memory initialization. zone_bootstrap is called - * before zone_init. - */ -void zone_bootstrap(void) -{ - simple_lock_init(&all_zones_lock); - first_zone = ZONE_NULL; - last_zone = &first_zone; - num_zones = 0; - - simple_lock_init(&zget_space_lock); - zalloc_next_space = zdata; - zalloc_end_of_space = zdata + zdata_size; - zalloc_wasted_space = 0; - - zone_zone = ZONE_NULL; - zone_zone = zinit(sizeof(struct zone), 0, 128 * sizeof(struct zone), - sizeof(struct zone), 0, "zones"); -} - -void zone_init(void) -{ - vm_offset_t zone_min; - vm_offset_t zone_max; - - vm_size_t zone_table_size; - - zone_map = kmem_suballoc(kernel_map, &zone_min, &zone_max, - zone_map_size, FALSE); - - /* - * Setup garbage collection information: - */ - - zone_table_size = atop(zone_max - zone_min) * - sizeof(struct zone_page_table_entry); - if (kmem_alloc_wired(zone_map, (vm_offset_t *) &zone_page_table, - zone_table_size) != KERN_SUCCESS) - panic("zone_init"); - zone_min = (vm_offset_t)zone_page_table + round_page(zone_table_size); - zone_pages = atop(zone_max - zone_min); - zone_map_min_address = zone_min; - zone_map_max_address = zone_max; - simple_lock_init(&zone_page_table_lock); - zone_page_init(zone_min, zone_max - zone_min, ZONE_PAGE_UNUSED); -} - - -/* - * zalloc returns an element from the specified zone. - */ -vm_offset_t zalloc(zone_t zone) -{ - vm_offset_t addr; - - if (zone == ZONE_NULL) - panic ("zalloc: null zone"); - - check_simple_locks(); - - zone_lock(zone); - REMOVE_FROM_ZONE(zone, addr, vm_offset_t); - while (addr == 0) { - /* - * If nothing was there, try to get more - */ - if (zone->doing_alloc) { - /* - * Someone is allocating memory for this zone. - * Wait for it to show up, then try again. - */ - assert_wait((event_t)&zone->doing_alloc, TRUE); - /* XXX say wakeup needed */ - zone_unlock(zone); - thread_block((void (*)()) 0); - zone_lock(zone); - } - else { - if ((zone->cur_size + (zone->type & ZONE_PAGEABLE ? - zone->alloc_size : zone->elem_size)) > - zone->max_size) { - if (zone->type & ZONE_EXHAUSTIBLE) - break; - /* - * Printf calls logwakeup, which calls - * select_wakeup which will do a zfree - * (which tries to take the select_zone - * lock... Hang. Release the lock now - * so it can be taken again later. - * NOTE: this used to be specific to - * the select_zone, but for - * cleanliness, we just unlock all - * zones before this. - */ - if (!(zone->type & ZONE_FIXED)) { - /* - * We're willing to overflow certain - * zones, but not without complaining. - * - * This is best used in conjunction - * with the collecatable flag. What we - * want is an assurance we can get the - * memory back, assuming there's no - * leak. - */ - zone->max_size += (zone->max_size >> 1); - } else if (!zone_ignore_overflow) { - zone_unlock(zone); - printf("zone \"%s\" empty.\n", - zone->zone_name); - panic("zalloc: zone %s exhausted", - zone->zone_name); - } - } - - if (zone->type & ZONE_PAGEABLE) - zone->doing_alloc = TRUE; - zone_unlock(zone); - - if (zone->type & ZONE_PAGEABLE) { - if (kmem_alloc_pageable(zone_map, &addr, - zone->alloc_size) - != KERN_SUCCESS) - panic("zalloc: zone %s exhausted", - zone->zone_name); - zcram(zone, addr, zone->alloc_size); - zone_lock(zone); - zone->doing_alloc = FALSE; - /* XXX check before doing this */ - thread_wakeup((event_t)&zone->doing_alloc); - - REMOVE_FROM_ZONE(zone, addr, vm_offset_t); - } else if (zone->type & ZONE_COLLECTABLE) { - if (kmem_alloc_wired(zone_map, - &addr, zone->alloc_size) - != KERN_SUCCESS) - panic("zalloc: zone %s exhausted", - zone->zone_name); - zone_page_init(addr, zone->alloc_size, - ZONE_PAGE_USED); - zcram(zone, addr, zone->alloc_size); - zone_lock(zone); - REMOVE_FROM_ZONE(zone, addr, vm_offset_t); - } else { - addr = zget_space(zone->elem_size, zone->align); - if (addr == 0) - panic("zalloc: zone %s exhausted", - zone->zone_name); - - zone_lock(zone); - zone_count_up(zone); - zone->cur_size += zone->elem_size; - zone_unlock(zone); - zone_page_alloc(addr, zone->elem_size); - return(addr); - } - } - } - - zone_unlock(zone); - return(addr); -} - - -/* - * zget returns an element from the specified zone - * and immediately returns nothing if there is nothing there. - * - * This form should be used when you can not block (like when - * processing an interrupt). - */ -vm_offset_t zget(zone_t zone) -{ - register vm_offset_t addr; - - if (zone == ZONE_NULL) - panic ("zalloc: null zone"); - - zone_lock(zone); - REMOVE_FROM_ZONE(zone, addr, vm_offset_t); - zone_unlock(zone); - - return(addr); -} - -boolean_t zone_check = FALSE; - -void zfree(zone_t zone, vm_offset_t elem) -{ - zone_lock(zone); - if (zone_check) { - vm_offset_t this; - - /* check the zone's consistency */ - - for (this = zone->free_elements; - this != 0; - this = * (vm_offset_t *) this) - if (this == elem) - panic("zfree"); - } - ADD_TO_ZONE(zone, elem); - zone_unlock(zone); -} - -/* - * Zone garbage collection subroutines - * - * These routines have in common the modification of entries in the - * zone_page_table. The latter contains one entry for every page - * in the zone_map. - * - * For each page table entry in the given range: - * - * zone_page_in_use - decrements in_free_list - * zone_page_free - increments in_free_list - * zone_page_init - initializes in_free_list and alloc_count - * zone_page_alloc - increments alloc_count - * zone_page_dealloc - decrements alloc_count - * zone_add_free_page_list - adds the page to the free list - * - * Two counts are maintained for each page, the in_free_list count and - * alloc_count. The alloc_count is how many zone elements have been - * allocated from a page. (Note that the page could contain elements - * that span page boundaries. The count includes these elements so - * one element may be counted in two pages.) In_free_list is a count - * of how many zone elements are currently free. If in_free_list is - * equal to alloc_count then the page is eligible for garbage - * collection. - * - * Alloc_count and in_free_list are initialized to the correct values - * for a particular zone when a page is zcram'ed into a zone. Subsequent - * gets and frees of zone elements will call zone_page_in_use and - * zone_page_free which modify the in_free_list count. When the zones - * garbage collector runs it will walk through a zones free element list, - * remove the elements that reside on collectable pages, and use - * zone_add_free_page_list to create a list of pages to be collected. - */ - -void zone_page_in_use(addr, size) -vm_offset_t addr; -vm_size_t size; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - zone_page_table[i].in_free_list--; - } - unlock_zone_page_table(); -} - -void zone_page_free(addr, size) -vm_offset_t addr; -vm_size_t size; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - /* Set in_free_list to (ZONE_PAGE_USED + 1) if - * it was previously set to ZONE_PAGE_UNUSED. - */ - if (zone_page_table[i].in_free_list == ZONE_PAGE_UNUSED) { - zone_page_table[i].in_free_list = 1; - } else { - zone_page_table[i].in_free_list++; - } - } - unlock_zone_page_table(); -} - -void zone_page_init(addr, size, value) - -vm_offset_t addr; -vm_size_t size; -int value; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - zone_page_table[i].alloc_count = value; - zone_page_table[i].in_free_list = 0; - } - unlock_zone_page_table(); -} - -void zone_page_alloc(addr, size) -vm_offset_t addr; -vm_size_t size; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - /* Set alloc_count to (ZONE_PAGE_USED + 1) if - * it was previously set to ZONE_PAGE_UNUSED. - */ - if (zone_page_table[i].alloc_count == ZONE_PAGE_UNUSED) { - zone_page_table[i].alloc_count = 1; - } else { - zone_page_table[i].alloc_count++; - } - } - unlock_zone_page_table(); -} - -void zone_page_dealloc(addr, size) -vm_offset_t addr; -vm_size_t size; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - zone_page_table[i].alloc_count--; - } - unlock_zone_page_table(); -} - -void -zone_add_free_page_list(free_list, addr, size) - struct zone_page_table_entry **free_list; - vm_offset_t addr; - vm_size_t size; -{ - int i, j; - if ((addr < zone_map_min_address) || - (addr+size > zone_map_max_address)) return; - i = atop(addr-zone_map_min_address); - j = atop((addr+size-1) - zone_map_min_address); - lock_zone_page_table(); - for (; i <= j; i++) { - if (zone_page_table[i].alloc_count == 0) { - zone_page_table[i].next = *free_list; - *free_list = &zone_page_table[i]; - zone_page_table[i].alloc_count = ZONE_PAGE_UNUSED; - zone_page_table[i].in_free_list = 0; - } - } - unlock_zone_page_table(); -} - - -/* This is used for walking through a zone's free element list. - */ -struct zone_free_entry { - struct zone_free_entry * next; -}; - - -/* Zone garbage collection - * - * zone_gc will walk through all the free elements in all the - * zones that are marked collectable looking for reclaimable - * pages. zone_gc is called by consider_zone_gc when the system - * begins to run out of memory. - */ -static void zone_gc(void) -{ - int max_zones; - zone_t z; - int i; - register spl_t s; - struct zone_page_table_entry *freep; - struct zone_page_table_entry *zone_free_page_list; - - simple_lock(&all_zones_lock); - max_zones = num_zones; - z = first_zone; - simple_unlock(&all_zones_lock); - - zone_free_page_list = (struct zone_page_table_entry *) 0; - - for (i = 0; i < max_zones; i++) { - struct zone_free_entry * last; - struct zone_free_entry * elt; - assert(z != ZONE_NULL); - /* run this at splhigh so that interupt routines that use zones - can not interupt while their zone is locked */ - s=splhigh(); - zone_lock(z); - - if ((z->type & (ZONE_PAGEABLE|ZONE_COLLECTABLE)) == ZONE_COLLECTABLE) { - - /* Count the free elements in each page. This loop - * requires that all in_free_list entries are zero. - */ - elt = (struct zone_free_entry *)(z->free_elements); - while ((elt != (struct zone_free_entry *)0)) { - zone_page_free((vm_offset_t)elt, z->elem_size); - elt = elt->next; - } - - /* Now determine which elements should be removed - * from the free list and, after all the elements - * on a page have been removed, add the element's - * page to a list of pages to be freed. - */ - elt = (struct zone_free_entry *)(z->free_elements); - last = elt; - while ((elt != (struct zone_free_entry *)0)) { - if (((vm_offset_t)elt>=zone_map_min_address)&& - ((vm_offset_t)elt<=zone_map_max_address)&& - (zone_page(elt)->in_free_list == - zone_page(elt)->alloc_count)) { - - z->cur_size -= z->elem_size; - zone_page_in_use((vm_offset_t)elt, z->elem_size); - zone_page_dealloc((vm_offset_t)elt, z->elem_size); - if (zone_page(elt)->alloc_count == 0 || - zone_page(elt+(z->elem_size-1))->alloc_count==0) { - zone_add_free_page_list( - &zone_free_page_list, - (vm_offset_t)elt, z->elem_size); - } - - - if (elt == last) { - elt = elt->next; - z->free_elements =(vm_offset_t)elt; - last = elt; - } else { - last->next = elt->next; - elt = elt->next; - } - } else { - /* This element is not eligible for collection - * so clear in_free_list in preparation for a - * subsequent garbage collection pass. - */ - if (((vm_offset_t)elt>=zone_map_min_address)&& - ((vm_offset_t)elt<=zone_map_max_address)) { - zone_page(elt)->in_free_list = 0; - } - last = elt; - elt = elt->next; - } - } - } - zone_unlock(z); - splx(s); - simple_lock(&all_zones_lock); - z = z->next_zone; - simple_unlock(&all_zones_lock); - } - - for (freep = zone_free_page_list; freep != 0; freep = freep->next) { - vm_offset_t free_addr; - - free_addr = zone_map_min_address + - PAGE_SIZE * (freep - zone_page_table); - - /* Hack Hack */ - /* Needed to make vm_map_delete's vm_map_clip_end always be - * able to get an element without having to call zget_space and - * hang because zone_map is already locked by vm_map_delete */ - - extern zone_t vm_map_kentry_zone; /* zone for kernel entry structures */ - vm_offset_t entry1 = zalloc(vm_map_kentry_zone), - entry2 = zalloc(vm_map_kentry_zone); - zfree(vm_map_kentry_zone, entry1); - zfree(vm_map_kentry_zone, entry2); - - kmem_free(zone_map, free_addr, PAGE_SIZE); - } -} - -boolean_t zone_gc_allowed = TRUE; -unsigned zone_gc_last_tick = 0; -unsigned zone_gc_max_rate = 0; /* in ticks */ - -/* - * consider_zone_gc: - * - * Called by the pageout daemon when the system needs more free pages. - */ - -void -consider_zone_gc(void) -{ - /* - * By default, don't attempt zone GC more frequently - * than once a second. - */ - - if (zone_gc_max_rate == 0) - zone_gc_max_rate = hz; - - if (zone_gc_allowed && - (sched_tick > (zone_gc_last_tick + zone_gc_max_rate))) { - zone_gc_last_tick = sched_tick; - zone_gc(); - } -} - -#if MACH_DEBUG -kern_return_t host_zone_info(host, namesp, namesCntp, infop, infoCntp) - host_t host; - zone_name_array_t *namesp; - unsigned int *namesCntp; - zone_info_array_t *infop; - unsigned int *infoCntp; -{ - zone_name_t *names; - vm_offset_t names_addr; - vm_size_t names_size = 0; /*'=0' to quiet gcc warnings */ - zone_info_t *info; - vm_offset_t info_addr; - vm_size_t info_size = 0; /*'=0' to quiet gcc warnings */ - unsigned int max_zones, i; - zone_t z; - kern_return_t kr; - - if (host == HOST_NULL) - return KERN_INVALID_HOST; - - /* - * We assume that zones aren't freed once allocated. - * We won't pick up any zones that are allocated later. - */ - - simple_lock(&all_zones_lock); - max_zones = num_zones; - z = first_zone; - simple_unlock(&all_zones_lock); - - if (max_zones <= *namesCntp) { - /* use in-line memory */ - - names = *namesp; - } else { - names_size = round_page(max_zones * sizeof *names); - kr = kmem_alloc_pageable(ipc_kernel_map, - &names_addr, names_size); - if (kr != KERN_SUCCESS) - return kr; - - names = (zone_name_t *) names_addr; - } - - if (max_zones <= *infoCntp) { - /* use in-line memory */ - - info = *infop; - } else { - info_size = round_page(max_zones * sizeof *info); - kr = kmem_alloc_pageable(ipc_kernel_map, - &info_addr, info_size); - if (kr != KERN_SUCCESS) { - if (names != *namesp) - kmem_free(ipc_kernel_map, - names_addr, names_size); - return kr; - } - - info = (zone_info_t *) info_addr; - } - - for (i = 0; i < max_zones; i++) { - zone_name_t *zn = &names[i]; - zone_info_t *zi = &info[i]; - struct zone zcopy; - - assert(z != ZONE_NULL); - - zone_lock(z); - zcopy = *z; - zone_unlock(z); - - simple_lock(&all_zones_lock); - z = z->next_zone; - simple_unlock(&all_zones_lock); - - /* assuming here the name data is static */ - (void) strncpy(zn->zn_name, zcopy.zone_name, - sizeof zn->zn_name); - -#ifdef ZONE_COUNT - zi->zi_count = zcopy.count; -#else - zi->zi_count = 0; -#endif - zi->zi_cur_size = zcopy.cur_size; - zi->zi_max_size = zcopy.max_size; - zi->zi_elem_size = zcopy.elem_size; - zi->zi_alloc_size = zcopy.alloc_size; - zi->zi_pageable = (zcopy.type & ZONE_PAGEABLE) != 0; - zi->zi_exhaustible = (zcopy.type & ZONE_EXHAUSTIBLE) != 0; - zi->zi_collectable = (zcopy.type & ZONE_COLLECTABLE) != 0; - } - - if (names != *namesp) { - vm_size_t used; - vm_map_copy_t copy; - - used = max_zones * sizeof *names; - - if (used != names_size) - memset((char *) (names_addr + used), 0, names_size - used); - - kr = vm_map_copyin(ipc_kernel_map, names_addr, names_size, - TRUE, ©); - assert(kr == KERN_SUCCESS); - - *namesp = (zone_name_t *) copy; - } - *namesCntp = max_zones; - - if (info != *infop) { - vm_size_t used; - vm_map_copy_t copy; - - used = max_zones * sizeof *info; - - if (used != info_size) - memset((char *) (info_addr + used), 0, info_size - used); - - kr = vm_map_copyin(ipc_kernel_map, info_addr, info_size, - TRUE, ©); - assert(kr == KERN_SUCCESS); - - *infop = (zone_info_t *) copy; - } - *infoCntp = max_zones; - - return KERN_SUCCESS; -} -#endif /* MACH_DEBUG */ diff --git a/kern/zalloc.h b/kern/zalloc.h deleted file mode 100644 index f1a1850b..00000000 --- a/kern/zalloc.h +++ /dev/null @@ -1,136 +0,0 @@ -/* - * Mach Operating System - * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University. - * Copyright (c) 1993,1994 The University of Utah and - * the Computer Systems Laboratory (CSL). - * All rights reserved. - * - * Permission to use, copy, modify and distribute this software and its - * documentation is hereby granted, provided that both the copyright - * notice and this permission notice appear in all copies of the - * software, derivative works or modified versions, and any portions - * thereof, and that both notices appear in supporting documentation. - * - * CARNEGIE MELLON, THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF - * THIS SOFTWARE IN ITS "AS IS" CONDITION, AND DISCLAIM ANY LIABILITY - * OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF - * THIS SOFTWARE. - * - * Carnegie Mellon requests users of this software to return to - * - * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU - * School of Computer Science - * Carnegie Mellon University - * Pittsburgh PA 15213-3890 - * - * any improvements or extensions that they make and grant Carnegie Mellon - * the rights to redistribute these changes. - */ -/* - * File: zalloc.h - * Author: Avadis Tevanian, Jr. - * Date: 1985 - * - */ - -#ifndef _KERN_ZALLOC_H_ -#define _KERN_ZALLOC_H_ - -#include <mach/machine/vm_types.h> -#include <kern/macro_help.h> -#include <kern/lock.h> -#include <kern/queue.h> -#include <machine/zalloc.h> - -/* - * A zone is a collection of fixed size blocks for which there - * is fast allocation/deallocation access. Kernel routines can - * use zones to manage data structures dynamically, creating a zone - * for each type of data structure to be managed. - * - */ - -struct zone { - decl_simple_lock_data(,lock) /* generic lock */ -#ifdef ZONE_COUNT - int count; /* Number of elements used now */ -#endif - vm_offset_t free_elements; - vm_size_t cur_size; /* current memory utilization */ - vm_size_t max_size; /* how large can this zone grow */ - vm_size_t elem_size; /* size of an element */ - vm_size_t align; /* alignment of elements */ - vm_size_t alloc_size; /* size used for more memory */ - boolean_t doing_alloc; /* is zone expanding now? */ - char *zone_name; /* a name for the zone */ - unsigned int type; /* type of memory */ - lock_data_t complex_lock; /* Lock for pageable zones */ - struct zone *next_zone; /* Link for all-zones list */ -}; -typedef struct zone *zone_t; - -#define ZONE_NULL ((zone_t) 0) - -/* Exported to everyone */ -zone_t zinit(vm_size_t size, vm_size_t align, vm_size_t max, - vm_size_t alloc, unsigned int memtype, char *name); -vm_offset_t zalloc(zone_t zone); -vm_offset_t zget(zone_t zone); -void zfree(zone_t zone, vm_offset_t elem); -void zcram(zone_t zone, vm_offset_t newmem, vm_size_t size); - -/* Exported only to vm/vm_init.c */ -void zone_bootstrap(void); -void zone_init(void); - -/* Exported only to vm/vm_pageout.c */ -void consider_zone_gc(void); - - -/* Memory type bits for zones */ -#define ZONE_PAGEABLE 0x00000001 -#define ZONE_COLLECTABLE 0x00000002 /* Garbage-collect this zone when memory runs low */ -#define ZONE_EXHAUSTIBLE 0x00000004 /* zalloc() on this zone is allowed to fail */ -#define ZONE_FIXED 0x00000008 /* Panic if zone is exhausted (XXX) */ - -/* Machine-dependent code can provide additional memory types. */ -#define ZONE_MACHINE_TYPES 0xffff0000 - - -#ifdef ZONE_COUNT -#define zone_count_up(zone) ((zone)->count++) -#define zone_count_down(zone) ((zone)->count--) -#else -#define zone_count_up(zone) -#define zone_count_down(zone) -#endif - - - -/* These quick inline versions only work for small, nonpageable zones (currently). */ - -static __inline vm_offset_t ZALLOC(zone_t zone) -{ - simple_lock(&zone->lock); - if (zone->free_elements == 0) { - simple_unlock(&zone->lock); - return zalloc(zone); - } else { - vm_offset_t element = zone->free_elements; - zone->free_elements = *((vm_offset_t *)(element)); - zone_count_up(zone); - simple_unlock(&zone->lock); - return element; - } -} - -static __inline void ZFREE(zone_t zone, vm_offset_t element) -{ - *((vm_offset_t *)(element)) = zone->free_elements; - zone->free_elements = (vm_offset_t) (element); - zone_count_down(zone); -} - - - -#endif /* _KERN_ZALLOC_H_ */ diff --git a/linux/Makefrag.am b/linux/Makefrag.am index fccb8070..7c7b432d 100644 --- a/linux/Makefrag.am +++ b/linux/Makefrag.am @@ -1,6 +1,6 @@ # Makefile fragment for Linux device drivers and the glue code. -# Copyright (C) 2006, 2007 Free Software Foundation, Inc. +# Copyright (C) 2006, 2007, 2011 Free Software Foundation, Inc. # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by the @@ -57,6 +57,7 @@ liblinux_a_SOURCES = \ linux/dev/kernel/sched.c \ linux/dev/glue/kmem.c \ linux/dev/glue/block.c \ + linux/dev/glue/glue.h \ linux/dev/arch/i386/kernel/setup.c liblinux_a_SOURCES += \ @@ -259,6 +260,14 @@ liblinux_a_SOURCES += \ linux/src/drivers/scsi/seagate.h endif +if device_driver_sym53c8xx +liblinux_a_SOURCES += \ + linux/src/drivers/scsi/sym53c8xx.c \ + linux/src/drivers/scsi/sym53c8xx_comm.h \ + linux/src/drivers/scsi/sym53c8xx.h \ + linux/src/drivers/scsi/sym53c8xx_defs.h +endif + if device_driver_t128 liblinux_a_SOURCES += \ linux/src/drivers/scsi/t128.c \ diff --git a/linux/configfrag.ac b/linux/configfrag.ac index cf671fcc..882af6bc 100644 --- a/linux/configfrag.ac +++ b/linux/configfrag.ac @@ -50,6 +50,11 @@ dnl USE OF THIS SOFTWARE. esac }] +AC_DEFUN([AC_OPTION_Linux_group], [ +AC_ARG_ENABLE([$1-group], + AS_HELP_STRING([--enable-$1-group], [$2])) +]) + # # AC_OPTION_Linux_ix86_at(name,description,option[,class]). Process # configuration option --enable-`name' (with description `description'). If @@ -62,7 +67,11 @@ AC_DEFUN([AC_OPTION_Linux_ix86_at], [ [unset enableval] AC_ARG_ENABLE([$1], AS_HELP_STRING([--enable-$1], [$2])) -[#TODO. Could use some M4 magic to avoid a lot of shell code. +[if test x$enable_$4_group = xno; +then + enableval=${enableval-no} +fi +#TODO. Could use some M4 magic to avoid a lot of shell code. case $host_platform:$host_cpu in at:i?86) case $enable_device_drivers:'$2' in @@ -140,9 +149,16 @@ AC_Linux_DRIVER_qemu([ide], [IDE disk controllers], [CONFIG_BLK_DEV_IDE]) +AC_ARG_ENABLE([ide-forcedma], + AS_HELP_STRING([--enable-ide-forcedma], [enable forced use of DMA on IDE]), + [test x"$enableval" = xno || + AC_DEFINE([CONFIG_BLK_DEV_FORCE_DMA], [1], [Force DMA on IDE block devices])]) + dnl SCSI controllers. +AC_OPTION_Linux_group([scsi], [SCSI drivers]) -AC_Linux_DRIVER([53c78xx], +# Disabled by default. +AC_Linux_DRIVER_nodef([53c78xx], [SCSI controller NCR 53C7,8xx], [CONFIG_SCSI_NCR53C7xx], [scsi]) @@ -181,7 +197,8 @@ AC_Linux_DRIVER([aha1740], [SCSI controller Adaptec AHA-1740], [CONFIG_SCSI_AHA1740], [scsi]) -AC_Linux_DRIVER([aic7xxx], +# Disabled by default. +AC_Linux_DRIVER_nodef([aic7xxx], [SCSI controller Adaptec AIC7xxx], [CONFIG_SCSI_AIC7XXX], [scsi]) @@ -212,7 +229,8 @@ AC_Linux_DRIVER_nodef([g_NCR5380], [SCSI controller Generic NCR5380/53c400 (ncr5380, ncr53c400)], [CONFIG_SCSI_GENERIC_NCR5380], [scsi]) -AC_Linux_DRIVER([gdth], +# Disabled by default. +AC_Linux_DRIVER_nodef([gdth], [GDT SCSI Disk Array Controller], [CONFIG_SCSI_GDTH], [scsi]) @@ -244,6 +262,10 @@ AC_Linux_DRIVER([seagate], [SCSI controller Seagate ST02, Future Domain TMC-8xx], [CONFIG_SCSI_SEAGATE], [scsi]) +AC_Linux_DRIVER([sym53c8xx], + [SCSI controller Symbios 53C8XX], + [CONFIG_SCSI_SYM53C8XX], + [scsi]) AC_Linux_DRIVER([t128], [SCSI controller Trantor T128/T128F/T228 (t128, t128f, t228)], [CONFIG_SCSI_T128], @@ -266,6 +288,7 @@ AC_Linux_DRIVER([wd7000], [scsi]) dnl Ethernet controllers. +AC_OPTION_Linux_group([net], [Network drivers]) AC_Linux_DRIVER([3c501], [Ethernet controller 3COM 501 (3c501) / Etherlink I], @@ -490,6 +513,7 @@ AC_Linux_DRIVER([znet], [net]) dnl PCMCIA device support. +AC_OPTION_Linux_group([pcmcia], [PCMCIA drivers]) AC_Linux_DRIVER([i82365], [Intel 82365 PC Card controller], @@ -537,6 +561,7 @@ AC_Linux_DRIVER([xirc2ps_cs], [pcmcia]) dnl Wireless device drivers. +AC_OPTION_Linux_group([wireless], [Wireless drivers]) AC_Linux_DRIVER([orinoco_cs], [Hermes or Prism 2 PCMCIA Wireless adapters (Orinoco)], diff --git a/linux/dev/arch/i386/kernel/irq.c b/linux/dev/arch/i386/kernel/irq.c index 7faaa62c..68bf0c4b 100644 --- a/linux/dev/arch/i386/kernel/irq.c +++ b/linux/dev/arch/i386/kernel/irq.c @@ -47,10 +47,8 @@ #include <asm/io.h> #include <asm/hardirq.h> -extern void linux_timer_intr (void); -extern spl_t splhigh (void); -extern spl_t spl0 (void); -extern void form_pic_mask (void); +#include <linux/dev/glue/glue.h> +#include <machine/machspl.h> #if 0 /* XXX: This is the way it's done in linux 2.2. GNU Mach currently uses intr_count. It should be made using local_{bh/irq}_count instead (through hardirq_enter/exit) for SMP support. */ @@ -74,7 +72,7 @@ spl_t linux_intr_pri; /* * Flag indicating an interrupt is being handled. */ -unsigned long intr_count = 0; +unsigned int intr_count = 0; /* * List of Linux interrupt handlers. @@ -95,12 +93,6 @@ static struct linux_action *irq_action[16] = NULL, NULL, NULL, NULL }; -extern spl_t curr_ipl; -extern int curr_pic_mask; -extern int pic_mask[]; - -extern void intnull (), prtnull (); - /* * Generic interrupt handler for Linux devices. * Set up a fake `struct pt_regs' then call the real handler. @@ -223,15 +215,15 @@ setup_x86_irq (int irq, struct linux_action *new) { /* Can't share interrupts unless both agree to */ if (!(old->flags & new->flags & SA_SHIRQ)) - return (-LINUX_EBUSY); + return (-EBUSY); /* Can't share interrupts unless both are same type */ if ((old->flags ^ new->flags) & SA_INTERRUPT) - return (-LINUX_EBUSY); + return (-EBUSY); /* Can't share at different levels */ if (intpri[irq] && linux_intr_pri != intpri[irq]) - return (-LINUX_EBUSY); + return (-EBUSY); /* add new interrupt at end of irq queue */ do @@ -271,7 +263,7 @@ request_irq (unsigned int irq, void (*handler) (int, void *, struct pt_regs *), assert (irq < 16); if (!handler) - return -LINUX_EINVAL; + return -EINVAL; /* * Hmm... Should I use `kalloc()' ? @@ -280,7 +272,7 @@ request_irq (unsigned int irq, void (*handler) (int, void *, struct pt_regs *), action = (struct linux_action *) linux_kmalloc (sizeof (struct linux_action), GFP_KERNEL); if (action == NULL) - return -LINUX_ENOMEM; + return -ENOMEM; action->handler = handler; action->next = NULL; @@ -446,7 +438,6 @@ static void show(char * str) int i; unsigned long *stack; int cpu = smp_processor_id(); - extern char *get_options(char *str, int *ints); printk("\n%s, CPU %d:\n", str, cpu); printk("irq: %d [%d %d]\n", diff --git a/linux/dev/drivers/block/floppy.c b/linux/dev/drivers/block/floppy.c index d3fcd6af..4c0977a3 100644 --- a/linux/dev/drivers/block/floppy.c +++ b/linux/dev/drivers/block/floppy.c @@ -177,6 +177,8 @@ static inline int __get_order(unsigned long size); #include <linux/blk.h> #include <linux/cdrom.h> /* for the compatibility eject ioctl */ +#include <linux/dev/glue/glue.h> + #ifndef FLOPPY_MOTOR_MASK #define FLOPPY_MOTOR_MASK 0xf0 @@ -437,7 +439,6 @@ static int probing = 0; static volatile int command_status = FD_COMMAND_NONE, fdc_busy = 0; static struct wait_queue *fdc_wait = NULL, *command_done = NULL; #ifdef MACH -extern int issig (void); #define NO_SIGNAL (! issig () || ! interruptible) #else #define NO_SIGNAL (!(current->signal & ~current->blocked) || !interruptible) @@ -1412,7 +1413,7 @@ static int interpret_errors(void) */ static void setup_rw_floppy(void) { - int i,ready_date,r, flags,dflags; + int i, ready_date, r, flags; timeout_fn function; flags = raw_cmd->flags; @@ -1435,7 +1436,6 @@ static void setup_rw_floppy(void) if (wait_for_completion(ready_date,function)) return; } - dflags = DRS->flags; if ((flags & FD_RAW_READ) || (flags & FD_RAW_WRITE)) setup_DMA(); @@ -3492,6 +3492,8 @@ static void config_types(void) printk("fd%d is unknown type %d",drive, UDP->cmos); } + else + allowed_drive_mask &= ~(1 << drive); } if (!first) printk("\n"); @@ -4172,8 +4174,6 @@ static void floppy_release_irq_and_dma(void) #ifdef MODULE -extern char *get_options(char *str, int *ints); - char *floppy=NULL; static void parse_floppy_cfg_string(char *cfg) @@ -4277,11 +4277,10 @@ void cleanup_module(void) * resource contention. */ void floppy_eject(void) { - int dummy; if(floppy_grab_irq_and_dma()==0) { lock_fdc(MAXTIMEOUT,0); - dummy=fd_eject(0); + fd_eject(0); process_fd_request(); floppy_release_irq_and_dma(); } diff --git a/linux/dev/drivers/block/genhd.c b/linux/dev/drivers/block/genhd.c index 3946b39c..95b499b1 100644 --- a/linux/dev/drivers/block/genhd.c +++ b/linux/dev/drivers/block/genhd.c @@ -39,6 +39,7 @@ #ifdef MACH #include <machine/spl.h> +#include <linux/dev/glue/glue.h> #endif #define SYS_IND(p) get_unaligned(&p->sys_ind) @@ -289,7 +290,7 @@ read_mbr: printk(" unable to read partition table\n"); return -1; } - data = bh->b_data; + data = (unsigned char *)bh->b_data; /* In some cases we modify the geometry */ /* of the drive (below), so ensure that */ /* nobody else tries to re-use this data. */ @@ -770,9 +771,7 @@ void device_setup(void) struct gendisk *p; int nr=0; #ifdef MACH - extern int linux_intr_pri; - - linux_intr_pri = SPL5; + linux_intr_pri = SPL6; #endif #ifndef MACH diff --git a/linux/dev/glue/block.c b/linux/dev/glue/block.c index b920fb62..8c41f088 100644 --- a/linux/dev/glue/block.c +++ b/linux/dev/glue/block.c @@ -49,6 +49,8 @@ #include <mach/vm_param.h> #include <mach/notify.h> +#include <kern/kalloc.h> + #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> @@ -61,6 +63,7 @@ #include <device/disk_status.h> #include <device/device_reply.user.h> #include <device/device_emul.h> +#include <device/ds_routines.h> /* TODO. This should be fixed to not be i386 specific. */ #include <i386at/disk.h> @@ -78,9 +81,7 @@ #include <linux/hdreg.h> #include <asm/io.h> -extern int linux_auto_config; -extern int linux_intr_pri; -extern int linux_to_mach_error (int); +#include <linux/dev/glue/glue.h> /* This task queue is not used in Mach: just for fixing undefined symbols. */ DECLARE_TASK_QUEUE (tq_disk); @@ -193,9 +194,6 @@ int read_ahead[MAX_BLKDEV] = {0, }; This is unused in Mach. It is here to make drivers compile. */ struct wait_queue *wait_for_request = NULL; -/* Map for allocating device memory. */ -extern vm_map_t device_io_map; - /* Initialize block drivers. */ int blk_dev_init () @@ -230,19 +228,17 @@ int register_blkdev (unsigned major, const char *name, struct file_operations *fops) { - int err = 0; - if (major == 0) { for (major = MAX_BLKDEV - 1; major > 0; major--) if (blkdevs[major].fops == NULL) goto out; - return -LINUX_EBUSY; + return -EBUSY; } if (major >= MAX_BLKDEV) - return -LINUX_EINVAL; + return -EINVAL; if (blkdevs[major].fops && blkdevs[major].fops != fops) - return -LINUX_EBUSY; + return -EBUSY; out: blkdevs[major].name = name; @@ -260,12 +256,10 @@ out: int unregister_blkdev (unsigned major, const char *name) { - int err; - if (major >= MAX_BLKDEV) - return -LINUX_EINVAL; + return -EINVAL; if (! blkdevs[major].fops || strcmp (blkdevs[major].name, name)) - return -LINUX_EINVAL; + return -EINVAL; blkdevs[major].fops = NULL; if (blkdevs[major].labels) { @@ -280,8 +274,6 @@ unregister_blkdev (unsigned major, const char *name) void set_blocksize (kdev_t dev, int size) { - extern int *blksize_size[]; - if (! blksize_size[MAJOR (dev)]) return; @@ -315,7 +307,7 @@ alloc_buffer (int size) d = current_thread ()->pcb->data; assert (d); queue_enter (&d->pages, m, vm_page_t, pageq); - return (void *) m->phys_addr; + return (void *) phystokv(m->phys_addr); } return (void *) __get_free_pages (GFP_KERNEL, 0, ~0UL); } @@ -324,7 +316,6 @@ alloc_buffer (int size) static void free_buffer (void *p, int size) { - int i; struct temp_data *d; vm_page_t m; @@ -336,7 +327,7 @@ free_buffer (void *p, int size) assert (d); queue_iterate (&d->pages, m, vm_page_t, pageq) { - if (m->phys_addr == (vm_offset_t) p) + if (phystokv(m->phys_addr) == (vm_offset_t) p) { queue_remove (&d->pages, m, vm_page_t, pageq); VM_PAGE_FREE (m); @@ -388,7 +379,6 @@ __brelse (struct buffer_head *bh) struct buffer_head * bread (kdev_t dev, int block, int size) { - int err; struct buffer_head *bh; bh = getblk (dev, block, size); @@ -537,7 +527,7 @@ rdwr_partial (int rw, kdev_t dev, loff_t *off, } bh->b_data = alloc_buffer (bh->b_size); if (! bh->b_data) - return -LINUX_ENOMEM; + return -ENOMEM; ll_rw_block (READ, 1, &bh); wait_on_buffer (bh); if (buffer_uptodate (bh)) @@ -556,7 +546,7 @@ rdwr_partial (int rw, kdev_t dev, loff_t *off, wait_on_buffer (bh); if (! buffer_uptodate (bh)) { - err = -LINUX_EIO; + err = -EIO; goto out; } } @@ -565,7 +555,7 @@ rdwr_partial (int rw, kdev_t dev, loff_t *off, *off += c; } else - err = -LINUX_EIO; + err = -EIO; out: free_buffer (bh->b_data, bh->b_size); return err; @@ -610,15 +600,15 @@ rdwr_full (int rw, kdev_t dev, loff_t *off, char **buf, int *resid, int bshift) if (cc > ((nbuf - nb) << bshift)) cc = (nbuf - nb) << bshift; if (! test_bit (BH_Bounce, &bh->b_state)) - bh->b_data = (char *) pmap_extract (vm_map_pmap (device_io_map), + bh->b_data = (char *) phystokv(pmap_extract (vm_map_pmap (device_io_map), (((vm_offset_t) *buf) - + (nb << bshift))); + + (nb << bshift)))); else { bh->b_data = alloc_buffer (cc); if (! bh->b_data) { - err = -LINUX_ENOMEM; + err = -ENOMEM; break; } if (rw == WRITE) @@ -642,7 +632,7 @@ rdwr_full (int rw, kdev_t dev, loff_t *off, char **buf, int *resid, int bshift) && rw == READ && test_bit (BH_Bounce, &bh->b_state)) memcpy (*buf + cc, bh->b_data, bh->b_size); else if (! err && ! buffer_uptodate (bh)) - err = -LINUX_EIO; + err = -EIO; if (test_bit (BH_Bounce, &bh->b_state)) free_buffer (bh->b_data, bh->b_size); } @@ -789,6 +779,7 @@ static struct block_data *open_list; extern struct device_emulation_ops linux_block_emulation_ops; static io_return_t device_close (void *); +static io_return_t device_close_forced (void *, int); /* Return a send right for block device BD. */ static ipc_port_t @@ -922,7 +913,7 @@ static kern_return_t init_partition (struct name_map *np, kdev_t *dev, struct device_struct *ds, int slice, int *part) { - int err, i, j; + int i, j; struct disklabel *lp; struct gendisk *gd = ds->gd; struct partition *p; @@ -947,7 +938,7 @@ init_partition (struct name_map *np, kdev_t *dev, if (gd->part[MINOR (d->inode.i_rdev)].nr_sects <= 0 || gd->part[MINOR (d->inode.i_rdev)].start_sect < 0) continue; - linux_intr_pri = SPL5; + linux_intr_pri = SPL6; d->file.f_flags = 0; d->file.f_mode = O_RDONLY; if (ds->fops->open && (*ds->fops->open) (&d->inode, &d->file)) @@ -1093,7 +1084,7 @@ device_open (ipc_port_t reply_port, mach_msg_type_name_t reply_port_type, if (ds->fops->open) { td.inode.i_rdev = dev; - linux_intr_pri = SPL5; + linux_intr_pri = SPL6; err = (*ds->fops->open) (&td.inode, &td.file); if (err) { @@ -1164,6 +1155,7 @@ out: { ipc_kobject_set (bd->port, IKO_NULL, IKOT_NONE); ipc_port_dealloc_kernel (bd->port); + *devp = IP_NULL; } kfree ((vm_offset_t) bd, sizeof (struct block_data)); bd = NULL; @@ -1174,18 +1166,16 @@ out: bd->open_count = 1; bd->next = open_list; open_list = bd; + *devp = &bd -> device; } - if (IP_VALID (reply_port)) - ds_device_open_reply (reply_port, reply_port_type, err, dev_to_port (bd)); - else if (! err) + if (!IP_VALID (reply_port) && ! err) device_close (bd); - - return MIG_NO_REPLY; + return err; } static io_return_t -device_close (void *d) +device_close_forced (void *d, int force) { struct block_data *bd = d, *bdp, **prev; struct device_struct *ds = bd->ds; @@ -1202,7 +1192,7 @@ device_close (void *d) } ds->busy = 1; - if (--bd->open_count == 0) + if (force || --bd->open_count == 0) { /* Wait for pending I/O to complete. */ while (bd->iocount > 0) @@ -1245,6 +1235,13 @@ device_close (void *d) return D_SUCCESS; } +static io_return_t +device_close (void *d) +{ + return device_close_forced (d, 0); +} + + #define MAX_COPY (VM_MAP_COPY_PAGE_LIST_MAX << PAGE_SHIFT) /* Check block BN and size COUNT for I/O validity @@ -1674,7 +1671,7 @@ device_get_status (void *d, dev_flavor_t flavor, dev_status_t status, INIT_DATA(); if ((*bd->ds->fops->ioctl) (&td.inode, &td.file, - HDIO_GETGEO, &hg)) + HDIO_GETGEO, (unsigned long)&hg)) return D_INVALID_OPERATION; dp->dp_type = DPT_WINI; /* XXX: It may be a floppy... */ @@ -1723,6 +1720,17 @@ device_set_status (void *d, dev_flavor_t flavor, dev_status_t status, return D_INVALID_OPERATION; } + +static void +device_no_senders (mach_no_senders_notification_t *ns) +{ + device_t dev; + + dev = dev_port_lookup((ipc_port_t) ns->not_header.msgh_remote_port); + assert(dev); + device_close_forced (dev->emul_data, 1); +} + struct device_emulation_ops linux_block_emulation_ops = { NULL, @@ -1738,7 +1746,7 @@ struct device_emulation_ops linux_block_emulation_ops = device_get_status, NULL, NULL, - NULL, + device_no_senders, NULL, NULL }; diff --git a/linux/dev/glue/glue.h b/linux/dev/glue/glue.h new file mode 100644 index 00000000..5d4f6d88 --- /dev/null +++ b/linux/dev/glue/glue.h @@ -0,0 +1,44 @@ +/* + * Copyright (C) 2011 Free Software Foundation + * + * This program is free software ; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation ; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY ; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with the program ; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#ifndef LINUX_DEV_GLUE_GLUE_H +#define LINUX_DEV_GLUE_GLUE_H + +#include <vm/vm_types.h> +#include <mach/machine/vm_types.h> + +extern int linux_auto_config; +extern int linux_intr_pri; + +extern void *alloc_contig_mem (unsigned, unsigned, unsigned, vm_page_t *); +extern void free_contig_mem (vm_page_t); +extern void init_IRQ (void); +extern void restore_IRQ (void); +extern void linux_kmem_init (void); +extern void linux_net_emulation_init (void); +extern void device_setup (void); +extern void linux_timer_intr (void); +extern void linux_bad_intr (int); +extern void linux_sched_init (void); +extern void pcmcia_init (void); +extern void linux_soft_intr (void); +extern int issig (void); +extern int linux_to_mach_error (int); +extern char *get_options(char *str, int *ints); + +#endif /* LINUX_DEV_GLUE_GLUE_H */ diff --git a/linux/dev/glue/kmem.c b/linux/dev/glue/kmem.c index 8c21ce7d..28321711 100644 --- a/linux/dev/glue/kmem.c +++ b/linux/dev/glue/kmem.c @@ -29,6 +29,7 @@ #include <kern/assert.h> #include <kern/kalloc.h> +#include <kern/printf.h> #include <vm/vm_page.h> #include <vm/vm_kern.h> @@ -40,14 +41,13 @@ #include <asm/system.h> -extern void *alloc_contig_mem (unsigned, unsigned, unsigned, vm_page_t *); -extern int printf (const char *, ...); +#include <linux/dev/glue/glue.h> /* Amount of memory to reserve for Linux memory allocator. We reserve 64K chunks to stay within DMA limits. Increase MEM_CHUNKS if the kernel is running out of memory. */ #define MEM_CHUNK_SIZE (64 * 1024) -#define MEM_CHUNKS 7 +#define MEM_CHUNKS 32 #define MEM_DMA_LIMIT (16 * 1024 * 1024) /* Mininum amount that linux_kmalloc will allocate. */ @@ -218,7 +218,7 @@ void * linux_kmalloc (unsigned int size, int priority) { int order, coalesced = 0; - unsigned flags; + unsigned long flags; struct pagehdr *ph; struct blkhdr *bh, *new_bh; @@ -310,7 +310,7 @@ again: void linux_kfree (void *p) { - unsigned flags; + unsigned long flags; struct blkhdr *bh; struct pagehdr *ph; @@ -385,7 +385,8 @@ unsigned long __get_free_pages (int priority, unsigned long order, int dma) { int i, pages_collected = 0; - unsigned flags, bits, off, j, len; + unsigned bits, off, j, len; + unsigned long flags; assert ((PAGE_SIZE << order) <= MEM_CHUNK_SIZE); @@ -444,7 +445,8 @@ void free_pages (unsigned long addr, unsigned long order) { int i; - unsigned flags, bits, len, j; + unsigned bits, len, j; + unsigned long flags; assert ((addr & PAGE_MASK) == 0); @@ -554,7 +556,7 @@ vfree (void *addr) if (!p) panic ("vmalloc_list_lookup failure"); - kmem_free (kernel_map, addr, p->size); + kmem_free (kernel_map, (vm_offset_t) addr, p->size); vmalloc_list_remove (p); } diff --git a/linux/dev/glue/misc.c b/linux/dev/glue/misc.c index d8ca3c24..77dc31dd 100644 --- a/linux/dev/glue/misc.c +++ b/linux/dev/glue/misc.c @@ -53,6 +53,7 @@ #include <sys/types.h> #include <mach/vm_param.h> #include <kern/thread.h> +#include <kern/printf.h> #include <vm/vm_map.h> #include <vm/vm_page.h> #include <device/device_types.h> @@ -66,10 +67,7 @@ #include <linux/blk.h> #include <linux/proc_fs.h> #include <linux/kernel_stat.h> - -extern boolean_t vm_map_lookup_entry (register vm_map_t, register vm_offset_t, - vm_map_entry_t *); -extern int printf (const char *, ...); +#include <linux/dev/glue/glue.h> int (*dispatch_scsi_info_ptr) (int ino, char *buffer, char **start, off_t offset, int length, int inout) = 0; @@ -84,34 +82,34 @@ linux_to_mach_error (int err) case 0: return D_SUCCESS; - case -LINUX_EPERM: + case -EPERM: return D_INVALID_OPERATION; - case -LINUX_EIO: + case -EIO: return D_IO_ERROR; - case -LINUX_ENXIO: + case -ENXIO: return D_NO_SUCH_DEVICE; - case -LINUX_EACCES: + case -EACCES: return D_INVALID_OPERATION; - case -LINUX_EFAULT: + case -EFAULT: return D_INVALID_SIZE; - case -LINUX_EBUSY: + case -EBUSY: return D_ALREADY_OPEN; - case -LINUX_EINVAL: + case -EINVAL: return D_INVALID_SIZE; - case -LINUX_EROFS: + case -EROFS: return D_READ_ONLY; - case -LINUX_EWOULDBLOCK: + case -EWOULDBLOCK: return D_WOULD_BLOCK; - case -LINUX_ENOMEM: + case -ENOMEM: return D_NO_MEMORY; default: @@ -123,6 +121,8 @@ linux_to_mach_error (int err) int issig () { + if (!current_thread()) + return 0; return current_thread ()->wait_result != THREAD_AWAKENED; } @@ -148,7 +148,7 @@ verify_area (int rw, const void *p, unsigned long size) || (entry->protection & prot) != prot) { vm_map_unlock_read (current_map ()); - return -LINUX_EFAULT; + return -EFAULT; } if (entry->vme_end - entry->vme_start >= len) break; @@ -229,7 +229,12 @@ add_blkdev_randomness (int major) void do_gettimeofday (struct timeval *tv) { - host_get_time (1, tv); + /* + * XXX: The first argument should be mach_host_self (), but that's too + * expensive, and the host argument is not used by host_get_time (), + * only checked not to be HOST_NULL. + */ + host_get_time ((host_t) 1, (time_value_t *) tv); } int diff --git a/linux/dev/glue/net.c b/linux/dev/glue/net.c index 095428d3..15732737 100644 --- a/linux/dev/glue/net.c +++ b/linux/dev/glue/net.c @@ -61,6 +61,7 @@ #include <sys/types.h> #include <machine/spl.h> +#include <machine/vm_param.h> #include <mach/mach_types.h> #include <mach/kern_return.h> @@ -69,6 +70,9 @@ #include <mach/vm_param.h> #include <mach/notify.h> +#include <kern/kalloc.h> +#include <kern/printf.h> + #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> @@ -84,6 +88,7 @@ #include <device/net_io.h> #include <device/device_reply.user.h> #include <device/device_emul.h> +#include <device/ds_routines.h> #define MACH_INCLUDE #include <linux/kernel.h> @@ -97,7 +102,7 @@ #include <linux/etherdevice.h> #include <linux/wireless.h> -extern int linux_intr_pri; +#include <linux/dev/glue/glue.h> /* One of these is associated with each instance of a device. */ struct net_data @@ -248,7 +253,6 @@ void dev_kfree_skb (struct sk_buff *skb, int mode) { unsigned flags; - extern void *io_done_list; /* Queue sk_buff on done list if there is a page list attached or we need to send a reply. @@ -425,7 +429,6 @@ device_write (void *d, ipc_port_t reply_port, { unsigned char *p; int i, amt, skblen, s; - io_return_t err = 0; vm_map_copy_t copy = (vm_map_copy_t) data; struct net_data *nd = d; struct linux_device *dev = nd->dev; @@ -447,7 +450,7 @@ device_write (void *d, ipc_port_t reply_port, assert (copy->cpy_npages == 1); skb->copy = copy; - skb->data = ((void *) copy->cpy_page_list[0]->phys_addr + skb->data = ((void *) phystokv(copy->cpy_page_list[0]->phys_addr) + (copy->offset & PAGE_MASK)); skb->len = count; skb->head = skb->data; @@ -461,7 +464,7 @@ device_write (void *d, ipc_port_t reply_port, skb->end = skb->tail; memcpy (skb->data, - ((void *) copy->cpy_page_list[0]->phys_addr + ((void *) phystokv(copy->cpy_page_list[0]->phys_addr) + (copy->offset & PAGE_MASK)), amt); count -= amt; @@ -471,7 +474,7 @@ device_write (void *d, ipc_port_t reply_port, amt = PAGE_SIZE; if (amt > count) amt = count; - memcpy (p, (void *) copy->cpy_page_list[i]->phys_addr, amt); + memcpy (p, (void *) phystokv(copy->cpy_page_list[i]->phys_addr), amt); count -= amt; p += amt; } diff --git a/linux/dev/include/asm-i386/errno.h b/linux/dev/include/asm-i386/errno.h deleted file mode 100644 index 1683367a..00000000 --- a/linux/dev/include/asm-i386/errno.h +++ /dev/null @@ -1,266 +0,0 @@ -#ifndef _I386_ERRNO_H -#define _I386_ERRNO_H - -#ifdef MACH_INCLUDE - -#define LINUX_EPERM 1 /* Operation not permitted */ -#define LINUX_ENOENT 2 /* No such file or directory */ -#define LINUX_ESRCH 3 /* No such process */ -#define LINUX_EINTR 4 /* Interrupted system call */ -#define LINUX_EIO 5 /* I/O error */ -#define LINUX_ENXIO 6 /* No such device or address */ -#define LINUX_E2BIG 7 /* Arg list too long */ -#define LINUX_ENOEXEC 8 /* Exec format error */ -#define LINUX_EBADF 9 /* Bad file number */ -#define LINUX_ECHILD 10 /* No child processes */ -#define LINUX_EAGAIN 11 /* Try again */ -#define LINUX_ENOMEM 12 /* Out of memory */ -#define LINUX_EACCES 13 /* Permission denied */ -#define LINUX_EFAULT 14 /* Bad address */ -#define LINUX_ENOTBLK 15 /* Block device required */ -#define LINUX_EBUSY 16 /* Device or resource busy */ -#define LINUX_EEXIST 17 /* File exists */ -#define LINUX_EXDEV 18 /* Cross-device link */ -#define LINUX_ENODEV 19 /* No such device */ -#define LINUX_ENOTDIR 20 /* Not a directory */ -#define LINUX_EISDIR 21 /* Is a directory */ -#define LINUX_EINVAL 22 /* Invalid argument */ -#define LINUX_ENFILE 23 /* File table overflow */ -#define LINUX_EMFILE 24 /* Too many open files */ -#define LINUX_ENOTTY 25 /* Not a typewriter */ -#define LINUX_ETXTBSY 26 /* Text file busy */ -#define LINUX_EFBIG 27 /* File too large */ -#define LINUX_ENOSPC 28 /* No space left on device */ -#define LINUX_ESPIPE 29 /* Illegal seek */ -#define LINUX_EROFS 30 /* Read-only file system */ -#define LINUX_EMLINK 31 /* Too many links */ -#define LINUX_EPIPE 32 /* Broken pipe */ -#define LINUX_EDOM 33 /* Math argument out of domain of func */ -#define LINUX_ERANGE 34 /* Math result not representable */ -#define LINUX_EDEADLK 35 /* Resource deadlock would occur */ -#define LINUX_ENAMETOOLONG 36 /* File name too long */ -#define LINUX_ENOLCK 37 /* No record locks available */ -#define LINUX_ENOSYS 38 /* Function not implemented */ -#define LINUX_ENOTEMPTY 39 /* Directory not empty */ -#define LINUX_ELOOP 40 /* Too many symbolic links encountered */ -#define LINUX_EWOULDBLOCK LINUX_EAGAIN /* Operation would block */ -#define LINUX_ENOMSG 42 /* No message of desired type */ -#define LINUX_EIDRM 43 /* Identifier removed */ -#define LINUX_ECHRNG 44 /* Channel number out of range */ -#define LINUX_EL2NSYNC 45 /* Level 2 not synchronized */ -#define LINUX_EL3HLT 46 /* Level 3 halted */ -#define LINUX_EL3RST 47 /* Level 3 reset */ -#define LINUX_ELNRNG 48 /* Link number out of range */ -#define LINUX_EUNATCH 49 /* Protocol driver not attached */ -#define LINUX_ENOCSI 50 /* No CSI structure available */ -#define LINUX_EL2HLT 51 /* Level 2 halted */ -#define LINUX_EBADE 52 /* Invalid exchange */ -#define LINUX_EBADR 53 /* Invalid request descriptor */ -#define LINUX_EXFULL 54 /* Exchange full */ -#define LINUX_ENOANO 55 /* No anode */ -#define LINUX_EBADRQC 56 /* Invalid request code */ -#define LINUX_EBADSLT 57 /* Invalid slot */ - -#define LINUX_EDEADLOCK LINUX_EDEADLK - -#define LINUX_EBFONT 59 /* Bad font file format */ -#define LINUX_ENOSTR 60 /* Device not a stream */ -#define LINUX_ENODATA 61 /* No data available */ -#define LINUX_ETIME 62 /* Timer expired */ -#define LINUX_ENOSR 63 /* Out of streams resources */ -#define LINUX_ENONET 64 /* Machine is not on the network */ -#define LINUX_ENOPKG 65 /* Package not installed */ -#define LINUX_EREMOTE 66 /* Object is remote */ -#define LINUX_ENOLINK 67 /* Link has been severed */ -#define LINUX_EADV 68 /* Advertise error */ -#define LINUX_ESRMNT 69 /* Srmount error */ -#define LINUX_ECOMM 70 /* Communication error on send */ -#define LINUX_EPROTO 71 /* Protocol error */ -#define LINUX_EMULTIHOP 72 /* Multihop attempted */ -#define LINUX_EDOTDOT 73 /* RFS specific error */ -#define LINUX_EBADMSG 74 /* Not a data message */ -#define LINUX_EOVERFLOW 75 /* Value too large for defined data type */ -#define LINUX_ENOTUNIQ 76 /* Name not unique on network */ -#define LINUX_EBADFD 77 /* File descriptor in bad state */ -#define LINUX_EREMCHG 78 /* Remote address changed */ -#define LINUX_ELIBACC 79 /* Can not access a needed shared library */ -#define LINUX_ELIBBAD 80 /* Accessing a corrupted shared library */ -#define LINUX_ELIBSCN 81 /* .lib section in a.out corrupted */ -#define LINUX_ELIBMAX 82 /* Attempting to link in too many shared libraries */ -#define LINUX_ELIBEXEC 83 /* Cannot exec a shared library directly */ -#define LINUX_EILSEQ 84 /* Illegal byte sequence */ -#define LINUX_ERESTART 85 /* Interrupted system call should be restarted */ -#define LINUX_ESTRPIPE 86 /* Streams pipe error */ -#define LINUX_EUSERS 87 /* Too many users */ -#define LINUX_ENOTSOCK 88 /* Socket operation on non-socket */ -#define LINUX_EDESTADDRREQ 89 /* Destination address required */ -#define LINUX_EMSGSIZE 90 /* Message too long */ -#define LINUX_EPROTOTYPE 91 /* Protocol wrong type for socket */ -#define LINUX_ENOPROTOOPT 92 /* Protocol not available */ -#define LINUX_EPROTONOSUPPORT 93 /* Protocol not supported */ -#define LINUX_ESOCKTNOSUPPORT 94 /* Socket type not supported */ -#define LINUX_EOPNOTSUPP 95 /* Operation not supported on transport endpoint */ -#define LINUX_EPFNOSUPPORT 96 /* Protocol family not supported */ -#define LINUX_EAFNOSUPPORT 97 /* Address family not supported by protocol */ -#define LINUX_EADDRINUSE 98 /* Address already in use */ -#define LINUX_EADDRNOTAVAIL 99 /* Cannot assign requested address */ -#define LINUX_ENETDOWN 100 /* Network is down */ -#define LINUX_ENETUNREACH 101 /* Network is unreachable */ -#define LINUX_ENETRESET 102 /* Network dropped connection because of reset */ -#define LINUX_ECONNABORTED 103 /* Software caused connection abort */ -#define LINUX_ECONNRESET 104 /* Connection reset by peer */ -#define LINUX_ENOBUFS 105 /* No buffer space available */ -#define LINUX_EISCONN 106 /* Transport endpoint is already connected */ -#define LINUX_ENOTCONN 107 /* Transport endpoint is not connected */ -#define LINUX_ESHUTDOWN 108 /* Cannot send after transport endpoint shutdown */ -#define LINUX_ETOOMANYREFS 109 /* Too many references: cannot splice */ -#define LINUX_ETIMEDOUT 110 /* Connection timed out */ -#define LINUX_ECONNREFUSED 111 /* Connection refused */ -#define LINUX_EHOSTDOWN 112 /* Host is down */ -#define LINUX_EHOSTUNREACH 113 /* No route to host */ -#define LINUX_EALREADY 114 /* Operation already in progress */ -#define LINUX_EINPROGRESS 115 /* Operation now in progress */ -#define LINUX_ESTALE 116 /* Stale NFS file handle */ -#define LINUX_EUCLEAN 117 /* Structure needs cleaning */ -#define LINUX_ENOTNAM 118 /* Not a XENIX named type file */ -#define LINUX_ENAVAIL 119 /* No XENIX semaphores available */ -#define LINUX_EISNAM 120 /* Is a named type file */ -#define LINUX_EREMOTEIO 121 /* Remote I/O error */ -#define LINUX_EDQUOT 122 /* Quota exceeded */ - -#define LINUX_ENOMEDIUM 123 /* No medium found */ -#define LINUX_EMEDIUMTYPE 124 /* Wrong medium type */ - -#else /* !MACH_INCLUDE */ - -#define EPERM 1 /* Operation not permitted */ -#define ENOENT 2 /* No such file or directory */ -#define ESRCH 3 /* No such process */ -#define EINTR 4 /* Interrupted system call */ -#define EIO 5 /* I/O error */ -#define ENXIO 6 /* No such device or address */ -#define E2BIG 7 /* Arg list too long */ -#define ENOEXEC 8 /* Exec format error */ -#define EBADF 9 /* Bad file number */ -#define ECHILD 10 /* No child processes */ -#define EAGAIN 11 /* Try again */ -#define ENOMEM 12 /* Out of memory */ -#define EACCES 13 /* Permission denied */ -#define EFAULT 14 /* Bad address */ -#define ENOTBLK 15 /* Block device required */ -#define EBUSY 16 /* Device or resource busy */ -#define EEXIST 17 /* File exists */ -#define EXDEV 18 /* Cross-device link */ -#define ENODEV 19 /* No such device */ -#define ENOTDIR 20 /* Not a directory */ -#define EISDIR 21 /* Is a directory */ -#define EINVAL 22 /* Invalid argument */ -#define ENFILE 23 /* File table overflow */ -#define EMFILE 24 /* Too many open files */ -#define ENOTTY 25 /* Not a typewriter */ -#define ETXTBSY 26 /* Text file busy */ -#define EFBIG 27 /* File too large */ -#define ENOSPC 28 /* No space left on device */ -#define ESPIPE 29 /* Illegal seek */ -#define EROFS 30 /* Read-only file system */ -#define EMLINK 31 /* Too many links */ -#define EPIPE 32 /* Broken pipe */ -#define EDOM 33 /* Math argument out of domain of func */ -#define ERANGE 34 /* Math result not representable */ -#define EDEADLK 35 /* Resource deadlock would occur */ -#define ENAMETOOLONG 36 /* File name too long */ -#define ENOLCK 37 /* No record locks available */ -#define ENOSYS 38 /* Function not implemented */ -#define ENOTEMPTY 39 /* Directory not empty */ -#define ELOOP 40 /* Too many symbolic links encountered */ -#define EWOULDBLOCK EAGAIN /* Operation would block */ -#define ENOMSG 42 /* No message of desired type */ -#define EIDRM 43 /* Identifier removed */ -#define ECHRNG 44 /* Channel number out of range */ -#define EL2NSYNC 45 /* Level 2 not synchronized */ -#define EL3HLT 46 /* Level 3 halted */ -#define EL3RST 47 /* Level 3 reset */ -#define ELNRNG 48 /* Link number out of range */ -#define EUNATCH 49 /* Protocol driver not attached */ -#define ENOCSI 50 /* No CSI structure available */ -#define EL2HLT 51 /* Level 2 halted */ -#define EBADE 52 /* Invalid exchange */ -#define EBADR 53 /* Invalid request descriptor */ -#define EXFULL 54 /* Exchange full */ -#define ENOANO 55 /* No anode */ -#define EBADRQC 56 /* Invalid request code */ -#define EBADSLT 57 /* Invalid slot */ - -#define EDEADLOCK EDEADLK - -#define EBFONT 59 /* Bad font file format */ -#define ENOSTR 60 /* Device not a stream */ -#define ENODATA 61 /* No data available */ -#define ETIME 62 /* Timer expired */ -#define ENOSR 63 /* Out of streams resources */ -#define ENONET 64 /* Machine is not on the network */ -#define ENOPKG 65 /* Package not installed */ -#define EREMOTE 66 /* Object is remote */ -#define ENOLINK 67 /* Link has been severed */ -#define EADV 68 /* Advertise error */ -#define ESRMNT 69 /* Srmount error */ -#define ECOMM 70 /* Communication error on send */ -#define EPROTO 71 /* Protocol error */ -#define EMULTIHOP 72 /* Multihop attempted */ -#define EDOTDOT 73 /* RFS specific error */ -#define EBADMSG 74 /* Not a data message */ -#define EOVERFLOW 75 /* Value too large for defined data type */ -#define ENOTUNIQ 76 /* Name not unique on network */ -#define EBADFD 77 /* File descriptor in bad state */ -#define EREMCHG 78 /* Remote address changed */ -#define ELIBACC 79 /* Can not access a needed shared library */ -#define ELIBBAD 80 /* Accessing a corrupted shared library */ -#define ELIBSCN 81 /* .lib section in a.out corrupted */ -#define ELIBMAX 82 /* Attempting to link in too many shared libraries */ -#define ELIBEXEC 83 /* Cannot exec a shared library directly */ -#define EILSEQ 84 /* Illegal byte sequence */ -#define ERESTART 85 /* Interrupted system call should be restarted */ -#define ESTRPIPE 86 /* Streams pipe error */ -#define EUSERS 87 /* Too many users */ -#define ENOTSOCK 88 /* Socket operation on non-socket */ -#define EDESTADDRREQ 89 /* Destination address required */ -#define EMSGSIZE 90 /* Message too long */ -#define EPROTOTYPE 91 /* Protocol wrong type for socket */ -#define ENOPROTOOPT 92 /* Protocol not available */ -#define EPROTONOSUPPORT 93 /* Protocol not supported */ -#define ESOCKTNOSUPPORT 94 /* Socket type not supported */ -#define EOPNOTSUPP 95 /* Operation not supported on transport endpoint */ -#define EPFNOSUPPORT 96 /* Protocol family not supported */ -#define EAFNOSUPPORT 97 /* Address family not supported by protocol */ -#define EADDRINUSE 98 /* Address already in use */ -#define EADDRNOTAVAIL 99 /* Cannot assign requested address */ -#define ENETDOWN 100 /* Network is down */ -#define ENETUNREACH 101 /* Network is unreachable */ -#define ENETRESET 102 /* Network dropped connection because of reset */ -#define ECONNABORTED 103 /* Software caused connection abort */ -#define ECONNRESET 104 /* Connection reset by peer */ -#define ENOBUFS 105 /* No buffer space available */ -#define EISCONN 106 /* Transport endpoint is already connected */ -#define ENOTCONN 107 /* Transport endpoint is not connected */ -#define ESHUTDOWN 108 /* Cannot send after transport endpoint shutdown */ -#define ETOOMANYREFS 109 /* Too many references: cannot splice */ -#define ETIMEDOUT 110 /* Connection timed out */ -#define ECONNREFUSED 111 /* Connection refused */ -#define EHOSTDOWN 112 /* Host is down */ -#define EHOSTUNREACH 113 /* No route to host */ -#define EALREADY 114 /* Operation already in progress */ -#define EINPROGRESS 115 /* Operation now in progress */ -#define ESTALE 116 /* Stale NFS file handle */ -#define EUCLEAN 117 /* Structure needs cleaning */ -#define ENOTNAM 118 /* Not a XENIX named type file */ -#define ENAVAIL 119 /* No XENIX semaphores available */ -#define EISNAM 120 /* Is a named type file */ -#define EREMOTEIO 121 /* Remote I/O error */ -#define EDQUOT 122 /* Quota exceeded */ - -#define ENOMEDIUM 123 /* No medium found */ -#define EMEDIUMTYPE 124 /* Wrong medium type */ - -#endif /* !MACH_INCLUDE */ - -#endif diff --git a/linux/dev/include/asm-i386/page.h b/linux/dev/include/asm-i386/page.h index 2bb6837e..be818481 100644 --- a/linux/dev/include/asm-i386/page.h +++ b/linux/dev/include/asm-i386/page.h @@ -1,12 +1,7 @@ #ifndef _I386_PAGE_H #define _I386_PAGE_H -#ifndef MACH_INCLUDE -/* PAGE_SHIFT determines the page size */ -#define PAGE_SHIFT 12 -#define PAGE_SIZE (1UL << PAGE_SHIFT) -#define PAGE_MASK (~(PAGE_SIZE-1)) -#endif +#include <mach/vm_param.h> #ifdef __KERNEL__ diff --git a/linux/dev/include/asm-i386/system.h b/linux/dev/include/asm-i386/system.h index 3e17d2d1..f26a33e7 100644 --- a/linux/dev/include/asm-i386/system.h +++ b/linux/dev/include/asm-i386/system.h @@ -224,9 +224,9 @@ static inline unsigned long __xchg(unsigned long x, void * ptr, int size) #define __sti() __asm__ __volatile__ ("sti": : :"memory") #define __cli() __asm__ __volatile__ ("cli": : :"memory") #define __save_flags(x) \ -__asm__ __volatile__("pushfl ; popl %0":"=g" (x): /* no input */ :"memory") +__asm__ __volatile__("pushf ; pop %0" : "=r" (x): /* no input */ :"memory") #define __restore_flags(x) \ -__asm__ __volatile__("pushl %0 ; popfl": /* no output */ :"g" (x):"memory") +__asm__ __volatile__("push %0 ; popf": /* no output */ :"g" (x):"memory") #ifdef __SMP__ diff --git a/linux/dev/include/linux/blk.h b/linux/dev/include/linux/blk.h index 852853c5..412b8641 100644 --- a/linux/dev/include/linux/blk.h +++ b/linux/dev/include/linux/blk.h @@ -385,7 +385,9 @@ static void end_request(int uptodate) { struct request *req = CURRENT; #endif /* IDE_DRIVER */ struct buffer_head * bh; +#ifndef MACH int nsect; +#endif req->errors = 0; if (!uptodate) { diff --git a/linux/dev/include/linux/kernel.h b/linux/dev/include/linux/kernel.h index aef73ac3..4db76acc 100644 --- a/linux/dev/include/linux/kernel.h +++ b/linux/dev/include/linux/kernel.h @@ -9,9 +9,7 @@ #include <stdarg.h> #include <linux/linkage.h> - -/* Optimization barrier */ -#define barrier() __asm__("": : :"memory") +#include <linux/compiler.h> #define INT_MAX ((int)(~0U>>1)) #define UINT_MAX (~0U) diff --git a/linux/dev/include/linux/notifier.h b/linux/dev/include/linux/notifier.h index eede20f0..b3c9ccf6 100644 --- a/linux/dev/include/linux/notifier.h +++ b/linux/dev/include/linux/notifier.h @@ -55,11 +55,7 @@ extern __inline__ int notifier_chain_unregister(struct notifier_block **nl, stru } nl=&((*nl)->next); } -#ifdef MACH_INCLUDE - return -LINUX_ENOENT; -#else return -ENOENT; -#endif } /* diff --git a/linux/dev/init/main.c b/linux/dev/init/main.c index 9ed35d3d..ecbd0b68 100644 --- a/linux/dev/init/main.c +++ b/linux/dev/init/main.c @@ -39,6 +39,7 @@ #include <machine/spl.h> #include <machine/pmap.h> #include <machine/vm_param.h> +#include <machine/model_dep.h> #define MACH_INCLUDE #include <linux/sched.h> @@ -47,6 +48,8 @@ #include <linux/delay.h> #include <linux/ioport.h> #include <linux/string.h> +#include <linux/pci.h> +#include <linux/dev/glue/glue.h> #include <asm/system.h> #include <asm/io.h> @@ -83,30 +86,6 @@ struct drive_info_struct */ static void calibrate_delay (void); -extern vm_offset_t phys_last_addr; - -extern void *alloc_contig_mem (unsigned, unsigned, unsigned, vm_page_t *); -extern void free_contig_mem (vm_page_t); -extern void init_IRQ (void); -extern void restore_IRQ (void); -extern void startrtclock (void); -extern void linux_version_init (void); -extern void linux_kmem_init (void); -extern unsigned long pci_init (unsigned long, unsigned long); -extern void linux_net_emulation_init (void); -extern void device_setup (void); -extern void linux_printk (char *,...); -extern int linux_timer_intr (void); -extern spl_t spl0 (void); -extern spl_t splhigh (void); -extern void form_pic_mask (void); -extern int linux_bad_intr (int); -extern int prtnull (); -extern int intnull (); -extern void linux_sched_init (void); -extern void pcmcia_init (void); - - /* * Amount of contiguous memory to allocate for initialization. */ @@ -170,7 +149,7 @@ linux_init (void) /* * Free unused memory. */ - while (pages && pages->phys_addr < round_page (memory_start)) + while (pages && phystokv(pages->phys_addr) < round_page (memory_start)) pages = (vm_page_t) pages->pageq.next; if (pages) free_contig_mem (pages); @@ -317,7 +296,7 @@ alloc_contig_mem (unsigned size, unsigned limit, kfree ((vm_offset_t) bits, bits_len); if (pages) *pages = page_list; - return (m); + return phystokv(m); } /* diff --git a/linux/dev/kernel/dma.c b/linux/dev/kernel/dma.c index 4b569780..bbda4bbf 100644 --- a/linux/dev/kernel/dma.c +++ b/linux/dev/kernel/dma.c @@ -80,10 +80,10 @@ int request_dma (unsigned int dmanr, const char *device_id) { if (dmanr >= MAX_DMA_CHANNELS) - return -LINUX_EINVAL; + return -EINVAL; if (xchg (&dma_chan_busy[dmanr].lock, 1) != 0) - return -LINUX_EBUSY; + return -EBUSY; dma_chan_busy[dmanr].device_id = device_id; diff --git a/linux/dev/kernel/printk.c b/linux/dev/kernel/printk.c index 1c45b245..7c65d303 100644 --- a/linux/dev/kernel/printk.c +++ b/linux/dev/kernel/printk.c @@ -27,6 +27,7 @@ #include <stdarg.h> #include <asm/system.h> #include <kern/assert.h> +#include <kern/printf.h> #include <device/cons.h> static char buf[2048]; @@ -40,7 +41,8 @@ int printk (char *fmt, ...) { va_list args; - int n, flags; + int n; + unsigned long flags; char *p, *msg, *buf_end; static int msg_level = -1; diff --git a/linux/dev/kernel/resource.c b/linux/dev/kernel/resource.c index 7a187552..ba107e8e 100644 --- a/linux/dev/kernel/resource.c +++ b/linux/dev/kernel/resource.c @@ -131,7 +131,7 @@ release_region (unsigned int from, unsigned int num) int check_region (unsigned int from, unsigned int num) { - return (find_gap (&iolist, from, num) == NULL) ? -LINUX_EBUSY : 0; + return (find_gap (&iolist, from, num) == NULL) ? -EBUSY : 0; } /* Called from init/main.c to reserve IO ports. */ diff --git a/linux/dev/kernel/sched.c b/linux/dev/kernel/sched.c index 792c2dae..87906a45 100644 --- a/linux/dev/kernel/sched.c +++ b/linux/dev/kernel/sched.c @@ -34,6 +34,9 @@ #include <kern/thread.h> #include <kern/sched_prim.h> +#include <kern/printf.h> + +#include <machine/machspl.h> #define MACH_INCLUDE #include <linux/sched.h> @@ -41,21 +44,13 @@ #include <linux/fs.h> #include <linux/blkdev.h> #include <linux/interrupt.h> +#include <linux/dev/glue/glue.h> #include <asm/system.h> #include <asm/atomic.h> int securelevel = 0; -extern void *alloc_contig_mem (unsigned, unsigned, unsigned, vm_page_t *); -extern void free_contig_mem (vm_page_t); -extern spl_t splhigh (void); -extern spl_t splx (spl_t); -extern void linux_soft_intr (void); -extern int issig (void); -extern int printf (const char *, ...); -extern int linux_auto_config; - static void timer_bh (void); DECLARE_TASK_QUEUE (tq_timer); @@ -165,7 +160,7 @@ __do_down (struct semaphore *sem, int task_state) if (task_state == TASK_INTERRUPTIBLE && issig ()) { - ret = -LINUX_EINTR; + ret = -EINTR; atomic_inc (&sem->count); break; } @@ -185,7 +180,7 @@ __do_down (struct semaphore *sem, int task_state) { if (task_state == TASK_INTERRUPTIBLE && issig ()) { - ret = -LINUX_EINTR; + ret = -EINTR; atomic_inc (&sem->count); break; } @@ -326,16 +321,6 @@ schedule (void) } void -cdrom_sleep (int t) -{ - int xxx; - - assert_wait ((event_t) &xxx, TRUE); - thread_set_timeout (t); - schedule (); -} - -void linux_sched_init (void) { /* diff --git a/linux/dev/kernel/softirq.c b/linux/dev/kernel/softirq.c index 96102a7a..ac95a7dd 100644 --- a/linux/dev/kernel/softirq.c +++ b/linux/dev/kernel/softirq.c @@ -13,16 +13,18 @@ #include <linux/interrupt.h> #include <asm/system.h> +#include <linux/dev/glue/glue.h> + int bh_mask_count[32]; -unsigned long bh_active = 0; -unsigned long bh_mask = 0; +unsigned int bh_active = 0; +unsigned int bh_mask = 0; void (*bh_base[32]) (void); void linux_soft_intr (void) { - unsigned long active; - unsigned long mask, left; + unsigned int active; + unsigned int mask, left; void (**bh) (void); sti (); @@ -42,5 +44,5 @@ linux_soft_intr (void) } return; bad_bh: - printk ("linux_soft_intr:bad interrupt handler entry %08lx\n", mask); + printk ("linux_soft_intr:bad interrupt handler entry %08x\n", mask); } diff --git a/linux/dev/lib/vsprintf.c b/linux/dev/lib/vsprintf.c index 0beb7471..541ec650 100644 --- a/linux/dev/lib/vsprintf.c +++ b/linux/dev/lib/vsprintf.c @@ -328,9 +328,9 @@ linux_vsprintf (char *buf, const char *fmt, va_list args) num = va_arg (args, unsigned long); else if (qualifier == 'h') if (flags & SIGN) - num = va_arg (args, short); + num = (short) va_arg (args, int); else - num = va_arg (args, unsigned short); + num = (unsigned short) va_arg (args, unsigned int); else if (flags & SIGN) num = va_arg (args, int); else diff --git a/linux/pcmcia-cs/glue/wireless_glue.h b/linux/pcmcia-cs/glue/wireless_glue.h index 05eef4ba..61006b4a 100644 --- a/linux/pcmcia-cs/glue/wireless_glue.h +++ b/linux/pcmcia-cs/glue/wireless_glue.h @@ -48,14 +48,7 @@ #endif -/* - * We compile everything directly into the GNU Mach kernel, there are no - * modules. - */ -#define SET_MODULE_OWNER(a) do{ } while(0) -#define EXPORT_SYMBOL(a) - - +#include <kern/debug.h> /* * We need some `schedule_task' replacement. This is defined in diff --git a/linux/src/arch/i386/kernel/bios32.c b/linux/src/arch/i386/kernel/bios32.c index 0b357be0..b069ce46 100644 --- a/linux/src/arch/i386/kernel/bios32.c +++ b/linux/src/arch/i386/kernel/bios32.c @@ -166,7 +166,7 @@ static unsigned long bios32_service(unsigned long service) unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%edi)" + __asm__("lcall *(%%edi); cld" : "=a" (return_code), "=b" (address), "=c" (length), @@ -206,10 +206,10 @@ static int check_pcibios(void) int pack; if ((pcibios_entry = bios32_service(PCI_SERVICE))) { - pci_indirect.address = pcibios_entry; + pci_indirect.address = phystokv(pcibios_entry); save_flags(flags); cli(); - __asm__("lcall *(%%edi)\n\t" + __asm__("lcall *(%%edi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:\tshl $8, %%eax\n\t" @@ -254,7 +254,7 @@ static int pci_bios_find_class (unsigned int class_code, unsigned short index, unsigned long flags; save_flags(flags); cli(); - __asm__ ("lcall *(%%edi)\n\t" + __asm__ ("lcall *(%%edi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -279,7 +279,7 @@ static int pci_bios_find_device (unsigned short vendor, unsigned short device_id unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%edi)\n\t" + __asm__("lcall *(%%edi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -304,7 +304,7 @@ static int pci_bios_read_config_byte(unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -326,7 +326,7 @@ static int pci_bios_read_config_word (unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -348,7 +348,7 @@ static int pci_bios_read_config_dword (unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -370,7 +370,7 @@ static int pci_bios_write_config_byte (unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -392,7 +392,7 @@ static int pci_bios_write_config_word (unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -414,7 +414,7 @@ static int pci_bios_write_config_dword (unsigned char bus, unsigned long flags; save_flags(flags); cli(); - __asm__("lcall *(%%esi)\n\t" + __asm__("lcall *(%%esi); cld\n\t" "jc 1f\n\t" "xor %%ah, %%ah\n" "1:" @@ -903,7 +903,7 @@ unsigned long pcibios_init(unsigned long memory_start, unsigned long memory_end) } else { bios32_entry = check->fields.entry; printk ("pcibios_init : BIOS32 Service Directory entry at 0x%lx\n", bios32_entry); - bios32_indirect.address = bios32_entry; + bios32_indirect.address = phystokv(bios32_entry); } } } diff --git a/linux/src/drivers/block/floppy.c b/linux/src/drivers/block/floppy.c index 0314a0b7..1b96c44f 100644 --- a/linux/src/drivers/block/floppy.c +++ b/linux/src/drivers/block/floppy.c @@ -177,6 +177,8 @@ static inline int __get_order(unsigned long size); #include <linux/blk.h> #include <linux/cdrom.h> /* for the compatibility eject ioctl */ +#include <linux/dev/glue/glue.h> + #ifndef FLOPPY_MOTOR_MASK #define FLOPPY_MOTOR_MASK 0xf0 @@ -4167,8 +4169,6 @@ static void floppy_release_irq_and_dma(void) #ifdef MODULE -extern char *get_options(char *str, int *ints); - char *floppy=NULL; static void parse_floppy_cfg_string(char *cfg) diff --git a/linux/src/drivers/block/ide-cd.c b/linux/src/drivers/block/ide-cd.c index 17f60b94..e4548f54 100644 --- a/linux/src/drivers/block/ide-cd.c +++ b/linux/src/drivers/block/ide-cd.c @@ -153,6 +153,10 @@ /***************************************************************************/ +#ifdef MACH +#include <kern/sched_prim.h> +#endif + #include <linux/types.h> #include <linux/kernel.h> #include <linux/delay.h> @@ -626,7 +630,7 @@ static void cdrom_queue_request_sense (ide_drive_t *drive, pc->c[0] = REQUEST_SENSE; pc->c[4] = len; - pc->buffer = (char *)reqbuf; + pc->buffer = (unsigned char *)reqbuf; pc->buflen = len; pc->sense_data = (struct atapi_request_sense *)failed_command; @@ -844,7 +848,7 @@ static int cdrom_start_packet_command (ide_drive_t *drive, int xferlen, HANDLER is the interrupt handler to call when the command completes or there's data ready. */ static int cdrom_transfer_packet_command (ide_drive_t *drive, - char *cmd_buf, int cmd_len, + unsigned char *cmd_buf, int cmd_len, ide_handler_t *handler) { if (CDROM_CONFIG_FLAGS (drive)->drq_interrupt) { @@ -1379,10 +1383,19 @@ static void cdrom_do_packet_command (ide_drive_t *drive) cdrom_start_packet_command (drive, len, cdrom_do_pc_continuation); } - -#ifndef MACH /* Sleep for TIME jiffies. Not to be called from an interrupt handler. */ +#ifdef MACH +static +void cdrom_sleep (int time) +{ + int xxx; + + assert_wait ((event_t) &xxx, TRUE); + thread_set_timeout (time); + schedule (); +} +#else static void cdrom_sleep (int time) { @@ -1660,7 +1673,7 @@ cdrom_read_capacity (ide_drive_t *drive, unsigned *capacity, pc.sense_data = reqbuf; pc.c[0] = READ_CAPACITY; - pc.buffer = (char *)&capbuf; + pc.buffer = (unsigned char *)&capbuf; pc.buflen = sizeof (capbuf); stat = cdrom_queue_packet_command (drive, &pc); @@ -1681,7 +1694,7 @@ cdrom_read_tocentry (ide_drive_t *drive, int trackno, int msf_flag, memset (&pc, 0, sizeof (pc)); pc.sense_data = reqbuf; - pc.buffer = buf; + pc.buffer = (unsigned char *)buf; pc.buflen = buflen; pc.c[0] = SCMD_READ_TOC; pc.c[6] = trackno; @@ -1813,7 +1826,7 @@ cdrom_read_subchannel (ide_drive_t *drive, int format, memset (&pc, 0, sizeof (pc)); pc.sense_data = reqbuf; - pc.buffer = buf; + pc.buffer = (unsigned char *) buf; pc.buflen = buflen; pc.c[0] = SCMD_READ_SUBCHANNEL; pc.c[1] = 2; /* MSF addressing */ @@ -1836,7 +1849,7 @@ cdrom_mode_sense (ide_drive_t *drive, int pageno, int modeflag, memset (&pc, 0, sizeof (pc)); pc.sense_data = reqbuf; - pc.buffer = buf; + pc.buffer = (unsigned char *)buf; pc.buflen = buflen; pc.c[0] = MODE_SENSE_10; pc.c[2] = pageno | (modeflag << 6); @@ -1855,7 +1868,7 @@ cdrom_mode_select (ide_drive_t *drive, int pageno, char *buf, int buflen, memset (&pc, 0, sizeof (pc)); pc.sense_data = reqbuf; - pc.buffer = buf; + pc.buffer = (unsigned char *)buf; pc.buflen = - buflen; pc.c[0] = MODE_SELECT_10; pc.c[1] = 0x10; @@ -1971,7 +1984,7 @@ cdrom_read_block (ide_drive_t *drive, int format, int lba, int nblocks, memset (&pc, 0, sizeof (pc)); pc.sense_data = reqbuf; - pc.buffer = buf; + pc.buffer = (unsigned char *)buf; pc.buflen = buflen; #if ! STANDARD_ATAPI @@ -2698,9 +2711,12 @@ void ide_cdrom_setup (ide_drive_t *drive) CDROM_CONFIG_FLAGS (drive)->subchan_as_bcd = 0; if (drive->id != NULL) { - if (strcmp (drive->id->model, "V003S0DS") == 0 && - drive->id->fw_rev[4] == '1' && - drive->id->fw_rev[6] <= '2') { + const char *model = (const char *)drive->id->model; + const char *fw_rev = (const char *)drive->id->fw_rev; + + if (strcmp (model, "V003S0DS") == 0 && + fw_rev[4] == '1' && + fw_rev[6] <= '2') { /* Vertos 300. Some versions of this drive like to talk BCD. */ CDROM_CONFIG_FLAGS (drive)->toctracks_as_bcd = 1; @@ -2709,27 +2725,27 @@ void ide_cdrom_setup (ide_drive_t *drive) CDROM_CONFIG_FLAGS (drive)->subchan_as_bcd = 1; } - else if (strcmp (drive->id->model, "V006E0DS") == 0 && - drive->id->fw_rev[4] == '1' && - drive->id->fw_rev[6] <= '2') { + else if (strcmp (model, "V006E0DS") == 0 && + fw_rev[4] == '1' && + fw_rev[6] <= '2') { /* Vertos 600 ESD. */ CDROM_CONFIG_FLAGS (drive)->toctracks_as_bcd = 1; } - else if (strcmp (drive->id->model, "GCD-R580B") == 0) + else if (strcmp (model, "GCD-R580B") == 0) drive->cdrom_info.max_sectors = 124; - else if (strcmp (drive->id->model, + else if (strcmp (model, "NEC CD-ROM DRIVE:260") == 0 && - strcmp (drive->id->fw_rev, "1.01") == 0) { + strcmp (fw_rev, "1.01") == 0) { /* Old NEC260 (not R). */ CDROM_CONFIG_FLAGS (drive)->tocaddr_as_bcd = 1; CDROM_CONFIG_FLAGS (drive)->playmsf_as_bcd = 1; CDROM_CONFIG_FLAGS (drive)->subchan_as_bcd = 1; } - else if (strcmp (drive->id->model, "WEARNES CDD-120") == 0 && - strcmp (drive->id->fw_rev, "A1.1") == 0) { + else if (strcmp (model, "WEARNES CDD-120") == 0 && + strcmp (fw_rev, "A1.1") == 0) { /* Wearnes */ CDROM_CONFIG_FLAGS (drive)->playmsf_as_bcd = 1; CDROM_CONFIG_FLAGS (drive)->subchan_as_bcd = 1; @@ -2737,9 +2753,9 @@ void ide_cdrom_setup (ide_drive_t *drive) /* Sanyo 3 CD changer uses a non-standard command for CD changing */ - else if ((strcmp(drive->id->model, "CD-ROM CDR-C3 G") == 0) || - (strcmp(drive->id->model, "CD-ROM CDR-C3G") == 0) || - (strcmp(drive->id->model, "CD-ROM CDR_C36") == 0)) { + else if ((strcmp(model, "CD-ROM CDR-C3 G") == 0) || + (strcmp(model, "CD-ROM CDR-C3G") == 0) || + (strcmp(model, "CD-ROM CDR_C36") == 0)) { /* uses CD in slot 0 when value is set to 3 */ CDROM_STATE_FLAGS (drive)->sanyo_slot = 3; } diff --git a/linux/src/drivers/block/ide.c b/linux/src/drivers/block/ide.c index 18f2e763..7ab790d4 100644 --- a/linux/src/drivers/block/ide.c +++ b/linux/src/drivers/block/ide.c @@ -604,6 +604,9 @@ void ide_set_handler (ide_drive_t *drive, ide_handler_t *handler, unsigned int t * * Returns: 1 if lba_capacity looks sensible * 0 otherwise + * + * Note: we must not change id->cyls here, otherwise a second call + * of this routine might no longer find lba_capacity ok. */ static int lba_capacity_is_ok (struct hd_driveid *id) { @@ -611,10 +614,16 @@ static int lba_capacity_is_ok (struct hd_driveid *id) unsigned long chs_sects = id->cyls * id->heads * id->sectors; unsigned long _10_percent = chs_sects / 10; - /* very large drives (8GB+) may lie about the number of cylinders */ - if (id->cyls == 16383 && id->heads == 16 && id->sectors == 63 && lba_sects > chs_sects) { + /* + * The ATA spec tells large drives to return + * C/H/S = 16383/16/63 independent of their size. + * Some drives can be jumpered to use 15 heads instead of 16. + */ + if (id->cyls == 16383 && id->sectors == 63 && + (id->heads == 15 || id->heads == 16) && + id->lba_capacity >= 16383*63*id->heads) return 1; /* lba_capacity is our only option */ - } + /* perform a rough sanity check on lba_sects: within 10% is "okay" */ if ((lba_sects - chs_sects) < _10_percent) return 1; /* lba_capacity is good */ @@ -631,11 +640,13 @@ static int lba_capacity_is_ok (struct hd_driveid *id) /* * current_capacity() returns the capacity (in sectors) of a drive * according to its current geometry/LBA settings. + * + * It also sets select.b.lba. */ static unsigned long current_capacity (ide_drive_t *drive) { struct hd_driveid *id = drive->id; - unsigned long capacity = drive->cyl * drive->head * drive->sect; + unsigned long capacity; if (!drive->present) return 0; @@ -645,8 +656,10 @@ static unsigned long current_capacity (ide_drive_t *drive) #endif /* CONFIG_BLK_DEV_IDEFLOPPY */ if (drive->media != ide_disk) return 0x7fffffff; /* cdrom or tape */ + drive->select.b.lba = 0; /* Determine capacity, and use LBA if the drive properly supports it */ + capacity = drive->cyl * drive->head * drive->sect; if (id != NULL && (id->capability & 2) && lba_capacity_is_ok(id)) { if (id->lba_capacity >= capacity) { capacity = id->lba_capacity; @@ -1525,7 +1538,7 @@ static inline void do_rw_disk (ide_drive_t *drive, struct request *rq, unsigned */ static void execute_drive_cmd (ide_drive_t *drive, struct request *rq) { - byte *args = rq->buffer; + byte *args = (byte *)rq->buffer; if (args) { #ifdef DEBUG printk("%s: DRIVE_CMD cmd=0x%02x sc=0x%02x fr=0x%02x xx=0x%02x\n", @@ -2020,7 +2033,7 @@ static int ide_open(struct inode * inode, struct file * filp) struct request rq; check_disk_change(inode->i_rdev); ide_init_drive_cmd (&rq); - rq.buffer = door_lock; + rq.buffer = (char *)door_lock; /* * Ignore the return code from door_lock, * since the open() has already succeeded, @@ -2071,7 +2084,7 @@ static void ide_release(struct inode * inode, struct file * file) struct request rq; invalidate_buffers(inode->i_rdev); ide_init_drive_cmd (&rq); - rq.buffer = door_unlock; + rq.buffer = (char *)door_unlock; (void) ide_do_drive_cmd(drive, &rq, ide_wait); } } @@ -2321,7 +2334,7 @@ static int ide_ioctl (struct inode *inode, struct file *file, argbuf[3] = args[3]; } if (!(err = verify_area(VERIFY_WRITE,(void *)arg, argsize))) { - rq.buffer = argbuf; + rq.buffer = (char *)argbuf; err = ide_do_drive_cmd(drive, &rq, ide_wait); memcpy_tofs((void *)arg, argbuf, argsize); } @@ -2455,7 +2468,7 @@ static inline void do_identify (ide_drive_t *drive, byte cmd) ide_fixstring (id->fw_rev, sizeof(id->fw_rev), bswap); ide_fixstring (id->serial_no, sizeof(id->serial_no), bswap); - if (strstr(id->model, "E X A B Y T E N E S T")) + if (strstr((char *)id->model, "E X A B Y T E N E S T")) return; #ifdef CONFIG_BLK_DEV_IDEATAPI @@ -2474,9 +2487,12 @@ static inline void do_identify (ide_drive_t *drive, byte cmd) #endif /* CONFIG_BLK_DEV_PROMISE */ if (!drive->ide_scsi) switch (type) { case 0: - if (!strstr(id->model, "oppy") && !strstr(id->model, "poyp") && !strstr(id->model, "ZIP")) + if (!strstr((char *)id->model, "oppy") && + !strstr((char *)id->model, "poyp") && + !strstr((char *)id->model, "ZIP")) printk("cdrom or floppy?, assuming "); - if (drive->media != ide_cdrom && !strstr(id->model, "CD-ROM")) { + if (drive->media != ide_cdrom && + !strstr((char *)id->model, "CD-ROM")) { #ifdef CONFIG_BLK_DEV_IDEFLOPPY printk("FLOPPY drive\n"); drive->media = ide_floppy; @@ -2565,8 +2581,7 @@ static inline void do_identify (ide_drive_t *drive, byte cmd) } /* Handle logical geometry translation by the drive */ if ((id->field_valid & 1) && id->cur_cyls && id->cur_heads - && (id->cur_heads <= 16) && id->cur_sectors) - { + && (id->cur_heads <= 16) && id->cur_sectors) { /* * Extract the physical drive geometry for our use. * Note that we purposely do *not* update the bios info. @@ -2591,7 +2606,8 @@ static inline void do_identify (ide_drive_t *drive, byte cmd) } } /* Use physical geometry if what we have still makes no sense */ - if ((!drive->head || drive->head > 16) && id->heads && id->heads <= 16) { + if ((!drive->head || drive->head > 16) && + id->heads && id->heads <= 16) { drive->cyl = id->cyls; drive->head = id->heads; drive->sect = id->sectors; @@ -2600,24 +2616,25 @@ static inline void do_identify (ide_drive_t *drive, byte cmd) /* calculate drive capacity, and select LBA if possible */ capacity = current_capacity (drive); - /* Correct the number of cyls if the bios value is too small */ - if (!drive->forced_geom && - capacity > drive->bios_cyl * drive->bios_sect * drive->bios_head) { - unsigned long cylsize; - cylsize = drive->bios_sect * drive->bios_head; - if (cylsize == 0 || capacity/cylsize > 65535) { - drive->bios_sect = 63; - drive->bios_head = 255; - cylsize = 63*255; - } - if (capacity/cylsize > 65535) - drive->bios_cyl = 65535; - else - drive->bios_cyl = capacity/cylsize; - } - - if (!strncmp(id->model, "BMI ", 4) && - strstr(id->model, " ENHANCED IDE ") && + /* + * if possible, give fdisk access to more of the drive, + * by correcting bios_cyls: + */ + if (capacity > drive->bios_cyl * drive->bios_head * drive->bios_sect + && !drive->forced_geom && drive->bios_sect && drive->bios_head) { + int cyl = (capacity / drive->bios_sect) / drive->bios_head; + if (cyl <= 65535) + drive->bios_cyl = cyl; + else { + /* OK until 539 GB */ + drive->bios_sect = 63; + drive->bios_head = 255; + drive->bios_cyl = capacity / (63*255); + } + } + + if (!strncmp((char *)id->model, "BMI ", 4) && + strstr((char *)id->model, " ENHANCED IDE ") && drive->select.b.lba) drive->no_geom = 1; @@ -2856,7 +2873,7 @@ static inline byte probe_for_drive (ide_drive_t *drive) (void) do_probe(drive, WIN_PIDENTIFY); /* look for ATAPI device */ #endif /* CONFIG_BLK_DEV_IDEATAPI */ } - if (drive->id && strstr(drive->id->model, "E X A B Y T E N E S T")) + if (drive->id && strstr((char *)drive->id->model, "E X A B Y T E N E S T")) enable_nest(drive); if (!drive->present) return 0; /* drive not found */ @@ -3343,7 +3360,7 @@ done: * This routine is called from the partition-table code in genhd.c * to "convert" a drive to a logical geometry with fewer than 1024 cyls. * - * The second parameter, "xparm", determines exactly how the translation + * The second parameter, "xparm", determines exactly how the translation * will be handled: * 0 = convert to CHS with fewer than 1024 cyls * using the same method as Ontrack DiskManager. @@ -3351,10 +3368,11 @@ done: * -1 = similar to "0", plus redirect sector 0 to sector 1. * >1 = convert to a CHS geometry with "xparm" heads. * - * Returns 0 if the translation was not possible, if the device was not + * Returns 0 if the translation was not possible, if the device was not * an IDE disk drive, or if a geometry was "forced" on the commandline. * Returns 1 if the geometry translation was successful. */ + int ide_xlate_1024 (kdev_t i_rdev, int xparm, const char *msg) { ide_drive_t *drive; @@ -3362,7 +3380,11 @@ int ide_xlate_1024 (kdev_t i_rdev, int xparm, const char *msg) const byte *heads = head_vals; unsigned long tracks; - if ((drive = get_info_ptr(i_rdev)) == NULL || drive->forced_geom) + drive = get_info_ptr(i_rdev); + if (!drive) + return 0; + + if (drive->forced_geom) return 0; if (xparm > 1 && xparm <= drive->bios_head && drive->bios_sect == 63) @@ -3370,6 +3392,10 @@ int ide_xlate_1024 (kdev_t i_rdev, int xparm, const char *msg) printk("%s ", msg); + if (xparm < 0 && (drive->bios_cyl * drive->bios_head * drive->bios_sect) < (1024 * 16 * 63)) { + return 0; /* small disk: no translation needed */ + } + if (drive->id) { drive->cyl = drive->id->cyls; drive->head = drive->id->heads; @@ -3608,9 +3634,6 @@ static void ide_probe_promise_20246(void) hwif->ctl_port = io[tmp + 1] + 2; hwif->noprobe = 0; } -#ifdef CONFIG_BLK_DEV_TRITON - ide_init_promise (bus, fn, &ide_hwifs[2], &ide_hwifs[3], io[4]); -#endif /* CONFIG_BLK_DEV_TRITON */ } #endif /* CONFIG_PCI */ @@ -3643,6 +3666,9 @@ static void probe_for_hwifs (void) ide_probe_pci (PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82371_0, &ide_init_triton, 1); ide_probe_pci (PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82371SB_1, &ide_init_triton, 0); ide_probe_pci (PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82371AB, &ide_init_triton, 0); + ide_probe_pci (PCI_VENDOR_ID_SI, PCI_DEVICE_ID_SI_5513, &ide_init_triton, 0); + ide_probe_pci (PCI_VENDOR_ID_VIA, PCI_DEVICE_ID_VIA_82C586_1, &ide_init_triton, 0); + ide_probe_pci (PCI_VENDOR_ID_AL, PCI_DEVICE_ID_AL_M5229, &ide_init_triton, 0); #endif /* CONFIG_BLK_DEV_TRITON */ ide_probe_promise_20246(); } diff --git a/linux/src/drivers/block/ide.h b/linux/src/drivers/block/ide.h index 44331607..edeedc97 100644 --- a/linux/src/drivers/block/ide.h +++ b/linux/src/drivers/block/ide.h @@ -152,7 +152,8 @@ typedef unsigned char byte; /* used everywhere */ /* * Timeouts for various operations: */ -#define WAIT_DRQ (5*HZ/100) /* 50msec - spec allows up to 20ms */ +#define WAIT_DRQ (1*HZ) /* 1s - spec allows up to 20ms, but CF + * cards and SSD drives need more */ #ifdef CONFIG_APM #define WAIT_READY (5*HZ) /* 5sec - some laptops are very slow */ #else @@ -198,7 +199,7 @@ struct atapi_request_sense { }; struct packet_command { - char *buffer; + unsigned char *buffer; int buflen; int stat; struct atapi_request_sense *sense_data; @@ -429,7 +430,8 @@ typedef void (ide_selectproc_t) (ide_drive_t *); typedef enum { ide_unknown, ide_generic, ide_triton, ide_cmd640, ide_dtc2278, ide_ali14xx, ide_qd6580, ide_umc8672, ide_ht6560b, - ide_promise, ide_promise_udma } + ide_promise, ide_hpt343, ide_udma, + ide_ultra66 } hwif_chipset_t; typedef struct hwif_s { diff --git a/linux/src/drivers/block/triton.c b/linux/src/drivers/block/triton.c index 3ed069d9..37eff2b3 100644 --- a/linux/src/drivers/block/triton.c +++ b/linux/src/drivers/block/triton.c @@ -1,107 +1,20 @@ /* * linux/drivers/block/triton.c Version 1.13 Aug 12, 1996 + * Version 1.13a June 1998 - new chipsets + * Version 1.13b July 1998 - DMA blacklist + * Version 1.14 June 22, 1999 * + * Copyright (c) 1998-1999 Andre Hedrick * Copyright (c) 1995-1996 Mark Lord * May be copied or modified under the terms of the GNU General Public License */ /* - * This module provides support for the Bus Master IDE DMA function - * of the Intel PCI Triton I/II chipsets (i82371FB or i82371SB). - * - * Pretty much the same code will work for the OPTi "Viper" chipset. - * Look for DMA support for this in linux kernel 2.1.xx, when it appears. - * - * DMA is currently supported only for hard disk drives (not cdroms). - * - * Support for cdroms will likely be added at a later date, - * after broader experience has been obtained with hard disks. - * - * Up to four drives may be enabled for DMA, and the Triton chipset will - * (hopefully) arbitrate the PCI bus among them. Note that the i82371 chip - * provides a single "line buffer" for the BM IDE function, so performance of - * multiple (two) drives doing DMA simultaneously will suffer somewhat, - * as they contest for that resource bottleneck. This is handled transparently - * inside the i82371 chip. - * - * By default, DMA support is prepared for use, but is currently enabled only - * for drives which support multi-word DMA mode2 (mword2), or which are - * recognized as "good" (see table below). Drives with only mode0 or mode1 - * (single or multi) DMA should also work with this chipset/driver (eg. MC2112A) - * but are not enabled by default. Use "hdparm -i" to view modes supported - * by a given drive. - * - * The hdparm-2.4 (or later) utility can be used for manually enabling/disabling - * DMA support, but must be (re-)compiled against this kernel version or later. - * - * To enable DMA, use "hdparm -d1 /dev/hd?" on a per-drive basis after booting. - * If problems arise, ide.c will disable DMA operation after a few retries. - * This error recovery mechanism works and has been extremely well exercised. - * - * IDE drives, depending on their vintage, may support several different modes - * of DMA operation. The boot-time modes are indicated with a "*" in - * the "hdparm -i" listing, and can be changed with *knowledgeable* use of - * the "hdparm -X" feature. There is seldom a need to do this, as drives - * normally power-up with their "best" PIO/DMA modes enabled. - * - * Testing was done with an ASUS P55TP4XE/100 system and the following drives: - * - * Quantum Fireball 1080A (1Gig w/83kB buffer), DMA mode2, PIO mode4. - * - DMA mode2 works well (7.4MB/sec), despite the tiny on-drive buffer. - * - This drive also does PIO mode4, at about the same speed as DMA mode2. - * An awesome drive for the price! - * - * Fujitsu M1606TA (1Gig w/256kB buffer), DMA mode2, PIO mode4. - * - DMA mode2 gives horrible performance (1.6MB/sec), despite the good - * size of the on-drive buffer and a boasted 10ms average access time. - * - PIO mode4 was better, but peaked at a mere 4.5MB/sec. - * - * Micropolis MC2112A (1Gig w/508kB buffer), drive pre-dates EIDE and ATA2. - * - DMA works fine (2.2MB/sec), probably due to the large on-drive buffer. - * - This older drive can also be tweaked for fastPIO (3.7MB/sec) by using - * maximum clock settings (5,4) and setting all flags except prefetch. - * - * Western Digital AC31000H (1Gig w/128kB buffer), DMA mode1, PIO mode3. - * - DMA does not work reliably. The drive appears to be somewhat tardy - * in deasserting DMARQ at the end of a sector. This is evident in - * the observation that WRITEs work most of the time, depending on - * cache-buffer occupancy, but multi-sector reads seldom work. - * - * Testing was done with a Gigabyte GA-586 ATE system and the following drive: - * (Uwe Bonnes - bon@elektron.ikp.physik.th-darmstadt.de) - * - * Western Digital AC31600H (1.6Gig w/128kB buffer), DMA mode2, PIO mode4. - * - much better than its 1Gig cousin, this drive is reported to work - * very well with DMA (7.3MB/sec). - * - * Other drives: - * - * Maxtor 7540AV (515Meg w/32kB buffer), DMA modes mword0/sword2, PIO mode3. - * - a budget drive, with budget performance, around 3MB/sec. - * - * Western Digital AC2850F (814Meg w/64kB buffer), DMA mode1, PIO mode3. - * - another "caviar" drive, similar to the AC31000, except that this one - * worked with DMA in at least one system. Throughput is about 3.8MB/sec - * for both DMA and PIO. - * - * Conner CFS850A (812Meg w/64kB buffer), DMA mode2, PIO mode4. - * - like most Conner models, this drive proves that even a fast interface - * cannot improve slow media. Both DMA and PIO peak around 3.5MB/sec. - * - * Maxtor 71260AT (1204Meg w/256kB buffer), DMA mword0/sword2, PIO mode3. - * - works with DMA, on some systems (but not always on others, eg. Dell), - * giving 3-4MB/sec performance, about the same as mode3. - * - * If you have any drive models to add, email your results to: mlord@pobox.com - * Keep an eye on /var/adm/messages for "DMA disabled" messages. - * - * Some people have reported trouble with Intel Zappa motherboards. - * This can be fixed by upgrading the AMI BIOS to version 1.00.04.BS0, - * available from ftp://ftp.intel.com/pub/bios/10004bs0.exe - * (thanks to Glen Morrell <glen@spin.Stanford.edu> for researching this). - * - * And, yes, Intel Zappa boards really *do* use the Triton IDE ports. + * This module provides support for Bus Master IDE DMA functions in various + * motherboard chipsets and PCI controller cards. + * Please check /Documentation/ide.txt and /Documentation/udma.txt for details. */ + #include <linux/config.h> #include <linux/types.h> #include <linux/kernel.h> @@ -116,10 +29,24 @@ #include <asm/io.h> #include <asm/dma.h> +#include <asm/irq.h> #include "ide.h" #undef DISPLAY_TRITON_TIMINGS /* define this to display timings */ +#undef DISPLAY_APOLLO_TIMINGS /* define this for extensive debugging information */ +#undef DISPLAY_ALI15X3_TIMINGS /* define this for extensive debugging information */ + +#if defined(CONFIG_PROC_FS) +#include <linux/stat.h> +#include <linux/proc_fs.h> +#ifdef DISPLAY_APOLLO_TIMINGS +#include <linux/via_ide_dma.h> +#endif +#ifdef DISPLAY_ALI15X3_TIMINGS +#include <linux/ali_ide_dma.h> +#endif +#endif /* * good_dma_drives() lists the model names (from "hdparm -i") @@ -133,6 +60,27 @@ const char *good_dma_drives[] = {"Micropolis 2112A", NULL}; /* + * bad_dma_drives() lists the model names (from "hdparm -i") + * of drives which supposedly support (U)DMA but which are + * known to corrupt data with this interface under Linux. + * + * Note: the list was generated by statistical analysis of problem + * reports. It's not clear if there are problems with the drives, + * or with some combination of drive/controller or what. + * + * You can forcibly override this if you wish. This is the kernel + * 'Tread carefully' list. + * + * Finally see http://www.wdc.com/quality/err-rec.html if you have + * one of the listed drives. + */ +const char *bad_dma_drives[] = {"WDC AC11000H", + "WDC AC22100H", + "WDC AC32500H", + "WDC AC33100H", + NULL}; + +/* * Our Physical Region Descriptor (PRD) table should be large enough * to handle the biggest I/O request we are likely to see. Since requests * can have no more than 256 sectors, and since the typical blocksize is @@ -151,6 +99,8 @@ const char *good_dma_drives[] = {"Micropolis 2112A", #define PRD_BYTES 8 #define PRD_ENTRIES (PAGE_SIZE / (2 * PRD_BYTES)) #define DEFAULT_BMIBA 0xe800 /* in case BIOS did not init it */ +#define DEFAULT_BMCRBA 0xcc00 /* VIA's default value */ +#define DEFAULT_BMALIBA 0xd400 /* ALI's default value */ /* * dma_intr() is the handler for disk read/write DMA interrupts @@ -245,23 +195,57 @@ static int build_dmatable (ide_drive_t *drive) return 1; /* let the PIO routines handle this weirdness */ } +/* + * We will only enable drives with multi-word (mode2) (U)DMA capabilities, + * and ignore the very rare cases of drives that can only do single-word + * (modes 0 & 1) (U)DMA transfers. We also discard "blacklisted" hard disks. + */ static int config_drive_for_dma (ide_drive_t *drive) { +#ifndef CONFIG_BLK_DEV_FORCE_DMA const char **list; struct hd_driveid *id = drive->id; +#endif + +#ifdef CONFIG_BLK_DEV_FORCE_DMA + drive->using_dma = 1; + return 0; +#else + if (HWIF(drive)->chipset == ide_hpt343) { + drive->using_dma = 0; /* no DMA */ + return 1; /* DMA disabled */ + } if (id && (id->capability & 1)) { - /* Enable DMA on any drive that has UltraDMA (mode 0/1/2) enabled */ - if (id->field_valid & 4) /* UltraDMA */ - if ((id->dma_ultra & (id->dma_ultra >> 8) & 7)) { + /* Consult the list of known "bad" drives */ + list = bad_dma_drives; + while (*list) { + if (!strcmp(*list++,id->model)) { + drive->using_dma = 0; /* no DMA */ + printk("ide: Disabling DMA modes on %s drive (%s).\n", drive->name, id->model); + return 1; /* DMA disabled */ + } + } + /* Enable DMA on any drive that has mode 4 or 2 UltraDMA enabled */ + if (id->field_valid & 4) { /* UltraDMA */ + /* Enable DMA on any drive that has mode 4 UltraDMA enabled */ + if (((id->dma_ultra & 0x1010) == 0x1010) && + (id->word93 & 0x2000) && + (HWIF(drive)->chipset == ide_ultra66)) { drive->using_dma = 1; - return 0; /* dma enabled */ + return 0; /* DMA enabled */ + } else + /* Enable DMA on any drive that has mode 2 UltraDMA enabled */ + if ((id->dma_ultra & 0x404) == 0x404) { + drive->using_dma = 1; + return 0; /* DMA enabled */ } - /* Enable DMA on any drive that has mode2 DMA (multi or single) enabled */ + } + /* Enable DMA on any drive that has mode2 DMA enabled */ if (id->field_valid & 2) /* regular DMA */ - if ((id->dma_mword & 0x404) == 0x404 || (id->dma_1word & 0x404) == 0x404) { + if ((id->dma_mword & 0x404) == 0x404) { drive->using_dma = 1; - return 0; /* dma enabled */ + return 0; /* DMA enabled */ } /* Consult the list of known "good" drives */ list = good_dma_drives; @@ -273,6 +257,7 @@ static int config_drive_for_dma (ide_drive_t *drive) } } return 1; /* DMA not enabled */ +#endif } /* @@ -388,22 +373,260 @@ static void init_triton_dma (ide_hwif_t *hwif, unsigned short base) } /* + * Set VIA Chipset Timings for (U)DMA modes enabled. + */ +static int set_via_timings (byte bus, byte fn, byte post, byte flush) +{ + byte via_config = 0; + int rc = 0; + + /* setting IDE read prefetch buffer and IDE post write buffer */ + if ((rc = pcibios_read_config_byte(bus, fn, 0x41, &via_config))) + return (1); + if ((rc = pcibios_write_config_byte(bus, fn, 0x41, via_config | post))) + return (1); + + /* setting Channel read and End-of-sector FIFO flush: */ + if ((rc = pcibios_read_config_byte(bus, fn, 0x46, &via_config))) + return (1); + if ((rc = pcibios_write_config_byte(bus, fn, 0x46, via_config | flush))) + return (1); + + return (0); +} + +static int setup_aladdin (byte bus, byte fn) +{ + byte confreg0 = 0, confreg1 = 0, progif = 0; + int errors = 0; + + if (pcibios_read_config_byte(bus, fn, 0x50, &confreg1)) + goto veryspecialsettingserror; + if (!(confreg1 & 0x02)) + if (pcibios_write_config_byte(bus, fn, 0x50, confreg1 | 0x02)) + goto veryspecialsettingserror; + + if (pcibios_read_config_byte(bus, fn, 0x09, &progif)) + goto veryspecialsettingserror; + if (!(progif & 0x40)) { + /* + * The way to enable them is to set progif + * writable at 0x4Dh register, and set bit 6 + * of progif to 1: + */ + if (pcibios_read_config_byte(bus, fn, 0x4d, &confreg0)) + goto veryspecialsettingserror; + if (confreg0 & 0x80) + if (pcibios_write_config_byte(bus, fn, 0x4d, confreg0 & ~0x80)) + goto veryspecialsettingserror; + if (pcibios_write_config_byte(bus, fn, 0x09, progif | 0x40)) + goto veryspecialsettingserror; + if (confreg0 & 0x80) + if (pcibios_write_config_byte(bus, fn, 0x4d, confreg0)) + errors++; + } + + if ((pcibios_read_config_byte(bus, fn, 0x09, &progif)) || (!(progif & 0x40))) + goto veryspecialsettingserror; + + printk("ide: ALI15X3: enabled read of IDE channels state (en/dis-abled) %s.\n", + errors ? "with Error(s)" : "Succeeded" ); + return 1; +veryspecialsettingserror: + printk("ide: ALI15X3: impossible to enable read of IDE channels state (en/dis-abled)!\n"); + return 0; +} + +void set_promise_hpt343_extra (unsigned short device, unsigned int bmiba) +{ + switch(device) { + case PCI_DEVICE_ID_PROMISE_20246: + if(!check_region((bmiba+16), 16)) + request_region((bmiba+16), 16, "PDC20246"); + break; + case PCI_DEVICE_ID_PROMISE_20262: + if (!check_region((bmiba+48), 48)) + request_region((bmiba+48), 48, "PDC20262"); + break; + case PCI_DEVICE_ID_TTI_HPT343: + if(!check_region((bmiba+16), 16)) + request_region((bmiba+16), 16, "HPT343"); + break; + default: + break; + } +} + +#define HPT343_PCI_INIT_REG 0x80 + +/* * ide_init_triton() prepares the IDE driver for DMA operation. * This routine is called once, from ide.c during driver initialization, - * for each triton chipset which is found (unlikely to be more than one). + * for each BM-DMA chipset which is found (rarely more than one). */ void ide_init_triton (byte bus, byte fn) { - int rc = 0, h; - int dma_enabled = 0; - unsigned short pcicmd; - unsigned int bmiba, timings; + byte bridgebus, bridgefn, bridgeset = 0, hpt34x_flag = 0; + unsigned char irq = 0; + int dma_enabled = 0, rc = 0, h; + unsigned short io[6], count = 0, step_count = 0, pass_count = 0; + unsigned short pcicmd, vendor, device, class; + unsigned int bmiba, timings, reg, tmp; + unsigned int addressbios = 0; + unsigned long flags; + unsigned index; + +#if defined(DISPLAY_APOLLO_TIMINGS) || defined(DISPLAY_ALI15X3_TIMINGS) + bmide_bus = bus; + bmide_fn = fn; +#endif /* DISPLAY_APOLLO_TIMINGS || DISPLAY_ALI15X3_TIMINGS */ + +/* + * We pick up the vendor, device, and class info for selecting the correct + * controller that is supported. Since we can access this routine more than + * once with the use of onboard and off-board EIDE controllers, a method + * of determining "who is who for what" is needed. + */ + + pcibios_read_config_word (bus, fn, PCI_VENDOR_ID, &vendor); + pcibios_read_config_word (bus, fn, PCI_DEVICE_ID, &device); + pcibios_read_config_word (bus, fn, PCI_CLASS_DEVICE, &class); + pcibios_read_config_byte (bus, fn, PCI_INTERRUPT_LINE, &irq); + + switch(vendor) { + case PCI_VENDOR_ID_INTEL: + printk("ide: Intel 82371 "); + switch(device) { + case PCI_DEVICE_ID_INTEL_82371_0: + printk("PIIX (single FIFO) "); + break; + case PCI_DEVICE_ID_INTEL_82371SB_1: + printk("PIIX3 (dual FIFO) "); + break; + case PCI_DEVICE_ID_INTEL_82371AB: + printk("PIIX4 (dual FIFO) "); + break; + default: + printk(" (unknown) 0x%04x ", device); + break; + } + printk("DMA Bus Mastering IDE "); + break; + case PCI_VENDOR_ID_SI: + printk("ide: SiS 5513 (dual FIFO) DMA Bus Mastering IDE "); + break; + case PCI_VENDOR_ID_VIA: + printk("ide: VIA VT82C586B (split FIFO) UDMA Bus Mastering IDE "); + break; + case PCI_VENDOR_ID_TTI: + /*PCI_CLASS_STORAGE_UNKNOWN == class */ + if (device == PCI_DEVICE_ID_TTI_HPT343) { + pcibios_write_config_byte(bus, fn, HPT343_PCI_INIT_REG, 0x00); + pcibios_read_config_word(bus, fn, PCI_COMMAND, &pcicmd); + hpt34x_flag = (pcicmd & PCI_COMMAND_MEMORY) ? 1 : 0; +#if 1 + if (!hpt34x_flag) { + save_flags(flags); + cli(); + pcibios_write_config_word(bus, fn, PCI_COMMAND, pcicmd & ~PCI_COMMAND_IO); + pcibios_read_config_dword(bus, fn, PCI_BASE_ADDRESS_4, &bmiba); + pcibios_write_config_dword(bus, fn, PCI_BASE_ADDRESS_0, bmiba | 0x20); + pcibios_write_config_dword(bus, fn, PCI_BASE_ADDRESS_1, bmiba | 0x34); + pcibios_write_config_dword(bus, fn, PCI_BASE_ADDRESS_2, bmiba | 0x28); + pcibios_write_config_dword(bus, fn, PCI_BASE_ADDRESS_3, bmiba | 0x3c); + pcibios_write_config_word(bus, fn, PCI_COMMAND, pcicmd); + bmiba = 0; + restore_flags(flags); + } +#endif + pcibios_write_config_byte(bus, fn, PCI_LATENCY_TIMER, 0x20); + goto hpt343_jump_in; + } else { + printk("ide: HPTXXX did == 0x%04X unsupport chipset error.\n", device); + return; + } + case PCI_VENDOR_ID_PROMISE: + /* + * I have been able to make my Promise Ultra33 UDMA card change class. + * It has reported as both PCI_CLASS_STORAGE_RAID and PCI_CLASS_STORAGE_IDE. + * Since the PCI_CLASS_STORAGE_RAID mode should automatically mirror the + * two halves of the PCI_CONFIG register data, but sometimes it forgets. + * Thus we guarantee that they are identical, with a quick check and + * correction if needed. + * PDC20246 (primary) PDC20247 (secondary) IDE hwif's. + * + * PDC20262 Promise Ultra66 UDMA. + * + * Note that Promise "stories,fibs,..." about this device not being + * capable of ATAPI and AT devices. + */ + if (class != PCI_CLASS_STORAGE_IDE) { + unsigned char irq_mirror = 0; + + pcibios_read_config_byte(bus, fn, (PCI_INTERRUPT_LINE)|0x80, &irq_mirror); + if (irq != irq_mirror) { + pcibios_write_config_byte(bus, fn, (PCI_INTERRUPT_LINE)|0x80, irq); + } + } + case PCI_VENDOR_ID_ARTOP: + /* PCI_CLASS_STORAGE_SCSI == class */ + /* + * I have found that by stroking rom_enable_bit on both the AEC6210U/UF and + * PDC20246 controller cards, the features desired are almost guaranteed + * to be enabled and compatible. This ROM may not be registered in the + * config data, but it can be turned on. Registration failure has only + * been observed if and only if Linux sets up the pci_io_address in the + * 0x6000 range. If they are setup in the 0xef00 range it is reported. + * WHY??? got me......... + */ +hpt343_jump_in: + printk("ide: %s UDMA Bus Mastering ", + (device == PCI_DEVICE_ID_ARTOP_ATP850UF) ? "AEC6210" : + (device == PCI_DEVICE_ID_PROMISE_20246) ? "PDC20246" : + (device == PCI_DEVICE_ID_PROMISE_20262) ? "PDC20262" : + (hpt34x_flag && (device == PCI_DEVICE_ID_TTI_HPT343)) ? "HPT345" : + (device == PCI_DEVICE_ID_TTI_HPT343) ? "HPT343" : "UNKNOWN"); + pcibios_read_config_dword(bus, fn, PCI_ROM_ADDRESS, &addressbios); + if (addressbios) { + pcibios_write_config_byte(bus, fn, PCI_ROM_ADDRESS, addressbios | PCI_ROM_ADDRESS_ENABLE); + printk("with ROM enabled at 0x%08x", addressbios); + } + /* + * This was stripped out of 2.1.XXX kernel code and parts from a patch called + * promise_update. This finds the PCI_BASE_ADDRESS spaces and makes them + * available for configuration later. + * PCI_BASE_ADDRESS_0 hwif0->io_base + * PCI_BASE_ADDRESS_1 hwif0->ctl_port + * PCI_BASE_ADDRESS_2 hwif1->io_base + * PCI_BASE_ADDRESS_3 hwif1->ctl_port + * PCI_BASE_ADDRESS_4 bmiba + */ + memset(io, 0, 6 * sizeof(unsigned short)); + for (reg = PCI_BASE_ADDRESS_0; reg <= PCI_BASE_ADDRESS_5; reg += 4) { + pcibios_read_config_dword(bus, fn, reg, &tmp); + if (tmp & PCI_BASE_ADDRESS_SPACE_IO) + io[count++] = tmp & PCI_BASE_ADDRESS_IO_MASK; + } + break; + case PCI_VENDOR_ID_AL: + save_flags(flags); + cli(); + for (index = 0; !pcibios_find_device (PCI_VENDOR_ID_AL, PCI_DEVICE_ID_AL_M1533, index, &bridgebus, &bridgefn); ++index) { + bridgeset = setup_aladdin(bus, fn); + } + restore_flags(flags); + printk("ide: ALI15X3 (dual FIFO) DMA Bus Mastering IDE "); + break; + default: + return; + } + + printk("\n Controller on PCI bus %d function %d\n", bus, fn); - printk("ide: i82371 PIIX (Triton) on PCI bus %d function %d\n", bus, fn); /* * See if IDE and BM-DMA features are enabled: */ - if ((rc = pcibios_read_config_word(bus, fn, 0x04, &pcicmd))) + if ((rc = pcibios_read_config_word(bus, fn, PCI_COMMAND, &pcicmd))) goto quit; if ((pcicmd & 1) == 0) { printk("ide: ports are not enabled (BIOS)\n"); @@ -427,21 +650,29 @@ void ide_init_triton (byte bus, byte fn) */ int try_again = 1; do { - if ((rc = pcibios_read_config_dword(bus, fn, 0x20, &bmiba))) + if ((rc = pcibios_read_config_dword(bus, fn, PCI_BASE_ADDRESS_4, &bmiba))) goto quit; bmiba &= 0xfff0; /* extract port base address */ if (bmiba) { dma_enabled = 1; break; } else { - printk("ide: BM-DMA base register is invalid (0x%04x, PnP BIOS problem)\n", bmiba); - if (inb(DEFAULT_BMIBA) != 0xff || !try_again) + printk("ide: BM-DMA base register is invalid (0x%04x, PnP BIOS problem)\n", bmiba); + if (inb(((vendor == PCI_VENDOR_ID_AL) ? DEFAULT_BMALIBA : + (vendor == PCI_VENDOR_ID_VIA) ? DEFAULT_BMCRBA : + DEFAULT_BMIBA)) != 0xff || !try_again) break; - printk("ide: setting BM-DMA base register to 0x%04x\n", DEFAULT_BMIBA); - if ((rc = pcibios_write_config_word(bus, fn, 0x04, pcicmd&~1))) + printk("ide: setting BM-DMA base register to 0x%04x\n", + ((vendor == PCI_VENDOR_ID_AL) ? DEFAULT_BMALIBA : + (vendor == PCI_VENDOR_ID_VIA) ? DEFAULT_BMCRBA : + DEFAULT_BMIBA)); + if ((rc = pcibios_write_config_word(bus, fn, PCI_COMMAND, pcicmd&~1))) goto quit; - rc = pcibios_write_config_dword(bus, fn, 0x20, DEFAULT_BMIBA|1); - if (pcibios_write_config_word(bus, fn, 0x04, pcicmd|5) || rc) + rc = pcibios_write_config_dword(bus, fn, 0x20, + ((vendor == PCI_VENDOR_ID_AL) ? DEFAULT_BMALIBA : + (vendor == PCI_VENDOR_ID_VIA) ? DEFAULT_BMCRBA : + DEFAULT_BMIBA)|1); + if (pcibios_write_config_word(bus, fn, PCI_COMMAND, pcicmd|5) || rc) goto quit; } } while (try_again--); @@ -450,89 +681,308 @@ void ide_init_triton (byte bus, byte fn) /* * See if ide port(s) are enabled */ - if ((rc = pcibios_read_config_dword(bus, fn, 0x40, &timings))) - goto quit; - if (!(timings & 0x80008000)) { - printk("ide: neither port is enabled\n"); + if ((rc = pcibios_read_config_dword(bus, fn, + (vendor == PCI_VENDOR_ID_PROMISE) ? 0x50 : + (vendor == PCI_VENDOR_ID_ARTOP) ? 0x54 : + (vendor == PCI_VENDOR_ID_SI) ? 0x48 : + (vendor == PCI_VENDOR_ID_AL) ? 0x08 : + 0x40, &timings))) goto quit; - } + /* + * We do a vendor check since the Ultra33/66 and AEC6210 + * holds their timings in a different location. + */ +#if 0 + printk("ide: timings == %08x\n", timings); +#endif + /* + * The switch preserves some stuff that was original. + */ + switch(vendor) { + case PCI_VENDOR_ID_INTEL: + if (!(timings & 0x80008000)) { + printk("ide: INTEL: neither port is enabled\n"); + goto quit; + } + break; + case PCI_VENDOR_ID_VIA: + if(!(timings & 0x03)) { + printk("ide: VIA: neither port is enabled\n"); + goto quit; + } + break; + case PCI_VENDOR_ID_AL: + timings <<= 16; + timings >>= 24; + if (!(timings & 0x30)) { + printk("ide: ALI15X3: neither port is enabled\n"); + goto quit; + } + break; + case PCI_VENDOR_ID_SI: + timings <<= 8; + timings >>= 24; + if (!(timings & 0x06)) { + printk("ide: SIS5513: neither port is enabled\n"); + goto quit; + } + break; + case PCI_VENDOR_ID_PROMISE: + printk(" (U)DMA Burst Bit %sABLED " \ + "Primary %s Mode " \ + "Secondary %s Mode.\n", + (inb(bmiba + 0x001f) & 1) ? "EN" : "DIS", + (inb(bmiba + 0x001a) & 1) ? "MASTER" : "PCI", + (inb(bmiba + 0x001b) & 1) ? "MASTER" : "PCI" ); +#if 0 + if (!(inb(bmiba + 0x001f) & 1)) { + outb(inb(bmiba + 0x001f)|0x01, (bmiba + 0x001f)); + printk(" (U)DMA Burst Bit Forced %sABLED.\n", + (inb(bmiba + 0x001f) & 1) ? "EN" : "DIS"); + } +#endif + break; + case PCI_VENDOR_ID_ARTOP: + case PCI_VENDOR_ID_TTI: + default: + break; + } /* * Save the dma_base port addr for each interface */ for (h = 0; h < MAX_HWIFS; ++h) { + ide_hwif_t *hwif = &ide_hwifs[h]; + byte channel = ((h == 1) || (h == 3) || (h == 5)) ? 1 : 0; + + /* + * This prevents the first contoller from accidentally + * initalizing the hwif's that it does not use and block + * an off-board ide-pci from getting in the game. + */ + if ((step_count >= 2) || (pass_count >= 2)) { + goto quit; + } + +#if 0 + if (hwif->chipset == ide_unknown) + printk("ide: index == %d channel(%d)\n", h, channel); +#endif + +#ifdef CONFIG_BLK_DEV_OFFBOARD + /* + * This is a forced override for the onboard ide controller + * to be enabled, if one chooses to have an offboard ide-pci + * card as the primary booting device. This beasty is + * for offboard UDMA upgrades with hard disks, but saving + * the onboard DMA2 controllers for CDROMS, TAPES, ZIPS, etc... + */ + if (((vendor == PCI_VENDOR_ID_INTEL) || + (vendor == PCI_VENDOR_ID_SI) || + (vendor == PCI_VENDOR_ID_VIA) || + (vendor == PCI_VENDOR_ID_AL)) && (h >= 2)) { + hwif->io_base = channel ? 0x170 : 0x1f0; + hwif->ctl_port = channel ? 0x376 : 0x3f6; + hwif->irq = channel ? 15 : 14; + hwif->noprobe = 0; + } +#endif /* CONFIG_BLK_DEV_OFFBOARD */ + /* + * If the chipset is listed as "ide_unknown", lets get a + * hwif while they last. This does the first check on + * the current availability of the ide_hwifs[h] in question. + */ + if (hwif->chipset != ide_unknown) { + continue; + } else if (vendor == PCI_VENDOR_ID_INTEL) { + unsigned short time; #ifdef DISPLAY_TRITON_TIMINGS - byte s_clks, r_clks; - unsigned short devid; + byte s_clks, r_clks; + unsigned short devid; #endif /* DISPLAY_TRITON_TIMINGS */ - ide_hwif_t *hwif = &ide_hwifs[h]; - unsigned short time; - if (hwif->io_base == 0x1f0) { - time = timings & 0xffff; - if ((time & 0x8000) == 0) /* interface enabled? */ - continue; - hwif->chipset = ide_triton; - if (dma_enabled) - init_triton_dma(hwif, bmiba); - } else if (hwif->io_base == 0x170) { - time = timings >> 16; - if ((time & 0x8000) == 0) /* interface enabled? */ + pass_count++; + if (hwif->io_base == 0x1f0) { + time = timings & 0xffff; + if ((time & 0x8000) == 0) /* interface enabled? */ + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba); + step_count++; + } else if (hwif->io_base == 0x170) { + time = timings >> 16; + if ((time & 0x8000) == 0) /* interface enabled? */ + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba + 8); + step_count++; + } else { continue; - hwif->chipset = ide_triton; - if (dma_enabled) - init_triton_dma(hwif, bmiba + 8); - } else - continue; + } #ifdef DISPLAY_TRITON_TIMINGS - s_clks = ((~time >> 12) & 3) + 2; - r_clks = ((~time >> 8) & 3) + 1; - printk(" %s timing: (0x%04x) sample_CLKs=%d, recovery_CLKs=%d\n", - hwif->name, time, s_clks, r_clks); - if ((time & 0x40) && !pcibios_read_config_word(bus, fn, 0x02, &devid) - && devid == PCI_DEVICE_ID_INTEL_82371SB_1) - { - byte stime; - if (pcibios_read_config_byte(bus, fn, 0x44, &stime)) { - if (hwif->io_base == 0x1f0) { - s_clks = ~stime >> 6; - r_clks = ~stime >> 4; + s_clks = ((~time >> 12) & 3) + 2; + r_clks = ((~time >> 8) & 3) + 1; + printk(" %s timing: (0x%04x) sample_CLKs=%d, recovery_CLKs=%d\n", + hwif->name, time, s_clks, r_clks); + if ((time & 0x40) && !pcibios_read_config_word(bus, fn, PCI_DEVICE_ID, &devid) + && devid == PCI_DEVICE_ID_INTEL_82371SB_1) { + byte stime; + if (pcibios_read_config_byte(bus, fn, 0x44, &stime)) { + if (hwif->io_base == 0x1f0) { + s_clks = ~stime >> 6; + r_clks = ~stime >> 4; + } else { + s_clks = ~stime >> 2; + r_clks = ~stime; + } + s_clks = (s_clks & 3) + 2; + r_clks = (r_clks & 3) + 1; + printk(" slave: sample_CLKs=%d, recovery_CLKs=%d\n", + s_clks, r_clks); + } + } + print_triton_drive_flags (0, time & 0xf); + print_triton_drive_flags (1, (time >> 4) & 0xf); +#endif /* DISPLAY_TRITON_TIMINGS */ + } else if (vendor == PCI_VENDOR_ID_SI) { + pass_count++; + if (hwif->io_base == 0x1f0) { + if ((timings & 0x02) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba); + step_count++; + } else if (hwif->io_base == 0x170) { + if ((timings & 0x04) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba + 8); + step_count++; + } else { + continue; + } + } else if (vendor == PCI_VENDOR_ID_VIA) { + pass_count++; + if (hwif->io_base == 0x1f0) { + if ((timings & 0x02) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba); + if (set_via_timings(bus, fn, 0xc0, 0xa0)) + goto quit; +#ifdef DISPLAY_APOLLO_TIMINGS + proc_register_dynamic(&proc_root, &via_proc_entry); +#endif /* DISPLAY_APOLLO_TIMINGS */ + step_count++; + } else if (hwif->io_base == 0x170) { + if ((timings & 0x01) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba + 8); + if (set_via_timings(bus, fn, 0x30, 0x50)) + goto quit; + step_count++; + } else { + continue; + } + } else if (vendor == PCI_VENDOR_ID_AL) { + byte ideic, inmir; + byte irq_routing_table[] = { -1, 9, 3, 10, 4, 5, 7, 6, + 1, 11, 0, 12, 0, 14, 0, 15 }; + + if (bridgeset) { + pcibios_read_config_byte(bridgebus, bridgefn, 0x58, &ideic); + ideic = ideic & 0x03; + if ((channel && ideic == 0x03) || (!channel && !ideic)) { + pcibios_read_config_byte(bridgebus, bridgefn, 0x44, &inmir); + inmir = inmir & 0x0f; + hwif->irq = irq_routing_table[inmir]; + } else if (channel && !(ideic & 0x01)) { + pcibios_read_config_byte(bridgebus, bridgefn, 0x75, &inmir); + inmir = inmir & 0x0f; + hwif->irq = irq_routing_table[inmir]; + } + } + pass_count++; + if (hwif->io_base == 0x1f0) { + if ((timings & 0x20) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba); + outb(inb(bmiba+2) & 0x60, bmiba+2); + if (inb(bmiba+2) & 0x80) + printk("ALI15X3: simplex device: DMA forced\n"); +#ifdef DISPLAY_ALI15X3_TIMINGS + proc_register_dynamic(&proc_root, &ali_proc_entry); +#endif /* DISPLAY_ALI15X3_TIMINGS */ + step_count++; + } else if (hwif->io_base == 0x170) { + if ((timings & 0x10) == 0) + continue; + hwif->chipset = ide_triton; + if (dma_enabled) + init_triton_dma(hwif, bmiba + 8); + outb(inb(bmiba+10) & 0x60, bmiba+10); + if (inb(bmiba+10) & 0x80) + printk("ALI15X3: simplex device: DMA forced\n"); + step_count++; + } else { + continue; + } + } else if ((vendor == PCI_VENDOR_ID_PROMISE) || + (vendor == PCI_VENDOR_ID_ARTOP) || + (vendor == PCI_VENDOR_ID_TTI)) { + pass_count++; + if (vendor == PCI_VENDOR_ID_TTI) { + if ((!hpt34x_flag) && (h < 2)) { + goto quit; + } else if (hpt34x_flag) { + hwif->io_base = channel ? (bmiba + 0x28) : (bmiba + 0x20); + hwif->ctl_port = channel ? (bmiba + 0x3e) : (bmiba + 0x36); + } else { + goto io_temps; + } + } else { +io_temps: + tmp = channel ? 2 : 0; + hwif->io_base = io[tmp]; + hwif->ctl_port = io[tmp + 1] + 2; + } + hwif->irq = irq; + hwif->noprobe = 0; + + if (device == PCI_DEVICE_ID_ARTOP_ATP850UF) { + hwif->serialized = 1; + } + + if ((vendor == PCI_VENDOR_ID_PROMISE) || + (vendor == PCI_VENDOR_ID_TTI)) { + set_promise_hpt343_extra(device, bmiba); + } + + if (dma_enabled) { + if ((!check_region(bmiba, 8)) && (!channel)) { + hwif->chipset = ((vendor == PCI_VENDOR_ID_TTI) && !hpt34x_flag) ? ide_hpt343 : + (device == PCI_DEVICE_ID_PROMISE_20262) ? ide_ultra66 : ide_udma; + init_triton_dma(hwif, bmiba); + step_count++; + } else if ((!check_region((bmiba + 0x08), 8)) && (channel)) { + hwif->chipset = ((vendor == PCI_VENDOR_ID_TTI) && !hpt34x_flag) ? ide_hpt343 : + (device == PCI_DEVICE_ID_PROMISE_20262) ? ide_ultra66 : ide_udma; + init_triton_dma(hwif, bmiba + 8); + step_count++; } else { - s_clks = ~stime >> 2; - r_clks = ~stime; + continue; } - s_clks = (s_clks & 3) + 2; - r_clks = (r_clks & 3) + 1; - printk(" slave: sample_CLKs=%d, recovery_CLKs=%d\n", - s_clks, r_clks); } } - print_triton_drive_flags (0, time & 0xf); - print_triton_drive_flags (1, (time >> 4) & 0xf); -#endif /* DISPLAY_TRITON_TIMINGS */ } -quit: if (rc) printk("ide: pcibios access failed - %s\n", pcibios_strerror(rc)); -} - -void ide_init_promise (byte bus, byte fn, ide_hwif_t *hwif0, ide_hwif_t *hwif1, unsigned short dma) -{ - int rc; - unsigned short pcicmd; - unsigned int bmiba = 0; - - printk("ide: Enabling DMA for Promise Technology IDE Ultra-DMA 33 on PCI bus %d function %d, port 0x%04x\n", bus, fn, dma); - if ((rc = pcibios_read_config_word(bus, fn, 0x04, &pcicmd)) || (pcicmd & 1) == 0 || (pcicmd & 4) == 0) - goto abort; - if ((rc = pcibios_read_config_dword(bus, fn, 0x20, &bmiba))) - goto abort; - bmiba &= 0xfff0; /* extract port base address */ - if (bmiba != dma || !bmiba) - goto abort; - hwif0->chipset = ide_promise_udma; - hwif1->chipset = ide_promise_udma; - init_triton_dma(hwif0, bmiba); - init_triton_dma(hwif1, bmiba + 0x08); - return; -abort: - printk(KERN_WARNING "ide: Promise/33 not configured correctly (BIOS)\n"); + quit: if (rc) printk("ide: pcibios access failed - %s\n", pcibios_strerror(rc)); } diff --git a/linux/src/drivers/net/3c503.c b/linux/src/drivers/net/3c503.c index a562db35..8ce488d8 100644 --- a/linux/src/drivers/net/3c503.c +++ b/linux/src/drivers/net/3c503.c @@ -192,7 +192,7 @@ el2_probe1(struct device *dev, int ioaddr) } if (ei_debug && version_printed++ == 0) - printk(version); + printk("%s", version); dev->base_addr = ioaddr; /* Allocate dev->priv and fill in 8390 specific dev fields. */ diff --git a/linux/src/drivers/net/3c505.c b/linux/src/drivers/net/3c505.c index b2163ecc..d78dad5b 100644 --- a/linux/src/drivers/net/3c505.c +++ b/linux/src/drivers/net/3c505.c @@ -1408,7 +1408,7 @@ static int elp_sense(struct device *dev) /* Wait for a while; the adapter may still be booting up */ if (elp_debug > 0) - printk(stilllooking_msg); + printk("%s", stilllooking_msg); if (orig_HCR & DIR) { /* If HCR.DIR is up, we pull it down. HSR.DIR should follow. */ outb_control(orig_HCR & ~DIR, addr); @@ -1439,7 +1439,7 @@ static int elp_sense(struct device *dev) * a hard reset. Also, do a hard reset if selected at the compile time. */ if (elp_debug > 0) - printk(found_msg); + printk("%s", found_msg); return 0; } diff --git a/linux/src/drivers/net/8390.c b/linux/src/drivers/net/8390.c index 9864106b..747ccb04 100644 --- a/linux/src/drivers/net/8390.c +++ b/linux/src/drivers/net/8390.c @@ -703,16 +703,13 @@ static void set_multicast_list(struct device *dev) int ethdev_init(struct device *dev) { if (ei_debug > 1) - printk(version); + printk("%s", version); if (dev->priv == NULL) { - struct ei_device *ei_local; - dev->priv = kmalloc(sizeof(struct ei_device), GFP_KERNEL); if (dev->priv == NULL) return -ENOMEM; memset(dev->priv, 0, sizeof(struct ei_device)); - ei_local = (struct ei_device *)dev->priv; } dev->hard_start_xmit = &ei_start_xmit; diff --git a/linux/src/drivers/net/at1700.c b/linux/src/drivers/net/at1700.c index 953c1889..9e42ab48 100644 --- a/linux/src/drivers/net/at1700.c +++ b/linux/src/drivers/net/at1700.c @@ -595,7 +595,9 @@ net_rx(struct device *dev) /* The inverse routine to net_open(). */ static int net_close(struct device *dev) { +#if 0 struct net_local *lp = (struct net_local *)dev->priv; +#endif int ioaddr = dev->base_addr; dev->tbusy = 1; diff --git a/linux/src/drivers/net/de4x5.c b/linux/src/drivers/net/de4x5.c index 114f6a74..a66f0564 100644 --- a/linux/src/drivers/net/de4x5.c +++ b/linux/src/drivers/net/de4x5.c @@ -1493,7 +1493,7 @@ de4x5_sw_reset(struct device *dev) status = -EIO; } - lp->tx_new = (++lp->tx_new) % lp->txRingSize; + lp->tx_new = (lp->tx_new + 1) % lp->txRingSize; lp->tx_old = lp->tx_new; return status; @@ -1553,7 +1553,7 @@ de4x5_queue_pkt(struct sk_buff *skb, struct device *dev) #endif outl(POLL_DEMAND, DE4X5_TPD);/* Start the TX */ - lp->tx_new = (++lp->tx_new) % lp->txRingSize; + lp->tx_new = (lp->tx_new + 1) % lp->txRingSize; dev->trans_start = jiffies; if (TX_BUFFS_AVAIL) { @@ -1708,7 +1708,7 @@ de4x5_rx(struct device *dev) } /* Change buffer ownership for this frame, back to the adapter */ - for (;lp->rx_old!=entry;lp->rx_old=(++lp->rx_old)%lp->rxRingSize) { + for (;lp->rx_old!=entry;lp->rx_old=(lp->rx_old+1)%lp->rxRingSize) { lp->rx_ring[lp->rx_old].status = cpu_to_le32(R_OWN); barrier(); } @@ -1719,7 +1719,7 @@ de4x5_rx(struct device *dev) /* ** Update entry information */ - lp->rx_new = (++lp->rx_new) % lp->rxRingSize; + lp->rx_new = (lp->rx_new + 1) % lp->rxRingSize; } return 0; @@ -1768,7 +1768,7 @@ de4x5_tx(struct device *dev) } /* Update all the pointers */ - lp->tx_old = (++lp->tx_old) % lp->txRingSize; + lp->tx_old = (lp->tx_old + 1) % lp->txRingSize; } if (TX_BUFFS_AVAIL && dev->tbusy) { /* Any resources available? */ @@ -1838,7 +1838,7 @@ de4x5_rx_ovfc(struct device *dev) for (; (s32)le32_to_cpu(lp->rx_ring[lp->rx_new].status)>=0;) { lp->rx_ring[lp->rx_new].status = cpu_to_le32(R_OWN); - lp->rx_new = (++lp->rx_new % lp->rxRingSize); + lp->rx_new = (lp->rx_new + 1) % lp->rxRingSize; } outl(omr, DE4X5_OMR); @@ -1963,7 +1963,7 @@ set_multicast_list(struct device *dev) load_packet(dev, lp->setup_frame, TD_IC | PERFECT_F | TD_SET | SETUP_FRAME_LEN, NULL); - lp->tx_new = (++lp->tx_new) % lp->txRingSize; + lp->tx_new = (lp->tx_new + 1) % lp->txRingSize; outl(POLL_DEMAND, DE4X5_TPD); /* Start the TX */ dev->trans_start = jiffies; } @@ -3576,7 +3576,7 @@ ping_media(struct device *dev, int msec) lp->tmp = lp->tx_new; /* Remember the ring position */ load_packet(dev, lp->frame, TD_LS | TD_FS | sizeof(lp->frame), NULL); - lp->tx_new = (++lp->tx_new) % lp->txRingSize; + lp->tx_new = (lp->tx_new + 1) % lp->txRingSize; outl(POLL_DEMAND, DE4X5_TPD); } @@ -5111,7 +5111,7 @@ mii_get_phy(struct device *dev) lp->useMII = TRUE; /* Search the MII address space for possible PHY devices */ - for (n=0, lp->mii_cnt=0, i=1; !((i==1) && (n==1)); i=(++i)%DE4X5_MAX_MII) { + for (n=0, lp->mii_cnt=0, i=1; !((i==1) && (n==1)); i=(i+1)%DE4X5_MAX_MII) { lp->phy[lp->active].addr = i; if (i==0) n++; /* Count cycles */ while (de4x5_reset_phy(dev)<0) udelay(100);/* Wait for reset */ @@ -5607,7 +5607,7 @@ de4x5_ioctl(struct device *dev, struct ifreq *rq, int cmd) while (test_and_set_bit(0, (void *)&dev->tbusy) != 0); load_packet(dev, lp->setup_frame, TD_IC | PERFECT_F | TD_SET | SETUP_FRAME_LEN, NULL); - lp->tx_new = (++lp->tx_new) % lp->txRingSize; + lp->tx_new = (lp->tx_new + 1) % lp->txRingSize; outl(POLL_DEMAND, DE4X5_TPD); /* Start the TX */ dev->tbusy = 0; /* Unlock the TX ring */ diff --git a/linux/src/drivers/net/depca.c b/linux/src/drivers/net/depca.c index 8cf6fc80..e1b03429 100644 --- a/linux/src/drivers/net/depca.c +++ b/linux/src/drivers/net/depca.c @@ -966,7 +966,7 @@ depca_rx(struct device *dev) } } /* Change buffer ownership for this last frame, back to the adapter */ - for (; lp->rx_old!=entry; lp->rx_old=(++lp->rx_old)&lp->rxRingMask) { + for (; lp->rx_old!=entry; lp->rx_old=(lp->rx_old+1)&lp->rxRingMask) { writel(readl(&lp->rx_ring[lp->rx_old].base) | R_OWN, &lp->rx_ring[lp->rx_old].base); } @@ -976,7 +976,7 @@ depca_rx(struct device *dev) /* ** Update entry information */ - lp->rx_new = (++lp->rx_new) & lp->rxRingMask; + lp->rx_new = (lp->rx_new + 1) & lp->rxRingMask; } return 0; @@ -1017,7 +1017,7 @@ depca_tx(struct device *dev) } /* Update all the pointers */ - lp->tx_old = (++lp->tx_old) & lp->txRingMask; + lp->tx_old = (lp->tx_old + 1) & lp->txRingMask; } return 0; @@ -1540,7 +1540,7 @@ static int load_packet(struct device *dev, struct sk_buff *skb) /* set up the buffer descriptors */ len = (skb->len < ETH_ZLEN) ? ETH_ZLEN : skb->len; - for (i = entry; i != end; i = (++i) & lp->txRingMask) { + for (i = entry; i != end; i = (i + 1) & lp->txRingMask) { /* clean out flags */ writel(readl(&lp->tx_ring[i].base) & ~T_FLAGS, &lp->tx_ring[i].base); writew(0x0000, &lp->tx_ring[i].misc); /* clears other error flags */ diff --git a/linux/src/drivers/net/pci-scan.c b/linux/src/drivers/net/pci-scan.c index 6187d5db..60525b76 100644 --- a/linux/src/drivers/net/pci-scan.c +++ b/linux/src/drivers/net/pci-scan.c @@ -347,7 +347,7 @@ int pci_drv_register(struct drv_id_info *drv_id, void *initial_device) pci_tbl[chip_idx].name, pciaddr, irq); if ( ! (pci_flags & PCI_UNUSED_IRQ) && - (irq == 0 || irq == 255)) { + (irq == 0 || irq >= 16)) { if (pci_bus == 32) /* Broken CardBus activation. */ printk(KERN_WARNING "Resources for CardBus device '%s' have" " not been allocated.\n" diff --git a/linux/src/drivers/scsi/FlashPoint.c b/linux/src/drivers/scsi/FlashPoint.c index a74c3c5f..aae35c03 100644 --- a/linux/src/drivers/scsi/FlashPoint.c +++ b/linux/src/drivers/scsi/FlashPoint.c @@ -2326,7 +2326,6 @@ void Debug_Load(UCHAR p_card, UCHAR p_bug_data); extern unsigned int SccbGlobalFlags; -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" /*---------------------------------------------------------------------- * * @@ -5352,7 +5351,7 @@ void Debug_Load(UCHAR p_card, UCHAR p_bug_data) } #endif -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * @@ -5424,7 +5423,7 @@ UCHAR debug_int[MAX_CARDS][debug_size] = { 0 }; UCHAR debug_index[MAX_CARDS] = { 0 }; UCHAR reserved_1[3] = { 0 }; #endif -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * @@ -7518,7 +7517,6 @@ void sinits(PSCCB p_sccb, UCHAR p_card) } -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" /*---------------------------------------------------------------------- * * @@ -8267,7 +8265,6 @@ void phaseBusFree(ULONG port, UCHAR p_card) -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" /*---------------------------------------------------------------------- * * @@ -8659,7 +8656,7 @@ void autoCmdCmplt(ULONG p_port, UCHAR p_card) queueCmdComplete(&BL_Card[p_card], currSCCB, p_card); } -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * @@ -9351,7 +9348,7 @@ void hostDataXferRestart(PSCCB currSCCB) currSCCB->Sccb_XferCnt = currSCCB->DataLength - currSCCB->Sccb_ATC; } } -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * @@ -10581,7 +10578,7 @@ void scsavdi(UCHAR p_card, ULONG p_port) utilEEWrite(p_port, sum_data, EEPROM_CHECK_SUM/2); utilEEWriteOnOff(p_port,0); /* Turn off write access */ } -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * @@ -11026,7 +11023,7 @@ void DiagEEPROM(ULONG p_port) } -#ident "$Id: FlashPoint.c,v 1.1 1999/04/26 05:53:56 tb Exp $" + /*---------------------------------------------------------------------- * * diff --git a/linux/src/drivers/scsi/advansys.c b/linux/src/drivers/scsi/advansys.c index 05cd0f6e..ef61fac9 100644 --- a/linux/src/drivers/scsi/advansys.c +++ b/linux/src/drivers/scsi/advansys.c @@ -2033,8 +2033,8 @@ STATIC void AscSetISAPNPWaitForKey(void); STATIC uchar AscGetChipIRQ(PortAddr, ushort); STATIC uchar AscSetChipIRQ(PortAddr, uchar, ushort); STATIC ushort AscGetChipBiosAddress(PortAddr, ushort); -STATIC int DvcEnterCritical(void); -STATIC void DvcLeaveCritical(int); +STATIC long DvcEnterCritical(void); +STATIC void DvcLeaveCritical(long); STATIC void DvcInPortWords(PortAddr, ushort *, int); STATIC void DvcOutPortWords(PortAddr, ushort *, int); STATIC void DvcOutPortDWords(PortAddr, ulong *, int); @@ -2870,8 +2870,8 @@ typedef struct adv_scsi_req_q { /* * Device drivers must define the following functions. */ -STATIC int DvcEnterCritical(void); -STATIC void DvcLeaveCritical(int); +STATIC long DvcEnterCritical(void); +STATIC void DvcLeaveCritical(long); STATIC void DvcSleepMilliSecond(ulong); STATIC uchar DvcAdvReadPCIConfigByte(ADV_DVC_VAR *, ushort); STATIC void DvcAdvWritePCIConfigByte(ADV_DVC_VAR *, ushort, uchar); @@ -5400,7 +5400,7 @@ advansys_queuecommand(Scsi_Cmnd *scp, void (*done)(Scsi_Cmnd *)) { struct Scsi_Host *shp; asc_board_t *boardp; - int flags; + long flags; Scsi_Cmnd *done_scp; shp = scp->host; @@ -5495,7 +5495,7 @@ advansys_abort(Scsi_Cmnd *scp) asc_board_t *boardp; ASC_DVC_VAR *asc_dvc_varp; ADV_DVC_VAR *adv_dvc_varp; - int flags; + long flags; int do_scsi_done; int scp_found; Scsi_Cmnd *done_scp = NULL; @@ -5734,7 +5734,7 @@ advansys_reset(Scsi_Cmnd *scp, unsigned int reset_flags) asc_board_t *boardp; ASC_DVC_VAR *asc_dvc_varp; ADV_DVC_VAR *adv_dvc_varp; - int flags; + long flags; Scsi_Cmnd *done_scp = NULL, *last_scp = NULL; Scsi_Cmnd *tscp, *new_last_scp; int do_scsi_done; @@ -6269,7 +6269,7 @@ advansys_interrupt(int irq, void *dev_id, struct pt_regs *regs) #endif /* version >= v1.3.70 */ { #if LINUX_VERSION_CODE < ASC_LINUX_VERSION(2,1,95) - int flags; + long flags; #else /* version >= v2.1.95 */ unsigned long flags; #endif /* version >= v2.1.95 */ @@ -9194,7 +9194,7 @@ asc_prt_line(char *buf, int buflen, char *fmt, ...) ret = vsprintf(s, fmt, args); ASC_ASSERT(ret < ASC_PRTLINE_SIZE); if (buf == NULL) { - (void) printk(s); + (void) printk("%s", s); ret = 0; } else { ret = ASC_MIN(buflen, ret); @@ -9227,10 +9227,10 @@ DvcSleepMilliSecond(ulong n) } } -STATIC int +STATIC long DvcEnterCritical(void) { - int flags; + long flags; save_flags(flags); cli(); @@ -9238,7 +9238,7 @@ DvcEnterCritical(void) } STATIC void -DvcLeaveCritical(int flags) +DvcLeaveCritical(long flags) { restore_flags(flags); } @@ -10198,7 +10198,7 @@ asc_prt_hex(char *f, uchar *s, int l) STATIC int interrupts_enabled(void) { - int flags; + long flags; save_flags(flags); if (flags & 0x0200) { @@ -15078,7 +15078,7 @@ AdvISR(ADV_DVC_VAR *asc_dvc) uchar int_stat; ushort next_done_loc, target_bit; int completed_q; - int flags; + long flags; ADV_SCSI_REQ_Q *scsiq; ASC_REQ_SENSE *sense_data; int ret; diff --git a/linux/src/drivers/scsi/aha152x.c b/linux/src/drivers/scsi/aha152x.c index a5724bf8..44fe1b0c 100644 --- a/linux/src/drivers/scsi/aha152x.c +++ b/linux/src/drivers/scsi/aha152x.c @@ -1539,7 +1539,7 @@ void aha152x_done(struct Scsi_Host *shpnt, int error) void aha152x_intr(int irqno, void *dev_id, struct pt_regs * regs) { struct Scsi_Host *shpnt = aha152x_host[irqno-IRQ_MIN]; - unsigned int flags; + unsigned long flags; int done=0, phase; #if defined(DEBUG_RACE) diff --git a/linux/src/drivers/scsi/aha1542.c b/linux/src/drivers/scsi/aha1542.c index 68954d75..cc27e5c0 100644 --- a/linux/src/drivers/scsi/aha1542.c +++ b/linux/src/drivers/scsi/aha1542.c @@ -356,7 +356,7 @@ static void aha1542_intr_handle(int irq, void *dev_id, struct pt_regs *regs) void (*my_done)(Scsi_Cmnd *) = NULL; int errstatus, mbi, mbo, mbistatus; int number_serviced; - unsigned int flags; + unsigned long flags; struct Scsi_Host * shost; Scsi_Cmnd * SCtmp; int flag; @@ -875,7 +875,7 @@ void aha1542_setup( char *str, int *ints) if (ints[0] < 1 || ints[0] > 4) { printk("aha1542: %s\n", str ); - printk(ahausage); + printk("%s", ahausage); printk("aha1542: Wrong parameters may cause system malfunction.. We try anyway..\n"); } @@ -905,7 +905,7 @@ void aha1542_setup( char *str, int *ints) break; default: printk("aha1542: %s\n", str ); - printk(ahausage); + printk("%s", ahausage); printk("aha1542: Valid values for DMASPEED are 5-8, 10 MB/s. Using jumper defaults.\n"); break; } diff --git a/linux/src/drivers/scsi/aic7xxx.c b/linux/src/drivers/scsi/aic7xxx.c index be817bb7..93bed411 100644 --- a/linux/src/drivers/scsi/aic7xxx.c +++ b/linux/src/drivers/scsi/aic7xxx.c @@ -2800,7 +2800,7 @@ aic7xxx_done_cmds_complete(struct aic7xxx_host *p) { Scsi_Cmnd *cmd; #if LINUX_VERSION_CODE < KERNEL_VERSION(2,1,95) - unsigned int cpu_flags = 0; + unsigned long cpu_flags = 0; #endif DRIVER_LOCK diff --git a/linux/src/drivers/scsi/eata.c b/linux/src/drivers/scsi/eata.c index ce859ced..49f08275 100644 --- a/linux/src/drivers/scsi/eata.c +++ b/linux/src/drivers/scsi/eata.c @@ -881,7 +881,7 @@ __initfunc (static inline int port_detect \ sprintf(name, "%s%d", driver_name, j); - printk("probing eata on %lx\n", port_base); + printk("\rprobing eata on %lx", port_base); if(check_region(port_base, REGION_SIZE)) { printk("%s: address 0x%03lx in use, skipping probe.\n", name, port_base); diff --git a/linux/src/drivers/scsi/gdth.c b/linux/src/drivers/scsi/gdth.c index 6cc2b780..0a4bef89 100644 --- a/linux/src/drivers/scsi/gdth.c +++ b/linux/src/drivers/scsi/gdth.c @@ -2851,8 +2851,9 @@ __initfunc (int gdth_detect(Scsi_Host_Template *shtp)) ushort eisa_slot,device_id,index; gdth_pci_str pcistr; int i,j,hanum; +#if LINUX_VERSION_CODE < 0x020000 unchar b; - +#endif #ifdef DEBUG_GDTH printk("GDT: This driver contains debugging information !! Trace level = %d\n", @@ -2878,7 +2879,7 @@ __initfunc (int gdth_detect(Scsi_Host_Template *shtp)) } /* initializations */ - gdth_polling = TRUE; b = 0; + gdth_polling = TRUE; for (i=0; i<GDTH_MAXCMDS; ++i) for (j=0; j<MAXHA; ++j) gdth_cmd_tab[i][j].cmnd = UNUSED_CMND; diff --git a/linux/src/drivers/scsi/hosts.c b/linux/src/drivers/scsi/hosts.c index 07646041..0f1beddd 100644 --- a/linux/src/drivers/scsi/hosts.c +++ b/linux/src/drivers/scsi/hosts.c @@ -133,6 +133,10 @@ #include "53c7,8xx.h" #endif +#ifdef CONFIG_SCSI_SYM53C8XX +#include "sym53c8xx.h" +#endif + #ifdef CONFIG_SCSI_NCR53C8XX #include "ncr53c8xx.h" #endif @@ -298,6 +302,9 @@ static Scsi_Host_Template builtin_scsi_hosts[] = #ifdef CONFIG_SCSI_NCR53C7xx NCR53c7xx, #endif +#ifdef CONFIG_SCSI_SYM53C8XX + SYM53C8XX, +#endif #ifdef CONFIG_SCSI_NCR53C8XX NCR53C8XX, #endif @@ -465,6 +472,7 @@ unsigned int scsi_init() * Initialize our semaphores. -1 is interpreted to mean * "inactive" - where as 0 will indicate a time out condition. */ + printk("\rprobing scsi %d/%d: %s \e[K", tpnt-builtin_scsi_hosts, MAX_SCSI_HOSTS, tpnt->name); pcount = next_scsi_host; if ((tpnt->detect) && @@ -489,6 +497,7 @@ unsigned int scsi_init() #endif } } + printk("\ndone\n"); for(shpnt=scsi_hostlist; shpnt; shpnt = shpnt->next) { diff --git a/linux/src/drivers/scsi/ncr53c8xx.c b/linux/src/drivers/scsi/ncr53c8xx.c index 1be3d9fe..0a584297 100644 --- a/linux/src/drivers/scsi/ncr53c8xx.c +++ b/linux/src/drivers/scsi/ncr53c8xx.c @@ -272,8 +272,10 @@ typedef u32 u_int32; #define u_int unsigned int #define u_long unsigned long +#ifndef MACH typedef u_long vm_offset_t; typedef int vm_size_t; +#endif #define bcopy(s, d, n) memcpy((d), (s), (n)) #define bzero(d, n) memset((d), 0, (n)) diff --git a/linux/src/drivers/scsi/ncr53c8xx.h b/linux/src/drivers/scsi/ncr53c8xx.h index cc009ca6..03424387 100644 --- a/linux/src/drivers/scsi/ncr53c8xx.h +++ b/linux/src/drivers/scsi/ncr53c8xx.h @@ -81,7 +81,7 @@ #endif #if LINUX_VERSION_CODE >= LinuxVersionCode(1,3,0) -# define SCSI_NCR_PROC_INFO_SUPPORT +//# define SCSI_NCR_PROC_INFO_SUPPORT #endif #if LINUX_VERSION_CODE >= LinuxVersionCode(1,3,72) diff --git a/linux/src/drivers/scsi/ppa.c b/linux/src/drivers/scsi/ppa.c index 5063a677..fd224f9e 100644 --- a/linux/src/drivers/scsi/ppa.c +++ b/linux/src/drivers/scsi/ppa.c @@ -259,7 +259,6 @@ int ppa_detect(Scsi_Host_Template * host) struct Scsi_Host *hreg; int ports; int i, nhosts; - unsigned short ppb; printk("ppa: Version %s\n", PPA_VERSION); nhosts = 0; @@ -267,7 +266,7 @@ int ppa_detect(Scsi_Host_Template * host) for (i = 0; i < parbus_no; i++) { if (parbus_base[i] == 0x0000) continue; - ppb = ppa_hosts[i].base = parbus_base[i]; + ppa_hosts[i].base = parbus_base[i]; /* sanity checks */ if (check_region(parbus_base[i], @@ -518,7 +517,7 @@ static inline int ppa_byte_in(unsigned short base, char *buffer, int len) " shrb $4,%%al\n" \ " orb %%al," #reg "\n" -static inline int ppa_nibble_in(unsigned short str_p, char *buffer, int len) +static inline int ppa_nibble_in(unsigned short base, char *buffer, int len) { for (; len; len--) { unsigned char h; diff --git a/linux/src/drivers/scsi/sym53c8xx.c b/linux/src/drivers/scsi/sym53c8xx.c new file mode 100644 index 00000000..f4969549 --- /dev/null +++ b/linux/src/drivers/scsi/sym53c8xx.c @@ -0,0 +1,14696 @@ +/****************************************************************************** +** High Performance device driver for the Symbios 53C896 controller. +** +** Copyright (C) 1998-2000 Gerard Roudier <groudier@club-internet.fr> +** +** This driver also supports all the Symbios 53C8XX controller family, +** except 53C810 revisions < 16, 53C825 revisions < 16 and all +** revisions of 53C815 controllers. +** +** This driver is based on the Linux port of the FreeBSD ncr driver. +** +** Copyright (C) 1994 Wolfgang Stanglmeier +** +**----------------------------------------------------------------------------- +** +** This program is free software; you can redistribute it and/or modify +** it under the terms of the GNU General Public License as published by +** the Free Software Foundation; either version 2 of the License, or +** (at your option) any later version. +** +** This program is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU General Public License for more details. +** +** You should have received a copy of the GNU General Public License +** along with this program; if not, write to the Free Software +** Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. +** +**----------------------------------------------------------------------------- +** +** The Linux port of the FreeBSD ncr driver has been achieved in +** november 1995 by: +** +** Gerard Roudier <groudier@club-internet.fr> +** +** Being given that this driver originates from the FreeBSD version, and +** in order to keep synergy on both, any suggested enhancements and corrections +** received on Linux are automatically a potential candidate for the FreeBSD +** version. +** +** The original driver has been written for 386bsd and FreeBSD by +** Wolfgang Stanglmeier <wolf@cologne.de> +** Stefan Esser <se@mi.Uni-Koeln.de> +** +**----------------------------------------------------------------------------- +** +** Major contributions: +** -------------------- +** +** NVRAM detection and reading. +** Copyright (C) 1997 Richard Waltham <dormouse@farsrobt.demon.co.uk> +** +******************************************************************************* +*/ + +/* +** Supported SCSI features: +** Synchronous data transfers +** Wide16 SCSI BUS +** Disconnection/Reselection +** Tagged command queuing +** SCSI Parity checking +** +** Supported NCR/SYMBIOS chips: +** 53C810A (8 bits, Fast 10, no rom BIOS) +** 53C825A (Wide, Fast 10, on-board rom BIOS) +** 53C860 (8 bits, Fast 20, no rom BIOS) +** 53C875 (Wide, Fast 20, on-board rom BIOS) +** 53C876 (Wide, Fast 20 Dual, on-board rom BIOS) +** 53C895 (Wide, Fast 40, on-board rom BIOS) +** 53C895A (Wide, Fast 40, on-board rom BIOS) +** 53C896 (Wide, Fast 40 Dual, on-board rom BIOS) +** 53C897 (Wide, Fast 40 Dual, on-board rom BIOS) +** 53C1510D (Wide, Fast 40 Dual, on-board rom BIOS) +** 53C1010 (Wide, Fast 80 Dual, on-board rom BIOS) +** 53C1010_66(Wide, Fast 80 Dual, on-board rom BIOS, 33/66MHz PCI) +** +** Other features: +** Memory mapped IO +** Module +** Shared IRQ +*/ + +/* +** Name and version of the driver +*/ +#define SCSI_NCR_DRIVER_NAME "sym53c8xx-1.7.1-20000726" + +#define SCSI_NCR_DEBUG_FLAGS (0) + +#define NAME53C "sym53c" +#define NAME53C8XX "sym53c8xx" + +/*========================================================== +** +** Include files +** +**========================================================== +*/ + +#define LinuxVersionCode(v, p, s) (((v)<<16)+((p)<<8)+(s)) + +#ifdef MODULE +#include <linux/module.h> +#endif + +#include <asm/dma.h> +#include <asm/io.h> +#include <asm/system.h> +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,17) +#include <linux/spinlock.h> +#elif LINUX_VERSION_CODE >= LinuxVersionCode(2,1,93) +#include <asm/spinlock.h> +#endif +#include <linux/delay.h> +#include <linux/signal.h> +#include <linux/sched.h> +#include <linux/errno.h> +#include <linux/pci.h> +#include <linux/string.h> +#include <linux/malloc.h> +#include <linux/mm.h> +#include <linux/ioport.h> +#include <linux/time.h> +#include <linux/timer.h> +#include <linux/stat.h> + +#include <linux/version.h> +#include <linux/blk.h> + +#ifdef CONFIG_ALL_PPC +#include <asm/prom.h> +#endif + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,35) +#include <linux/init.h> +#endif + +#ifndef __init +#define __init +#endif +#ifndef __initdata +#define __initdata +#endif + +#if LINUX_VERSION_CODE <= LinuxVersionCode(2,1,92) +#include <linux/bios32.h> +#endif + +#include "scsi.h" +#include "hosts.h" +#include "constants.h" +#include "sd.h" + +#include <linux/types.h> + +/* +** Define BITS_PER_LONG for earlier linux versions. +*/ +#ifndef BITS_PER_LONG +#if (~0UL) == 0xffffffffUL +#define BITS_PER_LONG 32 +#else +#define BITS_PER_LONG 64 +#endif +#endif + +/* +** Define the BSD style u_int32 and u_int64 type. +** Are in fact u_int32_t and u_int64_t :-) +*/ +typedef u32 u_int32; +typedef u64 u_int64; + +#include "sym53c8xx.h" + +/* +** Donnot compile integrity checking code for Linux-2.3.0 +** and above since SCSI data structures are not ready yet. +*/ +/* #if LINUX_VERSION_CODE < LinuxVersionCode(2,3,0) */ +#if 0 +#define SCSI_NCR_INTEGRITY_CHECKING +#endif + +#define MIN(a,b) (((a) < (b)) ? (a) : (b)) +#define MAX(a,b) (((a) > (b)) ? (a) : (b)) + +/* +** Hmmm... What complex some PCI-HOST bridges actually are, +** despite the fact that the PCI specifications are looking +** so smart and simple! ;-) +*/ +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,47) +#define SCSI_NCR_DYNAMIC_DMA_MAPPING +#endif + +/*========================================================== +** +** A la VMS/CAM-3 queue management. +** Implemented from linux list management. +** +**========================================================== +*/ + +typedef struct xpt_quehead { + struct xpt_quehead *flink; /* Forward pointer */ + struct xpt_quehead *blink; /* Backward pointer */ +} XPT_QUEHEAD; + +#define xpt_que_init(ptr) do { \ + (ptr)->flink = (ptr); (ptr)->blink = (ptr); \ +} while (0) + +static inline void __xpt_que_add(struct xpt_quehead * new, + struct xpt_quehead * blink, + struct xpt_quehead * flink) +{ + flink->blink = new; + new->flink = flink; + new->blink = blink; + blink->flink = new; +} + +static inline void __xpt_que_del(struct xpt_quehead * blink, + struct xpt_quehead * flink) +{ + flink->blink = blink; + blink->flink = flink; +} + +static inline int xpt_que_empty(struct xpt_quehead *head) +{ + return head->flink == head; +} + +static inline void xpt_que_splice(struct xpt_quehead *list, + struct xpt_quehead *head) +{ + struct xpt_quehead *first = list->flink; + + if (first != list) { + struct xpt_quehead *last = list->blink; + struct xpt_quehead *at = head->flink; + + first->blink = head; + head->flink = first; + + last->flink = at; + at->blink = last; + } +} + +#define xpt_que_entry(ptr, type, member) \ + ((type *)((char *)(ptr)-(unsigned long)(&((type *)0)->member))) + + +#define xpt_insque(new, pos) __xpt_que_add(new, pos, (pos)->flink) + +#define xpt_remque(el) __xpt_que_del((el)->blink, (el)->flink) + +#define xpt_insque_head(new, head) __xpt_que_add(new, head, (head)->flink) + +static inline struct xpt_quehead *xpt_remque_head(struct xpt_quehead *head) +{ + struct xpt_quehead *elem = head->flink; + + if (elem != head) + __xpt_que_del(head, elem->flink); + else + elem = 0; + return elem; +} + +#define xpt_insque_tail(new, head) __xpt_que_add(new, (head)->blink, head) + +static inline struct xpt_quehead *xpt_remque_tail(struct xpt_quehead *head) +{ + struct xpt_quehead *elem = head->blink; + + if (elem != head) + __xpt_que_del(elem->blink, head); + else + elem = 0; + return elem; +} + +/*========================================================== +** +** Configuration and Debugging +** +**========================================================== +*/ + +/* +** SCSI address of this device. +** The boot routines should have set it. +** If not, use this. +*/ + +#ifndef SCSI_NCR_MYADDR +#define SCSI_NCR_MYADDR (7) +#endif + +/* +** The maximum number of tags per logic unit. +** Used only for devices that support tags. +*/ + +#ifndef SCSI_NCR_MAX_TAGS +#define SCSI_NCR_MAX_TAGS (8) +#endif + +/* +** TAGS are actually unlimited (256 tags/lun). +** But Linux only supports 255. :) +*/ +#if SCSI_NCR_MAX_TAGS > 255 +#define MAX_TAGS 255 +#else +#define MAX_TAGS SCSI_NCR_MAX_TAGS +#endif + +/* +** Since the ncr chips only have a 8 bit ALU, we try to be clever +** about offset calculation in the TASK TABLE per LUN that is an +** array of DWORDS = 4 bytes. +*/ +#if MAX_TAGS > (512/4) +#define MAX_TASKS (1024/4) +#elif MAX_TAGS > (256/4) +#define MAX_TASKS (512/4) +#else +#define MAX_TASKS (256/4) +#endif + +/* +** This one means 'NO TAG for this job' +*/ +#define NO_TAG (256) + +/* +** Number of targets supported by the driver. +** n permits target numbers 0..n-1. +** Default is 16, meaning targets #0..#15. +** #7 .. is myself. +*/ + +#ifdef SCSI_NCR_MAX_TARGET +#define MAX_TARGET (SCSI_NCR_MAX_TARGET) +#else +#define MAX_TARGET (16) +#endif + +/* +** Number of logic units supported by the driver. +** n enables logic unit numbers 0..n-1. +** The common SCSI devices require only +** one lun, so take 1 as the default. +*/ + +#ifdef SCSI_NCR_MAX_LUN +#define MAX_LUN 64 +#else +#define MAX_LUN (1) +#endif + +/* +** Asynchronous pre-scaler (ns). Shall be 40 for +** the SCSI timings to be compliant. +*/ + +#ifndef SCSI_NCR_MIN_ASYNC +#define SCSI_NCR_MIN_ASYNC (40) +#endif + +/* +** The maximum number of jobs scheduled for starting. +** We allocate 4 entries more than the value we announce +** to the SCSI upper layer. Guess why ! :-) +*/ + +#ifdef SCSI_NCR_CAN_QUEUE +#define MAX_START (SCSI_NCR_CAN_QUEUE + 4) +#else +#define MAX_START (MAX_TARGET + 7 * MAX_TAGS) +#endif + +/* +** We donnot want to allocate more than 1 PAGE for the +** the start queue and the done queue. We hard-code entry +** size to 8 in order to let cpp do the checking. +** Allows 512-4=508 pending IOs for i386 but Linux seems for +** now not able to provide the driver with this amount of IOs. +*/ +#if MAX_START > PAGE_SIZE/8 +#undef MAX_START +#define MAX_START (PAGE_SIZE/8) +#endif + +/* +** The maximum number of segments a transfer is split into. +** We support up to 127 segments for both read and write. +*/ + +#define MAX_SCATTER (SCSI_NCR_MAX_SCATTER) +#define SCR_SG_SIZE (2) + +/* +** other +*/ + +#define NCR_SNOOP_TIMEOUT (1000000) + +/*========================================================== +** +** Miscallaneous BSDish defines. +** +**========================================================== +*/ + +#define u_char unsigned char +#define u_short unsigned short +#define u_int unsigned int +#define u_long unsigned long + +#ifndef bcopy +#define bcopy(s, d, n) memcpy((d), (s), (n)) +#endif + +#ifndef bzero +#define bzero(d, n) memset((d), 0, (n)) +#endif + +#ifndef offsetof +#define offsetof(t, m) ((size_t) (&((t *)0)->m)) +#endif + +/* +** Simple Wrapper to kernel PCI bus interface. +** +** This wrapper allows to get rid of old kernel PCI interface +** and still allows to preserve linux-2.0 compatibilty. +** In fact, it is mostly an incomplete emulation of the new +** PCI code for pre-2.2 kernels. When kernel-2.0 support +** will be dropped, we will just have to remove most of this +** code. +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,2,0) + +typedef struct pci_dev *pcidev_t; +#define PCIDEV_NULL (0) +#define PciBusNumber(d) (d)->bus->number +#define PciDeviceFn(d) (d)->devfn +#define PciVendorId(d) (d)->vendor +#define PciDeviceId(d) (d)->device +#define PciIrqLine(d) (d)->irq + +#if LINUX_VERSION_CODE > LinuxVersionCode(2,3,12) + +static int __init +pci_get_base_address(struct pci_dev *pdev, int index, u_long *base) +{ + *base = pdev->resource[index].start; + if ((pdev->resource[index].flags & 0x7) == 0x4) + ++index; + return ++index; +} +#else +static int __init +pci_get_base_address(struct pci_dev *pdev, int index, u_long *base) +{ + *base = pdev->base_address[index++]; + if ((*base & 0x7) == 0x4) { +#if BITS_PER_LONG > 32 + *base |= (((u_long)pdev->base_address[index]) << 32); +#endif + ++index; + } + return index; +} +#endif + +#else /* Incomplete emulation of current PCI code for pre-2.2 kernels */ + +typedef unsigned int pcidev_t; +#define PCIDEV_NULL (~0u) +#define PciBusNumber(d) ((d)>>8) +#define PciDeviceFn(d) ((d)&0xff) +#define __PciDev(busn, devfn) (((busn)<<8)+(devfn)) + +#define pci_present pcibios_present + +#define pci_read_config_byte(d, w, v) \ + pcibios_read_config_byte(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_read_config_word(d, w, v) \ + pcibios_read_config_word(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_read_config_dword(d, w, v) \ + pcibios_read_config_dword(PciBusNumber(d), PciDeviceFn(d), w, v) + +#define pci_write_config_byte(d, w, v) \ + pcibios_write_config_byte(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_write_config_word(d, w, v) \ + pcibios_write_config_word(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_write_config_dword(d, w, v) \ + pcibios_write_config_dword(PciBusNumber(d), PciDeviceFn(d), w, v) + +static pcidev_t __init +pci_find_device(unsigned int vendor, unsigned int device, pcidev_t prev) +{ + static unsigned short pci_index; + int retv; + unsigned char bus_number, device_fn; + + if (prev == PCIDEV_NULL) + pci_index = 0; + else + ++pci_index; + retv = pcibios_find_device (vendor, device, pci_index, + &bus_number, &device_fn); + return retv ? PCIDEV_NULL : __PciDev(bus_number, device_fn); +} + +static u_short __init PciVendorId(pcidev_t dev) +{ + u_short vendor_id; + pci_read_config_word(dev, PCI_VENDOR_ID, &vendor_id); + return vendor_id; +} + +static u_short __init PciDeviceId(pcidev_t dev) +{ + u_short device_id; + pci_read_config_word(dev, PCI_DEVICE_ID, &device_id); + return device_id; +} + +static u_int __init PciIrqLine(pcidev_t dev) +{ + u_char irq; + pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq); + return irq; +} + +static int __init +pci_get_base_address(pcidev_t dev, int offset, u_long *base) +{ + u_int32 tmp; + + pci_read_config_dword(dev, PCI_BASE_ADDRESS_0 + offset, &tmp); + *base = tmp; + offset += sizeof(u_int32); + if ((tmp & 0x7) == 0x4) { +#if BITS_PER_LONG > 32 + pci_read_config_dword(dev, PCI_BASE_ADDRESS_0 + offset, &tmp); + *base |= (((u_long)tmp) << 32); +#endif + offset += sizeof(u_int32); + } + return offset; +} + +#endif /* LINUX_VERSION_CODE >= LinuxVersionCode(2,2,0) */ + +/*========================================================== +** +** Debugging tags +** +**========================================================== +*/ + +#define DEBUG_ALLOC (0x0001) +#define DEBUG_PHASE (0x0002) +#define DEBUG_QUEUE (0x0008) +#define DEBUG_RESULT (0x0010) +#define DEBUG_POINTER (0x0020) +#define DEBUG_SCRIPT (0x0040) +#define DEBUG_TINY (0x0080) +#define DEBUG_TIMING (0x0100) +#define DEBUG_NEGO (0x0200) +#define DEBUG_TAGS (0x0400) +#define DEBUG_IC (0x0800) + +/* +** Enable/Disable debug messages. +** Can be changed at runtime too. +*/ + +#ifdef SCSI_NCR_DEBUG_INFO_SUPPORT +static int ncr_debug = SCSI_NCR_DEBUG_FLAGS; + #define DEBUG_FLAGS ncr_debug +#else + #define DEBUG_FLAGS SCSI_NCR_DEBUG_FLAGS +#endif + +/* +** SMP threading. +** +** Assuming that SMP systems are generally high end systems and may +** use several SCSI adapters, we are using one lock per controller +** instead of some global one. For the moment (linux-2.1.95), driver's +** entry points are called with the 'io_request_lock' lock held, so: +** - We are uselessly loosing a couple of micro-seconds to lock the +** controller data structure. +** - But the driver is not broken by design for SMP and so can be +** more resistant to bugs or bad changes in the IO sub-system code. +** - A small advantage could be that the interrupt code is grained as +** wished (e.g.: threaded by controller). +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,93) + +spinlock_t sym53c8xx_lock = SPIN_LOCK_UNLOCKED; +#define NCR_LOCK_DRIVER(flags) spin_lock_irqsave(&sym53c8xx_lock, flags) +#define NCR_UNLOCK_DRIVER(flags) spin_unlock_irqrestore(&sym53c8xx_lock,flags) + +#define NCR_INIT_LOCK_NCB(np) spin_lock_init(&np->smp_lock); +#define NCR_LOCK_NCB(np, flags) spin_lock_irqsave(&np->smp_lock, flags) +#define NCR_UNLOCK_NCB(np, flags) spin_unlock_irqrestore(&np->smp_lock, flags) + +#define NCR_LOCK_SCSI_DONE(np, flags) \ + spin_lock_irqsave(&io_request_lock, flags) +#define NCR_UNLOCK_SCSI_DONE(np, flags) \ + spin_unlock_irqrestore(&io_request_lock, flags) + +#else + +#define NCR_LOCK_DRIVER(flags) do { save_flags(flags); cli(); } while (0) +#define NCR_UNLOCK_DRIVER(flags) do { restore_flags(flags); } while (0) + +#define NCR_INIT_LOCK_NCB(np) do { } while (0) +#define NCR_LOCK_NCB(np, flags) do { save_flags(flags); cli(); } while (0) +#define NCR_UNLOCK_NCB(np, flags) do { restore_flags(flags); } while (0) + +#define NCR_LOCK_SCSI_DONE(np, flags) do {;} while (0) +#define NCR_UNLOCK_SCSI_DONE(np, flags) do {;} while (0) + +#endif + +/* +** Memory mapped IO +** +** Since linux-2.1, we must use ioremap() to map the io memory space. +** iounmap() to unmap it. That allows portability. +** Linux 1.3.X and 2.0.X allow to remap physical pages addresses greater +** than the highest physical memory address to kernel virtual pages with +** vremap() / vfree(). That was not portable but worked with i386 +** architecture. +*/ + +#if LINUX_VERSION_CODE < LinuxVersionCode(2,1,0) +#define ioremap vremap +#define iounmap vfree +#endif + +#ifdef __sparc__ +# include <asm/irq.h> +# if LINUX_VERSION_CODE < LinuxVersionCode(2,3,0) + /* ioremap/iounmap broken in 2.2.x on Sparc. -DaveM */ +# define ioremap(base, size) ((u_long) __va(base)) +# define iounmap(vaddr) +# endif +# define pcivtobus(p) bus_dvma_to_mem(p) +# define memcpy_to_pci(a, b, c) memcpy_toio((void *)(a), (const void *)(b), (c)) +#elif defined(__alpha__) +# define pcivtobus(p) ((p) & 0xfffffffful) +# define memcpy_to_pci(a, b, c) memcpy_toio((a), (b), (c)) +#else /* others */ +# define pcivtobus(p) (p) +# define memcpy_to_pci(a, b, c) memcpy_toio((a), (b), (c)) +#endif + +#ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED +static u_long __init remap_pci_mem(u_long base, u_long size) +{ + u_long page_base = ((u_long) base) & PAGE_MASK; + u_long page_offs = ((u_long) base) - page_base; + u_long page_remapped = (u_long) ioremap(page_base, page_offs+size); + + return page_remapped? (page_remapped + page_offs) : 0UL; +} + +static void __init unmap_pci_mem(u_long vaddr, u_long size) +{ + if (vaddr) + iounmap((void *) (vaddr & PAGE_MASK)); +} + +#endif /* not def SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ + +/* +** Insert a delay in micro-seconds and milli-seconds. +** ------------------------------------------------- +** Under Linux, udelay() is restricted to delay < 1 milli-second. +** In fact, it generally works for up to 1 second delay. +** Since 2.1.105, the mdelay() function is provided for delays +** in milli-seconds. +** Under 2.0 kernels, udelay() is an inline function that is very +** inaccurate on Pentium processors. +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,105) +#define UDELAY udelay +#define MDELAY mdelay +#else +static void UDELAY(long us) { udelay(us); } +static void MDELAY(long ms) { while (ms--) UDELAY(1000); } +#endif + +/* +** Simple power of two buddy-like allocator +** ---------------------------------------- +** This simple code is not intended to be fast, but to provide +** power of 2 aligned memory allocations. +** Since the SCRIPTS processor only supplies 8 bit arithmetic, +** this allocator allows simple and fast address calculations +** from the SCRIPTS code. In addition, cache line alignment +** is guaranteed for power of 2 cache line size. +** Enhanced in linux-2.3.44 to provide a memory pool per pcidev +** to support dynamic dma mapping. (I would have preferred a +** real bus astraction, btw). +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,0) +#define __GetFreePages(flags, order) __get_free_pages(flags, order) +#else +#define __GetFreePages(flags, order) __get_free_pages(flags, order, 0) +#endif + +#define MEMO_SHIFT 4 /* 16 bytes minimum memory chunk */ +#if PAGE_SIZE >= 8192 +#define MEMO_PAGE_ORDER 0 /* 1 PAGE maximum */ +#else +#define MEMO_PAGE_ORDER 1 /* 2 PAGES maximum */ +#endif +#define MEMO_FREE_UNUSED /* Free unused pages immediately */ +#define MEMO_WARN 1 +#define MEMO_GFP_FLAGS GFP_ATOMIC +#define MEMO_CLUSTER_SHIFT (PAGE_SHIFT+MEMO_PAGE_ORDER) +#define MEMO_CLUSTER_SIZE (1UL << MEMO_CLUSTER_SHIFT) +#define MEMO_CLUSTER_MASK (MEMO_CLUSTER_SIZE-1) + +typedef u_long m_addr_t; /* Enough bits to bit-hack addresses */ +typedef pcidev_t m_bush_t; /* Something that addresses DMAable */ + +typedef struct m_link { /* Link between free memory chunks */ + struct m_link *next; +} m_link_s; + +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING +typedef struct m_vtob { /* Virtual to Bus address translation */ + struct m_vtob *next; + m_addr_t vaddr; + m_addr_t baddr; +} m_vtob_s; +#define VTOB_HASH_SHIFT 5 +#define VTOB_HASH_SIZE (1UL << VTOB_HASH_SHIFT) +#define VTOB_HASH_MASK (VTOB_HASH_SIZE-1) +#define VTOB_HASH_CODE(m) \ + ((((m_addr_t) (m)) >> MEMO_CLUSTER_SHIFT) & VTOB_HASH_MASK) +#endif + +typedef struct m_pool { /* Memory pool of a given kind */ +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING + m_bush_t bush; + m_addr_t (*getp)(struct m_pool *); + void (*freep)(struct m_pool *, m_addr_t); +#define M_GETP() mp->getp(mp) +#define M_FREEP(p) mp->freep(mp, p) +#define GetPages() __GetFreePages(MEMO_GFP_FLAGS, MEMO_PAGE_ORDER) +#define FreePages(p) free_pages(p, MEMO_PAGE_ORDER) + int nump; + m_vtob_s *(vtob[VTOB_HASH_SIZE]); + struct m_pool *next; +#else +#define M_GETP() __GetFreePages(MEMO_GFP_FLAGS, MEMO_PAGE_ORDER) +#define M_FREEP(p) free_pages(p, MEMO_PAGE_ORDER) +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + struct m_link h[PAGE_SHIFT-MEMO_SHIFT+MEMO_PAGE_ORDER+1]; +} m_pool_s; + +static void *___m_alloc(m_pool_s *mp, int size) +{ + int i = 0; + int s = (1 << MEMO_SHIFT); + int j; + m_addr_t a; + m_link_s *h = mp->h; + + if (size > (PAGE_SIZE << MEMO_PAGE_ORDER)) + return 0; + + while (size > s) { + s <<= 1; + ++i; + } + + j = i; + while (!h[j].next) { + if (s == (PAGE_SIZE << MEMO_PAGE_ORDER)) { + h[j].next = (m_link_s *) M_GETP(); + if (h[j].next) + h[j].next->next = 0; + break; + } + ++j; + s <<= 1; + } + a = (m_addr_t) h[j].next; + if (a) { + h[j].next = h[j].next->next; + while (j > i) { + j -= 1; + s >>= 1; + h[j].next = (m_link_s *) (a+s); + h[j].next->next = 0; + } + } +#ifdef DEBUG + printk("___m_alloc(%d) = %p\n", size, (void *) a); +#endif + return (void *) a; +} + +static void ___m_free(m_pool_s *mp, void *ptr, int size) +{ + int i = 0; + int s = (1 << MEMO_SHIFT); + m_link_s *q; + m_addr_t a, b; + m_link_s *h = mp->h; + +#ifdef DEBUG + printk("___m_free(%p, %d)\n", ptr, size); +#endif + + if (size > (PAGE_SIZE << MEMO_PAGE_ORDER)) + return; + + while (size > s) { + s <<= 1; + ++i; + } + + a = (m_addr_t) ptr; + + while (1) { +#ifdef MEMO_FREE_UNUSED + if (s == (PAGE_SIZE << MEMO_PAGE_ORDER)) { + M_FREEP(a); + break; + } +#endif + b = a ^ s; + q = &h[i]; + while (q->next && q->next != (m_link_s *) b) { + q = q->next; + } + if (!q->next) { + ((m_link_s *) a)->next = h[i].next; + h[i].next = (m_link_s *) a; + break; + } + q->next = q->next->next; + a = a & b; + s <<= 1; + ++i; + } +} + +static void *__m_calloc2(m_pool_s *mp, int size, char *name, int uflags) +{ + void *p; + + p = ___m_alloc(mp, size); + + if (DEBUG_FLAGS & DEBUG_ALLOC) + printk ("new %-10s[%4d] @%p.\n", name, size, p); + + if (p) + bzero(p, size); + else if (uflags & MEMO_WARN) + printk (NAME53C8XX ": failed to allocate %s[%d]\n", name, size); + + return p; +} + +#define __m_calloc(mp, s, n) __m_calloc2(mp, s, n, MEMO_WARN) + +static void __m_free(m_pool_s *mp, void *ptr, int size, char *name) +{ + if (DEBUG_FLAGS & DEBUG_ALLOC) + printk ("freeing %-10s[%4d] @%p.\n", name, size, ptr); + + ___m_free(mp, ptr, size); + +} + +/* + * With pci bus iommu support, we use a default pool of unmapped memory + * for memory we donnot need to DMA from/to and one pool per pcidev for + * memory accessed by the PCI chip. `mp0' is the default not DMAable pool. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +static m_pool_s mp0; + +#else + +static m_addr_t ___mp0_getp(m_pool_s *mp) +{ + m_addr_t m = GetPages(); + if (m) + ++mp->nump; + return m; +} + +static void ___mp0_freep(m_pool_s *mp, m_addr_t m) +{ + FreePages(m); + --mp->nump; +} + +static m_pool_s mp0 = {0, ___mp0_getp, ___mp0_freep}; + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +static void *m_calloc(int size, char *name) +{ + u_long flags; + void *m; + NCR_LOCK_DRIVER(flags); + m = __m_calloc(&mp0, size, name); + NCR_UNLOCK_DRIVER(flags); + return m; +} + +static void m_free(void *ptr, int size, char *name) +{ + u_long flags; + NCR_LOCK_DRIVER(flags); + __m_free(&mp0, ptr, size, name); + NCR_UNLOCK_DRIVER(flags); +} + +/* + * DMAable pools. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +/* Without pci bus iommu support, all the memory is assumed DMAable */ + +#define __m_calloc_dma(b, s, n) m_calloc(s, n) +#define __m_free_dma(b, p, s, n) m_free(p, s, n) +#define __vtobus(b, p) virt_to_bus(p) + +#else + +/* + * With pci bus iommu support, we maintain one pool per pcidev and a + * hashed reverse table for virtual to bus physical address translations. + */ +static m_addr_t ___dma_getp(m_pool_s *mp) +{ + m_addr_t vp; + m_vtob_s *vbp; + + vbp = __m_calloc(&mp0, sizeof(*vbp), "VTOB"); + if (vbp) { + dma_addr_t daddr; + vp = (m_addr_t) pci_alloc_consistent(mp->bush, + PAGE_SIZE<<MEMO_PAGE_ORDER, + &daddr); + if (vp) { + int hc = VTOB_HASH_CODE(vp); + vbp->vaddr = vp; + vbp->baddr = daddr; + vbp->next = mp->vtob[hc]; + mp->vtob[hc] = vbp; + ++mp->nump; + return vp; + } + else + __m_free(&mp0, vbp, sizeof(*vbp), "VTOB"); + } + return 0; +} + +static void ___dma_freep(m_pool_s *mp, m_addr_t m) +{ + m_vtob_s **vbpp, *vbp; + int hc = VTOB_HASH_CODE(m); + + vbpp = &mp->vtob[hc]; + while (*vbpp && (*vbpp)->vaddr != m) + vbpp = &(*vbpp)->next; + if (*vbpp) { + vbp = *vbpp; + *vbpp = (*vbpp)->next; + pci_free_consistent(mp->bush, PAGE_SIZE<<MEMO_PAGE_ORDER, + (void *)vbp->vaddr, (dma_addr_t)vbp->baddr); + __m_free(&mp0, vbp, sizeof(*vbp), "VTOB"); + --mp->nump; + } +} + +static inline m_pool_s *___get_dma_pool(m_bush_t bush) +{ + m_pool_s *mp; + for (mp = mp0.next; mp && mp->bush != bush; mp = mp->next); + return mp; +} + +static m_pool_s *___cre_dma_pool(m_bush_t bush) +{ + m_pool_s *mp; + mp = __m_calloc(&mp0, sizeof(*mp), "MPOOL"); + if (mp) { + bzero(mp, sizeof(*mp)); + mp->bush = bush; + mp->getp = ___dma_getp; + mp->freep = ___dma_freep; + mp->next = mp0.next; + mp0.next = mp; + } + return mp; +} + +static void ___del_dma_pool(m_pool_s *p) +{ + struct m_pool **pp = &mp0.next; + + while (*pp && *pp != p) + pp = &(*pp)->next; + if (*pp) { + *pp = (*pp)->next; + __m_free(&mp0, p, sizeof(*p), "MPOOL"); + } +} + +static void *__m_calloc_dma(m_bush_t bush, int size, char *name) +{ + u_long flags; + struct m_pool *mp; + void *m = 0; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (!mp) + mp = ___cre_dma_pool(bush); + if (mp) + m = __m_calloc(mp, size, name); + if (mp && !mp->nump) + ___del_dma_pool(mp); + NCR_UNLOCK_DRIVER(flags); + + return m; +} + +static void __m_free_dma(m_bush_t bush, void *m, int size, char *name) +{ + u_long flags; + struct m_pool *mp; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (mp) + __m_free(mp, m, size, name); + if (mp && !mp->nump) + ___del_dma_pool(mp); + NCR_UNLOCK_DRIVER(flags); +} + +static m_addr_t __vtobus(m_bush_t bush, void *m) +{ + u_long flags; + m_pool_s *mp; + int hc = VTOB_HASH_CODE(m); + m_vtob_s *vp = 0; + m_addr_t a = ((m_addr_t) m) & ~MEMO_CLUSTER_MASK; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (mp) { + vp = mp->vtob[hc]; + while (vp && (m_addr_t) vp->vaddr != a) + vp = vp->next; + } + NCR_UNLOCK_DRIVER(flags); + return vp ? vp->baddr + (((m_addr_t) m) - a) : 0; +} + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +#define _m_calloc_dma(np, s, n) __m_calloc_dma(np->pdev, s, n) +#define _m_free_dma(np, p, s, n) __m_free_dma(np->pdev, p, s, n) +#define m_calloc_dma(s, n) _m_calloc_dma(np, s, n) +#define m_free_dma(p, s, n) _m_free_dma(np, p, s, n) +#define _vtobus(np, p) __vtobus(np->pdev, p) +#define vtobus(p) _vtobus(np, p) + +/* + * Deal with DMA mapping/unmapping. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +/* Linux versions prior to pci bus iommu kernel interface */ + +#define __unmap_scsi_data(pdev, cmd) do {; } while (0) +#define __map_scsi_single_data(pdev, cmd) (__vtobus(pdev,(cmd)->request_buffer)) +#define __map_scsi_sg_data(pdev, cmd) ((cmd)->use_sg) +#define __sync_scsi_data(pdev, cmd) do {; } while (0) + +#define scsi_sg_dma_address(sc) vtobus((sc)->address) +#define scsi_sg_dma_len(sc) ((sc)->length) + +#else + +/* Linux version with pci bus iommu kernel interface */ + +/* To keep track of the dma mapping (sg/single) that has been set */ +#define __data_mapped SCp.phase +#define __data_mapping SCp.have_data_in + +static void __unmap_scsi_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + switch(cmd->__data_mapped) { + case 2: + pci_unmap_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + break; + case 1: + pci_unmap_single(pdev, cmd->__data_mapping, + cmd->request_bufflen, dma_dir); + break; + } + cmd->__data_mapped = 0; +} + +static u_long __map_scsi_single_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + dma_addr_t mapping; + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + if (cmd->request_bufflen == 0) + return 0; + + mapping = pci_map_single(pdev, cmd->request_buffer, + cmd->request_bufflen, dma_dir); + cmd->__data_mapped = 1; + cmd->__data_mapping = mapping; + + return mapping; +} + +static int __map_scsi_sg_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int use_sg; + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + if (cmd->use_sg == 0) + return 0; + + use_sg = pci_map_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + cmd->__data_mapped = 2; + cmd->__data_mapping = use_sg; + + return use_sg; +} + +static void __sync_scsi_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + switch(cmd->__data_mapped) { + case 2: + pci_dma_sync_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + break; + case 1: + pci_dma_sync_single(pdev, cmd->__data_mapping, + cmd->request_bufflen, dma_dir); + break; + } +} + +#define scsi_sg_dma_address(sc) sg_dma_address(sc) +#define scsi_sg_dma_len(sc) sg_dma_len(sc) + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +#define unmap_scsi_data(np, cmd) __unmap_scsi_data(np->pdev, cmd) +#define map_scsi_single_data(np, cmd) __map_scsi_single_data(np->pdev, cmd) +#define map_scsi_sg_data(np, cmd) __map_scsi_sg_data(np->pdev, cmd) +#define sync_scsi_data(np, cmd) __sync_scsi_data(np->pdev, cmd) + + +/* + * Print out some buffer. + */ +static void ncr_print_hex(u_char *p, int n) +{ + while (n-- > 0) + printk (" %x", *p++); +} + +static void ncr_printl_hex(char *label, u_char *p, int n) +{ + printk("%s", label); + ncr_print_hex(p, n); + printk (".\n"); +} + +/* +** Transfer direction +** +** Until some linux kernel version near 2.3.40, low-level scsi +** drivers were not told about data transfer direction. +** We check the existence of this feature that has been expected +** for a _long_ time by all SCSI driver developers by just +** testing against the definition of SCSI_DATA_UNKNOWN. Indeed +** this is a hack, but testing against a kernel version would +** have been a shame. ;-) +*/ +#ifdef SCSI_DATA_UNKNOWN + +#define scsi_data_direction(cmd) (cmd->sc_data_direction) + +#else + +#define SCSI_DATA_UNKNOWN 0 +#define SCSI_DATA_WRITE 1 +#define SCSI_DATA_READ 2 +#define SCSI_DATA_NONE 3 + +static __inline__ int scsi_data_direction(Scsi_Cmnd *cmd) +{ + int direction; + + switch((int) cmd->cmnd[0]) { + case 0x08: /* READ(6) 08 */ + case 0x28: /* READ(10) 28 */ + case 0xA8: /* READ(12) A8 */ + direction = SCSI_DATA_READ; + break; + case 0x0A: /* WRITE(6) 0A */ + case 0x2A: /* WRITE(10) 2A */ + case 0xAA: /* WRITE(12) AA */ + direction = SCSI_DATA_WRITE; + break; + default: + direction = SCSI_DATA_UNKNOWN; + break; + } + + return direction; +} + +#endif /* SCSI_DATA_UNKNOWN */ + +/* +** Head of list of NCR boards +** +** For kernel version < 1.3.70, host is retrieved by its irq level. +** For later kernels, the internal host control block address +** (struct ncb) is used as device id parameter of the irq stuff. +*/ + +static struct Scsi_Host *first_host = NULL; + + +/* +** /proc directory entry and proc_info function +*/ +#ifdef SCSI_NCR_PROC_INFO_SUPPORT +#if LINUX_VERSION_CODE < LinuxVersionCode(2,3,27) +static struct proc_dir_entry proc_scsi_sym53c8xx = { + PROC_SCSI_SYM53C8XX, 9, NAME53C8XX, + S_IFDIR | S_IRUGO | S_IXUGO, 2 +}; +#endif +static int sym53c8xx_proc_info(char *buffer, char **start, off_t offset, + int length, int hostno, int func); +#endif + +/* +** Driver setup. +** +** This structure is initialized from linux config options. +** It can be overridden at boot-up by the boot command line. +*/ +static struct ncr_driver_setup + driver_setup = SCSI_NCR_DRIVER_SETUP; + +#ifdef SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT +static struct ncr_driver_setup + driver_safe_setup __initdata = SCSI_NCR_DRIVER_SAFE_SETUP; +# ifdef MODULE +char *sym53c8xx = 0; /* command line passed by insmod */ +# if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,30) +MODULE_PARM(sym53c8xx, "s"); +# endif +# endif +#endif + +/* +** Other Linux definitions +*/ +#define SetScsiResult(cmd, h_sts, s_sts) \ + cmd->result = (((h_sts) << 16) + ((s_sts) & 0x7f)) + +/* We may have to remind our amnesiac SCSI layer of the reason of the abort */ +#if 0 +#define SetScsiAbortResult(cmd) \ + SetScsiResult( \ + cmd, \ + (cmd)->abort_reason == DID_TIME_OUT ? DID_TIME_OUT : DID_ABORT, \ + 0xff) +#else +#define SetScsiAbortResult(cmd) SetScsiResult(cmd, DID_ABORT, 0xff) +#endif + +static void sym53c8xx_select_queue_depths( + struct Scsi_Host *host, struct scsi_device *devlist); +static void sym53c8xx_intr(int irq, void *dev_id, struct pt_regs * regs); +static void sym53c8xx_timeout(unsigned long np); + +#define initverbose (driver_setup.verbose) +#define bootverbose (np->verbose) + +#ifdef SCSI_NCR_NVRAM_SUPPORT +static u_char Tekram_sync[16] __initdata = + {25,31,37,43, 50,62,75,125, 12,15,18,21, 6,7,9,10}; +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/* +** Structures used by sym53c8xx_detect/sym53c8xx_pci_init to +** transmit device configuration to the ncr_attach() function. +*/ +typedef struct { + int bus; + u_char device_fn; + u_long base; + u_long base_2; + u_long io_port; + int irq; +/* port and reg fields to use INB, OUTB macros */ + u_long base_io; + volatile struct ncr_reg *reg; +} ncr_slot; + +typedef struct { + int type; +#define SCSI_NCR_SYMBIOS_NVRAM (1) +#define SCSI_NCR_TEKRAM_NVRAM (2) +#ifdef SCSI_NCR_NVRAM_SUPPORT + union { + Symbios_nvram Symbios; + Tekram_nvram Tekram; + } data; +#endif +} ncr_nvram; + +/* +** Structure used by sym53c8xx_detect/sym53c8xx_pci_init +** to save data on each detected board for ncr_attach(). +*/ +typedef struct { + pcidev_t pdev; + ncr_slot slot; + ncr_chip chip; + ncr_nvram *nvram; + u_char host_id; +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + u_char pqs_pds; +#endif + int attach_done; +} ncr_device; + +/*========================================================== +** +** assert () +** +**========================================================== +** +** modified copy from 386bsd:/usr/include/sys/assert.h +** +**---------------------------------------------------------- +*/ + +#define assert(expression) { \ + if (!(expression)) { \ + (void)panic( \ + "assertion \"%s\" failed: file \"%s\", line %d\n", \ + #expression, \ + __FILE__, __LINE__); \ + } \ +} + +/*========================================================== +** +** Command control block states. +** +**========================================================== +*/ + +#define HS_IDLE (0) +#define HS_BUSY (1) +#define HS_NEGOTIATE (2) /* sync/wide data transfer*/ +#define HS_DISCONNECT (3) /* Disconnected by target */ + +#define HS_DONEMASK (0x80) +#define HS_COMPLETE (4|HS_DONEMASK) +#define HS_SEL_TIMEOUT (5|HS_DONEMASK) /* Selection timeout */ +#define HS_RESET (6|HS_DONEMASK) /* SCSI reset */ +#define HS_ABORTED (7|HS_DONEMASK) /* Transfer aborted */ +#define HS_TIMEOUT (8|HS_DONEMASK) /* Software timeout */ +#define HS_FAIL (9|HS_DONEMASK) /* SCSI or PCI bus errors */ +#define HS_UNEXPECTED (10|HS_DONEMASK)/* Unexpected disconnect */ + +#define DSA_INVALID 0xffffffff + +/*========================================================== +** +** Software Interrupt Codes +** +**========================================================== +*/ + +#define SIR_BAD_STATUS (1) +#define SIR_SEL_ATN_NO_MSG_OUT (2) +#define SIR_MSG_RECEIVED (3) +#define SIR_MSG_WEIRD (4) +#define SIR_NEGO_FAILED (5) +#define SIR_NEGO_PROTO (6) +#define SIR_SCRIPT_STOPPED (7) +#define SIR_REJECT_TO_SEND (8) +#define SIR_SWIDE_OVERRUN (9) +#define SIR_SODL_UNDERRUN (10) +#define SIR_RESEL_NO_MSG_IN (11) +#define SIR_RESEL_NO_IDENTIFY (12) +#define SIR_RESEL_BAD_LUN (13) +#define SIR_TARGET_SELECTED (14) +#define SIR_RESEL_BAD_I_T_L (15) +#define SIR_RESEL_BAD_I_T_L_Q (16) +#define SIR_ABORT_SENT (17) +#define SIR_RESEL_ABORTED (18) +#define SIR_MSG_OUT_DONE (19) +#define SIR_AUTO_SENSE_DONE (20) +#define SIR_DUMMY_INTERRUPT (21) +#define SIR_DATA_OVERRUN (22) +#define SIR_BAD_PHASE (23) +#define SIR_MAX (23) + +/*========================================================== +** +** Extended error bits. +** xerr_status field of struct ccb. +** +**========================================================== +*/ + +#define XE_EXTRA_DATA (1) /* unexpected data phase */ +#define XE_BAD_PHASE (2) /* illegal phase (4/5) */ +#define XE_PARITY_ERR (4) /* unrecovered SCSI parity error */ +#define XE_SODL_UNRUN (1<<3) +#define XE_SWIDE_OVRUN (1<<4) + +/*========================================================== +** +** Negotiation status. +** nego_status field of struct ccb. +** +**========================================================== +*/ + +#define NS_NOCHANGE (0) +#define NS_SYNC (1) +#define NS_WIDE (2) +#define NS_PPR (4) + +/*========================================================== +** +** "Special features" of targets. +** quirks field of struct tcb. +** actualquirks field of struct ccb. +** +**========================================================== +*/ + +#define QUIRK_AUTOSAVE (0x01) + +/*========================================================== +** +** Capability bits in Inquire response byte 7. +** +**========================================================== +*/ + +#define INQ7_QUEUE (0x02) +#define INQ7_SYNC (0x10) +#define INQ7_WIDE16 (0x20) + +/*========================================================== +** +** A CCB hashed table is used to retrieve CCB address +** from DSA value. +** +**========================================================== +*/ + +#define CCB_HASH_SHIFT 8 +#define CCB_HASH_SIZE (1UL << CCB_HASH_SHIFT) +#define CCB_HASH_MASK (CCB_HASH_SIZE-1) +#define CCB_HASH_CODE(dsa) (((dsa) >> 11) & CCB_HASH_MASK) + +/*========================================================== +** +** Declaration of structs. +** +**========================================================== +*/ + +struct tcb; +struct lcb; +struct ccb; +struct ncb; +struct script; + +typedef struct ncb * ncb_p; +typedef struct tcb * tcb_p; +typedef struct lcb * lcb_p; +typedef struct ccb * ccb_p; + +struct link { + ncrcmd l_cmd; + ncrcmd l_paddr; +}; + +struct usrcmd { + u_long target; + u_long lun; + u_long data; + u_long cmd; +}; + +#define UC_SETSYNC 10 +#define UC_SETTAGS 11 +#define UC_SETDEBUG 12 +#define UC_SETORDER 13 +#define UC_SETWIDE 14 +#define UC_SETFLAG 15 +#define UC_SETVERBOSE 17 +#define UC_RESETDEV 18 +#define UC_CLEARDEV 19 + +#define UF_TRACE (0x01) +#define UF_NODISC (0x02) +#define UF_NOSCAN (0x04) + +/*======================================================================== +** +** Declaration of structs: target control block +** +**======================================================================== +*/ +struct tcb { + /*---------------------------------------------------------------- + ** LUN tables. + ** An array of bus addresses is used on reselection by + ** the SCRIPT. + **---------------------------------------------------------------- + */ + u_int32 *luntbl; /* lcbs bus address table */ + u_int32 b_luntbl; /* bus address of this table */ + u_int32 b_lun0; /* bus address of lun0 */ + lcb_p l0p; /* lcb of LUN #0 (normal case) */ +#if MAX_LUN > 1 + lcb_p *lmp; /* Other lcb's [1..MAX_LUN] */ +#endif + /*---------------------------------------------------------------- + ** Target capabilities. + **---------------------------------------------------------------- + */ + u_char inq_done; /* Target capabilities received */ + u_char inq_byte7; /* Contains these capabilities */ + + /*---------------------------------------------------------------- + ** Some flags. + **---------------------------------------------------------------- + */ + u_char to_reset; /* This target is to be reset */ + + /*---------------------------------------------------------------- + ** Pointer to the ccb used for negotiation. + ** Prevent from starting a negotiation for all queued commands + ** when tagged command queuing is enabled. + **---------------------------------------------------------------- + */ + ccb_p nego_cp; + + /*---------------------------------------------------------------- + ** negotiation of wide and synch transfer and device quirks. + ** sval, wval and uval are read from SCRIPTS and so have alignment + ** constraints. + **---------------------------------------------------------------- + */ +/*0*/ u_char minsync; +/*1*/ u_char sval; +/*2*/ u_short period; +/*0*/ u_char maxoffs; +/*1*/ u_char quirks; +/*2*/ u_char widedone; +/*3*/ u_char wval; +/*0*/ u_char uval; + +#ifdef SCSI_NCR_INTEGRITY_CHECKING + u_char ic_min_sync; + u_char ic_max_width; + u_char ic_done; +#endif + u_char ic_maximums_set; + u_char ppr_negotiation; + + /*---------------------------------------------------------------- + ** User settable limits and options. + ** These limits are read from the NVRAM if present. + **---------------------------------------------------------------- + */ + u_char usrsync; + u_char usrwide; + u_short usrtags; + u_char usrflag; +}; + +/*======================================================================== +** +** Declaration of structs: lun control block +** +**======================================================================== +*/ +struct lcb { + /*---------------------------------------------------------------- + ** On reselection, SCRIPTS use this value as a JUMP address + ** after the IDENTIFY has been successfully received. + ** This field is set to 'resel_tag' if TCQ is enabled and + ** to 'resel_notag' if TCQ is disabled. + ** (Must be at zero due to bad lun handling on reselection) + **---------------------------------------------------------------- + */ +/*0*/ u_int32 resel_task; + + /*---------------------------------------------------------------- + ** Task table used by the script processor to retrieve the + ** task corresponding to a reselected nexus. The TAG is used + ** as offset to determine the corresponding entry. + ** Each entry contains the associated CCB bus address. + **---------------------------------------------------------------- + */ + u_int32 tasktbl_0; /* Used if TCQ not enabled */ + u_int32 *tasktbl; + u_int32 b_tasktbl; + + /*---------------------------------------------------------------- + ** CCB queue management. + **---------------------------------------------------------------- + */ + XPT_QUEHEAD busy_ccbq; /* Queue of busy CCBs */ + XPT_QUEHEAD wait_ccbq; /* Queue of waiting for IO CCBs */ + u_short busyccbs; /* CCBs busy for this lun */ + u_short queuedccbs; /* CCBs queued to the controller*/ + u_short queuedepth; /* Queue depth for this lun */ + u_short scdev_depth; /* SCSI device queue depth */ + u_short maxnxs; /* Max possible nexuses */ + + /*---------------------------------------------------------------- + ** Control of tagged command queuing. + ** Tags allocation is performed using a circular buffer. + ** This avoids using a loop for tag allocation. + **---------------------------------------------------------------- + */ + u_short ia_tag; /* Tag allocation index */ + u_short if_tag; /* Tag release index */ + u_char *cb_tags; /* Circular tags buffer */ + u_char inq_byte7; /* Store unit CmdQ capability */ + u_char usetags; /* Command queuing is active */ + u_char to_clear; /* User wants to clear all tasks*/ + u_short maxtags; /* Max NR of tags asked by user */ + u_short numtags; /* Current number of tags */ + + /*---------------------------------------------------------------- + ** QUEUE FULL and ORDERED tag control. + **---------------------------------------------------------------- + */ + u_short num_good; /* Nr of GOOD since QUEUE FULL */ + u_short tags_sum[2]; /* Tags sum counters */ + u_char tags_si; /* Current index to tags sum */ + u_long tags_stime; /* Last time we switch tags_sum */ +}; + +/*======================================================================== +** +** Declaration of structs: actions for a task. +** +**======================================================================== +** +** It is part of the CCB and is called by the scripts processor to +** start or restart the data structure (nexus). +** +**------------------------------------------------------------------------ +*/ +struct action { + u_int32 start; + u_int32 restart; +}; + +/*======================================================================== +** +** Declaration of structs: Phase mismatch context. +** +**======================================================================== +** +** It is part of the CCB and is used as parameters for the DATA +** pointer. We need two contexts to handle correctly the SAVED +** DATA POINTER. +** +**------------------------------------------------------------------------ +*/ +struct pm_ctx { + struct scr_tblmove sg; /* Updated interrupted SG block */ + u_int32 ret; /* SCRIPT return address */ +}; + +/*======================================================================== +** +** Declaration of structs: global HEADER. +** +**======================================================================== +** +** In earlier driver versions, this substructure was copied from the +** ccb to a global address after selection (or reselection) and copied +** back before disconnect. Since we are now using LOAD/STORE DSA +** RELATIVE instructions, the script is able to access directly these +** fields, and so, this header is no more copied. +** +**------------------------------------------------------------------------ +*/ + +struct head { + /*---------------------------------------------------------------- + ** Start and restart SCRIPTS addresses (must be at 0). + **---------------------------------------------------------------- + */ + struct action go; + + /*---------------------------------------------------------------- + ** Saved data pointer. + ** Points to the position in the script responsible for the + ** actual transfer of data. + ** It's written after reception of a SAVE_DATA_POINTER message. + ** The goalpointer points after the last transfer command. + **---------------------------------------------------------------- + */ + u_int32 savep; + u_int32 lastp; + u_int32 goalp; + + /*---------------------------------------------------------------- + ** Alternate data pointer. + ** They are copied back to savep/lastp/goalp by the SCRIPTS + ** when the direction is unknown and the device claims data out. + **---------------------------------------------------------------- + */ + u_int32 wlastp; + u_int32 wgoalp; + + /*---------------------------------------------------------------- + ** Status fields. + **---------------------------------------------------------------- + */ + u_char status[4]; /* host status */ +}; + +/* +** LUN control block lookup. +** We use a direct pointer for LUN #0, and a table of pointers +** which is only allocated for devices that support LUN(s) > 0. +*/ +#if MAX_LUN <= 1 +#define ncr_lp(np, tp, lun) (!lun) ? (tp)->l0p : 0 +#else +#define ncr_lp(np, tp, lun) \ + (!lun) ? (tp)->l0p : (tp)->lmp ? (tp)->lmp[(lun)] : 0 +#endif + +/* +** The status bytes are used by the host and the script processor. +** +** The four bytes (status[4]) are copied to the scratchb register +** (declared as scr0..scr3 in ncr_reg.h) just after the select/reselect, +** and copied back just after disconnecting. +** Inside the script the XX_REG are used. +*/ + +/* +** Last four bytes (script) +*/ +#define QU_REG scr0 +#define HS_REG scr1 +#define HS_PRT nc_scr1 +#define SS_REG scr2 +#define SS_PRT nc_scr2 +#define HF_REG scr3 +#define HF_PRT nc_scr3 + +/* +** Last four bytes (host) +*/ +#define actualquirks phys.header.status[0] +#define host_status phys.header.status[1] +#define scsi_status phys.header.status[2] +#define host_flags phys.header.status[3] + +/* +** Host flags +*/ +#define HF_IN_PM0 1u +#define HF_IN_PM1 (1u<<1) +#define HF_ACT_PM (1u<<2) +#define HF_DP_SAVED (1u<<3) +#define HF_AUTO_SENSE (1u<<4) +#define HF_DATA_IN (1u<<5) +#define HF_PM_TO_C (1u<<6) +#define HF_EXT_ERR (1u<<7) + +#ifdef SCSI_NCR_IARB_SUPPORT +#define HF_HINT_IARB (1u<<7) +#endif + +/* +** This one is stolen from QU_REG.:) +*/ +#define HF_DATA_ST (1u<<7) + +/*========================================================== +** +** Declaration of structs: Data structure block +** +**========================================================== +** +** During execution of a ccb by the script processor, +** the DSA (data structure address) register points +** to this substructure of the ccb. +** This substructure contains the header with +** the script-processor-changable data and +** data blocks for the indirect move commands. +** +**---------------------------------------------------------- +*/ + +struct dsb { + + /* + ** Header. + */ + + struct head header; + + /* + ** Table data for Script + */ + + struct scr_tblsel select; + struct scr_tblmove smsg ; + struct scr_tblmove smsg_ext ; + struct scr_tblmove cmd ; + struct scr_tblmove sense ; + struct scr_tblmove wresid; + struct scr_tblmove data [MAX_SCATTER]; + + /* + ** Phase mismatch contexts. + ** We need two to handle correctly the + ** SAVED DATA POINTER. + */ + + struct pm_ctx pm0; + struct pm_ctx pm1; +}; + + +/*======================================================================== +** +** Declaration of structs: Command control block. +** +**======================================================================== +*/ +struct ccb { + /*---------------------------------------------------------------- + ** This is the data structure which is pointed by the DSA + ** register when it is executed by the script processor. + ** It must be the first entry. + **---------------------------------------------------------------- + */ + struct dsb phys; + + /*---------------------------------------------------------------- + ** The general SCSI driver provides a + ** pointer to a control block. + **---------------------------------------------------------------- + */ + Scsi_Cmnd *cmd; /* SCSI command */ + u_char cdb_buf[16]; /* Copy of CDB */ + u_char sense_buf[64]; + int data_len; /* Total data length */ + int segments; /* Number of SG segments */ + + /*---------------------------------------------------------------- + ** Message areas. + ** We prepare a message to be sent after selection. + ** We may use a second one if the command is rescheduled + ** due to CHECK_CONDITION or QUEUE FULL status. + ** Contents are IDENTIFY and SIMPLE_TAG. + ** While negotiating sync or wide transfer, + ** a SDTR or WDTR message is appended. + **---------------------------------------------------------------- + */ + u_char scsi_smsg [12]; + u_char scsi_smsg2[12]; + + /*---------------------------------------------------------------- + ** Miscellaneous status'. + **---------------------------------------------------------------- + */ + u_char nego_status; /* Negotiation status */ + u_char xerr_status; /* Extended error flags */ + u_int32 extra_bytes; /* Extraneous bytes transferred */ + + /*---------------------------------------------------------------- + ** Saved info for auto-sense + **---------------------------------------------------------------- + */ + u_char sv_scsi_status; + u_char sv_xerr_status; + + /*---------------------------------------------------------------- + ** Other fields. + **---------------------------------------------------------------- + */ + u_long p_ccb; /* BUS address of this CCB */ + u_char sensecmd[6]; /* Sense command */ + u_char to_abort; /* This CCB is to be aborted */ + u_short tag; /* Tag for this transfer */ + /* NO_TAG means no tag */ + u_char tags_si; /* Lun tags sum index (0,1) */ + + u_char target; + u_char lun; + u_short queued; + ccb_p link_ccb; /* Host adapter CCB chain */ + ccb_p link_ccbh; /* Host adapter CCB hash chain */ + XPT_QUEHEAD link_ccbq; /* Link to unit CCB queue */ + u_int32 startp; /* Initial data pointer */ + u_int32 lastp0; /* Initial 'lastp' */ + int ext_sg; /* Extreme data pointer, used */ + int ext_ofs; /* to calculate the residual. */ + int resid; +}; + +#define CCB_PHYS(cp,lbl) (cp->p_ccb + offsetof(struct ccb, lbl)) + + +/*======================================================================== +** +** Declaration of structs: NCR device descriptor +** +**======================================================================== +*/ +struct ncb { + /*---------------------------------------------------------------- + ** Idle task and invalid task actions and their bus + ** addresses. + **---------------------------------------------------------------- + */ + struct action idletask; + struct action notask; + struct action bad_i_t_l; + struct action bad_i_t_l_q; + u_long p_idletask; + u_long p_notask; + u_long p_bad_i_t_l; + u_long p_bad_i_t_l_q; + + /*---------------------------------------------------------------- + ** Dummy lun table to protect us against target returning bad + ** lun number on reselection. + **---------------------------------------------------------------- + */ + u_int32 *badluntbl; /* Table physical address */ + u_int32 resel_badlun; /* SCRIPT handler BUS address */ + + /*---------------------------------------------------------------- + ** Bit 32-63 of the on-chip RAM bus address in LE format. + ** The START_RAM64 script loads the MMRS and MMWS from this + ** field. + **---------------------------------------------------------------- + */ + u_int32 scr_ram_seg; + + /*---------------------------------------------------------------- + ** CCBs management queues. + **---------------------------------------------------------------- + */ + Scsi_Cmnd *waiting_list; /* Commands waiting for a CCB */ + /* when lcb is not allocated. */ + Scsi_Cmnd *done_list; /* Commands waiting for done() */ + /* callback to be invoked. */ +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,93) + spinlock_t smp_lock; /* Lock for SMP threading */ +#endif + + /*---------------------------------------------------------------- + ** Chip and controller indentification. + **---------------------------------------------------------------- + */ + int unit; /* Unit number */ + char chip_name[8]; /* Chip name */ + char inst_name[16]; /* ncb instance name */ + + /*---------------------------------------------------------------- + ** Initial value of some IO register bits. + ** These values are assumed to have been set by BIOS, and may + ** be used for probing adapter implementation differences. + **---------------------------------------------------------------- + */ + u_char sv_scntl0, sv_scntl3, sv_dmode, sv_dcntl, sv_ctest3, sv_ctest4, + sv_ctest5, sv_gpcntl, sv_stest2, sv_stest4, sv_stest1, sv_scntl4; + + /*---------------------------------------------------------------- + ** Actual initial value of IO register bits used by the + ** driver. They are loaded at initialisation according to + ** features that are to be enabled. + **---------------------------------------------------------------- + */ + u_char rv_scntl0, rv_scntl3, rv_dmode, rv_dcntl, rv_ctest3, rv_ctest4, + rv_ctest5, rv_stest2, rv_ccntl0, rv_ccntl1, rv_scntl4; + + /*---------------------------------------------------------------- + ** Target data. + ** Target control block bus address array used by the SCRIPT + ** on reselection. + **---------------------------------------------------------------- + */ + struct tcb target[MAX_TARGET]; + u_int32 *targtbl; + + /*---------------------------------------------------------------- + ** Virtual and physical bus addresses of the chip. + **---------------------------------------------------------------- + */ +#ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + u_long base_va; /* MMIO base virtual address */ + u_long base2_va; /* On-chip RAM virtual address */ +#endif + u_long base_ba; /* MMIO base bus address */ + u_long base_io; /* IO space base address */ + u_long base_ws; /* (MM)IO window size */ + u_long base2_ba; /* On-chip RAM bus address */ + u_long base2_ws; /* On-chip RAM window size */ + u_int irq; /* IRQ number */ + volatile /* Pointer to volatile for */ + struct ncr_reg *reg; /* memory mapped IO. */ + + /*---------------------------------------------------------------- + ** SCRIPTS virtual and physical bus addresses. + ** 'script' is loaded in the on-chip RAM if present. + ** 'scripth' stays in main memory for all chips except the + ** 53C895A and 53C896 that provide 8K on-chip RAM. + **---------------------------------------------------------------- + */ + struct script *script0; /* Copies of script and scripth */ + struct scripth *scripth0; /* relocated for this ncb. */ + u_long p_script; /* Actual script and scripth */ + u_long p_scripth; /* bus addresses. */ + u_long p_scripth0; + + /*---------------------------------------------------------------- + ** General controller parameters and configuration. + **---------------------------------------------------------------- + */ + pcidev_t pdev; + u_short device_id; /* PCI device id */ + u_char revision_id; /* PCI device revision id */ + u_char bus; /* PCI BUS number */ + u_char device_fn; /* PCI BUS device and function */ + u_char myaddr; /* SCSI id of the adapter */ + u_char maxburst; /* log base 2 of dwords burst */ + u_char maxwide; /* Maximum transfer width */ + u_char minsync; /* Minimum sync period factor */ + u_char maxsync; /* Maximum sync period factor */ + u_char maxoffs; /* Max scsi offset */ + u_char multiplier; /* Clock multiplier (1,2,4) */ + u_char clock_divn; /* Number of clock divisors */ + u_long clock_khz; /* SCSI clock frequency in KHz */ + u_int features; /* Chip features map */ + + /*---------------------------------------------------------------- + ** Range for the PCI clock frequency measurement result + ** that ensures the algorithm used by the driver can be + ** trusted for the SCSI clock frequency measurement. + ** (Assuming a PCI clock frequency of 33 MHz). + **---------------------------------------------------------------- + */ + u_int pciclock_min; + u_int pciclock_max; + + /*---------------------------------------------------------------- + ** Start queue management. + ** It is filled up by the host processor and accessed by the + ** SCRIPTS processor in order to start SCSI commands. + **---------------------------------------------------------------- + */ + u_long p_squeue; /* Start queue BUS address */ + u_int32 *squeue; /* Start queue virtual address */ + u_short squeueput; /* Next free slot of the queue */ + u_short actccbs; /* Number of allocated CCBs */ + u_short queuedepth; /* Start queue depth */ + + /*---------------------------------------------------------------- + ** Command completion queue. + ** It is the same size as the start queue to avoid overflow. + **---------------------------------------------------------------- + */ + u_short dqueueget; /* Next position to scan */ + u_int32 *dqueue; /* Completion (done) queue */ + + /*---------------------------------------------------------------- + ** Timeout handler. + **---------------------------------------------------------------- + */ + struct timer_list timer; /* Timer handler link header */ + u_long lasttime; + u_long settle_time; /* Resetting the SCSI BUS */ + + /*---------------------------------------------------------------- + ** Debugging and profiling. + **---------------------------------------------------------------- + */ + struct ncr_reg regdump; /* Register dump */ + u_long regtime; /* Time it has been done */ + + /*---------------------------------------------------------------- + ** Miscellaneous buffers accessed by the scripts-processor. + ** They shall be DWORD aligned, because they may be read or + ** written with a script command. + **---------------------------------------------------------------- + */ + u_char msgout[12]; /* Buffer for MESSAGE OUT */ + u_char msgin [12]; /* Buffer for MESSAGE IN */ + u_int32 lastmsg; /* Last SCSI message sent */ + u_char scratch; /* Scratch for SCSI receive */ + + /*---------------------------------------------------------------- + ** Miscellaneous configuration and status parameters. + **---------------------------------------------------------------- + */ + u_char scsi_mode; /* Current SCSI BUS mode */ + u_char order; /* Tag order to use */ + u_char verbose; /* Verbosity for this controller*/ + u_int32 ncr_cache; /* Used for cache test at init. */ + u_long p_ncb; /* BUS address of this NCB */ + + /*---------------------------------------------------------------- + ** CCB lists and queue. + **---------------------------------------------------------------- + */ + ccb_p ccbh[CCB_HASH_SIZE]; /* CCB hashed by DSA value */ + struct ccb *ccbc; /* CCB chain */ + XPT_QUEHEAD free_ccbq; /* Queue of available CCBs */ + + /*---------------------------------------------------------------- + ** IMMEDIATE ARBITRATION (IARB) control. + ** We keep track in 'last_cp' of the last CCB that has been + ** queued to the SCRIPTS processor and clear 'last_cp' when + ** this CCB completes. If last_cp is not zero at the moment + ** we queue a new CCB, we set a flag in 'last_cp' that is + ** used by the SCRIPTS as a hint for setting IARB. + ** We donnot set more than 'iarb_max' consecutive hints for + ** IARB in order to leave devices a chance to reselect. + ** By the way, any non zero value of 'iarb_max' is unfair. :) + **---------------------------------------------------------------- + */ +#ifdef SCSI_NCR_IARB_SUPPORT + struct ccb *last_cp; /* Last queud CCB used for IARB */ + u_short iarb_max; /* Max. # consecutive IARB hints*/ + u_short iarb_count; /* Actual # of these hints */ +#endif + + /*---------------------------------------------------------------- + ** We need the LCB in order to handle disconnections and + ** to count active CCBs for task management. So, we use + ** a unique CCB for LUNs we donnot have the LCB yet. + ** This queue normally should have at most 1 element. + **---------------------------------------------------------------- + */ + XPT_QUEHEAD b0_ccbq; + + /*---------------------------------------------------------------- + ** We use a different scatter function for 896 rev 1. + **---------------------------------------------------------------- + */ + int (*scatter) (ncb_p, ccb_p, Scsi_Cmnd *); + + /*---------------------------------------------------------------- + ** Command abort handling. + ** We need to synchronize tightly with the SCRIPTS + ** processor in order to handle things correctly. + **---------------------------------------------------------------- + */ + u_char abrt_msg[4]; /* Message to send buffer */ + struct scr_tblmove abrt_tbl; /* Table for the MOV of it */ + struct scr_tblsel abrt_sel; /* Sync params for selection */ + u_char istat_sem; /* Tells the chip to stop (SEM) */ + + /*---------------------------------------------------------------- + ** Fields that should be removed or changed. + **---------------------------------------------------------------- + */ + struct usrcmd user; /* Command from user */ + volatile u_char release_stage; /* Synchronisation stage on release */ + + /*---------------------------------------------------------------- + ** Fields that are used (primarily) for integrity check + **---------------------------------------------------------------- + */ + unsigned char check_integrity; /* Enable midlayer integ. check on + * bus scan. */ +#ifdef SCSI_NCR_INTEGRITY_CHECKING + unsigned char check_integ_par; /* Set if par or Init. Det. error + * used only during integ check */ +#endif +}; + +#define NCB_PHYS(np, lbl) (np->p_ncb + offsetof(struct ncb, lbl)) +#define NCB_SCRIPT_PHYS(np,lbl) (np->p_script + offsetof (struct script, lbl)) +#define NCB_SCRIPTH_PHYS(np,lbl) (np->p_scripth + offsetof (struct scripth,lbl)) +#define NCB_SCRIPTH0_PHYS(np,lbl) (np->p_scripth0+offsetof (struct scripth,lbl)) + +/*========================================================== +** +** +** Script for NCR-Processor. +** +** Use ncr_script_fill() to create the variable parts. +** Use ncr_script_copy_and_bind() to make a copy and +** bind to physical addresses. +** +** +**========================================================== +** +** We have to know the offsets of all labels before +** we reach them (for forward jumps). +** Therefore we declare a struct here. +** If you make changes inside the script, +** DONT FORGET TO CHANGE THE LENGTHS HERE! +** +**---------------------------------------------------------- +*/ + +/* +** Script fragments which are loaded into the on-chip RAM +** of 825A, 875, 876, 895, 895A and 896 chips. +*/ +struct script { + ncrcmd start [ 14]; + ncrcmd getjob_begin [ 4]; + ncrcmd getjob_end [ 4]; + ncrcmd select [ 8]; + ncrcmd wf_sel_done [ 2]; + ncrcmd send_ident [ 2]; +#ifdef SCSI_NCR_IARB_SUPPORT + ncrcmd select2 [ 8]; +#else + ncrcmd select2 [ 2]; +#endif + ncrcmd command [ 2]; + ncrcmd dispatch [ 28]; + ncrcmd sel_no_cmd [ 10]; + ncrcmd init [ 6]; + ncrcmd clrack [ 4]; + ncrcmd disp_status [ 4]; + ncrcmd datai_done [ 26]; + ncrcmd datao_done [ 12]; + ncrcmd ign_i_w_r_msg [ 4]; + ncrcmd datai_phase [ 2]; + ncrcmd datao_phase [ 4]; + ncrcmd msg_in [ 2]; + ncrcmd msg_in2 [ 10]; +#ifdef SCSI_NCR_IARB_SUPPORT + ncrcmd status [ 14]; +#else + ncrcmd status [ 10]; +#endif + ncrcmd complete [ 8]; +#ifdef SCSI_NCR_PCIQ_MAY_REORDER_WRITES + ncrcmd complete2 [ 12]; +#else + ncrcmd complete2 [ 10]; +#endif +#ifdef SCSI_NCR_PCIQ_SYNC_ON_INTR + ncrcmd done [ 18]; +#else + ncrcmd done [ 14]; +#endif + ncrcmd done_end [ 2]; + ncrcmd save_dp [ 8]; + ncrcmd restore_dp [ 4]; + ncrcmd disconnect [ 20]; +#ifdef SCSI_NCR_IARB_SUPPORT + ncrcmd idle [ 4]; +#else + ncrcmd idle [ 2]; +#endif +#ifdef SCSI_NCR_IARB_SUPPORT + ncrcmd ungetjob [ 6]; +#else + ncrcmd ungetjob [ 4]; +#endif + ncrcmd reselect [ 4]; + ncrcmd reselected [ 20]; + ncrcmd resel_scntl4 [ 30]; +#if MAX_TASKS*4 > 512 + ncrcmd resel_tag [ 18]; +#elif MAX_TASKS*4 > 256 + ncrcmd resel_tag [ 12]; +#else + ncrcmd resel_tag [ 8]; +#endif + ncrcmd resel_go [ 6]; + ncrcmd resel_notag [ 2]; + ncrcmd resel_dsa [ 8]; + ncrcmd data_in [MAX_SCATTER * SCR_SG_SIZE]; + ncrcmd data_in2 [ 4]; + ncrcmd data_out [MAX_SCATTER * SCR_SG_SIZE]; + ncrcmd data_out2 [ 4]; + ncrcmd pm0_data [ 12]; + ncrcmd pm0_data_out [ 6]; + ncrcmd pm0_data_end [ 6]; + ncrcmd pm1_data [ 12]; + ncrcmd pm1_data_out [ 6]; + ncrcmd pm1_data_end [ 6]; +}; + +/* +** Script fragments which stay in main memory for all chips +** except for the 895A and 896 that support 8K on-chip RAM. +*/ +struct scripth { + ncrcmd start64 [ 2]; + ncrcmd no_data [ 2]; + ncrcmd sel_for_abort [ 18]; + ncrcmd sel_for_abort_1 [ 2]; + ncrcmd select_no_atn [ 8]; + ncrcmd wf_sel_done_no_atn [ 4]; + + ncrcmd msg_in_etc [ 14]; + ncrcmd msg_received [ 4]; + ncrcmd msg_weird_seen [ 4]; + ncrcmd msg_extended [ 20]; + ncrcmd msg_bad [ 6]; + ncrcmd msg_weird [ 4]; + ncrcmd msg_weird1 [ 8]; + + ncrcmd wdtr_resp [ 6]; + ncrcmd send_wdtr [ 4]; + ncrcmd sdtr_resp [ 6]; + ncrcmd send_sdtr [ 4]; + ncrcmd ppr_resp [ 6]; + ncrcmd send_ppr [ 4]; + ncrcmd nego_bad_phase [ 4]; + ncrcmd msg_out [ 4]; + ncrcmd msg_out_done [ 4]; + ncrcmd data_ovrun [ 2]; + ncrcmd data_ovrun1 [ 22]; + ncrcmd data_ovrun2 [ 8]; + ncrcmd abort_resel [ 16]; + ncrcmd resend_ident [ 4]; + ncrcmd ident_break [ 4]; + ncrcmd ident_break_atn [ 4]; + ncrcmd sdata_in [ 6]; + ncrcmd data_io [ 2]; + ncrcmd data_io_com [ 8]; + ncrcmd data_io_out [ 12]; + ncrcmd resel_bad_lun [ 4]; + ncrcmd bad_i_t_l [ 4]; + ncrcmd bad_i_t_l_q [ 4]; + ncrcmd bad_status [ 6]; + ncrcmd tweak_pmj [ 12]; + ncrcmd pm_handle [ 20]; + ncrcmd pm_handle1 [ 4]; + ncrcmd pm_save [ 4]; + ncrcmd pm0_save [ 14]; + ncrcmd pm1_save [ 14]; + + /* WSR handling */ +#ifdef SYM_DEBUG_PM_WITH_WSR + ncrcmd pm_wsr_handle [ 44]; +#else + ncrcmd pm_wsr_handle [ 42]; +#endif + ncrcmd wsr_ma_helper [ 4]; + + /* Data area */ + ncrcmd zero [ 1]; + ncrcmd scratch [ 1]; + ncrcmd scratch1 [ 1]; + ncrcmd pm0_data_addr [ 1]; + ncrcmd pm1_data_addr [ 1]; + ncrcmd saved_dsa [ 1]; + ncrcmd saved_drs [ 1]; + ncrcmd done_pos [ 1]; + ncrcmd startpos [ 1]; + ncrcmd targtbl [ 1]; + /* End of data area */ + +#ifdef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + ncrcmd start_ram [ 1]; + ncrcmd script0_ba [ 4]; + ncrcmd start_ram64 [ 3]; + ncrcmd script0_ba64 [ 3]; + ncrcmd scripth0_ba64 [ 6]; + ncrcmd ram_seg64 [ 1]; +#endif + ncrcmd snooptest [ 6]; + ncrcmd snoopend [ 2]; +}; + +/*========================================================== +** +** +** Function headers. +** +** +**========================================================== +*/ + +static ccb_p ncr_alloc_ccb (ncb_p np); +static void ncr_complete (ncb_p np, ccb_p cp); +static void ncr_exception (ncb_p np); +static void ncr_free_ccb (ncb_p np, ccb_p cp); +static ccb_p ncr_ccb_from_dsa(ncb_p np, u_long dsa); +static void ncr_init_tcb (ncb_p np, u_char tn); +static lcb_p ncr_alloc_lcb (ncb_p np, u_char tn, u_char ln); +static lcb_p ncr_setup_lcb (ncb_p np, u_char tn, u_char ln, + u_char *inq_data); +static void ncr_getclock (ncb_p np, int mult); +static u_int ncr_getpciclock (ncb_p np); +static void ncr_selectclock (ncb_p np, u_char scntl3); +static ccb_p ncr_get_ccb (ncb_p np, u_char tn, u_char ln); +static void ncr_init (ncb_p np, int reset, char * msg, u_long code); +static void ncr_int_sbmc (ncb_p np); +static void ncr_int_par (ncb_p np, u_short sist); +static void ncr_int_ma (ncb_p np); +static void ncr_int_sir (ncb_p np); +static void ncr_int_sto (ncb_p np); +static void ncr_int_udc (ncb_p np); +static void ncr_negotiate (ncb_p np, tcb_p tp); +static int ncr_prepare_nego(ncb_p np, ccb_p cp, u_char *msgptr); +#ifdef SCSI_NCR_INTEGRITY_CHECKING +static int ncr_ic_nego(ncb_p np, ccb_p cp, Scsi_Cmnd *cmd, u_char *msgptr); +#endif +static void ncr_script_copy_and_bind + (ncb_p np, ncrcmd *src, ncrcmd *dst, int len); +static void ncr_script_fill (struct script * scr, struct scripth * scripth); +static int ncr_scatter_896R1 (ncb_p np, ccb_p cp, Scsi_Cmnd *cmd); +static int ncr_scatter (ncb_p np, ccb_p cp, Scsi_Cmnd *cmd); +static void ncr_getsync (ncb_p np, u_char sfac, u_char *fakp, u_char *scntl3p); +static void ncr_get_xfer_info(ncb_p np, tcb_p tp, u_char *factor, u_char *offset, u_char *width); +static void ncr_setsync (ncb_p np, ccb_p cp, u_char scntl3, u_char sxfer, u_char scntl4); +static void ncr_set_sync_wide_status (ncb_p np, u_char target); +static void ncr_setup_tags (ncb_p np, u_char tn, u_char ln); +static void ncr_setwide (ncb_p np, ccb_p cp, u_char wide, u_char ack); +static void ncr_setsyncwide (ncb_p np, ccb_p cp, u_char scntl3, u_char sxfer, u_char scntl4, u_char wide); +static int ncr_show_msg (u_char * msg); +static void ncr_print_msg (ccb_p cp, char *label, u_char * msg); +static int ncr_snooptest (ncb_p np); +static void ncr_timeout (ncb_p np); +static void ncr_wakeup (ncb_p np, u_long code); +static int ncr_wakeup_done (ncb_p np); +static void ncr_start_next_ccb (ncb_p np, lcb_p lp, int maxn); +static void ncr_put_start_queue(ncb_p np, ccb_p cp); +static void ncr_chip_reset (ncb_p np); +static void ncr_soft_reset (ncb_p np); +static void ncr_start_reset (ncb_p np); +static int ncr_reset_scsi_bus (ncb_p np, int enab_int, int settle_delay); +static int ncr_compute_residual (ncb_p np, ccb_p cp); + +#ifdef SCSI_NCR_USER_COMMAND_SUPPORT +static void ncr_usercmd (ncb_p np); +#endif + +static int ncr_attach (Scsi_Host_Template *tpnt, int unit, ncr_device *device); +static void ncr_free_resources(ncb_p np); + +static void insert_into_waiting_list(ncb_p np, Scsi_Cmnd *cmd); +static Scsi_Cmnd *retrieve_from_waiting_list(int to_remove, ncb_p np, Scsi_Cmnd *cmd); +static void process_waiting_list(ncb_p np, int sts); + +#define remove_from_waiting_list(np, cmd) \ + retrieve_from_waiting_list(1, (np), (cmd)) +#define requeue_waiting_list(np) process_waiting_list((np), DID_OK) +#define reset_waiting_list(np) process_waiting_list((np), DID_RESET) + +#ifdef SCSI_NCR_NVRAM_SUPPORT +static void ncr_get_nvram (ncr_device *devp, ncr_nvram *nvp); +static int sym_read_Tekram_nvram (ncr_slot *np, u_short device_id, + Tekram_nvram *nvram); +static int sym_read_Symbios_nvram (ncr_slot *np, Symbios_nvram *nvram); +#endif + +/*========================================================== +** +** +** Global static data. +** +** +**========================================================== +*/ + +static inline char *ncr_name (ncb_p np) +{ + return np->inst_name; +} + + +/*========================================================== +** +** +** Scripts for NCR-Processor. +** +** Use ncr_script_bind for binding to physical addresses. +** +** +**========================================================== +** +** NADDR generates a reference to a field of the controller data. +** PADDR generates a reference to another part of the script. +** RADDR generates a reference to a script processor register. +** FADDR generates a reference to a script processor register +** with offset. +** +**---------------------------------------------------------- +*/ + +#define RELOC_SOFTC 0x40000000 +#define RELOC_LABEL 0x50000000 +#define RELOC_REGISTER 0x60000000 +#if 0 +#define RELOC_KVAR 0x70000000 +#endif +#define RELOC_LABELH 0x80000000 +#define RELOC_MASK 0xf0000000 + +#define NADDR(label) (RELOC_SOFTC | offsetof(struct ncb, label)) +#define PADDR(label) (RELOC_LABEL | offsetof(struct script, label)) +#define PADDRH(label) (RELOC_LABELH | offsetof(struct scripth, label)) +#define RADDR(label) (RELOC_REGISTER | REG(label)) +#define FADDR(label,ofs)(RELOC_REGISTER | ((REG(label))+(ofs))) +#define KVAR(which) (RELOC_KVAR | (which)) + +#define SCR_DATA_ZERO 0xf00ff00f + +#ifdef RELOC_KVAR +#define SCRIPT_KVAR_JIFFIES (0) +#define SCRIPT_KVAR_FIRST SCRIPT_KVAR_JIFFIES +#define SCRIPT_KVAR_LAST SCRIPT_KVAR_JIFFIES +/* + * Kernel variables referenced in the scripts. + * THESE MUST ALL BE ALIGNED TO A 4-BYTE BOUNDARY. + */ +static void *script_kvars[] __initdata = + { (void *)&jiffies }; +#endif + +static struct script script0 __initdata = { +/*--------------------------< START >-----------------------*/ { + /* + ** This NOP will be patched with LED ON + ** SCR_REG_REG (gpreg, SCR_AND, 0xfe) + */ + SCR_NO_OP, + 0, + /* + ** Clear SIGP. + */ + SCR_FROM_REG (ctest2), + 0, + + /* + ** Stop here if the C code wants to perform + ** some error recovery procedure manually. + ** (Indicate this by setting SEM in ISTAT) + */ + SCR_FROM_REG (istat), + 0, + /* + ** Report to the C code the next position in + ** the start queue the SCRIPTS will schedule. + ** The C code must not change SCRATCHA. + */ + SCR_LOAD_ABS (scratcha, 4), + PADDRH (startpos), + SCR_INT ^ IFTRUE (MASK (SEM, SEM)), + SIR_SCRIPT_STOPPED, + + /* + ** Start the next job. + ** + ** @DSA = start point for this job. + ** SCRATCHA = address of this job in the start queue. + ** + ** We will restore startpos with SCRATCHA if we fails the + ** arbitration or if it is the idle job. + ** + ** The below GETJOB_BEGIN to GETJOB_END section of SCRIPTS + ** is a critical path. If it is partially executed, it then + ** may happen that the job address is not yet in the DSA + ** and the the next queue position points to the next JOB. + */ + SCR_LOAD_ABS (dsa, 4), + PADDRH (startpos), + SCR_LOAD_REL (temp, 4), + 4, +}/*-------------------------< GETJOB_BEGIN >------------------*/,{ + SCR_STORE_ABS (temp, 4), + PADDRH (startpos), + SCR_LOAD_REL (dsa, 4), + 0, +}/*-------------------------< GETJOB_END >--------------------*/,{ + SCR_LOAD_REL (temp, 4), + 0, + SCR_RETURN, + 0, + +}/*-------------------------< SELECT >----------------------*/,{ + /* + ** DSA contains the address of a scheduled + ** data structure. + ** + ** SCRATCHA contains the address of the start queue + ** entry which points to the next job. + ** + ** Set Initiator mode. + ** + ** (Target mode is left as an exercise for the reader) + */ + + SCR_CLR (SCR_TRG), + 0, + /* + ** And try to select this target. + */ + SCR_SEL_TBL_ATN ^ offsetof (struct dsb, select), + PADDR (ungetjob), + /* + ** Now there are 4 possibilities: + ** + ** (1) The ncr looses arbitration. + ** This is ok, because it will try again, + ** when the bus becomes idle. + ** (But beware of the timeout function!) + ** + ** (2) The ncr is reselected. + ** Then the script processor takes the jump + ** to the RESELECT label. + ** + ** (3) The ncr wins arbitration. + ** Then it will execute SCRIPTS instruction until + ** the next instruction that checks SCSI phase. + ** Then will stop and wait for selection to be + ** complete or selection time-out to occur. + ** + ** After having won arbitration, the ncr SCRIPTS + ** processor is able to execute instructions while + ** the SCSI core is performing SCSI selection. But + ** some script instruction that is not waiting for + ** a valid phase (or selection timeout) to occur + ** breaks the selection procedure, by probably + ** affecting timing requirements. + ** So we have to wait immediately for the next phase + ** or the selection to complete or time-out. + */ + + /* + ** load the savep (saved pointer) into + ** the actual data pointer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + /* + ** Initialize the status registers + */ + SCR_LOAD_REL (scr0, 4), + offsetof (struct ccb, phys.header.status), + +}/*-------------------------< WF_SEL_DONE >----------------------*/,{ + SCR_INT ^ IFFALSE (WHEN (SCR_MSG_OUT)), + SIR_SEL_ATN_NO_MSG_OUT, +}/*-------------------------< SEND_IDENT >----------------------*/,{ + /* + ** Selection complete. + ** Send the IDENTIFY and SIMPLE_TAG messages + ** (and the M_X_SYNC_REQ / M_X_WIDE_REQ message) + */ + SCR_MOVE_TBL ^ SCR_MSG_OUT, + offsetof (struct dsb, smsg), +}/*-------------------------< SELECT2 >----------------------*/,{ +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** Set IMMEDIATE ARBITRATION if we have been given + ** a hint to do so. (Some job to do after this one). + */ + SCR_FROM_REG (HF_REG), + 0, + SCR_JUMPR ^ IFFALSE (MASK (HF_HINT_IARB, HF_HINT_IARB)), + 8, + SCR_REG_REG (scntl1, SCR_OR, IARB), + 0, +#endif + /* + ** Anticipate the COMMAND phase. + ** This is the PHASE we expect at this point. + */ + SCR_JUMP ^ IFFALSE (WHEN (SCR_COMMAND)), + PADDR (sel_no_cmd), + +}/*-------------------------< COMMAND >--------------------*/,{ + /* + ** ... and send the command + */ + SCR_MOVE_TBL ^ SCR_COMMAND, + offsetof (struct dsb, cmd), + +}/*-----------------------< DISPATCH >----------------------*/,{ + /* + ** MSG_IN is the only phase that shall be + ** entered at least once for each (re)selection. + ** So we test it first. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_MSG_IN)), + PADDR (msg_in), + SCR_JUMP ^ IFTRUE (IF (SCR_DATA_OUT)), + PADDR (datao_phase), + SCR_JUMP ^ IFTRUE (IF (SCR_DATA_IN)), + PADDR (datai_phase), + SCR_JUMP ^ IFTRUE (IF (SCR_STATUS)), + PADDR (status), + SCR_JUMP ^ IFTRUE (IF (SCR_COMMAND)), + PADDR (command), + SCR_JUMP ^ IFTRUE (IF (SCR_MSG_OUT)), + PADDRH (msg_out), + /* + * Discard as many illegal phases as + * required and tell the C code about. + */ + SCR_JUMPR ^ IFFALSE (WHEN (SCR_ILG_OUT)), + 16, + SCR_MOVE_ABS (1) ^ SCR_ILG_OUT, + NADDR (scratch), + SCR_JUMPR ^ IFTRUE (WHEN (SCR_ILG_OUT)), + -16, + SCR_JUMPR ^ IFFALSE (WHEN (SCR_ILG_IN)), + 16, + SCR_MOVE_ABS (1) ^ SCR_ILG_IN, + NADDR (scratch), + SCR_JUMPR ^ IFTRUE (WHEN (SCR_ILG_IN)), + -16, + SCR_INT, + SIR_BAD_PHASE, + SCR_JUMP, + PADDR (dispatch), +}/*---------------------< SEL_NO_CMD >----------------------*/,{ + /* + ** The target does not switch to command + ** phase after IDENTIFY has been sent. + ** + ** If it stays in MSG OUT phase send it + ** the IDENTIFY again. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_MSG_OUT)), + PADDRH (resend_ident), + /* + ** If target does not switch to MSG IN phase + ** and we sent a negotiation, assert the + ** failure immediately. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_MSG_IN)), + PADDR (dispatch), + SCR_FROM_REG (HS_REG), + 0, + SCR_INT ^ IFTRUE (DATA (HS_NEGOTIATE)), + SIR_NEGO_FAILED, + /* + ** Jump to dispatcher. + */ + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< INIT >------------------------*/,{ + /* + ** Wait for the SCSI RESET signal to be + ** inactive before restarting operations, + ** since the chip may hang on SEL_ATN + ** if SCSI RESET is active. + */ + SCR_FROM_REG (sstat0), + 0, + SCR_JUMPR ^ IFTRUE (MASK (IRST, IRST)), + -16, + SCR_JUMP, + PADDR (start), +}/*-------------------------< CLRACK >----------------------*/,{ + /* + ** Terminate possible pending message phase. + */ + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< DISP_STATUS >----------------------*/,{ + /* + ** Anticipate STATUS phase. + ** + ** Does spare 3 SCRIPTS instructions when we have + ** completed the INPUT of the data. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_STATUS)), + PADDR (status), + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< DATAI_DONE >-------------------*/,{ + /* + * If the device wants us to send more data, + * we must count the extra bytes. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_DATA_IN)), + PADDRH (data_ovrun), + /* + ** If the SWIDE is not full, jump to dispatcher. + ** We anticipate a STATUS phase. + ** If we get later an IGNORE WIDE RESIDUE, we + ** will alias it as a MODIFY DP (-1). + */ + SCR_FROM_REG (scntl2), + 0, + SCR_JUMP ^ IFFALSE (MASK (WSR, WSR)), + PADDR (disp_status), + /* + ** The SWIDE is full. + ** Clear this condition. + */ + SCR_REG_REG (scntl2, SCR_OR, WSR), + 0, + /* + * We are expecting an IGNORE RESIDUE message + * from the device, otherwise we are in data + * overrun condition. Check against MSG_IN phase. + */ + SCR_INT ^ IFFALSE (WHEN (SCR_MSG_IN)), + SIR_SWIDE_OVERRUN, + SCR_JUMP ^ IFFALSE (WHEN (SCR_MSG_IN)), + PADDR (disp_status), + /* + * We are in MSG_IN phase, + * Read the first byte of the message. + * If it is not an IGNORE RESIDUE message, + * signal overrun and jump to message + * processing. + */ + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin[0]), + SCR_INT ^ IFFALSE (DATA (M_IGN_RESIDUE)), + SIR_SWIDE_OVERRUN, + SCR_JUMP ^ IFFALSE (DATA (M_IGN_RESIDUE)), + PADDR (msg_in2), + + /* + * We got the message we expected. + * Read the 2nd byte, and jump to dispatcher. + */ + SCR_CLR (SCR_ACK), + 0, + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin[1]), + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP, + PADDR (disp_status), + +}/*-------------------------< DATAO_DONE >-------------------*/,{ + /* + * If the device wants us to send more data, + * we must count the extra bytes. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_DATA_OUT)), + PADDRH (data_ovrun), + /* + ** If the SODL is not full jump to dispatcher. + ** We anticipate a MSG IN phase or a STATUS phase. + */ + SCR_FROM_REG (scntl2), + 0, + SCR_JUMP ^ IFFALSE (MASK (WSS, WSS)), + PADDR (disp_status), + /* + ** The SODL is full, clear this condition. + */ + SCR_REG_REG (scntl2, SCR_OR, WSS), + 0, + /* + ** And signal a DATA UNDERRUN condition + ** to the C code. + */ + SCR_INT, + SIR_SODL_UNDERRUN, + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< IGN_I_W_R_MSG >--------------*/,{ + /* + ** We jump here from the phase mismatch interrupt, + ** When we have a SWIDE and the device has presented + ** a IGNORE WIDE RESIDUE message on the BUS. + ** We just have to throw away this message and then + ** to jump to dispatcher. + */ + SCR_MOVE_ABS (2) ^ SCR_MSG_IN, + NADDR (scratch), + /* + ** Clear ACK and jump to dispatcher. + */ + SCR_JUMP, + PADDR (clrack), + +}/*-------------------------< DATAI_PHASE >------------------*/,{ + SCR_RETURN, + 0, +}/*-------------------------< DATAO_PHASE >------------------*/,{ + /* + ** Patch for 53c1010_66 only - to allow A0 part + ** to operate properly in a 33MHz PCI bus. + ** + ** SCR_REG_REG(scntl4, SCR_OR, 0x0c), + ** 0, + */ + SCR_NO_OP, + 0, + SCR_RETURN, + 0, +}/*-------------------------< MSG_IN >--------------------*/,{ + /* + ** Get the first byte of the message. + ** + ** The script processor doesn't negate the + ** ACK signal after this transfer. + */ + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin[0]), +}/*-------------------------< MSG_IN2 >--------------------*/,{ + /* + ** Check first against 1 byte messages + ** that we handle from SCRIPTS. + */ + SCR_JUMP ^ IFTRUE (DATA (M_COMPLETE)), + PADDR (complete), + SCR_JUMP ^ IFTRUE (DATA (M_DISCONNECT)), + PADDR (disconnect), + SCR_JUMP ^ IFTRUE (DATA (M_SAVE_DP)), + PADDR (save_dp), + SCR_JUMP ^ IFTRUE (DATA (M_RESTORE_DP)), + PADDR (restore_dp), + /* + ** We handle all other messages from the + ** C code, so no need to waste on-chip RAM + ** for those ones. + */ + SCR_JUMP, + PADDRH (msg_in_etc), + +}/*-------------------------< STATUS >--------------------*/,{ + /* + ** get the status + */ + SCR_MOVE_ABS (1) ^ SCR_STATUS, + NADDR (scratch), +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** If STATUS is not GOOD, clear IMMEDIATE ARBITRATION, + ** since we may have to tamper the start queue from + ** the C code. + */ + SCR_JUMPR ^ IFTRUE (DATA (S_GOOD)), + 8, + SCR_REG_REG (scntl1, SCR_AND, ~IARB), + 0, +#endif + /* + ** save status to scsi_status. + ** mark as complete. + */ + SCR_TO_REG (SS_REG), + 0, + SCR_LOAD_REG (HS_REG, HS_COMPLETE), + 0, + /* + ** Anticipate the MESSAGE PHASE for + ** the TASK COMPLETE message. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_MSG_IN)), + PADDR (msg_in), + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< COMPLETE >-----------------*/,{ + /* + ** Complete message. + ** + ** Copy the data pointer to LASTP in header. + */ + SCR_STORE_REL (temp, 4), + offsetof (struct ccb, phys.header.lastp), + /* + ** When we terminate the cycle by clearing ACK, + ** the target may disconnect immediately. + ** + ** We don't want to be told of an + ** "unexpected disconnect", + ** so we disable this feature. + */ + SCR_REG_REG (scntl2, SCR_AND, 0x7f), + 0, + /* + ** Terminate cycle ... + */ + SCR_CLR (SCR_ACK|SCR_ATN), + 0, + /* + ** ... and wait for the disconnect. + */ + SCR_WAIT_DISC, + 0, +}/*-------------------------< COMPLETE2 >-----------------*/,{ + /* + ** Save host status to header. + */ + SCR_STORE_REL (scr0, 4), + offsetof (struct ccb, phys.header.status), + +#ifdef SCSI_NCR_PCIQ_MAY_REORDER_WRITES + /* + ** Some bridges may reorder DMA writes to memory. + ** We donnot want the CPU to deal with completions + ** without all the posted write having been flushed + ** to memory. This DUMMY READ should flush posted + ** buffers prior to the CPU having to deal with + ** completions. + */ + SCR_LOAD_REL (scr0, 4), /* DUMMY READ */ + offsetof (struct ccb, phys.header.status), +#endif + /* + ** If command resulted in not GOOD status, + ** call the C code if needed. + */ + SCR_FROM_REG (SS_REG), + 0, + SCR_CALL ^ IFFALSE (DATA (S_GOOD)), + PADDRH (bad_status), + + /* + ** If we performed an auto-sense, call + ** the C code to synchronyze task aborts + ** with UNIT ATTENTION conditions. + */ + SCR_FROM_REG (HF_REG), + 0, + SCR_INT ^ IFTRUE (MASK (HF_AUTO_SENSE, HF_AUTO_SENSE)), + SIR_AUTO_SENSE_DONE, + +}/*------------------------< DONE >-----------------*/,{ +#ifdef SCSI_NCR_PCIQ_SYNC_ON_INTR + /* + ** It seems that some bridges flush everything + ** when the INTR line is raised. For these ones, + ** we can just ensure that the INTR line will be + ** raised before each completion. So, if it happens + ** that we have been faster that the CPU, we just + ** have to synchronize with it. A dummy programmed + ** interrupt will do the trick. + ** Note that we overlap at most 1 IO with the CPU + ** in this situation and that the IRQ line must not + ** be shared. + */ + SCR_FROM_REG (istat), + 0, + SCR_INT ^ IFTRUE (MASK (INTF, INTF)), + SIR_DUMMY_INTERRUPT, +#endif + /* + ** Copy the DSA to the DONE QUEUE and + ** signal completion to the host. + ** If we are interrupted between DONE + ** and DONE_END, we must reset, otherwise + ** the completed CCB will be lost. + */ + SCR_STORE_ABS (dsa, 4), + PADDRH (saved_dsa), + SCR_LOAD_ABS (dsa, 4), + PADDRH (done_pos), + SCR_LOAD_ABS (scratcha, 4), + PADDRH (saved_dsa), + SCR_STORE_REL (scratcha, 4), + 0, + /* + ** The instruction below reads the DONE QUEUE next + ** free position from memory. + ** In addition it ensures that all PCI posted writes + ** are flushed and so the DSA value of the done + ** CCB is visible by the CPU before INTFLY is raised. + */ + SCR_LOAD_REL (temp, 4), + 4, + SCR_INT_FLY, + 0, + SCR_STORE_ABS (temp, 4), + PADDRH (done_pos), +}/*------------------------< DONE_END >-----------------*/,{ + SCR_JUMP, + PADDR (start), + +}/*-------------------------< SAVE_DP >------------------*/,{ + /* + ** Clear ACK immediately. + ** No need to delay it. + */ + SCR_CLR (SCR_ACK), + 0, + /* + ** Keep track we received a SAVE DP, so + ** we will switch to the other PM context + ** on the next PM since the DP may point + ** to the current PM context. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_DP_SAVED), + 0, + /* + ** SAVE_DP message: + ** Copy the data pointer to SAVEP in header. + */ + SCR_STORE_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + SCR_JUMP, + PADDR (dispatch), +}/*-------------------------< RESTORE_DP >---------------*/,{ + /* + ** RESTORE_DP message: + ** Copy SAVEP in header to actual data pointer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + SCR_JUMP, + PADDR (clrack), + +}/*-------------------------< DISCONNECT >---------------*/,{ + /* + ** DISCONNECTing ... + ** + ** disable the "unexpected disconnect" feature, + ** and remove the ACK signal. + */ + SCR_REG_REG (scntl2, SCR_AND, 0x7f), + 0, + SCR_CLR (SCR_ACK|SCR_ATN), + 0, + /* + ** Wait for the disconnect. + */ + SCR_WAIT_DISC, + 0, + /* + ** Status is: DISCONNECTED. + */ + SCR_LOAD_REG (HS_REG, HS_DISCONNECT), + 0, + /* + ** Save host status to header. + */ + SCR_STORE_REL (scr0, 4), + offsetof (struct ccb, phys.header.status), + /* + ** If QUIRK_AUTOSAVE is set, + ** do an "save pointer" operation. + */ + SCR_FROM_REG (QU_REG), + 0, + SCR_JUMP ^ IFFALSE (MASK (QUIRK_AUTOSAVE, QUIRK_AUTOSAVE)), + PADDR (start), + /* + ** like SAVE_DP message: + ** Remember we saved the data pointer. + ** Copy data pointer to SAVEP in header. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_DP_SAVED), + 0, + SCR_STORE_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + SCR_JUMP, + PADDR (start), + +}/*-------------------------< IDLE >------------------------*/,{ + /* + ** Nothing to do? + ** Wait for reselect. + ** This NOP will be patched with LED OFF + ** SCR_REG_REG (gpreg, SCR_OR, 0x01) + */ + SCR_NO_OP, + 0, +#ifdef SCSI_NCR_IARB_SUPPORT + SCR_JUMPR, + 8, +#endif +}/*-------------------------< UNGETJOB >-----------------*/,{ +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** Set IMMEDIATE ARBITRATION, for the next time. + ** This will give us better chance to win arbitration + ** for the job we just wanted to do. + */ + SCR_REG_REG (scntl1, SCR_OR, IARB), + 0, +#endif + /* + ** We are not able to restart the SCRIPTS if we are + ** interrupted and these instruction haven't been + ** all executed. BTW, this is very unlikely to + ** happen, but we check that from the C code. + */ + SCR_LOAD_REG (dsa, 0xff), + 0, + SCR_STORE_ABS (scratcha, 4), + PADDRH (startpos), +}/*-------------------------< RESELECT >--------------------*/,{ + /* + ** make the host status invalid. + */ + SCR_CLR (SCR_TRG), + 0, + /* + ** Sleep waiting for a reselection. + ** If SIGP is set, special treatment. + ** + ** Zu allem bereit .. + */ + SCR_WAIT_RESEL, + PADDR(start), +}/*-------------------------< RESELECTED >------------------*/,{ + /* + ** This NOP will be patched with LED ON + ** SCR_REG_REG (gpreg, SCR_AND, 0xfe) + */ + SCR_NO_OP, + 0, + /* + ** load the target id into the sdid + */ + SCR_REG_SFBR (ssid, SCR_AND, 0x8F), + 0, + SCR_TO_REG (sdid), + 0, + /* + ** load the target control block address + */ + SCR_LOAD_ABS (dsa, 4), + PADDRH (targtbl), + SCR_SFBR_REG (dsa, SCR_SHL, 0), + 0, + SCR_REG_REG (dsa, SCR_SHL, 0), + 0, + SCR_REG_REG (dsa, SCR_AND, 0x3c), + 0, + SCR_LOAD_REL (dsa, 4), + 0, + /* + ** Load the synchronous transfer registers. + */ + SCR_LOAD_REL (scntl3, 1), + offsetof(struct tcb, wval), + SCR_LOAD_REL (sxfer, 1), + offsetof(struct tcb, sval), +}/*-------------------------< RESEL_SCNTL4 >------------------*/,{ + /* + ** Write with uval value. Patch if device + ** does not support Ultra3. + ** + ** SCR_LOAD_REL (scntl4, 1), + ** offsetof(struct tcb, uval), + */ + + SCR_NO_OP, + 0, + /* + * We expect MESSAGE IN phase. + * If not, get help from the C code. + */ + SCR_INT ^ IFFALSE (WHEN (SCR_MSG_IN)), + SIR_RESEL_NO_MSG_IN, + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin), + + /* + * If IDENTIFY LUN #0, use a faster path + * to find the LCB structure. + */ + SCR_JUMPR ^ IFTRUE (MASK (0x80, 0xbf)), + 56, + /* + * If message isn't an IDENTIFY, + * tell the C code about. + */ + SCR_INT ^ IFFALSE (MASK (0x80, 0x80)), + SIR_RESEL_NO_IDENTIFY, + /* + * It is an IDENTIFY message, + * Load the LUN control block address. + */ + SCR_LOAD_REL (dsa, 4), + offsetof(struct tcb, b_luntbl), + SCR_SFBR_REG (dsa, SCR_SHL, 0), + 0, + SCR_REG_REG (dsa, SCR_SHL, 0), + 0, + SCR_REG_REG (dsa, SCR_AND, 0xfc), + 0, + SCR_LOAD_REL (dsa, 4), + 0, + SCR_JUMPR, + 8, + /* + ** LUN 0 special case (but usual one :)) + */ + SCR_LOAD_REL (dsa, 4), + offsetof(struct tcb, b_lun0), + + /* + ** Load the reselect task action for this LUN. + ** Load the tasks DSA array for this LUN. + ** Call the action. + */ + SCR_LOAD_REL (temp, 4), + offsetof(struct lcb, resel_task), + SCR_LOAD_REL (dsa, 4), + offsetof(struct lcb, b_tasktbl), + SCR_RETURN, + 0, +}/*-------------------------< RESEL_TAG >-------------------*/,{ + /* + ** ACK the IDENTIFY or TAG previously received + */ + + SCR_CLR (SCR_ACK), + 0, + /* + ** Read IDENTIFY + SIMPLE + TAG using a single MOVE. + ** Agressive optimization, is'nt it? + ** No need to test the SIMPLE TAG message, since the + ** driver only supports conformant devices for tags. ;-) + */ + SCR_MOVE_ABS (2) ^ SCR_MSG_IN, + NADDR (msgin), + /* + ** Read the TAG from the SIDL. + ** Still an aggressive optimization. ;-) + ** Compute the CCB indirect jump address which + ** is (#TAG*2 & 0xfc) due to tag numbering using + ** 1,3,5..MAXTAGS*2+1 actual values. + */ + SCR_REG_SFBR (sidl, SCR_SHL, 0), + 0, +#if MAX_TASKS*4 > 512 + SCR_JUMPR ^ IFFALSE (CARRYSET), + 8, + SCR_REG_REG (dsa1, SCR_OR, 2), + 0, + SCR_REG_REG (sfbr, SCR_SHL, 0), + 0, + SCR_JUMPR ^ IFFALSE (CARRYSET), + 8, + SCR_REG_REG (dsa1, SCR_OR, 1), + 0, +#elif MAX_TASKS*4 > 256 + SCR_JUMPR ^ IFFALSE (CARRYSET), + 8, + SCR_REG_REG (dsa1, SCR_OR, 1), + 0, +#endif + /* + ** Retrieve the DSA of this task. + ** JUMP indirectly to the restart point of the CCB. + */ + SCR_SFBR_REG (dsa, SCR_AND, 0xfc), + 0, +}/*-------------------------< RESEL_GO >-------------------*/,{ + SCR_LOAD_REL (dsa, 4), + 0, + SCR_LOAD_REL (temp, 4), + offsetof(struct ccb, phys.header.go.restart), + SCR_RETURN, + 0, + /* In normal situations we branch to RESEL_DSA */ +}/*-------------------------< RESEL_NOTAG >-------------------*/,{ + /* + ** JUMP indirectly to the restart point of the CCB. + */ + SCR_JUMP, + PADDR (resel_go), + +}/*-------------------------< RESEL_DSA >-------------------*/,{ + /* + ** Ack the IDENTIFY or TAG previously received. + */ + SCR_CLR (SCR_ACK), + 0, + /* + ** load the savep (saved pointer) into + ** the actual data pointer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + /* + ** Initialize the status registers + */ + SCR_LOAD_REL (scr0, 4), + offsetof (struct ccb, phys.header.status), + /* + ** Jump to dispatcher. + */ + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< DATA_IN >--------------------*/,{ +/* +** Because the size depends on the +** #define MAX_SCATTER parameter, +** it is filled in at runtime. +** +** ##===========< i=0; i<MAX_SCATTER >========= +** || SCR_CHMOV_TBL ^ SCR_DATA_IN, +** || offsetof (struct dsb, data[ i]), +** ##========================================== +** +**--------------------------------------------------------- +*/ +0 +}/*-------------------------< DATA_IN2 >-------------------*/,{ + SCR_CALL, + PADDR (datai_done), + SCR_JUMP, + PADDRH (data_ovrun), +}/*-------------------------< DATA_OUT >--------------------*/,{ +/* +** Because the size depends on the +** #define MAX_SCATTER parameter, +** it is filled in at runtime. +** +** ##===========< i=0; i<MAX_SCATTER >========= +** || SCR_CHMOV_TBL ^ SCR_DATA_OUT, +** || offsetof (struct dsb, data[ i]), +** ##========================================== +** +**--------------------------------------------------------- +*/ +0 +}/*-------------------------< DATA_OUT2 >-------------------*/,{ + SCR_CALL, + PADDR (datao_done), + SCR_JUMP, + PADDRH (data_ovrun), + +}/*-------------------------< PM0_DATA >--------------------*/,{ + /* + ** Read our host flags to SFBR, so we will be able + ** to check against the data direction we expect. + */ + SCR_FROM_REG (HF_REG), + 0, + /* + ** Check against actual DATA PHASE. + */ + SCR_JUMP ^ IFFALSE (WHEN (SCR_DATA_IN)), + PADDR (pm0_data_out), + /* + ** Actual phase is DATA IN. + ** Check against expected direction. + */ + SCR_JUMP ^ IFFALSE (MASK (HF_DATA_IN, HF_DATA_IN)), + PADDRH (data_ovrun), + /* + ** Keep track we are moving data from the + ** PM0 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_IN_PM0), + 0, + /* + ** Move the data to memory. + */ + SCR_CHMOV_TBL ^ SCR_DATA_IN, + offsetof (struct ccb, phys.pm0.sg), + SCR_JUMP, + PADDR (pm0_data_end), +}/*-------------------------< PM0_DATA_OUT >----------------*/,{ + /* + ** Actual phase is DATA OUT. + ** Check against expected direction. + */ + SCR_JUMP ^ IFTRUE (MASK (HF_DATA_IN, HF_DATA_IN)), + PADDRH (data_ovrun), + /* + ** Keep track we are moving data from the + ** PM0 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_IN_PM0), + 0, + /* + ** Move the data from memory. + */ + SCR_CHMOV_TBL ^ SCR_DATA_OUT, + offsetof (struct ccb, phys.pm0.sg), +}/*-------------------------< PM0_DATA_END >----------------*/,{ + /* + ** Clear the flag that told we were moving + ** data from the PM0 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_AND, (~HF_IN_PM0)), + 0, + /* + ** Return to the previous DATA script which + ** is guaranteed by design (if no bug) to be + ** the main DATA script for this transfer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.pm0.ret), + SCR_RETURN, + 0, +}/*-------------------------< PM1_DATA >--------------------*/,{ + /* + ** Read our host flags to SFBR, so we will be able + ** to check against the data direction we expect. + */ + SCR_FROM_REG (HF_REG), + 0, + /* + ** Check against actual DATA PHASE. + */ + SCR_JUMP ^ IFFALSE (WHEN (SCR_DATA_IN)), + PADDR (pm1_data_out), + /* + ** Actual phase is DATA IN. + ** Check against expected direction. + */ + SCR_JUMP ^ IFFALSE (MASK (HF_DATA_IN, HF_DATA_IN)), + PADDRH (data_ovrun), + /* + ** Keep track we are moving data from the + ** PM1 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_IN_PM1), + 0, + /* + ** Move the data to memory. + */ + SCR_CHMOV_TBL ^ SCR_DATA_IN, + offsetof (struct ccb, phys.pm1.sg), + SCR_JUMP, + PADDR (pm1_data_end), +}/*-------------------------< PM1_DATA_OUT >----------------*/,{ + /* + ** Actual phase is DATA OUT. + ** Check against expected direction. + */ + SCR_JUMP ^ IFTRUE (MASK (HF_DATA_IN, HF_DATA_IN)), + PADDRH (data_ovrun), + /* + ** Keep track we are moving data from the + ** PM1 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_OR, HF_IN_PM1), + 0, + /* + ** Move the data from memory. + */ + SCR_CHMOV_TBL ^ SCR_DATA_OUT, + offsetof (struct ccb, phys.pm1.sg), +}/*-------------------------< PM1_DATA_END >----------------*/,{ + /* + ** Clear the flag that told we were moving + ** data from the PM1 DATA mini-script. + */ + SCR_REG_REG (HF_REG, SCR_AND, (~HF_IN_PM1)), + 0, + /* + ** Return to the previous DATA script which + ** is guaranteed by design (if no bug) to be + ** the main DATA script for this transfer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.pm1.ret), + SCR_RETURN, + 0, +}/*---------------------------------------------------------*/ +}; + + +static struct scripth scripth0 __initdata = { +/*------------------------< START64 >-----------------------*/{ + /* + ** SCRIPT entry point for the 895A and the 896. + ** For now, there is no specific stuff for that + ** chip at this point, but this may come. + */ + SCR_JUMP, + PADDR (init), +}/*-------------------------< NO_DATA >-------------------*/,{ + SCR_JUMP, + PADDRH (data_ovrun), +}/*-----------------------< SEL_FOR_ABORT >------------------*/,{ + /* + ** We are jumped here by the C code, if we have + ** some target to reset or some disconnected + ** job to abort. Since error recovery is a serious + ** busyness, we will really reset the SCSI BUS, if + ** case of a SCSI interrupt occurring in this path. + */ + + /* + ** Set initiator mode. + */ + SCR_CLR (SCR_TRG), + 0, + /* + ** And try to select this target. + */ + SCR_SEL_TBL_ATN ^ offsetof (struct ncb, abrt_sel), + PADDR (reselect), + + /* + ** Wait for the selection to complete or + ** the selection to time out. + */ + SCR_JUMPR ^ IFFALSE (WHEN (SCR_MSG_OUT)), + -8, + /* + ** Call the C code. + */ + SCR_INT, + SIR_TARGET_SELECTED, + /* + ** The C code should let us continue here. + ** Send the 'kiss of death' message. + ** We expect an immediate disconnect once + ** the target has eaten the message. + */ + SCR_REG_REG (scntl2, SCR_AND, 0x7f), + 0, + SCR_MOVE_TBL ^ SCR_MSG_OUT, + offsetof (struct ncb, abrt_tbl), + SCR_CLR (SCR_ACK|SCR_ATN), + 0, + SCR_WAIT_DISC, + 0, + /* + ** Tell the C code that we are done. + */ + SCR_INT, + SIR_ABORT_SENT, +}/*-----------------------< SEL_FOR_ABORT_1 >--------------*/,{ + /* + ** Jump at scheduler. + */ + SCR_JUMP, + PADDR (start), + +}/*------------------------< SELECT_NO_ATN >-----------------*/,{ + /* + ** Set Initiator mode. + ** And try to select this target without ATN. + */ + + SCR_CLR (SCR_TRG), + 0, + SCR_SEL_TBL ^ offsetof (struct dsb, select), + PADDR (ungetjob), + /* + ** load the savep (saved pointer) into + ** the actual data pointer. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + /* + ** Initialize the status registers + */ + SCR_LOAD_REL (scr0, 4), + offsetof (struct ccb, phys.header.status), + +}/*------------------------< WF_SEL_DONE_NO_ATN >-----------------*/,{ + /* + ** Wait immediately for the next phase or + ** the selection to complete or time-out. + */ + SCR_JUMPR ^ IFFALSE (WHEN (SCR_MSG_OUT)), + 0, + SCR_JUMP, + PADDR (select2), + +}/*-------------------------< MSG_IN_ETC >--------------------*/,{ + /* + ** If it is an EXTENDED (variable size message) + ** Handle it. + */ + SCR_JUMP ^ IFTRUE (DATA (M_EXTENDED)), + PADDRH (msg_extended), + /* + ** Let the C code handle any other + ** 1 byte message. + */ + SCR_JUMP ^ IFTRUE (MASK (0x00, 0xf0)), + PADDRH (msg_received), + SCR_JUMP ^ IFTRUE (MASK (0x10, 0xf0)), + PADDRH (msg_received), + /* + ** We donnot handle 2 bytes messages from SCRIPTS. + ** So, let the C code deal with these ones too. + */ + SCR_JUMP ^ IFFALSE (MASK (0x20, 0xf0)), + PADDRH (msg_weird_seen), + SCR_CLR (SCR_ACK), + 0, + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin[1]), + SCR_JUMP, + PADDRH (msg_received), + +}/*-------------------------< MSG_RECEIVED >--------------------*/,{ + SCR_LOAD_REL (scratcha, 4), /* DUMMY READ */ + 0, + SCR_INT, + SIR_MSG_RECEIVED, + +}/*-------------------------< MSG_WEIRD_SEEN >------------------*/,{ + SCR_LOAD_REL (scratcha1, 4), /* DUMMY READ */ + 0, + SCR_INT, + SIR_MSG_WEIRD, + +}/*-------------------------< MSG_EXTENDED >--------------------*/,{ + /* + ** Clear ACK and get the next byte + ** assumed to be the message length. + */ + SCR_CLR (SCR_ACK), + 0, + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (msgin[1]), + /* + ** Try to catch some unlikely situations as 0 length + ** or too large the length. + */ + SCR_JUMP ^ IFTRUE (DATA (0)), + PADDRH (msg_weird_seen), + SCR_TO_REG (scratcha), + 0, + SCR_REG_REG (sfbr, SCR_ADD, (256-8)), + 0, + SCR_JUMP ^ IFTRUE (CARRYSET), + PADDRH (msg_weird_seen), + /* + ** We donnot handle extended messages from SCRIPTS. + ** Read the amount of data correponding to the + ** message length and call the C code. + */ + SCR_STORE_REL (scratcha, 1), + offsetof (struct dsb, smsg_ext.size), + SCR_CLR (SCR_ACK), + 0, + SCR_MOVE_TBL ^ SCR_MSG_IN, + offsetof (struct dsb, smsg_ext), + SCR_JUMP, + PADDRH (msg_received), + +}/*-------------------------< MSG_BAD >------------------*/,{ + /* + ** unimplemented message - reject it. + */ + SCR_INT, + SIR_REJECT_TO_SEND, + SCR_SET (SCR_ATN), + 0, + SCR_JUMP, + PADDR (clrack), + +}/*-------------------------< MSG_WEIRD >--------------------*/,{ + /* + ** weird message received + ** ignore all MSG IN phases and reject it. + */ + SCR_INT, + SIR_REJECT_TO_SEND, + SCR_SET (SCR_ATN), + 0, +}/*-------------------------< MSG_WEIRD1 >--------------------*/,{ + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP ^ IFFALSE (WHEN (SCR_MSG_IN)), + PADDR (dispatch), + SCR_MOVE_ABS (1) ^ SCR_MSG_IN, + NADDR (scratch), + SCR_JUMP, + PADDRH (msg_weird1), +}/*-------------------------< WDTR_RESP >----------------*/,{ + /* + ** let the target fetch our answer. + */ + SCR_SET (SCR_ATN), + 0, + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP ^ IFFALSE (WHEN (SCR_MSG_OUT)), + PADDRH (nego_bad_phase), + +}/*-------------------------< SEND_WDTR >----------------*/,{ + /* + ** Send the M_X_WIDE_REQ + */ + SCR_MOVE_ABS (4) ^ SCR_MSG_OUT, + NADDR (msgout), + SCR_JUMP, + PADDRH (msg_out_done), + +}/*-------------------------< SDTR_RESP >-------------*/,{ + /* + ** let the target fetch our answer. + */ + SCR_SET (SCR_ATN), + 0, + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP ^ IFFALSE (WHEN (SCR_MSG_OUT)), + PADDRH (nego_bad_phase), + +}/*-------------------------< SEND_SDTR >-------------*/,{ + /* + ** Send the M_X_SYNC_REQ + */ + SCR_MOVE_ABS (5) ^ SCR_MSG_OUT, + NADDR (msgout), + SCR_JUMP, + PADDRH (msg_out_done), + +}/*-------------------------< PPR_RESP >-------------*/,{ + /* + ** let the target fetch our answer. + */ + SCR_SET (SCR_ATN), + 0, + SCR_CLR (SCR_ACK), + 0, + SCR_JUMP ^ IFFALSE (WHEN (SCR_MSG_OUT)), + PADDRH (nego_bad_phase), + +}/*-------------------------< SEND_PPR >-------------*/,{ + /* + ** Send the M_X_PPR_REQ + */ + SCR_MOVE_ABS (8) ^ SCR_MSG_OUT, + NADDR (msgout), + SCR_JUMP, + PADDRH (msg_out_done), + +}/*-------------------------< NEGO_BAD_PHASE >------------*/,{ + SCR_INT, + SIR_NEGO_PROTO, + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< MSG_OUT >-------------------*/,{ + /* + ** The target requests a message. + */ + SCR_MOVE_ABS (1) ^ SCR_MSG_OUT, + NADDR (msgout), + /* + ** ... wait for the next phase + ** if it's a message out, send it again, ... + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_MSG_OUT)), + PADDRH (msg_out), +}/*-------------------------< MSG_OUT_DONE >--------------*/,{ + /* + ** ... else clear the message ... + */ + SCR_INT, + SIR_MSG_OUT_DONE, + /* + ** ... and process the next phase + */ + SCR_JUMP, + PADDR (dispatch), + +}/*-------------------------< DATA_OVRUN >-----------------------*/,{ + /* + * Use scratcha to count the extra bytes. + */ + SCR_LOAD_ABS (scratcha, 4), + PADDRH (zero), +}/*-------------------------< DATA_OVRUN1 >----------------------*/,{ + /* + * The target may want to transfer too much data. + * + * If phase is DATA OUT write 1 byte and count it. + */ + SCR_JUMPR ^ IFFALSE (WHEN (SCR_DATA_OUT)), + 16, + SCR_CHMOV_ABS (1) ^ SCR_DATA_OUT, + NADDR (scratch), + SCR_JUMP, + PADDRH (data_ovrun2), + /* + * If WSR is set, clear this condition, and + * count this byte. + */ + SCR_FROM_REG (scntl2), + 0, + SCR_JUMPR ^ IFFALSE (MASK (WSR, WSR)), + 16, + SCR_REG_REG (scntl2, SCR_OR, WSR), + 0, + SCR_JUMP, + PADDRH (data_ovrun2), + /* + * Finally check against DATA IN phase. + * Signal data overrun to the C code + * and jump to dispatcher if not so. + * Read 1 byte otherwise and count it. + */ + SCR_JUMPR ^ IFTRUE (WHEN (SCR_DATA_IN)), + 16, + SCR_INT, + SIR_DATA_OVERRUN, + SCR_JUMP, + PADDR (dispatch), + SCR_CHMOV_ABS (1) ^ SCR_DATA_IN, + NADDR (scratch), +}/*-------------------------< DATA_OVRUN2 >----------------------*/,{ + /* + * Count this byte. + * This will allow to return a negative + * residual to user. + */ + SCR_REG_REG (scratcha, SCR_ADD, 0x01), + 0, + SCR_REG_REG (scratcha1, SCR_ADDC, 0), + 0, + SCR_REG_REG (scratcha2, SCR_ADDC, 0), + 0, + /* + * .. and repeat as required. + */ + SCR_JUMP, + PADDRH (data_ovrun1), + +}/*-------------------------< ABORT_RESEL >----------------*/,{ + SCR_SET (SCR_ATN), + 0, + SCR_CLR (SCR_ACK), + 0, + /* + ** send the abort/abortag/reset message + ** we expect an immediate disconnect + */ + SCR_REG_REG (scntl2, SCR_AND, 0x7f), + 0, + SCR_MOVE_ABS (1) ^ SCR_MSG_OUT, + NADDR (msgout), + SCR_CLR (SCR_ACK|SCR_ATN), + 0, + SCR_WAIT_DISC, + 0, + SCR_INT, + SIR_RESEL_ABORTED, + SCR_JUMP, + PADDR (start), +}/*-------------------------< RESEND_IDENT >-------------------*/,{ + /* + ** The target stays in MSG OUT phase after having acked + ** Identify [+ Tag [+ Extended message ]]. Targets shall + ** behave this way on parity error. + ** We must send it again all the messages. + */ + SCR_SET (SCR_ATN), /* Shall be asserted 2 deskew delays before the */ + 0, /* 1rst ACK = 90 ns. Hope the NCR is'nt too fast */ + SCR_JUMP, + PADDR (send_ident), +}/*-------------------------< IDENT_BREAK >-------------------*/,{ + SCR_CLR (SCR_ATN), + 0, + SCR_JUMP, + PADDR (select2), +}/*-------------------------< IDENT_BREAK_ATN >----------------*/,{ + SCR_SET (SCR_ATN), + 0, + SCR_JUMP, + PADDR (select2), +}/*-------------------------< SDATA_IN >-------------------*/,{ + SCR_CHMOV_TBL ^ SCR_DATA_IN, + offsetof (struct dsb, sense), + SCR_CALL, + PADDR (datai_done), + SCR_JUMP, + PADDRH (data_ovrun), +}/*-------------------------< DATA_IO >--------------------*/,{ + /* + ** We jump here if the data direction was unknown at the + ** time we had to queue the command to the scripts processor. + ** Pointers had been set as follow in this situation: + ** savep --> DATA_IO + ** lastp --> start pointer when DATA_IN + ** goalp --> goal pointer when DATA_IN + ** wlastp --> start pointer when DATA_OUT + ** wgoalp --> goal pointer when DATA_OUT + ** This script sets savep/lastp/goalp according to the + ** direction chosen by the target. + */ + SCR_JUMP ^ IFTRUE (WHEN (SCR_DATA_OUT)), + PADDRH(data_io_out), +}/*-------------------------< DATA_IO_COM >-----------------*/,{ + /* + ** Direction is DATA IN. + ** Warning: we jump here, even when phase is DATA OUT. + */ + SCR_LOAD_REL (scratcha, 4), + offsetof (struct ccb, phys.header.lastp), + SCR_STORE_REL (scratcha, 4), + offsetof (struct ccb, phys.header.savep), + + /* + ** Jump to the SCRIPTS according to actual direction. + */ + SCR_LOAD_REL (temp, 4), + offsetof (struct ccb, phys.header.savep), + SCR_RETURN, + 0, +}/*-------------------------< DATA_IO_OUT >-----------------*/,{ + /* + ** Direction is DATA OUT. + */ + SCR_REG_REG (HF_REG, SCR_AND, (~HF_DATA_IN)), + 0, + SCR_LOAD_REL (scratcha, 4), + offsetof (struct ccb, phys.header.wlastp), + SCR_STORE_REL (scratcha, 4), + offsetof (struct ccb, phys.header.lastp), + SCR_LOAD_REL (scratcha, 4), + offsetof (struct ccb, phys.header.wgoalp), + SCR_STORE_REL (scratcha, 4), + offsetof (struct ccb, phys.header.goalp), + SCR_JUMP, + PADDRH(data_io_com), + +}/*-------------------------< RESEL_BAD_LUN >---------------*/,{ + /* + ** Message is an IDENTIFY, but lun is unknown. + ** Signal problem to C code for logging the event. + ** Send a M_ABORT to clear all pending tasks. + */ + SCR_INT, + SIR_RESEL_BAD_LUN, + SCR_JUMP, + PADDRH (abort_resel), +}/*-------------------------< BAD_I_T_L >------------------*/,{ + /* + ** We donnot have a task for that I_T_L. + ** Signal problem to C code for logging the event. + ** Send a M_ABORT message. + */ + SCR_INT, + SIR_RESEL_BAD_I_T_L, + SCR_JUMP, + PADDRH (abort_resel), +}/*-------------------------< BAD_I_T_L_Q >----------------*/,{ + /* + ** We donnot have a task that matches the tag. + ** Signal problem to C code for logging the event. + ** Send a M_ABORTTAG message. + */ + SCR_INT, + SIR_RESEL_BAD_I_T_L_Q, + SCR_JUMP, + PADDRH (abort_resel), +}/*-------------------------< BAD_STATUS >-----------------*/,{ + /* + ** Anything different from INTERMEDIATE + ** CONDITION MET should be a bad SCSI status, + ** given that GOOD status has already been tested. + ** Call the C code. + */ + SCR_LOAD_ABS (scratcha, 4), + PADDRH (startpos), + SCR_INT ^ IFFALSE (DATA (S_COND_MET)), + SIR_BAD_STATUS, + SCR_RETURN, + 0, + +}/*-------------------------< TWEAK_PMJ >------------------*/,{ + /* + ** Disable PM handling from SCRIPTS for the data phase + ** and so force PM to be handled from C code if HF_PM_TO_C + ** flag is set. + */ + SCR_FROM_REG(HF_REG), + 0, + SCR_JUMPR ^ IFTRUE (MASK (HF_PM_TO_C, HF_PM_TO_C)), + 16, + SCR_REG_REG (ccntl0, SCR_OR, ENPMJ), + 0, + SCR_RETURN, + 0, + SCR_REG_REG (ccntl0, SCR_AND, (~ENPMJ)), + 0, + SCR_RETURN, + 0, + +}/*-------------------------< PM_HANDLE >------------------*/,{ + /* + ** Phase mismatch handling. + ** + ** Since we have to deal with 2 SCSI data pointers + ** (current and saved), we need at least 2 contexts. + ** Each context (pm0 and pm1) has a saved area, a + ** SAVE mini-script and a DATA phase mini-script. + */ + /* + ** Get the PM handling flags. + */ + SCR_FROM_REG (HF_REG), + 0, + /* + ** If no flags (1rst PM for example), avoid + ** all the below heavy flags testing. + ** This makes the normal case a bit faster. + */ + SCR_JUMP ^ IFTRUE (MASK (0, (HF_IN_PM0 | HF_IN_PM1 | HF_DP_SAVED))), + PADDRH (pm_handle1), + /* + ** If we received a SAVE DP, switch to the + ** other PM context since the savep may point + ** to the current PM context. + */ + SCR_JUMPR ^ IFFALSE (MASK (HF_DP_SAVED, HF_DP_SAVED)), + 8, + SCR_REG_REG (sfbr, SCR_XOR, HF_ACT_PM), + 0, + /* + ** If we have been interrupt in a PM DATA mini-script, + ** we take the return address from the corresponding + ** saved area. + ** This ensure the return address always points to the + ** main DATA script for this transfer. + */ + SCR_JUMP ^ IFTRUE (MASK (0, (HF_IN_PM0 | HF_IN_PM1))), + PADDRH (pm_handle1), + SCR_JUMPR ^ IFFALSE (MASK (HF_IN_PM0, HF_IN_PM0)), + 16, + SCR_LOAD_REL (ia, 4), + offsetof(struct ccb, phys.pm0.ret), + SCR_JUMP, + PADDRH (pm_save), + SCR_LOAD_REL (ia, 4), + offsetof(struct ccb, phys.pm1.ret), + SCR_JUMP, + PADDRH (pm_save), +}/*-------------------------< PM_HANDLE1 >-----------------*/,{ + /* + ** Normal case. + ** Update the return address so that it + ** will point after the interrupted MOVE. + */ + SCR_REG_REG (ia, SCR_ADD, 8), + 0, + SCR_REG_REG (ia1, SCR_ADDC, 0), + 0, +}/*-------------------------< PM_SAVE >--------------------*/,{ + /* + ** Clear all the flags that told us if we were + ** interrupted in a PM DATA mini-script and/or + ** we received a SAVE DP. + */ + SCR_SFBR_REG (HF_REG, SCR_AND, (~(HF_IN_PM0|HF_IN_PM1|HF_DP_SAVED))), + 0, + /* + ** Choose the current PM context. + */ + SCR_JUMP ^ IFTRUE (MASK (HF_ACT_PM, HF_ACT_PM)), + PADDRH (pm1_save), +}/*-------------------------< PM0_SAVE >-------------------*/,{ + SCR_STORE_REL (ia, 4), + offsetof(struct ccb, phys.pm0.ret), + /* + ** If WSR bit is set, either UA and RBC may + ** have to be changed whatever the device wants + ** to ignore this residue ot not. + */ + SCR_FROM_REG (scntl2), + 0, + SCR_CALL ^ IFTRUE (MASK (WSR, WSR)), + PADDRH (pm_wsr_handle), + /* + ** Save the remaining byte count, the updated + ** address and the return address. + */ + SCR_STORE_REL (rbc, 4), + offsetof(struct ccb, phys.pm0.sg.size), + SCR_STORE_REL (ua, 4), + offsetof(struct ccb, phys.pm0.sg.addr), + /* + ** Set the current pointer at the PM0 DATA mini-script. + */ + SCR_LOAD_ABS (temp, 4), + PADDRH (pm0_data_addr), + SCR_JUMP, + PADDR (dispatch), +}/*-------------------------< PM1_SAVE >-------------------*/,{ + SCR_STORE_REL (ia, 4), + offsetof(struct ccb, phys.pm1.ret), + /* + ** If WSR bit is set, either UA and RBC may + ** have been changed whatever the device wants + ** to ignore this residue or not. + */ + SCR_FROM_REG (scntl2), + 0, + SCR_CALL ^ IFTRUE (MASK (WSR, WSR)), + PADDRH (pm_wsr_handle), + /* + ** Save the remaining byte count, the updated + ** address and the return address. + */ + SCR_STORE_REL (rbc, 4), + offsetof(struct ccb, phys.pm1.sg.size), + SCR_STORE_REL (ua, 4), + offsetof(struct ccb, phys.pm1.sg.addr), + /* + ** Set the current pointer at the PM1 DATA mini-script. + */ + SCR_LOAD_ABS (temp, 4), + PADDRH (pm1_data_addr), + SCR_JUMP, + PADDR (dispatch), +}/*--------------------------< PM_WSR_HANDLE >-----------------------*/,{ + /* + * Phase mismatch handling from SCRIPT with WSR set. + * Such a condition can occur if the chip wants to + * execute a CHMOV(size > 1) when the WSR bit is + * set and the target changes PHASE. + */ +#ifdef SYM_DEBUG_PM_WITH_WSR + /* + * Some debugging may still be needed.:) + */ + SCR_INT, + SIR_PM_WITH_WSR, +#endif + /* + * We must move the residual byte to memory. + * + * UA contains bit 0..31 of the address to + * move the residual byte. + * Move it to the table indirect. + */ + SCR_STORE_REL (ua, 4), + offsetof (struct ccb, phys.wresid.addr), + /* + * Increment UA (move address to next position). + */ + SCR_REG_REG (ua, SCR_ADD, 1), + 0, + SCR_REG_REG (ua1, SCR_ADDC, 0), + 0, + SCR_REG_REG (ua2, SCR_ADDC, 0), + 0, + SCR_REG_REG (ua3, SCR_ADDC, 0), + 0, + /* + * Compute SCRATCHA as: + * - size to transfer = 1 byte. + * - bit 24..31 = high address bit [32...39]. + */ + SCR_LOAD_ABS (scratcha, 4), + PADDRH (zero), + SCR_REG_REG (scratcha, SCR_OR, 1), + 0, + SCR_FROM_REG (rbc3), + 0, + SCR_TO_REG (scratcha3), + 0, + /* + * Move this value to the table indirect. + */ + SCR_STORE_REL (scratcha, 4), + offsetof (struct ccb, phys.wresid.size), + /* + * Wait for a valid phase. + * While testing with bogus QUANTUM drives, the C1010 + * sometimes raised a spurious phase mismatch with + * WSR and the CHMOV(1) triggered another PM. + * Waiting explicitely for the PHASE seemed to avoid + * the nested phase mismatch. Btw, this didn't happen + * using my IBM drives. + */ + SCR_JUMPR ^ IFFALSE (WHEN (SCR_DATA_IN)), + 0, + /* + * Perform the move of the residual byte. + */ + SCR_CHMOV_TBL ^ SCR_DATA_IN, + offsetof (struct ccb, phys.wresid), + /* + * We can now handle the phase mismatch with UA fixed. + * RBC[0..23]=0 is a special case that does not require + * a PM context. The C code also checks against this. + */ + SCR_FROM_REG (rbc), + 0, + SCR_RETURN ^ IFFALSE (DATA (0)), + 0, + SCR_FROM_REG (rbc1), + 0, + SCR_RETURN ^ IFFALSE (DATA (0)), + 0, + SCR_FROM_REG (rbc2), + 0, + SCR_RETURN ^ IFFALSE (DATA (0)), + 0, + /* + * RBC[0..23]=0. + * Not only we donnot need a PM context, but this would + * lead to a bogus CHMOV(0). This condition means that + * the residual was the last byte to move from this CHMOV. + * So, we just have to move the current data script pointer + * (i.e. TEMP) to the SCRIPTS address following the + * interrupted CHMOV and jump to dispatcher. + */ + SCR_STORE_ABS (ia, 4), + PADDRH (scratch), + SCR_LOAD_ABS (temp, 4), + PADDRH (scratch), + SCR_JUMP, + PADDR (dispatch), +}/*--------------------------< WSR_MA_HELPER >-----------------------*/,{ + /* + * Helper for the C code when WSR bit is set. + * Perform the move of the residual byte. + */ + SCR_CHMOV_TBL ^ SCR_DATA_IN, + offsetof (struct ccb, phys.wresid), + SCR_JUMP, + PADDR (dispatch), +}/*-------------------------< ZERO >------------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< SCRATCH >---------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< SCRATCH1 >--------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< PM0_DATA_ADDR >---------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< PM1_DATA_ADDR >---------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< SAVED_DSA >-------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< SAVED_DRS >-------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< DONE_POS >--------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< STARTPOS >--------------------*/,{ + SCR_DATA_ZERO, +}/*-------------------------< TARGTBL >---------------------*/,{ + SCR_DATA_ZERO, + + +/* +** We may use MEMORY MOVE instructions to load the on chip-RAM, +** if it happens that mapping PCI memory is not possible. +** But writing the RAM from the CPU is the preferred method, +** since PCI 2.2 seems to disallow PCI self-mastering. +*/ + +#ifdef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + +}/*-------------------------< START_RAM >-------------------*/,{ + /* + ** Load the script into on-chip RAM, + ** and jump to start point. + */ + SCR_COPY (sizeof (struct script)), +}/*-------------------------< SCRIPT0_BA >--------------------*/,{ + 0, + PADDR (start), + SCR_JUMP, + PADDR (init), + +}/*-------------------------< START_RAM64 >--------------------*/,{ + /* + ** Load the RAM and start for 64 bit PCI (895A,896). + ** Both scripts (script and scripth) are loaded into + ** the RAM which is 8K (4K for 825A/875/895). + ** We also need to load some 32-63 bit segments + ** address of the SCRIPTS processor. + ** LOAD/STORE ABSOLUTE always refers to on-chip RAM + ** in our implementation. The main memory is + ** accessed using LOAD/STORE DSA RELATIVE. + */ + SCR_LOAD_REL (mmws, 4), + offsetof (struct ncb, scr_ram_seg), + SCR_COPY (sizeof(struct script)), +}/*-------------------------< SCRIPT0_BA64 >--------------------*/,{ + 0, + PADDR (start), + SCR_COPY (sizeof(struct scripth)), +}/*-------------------------< SCRIPTH0_BA64 >--------------------*/,{ + 0, + PADDRH (start64), + SCR_LOAD_REL (mmrs, 4), + offsetof (struct ncb, scr_ram_seg), + SCR_JUMP64, + PADDRH (start64), +}/*-------------------------< RAM_SEG64 >--------------------*/,{ + 0, + +#endif /* SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ + +}/*-------------------------< SNOOPTEST >-------------------*/,{ + /* + ** Read the variable. + */ + SCR_LOAD_REL (scratcha, 4), + offsetof(struct ncb, ncr_cache), + SCR_STORE_REL (temp, 4), + offsetof(struct ncb, ncr_cache), + SCR_LOAD_REL (temp, 4), + offsetof(struct ncb, ncr_cache), +}/*-------------------------< SNOOPEND >-------------------*/,{ + /* + ** And stop. + */ + SCR_INT, + 99, +}/*--------------------------------------------------------*/ +}; + +/*========================================================== +** +** +** Fill in #define dependent parts of the script +** +** +**========================================================== +*/ + +void __init ncr_script_fill (struct script * scr, struct scripth * scrh) +{ + int i; + ncrcmd *p; + + p = scr->data_in; + for (i=0; i<MAX_SCATTER; i++) { + *p++ =SCR_CHMOV_TBL ^ SCR_DATA_IN; + *p++ =offsetof (struct dsb, data[i]); + }; + + assert ((u_long)p == (u_long)&scr->data_in + sizeof (scr->data_in)); + + p = scr->data_out; + + for (i=0; i<MAX_SCATTER; i++) { + *p++ =SCR_CHMOV_TBL ^ SCR_DATA_OUT; + *p++ =offsetof (struct dsb, data[i]); + }; + + assert ((u_long)p == (u_long)&scr->data_out + sizeof (scr->data_out)); +} + +/*========================================================== +** +** +** Copy and rebind a script. +** +** +**========================================================== +*/ + +static void __init +ncr_script_copy_and_bind (ncb_p np,ncrcmd *src,ncrcmd *dst,int len) +{ + ncrcmd opcode, new, old, tmp1, tmp2; + ncrcmd *start, *end; + int relocs; + int opchanged = 0; + + start = src; + end = src + len/4; + + while (src < end) { + + opcode = *src++; + *dst++ = cpu_to_scr(opcode); + + /* + ** If we forget to change the length + ** in struct script, a field will be + ** padded with 0. This is an illegal + ** command. + */ + + if (opcode == 0) { + printk (KERN_INFO "%s: ERROR0 IN SCRIPT at %d.\n", + ncr_name(np), (int) (src-start-1)); + MDELAY (10000); + continue; + }; + + /* + ** We use the bogus value 0xf00ff00f ;-) + ** to reserve data area in SCRIPTS. + */ + if (opcode == SCR_DATA_ZERO) { + dst[-1] = 0; + continue; + } + + if (DEBUG_FLAGS & DEBUG_SCRIPT) + printk (KERN_INFO "%p: <%x>\n", + (src-1), (unsigned)opcode); + + /* + ** We don't have to decode ALL commands + */ + switch (opcode >> 28) { + + case 0xf: + /* + ** LOAD / STORE DSA relative, don't relocate. + */ + relocs = 0; + break; + case 0xe: + /* + ** LOAD / STORE absolute. + */ + relocs = 1; + break; + case 0xc: + /* + ** COPY has TWO arguments. + */ + relocs = 2; + tmp1 = src[0]; + tmp2 = src[1]; +#ifdef RELOC_KVAR + if ((tmp1 & RELOC_MASK) == RELOC_KVAR) + tmp1 = 0; + if ((tmp2 & RELOC_MASK) == RELOC_KVAR) + tmp2 = 0; +#endif + if ((tmp1 ^ tmp2) & 3) { + printk (KERN_ERR"%s: ERROR1 IN SCRIPT at %d.\n", + ncr_name(np), (int) (src-start-1)); + MDELAY (1000); + } + /* + ** If PREFETCH feature not enabled, remove + ** the NO FLUSH bit if present. + */ + if ((opcode & SCR_NO_FLUSH) && + !(np->features & FE_PFEN)) { + dst[-1] = cpu_to_scr(opcode & ~SCR_NO_FLUSH); + ++opchanged; + } + break; + + case 0x0: + /* + ** MOVE/CHMOV (absolute address) + */ + if (!(np->features & FE_WIDE)) + dst[-1] = cpu_to_scr(opcode | OPC_MOVE); + relocs = 1; + break; + + case 0x1: + /* + ** MOVE/CHMOV (table indirect) + */ + if (!(np->features & FE_WIDE)) + dst[-1] = cpu_to_scr(opcode | OPC_MOVE); + relocs = 0; + break; + + case 0x8: + /* + ** JUMP / CALL + ** dont't relocate if relative :-) + */ + if (opcode & 0x00800000) + relocs = 0; + else if ((opcode & 0xf8400000) == 0x80400000)/*JUMP64*/ + relocs = 2; + else + relocs = 1; + break; + + case 0x4: + case 0x5: + case 0x6: + case 0x7: + relocs = 1; + break; + + default: + relocs = 0; + break; + }; + + if (!relocs) { + *dst++ = cpu_to_scr(*src++); + continue; + } + while (relocs--) { + old = *src++; + + switch (old & RELOC_MASK) { + case RELOC_REGISTER: + new = (old & ~RELOC_MASK) + pcivtobus(np->base_ba); + break; + case RELOC_LABEL: + new = (old & ~RELOC_MASK) + np->p_script; + break; + case RELOC_LABELH: + new = (old & ~RELOC_MASK) + np->p_scripth; + break; + case RELOC_SOFTC: + new = (old & ~RELOC_MASK) + np->p_ncb; + break; +#ifdef RELOC_KVAR + case RELOC_KVAR: + new=0; + if (((old & ~RELOC_MASK) < SCRIPT_KVAR_FIRST) || + ((old & ~RELOC_MASK) > SCRIPT_KVAR_LAST)) + panic("ncr KVAR out of range"); + new = vtobus(script_kvars[old & ~RELOC_MASK]); +#endif + break; + case 0: + /* Don't relocate a 0 address. */ + if (old == 0) { + new = old; + break; + } + /* fall through */ + default: + new = 0; /* For 'cc' not to complain */ + panic("ncr_script_copy_and_bind: " + "weird relocation %x\n", old); + break; + } + + *dst++ = cpu_to_scr(new); + } + }; +} + +/*========================================================== +** +** +** Auto configuration: attach and init a host adapter. +** +** +**========================================================== +*/ + +/* +** Linux host data structure. +*/ + +struct host_data { + struct ncb *ncb; +}; + +/* +** Print something which allows to retrieve the controler type, unit, +** target, lun concerned by a kernel message. +*/ + +static void PRINT_TARGET(ncb_p np, int target) +{ + printk(KERN_INFO "%s-<%d,*>: ", ncr_name(np), target); +} + +static void PRINT_LUN(ncb_p np, int target, int lun) +{ + printk(KERN_INFO "%s-<%d,%d>: ", ncr_name(np), target, lun); +} + +static void PRINT_ADDR(Scsi_Cmnd *cmd) +{ + struct host_data *host_data = (struct host_data *) cmd->host->hostdata; + PRINT_LUN(host_data->ncb, cmd->target, cmd->lun); +} + +/*========================================================== +** +** NCR chip clock divisor table. +** Divisors are multiplied by 10,000,000 in order to make +** calculations more simple. +** +**========================================================== +*/ + +#define _5M 5000000 +static u_long div_10M[] = + {2*_5M, 3*_5M, 4*_5M, 6*_5M, 8*_5M, 12*_5M, 16*_5M}; + + +/*=============================================================== +** +** Prepare io register values used by ncr_init() according +** to selected and supported features. +** +** NCR/SYMBIOS chips allow burst lengths of 2, 4, 8, 16, 32, 64, +** 128 transfers. All chips support at least 16 transfers bursts. +** The 825A, 875 and 895 chips support bursts of up to 128 +** transfers and the 895A and 896 support bursts of up to 64 +** transfers. All other chips support up to 16 transfers bursts. +** +** For PCI 32 bit data transfers each transfer is a DWORD (4 bytes). +** It is a QUADWORD (8 bytes) for PCI 64 bit data transfers. +** Only the 896 is able to perform 64 bit data transfers. +** +** We use log base 2 (burst length) as internal code, with +** value 0 meaning "burst disabled". +** +**=============================================================== +*/ + +/* + * Burst length from burst code. + */ +#define burst_length(bc) (!(bc))? 0 : 1 << (bc) + +/* + * Burst code from io register bits. + */ +#define burst_code(dmode, ctest4, ctest5) \ + (ctest4) & 0x80? 0 : (((dmode) & 0xc0) >> 6) + ((ctest5) & 0x04) + 1 + +/* + * Set initial io register bits from burst code. + */ +static inline void ncr_init_burst(ncb_p np, u_char bc) +{ + np->rv_ctest4 &= ~0x80; + np->rv_dmode &= ~(0x3 << 6); + np->rv_ctest5 &= ~0x4; + + if (!bc) { + np->rv_ctest4 |= 0x80; + } + else { + --bc; + np->rv_dmode |= ((bc & 0x3) << 6); + np->rv_ctest5 |= (bc & 0x4); + } +} + +#ifdef SCSI_NCR_NVRAM_SUPPORT + +/* +** Get target set-up from Symbios format NVRAM. +*/ + +static void __init +ncr_Symbios_setup_target(ncb_p np, int target, Symbios_nvram *nvram) +{ + tcb_p tp = &np->target[target]; + Symbios_target *tn = &nvram->target[target]; + + tp->usrsync = tn->sync_period ? (tn->sync_period + 3) / 4 : 255; + tp->usrwide = tn->bus_width == 0x10 ? 1 : 0; + tp->usrtags = + (tn->flags & SYMBIOS_QUEUE_TAGS_ENABLED)? MAX_TAGS : 0; + + if (!(tn->flags & SYMBIOS_DISCONNECT_ENABLE)) + tp->usrflag |= UF_NODISC; + if (!(tn->flags & SYMBIOS_SCAN_AT_BOOT_TIME)) + tp->usrflag |= UF_NOSCAN; +} + +/* +** Get target set-up from Tekram format NVRAM. +*/ + +static void __init +ncr_Tekram_setup_target(ncb_p np, int target, Tekram_nvram *nvram) +{ + tcb_p tp = &np->target[target]; + struct Tekram_target *tn = &nvram->target[target]; + int i; + + if (tn->flags & TEKRAM_SYNC_NEGO) { + i = tn->sync_index & 0xf; + tp->usrsync = Tekram_sync[i]; + } + + tp->usrwide = (tn->flags & TEKRAM_WIDE_NEGO) ? 1 : 0; + + if (tn->flags & TEKRAM_TAGGED_COMMANDS) { + tp->usrtags = 2 << nvram->max_tags_index; + } + + if (!(tn->flags & TEKRAM_DISCONNECT_ENABLE)) + tp->usrflag = UF_NODISC; + + /* If any device does not support parity, we will not use this option */ + if (!(tn->flags & TEKRAM_PARITY_CHECK)) + np->rv_scntl0 &= ~0x0a; /* SCSI parity checking disabled */ +} +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/* +** Save initial settings of some IO registers. +** Assumed to have been set by BIOS. +*/ +static void __init ncr_save_initial_setting(ncb_p np) +{ + np->sv_scntl0 = INB(nc_scntl0) & 0x0a; + np->sv_dmode = INB(nc_dmode) & 0xce; + np->sv_dcntl = INB(nc_dcntl) & 0xa8; + np->sv_ctest3 = INB(nc_ctest3) & 0x01; + np->sv_ctest4 = INB(nc_ctest4) & 0x80; + np->sv_gpcntl = INB(nc_gpcntl); + np->sv_stest2 = INB(nc_stest2) & 0x20; + np->sv_stest4 = INB(nc_stest4); + np->sv_stest1 = INB(nc_stest1); + + np->sv_scntl3 = INB(nc_scntl3) & 0x07; + + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66) ){ + /* + ** C1010 always uses large fifo, bit 5 rsvd + ** scntl4 used ONLY with C1010 + */ + np->sv_ctest5 = INB(nc_ctest5) & 0x04 ; + np->sv_scntl4 = INB(nc_scntl4); + } + else { + np->sv_ctest5 = INB(nc_ctest5) & 0x24 ; + np->sv_scntl4 = 0; + } +} + +/* +** Prepare io register values used by ncr_init() +** according to selected and supported features. +*/ +static int __init ncr_prepare_setting(ncb_p np, ncr_nvram *nvram) +{ + u_char burst_max; + u_long period; + int i; + + /* + ** Wide ? + */ + + np->maxwide = (np->features & FE_WIDE)? 1 : 0; + + /* + ** Get the frequency of the chip's clock. + ** Find the right value for scntl3. + */ + + if (np->features & FE_QUAD) + np->multiplier = 4; + else if (np->features & FE_DBLR) + np->multiplier = 2; + else + np->multiplier = 1; + + np->clock_khz = (np->features & FE_CLK80)? 80000 : 40000; + np->clock_khz *= np->multiplier; + + if (np->clock_khz != 40000) + ncr_getclock(np, np->multiplier); + + /* + * Divisor to be used for async (timer pre-scaler). + * + * Note: For C1010 the async divisor is 2(8) if he + * quadrupler is disabled (enabled). + */ + + if ( (np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + + np->rv_scntl3 = 0; + } + else + { + i = np->clock_divn - 1; + while (--i >= 0) { + if (10ul * SCSI_NCR_MIN_ASYNC * np->clock_khz + > div_10M[i]) { + ++i; + break; + } + } + np->rv_scntl3 = i+1; + } + + + /* + * Save the ultra3 register for the C1010/C1010_66 + */ + + np->rv_scntl4 = np->sv_scntl4; + + /* + * Minimum synchronous period factor supported by the chip. + * Btw, 'period' is in tenths of nanoseconds. + */ + + period = (4 * div_10M[0] + np->clock_khz - 1) / np->clock_khz; + if (period <= 250) np->minsync = 10; + else if (period <= 303) np->minsync = 11; + else if (period <= 500) np->minsync = 12; + else np->minsync = (period + 40 - 1) / 40; + + /* + * Fix up. If sync. factor is 10 (160000Khz clock) and chip + * supports ultra3, then min. sync. period 12.5ns and the factor is 9 + */ + + if ((np->minsync == 10) && (np->features & FE_ULTRA3)) + np->minsync = 9; + + /* + * Check against chip SCSI standard support (SCSI-2,ULTRA,ULTRA2). + * + * Transfer period minimums: SCSI-1 200 (50); Fast 100 (25) + * Ultra 50 (12); Ultra2 (6); Ultra3 (3) + */ + + if (np->minsync < 25 && !(np->features & (FE_ULTRA|FE_ULTRA2|FE_ULTRA3))) + np->minsync = 25; + else if (np->minsync < 12 && (np->features & FE_ULTRA)) + np->minsync = 12; + else if (np->minsync < 10 && (np->features & FE_ULTRA2)) + np->minsync = 10; + else if (np->minsync < 9 && (np->features & FE_ULTRA3)) + np->minsync = 9; + + /* + * Maximum synchronous period factor supported by the chip. + */ + + period = (11 * div_10M[np->clock_divn - 1]) / (4 * np->clock_khz); + np->maxsync = period > 2540 ? 254 : period / 10; + + /* + ** 64 bit (53C895A or 53C896) ? + */ + if (np->features & FE_64BIT) +#ifdef SCSI_NCR_USE_64BIT_DAC + np->rv_ccntl1 |= (XTIMOD | EXTIBMV); +#else + np->rv_ccntl1 |= (DDAC); +#endif + + /* + ** Phase mismatch handled by SCRIPTS (53C895A, 53C896 or C1010) ? + */ + if (np->features & FE_NOPM) + np->rv_ccntl0 |= (ENPMJ); + + /* + ** Prepare initial value of other IO registers + */ +#if defined SCSI_NCR_TRUST_BIOS_SETTING + np->rv_scntl0 = np->sv_scntl0; + np->rv_dmode = np->sv_dmode; + np->rv_dcntl = np->sv_dcntl; + np->rv_ctest3 = np->sv_ctest3; + np->rv_ctest4 = np->sv_ctest4; + np->rv_ctest5 = np->sv_ctest5; + burst_max = burst_code(np->sv_dmode, np->sv_ctest4, np->sv_ctest5); +#else + + /* + ** Select burst length (dwords) + */ + burst_max = driver_setup.burst_max; + if (burst_max == 255) + burst_max = burst_code(np->sv_dmode, np->sv_ctest4, np->sv_ctest5); + if (burst_max > 7) + burst_max = 7; + if (burst_max > np->maxburst) + burst_max = np->maxburst; + + /* + ** DEL 352 - 53C810 Rev x11 - Part Number 609-0392140 - ITEM 2. + ** This chip and the 860 Rev 1 may wrongly use PCI cache line + ** based transactions on LOAD/STORE instructions. So we have + ** to prevent these chips from using such PCI transactions in + ** this driver. The generic sym53c8xx driver that does not use + ** LOAD/STORE instructions does not need this work-around. + */ + if ((np->device_id == PCI_DEVICE_ID_NCR_53C810 && + np->revision_id >= 0x10 && np->revision_id <= 0x11) || + (np->device_id == PCI_DEVICE_ID_NCR_53C860 && + np->revision_id <= 0x1)) + np->features &= ~(FE_WRIE|FE_ERL|FE_ERMP); + + /* + ** DEL ? - 53C1010 Rev 1 - Part Number 609-0393638 + ** 64-bit Slave Cycles must be disabled. + */ + if ( ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) && (np->revision_id < 0x02) ) + || (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66 ) ) + np->rv_ccntl1 |= 0x10; + + /* + ** Select all supported special features. + ** If we are using on-board RAM for scripts, prefetch (PFEN) + ** does not help, but burst op fetch (BOF) does. + ** Disabling PFEN makes sure BOF will be used. + */ + if (np->features & FE_ERL) + np->rv_dmode |= ERL; /* Enable Read Line */ + if (np->features & FE_BOF) + np->rv_dmode |= BOF; /* Burst Opcode Fetch */ + if (np->features & FE_ERMP) + np->rv_dmode |= ERMP; /* Enable Read Multiple */ +#if 1 + if ((np->features & FE_PFEN) && !np->base2_ba) +#else + if (np->features & FE_PFEN) +#endif + np->rv_dcntl |= PFEN; /* Prefetch Enable */ + if (np->features & FE_CLSE) + np->rv_dcntl |= CLSE; /* Cache Line Size Enable */ + if (np->features & FE_WRIE) + np->rv_ctest3 |= WRIE; /* Write and Invalidate */ + + + if ( (np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66) && + (np->features & FE_DFS)) + np->rv_ctest5 |= DFS; /* Dma Fifo Size */ + /* C1010/C1010_66 always large fifo */ + + /* + ** Select some other + */ + if (driver_setup.master_parity) + np->rv_ctest4 |= MPEE; /* Master parity checking */ + if (driver_setup.scsi_parity) + np->rv_scntl0 |= 0x0a; /* full arb., ena parity, par->ATN */ + +#ifdef SCSI_NCR_NVRAM_SUPPORT + /* + ** Get parity checking, host ID and verbose mode from NVRAM + **/ + if (nvram) { + switch(nvram->type) { + case SCSI_NCR_TEKRAM_NVRAM: + np->myaddr = nvram->data.Tekram.host_id & 0x0f; + break; + case SCSI_NCR_SYMBIOS_NVRAM: + if (!(nvram->data.Symbios.flags & SYMBIOS_PARITY_ENABLE)) + np->rv_scntl0 &= ~0x0a; + np->myaddr = nvram->data.Symbios.host_id & 0x0f; + if (nvram->data.Symbios.flags & SYMBIOS_VERBOSE_MSGS) + np->verbose += 1; + break; + } + } +#endif + /* + ** Get SCSI addr of host adapter (set by bios?). + */ + if (np->myaddr == 255) { + np->myaddr = INB(nc_scid) & 0x07; + if (!np->myaddr) + np->myaddr = SCSI_NCR_MYADDR; + } + +#endif /* SCSI_NCR_TRUST_BIOS_SETTING */ + + /* + * Prepare initial io register bits for burst length + */ + ncr_init_burst(np, burst_max); + + /* + ** Set SCSI BUS mode. + ** + ** - ULTRA2 chips (895/895A/896) + ** and ULTRA 3 chips (1010) report the current + ** BUS mode through the STEST4 IO register. + ** - For previous generation chips (825/825A/875), + ** user has to tell us how to check against HVD, + ** since a 100% safe algorithm is not possible. + */ + np->scsi_mode = SMODE_SE; + if (np->features & (FE_ULTRA2 | FE_ULTRA3)) + np->scsi_mode = (np->sv_stest4 & SMODE); + else if (np->features & FE_DIFF) { + switch(driver_setup.diff_support) { + case 4: /* Trust previous settings if present, then GPIO3 */ + if (np->sv_scntl3) { + if (np->sv_stest2 & 0x20) + np->scsi_mode = SMODE_HVD; + break; + } + case 3: /* SYMBIOS controllers report HVD through GPIO3 */ + if (nvram && nvram->type != SCSI_NCR_SYMBIOS_NVRAM) + break; + if (INB(nc_gpreg) & 0x08) + break; + case 2: /* Set HVD unconditionally */ + np->scsi_mode = SMODE_HVD; + case 1: /* Trust previous settings for HVD */ + if (np->sv_stest2 & 0x20) + np->scsi_mode = SMODE_HVD; + break; + default:/* Don't care about HVD */ + break; + } + } + if (np->scsi_mode == SMODE_HVD) + np->rv_stest2 |= 0x20; + + /* + ** Set LED support from SCRIPTS. + ** Ignore this feature for boards known to use a + ** specific GPIO wiring and for the 895A or 896 + ** that drive the LED directly. + ** Also probe initial setting of GPIO0 as output. + */ + if ((driver_setup.led_pin || + (nvram && nvram->type == SCSI_NCR_SYMBIOS_NVRAM)) && + !(np->features & FE_LEDC) && !(np->sv_gpcntl & 0x01)) + np->features |= FE_LED0; + + /* + ** Set irq mode. + */ + switch(driver_setup.irqm & 3) { + case 2: + np->rv_dcntl |= IRQM; + break; + case 1: + np->rv_dcntl |= (np->sv_dcntl & IRQM); + break; + default: + break; + } + + /* + ** Configure targets according to driver setup. + ** If NVRAM present get targets setup from NVRAM. + ** Allow to override sync, wide and NOSCAN from + ** boot command line. + */ + for (i = 0 ; i < MAX_TARGET ; i++) { + tcb_p tp = &np->target[i]; + + tp->usrsync = 255; +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (nvram) { + switch(nvram->type) { + case SCSI_NCR_TEKRAM_NVRAM: + ncr_Tekram_setup_target(np, i, &nvram->data.Tekram); + break; + case SCSI_NCR_SYMBIOS_NVRAM: + ncr_Symbios_setup_target(np, i, &nvram->data.Symbios); + break; + } + if (driver_setup.use_nvram & 0x2) + tp->usrsync = driver_setup.default_sync; + if (driver_setup.use_nvram & 0x4) + tp->usrwide = driver_setup.max_wide; + if (driver_setup.use_nvram & 0x8) + tp->usrflag &= ~UF_NOSCAN; + } + else { +#else + if (1) { +#endif + tp->usrsync = driver_setup.default_sync; + tp->usrwide = driver_setup.max_wide; + tp->usrtags = MAX_TAGS; + if (!driver_setup.disconnection) + np->target[i].usrflag = UF_NODISC; + } + } + + /* + ** Announce all that stuff to user. + */ + + i = nvram ? nvram->type : 0; + printk(KERN_INFO "%s: %sID %d, Fast-%d%s%s\n", ncr_name(np), + i == SCSI_NCR_SYMBIOS_NVRAM ? "Symbios format NVRAM, " : + (i == SCSI_NCR_TEKRAM_NVRAM ? "Tekram format NVRAM, " : ""), + np->myaddr, + np->minsync < 10 ? 80 : + (np->minsync < 12 ? 40 : (np->minsync < 25 ? 20 : 10) ), + (np->rv_scntl0 & 0xa) ? ", Parity Checking" : ", NO Parity", + (np->rv_stest2 & 0x20) ? ", Differential" : ""); + + if (bootverbose > 1) { + printk (KERN_INFO "%s: initial SCNTL3/DMODE/DCNTL/CTEST3/4/5 = " + "(hex) %02x/%02x/%02x/%02x/%02x/%02x\n", + ncr_name(np), np->sv_scntl3, np->sv_dmode, np->sv_dcntl, + np->sv_ctest3, np->sv_ctest4, np->sv_ctest5); + + printk (KERN_INFO "%s: final SCNTL3/DMODE/DCNTL/CTEST3/4/5 = " + "(hex) %02x/%02x/%02x/%02x/%02x/%02x\n", + ncr_name(np), np->rv_scntl3, np->rv_dmode, np->rv_dcntl, + np->rv_ctest3, np->rv_ctest4, np->rv_ctest5); + } + + if (bootverbose && np->base2_ba) + printk (KERN_INFO "%s: on-chip RAM at 0x%lx\n", + ncr_name(np), np->base2_ba); + + return 0; +} + + +#ifdef SCSI_NCR_DEBUG_NVRAM + +void __init ncr_display_Symbios_nvram(ncb_p np, Symbios_nvram *nvram) +{ + int i; + + /* display Symbios nvram host data */ + printk(KERN_DEBUG "%s: HOST ID=%d%s%s%s%s%s\n", + ncr_name(np), nvram->host_id & 0x0f, + (nvram->flags & SYMBIOS_SCAM_ENABLE) ? " SCAM" :"", + (nvram->flags & SYMBIOS_PARITY_ENABLE) ? " PARITY" :"", + (nvram->flags & SYMBIOS_VERBOSE_MSGS) ? " VERBOSE" :"", + (nvram->flags & SYMBIOS_CHS_MAPPING) ? " CHS_ALT" :"", + (nvram->flags1 & SYMBIOS_SCAN_HI_LO) ? " HI_LO" :""); + + /* display Symbios nvram drive data */ + for (i = 0 ; i < 15 ; i++) { + struct Symbios_target *tn = &nvram->target[i]; + printk(KERN_DEBUG "%s-%d:%s%s%s%s WIDTH=%d SYNC=%d TMO=%d\n", + ncr_name(np), i, + (tn->flags & SYMBIOS_DISCONNECT_ENABLE) ? " DISC" : "", + (tn->flags & SYMBIOS_SCAN_AT_BOOT_TIME) ? " SCAN_BOOT" : "", + (tn->flags & SYMBIOS_SCAN_LUNS) ? " SCAN_LUNS" : "", + (tn->flags & SYMBIOS_QUEUE_TAGS_ENABLED)? " TCQ" : "", + tn->bus_width, + tn->sync_period / 4, + tn->timeout); + } +} + +static u_char Tekram_boot_delay[7] __initdata = {3, 5, 10, 20, 30, 60, 120}; + +void __init ncr_display_Tekram_nvram(ncb_p np, Tekram_nvram *nvram) +{ + int i, tags, boot_delay; + char *rem; + + /* display Tekram nvram host data */ + tags = 2 << nvram->max_tags_index; + boot_delay = 0; + if (nvram->boot_delay_index < 6) + boot_delay = Tekram_boot_delay[nvram->boot_delay_index]; + switch((nvram->flags & TEKRAM_REMOVABLE_FLAGS) >> 6) { + default: + case 0: rem = ""; break; + case 1: rem = " REMOVABLE=boot device"; break; + case 2: rem = " REMOVABLE=all"; break; + } + + printk(KERN_DEBUG + "%s: HOST ID=%d%s%s%s%s%s%s%s%s%s BOOT DELAY=%d tags=%d\n", + ncr_name(np), nvram->host_id & 0x0f, + (nvram->flags1 & SYMBIOS_SCAM_ENABLE) ? " SCAM" :"", + (nvram->flags & TEKRAM_MORE_THAN_2_DRIVES) ? " >2DRIVES" :"", + (nvram->flags & TEKRAM_DRIVES_SUP_1GB) ? " >1GB" :"", + (nvram->flags & TEKRAM_RESET_ON_POWER_ON) ? " RESET" :"", + (nvram->flags & TEKRAM_ACTIVE_NEGATION) ? " ACT_NEG" :"", + (nvram->flags & TEKRAM_IMMEDIATE_SEEK) ? " IMM_SEEK" :"", + (nvram->flags & TEKRAM_SCAN_LUNS) ? " SCAN_LUNS" :"", + (nvram->flags1 & TEKRAM_F2_F6_ENABLED) ? " F2_F6" :"", + rem, boot_delay, tags); + + /* display Tekram nvram drive data */ + for (i = 0; i <= 15; i++) { + int sync, j; + struct Tekram_target *tn = &nvram->target[i]; + j = tn->sync_index & 0xf; + sync = Tekram_sync[j]; + printk(KERN_DEBUG "%s-%d:%s%s%s%s%s%s PERIOD=%d\n", + ncr_name(np), i, + (tn->flags & TEKRAM_PARITY_CHECK) ? " PARITY" : "", + (tn->flags & TEKRAM_SYNC_NEGO) ? " SYNC" : "", + (tn->flags & TEKRAM_DISCONNECT_ENABLE) ? " DISC" : "", + (tn->flags & TEKRAM_START_CMD) ? " START" : "", + (tn->flags & TEKRAM_TAGGED_COMMANDS) ? " TCQ" : "", + (tn->flags & TEKRAM_WIDE_NEGO) ? " WIDE" : "", + sync); + } +} +#endif /* SCSI_NCR_DEBUG_NVRAM */ + +/* +** Host attach and initialisations. +** +** Allocate host data and ncb structure. +** Request IO region and remap MMIO region. +** Do chip initialization. +** If all is OK, install interrupt handling and +** start the timer daemon. +*/ + +static int __init +ncr_attach (Scsi_Host_Template *tpnt, int unit, ncr_device *device) +{ + struct host_data *host_data; + ncb_p np = 0; + struct Scsi_Host *instance = 0; + u_long flags = 0; + ncr_nvram *nvram = device->nvram; + int i; + + printk(KERN_INFO NAME53C "%s-%d: rev 0x%x on pci bus %d device %d function %d " +#ifdef __sparc__ + "irq %s\n", +#else + "irq %d\n", +#endif + device->chip.name, unit, device->chip.revision_id, + device->slot.bus, (device->slot.device_fn & 0xf8) >> 3, + device->slot.device_fn & 7, +#ifdef __sparc__ + __irq_itoa(device->slot.irq)); +#else + device->slot.irq); +#endif + + /* + ** Allocate host_data structure + */ + if (!(instance = scsi_register(tpnt, sizeof(*host_data)))) + goto attach_error; + host_data = (struct host_data *) instance->hostdata; + + /* + ** Allocate the host control block. + */ + np = __m_calloc_dma(device->pdev, sizeof(struct ncb), "NCB"); + if (!np) + goto attach_error; + NCR_INIT_LOCK_NCB(np); + np->pdev = device->pdev; + np->p_ncb = vtobus(np); + host_data->ncb = np; + + /* + ** Store input informations in the host data structure. + */ + strncpy(np->chip_name, device->chip.name, sizeof(np->chip_name) - 1); + np->unit = unit; + np->verbose = driver_setup.verbose; + sprintf(np->inst_name, NAME53C "%s-%d", np->chip_name, np->unit); + np->device_id = device->chip.device_id; + np->revision_id = device->chip.revision_id; + np->bus = device->slot.bus; + np->device_fn = device->slot.device_fn; + np->features = device->chip.features; + np->clock_divn = device->chip.nr_divisor; + np->maxoffs = device->chip.offset_max; + np->maxburst = device->chip.burst_max; + np->myaddr = device->host_id; + + /* + ** Allocate the start queue. + */ + np->squeue = (ncrcmd *) + m_calloc_dma(sizeof(ncrcmd)*(MAX_START*2), "SQUEUE"); + if (!np->squeue) + goto attach_error; + np->p_squeue = vtobus(np->squeue); + + /* + ** Allocate the done queue. + */ + np->dqueue = (ncrcmd *) + m_calloc_dma(sizeof(ncrcmd)*(MAX_START*2), "DQUEUE"); + if (!np->dqueue) + goto attach_error; + + /* + ** Allocate the target bus address array. + */ + np->targtbl = (u_int32 *) m_calloc_dma(256, "TARGTBL"); + if (!np->targtbl) + goto attach_error; + + /* + ** Allocate SCRIPTS areas + */ + np->script0 = (struct script *) + m_calloc_dma(sizeof(struct script), "SCRIPT"); + if (!np->script0) + goto attach_error; + np->scripth0 = (struct scripth *) + m_calloc_dma(sizeof(struct scripth), "SCRIPTH"); + if (!np->scripth0) + goto attach_error; + + /* + ** Initialyze the CCB free queue and, + ** allocate some CCB. We need at least ONE. + */ + xpt_que_init(&np->free_ccbq); + xpt_que_init(&np->b0_ccbq); + if (!ncr_alloc_ccb(np)) + goto attach_error; + + /* + ** Initialize timer structure + ** + */ + init_timer(&np->timer); + np->timer.data = (unsigned long) np; + np->timer.function = sym53c8xx_timeout; + + /* + ** Try to map the controller chip to + ** virtual and physical memory. + */ + + np->base_ba = device->slot.base; + np->base_ws = (np->features & FE_IO256)? 256 : 128; + np->base2_ba = (np->features & FE_RAM)? device->slot.base_2 : 0; + +#ifndef SCSI_NCR_IOMAPPED + np->base_va = remap_pci_mem(np->base_ba, np->base_ws); + if (!np->base_va) { + printk(KERN_ERR "%s: can't map PCI MMIO region\n",ncr_name(np)); + goto attach_error; + } + else if (bootverbose > 1) + printk(KERN_INFO "%s: using memory mapped IO\n", ncr_name(np)); + + /* + ** Make the controller's registers available. + ** Now the INB INW INL OUTB OUTW OUTL macros + ** can be used safely. + */ + + np->reg = (struct ncr_reg *) np->base_va; + +#endif /* !defined SCSI_NCR_IOMAPPED */ + + /* + ** If on-chip RAM is used, make sure SCRIPTS isn't too large. + */ + if (np->base2_ba && sizeof(struct script) > 4096) { + printk(KERN_ERR "%s: script too large.\n", ncr_name(np)); + goto attach_error; + } + + /* + ** Try to map the controller chip into iospace. + */ + + if (device->slot.io_port) { + request_region(device->slot.io_port, np->base_ws, NAME53C8XX); + np->base_io = device->slot.io_port; + } + +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (nvram) { + switch(nvram->type) { + case SCSI_NCR_SYMBIOS_NVRAM: +#ifdef SCSI_NCR_DEBUG_NVRAM + ncr_display_Symbios_nvram(np, &nvram->data.Symbios); +#endif + break; + case SCSI_NCR_TEKRAM_NVRAM: +#ifdef SCSI_NCR_DEBUG_NVRAM + ncr_display_Tekram_nvram(np, &nvram->data.Tekram); +#endif + break; + default: + nvram = 0; +#ifdef SCSI_NCR_DEBUG_NVRAM + printk(KERN_DEBUG "%s: NVRAM: None or invalid data.\n", ncr_name(np)); +#endif + } + } +#endif + + /* + ** Save setting of some IO registers, so we will + ** be able to probe specific implementations. + */ + ncr_save_initial_setting (np); + + /* + ** Reset the chip now, since it has been reported + ** that SCSI clock calibration may not work properly + ** if the chip is currently active. + */ + ncr_chip_reset (np); + + /* + ** Do chip dependent initialization. + */ + (void) ncr_prepare_setting(np, nvram); + + /* + ** Check the PCI clock frequency if needed. + ** + ** Must be done after ncr_prepare_setting since it destroys + ** STEST1 that is used to probe for the clock multiplier. + ** + ** The range is currently [22688 - 45375 Khz], given + ** the values used by ncr_getclock(). + ** This calibration of the frequecy measurement + ** algorithm against the PCI clock frequency is only + ** performed if the driver has had to measure the SCSI + ** clock due to other heuristics not having been enough + ** to deduce the SCSI clock frequency. + ** + ** When the chip has been initialized correctly by the + ** SCSI BIOS, the driver deduces the presence of the + ** clock multiplier and the value of the SCSI clock from + ** initial values of IO registers, and therefore no + ** clock measurement is performed. + ** Normally the driver should never have to measure any + ** clock, unless the controller may use a 80 MHz clock + ** or has a clock multiplier and any of the following + ** condition is met: + ** + ** - No SCSI BIOS is present. + ** - SCSI BIOS did'nt enable the multiplier for some reason. + ** - User has disabled the controller from the SCSI BIOS. + ** - User booted the O/S from another O/S that did'nt enable + ** the multiplier for some reason. + ** + ** As a result, the driver may only have to measure some + ** frequency in very unusual situations. + ** + ** For this reality test against the PCI clock to really + ** protect against flaws in the udelay() calibration or + ** driver problem that affect the clock measurement + ** algorithm, the actual PCI clock frequency must be 33 MHz. + */ + i = np->pciclock_max ? ncr_getpciclock(np) : 0; + if (i && (i < np->pciclock_min || i > np->pciclock_max)) { + printk(KERN_ERR "%s: PCI clock (%u KHz) is out of range " + "[%u KHz - %u KHz].\n", + ncr_name(np), i, np->pciclock_min, np->pciclock_max); + goto attach_error; + } + + /* + ** Patch script to physical addresses + */ + ncr_script_fill (&script0, &scripth0); + + np->p_script = vtobus(np->script0); + np->p_scripth = vtobus(np->scripth0); + np->p_scripth0 = np->p_scripth; + + if (np->base2_ba) { + np->p_script = pcivtobus(np->base2_ba); + if (np->features & FE_RAM8K) { + np->base2_ws = 8192; + np->p_scripth = np->p_script + 4096; +#if BITS_PER_LONG > 32 + np->scr_ram_seg = cpu_to_scr(np->base2_ba >> 32); +#endif + } + else + np->base2_ws = 4096; +#ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + np->base2_va = remap_pci_mem(np->base2_ba, np->base2_ws); + if (!np->base2_va) { + printk(KERN_ERR "%s: can't map PCI MEMORY region\n", + ncr_name(np)); + goto attach_error; + } +#endif + } + + ncr_script_copy_and_bind (np, (ncrcmd *) &script0, (ncrcmd *) np->script0, sizeof(struct script)); + ncr_script_copy_and_bind (np, (ncrcmd *) &scripth0, (ncrcmd *) np->scripth0, sizeof(struct scripth)); + + /* + ** Patch some variables in SCRIPTS + */ + np->scripth0->pm0_data_addr[0] = + cpu_to_scr(NCB_SCRIPT_PHYS(np, pm0_data)); + np->scripth0->pm1_data_addr[0] = + cpu_to_scr(NCB_SCRIPT_PHYS(np, pm1_data)); + + /* + ** Patch if not Ultra 3 - Do not write to scntl4 + */ + if (np->features & FE_ULTRA3) { + np->script0->resel_scntl4[0] = cpu_to_scr(SCR_LOAD_REL (scntl4, 1)); + np->script0->resel_scntl4[1] = cpu_to_scr(offsetof(struct tcb, uval)); + } + + +#ifdef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + np->scripth0->script0_ba[0] = cpu_to_scr(vtobus(np->script0)); + np->scripth0->script0_ba64[0] = cpu_to_scr(vtobus(np->script0)); + np->scripth0->scripth0_ba64[0] = cpu_to_scr(vtobus(np->scripth0)); + np->scripth0->ram_seg64[0] = np->scr_ram_seg; +#endif + /* + ** Prepare the idle and invalid task actions. + */ + np->idletask.start = cpu_to_scr(NCB_SCRIPT_PHYS (np, idle)); + np->idletask.restart = cpu_to_scr(NCB_SCRIPTH_PHYS (np, bad_i_t_l)); + np->p_idletask = NCB_PHYS(np, idletask); + + np->notask.start = cpu_to_scr(NCB_SCRIPT_PHYS (np, idle)); + np->notask.restart = cpu_to_scr(NCB_SCRIPTH_PHYS (np, bad_i_t_l)); + np->p_notask = NCB_PHYS(np, notask); + + np->bad_i_t_l.start = cpu_to_scr(NCB_SCRIPT_PHYS (np, idle)); + np->bad_i_t_l.restart = cpu_to_scr(NCB_SCRIPTH_PHYS (np, bad_i_t_l)); + np->p_bad_i_t_l = NCB_PHYS(np, bad_i_t_l); + + np->bad_i_t_l_q.start = cpu_to_scr(NCB_SCRIPT_PHYS (np, idle)); + np->bad_i_t_l_q.restart = cpu_to_scr(NCB_SCRIPTH_PHYS (np,bad_i_t_l_q)); + np->p_bad_i_t_l_q = NCB_PHYS(np, bad_i_t_l_q); + + /* + ** Allocate and prepare the bad lun table. + */ + np->badluntbl = m_calloc_dma(256, "BADLUNTBL"); + if (!np->badluntbl) + goto attach_error; + + assert (offsetof(struct lcb, resel_task) == 0); + np->resel_badlun = cpu_to_scr(NCB_SCRIPTH_PHYS(np, resel_bad_lun)); + + for (i = 0 ; i < 64 ; i++) + np->badluntbl[i] = cpu_to_scr(NCB_PHYS(np, resel_badlun)); + + /* + ** Prepare the target bus address array. + */ + np->scripth0->targtbl[0] = cpu_to_scr(vtobus(np->targtbl)); + for (i = 0 ; i < MAX_TARGET ; i++) { + np->targtbl[i] = cpu_to_scr(NCB_PHYS(np, target[i])); + np->target[i].b_luntbl = cpu_to_scr(vtobus(np->badluntbl)); + np->target[i].b_lun0 = cpu_to_scr(NCB_PHYS(np, resel_badlun)); + } + + /* + ** Patch the script for LED support. + */ + + if (np->features & FE_LED0) { + np->script0->idle[0] = + cpu_to_scr(SCR_REG_REG(gpreg, SCR_OR, 0x01)); + np->script0->reselected[0] = + cpu_to_scr(SCR_REG_REG(gpreg, SCR_AND, 0xfe)); + np->script0->start[0] = + cpu_to_scr(SCR_REG_REG(gpreg, SCR_AND, 0xfe)); + } + + /* + ** Patch the script to provide an extra clock cycle on + ** data out phase - 53C1010_66MHz part only. + */ + if (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66){ + np->script0->datao_phase[0] = + cpu_to_scr(SCR_REG_REG(scntl4, SCR_OR, 0x0c)); + } + +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** If user does not want to use IMMEDIATE ARBITRATION + ** when we are reselected while attempting to arbitrate, + ** patch the SCRIPTS accordingly with a SCRIPT NO_OP. + */ + if (!(driver_setup.iarb & 1)) + np->script0->ungetjob[0] = cpu_to_scr(SCR_NO_OP); + /* + ** If user wants IARB to be set when we win arbitration + ** and have other jobs, compute the max number of consecutive + ** settings of IARB hint before we leave devices a chance to + ** arbitrate for reselection. + */ + np->iarb_max = (driver_setup.iarb >> 4); +#endif + + /* + ** DEL 472 - 53C896 Rev 1 - Part Number 609-0393055 - ITEM 5. + */ + if (np->device_id == PCI_DEVICE_ID_NCR_53C896 && + np->revision_id <= 0x1 && (np->features & FE_NOPM)) { + np->scatter = ncr_scatter_896R1; + np->script0->datai_phase[0] = cpu_to_scr(SCR_JUMP); + np->script0->datai_phase[1] = + cpu_to_scr(NCB_SCRIPTH_PHYS (np, tweak_pmj)); + np->script0->datao_phase[0] = cpu_to_scr(SCR_JUMP); + np->script0->datao_phase[1] = + cpu_to_scr(NCB_SCRIPTH_PHYS (np, tweak_pmj)); + } + else +#ifdef DEBUG_896R1 + np->scatter = ncr_scatter_896R1; +#else + np->scatter = ncr_scatter; +#endif + + /* + ** Reset chip. + ** We should use ncr_soft_reset(), but we donnot want to do + ** so, since we may not be safe if ABRT interrupt occurs due + ** to the BIOS or previous O/S having enable this interrupt. + ** + ** For C1010 need to set ABRT bit prior to SRST if SCRIPTs + ** are running. Not true in this case. + */ + ncr_chip_reset(np); + + /* + ** Now check the cache handling of the pci chipset. + */ + + if (ncr_snooptest (np)) { + printk (KERN_ERR "CACHE INCORRECTLY CONFIGURED.\n"); + goto attach_error; + }; + + /* + ** Install the interrupt handler. + ** If we synchonize the C code with SCRIPTS on interrupt, + ** we donnot want to share the INTR line at all. + */ + if (request_irq(device->slot.irq, sym53c8xx_intr, +#ifdef SCSI_NCR_PCIQ_SYNC_ON_INTR + ((driver_setup.irqm & 0x20) ? 0 : SA_INTERRUPT), +#else + ((driver_setup.irqm & 0x10) ? 0 : SA_SHIRQ) | + +#if 0 && LINUX_VERSION_CODE < LinuxVersionCode(2,2,0) + ((driver_setup.irqm & 0x20) ? 0 : SA_INTERRUPT), +#else + 0, +#endif +#endif + NAME53C8XX, np)) { + printk(KERN_ERR "%s: request irq %d failure\n", + ncr_name(np), device->slot.irq); + goto attach_error; + } + np->irq = device->slot.irq; + + /* + ** After SCSI devices have been opened, we cannot + ** reset the bus safely, so we do it here. + ** Interrupt handler does the real work. + ** Process the reset exception, + ** if interrupts are not enabled yet. + ** Then enable disconnects. + */ + NCR_LOCK_NCB(np, flags); + if (ncr_reset_scsi_bus(np, 0, driver_setup.settle_delay) != 0) { + printk(KERN_ERR "%s: FATAL ERROR: CHECK SCSI BUS - CABLES, TERMINATION, DEVICE POWER etc.!\n", ncr_name(np)); + + NCR_UNLOCK_NCB(np, flags); + goto attach_error; + } + ncr_exception (np); + + /* + ** The middle-level SCSI driver does not + ** wait for devices to settle. + ** Wait synchronously if more than 2 seconds. + */ + if (driver_setup.settle_delay > 2) { + printk(KERN_INFO "%s: waiting %d seconds for scsi devices to settle...\n", + ncr_name(np), driver_setup.settle_delay); + MDELAY (1000 * driver_setup.settle_delay); + } + + /* + ** start the timeout daemon + */ + np->lasttime=0; + ncr_timeout (np); + + /* + ** use SIMPLE TAG messages by default + */ +#ifdef SCSI_NCR_ALWAYS_SIMPLE_TAG + np->order = M_SIMPLE_TAG; +#endif + + /* + ** Done. + */ + if (!first_host) + first_host = instance; + + /* + ** Fill Linux host instance structure + ** and return success. + */ + instance->max_channel = 0; + instance->this_id = np->myaddr; + instance->max_id = np->maxwide ? 16 : 8; + instance->max_lun = MAX_LUN; +#ifndef SCSI_NCR_IOMAPPED +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,29) + instance->base = (unsigned long) np->reg; +#else + instance->base = (char *) np->reg; +#endif +#endif + instance->irq = np->irq; + instance->unique_id = np->base_io; + instance->io_port = np->base_io; + instance->n_io_port = np->base_ws; + instance->dma_channel = 0; + instance->cmd_per_lun = MAX_TAGS; + instance->can_queue = (MAX_START-4); + + np->check_integrity = 0; + +#ifdef SCSI_NCR_INTEGRITY_CHECKING + instance->check_integrity = 0; + +#ifdef SCSI_NCR_ENABLE_INTEGRITY_CHECK + if ( !(driver_setup.bus_check & 0x04) ) { + np->check_integrity = 1; + instance->check_integrity = 1; + } +#endif +#endif + + instance->select_queue_depths = sym53c8xx_select_queue_depths; + + NCR_UNLOCK_NCB(np, flags); + + /* + ** Now let the generic SCSI driver + ** look for the SCSI devices on the bus .. + */ + return 0; + +attach_error: + if (!instance) return -1; + printk(KERN_INFO "%s: giving up ...\n", ncr_name(np)); + if (np) + ncr_free_resources(np); + scsi_unregister(instance); + + return -1; + } + + +/* +** Free controller resources. +*/ +static void ncr_free_resources(ncb_p np) +{ + ccb_p cp; + tcb_p tp; + lcb_p lp; + int target, lun; + + if (np->irq) + free_irq(np->irq, np); + if (np->base_io) + release_region(np->base_io, np->base_ws); +#ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + if (np->base_va) + unmap_pci_mem(np->base_va, np->base_ws); + if (np->base2_va) + unmap_pci_mem(np->base2_va, np->base2_ws); +#endif + if (np->scripth0) + m_free_dma(np->scripth0, sizeof(struct scripth), "SCRIPTH"); + if (np->script0) + m_free_dma(np->script0, sizeof(struct script), "SCRIPT"); + if (np->squeue) + m_free_dma(np->squeue, sizeof(ncrcmd)*(MAX_START*2), "SQUEUE"); + if (np->dqueue) + m_free_dma(np->dqueue, sizeof(ncrcmd)*(MAX_START*2),"DQUEUE"); + + while ((cp = np->ccbc) != NULL) { + np->ccbc = cp->link_ccb; + m_free_dma(cp, sizeof(*cp), "CCB"); + } + + if (np->badluntbl) + m_free_dma(np->badluntbl, 256,"BADLUNTBL"); + + for (target = 0; target < MAX_TARGET ; target++) { + tp = &np->target[target]; + for (lun = 0 ; lun < MAX_LUN ; lun++) { + lp = ncr_lp(np, tp, lun); + if (!lp) + continue; + if (lp->tasktbl != &lp->tasktbl_0) + m_free_dma(lp->tasktbl, MAX_TASKS*4, "TASKTBL"); + if (lp->cb_tags) + m_free(lp->cb_tags, MAX_TAGS, "CB_TAGS"); + m_free_dma(lp, sizeof(*lp), "LCB"); + } +#if MAX_LUN > 1 + if (tp->lmp) + m_free(tp->lmp, MAX_LUN * sizeof(lcb_p), "LMP"); + if (tp->luntbl) + m_free_dma(tp->luntbl, 256, "LUNTBL"); +#endif + } + + if (np->targtbl) + m_free_dma(np->targtbl, 256, "TARGTBL"); + + m_free_dma(np, sizeof(*np), "NCB"); +} + + +/*========================================================== +** +** +** Done SCSI commands list management. +** +** We donnot enter the scsi_done() callback immediately +** after a command has been seen as completed but we +** insert it into a list which is flushed outside any kind +** of driver critical section. +** This allows to do minimal stuff under interrupt and +** inside critical sections and to also avoid locking up +** on recursive calls to driver entry points under SMP. +** In fact, the only kernel point which is entered by the +** driver with a driver lock set is get_free_pages(GFP_ATOMIC...) +** that shall not reenter the driver under any circumstance. +** +**========================================================== +*/ +static inline void ncr_queue_done_cmd(ncb_p np, Scsi_Cmnd *cmd) +{ + unmap_scsi_data(np, cmd); + cmd->host_scribble = (char *) np->done_list; + np->done_list = cmd; +} + +static inline void ncr_flush_done_cmds(Scsi_Cmnd *lcmd) +{ + Scsi_Cmnd *cmd; + + while (lcmd) { + cmd = lcmd; + lcmd = (Scsi_Cmnd *) cmd->host_scribble; + cmd->scsi_done(cmd); + } +} + +/*========================================================== +** +** +** Prepare the next negotiation message for integrity check, +** if needed. +** +** Fill in the part of message buffer that contains the +** negotiation and the nego_status field of the CCB. +** Returns the size of the message in bytes. +** +** If tp->ppr_negotiation is 1 and a M_REJECT occurs, then +** we disable ppr_negotiation. If the first ppr_negotiation is +** successful, set this flag to 2. +** +**========================================================== +*/ +#ifdef SCSI_NCR_INTEGRITY_CHECKING +static int ncr_ic_nego(ncb_p np, ccb_p cp, Scsi_Cmnd *cmd, u_char *msgptr) +{ + tcb_p tp = &np->target[cp->target]; + int msglen = 0; + int nego = 0; + u_char new_width, new_offset, new_period; + u_char no_increase; + + if (tp->ppr_negotiation == 1) /* PPR message successful */ + tp->ppr_negotiation = 2; + + if (tp->inq_done) { + + if (!tp->ic_maximums_set) { + tp->ic_maximums_set = 1; + + /* + * Check against target, host and user limits + */ + if ( (tp->inq_byte7 & INQ7_WIDE16) && + np->maxwide && tp->usrwide) + tp->ic_max_width = 1; + else + tp->ic_max_width = 0; + + + if ((tp->inq_byte7 & INQ7_SYNC) && tp->maxoffs) + tp->ic_min_sync = (tp->minsync < np->minsync) ? + np->minsync : tp->minsync; + else + tp->ic_min_sync = 255; + + tp->period = 1; + tp->widedone = 1; + + /* + * Enable PPR negotiation - only if Ultra3 support + * is accessible. + */ + +#if 0 + if (tp->ic_max_width && (tp->ic_min_sync != 255 )) + tp->ppr_negotiation = 1; +#endif + tp->ppr_negotiation = 0; + if (np->features & FE_ULTRA3) { + if (tp->ic_max_width && (tp->ic_min_sync == 0x09)) + tp->ppr_negotiation = 1; + } + + if (!tp->ppr_negotiation) + cmd->ic_nego &= ~NS_PPR; + } + + if (DEBUG_FLAGS & DEBUG_IC) { + printk("%s: cmd->ic_nego %d, 1st byte 0x%2X\n", + ncr_name(np), cmd->ic_nego, cmd->cmnd[0]); + } + + /* Previous command recorded a parity or an initiator + * detected error condition. Force bus to narrow for this + * target. Clear flag. Negotation on request sense. + * Note: kernel forces 2 bus resets :o( but clears itself out. + * Minor bug? in scsi_obsolete.c (ugly) + */ + if (np->check_integ_par) { + printk("%s: Parity Error. Target set to narrow.\n", + ncr_name(np)); + tp->ic_max_width = 0; + tp->widedone = tp->period = 0; + } + + /* Initializing: + * If ic_nego == NS_PPR, we are in the initial test for + * PPR messaging support. If driver flag is clear, then + * either we don't support PPR nego (narrow or async device) + * or this is the second TUR and we have had a M. REJECT + * or unexpected disconnect on the first PPR negotiation. + * Do not negotiate, reset nego flags (in case a reset has + * occurred), clear ic_nego and return. + * General case: Kernel will clear flag on a fallback. + * Do only SDTR or WDTR in the future. + */ + if (!tp->ppr_negotiation && (cmd->ic_nego == NS_PPR )) { + tp->ppr_negotiation = 0; + cmd->ic_nego &= ~NS_PPR; + tp->widedone = tp->period = 1; + return msglen; + } + else if (( tp->ppr_negotiation && !(cmd->ic_nego & NS_PPR )) || + (!tp->ppr_negotiation && (cmd->ic_nego & NS_PPR )) ) { + tp->ppr_negotiation = 0; + cmd->ic_nego &= ~NS_PPR; + } + + /* + * Always check the PPR nego. flag bit if ppr_negotiation + * is set. If the ic_nego PPR bit is clear, + * there must have been a fallback. Do only + * WDTR / SDTR in the future. + */ + if ((tp->ppr_negotiation) && (!(cmd->ic_nego & NS_PPR))) + tp->ppr_negotiation = 0; + + /* In case of a bus reset, ncr_negotiate will reset + * the flags tp->widedone and tp->period to 0, forcing + * a new negotiation. Do WDTR then SDTR. If PPR, do both. + * Do NOT increase the period. It is possible for the Scsi_Cmnd + * flags to be set to increase the period when a bus reset + * occurs - we don't want to change anything. + */ + + no_increase = 0; + + if (tp->ppr_negotiation && (!tp->widedone) && (!tp->period) ) { + cmd->ic_nego = NS_PPR; + tp->widedone = tp->period = 1; + no_increase = 1; + } + else if (!tp->widedone) { + cmd->ic_nego = NS_WIDE; + tp->widedone = 1; + no_increase = 1; + } + else if (!tp->period) { + cmd->ic_nego = NS_SYNC; + tp->period = 1; + no_increase = 1; + } + + new_width = cmd->ic_nego_width & tp->ic_max_width; + + switch (cmd->ic_nego_sync) { + case 2: /* increase the period */ + if (!no_increase) { + if (tp->ic_min_sync <= 0x09) + tp->ic_min_sync = 0x0A; + else if (tp->ic_min_sync <= 0x0A) + tp->ic_min_sync = 0x0C; + else if (tp->ic_min_sync <= 0x0C) + tp->ic_min_sync = 0x19; + else if (tp->ic_min_sync <= 0x19) + tp->ic_min_sync *= 2; + else { + tp->ic_min_sync = 255; + cmd->ic_nego_sync = 0; + tp->maxoffs = 0; + } + } + new_period = tp->maxoffs?tp->ic_min_sync:0; + new_offset = tp->maxoffs; + break; + + case 1: /* nego. to maximum */ + new_period = tp->maxoffs?tp->ic_min_sync:0; + new_offset = tp->maxoffs; + break; + + case 0: /* nego to async */ + default: + new_period = 0; + new_offset = 0; + break; + }; + + + nego = NS_NOCHANGE; + if (tp->ppr_negotiation) { + u_char options_byte = 0; + + /* + ** Must make sure data is consistent. + ** If period is 9 and sync, must be wide and DT bit set. + ** else period must be larger. If the width is 0, + ** reset bus to wide but increase the period to 0x0A. + ** Note: The strange else clause is due to the integrity check. + ** If fails at 0x09, wide, the I.C. code will redo at the same + ** speed but a narrow bus. The driver must take care of slowing + ** the bus speed down. + ** + ** The maximum offset in ST mode is 31, in DT mode 62 (1010/1010_66 only) + */ + if ( (new_period==0x09) && new_offset) { + if (new_width) + options_byte = 0x02; + else { + tp->ic_min_sync = 0x0A; + new_period = 0x0A; + cmd->ic_nego_width = 1; + new_width = 1; + new_offset &= 0x1f; + } + } + else if (new_period > 0x09) + new_offset &= 0x1f; + + nego = NS_PPR; + + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 6; + msgptr[msglen++] = M_X_PPR_REQ; + msgptr[msglen++] = new_period; + msgptr[msglen++] = 0; + msgptr[msglen++] = new_offset; + msgptr[msglen++] = new_width; + msgptr[msglen++] = options_byte; + + } + else { + switch (cmd->ic_nego & ~NS_PPR) { + case NS_WIDE: + /* + ** WDTR negotiation on if device supports + ** wide or if wide device forced narrow + ** due to a parity error. + */ + + cmd->ic_nego_width &= tp->ic_max_width; + + if (tp->ic_max_width | np->check_integ_par) { + nego = NS_WIDE; + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 2; + msgptr[msglen++] = M_X_WIDE_REQ; + msgptr[msglen++] = new_width; + } + break; + + case NS_SYNC: + /* + ** negotiate synchronous transfers + ** Target must support sync transfers. + ** Min. period = 0x0A, maximum offset of 31=0x1f. + */ + + if (tp->inq_byte7 & INQ7_SYNC) { + + if (new_offset && (new_period < 0x0A)) { + tp->ic_min_sync = 0x0A; + new_period = 0x0A; + } + nego = NS_SYNC; + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 3; + msgptr[msglen++] = M_X_SYNC_REQ; + msgptr[msglen++] = new_period; + msgptr[msglen++] = new_offset & 0x1f; + } + else + cmd->ic_nego_sync = 0; + break; + + case NS_NOCHANGE: + break; + } + } + + }; + + cp->nego_status = nego; + np->check_integ_par = 0; + + if (nego) { + tp->nego_cp = cp; + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, nego == NS_WIDE ? + "wide/narrow msgout": + (nego == NS_SYNC ? "sync/async msgout" : "ppr msgout"), + msgptr); + }; + }; + + return msglen; +} +#endif /* SCSI_NCR_INTEGRITY_CHECKING */ + +/*========================================================== +** +** +** Prepare the next negotiation message if needed. +** +** Fill in the part of message buffer that contains the +** negotiation and the nego_status field of the CCB. +** Returns the size of the message in bytes. +** +** +**========================================================== +*/ + + +static int ncr_prepare_nego(ncb_p np, ccb_p cp, u_char *msgptr) +{ + tcb_p tp = &np->target[cp->target]; + int msglen = 0; + int nego = 0; + u_char width, offset, factor, last_byte; + + if (!np->check_integrity) { + /* If integrity checking disabled, enable PPR messaging + * if device supports wide, sync and ultra 3 + */ + if (tp->ppr_negotiation == 1) /* PPR message successful */ + tp->ppr_negotiation = 2; + + if ((tp->inq_done) && (!tp->ic_maximums_set)) { + tp->ic_maximums_set = 1; + + /* + * Issue PPR only if board is capable + * and set-up for Ultra3 transfers. + */ + tp->ppr_negotiation = 0; + if ( (np->features & FE_ULTRA3) && + (tp->usrwide) && (tp->maxoffs) && + (tp->minsync == 0x09) ) + tp->ppr_negotiation = 1; + } + } + + if (tp->inq_done) { + /* + * Get the current width, offset and period + */ + ncr_get_xfer_info( np, tp, &factor, + &offset, &width); + + /* + ** negotiate wide transfers ? + */ + + if (!tp->widedone) { + if (tp->inq_byte7 & INQ7_WIDE16) { + if (tp->ppr_negotiation) + nego = NS_PPR; + else + nego = NS_WIDE; + + width = tp->usrwide; +#ifdef SCSI_NCR_INTEGRITY_CHECKING + if (tp->ic_done) + width &= tp->ic_max_width; +#endif + } else + tp->widedone=1; + + }; + + /* + ** negotiate synchronous transfers? + */ + + if ((nego != NS_WIDE) && !tp->period) { + if (tp->inq_byte7 & INQ7_SYNC) { + if (tp->ppr_negotiation) + nego = NS_PPR; + else + nego = NS_SYNC; + + /* Check for async flag */ + if (tp->maxoffs == 0) { + offset = 0; + factor = 0; + } + else { + offset = tp->maxoffs; + factor = tp->minsync; +#ifdef SCSI_NCR_INTEGRITY_CHECKING + if ((tp->ic_done) && + (factor < tp->ic_min_sync)) + factor = tp->ic_min_sync; +#endif + } + + } else { + offset = 0; + factor = 0; + tp->period =0xffff; + PRINT_TARGET(np, cp->target); + printk ("target did not report SYNC.\n"); + }; + }; + }; + + switch (nego) { + case NS_PPR: + /* + ** Must make sure data is consistent. + ** If period is 9 and sync, must be wide and DT bit set + ** else period must be larger. + ** Maximum offset is 31=0x1f is ST mode, 62 if DT mode + */ + last_byte = 0; + if ( (factor==9) && offset) { + if (!width) { + factor = 0x0A; + offset &= 0x1f; + } + else + last_byte = 0x02; + } + else if (factor > 0x09) + offset &= 0x1f; + + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 6; + msgptr[msglen++] = M_X_PPR_REQ; + msgptr[msglen++] = factor; + msgptr[msglen++] = 0; + msgptr[msglen++] = offset; + msgptr[msglen++] = width; + msgptr[msglen++] = last_byte; + break; + case NS_SYNC: + /* + ** Never negotiate faster than Ultra 2 (25ns periods) + */ + if (offset && (factor < 0x0A)) { + factor = 0x0A; + tp->minsync = 0x0A; + } + + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 3; + msgptr[msglen++] = M_X_SYNC_REQ; + msgptr[msglen++] = factor; + msgptr[msglen++] = offset & 0x1f; + break; + case NS_WIDE: + msgptr[msglen++] = M_EXTENDED; + msgptr[msglen++] = 2; + msgptr[msglen++] = M_X_WIDE_REQ; + msgptr[msglen++] = width; + break; + }; + + cp->nego_status = nego; + + if (nego) { + tp->nego_cp = cp; + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, nego == NS_WIDE ? + "wide msgout": + (nego == NS_SYNC ? "sync msgout" : "ppr msgout"), + msgptr); + }; + }; + + return msglen; +} + +/*========================================================== +** +** +** Start execution of a SCSI command. +** This is called from the generic SCSI driver. +** +** +**========================================================== +*/ +static int ncr_queue_command (ncb_p np, Scsi_Cmnd *cmd) +{ +/* Scsi_Device *device = cmd->device; */ + tcb_p tp = &np->target[cmd->target]; + lcb_p lp = ncr_lp(np, tp, cmd->lun); + ccb_p cp; + + u_char idmsg, *msgptr; + u_int msglen; + int direction; + u_int32 lastp, goalp; + + /*--------------------------------------------- + ** + ** Some shortcuts ... + ** + **--------------------------------------------- + */ + if ((cmd->target == np->myaddr ) || + (cmd->target >= MAX_TARGET) || + (cmd->lun >= MAX_LUN )) { + return(DID_BAD_TARGET); + } + + /*--------------------------------------------- + ** + ** Complete the 1st TEST UNIT READY command + ** with error condition if the device is + ** flagged NOSCAN, in order to speed up + ** the boot. + ** + **--------------------------------------------- + */ + if (cmd->cmnd[0] == 0 && (tp->usrflag & UF_NOSCAN)) { + tp->usrflag &= ~UF_NOSCAN; + return DID_BAD_TARGET; + } + + if (DEBUG_FLAGS & DEBUG_TINY) { + PRINT_ADDR(cmd); + printk ("CMD=%x ", cmd->cmnd[0]); + } + + /*--------------------------------------------------- + ** + ** Assign a ccb / bind cmd. + ** If resetting, shorten settle_time if necessary + ** in order to avoid spurious timeouts. + ** If resetting or no free ccb, + ** insert cmd into the waiting list. + ** + **---------------------------------------------------- + */ + if (np->settle_time && cmd->timeout_per_command >= HZ) { + u_long tlimit = ktime_get(cmd->timeout_per_command - HZ); + if (ktime_dif(np->settle_time, tlimit) > 0) + np->settle_time = tlimit; + } + + if (np->settle_time || !(cp=ncr_get_ccb (np, cmd->target, cmd->lun))) { + insert_into_waiting_list(np, cmd); + return(DID_OK); + } + cp->cmd = cmd; + + /*--------------------------------------------------- + ** + ** Enable tagged queue if asked by scsi ioctl + ** + **---------------------------------------------------- + */ +#if 0 /* This stuff was only usefull for linux-1.2.13 */ + if (lp && !lp->numtags && cmd->device && cmd->device->tagged_queue) { + lp->numtags = tp->usrtags; + ncr_setup_tags (np, cp->target, cp->lun); + } +#endif + + /*---------------------------------------------------- + ** + ** Build the identify / tag / sdtr message + ** + **---------------------------------------------------- + */ + + idmsg = M_IDENTIFY | cp->lun; + + if (cp ->tag != NO_TAG || (lp && !(tp->usrflag & UF_NODISC))) + idmsg |= 0x40; + + msgptr = cp->scsi_smsg; + msglen = 0; + msgptr[msglen++] = idmsg; + + if (cp->tag != NO_TAG) { + char order = np->order; + + /* + ** Force ordered tag if necessary to avoid timeouts + ** and to preserve interactivity. + */ + if (lp && ktime_exp(lp->tags_stime)) { + lp->tags_si = !(lp->tags_si); + if (lp->tags_sum[lp->tags_si]) { + order = M_ORDERED_TAG; + if ((DEBUG_FLAGS & DEBUG_TAGS)||bootverbose>0){ + PRINT_ADDR(cmd); + printk("ordered tag forced.\n"); + } + } + lp->tags_stime = ktime_get(3*HZ); + } + + if (order == 0) { + /* + ** Ordered write ops, unordered read ops. + */ + switch (cmd->cmnd[0]) { + case 0x08: /* READ_SMALL (6) */ + case 0x28: /* READ_BIG (10) */ + case 0xa8: /* READ_HUGE (12) */ + order = M_SIMPLE_TAG; + break; + default: + order = M_ORDERED_TAG; + } + } + msgptr[msglen++] = order; + /* + ** For less than 128 tags, actual tags are numbered + ** 1,3,5,..2*MAXTAGS+1,since we may have to deal + ** with devices that have problems with #TAG 0 or too + ** great #TAG numbers. For more tags (up to 256), + ** we use directly our tag number. + */ +#if MAX_TASKS > (512/4) + msgptr[msglen++] = cp->tag; +#else + msgptr[msglen++] = (cp->tag << 1) + 1; +#endif + } + + cp->host_flags = 0; + + /*---------------------------------------------------- + ** + ** Build the data descriptors + ** + **---------------------------------------------------- + */ + + direction = scsi_data_direction(cmd); + if (direction != SCSI_DATA_NONE) { + cp->segments = np->scatter (np, cp, cp->cmd); + if (cp->segments < 0) { + ncr_free_ccb(np, cp); + return(DID_ERROR); + } + } + else { + cp->data_len = 0; + cp->segments = 0; + } + + /*--------------------------------------------------- + ** + ** negotiation required? + ** + ** (nego_status is filled by ncr_prepare_nego()) + ** + **--------------------------------------------------- + */ + + cp->nego_status = 0; + +#ifdef SCSI_NCR_INTEGRITY_CHECKING + if ((np->check_integrity && tp->ic_done) || !np->check_integrity) { + if ((!tp->widedone || !tp->period) && !tp->nego_cp && lp) { + msglen += ncr_prepare_nego (np, cp, msgptr + msglen); + } + } + else if (np->check_integrity && (cmd->ic_in_progress)) { + msglen += ncr_ic_nego (np, cp, cmd, msgptr + msglen); + } + else if (np->check_integrity && cmd->ic_complete) { + u_long current_period; + u_char current_offset, current_width, current_factor; + + ncr_get_xfer_info (np, tp, ¤t_factor, + ¤t_offset, ¤t_width); + + tp->ic_max_width = current_width; + tp->ic_min_sync = current_factor; + + if (current_factor == 9) current_period = 125; + else if (current_factor == 10) current_period = 250; + else if (current_factor == 11) current_period = 303; + else if (current_factor == 12) current_period = 500; + else current_period = current_factor * 40; + + /* + * Negotiation for this target is complete. Update flags. + */ + tp->period = current_period; + tp->widedone = 1; + tp->ic_done = 1; + + printk("%s: Integrity Check Complete: \n", ncr_name(np)); + + printk("%s: %s %s SCSI", ncr_name(np), + current_offset?"SYNC":"ASYNC", + tp->ic_max_width?"WIDE":"NARROW"); + if (current_offset) { + u_long mbs = 10000 * (tp->ic_max_width + 1); + + printk(" %d.%d MB/s", + (int) (mbs / current_period), (int) (mbs % current_period)); + + printk(" (%d ns, %d offset)\n", + (int) current_period/10, current_offset); + } + else + printk(" %d MB/s. \n ", (tp->ic_max_width+1)*5); + } +#else + if ((!tp->widedone || !tp->period) && !tp->nego_cp && lp) { + msglen += ncr_prepare_nego (np, cp, msgptr + msglen); + } +#endif /* SCSI_NCR_INTEGRITY_CHECKING */ + + + /*---------------------------------------------------- + ** + ** Determine xfer direction. + ** + **---------------------------------------------------- + */ + if (!cp->data_len) + direction = SCSI_DATA_NONE; + + /* + ** If data direction is UNKNOWN, speculate DATA_READ + ** but prepare alternate pointers for WRITE in case + ** of our speculation will be just wrong. + ** SCRIPTS will swap values if needed. + */ + switch(direction) { + case SCSI_DATA_UNKNOWN: + case SCSI_DATA_WRITE: + goalp = NCB_SCRIPT_PHYS (np, data_out2) + 8; + lastp = goalp - 8 - (cp->segments * (SCR_SG_SIZE*4)); + if (direction != SCSI_DATA_UNKNOWN) + break; + cp->phys.header.wgoalp = cpu_to_scr(goalp); + cp->phys.header.wlastp = cpu_to_scr(lastp); + /* fall through */ + case SCSI_DATA_READ: + cp->host_flags |= HF_DATA_IN; + goalp = NCB_SCRIPT_PHYS (np, data_in2) + 8; + lastp = goalp - 8 - (cp->segments * (SCR_SG_SIZE*4)); + break; + default: + case SCSI_DATA_NONE: + lastp = goalp = NCB_SCRIPTH_PHYS (np, no_data); + break; + } + + /* + ** Set all pointers values needed by SCRIPTS. + ** If direction is unknown, start at data_io. + */ + cp->phys.header.lastp = cpu_to_scr(lastp); + cp->phys.header.goalp = cpu_to_scr(goalp); + + if (direction == SCSI_DATA_UNKNOWN) + cp->phys.header.savep = + cpu_to_scr(NCB_SCRIPTH_PHYS (np, data_io)); + else + cp->phys.header.savep= cpu_to_scr(lastp); + + /* + ** Save the initial data pointer in order to be able + ** to redo the command. + ** We also have to save the initial lastp, since it + ** will be changed to DATA_IO if we don't know the data + ** direction and the device completes the command with + ** QUEUE FULL status (without entering the data phase). + */ + cp->startp = cp->phys.header.savep; + cp->lastp0 = cp->phys.header.lastp; + + /*---------------------------------------------------- + ** + ** fill in ccb + ** + **---------------------------------------------------- + ** + ** + ** physical -> virtual backlink + ** Generic SCSI command + */ + + /* + ** Startqueue + */ + cp->phys.header.go.start = cpu_to_scr(NCB_SCRIPT_PHYS (np,select)); + cp->phys.header.go.restart = cpu_to_scr(NCB_SCRIPT_PHYS (np,resel_dsa)); + /* + ** select + */ + cp->phys.select.sel_id = cp->target; + cp->phys.select.sel_scntl3 = tp->wval; + cp->phys.select.sel_sxfer = tp->sval; + cp->phys.select.sel_scntl4 = tp->uval; + /* + ** message + */ + cp->phys.smsg.addr = cpu_to_scr(CCB_PHYS (cp, scsi_smsg)); + cp->phys.smsg.size = cpu_to_scr(msglen); + + /* + ** command + */ + memcpy(cp->cdb_buf, cmd->cmnd, MIN(cmd->cmd_len, sizeof(cp->cdb_buf))); + cp->phys.cmd.addr = cpu_to_scr(CCB_PHYS (cp, cdb_buf[0])); + cp->phys.cmd.size = cpu_to_scr(cmd->cmd_len); + + /* + ** status + */ + cp->actualquirks = tp->quirks; + cp->host_status = cp->nego_status ? HS_NEGOTIATE : HS_BUSY; + cp->scsi_status = S_ILLEGAL; + cp->xerr_status = 0; + cp->extra_bytes = 0; + + /* + ** extreme data pointer. + ** shall be positive, so -1 is lower than lowest.:) + */ + cp->ext_sg = -1; + cp->ext_ofs = 0; + + /*---------------------------------------------------- + ** + ** Critical region: start this job. + ** + **---------------------------------------------------- + */ + + /* + ** activate this job. + */ + + /* + ** insert next CCBs into start queue. + ** 2 max at a time is enough to flush the CCB wait queue. + */ + if (lp) + ncr_start_next_ccb(np, lp, 2); + else + ncr_put_start_queue(np, cp); + + /* + ** Command is successfully queued. + */ + + return(DID_OK); +} + + +/*========================================================== +** +** +** Insert a CCB into the start queue and wake up the +** SCRIPTS processor. +** +** +**========================================================== +*/ + +static void ncr_start_next_ccb(ncb_p np, lcb_p lp, int maxn) +{ + XPT_QUEHEAD *qp; + ccb_p cp; + + while (maxn-- && lp->queuedccbs < lp->queuedepth) { + qp = xpt_remque_head(&lp->wait_ccbq); + if (!qp) + break; + ++lp->queuedccbs; + cp = xpt_que_entry(qp, struct ccb, link_ccbq); + xpt_insque_tail(qp, &lp->busy_ccbq); + lp->tasktbl[cp->tag == NO_TAG ? 0 : cp->tag] = + cpu_to_scr(cp->p_ccb); + ncr_put_start_queue(np, cp); + } +} + +static void ncr_put_start_queue(ncb_p np, ccb_p cp) +{ + u_short qidx; + +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** If the previously queued CCB is not yet done, + ** set the IARB hint. The SCRIPTS will go with IARB + ** for this job when starting the previous one. + ** We leave devices a chance to win arbitration by + ** not using more than 'iarb_max' consecutive + ** immediate arbitrations. + */ + if (np->last_cp && np->iarb_count < np->iarb_max) { + np->last_cp->host_flags |= HF_HINT_IARB; + ++np->iarb_count; + } + else + np->iarb_count = 0; + np->last_cp = cp; +#endif + + /* + ** insert into start queue. + */ + qidx = np->squeueput + 2; + if (qidx >= MAX_START*2) qidx = 0; + + np->squeue [qidx] = cpu_to_scr(np->p_idletask); + MEMORY_BARRIER(); + np->squeue [np->squeueput] = cpu_to_scr(cp->p_ccb); + + np->squeueput = qidx; + cp->queued = 1; + + if (DEBUG_FLAGS & DEBUG_QUEUE) + printk ("%s: queuepos=%d.\n", ncr_name (np), np->squeueput); + + /* + ** Script processor may be waiting for reselect. + ** Wake it up. + */ + MEMORY_BARRIER(); + OUTB (nc_istat, SIGP|np->istat_sem); +} + + +/*========================================================== +** +** Soft reset the chip. +** +** Some 896 and 876 chip revisions may hang-up if we set +** the SRST (soft reset) bit at the wrong time when SCRIPTS +** are running. +** So, we need to abort the current operation prior to +** soft resetting the chip. +** +**========================================================== +*/ + +static void ncr_chip_reset (ncb_p np) +{ + OUTB (nc_istat, SRST); + UDELAY (10); + OUTB (nc_istat, 0); +} + +static void ncr_soft_reset(ncb_p np) +{ + u_char istat; + int i; + + OUTB (nc_istat, CABRT); + for (i = 1000000 ; i ; --i) { + istat = INB (nc_istat); + if (istat & SIP) { + INW (nc_sist); + continue; + } + if (istat & DIP) { + OUTB (nc_istat, 0); + INB (nc_dstat); + break; + } + } + if (!i) + printk("%s: unable to abort current chip operation.\n", + ncr_name(np)); + ncr_chip_reset(np); +} + +/*========================================================== +** +** +** Start reset process. +** The interrupt handler will reinitialize the chip. +** The timeout handler will wait for settle_time before +** clearing it and so resuming command processing. +** +** +**========================================================== +*/ +static void ncr_start_reset(ncb_p np) +{ + (void) ncr_reset_scsi_bus(np, 1, driver_setup.settle_delay); +} + +static int ncr_reset_scsi_bus(ncb_p np, int enab_int, int settle_delay) +{ + u_int32 term; + int retv = 0; + + np->settle_time = ktime_get(settle_delay * HZ); + + if (bootverbose > 1) + printk("%s: resetting, " + "command processing suspended for %d seconds\n", + ncr_name(np), settle_delay); + + ncr_soft_reset(np); /* Soft reset the chip */ + UDELAY (2000); /* The 895/6 need time for the bus mode to settle */ + if (enab_int) + OUTW (nc_sien, RST); + /* + ** Enable Tolerant, reset IRQD if present and + ** properly set IRQ mode, prior to resetting the bus. + */ + OUTB (nc_stest3, TE); + OUTB (nc_dcntl, (np->rv_dcntl & IRQM)); + OUTB (nc_scntl1, CRST); + UDELAY (200); + + if (!driver_setup.bus_check) + goto out; + /* + ** Check for no terminators or SCSI bus shorts to ground. + ** Read SCSI data bus, data parity bits and control signals. + ** We are expecting RESET to be TRUE and other signals to be + ** FALSE. + */ + term = INB(nc_sstat0); + term = ((term & 2) << 7) + ((term & 1) << 17); /* rst sdp0 */ + term |= ((INB(nc_sstat2) & 0x01) << 26) | /* sdp1 */ + ((INW(nc_sbdl) & 0xff) << 9) | /* d7-0 */ + ((INW(nc_sbdl) & 0xff00) << 10) | /* d15-8 */ + INB(nc_sbcl); /* req ack bsy sel atn msg cd io */ + + if (!(np->features & FE_WIDE)) + term &= 0x3ffff; + + if (term != (2<<7)) { + printk("%s: suspicious SCSI data while resetting the BUS.\n", + ncr_name(np)); + printk("%s: %sdp0,d7-0,rst,req,ack,bsy,sel,atn,msg,c/d,i/o = " + "0x%lx, expecting 0x%lx\n", + ncr_name(np), + (np->features & FE_WIDE) ? "dp1,d15-8," : "", + (u_long)term, (u_long)(2<<7)); + if (driver_setup.bus_check == 1) + retv = 1; + } +out: + OUTB (nc_scntl1, 0); + return retv; +} + +/*========================================================== +** +** +** Reset the SCSI BUS. +** This is called from the generic SCSI driver. +** +** +**========================================================== +*/ +static int ncr_reset_bus (ncb_p np, Scsi_Cmnd *cmd, int sync_reset) +{ +/* Scsi_Device *device = cmd->device; */ + ccb_p cp; + int found; + +/* + * Return immediately if reset is in progress. + */ + if (np->settle_time) { + return SCSI_RESET_PUNT; + } +/* + * Start the reset process. + * The script processor is then assumed to be stopped. + * Commands will now be queued in the waiting list until a settle + * delay of 2 seconds will be completed. + */ + ncr_start_reset(np); +/* + * First, look in the wakeup list + */ + for (found=0, cp=np->ccbc; cp; cp=cp->link_ccb) { + /* + ** look for the ccb of this command. + */ + if (cp->host_status == HS_IDLE) continue; + if (cp->cmd == cmd) { + found = 1; + break; + } + } +/* + * Then, look in the waiting list + */ + if (!found && retrieve_from_waiting_list(0, np, cmd)) + found = 1; +/* + * Wake-up all awaiting commands with DID_RESET. + */ + reset_waiting_list(np); +/* + * Wake-up all pending commands with HS_RESET -> DID_RESET. + */ + ncr_wakeup(np, HS_RESET); +/* + * If the involved command was not in a driver queue, and the + * scsi driver told us reset is synchronous, and the command is not + * currently in the waiting list, complete it with DID_RESET status, + * in order to keep it alive. + */ + if (!found && sync_reset && !retrieve_from_waiting_list(0, np, cmd)) { + SetScsiResult(cmd, DID_RESET, 0); + ncr_queue_done_cmd(np, cmd); + } + + return SCSI_RESET_SUCCESS; +} + +/*========================================================== +** +** +** Abort an SCSI command. +** This is called from the generic SCSI driver. +** +** +**========================================================== +*/ +static int ncr_abort_command (ncb_p np, Scsi_Cmnd *cmd) +{ +/* Scsi_Device *device = cmd->device; */ + ccb_p cp; + +/* + * First, look for the scsi command in the waiting list + */ + if (remove_from_waiting_list(np, cmd)) { + SetScsiAbortResult(cmd); + ncr_queue_done_cmd(np, cmd); + return SCSI_ABORT_SUCCESS; + } + +/* + * Then, look in the wakeup list + */ + for (cp=np->ccbc; cp; cp=cp->link_ccb) { + /* + ** look for the ccb of this command. + */ + if (cp->host_status == HS_IDLE) continue; + if (cp->cmd == cmd) + break; + } + + if (!cp) { + return SCSI_ABORT_NOT_RUNNING; + } + + /* + ** Keep track we have to abort this job. + */ + cp->to_abort = 1; + + /* + ** Tell the SCRIPTS processor to stop + ** and synchronize with us. + */ + np->istat_sem = SEM; + + /* + ** If there are no requests, the script + ** processor will sleep on SEL_WAIT_RESEL. + ** Let's wake it up, since it may have to work. + */ + OUTB (nc_istat, SIGP|SEM); + + /* + ** Tell user we are working for him. + */ + return SCSI_ABORT_PENDING; +} + +/*========================================================== +** +** Linux release module stuff. +** +** Called before unloading the module +** Detach the host. +** We have to free resources and halt the NCR chip +** +**========================================================== +*/ + +#ifdef MODULE +static int ncr_detach(ncb_p np) +{ + int i; + + printk("%s: detaching ...\n", ncr_name(np)); + +/* +** Stop the ncr_timeout process +** Set release_stage to 1 and wait that ncr_timeout() set it to 2. +*/ + np->release_stage = 1; + for (i = 50 ; i && np->release_stage != 2 ; i--) MDELAY (100); + if (np->release_stage != 2) + printk("%s: the timer seems to be already stopped\n", + ncr_name(np)); + else np->release_stage = 2; + +/* +** Reset NCR chip. +** We should use ncr_soft_reset(), but we donnot want to do +** so, since we may not be safe if interrupts occur. +*/ + + printk("%s: resetting chip\n", ncr_name(np)); + ncr_chip_reset(np); + +/* +** Restore bios setting for automatic clock detection. +*/ + OUTB(nc_dmode, np->sv_dmode); + OUTB(nc_dcntl, np->sv_dcntl); + OUTB(nc_ctest3, np->sv_ctest3); + OUTB(nc_ctest4, np->sv_ctest4); + OUTB(nc_ctest5, np->sv_ctest5); + OUTB(nc_gpcntl, np->sv_gpcntl); + OUTB(nc_stest2, np->sv_stest2); + + ncr_selectclock(np, np->sv_scntl3); +/* +** Free host resources +*/ + ncr_free_resources(np); + + return 1; +} +#endif + +/*========================================================== +** +** +** Complete execution of a SCSI command. +** Signal completion to the generic SCSI driver. +** +** +**========================================================== +*/ + +void ncr_complete (ncb_p np, ccb_p cp) +{ + Scsi_Cmnd *cmd; + tcb_p tp; + lcb_p lp; + + /* + ** Sanity check + */ + if (!cp || !cp->cmd) + return; + + /* + ** Print some debugging info. + */ + + if (DEBUG_FLAGS & DEBUG_TINY) + printk ("CCB=%lx STAT=%x/%x\n", (unsigned long)cp, + cp->host_status,cp->scsi_status); + + /* + ** Get command, target and lun pointers. + */ + + cmd = cp->cmd; + cp->cmd = NULL; + tp = &np->target[cp->target]; + lp = ncr_lp(np, tp, cp->lun); + + /* + ** We donnot queue more than 1 ccb per target + ** with negotiation at any time. If this ccb was + ** used for negotiation, clear this info in the tcb. + */ + + if (cp == tp->nego_cp) + tp->nego_cp = 0; + +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** We just complete the last queued CCB. + ** Clear this info that is no more relevant. + */ + if (cp == np->last_cp) + np->last_cp = 0; +#endif + + /* + ** If auto-sense performed, change scsi status, + ** Otherwise, compute the residual. + */ + if (cp->host_flags & HF_AUTO_SENSE) { + cp->scsi_status = cp->sv_scsi_status; + cp->xerr_status = cp->sv_xerr_status; + } + else { + cp->resid = 0; + if (cp->xerr_status || + cp->phys.header.lastp != cp->phys.header.goalp) + cp->resid = ncr_compute_residual(np, cp); + } + + /* + ** Check for extended errors. + */ + + if (cp->xerr_status) { + if (cp->xerr_status & XE_PARITY_ERR) { + PRINT_ADDR(cmd); + printk ("unrecovered SCSI parity error.\n"); + } + if (cp->xerr_status & XE_EXTRA_DATA) { + PRINT_ADDR(cmd); + printk ("extraneous data discarded.\n"); + } + if (cp->xerr_status & XE_BAD_PHASE) { + PRINT_ADDR(cmd); + printk ("illegal scsi phase (4/5).\n"); + } + if (cp->xerr_status & XE_SODL_UNRUN) { + PRINT_ADDR(cmd); + printk ("ODD transfer in DATA OUT phase.\n"); + } + if (cp->xerr_status & XE_SWIDE_OVRUN){ + PRINT_ADDR(cmd); + printk ("ODD transfer in DATA IN phase.\n"); + } + + if (cp->host_status==HS_COMPLETE) + cp->host_status = HS_FAIL; + } + + /* + ** Print out any error for debugging purpose. + */ + if (DEBUG_FLAGS & (DEBUG_RESULT|DEBUG_TINY)) { + if (cp->host_status!=HS_COMPLETE || cp->scsi_status!=S_GOOD || + cp->resid) { + PRINT_ADDR(cmd); + printk ("ERROR: cmd=%x host_status=%x scsi_status=%x " + "data_len=%d residual=%d\n", + cmd->cmnd[0], cp->host_status, cp->scsi_status, + cp->data_len, cp->resid); + } + } + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,99) + /* + ** Move residual byte count to user structure. + */ + cmd->resid = cp->resid; +#endif + /* + ** Check the status. + */ + if ( (cp->host_status == HS_COMPLETE) + && (cp->scsi_status == S_GOOD || + cp->scsi_status == S_COND_MET)) { + /* + ** All went well (GOOD status). + ** CONDITION MET status is returned on + ** `Pre-Fetch' or `Search data' success. + */ + SetScsiResult(cmd, DID_OK, cp->scsi_status); + + /* + ** Allocate the lcb if not yet. + */ + if (!lp) + ncr_alloc_lcb (np, cp->target, cp->lun); + + /* + ** On standard INQUIRY response (EVPD and CmDt + ** not set), setup logical unit according to + ** announced capabilities (we need the 1rst 7 bytes). + */ + if (cmd->cmnd[0] == 0x12 && !(cmd->cmnd[1] & 0x3) && + cmd->cmnd[4] >= 7 && !cmd->use_sg) { + sync_scsi_data(np, cmd); /* SYNC the data */ + ncr_setup_lcb (np, cp->target, cp->lun, + (char *) cmd->request_buffer); + } + + /* + ** If tags was reduced due to queue full, + ** increase tags if 1000 good status received. + */ + if (lp && lp->usetags && lp->numtags < lp->maxtags) { + ++lp->num_good; + if (lp->num_good >= 1000) { + lp->num_good = 0; + ++lp->numtags; + ncr_setup_tags (np, cp->target, cp->lun); + } + } + } else if ((cp->host_status == HS_COMPLETE) + && (cp->scsi_status == S_CHECK_COND)) { + /* + ** Check condition code + */ + SetScsiResult(cmd, DID_OK, S_CHECK_COND); + + if (DEBUG_FLAGS & (DEBUG_RESULT|DEBUG_TINY)) { + PRINT_ADDR(cmd); + ncr_printl_hex("sense data:", cmd->sense_buffer, 14); + } + } else if ((cp->host_status == HS_COMPLETE) + && (cp->scsi_status == S_CONFLICT)) { + /* + ** Reservation Conflict condition code + */ + SetScsiResult(cmd, DID_OK, S_CONFLICT); + + } else if ((cp->host_status == HS_COMPLETE) + && (cp->scsi_status == S_BUSY || + cp->scsi_status == S_QUEUE_FULL)) { + + /* + ** Target is busy. + */ + SetScsiResult(cmd, DID_OK, cp->scsi_status); + + } else if ((cp->host_status == HS_SEL_TIMEOUT) + || (cp->host_status == HS_TIMEOUT)) { + + /* + ** No response + */ + SetScsiResult(cmd, DID_TIME_OUT, cp->scsi_status); + + } else if (cp->host_status == HS_RESET) { + + /* + ** SCSI bus reset + */ + SetScsiResult(cmd, DID_RESET, cp->scsi_status); + + } else if (cp->host_status == HS_ABORTED) { + + /* + ** Transfer aborted + */ + SetScsiAbortResult(cmd); + + } else { + int did_status; + + /* + ** Other protocol messes + */ + PRINT_ADDR(cmd); + printk ("COMMAND FAILED (%x %x) @%p.\n", + cp->host_status, cp->scsi_status, cp); + + did_status = DID_ERROR; + if (cp->xerr_status & XE_PARITY_ERR) + did_status = DID_PARITY; + + SetScsiResult(cmd, did_status, cp->scsi_status); + } + + /* + ** trace output + */ + + if (tp->usrflag & UF_TRACE) { + PRINT_ADDR(cmd); + printk (" CMD:"); + ncr_print_hex(cmd->cmnd, cmd->cmd_len); + + if (cp->host_status==HS_COMPLETE) { + switch (cp->scsi_status) { + case S_GOOD: + printk (" GOOD"); + break; + case S_CHECK_COND: + printk (" SENSE:"); + ncr_print_hex(cmd->sense_buffer, 14); + break; + default: + printk (" STAT: %x\n", cp->scsi_status); + break; + } + } else printk (" HOSTERROR: %x", cp->host_status); + printk ("\n"); + } + + /* + ** Free this ccb + */ + ncr_free_ccb (np, cp); + + /* + ** requeue awaiting scsi commands for this lun. + */ + if (lp && lp->queuedccbs < lp->queuedepth && + !xpt_que_empty(&lp->wait_ccbq)) + ncr_start_next_ccb(np, lp, 2); + + /* + ** requeue awaiting scsi commands for this controller. + */ + if (np->waiting_list) + requeue_waiting_list(np); + + /* + ** signal completion to generic driver. + */ + ncr_queue_done_cmd(np, cmd); +} + +/*========================================================== +** +** +** Signal all (or one) control block done. +** +** +**========================================================== +*/ + +/* +** The NCR has completed CCBs. +** Look at the DONE QUEUE. +** +** On architectures that may reorder LOAD/STORE operations, +** a memory barrier may be needed after the reading of the +** so-called `flag' and prior to dealing with the data. +*/ +int ncr_wakeup_done (ncb_p np) +{ + ccb_p cp; + int i, n; + u_long dsa; + + n = 0; + i = np->dqueueget; + while (1) { + dsa = scr_to_cpu(np->dqueue[i]); + if (!dsa) + break; + np->dqueue[i] = 0; + if ((i = i+2) >= MAX_START*2) + i = 0; + + cp = ncr_ccb_from_dsa(np, dsa); + if (cp) { + MEMORY_BARRIER(); + ncr_complete (np, cp); + ++n; + } + else + printk (KERN_ERR "%s: bad DSA (%lx) in done queue.\n", + ncr_name(np), dsa); + } + np->dqueueget = i; + + return n; +} + +/* +** Complete all active CCBs. +*/ +void ncr_wakeup (ncb_p np, u_long code) +{ + ccb_p cp = np->ccbc; + + while (cp) { + if (cp->host_status != HS_IDLE) { + cp->host_status = code; + ncr_complete (np, cp); + } + cp = cp->link_ccb; + } +} + +/*========================================================== +** +** +** Start NCR chip. +** +** +**========================================================== +*/ + +void ncr_init (ncb_p np, int reset, char * msg, u_long code) +{ + int i; + u_long phys; + + /* + ** Reset chip if asked, otherwise just clear fifos. + */ + + if (reset) + ncr_soft_reset(np); + else { + OUTB (nc_stest3, TE|CSF); + OUTONB (nc_ctest3, CLF); + } + + /* + ** Message. + */ + + if (msg) printk (KERN_INFO "%s: restart (%s).\n", ncr_name (np), msg); + + /* + ** Clear Start Queue + */ + phys = np->p_squeue; + np->queuedepth = MAX_START - 1; /* 1 entry needed as end marker */ + for (i = 0; i < MAX_START*2; i += 2) { + np->squeue[i] = cpu_to_scr(np->p_idletask); + np->squeue[i+1] = cpu_to_scr(phys + (i+2)*4); + } + np->squeue[MAX_START*2-1] = cpu_to_scr(phys); + + + /* + ** Start at first entry. + */ + np->squeueput = 0; + np->scripth0->startpos[0] = cpu_to_scr(phys); + + /* + ** Clear Done Queue + */ + phys = vtobus(np->dqueue); + for (i = 0; i < MAX_START*2; i += 2) { + np->dqueue[i] = 0; + np->dqueue[i+1] = cpu_to_scr(phys + (i+2)*4); + } + np->dqueue[MAX_START*2-1] = cpu_to_scr(phys); + + /* + ** Start at first entry. + */ + np->scripth0->done_pos[0] = cpu_to_scr(phys); + np->dqueueget = 0; + + /* + ** Wakeup all pending jobs. + */ + ncr_wakeup (np, code); + + /* + ** Init chip. + */ + + OUTB (nc_istat, 0x00 ); /* Remove Reset, abort */ + UDELAY (2000); /* The 895 needs time for the bus mode to settle */ + + OUTB (nc_scntl0, np->rv_scntl0 | 0xc0); + /* full arb., ena parity, par->ATN */ + OUTB (nc_scntl1, 0x00); /* odd parity, and remove CRST!! */ + + ncr_selectclock(np, np->rv_scntl3); /* Select SCSI clock */ + + OUTB (nc_scid , RRE|np->myaddr); /* Adapter SCSI address */ + OUTW (nc_respid, 1ul<<np->myaddr); /* Id to respond to */ + OUTB (nc_istat , SIGP ); /* Signal Process */ + OUTB (nc_dmode , np->rv_dmode); /* Burst length, dma mode */ + OUTB (nc_ctest5, np->rv_ctest5); /* Large fifo + large burst */ + + OUTB (nc_dcntl , NOCOM|np->rv_dcntl); /* Protect SFBR */ + OUTB (nc_ctest3, np->rv_ctest3); /* Write and invalidate */ + OUTB (nc_ctest4, np->rv_ctest4); /* Master parity checking */ + + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)){ + OUTB (nc_stest2, EXT|np->rv_stest2); + /* Extended Sreq/Sack filtering, not supported in C1010/C1010_66 */ + } + OUTB (nc_stest3, TE); /* TolerANT enable */ + OUTB (nc_stime0, 0x0c); /* HTH disabled STO 0.25 sec */ + + /* + ** DEL 441 - 53C876 Rev 5 - Part Number 609-0392787/2788 - ITEM 2. + ** Disable overlapped arbitration for all dual-function + ** devices, regardless revision id. + ** We may consider it is a post-chip-design feature. ;-) + ** + ** Errata applies to all 896 and 1010 parts. + */ + if (np->device_id == PCI_DEVICE_ID_NCR_53C875) + OUTB (nc_ctest0, (1<<5)); + else if (np->device_id == PCI_DEVICE_ID_NCR_53C896 || + np->device_id == PCI_DEVICE_ID_LSI_53C1010 || + np->device_id == PCI_DEVICE_ID_LSI_53C1010_66 ) + np->rv_ccntl0 |= DPR; + + /* + ** C1010_66MHz rev 0 part requies AIPCNTL1 bit 3 to be set. + */ + if (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66) + OUTB(nc_aipcntl1, (1<<3)); + + /* + ** If 64 bit (895A/896/1010/1010_66) write the CCNTL1 register to + ** enable 40 bit address table indirect addressing for MOVE. + ** Also write CCNTL0 if 64 bit chip, since this register seems + ** to only be used by 64 bit cores. + */ + if (np->features & FE_64BIT) { + OUTB (nc_ccntl0, np->rv_ccntl0); + OUTB (nc_ccntl1, np->rv_ccntl1); + } + + /* + ** If phase mismatch handled by scripts (53C895A or 53C896 + ** or 53C1010 or 53C1010_66), set PM jump addresses. + */ + + if (np->features & FE_NOPM) { + printk(KERN_INFO "%s: handling phase mismatch from SCRIPTS.\n", + ncr_name(np)); + OUTL (nc_pmjad1, NCB_SCRIPTH_PHYS (np, pm_handle)); + OUTL (nc_pmjad2, NCB_SCRIPTH_PHYS (np, pm_handle)); + } + + /* + ** Enable GPIO0 pin for writing if LED support from SCRIPTS. + ** Also set GPIO5 and clear GPIO6 if hardware LED control. + */ + + if (np->features & FE_LED0) + OUTB(nc_gpcntl, INB(nc_gpcntl) & ~0x01); + else if (np->features & FE_LEDC) + OUTB(nc_gpcntl, (INB(nc_gpcntl) & ~0x41) | 0x20); + + + /* + ** enable ints + */ + + OUTW (nc_sien , STO|HTH|MA|SGE|UDC|RST|PAR); + OUTB (nc_dien , MDPE|BF|SSI|SIR|IID); + + /* + ** For 895/895A/896/c1010 + ** Enable SBMC interrupt and save current SCSI bus mode. + */ + if ( (np->features & FE_ULTRA2) || (np->features & FE_ULTRA3) ) { + OUTONW (nc_sien, SBMC); + np->scsi_mode = INB (nc_stest4) & SMODE; + } + + /* + ** Fill in target structure. + ** Reinitialize usrsync. + ** Reinitialize usrwide. + ** Prepare sync negotiation according to actual SCSI bus mode. + */ + + for (i=0;i<MAX_TARGET;i++) { + tcb_p tp = &np->target[i]; + + tp->to_reset = 0; + + tp->sval = 0; + tp->wval = np->rv_scntl3; + tp->uval = np->rv_scntl4; + + if (tp->usrsync != 255) { + if (tp->usrsync <= np->maxsync) { + if (tp->usrsync < np->minsync) { + tp->usrsync = np->minsync; + } + } + else + tp->usrsync = 255; + }; + + if (tp->usrwide > np->maxwide) + tp->usrwide = np->maxwide; + + ncr_negotiate (np, tp); + } + + /* + ** Download SCSI SCRIPTS to on-chip RAM if present, + ** and start script processor. + ** We do the download preferently from the CPU. + ** For platforms that may not support PCI memory mapping, + ** we use a simple SCRIPTS that performs MEMORY MOVEs. + */ + if (np->base2_ba) { + if (bootverbose) + printk ("%s: Downloading SCSI SCRIPTS.\n", + ncr_name(np)); +#ifdef SCSI_NCR_PCI_MEM_NOT_SUPPORTED + if (np->base2_ws == 8192) + phys = NCB_SCRIPTH0_PHYS (np, start_ram64); + else + phys = NCB_SCRIPTH_PHYS (np, start_ram); +#else + if (np->base2_ws == 8192) { + memcpy_to_pci(np->base2_va + 4096, + np->scripth0, sizeof(struct scripth)); + OUTL (nc_mmws, np->scr_ram_seg); + OUTL (nc_mmrs, np->scr_ram_seg); + OUTL (nc_sfs, np->scr_ram_seg); + phys = NCB_SCRIPTH_PHYS (np, start64); + } + else + phys = NCB_SCRIPT_PHYS (np, init); + memcpy_to_pci(np->base2_va, np->script0, sizeof(struct script)); +#endif /* SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ + } + else + phys = NCB_SCRIPT_PHYS (np, init); + + np->istat_sem = 0; + + OUTL (nc_dsa, np->p_ncb); + OUTL_DSP (phys); +} + +/*========================================================== +** +** Prepare the negotiation values for wide and +** synchronous transfers. +** +**========================================================== +*/ + +static void ncr_negotiate (struct ncb* np, struct tcb* tp) +{ + /* + ** minsync unit is 4ns ! + */ + + u_long minsync = tp->usrsync; + + /* + ** SCSI bus mode limit + */ + + if (np->scsi_mode && np->scsi_mode == SMODE_SE) { + if (minsync < 12) minsync = 12; + } + + /* + ** our limit .. + */ + + if (minsync < np->minsync) + minsync = np->minsync; + + /* + ** divider limit + */ + + if (minsync > np->maxsync) + minsync = 255; + + tp->minsync = minsync; + tp->maxoffs = (minsync<255 ? np->maxoffs : 0); + + /* + ** period=0: has to negotiate sync transfer + */ + + tp->period=0; + + /* + ** widedone=0: has to negotiate wide transfer + */ + tp->widedone=0; +} + +/*========================================================== +** +** Get clock factor and sync divisor for a given +** synchronous factor period. +** Returns the clock factor (in sxfer) and scntl3 +** synchronous divisor field. +** +**========================================================== +*/ + +static void ncr_getsync(ncb_p np, u_char sfac, u_char *fakp, u_char *scntl3p) +{ + u_long clk = np->clock_khz; /* SCSI clock frequency in kHz */ + int div = np->clock_divn; /* Number of divisors supported */ + u_long fak; /* Sync factor in sxfer */ + u_long per; /* Period in tenths of ns */ + u_long kpc; /* (per * clk) */ + + /* + ** Compute the synchronous period in tenths of nano-seconds + ** from sfac. + ** + ** Note, if sfac == 9, DT is being used. Double the period of 125 + ** to 250. + */ + if (sfac <= 10) per = 250; + else if (sfac == 11) per = 303; + else if (sfac == 12) per = 500; + else per = 40 * sfac; + + /* + ** Look for the greatest clock divisor that allows an + ** input speed faster than the period. + */ + kpc = per * clk; + while (--div >= 0) + if (kpc >= (div_10M[div] << 2)) break; + + /* + ** Calculate the lowest clock factor that allows an output + ** speed not faster than the period. + */ + fak = (kpc - 1) / div_10M[div] + 1; + +#if 0 /* This optimization does not seem very usefull */ + + per = (fak * div_10M[div]) / clk; + + /* + ** Why not to try the immediate lower divisor and to choose + ** the one that allows the fastest output speed ? + ** We dont want input speed too much greater than output speed. + */ + if (div >= 1 && fak < 8) { + u_long fak2, per2; + fak2 = (kpc - 1) / div_10M[div-1] + 1; + per2 = (fak2 * div_10M[div-1]) / clk; + if (per2 < per && fak2 <= 8) { + fak = fak2; + per = per2; + --div; + } + } +#endif + + if (fak < 4) fak = 4; /* Should never happen, too bad ... */ + + /* + ** Compute and return sync parameters for the ncr + */ + *fakp = fak - 4; + + /* + ** If sfac < 25, and 8xx parts, desire that the chip operate at + ** least at Ultra speeds. Must set bit 7 of scntl3. + ** For C1010, do not set this bit. If operating at Ultra3 speeds, + ** set the U3EN bit instead. + */ + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + *scntl3p = (div+1) << 4; + *fakp = 0; + } + else { + *scntl3p = ((div+1) << 4) + (sfac < 25 ? 0x80 : 0); + *fakp = fak - 4; + } +} + +/*========================================================== +** +** Utility routine to return the current bus width +** synchronous period and offset. +** Utilizes target sval, wval and uval +** +**========================================================== +*/ +static void ncr_get_xfer_info(ncb_p np, tcb_p tp, u_char *factor, + u_char *offset, u_char *width) +{ + + u_char idiv; + u_long period; + + *width = (tp->wval & EWS) ? 1 : 0; + + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) + *offset = (tp->sval & 0x3f); + else + *offset = (tp->sval & 0x1f); + + /* + * Midlayer signal to the driver that all of the scsi commands + * for the integrity check have completed. Save the negotiated + * parameters (extracted from sval, wval and uval). + * See ncr_setsync for alg. details. + */ + + idiv = (tp->wval>>4) & 0x07; + + if ( *offset && idiv ) { + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)){ + if (tp->uval & 0x80) + period = (2*div_10M[idiv-1])/np->clock_khz; + else + period = (4*div_10M[idiv-1])/np->clock_khz; + } + else + period = (((tp->sval>>5)+4)*div_10M[idiv-1])/np->clock_khz; + } + else + period = 0xffff; + + if (period <= 125) *factor = 9; + else if (period <= 250) *factor = 10; + else if (period <= 303) *factor = 11; + else if (period <= 500) *factor = 12; + else *factor = (period + 40 - 1) / 40; + +} + + +/*========================================================== +** +** Set actual values, sync status and patch all ccbs of +** a target according to new sync/wide agreement. +** +**========================================================== +*/ + +static void ncr_set_sync_wide_status (ncb_p np, u_char target) +{ + ccb_p cp = np->ccbc; + tcb_p tp = &np->target[target]; + + /* + ** set actual value and sync_status + ** + ** TEMP register contains current scripts address + ** which is data type/direction/dependent. + */ + OUTB (nc_sxfer, tp->sval); + OUTB (nc_scntl3, tp->wval); + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) + OUTB (nc_scntl4, tp->uval); + + /* + ** patch ALL ccbs of this target. + */ + for (cp = np->ccbc; cp; cp = cp->link_ccb) { + if (cp->host_status == HS_IDLE) + continue; + if (cp->target != target) + continue; + cp->phys.select.sel_scntl3 = tp->wval; + cp->phys.select.sel_sxfer = tp->sval; + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) + cp->phys.select.sel_scntl4 = tp->uval; + }; +} + +/*========================================================== +** +** Switch sync mode for current job and it's target +** +**========================================================== +*/ + +static void ncr_setsync (ncb_p np, ccb_p cp, u_char scntl3, u_char sxfer, + u_char scntl4) +{ + tcb_p tp; + u_char target = INB (nc_sdid) & 0x0f; + u_char idiv; + u_char offset; + + assert (cp); + if (!cp) return; + + assert (target == (cp->target & 0xf)); + + tp = &np->target[target]; + + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + offset = sxfer & 0x3f; /* bits 5-0 */ + scntl3 = (scntl3 & 0xf0) | (tp->wval & EWS); + scntl4 = (scntl4 & 0x80); + } + else { + offset = sxfer & 0x1f; /* bits 4-0 */ + if (!scntl3 || !offset) + scntl3 = np->rv_scntl3; + + scntl3 = (scntl3 & 0xf0) | (tp->wval & EWS) | + (np->rv_scntl3 & 0x07); + } + + + /* + ** Deduce the value of controller sync period from scntl3. + ** period is in tenths of nano-seconds. + */ + + idiv = ((scntl3 >> 4) & 0x7); + if ( offset && idiv) { + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + /* Note: If extra data hold clocks are used, + * the formulas below must be modified. + * When scntl4 == 0, ST mode. + */ + if (scntl4 & 0x80) + tp->period = (2*div_10M[idiv-1])/np->clock_khz; + else + tp->period = (4*div_10M[idiv-1])/np->clock_khz; + } + else + tp->period = (((sxfer>>5)+4)*div_10M[idiv-1])/np->clock_khz; + } + else + tp->period = 0xffff; + + + /* + ** Stop there if sync parameters are unchanged + */ + if (tp->sval == sxfer && tp->wval == scntl3 && tp->uval == scntl4) return; + tp->sval = sxfer; + tp->wval = scntl3; + tp->uval = scntl4; + + /* + ** Bells and whistles ;-) + ** Donnot announce negotiations due to auto-sense, + ** unless user really want us to be verbose. :) + */ + if ( bootverbose < 2 && (cp->host_flags & HF_AUTO_SENSE)) + goto next; + PRINT_TARGET(np, target); + if (offset) { + unsigned f10 = 100000 << (tp->widedone ? tp->widedone -1 : 0); + unsigned mb10 = (f10 + tp->period/2) / tp->period; + char *scsi; + + /* + ** Disable extended Sreq/Sack filtering + */ + if ((tp->period <= 2000) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + OUTOFFB (nc_stest2, EXT); + + /* + ** Bells and whistles ;-) + */ + if (tp->period < 250) scsi = "FAST-80"; + else if (tp->period < 500) scsi = "FAST-40"; + else if (tp->period < 1000) scsi = "FAST-20"; + else if (tp->period < 2000) scsi = "FAST-10"; + else scsi = "FAST-5"; + + printk ("%s %sSCSI %d.%d MB/s (%d ns, offset %d)\n", scsi, + tp->widedone > 1 ? "WIDE " : "", + mb10 / 10, mb10 % 10, tp->period / 10, offset); + } else + printk ("%sasynchronous.\n", tp->widedone > 1 ? "wide " : ""); +next: + /* + ** set actual value and sync_status + ** patch ALL ccbs of this target. + */ + ncr_set_sync_wide_status(np, target); +} + + +/*========================================================== +** +** Switch wide mode for current job and it's target +** SCSI specs say: a SCSI device that accepts a WDTR +** message shall reset the synchronous agreement to +** asynchronous mode. +** +**========================================================== +*/ + +static void ncr_setwide (ncb_p np, ccb_p cp, u_char wide, u_char ack) +{ + u_short target = INB (nc_sdid) & 0x0f; + tcb_p tp; + u_char scntl3; + u_char sxfer; + + assert (cp); + if (!cp) return; + + assert (target == (cp->target & 0xf)); + + tp = &np->target[target]; + tp->widedone = wide+1; + scntl3 = (tp->wval & (~EWS)) | (wide ? EWS : 0); + + sxfer = ack ? 0 : tp->sval; + + /* + ** Stop there if sync/wide parameters are unchanged + */ + if (tp->sval == sxfer && tp->wval == scntl3) return; + tp->sval = sxfer; + tp->wval = scntl3; + + /* + ** Bells and whistles ;-) + */ + if (bootverbose >= 2) { + PRINT_TARGET(np, target); + if (scntl3 & EWS) + printk ("WIDE SCSI (16 bit) enabled.\n"); + else + printk ("WIDE SCSI disabled.\n"); + } + + /* + ** set actual value and sync_status + ** patch ALL ccbs of this target. + */ + ncr_set_sync_wide_status(np, target); +} + + +/*========================================================== +** +** Switch sync/wide mode for current job and it's target +** PPR negotiations only +** +**========================================================== +*/ + +static void ncr_setsyncwide (ncb_p np, ccb_p cp, u_char scntl3, u_char sxfer, + u_char scntl4, u_char wide) +{ + tcb_p tp; + u_char target = INB (nc_sdid) & 0x0f; + u_char idiv; + u_char offset; + + assert (cp); + if (!cp) return; + + assert (target == (cp->target & 0xf)); + + tp = &np->target[target]; + tp->widedone = wide+1; + + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + offset = sxfer & 0x3f; /* bits 5-0 */ + scntl3 = (scntl3 & 0xf0) | (wide ? EWS : 0); + scntl4 = (scntl4 & 0x80); + } + else { + offset = sxfer & 0x1f; /* bits 4-0 */ + if (!scntl3 || !offset) + scntl3 = np->rv_scntl3; + + scntl3 = (scntl3 & 0xf0) | (wide ? EWS : 0) | + (np->rv_scntl3 & 0x07); + } + + + /* + ** Deduce the value of controller sync period from scntl3. + ** period is in tenths of nano-seconds. + */ + + idiv = ((scntl3 >> 4) & 0x7); + if ( offset && idiv) { + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + /* Note: If extra data hold clocks are used, + * the formulas below must be modified. + * When scntl4 == 0, ST mode. + */ + if (scntl4 & 0x80) + tp->period = (2*div_10M[idiv-1])/np->clock_khz; + else + tp->period = (4*div_10M[idiv-1])/np->clock_khz; + } + else + tp->period = (((sxfer>>5)+4)*div_10M[idiv-1])/np->clock_khz; + } + else + tp->period = 0xffff; + + + /* + ** Stop there if sync parameters are unchanged + */ + if (tp->sval == sxfer && tp->wval == scntl3 && tp->uval == scntl4) return; + tp->sval = sxfer; + tp->wval = scntl3; + tp->uval = scntl4; + + /* + ** Bells and whistles ;-) + ** Donnot announce negotiations due to auto-sense, + ** unless user really want us to be verbose. :) + */ + if ( bootverbose < 2 && (cp->host_flags & HF_AUTO_SENSE)) + goto next; + PRINT_TARGET(np, target); + if (offset) { + unsigned f10 = 100000 << (tp->widedone ? tp->widedone -1 : 0); + unsigned mb10 = (f10 + tp->period/2) / tp->period; + char *scsi; + + /* + ** Disable extended Sreq/Sack filtering + */ + if ((tp->period <= 2000) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + OUTOFFB (nc_stest2, EXT); + + /* + ** Bells and whistles ;-) + */ + if (tp->period < 250) scsi = "FAST-80"; + else if (tp->period < 500) scsi = "FAST-40"; + else if (tp->period < 1000) scsi = "FAST-20"; + else if (tp->period < 2000) scsi = "FAST-10"; + else scsi = "FAST-5"; + + printk ("%s %sSCSI %d.%d MB/s (%d ns, offset %d)\n", scsi, + tp->widedone > 1 ? "WIDE " : "", + mb10 / 10, mb10 % 10, tp->period / 10, offset); + } else + printk ("%sasynchronous.\n", tp->widedone > 1 ? "wide " : ""); +next: + /* + ** set actual value and sync_status + ** patch ALL ccbs of this target. + */ + ncr_set_sync_wide_status(np, target); +} + + + + +/*========================================================== +** +** Switch tagged mode for a target. +** +**========================================================== +*/ + +static void ncr_setup_tags (ncb_p np, u_char tn, u_char ln) +{ + tcb_p tp = &np->target[tn]; + lcb_p lp = ncr_lp(np, tp, ln); + u_short reqtags, maxdepth; + + /* + ** Just in case ... + */ + if ((!tp) || (!lp)) + return; + + /* + ** If SCSI device queue depth is not yet set, leave here. + */ + if (!lp->scdev_depth) + return; + + /* + ** Donnot allow more tags than the SCSI driver can queue + ** for this device. + ** Donnot allow more tags than we can handle. + */ + maxdepth = lp->scdev_depth; + if (maxdepth > lp->maxnxs) maxdepth = lp->maxnxs; + if (lp->maxtags > maxdepth) lp->maxtags = maxdepth; + if (lp->numtags > maxdepth) lp->numtags = maxdepth; + + /* + ** only devices conformant to ANSI Version >= 2 + ** only devices capable of tagged commands + ** only if enabled by user .. + */ + if ((lp->inq_byte7 & INQ7_QUEUE) && lp->numtags > 1) { + reqtags = lp->numtags; + } else { + reqtags = 1; + }; + + /* + ** Update max number of tags + */ + lp->numtags = reqtags; + if (lp->numtags > lp->maxtags) + lp->maxtags = lp->numtags; + + /* + ** If we want to switch tag mode, we must wait + ** for no CCB to be active. + */ + if (reqtags > 1 && lp->usetags) { /* Stay in tagged mode */ + if (lp->queuedepth == reqtags) /* Already announced */ + return; + lp->queuedepth = reqtags; + } + else if (reqtags <= 1 && !lp->usetags) { /* Stay in untagged mode */ + lp->queuedepth = reqtags; + return; + } + else { /* Want to switch tag mode */ + if (lp->busyccbs) /* If not yet safe, return */ + return; + lp->queuedepth = reqtags; + lp->usetags = reqtags > 1 ? 1 : 0; + } + + /* + ** Patch the lun mini-script, according to tag mode. + */ + lp->resel_task = lp->usetags? + cpu_to_scr(NCB_SCRIPT_PHYS(np, resel_tag)) : + cpu_to_scr(NCB_SCRIPT_PHYS(np, resel_notag)); + + /* + ** Announce change to user. + */ + if (bootverbose) { + PRINT_LUN(np, tn, ln); + if (lp->usetags) + printk("tagged command queue depth set to %d\n", reqtags); + else + printk("tagged command queueing disabled\n"); + } +} + +/*---------------------------------------------------- +** +** handle user commands +** +**---------------------------------------------------- +*/ + +#ifdef SCSI_NCR_USER_COMMAND_SUPPORT + +static void ncr_usercmd (ncb_p np) +{ + u_char t; + tcb_p tp; + int ln; + u_long size; + + switch (np->user.cmd) { + case 0: return; + + case UC_SETDEBUG: +#ifdef SCSI_NCR_DEBUG_INFO_SUPPORT + ncr_debug = np->user.data; +#endif + break; + + case UC_SETORDER: + np->order = np->user.data; + break; + + case UC_SETVERBOSE: + np->verbose = np->user.data; + break; + + default: + /* + ** We assume that other commands apply to targets. + ** This should always be the case and avoid the below + ** 4 lines to be repeated 5 times. + */ + for (t = 0; t < MAX_TARGET; t++) { + if (!((np->user.target >> t) & 1)) + continue; + tp = &np->target[t]; + + switch (np->user.cmd) { + + case UC_SETSYNC: + tp->usrsync = np->user.data; + ncr_negotiate (np, tp); + break; + + case UC_SETWIDE: + size = np->user.data; + if (size > np->maxwide) + size=np->maxwide; + tp->usrwide = size; + ncr_negotiate (np, tp); + break; + + case UC_SETTAGS: + tp->usrtags = np->user.data; + for (ln = 0; ln < MAX_LUN; ln++) { + lcb_p lp; + lp = ncr_lp(np, tp, ln); + if (!lp) + continue; + lp->numtags = np->user.data; + lp->maxtags = lp->numtags; + ncr_setup_tags (np, t, ln); + } + break; + + case UC_RESETDEV: + tp->to_reset = 1; + np->istat_sem = SEM; + OUTB (nc_istat, SIGP|SEM); + break; + + case UC_CLEARDEV: + for (ln = 0; ln < MAX_LUN; ln++) { + lcb_p lp; + lp = ncr_lp(np, tp, ln); + if (lp) + lp->to_clear = 1; + } + np->istat_sem = SEM; + OUTB (nc_istat, SIGP|SEM); + break; + + case UC_SETFLAG: + tp->usrflag = np->user.data; + break; + } + } + break; + } + np->user.cmd=0; +} +#endif + +/*========================================================== +** +** +** ncr timeout handler. +** +** +**========================================================== +** +** Misused to keep the driver running when +** interrupts are not configured correctly. +** +**---------------------------------------------------------- +*/ + +static void ncr_timeout (ncb_p np) +{ + u_long thistime = ktime_get(0); + + /* + ** If release process in progress, let's go + ** Set the release stage from 1 to 2 to synchronize + ** with the release process. + */ + + if (np->release_stage) { + if (np->release_stage == 1) np->release_stage = 2; + return; + } + +#ifdef SCSI_NCR_PCIQ_BROKEN_INTR + np->timer.expires = ktime_get((HZ+9)/10); +#else + np->timer.expires = ktime_get(SCSI_NCR_TIMER_INTERVAL); +#endif + add_timer(&np->timer); + + /* + ** If we are resetting the ncr, wait for settle_time before + ** clearing it. Then command processing will be resumed. + */ + if (np->settle_time) { + if (np->settle_time <= thistime) { + if (bootverbose > 1) + printk("%s: command processing resumed\n", ncr_name(np)); + np->settle_time = 0; + requeue_waiting_list(np); + } + return; + } + + /* + ** Nothing to do for now, but that may come. + */ + if (np->lasttime + 4*HZ < thistime) { + np->lasttime = thistime; + } + +#ifdef SCSI_NCR_PCIQ_MAY_MISS_COMPLETIONS + /* + ** Some way-broken PCI bridges may lead to + ** completions being lost when the clearing + ** of the INTFLY flag by the CPU occurs + ** concurrently with the chip raising this flag. + ** If this ever happen, lost completions will + ** be reaped here. + */ + ncr_wakeup_done(np); +#endif + +#ifdef SCSI_NCR_PCIQ_BROKEN_INTR + if (INB(nc_istat) & (INTF|SIP|DIP)) { + + /* + ** Process pending interrupts. + */ + if (DEBUG_FLAGS & DEBUG_TINY) printk ("{"); + ncr_exception (np); + if (DEBUG_FLAGS & DEBUG_TINY) printk ("}"); + } +#endif /* SCSI_NCR_PCIQ_BROKEN_INTR */ +} + +/*========================================================== +** +** log message for real hard errors +** +** "ncr0 targ 0?: ERROR (ds:si) (so-si-sd) (sxfer/scntl3) @ name (dsp:dbc)." +** " reg: r0 r1 r2 r3 r4 r5 r6 ..... rf." +** +** exception register: +** ds: dstat +** si: sist +** +** SCSI bus lines: +** so: control lines as driven by NCR. +** si: control lines as seen by NCR. +** sd: scsi data lines as seen by NCR. +** +** wide/fastmode: +** sxfer: (see the manual) +** scntl3: (see the manual) +** +** current script command: +** dsp: script address (relative to start of script). +** dbc: first word of script command. +** +** First 24 register of the chip: +** r0..rf +** +**========================================================== +*/ + +static void ncr_log_hard_error(ncb_p np, u_short sist, u_char dstat) +{ + u_int32 dsp; + int script_ofs; + int script_size; + char *script_name; + u_char *script_base; + int i; + + dsp = INL (nc_dsp); + + if (dsp > np->p_script && dsp <= np->p_script + sizeof(struct script)) { + script_ofs = dsp - np->p_script; + script_size = sizeof(struct script); + script_base = (u_char *) np->script0; + script_name = "script"; + } + else if (np->p_scripth < dsp && + dsp <= np->p_scripth + sizeof(struct scripth)) { + script_ofs = dsp - np->p_scripth; + script_size = sizeof(struct scripth); + script_base = (u_char *) np->scripth0; + script_name = "scripth"; + } else { + script_ofs = dsp; + script_size = 0; + script_base = 0; + script_name = "mem"; + } + + printk ("%s:%d: ERROR (%x:%x) (%x-%x-%x) (%x/%x) @ (%s %x:%08x).\n", + ncr_name (np), (unsigned)INB (nc_sdid)&0x0f, dstat, sist, + (unsigned)INB (nc_socl), (unsigned)INB (nc_sbcl), (unsigned)INB (nc_sbdl), + (unsigned)INB (nc_sxfer),(unsigned)INB (nc_scntl3), script_name, script_ofs, + (unsigned)INL (nc_dbc)); + + if (((script_ofs & 3) == 0) && + (unsigned)script_ofs < script_size) { + printk ("%s: script cmd = %08x\n", ncr_name(np), + scr_to_cpu((int) *(ncrcmd *)(script_base + script_ofs))); + } + + printk ("%s: regdump:", ncr_name(np)); + for (i=0; i<24;i++) + printk (" %02x", (unsigned)INB_OFF(i)); + printk (".\n"); +} + +/*============================================================ +** +** ncr chip exception handler. +** +**============================================================ +** +** In normal situations, interrupt conditions occur one at +** a time. But when something bad happens on the SCSI BUS, +** the chip may raise several interrupt flags before +** stopping and interrupting the CPU. The additionnal +** interrupt flags are stacked in some extra registers +** after the SIP and/or DIP flag has been raised in the +** ISTAT. After the CPU has read the interrupt condition +** flag from SIST or DSTAT, the chip unstacks the other +** interrupt flags and sets the corresponding bits in +** SIST or DSTAT. Since the chip starts stacking once the +** SIP or DIP flag is set, there is a small window of time +** where the stacking does not occur. +** +** Typically, multiple interrupt conditions may happen in +** the following situations: +** +** - SCSI parity error + Phase mismatch (PAR|MA) +** When an parity error is detected in input phase +** and the device switches to msg-in phase inside a +** block MOV. +** - SCSI parity error + Unexpected disconnect (PAR|UDC) +** When a stupid device does not want to handle the +** recovery of an SCSI parity error. +** - Some combinations of STO, PAR, UDC, ... +** When using non compliant SCSI stuff, when user is +** doing non compliant hot tampering on the BUS, when +** something really bad happens to a device, etc ... +** +** The heuristic suggested by SYMBIOS to handle +** multiple interrupts is to try unstacking all +** interrupts conditions and to handle them on some +** priority based on error severity. +** This will work when the unstacking has been +** successful, but we cannot be 100 % sure of that, +** since the CPU may have been faster to unstack than +** the chip is able to stack. Hmmm ... But it seems that +** such a situation is very unlikely to happen. +** +** If this happen, for example STO catched by the CPU +** then UDC happenning before the CPU have restarted +** the SCRIPTS, the driver may wrongly complete the +** same command on UDC, since the SCRIPTS didn't restart +** and the DSA still points to the same command. +** We avoid this situation by setting the DSA to an +** invalid value when the CCB is completed and before +** restarting the SCRIPTS. +** +** Another issue is that we need some section of our +** recovery procedures to be somehow uninterruptible and +** that the SCRIPTS processor does not provides such a +** feature. For this reason, we handle recovery preferently +** from the C code and check against some SCRIPTS +** critical sections from the C code. +** +** Hopefully, the interrupt handling of the driver is now +** able to resist to weird BUS error conditions, but donnot +** ask me for any guarantee that it will never fail. :-) +** Use at your own decision and risk. +** +**============================================================ +*/ + +void ncr_exception (ncb_p np) +{ + u_char istat, istatc; + u_char dstat; + u_short sist; + int i; + + /* + ** interrupt on the fly ? + ** + ** A `dummy read' is needed to ensure that the + ** clear of the INTF flag reaches the device + ** before the scanning of the DONE queue. + */ + istat = INB (nc_istat); + if (istat & INTF) { + OUTB (nc_istat, (istat & SIGP) | INTF | np->istat_sem); + istat = INB (nc_istat); /* DUMMY READ */ + if (DEBUG_FLAGS & DEBUG_TINY) printk ("F "); + (void)ncr_wakeup_done (np); + }; + + if (!(istat & (SIP|DIP))) + return; + +#if 0 /* We should never get this one */ + if (istat & CABRT) + OUTB (nc_istat, CABRT); +#endif + + /* + ** Steinbach's Guideline for Systems Programming: + ** Never test for an error condition you don't know how to handle. + */ + + /*======================================================== + ** PAR and MA interrupts may occur at the same time, + ** and we need to know of both in order to handle + ** this situation properly. We try to unstack SCSI + ** interrupts for that reason. BTW, I dislike a LOT + ** such a loop inside the interrupt routine. + ** Even if DMA interrupt stacking is very unlikely to + ** happen, we also try unstacking these ones, since + ** this has no performance impact. + **========================================================= + */ + sist = 0; + dstat = 0; + istatc = istat; + do { + if (istatc & SIP) + sist |= INW (nc_sist); + if (istatc & DIP) + dstat |= INB (nc_dstat); + istatc = INB (nc_istat); + istat |= istatc; + } while (istatc & (SIP|DIP)); + + if (DEBUG_FLAGS & DEBUG_TINY) + printk ("<%d|%x:%x|%x:%x>", + (int)INB(nc_scr0), + dstat,sist, + (unsigned)INL(nc_dsp), + (unsigned)INL(nc_dbc)); + + /* + ** On paper, a memory barrier may be needed here. + ** And since we are paranoid ... :) + */ + MEMORY_BARRIER(); + + /*======================================================== + ** First, interrupts we want to service cleanly. + ** + ** Phase mismatch (MA) is the most frequent interrupt + ** for chip earlier than the 896 and so we have to service + ** it as quickly as possible. + ** A SCSI parity error (PAR) may be combined with a phase + ** mismatch condition (MA). + ** Programmed interrupts (SIR) are used to call the C code + ** from SCRIPTS. + ** The single step interrupt (SSI) is not used in this + ** driver. + **========================================================= + */ + + if (!(sist & (STO|GEN|HTH|SGE|UDC|SBMC|RST)) && + !(dstat & (MDPE|BF|ABRT|IID))) { + if (sist & PAR) ncr_int_par (np, sist); + else if (sist & MA) ncr_int_ma (np); + else if (dstat & SIR) ncr_int_sir (np); + else if (dstat & SSI) OUTONB_STD (); + else goto unknown_int; + return; + }; + + /*======================================================== + ** Now, interrupts that donnot happen in normal + ** situations and that we may need to recover from. + ** + ** On SCSI RESET (RST), we reset everything. + ** On SCSI BUS MODE CHANGE (SBMC), we complete all + ** active CCBs with RESET status, prepare all devices + ** for negotiating again and restart the SCRIPTS. + ** On STO and UDC, we complete the CCB with the corres- + ** ponding status and restart the SCRIPTS. + **========================================================= + */ + + if (sist & RST) { + ncr_init (np, 1, bootverbose ? "scsi reset" : NULL, HS_RESET); + return; + }; + + OUTB (nc_ctest3, np->rv_ctest3 | CLF); /* clear dma fifo */ + OUTB (nc_stest3, TE|CSF); /* clear scsi fifo */ + + if (!(sist & (GEN|HTH|SGE)) && + !(dstat & (MDPE|BF|ABRT|IID))) { + if (sist & SBMC) ncr_int_sbmc (np); + else if (sist & STO) ncr_int_sto (np); + else if (sist & UDC) ncr_int_udc (np); + else goto unknown_int; + return; + }; + + /*========================================================= + ** Now, interrupts we are not able to recover cleanly. + ** + ** Do the register dump. + ** Log message for hard errors. + ** Reset everything. + **========================================================= + */ + if (ktime_exp(np->regtime)) { + np->regtime = ktime_get(10*HZ); + for (i = 0; i<sizeof(np->regdump); i++) + ((char*)&np->regdump)[i] = INB_OFF(i); + np->regdump.nc_dstat = dstat; + np->regdump.nc_sist = sist; + }; + + ncr_log_hard_error(np, sist, dstat); + + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + u_char ctest4_o, ctest4_m; + u_char shadow; + + /* + * Get shadow register data + * Write 1 to ctest4 + */ + ctest4_o = INB(nc_ctest4); + + OUTB(nc_ctest4, ctest4_o | 0x10); + + ctest4_m = INB(nc_ctest4); + shadow = INW_OFF(0x42); + + OUTB(nc_ctest4, ctest4_o); + + printk("%s: ctest4/sist original 0x%x/0x%X mod: 0x%X/0x%x\n", + ncr_name(np), ctest4_o, sist, ctest4_m, shadow); + } + + if ((sist & (GEN|HTH|SGE)) || + (dstat & (MDPE|BF|ABRT|IID))) { + ncr_start_reset(np); + return; + }; + +unknown_int: + /*========================================================= + ** We just miss the cause of the interrupt. :( + ** Print a message. The timeout will do the real work. + **========================================================= + */ + printk( "%s: unknown interrupt(s) ignored, " + "ISTAT=0x%x DSTAT=0x%x SIST=0x%x\n", + ncr_name(np), istat, dstat, sist); +} + + +/*========================================================== +** +** generic recovery from scsi interrupt +** +**========================================================== +** +** The doc says that when the chip gets an SCSI interrupt, +** it tries to stop in an orderly fashion, by completing +** an instruction fetch that had started or by flushing +** the DMA fifo for a write to memory that was executing. +** Such a fashion is not enough to know if the instruction +** that was just before the current DSP value has been +** executed or not. +** +** There are 3 small SCRIPTS sections that deal with the +** start queue and the done queue that may break any +** assomption from the C code if we are interrupted +** inside, so we reset if it happens. Btw, since these +** SCRIPTS sections are executed while the SCRIPTS hasn't +** started SCSI operations, it is very unlikely to happen. +** +** All the driver data structures are supposed to be +** allocated from the same 4 GB memory window, so there +** is a 1 to 1 relationship between DSA and driver data +** structures. Since we are careful :) to invalidate the +** DSA when we complete a command or when the SCRIPTS +** pushes a DSA into a queue, we can trust it when it +** points to a CCB. +** +**---------------------------------------------------------- +*/ +static void ncr_recover_scsi_int (ncb_p np, u_char hsts) +{ + u_int32 dsp = INL (nc_dsp); + u_int32 dsa = INL (nc_dsa); + ccb_p cp = ncr_ccb_from_dsa(np, dsa); + + /* + ** If we haven't been interrupted inside the SCRIPTS + ** critical pathes, we can safely restart the SCRIPTS + ** and trust the DSA value if it matches a CCB. + */ + if ((!(dsp > NCB_SCRIPT_PHYS (np, getjob_begin) && + dsp < NCB_SCRIPT_PHYS (np, getjob_end) + 1)) && + (!(dsp > NCB_SCRIPT_PHYS (np, ungetjob) && + dsp < NCB_SCRIPT_PHYS (np, reselect) + 1)) && + (!(dsp > NCB_SCRIPTH_PHYS (np, sel_for_abort) && + dsp < NCB_SCRIPTH_PHYS (np, sel_for_abort_1) + 1)) && + (!(dsp > NCB_SCRIPT_PHYS (np, done) && + dsp < NCB_SCRIPT_PHYS (np, done_end) + 1))) { + if (cp) { + cp->host_status = hsts; + ncr_complete (np, cp); + } + OUTL (nc_dsa, DSA_INVALID); + OUTB (nc_ctest3, np->rv_ctest3 | CLF); /* clear dma fifo */ + OUTB (nc_stest3, TE|CSF); /* clear scsi fifo */ + OUTL_DSP (NCB_SCRIPT_PHYS (np, start)); + } + else + goto reset_all; + + return; + +reset_all: + ncr_start_reset(np); +} + +/*========================================================== +** +** ncr chip exception handler for selection timeout +** +**========================================================== +** +** There seems to be a bug in the 53c810. +** Although a STO-Interrupt is pending, +** it continues executing script commands. +** But it will fail and interrupt (IID) on +** the next instruction where it's looking +** for a valid phase. +** +**---------------------------------------------------------- +*/ + +void ncr_int_sto (ncb_p np) +{ + u_int32 dsp = INL (nc_dsp); + + if (DEBUG_FLAGS & DEBUG_TINY) printk ("T"); + + if (dsp == NCB_SCRIPT_PHYS (np, wf_sel_done) + 8 || + !(driver_setup.recovery & 1)) + ncr_recover_scsi_int(np, HS_SEL_TIMEOUT); + else + ncr_start_reset(np); +} + +/*========================================================== +** +** ncr chip exception handler for unexpected disconnect +** +**========================================================== +** +**---------------------------------------------------------- +*/ +void ncr_int_udc (ncb_p np) +{ + u_int32 dsa = INL (nc_dsa); + ccb_p cp = ncr_ccb_from_dsa(np, dsa); + tcb_p tp = &np->target[cp->target]; + + /* + * Fix Up. Some disks respond to a PPR negotation with + * a bus free instead of a message reject. + * Disable ppr negotiation if this is first time + * tried ppr negotiation. + */ + + if (tp->ppr_negotiation == 1) + tp->ppr_negotiation = 0; + + printk ("%s: unexpected disconnect\n", ncr_name(np)); + ncr_recover_scsi_int(np, HS_UNEXPECTED); +} + +/*========================================================== +** +** ncr chip exception handler for SCSI bus mode change +** +**========================================================== +** +** spi2-r12 11.2.3 says a transceiver mode change must +** generate a reset event and a device that detects a reset +** event shall initiate a hard reset. It says also that a +** device that detects a mode change shall set data transfer +** mode to eight bit asynchronous, etc... +** So, just resetting should be enough. +** +** +**---------------------------------------------------------- +*/ + +static void ncr_int_sbmc (ncb_p np) +{ + u_char scsi_mode = INB (nc_stest4) & SMODE; + + printk("%s: SCSI bus mode change from %x to %x.\n", + ncr_name(np), np->scsi_mode, scsi_mode); + + np->scsi_mode = scsi_mode; + + + /* + ** Suspend command processing for 1 second and + ** reinitialize all except the chip. + */ + np->settle_time = ktime_get(1*HZ); + ncr_init (np, 0, bootverbose ? "scsi mode change" : NULL, HS_RESET); +} + +/*========================================================== +** +** ncr chip exception handler for SCSI parity error. +** +**========================================================== +** +** When the chip detects a SCSI parity error and is +** currently executing a (CH)MOV instruction, it does +** not interrupt immediately, but tries to finish the +** transfer of the current scatter entry before +** interrupting. The following situations may occur: +** +** - The complete scatter entry has been transferred +** without the device having changed phase. +** The chip will then interrupt with the DSP pointing +** to the instruction that follows the MOV. +** +** - A phase mismatch occurs before the MOV finished +** and phase errors are to be handled by the C code. +** The chip will then interrupt with both PAR and MA +** conditions set. +** +** - A phase mismatch occurs before the MOV finished and +** phase errors are to be handled by SCRIPTS (895A or 896). +** The chip will load the DSP with the phase mismatch +** JUMP address and interrupt the host processor. +** +**---------------------------------------------------------- +*/ + +static void ncr_int_par (ncb_p np, u_short sist) +{ + u_char hsts = INB (HS_PRT); + u_int32 dsp = INL (nc_dsp); + u_int32 dbc = INL (nc_dbc); + u_int32 dsa = INL (nc_dsa); + u_char sbcl = INB (nc_sbcl); + u_char cmd = dbc >> 24; + int phase = cmd & 7; + ccb_p cp = ncr_ccb_from_dsa(np, dsa); + + printk("%s: SCSI parity error detected: SCR1=%d DBC=%x SBCL=%x\n", + ncr_name(np), hsts, dbc, sbcl); + + /* + ** Check that the chip is connected to the SCSI BUS. + */ + if (!(INB (nc_scntl1) & ISCON)) { + if (!(driver_setup.recovery & 1)) { + ncr_recover_scsi_int(np, HS_FAIL); + return; + } + goto reset_all; + } + + /* + ** If the nexus is not clearly identified, reset the bus. + ** We will try to do better later. + */ + if (!cp) + goto reset_all; + + /* + ** Check instruction was a MOV, direction was INPUT and + ** ATN is asserted. + */ + if ((cmd & 0xc0) || !(phase & 1) || !(sbcl & 0x8)) + goto reset_all; + + /* + ** Keep track of the parity error. + */ + OUTONB (HF_PRT, HF_EXT_ERR); + cp->xerr_status |= XE_PARITY_ERR; + + /* + ** Prepare the message to send to the device. + */ + np->msgout[0] = (phase == 7) ? M_PARITY : M_ID_ERROR; + +#ifdef SCSI_NCR_INTEGRITY_CHECKING + /* + ** Save error message. For integrity check use only. + */ + if (np->check_integrity) + np->check_integ_par = np->msgout[0]; +#endif + + /* + ** If the old phase was DATA IN or DT DATA IN phase, + ** we have to deal with the 3 situations described above. + ** For other input phases (MSG IN and STATUS), the device + ** must resend the whole thing that failed parity checking + ** or signal error. So, jumping to dispatcher should be OK. + */ + if ((phase == 1) || (phase == 5)) { + /* Phase mismatch handled by SCRIPTS */ + if (dsp == NCB_SCRIPTH_PHYS (np, pm_handle)) + OUTL_DSP (dsp); + /* Phase mismatch handled by the C code */ + else if (sist & MA) + ncr_int_ma (np); + /* No phase mismatch occurred */ + else { + OUTL (nc_temp, dsp); + OUTL_DSP (NCB_SCRIPT_PHYS (np, dispatch)); + } + } + else + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + return; + +reset_all: + ncr_start_reset(np); + return; +} + +/*========================================================== +** +** +** ncr chip exception handler for phase errors. +** +** +**========================================================== +** +** We have to construct a new transfer descriptor, +** to transfer the rest of the current block. +** +**---------------------------------------------------------- +*/ + +static void ncr_int_ma (ncb_p np) +{ + u_int32 dbc; + u_int32 rest; + u_int32 dsp; + u_int32 dsa; + u_int32 nxtdsp; + u_int32 *vdsp; + u_int32 oadr, olen; + u_int32 *tblp; + u_int32 newcmd; + u_int delta; + u_char cmd; + u_char hflags, hflags0; + struct pm_ctx *pm; + ccb_p cp; + + dsp = INL (nc_dsp); + dbc = INL (nc_dbc); + dsa = INL (nc_dsa); + + cmd = dbc >> 24; + rest = dbc & 0xffffff; + delta = 0; + + /* + ** locate matching cp. + */ + cp = ncr_ccb_from_dsa(np, dsa); + + if (DEBUG_FLAGS & DEBUG_PHASE) + printk("CCB = %2x %2x %2x %2x %2x %2x\n", + cp->cmd->cmnd[0], cp->cmd->cmnd[1], cp->cmd->cmnd[2], + cp->cmd->cmnd[3], cp->cmd->cmnd[4], cp->cmd->cmnd[5]); + + /* + ** Donnot take into account dma fifo and various buffers in + ** INPUT phase since the chip flushes everything before + ** raising the MA interrupt for interrupted INPUT phases. + ** For DATA IN phase, we will check for the SWIDE later. + */ + if ((cmd & 7) != 1 && (cmd & 7) != 5) { + u_int32 dfifo; + u_char ss0, ss2; + + /* + ** If C1010, DFBC contains number of bytes in DMA fifo. + ** else read DFIFO, CTEST[4-6] using 1 PCI bus ownership. + */ + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) + delta = INL(nc_dfbc) & 0xffff; + else { + dfifo = INL(nc_dfifo); + + /* + ** Calculate remaining bytes in DMA fifo. + ** C1010 - always large fifo, value in dfbc + ** Otherwise, (CTEST5 = dfifo >> 16) + */ + if (dfifo & (DFS << 16)) + delta = ((((dfifo >> 8) & 0x300) | + (dfifo & 0xff)) - rest) & 0x3ff; + else + delta = ((dfifo & 0xff) - rest) & 0x7f; + + /* + ** The data in the dma fifo has not been + ** transferred to the target -> add the amount + ** to the rest and clear the data. + ** Check the sstat2 register in case of wide + ** transfer. + */ + + } + + rest += delta; + ss0 = INB (nc_sstat0); + if (ss0 & OLF) rest++; + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66) && (ss0 & ORF)) + rest++; + if (cp && (cp->phys.select.sel_scntl3 & EWS)) { + ss2 = INB (nc_sstat2); + if (ss2 & OLF1) rest++; + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66) && (ss2 & ORF)) + rest++; + }; + + /* + ** Clear fifos. + */ + OUTB (nc_ctest3, np->rv_ctest3 | CLF); /* dma fifo */ + OUTB (nc_stest3, TE|CSF); /* scsi fifo */ + } + + /* + ** log the information + */ + + if (DEBUG_FLAGS & (DEBUG_TINY|DEBUG_PHASE)) + printk ("P%x%x RL=%d D=%d ", cmd&7, INB(nc_sbcl)&7, + (unsigned) rest, (unsigned) delta); + + /* + ** try to find the interrupted script command, + ** and the address at which to continue. + */ + vdsp = 0; + nxtdsp = 0; + if (dsp > np->p_script && + dsp <= np->p_script + sizeof(struct script)) { + vdsp = (u_int32 *)((char*)np->script0 + (dsp-np->p_script-8)); + nxtdsp = dsp; + } + else if (dsp > np->p_scripth && + dsp <= np->p_scripth + sizeof(struct scripth)) { + vdsp = (u_int32 *)((char*)np->scripth0 + (dsp-np->p_scripth-8)); + nxtdsp = dsp; + } + + /* + ** log the information + */ + if (DEBUG_FLAGS & DEBUG_PHASE) { + printk ("\nCP=%p DSP=%x NXT=%x VDSP=%p CMD=%x ", + cp, (unsigned)dsp, (unsigned)nxtdsp, vdsp, cmd); + }; + + if (!vdsp) { + printk ("%s: interrupted SCRIPT address not found.\n", + ncr_name (np)); + goto reset_all; + } + + if (!cp) { + printk ("%s: SCSI phase error fixup: CCB already dequeued.\n", + ncr_name (np)); + goto reset_all; + } + + /* + ** get old startaddress and old length. + */ + + oadr = scr_to_cpu(vdsp[1]); + + if (cmd & 0x10) { /* Table indirect */ + tblp = (u_int32 *) ((char*) &cp->phys + oadr); + olen = scr_to_cpu(tblp[0]); + oadr = scr_to_cpu(tblp[1]); + } else { + tblp = (u_int32 *) 0; + olen = scr_to_cpu(vdsp[0]) & 0xffffff; + }; + + if (DEBUG_FLAGS & DEBUG_PHASE) { + printk ("OCMD=%x\nTBLP=%p OLEN=%x OADR=%x\n", + (unsigned) (scr_to_cpu(vdsp[0]) >> 24), + tblp, + (unsigned) olen, + (unsigned) oadr); + }; + + /* + ** check cmd against assumed interrupted script command. + ** If dt data phase, the MOVE instruction hasn't bit 4 of + ** the phase. + */ + + if (((cmd & 2) ? cmd : (cmd & ~4)) != (scr_to_cpu(vdsp[0]) >> 24)) { + PRINT_ADDR(cp->cmd); + printk ("internal error: cmd=%02x != %02x=(vdsp[0] >> 24)\n", + (unsigned)cmd, (unsigned)scr_to_cpu(vdsp[0]) >> 24); + + goto reset_all; + }; + + /* + ** if old phase not dataphase, leave here. + ** C/D line is low if data. + */ + + if (cmd & 0x02) { + PRINT_ADDR(cp->cmd); + printk ("phase change %x-%x %d@%08x resid=%d.\n", + cmd&7, INB(nc_sbcl)&7, (unsigned)olen, + (unsigned)oadr, (unsigned)rest); + goto unexpected_phase; + }; + + /* + ** Choose the correct PM save area. + ** + ** Look at the PM_SAVE SCRIPT if you want to understand + ** this stuff. The equivalent code is implemented in + ** SCRIPTS for the 895A and 896 that are able to handle + ** PM from the SCRIPTS processor. + */ + + hflags0 = INB (HF_PRT); + hflags = hflags0; + + if (hflags & (HF_IN_PM0 | HF_IN_PM1 | HF_DP_SAVED)) { + if (hflags & HF_IN_PM0) + nxtdsp = scr_to_cpu(cp->phys.pm0.ret); + else if (hflags & HF_IN_PM1) + nxtdsp = scr_to_cpu(cp->phys.pm1.ret); + + if (hflags & HF_DP_SAVED) + hflags ^= HF_ACT_PM; + } + + if (!(hflags & HF_ACT_PM)) { + pm = &cp->phys.pm0; + newcmd = NCB_SCRIPT_PHYS(np, pm0_data); + } + else { + pm = &cp->phys.pm1; + newcmd = NCB_SCRIPT_PHYS(np, pm1_data); + } + + hflags &= ~(HF_IN_PM0 | HF_IN_PM1 | HF_DP_SAVED); + if (hflags != hflags0) + OUTB (HF_PRT, hflags); + + /* + ** fillin the phase mismatch context + */ + + pm->sg.addr = cpu_to_scr(oadr + olen - rest); + pm->sg.size = cpu_to_scr(rest); + pm->ret = cpu_to_scr(nxtdsp); + + /* + ** If we have a SWIDE, + ** - prepare the address to write the SWIDE from SCRIPTS, + ** - compute the SCRIPTS address to restart from, + ** - move current data pointer context by one byte. + */ + nxtdsp = NCB_SCRIPT_PHYS (np, dispatch); + if ( ((cmd & 7) == 1 || (cmd & 7) == 5) + && cp && (cp->phys.select.sel_scntl3 & EWS) && + (INB (nc_scntl2) & WSR)) { + u32 tmp; + +#ifdef SYM_DEBUG_PM_WITH_WSR + PRINT_ADDR(cp); + printf ("MA interrupt with WSR set - " + "pm->sg.addr=%x - pm->sg.size=%d\n", + pm->sg.addr, pm->sg.size); +#endif + /* + * Set up the table indirect for the MOVE + * of the residual byte and adjust the data + * pointer context. + */ + tmp = scr_to_cpu(pm->sg.addr); + cp->phys.wresid.addr = cpu_to_scr(tmp); + pm->sg.addr = cpu_to_scr(tmp + 1); + tmp = scr_to_cpu(pm->sg.size); + cp->phys.wresid.size = cpu_to_scr((tmp&0xff000000) | 1); + pm->sg.size = cpu_to_scr(tmp - 1); + + /* + * If only the residual byte is to be moved, + * no PM context is needed. + */ + if ((tmp&0xffffff) == 1) + newcmd = pm->ret; + + /* + * Prepare the address of SCRIPTS that will + * move the residual byte to memory. + */ + nxtdsp = NCB_SCRIPTH_PHYS (np, wsr_ma_helper); + } + + if (DEBUG_FLAGS & DEBUG_PHASE) { + PRINT_ADDR(cp->cmd); + printk ("PM %x %x %x / %x %x %x.\n", + hflags0, hflags, newcmd, + (unsigned)scr_to_cpu(pm->sg.addr), + (unsigned)scr_to_cpu(pm->sg.size), + (unsigned)scr_to_cpu(pm->ret)); + } + + /* + ** Restart the SCRIPTS processor. + */ + + OUTL (nc_temp, newcmd); + OUTL_DSP (nxtdsp); + return; + + /* + ** Unexpected phase changes that occurs when the current phase + ** is not a DATA IN or DATA OUT phase are due to error conditions. + ** Such event may only happen when the SCRIPTS is using a + ** multibyte SCSI MOVE. + ** + ** Phase change Some possible cause + ** + ** COMMAND --> MSG IN SCSI parity error detected by target. + ** COMMAND --> STATUS Bad command or refused by target. + ** MSG OUT --> MSG IN Message rejected by target. + ** MSG OUT --> COMMAND Bogus target that discards extended + ** negotiation messages. + ** + ** The code below does not care of the new phase and so + ** trusts the target. Why to annoy it ? + ** If the interrupted phase is COMMAND phase, we restart at + ** dispatcher. + ** If a target does not get all the messages after selection, + ** the code assumes blindly that the target discards extended + ** messages and clears the negotiation status. + ** If the target does not want all our response to negotiation, + ** we force a SIR_NEGO_PROTO interrupt (it is a hack that avoids + ** bloat for such a should_not_happen situation). + ** In all other situation, we reset the BUS. + ** Are these assumptions reasonnable ? (Wait and see ...) + */ +unexpected_phase: + dsp -= 8; + nxtdsp = 0; + + switch (cmd & 7) { + case 2: /* COMMAND phase */ + nxtdsp = NCB_SCRIPT_PHYS (np, dispatch); + break; +#if 0 + case 3: /* STATUS phase */ + nxtdsp = NCB_SCRIPT_PHYS (np, dispatch); + break; +#endif + case 6: /* MSG OUT phase */ + /* + ** If the device may want to use untagged when we want + ** tagged, we prepare an IDENTIFY without disc. granted, + ** since we will not be able to handle reselect. + ** Otherwise, we just don't care. + */ + if (dsp == NCB_SCRIPT_PHYS (np, send_ident)) { + if (cp->tag != NO_TAG && olen - rest <= 3) { + cp->host_status = HS_BUSY; + np->msgout[0] = M_IDENTIFY | cp->lun; + nxtdsp = NCB_SCRIPTH_PHYS (np, ident_break_atn); + } + else + nxtdsp = NCB_SCRIPTH_PHYS (np, ident_break); + } + else if (dsp == NCB_SCRIPTH_PHYS (np, send_wdtr) || + dsp == NCB_SCRIPTH_PHYS (np, send_sdtr) || + dsp == NCB_SCRIPTH_PHYS (np, send_ppr)) { + nxtdsp = NCB_SCRIPTH_PHYS (np, nego_bad_phase); + } + break; +#if 0 + case 7: /* MSG IN phase */ + nxtdsp = NCB_SCRIPT_PHYS (np, clrack); + break; +#endif + } + + if (nxtdsp) { + OUTL_DSP (nxtdsp); + return; + } + +reset_all: + ncr_start_reset(np); +} + +/*========================================================== +** +** ncr chip handler for QUEUE FULL and CHECK CONDITION +** +**========================================================== +** +** On QUEUE FULL status, we set the actual tagged command +** queue depth to the number of disconnected CCBs that is +** hopefully a good value to avoid further QUEUE FULL. +** +** On CHECK CONDITION or COMMAND TERMINATED, we use the +** CCB of the failed command for performing a REQUEST +** SENSE SCSI command. +** +** We do not want to change the order commands will be +** actually queued to the device after we received a +** QUEUE FULL status. We also want to properly deal with +** contingent allegiance condition. For these reasons, +** we remove from the start queue all commands for this +** LUN that haven't been yet queued to the device and +** put them back in the correponding LUN queue, then +** requeue the CCB that failed in front of the LUN queue. +** I just hope this not to be performed too often. :) +** +** If we are using IMMEDIATE ARBITRATION, we clear the +** IARB hint for every commands we encounter in order not +** to be stuck with a won arbitration and no job to queue +** to a device. +**---------------------------------------------------------- +*/ + +static void ncr_sir_to_redo(ncb_p np, int num, ccb_p cp) +{ + Scsi_Cmnd *cmd = cp->cmd; + tcb_p tp = &np->target[cp->target]; + lcb_p lp = ncr_lp(np, tp, cp->lun); + ccb_p cp2; + int busyccbs = 1; + u_int32 startp; + u_char s_status = INB (SS_PRT); + int msglen; + int i, j; + + + /* + ** If the LCB is not yet available, then only + ** 1 IO is accepted, so we should have it. + */ + if (!lp) + goto next; + /* + ** Remove all CCBs queued to the chip for that LUN and put + ** them back in the LUN CCB wait queue. + */ + busyccbs = lp->queuedccbs; + i = (INL (nc_scratcha) - np->p_squeue) / 4; + j = i; + while (i != np->squeueput) { + cp2 = ncr_ccb_from_dsa(np, scr_to_cpu(np->squeue[i])); + assert(cp2); +#ifdef SCSI_NCR_IARB_SUPPORT + /* IARB hints may not be relevant any more. Forget them. */ + cp2->host_flags &= ~HF_HINT_IARB; +#endif + if (cp2 && cp2->target == cp->target && cp2->lun == cp->lun) { + xpt_remque(&cp2->link_ccbq); + xpt_insque_head(&cp2->link_ccbq, &lp->wait_ccbq); + --lp->queuedccbs; + cp2->queued = 0; + } + else { + if (i != j) + np->squeue[j] = np->squeue[i]; + if ((j += 2) >= MAX_START*2) j = 0; + } + if ((i += 2) >= MAX_START*2) i = 0; + } + if (i != j) /* Copy back the idle task if needed */ + np->squeue[j] = np->squeue[i]; + np->squeueput = j; /* Update our current start queue pointer */ + + /* + ** Requeue the interrupted CCB in front of the + ** LUN CCB wait queue to preserve ordering. + */ + xpt_remque(&cp->link_ccbq); + xpt_insque_head(&cp->link_ccbq, &lp->wait_ccbq); + --lp->queuedccbs; + cp->queued = 0; + +next: + +#ifdef SCSI_NCR_IARB_SUPPORT + /* IARB hint may not be relevant any more. Forget it. */ + cp->host_flags &= ~HF_HINT_IARB; + if (np->last_cp) + np->last_cp = 0; +#endif + + /* + ** Now we can restart the SCRIPTS processor safely. + */ + OUTL_DSP (NCB_SCRIPT_PHYS (np, start)); + + switch(s_status) { + default: + case S_BUSY: + ncr_complete(np, cp); + break; + case S_QUEUE_FULL: + if (!lp || !lp->queuedccbs) { + ncr_complete(np, cp); + break; + } + if (bootverbose >= 1) { + PRINT_ADDR(cmd); + printk ("QUEUE FULL! %d busy, %d disconnected CCBs\n", + busyccbs, lp->queuedccbs); + } + /* + ** Decrease number of tags to the number of + ** disconnected commands. + */ + if (lp->queuedccbs < lp->numtags) { + lp->numtags = lp->queuedccbs; + lp->num_good = 0; + ncr_setup_tags (np, cp->target, cp->lun); + } + /* + ** Repair the offending CCB. + */ + cp->phys.header.savep = cp->startp; + cp->phys.header.lastp = cp->lastp0; + cp->host_status = HS_BUSY; + cp->scsi_status = S_ILLEGAL; + cp->xerr_status = 0; + cp->extra_bytes = 0; + cp->host_flags &= (HF_PM_TO_C|HF_DATA_IN); + + break; + + case S_TERMINATED: + case S_CHECK_COND: + /* + ** If we were requesting sense, give up. + */ + if (cp->host_flags & HF_AUTO_SENSE) { + ncr_complete(np, cp); + break; + } + + /* + ** Save SCSI status and extended error. + ** Compute the data residual now. + */ + cp->sv_scsi_status = cp->scsi_status; + cp->sv_xerr_status = cp->xerr_status; + cp->resid = ncr_compute_residual(np, cp); + + /* + ** Device returned CHECK CONDITION status. + ** Prepare all needed data strutures for getting + ** sense data. + */ + + /* + ** identify message + */ + cp->scsi_smsg2[0] = M_IDENTIFY | cp->lun; + msglen = 1; + + /* + ** If we are currently using anything different from + ** async. 8 bit data transfers with that target, + ** start a negotiation, since the device may want + ** to report us a UNIT ATTENTION condition due to + ** a cause we currently ignore, and we donnot want + ** to be stuck with WIDE and/or SYNC data transfer. + ** + ** cp->nego_status is filled by ncr_prepare_nego(). + ** + ** Do NOT negotiate if performing integrity check + ** or if integrity check has completed, all check + ** conditions will have been cleared. + */ + +#ifdef SCSI_NCR_INTEGRITY_CHECKING + if (DEBUG_FLAGS & DEBUG_IC) { + printk("%s: ncr_sir_to_redo: ic_done %2X, in_progress %2X\n", + ncr_name(np), tp->ic_done, cp->cmd->ic_in_progress); + } + + /* + ** If parity error during integrity check, + ** set the target width to narrow. Otherwise, + ** do not negotiate on a request sense. + */ + if ( np->check_integ_par && np->check_integrity + && cp->cmd->ic_in_progress ) { + cp->nego_status = 0; + msglen += + ncr_ic_nego (np, cp, cmd ,&cp->scsi_smsg2[msglen]); + } + + if (!np->check_integrity || + (np->check_integrity && + (!cp->cmd->ic_in_progress && !tp->ic_done)) ) { + ncr_negotiate(np, tp); + cp->nego_status = 0; + { + u_char sync_offset; + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) + sync_offset = tp->sval & 0x3f; + else + sync_offset = tp->sval & 0x1f; + + if ((tp->wval & EWS) || sync_offset) + msglen += + ncr_prepare_nego (np, cp, &cp->scsi_smsg2[msglen]); + } + + } +#else + ncr_negotiate(np, tp); + cp->nego_status = 0; + if ((tp->wval & EWS) || (tp->sval & 0x1f)) + msglen += + ncr_prepare_nego (np, cp, &cp->scsi_smsg2[msglen]); +#endif /* SCSI_NCR_INTEGRITY_CHECKING */ + + /* + ** Message table indirect structure. + */ + cp->phys.smsg.addr = cpu_to_scr(CCB_PHYS (cp, scsi_smsg2)); + cp->phys.smsg.size = cpu_to_scr(msglen); + + /* + ** sense command + */ + cp->phys.cmd.addr = cpu_to_scr(CCB_PHYS (cp, sensecmd)); + cp->phys.cmd.size = cpu_to_scr(6); + + /* + ** patch requested size into sense command + */ + cp->sensecmd[0] = 0x03; + cp->sensecmd[1] = cp->lun << 5; + cp->sensecmd[4] = sizeof(cp->sense_buf); + + /* + ** sense data + */ + bzero(cp->sense_buf, sizeof(cp->sense_buf)); + cp->phys.sense.addr = cpu_to_scr(CCB_PHYS(cp,sense_buf[0])); + cp->phys.sense.size = cpu_to_scr(sizeof(cp->sense_buf)); + + /* + ** requeue the command. + */ + startp = NCB_SCRIPTH_PHYS (np, sdata_in); + + cp->phys.header.savep = cpu_to_scr(startp); + cp->phys.header.goalp = cpu_to_scr(startp + 16); + cp->phys.header.lastp = cpu_to_scr(startp); + cp->phys.header.wgoalp = cpu_to_scr(startp + 16); + cp->phys.header.wlastp = cpu_to_scr(startp); + + cp->host_status = cp->nego_status ? HS_NEGOTIATE : HS_BUSY; + cp->scsi_status = S_ILLEGAL; + cp->host_flags = (HF_AUTO_SENSE|HF_DATA_IN); + + cp->phys.header.go.start = + cpu_to_scr(NCB_SCRIPT_PHYS (np, select)); + + /* + ** If lp not yet allocated, requeue the command. + */ + if (!lp) + ncr_put_start_queue(np, cp); + break; + } + + /* + ** requeue awaiting scsi commands for this lun. + */ + if (lp) + ncr_start_next_ccb(np, lp, 1); + + return; +} + +/*---------------------------------------------------------- +** +** After a device has accepted some management message +** as BUS DEVICE RESET, ABORT TASK, etc ..., or when +** a device signals a UNIT ATTENTION condition, some +** tasks are thrown away by the device. We are required +** to reflect that on our tasks list since the device +** will never complete these tasks. +** +** This function completes all disconnected CCBs for a +** given target that matches the following criteria: +** - lun=-1 means any logical UNIT otherwise a given one. +** - task=-1 means any task, otherwise a given one. +**---------------------------------------------------------- +*/ +static int ncr_clear_tasks(ncb_p np, u_char hsts, + int target, int lun, int task) +{ + int i = 0; + ccb_p cp; + + for (cp = np->ccbc; cp; cp = cp->link_ccb) { + if (cp->host_status != HS_DISCONNECT) + continue; + if (cp->target != target) + continue; + if (lun != -1 && cp->lun != lun) + continue; + if (task != -1 && cp->tag != NO_TAG && cp->scsi_smsg[2] != task) + continue; + cp->host_status = hsts; + cp->scsi_status = S_ILLEGAL; + ncr_complete(np, cp); + ++i; + } + return i; +} + +/*========================================================== +** +** ncr chip handler for TASKS recovery. +** +**========================================================== +** +** We cannot safely abort a command, while the SCRIPTS +** processor is running, since we just would be in race +** with it. +** +** As long as we have tasks to abort, we keep the SEM +** bit set in the ISTAT. When this bit is set, the +** SCRIPTS processor interrupts (SIR_SCRIPT_STOPPED) +** each time it enters the scheduler. +** +** If we have to reset a target, clear tasks of a unit, +** or to perform the abort of a disconnected job, we +** restart the SCRIPTS for selecting the target. Once +** selected, the SCRIPTS interrupts (SIR_TARGET_SELECTED). +** If it loses arbitration, the SCRIPTS will interrupt again +** the next time it will enter its scheduler, and so on ... +** +** On SIR_TARGET_SELECTED, we scan for the more +** appropriate thing to do: +** +** - If nothing, we just sent a M_ABORT message to the +** target to get rid of the useless SCSI bus ownership. +** According to the specs, no tasks shall be affected. +** - If the target is to be reset, we send it a M_RESET +** message. +** - If a logical UNIT is to be cleared , we send the +** IDENTIFY(lun) + M_ABORT. +** - If an untagged task is to be aborted, we send the +** IDENTIFY(lun) + M_ABORT. +** - If a tagged task is to be aborted, we send the +** IDENTIFY(lun) + task attributes + M_ABORT_TAG. +** +** Once our 'kiss of death' :) message has been accepted +** by the target, the SCRIPTS interrupts again +** (SIR_ABORT_SENT). On this interrupt, we complete +** all the CCBs that should have been aborted by the +** target according to our message. +** +**---------------------------------------------------------- +*/ +static void ncr_sir_task_recovery(ncb_p np, int num) +{ + ccb_p cp; + tcb_p tp; + int target=-1, lun=-1, task; + int i, k; + u_char *p; + + switch(num) { + /* + ** The SCRIPTS processor stopped before starting + ** the next command in order to allow us to perform + ** some task recovery. + */ + case SIR_SCRIPT_STOPPED: + + /* + ** Do we have any target to reset or unit to clear ? + */ + for (i = 0 ; i < MAX_TARGET ; i++) { + tp = &np->target[i]; + if (tp->to_reset || (tp->l0p && tp->l0p->to_clear)) { + target = i; + break; + } + if (!tp->lmp) + continue; + for (k = 1 ; k < MAX_LUN ; k++) { + if (tp->lmp[k] && tp->lmp[k]->to_clear) { + target = i; + break; + } + } + if (target != -1) + break; + } + + /* + ** If not, look at the CCB list for any + ** disconnected CCB to be aborted. + */ + if (target == -1) { + for (cp = np->ccbc; cp; cp = cp->link_ccb) { + if (cp->host_status != HS_DISCONNECT) + continue; + if (cp->to_abort) { + target = cp->target; + break; + } + } + } + + /* + ** If some target is to be selected, + ** prepare and start the selection. + */ + if (target != -1) { + tp = &np->target[target]; + np->abrt_sel.sel_id = target; + np->abrt_sel.sel_scntl3 = tp->wval; + np->abrt_sel.sel_sxfer = tp->sval; + np->abrt_sel.sel_scntl4 = tp->uval; + OUTL(nc_dsa, np->p_ncb); + OUTL_DSP (NCB_SCRIPTH_PHYS (np, sel_for_abort)); + return; + } + + /* + ** Nothing is to be selected, so we donnot need + ** to synchronize with the SCRIPTS anymore. + ** Remove the SEM flag from the ISTAT. + */ + np->istat_sem = 0; + OUTB (nc_istat, SIGP); + + /* + ** Now look at CCBs to abort that haven't started yet. + ** Remove all those CCBs from the start queue and + ** complete them with appropriate status. + ** Btw, the SCRIPTS processor is still stopped, so + ** we are not in race. + */ + for (cp = np->ccbc; cp; cp = cp->link_ccb) { + if (cp->host_status != HS_BUSY && + cp->host_status != HS_NEGOTIATE) + continue; + if (!cp->to_abort) + continue; +#ifdef SCSI_NCR_IARB_SUPPORT + /* + ** If we are using IMMEDIATE ARBITRATION, we donnot + ** want to cancel the last queued CCB, since the + ** SCRIPTS may have anticipated the selection. + */ + if (cp == np->last_cp) { + cp->to_abort = 0; + continue; + } +#endif + /* + ** Compute index of next position in the start + ** queue the SCRIPTS will schedule. + */ + i = (INL (nc_scratcha) - np->p_squeue) / 4; + + /* + ** Remove the job from the start queue. + */ + k = -1; + while (1) { + if (i == np->squeueput) + break; + if (k == -1) { /* Not found yet */ + if (cp == ncr_ccb_from_dsa(np, + scr_to_cpu(np->squeue[i]))) + k = i; /* Found */ + } + else { + /* + ** Once found, we have to move + ** back all jobs by 1 position. + */ + np->squeue[k] = np->squeue[i]; + k += 2; + if (k >= MAX_START*2) + k = 0; + } + + i += 2; + if (i >= MAX_START*2) + i = 0; + } + if (k != -1) { + np->squeue[k] = np->squeue[i]; /* Idle task */ + np->squeueput = k; /* Start queue pointer */ + } + cp->host_status = HS_ABORTED; + cp->scsi_status = S_ILLEGAL; + ncr_complete(np, cp); + } + break; + /* + ** The SCRIPTS processor has selected a target + ** we may have some manual recovery to perform for. + */ + case SIR_TARGET_SELECTED: + target = (INB (nc_sdid) & 0xf); + tp = &np->target[target]; + + np->abrt_tbl.addr = vtobus(np->abrt_msg); + + /* + ** If the target is to be reset, prepare a + ** M_RESET message and clear the to_reset flag + ** since we donnot expect this operation to fail. + */ + if (tp->to_reset) { + np->abrt_msg[0] = M_RESET; + np->abrt_tbl.size = 1; + tp->to_reset = 0; + break; + } + + /* + ** Otherwise, look for some logical unit to be cleared. + */ + if (tp->l0p && tp->l0p->to_clear) + lun = 0; + else if (tp->lmp) { + for (k = 1 ; k < MAX_LUN ; k++) { + if (tp->lmp[k] && tp->lmp[k]->to_clear) { + lun = k; + break; + } + } + } + + /* + ** If a logical unit is to be cleared, prepare + ** an IDENTIFY(lun) + ABORT MESSAGE. + */ + if (lun != -1) { + lcb_p lp = ncr_lp(np, tp, lun); + lp->to_clear = 0; /* We donnot expect to fail here */ + np->abrt_msg[0] = M_IDENTIFY | lun; + np->abrt_msg[1] = M_ABORT; + np->abrt_tbl.size = 2; + break; + } + + /* + ** Otherwise, look for some disconnected job to + ** abort for this target. + */ + for (cp = np->ccbc; cp; cp = cp->link_ccb) { + if (cp->host_status != HS_DISCONNECT) + continue; + if (cp->target != target) + continue; + if (cp->to_abort) + break; + } + + /* + ** If we have none, probably since the device has + ** completed the command before we won abitration, + ** send a M_ABORT message without IDENTIFY. + ** According to the specs, the device must just + ** disconnect the BUS and not abort any task. + */ + if (!cp) { + np->abrt_msg[0] = M_ABORT; + np->abrt_tbl.size = 1; + break; + } + + /* + ** We have some task to abort. + ** Set the IDENTIFY(lun) + */ + np->abrt_msg[0] = M_IDENTIFY | cp->lun; + + /* + ** If we want to abort an untagged command, we + ** will send a IDENTIFY + M_ABORT. + ** Otherwise (tagged command), we will send + ** a IDENTITFY + task attributes + ABORT TAG. + */ + if (cp->tag == NO_TAG) { + np->abrt_msg[1] = M_ABORT; + np->abrt_tbl.size = 2; + } + else { + np->abrt_msg[1] = cp->scsi_smsg[1]; + np->abrt_msg[2] = cp->scsi_smsg[2]; + np->abrt_msg[3] = M_ABORT_TAG; + np->abrt_tbl.size = 4; + } + cp->to_abort = 0; /* We donnot expect to fail here */ + break; + + /* + ** The target has accepted our message and switched + ** to BUS FREE phase as we expected. + */ + case SIR_ABORT_SENT: + target = (INB (nc_sdid) & 0xf); + tp = &np->target[target]; + + /* + ** If we didn't abort anything, leave here. + */ + if (np->abrt_msg[0] == M_ABORT) + break; + + /* + ** If we sent a M_RESET, then a hardware reset has + ** been performed by the target. + ** - Reset everything to async 8 bit + ** - Tell ourself to negotiate next time :-) + ** - Prepare to clear all disconnected CCBs for + ** this target from our task list (lun=task=-1) + */ + lun = -1; + task = -1; + if (np->abrt_msg[0] == M_RESET) { + tp->sval = 0; + tp->wval = np->rv_scntl3; + tp->uval = np->rv_scntl4; + ncr_set_sync_wide_status(np, target); + ncr_negotiate(np, tp); + } + + /* + ** Otherwise, check for the LUN and TASK(s) + ** concerned by the cancelation. + ** If it is not ABORT_TAG then it is CLEAR_QUEUE + ** or an ABORT message :-) + */ + else { + lun = np->abrt_msg[0] & 0x3f; + if (np->abrt_msg[1] == M_ABORT_TAG) + task = np->abrt_msg[2]; + } + + /* + ** Complete all the CCBs the device should have + ** aborted due to our 'kiss of death' message. + */ + (void) ncr_clear_tasks(np, HS_ABORTED, target, lun, task); + break; + + /* + ** We have performed a auto-sense that succeeded. + ** If the device reports a UNIT ATTENTION condition + ** due to a RESET condition, we must complete all + ** disconnect CCBs for this unit since the device + ** shall have thrown them away. + ** Since I haven't time to guess what the specs are + ** expecting for other UNIT ATTENTION conditions, I + ** decided to only care about RESET conditions. :) + */ + case SIR_AUTO_SENSE_DONE: + cp = ncr_ccb_from_dsa(np, INL (nc_dsa)); + if (!cp) + break; + memcpy(cp->cmd->sense_buffer, cp->sense_buf, + sizeof(cp->cmd->sense_buffer)); + p = &cp->cmd->sense_buffer[0]; + + if (p[0] != 0x70 || p[2] != 0x6 || p[12] != 0x29) + break; +#if 0 + (void) ncr_clear_tasks(np, HS_RESET, cp->target, cp->lun, -1); +#endif + break; + } + + /* + ** Print to the log the message we intend to send. + */ + if (num == SIR_TARGET_SELECTED) { + PRINT_TARGET(np, target); + ncr_printl_hex("control msgout:", np->abrt_msg, + np->abrt_tbl.size); + np->abrt_tbl.size = cpu_to_scr(np->abrt_tbl.size); + } + + /* + ** Let the SCRIPTS processor continue. + */ + OUTONB_STD (); +} + + +/*========================================================== +** +** Gérard's alchemy:) that deals with with the data +** pointer for both MDP and the residual calculation. +** +**========================================================== +** +** I didn't want to bloat the code by more than 200 +** lignes for the handling of both MDP and the residual. +** This has been achieved by using a data pointer +** representation consisting in an index in the data +** array (dp_sg) and a negative offset (dp_ofs) that +** have the following meaning: +** +** - dp_sg = MAX_SCATTER +** we are at the end of the data script. +** - dp_sg < MAX_SCATTER +** dp_sg points to the next entry of the scatter array +** we want to transfer. +** - dp_ofs < 0 +** dp_ofs represents the residual of bytes of the +** previous entry scatter entry we will send first. +** - dp_ofs = 0 +** no residual to send first. +** +** The function ncr_evaluate_dp() accepts an arbitray +** offset (basically from the MDP message) and returns +** the corresponding values of dp_sg and dp_ofs. +** +**---------------------------------------------------------- +*/ + +static int ncr_evaluate_dp(ncb_p np, ccb_p cp, u_int32 scr, int *ofs) +{ + u_int32 dp_scr; + int dp_ofs, dp_sg, dp_sgmin; + int tmp; + struct pm_ctx *pm; + + /* + ** Compute the resulted data pointer in term of a script + ** address within some DATA script and a signed byte offset. + */ + dp_scr = scr; + dp_ofs = *ofs; + if (dp_scr == NCB_SCRIPT_PHYS (np, pm0_data)) + pm = &cp->phys.pm0; + else if (dp_scr == NCB_SCRIPT_PHYS (np, pm1_data)) + pm = &cp->phys.pm1; + else + pm = 0; + + if (pm) { + dp_scr = scr_to_cpu(pm->ret); + dp_ofs -= scr_to_cpu(pm->sg.size); + } + + /* + ** Deduce the index of the sg entry. + ** Keep track of the index of the first valid entry. + ** If result is dp_sg = MAX_SCATTER, then we are at the + ** end of the data and vice-versa. + */ + tmp = scr_to_cpu(cp->phys.header.goalp); + dp_sg = MAX_SCATTER; + if (dp_scr != tmp) + dp_sg -= (tmp - 8 - (int)dp_scr) / (SCR_SG_SIZE*4); + dp_sgmin = MAX_SCATTER - cp->segments; + + /* + ** Move to the sg entry the data pointer belongs to. + ** + ** If we are inside the data area, we expect result to be: + ** + ** Either, + ** dp_ofs = 0 and dp_sg is the index of the sg entry + ** the data pointer belongs to (or the end of the data) + ** Or, + ** dp_ofs < 0 and dp_sg is the index of the sg entry + ** the data pointer belongs to + 1. + */ + if (dp_ofs < 0) { + int n; + while (dp_sg > dp_sgmin) { + --dp_sg; + tmp = scr_to_cpu(cp->phys.data[dp_sg].size); + n = dp_ofs + (tmp & 0xffffff); + if (n > 0) { + ++dp_sg; + break; + } + dp_ofs = n; + } + } + else if (dp_ofs > 0) { + while (dp_sg < MAX_SCATTER) { + tmp = scr_to_cpu(cp->phys.data[dp_sg].size); + dp_ofs -= (tmp & 0xffffff); + ++dp_sg; + if (dp_ofs <= 0) + break; + } + } + + /* + ** Make sure the data pointer is inside the data area. + ** If not, return some error. + */ + if (dp_sg < dp_sgmin || (dp_sg == dp_sgmin && dp_ofs < 0)) + goto out_err; + else if (dp_sg > MAX_SCATTER || (dp_sg == MAX_SCATTER && dp_ofs > 0)) + goto out_err; + + /* + ** Save the extreme pointer if needed. + */ + if (dp_sg > cp->ext_sg || + (dp_sg == cp->ext_sg && dp_ofs > cp->ext_ofs)) { + cp->ext_sg = dp_sg; + cp->ext_ofs = dp_ofs; + } + + /* + ** Return data. + */ + *ofs = dp_ofs; + return dp_sg; + +out_err: + return -1; +} + +/*========================================================== +** +** ncr chip handler for MODIFY DATA POINTER MESSAGE +** +**========================================================== +** +** We also call this function on IGNORE WIDE RESIDUE +** messages that do not match a SWIDE full condition. +** Btw, we assume in that situation that such a message +** is equivalent to a MODIFY DATA POINTER (offset=-1). +** +**---------------------------------------------------------- +*/ + +static void ncr_modify_dp(ncb_p np, tcb_p tp, ccb_p cp, int ofs) +{ + int dp_ofs = ofs; + u_int32 dp_scr = INL (nc_temp); + u_int32 dp_ret; + u_int32 tmp; + u_char hflags; + int dp_sg; + struct pm_ctx *pm; + + /* + ** Not supported for auto_sense; + */ + if (cp->host_flags & HF_AUTO_SENSE) + goto out_reject; + + /* + ** Apply our alchemy:) (see comments in ncr_evaluate_dp()), + ** to the resulted data pointer. + */ + dp_sg = ncr_evaluate_dp(np, cp, dp_scr, &dp_ofs); + if (dp_sg < 0) + goto out_reject; + + /* + ** And our alchemy:) allows to easily calculate the data + ** script address we want to return for the next data phase. + */ + dp_ret = cpu_to_scr(cp->phys.header.goalp); + dp_ret = dp_ret - 8 - (MAX_SCATTER - dp_sg) * (SCR_SG_SIZE*4); + + /* + ** If offset / scatter entry is zero we donnot need + ** a context for the new current data pointer. + */ + if (dp_ofs == 0) { + dp_scr = dp_ret; + goto out_ok; + } + + /* + ** Get a context for the new current data pointer. + */ + hflags = INB (HF_PRT); + + if (hflags & HF_DP_SAVED) + hflags ^= HF_ACT_PM; + + if (!(hflags & HF_ACT_PM)) { + pm = &cp->phys.pm0; + dp_scr = NCB_SCRIPT_PHYS (np, pm0_data); + } + else { + pm = &cp->phys.pm1; + dp_scr = NCB_SCRIPT_PHYS (np, pm1_data); + } + + hflags &= ~(HF_DP_SAVED); + + OUTB (HF_PRT, hflags); + + /* + ** Set up the new current data pointer. + ** ofs < 0 there, and for the next data phase, we + ** want to transfer part of the data of the sg entry + ** corresponding to index dp_sg-1 prior to returning + ** to the main data script. + */ + pm->ret = cpu_to_scr(dp_ret); + tmp = scr_to_cpu(cp->phys.data[dp_sg-1].addr); + tmp += scr_to_cpu(cp->phys.data[dp_sg-1].size) + dp_ofs; + pm->sg.addr = cpu_to_scr(tmp); + pm->sg.size = cpu_to_scr(-dp_ofs); + +out_ok: + OUTL (nc_temp, dp_scr); + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + return; + +out_reject: + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); +} + + +/*========================================================== +** +** ncr chip calculation of the data residual. +** +**========================================================== +** +** As I used to say, the requirement of data residual +** in SCSI is broken, useless and cannot be achieved +** without huge complexity. +** But most OSes and even the official CAM require it. +** When stupidity happens to be so widely spread inside +** a community, it gets hard to convince. +** +** Anyway, I don't care, since I am not going to use +** any software that considers this data residual as +** a relevant information. :) +** +**---------------------------------------------------------- +*/ + +static int ncr_compute_residual(ncb_p np, ccb_p cp) +{ + int dp_sg, dp_sgmin, tmp; + int resid=0; + int dp_ofs = 0; + + /* + * Check for some data lost or just thrown away. + * We are not required to be quite accurate in this + * situation. Btw, if we are odd for output and the + * device claims some more data, it may well happen + * than our residual be zero. :-) + */ + if (cp->xerr_status & (XE_EXTRA_DATA|XE_SODL_UNRUN|XE_SWIDE_OVRUN)) { + if (cp->xerr_status & XE_EXTRA_DATA) + resid -= cp->extra_bytes; + if (cp->xerr_status & XE_SODL_UNRUN) + ++resid; + if (cp->xerr_status & XE_SWIDE_OVRUN) + --resid; + } + + + /* + ** If SCRIPTS reaches its goal point, then + ** there is no additionnal residual. + */ + if (cp->phys.header.lastp == cp->phys.header.goalp) + return resid; + + /* + ** If the last data pointer is data_io (direction + ** unknown), then no data transfer should have + ** taken place. + */ + if (cp->phys.header.lastp == NCB_SCRIPTH_PHYS (np, data_io)) + return cp->data_len; + + /* + ** If no data transfer occurs, or if the data + ** pointer is weird, return full residual. + */ + if (cp->startp == cp->phys.header.lastp || + ncr_evaluate_dp(np, cp, scr_to_cpu(cp->phys.header.lastp), + &dp_ofs) < 0) { + return cp->data_len; + } + + /* + ** We are now full comfortable in the computation + ** of the data residual (2's complement). + */ + dp_sgmin = MAX_SCATTER - cp->segments; + resid = -cp->ext_ofs; + for (dp_sg = cp->ext_sg; dp_sg < MAX_SCATTER; ++dp_sg) { + tmp = scr_to_cpu(cp->phys.data[dp_sg].size); + resid += (tmp & 0xffffff); + } + + /* + ** Hopefully, the result is not too wrong. + */ + return resid; +} + +/*========================================================== +** +** Print out the containt of a SCSI message. +** +**========================================================== +*/ + +static int ncr_show_msg (u_char * msg) +{ + u_char i; + printk ("%x",*msg); + if (*msg==M_EXTENDED) { + for (i=1;i<8;i++) { + if (i-1>msg[1]) break; + printk ("-%x",msg[i]); + }; + return (i+1); + } else if ((*msg & 0xf0) == 0x20) { + printk ("-%x",msg[1]); + return (2); + }; + return (1); +} + +static void ncr_print_msg (ccb_p cp, char *label, u_char *msg) +{ + if (cp) + PRINT_ADDR(cp->cmd); + if (label) + printk ("%s: ", label); + + (void) ncr_show_msg (msg); + printk (".\n"); +} + +/*=================================================================== +** +** Negotiation for WIDE and SYNCHRONOUS DATA TRANSFER. +** +**=================================================================== +** +** Was Sie schon immer ueber transfermode negotiation wissen wollten ... +** +** We try to negotiate sync and wide transfer only after +** a successfull inquire command. We look at byte 7 of the +** inquire data to determine the capabilities of the target. +** +** When we try to negotiate, we append the negotiation message +** to the identify and (maybe) simple tag message. +** The host status field is set to HS_NEGOTIATE to mark this +** situation. +** +** If the target doesn't answer this message immediately +** (as required by the standard), the SIR_NEGO_FAILED interrupt +** will be raised eventually. +** The handler removes the HS_NEGOTIATE status, and sets the +** negotiated value to the default (async / nowide). +** +** If we receive a matching answer immediately, we check it +** for validity, and set the values. +** +** If we receive a Reject message immediately, we assume the +** negotiation has failed, and fall back to standard values. +** +** If we receive a negotiation message while not in HS_NEGOTIATE +** state, it's a target initiated negotiation. We prepare a +** (hopefully) valid answer, set our parameters, and send back +** this answer to the target. +** +** If the target doesn't fetch the answer (no message out phase), +** we assume the negotiation has failed, and fall back to default +** settings (SIR_NEGO_PROTO interrupt). +** +** When we set the values, we adjust them in all ccbs belonging +** to this target, in the controller's register, and in the "phys" +** field of the controller's struct ncb. +** +**--------------------------------------------------------------------- +*/ + +/*========================================================== +** +** ncr chip handler for SYNCHRONOUS DATA TRANSFER +** REQUEST (SDTR) message. +** +**========================================================== +** +** Read comments above. +** +**---------------------------------------------------------- +*/ +static void ncr_sync_nego(ncb_p np, tcb_p tp, ccb_p cp) +{ + u_char scntl3, scntl4; + u_char chg, ofs, per, fak; + + /* + ** Synchronous request message received. + */ + + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "sync msg in", np->msgin); + }; + + /* + ** get requested values. + */ + + chg = 0; + per = np->msgin[3]; + ofs = np->msgin[4]; + if (ofs==0) per=255; + + /* + ** if target sends SDTR message, + ** it CAN transfer synch. + */ + + if (ofs) + tp->inq_byte7 |= INQ7_SYNC; + + /* + ** check values against driver limits. + */ + + if (per < np->minsync) + {chg = 1; per = np->minsync;} + if (per < tp->minsync) + {chg = 1; per = tp->minsync;} + if (ofs > tp->maxoffs) + {chg = 1; ofs = tp->maxoffs;} + + /* + ** Check against controller limits. + */ + fak = 7; + scntl3 = 0; + scntl4 = 0; + if (ofs != 0) { + ncr_getsync(np, per, &fak, &scntl3); + if (fak > 7) { + chg = 1; + ofs = 0; + } + } + if (ofs == 0) { + fak = 7; + per = 0; + scntl3 = 0; + scntl4 = 0; + tp->minsync = 0; + } + + if (DEBUG_FLAGS & DEBUG_NEGO) { + PRINT_ADDR(cp->cmd); + printk ("sync: per=%d scntl3=0x%x scntl4=0x%x ofs=%d fak=%d chg=%d.\n", + per, scntl3, scntl4, ofs, fak, chg); + } + + if (INB (HS_PRT) == HS_NEGOTIATE) { + OUTB (HS_PRT, HS_BUSY); + switch (cp->nego_status) { + case NS_SYNC: + /* + ** This was an answer message + */ + if (chg) { + /* + ** Answer wasn't acceptable. + */ + ncr_setsync (np, cp, 0, 0xe0, 0); + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + } else { + /* + ** Answer is ok. + */ + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + ncr_setsync (np, cp, scntl3, (fak<<5)|ofs,0); + else + ncr_setsync (np, cp, scntl3, ofs, scntl4); + + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + }; + return; + + case NS_WIDE: + ncr_setwide (np, cp, 0, 0); + break; + }; + }; + + /* + ** It was a request. Set value and + ** prepare an answer message + */ + + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + ncr_setsync (np, cp, scntl3, (fak<<5)|ofs,0); + else + ncr_setsync (np, cp, scntl3, ofs, scntl4); + + np->msgout[0] = M_EXTENDED; + np->msgout[1] = 3; + np->msgout[2] = M_X_SYNC_REQ; + np->msgout[3] = per; + np->msgout[4] = ofs; + + cp->nego_status = NS_SYNC; + + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "sync msgout", np->msgout); + } + + np->msgin [0] = M_NOOP; + + if (!ofs) + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + else + OUTL_DSP (NCB_SCRIPTH_PHYS (np, sdtr_resp)); +} + +/*========================================================== +** +** ncr chip handler for WIDE DATA TRANSFER REQUEST +** (WDTR) message. +** +**========================================================== +** +** Read comments above. +** +**---------------------------------------------------------- +*/ +static void ncr_wide_nego(ncb_p np, tcb_p tp, ccb_p cp) +{ + u_char chg, wide; + + /* + ** Wide request message received. + */ + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "wide msgin", np->msgin); + }; + + /* + ** get requested values. + */ + + chg = 0; + wide = np->msgin[3]; + + /* + ** if target sends WDTR message, + ** it CAN transfer wide. + */ + + if (wide) + tp->inq_byte7 |= INQ7_WIDE16; + + /* + ** check values against driver limits. + */ + + if (wide > tp->usrwide) + {chg = 1; wide = tp->usrwide;} + + if (DEBUG_FLAGS & DEBUG_NEGO) { + PRINT_ADDR(cp->cmd); + printk ("wide: wide=%d chg=%d.\n", wide, chg); + } + + if (INB (HS_PRT) == HS_NEGOTIATE) { + OUTB (HS_PRT, HS_BUSY); + switch (cp->nego_status) { + case NS_WIDE: + /* + ** This was an answer message + */ + if (chg) { + /* + ** Answer wasn't acceptable. + */ + ncr_setwide (np, cp, 0, 1); + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + } else { + /* + ** Answer is ok. + */ + ncr_setwide (np, cp, wide, 1); + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + }; + return; + + case NS_SYNC: + ncr_setsync (np, cp, 0, 0xe0, 0); + break; + }; + }; + + /* + ** It was a request, set value and + ** prepare an answer message + */ + + ncr_setwide (np, cp, wide, 1); + + np->msgout[0] = M_EXTENDED; + np->msgout[1] = 2; + np->msgout[2] = M_X_WIDE_REQ; + np->msgout[3] = wide; + + np->msgin [0] = M_NOOP; + + cp->nego_status = NS_WIDE; + + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "wide msgout", np->msgout); + } + + OUTL_DSP (NCB_SCRIPTH_PHYS (np, wdtr_resp)); +} +/*========================================================== +** +** ncr chip handler for PARALLEL PROTOCOL REQUEST +** (PPR) message. +** +**========================================================== +** +** Read comments above. +** +**---------------------------------------------------------- +*/ +static void ncr_ppr_nego(ncb_p np, tcb_p tp, ccb_p cp) +{ + u_char scntl3, scntl4; + u_char chg, ofs, per, fak, wth, dt; + + /* + ** PPR message received. + */ + + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "ppr msg in", np->msgin); + }; + + /* + ** get requested values. + */ + + chg = 0; + per = np->msgin[3]; + ofs = np->msgin[5]; + wth = np->msgin[6]; + dt = np->msgin[7]; + if (ofs==0) per=255; + + /* + ** if target sends sync (wide), + ** it CAN transfer synch (wide). + */ + + if (ofs) + tp->inq_byte7 |= INQ7_SYNC; + + if (wth) + tp->inq_byte7 |= INQ7_WIDE16; + + /* + ** check values against driver limits. + */ + + if (wth > tp->usrwide) + {chg = 1; wth = tp->usrwide;} + if (per < np->minsync) + {chg = 1; per = np->minsync;} + if (per < tp->minsync) + {chg = 1; per = tp->minsync;} + if (ofs > tp->maxoffs) + {chg = 1; ofs = tp->maxoffs;} + + /* + ** Check against controller limits. + */ + fak = 7; + scntl3 = 0; + scntl4 = 0; + if (ofs != 0) { + scntl4 = dt ? 0x80 : 0; + ncr_getsync(np, per, &fak, &scntl3); + if (fak > 7) { + chg = 1; + ofs = 0; + } + } + if (ofs == 0) { + fak = 7; + per = 0; + scntl3 = 0; + scntl4 = 0; + tp->minsync = 0; + } + + /* + ** If target responds with Ultra 3 speed + ** but narrow or not DT, reject. + ** If target responds with DT request + ** but not Ultra3 speeds, reject message, + ** reset min sync for target to 0x0A and + ** set flags to re-negotiate. + */ + + if ((per == 0x09) && ofs && (!wth || !dt)) + chg = 1; + else if (( (per > 0x09) && dt) ) + chg = 2; + + + if (DEBUG_FLAGS & DEBUG_NEGO) { + PRINT_ADDR(cp->cmd); + printk ("ppr: wth=%d per=%d scntl3=0x%x scntl4=0x%x ofs=%d fak=%d chg=%d.\n", + wth, per, scntl3, scntl4, ofs, fak, chg); + } + + if (INB (HS_PRT) == HS_NEGOTIATE) { + OUTB (HS_PRT, HS_BUSY); + switch (cp->nego_status) { + case NS_PPR: + /* + ** This was an answer message + */ + if (chg) { + /* + ** Answer wasn't acceptable. + */ + if (chg == 2) { + /* Send message reject and reset flags for + ** host to re-negotiate with min period 0x0A. + */ + tp->minsync = 0x0A; + tp->period = 0; + tp->widedone = 0; + } + ncr_setsyncwide (np, cp, 0, 0xe0, 0, 0); + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + } else { + /* + ** Answer is ok. + */ + + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + ncr_setsyncwide (np, cp, scntl3, (fak<<5)|ofs,0, wth); + else + ncr_setsyncwide (np, cp, scntl3, ofs, scntl4, wth); + + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + + }; + return; + + case NS_SYNC: + ncr_setsync (np, cp, 0, 0xe0, 0); + break; + + case NS_WIDE: + ncr_setwide (np, cp, 0, 0); + break; + }; + }; + + /* + ** It was a request. Set value and + ** prepare an answer message + ** + ** If narrow or not DT and requesting Ultra3 + ** slow the bus down and force ST. If not + ** requesting Ultra3, force ST. + ** Max offset is 31=0x1f if ST mode. + */ + + if ((per == 0x09) && ofs && (!wth || !dt)) { + per = 0x0A; + dt = 0; + ofs &= 0x1f; + } + else if ( (per > 0x09) && dt) { + dt = 0; + ofs &= 0x1f; + } + + if ((np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66)) + ncr_setsyncwide (np, cp, scntl3, (fak<<5)|ofs,0, wth); + else + ncr_setsyncwide (np, cp, scntl3, ofs, scntl4, wth); + + np->msgout[0] = M_EXTENDED; + np->msgout[1] = 6; + np->msgout[2] = M_X_PPR_REQ; + np->msgout[3] = per; + np->msgout[4] = 0; + np->msgout[5] = ofs; + np->msgout[6] = wth; + np->msgout[7] = dt; + + cp->nego_status = NS_PPR; + + if (DEBUG_FLAGS & DEBUG_NEGO) { + ncr_print_msg(cp, "ppr msgout", np->msgout); + } + + np->msgin [0] = M_NOOP; + + if (!ofs) + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + else + OUTL_DSP (NCB_SCRIPTH_PHYS (np, ppr_resp)); +} + + + +/* +** Reset SYNC or WIDE to default settings. +** Called when a negotiation does not succeed either +** on rejection or on protocol error. +*/ +static void ncr_nego_default(ncb_p np, tcb_p tp, ccb_p cp) +{ + /* + ** any error in negotiation: + ** fall back to default mode. + */ + switch (cp->nego_status) { + + case NS_SYNC: + ncr_setsync (np, cp, 0, 0xe0, 0); + break; + + case NS_WIDE: + ncr_setwide (np, cp, 0, 0); + break; + + case NS_PPR: + /* + * ppr_negotiation is set to 1 on the first ppr nego command. + * If ppr is successful, it is reset to 2. + * If unsuccessful it is reset to 0. + */ + if (DEBUG_FLAGS & DEBUG_NEGO) { + tcb_p tp=&np->target[cp->target]; + u_char factor, offset, width; + + ncr_get_xfer_info ( np, tp, &factor, &offset, &width); + + printk("Current factor %d offset %d width %d\n", + factor, offset, width); + } + if (tp->ppr_negotiation == 2) + ncr_setsyncwide (np, cp, 0, 0xe0, 0, 0); + else if (tp->ppr_negotiation == 1) { + + /* First ppr command has received a M REJECT. + * Do not change the existing wide/sync parameter + * values (asyn/narrow if this as the first nego; + * may be different if target initiates nego.). + */ + tp->ppr_negotiation = 0; + } + else + { + tp->ppr_negotiation = 0; + ncr_setwide (np, cp, 0, 0); + } + break; + }; + np->msgin [0] = M_NOOP; + np->msgout[0] = M_NOOP; + cp->nego_status = 0; +} + +/*========================================================== +** +** ncr chip handler for MESSAGE REJECT received for +** a WIDE or SYNCHRONOUS negotiation. +** +** clear the PPR negotiation flag, all future nego. +** will be SDTR and WDTR +** +**========================================================== +** +** Read comments above. +** +**---------------------------------------------------------- +*/ +static void ncr_nego_rejected(ncb_p np, tcb_p tp, ccb_p cp) +{ + ncr_nego_default(np, tp, cp); + OUTB (HS_PRT, HS_BUSY); +} + + +/*========================================================== +** +** +** ncr chip exception handler for programmed interrupts. +** +** +**========================================================== +*/ + +void ncr_int_sir (ncb_p np) +{ + u_char num = INB (nc_dsps); + u_long dsa = INL (nc_dsa); + ccb_p cp = ncr_ccb_from_dsa(np, dsa); + u_char target = INB (nc_sdid) & 0x0f; + tcb_p tp = &np->target[target]; + int tmp; + + if (DEBUG_FLAGS & DEBUG_TINY) printk ("I#%d", num); + + switch (num) { + /* + ** See comments in the SCRIPTS code. + */ +#ifdef SCSI_NCR_PCIQ_SYNC_ON_INTR + case SIR_DUMMY_INTERRUPT: + goto out; +#endif + + /* + ** The C code is currently trying to recover from something. + ** Typically, user want to abort some command. + */ + case SIR_SCRIPT_STOPPED: + case SIR_TARGET_SELECTED: + case SIR_ABORT_SENT: + case SIR_AUTO_SENSE_DONE: + ncr_sir_task_recovery(np, num); + return; + /* + ** The device didn't go to MSG OUT phase after having + ** been selected with ATN. We donnot want to handle + ** that. + */ + case SIR_SEL_ATN_NO_MSG_OUT: + printk ("%s:%d: No MSG OUT phase after selection with ATN.\n", + ncr_name (np), target); + goto out_stuck; + /* + ** The device didn't switch to MSG IN phase after + ** having reseleted the initiator. + */ + case SIR_RESEL_NO_MSG_IN: + /* + ** After reselection, the device sent a message that wasn't + ** an IDENTIFY. + */ + case SIR_RESEL_NO_IDENTIFY: + /* + ** If devices reselecting without sending an IDENTIFY + ** message still exist, this should help. + ** We just assume lun=0, 1 CCB, no tag. + */ + if (tp->l0p) { + OUTL (nc_dsa, scr_to_cpu(tp->l0p->tasktbl[0])); + OUTL_DSP (NCB_SCRIPT_PHYS (np, resel_go)); + return; + } + /* + ** The device reselected a LUN we donnot know of. + */ + case SIR_RESEL_BAD_LUN: + np->msgout[0] = M_RESET; + goto out; + /* + ** The device reselected for an untagged nexus and we + ** haven't any. + */ + case SIR_RESEL_BAD_I_T_L: + np->msgout[0] = M_ABORT; + goto out; + /* + ** The device reselected for a tagged nexus that we donnot + ** have. + */ + case SIR_RESEL_BAD_I_T_L_Q: + np->msgout[0] = M_ABORT_TAG; + goto out; + /* + ** The SCRIPTS let us know that the device has grabbed + ** our message and will abort the job. + */ + case SIR_RESEL_ABORTED: + np->lastmsg = np->msgout[0]; + np->msgout[0] = M_NOOP; + printk ("%s:%d: message %x sent on bad reselection.\n", + ncr_name (np), target, np->lastmsg); + goto out; + /* + ** The SCRIPTS let us know that a message has been + ** successfully sent to the device. + */ + case SIR_MSG_OUT_DONE: + np->lastmsg = np->msgout[0]; + np->msgout[0] = M_NOOP; + /* Should we really care of that */ + if (np->lastmsg == M_PARITY || np->lastmsg == M_ID_ERROR) { + if (cp) { + cp->xerr_status &= ~XE_PARITY_ERR; + if (!cp->xerr_status) + OUTOFFB (HF_PRT, HF_EXT_ERR); + } + } + goto out; + /* + ** The device didn't send a GOOD SCSI status. + ** We may have some work to do prior to allow + ** the SCRIPTS processor to continue. + */ + case SIR_BAD_STATUS: + if (!cp) + goto out; + ncr_sir_to_redo(np, num, cp); + return; + /* + ** We are asked by the SCRIPTS to prepare a + ** REJECT message. + */ + case SIR_REJECT_TO_SEND: + ncr_print_msg(cp, "M_REJECT to send for ", np->msgin); + np->msgout[0] = M_REJECT; + goto out; + /* + ** We have been ODD at the end of a DATA IN + ** transfer and the device didn't send a + ** IGNORE WIDE RESIDUE message. + ** It is a data overrun condition. + */ + case SIR_SWIDE_OVERRUN: + if (cp) { + OUTONB (HF_PRT, HF_EXT_ERR); + cp->xerr_status |= XE_SWIDE_OVRUN; + } + goto out; + /* + ** We have been ODD at the end of a DATA OUT + ** transfer. + ** It is a data underrun condition. + */ + case SIR_SODL_UNDERRUN: + if (cp) { + OUTONB (HF_PRT, HF_EXT_ERR); + cp->xerr_status |= XE_SODL_UNRUN; + } + goto out; + /* + ** The device wants us to tranfer more data than + ** expected or in the wrong direction. + ** The number of extra bytes is in scratcha. + ** It is a data overrun condition. + */ + case SIR_DATA_OVERRUN: + if (cp) { + OUTONB (HF_PRT, HF_EXT_ERR); + cp->xerr_status |= XE_EXTRA_DATA; + cp->extra_bytes += INL (nc_scratcha); + } + goto out; + /* + ** The device switched to an illegal phase (4/5). + */ + case SIR_BAD_PHASE: + if (cp) { + OUTONB (HF_PRT, HF_EXT_ERR); + cp->xerr_status |= XE_BAD_PHASE; + } + goto out; + /* + ** We received a message. + */ + case SIR_MSG_RECEIVED: + if (!cp) + goto out_stuck; + switch (np->msgin [0]) { + /* + ** We received an extended message. + ** We handle MODIFY DATA POINTER, SDTR, WDTR + ** and reject all other extended messages. + */ + case M_EXTENDED: + switch (np->msgin [2]) { + case M_X_MODIFY_DP: + if (DEBUG_FLAGS & DEBUG_POINTER) + ncr_print_msg(cp,"modify DP",np->msgin); + tmp = (np->msgin[3]<<24) + (np->msgin[4]<<16) + + (np->msgin[5]<<8) + (np->msgin[6]); + ncr_modify_dp(np, tp, cp, tmp); + return; + case M_X_SYNC_REQ: + ncr_sync_nego(np, tp, cp); + return; + case M_X_WIDE_REQ: + ncr_wide_nego(np, tp, cp); + return; + case M_X_PPR_REQ: + ncr_ppr_nego(np, tp, cp); + return; + default: + goto out_reject; + } + break; + /* + ** We received a 1/2 byte message not handled from SCRIPTS. + ** We are only expecting MESSAGE REJECT and IGNORE WIDE + ** RESIDUE messages that haven't been anticipated by + ** SCRIPTS on SWIDE full condition. Unanticipated IGNORE + ** WIDE RESIDUE messages are aliased as MODIFY DP (-1). + */ + case M_IGN_RESIDUE: + if (DEBUG_FLAGS & DEBUG_POINTER) + ncr_print_msg(cp,"ign wide residue", np->msgin); + ncr_modify_dp(np, tp, cp, -1); + return; + case M_REJECT: + if (INB (HS_PRT) == HS_NEGOTIATE) + ncr_nego_rejected(np, tp, cp); + else { + PRINT_ADDR(cp->cmd); + printk ("M_REJECT received (%x:%x).\n", + scr_to_cpu(np->lastmsg), np->msgout[0]); + } + goto out_clrack; + break; + default: + goto out_reject; + } + break; + /* + ** We received an unknown message. + ** Ignore all MSG IN phases and reject it. + */ + case SIR_MSG_WEIRD: + ncr_print_msg(cp, "WEIRD message received", np->msgin); + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_weird)); + return; + /* + ** Negotiation failed. + ** Target does not send us the reply. + ** Remove the HS_NEGOTIATE status. + */ + case SIR_NEGO_FAILED: + OUTB (HS_PRT, HS_BUSY); + /* + ** Negotiation failed. + ** Target does not want answer message. + */ + case SIR_NEGO_PROTO: + ncr_nego_default(np, tp, cp); + goto out; + }; + +out: + OUTONB_STD (); + return; +out_reject: + OUTL_DSP (NCB_SCRIPTH_PHYS (np, msg_bad)); + return; +out_clrack: + OUTL_DSP (NCB_SCRIPT_PHYS (np, clrack)); + return; +out_stuck: + return; +} + + +/*========================================================== +** +** +** Aquire a control block +** +** +**========================================================== +*/ + +static ccb_p ncr_get_ccb (ncb_p np, u_char tn, u_char ln) +{ + tcb_p tp = &np->target[tn]; + lcb_p lp = ncr_lp(np, tp, ln); + u_short tag = NO_TAG; + XPT_QUEHEAD *qp; + ccb_p cp = (ccb_p) 0; + + /* + ** Allocate a new CCB if needed. + */ + if (xpt_que_empty(&np->free_ccbq)) + (void) ncr_alloc_ccb(np); + + /* + ** Look for a free CCB + */ + qp = xpt_remque_head(&np->free_ccbq); + if (!qp) + goto out; + cp = xpt_que_entry(qp, struct ccb, link_ccbq); + + /* + ** If the LCB is not yet available and we already + ** have queued a CCB for a LUN without LCB, + ** give up. Otherwise all is fine. :-) + */ + if (!lp) { + if (xpt_que_empty(&np->b0_ccbq)) + xpt_insque_head(&cp->link_ccbq, &np->b0_ccbq); + else + goto out_free; + } else { + /* + ** Tune tag mode if asked by user. + */ + if (lp->queuedepth != lp->numtags) { + ncr_setup_tags(np, tn, ln); + } + + /* + ** Get a tag for this nexus if required. + ** Keep from using more tags than we can handle. + */ + if (lp->usetags) { + if (lp->busyccbs < lp->maxnxs) { + tag = lp->cb_tags[lp->ia_tag]; + ++lp->ia_tag; + if (lp->ia_tag == MAX_TAGS) + lp->ia_tag = 0; + cp->tags_si = lp->tags_si; + ++lp->tags_sum[cp->tags_si]; + } + else + goto out_free; + } + + /* + ** Put the CCB in the LUN wait queue and + ** count it as busy. + */ + xpt_insque_tail(&cp->link_ccbq, &lp->wait_ccbq); + ++lp->busyccbs; + } + + /* + ** Remember all informations needed to free this CCB. + */ + cp->to_abort = 0; + cp->tag = tag; + cp->target = tn; + cp->lun = ln; + + if (DEBUG_FLAGS & DEBUG_TAGS) { + PRINT_LUN(np, tn, ln); + printk ("ccb @%p using tag %d.\n", cp, tag); + } + +out: + return cp; +out_free: + xpt_insque_head(&cp->link_ccbq, &np->free_ccbq); + return (ccb_p) 0; +} + +/*========================================================== +** +** +** Release one control block +** +** +**========================================================== +*/ + +static void ncr_free_ccb (ncb_p np, ccb_p cp) +{ + tcb_p tp = &np->target[cp->target]; + lcb_p lp = ncr_lp(np, tp, cp->lun); + + if (DEBUG_FLAGS & DEBUG_TAGS) { + PRINT_LUN(np, cp->target, cp->lun); + printk ("ccb @%p freeing tag %d.\n", cp, cp->tag); + } + + /* + ** If lun control block available, make available + ** the task slot and the tag if any. + ** Decrement counters. + */ + if (lp) { + if (cp->tag != NO_TAG) { + lp->cb_tags[lp->if_tag++] = cp->tag; + if (lp->if_tag == MAX_TAGS) + lp->if_tag = 0; + --lp->tags_sum[cp->tags_si]; + lp->tasktbl[cp->tag] = cpu_to_scr(np->p_bad_i_t_l_q); + } else { + lp->tasktbl[0] = cpu_to_scr(np->p_bad_i_t_l); + } + --lp->busyccbs; + if (cp->queued) { + --lp->queuedccbs; + } + } + + /* + ** Make this CCB available. + */ + xpt_remque(&cp->link_ccbq); + xpt_insque_head(&cp->link_ccbq, &np->free_ccbq); + cp -> host_status = HS_IDLE; + cp -> queued = 0; +} + +/*------------------------------------------------------------------------ +** Allocate a CCB and initialize its fixed part. +**------------------------------------------------------------------------ +**------------------------------------------------------------------------ +*/ +static ccb_p ncr_alloc_ccb(ncb_p np) +{ + ccb_p cp = 0; + int hcode; + + /* + ** Allocate memory for this CCB. + */ + cp = m_calloc_dma(sizeof(struct ccb), "CCB"); + if (!cp) + return 0; + + /* + ** Count it and initialyze it. + */ + np->actccbs++; + + /* + ** Remember virtual and bus address of this ccb. + */ + cp->p_ccb = vtobus(cp); + + /* + ** Insert this ccb into the hashed list. + */ + hcode = CCB_HASH_CODE(cp->p_ccb); + cp->link_ccbh = np->ccbh[hcode]; + np->ccbh[hcode] = cp; + + /* + ** Initialyze the start and restart actions. + */ + cp->phys.header.go.start = cpu_to_scr(NCB_SCRIPT_PHYS (np, idle)); + cp->phys.header.go.restart = cpu_to_scr(NCB_SCRIPTH_PHYS(np,bad_i_t_l)); + + /* + ** Initilialyze some other fields. + */ + cp->phys.smsg_ext.addr = cpu_to_scr(NCB_PHYS(np, msgin[2])); + + /* + ** Chain into wakeup list and free ccb queue. + */ + cp->link_ccb = np->ccbc; + np->ccbc = cp; + + xpt_insque_head(&cp->link_ccbq, &np->free_ccbq); + + return cp; +} + +/*------------------------------------------------------------------------ +** Look up a CCB from a DSA value. +**------------------------------------------------------------------------ +**------------------------------------------------------------------------ +*/ +static ccb_p ncr_ccb_from_dsa(ncb_p np, u_long dsa) +{ + int hcode; + ccb_p cp; + + hcode = CCB_HASH_CODE(dsa); + cp = np->ccbh[hcode]; + while (cp) { + if (cp->p_ccb == dsa) + break; + cp = cp->link_ccbh; + } + + return cp; +} + +/*========================================================== +** +** +** Allocation of resources for Targets/Luns/Tags. +** +** +**========================================================== +*/ + + +/*------------------------------------------------------------------------ +** Target control block initialisation. +**------------------------------------------------------------------------ +** This data structure is fully initialized after a SCSI command +** has been successfully completed for this target. +**------------------------------------------------------------------------ +*/ +static void ncr_init_tcb (ncb_p np, u_char tn) +{ + /* + ** Check some alignments required by the chip. + */ + assert (( (offsetof(struct ncr_reg, nc_sxfer) ^ + offsetof(struct tcb , sval )) &3) == 0); + assert (( (offsetof(struct ncr_reg, nc_scntl3) ^ + offsetof(struct tcb , wval )) &3) == 0); + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)){ + assert (( (offsetof(struct ncr_reg, nc_scntl4) ^ + offsetof(struct tcb , uval )) &3) == 0); + } +} + +/*------------------------------------------------------------------------ +** Lun control block allocation and initialization. +**------------------------------------------------------------------------ +** This data structure is allocated and initialized after a SCSI +** command has been successfully completed for this target/lun. +**------------------------------------------------------------------------ +*/ +static lcb_p ncr_alloc_lcb (ncb_p np, u_char tn, u_char ln) +{ + tcb_p tp = &np->target[tn]; + lcb_p lp = ncr_lp(np, tp, ln); + + /* + ** Already done, return. + */ + if (lp) + return lp; + + /* + ** Initialize the target control block if not yet. + */ + ncr_init_tcb(np, tn); + + /* + ** Allocate the lcb bus address array. + ** Compute the bus address of this table. + */ + if (ln && !tp->luntbl) { + int i; + + tp->luntbl = m_calloc_dma(256, "LUNTBL"); + if (!tp->luntbl) + goto fail; + for (i = 0 ; i < 64 ; i++) + tp->luntbl[i] = cpu_to_scr(NCB_PHYS(np, resel_badlun)); + tp->b_luntbl = cpu_to_scr(vtobus(tp->luntbl)); + } + + /* + ** Allocate the table of pointers for LUN(s) > 0, if needed. + */ + if (ln && !tp->lmp) { + tp->lmp = m_calloc(MAX_LUN * sizeof(lcb_p), "LMP"); + if (!tp->lmp) + goto fail; + } + + /* + ** Allocate the lcb. + ** Make it available to the chip. + */ + lp = m_calloc_dma(sizeof(struct lcb), "LCB"); + if (!lp) + goto fail; + if (ln) { + tp->lmp[ln] = lp; + tp->luntbl[ln] = cpu_to_scr(vtobus(lp)); + } + else { + tp->l0p = lp; + tp->b_lun0 = cpu_to_scr(vtobus(lp)); + } + + /* + ** Initialize the CCB queue headers. + */ + xpt_que_init(&lp->busy_ccbq); + xpt_que_init(&lp->wait_ccbq); + + /* + ** Set max CCBs to 1 and use the default task array + ** by default. + */ + lp->maxnxs = 1; + lp->tasktbl = &lp->tasktbl_0; + lp->b_tasktbl = cpu_to_scr(vtobus(lp->tasktbl)); + lp->tasktbl[0] = cpu_to_scr(np->p_notask); + lp->resel_task = cpu_to_scr(NCB_SCRIPT_PHYS(np, resel_notag)); + + /* + ** Initialize command queuing control. + */ + lp->busyccbs = 1; + lp->queuedccbs = 1; + lp->queuedepth = 1; +fail: + return lp; +} + + +/*------------------------------------------------------------------------ +** Lun control block setup on INQUIRY data received. +**------------------------------------------------------------------------ +** We only support WIDE, SYNC for targets and CMDQ for logical units. +** This setup is done on each INQUIRY since we are expecting user +** will play with CHANGE DEFINITION commands. :-) +**------------------------------------------------------------------------ +*/ +static lcb_p ncr_setup_lcb (ncb_p np, u_char tn, u_char ln, u_char *inq_data) +{ + tcb_p tp = &np->target[tn]; + lcb_p lp = ncr_lp(np, tp, ln); + u_char inq_byte7; + int i; + + /* + ** If no lcb, try to allocate it. + */ + if (!lp && !(lp = ncr_alloc_lcb(np, tn, ln))) + goto fail; + +#if 0 /* No more used. Left here as provision */ + /* + ** Get device quirks. + */ + tp->quirks = 0; + if (tp->quirks && bootverbose) { + PRINT_LUN(np, tn, ln); + printk ("quirks=%x.\n", tp->quirks); + } +#endif + + /* + ** Evaluate trustable target/unit capabilities. + ** We only believe device version >= SCSI-2 that + ** use appropriate response data format (2). + ** But it seems that some CCS devices also + ** support SYNC and I donnot want to frustrate + ** anybody. ;-) + */ + inq_byte7 = 0; + if ((inq_data[2] & 0x7) >= 2 && (inq_data[3] & 0xf) == 2) + inq_byte7 = inq_data[7]; + else if ((inq_data[2] & 0x7) == 1 && (inq_data[3] & 0xf) == 1) + inq_byte7 = INQ7_SYNC; + + /* + ** Throw away announced LUN capabilities if we are told + ** that there is no real device supported by the logical unit. + */ + if ((inq_data[0] & 0xe0) > 0x20 || (inq_data[0] & 0x1f) == 0x1f) + inq_byte7 &= (INQ7_SYNC | INQ7_WIDE16); + + /* + ** If user is wanting SYNC, force this feature. + */ + if (driver_setup.force_sync_nego) + inq_byte7 |= INQ7_SYNC; + + /* + ** Prepare negotiation if SIP capabilities have changed. + */ + tp->inq_done = 1; + if ((inq_byte7 ^ tp->inq_byte7) & (INQ7_SYNC | INQ7_WIDE16)) { + tp->inq_byte7 = inq_byte7; + ncr_negotiate(np, tp); + } + + /* + ** If unit supports tagged commands, allocate and + ** initialyze the task table if not yet. + */ + if ((inq_byte7 & INQ7_QUEUE) && lp->tasktbl == &lp->tasktbl_0) { + lp->tasktbl = m_calloc_dma(MAX_TASKS*4, "TASKTBL"); + if (!lp->tasktbl) { + lp->tasktbl = &lp->tasktbl_0; + goto fail; + } + lp->b_tasktbl = cpu_to_scr(vtobus(lp->tasktbl)); + for (i = 0 ; i < MAX_TASKS ; i++) + lp->tasktbl[i] = cpu_to_scr(np->p_notask); + + lp->cb_tags = m_calloc(MAX_TAGS, "CB_TAGS"); + if (!lp->cb_tags) + goto fail; + for (i = 0 ; i < MAX_TAGS ; i++) + lp->cb_tags[i] = i; + + lp->maxnxs = MAX_TAGS; + lp->tags_stime = ktime_get(3*HZ); + } + + /* + ** Adjust tagged queueing status if needed. + */ + if ((inq_byte7 ^ lp->inq_byte7) & INQ7_QUEUE) { + lp->inq_byte7 = inq_byte7; + lp->numtags = lp->maxtags; + ncr_setup_tags (np, tn, ln); + } + +fail: + return lp; +} + +/*========================================================== +** +** +** Build Scatter Gather Block +** +** +**========================================================== +** +** The transfer area may be scattered among +** several non adjacent physical pages. +** +** We may use MAX_SCATTER blocks. +** +**---------------------------------------------------------- +*/ + +/* +** We try to reduce the number of interrupts caused +** by unexpected phase changes due to disconnects. +** A typical harddisk may disconnect before ANY block. +** If we wanted to avoid unexpected phase changes at all +** we had to use a break point every 512 bytes. +** Of course the number of scatter/gather blocks is +** limited. +** Under Linux, the scatter/gatter blocks are provided by +** the generic driver. We just have to copy addresses and +** sizes to the data segment array. +*/ + +/* +** For 64 bit systems, we use the 8 upper bits of the size field +** to provide bus address bits 32-39 to the SCRIPTS processor. +** This allows the 895A and 896 to address up to 1 TB of memory. +** For 32 bit chips on 64 bit systems, we must be provided with +** memory addresses that fit into the first 32 bit bus address +** range and so, this does not matter and we expect an error from +** the chip if this ever happen. +** +** We use a separate function for the case Linux does not provide +** a scatter list in order to allow better code optimization +** for the case we have a scatter list (BTW, for now this just wastes +** about 40 bytes of code for x86, but my guess is that the scatter +** code will get more complex later). +*/ + +#ifdef SCSI_NCR_USE_64BIT_DAC +#define SCATTER_ONE(data, badd, len) \ + (data)->addr = cpu_to_scr(badd); \ + (data)->size = cpu_to_scr((((badd) >> 8) & 0xff000000) + len); +#else +#define SCATTER_ONE(data, badd, len) \ + (data)->addr = cpu_to_scr(badd); \ + (data)->size = cpu_to_scr(len); +#endif + +#define CROSS_16MB(p, n) (((((u_long) p) + n - 1) ^ ((u_long) p)) & ~0xffffff) + +static int ncr_scatter_no_sglist(ncb_p np, ccb_p cp, Scsi_Cmnd *cmd) +{ + struct scr_tblmove *data = &cp->phys.data[MAX_SCATTER-1]; + int segment; + + cp->data_len = cmd->request_bufflen; + + if (cmd->request_bufflen) { + u_long baddr = map_scsi_single_data(np, cmd); + + SCATTER_ONE(data, baddr, cmd->request_bufflen); + if (CROSS_16MB(baddr, cmd->request_bufflen)) { + cp->host_flags |= HF_PM_TO_C; +#ifdef DEBUG_896R1 +printk("He! we are crossing a 16 MB boundary (0x%lx, 0x%x)\n", + baddr, cmd->request_bufflen); +#endif + } + segment = 1; + } + else + segment = 0; + + return segment; +} + +/* +** DEL 472 - 53C896 Rev 1 - Part Number 609-0393055 - ITEM 5. +** +** We disable data phase mismatch handling from SCRIPTS for data +** transfers that contains scatter/gather entries that cross +** a 16 MB boundary. +** We use a different scatter function for 896 rev. 1 that needs +** such a work-around. Doing so, we do not affect performance for +** other chips. +** This problem should not be triggered for disk IOs under Linux, +** since such IOs are performed using pages and buffers that are +** nicely power-of-two sized and aligned. But, since this may change +** at any time, a work-around was required. +*/ +static int ncr_scatter_896R1(ncb_p np, ccb_p cp, Scsi_Cmnd *cmd) +{ + int segn; + int use_sg = (int) cmd->use_sg; + + cp->data_len = 0; + + if (!use_sg) + segn = ncr_scatter_no_sglist(np, cp, cmd); + else if (use_sg > MAX_SCATTER) + segn = -1; + else { + struct scatterlist *scatter = (struct scatterlist *)cmd->buffer; + struct scr_tblmove *data; + + use_sg = map_scsi_sg_data(np, cmd); + data = &cp->phys.data[MAX_SCATTER - use_sg]; + + for (segn = 0; segn < use_sg; segn++) { + u_long baddr = scsi_sg_dma_address(&scatter[segn]); + unsigned int len = scsi_sg_dma_len(&scatter[segn]); + + SCATTER_ONE(&data[segn], + baddr, + len); + if (CROSS_16MB(baddr, scatter[segn].length)) { + cp->host_flags |= HF_PM_TO_C; +#ifdef DEBUG_896R1 +printk("He! we are crossing a 16 MB boundary (0x%lx, 0x%x)\n", + baddr, scatter[segn].length); +#endif + } + cp->data_len += len; + } + } + + return segn; +} + +static int ncr_scatter(ncb_p np, ccb_p cp, Scsi_Cmnd *cmd) +{ + int segment; + int use_sg = (int) cmd->use_sg; + + cp->data_len = 0; + + if (!use_sg) + segment = ncr_scatter_no_sglist(np, cp, cmd); + else if (use_sg > MAX_SCATTER) + segment = -1; + else { + struct scatterlist *scatter = (struct scatterlist *)cmd->buffer; + struct scr_tblmove *data; + + use_sg = map_scsi_sg_data(np, cmd); + data = &cp->phys.data[MAX_SCATTER - use_sg]; + + for (segment = 0; segment < use_sg; segment++) { + u_long baddr = scsi_sg_dma_address(&scatter[segment]); + unsigned int len = scsi_sg_dma_len(&scatter[segment]); + + SCATTER_ONE(&data[segment], + baddr, + len); + cp->data_len += len; + } + } + + return segment; +} + +/*========================================================== +** +** +** Test the pci bus snoop logic :-( +** +** Has to be called with interrupts disabled. +** +** +**========================================================== +*/ + +#ifndef SCSI_NCR_IOMAPPED +static int __init ncr_regtest (struct ncb* np) +{ + register volatile u_int32 data; + /* + ** ncr registers may NOT be cached. + ** write 0xffffffff to a read only register area, + ** and try to read it back. + */ + data = 0xffffffff; + OUTL_OFF(offsetof(struct ncr_reg, nc_dstat), data); + data = INL_OFF(offsetof(struct ncr_reg, nc_dstat)); +#if 1 + if (data == 0xffffffff) { +#else + if ((data & 0xe2f0fffd) != 0x02000080) { +#endif + printk ("CACHE TEST FAILED: reg dstat-sstat2 readback %x.\n", + (unsigned) data); + return (0x10); + }; + return (0); +} +#endif + +static int __init ncr_snooptest (struct ncb* np) +{ + u_int32 ncr_rd, ncr_wr, ncr_bk, host_rd, host_wr, pc; + int i, err=0; +#ifndef SCSI_NCR_IOMAPPED + if (np->reg) { + err |= ncr_regtest (np); + if (err) return (err); + } +#endif + /* + ** init + */ + pc = NCB_SCRIPTH0_PHYS (np, snooptest); + host_wr = 1; + ncr_wr = 2; + /* + ** Set memory and register. + */ + np->ncr_cache = cpu_to_scr(host_wr); + OUTL (nc_temp, ncr_wr); + /* + ** Start script (exchange values) + */ + OUTL (nc_dsa, np->p_ncb); + OUTL_DSP (pc); + /* + ** Wait 'til done (with timeout) + */ + for (i=0; i<NCR_SNOOP_TIMEOUT; i++) + if (INB(nc_istat) & (INTF|SIP|DIP)) + break; + /* + ** Save termination position. + */ + pc = INL (nc_dsp); + /* + ** Read memory and register. + */ + host_rd = scr_to_cpu(np->ncr_cache); + ncr_rd = INL (nc_scratcha); + ncr_bk = INL (nc_temp); + + /* + ** check for timeout + */ + if (i>=NCR_SNOOP_TIMEOUT) { + printk ("CACHE TEST FAILED: timeout.\n"); + return (0x20); + }; + /* + ** Check termination position. + */ + if (pc != NCB_SCRIPTH0_PHYS (np, snoopend)+8) { + printk ("CACHE TEST FAILED: script execution failed.\n"); + printk ("start=%08lx, pc=%08lx, end=%08lx\n", + (u_long) NCB_SCRIPTH0_PHYS (np, snooptest), (u_long) pc, + (u_long) NCB_SCRIPTH0_PHYS (np, snoopend) +8); + return (0x40); + }; + /* + ** Show results. + */ + if (host_wr != ncr_rd) { + printk ("CACHE TEST FAILED: host wrote %d, ncr read %d.\n", + (int) host_wr, (int) ncr_rd); + err |= 1; + }; + if (host_rd != ncr_wr) { + printk ("CACHE TEST FAILED: ncr wrote %d, host read %d.\n", + (int) ncr_wr, (int) host_rd); + err |= 2; + }; + if (ncr_bk != ncr_wr) { + printk ("CACHE TEST FAILED: ncr wrote %d, read back %d.\n", + (int) ncr_wr, (int) ncr_bk); + err |= 4; + }; + return (err); +} + +/*========================================================== +** +** Determine the ncr's clock frequency. +** This is essential for the negotiation +** of the synchronous transfer rate. +** +**========================================================== +** +** Note: we have to return the correct value. +** THERE IS NO SAFE DEFAULT VALUE. +** +** Most NCR/SYMBIOS boards are delivered with a 40 Mhz clock. +** 53C860 and 53C875 rev. 1 support fast20 transfers but +** do not have a clock doubler and so are provided with a +** 80 MHz clock. All other fast20 boards incorporate a doubler +** and so should be delivered with a 40 MHz clock. +** The recent fast40 chips (895/896/895A) and the +** fast80 chip (C1010) use a 40 Mhz base clock +** and provide a clock quadrupler (160 Mhz). The code below +** tries to deal as cleverly as possible with all this stuff. +** +**---------------------------------------------------------- +*/ + +/* + * Select NCR SCSI clock frequency + */ +static void ncr_selectclock(ncb_p np, u_char scntl3) +{ + if (np->multiplier < 2) { + OUTB(nc_scntl3, scntl3); + return; + } + + if (bootverbose >= 2) + printk ("%s: enabling clock multiplier\n", ncr_name(np)); + + OUTB(nc_stest1, DBLEN); /* Enable clock multiplier */ + + if ( (np->device_id != PCI_DEVICE_ID_LSI_53C1010) && + (np->device_id != PCI_DEVICE_ID_LSI_53C1010_66) && + (np->multiplier > 2)) { + int i = 20; /* Poll bit 5 of stest4 for quadrupler */ + while (!(INB(nc_stest4) & LCKFRQ) && --i > 0) + UDELAY (20); + if (!i) + printk("%s: the chip cannot lock the frequency\n", + ncr_name(np)); + + } else /* Wait 120 micro-seconds for multiplier*/ + UDELAY (120); + + OUTB(nc_stest3, HSC); /* Halt the scsi clock */ + OUTB(nc_scntl3, scntl3); + OUTB(nc_stest1, (DBLEN|DBLSEL));/* Select clock multiplier */ + OUTB(nc_stest3, 0x00); /* Restart scsi clock */ +} + + +/* + * calculate NCR SCSI clock frequency (in KHz) + */ +static unsigned __init ncrgetfreq (ncb_p np, int gen) +{ + unsigned int ms = 0; + unsigned int f; + int count; + + /* + * Measure GEN timer delay in order + * to calculate SCSI clock frequency + * + * This code will never execute too + * many loop iterations (if DELAY is + * reasonably correct). It could get + * too low a delay (too high a freq.) + * if the CPU is slow executing the + * loop for some reason (an NMI, for + * example). For this reason we will + * if multiple measurements are to be + * performed trust the higher delay + * (lower frequency returned). + */ + OUTW (nc_sien , 0x0);/* mask all scsi interrupts */ + /* enable general purpose timer */ + (void) INW (nc_sist); /* clear pending scsi interrupt */ + OUTB (nc_dien , 0); /* mask all dma interrupts */ + (void) INW (nc_sist); /* another one, just to be sure :) */ + OUTB (nc_scntl3, 4); /* set pre-scaler to divide by 3 */ + OUTB (nc_stime1, 0); /* disable general purpose timer */ + OUTB (nc_stime1, gen); /* set to nominal delay of 1<<gen * 125us */ + /* Temporary fix for udelay issue with Alpha + platform */ + while (!(INW(nc_sist) & GEN) && ms++ < 100000) { + /* count 1ms */ + for (count = 0; count < 10; count++) + UDELAY (100); + } + OUTB (nc_stime1, 0); /* disable general purpose timer */ + /* + * set prescaler to divide by whatever 0 means + * 0 ought to choose divide by 2, but appears + * to set divide by 3.5 mode in my 53c810 ... + */ + OUTB (nc_scntl3, 0); + + /* + * adjust for prescaler, and convert into KHz + * scale values derived empirically. C1010 uses + * different dividers + */ +#if 0 + if (np->device_id == PCI_DEVICE_ID_LSI_53C1010) + f = ms ? ((1 << gen) * 2866 ) / ms : 0; + else +#endif + f = ms ? ((1 << gen) * 4340) / ms : 0; + + if (bootverbose >= 2) + printk ("%s: Delay (GEN=%d): %u msec, %u KHz\n", + ncr_name(np), gen, ms, f); + + return f; +} + +static unsigned __init ncr_getfreq (ncb_p np) +{ + u_int f1, f2; + int gen = 11; + + (void) ncrgetfreq (np, gen); /* throw away first result */ + f1 = ncrgetfreq (np, gen); + f2 = ncrgetfreq (np, gen); + if (f1 > f2) f1 = f2; /* trust lower result */ + return f1; +} + +/* + * Get/probe NCR SCSI clock frequency + */ +static void __init ncr_getclock (ncb_p np, int mult) +{ + unsigned char scntl3 = np->sv_scntl3; + unsigned char stest1 = np->sv_stest1; + unsigned f1; + + np->multiplier = 1; + f1 = 40000; + + /* + ** True with 875/895/896/895A with clock multiplier selected + */ + if (mult > 1 && (stest1 & (DBLEN+DBLSEL)) == DBLEN+DBLSEL) { + if (bootverbose >= 2) + printk ("%s: clock multiplier found\n", ncr_name(np)); + np->multiplier = mult; + } + + /* + ** If multiplier not found but a C1010, assume a mult of 4. + ** If multiplier not found or scntl3 not 7,5,3, + ** reset chip and get frequency from general purpose timer. + ** Otherwise trust scntl3 BIOS setting. + */ + if ((np->device_id == PCI_DEVICE_ID_LSI_53C1010) || + (np->device_id == PCI_DEVICE_ID_LSI_53C1010_66)) { + f1=40000; + np->multiplier = mult; + if (bootverbose >= 2) + printk ("%s: clock multiplier assumed\n", ncr_name(np)); + } + else if (np->multiplier != mult || (scntl3 & 7) < 3 || !(scntl3 & 1)) { + OUTB (nc_stest1, 0); /* make sure doubler is OFF */ + f1 = ncr_getfreq (np); + + if (bootverbose) + printk ("%s: NCR clock is %uKHz\n", ncr_name(np), f1); + + if (f1 < 55000) f1 = 40000; + else f1 = 80000; + + /* + ** Suggest to also check the PCI clock frequency + ** to make sure our frequency calculation algorithm + ** is not too biased. + */ + if (np->features & FE_66MHZ) { + np->pciclock_min = (66000*55+80-1)/80; + np->pciclock_max = (66000*55)/40; + } + else { + np->pciclock_min = (33000*55+80-1)/80; + np->pciclock_max = (33000*55)/40; + } + + if (f1 == 40000 && mult > 1) { + if (bootverbose >= 2) + printk ("%s: clock multiplier assumed\n", ncr_name(np)); + np->multiplier = mult; + } + } else { + if ((scntl3 & 7) == 3) f1 = 40000; + else if ((scntl3 & 7) == 5) f1 = 80000; + else f1 = 160000; + + f1 /= np->multiplier; + } + + /* + ** Compute controller synchronous parameters. + */ + f1 *= np->multiplier; + np->clock_khz = f1; +} + +/* + * Get/probe PCI clock frequency + */ +static u_int __init ncr_getpciclock (ncb_p np) +{ + static u_int f; + + OUTB (nc_stest1, SCLK); /* Use the PCI clock as SCSI clock */ + f = ncr_getfreq (np); + OUTB (nc_stest1, 0); + + return f; +} + +/*===================== LINUX ENTRY POINTS SECTION ==========================*/ + +#ifndef uchar +#define uchar unsigned char +#endif + +#ifndef ushort +#define ushort unsigned short +#endif + +#ifndef ulong +#define ulong unsigned long +#endif + +/* --------------------------------------------------------------------- +** +** Driver setup from the boot command line +** +** --------------------------------------------------------------------- +*/ + +#ifdef MODULE +#define ARG_SEP ' ' +#else +#define ARG_SEP ',' +#endif + +#define OPT_TAGS 1 +#define OPT_MASTER_PARITY 2 +#define OPT_SCSI_PARITY 3 +#define OPT_DISCONNECTION 4 +#define OPT_SPECIAL_FEATURES 5 +#define OPT_ULTRA_SCSI 6 +#define OPT_FORCE_SYNC_NEGO 7 +#define OPT_REVERSE_PROBE 8 +#define OPT_DEFAULT_SYNC 9 +#define OPT_VERBOSE 10 +#define OPT_DEBUG 11 +#define OPT_BURST_MAX 12 +#define OPT_LED_PIN 13 +#define OPT_MAX_WIDE 14 +#define OPT_SETTLE_DELAY 15 +#define OPT_DIFF_SUPPORT 16 +#define OPT_IRQM 17 +#define OPT_PCI_FIX_UP 18 +#define OPT_BUS_CHECK 19 +#define OPT_OPTIMIZE 20 +#define OPT_RECOVERY 21 +#define OPT_SAFE_SETUP 22 +#define OPT_USE_NVRAM 23 +#define OPT_EXCLUDE 24 +#define OPT_HOST_ID 25 + +#ifdef SCSI_NCR_IARB_SUPPORT +#define OPT_IARB 26 +#endif + +static char setup_token[] __initdata = + "tags:" "mpar:" + "spar:" "disc:" + "specf:" "ultra:" + "fsn:" "revprob:" + "sync:" "verb:" + "debug:" "burst:" + "led:" "wide:" + "settle:" "diff:" + "irqm:" "pcifix:" + "buschk:" "optim:" + "recovery:" + "safe:" "nvram:" + "excl:" "hostid:" +#ifdef SCSI_NCR_IARB_SUPPORT + "iarb:" +#endif + ; /* DONNOT REMOVE THIS ';' */ + +#ifdef MODULE +#define ARG_SEP ' ' +#else +#define ARG_SEP ',' +#endif + +static int __init get_setup_token(char *p) +{ + char *cur = setup_token; + char *pc; + int i = 0; + + while (cur != NULL && (pc = strchr(cur, ':')) != NULL) { + ++pc; + ++i; + if (!strncmp(p, cur, pc - cur)) + return i; + cur = pc; + } + return 0; +} + + +int __init sym53c8xx_setup(char *str) +{ +#ifdef SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT + char *cur = str; + char *pc, *pv; + unsigned long val; + int i, c; + int xi = 0; + + while (cur != NULL && (pc = strchr(cur, ':')) != NULL) { + char *pe; + + val = 0; + pv = pc; + c = *++pv; + + if (c == 'n') + val = 0; + else if (c == 'y') + val = 1; + else + val = (int) simple_strtoul(pv, &pe, 0); + + switch (get_setup_token(cur)) { + case OPT_TAGS: + driver_setup.default_tags = val; + if (pe && *pe == '/') { + i = 0; + while (*pe && *pe != ARG_SEP && + i < sizeof(driver_setup.tag_ctrl)-1) { + driver_setup.tag_ctrl[i++] = *pe++; + } + driver_setup.tag_ctrl[i] = '\0'; + } + break; + case OPT_MASTER_PARITY: + driver_setup.master_parity = val; + break; + case OPT_SCSI_PARITY: + driver_setup.scsi_parity = val; + break; + case OPT_DISCONNECTION: + driver_setup.disconnection = val; + break; + case OPT_SPECIAL_FEATURES: + driver_setup.special_features = val; + break; + case OPT_ULTRA_SCSI: + driver_setup.ultra_scsi = val; + break; + case OPT_FORCE_SYNC_NEGO: + driver_setup.force_sync_nego = val; + break; + case OPT_REVERSE_PROBE: + driver_setup.reverse_probe = val; + break; + case OPT_DEFAULT_SYNC: + driver_setup.default_sync = val; + break; + case OPT_VERBOSE: + driver_setup.verbose = val; + break; + case OPT_DEBUG: + driver_setup.debug = val; + break; + case OPT_BURST_MAX: + driver_setup.burst_max = val; + break; + case OPT_LED_PIN: + driver_setup.led_pin = val; + break; + case OPT_MAX_WIDE: + driver_setup.max_wide = val? 1:0; + break; + case OPT_SETTLE_DELAY: + driver_setup.settle_delay = val; + break; + case OPT_DIFF_SUPPORT: + driver_setup.diff_support = val; + break; + case OPT_IRQM: + driver_setup.irqm = val; + break; + case OPT_PCI_FIX_UP: + driver_setup.pci_fix_up = val; + break; + case OPT_BUS_CHECK: + driver_setup.bus_check = val; + break; + case OPT_OPTIMIZE: + driver_setup.optimize = val; + break; + case OPT_RECOVERY: + driver_setup.recovery = val; + break; + case OPT_USE_NVRAM: + driver_setup.use_nvram = val; + break; + case OPT_SAFE_SETUP: + memcpy(&driver_setup, &driver_safe_setup, + sizeof(driver_setup)); + break; + case OPT_EXCLUDE: + if (xi < SCSI_NCR_MAX_EXCLUDES) + driver_setup.excludes[xi++] = val; + break; + case OPT_HOST_ID: + driver_setup.host_id = val; + break; +#ifdef SCSI_NCR_IARB_SUPPORT + case OPT_IARB: + driver_setup.iarb = val; + break; +#endif + default: + printk("sym53c8xx_setup: unexpected boot option '%.*s' ignored\n", (int)(pc-cur+1), cur); + break; + } + + if ((cur = strchr(cur, ARG_SEP)) != NULL) + ++cur; + } +#endif /* SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT */ + return 1; +} + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,13) +#ifndef MODULE +__setup("sym53c8xx=", sym53c8xx_setup); +#endif +#endif + +static int +sym53c8xx_pci_init(Scsi_Host_Template *tpnt, pcidev_t pdev, ncr_device *device); + +/* +** Linux entry point for SYM53C8XX devices detection routine. +** +** Called by the middle-level scsi drivers at initialization time, +** or at module installation. +** +** Read the PCI configuration and try to attach each +** detected NCR board. +** +** If NVRAM is present, try to attach boards according to +** the used defined boot order. +** +** Returns the number of boards successfully attached. +*/ + +static void __init ncr_print_driver_setup(void) +{ +#define YesNo(y) y ? 'y' : 'n' + printk (NAME53C8XX ": setup=disc:%c,specf:%d,ultra:%d,tags:%d,sync:%d," + "burst:%d,wide:%c,diff:%d,revprob:%c,buschk:0x%x\n", + YesNo(driver_setup.disconnection), + driver_setup.special_features, + driver_setup.ultra_scsi, + driver_setup.default_tags, + driver_setup.default_sync, + driver_setup.burst_max, + YesNo(driver_setup.max_wide), + driver_setup.diff_support, + YesNo(driver_setup.reverse_probe), + driver_setup.bus_check); + + printk (NAME53C8XX ": setup=mpar:%c,spar:%c,fsn=%c,verb:%d,debug:0x%x," + "led:%c,settle:%d,irqm:0x%x,nvram:0x%x,pcifix:0x%x\n", + YesNo(driver_setup.master_parity), + YesNo(driver_setup.scsi_parity), + YesNo(driver_setup.force_sync_nego), + driver_setup.verbose, + driver_setup.debug, + YesNo(driver_setup.led_pin), + driver_setup.settle_delay, + driver_setup.irqm, + driver_setup.use_nvram, + driver_setup.pci_fix_up); +#undef YesNo +} + +/*=================================================================== +** SYM53C8XX devices description table and chip ids list. +**=================================================================== +*/ + +static ncr_chip ncr_chip_table[] __initdata = SCSI_NCR_CHIP_TABLE; +static ushort ncr_chip_ids[] __initdata = SCSI_NCR_CHIP_IDS; + +#ifdef SCSI_NCR_PQS_PDS_SUPPORT +/*=================================================================== +** Detect all NCR PQS/PDS boards and keep track of their bus nr. +** +** The NCR PQS or PDS card is constructed as a DEC bridge +** behind which sit a proprietary NCR memory controller and +** four or two 53c875s as separate devices. In its usual mode +** of operation, the 875s are slaved to the memory controller +** for all transfers. We can tell if an 875 is part of a +** PQS/PDS or not since if it is, it will be on the same bus +** as the memory controller. To operate with the Linux +** driver, the memory controller is disabled and the 875s +** freed to function independently. The only wrinkle is that +** the preset SCSI ID (which may be zero) must be read in from +** a special configuration space register of the 875 +**=================================================================== +*/ +#define SCSI_NCR_MAX_PQS_BUS 16 +static int pqs_bus[SCSI_NCR_MAX_PQS_BUS] __initdata = { 0 }; + +static void __init ncr_detect_pqs_pds(void) +{ + short index; + pcidev_t dev = PCIDEV_NULL; + + for(index=0; index < SCSI_NCR_MAX_PQS_BUS; index++) { + u_char tmp; + + dev = pci_find_device(0x101a, 0x0009, dev); + if (dev == PCIDEV_NULL) { + pqs_bus[index] = -1; + break; + } + printk(KERN_INFO NAME53C8XX ": NCR PQS/PDS memory controller detected on bus %d\n", PciBusNumber(dev)); + pci_read_config_byte(dev, 0x44, &tmp); + /* bit 1: allow individual 875 configuration */ + tmp |= 0x2; + pci_write_config_byte(dev, 0x44, tmp); + pci_read_config_byte(dev, 0x45, &tmp); + /* bit 2: drive individual 875 interrupts to the bus */ + tmp |= 0x4; + pci_write_config_byte(dev, 0x45, tmp); + + pqs_bus[index] = PciBusNumber(dev); + } +} +#endif /* SCSI_NCR_PQS_PDS_SUPPORT */ + +/*=================================================================== +** Detect all 53c8xx hosts and then attach them. +** +** If we are using NVRAM, once all hosts are detected, we need to +** check any NVRAM for boot order in case detect and boot order +** differ and attach them using the order in the NVRAM. +** +** If no NVRAM is found or data appears invalid attach boards in +** the the order they are detected. +**=================================================================== +*/ +int __init sym53c8xx_detect(Scsi_Host_Template *tpnt) +{ + pcidev_t pcidev; + int i, j, chips, hosts, count; + int attach_count = 0; + ncr_device *devtbl, *devp; +#ifdef SCSI_NCR_NVRAM_SUPPORT + ncr_nvram nvram0, nvram, *nvp; +#endif + + /* + ** PCI is required. + */ + if (!pci_present()) + return 0; + + /* + ** Initialize driver general stuff. + */ +#ifdef SCSI_NCR_PROC_INFO_SUPPORT +#if LINUX_VERSION_CODE < LinuxVersionCode(2,3,27) + tpnt->proc_dir = &proc_scsi_sym53c8xx; +#else + tpnt->proc_name = NAME53C8XX; +#endif + tpnt->proc_info = sym53c8xx_proc_info; +#endif + +#if defined(SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT) && defined(MODULE) +if (sym53c8xx) + sym53c8xx_setup(sym53c8xx); +#endif +#ifdef SCSI_NCR_DEBUG_INFO_SUPPORT + ncr_debug = driver_setup.debug; +#endif + + if (initverbose >= 2) + ncr_print_driver_setup(); + + /* + ** Allocate the device table since we donnot want to + ** overflow the kernel stack. + ** 1 x 4K PAGE is enough for more than 40 devices for i386. + */ + devtbl = m_calloc(PAGE_SIZE, "devtbl"); + if (!devtbl) + return 0; + + /* + ** Detect all NCR PQS/PDS memory controllers. + */ +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + ncr_detect_pqs_pds(); +#endif + + /* + ** Detect all 53c8xx hosts. + ** Save the first Symbios NVRAM content if any + ** for the boot order. + */ + chips = sizeof(ncr_chip_ids) / sizeof(ncr_chip_ids[0]); + hosts = PAGE_SIZE / sizeof(*devtbl); +#ifdef SCSI_NCR_NVRAM_SUPPORT + nvp = (driver_setup.use_nvram & 0x1) ? &nvram0 : 0; +#endif + j = 0; + count = 0; + pcidev = PCIDEV_NULL; + while (1) { + char *msg = ""; + if (count >= hosts) + break; + if (j >= chips) + break; + i = driver_setup.reverse_probe ? chips - 1 - j : j; + pcidev = pci_find_device(PCI_VENDOR_ID_NCR, ncr_chip_ids[i], + pcidev); + if (pcidev == PCIDEV_NULL) { + ++j; + continue; + } + /* Some HW as the HP LH4 may report twice PCI devices */ + for (i = 0; i < count ; i++) { + if (devtbl[i].slot.bus == PciBusNumber(pcidev) && + devtbl[i].slot.device_fn == PciDeviceFn(pcidev)) + break; + } + if (i != count) /* Ignore this device if we already have it */ + continue; + devp = &devtbl[count]; + devp->host_id = driver_setup.host_id; + devp->attach_done = 0; + if (sym53c8xx_pci_init(tpnt, pcidev, devp)) { + continue; + } + ++count; +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (nvp) { + ncr_get_nvram(devp, nvp); + switch(nvp->type) { + case SCSI_NCR_SYMBIOS_NVRAM: + /* + * Switch to the other nvram buffer, so that + * nvram0 will contain the first Symbios + * format NVRAM content with boot order. + */ + nvp = &nvram; + msg = "with Symbios NVRAM"; + break; + case SCSI_NCR_TEKRAM_NVRAM: + msg = "with Tekram NVRAM"; + break; + } + } +#endif +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + if (devp->pqs_pds) + msg = "(NCR PQS/PDS)"; +#endif + printk(KERN_INFO NAME53C8XX ": 53c%s detected %s\n", + devp->chip.name, msg); + } + + /* + ** If we have found a SYMBIOS NVRAM, use first the NVRAM boot + ** sequence as device boot order. + ** check devices in the boot record against devices detected. + ** attach devices if we find a match. boot table records that + ** do not match any detected devices will be ignored. + ** devices that do not match any boot table will not be attached + ** here but will attempt to be attached during the device table + ** rescan. + */ +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (!nvp || nvram0.type != SCSI_NCR_SYMBIOS_NVRAM) + goto next; + for (i = 0; i < 4; i++) { + Symbios_host *h = &nvram0.data.Symbios.host[i]; + for (j = 0 ; j < count ; j++) { + devp = &devtbl[j]; + if (h->device_fn != devp->slot.device_fn || + h->bus_nr != devp->slot.bus || + h->device_id != devp->chip.device_id) + continue; + if (devp->attach_done) + continue; + if (h->flags & SYMBIOS_INIT_SCAN_AT_BOOT) { + ncr_get_nvram(devp, nvp); + if (!ncr_attach (tpnt, attach_count, devp)) + attach_count++; + } + else if (!(driver_setup.use_nvram & 0x80)) + printk(KERN_INFO NAME53C8XX + ": 53c%s state OFF thus not attached\n", + devp->chip.name); + else + continue; + + devp->attach_done = 1; + break; + } + } +next: +#endif + + /* + ** Rescan device list to make sure all boards attached. + ** Devices without boot records will not be attached yet + ** so try to attach them here. + */ + for (i= 0; i < count; i++) { + devp = &devtbl[i]; + if (!devp->attach_done) { +#ifdef SCSI_NCR_NVRAM_SUPPORT + ncr_get_nvram(devp, nvp); +#endif + if (!ncr_attach (tpnt, attach_count, devp)) + attach_count++; + } + } + + m_free(devtbl, PAGE_SIZE, "devtbl"); + + return attach_count; +} + +/*=================================================================== +** Read and check the PCI configuration for any detected NCR +** boards and save data for attaching after all boards have +** been detected. +**=================================================================== +*/ +static int __init +sym53c8xx_pci_init(Scsi_Host_Template *tpnt, pcidev_t pdev, ncr_device *device) +{ + u_short vendor_id, device_id, command, status_reg; + u_char cache_line_size, latency_timer; + u_char suggested_cache_line_size = 0; + u_char pci_fix_up = driver_setup.pci_fix_up; + u_char revision; + u_int irq; + u_long base, base_2, io_port; + int i; + ncr_chip *chip; + + printk(KERN_INFO NAME53C8XX ": at PCI bus %d, device %d, function %d\n", + PciBusNumber(pdev), + (int) (PciDeviceFn(pdev) & 0xf8) >> 3, + (int) (PciDeviceFn(pdev) & 7)); + +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING + if (!pci_dma_supported(pdev, (dma_addr_t) (0xffffffffUL))) { + printk(KERN_WARNING NAME53C8XX + "32 BIT PCI BUS DMA ADDRESSING NOT SUPPORTED\n"); + return -1; + } +#endif + + /* + ** Read info from the PCI config space. + ** pci_read_config_xxx() functions are assumed to be used for + ** successfully detected PCI devices. + */ + vendor_id = PciVendorId(pdev); + device_id = PciDeviceId(pdev); + irq = PciIrqLine(pdev); + i = 0; + i = pci_get_base_address(pdev, i, &io_port); + i = pci_get_base_address(pdev, i, &base); + (void) pci_get_base_address(pdev, i, &base_2); + + pci_read_config_word(pdev, PCI_COMMAND, &command); + pci_read_config_byte(pdev, PCI_CLASS_REVISION, &revision); + pci_read_config_byte(pdev, PCI_CACHE_LINE_SIZE, &cache_line_size); + pci_read_config_byte(pdev, PCI_LATENCY_TIMER, &latency_timer); + pci_read_config_word(pdev, PCI_STATUS, &status_reg); + +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + /* + ** Match the BUS number for PQS/PDS devices. + ** Read the SCSI ID from a special register mapped + ** into the configuration space of the individual + ** 875s. This register is set up by the PQS bios + */ + for(i = 0; i < SCSI_NCR_MAX_PQS_BUS && pqs_bus[i] != -1; i++) { + u_char tmp; + if (pqs_bus[i] == PciBusNumber(pdev)) { + pci_read_config_byte(pdev, 0x84, &tmp); + device->pqs_pds = 1; + device->host_id = tmp; + break; + } + } +#endif /* SCSI_NCR_PQS_PDS_SUPPORT */ + + /* + ** If user excludes this chip, donnot initialize it. + */ + for (i = 0 ; i < SCSI_NCR_MAX_EXCLUDES ; i++) { + if (driver_setup.excludes[i] == + (io_port & PCI_BASE_ADDRESS_IO_MASK)) + return -1; + } + /* + ** Check if the chip is supported + */ + chip = 0; + for (i = 0; i < sizeof(ncr_chip_table)/sizeof(ncr_chip_table[0]); i++) { + if (device_id != ncr_chip_table[i].device_id) + continue; + if (revision > ncr_chip_table[i].revision_id) + continue; + if (!(ncr_chip_table[i].features & FE_LDSTR)) + break; + chip = &device->chip; + memcpy(chip, &ncr_chip_table[i], sizeof(*chip)); + chip->revision_id = revision; + break; + } + + /* + ** Ignore Symbios chips controlled by SISL RAID controller. + ** This controller sets value 0x52414944 at RAM end - 16. + */ +#if defined(__i386__) && !defined(SCSI_NCR_PCI_MEM_NOT_SUPPORTED) + if (chip && (base_2 & PCI_BASE_ADDRESS_MEM_MASK)) { + unsigned int ram_size, ram_val; + u_long ram_ptr; + + if (chip->features & FE_RAM8K) + ram_size = 8192; + else + ram_size = 4096; + + ram_ptr = remap_pci_mem(base_2 & PCI_BASE_ADDRESS_MEM_MASK, + ram_size); + if (ram_ptr) { + ram_val = readl_raw(ram_ptr + ram_size - 16); + unmap_pci_mem(ram_ptr, ram_size); + if (ram_val == 0x52414944) { + printk(NAME53C8XX": not initializing, " + "driven by SISL RAID controller.\n"); + return -1; + } + } + } +#endif /* i386 and PCI MEMORY accessible */ + + if (!chip) { + printk(NAME53C8XX ": not initializing, device not supported\n"); + return -1; + } + +#ifdef __powerpc__ + /* + ** Fix-up for power/pc. + ** Should not be performed by the driver. + */ + if ((command & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) + != (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) { + printk(NAME53C8XX ": setting%s%s...\n", + (command & PCI_COMMAND_IO) ? "" : " PCI_COMMAND_IO", + (command & PCI_COMMAND_MEMORY) ? "" : " PCI_COMMAND_MEMORY"); + command |= (PCI_COMMAND_IO | PCI_COMMAND_MEMORY); + pci_write_config_word(pdev, PCI_COMMAND, command); + } + +#if LINUX_VERSION_CODE < LinuxVersionCode(2,2,0) + if ( is_prep ) { + if (io_port >= 0x10000000) { + printk(NAME53C8XX ": reallocating io_port (Wacky IBM)"); + io_port = (io_port & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_0, io_port); + } + if (base >= 0x10000000) { + printk(NAME53C8XX ": reallocating base (Wacky IBM)"); + base = (base & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_1, base); + } + if (base_2 >= 0x10000000) { + printk(NAME53C8XX ": reallocating base2 (Wacky IBM)"); + base_2 = (base_2 & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_2, base_2); + } + } +#endif +#endif /* __powerpc__ */ + +#if defined(__sparc__) && (LINUX_VERSION_CODE < LinuxVersionCode(2,3,0)) + /* + ** Fix-ups for sparc. + ** + ** I wrote: Should not be performed by the driver, + ** Guy wrote: but how can OBP know each and every PCI card, + ** if they don't use Fcode? + ** I replied: no need to know each and every PCI card, just + ** be skilled enough to understand the PCI specs. + */ + + /* + ** PCI configuration is based on configuration registers being + ** coherent with hardware and software resource identifications. + ** This is fairly simple, but seems still too complex for Sparc. + */ + base = __pa(base); + base_2 = __pa(base_2); + + if (!cache_line_size) + suggested_cache_line_size = 16; + + driver_setup.pci_fix_up |= 0x7; + +#endif /* __sparc__ */ + +#if defined(__i386__) && !defined(MODULE) + if (!cache_line_size) { +#if LINUX_VERSION_CODE < LinuxVersionCode(2,1,75) + extern char x86; + switch(x86) { +#else + switch(boot_cpu_data.x86) { +#endif + case 4: suggested_cache_line_size = 4; break; + case 6: + case 5: suggested_cache_line_size = 8; break; + } + } +#endif /* __i386__ */ + + /* + ** Check availability of IO space, memory space. + ** Enable master capability if not yet. + ** + ** We shouldn't have to care about the IO region when + ** we are using MMIO. But calling check_region() from + ** both the ncr53c8xx and the sym53c8xx drivers prevents + ** from attaching devices from the both drivers. + ** If you have a better idea, let me know. + */ +/* #ifdef SCSI_NCR_IOMAPPED */ +#if 1 + if (!(command & PCI_COMMAND_IO)) { + printk(NAME53C8XX ": I/O base address (0x%lx) disabled.\n", + (long) io_port); + io_port = 0; + } +#endif + if (!(command & PCI_COMMAND_MEMORY)) { + printk(NAME53C8XX ": PCI_COMMAND_MEMORY not set.\n"); + base = 0; + base_2 = 0; + } + io_port &= PCI_BASE_ADDRESS_IO_MASK; + base &= PCI_BASE_ADDRESS_MEM_MASK; + base_2 &= PCI_BASE_ADDRESS_MEM_MASK; + +/* #ifdef SCSI_NCR_IOMAPPED */ +#if 1 + if (io_port && check_region (io_port, 128)) { + printk(NAME53C8XX ": IO region 0x%lx[0..127] is in use\n", + (long) io_port); + io_port = 0; + } + if (!io_port) + return -1; +#endif +#ifndef SCSI_NCR_IOMAPPED + if (!base) { + printk(NAME53C8XX ": MMIO base address disabled.\n"); + return -1; + } +#endif + + /* + ** Set MASTER capable and PARITY bit, if not yet. + */ + if ((command & (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY)) + != (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY)) { + printk(NAME53C8XX ": setting%s%s...(fix-up)\n", + (command & PCI_COMMAND_MASTER) ? "" : " PCI_COMMAND_MASTER", + (command & PCI_COMMAND_PARITY) ? "" : " PCI_COMMAND_PARITY"); + command |= (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY); + pci_write_config_word(pdev, PCI_COMMAND, command); + } + + /* + ** Fix some features according to driver setup. + */ + if (!(driver_setup.special_features & 1)) + chip->features &= ~FE_SPECIAL_SET; + else { + if (driver_setup.special_features & 2) + chip->features &= ~FE_WRIE; + if (driver_setup.special_features & 4) + chip->features &= ~FE_NOPM; + } + + /* + ** Work around for errant bit in 895A. The 66Mhz + ** capable bit is set erroneously. Clear this bit. + ** (Item 1 DEL 533) + ** + ** Make sure Config space and Features agree. + ** + ** Recall: writes are not normal to status register - + ** write a 1 to clear and a 0 to leave unchanged. + ** Can only reset bits. + */ + if (chip->features & FE_66MHZ) { + if (!(status_reg & PCI_STATUS_66MHZ)) + chip->features &= ~FE_66MHZ; + } + else { + if (status_reg & PCI_STATUS_66MHZ) { + status_reg = PCI_STATUS_66MHZ; + pci_write_config_word(pdev, PCI_STATUS, status_reg); + pci_read_config_word(pdev, PCI_STATUS, &status_reg); + } + } + + if (driver_setup.ultra_scsi < 3 && (chip->features & FE_ULTRA3)) { + chip->features |= FE_ULTRA2; + chip->features &= ~FE_ULTRA3; + } + if (driver_setup.ultra_scsi < 2 && (chip->features & FE_ULTRA2)) { + chip->features |= FE_ULTRA; + chip->features &= ~FE_ULTRA2; + } + if (driver_setup.ultra_scsi < 1) + chip->features &= ~FE_ULTRA; + + if (!driver_setup.max_wide) + chip->features &= ~FE_WIDE; + + /* + * C1010 Ultra3 support requires 16 bit data transfers. + */ + if (!driver_setup.max_wide && (chip->features & FE_ULTRA3)) { + chip->features |= FE_ULTRA2; + chip->features |= ~FE_ULTRA3; + } + + /* + ** Some features are required to be enabled in order to + ** work around some chip problems. :) ;) + ** (ITEM 12 of a DEL about the 896 I haven't yet). + ** We must ensure the chip will use WRITE AND INVALIDATE. + ** The revision number limit is for now arbitrary. + */ + if (device_id == PCI_DEVICE_ID_NCR_53C896 && revision <= 0x10) { + chip->features |= (FE_WRIE | FE_CLSE); + pci_fix_up |= 3; /* Force appropriate PCI fix-up */ + } + +#ifdef SCSI_NCR_PCI_FIX_UP_SUPPORT + /* + ** Try to fix up PCI config according to wished features. + */ + if ((pci_fix_up & 1) && (chip->features & FE_CLSE) && + !cache_line_size && suggested_cache_line_size) { + cache_line_size = suggested_cache_line_size; + pci_write_config_byte(pdev, + PCI_CACHE_LINE_SIZE, cache_line_size); + printk(NAME53C8XX ": PCI_CACHE_LINE_SIZE set to %d (fix-up).\n", + cache_line_size); + } + + if ((pci_fix_up & 2) && cache_line_size && + (chip->features & FE_WRIE) && !(command & PCI_COMMAND_INVALIDATE)) { + printk(NAME53C8XX": setting PCI_COMMAND_INVALIDATE (fix-up)\n"); + command |= PCI_COMMAND_INVALIDATE; + pci_write_config_word(pdev, PCI_COMMAND, command); + } + + /* + ** Tune PCI LATENCY TIMER according to burst max length transfer. + ** (latency timer >= burst length + 6, we add 10 to be quite sure) + */ + + if (chip->burst_max && (latency_timer == 0 || (pci_fix_up & 4))) { + uchar lt = (1 << chip->burst_max) + 6 + 10; + if (latency_timer < lt) { + printk(NAME53C8XX + ": changing PCI_LATENCY_TIMER from %d to %d.\n", + (int) latency_timer, (int) lt); + latency_timer = lt; + pci_write_config_byte(pdev, + PCI_LATENCY_TIMER, latency_timer); + } + } + +#endif /* SCSI_NCR_PCI_FIX_UP_SUPPORT */ + + /* + ** Initialise ncr_device structure with items required by ncr_attach. + */ + device->pdev = pdev; + device->slot.bus = PciBusNumber(pdev); + device->slot.device_fn = PciDeviceFn(pdev); + device->slot.base = base; + device->slot.base_2 = base_2; + device->slot.io_port = io_port; + device->slot.irq = irq; + device->attach_done = 0; + + return 0; +} + + +/*=================================================================== +** Detect and try to read SYMBIOS and TEKRAM NVRAM. +** +** Data can be used to order booting of boards. +** +** Data is saved in ncr_device structure if NVRAM found. This +** is then used to find drive boot order for ncr_attach(). +** +** NVRAM data is passed to Scsi_Host_Template later during +** ncr_attach() for any device set up. +*=================================================================== +*/ +#ifdef SCSI_NCR_NVRAM_SUPPORT +static void __init ncr_get_nvram(ncr_device *devp, ncr_nvram *nvp) +{ + devp->nvram = nvp; + if (!nvp) + return; + /* + ** Get access to chip IO registers + */ +#ifdef SCSI_NCR_IOMAPPED + request_region(devp->slot.io_port, 128, NAME53C8XX); + devp->slot.base_io = devp->slot.io_port; +#else + devp->slot.reg = (struct ncr_reg *) remap_pci_mem(devp->slot.base, 128); + if (!devp->slot.reg) + return; +#endif + + /* + ** Try to read SYMBIOS nvram. + ** Try to read TEKRAM nvram if Symbios nvram not found. + */ + if (!sym_read_Symbios_nvram(&devp->slot, &nvp->data.Symbios)) + nvp->type = SCSI_NCR_SYMBIOS_NVRAM; + else if (!sym_read_Tekram_nvram(&devp->slot, devp->chip.device_id, + &nvp->data.Tekram)) + nvp->type = SCSI_NCR_TEKRAM_NVRAM; + else { + nvp->type = 0; + devp->nvram = 0; + } + + /* + ** Release access to chip IO registers + */ +#ifdef SCSI_NCR_IOMAPPED + release_region(devp->slot.base_io, 128); +#else + unmap_pci_mem((u_long) devp->slot.reg, 128ul); +#endif + +} +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/* +** Linux select queue depths function +*/ + +#define DEF_DEPTH (driver_setup.default_tags) +#define ALL_TARGETS -2 +#define NO_TARGET -1 +#define ALL_LUNS -2 +#define NO_LUN -1 + +static int device_queue_depth(ncb_p np, int target, int lun) +{ + int c, h, t, u, v; + char *p = driver_setup.tag_ctrl; + char *ep; + + h = -1; + t = NO_TARGET; + u = NO_LUN; + while ((c = *p++) != 0) { + v = simple_strtoul(p, &ep, 0); + switch(c) { + case '/': + ++h; + t = ALL_TARGETS; + u = ALL_LUNS; + break; + case 't': + if (t != target) + t = (target == v) ? v : NO_TARGET; + u = ALL_LUNS; + break; + case 'u': + if (u != lun) + u = (lun == v) ? v : NO_LUN; + break; + case 'q': + if (h == np->unit && + (t == ALL_TARGETS || t == target) && + (u == ALL_LUNS || u == lun)) + return v; + break; + case '-': + t = ALL_TARGETS; + u = ALL_LUNS; + break; + default: + break; + } + p = ep; + } + return DEF_DEPTH; +} + +static void sym53c8xx_select_queue_depths(struct Scsi_Host *host, struct scsi_device *devlist) +{ + struct scsi_device *device; + + for (device = devlist; device; device = device->next) { + ncb_p np; + tcb_p tp; + lcb_p lp; + int numtags; + + if (device->host != host) + continue; + + np = ((struct host_data *) host->hostdata)->ncb; + tp = &np->target[device->id]; + lp = ncr_lp(np, tp, device->lun); + + /* + ** Select queue depth from driver setup. + ** Donnot use more than configured by user. + ** Use at least 2. + ** Donnot use more than our maximum. + */ + numtags = device_queue_depth(np, device->id, device->lun); + if (numtags > tp->usrtags) + numtags = tp->usrtags; + if (!device->tagged_supported) + numtags = 1; + device->queue_depth = numtags; + if (device->queue_depth < 2) + device->queue_depth = 2; + if (device->queue_depth > MAX_TAGS) + device->queue_depth = MAX_TAGS; + + /* + ** Since the queue depth is not tunable under Linux, + ** we need to know this value in order not to + ** announce stupid things to user. + */ + if (lp) { + lp->numtags = lp->maxtags = numtags; + lp->scdev_depth = device->queue_depth; + } + ncr_setup_tags (np, device->id, device->lun); + +#ifdef DEBUG_SYM53C8XX +printk("sym53c8xx_select_queue_depth: host=%d, id=%d, lun=%d, depth=%d\n", + np->unit, device->id, device->lun, device->queue_depth); +#endif + } +} + +/* +** Linux entry point for info() function +*/ +const char *sym53c8xx_info (struct Scsi_Host *host) +{ + return SCSI_NCR_DRIVER_NAME; +} + +/* +** Linux entry point of queuecommand() function +*/ + +int sym53c8xx_queue_command (Scsi_Cmnd *cmd, void (* done)(Scsi_Cmnd *)) +{ + ncb_p np = ((struct host_data *) cmd->host->hostdata)->ncb; + unsigned long flags; + int sts; + +#ifdef DEBUG_SYM53C8XX +printk("sym53c8xx_queue_command\n"); +#endif + + cmd->scsi_done = done; + cmd->host_scribble = NULL; + cmd->SCp.ptr = NULL; + cmd->SCp.buffer = NULL; +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING + cmd->__data_mapped = 0; + cmd->__data_mapping = 0; +#endif + + NCR_LOCK_NCB(np, flags); + + if ((sts = ncr_queue_command(np, cmd)) != DID_OK) { + SetScsiResult(cmd, sts, 0); +#ifdef DEBUG_SYM53C8XX +printk("sym53c8xx : command not queued - result=%d\n", sts); +#endif + } +#ifdef DEBUG_SYM53C8XX + else +printk("sym53c8xx : command successfully queued\n"); +#endif + + NCR_UNLOCK_NCB(np, flags); + + if (sts != DID_OK) { + unmap_scsi_data(np, cmd); + done(cmd); + } + + return sts; +} + +/* +** Linux entry point of the interrupt handler. +** Since linux versions > 1.3.70, we trust the kernel for +** passing the internal host descriptor as 'dev_id'. +** Otherwise, we scan the host list and call the interrupt +** routine for each host that uses this IRQ. +*/ + +static void sym53c8xx_intr(int irq, void *dev_id, struct pt_regs * regs) +{ + unsigned long flags; + ncb_p np = (ncb_p) dev_id; + Scsi_Cmnd *done_list; + +#ifdef DEBUG_SYM53C8XX + printk("sym53c8xx : interrupt received\n"); +#endif + + if (DEBUG_FLAGS & DEBUG_TINY) printk ("["); + + NCR_LOCK_NCB(np, flags); + ncr_exception(np); + done_list = np->done_list; + np->done_list = 0; + NCR_UNLOCK_NCB(np, flags); + + if (DEBUG_FLAGS & DEBUG_TINY) printk ("]\n"); + + if (done_list) { + NCR_LOCK_SCSI_DONE(np, flags); + ncr_flush_done_cmds(done_list); + NCR_UNLOCK_SCSI_DONE(np, flags); + } +} + +/* +** Linux entry point of the timer handler +*/ + +static void sym53c8xx_timeout(unsigned long npref) +{ + ncb_p np = (ncb_p) npref; + unsigned long flags; + Scsi_Cmnd *done_list; + + NCR_LOCK_NCB(np, flags); + ncr_timeout((ncb_p) np); + done_list = np->done_list; + np->done_list = 0; + NCR_UNLOCK_NCB(np, flags); + + if (done_list) { + NCR_LOCK_SCSI_DONE(np, flags); + ncr_flush_done_cmds(done_list); + NCR_UNLOCK_SCSI_DONE(np, flags); + } +} + +/* +** Linux entry point of reset() function +*/ + +#if defined SCSI_RESET_SYNCHRONOUS && defined SCSI_RESET_ASYNCHRONOUS +int sym53c8xx_reset(Scsi_Cmnd *cmd, unsigned int reset_flags) +#else +int sym53c8xx_reset(Scsi_Cmnd *cmd) +#endif +{ + ncb_p np = ((struct host_data *) cmd->host->hostdata)->ncb; + int sts; + unsigned long flags; + Scsi_Cmnd *done_list; + +#if defined SCSI_RESET_SYNCHRONOUS && defined SCSI_RESET_ASYNCHRONOUS + printk("sym53c8xx_reset: pid=%lu reset_flags=%x serial_number=%ld serial_number_at_timeout=%ld\n", + cmd->pid, reset_flags, cmd->serial_number, cmd->serial_number_at_timeout); +#else + printk("sym53c8xx_reset: command pid %lu\n", cmd->pid); +#endif + + NCR_LOCK_NCB(np, flags); + + /* + * We have to just ignore reset requests in some situations. + */ +#if defined SCSI_RESET_NOT_RUNNING + if (cmd->serial_number != cmd->serial_number_at_timeout) { + sts = SCSI_RESET_NOT_RUNNING; + goto out; + } +#endif + /* + * If the mid-level driver told us reset is synchronous, it seems + * that we must call the done() callback for the involved command, + * even if this command was not queued to the low-level driver, + * before returning SCSI_RESET_SUCCESS. + */ + +#if defined SCSI_RESET_SYNCHRONOUS && defined SCSI_RESET_ASYNCHRONOUS + sts = ncr_reset_bus(np, cmd, + (reset_flags & (SCSI_RESET_SYNCHRONOUS | SCSI_RESET_ASYNCHRONOUS)) == SCSI_RESET_SYNCHRONOUS); +#else + sts = ncr_reset_bus(np, cmd, 0); +#endif + + /* + * Since we always reset the controller, when we return success, + * we add this information to the return code. + */ +#if defined SCSI_RESET_HOST_RESET + if (sts == SCSI_RESET_SUCCESS) + sts |= SCSI_RESET_HOST_RESET; +#endif + +out: + done_list = np->done_list; + np->done_list = 0; + NCR_UNLOCK_NCB(np, flags); + + ncr_flush_done_cmds(done_list); + + return sts; +} + +/* +** Linux entry point of abort() function +*/ + +int sym53c8xx_abort(Scsi_Cmnd *cmd) +{ + ncb_p np = ((struct host_data *) cmd->host->hostdata)->ncb; + int sts; + unsigned long flags; + Scsi_Cmnd *done_list; + +#if defined SCSI_RESET_SYNCHRONOUS && defined SCSI_RESET_ASYNCHRONOUS + printk("sym53c8xx_abort: pid=%lu serial_number=%ld serial_number_at_timeout=%ld\n", + cmd->pid, cmd->serial_number, cmd->serial_number_at_timeout); +#else + printk("sym53c8xx_abort: command pid %lu\n", cmd->pid); +#endif + + NCR_LOCK_NCB(np, flags); + +#if defined SCSI_RESET_SYNCHRONOUS && defined SCSI_RESET_ASYNCHRONOUS + /* + * We have to just ignore abort requests in some situations. + */ + if (cmd->serial_number != cmd->serial_number_at_timeout) { + sts = SCSI_ABORT_NOT_RUNNING; + goto out; + } +#endif + + sts = ncr_abort_command(np, cmd); +out: + done_list = np->done_list; + np->done_list = 0; + NCR_UNLOCK_NCB(np, flags); + + ncr_flush_done_cmds(done_list); + + return sts; +} + + +#ifdef MODULE +int sym53c8xx_release(struct Scsi_Host *host) +{ +#ifdef DEBUG_SYM53C8XX +printk("sym53c8xx : release\n"); +#endif + ncr_detach(((struct host_data *) host->hostdata)->ncb); + + return 1; +} +#endif + + +/* +** Scsi command waiting list management. +** +** It may happen that we cannot insert a scsi command into the start queue, +** in the following circumstances. +** Too few preallocated ccb(s), +** maxtags < cmd_per_lun of the Linux host control block, +** etc... +** Such scsi commands are inserted into a waiting list. +** When a scsi command complete, we try to requeue the commands of the +** waiting list. +*/ + +#define next_wcmd host_scribble + +static void insert_into_waiting_list(ncb_p np, Scsi_Cmnd *cmd) +{ + Scsi_Cmnd *wcmd; + +#ifdef DEBUG_WAITING_LIST + printk("%s: cmd %lx inserted into waiting list\n", ncr_name(np), (u_long) cmd); +#endif + cmd->next_wcmd = 0; + if (!(wcmd = np->waiting_list)) np->waiting_list = cmd; + else { + while ((wcmd->next_wcmd) != 0) + wcmd = (Scsi_Cmnd *) wcmd->next_wcmd; + wcmd->next_wcmd = (char *) cmd; + } +} + +static Scsi_Cmnd *retrieve_from_waiting_list(int to_remove, ncb_p np, Scsi_Cmnd *cmd) +{ + Scsi_Cmnd **pcmd = &np->waiting_list; + + while (*pcmd) { + if (cmd == *pcmd) { + if (to_remove) { + *pcmd = (Scsi_Cmnd *) cmd->next_wcmd; + cmd->next_wcmd = 0; + } +#ifdef DEBUG_WAITING_LIST + printk("%s: cmd %lx retrieved from waiting list\n", ncr_name(np), (u_long) cmd); +#endif + return cmd; + } + pcmd = (Scsi_Cmnd **) &(*pcmd)->next_wcmd; + } + return 0; +} + +static void process_waiting_list(ncb_p np, int sts) +{ + Scsi_Cmnd *waiting_list, *wcmd; + + waiting_list = np->waiting_list; + np->waiting_list = 0; + +#ifdef DEBUG_WAITING_LIST + if (waiting_list) printk("%s: waiting_list=%lx processing sts=%d\n", ncr_name(np), (u_long) waiting_list, sts); +#endif + while ((wcmd = waiting_list) != 0) { + waiting_list = (Scsi_Cmnd *) wcmd->next_wcmd; + wcmd->next_wcmd = 0; + if (sts == DID_OK) { +#ifdef DEBUG_WAITING_LIST + printk("%s: cmd %lx trying to requeue\n", ncr_name(np), (u_long) wcmd); +#endif + sts = ncr_queue_command(np, wcmd); + } + if (sts != DID_OK) { +#ifdef DEBUG_WAITING_LIST + printk("%s: cmd %lx done forced sts=%d\n", ncr_name(np), (u_long) wcmd, sts); +#endif + SetScsiResult(wcmd, sts, 0); + ncr_queue_done_cmd(np, wcmd); + } + } +} + +#undef next_wcmd + +#ifdef SCSI_NCR_PROC_INFO_SUPPORT + +/*========================================================================= +** Proc file system stuff +** +** A read operation returns adapter information. +** A write operation is a control command. +** The string is parsed in the driver code and the command is passed +** to the ncr_usercmd() function. +**========================================================================= +*/ + +#ifdef SCSI_NCR_USER_COMMAND_SUPPORT + +#define is_digit(c) ((c) >= '0' && (c) <= '9') +#define digit_to_bin(c) ((c) - '0') +#define is_space(c) ((c) == ' ' || (c) == '\t') + +static int skip_spaces(char *ptr, int len) +{ + int cnt, c; + + for (cnt = len; cnt > 0 && (c = *ptr++) && is_space(c); cnt--); + + return (len - cnt); +} + +static int get_int_arg(char *ptr, int len, u_long *pv) +{ + int cnt, c; + u_long v; + + for (v = 0, cnt = len; cnt > 0 && (c = *ptr++) && is_digit(c); cnt--) { + v = (v * 10) + digit_to_bin(c); + } + + if (pv) + *pv = v; + + return (len - cnt); +} + +static int is_keyword(char *ptr, int len, char *verb) +{ + int verb_len = strlen(verb); + + if (len >= strlen(verb) && !memcmp(verb, ptr, verb_len)) + return verb_len; + else + return 0; + +} + +#define SKIP_SPACES(min_spaces) \ + if ((arg_len = skip_spaces(ptr, len)) < (min_spaces)) \ + return -EINVAL; \ + ptr += arg_len; len -= arg_len; + +#define GET_INT_ARG(v) \ + if (!(arg_len = get_int_arg(ptr, len, &(v)))) \ + return -EINVAL; \ + ptr += arg_len; len -= arg_len; + + +/* +** Parse a control command +*/ + +static int ncr_user_command(ncb_p np, char *buffer, int length) +{ + char *ptr = buffer; + int len = length; + struct usrcmd *uc = &np->user; + int arg_len; + u_long target; + + bzero(uc, sizeof(*uc)); + + if (len > 0 && ptr[len-1] == '\n') + --len; + + if ((arg_len = is_keyword(ptr, len, "setsync")) != 0) + uc->cmd = UC_SETSYNC; + else if ((arg_len = is_keyword(ptr, len, "settags")) != 0) + uc->cmd = UC_SETTAGS; + else if ((arg_len = is_keyword(ptr, len, "setorder")) != 0) + uc->cmd = UC_SETORDER; + else if ((arg_len = is_keyword(ptr, len, "setverbose")) != 0) + uc->cmd = UC_SETVERBOSE; + else if ((arg_len = is_keyword(ptr, len, "setwide")) != 0) + uc->cmd = UC_SETWIDE; + else if ((arg_len = is_keyword(ptr, len, "setdebug")) != 0) + uc->cmd = UC_SETDEBUG; + else if ((arg_len = is_keyword(ptr, len, "setflag")) != 0) + uc->cmd = UC_SETFLAG; + else if ((arg_len = is_keyword(ptr, len, "resetdev")) != 0) + uc->cmd = UC_RESETDEV; + else if ((arg_len = is_keyword(ptr, len, "cleardev")) != 0) + uc->cmd = UC_CLEARDEV; + else + arg_len = 0; + +#ifdef DEBUG_PROC_INFO +printk("ncr_user_command: arg_len=%d, cmd=%ld\n", arg_len, uc->cmd); +#endif + + if (!arg_len) + return -EINVAL; + ptr += arg_len; len -= arg_len; + + switch(uc->cmd) { + case UC_SETSYNC: + case UC_SETTAGS: + case UC_SETWIDE: + case UC_SETFLAG: + case UC_RESETDEV: + case UC_CLEARDEV: + SKIP_SPACES(1); + if ((arg_len = is_keyword(ptr, len, "all")) != 0) { + ptr += arg_len; len -= arg_len; + uc->target = ~0; + } else { + GET_INT_ARG(target); + uc->target = (1<<target); +#ifdef DEBUG_PROC_INFO +printk("ncr_user_command: target=%ld\n", target); +#endif + } + break; + } + + switch(uc->cmd) { + case UC_SETVERBOSE: + case UC_SETSYNC: + case UC_SETTAGS: + case UC_SETWIDE: + SKIP_SPACES(1); + GET_INT_ARG(uc->data); +#ifdef DEBUG_PROC_INFO +printk("ncr_user_command: data=%ld\n", uc->data); +#endif + break; + case UC_SETORDER: + SKIP_SPACES(1); + if ((arg_len = is_keyword(ptr, len, "simple"))) + uc->data = M_SIMPLE_TAG; + else if ((arg_len = is_keyword(ptr, len, "ordered"))) + uc->data = M_ORDERED_TAG; + else if ((arg_len = is_keyword(ptr, len, "default"))) + uc->data = 0; + else + return -EINVAL; + break; + case UC_SETDEBUG: + while (len > 0) { + SKIP_SPACES(1); + if ((arg_len = is_keyword(ptr, len, "alloc"))) + uc->data |= DEBUG_ALLOC; + else if ((arg_len = is_keyword(ptr, len, "phase"))) + uc->data |= DEBUG_PHASE; + else if ((arg_len = is_keyword(ptr, len, "queue"))) + uc->data |= DEBUG_QUEUE; + else if ((arg_len = is_keyword(ptr, len, "result"))) + uc->data |= DEBUG_RESULT; + else if ((arg_len = is_keyword(ptr, len, "pointer"))) + uc->data |= DEBUG_POINTER; + else if ((arg_len = is_keyword(ptr, len, "script"))) + uc->data |= DEBUG_SCRIPT; + else if ((arg_len = is_keyword(ptr, len, "tiny"))) + uc->data |= DEBUG_TINY; + else if ((arg_len = is_keyword(ptr, len, "timing"))) + uc->data |= DEBUG_TIMING; + else if ((arg_len = is_keyword(ptr, len, "nego"))) + uc->data |= DEBUG_NEGO; + else if ((arg_len = is_keyword(ptr, len, "tags"))) + uc->data |= DEBUG_TAGS; + else + return -EINVAL; + ptr += arg_len; len -= arg_len; + } +#ifdef DEBUG_PROC_INFO +printk("ncr_user_command: data=%ld\n", uc->data); +#endif + break; + case UC_SETFLAG: + while (len > 0) { + SKIP_SPACES(1); + if ((arg_len = is_keyword(ptr, len, "trace"))) + uc->data |= UF_TRACE; + else if ((arg_len = is_keyword(ptr, len, "no_disc"))) + uc->data |= UF_NODISC; + else + return -EINVAL; + ptr += arg_len; len -= arg_len; + } + break; + default: + break; + } + + if (len) + return -EINVAL; + else { + long flags; + + NCR_LOCK_NCB(np, flags); + ncr_usercmd (np); + NCR_UNLOCK_NCB(np, flags); + } + return length; +} + +#endif /* SCSI_NCR_USER_COMMAND_SUPPORT */ + +#ifdef SCSI_NCR_USER_INFO_SUPPORT + +struct info_str +{ + char *buffer; + int length; + int offset; + int pos; +}; + +static void copy_mem_info(struct info_str *info, char *data, int len) +{ + if (info->pos + len > info->length) + len = info->length - info->pos; + + if (info->pos + len < info->offset) { + info->pos += len; + return; + } + if (info->pos < info->offset) { + data += (info->offset - info->pos); + len -= (info->offset - info->pos); + } + + if (len > 0) { + memcpy(info->buffer + info->pos, data, len); + info->pos += len; + } +} + +static int copy_info(struct info_str *info, char *fmt, ...) +{ + va_list args; + char buf[81]; + int len; + + va_start(args, fmt); + len = vsprintf(buf, fmt, args); + va_end(args); + + copy_mem_info(info, buf, len); + return len; +} + +/* +** Copy formatted information into the input buffer. +*/ + +static int ncr_host_info(ncb_p np, char *ptr, off_t offset, int len) +{ + struct info_str info; +#ifdef CONFIG_ALL_PPC + struct device_node* of_node; +#endif + + info.buffer = ptr; + info.length = len; + info.offset = offset; + info.pos = 0; + + copy_info(&info, "General information:\n"); + copy_info(&info, " Chip " NAME53C "%s, device id 0x%x, " + "revision id 0x%x\n", + np->chip_name, np->device_id, np->revision_id); + copy_info(&info, " On PCI bus %d, device %d, function %d, " +#ifdef __sparc__ + "IRQ %s\n", +#else + "IRQ %d\n", +#endif + np->bus, (np->device_fn & 0xf8) >> 3, np->device_fn & 7, +#ifdef __sparc__ + __irq_itoa(np->irq)); +#else + (int) np->irq); +#endif +#ifdef CONFIG_ALL_PPC + of_node = find_pci_device_OFnode(np->bus, np->device_fn); + if (of_node && of_node->full_name) + copy_info(&info, "PPC OpenFirmware path : %s\n", of_node->full_name); +#endif + copy_info(&info, " Synchronous period factor %d, " + "max commands per lun %d\n", + (int) np->minsync, MAX_TAGS); + + if (driver_setup.debug || driver_setup.verbose > 1) { + copy_info(&info, " Debug flags 0x%x, verbosity level %d\n", + driver_setup.debug, driver_setup.verbose); + } + + return info.pos > info.offset? info.pos - info.offset : 0; +} + +#endif /* SCSI_NCR_USER_INFO_SUPPORT */ + +/* +** Entry point of the scsi proc fs of the driver. +** - func = 0 means read (returns adapter infos) +** - func = 1 means write (parse user control command) +*/ + +static int sym53c8xx_proc_info(char *buffer, char **start, off_t offset, + int length, int hostno, int func) +{ + struct Scsi_Host *host; + struct host_data *host_data; + ncb_p ncb = 0; + int retv; + +#ifdef DEBUG_PROC_INFO +printk("sym53c8xx_proc_info: hostno=%d, func=%d\n", hostno, func); +#endif + + for (host = first_host; host; host = host->next) { + if (host->hostt != first_host->hostt) + continue; + if (host->host_no == hostno) { + host_data = (struct host_data *) host->hostdata; + ncb = host_data->ncb; + break; + } + } + + if (!ncb) + return -EINVAL; + + if (func) { +#ifdef SCSI_NCR_USER_COMMAND_SUPPORT + retv = ncr_user_command(ncb, buffer, length); +#else + retv = -EINVAL; +#endif + } + else { + if (start) + *start = buffer; +#ifdef SCSI_NCR_USER_INFO_SUPPORT + retv = ncr_host_info(ncb, buffer, offset, length); +#else + retv = -EINVAL; +#endif + } + + return retv; +} + + +/*========================================================================= +** End of proc file system stuff +**========================================================================= +*/ +#endif + + +#ifdef SCSI_NCR_NVRAM_SUPPORT + +/* + * 24C16 EEPROM reading. + * + * GPOI0 - data in/data out + * GPIO1 - clock + * Symbios NVRAM wiring now also used by Tekram. + */ + +#define SET_BIT 0 +#define CLR_BIT 1 +#define SET_CLK 2 +#define CLR_CLK 3 + +/* + * Set/clear data/clock bit in GPIO0 + */ +static void __init +S24C16_set_bit(ncr_slot *np, u_char write_bit, u_char *gpreg, int bit_mode) +{ + UDELAY (5); + switch (bit_mode){ + case SET_BIT: + *gpreg |= write_bit; + break; + case CLR_BIT: + *gpreg &= 0xfe; + break; + case SET_CLK: + *gpreg |= 0x02; + break; + case CLR_CLK: + *gpreg &= 0xfd; + break; + + } + OUTB (nc_gpreg, *gpreg); + UDELAY (5); +} + +/* + * Send START condition to NVRAM to wake it up. + */ +static void __init S24C16_start(ncr_slot *np, u_char *gpreg) +{ + S24C16_set_bit(np, 1, gpreg, SET_BIT); + S24C16_set_bit(np, 0, gpreg, SET_CLK); + S24C16_set_bit(np, 0, gpreg, CLR_BIT); + S24C16_set_bit(np, 0, gpreg, CLR_CLK); +} + +/* + * Send STOP condition to NVRAM - puts NVRAM to sleep... ZZzzzz!! + */ +static void __init S24C16_stop(ncr_slot *np, u_char *gpreg) +{ + S24C16_set_bit(np, 0, gpreg, SET_CLK); + S24C16_set_bit(np, 1, gpreg, SET_BIT); +} + +/* + * Read or write a bit to the NVRAM, + * read if GPIO0 input else write if GPIO0 output + */ +static void __init +S24C16_do_bit(ncr_slot *np, u_char *read_bit, u_char write_bit, u_char *gpreg) +{ + S24C16_set_bit(np, write_bit, gpreg, SET_BIT); + S24C16_set_bit(np, 0, gpreg, SET_CLK); + if (read_bit) + *read_bit = INB (nc_gpreg); + S24C16_set_bit(np, 0, gpreg, CLR_CLK); + S24C16_set_bit(np, 0, gpreg, CLR_BIT); +} + +/* + * Output an ACK to the NVRAM after reading, + * change GPIO0 to output and when done back to an input + */ +static void __init +S24C16_write_ack(ncr_slot *np, u_char write_bit, u_char *gpreg, u_char *gpcntl) +{ + OUTB (nc_gpcntl, *gpcntl & 0xfe); + S24C16_do_bit(np, 0, write_bit, gpreg); + OUTB (nc_gpcntl, *gpcntl); +} + +/* + * Input an ACK from NVRAM after writing, + * change GPIO0 to input and when done back to an output + */ +static void __init +S24C16_read_ack(ncr_slot *np, u_char *read_bit, u_char *gpreg, u_char *gpcntl) +{ + OUTB (nc_gpcntl, *gpcntl | 0x01); + S24C16_do_bit(np, read_bit, 1, gpreg); + OUTB (nc_gpcntl, *gpcntl); +} + +/* + * WRITE a byte to the NVRAM and then get an ACK to see it was accepted OK, + * GPIO0 must already be set as an output + */ +static void __init +S24C16_write_byte(ncr_slot *np, u_char *ack_data, u_char write_data, + u_char *gpreg, u_char *gpcntl) +{ + int x; + + for (x = 0; x < 8; x++) + S24C16_do_bit(np, 0, (write_data >> (7 - x)) & 0x01, gpreg); + + S24C16_read_ack(np, ack_data, gpreg, gpcntl); +} + +/* + * READ a byte from the NVRAM and then send an ACK to say we have got it, + * GPIO0 must already be set as an input + */ +static void __init +S24C16_read_byte(ncr_slot *np, u_char *read_data, u_char ack_data, + u_char *gpreg, u_char *gpcntl) +{ + int x; + u_char read_bit; + + *read_data = 0; + for (x = 0; x < 8; x++) { + S24C16_do_bit(np, &read_bit, 1, gpreg); + *read_data |= ((read_bit & 0x01) << (7 - x)); + } + + S24C16_write_ack(np, ack_data, gpreg, gpcntl); +} + +/* + * Read 'len' bytes starting at 'offset'. + */ +static int __init +sym_read_S24C16_nvram (ncr_slot *np, int offset, u_char *data, int len) +{ + u_char gpcntl, gpreg; + u_char old_gpcntl, old_gpreg; + u_char ack_data; + int retv = 1; + int x; + + /* save current state of GPCNTL and GPREG */ + old_gpreg = INB (nc_gpreg); + old_gpcntl = INB (nc_gpcntl); + gpcntl = old_gpcntl & 0xfc; + + /* set up GPREG & GPCNTL to set GPIO0 and GPIO1 in to known state */ + OUTB (nc_gpreg, old_gpreg); + OUTB (nc_gpcntl, gpcntl); + + /* this is to set NVRAM into a known state with GPIO0/1 both low */ + gpreg = old_gpreg; + S24C16_set_bit(np, 0, &gpreg, CLR_CLK); + S24C16_set_bit(np, 0, &gpreg, CLR_BIT); + + /* now set NVRAM inactive with GPIO0/1 both high */ + S24C16_stop(np, &gpreg); + + /* activate NVRAM */ + S24C16_start(np, &gpreg); + + /* write device code and random address MSB */ + S24C16_write_byte(np, &ack_data, + 0xa0 | ((offset >> 7) & 0x0e), &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* write random address LSB */ + S24C16_write_byte(np, &ack_data, + offset & 0xff, &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* regenerate START state to set up for reading */ + S24C16_start(np, &gpreg); + + /* rewrite device code and address MSB with read bit set (lsb = 0x01) */ + S24C16_write_byte(np, &ack_data, + 0xa1 | ((offset >> 7) & 0x0e), &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* now set up GPIO0 for inputting data */ + gpcntl |= 0x01; + OUTB (nc_gpcntl, gpcntl); + + /* input all requested data - only part of total NVRAM */ + for (x = 0; x < len; x++) + S24C16_read_byte(np, &data[x], (x == (len-1)), &gpreg, &gpcntl); + + /* finally put NVRAM back in inactive mode */ + gpcntl &= 0xfe; + OUTB (nc_gpcntl, gpcntl); + S24C16_stop(np, &gpreg); + retv = 0; +out: + /* return GPIO0/1 to original states after having accessed NVRAM */ + OUTB (nc_gpcntl, old_gpcntl); + OUTB (nc_gpreg, old_gpreg); + + return retv; +} + +#undef SET_BIT +#undef CLR_BIT +#undef SET_CLK +#undef CLR_CLK + +/* + * Try reading Symbios NVRAM. + * Return 0 if OK. + */ +static int __init sym_read_Symbios_nvram (ncr_slot *np, Symbios_nvram *nvram) +{ + static u_char Symbios_trailer[6] = {0xfe, 0xfe, 0, 0, 0, 0}; + u_char *data = (u_char *) nvram; + int len = sizeof(*nvram); + u_short csum; + int x; + + /* probe the 24c16 and read the SYMBIOS 24c16 area */ + if (sym_read_S24C16_nvram (np, SYMBIOS_NVRAM_ADDRESS, data, len)) + return 1; + + /* check valid NVRAM signature, verify byte count and checksum */ + if (nvram->type != 0 || + memcmp(nvram->trailer, Symbios_trailer, 6) || + nvram->byte_count != len - 12) + return 1; + + /* verify checksum */ + for (x = 6, csum = 0; x < len - 6; x++) + csum += data[x]; + if (csum != nvram->checksum) + return 1; + + return 0; +} + +/* + * 93C46 EEPROM reading. + * + * GPOI0 - data in + * GPIO1 - data out + * GPIO2 - clock + * GPIO4 - chip select + * + * Used by Tekram. + */ + +/* + * Pulse clock bit in GPIO0 + */ +static void __init T93C46_Clk(ncr_slot *np, u_char *gpreg) +{ + OUTB (nc_gpreg, *gpreg | 0x04); + UDELAY (2); + OUTB (nc_gpreg, *gpreg); +} + +/* + * Read bit from NVRAM + */ +static void __init T93C46_Read_Bit(ncr_slot *np, u_char *read_bit, u_char *gpreg) +{ + UDELAY (2); + T93C46_Clk(np, gpreg); + *read_bit = INB (nc_gpreg); +} + +/* + * Write bit to GPIO0 + */ +static void __init T93C46_Write_Bit(ncr_slot *np, u_char write_bit, u_char *gpreg) +{ + if (write_bit & 0x01) + *gpreg |= 0x02; + else + *gpreg &= 0xfd; + + *gpreg |= 0x10; + + OUTB (nc_gpreg, *gpreg); + UDELAY (2); + + T93C46_Clk(np, gpreg); +} + +/* + * Send STOP condition to NVRAM - puts NVRAM to sleep... ZZZzzz!! + */ +static void __init T93C46_Stop(ncr_slot *np, u_char *gpreg) +{ + *gpreg &= 0xef; + OUTB (nc_gpreg, *gpreg); + UDELAY (2); + + T93C46_Clk(np, gpreg); +} + +/* + * Send read command and address to NVRAM + */ +static void __init +T93C46_Send_Command(ncr_slot *np, u_short write_data, + u_char *read_bit, u_char *gpreg) +{ + int x; + + /* send 9 bits, start bit (1), command (2), address (6) */ + for (x = 0; x < 9; x++) + T93C46_Write_Bit(np, (u_char) (write_data >> (8 - x)), gpreg); + + *read_bit = INB (nc_gpreg); +} + +/* + * READ 2 bytes from the NVRAM + */ +static void __init +T93C46_Read_Word(ncr_slot *np, u_short *nvram_data, u_char *gpreg) +{ + int x; + u_char read_bit; + + *nvram_data = 0; + for (x = 0; x < 16; x++) { + T93C46_Read_Bit(np, &read_bit, gpreg); + + if (read_bit & 0x01) + *nvram_data |= (0x01 << (15 - x)); + else + *nvram_data &= ~(0x01 << (15 - x)); + } +} + +/* + * Read Tekram NvRAM data. + */ +static int __init +T93C46_Read_Data(ncr_slot *np, u_short *data,int len,u_char *gpreg) +{ + u_char read_bit; + int x; + + for (x = 0; x < len; x++) { + + /* output read command and address */ + T93C46_Send_Command(np, 0x180 | x, &read_bit, gpreg); + if (read_bit & 0x01) + return 1; /* Bad */ + T93C46_Read_Word(np, &data[x], gpreg); + T93C46_Stop(np, gpreg); + } + + return 0; +} + +/* + * Try reading 93C46 Tekram NVRAM. + */ +static int __init +sym_read_T93C46_nvram (ncr_slot *np, Tekram_nvram *nvram) +{ + u_char gpcntl, gpreg; + u_char old_gpcntl, old_gpreg; + int retv = 1; + + /* save current state of GPCNTL and GPREG */ + old_gpreg = INB (nc_gpreg); + old_gpcntl = INB (nc_gpcntl); + + /* set up GPREG & GPCNTL to set GPIO0/1/2/4 in to known state, 0 in, + 1/2/4 out */ + gpreg = old_gpreg & 0xe9; + OUTB (nc_gpreg, gpreg); + gpcntl = (old_gpcntl & 0xe9) | 0x09; + OUTB (nc_gpcntl, gpcntl); + + /* input all of NVRAM, 64 words */ + retv = T93C46_Read_Data(np, (u_short *) nvram, + sizeof(*nvram) / sizeof(short), &gpreg); + + /* return GPIO0/1/2/4 to original states after having accessed NVRAM */ + OUTB (nc_gpcntl, old_gpcntl); + OUTB (nc_gpreg, old_gpreg); + + return retv; +} + +/* + * Try reading Tekram NVRAM. + * Return 0 if OK. + */ +static int __init +sym_read_Tekram_nvram (ncr_slot *np, u_short device_id, Tekram_nvram *nvram) +{ + u_char *data = (u_char *) nvram; + int len = sizeof(*nvram); + u_short csum; + int x; + + switch (device_id) { + case PCI_DEVICE_ID_NCR_53C885: + case PCI_DEVICE_ID_NCR_53C895: + case PCI_DEVICE_ID_NCR_53C896: + x = sym_read_S24C16_nvram(np, TEKRAM_24C16_NVRAM_ADDRESS, + data, len); + break; + case PCI_DEVICE_ID_NCR_53C875: + x = sym_read_S24C16_nvram(np, TEKRAM_24C16_NVRAM_ADDRESS, + data, len); + if (!x) + break; + default: + x = sym_read_T93C46_nvram(np, nvram); + break; + } + if (x) + return 1; + + /* verify checksum */ + for (x = 0, csum = 0; x < len - 1; x += 2) + csum += data[x] + (data[x+1] << 8); + if (csum != 0x1234) + return 1; + + return 0; +} + +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/* +** Module stuff +*/ + +#ifdef MODULE +Scsi_Host_Template driver_template = SYM53C8XX; +#include "scsi_module.c" +#endif diff --git a/linux/src/drivers/scsi/sym53c8xx.h b/linux/src/drivers/scsi/sym53c8xx.h new file mode 100644 index 00000000..128fe165 --- /dev/null +++ b/linux/src/drivers/scsi/sym53c8xx.h @@ -0,0 +1,116 @@ +/****************************************************************************** +** High Performance device driver for the Symbios 53C896 controller. +** +** Copyright (C) 1998-2000 Gerard Roudier <groudier@club-internet.fr> +** +** This driver also supports all the Symbios 53C8XX controller family, +** except 53C810 revisions < 16, 53C825 revisions < 16 and all +** revisions of 53C815 controllers. +** +** This driver is based on the Linux port of the FreeBSD ncr driver. +** +** Copyright (C) 1994 Wolfgang Stanglmeier +** +**----------------------------------------------------------------------------- +** +** This program is free software; you can redistribute it and/or modify +** it under the terms of the GNU General Public License as published by +** the Free Software Foundation; either version 2 of the License, or +** (at your option) any later version. +** +** This program is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU General Public License for more details. +** +** You should have received a copy of the GNU General Public License +** along with this program; if not, write to the Free Software +** Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. +** +**----------------------------------------------------------------------------- +** +** The Linux port of the FreeBSD ncr driver has been achieved in +** november 1995 by: +** +** Gerard Roudier <groudier@club-internet.fr> +** +** Being given that this driver originates from the FreeBSD version, and +** in order to keep synergy on both, any suggested enhancements and corrections +** received on Linux are automatically a potential candidate for the FreeBSD +** version. +** +** The original driver has been written for 386bsd and FreeBSD by +** Wolfgang Stanglmeier <wolf@cologne.de> +** Stefan Esser <se@mi.Uni-Koeln.de> +** +**----------------------------------------------------------------------------- +** +** Major contributions: +** -------------------- +** +** NVRAM detection and reading. +** Copyright (C) 1997 Richard Waltham <dormouse@farsrobt.demon.co.uk> +** +******************************************************************************* +*/ + +#ifndef SYM53C8XX_H +#define SYM53C8XX_H + +#include "sym53c8xx_defs.h" + +/* +** Define Scsi_Host_Template parameters +** +** Used by hosts.c and sym53c8xx.c with module configuration. +*/ + +#if defined(HOSTS_C) || defined(MODULE) + +#include <scsi/scsicam.h> + +int sym53c8xx_abort(Scsi_Cmnd *); +int sym53c8xx_detect(Scsi_Host_Template *tpnt); +const char *sym53c8xx_info(struct Scsi_Host *host); +int sym53c8xx_queue_command(Scsi_Cmnd *, void (*done)(Scsi_Cmnd *)); +int sym53c8xx_reset(Scsi_Cmnd *, unsigned int); + +#ifdef MODULE +int sym53c8xx_release(struct Scsi_Host *); +#else +#define sym53c8xx_release NULL +#endif + + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,75) + +#define SYM53C8XX { name: "", \ + detect: sym53c8xx_detect, \ + release: sym53c8xx_release, \ + info: sym53c8xx_info, \ + queuecommand: sym53c8xx_queue_command,\ + abort: sym53c8xx_abort, \ + reset: sym53c8xx_reset, \ + bios_param: scsicam_bios_param, \ + can_queue: SCSI_NCR_CAN_QUEUE, \ + this_id: 7, \ + sg_tablesize: SCSI_NCR_SG_TABLESIZE, \ + cmd_per_lun: SCSI_NCR_CMD_PER_LUN, \ + use_clustering: DISABLE_CLUSTERING} + +#else + +#define SYM53C8XX { NULL, NULL, NULL, NULL, \ + NULL, sym53c8xx_detect, \ + sym53c8xx_release, sym53c8xx_info, NULL, \ + sym53c8xx_queue_command,sym53c8xx_abort, \ + sym53c8xx_reset, NULL, scsicam_bios_param, \ + SCSI_NCR_CAN_QUEUE, 7, \ + SCSI_NCR_SG_TABLESIZE, SCSI_NCR_CMD_PER_LUN, \ + 0, 0, DISABLE_CLUSTERING} + +#endif /* LINUX_VERSION_CODE */ + +#endif /* defined(HOSTS_C) || defined(MODULE) */ + +#endif /* SYM53C8XX_H */ diff --git a/linux/src/drivers/scsi/sym53c8xx_comm.h b/linux/src/drivers/scsi/sym53c8xx_comm.h new file mode 100644 index 00000000..ba961db0 --- /dev/null +++ b/linux/src/drivers/scsi/sym53c8xx_comm.h @@ -0,0 +1,2717 @@ +/****************************************************************************** +** High Performance device driver for the Symbios 53C896 controller. +** +** Copyright (C) 1998-2000 Gerard Roudier <groudier@club-internet.fr> +** +** This driver also supports all the Symbios 53C8XX controller family, +** except 53C810 revisions < 16, 53C825 revisions < 16 and all +** revisions of 53C815 controllers. +** +** This driver is based on the Linux port of the FreeBSD ncr driver. +** +** Copyright (C) 1994 Wolfgang Stanglmeier +** +**----------------------------------------------------------------------------- +** +** This program is free software; you can redistribute it and/or modify +** it under the terms of the GNU General Public License as published by +** the Free Software Foundation; either version 2 of the License, or +** (at your option) any later version. +** +** This program is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU General Public License for more details. +** +** You should have received a copy of the GNU General Public License +** along with this program; if not, write to the Free Software +** Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. +** +**----------------------------------------------------------------------------- +** +** The Linux port of the FreeBSD ncr driver has been achieved in +** november 1995 by: +** +** Gerard Roudier <groudier@club-internet.fr> +** +** Being given that this driver originates from the FreeBSD version, and +** in order to keep synergy on both, any suggested enhancements and corrections +** received on Linux are automatically a potential candidate for the FreeBSD +** version. +** +** The original driver has been written for 386bsd and FreeBSD by +** Wolfgang Stanglmeier <wolf@cologne.de> +** Stefan Esser <se@mi.Uni-Koeln.de> +** +**----------------------------------------------------------------------------- +** +** Major contributions: +** -------------------- +** +** NVRAM detection and reading. +** Copyright (C) 1997 Richard Waltham <dormouse@farsrobt.demon.co.uk> +** +******************************************************************************* +*/ + +/* +** This file contains definitions and code that the +** sym53c8xx and ncr53c8xx drivers should share. +** The sharing will be achieved in a further version +** of the driver bundle. For now, only the ncr53c8xx +** driver includes this file. +*/ + +#define MIN(a,b) (((a) < (b)) ? (a) : (b)) +#define MAX(a,b) (((a) > (b)) ? (a) : (b)) + +/*========================================================== +** +** Hmmm... What complex some PCI-HOST bridges actually +** are, despite the fact that the PCI specifications +** are looking so smart and simple! ;-) +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,3,47) +#define SCSI_NCR_DYNAMIC_DMA_MAPPING +#endif + +/*========================================================== +** +** Miscallaneous defines. +** +**========================================================== +*/ + +#define u_char unsigned char +#define u_short unsigned short +#define u_int unsigned int +#define u_long unsigned long + +#ifndef bcopy +#define bcopy(s, d, n) memcpy((d), (s), (n)) +#endif + +#ifndef bcmp +#define bcmp(s, d, n) memcmp((d), (s), (n)) +#endif + +#ifndef bzero +#define bzero(d, n) memset((d), 0, (n)) +#endif + +#ifndef offsetof +#define offsetof(t, m) ((size_t) (&((t *)0)->m)) +#endif + +/*========================================================== +** +** assert () +** +**========================================================== +** +** modified copy from 386bsd:/usr/include/sys/assert.h +** +**---------------------------------------------------------- +*/ + +#define assert(expression) { \ + if (!(expression)) { \ + (void)panic( \ + "assertion \"%s\" failed: file \"%s\", line %d\n", \ + #expression, \ + __FILE__, __LINE__); \ + } \ +} + +/*========================================================== +** +** Debugging tags +** +**========================================================== +*/ + +#define DEBUG_ALLOC (0x0001) +#define DEBUG_PHASE (0x0002) +#define DEBUG_QUEUE (0x0008) +#define DEBUG_RESULT (0x0010) +#define DEBUG_POINTER (0x0020) +#define DEBUG_SCRIPT (0x0040) +#define DEBUG_TINY (0x0080) +#define DEBUG_TIMING (0x0100) +#define DEBUG_NEGO (0x0200) +#define DEBUG_TAGS (0x0400) +#define DEBUG_SCATTER (0x0800) +#define DEBUG_IC (0x1000) + +/* +** Enable/Disable debug messages. +** Can be changed at runtime too. +*/ + +#ifdef SCSI_NCR_DEBUG_INFO_SUPPORT +static int ncr_debug = SCSI_NCR_DEBUG_FLAGS; + #define DEBUG_FLAGS ncr_debug +#else + #define DEBUG_FLAGS SCSI_NCR_DEBUG_FLAGS +#endif + +/*========================================================== +** +** A la VMS/CAM-3 queue management. +** Implemented from linux list management. +** +**========================================================== +*/ + +typedef struct xpt_quehead { + struct xpt_quehead *flink; /* Forward pointer */ + struct xpt_quehead *blink; /* Backward pointer */ +} XPT_QUEHEAD; + +#define xpt_que_init(ptr) do { \ + (ptr)->flink = (ptr); (ptr)->blink = (ptr); \ +} while (0) + +static inline void __xpt_que_add(struct xpt_quehead * new, + struct xpt_quehead * blink, + struct xpt_quehead * flink) +{ + flink->blink = new; + new->flink = flink; + new->blink = blink; + blink->flink = new; +} + +static inline void __xpt_que_del(struct xpt_quehead * blink, + struct xpt_quehead * flink) +{ + flink->blink = blink; + blink->flink = flink; +} + +static inline int xpt_que_empty(struct xpt_quehead *head) +{ + return head->flink == head; +} + +static inline void xpt_que_splice(struct xpt_quehead *list, + struct xpt_quehead *head) +{ + struct xpt_quehead *first = list->flink; + + if (first != list) { + struct xpt_quehead *last = list->blink; + struct xpt_quehead *at = head->flink; + + first->blink = head; + head->flink = first; + + last->flink = at; + at->blink = last; + } +} + +#define xpt_que_entry(ptr, type, member) \ + ((type *)((char *)(ptr)-(unsigned long)(&((type *)0)->member))) + + +#define xpt_insque(new, pos) __xpt_que_add(new, pos, (pos)->flink) + +#define xpt_remque(el) __xpt_que_del((el)->blink, (el)->flink) + +#define xpt_insque_head(new, head) __xpt_que_add(new, head, (head)->flink) + +static inline struct xpt_quehead *xpt_remque_head(struct xpt_quehead *head) +{ + struct xpt_quehead *elem = head->flink; + + if (elem != head) + __xpt_que_del(head, elem->flink); + else + elem = 0; + return elem; +} + +#define xpt_insque_tail(new, head) __xpt_que_add(new, (head)->blink, head) + +static inline struct xpt_quehead *xpt_remque_tail(struct xpt_quehead *head) +{ + struct xpt_quehead *elem = head->blink; + + if (elem != head) + __xpt_que_del(elem->blink, head); + else + elem = 0; + return elem; +} + +/*========================================================== +** +** Simple Wrapper to kernel PCI bus interface. +** +** This wrapper allows to get rid of old kernel PCI +** interface and still allows to preserve linux-2.0 +** compatibilty. In fact, it is mostly an incomplete +** emulation of the new PCI code for pre-2.2 kernels. +** When kernel-2.0 support will be dropped, we will +** just have to remove most of this code. +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,2,0) + +typedef struct pci_dev *pcidev_t; +#define PCIDEV_NULL (0) +#define PciBusNumber(d) (d)->bus->number +#define PciDeviceFn(d) (d)->devfn +#define PciVendorId(d) (d)->vendor +#define PciDeviceId(d) (d)->device +#define PciIrqLine(d) (d)->irq + +#if LINUX_VERSION_CODE > LinuxVersionCode(2,3,12) + +static int __init +pci_get_base_address(struct pci_dev *pdev, int index, u_long *base) +{ + *base = pdev->resource[index].start; + if ((pdev->resource[index].flags & 0x7) == 0x4) + ++index; + return ++index; +} +#else +static int __init +pci_get_base_address(struct pci_dev *pdev, int index, u_long *base) +{ + *base = pdev->base_address[index++]; + if ((*base & 0x7) == 0x4) { +#if BITS_PER_LONG > 32 + *base |= (((u_long)pdev->base_address[index]) << 32); +#endif + ++index; + } + return index; +} +#endif + +#else /* Incomplete emulation of current PCI code for pre-2.2 kernels */ + +typedef unsigned int pcidev_t; +#define PCIDEV_NULL (~0u) +#define PciBusNumber(d) ((d)>>8) +#define PciDeviceFn(d) ((d)&0xff) +#define __PciDev(busn, devfn) (((busn)<<8)+(devfn)) + +#define pci_present pcibios_present + +#define pci_read_config_byte(d, w, v) \ + pcibios_read_config_byte(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_read_config_word(d, w, v) \ + pcibios_read_config_word(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_read_config_dword(d, w, v) \ + pcibios_read_config_dword(PciBusNumber(d), PciDeviceFn(d), w, v) + +#define pci_write_config_byte(d, w, v) \ + pcibios_write_config_byte(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_write_config_word(d, w, v) \ + pcibios_write_config_word(PciBusNumber(d), PciDeviceFn(d), w, v) +#define pci_write_config_dword(d, w, v) \ + pcibios_write_config_dword(PciBusNumber(d), PciDeviceFn(d), w, v) + +static pcidev_t __init +pci_find_device(unsigned int vendor, unsigned int device, pcidev_t prev) +{ + static unsigned short pci_index; + int retv; + unsigned char bus_number, device_fn; + + if (prev == PCIDEV_NULL) + pci_index = 0; + else + ++pci_index; + retv = pcibios_find_device (vendor, device, pci_index, + &bus_number, &device_fn); + return retv ? PCIDEV_NULL : __PciDev(bus_number, device_fn); +} + +static u_short __init PciVendorId(pcidev_t dev) +{ + u_short vendor_id; + pci_read_config_word(dev, PCI_VENDOR_ID, &vendor_id); + return vendor_id; +} + +static u_short __init PciDeviceId(pcidev_t dev) +{ + u_short device_id; + pci_read_config_word(dev, PCI_DEVICE_ID, &device_id); + return device_id; +} + +static u_int __init PciIrqLine(pcidev_t dev) +{ + u_char irq; + pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &irq); + return irq; +} + +static int __init +pci_get_base_address(pcidev_t dev, int offset, u_long *base) +{ + u_int32 tmp; + + pci_read_config_dword(dev, PCI_BASE_ADDRESS_0 + offset, &tmp); + *base = tmp; + offset += sizeof(u_int32); + if ((tmp & 0x7) == 0x4) { +#if BITS_PER_LONG > 32 + pci_read_config_dword(dev, PCI_BASE_ADDRESS_0 + offset, &tmp); + *base |= (((u_long)tmp) << 32); +#endif + offset += sizeof(u_int32); + } + return offset; +} + +#endif /* LINUX_VERSION_CODE >= LinuxVersionCode(2,2,0) */ + +/*========================================================== +** +** SMP threading. +** +** Assuming that SMP systems are generally high end +** systems and may use several SCSI adapters, we are +** using one lock per controller instead of some global +** one. For the moment (linux-2.1.95), driver's entry +** points are called with the 'io_request_lock' lock +** held, so: +** - We are uselessly loosing a couple of micro-seconds +** to lock the controller data structure. +** - But the driver is not broken by design for SMP and +** so can be more resistant to bugs or bad changes in +** the IO sub-system code. +** - A small advantage could be that the interrupt code +** is grained as wished (e.g.: by controller). +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,93) +spinlock_t DRIVER_SMP_LOCK = SPIN_LOCK_UNLOCKED; +#define NCR_LOCK_DRIVER(flags) spin_lock_irqsave(&DRIVER_SMP_LOCK, flags) +#define NCR_UNLOCK_DRIVER(flags) \ + spin_unlock_irqrestore(&DRIVER_SMP_LOCK, flags) + +#define NCR_INIT_LOCK_NCB(np) spin_lock_init(&np->smp_lock) +#define NCR_LOCK_NCB(np, flags) spin_lock_irqsave(&np->smp_lock, flags) +#define NCR_UNLOCK_NCB(np, flags) spin_unlock_irqrestore(&np->smp_lock, flags) + +#define NCR_LOCK_SCSI_DONE(np, flags) \ + spin_lock_irqsave(&io_request_lock, flags) +#define NCR_UNLOCK_SCSI_DONE(np, flags) \ + spin_unlock_irqrestore(&io_request_lock, flags) + +#else + +#define NCR_LOCK_DRIVER(flags) do { save_flags(flags); cli(); } while (0) +#define NCR_UNLOCK_DRIVER(flags) do { restore_flags(flags); } while (0) + +#define NCR_INIT_LOCK_NCB(np) do { } while (0) +#define NCR_LOCK_NCB(np, flags) do { save_flags(flags); cli(); } while (0) +#define NCR_UNLOCK_NCB(np, flags) do { restore_flags(flags); } while (0) + +#define NCR_LOCK_SCSI_DONE(np, flags) do {;} while (0) +#define NCR_UNLOCK_SCSI_DONE(np, flags) do {;} while (0) + +#endif + +/*========================================================== +** +** Memory mapped IO +** +** Since linux-2.1, we must use ioremap() to map the io +** memory space and iounmap() to unmap it. This allows +** portability. Linux 1.3.X and 2.0.X allow to remap +** physical pages addresses greater than the highest +** physical memory address to kernel virtual pages with +** vremap() / vfree(). That was not portable but worked +** with i386 architecture. +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE < LinuxVersionCode(2,1,0) +#define ioremap vremap +#define iounmap vfree +#endif + +#ifdef __sparc__ +# include <asm/irq.h> +# define pcivtobus(p) bus_dvma_to_mem(p) +# define memcpy_to_pci(a, b, c) memcpy_toio((a), (b), (c)) +#elif defined(__alpha__) +# define pcivtobus(p) ((p) & 0xfffffffful) +# define memcpy_to_pci(a, b, c) memcpy_toio((a), (b), (c)) +#else /* others */ +# define pcivtobus(p) (p) +# define memcpy_to_pci(a, b, c) memcpy_toio((a), (b), (c)) +#endif + +#ifndef SCSI_NCR_PCI_MEM_NOT_SUPPORTED +static u_long __init remap_pci_mem(u_long base, u_long size) +{ + u_long page_base = ((u_long) base) & PAGE_MASK; + u_long page_offs = ((u_long) base) - page_base; + u_long page_remapped = (u_long) ioremap(page_base, page_offs+size); + + return page_remapped? (page_remapped + page_offs) : 0UL; +} + +static void __init unmap_pci_mem(u_long vaddr, u_long size) +{ + if (vaddr) + iounmap((void *) (vaddr & PAGE_MASK)); +} + +#endif /* not def SCSI_NCR_PCI_MEM_NOT_SUPPORTED */ + +/*========================================================== +** +** Insert a delay in micro-seconds and milli-seconds. +** +** Under Linux, udelay() is restricted to delay < +** 1 milli-second. In fact, it generally works for up +** to 1 second delay. Since 2.1.105, the mdelay() function +** is provided for delays in milli-seconds. +** Under 2.0 kernels, udelay() is an inline function +** that is very inaccurate on Pentium processors. +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,105) +#define UDELAY udelay +#define MDELAY mdelay +#else +static void UDELAY(long us) { udelay(us); } +static void MDELAY(long ms) { while (ms--) UDELAY(1000); } +#endif + +/*========================================================== +** +** Simple power of two buddy-like allocator. +** +** This simple code is not intended to be fast, but to +** provide power of 2 aligned memory allocations. +** Since the SCRIPTS processor only supplies 8 bit +** arithmetic, this allocator allows simple and fast +** address calculations from the SCRIPTS code. +** In addition, cache line alignment is guaranteed for +** power of 2 cache line size. +** Enhanced in linux-2.3.44 to provide a memory pool +** per pcidev to support dynamic dma mapping. (I would +** have preferred a real bus astraction, btw). +** +**========================================================== +*/ + +#if LINUX_VERSION_CODE >= LinuxVersionCode(2,1,0) +#define __GetFreePages(flags, order) __get_free_pages(flags, order) +#else +#define __GetFreePages(flags, order) __get_free_pages(flags, order, 0) +#endif + +#define MEMO_SHIFT 4 /* 16 bytes minimum memory chunk */ +#if PAGE_SIZE >= 8192 +#define MEMO_PAGE_ORDER 0 /* 1 PAGE maximum */ +#else +#define MEMO_PAGE_ORDER 1 /* 2 PAGES maximum */ +#endif +#define MEMO_FREE_UNUSED /* Free unused pages immediately */ +#define MEMO_WARN 1 +#define MEMO_GFP_FLAGS GFP_ATOMIC +#define MEMO_CLUSTER_SHIFT (PAGE_SHIFT+MEMO_PAGE_ORDER) +#define MEMO_CLUSTER_SIZE (1UL << MEMO_CLUSTER_SHIFT) +#define MEMO_CLUSTER_MASK (MEMO_CLUSTER_SIZE-1) + +typedef u_long m_addr_t; /* Enough bits to bit-hack addresses */ +typedef pcidev_t m_bush_t; /* Something that addresses DMAable */ + +typedef struct m_link { /* Link between free memory chunks */ + struct m_link *next; +} m_link_s; + +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING +typedef struct m_vtob { /* Virtual to Bus address translation */ + struct m_vtob *next; + m_addr_t vaddr; + m_addr_t baddr; +} m_vtob_s; +#define VTOB_HASH_SHIFT 5 +#define VTOB_HASH_SIZE (1UL << VTOB_HASH_SHIFT) +#define VTOB_HASH_MASK (VTOB_HASH_SIZE-1) +#define VTOB_HASH_CODE(m) \ + ((((m_addr_t) (m)) >> MEMO_CLUSTER_SHIFT) & VTOB_HASH_MASK) +#endif + +typedef struct m_pool { /* Memory pool of a given kind */ +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING + m_bush_t bush; + m_addr_t (*getp)(struct m_pool *); + void (*freep)(struct m_pool *, m_addr_t); +#define M_GETP() mp->getp(mp) +#define M_FREEP(p) mp->freep(mp, p) +#define GetPages() __GetFreePages(MEMO_GFP_FLAGS, MEMO_PAGE_ORDER) +#define FreePages(p) free_pages(p, MEMO_PAGE_ORDER) + int nump; + m_vtob_s *(vtob[VTOB_HASH_SIZE]); + struct m_pool *next; +#else +#define M_GETP() __GetFreePages(MEMO_GFP_FLAGS, MEMO_PAGE_ORDER) +#define M_FREEP(p) free_pages(p, MEMO_PAGE_ORDER) +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + struct m_link h[PAGE_SHIFT-MEMO_SHIFT+MEMO_PAGE_ORDER+1]; +} m_pool_s; + +static void *___m_alloc(m_pool_s *mp, int size) +{ + int i = 0; + int s = (1 << MEMO_SHIFT); + int j; + m_addr_t a; + m_link_s *h = mp->h; + + if (size > (PAGE_SIZE << MEMO_PAGE_ORDER)) + return 0; + + while (size > s) { + s <<= 1; + ++i; + } + + j = i; + while (!h[j].next) { + if (s == (PAGE_SIZE << MEMO_PAGE_ORDER)) { + h[j].next = (m_link_s *) M_GETP(); + if (h[j].next) + h[j].next->next = 0; + break; + } + ++j; + s <<= 1; + } + a = (m_addr_t) h[j].next; + if (a) { + h[j].next = h[j].next->next; + while (j > i) { + j -= 1; + s >>= 1; + h[j].next = (m_link_s *) (a+s); + h[j].next->next = 0; + } + } +#ifdef DEBUG + printk("___m_alloc(%d) = %p\n", size, (void *) a); +#endif + return (void *) a; +} + +static void ___m_free(m_pool_s *mp, void *ptr, int size) +{ + int i = 0; + int s = (1 << MEMO_SHIFT); + m_link_s *q; + m_addr_t a, b; + m_link_s *h = mp->h; + +#ifdef DEBUG + printk("___m_free(%p, %d)\n", ptr, size); +#endif + + if (size > (PAGE_SIZE << MEMO_PAGE_ORDER)) + return; + + while (size > s) { + s <<= 1; + ++i; + } + + a = (m_addr_t) ptr; + + while (1) { +#ifdef MEMO_FREE_UNUSED + if (s == (PAGE_SIZE << MEMO_PAGE_ORDER)) { + M_FREEP(a); + break; + } +#endif + b = a ^ s; + q = &h[i]; + while (q->next && q->next != (m_link_s *) b) { + q = q->next; + } + if (!q->next) { + ((m_link_s *) a)->next = h[i].next; + h[i].next = (m_link_s *) a; + break; + } + q->next = q->next->next; + a = a & b; + s <<= 1; + ++i; + } +} + +static void *__m_calloc2(m_pool_s *mp, int size, char *name, int uflags) +{ + void *p; + + p = ___m_alloc(mp, size); + + if (DEBUG_FLAGS & DEBUG_ALLOC) + printk ("new %-10s[%4d] @%p.\n", name, size, p); + + if (p) + bzero(p, size); + else if (uflags & MEMO_WARN) + printk (NAME53C8XX ": failed to allocate %s[%d]\n", name, size); + + return p; +} + +#define __m_calloc(mp, s, n) __m_calloc2(mp, s, n, MEMO_WARN) + +static void __m_free(m_pool_s *mp, void *ptr, int size, char *name) +{ + if (DEBUG_FLAGS & DEBUG_ALLOC) + printk ("freeing %-10s[%4d] @%p.\n", name, size, ptr); + + ___m_free(mp, ptr, size); + +} + +/* + * With pci bus iommu support, we use a default pool of unmapped memory + * for memory we donnot need to DMA from/to and one pool per pcidev for + * memory accessed by the PCI chip. `mp0' is the default not DMAable pool. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +static m_pool_s mp0; + +#else + +static m_addr_t ___mp0_getp(m_pool_s *mp) +{ + m_addr_t m = GetPages(); + if (m) + ++mp->nump; + return m; +} + +static void ___mp0_freep(m_pool_s *mp, m_addr_t m) +{ + FreePages(m); + --mp->nump; +} + +static m_pool_s mp0 = {0, ___mp0_getp, ___mp0_freep}; + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +static void *m_calloc(int size, char *name) +{ + u_long flags; + void *m; + NCR_LOCK_DRIVER(flags); + m = __m_calloc(&mp0, size, name); + NCR_UNLOCK_DRIVER(flags); + return m; +} + +static void m_free(void *ptr, int size, char *name) +{ + u_long flags; + NCR_LOCK_DRIVER(flags); + __m_free(&mp0, ptr, size, name); + NCR_UNLOCK_DRIVER(flags); +} + +/* + * DMAable pools. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +/* Without pci bus iommu support, all the memory is assumed DMAable */ + +#define __m_calloc_dma(b, s, n) m_calloc(s, n) +#define __m_free_dma(b, p, s, n) m_free(p, s, n) +#define __vtobus(b, p) virt_to_bus(p) + +#else + +/* + * With pci bus iommu support, we maintain one pool per pcidev and a + * hashed reverse table for virtual to bus physical address translations. + */ +static m_addr_t ___dma_getp(m_pool_s *mp) +{ + m_addr_t vp; + m_vtob_s *vbp; + + vbp = __m_calloc(&mp0, sizeof(*vbp), "VTOB"); + if (vbp) { + dma_addr_t daddr; + vp = (m_addr_t) pci_alloc_consistent(mp->bush, + PAGE_SIZE<<MEMO_PAGE_ORDER, + &daddr); + if (vp) { + int hc = VTOB_HASH_CODE(vp); + vbp->vaddr = vp; + vbp->baddr = daddr; + vbp->next = mp->vtob[hc]; + mp->vtob[hc] = vbp; + ++mp->nump; + return vp; + } + } + if (vbp) + __m_free(&mp0, vbp, sizeof(*vbp), "VTOB"); + return 0; +} + +static void ___dma_freep(m_pool_s *mp, m_addr_t m) +{ + m_vtob_s **vbpp, *vbp; + int hc = VTOB_HASH_CODE(m); + + vbpp = &mp->vtob[hc]; + while (*vbpp && (*vbpp)->vaddr != m) + vbpp = &(*vbpp)->next; + if (*vbpp) { + vbp = *vbpp; + *vbpp = (*vbpp)->next; + pci_free_consistent(mp->bush, PAGE_SIZE<<MEMO_PAGE_ORDER, + (void *)vbp->vaddr, (dma_addr_t)vbp->baddr); + __m_free(&mp0, vbp, sizeof(*vbp), "VTOB"); + --mp->nump; + } +} + +static inline m_pool_s *___get_dma_pool(m_bush_t bush) +{ + m_pool_s *mp; + for (mp = mp0.next; mp && mp->bush != bush; mp = mp->next); + return mp; +} + +static m_pool_s *___cre_dma_pool(m_bush_t bush) +{ + m_pool_s *mp; + mp = __m_calloc(&mp0, sizeof(*mp), "MPOOL"); + if (mp) { + bzero(mp, sizeof(*mp)); + mp->bush = bush; + mp->getp = ___dma_getp; + mp->freep = ___dma_freep; + mp->next = mp0.next; + mp0.next = mp; + } + return mp; +} + +static void ___del_dma_pool(m_pool_s *p) +{ + struct m_pool **pp = &mp0.next; + + while (*pp && *pp != p) + pp = &(*pp)->next; + if (*pp) { + *pp = (*pp)->next; + __m_free(&mp0, p, sizeof(*p), "MPOOL"); + } +} + +static void *__m_calloc_dma(m_bush_t bush, int size, char *name) +{ + u_long flags; + struct m_pool *mp; + void *m = 0; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (!mp) + mp = ___cre_dma_pool(bush); + if (mp) + m = __m_calloc(mp, size, name); + if (mp && !mp->nump) + ___del_dma_pool(mp); + NCR_UNLOCK_DRIVER(flags); + + return m; +} + +static void __m_free_dma(m_bush_t bush, void *m, int size, char *name) +{ + u_long flags; + struct m_pool *mp; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (mp) + __m_free(mp, m, size, name); + if (mp && !mp->nump) + ___del_dma_pool(mp); + NCR_UNLOCK_DRIVER(flags); +} + +static m_addr_t __vtobus(m_bush_t bush, void *m) +{ + u_long flags; + m_pool_s *mp; + int hc = VTOB_HASH_CODE(m); + m_vtob_s *vp = 0; + m_addr_t a = ((m_addr_t) m) & ~MEMO_CLUSTER_MASK; + + NCR_LOCK_DRIVER(flags); + mp = ___get_dma_pool(bush); + if (mp) { + vp = mp->vtob[hc]; + while (vp && (m_addr_t) vp->vaddr != a) + vp = vp->next; + } + NCR_UNLOCK_DRIVER(flags); + return vp ? vp->baddr + (((m_addr_t) m) - a) : 0; +} + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +#define _m_calloc_dma(np, s, n) __m_calloc_dma(np->pdev, s, n) +#define _m_free_dma(np, p, s, n) __m_free_dma(np->pdev, p, s, n) +#define m_calloc_dma(s, n) _m_calloc_dma(np, s, n) +#define m_free_dma(p, s, n) _m_free_dma(np, p, s, n) +#define _vtobus(np, p) __vtobus(np->pdev, p) +#define vtobus(p) _vtobus(np, p) + +/* + * Deal with DMA mapping/unmapping. + */ + +#ifndef SCSI_NCR_DYNAMIC_DMA_MAPPING + +/* Linux versions prior to pci bus iommu kernel interface */ + +#define __unmap_scsi_data(pdev, cmd) do {; } while (0) +#define __map_scsi_single_data(pdev, cmd) (__vtobus(pdev,(cmd)->request_buffer)) +#define __map_scsi_sg_data(pdev, cmd) ((cmd)->use_sg) +#define __sync_scsi_data(pdev, cmd) do {; } while (0) + +#define scsi_sg_dma_address(sc) vtobus((sc)->address) +#define scsi_sg_dma_len(sc) ((sc)->length) + +#else + +/* Linux version with pci bus iommu kernel interface */ + +/* To keep track of the dma mapping (sg/single) that has been set */ +#define __data_mapped SCp.phase +#define __data_mapping SCp.have_data_in + +static void __unmap_scsi_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + switch(cmd->__data_mapped) { + case 2: + pci_unmap_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + break; + case 1: + pci_unmap_single(pdev, cmd->__data_mapping, + cmd->request_bufflen, dma_dir); + break; + } + cmd->__data_mapped = 0; +} + +static u_long __map_scsi_single_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + dma_addr_t mapping; + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + if (cmd->request_bufflen == 0) + return 0; + + mapping = pci_map_single(pdev, cmd->request_buffer, + cmd->request_bufflen, dma_dir); + cmd->__data_mapped = 1; + cmd->__data_mapping = mapping; + + return mapping; +} + +static int __map_scsi_sg_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int use_sg; + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + if (cmd->use_sg == 0) + return 0; + + use_sg = pci_map_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + cmd->__data_mapped = 2; + cmd->__data_mapping = use_sg; + + return use_sg; +} + +static void __sync_scsi_data(pcidev_t pdev, Scsi_Cmnd *cmd) +{ + int dma_dir = scsi_to_pci_dma_dir(cmd->sc_data_direction); + + switch(cmd->__data_mapped) { + case 2: + pci_dma_sync_sg(pdev, cmd->buffer, cmd->use_sg, dma_dir); + break; + case 1: + pci_dma_sync_single(pdev, cmd->__data_mapping, + cmd->request_bufflen, dma_dir); + break; + } +} + +#define scsi_sg_dma_address(sc) sg_dma_address(sc) +#define scsi_sg_dma_len(sc) sg_dma_len(sc) + +#endif /* SCSI_NCR_DYNAMIC_DMA_MAPPING */ + +#define unmap_scsi_data(np, cmd) __unmap_scsi_data(np->pdev, cmd) +#define map_scsi_single_data(np, cmd) __map_scsi_single_data(np->pdev, cmd) +#define map_scsi_sg_data(np, cmd) __map_scsi_sg_data(np->pdev, cmd) +#define sync_scsi_data(np, cmd) __sync_scsi_data(np->pdev, cmd) + +/*========================================================== +** +** SCSI data transfer direction +** +** Until some linux kernel version near 2.3.40, +** low-level scsi drivers were not told about data +** transfer direction. We check the existence of this +** feature that has been expected for a _long_ time by +** all SCSI driver developers by just testing against +** the definition of SCSI_DATA_UNKNOWN. Indeed this is +** a hack, but testing against a kernel version would +** have been a shame. ;-) +** +**========================================================== +*/ +#ifdef SCSI_DATA_UNKNOWN + +#define scsi_data_direction(cmd) (cmd->sc_data_direction) + +#else + +#define SCSI_DATA_UNKNOWN 0 +#define SCSI_DATA_WRITE 1 +#define SCSI_DATA_READ 2 +#define SCSI_DATA_NONE 3 + +static __inline__ int scsi_data_direction(Scsi_Cmnd *cmd) +{ + int direction; + + switch((int) cmd->cmnd[0]) { + case 0x08: /* READ(6) 08 */ + case 0x28: /* READ(10) 28 */ + case 0xA8: /* READ(12) A8 */ + direction = SCSI_DATA_READ; + break; + case 0x0A: /* WRITE(6) 0A */ + case 0x2A: /* WRITE(10) 2A */ + case 0xAA: /* WRITE(12) AA */ + direction = SCSI_DATA_WRITE; + break; + default: + direction = SCSI_DATA_UNKNOWN; + break; + } + + return direction; +} + +#endif /* SCSI_DATA_UNKNOWN */ + +/*========================================================== +** +** Driver setup. +** +** This structure is initialized from linux config +** options. It can be overridden at boot-up by the boot +** command line. +** +**========================================================== +*/ +static struct ncr_driver_setup + driver_setup = SCSI_NCR_DRIVER_SETUP; + +#ifdef SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT +static struct ncr_driver_setup + driver_safe_setup __initdata = SCSI_NCR_DRIVER_SAFE_SETUP; +#endif + +#define initverbose (driver_setup.verbose) +#define bootverbose (np->verbose) + + +/*========================================================== +** +** Structures used by the detection routine to transmit +** device configuration to the attach function. +** +**========================================================== +*/ +typedef struct { + int bus; + u_char device_fn; + u_long base; + u_long base_2; + u_long io_port; + int irq; +/* port and reg fields to use INB, OUTB macros */ + u_long base_io; + volatile struct ncr_reg *reg; +} ncr_slot; + +/*========================================================== +** +** Structure used to store the NVRAM content. +** +**========================================================== +*/ +typedef struct { + int type; +#define SCSI_NCR_SYMBIOS_NVRAM (1) +#define SCSI_NCR_TEKRAM_NVRAM (2) +#ifdef SCSI_NCR_NVRAM_SUPPORT + union { + Symbios_nvram Symbios; + Tekram_nvram Tekram; + } data; +#endif +} ncr_nvram; + +/*========================================================== +** +** Structure used by detection routine to save data on +** each detected board for attach. +** +**========================================================== +*/ +typedef struct { + pcidev_t pdev; + ncr_slot slot; + ncr_chip chip; + ncr_nvram *nvram; + u_char host_id; +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + u_char pqs_pds; +#endif + int attach_done; +} ncr_device; + +static int ncr_attach (Scsi_Host_Template *tpnt, int unit, ncr_device *device); + +/*========================================================== +** +** NVRAM detection and reading. +** +** Currently supported: +** - 24C16 EEPROM with both Symbios and Tekram layout. +** - 93C46 EEPROM with Tekram layout. +** +**========================================================== +*/ + +#ifdef SCSI_NCR_NVRAM_SUPPORT +/* + * 24C16 EEPROM reading. + * + * GPOI0 - data in/data out + * GPIO1 - clock + * Symbios NVRAM wiring now also used by Tekram. + */ + +#define SET_BIT 0 +#define CLR_BIT 1 +#define SET_CLK 2 +#define CLR_CLK 3 + +/* + * Set/clear data/clock bit in GPIO0 + */ +static void __init +S24C16_set_bit(ncr_slot *np, u_char write_bit, u_char *gpreg, int bit_mode) +{ + UDELAY (5); + switch (bit_mode){ + case SET_BIT: + *gpreg |= write_bit; + break; + case CLR_BIT: + *gpreg &= 0xfe; + break; + case SET_CLK: + *gpreg |= 0x02; + break; + case CLR_CLK: + *gpreg &= 0xfd; + break; + + } + OUTB (nc_gpreg, *gpreg); + UDELAY (5); +} + +/* + * Send START condition to NVRAM to wake it up. + */ +static void __init S24C16_start(ncr_slot *np, u_char *gpreg) +{ + S24C16_set_bit(np, 1, gpreg, SET_BIT); + S24C16_set_bit(np, 0, gpreg, SET_CLK); + S24C16_set_bit(np, 0, gpreg, CLR_BIT); + S24C16_set_bit(np, 0, gpreg, CLR_CLK); +} + +/* + * Send STOP condition to NVRAM - puts NVRAM to sleep... ZZzzzz!! + */ +static void __init S24C16_stop(ncr_slot *np, u_char *gpreg) +{ + S24C16_set_bit(np, 0, gpreg, SET_CLK); + S24C16_set_bit(np, 1, gpreg, SET_BIT); +} + +/* + * Read or write a bit to the NVRAM, + * read if GPIO0 input else write if GPIO0 output + */ +static void __init +S24C16_do_bit(ncr_slot *np, u_char *read_bit, u_char write_bit, u_char *gpreg) +{ + S24C16_set_bit(np, write_bit, gpreg, SET_BIT); + S24C16_set_bit(np, 0, gpreg, SET_CLK); + if (read_bit) + *read_bit = INB (nc_gpreg); + S24C16_set_bit(np, 0, gpreg, CLR_CLK); + S24C16_set_bit(np, 0, gpreg, CLR_BIT); +} + +/* + * Output an ACK to the NVRAM after reading, + * change GPIO0 to output and when done back to an input + */ +static void __init +S24C16_write_ack(ncr_slot *np, u_char write_bit, u_char *gpreg, u_char *gpcntl) +{ + OUTB (nc_gpcntl, *gpcntl & 0xfe); + S24C16_do_bit(np, 0, write_bit, gpreg); + OUTB (nc_gpcntl, *gpcntl); +} + +/* + * Input an ACK from NVRAM after writing, + * change GPIO0 to input and when done back to an output + */ +static void __init +S24C16_read_ack(ncr_slot *np, u_char *read_bit, u_char *gpreg, u_char *gpcntl) +{ + OUTB (nc_gpcntl, *gpcntl | 0x01); + S24C16_do_bit(np, read_bit, 1, gpreg); + OUTB (nc_gpcntl, *gpcntl); +} + +/* + * WRITE a byte to the NVRAM and then get an ACK to see it was accepted OK, + * GPIO0 must already be set as an output + */ +static void __init +S24C16_write_byte(ncr_slot *np, u_char *ack_data, u_char write_data, + u_char *gpreg, u_char *gpcntl) +{ + int x; + + for (x = 0; x < 8; x++) + S24C16_do_bit(np, 0, (write_data >> (7 - x)) & 0x01, gpreg); + + S24C16_read_ack(np, ack_data, gpreg, gpcntl); +} + +/* + * READ a byte from the NVRAM and then send an ACK to say we have got it, + * GPIO0 must already be set as an input + */ +static void __init +S24C16_read_byte(ncr_slot *np, u_char *read_data, u_char ack_data, + u_char *gpreg, u_char *gpcntl) +{ + int x; + u_char read_bit; + + *read_data = 0; + for (x = 0; x < 8; x++) { + S24C16_do_bit(np, &read_bit, 1, gpreg); + *read_data |= ((read_bit & 0x01) << (7 - x)); + } + + S24C16_write_ack(np, ack_data, gpreg, gpcntl); +} + +/* + * Read 'len' bytes starting at 'offset'. + */ +static int __init +sym_read_S24C16_nvram (ncr_slot *np, int offset, u_char *data, int len) +{ + u_char gpcntl, gpreg; + u_char old_gpcntl, old_gpreg; + u_char ack_data; + int retv = 1; + int x; + + /* save current state of GPCNTL and GPREG */ + old_gpreg = INB (nc_gpreg); + old_gpcntl = INB (nc_gpcntl); + gpcntl = old_gpcntl & 0xfc; + + /* set up GPREG & GPCNTL to set GPIO0 and GPIO1 in to known state */ + OUTB (nc_gpreg, old_gpreg); + OUTB (nc_gpcntl, gpcntl); + + /* this is to set NVRAM into a known state with GPIO0/1 both low */ + gpreg = old_gpreg; + S24C16_set_bit(np, 0, &gpreg, CLR_CLK); + S24C16_set_bit(np, 0, &gpreg, CLR_BIT); + + /* now set NVRAM inactive with GPIO0/1 both high */ + S24C16_stop(np, &gpreg); + + /* activate NVRAM */ + S24C16_start(np, &gpreg); + + /* write device code and random address MSB */ + S24C16_write_byte(np, &ack_data, + 0xa0 | ((offset >> 7) & 0x0e), &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* write random address LSB */ + S24C16_write_byte(np, &ack_data, + offset & 0xff, &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* regenerate START state to set up for reading */ + S24C16_start(np, &gpreg); + + /* rewrite device code and address MSB with read bit set (lsb = 0x01) */ + S24C16_write_byte(np, &ack_data, + 0xa1 | ((offset >> 7) & 0x0e), &gpreg, &gpcntl); + if (ack_data & 0x01) + goto out; + + /* now set up GPIO0 for inputting data */ + gpcntl |= 0x01; + OUTB (nc_gpcntl, gpcntl); + + /* input all requested data - only part of total NVRAM */ + for (x = 0; x < len; x++) + S24C16_read_byte(np, &data[x], (x == (len-1)), &gpreg, &gpcntl); + + /* finally put NVRAM back in inactive mode */ + gpcntl &= 0xfe; + OUTB (nc_gpcntl, gpcntl); + S24C16_stop(np, &gpreg); + retv = 0; +out: + /* return GPIO0/1 to original states after having accessed NVRAM */ + OUTB (nc_gpcntl, old_gpcntl); + OUTB (nc_gpreg, old_gpreg); + + return retv; +} + +#undef SET_BIT 0 +#undef CLR_BIT 1 +#undef SET_CLK 2 +#undef CLR_CLK 3 + +/* + * Try reading Symbios NVRAM. + * Return 0 if OK. + */ +static int __init sym_read_Symbios_nvram (ncr_slot *np, Symbios_nvram *nvram) +{ + static u_char Symbios_trailer[6] = {0xfe, 0xfe, 0, 0, 0, 0}; + u_char *data = (u_char *) nvram; + int len = sizeof(*nvram); + u_short csum; + int x; + + /* probe the 24c16 and read the SYMBIOS 24c16 area */ + if (sym_read_S24C16_nvram (np, SYMBIOS_NVRAM_ADDRESS, data, len)) + return 1; + + /* check valid NVRAM signature, verify byte count and checksum */ + if (nvram->type != 0 || + memcmp(nvram->trailer, Symbios_trailer, 6) || + nvram->byte_count != len - 12) + return 1; + + /* verify checksum */ + for (x = 6, csum = 0; x < len - 6; x++) + csum += data[x]; + if (csum != nvram->checksum) + return 1; + + return 0; +} + +/* + * 93C46 EEPROM reading. + * + * GPOI0 - data in + * GPIO1 - data out + * GPIO2 - clock + * GPIO4 - chip select + * + * Used by Tekram. + */ + +/* + * Pulse clock bit in GPIO0 + */ +static void __init T93C46_Clk(ncr_slot *np, u_char *gpreg) +{ + OUTB (nc_gpreg, *gpreg | 0x04); + UDELAY (2); + OUTB (nc_gpreg, *gpreg); +} + +/* + * Read bit from NVRAM + */ +static void __init T93C46_Read_Bit(ncr_slot *np, u_char *read_bit, u_char *gpreg) +{ + UDELAY (2); + T93C46_Clk(np, gpreg); + *read_bit = INB (nc_gpreg); +} + +/* + * Write bit to GPIO0 + */ +static void __init T93C46_Write_Bit(ncr_slot *np, u_char write_bit, u_char *gpreg) +{ + if (write_bit & 0x01) + *gpreg |= 0x02; + else + *gpreg &= 0xfd; + + *gpreg |= 0x10; + + OUTB (nc_gpreg, *gpreg); + UDELAY (2); + + T93C46_Clk(np, gpreg); +} + +/* + * Send STOP condition to NVRAM - puts NVRAM to sleep... ZZZzzz!! + */ +static void __init T93C46_Stop(ncr_slot *np, u_char *gpreg) +{ + *gpreg &= 0xef; + OUTB (nc_gpreg, *gpreg); + UDELAY (2); + + T93C46_Clk(np, gpreg); +} + +/* + * Send read command and address to NVRAM + */ +static void __init +T93C46_Send_Command(ncr_slot *np, u_short write_data, + u_char *read_bit, u_char *gpreg) +{ + int x; + + /* send 9 bits, start bit (1), command (2), address (6) */ + for (x = 0; x < 9; x++) + T93C46_Write_Bit(np, (u_char) (write_data >> (8 - x)), gpreg); + + *read_bit = INB (nc_gpreg); +} + +/* + * READ 2 bytes from the NVRAM + */ +static void __init +T93C46_Read_Word(ncr_slot *np, u_short *nvram_data, u_char *gpreg) +{ + int x; + u_char read_bit; + + *nvram_data = 0; + for (x = 0; x < 16; x++) { + T93C46_Read_Bit(np, &read_bit, gpreg); + + if (read_bit & 0x01) + *nvram_data |= (0x01 << (15 - x)); + else + *nvram_data &= ~(0x01 << (15 - x)); + } +} + +/* + * Read Tekram NvRAM data. + */ +static int __init +T93C46_Read_Data(ncr_slot *np, u_short *data,int len,u_char *gpreg) +{ + u_char read_bit; + int x; + + for (x = 0; x < len; x++) { + + /* output read command and address */ + T93C46_Send_Command(np, 0x180 | x, &read_bit, gpreg); + if (read_bit & 0x01) + return 1; /* Bad */ + T93C46_Read_Word(np, &data[x], gpreg); + T93C46_Stop(np, gpreg); + } + + return 0; +} + +/* + * Try reading 93C46 Tekram NVRAM. + */ +static int __init +sym_read_T93C46_nvram (ncr_slot *np, Tekram_nvram *nvram) +{ + u_char gpcntl, gpreg; + u_char old_gpcntl, old_gpreg; + int retv = 1; + + /* save current state of GPCNTL and GPREG */ + old_gpreg = INB (nc_gpreg); + old_gpcntl = INB (nc_gpcntl); + + /* set up GPREG & GPCNTL to set GPIO0/1/2/4 in to known state, 0 in, + 1/2/4 out */ + gpreg = old_gpreg & 0xe9; + OUTB (nc_gpreg, gpreg); + gpcntl = (old_gpcntl & 0xe9) | 0x09; + OUTB (nc_gpcntl, gpcntl); + + /* input all of NVRAM, 64 words */ + retv = T93C46_Read_Data(np, (u_short *) nvram, + sizeof(*nvram) / sizeof(short), &gpreg); + + /* return GPIO0/1/2/4 to original states after having accessed NVRAM */ + OUTB (nc_gpcntl, old_gpcntl); + OUTB (nc_gpreg, old_gpreg); + + return retv; +} + +/* + * Try reading Tekram NVRAM. + * Return 0 if OK. + */ +static int __init +sym_read_Tekram_nvram (ncr_slot *np, u_short device_id, Tekram_nvram *nvram) +{ + u_char *data = (u_char *) nvram; + int len = sizeof(*nvram); + u_short csum; + int x; + + switch (device_id) { + case PCI_DEVICE_ID_NCR_53C885: + case PCI_DEVICE_ID_NCR_53C895: + case PCI_DEVICE_ID_NCR_53C896: + x = sym_read_S24C16_nvram(np, TEKRAM_24C16_NVRAM_ADDRESS, + data, len); + break; + case PCI_DEVICE_ID_NCR_53C875: + x = sym_read_S24C16_nvram(np, TEKRAM_24C16_NVRAM_ADDRESS, + data, len); + if (!x) + break; + default: + x = sym_read_T93C46_nvram(np, nvram); + break; + } + if (x) + return 1; + + /* verify checksum */ + for (x = 0, csum = 0; x < len - 1; x += 2) + csum += data[x] + (data[x+1] << 8); + if (csum != 0x1234) + return 1; + + return 0; +} + +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/*=================================================================== +** +** Detect and try to read SYMBIOS and TEKRAM NVRAM. +** +** Data can be used to order booting of boards. +** +** Data is saved in ncr_device structure if NVRAM found. This +** is then used to find drive boot order for ncr_attach(). +** +** NVRAM data is passed to Scsi_Host_Template later during +** ncr_attach() for any device set up. +** +**=================================================================== +*/ +#ifdef SCSI_NCR_NVRAM_SUPPORT +static void __init ncr_get_nvram(ncr_device *devp, ncr_nvram *nvp) +{ + devp->nvram = nvp; + if (!nvp) + return; + /* + ** Get access to chip IO registers + */ +#ifdef SCSI_NCR_IOMAPPED + request_region(devp->slot.io_port, 128, NAME53C8XX); + devp->slot.base_io = devp->slot.io_port; +#else + devp->slot.reg = (struct ncr_reg *) remap_pci_mem(devp->slot.base, 128); + if (!devp->slot.reg) + return; +#endif + + /* + ** Try to read SYMBIOS nvram. + ** Try to read TEKRAM nvram if Symbios nvram not found. + */ + if (!sym_read_Symbios_nvram(&devp->slot, &nvp->data.Symbios)) + nvp->type = SCSI_NCR_SYMBIOS_NVRAM; + else if (!sym_read_Tekram_nvram(&devp->slot, devp->chip.device_id, + &nvp->data.Tekram)) + nvp->type = SCSI_NCR_TEKRAM_NVRAM; + else { + nvp->type = 0; + devp->nvram = 0; + } + + /* + ** Release access to chip IO registers + */ +#ifdef SCSI_NCR_IOMAPPED + release_region(devp->slot.base_io, 128); +#else + unmap_pci_mem((u_long) devp->slot.reg, 128ul); +#endif + +} + +/*=================================================================== +** +** Display the content of NVRAM for debugging purpose. +** +**=================================================================== +*/ +#ifdef SCSI_NCR_DEBUG_NVRAM +static void __init ncr_display_Symbios_nvram(Symbios_nvram *nvram) +{ + int i; + + /* display Symbios nvram host data */ + printk(KERN_DEBUG NAME53C8XX ": HOST ID=%d%s%s%s%s%s\n", + nvram->host_id & 0x0f, + (nvram->flags & SYMBIOS_SCAM_ENABLE) ? " SCAM" :"", + (nvram->flags & SYMBIOS_PARITY_ENABLE) ? " PARITY" :"", + (nvram->flags & SYMBIOS_VERBOSE_MSGS) ? " VERBOSE" :"", + (nvram->flags & SYMBIOS_CHS_MAPPING) ? " CHS_ALT" :"", + (nvram->flags1 & SYMBIOS_SCAN_HI_LO) ? " HI_LO" :""); + + /* display Symbios nvram drive data */ + for (i = 0 ; i < 15 ; i++) { + struct Symbios_target *tn = &nvram->target[i]; + printk(KERN_DEBUG NAME53C8XX + "-%d:%s%s%s%s WIDTH=%d SYNC=%d TMO=%d\n", + i, + (tn->flags & SYMBIOS_DISCONNECT_ENABLE) ? " DISC" : "", + (tn->flags & SYMBIOS_SCAN_AT_BOOT_TIME) ? " SCAN_BOOT" : "", + (tn->flags & SYMBIOS_SCAN_LUNS) ? " SCAN_LUNS" : "", + (tn->flags & SYMBIOS_QUEUE_TAGS_ENABLED)? " TCQ" : "", + tn->bus_width, + tn->sync_period / 4, + tn->timeout); + } +} + +static u_char Tekram_boot_delay[7] __initdata = {3, 5, 10, 20, 30, 60, 120}; + +static void __init ncr_display_Tekram_nvram(Tekram_nvram *nvram) +{ + int i, tags, boot_delay; + char *rem; + + /* display Tekram nvram host data */ + tags = 2 << nvram->max_tags_index; + boot_delay = 0; + if (nvram->boot_delay_index < 6) + boot_delay = Tekram_boot_delay[nvram->boot_delay_index]; + switch((nvram->flags & TEKRAM_REMOVABLE_FLAGS) >> 6) { + default: + case 0: rem = ""; break; + case 1: rem = " REMOVABLE=boot device"; break; + case 2: rem = " REMOVABLE=all"; break; + } + + printk(KERN_DEBUG NAME53C8XX + ": HOST ID=%d%s%s%s%s%s%s%s%s%s BOOT DELAY=%d tags=%d\n", + nvram->host_id & 0x0f, + (nvram->flags1 & SYMBIOS_SCAM_ENABLE) ? " SCAM" :"", + (nvram->flags & TEKRAM_MORE_THAN_2_DRIVES) ? " >2DRIVES":"", + (nvram->flags & TEKRAM_DRIVES_SUP_1GB) ? " >1GB" :"", + (nvram->flags & TEKRAM_RESET_ON_POWER_ON) ? " RESET" :"", + (nvram->flags & TEKRAM_ACTIVE_NEGATION) ? " ACT_NEG" :"", + (nvram->flags & TEKRAM_IMMEDIATE_SEEK) ? " IMM_SEEK" :"", + (nvram->flags & TEKRAM_SCAN_LUNS) ? " SCAN_LUNS" :"", + (nvram->flags1 & TEKRAM_F2_F6_ENABLED) ? " F2_F6" :"", + rem, boot_delay, tags); + + /* display Tekram nvram drive data */ + for (i = 0; i <= 15; i++) { + int sync, j; + struct Tekram_target *tn = &nvram->target[i]; + j = tn->sync_index & 0xf; + sync = Tekram_sync[j]; + printk(KERN_DEBUG NAME53C8XX "-%d:%s%s%s%s%s%s PERIOD=%d\n", + i, + (tn->flags & TEKRAM_PARITY_CHECK) ? " PARITY" : "", + (tn->flags & TEKRAM_SYNC_NEGO) ? " SYNC" : "", + (tn->flags & TEKRAM_DISCONNECT_ENABLE) ? " DISC" : "", + (tn->flags & TEKRAM_START_CMD) ? " START" : "", + (tn->flags & TEKRAM_TAGGED_COMMANDS) ? " TCQ" : "", + (tn->flags & TEKRAM_WIDE_NEGO) ? " WIDE" : "", + sync); + } +} +#endif /* SCSI_NCR_DEBUG_NVRAM */ +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + + +/*=================================================================== +** +** Utility routines that protperly return data through /proc FS. +** +**=================================================================== +*/ +#ifdef SCSI_NCR_USER_INFO_SUPPORT + +struct info_str +{ + char *buffer; + int length; + int offset; + int pos; +}; + +static void copy_mem_info(struct info_str *info, char *data, int len) +{ + if (info->pos + len > info->length) + len = info->length - info->pos; + + if (info->pos + len < info->offset) { + info->pos += len; + return; + } + if (info->pos < info->offset) { + data += (info->offset - info->pos); + len -= (info->offset - info->pos); + } + + if (len > 0) { + memcpy(info->buffer + info->pos, data, len); + info->pos += len; + } +} + +static int copy_info(struct info_str *info, char *fmt, ...) +{ + va_list args; + char buf[81]; + int len; + + va_start(args, fmt); + len = vsprintf(buf, fmt, args); + va_end(args); + + copy_mem_info(info, buf, len); + return len; +} + +#endif + +/*=================================================================== +** +** Driver setup from the boot command line +** +**=================================================================== +*/ + +#ifdef MODULE +#define ARG_SEP ' ' +#else +#define ARG_SEP ',' +#endif + +#define OPT_TAGS 1 +#define OPT_MASTER_PARITY 2 +#define OPT_SCSI_PARITY 3 +#define OPT_DISCONNECTION 4 +#define OPT_SPECIAL_FEATURES 5 +#define OPT_ULTRA_SCSI 6 +#define OPT_FORCE_SYNC_NEGO 7 +#define OPT_REVERSE_PROBE 8 +#define OPT_DEFAULT_SYNC 9 +#define OPT_VERBOSE 10 +#define OPT_DEBUG 11 +#define OPT_BURST_MAX 12 +#define OPT_LED_PIN 13 +#define OPT_MAX_WIDE 14 +#define OPT_SETTLE_DELAY 15 +#define OPT_DIFF_SUPPORT 16 +#define OPT_IRQM 17 +#define OPT_PCI_FIX_UP 18 +#define OPT_BUS_CHECK 19 +#define OPT_OPTIMIZE 20 +#define OPT_RECOVERY 21 +#define OPT_SAFE_SETUP 22 +#define OPT_USE_NVRAM 23 +#define OPT_EXCLUDE 24 +#define OPT_HOST_ID 25 + +#ifdef SCSI_NCR_IARB_SUPPORT +#define OPT_IARB 26 +#endif + +static char setup_token[] __initdata = + "tags:" "mpar:" + "spar:" "disc:" + "specf:" "ultra:" + "fsn:" "revprob:" + "sync:" "verb:" + "debug:" "burst:" + "led:" "wide:" + "settle:" "diff:" + "irqm:" "pcifix:" + "buschk:" "optim:" + "recovery:" + "safe:" "nvram:" + "excl:" "hostid:" +#ifdef SCSI_NCR_IARB_SUPPORT + "iarb:" +#endif + ; /* DONNOT REMOVE THIS ';' */ + +#ifdef MODULE +#define ARG_SEP ' ' +#else +#define ARG_SEP ',' +#endif + +static int __init get_setup_token(char *p) +{ + char *cur = setup_token; + char *pc; + int i = 0; + + while (cur != NULL && (pc = strchr(cur, ':')) != NULL) { + ++pc; + ++i; + if (!strncmp(p, cur, pc - cur)) + return i; + cur = pc; + } + return 0; +} + + +static int __init sym53c8xx__setup(char *str) +{ +#ifdef SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT + char *cur = str; + char *pc, *pv; + int i, val, c; + int xi = 0; + + while (cur != NULL && (pc = strchr(cur, ':')) != NULL) { + char *pe; + + val = 0; + pv = pc; + c = *++pv; + + if (c == 'n') + val = 0; + else if (c == 'y') + val = 1; + else + val = (int) simple_strtoul(pv, &pe, 0); + + switch (get_setup_token(cur)) { + case OPT_TAGS: + driver_setup.default_tags = val; + if (pe && *pe == '/') { + i = 0; + while (*pe && *pe != ARG_SEP && + i < sizeof(driver_setup.tag_ctrl)-1) { + driver_setup.tag_ctrl[i++] = *pe++; + } + driver_setup.tag_ctrl[i] = '\0'; + } + break; + case OPT_MASTER_PARITY: + driver_setup.master_parity = val; + break; + case OPT_SCSI_PARITY: + driver_setup.scsi_parity = val; + break; + case OPT_DISCONNECTION: + driver_setup.disconnection = val; + break; + case OPT_SPECIAL_FEATURES: + driver_setup.special_features = val; + break; + case OPT_ULTRA_SCSI: + driver_setup.ultra_scsi = val; + break; + case OPT_FORCE_SYNC_NEGO: + driver_setup.force_sync_nego = val; + break; + case OPT_REVERSE_PROBE: + driver_setup.reverse_probe = val; + break; + case OPT_DEFAULT_SYNC: + driver_setup.default_sync = val; + break; + case OPT_VERBOSE: + driver_setup.verbose = val; + break; + case OPT_DEBUG: + driver_setup.debug = val; + break; + case OPT_BURST_MAX: + driver_setup.burst_max = val; + break; + case OPT_LED_PIN: + driver_setup.led_pin = val; + break; + case OPT_MAX_WIDE: + driver_setup.max_wide = val? 1:0; + break; + case OPT_SETTLE_DELAY: + driver_setup.settle_delay = val; + break; + case OPT_DIFF_SUPPORT: + driver_setup.diff_support = val; + break; + case OPT_IRQM: + driver_setup.irqm = val; + break; + case OPT_PCI_FIX_UP: + driver_setup.pci_fix_up = val; + break; + case OPT_BUS_CHECK: + driver_setup.bus_check = val; + break; + case OPT_OPTIMIZE: + driver_setup.optimize = val; + break; + case OPT_RECOVERY: + driver_setup.recovery = val; + break; + case OPT_USE_NVRAM: + driver_setup.use_nvram = val; + break; + case OPT_SAFE_SETUP: + memcpy(&driver_setup, &driver_safe_setup, + sizeof(driver_setup)); + break; + case OPT_EXCLUDE: + if (xi < SCSI_NCR_MAX_EXCLUDES) + driver_setup.excludes[xi++] = val; + break; + case OPT_HOST_ID: + driver_setup.host_id = val; + break; +#ifdef SCSI_NCR_IARB_SUPPORT + case OPT_IARB: + driver_setup.iarb = val; + break; +#endif + default: + printk("sym53c8xx_setup: unexpected boot option '%.*s' ignored\n", (int)(pc-cur+1), cur); + break; + } + + if ((cur = strchr(cur, ARG_SEP)) != NULL) + ++cur; + } +#endif /* SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT */ + return 1; +} + +/*=================================================================== +** +** Get device queue depth from boot command line. +** +**=================================================================== +*/ +#define DEF_DEPTH (driver_setup.default_tags) +#define ALL_TARGETS -2 +#define NO_TARGET -1 +#define ALL_LUNS -2 +#define NO_LUN -1 + +static int device_queue_depth(int unit, int target, int lun) +{ + int c, h, t, u, v; + char *p = driver_setup.tag_ctrl; + char *ep; + + h = -1; + t = NO_TARGET; + u = NO_LUN; + while ((c = *p++) != 0) { + v = simple_strtoul(p, &ep, 0); + switch(c) { + case '/': + ++h; + t = ALL_TARGETS; + u = ALL_LUNS; + break; + case 't': + if (t != target) + t = (target == v) ? v : NO_TARGET; + u = ALL_LUNS; + break; + case 'u': + if (u != lun) + u = (lun == v) ? v : NO_LUN; + break; + case 'q': + if (h == unit && + (t == ALL_TARGETS || t == target) && + (u == ALL_LUNS || u == lun)) + return v; + break; + case '-': + t = ALL_TARGETS; + u = ALL_LUNS; + break; + default: + break; + } + p = ep; + } + return DEF_DEPTH; +} + +/*=================================================================== +** +** Print out information about driver configuration. +** +**=================================================================== +*/ +static void __init ncr_print_driver_setup(void) +{ +#define YesNo(y) y ? 'y' : 'n' + printk (NAME53C8XX ": setup=disc:%c,specf:%d,ultra:%d,tags:%d,sync:%d," + "burst:%d,wide:%c,diff:%d,revprob:%c,buschk:0x%x\n", + YesNo(driver_setup.disconnection), + driver_setup.special_features, + driver_setup.ultra_scsi, + driver_setup.default_tags, + driver_setup.default_sync, + driver_setup.burst_max, + YesNo(driver_setup.max_wide), + driver_setup.diff_support, + YesNo(driver_setup.reverse_probe), + driver_setup.bus_check); + + printk (NAME53C8XX ": setup=mpar:%c,spar:%c,fsn=%c,verb:%d,debug:0x%x," + "led:%c,settle:%d,irqm:0x%x,nvram:0x%x,pcifix:0x%x\n", + YesNo(driver_setup.master_parity), + YesNo(driver_setup.scsi_parity), + YesNo(driver_setup.force_sync_nego), + driver_setup.verbose, + driver_setup.debug, + YesNo(driver_setup.led_pin), + driver_setup.settle_delay, + driver_setup.irqm, + driver_setup.use_nvram, + driver_setup.pci_fix_up); +#undef YesNo +} + +/*=================================================================== +** +** SYM53C8XX devices description table. +** +**=================================================================== +*/ + +static ncr_chip ncr_chip_table[] __initdata = SCSI_NCR_CHIP_TABLE; + +#ifdef SCSI_NCR_PQS_PDS_SUPPORT +/*=================================================================== +** +** Detect all NCR PQS/PDS boards and keep track of their bus nr. +** +** The NCR PQS or PDS card is constructed as a DEC bridge +** behind which sit a proprietary NCR memory controller and +** four or two 53c875s as separate devices. In its usual mode +** of operation, the 875s are slaved to the memory controller +** for all transfers. We can tell if an 875 is part of a +** PQS/PDS or not since if it is, it will be on the same bus +** as the memory controller. To operate with the Linux +** driver, the memory controller is disabled and the 875s +** freed to function independently. The only wrinkle is that +** the preset SCSI ID (which may be zero) must be read in from +** a special configuration space register of the 875. +** +**=================================================================== +*/ +#define SCSI_NCR_MAX_PQS_BUS 16 +static int pqs_bus[SCSI_NCR_MAX_PQS_BUS] __initdata = { 0 }; + +static void __init ncr_detect_pqs_pds(void) +{ + short index; + pcidev_t dev = PCIDEV_NULL; + + for(index=0; index < SCSI_NCR_MAX_PQS_BUS; index++) { + u_char tmp; + + dev = pci_find_device(0x101a, 0x0009, dev); + if (dev == PCIDEV_NULL) { + pqs_bus[index] = -1; + break; + } + printk(KERN_INFO NAME53C8XX ": NCR PQS/PDS memory controller detected on bus %d\n", PciBusNumber(dev)); + pci_read_config_byte(dev, 0x44, &tmp); + /* bit 1: allow individual 875 configuration */ + tmp |= 0x2; + pci_write_config_byte(dev, 0x44, tmp); + pci_read_config_byte(dev, 0x45, &tmp); + /* bit 2: drive individual 875 interrupts to the bus */ + tmp |= 0x4; + pci_write_config_byte(dev, 0x45, tmp); + + pqs_bus[index] = PciBusNumber(dev); + } +} +#endif /* SCSI_NCR_PQS_PDS_SUPPORT */ + +/*=================================================================== +** +** Read and check the PCI configuration for any detected NCR +** boards and save data for attaching after all boards have +** been detected. +** +**=================================================================== +*/ +static int __init +sym53c8xx_pci_init(Scsi_Host_Template *tpnt, pcidev_t pdev, ncr_device *device) +{ + u_short vendor_id, device_id, command; + u_char cache_line_size, latency_timer; + u_char suggested_cache_line_size = 0; + u_char pci_fix_up = driver_setup.pci_fix_up; + u_char revision; + u_int irq; + u_long base, base_2, io_port; + int i; + ncr_chip *chip; + + printk(KERN_INFO NAME53C8XX ": at PCI bus %d, device %d, function %d\n", + PciBusNumber(pdev), + (int) (PciDeviceFn(pdev) & 0xf8) >> 3, + (int) (PciDeviceFn(pdev) & 7)); + +#ifdef SCSI_NCR_DYNAMIC_DMA_MAPPING + if (!pci_dma_supported(pdev, (dma_addr_t) (0xffffffffUL))) { + printk(KERN_WARNING NAME53C8XX + "32 BIT PCI BUS DMA ADDRESSING NOT SUPPORTED\n"); + return -1; + } +#endif + + /* + ** Read info from the PCI config space. + ** pci_read_config_xxx() functions are assumed to be used for + ** successfully detected PCI devices. + */ + vendor_id = PciVendorId(pdev); + device_id = PciDeviceId(pdev); + irq = PciIrqLine(pdev); + i = 0; + i = pci_get_base_address(pdev, i, &io_port); + i = pci_get_base_address(pdev, i, &base); + (void) pci_get_base_address(pdev, i, &base_2); + + pci_read_config_word(pdev, PCI_COMMAND, &command); + pci_read_config_byte(pdev, PCI_CLASS_REVISION, &revision); + pci_read_config_byte(pdev, PCI_CACHE_LINE_SIZE, &cache_line_size); + pci_read_config_byte(pdev, PCI_LATENCY_TIMER, &latency_timer); + +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + /* + ** Match the BUS number for PQS/PDS devices. + ** Read the SCSI ID from a special register mapped + ** into the configuration space of the individual + ** 875s. This register is set up by the PQS bios + */ + for(i = 0; i < SCSI_NCR_MAX_PQS_BUS && pqs_bus[i] != -1; i++) { + u_char tmp; + if (pqs_bus[i] == PciBusNumber(pdev)) { + pci_read_config_byte(pdev, 0x84, &tmp); + device->pqs_pds = 1; + device->host_id = tmp; + break; + } + } +#endif /* SCSI_NCR_PQS_PDS_SUPPORT */ + + /* + ** If user excludes this chip, donnot initialize it. + */ + for (i = 0 ; i < SCSI_NCR_MAX_EXCLUDES ; i++) { + if (driver_setup.excludes[i] == + (io_port & PCI_BASE_ADDRESS_IO_MASK)) + return -1; + } + /* + ** Check if the chip is supported + */ + if ((device_id == PCI_DEVICE_ID_LSI_53C1010) || + (device_id == PCI_DEVICE_ID_LSI_53C1010_66)){ + printk(NAME53C8XX ": not initializing, device not supported\n"); + return -1; + } + chip = 0; + for (i = 0; i < sizeof(ncr_chip_table)/sizeof(ncr_chip_table[0]); i++) { + if (device_id != ncr_chip_table[i].device_id) + continue; + if (revision > ncr_chip_table[i].revision_id) + continue; + chip = &device->chip; + memcpy(chip, &ncr_chip_table[i], sizeof(*chip)); + chip->revision_id = revision; + break; + } + + /* + ** Ignore Symbios chips controlled by SISL RAID controller. + ** This controller sets value 0x52414944 at RAM end - 16. + */ +#if defined(__i386__) && !defined(SCSI_NCR_PCI_MEM_NOT_SUPPORTED) + if (chip && (base_2 & PCI_BASE_ADDRESS_MEM_MASK)) { + unsigned int ram_size, ram_val; + u_long ram_ptr; + + if (chip->features & FE_RAM8K) + ram_size = 8192; + else + ram_size = 4096; + + ram_ptr = remap_pci_mem(base_2 & PCI_BASE_ADDRESS_MEM_MASK, + ram_size); + if (ram_ptr) { + ram_val = readl_raw(ram_ptr + ram_size - 16); + unmap_pci_mem(ram_ptr, ram_size); + if (ram_val == 0x52414944) { + printk(NAME53C8XX": not initializing, " + "driven by SISL RAID controller.\n"); + return -1; + } + } + } +#endif /* i386 and PCI MEMORY accessible */ + + if (!chip) { + printk(NAME53C8XX ": not initializing, device not supported\n"); + return -1; + } + +#ifdef __powerpc__ + /* + ** Fix-up for power/pc. + ** Should not be performed by the driver. + */ + if ((command & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) + != (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) { + printk(NAME53C8XX ": setting%s%s...\n", + (command & PCI_COMMAND_IO) ? "" : " PCI_COMMAND_IO", + (command & PCI_COMMAND_MEMORY) ? "" : " PCI_COMMAND_MEMORY"); + command |= (PCI_COMMAND_IO | PCI_COMMAND_MEMORY); + pci_write_config_word(pdev, PCI_COMMAND, command); + } + +#if LINUX_VERSION_CODE < LinuxVersionCode(2,2,0) + if ( is_prep ) { + if (io_port >= 0x10000000) { + printk(NAME53C8XX ": reallocating io_port (Wacky IBM)"); + io_port = (io_port & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_0, io_port); + } + if (base >= 0x10000000) { + printk(NAME53C8XX ": reallocating base (Wacky IBM)"); + base = (base & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_1, base); + } + if (base_2 >= 0x10000000) { + printk(NAME53C8XX ": reallocating base2 (Wacky IBM)"); + base_2 = (base_2 & 0x00FFFFFF) | 0x01000000; + pci_write_config_dword(pdev, + PCI_BASE_ADDRESS_2, base_2); + } + } +#endif +#endif /* __powerpc__ */ + +#if defined(__sparc__) && (LINUX_VERSION_CODE < LinuxVersionCode(2,3,0)) + /* + * Severall fix-ups for sparc. + * + * Should not be performed by the driver, which is why all + * this crap is cleaned up in 2.4.x + */ + + base = __pa(base); + base_2 = __pa(base_2); + + if (!(command & PCI_COMMAND_MASTER)) { + if (initverbose >= 2) + printk("ncr53c8xx: setting PCI_COMMAND_MASTER bit (fixup)\n"); + command |= PCI_COMMAND_MASTER; + pcibios_write_config_word(bus, device_fn, PCI_COMMAND, command); + pcibios_read_config_word(bus, device_fn, PCI_COMMAND, &command); + } + + if ((chip->features & FE_WRIE) && !(command & PCI_COMMAND_INVALIDATE)) { + if (initverbose >= 2) + printk("ncr53c8xx: setting PCI_COMMAND_INVALIDATE bit (fixup)\n"); + command |= PCI_COMMAND_INVALIDATE; + pcibios_write_config_word(bus, device_fn, PCI_COMMAND, command); + pcibios_read_config_word(bus, device_fn, PCI_COMMAND, &command); + } + + if ((chip->features & FE_CLSE) && !cache_line_size) { + /* PCI_CACHE_LINE_SIZE value is in 32-bit words. */ + cache_line_size = 64 / sizeof(u_int32); + if (initverbose >= 2) + printk("ncr53c8xx: setting PCI_CACHE_LINE_SIZE to %d (fixup)\n", + cache_line_size); + pcibios_write_config_byte(bus, device_fn, + PCI_CACHE_LINE_SIZE, cache_line_size); + pcibios_read_config_byte(bus, device_fn, + PCI_CACHE_LINE_SIZE, &cache_line_size); + } + + if (!latency_timer) { + unsigned char min_gnt; + + pcibios_read_config_byte(bus, device_fn, + PCI_MIN_GNT, &min_gnt); + if (min_gnt == 0) + latency_timer = 128; + else + latency_timer = ((min_gnt << 3) & 0xff); + printk("ncr53c8xx: setting PCI_LATENCY_TIMER to %d bus clocks (fixup)\n", latency_timer); + pcibios_write_config_byte(bus, device_fn, + PCI_LATENCY_TIMER, latency_timer); + pcibios_read_config_byte(bus, device_fn, + PCI_LATENCY_TIMER, &latency_timer); + } +#endif /* __sparc__ && (LINUX_VERSION_CODE < LinuxVersionCode(2,3,0)) */ + +#if defined(__i386__) && !defined(MODULE) + if (!cache_line_size) { +#if LINUX_VERSION_CODE < LinuxVersionCode(2,1,75) + extern char x86; + switch(x86) { +#else + switch(boot_cpu_data.x86) { +#endif + case 4: suggested_cache_line_size = 4; break; + case 6: + case 5: suggested_cache_line_size = 8; break; + } + } +#endif /* __i386__ */ + + /* + ** Check availability of IO space, memory space. + ** Enable master capability if not yet. + ** + ** We shouldn't have to care about the IO region when + ** we are using MMIO. But calling check_region() from + ** both the ncr53c8xx and the sym53c8xx drivers prevents + ** from attaching devices from the both drivers. + ** If you have a better idea, let me know. + */ +/* #ifdef SCSI_NCR_IOMAPPED */ +#if 1 + if (!(command & PCI_COMMAND_IO)) { + printk(NAME53C8XX ": I/O base address (0x%lx) disabled.\n", + (long) io_port); + io_port = 0; + } +#endif + if (!(command & PCI_COMMAND_MEMORY)) { + printk(NAME53C8XX ": PCI_COMMAND_MEMORY not set.\n"); + base = 0; + base_2 = 0; + } + io_port &= PCI_BASE_ADDRESS_IO_MASK; + base &= PCI_BASE_ADDRESS_MEM_MASK; + base_2 &= PCI_BASE_ADDRESS_MEM_MASK; + +/* #ifdef SCSI_NCR_IOMAPPED */ +#if 1 + if (io_port && check_region (io_port, 128)) { + printk(NAME53C8XX ": IO region 0x%lx[0..127] is in use\n", + (long) io_port); + io_port = 0; + } + if (!io_port) + return -1; +#endif +#ifndef SCSI_NCR_IOMAPPED + if (!base) { + printk(NAME53C8XX ": MMIO base address disabled.\n"); + return -1; + } +#endif + +/* The ncr53c8xx driver never did set the PCI parity bit. */ +/* Since setting this bit is known to trigger spurious MDPE */ +/* errors on some 895 controllers when noise on power lines is */ +/* too high, I donnot want to change previous ncr53c8xx driver */ +/* behaviour on that point (the sym53c8xx driver set this bit). */ +#if 0 + /* + ** Set MASTER capable and PARITY bit, if not yet. + */ + if ((command & (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY)) + != (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY)) { + printk(NAME53C8XX ": setting%s%s...(fix-up)\n", + (command & PCI_COMMAND_MASTER) ? "" : " PCI_COMMAND_MASTER", + (command & PCI_COMMAND_PARITY) ? "" : " PCI_COMMAND_PARITY"); + command |= (PCI_COMMAND_MASTER | PCI_COMMAND_PARITY); + pci_write_config_word(pdev, PCI_COMMAND, command); + } +#else + /* + ** Set MASTER capable if not yet. + */ + if ((command & PCI_COMMAND_MASTER) != PCI_COMMAND_MASTER) { + printk(NAME53C8XX ": setting PCI_COMMAND_MASTER...(fix-up)\n"); + command |= PCI_COMMAND_MASTER; + pci_write_config_word(pdev, PCI_COMMAND, command); + } +#endif + + /* + ** Fix some features according to driver setup. + */ + if (!(driver_setup.special_features & 1)) + chip->features &= ~FE_SPECIAL_SET; + else { + if (driver_setup.special_features & 2) + chip->features &= ~FE_WRIE; + if (driver_setup.special_features & 4) + chip->features &= ~FE_NOPM; + } + if (driver_setup.ultra_scsi < 2 && (chip->features & FE_ULTRA2)) { + chip->features |= FE_ULTRA; + chip->features &= ~FE_ULTRA2; + } + if (driver_setup.ultra_scsi < 1) + chip->features &= ~FE_ULTRA; + if (!driver_setup.max_wide) + chip->features &= ~FE_WIDE; + + /* + ** Some features are required to be enabled in order to + ** work around some chip problems. :) ;) + ** (ITEM 12 of a DEL about the 896 I haven't yet). + ** We must ensure the chip will use WRITE AND INVALIDATE. + ** The revision number limit is for now arbitrary. + */ + if (device_id == PCI_DEVICE_ID_NCR_53C896 && revision <= 0x10) { + chip->features |= (FE_WRIE | FE_CLSE); + pci_fix_up |= 3; /* Force appropriate PCI fix-up */ + } + +#ifdef SCSI_NCR_PCI_FIX_UP_SUPPORT + /* + ** Try to fix up PCI config according to wished features. + */ + if ((pci_fix_up & 1) && (chip->features & FE_CLSE) && + !cache_line_size && suggested_cache_line_size) { + cache_line_size = suggested_cache_line_size; + pci_write_config_byte(pdev, + PCI_CACHE_LINE_SIZE, cache_line_size); + printk(NAME53C8XX ": PCI_CACHE_LINE_SIZE set to %d (fix-up).\n", + cache_line_size); + } + + if ((pci_fix_up & 2) && cache_line_size && + (chip->features & FE_WRIE) && !(command & PCI_COMMAND_INVALIDATE)) { + printk(NAME53C8XX": setting PCI_COMMAND_INVALIDATE (fix-up)\n"); + command |= PCI_COMMAND_INVALIDATE; + pci_write_config_word(pdev, PCI_COMMAND, command); + } + + /* + ** Tune PCI LATENCY TIMER according to burst max length transfer. + ** (latency timer >= burst length + 6, we add 10 to be quite sure) + */ + + if (chip->burst_max && (latency_timer == 0 || (pci_fix_up & 4))) { + u_char lt = (1 << chip->burst_max) + 6 + 10; + if (latency_timer < lt) { + printk(NAME53C8XX + ": changing PCI_LATENCY_TIMER from %d to %d.\n", + (int) latency_timer, (int) lt); + latency_timer = lt; + pci_write_config_byte(pdev, + PCI_LATENCY_TIMER, latency_timer); + } + } + +#endif /* SCSI_NCR_PCI_FIX_UP_SUPPORT */ + + /* + ** Initialise ncr_device structure with items required by ncr_attach. + */ + device->pdev = pdev; + device->slot.bus = PciBusNumber(pdev); + device->slot.device_fn = PciDeviceFn(pdev); + device->slot.base = base; + device->slot.base_2 = base_2; + device->slot.io_port = io_port; + device->slot.irq = irq; + device->attach_done = 0; + + return 0; +} + +/*=================================================================== +** +** Detect all 53c8xx hosts and then attach them. +** +** If we are using NVRAM, once all hosts are detected, we need to +** check any NVRAM for boot order in case detect and boot order +** differ and attach them using the order in the NVRAM. +** +** If no NVRAM is found or data appears invalid attach boards in +** the the order they are detected. +** +**=================================================================== +*/ +static int __init +sym53c8xx__detect(Scsi_Host_Template *tpnt, u_short ncr_chip_ids[], int chips) +{ + pcidev_t pcidev; + int i, j, hosts, count; + int attach_count = 0; + ncr_device *devtbl, *devp; +#ifdef SCSI_NCR_NVRAM_SUPPORT + ncr_nvram nvram0, nvram, *nvp; +#endif + + /* + ** PCI is required. + */ + if (!pci_present()) + return 0; + +#ifdef SCSI_NCR_DEBUG_INFO_SUPPORT + ncr_debug = driver_setup.debug; +#endif + if (initverbose >= 2) + ncr_print_driver_setup(); + + /* + ** Allocate the device table since we donnot want to + ** overflow the kernel stack. + ** 1 x 4K PAGE is enough for more than 40 devices for i386. + */ + devtbl = m_calloc(PAGE_SIZE, "devtbl"); + if (!devtbl) + return 0; + + /* + ** Detect all NCR PQS/PDS memory controllers. + */ +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + ncr_detect_pqs_pds(); +#endif + + /* + ** Detect all 53c8xx hosts. + ** Save the first Symbios NVRAM content if any + ** for the boot order. + */ + hosts = PAGE_SIZE / sizeof(*devtbl); +#ifdef SCSI_NCR_NVRAM_SUPPORT + nvp = (driver_setup.use_nvram & 0x1) ? &nvram0 : 0; +#endif + j = 0; + count = 0; + pcidev = PCIDEV_NULL; + while (1) { + char *msg = ""; + if (count >= hosts) + break; + if (j >= chips) + break; + i = driver_setup.reverse_probe ? chips - 1 - j : j; + pcidev = pci_find_device(PCI_VENDOR_ID_NCR, ncr_chip_ids[i], + pcidev); + if (pcidev == PCIDEV_NULL) { + ++j; + continue; + } + /* Some HW as the HP LH4 may report twice PCI devices */ + for (i = 0; i < count ; i++) { + if (devtbl[i].slot.bus == PciBusNumber(pcidev) && + devtbl[i].slot.device_fn == PciDeviceFn(pcidev)) + break; + } + if (i != count) /* Ignore this device if we already have it */ + continue; + devp = &devtbl[count]; + devp->host_id = driver_setup.host_id; + devp->attach_done = 0; + if (sym53c8xx_pci_init(tpnt, pcidev, devp)) { + continue; + } + ++count; +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (nvp) { + ncr_get_nvram(devp, nvp); + switch(nvp->type) { + case SCSI_NCR_SYMBIOS_NVRAM: + /* + * Switch to the other nvram buffer, so that + * nvram0 will contain the first Symbios + * format NVRAM content with boot order. + */ + nvp = &nvram; + msg = "with Symbios NVRAM"; + break; + case SCSI_NCR_TEKRAM_NVRAM: + msg = "with Tekram NVRAM"; + break; + } + } +#endif +#ifdef SCSI_NCR_PQS_PDS_SUPPORT + if (devp->pqs_pds) + msg = "(NCR PQS/PDS)"; +#endif + printk(KERN_INFO NAME53C8XX ": 53c%s detected %s\n", + devp->chip.name, msg); + } + + /* + ** If we have found a SYMBIOS NVRAM, use first the NVRAM boot + ** sequence as device boot order. + ** check devices in the boot record against devices detected. + ** attach devices if we find a match. boot table records that + ** do not match any detected devices will be ignored. + ** devices that do not match any boot table will not be attached + ** here but will attempt to be attached during the device table + ** rescan. + */ +#ifdef SCSI_NCR_NVRAM_SUPPORT + if (!nvp || nvram0.type != SCSI_NCR_SYMBIOS_NVRAM) + goto next; + for (i = 0; i < 4; i++) { + Symbios_host *h = &nvram0.data.Symbios.host[i]; + for (j = 0 ; j < count ; j++) { + devp = &devtbl[j]; + if (h->device_fn != devp->slot.device_fn || + h->bus_nr != devp->slot.bus || + h->device_id != devp->chip.device_id) + continue; + if (devp->attach_done) + continue; + if (h->flags & SYMBIOS_INIT_SCAN_AT_BOOT) { + ncr_get_nvram(devp, nvp); + if (!ncr_attach (tpnt, attach_count, devp)) + attach_count++; + } +#if 0 /* Restore previous behaviour of ncr53c8xx driver */ + else if (!(driver_setup.use_nvram & 0x80)) + printk(KERN_INFO NAME53C8XX + ": 53c%s state OFF thus not attached\n", + devp->chip.name); +#endif + else + continue; + + devp->attach_done = 1; + break; + } + } +next: +#endif + + /* + ** Rescan device list to make sure all boards attached. + ** Devices without boot records will not be attached yet + ** so try to attach them here. + */ + for (i= 0; i < count; i++) { + devp = &devtbl[i]; + if (!devp->attach_done) { +#ifdef SCSI_NCR_NVRAM_SUPPORT + ncr_get_nvram(devp, nvp); +#endif + if (!ncr_attach (tpnt, attach_count, devp)) + attach_count++; + } + } + + m_free(devtbl, PAGE_SIZE, "devtbl"); + + return attach_count; +} diff --git a/linux/src/drivers/scsi/sym53c8xx_defs.h b/linux/src/drivers/scsi/sym53c8xx_defs.h new file mode 100644 index 00000000..10acf780 --- /dev/null +++ b/linux/src/drivers/scsi/sym53c8xx_defs.h @@ -0,0 +1,1767 @@ +/****************************************************************************** +** High Performance device driver for the Symbios 53C896 controller. +** +** Copyright (C) 1998-2000 Gerard Roudier <groudier@club-internet.fr> +** +** This driver also supports all the Symbios 53C8XX controller family, +** except 53C810 revisions < 16, 53C825 revisions < 16 and all +** revisions of 53C815 controllers. +** +** This driver is based on the Linux port of the FreeBSD ncr driver. +** +** Copyright (C) 1994 Wolfgang Stanglmeier +** +**----------------------------------------------------------------------------- +** +** This program is free software; you can redistribute it and/or modify +** it under the terms of the GNU General Public License as published by +** the Free Software Foundation; either version 2 of the License, or +** (at your option) any later version. +** +** This program is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU General Public License for more details. +** +** You should have received a copy of the GNU General Public License +** along with this program; if not, write to the Free Software +** Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. +** +**----------------------------------------------------------------------------- +** +** The Linux port of the FreeBSD ncr driver has been achieved in +** november 1995 by: +** +** Gerard Roudier <groudier@club-internet.fr> +** +** Being given that this driver originates from the FreeBSD version, and +** in order to keep synergy on both, any suggested enhancements and corrections +** received on Linux are automatically a potential candidate for the FreeBSD +** version. +** +** The original driver has been written for 386bsd and FreeBSD by +** Wolfgang Stanglmeier <wolf@cologne.de> +** Stefan Esser <se@mi.Uni-Koeln.de> +** +**----------------------------------------------------------------------------- +** +** Major contributions: +** -------------------- +** +** NVRAM detection and reading. +** Copyright (C) 1997 Richard Waltham <dormouse@farsrobt.demon.co.uk> +** +******************************************************************************* +*/ + +#ifndef SYM53C8XX_DEFS_H +#define SYM53C8XX_DEFS_H + +/* +** Check supported Linux versions +*/ + +#if !defined(LINUX_VERSION_CODE) +#include <linux/version.h> +#endif +#include <linux/config.h> + +#ifndef LinuxVersionCode +#define LinuxVersionCode(v, p, s) (((v)<<16)+((p)<<8)+(s)) +#endif + +/* + * NCR PQS/PDS special device support. + */ +#ifdef CONFIG_SCSI_NCR53C8XX_PQS_PDS +#define SCSI_NCR_PQS_PDS_SUPPORT +#endif + +/* + * No more an option, enabled by default. + */ +#ifndef CONFIG_SCSI_NCR53C8XX_NVRAM_DETECT +#define CONFIG_SCSI_NCR53C8XX_NVRAM_DETECT +#endif + +/* +** These options are not tunable from 'make config' +*/ +//#define SCSI_NCR_PROC_INFO_SUPPORT + +/* +** If you want a driver as small as possible, donnot define the +** following options. +*/ +#define SCSI_NCR_BOOT_COMMAND_LINE_SUPPORT +#define SCSI_NCR_DEBUG_INFO_SUPPORT +#define SCSI_NCR_PCI_FIX_UP_SUPPORT +#ifdef SCSI_NCR_PROC_INFO_SUPPORT +# define SCSI_NCR_USER_COMMAND_SUPPORT +# define SCSI_NCR_USER_INFO_SUPPORT +#endif + +/* +** To disable integrity checking, do not define the +** following option. +*/ +#ifdef CONFIG_SCSI_NCR53C8XX_INTEGRITY_CHECK +# define SCSI_NCR_ENABLE_INTEGRITY_CHECK +#endif + +/*========================================================== +** +** nvram settings - #define SCSI_NCR_NVRAM_SUPPORT to enable +** +**========================================================== +*/ + +#ifdef CONFIG_SCSI_NCR53C8XX_NVRAM_DETECT +#define SCSI_NCR_NVRAM_SUPPORT +/* #define SCSI_NCR_DEBUG_NVRAM */ +#endif + +/* --------------------------------------------------------------------- +** Take into account kernel configured parameters. +** Most of these options can be overridden at startup by a command line. +** --------------------------------------------------------------------- +*/ + +/* + * For Ultra2 and Ultra3 SCSI support option, use special features. + * + * Value (default) means: + * bit 0 : all features enabled, except: + * bit 1 : PCI Write And Invalidate. + * bit 2 : Data Phase Mismatch handling from SCRIPTS. + * + * Use boot options ncr53c8xx=specf:1 if you want all chip features to be + * enabled by the driver. + */ +#define SCSI_NCR_SETUP_SPECIAL_FEATURES (3) + +/* + * For Ultra2 and Ultra3 SCSI support allow 80Mhz synchronous data transfers. + * Value means: + * 0 - Ultra speeds disabled + * 1 - Ultra enabled (Maximum 20Mtrans/sec) + * 2 - Ultra2 enabled (Maximum 40Mtrans/sec) + * 3 - Ultra3 enabled (Maximum 80Mtrans/sec) + * + * Use boot options sym53c8xx=ultra:3 to enable Ultra3 support. + */ + +#define SCSI_NCR_SETUP_ULTRA_SCSI (3) +#define SCSI_NCR_MAX_SYNC (80) + +/* + * Allow tags from 2 to 256, default 8 + */ +#ifdef CONFIG_SCSI_NCR53C8XX_MAX_TAGS +#if CONFIG_SCSI_NCR53C8XX_MAX_TAGS < 2 +#define SCSI_NCR_MAX_TAGS (2) +#elif CONFIG_SCSI_NCR53C8XX_MAX_TAGS > 256 +#define SCSI_NCR_MAX_TAGS (256) +#else +#define SCSI_NCR_MAX_TAGS CONFIG_SCSI_NCR53C8XX_MAX_TAGS +#endif +#else +#define SCSI_NCR_MAX_TAGS (8) +#endif + +/* + * Allow tagged command queuing support if configured with default number + * of tags set to max (see above). + */ +#ifdef CONFIG_SCSI_NCR53C8XX_DEFAULT_TAGS +#define SCSI_NCR_SETUP_DEFAULT_TAGS CONFIG_SCSI_NCR53C8XX_DEFAULT_TAGS +#elif defined CONFIG_SCSI_NCR53C8XX_TAGGED_QUEUE +#define SCSI_NCR_SETUP_DEFAULT_TAGS SCSI_NCR_MAX_TAGS +#else +#define SCSI_NCR_SETUP_DEFAULT_TAGS (0) +#endif + +/* + * Use normal IO if configured. Forced for alpha and powerpc. + * Powerpc fails copying to on-chip RAM using memcpy_toio(). + */ +#if defined(CONFIG_SCSI_NCR53C8XX_IOMAPPED) +#define SCSI_NCR_IOMAPPED +#elif defined(__alpha__) +#define SCSI_NCR_IOMAPPED +#elif defined(__powerpc__) +#define SCSI_NCR_IOMAPPED +#define SCSI_NCR_PCI_MEM_NOT_SUPPORTED +#elif defined(__sparc__) +#undef SCSI_NCR_IOMAPPED +#endif + +/* + * Should we enable DAC cycles on this platform? + * Until further investigation we do not enable it + * anywhere at the moment. + */ +#undef SCSI_NCR_USE_64BIT_DAC + +/* + * Immediate arbitration + */ +#if defined(CONFIG_SCSI_NCR53C8XX_IARB) +#define SCSI_NCR_IARB_SUPPORT +#endif + +/* + * Should we enable DAC cycles on sparc64 platforms? + * Until further investigation we do not enable it + * anywhere at the moment. + */ +#undef SCSI_NCR_USE_64BIT_DAC + +/* + * Sync transfer frequency at startup. + * Allow from 5Mhz to 80Mhz default 20 Mhz. + */ +#ifndef CONFIG_SCSI_NCR53C8XX_SYNC +#define CONFIG_SCSI_NCR53C8XX_SYNC (20) +#elif CONFIG_SCSI_NCR53C8XX_SYNC > SCSI_NCR_MAX_SYNC +#undef CONFIG_SCSI_NCR53C8XX_SYNC +#define CONFIG_SCSI_NCR53C8XX_SYNC SCSI_NCR_MAX_SYNC +#endif + +#if CONFIG_SCSI_NCR53C8XX_SYNC == 0 +#define SCSI_NCR_SETUP_DEFAULT_SYNC (255) +#elif CONFIG_SCSI_NCR53C8XX_SYNC <= 5 +#define SCSI_NCR_SETUP_DEFAULT_SYNC (50) +#elif CONFIG_SCSI_NCR53C8XX_SYNC <= 20 +#define SCSI_NCR_SETUP_DEFAULT_SYNC (250/(CONFIG_SCSI_NCR53C8XX_SYNC)) +#elif CONFIG_SCSI_NCR53C8XX_SYNC <= 33 +#define SCSI_NCR_SETUP_DEFAULT_SYNC (11) +#elif CONFIG_SCSI_NCR53C8XX_SYNC <= 40 +#define SCSI_NCR_SETUP_DEFAULT_SYNC (10) +#else +#define SCSI_NCR_SETUP_DEFAULT_SYNC (9) +#endif + +/* + * Disallow disconnections at boot-up + */ +#ifdef CONFIG_SCSI_NCR53C8XX_NO_DISCONNECT +#define SCSI_NCR_SETUP_DISCONNECTION (0) +#else +#define SCSI_NCR_SETUP_DISCONNECTION (1) +#endif + +/* + * Force synchronous negotiation for all targets + */ +#ifdef CONFIG_SCSI_NCR53C8XX_FORCE_SYNC_NEGO +#define SCSI_NCR_SETUP_FORCE_SYNC_NEGO (1) +#else +#define SCSI_NCR_SETUP_FORCE_SYNC_NEGO (0) +#endif + +/* + * Disable master parity checking (flawed hardwares need that) + */ +#ifdef CONFIG_SCSI_NCR53C8XX_DISABLE_MPARITY_CHECK +#define SCSI_NCR_SETUP_MASTER_PARITY (0) +#else +#define SCSI_NCR_SETUP_MASTER_PARITY (1) +#endif + +/* + * Disable scsi parity checking (flawed devices may need that) + */ +#ifdef CONFIG_SCSI_NCR53C8XX_DISABLE_PARITY_CHECK +#define SCSI_NCR_SETUP_SCSI_PARITY (0) +#else +#define SCSI_NCR_SETUP_SCSI_PARITY (1) +#endif + +/* + * Vendor specific stuff + */ +#ifdef CONFIG_SCSI_NCR53C8XX_SYMBIOS_COMPAT +#define SCSI_NCR_SETUP_LED_PIN (1) +#define SCSI_NCR_SETUP_DIFF_SUPPORT (4) +#else +#define SCSI_NCR_SETUP_LED_PIN (0) +#define SCSI_NCR_SETUP_DIFF_SUPPORT (0) +#endif + +/* + * Settle time after reset at boot-up + */ +#define SCSI_NCR_SETUP_SETTLE_TIME (2) + +/* +** Bridge quirks work-around option defaulted to 1. +*/ +#ifndef SCSI_NCR_PCIQ_WORK_AROUND_OPT +#define SCSI_NCR_PCIQ_WORK_AROUND_OPT 1 +#endif + +/* +** Work-around common bridge misbehaviour. +** +** - Do not flush posted writes in the opposite +** direction on read. +** - May reorder DMA writes to memory. +** +** This option should not affect performances +** significantly, so it is the default. +*/ +#if SCSI_NCR_PCIQ_WORK_AROUND_OPT == 1 +#define SCSI_NCR_PCIQ_MAY_NOT_FLUSH_PW_UPSTREAM +#define SCSI_NCR_PCIQ_MAY_REORDER_WRITES +#define SCSI_NCR_PCIQ_MAY_MISS_COMPLETIONS + +/* +** Same as option 1, but also deal with +** misconfigured interrupts. +** +** - Edge triggerred instead of level sensitive. +** - No interrupt line connected. +** - IRQ number misconfigured. +** +** If no interrupt is delivered, the driver will +** catch the interrupt conditions 10 times per +** second. No need to say that this option is +** not recommended. +*/ +#elif SCSI_NCR_PCIQ_WORK_AROUND_OPT == 2 +#define SCSI_NCR_PCIQ_MAY_NOT_FLUSH_PW_UPSTREAM +#define SCSI_NCR_PCIQ_MAY_REORDER_WRITES +#define SCSI_NCR_PCIQ_MAY_MISS_COMPLETIONS +#define SCSI_NCR_PCIQ_BROKEN_INTR + +/* +** Some bridge designers decided to flush +** everything prior to deliver the interrupt. +** This option tries to deal with such a +** behaviour. +*/ +#elif SCSI_NCR_PCIQ_WORK_AROUND_OPT == 3 +#define SCSI_NCR_PCIQ_SYNC_ON_INTR +#endif + +/* +** Other parameters not configurable with "make config" +** Avoid to change these constants, unless you know what you are doing. +*/ + +#define SCSI_NCR_ALWAYS_SIMPLE_TAG +#define SCSI_NCR_MAX_SCATTER (127) +#define SCSI_NCR_MAX_TARGET (16) + +/* +** Compute some desirable value for CAN_QUEUE +** and CMD_PER_LUN. +** The driver will use lower values if these +** ones appear to be too large. +*/ +#define SCSI_NCR_CAN_QUEUE (8*SCSI_NCR_MAX_TAGS + 2*SCSI_NCR_MAX_TARGET) +#define SCSI_NCR_CMD_PER_LUN (SCSI_NCR_MAX_TAGS) + +#define SCSI_NCR_SG_TABLESIZE (SCSI_NCR_MAX_SCATTER) +#define SCSI_NCR_TIMER_INTERVAL (HZ) + +#if 1 /* defined CONFIG_SCSI_MULTI_LUN */ +#define SCSI_NCR_MAX_LUN (16) +#else +#define SCSI_NCR_MAX_LUN (1) +#endif + +#ifndef HOSTS_C + +/* +** These simple macros limit expression involving +** kernel time values (jiffies) to some that have +** chance not to be too much incorrect. :-) +*/ +#define ktime_get(o) (jiffies + (u_long) o) +#define ktime_exp(b) ((long)(jiffies) - (long)(b) >= 0) +#define ktime_dif(a, b) ((long)(a) - (long)(b)) +/* These ones are not used in this driver */ +#define ktime_add(a, o) ((a) + (u_long)(o)) +#define ktime_sub(a, o) ((a) - (u_long)(o)) + + +/* + * IO functions definition for big/little endian CPU support. + * For now, the NCR is only supported in little endian addressing mode, + */ + +#ifdef __BIG_ENDIAN + +#if LINUX_VERSION_CODE < LinuxVersionCode(2,1,0) +#error "BIG ENDIAN byte ordering needs kernel version >= 2.1.0" +#endif + +#define inw_l2b inw +#define inl_l2b inl +#define outw_b2l outw +#define outl_b2l outl +#define readw_l2b readw +#define readl_l2b readl +#define writew_b2l writew +#define writel_b2l writel + +#else /* little endian */ + +#if defined(__i386__) /* i386 implements full FLAT memory/MMIO model */ +#define inw_raw inw +#define inl_raw inl +#define outw_raw outw +#define outl_raw outl +#define readb_raw(a) (*(volatile unsigned char *) (a)) +#define readw_raw(a) (*(volatile unsigned short *) (a)) +#define readl_raw(a) (*(volatile unsigned int *) (a)) +#define writeb_raw(b,a) ((*(volatile unsigned char *) (a)) = (b)) +#define writew_raw(b,a) ((*(volatile unsigned short *) (a)) = (b)) +#define writel_raw(b,a) ((*(volatile unsigned int *) (a)) = (b)) + +#else /* Other little-endian */ +#define inw_raw inw +#define inl_raw inl +#define outw_raw outw +#define outl_raw outl +#define readw_raw readw +#define readl_raw readl +#define writew_raw writew +#define writel_raw writel + +#endif +#endif + +#ifdef SCSI_NCR_BIG_ENDIAN +#error "The NCR in BIG ENDIAN addressing mode is not (yet) supported" +#endif + + +/* + * IA32 architecture does not reorder STORES and prevents + * LOADS from passing STORES. It is called `program order' + * by Intel and allows device drivers to deal with memory + * ordering by only ensuring that the code is not reordered + * by the compiler when ordering is required. + * Other architectures implement a weaker ordering that + * requires memory barriers (and also IO barriers when they + * make sense) to be used. + * We want to be paranoid for ppc and ia64. :) + */ + +#if defined __i386__ +#define MEMORY_BARRIER() do { ; } while(0) +#elif defined __powerpc__ +#define MEMORY_BARRIER() __asm__ volatile("eieio; sync" : : : "memory") +#elif defined __ia64__ +#define MEMORY_BARRIER() __asm__ volatile("mf.a; mf" : : : "memory") +#else +#define MEMORY_BARRIER() mb() +#endif + + +/* + * If the NCR uses big endian addressing mode over the + * PCI, actual io register addresses for byte and word + * accesses must be changed according to lane routing. + * Btw, ncr_offb() and ncr_offw() macros only apply to + * constants and so donnot generate bloated code. + */ + +#if defined(SCSI_NCR_BIG_ENDIAN) + +#define ncr_offb(o) (((o)&~3)+((~((o)&3))&3)) +#define ncr_offw(o) (((o)&~3)+((~((o)&3))&2)) + +#else + +#define ncr_offb(o) (o) +#define ncr_offw(o) (o) + +#endif + +/* + * If the CPU and the NCR use same endian-ness addressing, + * no byte reordering is needed for script patching. + * Macro cpu_to_scr() is to be used for script patching. + * Macro scr_to_cpu() is to be used for getting a DWORD + * from the script. + */ + +#if defined(__BIG_ENDIAN) && !defined(SCSI_NCR_BIG_ENDIAN) + +#define cpu_to_scr(dw) cpu_to_le32(dw) +#define scr_to_cpu(dw) le32_to_cpu(dw) + +#elif defined(__LITTLE_ENDIAN) && defined(SCSI_NCR_BIG_ENDIAN) + +#define cpu_to_scr(dw) cpu_to_be32(dw) +#define scr_to_cpu(dw) be32_to_cpu(dw) + +#else + +#define cpu_to_scr(dw) (dw) +#define scr_to_cpu(dw) (dw) + +#endif + +/* + * Access to the controller chip. + * + * If SCSI_NCR_IOMAPPED is defined, the driver will use + * normal IOs instead of the MEMORY MAPPED IO method + * recommended by PCI specifications. + * If all PCI bridges, host brigdes and architectures + * would have been correctly designed for PCI, this + * option would be useless. + * + * If the CPU and the NCR use same endian-ness addressing, + * no byte reordering is needed for accessing chip io + * registers. Functions suffixed by '_raw' are assumed + * to access the chip over the PCI without doing byte + * reordering. Functions suffixed by '_l2b' are + * assumed to perform little-endian to big-endian byte + * reordering, those suffixed by '_b2l' blah, blah, + * blah, ... + */ + +#if defined(SCSI_NCR_IOMAPPED) + +/* + * IO mapped only input / ouput + */ + +#define INB_OFF(o) inb (np->base_io + ncr_offb(o)) +#define OUTB_OFF(o, val) outb ((val), np->base_io + ncr_offb(o)) + +#if defined(__BIG_ENDIAN) && !defined(SCSI_NCR_BIG_ENDIAN) + +#define INW_OFF(o) inw_l2b (np->base_io + ncr_offw(o)) +#define INL_OFF(o) inl_l2b (np->base_io + (o)) + +#define OUTW_OFF(o, val) outw_b2l ((val), np->base_io + ncr_offw(o)) +#define OUTL_OFF(o, val) outl_b2l ((val), np->base_io + (o)) + +#elif defined(__LITTLE_ENDIAN) && defined(SCSI_NCR_BIG_ENDIAN) + +#define INW_OFF(o) inw_b2l (np->base_io + ncr_offw(o)) +#define INL_OFF(o) inl_b2l (np->base_io + (o)) + +#define OUTW_OFF(o, val) outw_l2b ((val), np->base_io + ncr_offw(o)) +#define OUTL_OFF(o, val) outl_l2b ((val), np->base_io + (o)) + +#else + +#define INW_OFF(o) inw_raw (np->base_io + ncr_offw(o)) +#define INL_OFF(o) inl_raw (np->base_io + (o)) + +#define OUTW_OFF(o, val) outw_raw ((val), np->base_io + ncr_offw(o)) +#define OUTL_OFF(o, val) outl_raw ((val), np->base_io + (o)) + +#endif /* ENDIANs */ + +#else /* defined SCSI_NCR_IOMAPPED */ + +/* + * MEMORY mapped IO input / output + */ + +#define INB_OFF(o) readb((char *)np->reg + ncr_offb(o)) +#define OUTB_OFF(o, val) writeb((val), (char *)np->reg + ncr_offb(o)) + +#if defined(__BIG_ENDIAN) && !defined(SCSI_NCR_BIG_ENDIAN) + +#define INW_OFF(o) readw_l2b((char *)np->reg + ncr_offw(o)) +#define INL_OFF(o) readl_l2b((char *)np->reg + (o)) + +#define OUTW_OFF(o, val) writew_b2l((val), (char *)np->reg + ncr_offw(o)) +#define OUTL_OFF(o, val) writel_b2l((val), (char *)np->reg + (o)) + +#elif defined(__LITTLE_ENDIAN) && defined(SCSI_NCR_BIG_ENDIAN) + +#define INW_OFF(o) readw_b2l((char *)np->reg + ncr_offw(o)) +#define INL_OFF(o) readl_b2l((char *)np->reg + (o)) + +#define OUTW_OFF(o, val) writew_l2b((val), (char *)np->reg + ncr_offw(o)) +#define OUTL_OFF(o, val) writel_l2b((val), (char *)np->reg + (o)) + +#else + +#define INW_OFF(o) readw_raw((char *)np->reg + ncr_offw(o)) +#define INL_OFF(o) readl_raw((char *)np->reg + (o)) + +#define OUTW_OFF(o, val) writew_raw((val), (char *)np->reg + ncr_offw(o)) +#define OUTL_OFF(o, val) writel_raw((val), (char *)np->reg + (o)) + +#endif + +#endif /* defined SCSI_NCR_IOMAPPED */ + +#define INB(r) INB_OFF (offsetof(struct ncr_reg,r)) +#define INW(r) INW_OFF (offsetof(struct ncr_reg,r)) +#define INL(r) INL_OFF (offsetof(struct ncr_reg,r)) + +#define OUTB(r, val) OUTB_OFF (offsetof(struct ncr_reg,r), (val)) +#define OUTW(r, val) OUTW_OFF (offsetof(struct ncr_reg,r), (val)) +#define OUTL(r, val) OUTL_OFF (offsetof(struct ncr_reg,r), (val)) + +/* + * Set bit field ON, OFF + */ + +#define OUTONB(r, m) OUTB(r, INB(r) | (m)) +#define OUTOFFB(r, m) OUTB(r, INB(r) & ~(m)) +#define OUTONW(r, m) OUTW(r, INW(r) | (m)) +#define OUTOFFW(r, m) OUTW(r, INW(r) & ~(m)) +#define OUTONL(r, m) OUTL(r, INL(r) | (m)) +#define OUTOFFL(r, m) OUTL(r, INL(r) & ~(m)) + +/* + * We normally want the chip to have a consistent view + * of driver internal data structures when we restart it. + * Thus these macros. + */ +#define OUTL_DSP(v) \ + do { \ + MEMORY_BARRIER(); \ + OUTL (nc_dsp, (v)); \ + } while (0) + +#define OUTONB_STD() \ + do { \ + MEMORY_BARRIER(); \ + OUTONB (nc_dcntl, (STD|NOCOM)); \ + } while (0) + + +/* +** NCR53C8XX Device Ids +*/ + +#ifndef PCI_DEVICE_ID_NCR_53C810 +#define PCI_DEVICE_ID_NCR_53C810 1 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C810AP +#define PCI_DEVICE_ID_NCR_53C810AP 5 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C815 +#define PCI_DEVICE_ID_NCR_53C815 4 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C820 +#define PCI_DEVICE_ID_NCR_53C820 2 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C825 +#define PCI_DEVICE_ID_NCR_53C825 3 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C860 +#define PCI_DEVICE_ID_NCR_53C860 6 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C875 +#define PCI_DEVICE_ID_NCR_53C875 0xf +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C875J +#define PCI_DEVICE_ID_NCR_53C875J 0x8f +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C885 +#define PCI_DEVICE_ID_NCR_53C885 0xd +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C895 +#define PCI_DEVICE_ID_NCR_53C895 0xc +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C896 +#define PCI_DEVICE_ID_NCR_53C896 0xb +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C895A +#define PCI_DEVICE_ID_NCR_53C895A 0x12 +#endif + +#ifndef PCI_DEVICE_ID_NCR_53C1510D +#define PCI_DEVICE_ID_NCR_53C1510D 0xa +#endif + +#ifndef PCI_DEVICE_ID_LSI_53C1010 +#define PCI_DEVICE_ID_LSI_53C1010 0x20 +#endif + +#ifndef PCI_DEVICE_ID_LSI_53C1010_66 +#define PCI_DEVICE_ID_LSI_53C1010_66 0x21 +#endif + + +/* +** NCR53C8XX devices features table. +*/ +typedef struct { + unsigned short device_id; + unsigned short revision_id; + char *name; + unsigned char burst_max; /* log-base-2 of max burst */ + unsigned char offset_max; + unsigned char nr_divisor; + unsigned int features; +#define FE_LED0 (1<<0) +#define FE_WIDE (1<<1) /* Wide data transfers */ +#define FE_ULTRA (1<<2) /* Ultra speed 20Mtrans/sec */ +#define FE_ULTRA2 (1<<3) /* Ultra 2 - 40 Mtrans/sec */ +#define FE_DBLR (1<<4) /* Clock doubler present */ +#define FE_QUAD (1<<5) /* Clock quadrupler present */ +#define FE_ERL (1<<6) /* Enable read line */ +#define FE_CLSE (1<<7) /* Cache line size enable */ +#define FE_WRIE (1<<8) /* Write & Invalidate enable */ +#define FE_ERMP (1<<9) /* Enable read multiple */ +#define FE_BOF (1<<10) /* Burst opcode fetch */ +#define FE_DFS (1<<11) /* DMA fifo size */ +#define FE_PFEN (1<<12) /* Prefetch enable */ +#define FE_LDSTR (1<<13) /* Load/Store supported */ +#define FE_RAM (1<<14) /* On chip RAM present */ +#define FE_CLK80 (1<<15) /* Board clock is 80 MHz */ +#define FE_RAM8K (1<<16) /* On chip RAM sized 8Kb */ +#define FE_64BIT (1<<17) /* Supports 64-bit addressing */ +#define FE_IO256 (1<<18) /* Requires full 256 bytes in PCI space */ +#define FE_NOPM (1<<19) /* Scripts handles phase mismatch */ +#define FE_LEDC (1<<20) /* Hardware control of LED */ +#define FE_DIFF (1<<21) /* Support Differential SCSI */ +#define FE_ULTRA3 (1<<22) /* Ultra-3 80Mtrans/sec */ +#define FE_66MHZ (1<<23) /* 66MHz PCI Support */ + +#define FE_CACHE_SET (FE_ERL|FE_CLSE|FE_WRIE|FE_ERMP) +#define FE_SCSI_SET (FE_WIDE|FE_ULTRA|FE_ULTRA2|FE_DBLR|FE_QUAD|F_CLK80) +#define FE_SPECIAL_SET (FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN|FE_RAM) +} ncr_chip; + +/* +** DEL 397 - 53C875 Rev 3 - Part Number 609-0392410 - ITEM 3. +** Memory Read transaction terminated by a retry followed by +** Memory Read Line command. +*/ +#define FE_CACHE0_SET (FE_CACHE_SET & ~FE_ERL) + +/* +** DEL 397 - 53C875 Rev 3 - Part Number 609-0392410 - ITEM 5. +** On paper, this errata is harmless. But it is a good reason for +** using a shorter programmed burst length (64 DWORDS instead of 128). +*/ + +#define SCSI_NCR_CHIP_TABLE \ +{ \ + {PCI_DEVICE_ID_NCR_53C810, 0x0f, "810", 4, 8, 4, \ + FE_ERL} \ + , \ + {PCI_DEVICE_ID_NCR_53C810, 0xff, "810a", 4, 8, 4, \ + FE_CACHE_SET|FE_LDSTR|FE_PFEN|FE_BOF} \ + , \ + {PCI_DEVICE_ID_NCR_53C815, 0xff, "815", 4, 8, 4, \ + FE_ERL|FE_BOF} \ + , \ + {PCI_DEVICE_ID_NCR_53C820, 0xff, "820", 4, 8, 4, \ + FE_WIDE|FE_ERL} \ + , \ + {PCI_DEVICE_ID_NCR_53C825, 0x0f, "825", 4, 8, 4, \ + FE_WIDE|FE_ERL|FE_BOF|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C825, 0xff, "825a", 6, 8, 4, \ + FE_WIDE|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN|FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C860, 0xff, "860", 4, 8, 5, \ + FE_ULTRA|FE_CLK80|FE_CACHE_SET|FE_BOF|FE_LDSTR|FE_PFEN} \ + , \ + {PCI_DEVICE_ID_NCR_53C875, 0x01, "875", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_CLK80|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN|\ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C875, 0x0f, "875", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C875, 0x1f, "876", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C875, 0x2f, "875E", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C875, 0xff, "876", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C875J,0xff, "875J", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM} \ + , \ + {PCI_DEVICE_ID_NCR_53C885, 0xff, "885", 6, 16, 5, \ + FE_WIDE|FE_ULTRA|FE_DBLR|FE_CACHE0_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_DIFF} \ + , \ + {PCI_DEVICE_ID_NCR_53C895, 0xff, "895", 6, 31, 7, \ + FE_WIDE|FE_ULTRA2|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM} \ + , \ + {PCI_DEVICE_ID_NCR_53C896, 0xff, "896", 6, 31, 7, \ + FE_WIDE|FE_ULTRA2|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_RAM8K|FE_64BIT|FE_IO256|FE_NOPM|FE_LEDC} \ + , \ + {PCI_DEVICE_ID_NCR_53C895A, 0xff, "895a", 6, 31, 7, \ + FE_WIDE|FE_ULTRA2|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_RAM8K|FE_64BIT|FE_IO256|FE_NOPM|FE_LEDC} \ + , \ + {PCI_DEVICE_ID_NCR_53C1510D, 0xff, "1510D", 7, 31, 7, \ + FE_WIDE|FE_ULTRA2|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_IO256} \ + , \ + {PCI_DEVICE_ID_LSI_53C1010, 0xff, "1010", 6, 31, 7, \ + FE_WIDE|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_RAM8K|FE_64BIT|FE_IO256|FE_NOPM|FE_LEDC|FE_ULTRA3} \ + , \ + {PCI_DEVICE_ID_LSI_53C1010_66, 0xff, "1010_66", 6, 31, 7, \ + FE_WIDE|FE_QUAD|FE_CACHE_SET|FE_BOF|FE_DFS|FE_LDSTR|FE_PFEN| \ + FE_RAM|FE_RAM8K|FE_64BIT|FE_IO256|FE_NOPM|FE_LEDC|FE_ULTRA3|FE_66MHZ} \ +} + +/* + * List of supported NCR chip ids + */ +#define SCSI_NCR_CHIP_IDS \ +{ \ + PCI_DEVICE_ID_NCR_53C810, \ + PCI_DEVICE_ID_NCR_53C815, \ + PCI_DEVICE_ID_NCR_53C820, \ + PCI_DEVICE_ID_NCR_53C825, \ + PCI_DEVICE_ID_NCR_53C860, \ + PCI_DEVICE_ID_NCR_53C875, \ + PCI_DEVICE_ID_NCR_53C875J, \ + PCI_DEVICE_ID_NCR_53C885, \ + PCI_DEVICE_ID_NCR_53C895, \ + PCI_DEVICE_ID_NCR_53C896, \ + PCI_DEVICE_ID_NCR_53C895A, \ + PCI_DEVICE_ID_NCR_53C1510D, \ + PCI_DEVICE_ID_LSI_53C1010, \ + PCI_DEVICE_ID_LSI_53C1010_66 \ +} + +/* +** Driver setup structure. +** +** This structure is initialized from linux config options. +** It can be overridden at boot-up by the boot command line. +*/ +#define SCSI_NCR_MAX_EXCLUDES 8 +struct ncr_driver_setup { + u_char master_parity; + u_char scsi_parity; + u_char disconnection; + u_char special_features; + u_char ultra_scsi; + u_char force_sync_nego; + u_char reverse_probe; + u_char pci_fix_up; + u_char use_nvram; + u_char verbose; + u_char default_tags; + u_short default_sync; + u_short debug; + u_char burst_max; + u_char led_pin; + u_char max_wide; + u_char settle_delay; + u_char diff_support; + u_char irqm; + u_char bus_check; + u_char optimize; + u_char recovery; + u_char host_id; + u_short iarb; + u_long excludes[SCSI_NCR_MAX_EXCLUDES]; + char tag_ctrl[100]; +}; + +/* +** Initial setup. +** Can be overriden at startup by a command line. +*/ +#define SCSI_NCR_DRIVER_SETUP \ +{ \ + SCSI_NCR_SETUP_MASTER_PARITY, \ + SCSI_NCR_SETUP_SCSI_PARITY, \ + SCSI_NCR_SETUP_DISCONNECTION, \ + SCSI_NCR_SETUP_SPECIAL_FEATURES, \ + SCSI_NCR_SETUP_ULTRA_SCSI, \ + SCSI_NCR_SETUP_FORCE_SYNC_NEGO, \ + 0, \ + 0, \ + 1, \ + 0, \ + SCSI_NCR_SETUP_DEFAULT_TAGS, \ + SCSI_NCR_SETUP_DEFAULT_SYNC, \ + 0x00, \ + 7, \ + SCSI_NCR_SETUP_LED_PIN, \ + 1, \ + SCSI_NCR_SETUP_SETTLE_TIME, \ + SCSI_NCR_SETUP_DIFF_SUPPORT, \ + 0, \ + 1, \ + 0, \ + 0, \ + 255, \ + 0x00 \ +} + +/* +** Boot fail safe setup. +** Override initial setup from boot command line: +** ncr53c8xx=safe:y +*/ +#define SCSI_NCR_DRIVER_SAFE_SETUP \ +{ \ + 0, \ + 1, \ + 0, \ + 0, \ + 0, \ + 0, \ + 0, \ + 0, \ + 1, \ + 2, \ + 0, \ + 255, \ + 0x00, \ + 255, \ + 0, \ + 0, \ + 10, \ + 1, \ + 1, \ + 1, \ + 0, \ + 0, \ + 255 \ +} + +#ifdef SCSI_NCR_NVRAM_SUPPORT +/* +** Symbios NvRAM data format +*/ +#define SYMBIOS_NVRAM_SIZE 368 +#define SYMBIOS_NVRAM_ADDRESS 0x100 + +struct Symbios_nvram { +/* Header 6 bytes */ + u_short type; /* 0x0000 */ + u_short byte_count; /* excluding header/trailer */ + u_short checksum; + +/* Controller set up 20 bytes */ + u_char v_major; /* 0x00 */ + u_char v_minor; /* 0x30 */ + u_int32 boot_crc; + u_short flags; +#define SYMBIOS_SCAM_ENABLE (1) +#define SYMBIOS_PARITY_ENABLE (1<<1) +#define SYMBIOS_VERBOSE_MSGS (1<<2) +#define SYMBIOS_CHS_MAPPING (1<<3) +#define SYMBIOS_NO_NVRAM (1<<3) /* ??? */ + u_short flags1; +#define SYMBIOS_SCAN_HI_LO (1) + u_short term_state; +#define SYMBIOS_TERM_CANT_PROGRAM (0) +#define SYMBIOS_TERM_ENABLED (1) +#define SYMBIOS_TERM_DISABLED (2) + u_short rmvbl_flags; +#define SYMBIOS_RMVBL_NO_SUPPORT (0) +#define SYMBIOS_RMVBL_BOOT_DEVICE (1) +#define SYMBIOS_RMVBL_MEDIA_INSTALLED (2) + u_char host_id; + u_char num_hba; /* 0x04 */ + u_char num_devices; /* 0x10 */ + u_char max_scam_devices; /* 0x04 */ + u_char num_valid_scam_devives; /* 0x00 */ + u_char rsvd; + +/* Boot order 14 bytes * 4 */ + struct Symbios_host{ + u_short type; /* 4:8xx / 0:nok */ + u_short device_id; /* PCI device id */ + u_short vendor_id; /* PCI vendor id */ + u_char bus_nr; /* PCI bus number */ + u_char device_fn; /* PCI device/function number << 3*/ + u_short word8; + u_short flags; +#define SYMBIOS_INIT_SCAN_AT_BOOT (1) + u_short io_port; /* PCI io_port address */ + } host[4]; + +/* Targets 8 bytes * 16 */ + struct Symbios_target { + u_char flags; +#define SYMBIOS_DISCONNECT_ENABLE (1) +#define SYMBIOS_SCAN_AT_BOOT_TIME (1<<1) +#define SYMBIOS_SCAN_LUNS (1<<2) +#define SYMBIOS_QUEUE_TAGS_ENABLED (1<<3) + u_char rsvd; + u_char bus_width; /* 0x08/0x10 */ + u_char sync_offset; + u_short sync_period; /* 4*period factor */ + u_short timeout; + } target[16]; +/* Scam table 8 bytes * 4 */ + struct Symbios_scam { + u_short id; + u_short method; +#define SYMBIOS_SCAM_DEFAULT_METHOD (0) +#define SYMBIOS_SCAM_DONT_ASSIGN (1) +#define SYMBIOS_SCAM_SET_SPECIFIC_ID (2) +#define SYMBIOS_SCAM_USE_ORDER_GIVEN (3) + u_short status; +#define SYMBIOS_SCAM_UNKNOWN (0) +#define SYMBIOS_SCAM_DEVICE_NOT_FOUND (1) +#define SYMBIOS_SCAM_ID_NOT_SET (2) +#define SYMBIOS_SCAM_ID_VALID (3) + u_char target_id; + u_char rsvd; + } scam[4]; + + u_char spare_devices[15*8]; + u_char trailer[6]; /* 0xfe 0xfe 0x00 0x00 0x00 0x00 */ +}; +typedef struct Symbios_nvram Symbios_nvram; +typedef struct Symbios_host Symbios_host; +typedef struct Symbios_target Symbios_target; +typedef struct Symbios_scam Symbios_scam; + +/* +** Tekram NvRAM data format. +*/ +#define TEKRAM_NVRAM_SIZE 64 +#define TEKRAM_93C46_NVRAM_ADDRESS 0 +#define TEKRAM_24C16_NVRAM_ADDRESS 0x40 + +struct Tekram_nvram { + struct Tekram_target { + u_char flags; +#define TEKRAM_PARITY_CHECK (1) +#define TEKRAM_SYNC_NEGO (1<<1) +#define TEKRAM_DISCONNECT_ENABLE (1<<2) +#define TEKRAM_START_CMD (1<<3) +#define TEKRAM_TAGGED_COMMANDS (1<<4) +#define TEKRAM_WIDE_NEGO (1<<5) + u_char sync_index; + u_short word2; + } target[16]; + u_char host_id; + u_char flags; +#define TEKRAM_MORE_THAN_2_DRIVES (1) +#define TEKRAM_DRIVES_SUP_1GB (1<<1) +#define TEKRAM_RESET_ON_POWER_ON (1<<2) +#define TEKRAM_ACTIVE_NEGATION (1<<3) +#define TEKRAM_IMMEDIATE_SEEK (1<<4) +#define TEKRAM_SCAN_LUNS (1<<5) +#define TEKRAM_REMOVABLE_FLAGS (3<<6) /* 0: disable; 1: boot device; 2:all */ + u_char boot_delay_index; + u_char max_tags_index; + u_short flags1; +#define TEKRAM_F2_F6_ENABLED (1) + u_short spare[29]; +}; +typedef struct Tekram_nvram Tekram_nvram; +typedef struct Tekram_target Tekram_target; + +#endif /* SCSI_NCR_NVRAM_SUPPORT */ + +/**************** ORIGINAL CONTENT of ncrreg.h from FreeBSD ******************/ + +/*----------------------------------------------------------------- +** +** The ncr 53c810 register structure. +** +**----------------------------------------------------------------- +*/ + +struct ncr_reg { +/*00*/ u_char nc_scntl0; /* full arb., ena parity, par->ATN */ + +/*01*/ u_char nc_scntl1; /* no reset */ + #define ISCON 0x10 /* connected to scsi */ + #define CRST 0x08 /* force reset */ + #define IARB 0x02 /* immediate arbitration */ + +/*02*/ u_char nc_scntl2; /* no disconnect expected */ + #define SDU 0x80 /* cmd: disconnect will raise error */ + #define CHM 0x40 /* sta: chained mode */ + #define WSS 0x08 /* sta: wide scsi send [W]*/ + #define WSR 0x01 /* sta: wide scsi received [W]*/ + +/*03*/ u_char nc_scntl3; /* cnf system clock dependent */ + #define EWS 0x08 /* cmd: enable wide scsi [W]*/ + #define ULTRA 0x80 /* cmd: ULTRA enable */ + /* bits 0-2, 7 rsvd for C1010 */ + +/*04*/ u_char nc_scid; /* cnf host adapter scsi address */ + #define RRE 0x40 /* r/w:e enable response to resel. */ + #define SRE 0x20 /* r/w:e enable response to select */ + +/*05*/ u_char nc_sxfer; /* ### Sync speed and count */ + /* bits 6-7 rsvd for C1010 */ + +/*06*/ u_char nc_sdid; /* ### Destination-ID */ + +/*07*/ u_char nc_gpreg; /* ??? IO-Pins */ + +/*08*/ u_char nc_sfbr; /* ### First byte in phase */ + +/*09*/ u_char nc_socl; + #define CREQ 0x80 /* r/w: SCSI-REQ */ + #define CACK 0x40 /* r/w: SCSI-ACK */ + #define CBSY 0x20 /* r/w: SCSI-BSY */ + #define CSEL 0x10 /* r/w: SCSI-SEL */ + #define CATN 0x08 /* r/w: SCSI-ATN */ + #define CMSG 0x04 /* r/w: SCSI-MSG */ + #define CC_D 0x02 /* r/w: SCSI-C_D */ + #define CI_O 0x01 /* r/w: SCSI-I_O */ + +/*0a*/ u_char nc_ssid; + +/*0b*/ u_char nc_sbcl; + +/*0c*/ u_char nc_dstat; + #define DFE 0x80 /* sta: dma fifo empty */ + #define MDPE 0x40 /* int: master data parity error */ + #define BF 0x20 /* int: script: bus fault */ + #define ABRT 0x10 /* int: script: command aborted */ + #define SSI 0x08 /* int: script: single step */ + #define SIR 0x04 /* int: script: interrupt instruct. */ + #define IID 0x01 /* int: script: illegal instruct. */ + +/*0d*/ u_char nc_sstat0; + #define ILF 0x80 /* sta: data in SIDL register lsb */ + #define ORF 0x40 /* sta: data in SODR register lsb */ + #define OLF 0x20 /* sta: data in SODL register lsb */ + #define AIP 0x10 /* sta: arbitration in progress */ + #define LOA 0x08 /* sta: arbitration lost */ + #define WOA 0x04 /* sta: arbitration won */ + #define IRST 0x02 /* sta: scsi reset signal */ + #define SDP 0x01 /* sta: scsi parity signal */ + +/*0e*/ u_char nc_sstat1; + #define FF3210 0xf0 /* sta: bytes in the scsi fifo */ + +/*0f*/ u_char nc_sstat2; + #define ILF1 0x80 /* sta: data in SIDL register msb[W]*/ + #define ORF1 0x40 /* sta: data in SODR register msb[W]*/ + #define OLF1 0x20 /* sta: data in SODL register msb[W]*/ + #define DM 0x04 /* sta: DIFFSENS mismatch (895/6 only) */ + #define LDSC 0x02 /* sta: disconnect & reconnect */ + +/*10*/ u_char nc_dsa; /* --> Base page */ +/*11*/ u_char nc_dsa1; +/*12*/ u_char nc_dsa2; +/*13*/ u_char nc_dsa3; + +/*14*/ u_char nc_istat; /* --> Main Command and status */ + #define CABRT 0x80 /* cmd: abort current operation */ + #define SRST 0x40 /* mod: reset chip */ + #define SIGP 0x20 /* r/w: message from host to ncr */ + #define SEM 0x10 /* r/w: message between host + ncr */ + #define CON 0x08 /* sta: connected to scsi */ + #define INTF 0x04 /* sta: int on the fly (reset by wr)*/ + #define SIP 0x02 /* sta: scsi-interrupt */ + #define DIP 0x01 /* sta: host/script interrupt */ + +/*15*/ u_char nc_istat1; /* 896 only */ +/*16*/ u_char nc_mbox0; /* 896 only */ +/*17*/ u_char nc_mbox1; /* 896 only */ + +/*18*/ u_char nc_ctest0; +/*19*/ u_char nc_ctest1; + +/*1a*/ u_char nc_ctest2; + #define CSIGP 0x40 + /* bits 0-2,7 rsvd for C1010 */ + +/*1b*/ u_char nc_ctest3; + #define FLF 0x08 /* cmd: flush dma fifo */ + #define CLF 0x04 /* cmd: clear dma fifo */ + #define FM 0x02 /* mod: fetch pin mode */ + #define WRIE 0x01 /* mod: write and invalidate enable */ + /* bits 4-7 rsvd for C1010 */ + +/*1c*/ u_int32 nc_temp; /* ### Temporary stack */ + +/*20*/ u_char nc_dfifo; +/*21*/ u_char nc_ctest4; + #define BDIS 0x80 /* mod: burst disable */ + #define MPEE 0x08 /* mod: master parity error enable */ + +/*22*/ u_char nc_ctest5; + #define DFS 0x20 /* mod: dma fifo size */ + /* bits 0-1, 3-7 rsvd for C1010 */ +/*23*/ u_char nc_ctest6; + +/*24*/ u_int32 nc_dbc; /* ### Byte count and command */ +/*28*/ u_int32 nc_dnad; /* ### Next command register */ +/*2c*/ u_int32 nc_dsp; /* --> Script Pointer */ +/*30*/ u_int32 nc_dsps; /* --> Script pointer save/opcode#2 */ + +/*34*/ u_char nc_scratcha; /* Temporary register a */ +/*35*/ u_char nc_scratcha1; +/*36*/ u_char nc_scratcha2; +/*37*/ u_char nc_scratcha3; + +/*38*/ u_char nc_dmode; + #define BL_2 0x80 /* mod: burst length shift value +2 */ + #define BL_1 0x40 /* mod: burst length shift value +1 */ + #define ERL 0x08 /* mod: enable read line */ + #define ERMP 0x04 /* mod: enable read multiple */ + #define BOF 0x02 /* mod: burst op code fetch */ + +/*39*/ u_char nc_dien; +/*3a*/ u_char nc_sbr; + +/*3b*/ u_char nc_dcntl; /* --> Script execution control */ + #define CLSE 0x80 /* mod: cache line size enable */ + #define PFF 0x40 /* cmd: pre-fetch flush */ + #define PFEN 0x20 /* mod: pre-fetch enable */ + #define SSM 0x10 /* mod: single step mode */ + #define IRQM 0x08 /* mod: irq mode (1 = totem pole !) */ + #define STD 0x04 /* cmd: start dma mode */ + #define IRQD 0x02 /* mod: irq disable */ + #define NOCOM 0x01 /* cmd: protect sfbr while reselect */ + /* bits 0-1 rsvd for C1010 */ + +/*3c*/ u_int32 nc_adder; + +/*40*/ u_short nc_sien; /* -->: interrupt enable */ +/*42*/ u_short nc_sist; /* <--: interrupt status */ + #define SBMC 0x1000/* sta: SCSI Bus Mode Change (895/6 only) */ + #define STO 0x0400/* sta: timeout (select) */ + #define GEN 0x0200/* sta: timeout (general) */ + #define HTH 0x0100/* sta: timeout (handshake) */ + #define MA 0x80 /* sta: phase mismatch */ + #define CMP 0x40 /* sta: arbitration complete */ + #define SEL 0x20 /* sta: selected by another device */ + #define RSL 0x10 /* sta: reselected by another device*/ + #define SGE 0x08 /* sta: gross error (over/underflow)*/ + #define UDC 0x04 /* sta: unexpected disconnect */ + #define RST 0x02 /* sta: scsi bus reset detected */ + #define PAR 0x01 /* sta: scsi parity error */ + +/*44*/ u_char nc_slpar; +/*45*/ u_char nc_swide; +/*46*/ u_char nc_macntl; +/*47*/ u_char nc_gpcntl; +/*48*/ u_char nc_stime0; /* cmd: timeout for select&handshake*/ +/*49*/ u_char nc_stime1; /* cmd: timeout user defined */ +/*4a*/ u_short nc_respid; /* sta: Reselect-IDs */ + +/*4c*/ u_char nc_stest0; + +/*4d*/ u_char nc_stest1; + #define SCLK 0x80 /* Use the PCI clock as SCSI clock */ + #define DBLEN 0x08 /* clock doubler running */ + #define DBLSEL 0x04 /* clock doubler selected */ + + +/*4e*/ u_char nc_stest2; + #define ROF 0x40 /* reset scsi offset (after gross error!) */ + #define EXT 0x02 /* extended filtering */ + +/*4f*/ u_char nc_stest3; + #define TE 0x80 /* c: tolerAnt enable */ + #define HSC 0x20 /* c: Halt SCSI Clock */ + #define CSF 0x02 /* c: clear scsi fifo */ + +/*50*/ u_short nc_sidl; /* Lowlevel: latched from scsi data */ +/*52*/ u_char nc_stest4; + #define SMODE 0xc0 /* SCSI bus mode (895/6 only) */ + #define SMODE_HVD 0x40 /* High Voltage Differential */ + #define SMODE_SE 0x80 /* Single Ended */ + #define SMODE_LVD 0xc0 /* Low Voltage Differential */ + #define LCKFRQ 0x20 /* Frequency Lock (895/6 only) */ + /* bits 0-5 rsvd for C1010 */ + +/*53*/ u_char nc_53_; +/*54*/ u_short nc_sodl; /* Lowlevel: data out to scsi data */ +/*56*/ u_char nc_ccntl0; /* Chip Control 0 (896) */ + #define ENPMJ 0x80 /* Enable Phase Mismatch Jump */ + #define PMJCTL 0x40 /* Phase Mismatch Jump Control */ + #define ENNDJ 0x20 /* Enable Non Data PM Jump */ + #define DISFC 0x10 /* Disable Auto FIFO Clear */ + #define DILS 0x02 /* Disable Internal Load/Store */ + #define DPR 0x01 /* Disable Pipe Req */ + +/*57*/ u_char nc_ccntl1; /* Chip Control 1 (896) */ + #define ZMOD 0x80 /* High Impedance Mode */ + #define DIC 0x10 /* Disable Internal Cycles */ + #define DDAC 0x08 /* Disable Dual Address Cycle */ + #define XTIMOD 0x04 /* 64-bit Table Ind. Indexing Mode */ + #define EXTIBMV 0x02 /* Enable 64-bit Table Ind. BMOV */ + #define EXDBMV 0x01 /* Enable 64-bit Direct BMOV */ + +/*58*/ u_short nc_sbdl; /* Lowlevel: data from scsi data */ +/*5a*/ u_short nc_5a_; + +/*5c*/ u_char nc_scr0; /* Working register B */ +/*5d*/ u_char nc_scr1; /* */ +/*5e*/ u_char nc_scr2; /* */ +/*5f*/ u_char nc_scr3; /* */ + +/*60*/ u_char nc_scrx[64]; /* Working register C-R */ +/*a0*/ u_int32 nc_mmrs; /* Memory Move Read Selector */ +/*a4*/ u_int32 nc_mmws; /* Memory Move Write Selector */ +/*a8*/ u_int32 nc_sfs; /* Script Fetch Selector */ +/*ac*/ u_int32 nc_drs; /* DSA Relative Selector */ +/*b0*/ u_int32 nc_sbms; /* Static Block Move Selector */ +/*b4*/ u_int32 nc_dbms; /* Dynamic Block Move Selector */ +/*b8*/ u_int32 nc_dnad64; /* DMA Next Address 64 */ +/*bc*/ u_short nc_scntl4; /* C1010 only */ + #define U3EN 0x80 /* Enable Ultra 3 */ + #define AIPEN 0x40 /* Allow check upper byte lanes */ + #define XCLKH_DT 0x08 /* Extra clock of data hold on DT + transfer edge */ + #define XCLKH_ST 0x04 /* Extra clock of data hold on ST + transfer edge */ + +/*be*/ u_char nc_aipcntl0; /* Epat Control 1 C1010 only */ +/*bf*/ u_char nc_aipcntl1; /* AIP Control C1010_66 Only */ + +/*c0*/ u_int32 nc_pmjad1; /* Phase Mismatch Jump Address 1 */ +/*c4*/ u_int32 nc_pmjad2; /* Phase Mismatch Jump Address 2 */ +/*c8*/ u_char nc_rbc; /* Remaining Byte Count */ +/*c9*/ u_char nc_rbc1; /* */ +/*ca*/ u_char nc_rbc2; /* */ +/*cb*/ u_char nc_rbc3; /* */ + +/*cc*/ u_char nc_ua; /* Updated Address */ +/*cd*/ u_char nc_ua1; /* */ +/*ce*/ u_char nc_ua2; /* */ +/*cf*/ u_char nc_ua3; /* */ +/*d0*/ u_int32 nc_esa; /* Entry Storage Address */ +/*d4*/ u_char nc_ia; /* Instruction Address */ +/*d5*/ u_char nc_ia1; +/*d6*/ u_char nc_ia2; +/*d7*/ u_char nc_ia3; +/*d8*/ u_int32 nc_sbc; /* SCSI Byte Count (3 bytes only) */ +/*dc*/ u_int32 nc_csbc; /* Cumulative SCSI Byte Count */ + + /* Following for C1010 only */ +/*e0*/ u_short nc_crcpad; /* CRC Value */ +/*e2*/ u_char nc_crccntl0; /* CRC control register */ + #define SNDCRC 0x10 /* Send CRC Request */ +/*e3*/ u_char nc_crccntl1; /* CRC control register */ +/*e4*/ u_int32 nc_crcdata; /* CRC data register */ +/*e8*/ u_int32 nc_e8_; /* rsvd */ +/*ec*/ u_int32 nc_ec_; /* rsvd */ +/*f0*/ u_short nc_dfbc; /* DMA FIFO byte count */ + +}; + +/*----------------------------------------------------------- +** +** Utility macros for the script. +** +**----------------------------------------------------------- +*/ + +#define REGJ(p,r) (offsetof(struct ncr_reg, p ## r)) +#define REG(r) REGJ (nc_, r) + +typedef u_int32 ncrcmd; + +/*----------------------------------------------------------- +** +** SCSI phases +** +** DT phases illegal for ncr driver. +** +**----------------------------------------------------------- +*/ + +#define SCR_DATA_OUT 0x00000000 +#define SCR_DATA_IN 0x01000000 +#define SCR_COMMAND 0x02000000 +#define SCR_STATUS 0x03000000 +#define SCR_DT_DATA_OUT 0x04000000 +#define SCR_DT_DATA_IN 0x05000000 +#define SCR_MSG_OUT 0x06000000 +#define SCR_MSG_IN 0x07000000 + +#define SCR_ILG_OUT 0x04000000 +#define SCR_ILG_IN 0x05000000 + +/*----------------------------------------------------------- +** +** Data transfer via SCSI. +** +**----------------------------------------------------------- +** +** MOVE_ABS (LEN) +** <<start address>> +** +** MOVE_IND (LEN) +** <<dnad_offset>> +** +** MOVE_TBL +** <<dnad_offset>> +** +**----------------------------------------------------------- +*/ + +#define OPC_MOVE 0x08000000 + +#define SCR_MOVE_ABS(l) ((0x00000000 | OPC_MOVE) | (l)) +#define SCR_MOVE_IND(l) ((0x20000000 | OPC_MOVE) | (l)) +#define SCR_MOVE_TBL (0x10000000 | OPC_MOVE) + +#define SCR_CHMOV_ABS(l) ((0x00000000) | (l)) +#define SCR_CHMOV_IND(l) ((0x20000000) | (l)) +#define SCR_CHMOV_TBL (0x10000000) + +struct scr_tblmove { + u_int32 size; + u_int32 addr; +}; + +/*----------------------------------------------------------- +** +** Selection +** +**----------------------------------------------------------- +** +** SEL_ABS | SCR_ID (0..15) [ | REL_JMP] +** <<alternate_address>> +** +** SEL_TBL | << dnad_offset>> [ | REL_JMP] +** <<alternate_address>> +** +**----------------------------------------------------------- +*/ + +#define SCR_SEL_ABS 0x40000000 +#define SCR_SEL_ABS_ATN 0x41000000 +#define SCR_SEL_TBL 0x42000000 +#define SCR_SEL_TBL_ATN 0x43000000 + +struct scr_tblsel { + u_char sel_scntl4; + u_char sel_sxfer; + u_char sel_id; + u_char sel_scntl3; +}; + +#define SCR_JMP_REL 0x04000000 +#define SCR_ID(id) (((u_int32)(id)) << 16) + +/*----------------------------------------------------------- +** +** Waiting for Disconnect or Reselect +** +**----------------------------------------------------------- +** +** WAIT_DISC +** dummy: <<alternate_address>> +** +** WAIT_RESEL +** <<alternate_address>> +** +**----------------------------------------------------------- +*/ + +#define SCR_WAIT_DISC 0x48000000 +#define SCR_WAIT_RESEL 0x50000000 + +/*----------------------------------------------------------- +** +** Bit Set / Reset +** +**----------------------------------------------------------- +** +** SET (flags {|.. }) +** +** CLR (flags {|.. }) +** +**----------------------------------------------------------- +*/ + +#define SCR_SET(f) (0x58000000 | (f)) +#define SCR_CLR(f) (0x60000000 | (f)) + +#define SCR_CARRY 0x00000400 +#define SCR_TRG 0x00000200 +#define SCR_ACK 0x00000040 +#define SCR_ATN 0x00000008 + + + + +/*----------------------------------------------------------- +** +** Memory to memory move +** +**----------------------------------------------------------- +** +** COPY (bytecount) +** << source_address >> +** << destination_address >> +** +** SCR_COPY sets the NO FLUSH option by default. +** SCR_COPY_F does not set this option. +** +** For chips which do not support this option, +** ncr_copy_and_bind() will remove this bit. +**----------------------------------------------------------- +*/ + +#define SCR_NO_FLUSH 0x01000000 + +#define SCR_COPY(n) (0xc0000000 | SCR_NO_FLUSH | (n)) +#define SCR_COPY_F(n) (0xc0000000 | (n)) + +/*----------------------------------------------------------- +** +** Register move and binary operations +** +**----------------------------------------------------------- +** +** SFBR_REG (reg, op, data) reg = SFBR op data +** << 0 >> +** +** REG_SFBR (reg, op, data) SFBR = reg op data +** << 0 >> +** +** REG_REG (reg, op, data) reg = reg op data +** << 0 >> +** +**----------------------------------------------------------- +** On 810A, 860, 825A, 875, 895 and 896 chips the content +** of SFBR register can be used as data (SCR_SFBR_DATA). +** The 896 has additionnal IO registers starting at +** offset 0x80. Bit 7 of register offset is stored in +** bit 7 of the SCRIPTS instruction first DWORD. +**----------------------------------------------------------- +*/ + +#define SCR_REG_OFS(ofs) ((((ofs) & 0x7f) << 16ul) + ((ofs) & 0x80)) + +#define SCR_SFBR_REG(reg,op,data) \ + (0x68000000 | (SCR_REG_OFS(REG(reg))) | (op) | (((data)&0xff)<<8ul)) + +#define SCR_REG_SFBR(reg,op,data) \ + (0x70000000 | (SCR_REG_OFS(REG(reg))) | (op) | (((data)&0xff)<<8ul)) + +#define SCR_REG_REG(reg,op,data) \ + (0x78000000 | (SCR_REG_OFS(REG(reg))) | (op) | (((data)&0xff)<<8ul)) + + +#define SCR_LOAD 0x00000000 +#define SCR_SHL 0x01000000 +#define SCR_OR 0x02000000 +#define SCR_XOR 0x03000000 +#define SCR_AND 0x04000000 +#define SCR_SHR 0x05000000 +#define SCR_ADD 0x06000000 +#define SCR_ADDC 0x07000000 + +#define SCR_SFBR_DATA (0x00800000>>8ul) /* Use SFBR as data */ + +/*----------------------------------------------------------- +** +** FROM_REG (reg) SFBR = reg +** << 0 >> +** +** TO_REG (reg) reg = SFBR +** << 0 >> +** +** LOAD_REG (reg, data) reg = <data> +** << 0 >> +** +** LOAD_SFBR(data) SFBR = <data> +** << 0 >> +** +**----------------------------------------------------------- +*/ + +#define SCR_FROM_REG(reg) \ + SCR_REG_SFBR(reg,SCR_OR,0) + +#define SCR_TO_REG(reg) \ + SCR_SFBR_REG(reg,SCR_OR,0) + +#define SCR_LOAD_REG(reg,data) \ + SCR_REG_REG(reg,SCR_LOAD,data) + +#define SCR_LOAD_SFBR(data) \ + (SCR_REG_SFBR (gpreg, SCR_LOAD, data)) + +/*----------------------------------------------------------- +** +** LOAD from memory to register. +** STORE from register to memory. +** +** Only supported by 810A, 860, 825A, 875, 895 and 896. +** +**----------------------------------------------------------- +** +** LOAD_ABS (LEN) +** <<start address>> +** +** LOAD_REL (LEN) (DSA relative) +** <<dsa_offset>> +** +**----------------------------------------------------------- +*/ + +#define SCR_REG_OFS2(ofs) (((ofs) & 0xff) << 16ul) +#define SCR_NO_FLUSH2 0x02000000 +#define SCR_DSA_REL2 0x10000000 + +#define SCR_LOAD_R(reg, how, n) \ + (0xe1000000 | how | (SCR_REG_OFS2(REG(reg))) | (n)) + +#define SCR_STORE_R(reg, how, n) \ + (0xe0000000 | how | (SCR_REG_OFS2(REG(reg))) | (n)) + +#define SCR_LOAD_ABS(reg, n) SCR_LOAD_R(reg, SCR_NO_FLUSH2, n) +#define SCR_LOAD_REL(reg, n) SCR_LOAD_R(reg, SCR_NO_FLUSH2|SCR_DSA_REL2, n) +#define SCR_LOAD_ABS_F(reg, n) SCR_LOAD_R(reg, 0, n) +#define SCR_LOAD_REL_F(reg, n) SCR_LOAD_R(reg, SCR_DSA_REL2, n) + +#define SCR_STORE_ABS(reg, n) SCR_STORE_R(reg, SCR_NO_FLUSH2, n) +#define SCR_STORE_REL(reg, n) SCR_STORE_R(reg, SCR_NO_FLUSH2|SCR_DSA_REL2,n) +#define SCR_STORE_ABS_F(reg, n) SCR_STORE_R(reg, 0, n) +#define SCR_STORE_REL_F(reg, n) SCR_STORE_R(reg, SCR_DSA_REL2, n) + + +/*----------------------------------------------------------- +** +** Waiting for Disconnect or Reselect +** +**----------------------------------------------------------- +** +** JUMP [ | IFTRUE/IFFALSE ( ... ) ] +** <<address>> +** +** JUMPR [ | IFTRUE/IFFALSE ( ... ) ] +** <<distance>> +** +** CALL [ | IFTRUE/IFFALSE ( ... ) ] +** <<address>> +** +** CALLR [ | IFTRUE/IFFALSE ( ... ) ] +** <<distance>> +** +** RETURN [ | IFTRUE/IFFALSE ( ... ) ] +** <<dummy>> +** +** INT [ | IFTRUE/IFFALSE ( ... ) ] +** <<ident>> +** +** INT_FLY [ | IFTRUE/IFFALSE ( ... ) ] +** <<ident>> +** +** Conditions: +** WHEN (phase) +** IF (phase) +** CARRYSET +** DATA (data, mask) +** +**----------------------------------------------------------- +*/ + +#define SCR_NO_OP 0x80000000 +#define SCR_JUMP 0x80080000 +#define SCR_JUMP64 0x80480000 +#define SCR_JUMPR 0x80880000 +#define SCR_CALL 0x88080000 +#define SCR_CALLR 0x88880000 +#define SCR_RETURN 0x90080000 +#define SCR_INT 0x98080000 +#define SCR_INT_FLY 0x98180000 + +#define IFFALSE(arg) (0x00080000 | (arg)) +#define IFTRUE(arg) (0x00000000 | (arg)) + +#define WHEN(phase) (0x00030000 | (phase)) +#define IF(phase) (0x00020000 | (phase)) + +#define DATA(D) (0x00040000 | ((D) & 0xff)) +#define MASK(D,M) (0x00040000 | (((M ^ 0xff) & 0xff) << 8ul)|((D) & 0xff)) + +#define CARRYSET (0x00200000) + +/*----------------------------------------------------------- +** +** SCSI constants. +** +**----------------------------------------------------------- +*/ + +/* +** Messages +*/ + +#define M_COMPLETE (0x00) +#define M_EXTENDED (0x01) +#define M_SAVE_DP (0x02) +#define M_RESTORE_DP (0x03) +#define M_DISCONNECT (0x04) +#define M_ID_ERROR (0x05) +#define M_ABORT (0x06) +#define M_REJECT (0x07) +#define M_NOOP (0x08) +#define M_PARITY (0x09) +#define M_LCOMPLETE (0x0a) +#define M_FCOMPLETE (0x0b) +#define M_RESET (0x0c) +#define M_ABORT_TAG (0x0d) +#define M_CLEAR_QUEUE (0x0e) +#define M_INIT_REC (0x0f) +#define M_REL_REC (0x10) +#define M_TERMINATE (0x11) +#define M_SIMPLE_TAG (0x20) +#define M_HEAD_TAG (0x21) +#define M_ORDERED_TAG (0x22) +#define M_IGN_RESIDUE (0x23) +#define M_IDENTIFY (0x80) + +#define M_X_MODIFY_DP (0x00) +#define M_X_SYNC_REQ (0x01) +#define M_X_WIDE_REQ (0x03) +#define M_X_PPR_REQ (0x04) + +/* +** Status +*/ + +#define S_GOOD (0x00) +#define S_CHECK_COND (0x02) +#define S_COND_MET (0x04) +#define S_BUSY (0x08) +#define S_INT (0x10) +#define S_INT_COND_MET (0x14) +#define S_CONFLICT (0x18) +#define S_TERMINATED (0x20) +#define S_QUEUE_FULL (0x28) +#define S_ILLEGAL (0xff) +#define S_SENSE (0x80) + +/* + * End of ncrreg from FreeBSD + */ + +#endif /* !defined HOSTS_C */ + +#endif /* defined SYM53C8XX_DEFS_H */ diff --git a/linux/src/drivers/scsi/ultrastor.c b/linux/src/drivers/scsi/ultrastor.c index a1a1982d..de824726 100644 --- a/linux/src/drivers/scsi/ultrastor.c +++ b/linux/src/drivers/scsi/ultrastor.c @@ -333,7 +333,7 @@ static void log_ultrastor_abort(register struct ultrastor_config *config, { static char fmt[80] = "abort %d (%x); MSCP free pool: %x;"; register int i; - int flags; + unsigned long flags; save_flags(flags); cli(); @@ -681,7 +681,7 @@ int ultrastor_queuecommand(Scsi_Cmnd *SCpnt, void (*done)(Scsi_Cmnd *)) int mscp_index; #endif unsigned int status; - int flags; + unsigned long flags; /* Next test is for debugging; "can't happen" */ if ((config.mscp_free & ((1U << ULTRASTOR_MAX_CMDS) - 1)) == 0) @@ -853,7 +853,7 @@ int ultrastor_abort(Scsi_Cmnd *SCpnt) { int port0 = (config.slot << 12) | 0xc80; int i; - int flags; + unsigned long flags; save_flags(flags); cli(); strcpy(out, "OGM %d:%x ICM %d:%x ports: "); @@ -879,7 +879,7 @@ int ultrastor_abort(Scsi_Cmnd *SCpnt) if (config.slot ? inb(config.icm_address - 1) == 2 : (inb(SYS_DOORBELL_INTR(config.doorbell_address)) & 1)) { - int flags; + unsigned long flags; save_flags(flags); printk("Ux4F: abort while completed command pending\n"); restore_flags(flags); @@ -901,7 +901,7 @@ int ultrastor_abort(Scsi_Cmnd *SCpnt) and the interrupt handler will call done. */ if (config.slot && inb(config.ogm_address - 1) == 0) { - int flags; + unsigned long flags; save_flags(flags); cli(); @@ -953,7 +953,7 @@ int ultrastor_abort(Scsi_Cmnd *SCpnt) int ultrastor_reset(Scsi_Cmnd * SCpnt, unsigned int reset_flags) { - int flags; + unsigned long flags; register int i; #if (ULTRASTOR_DEBUG & UD_RESET) printk("US14F: reset: called\n"); diff --git a/linux/src/drivers/scsi/wd7000.c b/linux/src/drivers/scsi/wd7000.c index d910e27b..08d3ac3c 100644 --- a/linux/src/drivers/scsi/wd7000.c +++ b/linux/src/drivers/scsi/wd7000.c @@ -573,7 +573,7 @@ static int mail_out (Adapter *host, Scb *scbptr) break; } else - ogmb = (++ogmb) % OGMB_CNT; + ogmb = (ogmb + 1) % OGMB_CNT; } restore_flags (flags); diff --git a/linux/src/include/asm-i386/io.h b/linux/src/include/asm-i386/io.h index 98e32ce6..f961f1d2 100644 --- a/linux/src/include/asm-i386/io.h +++ b/linux/src/include/asm-i386/io.h @@ -25,6 +25,8 @@ * Linus */ +#include <machine/vm_param.h> + #ifdef SLOW_IO_BY_JUMPING #define __SLOW_DOWN_IO __asm__ __volatile__("jmp 1f\n1:\tjmp 1f\n1:") #else @@ -45,12 +47,12 @@ */ extern inline unsigned long virt_to_phys(volatile void * address) { - return (unsigned long) address; + return (unsigned long) _kvtophys(address); } extern inline void * phys_to_virt(unsigned long address) { - return (void *) address; + return (void *) phystokv(address); } /* diff --git a/linux/src/include/asm-i386/posix_types.h b/linux/src/include/asm-i386/posix_types.h index 712ef70c..6a04605a 100644 --- a/linux/src/include/asm-i386/posix_types.h +++ b/linux/src/include/asm-i386/posix_types.h @@ -15,8 +15,8 @@ typedef long __kernel_off_t; typedef int __kernel_pid_t; typedef unsigned short __kernel_uid_t; typedef unsigned short __kernel_gid_t; -typedef unsigned int __kernel_size_t; -typedef int __kernel_ssize_t; +typedef unsigned long __kernel_size_t; +typedef long __kernel_ssize_t; typedef int __kernel_ptrdiff_t; typedef long __kernel_time_t; typedef long __kernel_clock_t; diff --git a/linux/src/include/asm-i386/string-486.h b/linux/src/include/asm-i386/string-486.h deleted file mode 100644 index 1a337ba7..00000000 --- a/linux/src/include/asm-i386/string-486.h +++ /dev/null @@ -1,702 +0,0 @@ -#ifndef _I386_STRING_I486_H_ -#define _I386_STRING_I486_H_ - -/* - * This string-include defines all string functions as inline - * functions. Use gcc. It also assumes ds=es=data space, this should be - * normal. Most of the string-functions are rather heavily hand-optimized, - * see especially strtok,strstr,str[c]spn. They should work, but are not - * very easy to understand. Everything is done entirely within the register - * set, making the functions fast and clean. - * - * Copyright (C) 1991, 1992 Linus Torvalds - * Revised and optimized for i486/pentium - * 1994/03/15 by Alberto Vignani/Davide Parodi @crf.it - * - * Split into 2 CPU specific files by Alan Cox to keep #ifdef noise down. - */ - -#define __HAVE_ARCH_STRCPY -extern inline char * strcpy(char * dest,const char *src) -{ -register char *tmp= (char *)dest; -register char dummy; -__asm__ __volatile__( - "\n1:\t" - "movb (%0),%b2\n\t" - "incl %0\n\t" - "movb %b2,(%1)\n\t" - "incl %1\n\t" - "testb %b2,%b2\n\t" - "jne 1b" - :"=r" (src), "=r" (tmp), "=q" (dummy) - :"0" (src), "1" (tmp) - :"memory"); -return dest; -} - -#define __HAVE_ARCH_STRNCPY -extern inline char * strncpy(char * dest,const char *src,size_t count) -{ -register char *tmp= (char *)dest; -register char dummy; -if (count) { -__asm__ __volatile__( - "\n1:\t" - "movb (%0),%b2\n\t" - "incl %0\n\t" - "movb %b2,(%1)\n\t" - "incl %1\n\t" - "decl %3\n\t" - "je 3f\n\t" - "testb %b2,%b2\n\t" - "jne 1b\n\t" - "2:\tmovb %b2,(%1)\n\t" - "incl %1\n\t" - "decl %3\n\t" - "jne 2b\n\t" - "3:" - :"=r" (src), "=r" (tmp), "=q" (dummy), "=r" (count) - :"0" (src), "1" (tmp), "3" (count) - :"memory"); - } /* if (count) */ -return dest; -} - -#define __HAVE_ARCH_STRCAT -extern inline char * strcat(char * dest,const char * src) -{ -register char *tmp = (char *)(dest-1); -register char dummy; -__asm__ __volatile__( - "\n1:\tincl %1\n\t" - "cmpb $0,(%1)\n\t" - "jne 1b\n" - "2:\tmovb (%2),%b0\n\t" - "incl %2\n\t" - "movb %b0,(%1)\n\t" - "incl %1\n\t" - "testb %b0,%b0\n\t" - "jne 2b\n" - :"=q" (dummy), "=r" (tmp), "=r" (src) - :"1" (tmp), "2" (src) - :"memory"); -return dest; -} - -#define __HAVE_ARCH_STRNCAT -extern inline char * strncat(char * dest,const char * src,size_t count) -{ -register char *tmp = (char *)(dest-1); -register char dummy; -__asm__ __volatile__( - "\n1:\tincl %1\n\t" - "cmpb $0,(%1)\n\t" - "jne 1b\n" - "2:\tdecl %3\n\t" - "js 3f\n\t" - "movb (%2),%b0\n\t" - "incl %2\n\t" - "movb %b0,(%1)\n\t" - "incl %1\n\t" - "testb %b0,%b0\n\t" - "jne 2b\n" - "3:\txorb %b0,%b0\n\t" - "movb %b0,(%1)\n\t" - :"=q" (dummy), "=r" (tmp), "=r" (src), "=r" (count) - :"1" (tmp), "2" (src), "3" (count) - :"memory"); -return dest; -} - -#define __HAVE_ARCH_STRCMP -extern inline int strcmp(const char * cs,const char * ct) -{ -register int __res; -__asm__ __volatile__( - "\n1:\tmovb (%1),%b0\n\t" - "incl %1\n\t" - "cmpb %b0,(%2)\n\t" - "jne 2f\n\t" - "incl %2\n\t" - "testb %b0,%b0\n\t" - "jne 1b\n\t" - "xorl %k0,%k0\n\t" - "jmp 3f\n" - "2:\tmovl $1,%k0\n\t" - "jb 3f\n\t" - "negl %k0\n" - "3:" - :"=q" (__res), "=r" (cs), "=r" (ct) - :"1" (cs), "2" (ct) - : "memory" ); -return __res; -} - -#define __HAVE_ARCH_STRNCMP -extern inline int strncmp(const char * cs,const char * ct,size_t count) -{ -register int __res; -__asm__ __volatile__( - "\n1:\tdecl %3\n\t" - "js 2f\n\t" - "movb (%1),%b0\n\t" - "incl %1\n\t" - "cmpb %b0,(%2)\n\t" - "jne 3f\n\t" - "incl %2\n\t" - "testb %b0,%b0\n\t" - "jne 1b\n" - "2:\txorl %k0,%k0\n\t" - "jmp 4f\n" - "3:\tmovl $1,%k0\n\t" - "jb 4f\n\t" - "negl %k0\n" - "4:" - :"=q" (__res), "=r" (cs), "=r" (ct), "=r" (count) - :"1" (cs), "2" (ct), "3" (count)); -return __res; -} - -#define __HAVE_ARCH_STRCHR -extern inline char * strchr(const char * s, int c) -{ -register char * __res; -__asm__ __volatile__( - "movb %%al,%%ah\n" - "1:\tmovb (%1),%%al\n\t" - "cmpb %%ah,%%al\n\t" - "je 2f\n\t" - "incl %1\n\t" - "testb %%al,%%al\n\t" - "jne 1b\n\t" - "xorl %1,%1\n" - "2:\tmovl %1,%0\n\t" - :"=a" (__res), "=r" (s) - :"0" (c), "1" (s)); -return __res; -} - -#define __HAVE_ARCH_STRRCHR -extern inline char * strrchr(const char * s, int c) -{ -register char * __res; -__asm__ __volatile__( - "cld\n\t" - "movb %%al,%%ah\n" - "1:\tlodsb\n\t" - "cmpb %%ah,%%al\n\t" - "jne 2f\n\t" - "leal -1(%%esi),%0\n" - "2:\ttestb %%al,%%al\n\t" - "jne 1b" - :"=d" (__res):"0" (0),"S" (s),"a" (c):"ax","si"); -return __res; -} - -#define __HAVE_ARCH_STRSPN -extern inline size_t strspn(const char * cs, const char * ct) -{ -register char * __res; -__asm__ __volatile__( - "cld\n\t" - "movl %4,%%edi\n\t" - "repne\n\t" - "scasb\n\t" - "notl %%ecx\n\t" - "decl %%ecx\n\t" - "movl %%ecx,%%edx\n" - "1:\tlodsb\n\t" - "testb %%al,%%al\n\t" - "je 2f\n\t" - "movl %4,%%edi\n\t" - "movl %%edx,%%ecx\n\t" - "repne\n\t" - "scasb\n\t" - "je 1b\n" - "2:\tdecl %0" - :"=S" (__res):"a" (0),"c" (0xffffffff),"0" (cs),"g" (ct) - :"ax","cx","dx","di"); -return __res-cs; -} - -#define __HAVE_ARCH_STRCSPN -extern inline size_t strcspn(const char * cs, const char * ct) -{ -register char * __res; -__asm__ __volatile__( - "cld\n\t" - "movl %4,%%edi\n\t" - "repne\n\t" - "scasb\n\t" - "notl %%ecx\n\t" - "decl %%ecx\n\t" - "movl %%ecx,%%edx\n" - "1:\tlodsb\n\t" - "testb %%al,%%al\n\t" - "je 2f\n\t" - "movl %4,%%edi\n\t" - "movl %%edx,%%ecx\n\t" - "repne\n\t" - "scasb\n\t" - "jne 1b\n" - "2:\tdecl %0" - :"=S" (__res):"a" (0),"c" (0xffffffff),"0" (cs),"g" (ct) - :"ax","cx","dx","di"); -return __res-cs; -} - -#define __HAVE_ARCH_STRPBRK -extern inline char * strpbrk(const char * cs,const char * ct) -{ -register char * __res; -__asm__ __volatile__( - "cld\n\t" - "movl %4,%%edi\n\t" - "repne\n\t" - "scasb\n\t" - "notl %%ecx\n\t" - "decl %%ecx\n\t" - "movl %%ecx,%%edx\n" - "1:\tlodsb\n\t" - "testb %%al,%%al\n\t" - "je 2f\n\t" - "movl %4,%%edi\n\t" - "movl %%edx,%%ecx\n\t" - "repne\n\t" - "scasb\n\t" - "jne 1b\n\t" - "decl %0\n\t" - "jmp 3f\n" - "2:\txorl %0,%0\n" - "3:" - :"=S" (__res):"a" (0),"c" (0xffffffff),"0" (cs),"g" (ct) - :"ax","cx","dx","di"); -return __res; -} - -#define __HAVE_ARCH_STRSTR -extern inline char * strstr(const char * cs,const char * ct) -{ -register char * __res; -__asm__ __volatile__( - "cld\n\t" \ - "movl %4,%%edi\n\t" - "repne\n\t" - "scasb\n\t" - "notl %%ecx\n\t" - "decl %%ecx\n\t" /* NOTE! This also sets Z if searchstring='' */ - "movl %%ecx,%%edx\n" - "1:\tmovl %4,%%edi\n\t" - "movl %%esi,%%eax\n\t" - "movl %%edx,%%ecx\n\t" - "repe\n\t" - "cmpsb\n\t" - "je 2f\n\t" /* also works for empty string, see above */ - "xchgl %%eax,%%esi\n\t" - "incl %%esi\n\t" - "cmpb $0,-1(%%eax)\n\t" - "jne 1b\n\t" - "xorl %%eax,%%eax\n\t" - "2:" - :"=a" (__res):"0" (0),"c" (0xffffffff),"S" (cs),"g" (ct) - :"cx","dx","di","si"); -return __res; -} - -#define __HAVE_ARCH_STRLEN -extern inline size_t strlen(const char * s) -{ -/* - * slightly slower on a 486, but with better chances of - * register allocation - */ -register char dummy, *tmp= (char *)s; -__asm__ __volatile__( - "\n1:\t" - "movb\t(%0),%1\n\t" - "incl\t%0\n\t" - "testb\t%1,%1\n\t" - "jne\t1b" - :"=r" (tmp),"=q" (dummy) - :"0" (s) - : "memory" ); -return (tmp-s-1); -} - -/* Added by Gertjan van Wingerde to make minix and sysv module work */ -#define __HAVE_ARCH_STRNLEN -extern inline size_t strnlen(const char * s, size_t count) -{ -register int __res; -__asm__ __volatile__( - "movl %1,%0\n\t" - "jmp 2f\n" - "1:\tcmpb $0,(%0)\n\t" - "je 3f\n\t" - "incl %0\n" - "2:\tdecl %2\n\t" - "cmpl $-1,%2\n\t" - "jne 1b\n" - "3:\tsubl %1,%0" - :"=a" (__res) - :"c" (s),"d" (count) - :"dx"); -return __res; -} -/* end of additional stuff */ - -#define __HAVE_ARCH_STRTOK -extern inline char * strtok(char * s,const char * ct) -{ -register char * __res; -__asm__ __volatile__( - "testl %1,%1\n\t" - "jne 1f\n\t" - "testl %0,%0\n\t" - "je 8f\n\t" - "movl %0,%1\n" - "1:\txorl %0,%0\n\t" - "movl $-1,%%ecx\n\t" - "xorl %%eax,%%eax\n\t" - "cld\n\t" - "movl %4,%%edi\n\t" - "repne\n\t" - "scasb\n\t" - "notl %%ecx\n\t" - "decl %%ecx\n\t" - "je 7f\n\t" /* empty delimiter-string */ - "movl %%ecx,%%edx\n" - "2:\tlodsb\n\t" - "testb %%al,%%al\n\t" - "je 7f\n\t" - "movl %4,%%edi\n\t" - "movl %%edx,%%ecx\n\t" - "repne\n\t" - "scasb\n\t" - "je 2b\n\t" - "decl %1\n\t" - "cmpb $0,(%1)\n\t" - "je 7f\n\t" - "movl %1,%0\n" - "3:\tlodsb\n\t" - "testb %%al,%%al\n\t" - "je 5f\n\t" - "movl %4,%%edi\n\t" - "movl %%edx,%%ecx\n\t" - "repne\n\t" - "scasb\n\t" - "jne 3b\n\t" - "decl %1\n\t" - "cmpb $0,(%1)\n\t" - "je 5f\n\t" - "movb $0,(%1)\n\t" - "incl %1\n\t" - "jmp 6f\n" - "5:\txorl %1,%1\n" - "6:\tcmpb $0,(%0)\n\t" - "jne 7f\n\t" - "xorl %0,%0\n" - "7:\ttestl %0,%0\n\t" - "jne 8f\n\t" - "movl %0,%1\n" - "8:" - :"=b" (__res),"=S" (___strtok) - :"0" (___strtok),"1" (s),"g" (ct) - :"ax","cx","dx","di","memory"); -return __res; -} - -#define __memcpy_c(d,s,count) \ -((count%4==0) ? \ - __memcpy_by4((d),(s),(count)) : \ - ((count%2==0) ? \ - __memcpy_by2((d),(s),(count)) : \ - __memcpy_g((d),(s),(count)))) - -#define __HAVE_ARCH_MEMCPY -#define memcpy(d,s,count) \ -(count == 0 \ - ? d \ - : __builtin_constant_p(count) \ - ? __memcpy_c((d),(s),(count)) \ - : __memcpy_g((d),(s),(count))) - -/* - * These ought to get tweaked to do some cache priming. - */ - -extern inline void * __memcpy_by4(void * to, const void * from, size_t n) -{ -register void *tmp = (void *)to; -register int dummy1,dummy2; -__asm__ __volatile__ ( - "\n1:\tmovl (%2),%0\n\t" - "addl $4,%2\n\t" - "movl %0,(%1)\n\t" - "addl $4,%1\n\t" - "decl %3\n\t" - "jnz 1b" - :"=r" (dummy1), "=r" (tmp), "=r" (from), "=r" (dummy2) - :"1" (tmp), "2" (from), "3" (n/4) - :"memory"); -return to; -} - -extern inline void * __memcpy_by2(void * to, const void * from, size_t n) -{ -register void *tmp = (void *)to; -register int dummy1,dummy2; -__asm__ __volatile__ ( - "shrl $1,%3\n\t" - "jz 2f\n" /* only a word */ - "1:\tmovl (%2),%0\n\t" - "addl $4,%2\n\t" - "movl %0,(%1)\n\t" - "addl $4,%1\n\t" - "decl %3\n\t" - "jnz 1b\n" - "2:\tmovw (%2),%w0\n\t" - "movw %w0,(%1)" - :"=r" (dummy1), "=r" (tmp), "=r" (from), "=r" (dummy2) - :"1" (tmp), "2" (from), "3" (n/2) - :"memory"); -return to; -} - -extern inline void * __memcpy_g(void * to, const void * from, size_t n) -{ -register void *tmp = (void *)to; -__asm__ __volatile__ ( - "cld\n\t" - "shrl $1,%%ecx\n\t" - "jnc 1f\n\t" - "movsb\n" - "1:\tshrl $1,%%ecx\n\t" - "jnc 2f\n\t" - "movsw\n" - "2:\trep\n\t" - "movsl" - : /* no output */ - :"c" (n),"D" ((long) tmp),"S" ((long) from) - :"cx","di","si","memory"); -return to; -} - - -#define __HAVE_ARCH_MEMMOVE -extern inline void * memmove(void * dest,const void * src, size_t n) -{ -register void *tmp = (void *)dest; -if (dest<src) -__asm__ __volatile__ ( - "cld\n\t" - "rep\n\t" - "movsb" - : /* no output */ - :"c" (n),"S" (src),"D" (tmp) - :"cx","si","di"); -else -__asm__ __volatile__ ( - "std\n\t" - "rep\n\t" - "movsb\n\t" - "cld\n\t" - : /* no output */ - :"c" (n), "S" (n-1+(const char *)src), "D" (n-1+(char *)tmp) - :"cx","si","di","memory"); -return dest; -} - -extern inline int memcmp(const void * cs,const void * ct,size_t count) -{ -register int __res; -__asm__ __volatile__( - "cld\n\t" - "repe\n\t" - "cmpsb\n\t" - "je 1f\n\t" - "sbbl %0,%0\n\t" - "orb $1,%b0\n" - "1:" - :"=abd" (__res):"0" (0),"S" (cs),"D" (ct),"c" (count) - :"si","di","cx"); -return __res; -} - -#define __HAVE_ARCH_MEMCHR -extern inline void * memchr(const void * cs,int c,size_t count) -{ -register void * __res; -if (!count) - return NULL; -__asm__ __volatile__( - "cld\n\t" - "repne\n\t" - "scasb\n\t" - "je 1f\n\t" - "movl $1,%0\n" - "1:\tdecl %0" - :"=D" (__res):"a" (c),"D" (cs),"c" (count) - :"cx"); -return __res; -} - -#define __memset_cc(s,c,count) \ -((count%4==0) ? \ - __memset_cc_by4((s),(c),(count)) : \ - ((count%2==0) ? \ - __memset_cc_by2((s),(c),(count)) : \ - __memset_cg((s),(c),(count)))) - -#define __memset_gc(s,c,count) \ -((count%4==0) ? \ - __memset_gc_by4((s),(c),(count)) : \ - ((count%2==0) ? \ - __memset_gc_by2((s),(c),(count)) : \ - __memset_gg((s),(c),(count)))) - -#define __HAVE_ARCH_MEMSET -#define memset(s,c,count) \ -(count == 0 \ - ? s \ - : __builtin_constant_p(c) \ - ? __builtin_constant_p(count) \ - ? __memset_cc((s),(c),(count)) \ - : __memset_cg((s),(c),(count)) \ - : __builtin_constant_p(count) \ - ? __memset_gc((s),(c),(count)) \ - : __memset_gg((s),(c),(count))) - -extern inline void * __memset_cc_by4(void * s, char c, size_t count) -{ -/* - * register char *tmp = s; - */ -register char *tmp = (char *)s; -register int dummy; -__asm__ __volatile__ ( - "\n1:\tmovl %2,(%0)\n\t" - "addl $4,%0\n\t" - "decl %1\n\t" - "jnz 1b" - :"=r" (tmp), "=r" (dummy) - :"r" (0x01010101UL * (unsigned char) c), "0" (tmp), "1" (count/4) - :"memory"); -return s; -} - -extern inline void * __memset_cc_by2(void * s, char c, size_t count) -{ -register void *tmp = (void *)s; -register int dummy; -__asm__ __volatile__ ( - "shrl $1,%1\n\t" /* may be divisible also by 4 */ - "jz 2f\n" - "\n1:\tmovl %2,(%0)\n\t" - "addl $4,%0\n\t" - "decl %1\n\t" - "jnz 1b\n" - "2:\tmovw %w2,(%0)" - :"=r" (tmp), "=r" (dummy) - :"r" (0x01010101UL * (unsigned char) c), "0" (tmp), "1" (count/2) - :"memory"); -return s; -} - -extern inline void * __memset_gc_by4(void * s, char c, size_t count) -{ -register void *tmp = (void *)s; -register int dummy; -__asm__ __volatile__ ( - "movb %b0,%h0\n" - "pushw %w0\n\t" - "shll $16,%k0\n\t" - "popw %w0\n" - "1:\tmovl %k0,(%1)\n\t" - "addl $4,%1\n\t" - "decl %2\n\t" - "jnz 1b\n" - :"=q" (c), "=r" (tmp), "=r" (dummy) - :"0" ((unsigned) c), "1" (tmp), "2" (count/4) - :"memory"); -return s; -} - -extern inline void * __memset_gc_by2(void * s, char c, size_t count) -{ -register void *tmp = (void *)s; -register int dummy1,dummy2; -__asm__ __volatile__ ( - "movb %b0,%h0\n\t" - "shrl $1,%2\n\t" /* may be divisible also by 4 */ - "jz 2f\n\t" - "pushw %w0\n\t" - "shll $16,%k0\n\t" - "popw %w0\n" - "1:\tmovl %k0,(%1)\n\t" - "addl $4,%1\n\t" - "decl %2\n\t" - "jnz 1b\n" - "2:\tmovw %w0,(%1)" - :"=q" (dummy1), "=r" (tmp), "=r" (dummy2) - :"0" ((unsigned) c), "1" (tmp), "2" (count/2) - :"memory"); -return s; -} - -extern inline void * __memset_cg(void * s, char c, size_t count) -{ -register void *tmp = (void *)s; -__asm__ __volatile__ ( - "shrl $1,%%ecx\n\t" - "cld\n\t" - "rep\n\t" - "stosw\n\t" - "jnc 1f\n\t" - "movb %%al,(%%edi)\n" - "1:" - : /* no output */ - :"c" (count),"D" (tmp), "a" (0x0101U * (unsigned char) c) - :"cx","di","memory"); -return s; -} - -extern inline void * __memset_gg(void * s,char c,size_t count) -{ -register void *tmp = (void *)s; -__asm__ __volatile__ ( - "movb %%al,%%ah\n\t" - "shrl $1,%%ecx\n\t" - "cld\n\t" - "rep\n\t" - "stosw\n\t" - "jnc 1f\n\t" - "movb %%al,(%%edi)\n" - "1:" - : /* no output */ - :"c" (count),"D" (tmp), "a" (c) - :"cx","di","memory"); -return s; -} - - -/* - * find the first occurrence of byte 'c', or 1 past the area if none - */ -#define __HAVE_ARCH_MEMSCAN -extern inline void * memscan(void * addr, int c, size_t size) -{ - if (!size) - return addr; - __asm__("cld - repnz; scasb - jnz 1f - dec %%edi -1: " - : "=D" (addr), "=c" (size) - : "0" (addr), "1" (size), "a" (c)); - return addr; -} - -#endif diff --git a/linux/src/include/asm-i386/types.h b/linux/src/include/asm-i386/types.h index 71309ac8..d792546f 100644 --- a/linux/src/include/asm-i386/types.h +++ b/linux/src/include/asm-i386/types.h @@ -39,6 +39,8 @@ typedef unsigned int u32; typedef signed long long s64; typedef unsigned long long u64; +#define BITS_PER_LONG 32 + #endif /* __KERNEL__ */ #endif diff --git a/linux/src/include/linux/compatmac.h b/linux/src/include/linux/compatmac.h index b9a41127..95370702 100644 --- a/linux/src/include/linux/compatmac.h +++ b/linux/src/include/linux/compatmac.h @@ -47,6 +47,7 @@ #define COMPATMAC_H #include <linux/version.h> +#include <asm/io.h> #if LINUX_VERSION_CODE < 0x020100 /* Less than 2.1.0 */ #define TWO_ZERO @@ -96,11 +97,11 @@ static inline unsigned char get_irq (unsigned char bus, unsigned char fn) static inline void *ioremap(unsigned long base, long length) { - if (base < 0x100000) return (void *)base; + if (base < 0x100000) return phys_to_virt(base); return vremap (base, length); } -#define my_iounmap(x, b) (((long)x<0x100000)?0:vfree ((void*)x)) +#define my_iounmap(x, b) (((long)x<(long)phys_to_virt(0x100000))?0:vfree ((void*)x)) #define capable(x) suser() diff --git a/linux/src/include/linux/compiler-gcc.h b/linux/src/include/linux/compiler-gcc.h new file mode 100644 index 00000000..59e4028e --- /dev/null +++ b/linux/src/include/linux/compiler-gcc.h @@ -0,0 +1,106 @@ +#ifndef __LINUX_COMPILER_H +#error "Please don't include <linux/compiler-gcc.h> directly, include <linux/compiler.h> instead." +#endif + +/* + * Common definitions for all gcc versions go here. + */ + + +/* Optimization barrier */ +/* The "volatile" is due to gcc bugs */ +#define barrier() __asm__ __volatile__("": : :"memory") + +/* + * This macro obfuscates arithmetic on a variable address so that gcc + * shouldn't recognize the original var, and make assumptions about it. + * + * This is needed because the C standard makes it undefined to do + * pointer arithmetic on "objects" outside their boundaries and the + * gcc optimizers assume this is the case. In particular they + * assume such arithmetic does not wrap. + * + * A miscompilation has been observed because of this on PPC. + * To work around it we hide the relationship of the pointer and the object + * using this macro. + * + * Versions of the ppc64 compiler before 4.1 had a bug where use of + * RELOC_HIDE could trash r30. The bug can be worked around by changing + * the inline assembly constraint from =g to =r, in this particular + * case either is valid. + */ +#define RELOC_HIDE(ptr, off) \ + ({ unsigned long __ptr; \ + __asm__ ("" : "=r"(__ptr) : "0"(ptr)); \ + (typeof(ptr)) (__ptr + (off)); }) + +#ifdef __CHECKER__ +#define __must_be_array(arr) 0 +#else +/* &a[0] degrades to a pointer: a different type from an array */ +#define __must_be_array(a) BUILD_BUG_ON_ZERO(__same_type((a), &(a)[0])) +#endif + +/* + * Force always-inline if the user requests it so via the .config, + * or if gcc is too old: + */ +#if !defined(CONFIG_ARCH_SUPPORTS_OPTIMIZED_INLINING) || \ + !defined(CONFIG_OPTIMIZE_INLINING) || (__GNUC__ < 4) +# define inline inline __attribute__((always_inline)) +# define __inline__ __inline__ __attribute__((always_inline)) +# define __inline __inline __attribute__((always_inline)) +#endif + +#define __deprecated __attribute__((deprecated)) +#define __packed __attribute__((packed)) +#define __weak __attribute__((weak)) + +/* + * it doesn't make sense on ARM (currently the only user of __naked) to trace + * naked functions because then mcount is called without stack and frame pointer + * being set up and there is no chance to restore the lr register to the value + * before mcount was called. + * + * The asm() bodies of naked functions often depend on standard calling conventions, + * therefore they must be noinline and noclone. GCC 4.[56] currently fail to enforce + * this, so we must do so ourselves. See GCC PR44290. + */ +#define __naked __attribute__((naked)) noinline __noclone notrace + +#define __noreturn __attribute__((noreturn)) + +/* + * From the GCC manual: + * + * Many functions have no effects except the return value and their + * return value depends only on the parameters and/or global + * variables. Such a function can be subject to common subexpression + * elimination and loop optimization just as an arithmetic operator + * would be. + * [...] + */ +#define __pure __attribute__((pure)) +#define __aligned(x) __attribute__((aligned(x))) +#define __printf(a,b) __attribute__((format(printf,a,b))) +#define noinline __attribute__((noinline)) +#define __attribute_const__ __attribute__((__const__)) +#define __maybe_unused __attribute__((unused)) +#define __always_unused __attribute__((unused)) + +#define __gcc_header(x) #x +#define _gcc_header(x) __gcc_header(linux/compiler-gcc##x.h) +#define gcc_header(x) _gcc_header(x) +#include gcc_header(__GNUC__) + +#if !defined(__noclone) +#define __noclone /* not needed */ +#endif + +/* + * A trick to suppress uninitialized variable warning without generating any + * code + */ +#define uninitialized_var(x) x = x + +#define __always_inline inline __attribute__((always_inline)) diff --git a/linux/src/include/linux/compiler-gcc3.h b/linux/src/include/linux/compiler-gcc3.h new file mode 100644 index 00000000..37d41243 --- /dev/null +++ b/linux/src/include/linux/compiler-gcc3.h @@ -0,0 +1,23 @@ +#ifndef __LINUX_COMPILER_H +#error "Please don't include <linux/compiler-gcc3.h> directly, include <linux/compiler.h> instead." +#endif + +#if __GNUC_MINOR__ < 2 +# error Sorry, your compiler is too old - please upgrade it. +#endif + +#if __GNUC_MINOR__ >= 3 +# define __used __attribute__((__used__)) +#else +# define __used __attribute__((__unused__)) +#endif + +#if __GNUC_MINOR__ >= 4 +#define __must_check __attribute__((warn_unused_result)) +#endif + +#ifdef CONFIG_GCOV_KERNEL +# if __GNUC_MINOR__ < 4 +# error "GCOV profiling support for gcc versions below 3.4 not included" +# endif /* __GNUC_MINOR__ */ +#endif /* CONFIG_GCOV_KERNEL */ diff --git a/linux/src/include/linux/compiler-gcc4.h b/linux/src/include/linux/compiler-gcc4.h new file mode 100644 index 00000000..dfadc96e --- /dev/null +++ b/linux/src/include/linux/compiler-gcc4.h @@ -0,0 +1,57 @@ +#ifndef __LINUX_COMPILER_H +#error "Please don't include <linux/compiler-gcc4.h> directly, include <linux/compiler.h> instead." +#endif + +/* GCC 4.1.[01] miscompiles __weak */ +#ifdef __KERNEL__ +# if __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ <= 1 +# error Your version of gcc miscompiles the __weak directive +# endif +#endif + +#define __used __attribute__((__used__)) +#define __must_check __attribute__((warn_unused_result)) +#define __compiler_offsetof(a,b) __builtin_offsetof(a,b) + +#if __GNUC_MINOR__ >= 3 +/* Mark functions as cold. gcc will assume any path leading to a call + to them will be unlikely. This means a lot of manual unlikely()s + are unnecessary now for any paths leading to the usual suspects + like BUG(), printk(), panic() etc. [but let's keep them for now for + older compilers] + + Early snapshots of gcc 4.3 don't support this and we can't detect this + in the preprocessor, but we can live with this because they're unreleased. + Maketime probing would be overkill here. + + gcc also has a __attribute__((__hot__)) to move hot functions into + a special section, but I don't see any sense in this right now in + the kernel context */ +#define __cold __attribute__((__cold__)) + + +#if __GNUC_MINOR__ >= 5 +/* + * Mark a position in code as unreachable. This can be used to + * suppress control flow warnings after asm blocks that transfer + * control elsewhere. + * + * Early snapshots of gcc 4.5 don't support this and we can't detect + * this in the preprocessor, but we can live with this because they're + * unreleased. Really, we need to have autoconf for the kernel. + */ +#define unreachable() __builtin_unreachable() + +/* Mark a function definition as prohibited from being cloned. */ +#define __noclone __attribute__((__noclone__)) + +#endif +#endif + +#if __GNUC_MINOR__ > 0 +#define __compiletime_object_size(obj) __builtin_object_size(obj, 0) +#endif +#if __GNUC_MINOR__ >= 4 && !defined(__CHECKER__) +#define __compiletime_warning(message) __attribute__((warning(message))) +#define __compiletime_error(message) __attribute__((error(message))) +#endif diff --git a/linux/src/include/linux/compiler.h b/linux/src/include/linux/compiler.h new file mode 100644 index 00000000..320d6c94 --- /dev/null +++ b/linux/src/include/linux/compiler.h @@ -0,0 +1,311 @@ +#ifndef __LINUX_COMPILER_H +#define __LINUX_COMPILER_H + +#ifndef __ASSEMBLY__ + +#ifdef __CHECKER__ +# define __user __attribute__((noderef, address_space(1))) +# define __kernel __attribute__((address_space(0))) +# define __safe __attribute__((safe)) +# define __force __attribute__((force)) +# define __nocast __attribute__((nocast)) +# define __iomem __attribute__((noderef, address_space(2))) +# define __acquires(x) __attribute__((context(x,0,1))) +# define __releases(x) __attribute__((context(x,1,0))) +# define __acquire(x) __context__(x,1) +# define __release(x) __context__(x,-1) +# define __cond_lock(x,c) ((c) ? ({ __acquire(x); 1; }) : 0) +# define __percpu __attribute__((noderef, address_space(3))) +#ifdef CONFIG_SPARSE_RCU_POINTER +# define __rcu __attribute__((noderef, address_space(4))) +#else +# define __rcu +#endif +extern void __chk_user_ptr(const volatile void __user *); +extern void __chk_io_ptr(const volatile void __iomem *); +#else +# define __user +# define __kernel +# define __safe +# define __force +# define __nocast +# define __iomem +# define __chk_user_ptr(x) (void)0 +# define __chk_io_ptr(x) (void)0 +# define __builtin_warning(x, y...) (1) +# define __acquires(x) +# define __releases(x) +# define __acquire(x) (void)0 +# define __release(x) (void)0 +# define __cond_lock(x,c) (c) +# define __percpu +# define __rcu +#endif + +#ifdef __KERNEL__ + +#ifdef __GNUC__ +#include <linux/compiler-gcc.h> +#endif + +#define notrace __attribute__((no_instrument_function)) + +/* Intel compiler defines __GNUC__. So we will overwrite implementations + * coming from above header files here + */ +#ifdef __INTEL_COMPILER +# include <linux/compiler-intel.h> +#endif + +/* + * Generic compiler-dependent macros required for kernel + * build go below this comment. Actual compiler/compiler version + * specific implementations come from the above header files + */ + +struct ftrace_branch_data { + const char *func; + const char *file; + unsigned line; + union { + struct { + unsigned long correct; + unsigned long incorrect; + }; + struct { + unsigned long miss; + unsigned long hit; + }; + unsigned long miss_hit[2]; + }; +}; + +/* + * Note: DISABLE_BRANCH_PROFILING can be used by special lowlevel code + * to disable branch tracing on a per file basis. + */ +#if defined(CONFIG_TRACE_BRANCH_PROFILING) \ + && !defined(DISABLE_BRANCH_PROFILING) && !defined(__CHECKER__) +void ftrace_likely_update(struct ftrace_branch_data *f, int val, int expect); + +#define likely_notrace(x) __builtin_expect(!!(x), 1) +#define unlikely_notrace(x) __builtin_expect(!!(x), 0) + +#define __branch_check__(x, expect) ({ \ + int ______r; \ + static struct ftrace_branch_data \ + __attribute__((__aligned__(4))) \ + __attribute__((section("_ftrace_annotated_branch"))) \ + ______f = { \ + .func = __func__, \ + .file = __FILE__, \ + .line = __LINE__, \ + }; \ + ______r = likely_notrace(x); \ + ftrace_likely_update(&______f, ______r, expect); \ + ______r; \ + }) + +/* + * Using __builtin_constant_p(x) to ignore cases where the return + * value is always the same. This idea is taken from a similar patch + * written by Daniel Walker. + */ +# ifndef likely +# define likely(x) (__builtin_constant_p(x) ? !!(x) : __branch_check__(x, 1)) +# endif +# ifndef unlikely +# define unlikely(x) (__builtin_constant_p(x) ? !!(x) : __branch_check__(x, 0)) +# endif + +#ifdef CONFIG_PROFILE_ALL_BRANCHES +/* + * "Define 'is'", Bill Clinton + * "Define 'if'", Steven Rostedt + */ +#define if(cond, ...) __trace_if( (cond , ## __VA_ARGS__) ) +#define __trace_if(cond) \ + if (__builtin_constant_p((cond)) ? !!(cond) : \ + ({ \ + int ______r; \ + static struct ftrace_branch_data \ + __attribute__((__aligned__(4))) \ + __attribute__((section("_ftrace_branch"))) \ + ______f = { \ + .func = __func__, \ + .file = __FILE__, \ + .line = __LINE__, \ + }; \ + ______r = !!(cond); \ + ______f.miss_hit[______r]++; \ + ______r; \ + })) +#endif /* CONFIG_PROFILE_ALL_BRANCHES */ + +#else +# define likely(x) __builtin_expect(!!(x), 1) +# define unlikely(x) __builtin_expect(!!(x), 0) +#endif + +/* Optimization barrier */ +#ifndef barrier +# define barrier() __memory_barrier() +#endif + +/* Unreachable code */ +#ifndef unreachable +# define unreachable() do { } while (1) +#endif + +#ifndef RELOC_HIDE +# define RELOC_HIDE(ptr, off) \ + ({ unsigned long __ptr; \ + __ptr = (unsigned long) (ptr); \ + (typeof(ptr)) (__ptr + (off)); }) +#endif + +#endif /* __KERNEL__ */ + +#endif /* __ASSEMBLY__ */ + +#ifdef __KERNEL__ +/* + * Allow us to mark functions as 'deprecated' and have gcc emit a nice + * warning for each use, in hopes of speeding the functions removal. + * Usage is: + * int __deprecated foo(void) + */ +#ifndef __deprecated +# define __deprecated /* unimplemented */ +#endif + +#ifdef MODULE +#define __deprecated_for_modules __deprecated +#else +#define __deprecated_for_modules +#endif + +#ifndef __must_check +#define __must_check +#endif + +#ifndef CONFIG_ENABLE_MUST_CHECK +#undef __must_check +#define __must_check +#endif +#ifndef CONFIG_ENABLE_WARN_DEPRECATED +#undef __deprecated +#undef __deprecated_for_modules +#define __deprecated +#define __deprecated_for_modules +#endif + +/* + * Allow us to avoid 'defined but not used' warnings on functions and data, + * as well as force them to be emitted to the assembly file. + * + * As of gcc 3.4, static functions that are not marked with attribute((used)) + * may be elided from the assembly file. As of gcc 3.4, static data not so + * marked will not be elided, but this may change in a future gcc version. + * + * NOTE: Because distributions shipped with a backported unit-at-a-time + * compiler in gcc 3.3, we must define __used to be __attribute__((used)) + * for gcc >=3.3 instead of 3.4. + * + * In prior versions of gcc, such functions and data would be emitted, but + * would be warned about except with attribute((unused)). + * + * Mark functions that are referenced only in inline assembly as __used so + * the code is emitted even though it appears to be unreferenced. + */ +#ifndef __used +# define __used /* unimplemented */ +#endif + +#ifndef __maybe_unused +# define __maybe_unused /* unimplemented */ +#endif + +#ifndef __always_unused +# define __always_unused /* unimplemented */ +#endif + +#ifndef noinline +#define noinline +#endif + +/* + * Rather then using noinline to prevent stack consumption, use + * noinline_for_stack instead. For documentaiton reasons. + */ +#define noinline_for_stack noinline + +#ifndef __always_inline +#define __always_inline inline +#endif + +#endif /* __KERNEL__ */ + +/* + * From the GCC manual: + * + * Many functions do not examine any values except their arguments, + * and have no effects except the return value. Basically this is + * just slightly more strict class than the `pure' attribute above, + * since function is not allowed to read global memory. + * + * Note that a function that has pointer arguments and examines the + * data pointed to must _not_ be declared `const'. Likewise, a + * function that calls a non-`const' function usually must not be + * `const'. It does not make sense for a `const' function to return + * `void'. + */ +#ifndef __attribute_const__ +# define __attribute_const__ /* unimplemented */ +#endif + +/* + * Tell gcc if a function is cold. The compiler will assume any path + * directly leading to the call is unlikely. + */ + +#ifndef __cold +#define __cold +#endif + +/* Simple shorthand for a section definition */ +#ifndef __section +# define __section(S) __attribute__ ((__section__(#S))) +#endif + +/* Are two types/vars the same type (ignoring qualifiers)? */ +#ifndef __same_type +# define __same_type(a, b) __builtin_types_compatible_p(typeof(a), typeof(b)) +#endif + +/* Compile time object size, -1 for unknown */ +#ifndef __compiletime_object_size +# define __compiletime_object_size(obj) -1 +#endif +#ifndef __compiletime_warning +# define __compiletime_warning(message) +#endif +#ifndef __compiletime_error +# define __compiletime_error(message) +#endif + +/* + * Prevent the compiler from merging or refetching accesses. The compiler + * is also forbidden from reordering successive instances of ACCESS_ONCE(), + * but only when the compiler is aware of some particular ordering. One way + * to make the compiler aware of ordering is to put the two invocations of + * ACCESS_ONCE() in different C statements. + * + * This macro does absolutely -nothing- to prevent the CPU from reordering, + * merging, or refetching absolutely anything at any time. Its main intended + * use is to mediate communication between process-level code and irq/NMI + * handlers, all running on the same CPU. + */ +#define ACCESS_ONCE(x) (*(volatile typeof(x) *)&(x)) + +#endif /* __LINUX_COMPILER_H */ diff --git a/linux/src/include/linux/ctype.h b/linux/src/include/linux/ctype.h index 838ef933..8acfe312 100644 --- a/linux/src/include/linux/ctype.h +++ b/linux/src/include/linux/ctype.h @@ -1,6 +1,11 @@ #ifndef _LINUX_CTYPE_H #define _LINUX_CTYPE_H +/* + * NOTE! This ctype does not handle EOF like the standard C + * library is required to. + */ + #define _U 0x01 /* upper */ #define _L 0x02 /* lower */ #define _D 0x04 /* digit */ @@ -10,25 +15,50 @@ #define _X 0x40 /* hex digit */ #define _SP 0x80 /* hard space (0x20) */ -extern unsigned char _ctype[]; -extern char _ctmp; - -#define isalnum(c) ((_ctype+1)[c]&(_U|_L|_D)) -#define isalpha(c) ((_ctype+1)[c]&(_U|_L)) -#define iscntrl(c) ((_ctype+1)[c]&(_C)) -#define isdigit(c) ((_ctype+1)[c]&(_D)) -#define isgraph(c) ((_ctype+1)[c]&(_P|_U|_L|_D)) -#define islower(c) ((_ctype+1)[c]&(_L)) -#define isprint(c) ((_ctype+1)[c]&(_P|_U|_L|_D|_SP)) -#define ispunct(c) ((_ctype+1)[c]&(_P)) -#define isspace(c) ((_ctype+1)[c]&(_S)) -#define isupper(c) ((_ctype+1)[c]&(_U)) -#define isxdigit(c) ((_ctype+1)[c]&(_D|_X)) - -#define isascii(c) (((unsigned) c)<=0x7f) -#define toascii(c) (((unsigned) c)&0x7f) - -#define tolower(c) (_ctmp=c,isupper(_ctmp)?_ctmp-('A'-'a'):_ctmp) -#define toupper(c) (_ctmp=c,islower(_ctmp)?_ctmp-('a'-'A'):_ctmp) +extern const unsigned char _ctype[]; + +#define __ismask(x) (_ctype[(int)(unsigned char)(x)]) + +#define isalnum(c) ((__ismask(c)&(_U|_L|_D)) != 0) +#define isalpha(c) ((__ismask(c)&(_U|_L)) != 0) +#define iscntrl(c) ((__ismask(c)&(_C)) != 0) +#define isdigit(c) ((__ismask(c)&(_D)) != 0) +#define isgraph(c) ((__ismask(c)&(_P|_U|_L|_D)) != 0) +#define islower(c) ((__ismask(c)&(_L)) != 0) +#define isprint(c) ((__ismask(c)&(_P|_U|_L|_D|_SP)) != 0) +#define ispunct(c) ((__ismask(c)&(_P)) != 0) +/* Note: isspace() must return false for %NUL-terminator */ +#define isspace(c) ((__ismask(c)&(_S)) != 0) +#define isupper(c) ((__ismask(c)&(_U)) != 0) +#define isxdigit(c) ((__ismask(c)&(_D|_X)) != 0) + +#define isascii(c) (((unsigned char)(c))<=0x7f) +#define toascii(c) (((unsigned char)(c))&0x7f) + +static inline unsigned char __tolower(unsigned char c) +{ + if (isupper(c)) + c -= 'A'-'a'; + return c; +} + +static inline unsigned char __toupper(unsigned char c) +{ + if (islower(c)) + c -= 'a'-'A'; + return c; +} + +#define tolower(c) __tolower(c) +#define toupper(c) __toupper(c) + +/* + * Fast implementation of tolower() for internal usage. Do not use in your + * code. + */ +static inline char _tolower(const char c) +{ + return c | 0x20; +} #endif diff --git a/linux/src/include/linux/hdreg.h b/linux/src/include/linux/hdreg.h index 2645deea..e223480d 100644 --- a/linux/src/include/linux/hdreg.h +++ b/linux/src/include/linux/hdreg.h @@ -92,11 +92,12 @@ struct hd_geometry { #define HDIO_GETGEO 0x0301 /* get device geometry */ #define HDIO_GET_UNMASKINTR 0x0302 /* get current unmask setting */ #define HDIO_GET_MULTCOUNT 0x0304 /* get current IDE blockmode setting */ -#define HDIO_GET_IDENTITY 0x0307 /* get IDE identification info */ +#define HDIO_OBSOLETE_IDENTITY 0x0307 /* OBSOLETE, DO NOT USE: returns 142 bytes */ #define HDIO_GET_KEEPSETTINGS 0x0308 /* get keep-settings-on-reset flag */ #define HDIO_GET_32BIT 0x0309 /* get current io_32bit setting */ #define HDIO_GET_NOWERR 0x030a /* get ignore-write-error flag */ #define HDIO_GET_DMA 0x030b /* get use-dma flag */ +#define HDIO_GET_IDENTITY 0x030d /* get IDE identification info */ #define HDIO_DRIVE_CMD 0x031f /* execute a special drive command */ /* hd/ide ctl's that pass (arg) non-ptr values are numbered 0x032n/0x033n */ @@ -166,14 +167,54 @@ struct hd_driveid { unsigned short word79; unsigned short word80; unsigned short word81; - unsigned short word82; - unsigned short word83; + unsigned short command_sets; /* bits 0:Smart 1:Security 2:Removable 3:PM */ + unsigned short word83; /* bits 14:Smart Enabled 13:0 zero */ unsigned short word84; unsigned short word85; unsigned short word86; unsigned short word87; unsigned short dma_ultra; - unsigned short reserved[167]; + unsigned short word89; /* reserved (word 89) */ + unsigned short word90; /* reserved (word 90) */ + unsigned short word91; /* reserved (word 91) */ + unsigned short word92; /* reserved (word 92) */ + unsigned short word93; /* reserved (word 93) */ + unsigned short word94; /* reserved (word 94) */ + unsigned short word95; /* reserved (word 95) */ + unsigned short word96; /* reserved (word 96) */ + unsigned short word97; /* reserved (word 97) */ + unsigned short word98; /* reserved (word 98) */ + unsigned short word99; /* reserved (word 99) */ + unsigned short word100; /* reserved (word 100) */ + unsigned short word101; /* reserved (word 101) */ + unsigned short word102; /* reserved (word 102) */ + unsigned short word103; /* reserved (word 103) */ + unsigned short word104; /* reserved (word 104) */ + unsigned short word105; /* reserved (word 105) */ + unsigned short word106; /* reserved (word 106) */ + unsigned short word107; /* reserved (word 107) */ + unsigned short word108; /* reserved (word 108) */ + unsigned short word109; /* reserved (word 109) */ + unsigned short word110; /* reserved (word 110) */ + unsigned short word111; /* reserved (word 111) */ + unsigned short word112; /* reserved (word 112) */ + unsigned short word113; /* reserved (word 113) */ + unsigned short word114; /* reserved (word 114) */ + unsigned short word115; /* reserved (word 115) */ + unsigned short word116; /* reserved (word 116) */ + unsigned short word117; /* reserved (word 117) */ + unsigned short word118; /* reserved (word 118) */ + unsigned short word119; /* reserved (word 119) */ + unsigned short word120; /* reserved (word 120) */ + unsigned short word121; /* reserved (word 121) */ + unsigned short word122; /* reserved (word 122) */ + unsigned short word123; /* reserved (word 123) */ + unsigned short word124; /* reserved (word 124) */ + unsigned short word125; /* reserved (word 125) */ + unsigned short word126; /* reserved (word 126) */ + unsigned short word127; /* reserved (word 127) */ + unsigned short security; /* bits 0:suuport 1:enabled 2:locked 3:frozen */ + unsigned short reserved[127]; }; #ifdef __KERNEL__ diff --git a/linux/pcmcia-cs/include/linux/init.h b/linux/src/include/linux/init.h index 06da72d0..d4798b25 100644 --- a/linux/pcmcia-cs/include/linux/init.h +++ b/linux/src/include/linux/init.h @@ -1,9 +1,17 @@ #ifndef _COMPAT_INIT_H #define _COMPAT_INIT_H +#include <linux/compiler.h> + +#ifdef MODULE +#define __exitused +#else +#define __exitused __used +#endif + #define __init #define __initdata -#define __exit +#define __exit __exitused __cold notrace #define __exitdata #define __devinit #define __devinitdata diff --git a/linux/src/include/linux/interrupt.h b/linux/src/include/linux/interrupt.h index 8654a4f2..5765260d 100644 --- a/linux/src/include/linux/interrupt.h +++ b/linux/src/include/linux/interrupt.h @@ -14,11 +14,11 @@ struct irqaction { struct irqaction *next; }; -extern unsigned long intr_count; +extern unsigned int intr_count; extern int bh_mask_count[32]; -extern unsigned long bh_active; -extern unsigned long bh_mask; +extern unsigned int bh_active; +extern unsigned int bh_mask; extern void (*bh_base[32])(void); asmlinkage void do_bottom_half(void); diff --git a/linux/src/include/linux/kcomp.h b/linux/src/include/linux/kcomp.h index 1f7344a6..5e06d7e7 100644 --- a/linux/src/include/linux/kcomp.h +++ b/linux/src/include/linux/kcomp.h @@ -7,8 +7,6 @@ #include <linux/netdevice.h> #include <linux/pagemap.h> -#define __exit - #define pci_enable_device(x) 0 #define page_address(x) (x | PAGE_OFFSET) diff --git a/linux/src/include/linux/module.h b/linux/src/include/linux/module.h index bc364f79..acc25403 100644 --- a/linux/src/include/linux/module.h +++ b/linux/src/include/linux/module.h @@ -101,9 +101,12 @@ int Using_Versions; /* gcc will handle this global (used as a flag) correctly */ #else +#define EXPORT_SYMBOL(sym) + #define MOD_INC_USE_COUNT do { } while (0) #define MOD_DEC_USE_COUNT do { } while (0) #define MOD_IN_USE 1 +#define SET_MODULE_OWNER(dev) do{ } while(0) #endif diff --git a/linux/src/include/linux/stddef.h b/linux/src/include/linux/stddef.h index c6221e71..488d49c0 100644 --- a/linux/src/include/linux/stddef.h +++ b/linux/src/include/linux/stddef.h @@ -3,7 +3,7 @@ #ifndef _SIZE_T #define _SIZE_T -typedef unsigned int size_t; +typedef unsigned long size_t; #endif #undef NULL diff --git a/linux/src/init/main.c b/linux/src/init/main.c index d41ec60c..1aa15b97 100644 --- a/linux/src/init/main.c +++ b/linux/src/init/main.c @@ -42,6 +42,8 @@ #include <asm/bugs.h> +#include <linux/dev/glue/glue.h> + /* * Versions of gcc older than that listed below may actually compile * and link okay, but the end product can have subtle run time bugs. @@ -64,7 +66,6 @@ extern int bdflush(void *); extern int kswapd(void *); extern void kswapd_setup(void); -extern void init_IRQ(void); extern void init_modules(void); extern long console_init(long, long); extern long kmalloc_init(long,long); diff --git a/linux/src/kernel/softirq.c b/linux/src/kernel/softirq.c index 022b5535..32038b15 100644 --- a/linux/src/kernel/softirq.c +++ b/linux/src/kernel/softirq.c @@ -21,18 +21,18 @@ #include <asm/irq.h> #include <asm/bitops.h> -unsigned long intr_count = 0; +unsigned int intr_count = 0; int bh_mask_count[32]; -unsigned long bh_active = 0; -unsigned long bh_mask = 0; +unsigned int bh_active = 0; +unsigned int bh_mask = 0; void (*bh_base[32])(void); asmlinkage void do_bottom_half(void) { - unsigned long active; - unsigned long mask, left; + unsigned int active; + unsigned int mask, left; void (**bh)(void); sti(); diff --git a/linux/src/lib/ctype.c b/linux/src/lib/ctype.c index c5be2641..26baa620 100644 --- a/linux/src/lib/ctype.c +++ b/linux/src/lib/ctype.c @@ -5,32 +5,32 @@ */ #include <linux/ctype.h> +#include <linux/module.h> -char _ctmp; -unsigned char _ctype[] = {0x00, /* EOF */ -_C,_C,_C,_C,_C,_C,_C,_C, /* 0-7 */ -_C,_C|_S,_C|_S,_C|_S,_C|_S,_C|_S,_C,_C, /* 8-15 */ -_C,_C,_C,_C,_C,_C,_C,_C, /* 16-23 */ -_C,_C,_C,_C,_C,_C,_C,_C, /* 24-31 */ -_S|_SP,_P,_P,_P,_P,_P,_P,_P, /* 32-39 */ -_P,_P,_P,_P,_P,_P,_P,_P, /* 40-47 */ -_D,_D,_D,_D,_D,_D,_D,_D, /* 48-55 */ -_D,_D,_P,_P,_P,_P,_P,_P, /* 56-63 */ -_P,_U|_X,_U|_X,_U|_X,_U|_X,_U|_X,_U|_X,_U, /* 64-71 */ -_U,_U,_U,_U,_U,_U,_U,_U, /* 72-79 */ -_U,_U,_U,_U,_U,_U,_U,_U, /* 80-87 */ -_U,_U,_U,_P,_P,_P,_P,_P, /* 88-95 */ -_P,_L|_X,_L|_X,_L|_X,_L|_X,_L|_X,_L|_X,_L, /* 96-103 */ -_L,_L,_L,_L,_L,_L,_L,_L, /* 104-111 */ -_L,_L,_L,_L,_L,_L,_L,_L, /* 112-119 */ -_L,_L,_L,_P,_P,_P,_P,_C, /* 120-127 */ -0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 128-143 */ -0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 144-159 */ -_S|_SP,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P, /* 160-175 */ -_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P, /* 176-191 */ -_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U, /* 192-207 */ -_U,_U,_U,_U,_U,_U,_U,_P,_U,_U,_U,_U,_U,_U,_U,_L, /* 208-223 */ -_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L, /* 224-239 */ -_L,_L,_L,_L,_L,_L,_L,_P,_L,_L,_L,_L,_L,_L,_L,_L}; /* 240-255 */ - +const unsigned char _ctype[] = { +_C,_C,_C,_C,_C,_C,_C,_C, /* 0-7 */ +_C,_C|_S,_C|_S,_C|_S,_C|_S,_C|_S,_C,_C, /* 8-15 */ +_C,_C,_C,_C,_C,_C,_C,_C, /* 16-23 */ +_C,_C,_C,_C,_C,_C,_C,_C, /* 24-31 */ +_S|_SP,_P,_P,_P,_P,_P,_P,_P, /* 32-39 */ +_P,_P,_P,_P,_P,_P,_P,_P, /* 40-47 */ +_D,_D,_D,_D,_D,_D,_D,_D, /* 48-55 */ +_D,_D,_P,_P,_P,_P,_P,_P, /* 56-63 */ +_P,_U|_X,_U|_X,_U|_X,_U|_X,_U|_X,_U|_X,_U, /* 64-71 */ +_U,_U,_U,_U,_U,_U,_U,_U, /* 72-79 */ +_U,_U,_U,_U,_U,_U,_U,_U, /* 80-87 */ +_U,_U,_U,_P,_P,_P,_P,_P, /* 88-95 */ +_P,_L|_X,_L|_X,_L|_X,_L|_X,_L|_X,_L|_X,_L, /* 96-103 */ +_L,_L,_L,_L,_L,_L,_L,_L, /* 104-111 */ +_L,_L,_L,_L,_L,_L,_L,_L, /* 112-119 */ +_L,_L,_L,_P,_P,_P,_P,_C, /* 120-127 */ +0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 128-143 */ +0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 144-159 */ +_S|_SP,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P, /* 160-175 */ +_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P,_P, /* 176-191 */ +_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U,_U, /* 192-207 */ +_U,_U,_U,_U,_U,_U,_U,_P,_U,_U,_U,_U,_U,_U,_U,_L, /* 208-223 */ +_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L, /* 224-239 */ +_L,_L,_L,_L,_L,_L,_L,_P,_L,_L,_L,_L,_L,_L,_L,_L}; /* 240-255 */ +EXPORT_SYMBOL(_ctype); diff --git a/tests/.gitignore b/tests/.gitignore new file mode 100644 index 00000000..6e86af1f --- /dev/null +++ b/tests/.gitignore @@ -0,0 +1 @@ +/test-mbchk diff --git a/vm/memory_object.c b/vm/memory_object.c index 57dde76b..e281c6a3 100644 --- a/vm/memory_object.c +++ b/vm/memory_object.c @@ -891,34 +891,6 @@ MACRO_END return (KERN_SUCCESS); } -/* - * Old version of memory_object_lock_request. - */ -kern_return_t -xxx_memory_object_lock_request(object, offset, size, - should_clean, should_flush, prot, - reply_to, reply_to_type) - register vm_object_t object; - register vm_offset_t offset; - register vm_size_t size; - boolean_t should_clean; - boolean_t should_flush; - vm_prot_t prot; - ipc_port_t reply_to; - mach_msg_type_name_t reply_to_type; -{ - register int should_return; - - if (should_clean) - should_return = MEMORY_OBJECT_RETURN_DIRTY; - else - should_return = MEMORY_OBJECT_RETURN_NONE; - - return(memory_object_lock_request(object,offset,size, - should_return, should_flush, prot, - reply_to, reply_to_type)); -} - kern_return_t memory_object_set_attributes_common(object, object_ready, may_cache, copy_strategy, use_old_pageout) diff --git a/vm/memory_object_proxy.c b/vm/memory_object_proxy.c index 4d81a687..4fed312e 100644 --- a/vm/memory_object_proxy.c +++ b/vm/memory_object_proxy.c @@ -40,13 +40,15 @@ #include <mach/kern_return.h> #include <mach/notify.h> #include <mach/vm_prot.h> -#include <kern/zalloc.h> -#include <kern/mach_param.h> +#include <kern/printf.h> +#include <kern/slab.h> #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> -/* The zone which holds our proxy memory objects. */ -static zone_t memory_object_proxy_zone; +#include <vm/memory_object_proxy.h> + +/* The cache which holds our proxy memory objects. */ +static struct kmem_cache memory_object_proxy_cache; struct memory_object_proxy { @@ -61,13 +63,8 @@ typedef struct memory_object_proxy *memory_object_proxy_t; void memory_object_proxy_init (void) { - /* For limit, see PORT_MAX. */ - memory_object_proxy_zone = zinit (sizeof (struct memory_object_proxy), 0, - (TASK_MAX * 3 + THREAD_MAX) - * sizeof (struct memory_object_proxy), - 256 * sizeof (struct memory_object_proxy), - ZONE_EXHAUSTIBLE, - "proxy memory object zone"); + kmem_cache_init (&memory_object_proxy_cache, "memory_object_proxy", + sizeof (struct memory_object_proxy), 0, NULL, NULL, NULL, 0); } /* Lookup a proxy memory object by its port. */ @@ -125,7 +122,6 @@ memory_object_create_proxy (ipc_space_t space, vm_prot_t max_protection, vm_offset_t *len, natural_t len_count, ipc_port_t *port) { - kern_return_t kr; memory_object_proxy_t proxy; ipc_port_t notify; @@ -151,13 +147,13 @@ memory_object_create_proxy (ipc_space_t space, vm_prot_t max_protection, if (start[0] != 0 || len[0] != (vm_offset_t) ~0) return KERN_INVALID_ARGUMENT; - proxy = (memory_object_proxy_t) zalloc (memory_object_proxy_zone); + proxy = (memory_object_proxy_t) kmem_cache_alloc (&memory_object_proxy_cache); /* Allocate port, keeping a reference for it. */ proxy->port = ipc_port_alloc_kernel (); if (proxy->port == IP_NULL) { - zfree (memory_object_proxy_zone, (vm_offset_t) proxy); + kmem_cache_free (&memory_object_proxy_cache, (vm_offset_t) proxy); return KERN_RESOURCE_SHORTAGE; } /* Associate the port with the proxy memory object. */ diff --git a/vm/memory_object_proxy.h b/vm/memory_object_proxy.h new file mode 100644 index 00000000..f4be0d0d --- /dev/null +++ b/vm/memory_object_proxy.h @@ -0,0 +1,48 @@ +/* memory_object_proxy.h - Proxy memory objects for Mach. + Copyright (C) 2005, 2011 Free Software Foundation, Inc. + Written by Marcus Brinkmann. + + This file is part of GNU Mach. + + GNU Mach is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2, or (at your option) + any later version. + + GNU Mach is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA. */ + +#ifndef _VM_MEMORY_OBJECT_PROXY_H_ +#define _VM_MEMORY_OBJECT_PROXT_H_ + +#include <ipc/ipc_types.h> +#include <mach/boolean.h> +#include <mach/machine/kern_return.h> +#include <mach/machine/vm_types.h> +#include <mach/message.h> +#include <mach/vm_prot.h> + +extern void memory_object_proxy_init (void); +extern boolean_t memory_object_proxy_notify (mach_msg_header_t *msg); +extern kern_return_t memory_object_create_proxy (ipc_space_t space, + vm_prot_t max_protection, + ipc_port_t *object, + natural_t object_count, + vm_offset_t *offset, + natural_t offset_count, + vm_offset_t *start, + natural_t start_count, + vm_offset_t *len, + natural_t len_count, + ipc_port_t *port); +extern kern_return_t memory_object_proxy_lookup (ipc_port_t port, + ipc_port_t *object, + vm_prot_t *max_protection); + +#endif /* _VM_MEMORY_OBJECT_PROXT_H_ */ diff --git a/vm/vm_external.c b/vm/vm_external.c index ac47faa3..e9643ffc 100644 --- a/vm/vm_external.c +++ b/vm/vm_external.c @@ -31,7 +31,7 @@ */ #include <mach/boolean.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <vm/vm_external.h> #include <mach/vm_param.h> #include <kern/assert.h> @@ -40,7 +40,7 @@ boolean_t vm_external_unsafe = FALSE; -zone_t vm_external_zone = ZONE_NULL; +struct kmem_cache vm_external_cache; /* * The implementation uses bit arrays to record whether @@ -52,8 +52,8 @@ zone_t vm_external_zone = ZONE_NULL; #define SMALL_SIZE (VM_EXTERNAL_SMALL_SIZE/8) #define LARGE_SIZE (VM_EXTERNAL_LARGE_SIZE/8) -zone_t vm_object_small_existence_map_zone; -zone_t vm_object_large_existence_map_zone; +struct kmem_cache vm_object_small_existence_map_cache; +struct kmem_cache vm_object_large_existence_map_cache; vm_external_t vm_external_create(size) @@ -62,20 +62,17 @@ vm_external_t vm_external_create(size) vm_external_t result; vm_size_t bytes; - if (vm_external_zone == ZONE_NULL) - return(VM_EXTERNAL_NULL); - - result = (vm_external_t) zalloc(vm_external_zone); + result = (vm_external_t) kmem_cache_alloc(&vm_external_cache); result->existence_map = (char *) 0; bytes = (atop(size) + 07) >> 3; if (bytes <= SMALL_SIZE) { result->existence_map = - (char *) zalloc(vm_object_small_existence_map_zone); + (char *) kmem_cache_alloc(&vm_object_small_existence_map_cache); result->existence_size = SMALL_SIZE; } else if (bytes <= LARGE_SIZE) { result->existence_map = - (char *) zalloc(vm_object_large_existence_map_zone); + (char *) kmem_cache_alloc(&vm_object_large_existence_map_cache); result->existence_size = LARGE_SIZE; } return(result); @@ -89,14 +86,14 @@ void vm_external_destroy(e) if (e->existence_map != (char *) 0) { if (e->existence_size <= SMALL_SIZE) { - zfree(vm_object_small_existence_map_zone, + kmem_cache_free(&vm_object_small_existence_map_cache, (vm_offset_t) e->existence_map); } else { - zfree(vm_object_large_existence_map_zone, + kmem_cache_free(&vm_object_large_existence_map_cache, (vm_offset_t) e->existence_map); } } - zfree(vm_external_zone, (vm_offset_t) e); + kmem_cache_free(&vm_external_cache, (vm_offset_t) e); } vm_external_state_t _vm_external_state_get(e, offset) @@ -142,18 +139,14 @@ void vm_external_module_initialize(void) { vm_size_t size = (vm_size_t) sizeof(struct vm_external); - vm_external_zone = zinit(size, 0, 16*1024*size, size, - 0, "external page bitmaps"); + kmem_cache_init(&vm_external_cache, "vm_external", size, 0, + NULL, NULL, NULL, 0); - vm_object_small_existence_map_zone = zinit(SMALL_SIZE, 0, - round_page(LARGE_SIZE * SMALL_SIZE), - round_page(SMALL_SIZE), - ZONE_EXHAUSTIBLE, - "object small existence maps"); + kmem_cache_init(&vm_object_small_existence_map_cache, + "small_existence_map", SMALL_SIZE, 0, + NULL, NULL, NULL, 0); - vm_object_large_existence_map_zone = zinit(LARGE_SIZE, 0, - round_page(8 * LARGE_SIZE), - round_page(LARGE_SIZE), - ZONE_EXHAUSTIBLE, - "object large existence maps"); + kmem_cache_init(&vm_object_large_existence_map_cache, + "large_existence_map", LARGE_SIZE, 0, + NULL, NULL, NULL, 0); } diff --git a/vm/vm_fault.c b/vm/vm_fault.c index cce043a1..7e849616 100644 --- a/vm/vm_fault.c +++ b/vm/vm_fault.c @@ -51,9 +51,8 @@ #include <mach/memory_object.h> #include <vm/memory_object_user.user.h> /* For memory_object_data_{request,unlock} */ -#include <kern/mach_param.h> #include <kern/macro_help.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #if MACH_PCSAMPLE #include <kern/pc_sample.h> @@ -85,7 +84,7 @@ typedef struct vm_fault_state { vm_prot_t vmfp_access; } vm_fault_state_t; -zone_t vm_fault_state_zone = 0; +struct kmem_cache vm_fault_state_cache; int vm_object_absent_max = 50; @@ -107,10 +106,8 @@ extern struct db_watchpoint *db_watchpoint_list; */ void vm_fault_init(void) { - vm_fault_state_zone = zinit(sizeof(vm_fault_state_t), 0, - THREAD_MAX * sizeof(vm_fault_state_t), - sizeof(vm_fault_state_t), - 0, "vm fault state"); + kmem_cache_init(&vm_fault_state_cache, "vm_fault_state", + sizeof(vm_fault_state_t), 0, NULL, NULL, NULL, 0); } /* @@ -257,6 +254,7 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, vm_stat_sample(SAMPLED_PC_VM_FAULTS_ANY); vm_stat.faults++; /* needs lock XXX */ + current_task()->faults++; /* * Recovery actions @@ -474,6 +472,7 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, vm_stat_sample(SAMPLED_PC_VM_ZFILL_FAULTS); vm_stat.zero_fill_count++; + current_task()->zero_fills++; vm_object_lock(object); pmap_clear_modify(m->phys_addr); break; @@ -555,6 +554,7 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, if (m->inactive) { vm_stat_sample(SAMPLED_PC_VM_REACTIVATION_FAULTS); vm_stat.reactivations++; + current_task()->reactivations++; } VM_PAGE_QUEUES_REMOVE(m); @@ -654,13 +654,14 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, vm_stat.pageins++; vm_stat_sample(SAMPLED_PC_VM_PAGEIN_FAULTS); + current_task()->pageins++; if ((rc = memory_object_data_request(object->pager, object->pager_request, m->offset + object->paging_offset, PAGE_SIZE, access_required)) != KERN_SUCCESS) { if (rc != MACH_SEND_INTERRUPTED) - printf("%s(0x%p, 0x%p, 0x%x, 0x%x, 0x%x) failed, %x\n", + printf("%s(0x%p, 0x%p, 0x%lx, 0x%x, 0x%x) failed, %x\n", "memory_object_data_request", object->pager, object->pager_request, @@ -743,6 +744,7 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, vm_page_zero_fill(m); vm_stat_sample(SAMPLED_PC_VM_ZFILL_FAULTS); vm_stat.zero_fill_count++; + current_task()->zero_fills++; vm_object_lock(object); pmap_clear_modify(m->phys_addr); break; @@ -858,6 +860,7 @@ vm_fault_return_t vm_fault_page(first_object, first_offset, vm_stat.cow_faults++; vm_stat_sample(SAMPLED_PC_VM_COW_FAULTS); + current_task()->cow_faults++; object = first_object; offset = first_offset; @@ -1206,12 +1209,12 @@ kern_return_t vm_fault(map, vaddr, fault_type, change_wiring, /* * if this assignment stmt is written as - * 'active_threads[cpu_number()] = zalloc()', - * cpu_number may be evaluated before zalloc; - * if zalloc blocks, cpu_number will be wrong + * 'active_threads[cpu_number()] = kmem_cache_alloc()', + * cpu_number may be evaluated before kmem_cache_alloc; + * if kmem_cache_alloc blocks, cpu_number will be wrong */ - state = (char *) zalloc(vm_fault_state_zone); + state = (char *) kmem_cache_alloc(&vm_fault_state_cache); current_thread()->ith_other = state; } @@ -1490,7 +1493,7 @@ kern_return_t vm_fault(map, vaddr, fault_type, change_wiring, register vm_fault_state_t *state = (vm_fault_state_t *) current_thread()->ith_other; - zfree(vm_fault_state_zone, (vm_offset_t) state); + kmem_cache_free(&vm_fault_state_cache, (vm_offset_t) state); (*continuation)(kr); /*NOTREACHED*/ } @@ -1641,6 +1644,7 @@ kern_return_t vm_fault_wire_fast(map, va, entry) vm_prot_t prot; vm_stat.faults++; /* needs lock XXX */ + current_task()->faults++; /* * Recovery actions */ diff --git a/vm/vm_init.c b/vm/vm_init.c index f6a40602..89eb0984 100644 --- a/vm/vm_init.c +++ b/vm/vm_init.c @@ -35,7 +35,7 @@ */ #include <mach/machine/vm_types.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <kern/kalloc.h> #include <vm/vm_fault.h> #include <vm/vm_object.h> @@ -43,7 +43,7 @@ #include <vm/vm_page.h> #include <vm/vm_kern.h> #include <vm/memory_object.h> - +#include <vm/memory_object_proxy.h> /* @@ -67,12 +67,12 @@ void vm_mem_bootstrap() * Initialize other VM packages */ - zone_bootstrap(); + slab_bootstrap(); vm_object_bootstrap(); vm_map_init(); kmem_init(start, end); pmap_init(); - zone_init(); + slab_init(); kalloc_init(); vm_fault_init(); vm_page_module_init(); diff --git a/vm/vm_kern.c b/vm/vm_kern.c index cfa33ff8..fd46e982 100644 --- a/vm/vm_kern.c +++ b/vm/vm_kern.c @@ -58,7 +58,8 @@ * Variables exported by this module. */ -vm_map_t kernel_map; +static struct vm_map kernel_map_store; +vm_map_t kernel_map = &kernel_map_store; vm_map_t kernel_pageable_map; extern void kmem_alloc_pages(); @@ -811,27 +812,27 @@ kmem_remap_pages(object, offset, start, end, protection) } /* - * kmem_suballoc: + * kmem_submap: * - * Allocates a map to manage a subrange + * Initializes a map to manage a subrange * of the kernel virtual address space. * * Arguments are as follows: * + * map Map to initialize * parent Map to take range from * size Size of range to find * min, max Returned endpoints of map * pageable Can the region be paged */ -vm_map_t -kmem_suballoc(parent, min, max, size, pageable) - vm_map_t parent; +void +kmem_submap(map, parent, min, max, size, pageable) + vm_map_t map, parent; vm_offset_t *min, *max; vm_size_t size; boolean_t pageable; { - vm_map_t map; vm_offset_t addr; kern_return_t kr; @@ -850,20 +851,16 @@ kmem_suballoc(parent, min, max, size, pageable) vm_submap_object, (vm_offset_t) 0, FALSE, VM_PROT_DEFAULT, VM_PROT_ALL, VM_INHERIT_DEFAULT); if (kr != KERN_SUCCESS) - panic("kmem_suballoc"); + panic("kmem_submap"); pmap_reference(vm_map_pmap(parent)); - map = vm_map_create(vm_map_pmap(parent), addr, addr + size, pageable); - if (map == VM_MAP_NULL) - panic("kmem_suballoc"); - + vm_map_setup(map, vm_map_pmap(parent), addr, addr + size, pageable); kr = vm_map_submap(parent, addr, addr + size, map); if (kr != KERN_SUCCESS) - panic("kmem_suballoc"); + panic("kmem_submap"); *min = addr; *max = addr + size; - return map; } /* @@ -876,9 +873,8 @@ void kmem_init(start, end) vm_offset_t start; vm_offset_t end; { - kernel_map = vm_map_create(pmap_kernel(), - VM_MIN_KERNEL_ADDRESS, end, - FALSE); + vm_map_setup(kernel_map, pmap_kernel(), VM_MIN_KERNEL_ADDRESS, end, + FALSE); /* * Reserve virtual memory allocated up to this time. diff --git a/vm/vm_kern.h b/vm/vm_kern.h index ca93d7a4..22b7c123 100644 --- a/vm/vm_kern.h +++ b/vm/vm_kern.h @@ -58,8 +58,8 @@ extern kern_return_t kmem_realloc(vm_map_t, vm_offset_t, vm_size_t, vm_offset_t *, vm_size_t); extern void kmem_free(vm_map_t, vm_offset_t, vm_size_t); -extern vm_map_t kmem_suballoc(vm_map_t, vm_offset_t *, vm_offset_t *, - vm_size_t, boolean_t); +extern void kmem_submap(vm_map_t, vm_map_t, vm_offset_t *, + vm_offset_t *, vm_size_t, boolean_t); extern kern_return_t kmem_io_map_copyout(vm_map_t, vm_offset_t *, vm_offset_t *, vm_size_t *, diff --git a/vm/vm_map.c b/vm/vm_map.c index 751e0314..47db118f 100644 --- a/vm/vm_map.c +++ b/vm/vm_map.c @@ -41,7 +41,9 @@ #include <mach/vm_param.h> #include <kern/assert.h> #include <kern/debug.h> -#include <kern/zalloc.h> +#include <kern/kalloc.h> +#include <kern/rbtree.h> +#include <kern/slab.h> #include <vm/pmap.h> #include <vm/vm_fault.h> #include <vm/vm_map.h> @@ -51,6 +53,11 @@ #include <vm/vm_kern.h> #include <ipc/ipc_port.h> +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ + + /* Forward declarations */ kern_return_t vm_map_delete( vm_map_t map, @@ -70,7 +77,7 @@ void vm_map_copy_page_discard (vm_map_copy_t copy); * map entry to the same memory - the wired count in the new entry * must be set to zero. vm_map_entry_copy_full() creates a new * entry that is identical to the old entry. This preserves the - * wire count; it's used for map splitting and zone changing in + * wire count; it's used for map splitting and cache changing in * vm_map_copyout. */ #define vm_map_entry_copy(NEW,OLD) \ @@ -94,7 +101,7 @@ MACRO_END * Synchronization is required prior to most operations. * * Maps consist of an ordered doubly-linked list of simple - * entries; a single hint is used to speed up lookups. + * entries; a hint and a red-black tree are used to speed up lookups. * * Sharing maps have been deleted from this version of Mach. * All shared objects are now mapped directly into the respective @@ -130,10 +137,10 @@ MACRO_END * vm_object_copy_strategically() in vm_object.c. */ -zone_t vm_map_zone; /* zone for vm_map structures */ -zone_t vm_map_entry_zone; /* zone for vm_map_entry structures */ -zone_t vm_map_kentry_zone; /* zone for kernel entry structures */ -zone_t vm_map_copy_zone; /* zone for vm_map_copy structures */ +struct kmem_cache vm_map_cache; /* cache for vm_map structures */ +struct kmem_cache vm_map_entry_cache; /* cache for vm_map_entry structures */ +struct kmem_cache vm_map_kentry_cache; /* cache for kernel entry structures */ +struct kmem_cache vm_map_copy_cache; /* cache for vm_map_copy structures */ boolean_t vm_map_lookup_entry(); /* forward declaration */ @@ -143,7 +150,8 @@ boolean_t vm_map_lookup_entry(); /* forward declaration */ * vm_map_submap creates the submap. */ -vm_object_t vm_submap_object; +static struct vm_object vm_submap_object_store; +vm_object_t vm_submap_object = &vm_submap_object_store; /* * vm_map_init: @@ -151,51 +159,82 @@ vm_object_t vm_submap_object; * Initialize the vm_map module. Must be called before * any other vm_map routines. * - * Map and entry structures are allocated from zones -- we must - * initialize those zones. + * Map and entry structures are allocated from caches -- we must + * initialize those caches. * - * There are three zones of interest: + * There are three caches of interest: * - * vm_map_zone: used to allocate maps. - * vm_map_entry_zone: used to allocate map entries. - * vm_map_kentry_zone: used to allocate map entries for the kernel. + * vm_map_cache: used to allocate maps. + * vm_map_entry_cache: used to allocate map entries. + * vm_map_kentry_cache: used to allocate map entries for the kernel. * - * The kernel allocates map entries from a special zone that is initially - * "crammed" with memory. It would be difficult (perhaps impossible) for - * the kernel to allocate more memory to a entry zone when it became - * empty since the very act of allocating memory implies the creation - * of a new entry. + * Kernel map entries are allocated from a special cache, using a custom + * page allocation function to avoid recursion. It would be difficult + * (perhaps impossible) for the kernel to allocate more memory to an entry + * cache when it became empty since the very act of allocating memory + * implies the creation of a new entry. */ vm_offset_t kentry_data; -vm_size_t kentry_data_size; -int kentry_count = 256; /* to init kentry_data_size */ +vm_size_t kentry_data_size = KENTRY_DATA_SIZE; -void vm_map_init(void) +static vm_offset_t kentry_pagealloc(vm_size_t size) { - vm_map_zone = zinit((vm_size_t) sizeof(struct vm_map), 0, 40*1024, - PAGE_SIZE, 0, "maps"); - vm_map_entry_zone = zinit((vm_size_t) sizeof(struct vm_map_entry), - 0, 1024*1024, PAGE_SIZE*5, - 0, "non-kernel map entries"); - vm_map_kentry_zone = zinit((vm_size_t) sizeof(struct vm_map_entry), 0, - kentry_data_size, kentry_data_size, - ZONE_FIXED /* XXX */, "kernel map entries"); - - vm_map_copy_zone = zinit((vm_size_t) sizeof(struct vm_map_copy), - 0, 16*1024, PAGE_SIZE, 0, - "map copies"); + vm_offset_t result; - /* - * Cram the kentry zone with initial data. - */ - zcram(vm_map_kentry_zone, kentry_data, kentry_data_size); + if (size > kentry_data_size) + panic("vm_map: kentry memory exhausted"); + + result = kentry_data; + kentry_data += size; + kentry_data_size -= size; + return result; +} + +void vm_map_init(void) +{ + kmem_cache_init(&vm_map_cache, "vm_map", sizeof(struct vm_map), 0, + NULL, NULL, NULL, 0); + kmem_cache_init(&vm_map_entry_cache, "vm_map_entry", + sizeof(struct vm_map_entry), 0, NULL, NULL, NULL, 0); + kmem_cache_init(&vm_map_kentry_cache, "vm_map_kentry", + sizeof(struct vm_map_entry), 0, NULL, kentry_pagealloc, + NULL, KMEM_CACHE_NOCPUPOOL | KMEM_CACHE_NOOFFSLAB + | KMEM_CACHE_NORECLAIM); + kmem_cache_init(&vm_map_copy_cache, "vm_map_copy", + sizeof(struct vm_map_copy), 0, NULL, NULL, NULL, 0); /* * Submap object is initialized by vm_object_init. */ } +void vm_map_setup(map, pmap, min, max, pageable) + vm_map_t map; + pmap_t pmap; + vm_offset_t min, max; + boolean_t pageable; +{ + vm_map_first_entry(map) = vm_map_to_entry(map); + vm_map_last_entry(map) = vm_map_to_entry(map); + map->hdr.nentries = 0; + map->hdr.entries_pageable = pageable; + rbtree_init(&map->hdr.tree); + + map->size = 0; + map->ref_count = 1; + map->pmap = pmap; + map->min_offset = min; + map->max_offset = max; + map->wiring_required = FALSE; + map->wait_for_space = FALSE; + map->first_free = vm_map_to_entry(map); + map->hint = vm_map_to_entry(map); + vm_map_lock_init(map); + simple_lock_init(&map->ref_lock); + simple_lock_init(&map->hint_lock); +} + /* * vm_map_create: * @@ -210,27 +249,11 @@ vm_map_t vm_map_create(pmap, min, max, pageable) { register vm_map_t result; - result = (vm_map_t) zalloc(vm_map_zone); + result = (vm_map_t) kmem_cache_alloc(&vm_map_cache); if (result == VM_MAP_NULL) panic("vm_map_create"); - vm_map_first_entry(result) = vm_map_to_entry(result); - vm_map_last_entry(result) = vm_map_to_entry(result); - result->hdr.nentries = 0; - result->hdr.entries_pageable = pageable; - - result->size = 0; - result->ref_count = 1; - result->pmap = pmap; - result->min_offset = min; - result->max_offset = max; - result->wiring_required = FALSE; - result->wait_for_space = FALSE; - result->first_free = vm_map_to_entry(result); - result->hint = vm_map_to_entry(result); - vm_map_lock_init(result); - simple_lock_init(&result->ref_lock); - simple_lock_init(&result->hint_lock); + vm_map_setup(result, pmap, min, max, pageable); return(result); } @@ -250,15 +273,15 @@ vm_map_t vm_map_create(pmap, min, max, pageable) vm_map_entry_t _vm_map_entry_create(map_header) register struct vm_map_header *map_header; { - register zone_t zone; + register kmem_cache_t cache; register vm_map_entry_t entry; if (map_header->entries_pageable) - zone = vm_map_entry_zone; + cache = &vm_map_entry_cache; else - zone = vm_map_kentry_zone; + cache = &vm_map_kentry_cache; - entry = (vm_map_entry_t) zalloc(zone); + entry = (vm_map_entry_t) kmem_cache_alloc(cache); if (entry == VM_MAP_ENTRY_NULL) panic("vm_map_entry_create"); @@ -280,20 +303,50 @@ void _vm_map_entry_dispose(map_header, entry) register struct vm_map_header *map_header; register vm_map_entry_t entry; { - register zone_t zone; + register kmem_cache_t cache; if (map_header->entries_pageable) - zone = vm_map_entry_zone; + cache = &vm_map_entry_cache; else - zone = vm_map_kentry_zone; + cache = &vm_map_kentry_cache; - zfree(zone, (vm_offset_t) entry); + kmem_cache_free(cache, (vm_offset_t) entry); +} + +/* + * Red-black tree lookup/insert comparison functions + */ +static inline int vm_map_entry_cmp_lookup(vm_offset_t addr, + const struct rbtree_node *node) +{ + struct vm_map_entry *entry; + + entry = rbtree_entry(node, struct vm_map_entry, tree_node); + + if (addr < entry->vme_start) + return -1; + else if (addr < entry->vme_end) + return 0; + else + return 1; +} + +static inline int vm_map_entry_cmp_insert(const struct rbtree_node *a, + const struct rbtree_node *b) +{ + struct vm_map_entry *entry; + + entry = rbtree_entry(a, struct vm_map_entry, tree_node); + return vm_map_entry_cmp_lookup(entry->vme_start, b); } /* * vm_map_entry_{un,}link: * * Insert/remove entries from maps (or map copies). + * + * The start and end addresses of the entries must be properly set + * before using these macros. */ #define vm_map_entry_link(map, after_where, entry) \ _vm_map_entry_link(&(map)->hdr, after_where, entry) @@ -308,6 +361,8 @@ void _vm_map_entry_dispose(map_header, entry) (entry)->vme_next = (after_where)->vme_next; \ (entry)->vme_prev->vme_next = \ (entry)->vme_next->vme_prev = (entry); \ + rbtree_insert(&(hdr)->tree, &(entry)->tree_node, \ + vm_map_entry_cmp_insert); \ MACRO_END #define vm_map_entry_unlink(map, entry) \ @@ -321,6 +376,7 @@ void _vm_map_entry_dispose(map_header, entry) (hdr)->nentries--; \ (entry)->vme_next->vme_prev = (entry)->vme_prev; \ (entry)->vme_prev->vme_next = (entry)->vme_next; \ + rbtree_remove(&(hdr)->tree, &(entry)->tree_node); \ MACRO_END /* @@ -368,7 +424,7 @@ void vm_map_deallocate(map) pmap_destroy(map->pmap); - zfree(vm_map_zone, (vm_offset_t) map); + kmem_cache_free(&vm_map_cache, (vm_offset_t) map); } /* @@ -397,70 +453,49 @@ boolean_t vm_map_lookup_entry(map, address, entry) register vm_offset_t address; vm_map_entry_t *entry; /* OUT */ { - register vm_map_entry_t cur; - register vm_map_entry_t last; + register struct rbtree_node *node; + register vm_map_entry_t hint; /* - * Start looking either from the head of the - * list, or from the hint. + * First, make a quick check to see if we are already + * looking at the entry we want (which is often the case). */ simple_lock(&map->hint_lock); - cur = map->hint; + hint = map->hint; simple_unlock(&map->hint_lock); - if (cur == vm_map_to_entry(map)) - cur = cur->vme_next; - - if (address >= cur->vme_start) { - /* - * Go from hint to end of list. - * - * But first, make a quick check to see if - * we are already looking at the entry we - * want (which is usually the case). - * Note also that we don't need to save the hint - * here... it is the same hint (unless we are - * at the header, in which case the hint didn't - * buy us anything anyway). - */ - last = vm_map_to_entry(map); - if ((cur != last) && (cur->vme_end > address)) { - *entry = cur; + if ((hint != vm_map_to_entry(map)) && (address >= hint->vme_start)) { + if (address < hint->vme_end) { + *entry = hint; return(TRUE); + } else { + vm_map_entry_t next = hint->vme_next; + + if ((next == vm_map_to_entry(map)) + || (address < next->vme_start)) { + *entry = hint; + return(FALSE); + } } } - else { - /* - * Go from start to hint, *inclusively* - */ - last = cur->vme_next; - cur = vm_map_first_entry(map); - } /* - * Search linearly + * If the hint didn't help, use the red-black tree. */ - while (cur != last) { - if (cur->vme_end > address) { - if (address >= cur->vme_start) { - /* - * Save this lookup for future - * hints, and return - */ + node = rbtree_lookup_nearest(&map->hdr.tree, address, + vm_map_entry_cmp_lookup, RBTREE_LEFT); - *entry = cur; - SAVE_HINT(map, cur); - return(TRUE); - } - break; - } - cur = cur->vme_next; + if (node == NULL) { + *entry = vm_map_to_entry(map); + SAVE_HINT(map, *entry); + return(FALSE); + } else { + *entry = rbtree_entry(node, struct vm_map_entry, tree_node); + SAVE_HINT(map, *entry); + return((address < (*entry)->vme_end) ? TRUE : FALSE); } - *entry = cur->vme_prev; - SAVE_HINT(map, *entry); - return(FALSE); } /* @@ -695,7 +730,7 @@ vm_map_pmap_enter(map, addr, end_addr, object, offset, protection) if (vm_map_pmap_enter_print) { printf("vm_map_pmap_enter:"); - printf("map: %p, addr: %x, object: %p, offset: %x\n", + printf("map: %p, addr: %lx, object: %p, offset: %lx\n", map, addr, object, offset); } @@ -754,6 +789,9 @@ kern_return_t vm_map_enter( #define RETURN(value) { result = value; goto BailOut; } + if (size == 0) + return KERN_INVALID_ARGUMENT; + StartAgain: ; start = *address; @@ -1907,7 +1945,7 @@ free_next_copy: register vm_map_copy_t new_copy; new_copy = (vm_map_copy_t) copy->cpy_cont_args; - zfree(vm_map_copy_zone, (vm_offset_t) copy); + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) copy); copy = new_copy; goto free_next_copy; } @@ -1918,7 +1956,7 @@ free_next_copy: break; } - zfree(vm_map_copy_zone, (vm_offset_t) copy); + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) copy); } /* @@ -1952,7 +1990,7 @@ vm_map_copy_copy(copy) * from the old one into it. */ - new_copy = (vm_map_copy_t) zalloc(vm_map_copy_zone); + new_copy = (vm_map_copy_t) kmem_cache_alloc(&vm_map_copy_cache); *new_copy = *copy; if (copy->type == VM_MAP_COPY_ENTRY_LIST) { @@ -2160,7 +2198,7 @@ start_pass_1: /* * XXXO If there are no permanent objects in the destination, - * XXXO and the source and destination map entry zones match, + * XXXO and the source and destination map entry caches match, * XXXO and the destination map entry is not shared, * XXXO then the map entries can be deleted and replaced * XXXO with those from the copy. The following code is the @@ -2398,12 +2436,16 @@ start_pass_1: */ #define vm_map_copy_insert(map, where, copy) \ MACRO_BEGIN \ + struct rbtree_node *node, *tmp; \ + rbtree_for_each_remove(&(copy)->cpy_hdr.tree, node, tmp) \ + rbtree_insert(&(map)->hdr.tree, node, \ + vm_map_entry_cmp_insert); \ (((where)->vme_next)->vme_prev = vm_map_copy_last_entry(copy)) \ ->vme_next = ((where)->vme_next); \ ((where)->vme_next = vm_map_copy_first_entry(copy)) \ ->vme_prev = (where); \ (map)->hdr.nentries += (copy)->cpy_hdr.nentries; \ - zfree(vm_map_copy_zone, (vm_offset_t) copy); \ + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) copy); \ MACRO_END /* @@ -2459,7 +2501,7 @@ kern_return_t vm_map_copyout(dst_map, dst_addr, copy) VM_INHERIT_DEFAULT); if (kr != KERN_SUCCESS) return(kr); - zfree(vm_map_copy_zone, (vm_offset_t) copy); + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) copy); return(KERN_SUCCESS); } @@ -2516,15 +2558,15 @@ kern_return_t vm_map_copyout(dst_map, dst_addr, copy) * Mismatches occur when dealing with the default * pager. */ - zone_t old_zone; + kmem_cache_t old_cache; vm_map_entry_t next, new; /* - * Find the zone that the copies were allocated from + * Find the cache that the copies were allocated from */ - old_zone = (copy->cpy_hdr.entries_pageable) - ? vm_map_entry_zone - : vm_map_kentry_zone; + old_cache = (copy->cpy_hdr.entries_pageable) + ? &vm_map_entry_cache + : &vm_map_kentry_cache; entry = vm_map_copy_first_entry(copy); /* @@ -2547,7 +2589,7 @@ kern_return_t vm_map_copyout(dst_map, dst_addr, copy) vm_map_copy_last_entry(copy), new); next = entry->vme_next; - zfree(old_zone, (vm_offset_t) entry); + kmem_cache_free(old_cache, (vm_offset_t) entry); entry = next; } } @@ -3036,10 +3078,10 @@ error: * Consume on success logic. */ if (copy != orig_copy) { - zfree(vm_map_copy_zone, (vm_offset_t) copy); + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) copy); } if (result == KERN_SUCCESS) { - zfree(vm_map_copy_zone, (vm_offset_t) orig_copy); + kmem_cache_free(&vm_map_copy_cache, (vm_offset_t) orig_copy); } return(result); @@ -3116,12 +3158,13 @@ kern_return_t vm_map_copyin(src_map, src_addr, len, src_destroy, copy_result) * remember the endpoints prior to rounding. */ - copy = (vm_map_copy_t) zalloc(vm_map_copy_zone); + copy = (vm_map_copy_t) kmem_cache_alloc(&vm_map_copy_cache); vm_map_copy_first_entry(copy) = vm_map_copy_last_entry(copy) = vm_map_copy_to_entry(copy); copy->type = VM_MAP_COPY_ENTRY_LIST; copy->cpy_hdr.nentries = 0; copy->cpy_hdr.entries_pageable = TRUE; + rbtree_init(©->cpy_hdr.tree); copy->offset = src_addr; copy->size = len; @@ -3443,7 +3486,7 @@ kern_return_t vm_map_copyin_object(object, offset, size, copy_result) * and null links. */ - copy = (vm_map_copy_t) zalloc(vm_map_copy_zone); + copy = (vm_map_copy_t) kmem_cache_alloc(&vm_map_copy_cache); vm_map_copy_first_entry(copy) = vm_map_copy_last_entry(copy) = VM_MAP_ENTRY_NULL; copy->type = VM_MAP_COPY_OBJECT; @@ -3598,7 +3641,7 @@ kern_return_t vm_map_copyin_page_list(src_map, src_addr, len, src_destroy, * be page-aligned. */ - copy = (vm_map_copy_t) zalloc(vm_map_copy_zone); + copy = (vm_map_copy_t) kmem_cache_alloc(&vm_map_copy_cache); copy->type = VM_MAP_COPY_PAGE_LIST; copy->cpy_npages = 0; copy->offset = src_addr; @@ -4157,6 +4200,8 @@ vm_map_t vm_map_fork(old_map) object->ref_count++; vm_object_unlock(object); + new_entry = vm_map_entry_create(new_map); + if (old_entry->projected_on != 0) { /* * If entry is projected buffer, clone the @@ -4171,7 +4216,6 @@ vm_map_t vm_map_fork(old_map) * Mark both entries as shared. */ - new_entry = vm_map_entry_create(new_map); vm_map_entry_copy(new_entry, old_entry); old_entry->is_shared = TRUE; new_entry->is_shared = TRUE; diff --git a/vm/vm_map.h b/vm/vm_map.h index 567fe933..a15e681b 100644 --- a/vm/vm_map.h +++ b/vm/vm_map.h @@ -51,8 +51,12 @@ #include <vm/vm_page.h> #include <vm/vm_types.h> #include <kern/lock.h> +#include <kern/rbtree.h> #include <kern/macro_help.h> +/* TODO: make it dynamic */ +#define KENTRY_DATA_SIZE (64*PAGE_SIZE) + /* * Types defined: * @@ -100,6 +104,7 @@ struct vm_map_entry { #define vme_next links.next #define vme_start links.start #define vme_end links.end + struct rbtree_node tree_node; /* links to other entries in tree */ union vm_map_object object; /* object I point to */ vm_offset_t offset; /* offset into object */ unsigned int @@ -135,6 +140,7 @@ typedef struct vm_map_entry *vm_map_entry_t; */ struct vm_map_header { struct vm_map_links links; /* first, last, min, max */ + struct rbtree tree; /* Sorted tree of entries */ int nentries; /* Number of entries */ boolean_t entries_pageable; /* are map entries pageable? */ @@ -150,9 +156,11 @@ struct vm_map_header { * * Implementation: * Maps are doubly-linked lists of map entries, sorted - * by address. One hint is used to start - * searches again from the last successful search, - * insertion, or removal. Another hint is used to + * by address. They're also contained in a red-black tree. + * One hint is used to start searches again at the last + * successful search, insertion, or removal. If the hint + * lookup failed (i.e. the hint didn't refer to the requested + * entry), a BST lookup is performed. Another hint is used to * quickly find free space. */ struct vm_map { @@ -352,11 +360,14 @@ MACRO_END */ extern vm_offset_t kentry_data; -extern vm_offset_t kentry_data_size; +extern vm_size_t kentry_data_size; extern int kentry_count; /* Initialize the module */ extern void vm_map_init(void); +/* Initialize an empty map */ +extern void vm_map_setup(vm_map_t, pmap_t, vm_offset_t, vm_offset_t, + boolean_t); /* Create an empty map */ extern vm_map_t vm_map_create(pmap_t, vm_offset_t, vm_offset_t, boolean_t); diff --git a/vm/vm_object.c b/vm/vm_object.c index 9057973d..d83c39fe 100644 --- a/vm/vm_object.c +++ b/vm/vm_object.c @@ -47,7 +47,7 @@ #include <kern/lock.h> #include <kern/queue.h> #include <kern/xpr.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <vm/memory_object.h> #include <vm/vm_fault.h> #include <vm/vm_map.h> @@ -55,6 +55,10 @@ #include <vm/vm_page.h> #include <vm/vm_pageout.h> +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ + void memory_object_release( ipc_port_t pager, @@ -141,13 +145,14 @@ void vm_object_deactivate_pages(vm_object_t); * ZZZ Continue this comment. */ -zone_t vm_object_zone; /* vm backing store zone */ +struct kmem_cache vm_object_cache; /* vm backing store cache */ /* * All wired-down kernel memory belongs to a single virtual * memory object (kernel_object) to avoid wasting data structures. */ -vm_object_t kernel_object; +static struct vm_object kernel_object_store; +vm_object_t kernel_object = &kernel_object_store; /* * Virtual memory objects that are not referenced by @@ -191,6 +196,15 @@ decl_simple_lock_data(,vm_object_cached_lock_data) simple_unlock(&vm_object_cached_lock_data) /* + * Number of physical pages referenced by cached objects. + * This counter is protected by its own lock to work around + * lock ordering issues. + */ +int vm_object_cached_pages; + +decl_simple_lock_data(,vm_object_cached_pages_lock_data) + +/* * Virtual memory objects are initialized from * a template (see vm_object_allocate). * @@ -198,7 +212,7 @@ decl_simple_lock_data(,vm_object_cached_lock_data) * object structure, be sure to add initialization * (see vm_object_init). */ -vm_object_t vm_object_template; +struct vm_object vm_object_template; /* * vm_object_allocate: @@ -206,17 +220,24 @@ vm_object_t vm_object_template; * Returns a new object with the given size. */ +static void _vm_object_setup( + vm_object_t object, + vm_size_t size) +{ + *object = vm_object_template; + queue_init(&object->memq); + vm_object_lock_init(object); + object->size = size; +} + vm_object_t _vm_object_allocate( vm_size_t size) { register vm_object_t object; - object = (vm_object_t) zalloc(vm_object_zone); + object = (vm_object_t) kmem_cache_alloc(&vm_object_cache); - *object = *vm_object_template; - queue_init(&object->memq); - vm_object_lock_init(object); - object->size = size; + _vm_object_setup(object, size); return object; } @@ -244,10 +265,8 @@ vm_object_t vm_object_allocate( */ void vm_object_bootstrap(void) { - vm_object_zone = zinit((vm_size_t) sizeof(struct vm_object), 0, - round_page(512*1024), - round_page(12*1024), - 0, "objects"); + kmem_cache_init(&vm_object_cache, "vm_object", + sizeof(struct vm_object), 0, NULL, NULL, NULL, 0); queue_init(&vm_object_cached_list); simple_lock_init(&vm_object_cached_lock_data); @@ -256,53 +275,50 @@ void vm_object_bootstrap(void) * Fill in a template object, for quick initialization */ - vm_object_template = (vm_object_t) zalloc(vm_object_zone); - memset(vm_object_template, 0, sizeof *vm_object_template); - - vm_object_template->ref_count = 1; - vm_object_template->size = 0; - vm_object_template->resident_page_count = 0; - vm_object_template->copy = VM_OBJECT_NULL; - vm_object_template->shadow = VM_OBJECT_NULL; - vm_object_template->shadow_offset = (vm_offset_t) 0; + vm_object_template.ref_count = 1; + vm_object_template.size = 0; + vm_object_template.resident_page_count = 0; + vm_object_template.copy = VM_OBJECT_NULL; + vm_object_template.shadow = VM_OBJECT_NULL; + vm_object_template.shadow_offset = (vm_offset_t) 0; - vm_object_template->pager = IP_NULL; - vm_object_template->paging_offset = 0; - vm_object_template->pager_request = PAGER_REQUEST_NULL; - vm_object_template->pager_name = IP_NULL; + vm_object_template.pager = IP_NULL; + vm_object_template.paging_offset = 0; + vm_object_template.pager_request = PAGER_REQUEST_NULL; + vm_object_template.pager_name = IP_NULL; - vm_object_template->pager_created = FALSE; - vm_object_template->pager_initialized = FALSE; - vm_object_template->pager_ready = FALSE; + vm_object_template.pager_created = FALSE; + vm_object_template.pager_initialized = FALSE; + vm_object_template.pager_ready = FALSE; - vm_object_template->copy_strategy = MEMORY_OBJECT_COPY_NONE; + vm_object_template.copy_strategy = MEMORY_OBJECT_COPY_NONE; /* ignored if temporary, will be reset before * permanent object becomes ready */ - vm_object_template->use_shared_copy = FALSE; - vm_object_template->shadowed = FALSE; - - vm_object_template->absent_count = 0; - vm_object_template->all_wanted = 0; /* all bits FALSE */ - - vm_object_template->paging_in_progress = 0; - vm_object_template->can_persist = FALSE; - vm_object_template->internal = TRUE; - vm_object_template->temporary = TRUE; - vm_object_template->alive = TRUE; - vm_object_template->lock_in_progress = FALSE; - vm_object_template->lock_restart = FALSE; - vm_object_template->use_old_pageout = TRUE; /* XXX change later */ - vm_object_template->last_alloc = (vm_offset_t) 0; + vm_object_template.use_shared_copy = FALSE; + vm_object_template.shadowed = FALSE; + + vm_object_template.absent_count = 0; + vm_object_template.all_wanted = 0; /* all bits FALSE */ + + vm_object_template.paging_in_progress = 0; + vm_object_template.can_persist = FALSE; + vm_object_template.internal = TRUE; + vm_object_template.temporary = TRUE; + vm_object_template.alive = TRUE; + vm_object_template.lock_in_progress = FALSE; + vm_object_template.lock_restart = FALSE; + vm_object_template.use_old_pageout = TRUE; /* XXX change later */ + vm_object_template.last_alloc = (vm_offset_t) 0; #if MACH_PAGEMAP - vm_object_template->existence_info = VM_EXTERNAL_NULL; + vm_object_template.existence_info = VM_EXTERNAL_NULL; #endif /* MACH_PAGEMAP */ /* * Initialize the "kernel object" */ - kernel_object = _vm_object_allocate( + _vm_object_setup(kernel_object, VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS); /* @@ -310,7 +326,7 @@ void vm_object_bootstrap(void) * kernel object so that no limit is imposed on submap sizes. */ - vm_submap_object = _vm_object_allocate( + _vm_object_setup(vm_submap_object, VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS); #if MACH_PAGEMAP @@ -407,6 +423,7 @@ void vm_object_deallocate( queue_enter(&vm_object_cached_list, object, vm_object_t, cached_list); overflow = (++vm_object_cached_count > vm_object_cached_max); + vm_object_cached_pages_update(object->resident_page_count); vm_object_cache_unlock(); vm_object_deactivate_pages(object); @@ -660,7 +677,7 @@ void vm_object_terminate( * Free the space for the object. */ - zfree(vm_object_zone, (vm_offset_t) object); + kmem_cache_free(&vm_object_cache, (vm_offset_t) object); } /* @@ -916,7 +933,7 @@ boolean_t vm_object_pmap_protect_by_page = FALSE; void vm_object_pmap_protect( register vm_object_t object, register vm_offset_t offset, - vm_offset_t size, + vm_size_t size, pmap_t pmap, vm_offset_t pmap_start, vm_prot_t prot) @@ -1857,6 +1874,7 @@ vm_object_t vm_object_lookup( queue_remove(&vm_object_cached_list, object, vm_object_t, cached_list); vm_object_cached_count--; + vm_object_cached_pages_update(-object->resident_page_count); } object->ref_count++; @@ -1888,6 +1906,7 @@ vm_object_t vm_object_lookup_name( queue_remove(&vm_object_cached_list, object, vm_object_t, cached_list); vm_object_cached_count--; + vm_object_cached_pages_update(-object->resident_page_count); } object->ref_count++; @@ -1924,6 +1943,7 @@ void vm_object_destroy( queue_remove(&vm_object_cached_list, object, vm_object_t, cached_list); vm_object_cached_count--; + vm_object_cached_pages_update(-object->resident_page_count); } object->ref_count++; @@ -2077,6 +2097,7 @@ restart: queue_remove(&vm_object_cached_list, object, vm_object_t, cached_list); vm_object_cached_count--; + vm_object_cached_pages_update(-object->resident_page_count); } object->ref_count++; vm_object_unlock(object); @@ -2618,7 +2639,7 @@ void vm_object_collapse( vm_object_unlock(object); if (old_name_port != IP_NULL) ipc_port_dealloc_kernel(old_name_port); - zfree(vm_object_zone, (vm_offset_t) backing_object); + kmem_cache_free(&vm_object_cache, (vm_offset_t) backing_object); vm_object_lock(object); object_collapses++; diff --git a/vm/vm_object.h b/vm/vm_object.h index c9925709..4e4c9498 100644 --- a/vm/vm_object.h +++ b/vm/vm_object.h @@ -71,8 +71,8 @@ struct vm_object { * if internal) */ - short ref_count; /* Number of references */ - short resident_page_count; + int ref_count; /* Number of references */ + int resident_page_count; /* number of resident pages */ struct vm_object *copy; /* Object that should receive @@ -370,4 +370,23 @@ MACRO_END #define vm_object_lock_taken(object) simple_lock_taken(&(object)->Lock) #endif /* VM_OBJECT_DEBUG */ +/* + * Page cache accounting. + * + * The number of cached objects and pages can be read + * without holding any lock. + */ + +extern int vm_object_cached_count; + +extern int vm_object_cached_pages; +decl_simple_lock_data(extern,vm_object_cached_pages_lock_data) + +#define vm_object_cached_pages_update(page_count) \ + MACRO_BEGIN \ + simple_lock(&vm_object_cached_pages_lock_data); \ + vm_object_cached_pages += (page_count); \ + simple_unlock(&vm_object_cached_pages_lock_data); \ + MACRO_END + #endif /* _VM_VM_OBJECT_H_ */ diff --git a/vm/vm_page.h b/vm/vm_page.h index f13b0af9..4536d1c5 100644 --- a/vm/vm_page.h +++ b/vm/vm_page.h @@ -41,7 +41,6 @@ #include <vm/vm_types.h> #include <kern/queue.h> #include <kern/lock.h> -#include <kern/zalloc.h> #include <kern/macro_help.h> #include <kern/sched_prim.h> /* definitions of wait/wakeup */ diff --git a/vm/vm_pageout.c b/vm/vm_pageout.c index 7a755bf4..661675f0 100644 --- a/vm/vm_pageout.c +++ b/vm/vm_pageout.c @@ -43,6 +43,7 @@ #include <mach/vm_statistics.h> #include <kern/counters.h> #include <kern/debug.h> +#include <kern/slab.h> #include <kern/task.h> #include <kern/thread.h> #include <vm/pmap.h> @@ -544,8 +545,8 @@ void vm_pageout_scan() * into an internal object and then immediately double-page it, * sending it to the default pager. * - * consider_zone_gc should be last, because the other operations - * might return memory to zones. When we pause we use + * slab_collect should be last, because the other operations + * might return memory to caches. When we pause we use * vm_pageout_scan_continue as our continuation, so we will * reenter vm_pageout_scan periodically and attempt to reclaim * internal memory even if we never reach vm_page_free_target. @@ -555,7 +556,7 @@ void vm_pageout_scan() net_kmsg_collect(); consider_task_collect(); consider_thread_collect(); - consider_zone_gc(); + slab_collect(); for (burst_count = 0;;) { register vm_page_t m; @@ -763,6 +764,7 @@ void vm_pageout_scan() vm_object_unlock(object); vm_page_activate(m); vm_stat.reactivations++; + current_task()->reactivations++; vm_page_unlock_queues(); vm_pageout_inactive_used++; continue; diff --git a/vm/vm_resident.c b/vm/vm_resident.c index a6334989..7cf4fb16 100644 --- a/vm/vm_resident.c +++ b/vm/vm_resident.c @@ -45,7 +45,7 @@ #include <mach/vm_statistics.h> #include <machine/vm_param.h> #include <kern/xpr.h> -#include <kern/zalloc.h> +#include <kern/slab.h> #include <vm/pmap.h> #include <vm/vm_map.h> #include <vm/vm_page.h> @@ -58,9 +58,10 @@ #include <vm/vm_user.h> #endif -/* in zalloc.c XXX */ -extern vm_offset_t zdata; -extern vm_size_t zdata_size; +#if MACH_KDB +#include <ddb/db_output.h> +#endif /* MACH_KDB */ + /* * Associated with eacn page of user-allocatable memory is a @@ -126,7 +127,7 @@ unsigned int vm_page_free_count_minimum; /* debugging */ * These page structures are allocated the way * most other kernel structures are. */ -zone_t vm_page_zone; +struct kmem_cache vm_page_cache; /* * Fictitious pages don't have a physical address, @@ -239,14 +240,11 @@ void vm_page_bootstrap( vm_page_free_wanted = 0; /* - * Steal memory for the zone system. + * Steal memory for the kernel map entries. */ - kentry_data_size = kentry_count * sizeof(struct vm_map_entry); kentry_data = pmap_steal_memory(kentry_data_size); - zdata = pmap_steal_memory(zdata_size); - /* * Allocate (and initialize) the virtual-to-physical * table hash buckets. @@ -393,6 +391,11 @@ void pmap_startup( vm_page_init(&pages[i], paddr); pages_initialized++; } + i = 0; + while (pmap_next_page(&paddr)) + i++; + if (i) + printf("%d memory page(s) left away\n", i); /* * Release pages in reverse order so that physical pages @@ -425,10 +428,8 @@ void pmap_startup( */ void vm_page_module_init(void) { - vm_page_zone = zinit((vm_size_t) sizeof(struct vm_page), 0, - VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS, - PAGE_SIZE, - 0, "vm pages"); + kmem_cache_init(&vm_page_cache, "vm_page", sizeof(struct vm_page), 0, + NULL, NULL, NULL, 0); } /* @@ -450,7 +451,7 @@ void vm_page_create( for (paddr = round_page(start); paddr < trunc_page(end); paddr += PAGE_SIZE) { - m = (vm_page_t) zalloc(vm_page_zone); + m = (vm_page_t) kmem_cache_alloc(&vm_page_cache); if (m == VM_PAGE_NULL) panic("vm_page_create"); @@ -521,6 +522,10 @@ void vm_page_insert( */ object->resident_page_count++; + assert(object->resident_page_count >= 0); + + if (object->can_persist && (object->ref_count == 0)) + vm_object_cached_pages_update(1); /* * Detect sequential access and inactivate previous page. @@ -589,6 +594,10 @@ void vm_page_replace( m->tabled = FALSE; object->resident_page_count--; + if (object->can_persist + && (object->ref_count == 0)) + vm_object_cached_pages_update(-1); + /* * Return page to the free list. * Note the page is not tabled now, so this @@ -620,6 +629,10 @@ void vm_page_replace( */ object->resident_page_count++; + assert(object->resident_page_count >= 0); + + if (object->can_persist && (object->ref_count == 0)) + vm_object_cached_pages_update(1); } /* @@ -675,6 +688,9 @@ void vm_page_remove( mem->object->resident_page_count--; mem->tabled = FALSE; + + if (mem->object->can_persist && (mem->object->ref_count == 0)) + vm_object_cached_pages_update(-1); } /* @@ -805,7 +821,7 @@ void vm_page_more_fictitious(void) int i; for (i = 0; i < vm_page_fictitious_quantum; i++) { - m = (vm_page_t) zalloc(vm_page_zone); + m = (vm_page_t) kmem_cache_alloc(&vm_page_cache); if (m == VM_PAGE_NULL) panic("vm_page_more_fictitious"); diff --git a/vm/vm_user.c b/vm/vm_user.c index 672daab8..6fe398e0 100644 --- a/vm/vm_user.c +++ b/vm/vm_user.c @@ -38,12 +38,14 @@ #include <mach/vm_attributes.h> #include <mach/vm_param.h> #include <mach/vm_statistics.h> +#include <mach/vm_cache_statistics.h> #include <kern/host.h> #include <kern/task.h> #include <vm/vm_fault.h> #include <vm/vm_kern.h> #include <vm/vm_map.h> #include <vm/vm_object.h> +#include <vm/memory_object_proxy.h> #include <vm/vm_page.h> @@ -188,6 +190,29 @@ kern_return_t vm_statistics(map, stat) return(KERN_SUCCESS); } +kern_return_t vm_cache_statistics( + vm_map_t map, + vm_cache_statistics_data_t *stats) +{ + if (map == VM_MAP_NULL) + return KERN_INVALID_ARGUMENT; + + stats->cache_object_count = vm_object_cached_count; + stats->cache_count = vm_object_cached_pages; + + /* XXX Not implemented yet */ + stats->active_tmp_count = 0; + stats->inactive_tmp_count = 0; + stats->active_perm_count = 0; + stats->inactive_perm_count = 0; + stats->dirty_count = 0; + stats->laundry_count = 0; + stats->writeback_count = 0; + stats->slab_count = 0; + stats->slab_reclaim_count = 0; + return KERN_SUCCESS; +} + /* * Handle machine-specific attributes for a mapping, such * as cachability, migrability, etc. @@ -277,11 +302,6 @@ kern_return_t vm_copy(map, source_address, size, dest_address) } -/* XXX From memory_object_proxy.c */ -kern_return_t -memory_object_proxy_lookup (ipc_port_t proxy_object, ipc_port_t *object, - vm_prot_t *max_protection); - /* * Routine: vm_map */ @@ -322,6 +342,9 @@ kern_return_t vm_map( return(KERN_INVALID_ARGUMENT); } + if (size == 0) + return KERN_INVALID_ARGUMENT; + *address = trunc_page(*address); size = round_page(size); diff --git a/vm/vm_user.h b/vm/vm_user.h index 3f15e5ed..c6f20a82 100644 --- a/vm/vm_user.h +++ b/vm/vm_user.h @@ -37,6 +37,7 @@ #include <mach/kern_return.h> #include <mach/std_types.h> +#include <mach/mach_types.h> extern kern_return_t vm_allocate(vm_map_t, vm_offset_t *, vm_size_t, boolean_t); @@ -46,6 +47,7 @@ extern kern_return_t vm_inherit(vm_map_t, vm_offset_t, vm_size_t, extern kern_return_t vm_protect(vm_map_t, vm_offset_t, vm_size_t, boolean_t, vm_prot_t); extern kern_return_t vm_statistics(vm_map_t, vm_statistics_data_t *); +extern kern_return_t vm_cache_statistics(vm_map_t, vm_cache_statistics_data_t *); extern kern_return_t vm_read(vm_map_t, vm_address_t, vm_size_t, pointer_t *, vm_size_t *); extern kern_return_t vm_write(vm_map_t, vm_address_t, pointer_t, vm_size_t); diff --git a/xen/block.c b/xen/block.c index 4718891d..4253ef04 100644 --- a/xen/block.c +++ b/xen/block.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006-2009, 2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009, 2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -18,6 +18,7 @@ #include <sys/types.h> #include <mach/mig_errors.h> +#include <kern/kalloc.h> #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> #include <vm/vm_kern.h> @@ -229,12 +230,12 @@ void hyp_block_init(void) { t = hyp_store_transaction_start(); /* Get a page for ring */ - if (kmem_alloc_wired(kernel_map, &addr, PAGE_SIZE) != KERN_SUCCESS) + if ((addr = vm_page_grab_phys_addr()) == -1) panic("%s: couldn't allocate space for store ring\n", device_name); - ring = (void*) addr; + ring = (void*) phystokv(addr); SHARED_RING_INIT(ring); FRONT_RING_INIT(&bd->ring, ring, PAGE_SIZE); - grant = hyp_grant_give(domid, atop(kvtophys(addr)), 0); + grant = hyp_grant_give(domid, atop(addr), 0); /* and give it to backend. */ i = sprintf(port_name, "%u", grant); @@ -516,7 +517,7 @@ device_read (void *d, ipc_port_t reply_port, thread_block(NULL); if (err) - printf("error reading %d bytes at sector %d\n", amt, + printf("error reading %d bytes at sector %ld\n", amt, bn + offset / 512); for (i = 0; i < nbpages; i++) diff --git a/xen/block.h b/xen/block.h index 5955968a..82488855 100644 --- a/xen/block.h +++ b/xen/block.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by diff --git a/xen/configfrag.ac b/xen/configfrag.ac index eb689960..3745a31f 100644 --- a/xen/configfrag.ac +++ b/xen/configfrag.ac @@ -25,8 +25,10 @@ dnl 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. AC_DEFINE([MACH_HYP], [], [be a hypervisor guest]) AM_CONDITIONAL([PLATFORM_xen], [true]) +dnl These are experimental + AC_ARG_ENABLE([pseudo-phys], - AS_HELP_STRING([--enable-pseudo-phys], [Pseudo physical support])) + AS_HELP_STRING([--disable-pseudo-phys], [Pseudo physical pages support])) [if [ x"$enable_pseudo_phys" = xno ]; then] AM_CONDITIONAL([enable_pseudo_phys], [false]) [else] @@ -34,9 +36,39 @@ dnl 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. AM_CONDITIONAL([enable_pseudo_phys], [true]) [fi] + AC_ARG_ENABLE([pv-pagetables], + AS_HELP_STRING([--disable-pv-pagetables], [Paravirtualized page tables support])) + [if [ x"$enable_pv_pagetables" = xno ]; then] + AM_CONDITIONAL([enable_pv_pagetables], [false]) + [else] + AC_DEFINE([MACH_PV_PAGETABLES], [], [Enable paravirtualized page tables support]) + AM_CONDITIONAL([enable_pv_pagetables], [true]) + [fi] + + AC_ARG_ENABLE([pv-descriptors], + AS_HELP_STRING([--disable-pv-descriptors], [Paravirtualized segment descriptors support])) + [if [ x"$enable_pv_descriptors" = xno ]; then] + AM_CONDITIONAL([enable_pv_descriptors], [false]) + [else] + AC_DEFINE([MACH_PV_DESCRIPTORS], [], [Enable paravirtualized segment descriptors support]) + AM_CONDITIONAL([enable_pv_descriptors], [true]) + [fi] + + AC_ARG_ENABLE([ring1], + AS_HELP_STRING([--disable-ring1], [ring1 kernel support])) + [if [ x"$enable_ring1" = xno ]; then] + AM_CONDITIONAL([enable_ring1], [false]) + [else] + AC_DEFINE([MACH_RING1], [], [Enable ring1 kernel support]) + AM_CONDITIONAL([enable_ring1], [true]) + [fi] + [else] AM_CONDITIONAL([PLATFORM_xen], [false]) AM_CONDITIONAL([enable_pseudo_phys], [false]) + AM_CONDITIONAL([enable_pv_pagetables], [false]) + AM_CONDITIONAL([enable_pv_descriptors], [false]) + AM_CONDITIONAL([enable_ring1], [false]) [fi] dnl Local Variables: diff --git a/xen/console.c b/xen/console.c index 9798ec0a..884376ff 100644 --- a/xen/console.c +++ b/xen/console.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -47,9 +47,13 @@ int hypputc(int c) hyp_console_io(CONSOLEIO_write, 1, kvtolin(&d)); } else { spl_t spl = splhigh(); + int complain; simple_lock(&outlock); while (hyp_ring_smash(console->out, console->out_prod, console->out_cons)) { - hyp_console_put("ring smash\n"); + if (!complain) { + complain = 1; + hyp_console_put("ring smash\n"); + } /* TODO: are we allowed to sleep in putc? */ hyp_yield(); } @@ -216,9 +220,6 @@ int hypcnclose(int dev, int flag) int hypcnprobe(struct consdev *cp) { - struct xencons_interface *my_console; - my_console = (void*) mfn_to_kv(boot_info.console_mfn); - cp->cn_dev = makedev(0, 0); cp->cn_pri = CN_INTERNAL; return 0; @@ -231,7 +232,9 @@ int hypcninit(struct consdev *cp) simple_lock_init(&outlock); simple_lock_init(&inlock); console = (void*) mfn_to_kv(boot_info.console_mfn); +#ifdef MACH_PV_PAGETABLES pmap_set_page_readwrite(console); +#endif /* MACH_PV_PAGETABLES */ hyp_evt_handler(boot_info.console_evtchn, hypcnintr, 0, SPL6); return 0; } diff --git a/xen/console.h b/xen/console.h index fa13dc0f..ad171a47 100644 --- a/xen/console.h +++ b/xen/console.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009, 2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -21,6 +21,8 @@ #include <machine/xen.h> #include <string.h> +#include <device/cons.h> + #define hyp_console_write(str, len) hyp_console_io (CONSOLEIO_write, (len), kvtolin(str)) #define hyp_console_put(str) ({ \ @@ -30,4 +32,9 @@ extern void hyp_console_init(void); +extern int hypcnputc(dev_t dev, int c); +extern int hypcngetc(dev_t dev, int wait); +extern int hypcnprobe(struct consdev *cp); +extern int hypcninit(struct consdev *cp); + #endif /* XEN_CONSOLE_H */ @@ -1,5 +1,5 @@ /* - * Copyright (C) 2007 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2007-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -62,12 +62,12 @@ void hyp_c_callback(void *ret_addr, void *regs) if (ivect[n]) { spl_t spl = splx(intpri[n]); - asm ("lock; andl %1,%0":"=m"(hyp_shared_info.evtchn_pending[i]):"r"(~(1<<j))); + asm ("lock; and %1,%0":"=m"(hyp_shared_info.evtchn_pending[i]):"r"(~(1UL<<j))); ivect[n](iunit[n], spl, ret_addr, regs); splx_cli(spl); } else { printf("warning: lost unbound event %d\n", n); - asm ("lock; andl %1,%0":"=m"(hyp_shared_info.evtchn_pending[i]):"r"(~(1<<j))); + asm ("lock; and %1,%0":"=m"(hyp_shared_info.evtchn_pending[i]):"r"(~(1UL<<j))); } } } @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by diff --git a/xen/grant.c b/xen/grant.c index 505d2026..3d5c3fe7 100644 --- a/xen/grant.c +++ b/xen/grant.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by diff --git a/xen/grant.h b/xen/grant.h index ff8617d4..30ce344d 100644 --- a/xen/grant.h +++ b/xen/grant.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006-2009, 2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009, 2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -18,6 +18,7 @@ #include <sys/types.h> #include <mach/mig_errors.h> +#include <kern/kalloc.h> #include <ipc/ipc_port.h> #include <ipc/ipc_space.h> #include <vm/vm_kern.h> @@ -117,6 +118,61 @@ static void enqueue_rx_buf(struct net_data *nd, int number) { req->gref = nd->rx_buf_gnt[number] = gref; } +static int recompute_checksum(void *data, int len) { + unsigned16_t *header16 = data; + unsigned8_t *header8 = data; + unsigned length, i; + unsigned32_t checksum = 0; + + /* IPv4 header length */ + length = (header8[0] & 0xf) * 4; + if (length < 20) + /* Too small for an IP header16 */ + return -1; + if (length > len) + /* Does not fit in the ethernet frame */ + return -1; + + /* Compute IP header checksum */ + header16[5] = 0; + for (i = 0; i < length/2; i++) + checksum += ntohs(header16[i]); + + while (checksum >> 16) + checksum = (checksum & 0xffff) + (checksum >> 16); + + header16[5] = htons(~checksum); + + if (header8[9] == 6) { + /* Need to fix TCP checksum as well */ + unsigned16_t *tcp_header16 = header16 + length/2; + unsigned8_t *tcp_header8 = header8 + length; + unsigned tcp_length = ntohs(header16[1]) - length; + + /* Pseudo IP header */ + checksum = ntohs(header16[6]) + ntohs(header16[7]) + + ntohs(header16[8]) + ntohs(header16[9]) + + header8[9] + tcp_length; + + tcp_header16[8] = 0; + for (i = 0; i < tcp_length / 2; i++) + checksum += ntohs(tcp_header16[i]); + if (tcp_length & 1) + checksum += tcp_header8[tcp_length-1] << 8; + + while (checksum >> 16) + checksum = (checksum & 0xffff) + (checksum >> 16); + + tcp_header16[8] = htons(~checksum); + } else if (header8[9] == 17) { + /* Drop any bogus checksum */ + unsigned16_t *udp_header16 = header16 + length/2; + udp_header16[3] = 0; + } + + return 0; +} + static void hyp_net_intr(int unit) { ipc_kmsg_t kmsg; struct ether_header *eh; @@ -172,6 +228,25 @@ static void hyp_net_intr(int unit) { data = nd->rx_buf[number] + rx_rsp->offset; len = rx_rsp->status; + if (rx_rsp->flags & NETRXF_csum_blank) { + struct ether_header *ether = data; + + if (!(rx_rsp->flags & NETRXF_data_validated)) { + printf("packet with no checksum and not validated, dropping it\n"); + goto drop_kmsg; + } + + /* TODO: rather tell pfinet to ignore checksums */ + + if (ntohs(ether->ether_type) != 0x0800) { + printf("packet with no checksum and not IPv4, dropping it\n"); + goto drop_kmsg; + } + + if (recompute_checksum(data + sizeof(*ether), len - sizeof(*ether))) + goto drop_kmsg; + } + eh = (void*) (net_kmsg(kmsg)->header); ph = (void*) (net_kmsg(kmsg)->packet); memcpy(eh, data, sizeof (struct ether_header)); @@ -283,12 +358,12 @@ void hyp_net_init(void) { t = hyp_store_transaction_start(); /* Get a page for tx_ring */ - if (kmem_alloc_wired(kernel_map, &addr, PAGE_SIZE) != KERN_SUCCESS) + if ((addr = vm_page_grab_phys_addr()) == -1) panic("eth: couldn't allocate space for store tx_ring"); - tx_ring = (void*) addr; + tx_ring = (void*) phystokv(addr); SHARED_RING_INIT(tx_ring); FRONT_RING_INIT(&nd->tx, tx_ring, PAGE_SIZE); - grant = hyp_grant_give(domid, atop(kvtophys(addr)), 0); + grant = hyp_grant_give(domid, atop(addr), 0); /* and give it to backend. */ i = sprintf(port_name, "%u", grant); @@ -298,12 +373,12 @@ void hyp_net_init(void) { kfree((vm_offset_t) c, strlen(c)+1); /* Get a page for rx_ring */ - if (kmem_alloc_wired(kernel_map, &addr, PAGE_SIZE) != KERN_SUCCESS) + if ((addr = vm_page_grab_phys_addr()) == -1) panic("eth: couldn't allocate space for store tx_ring"); - rx_ring = (void*) addr; + rx_ring = (void*) phystokv(addr); SHARED_RING_INIT(rx_ring); FRONT_RING_INIT(&nd->rx, rx_ring, PAGE_SIZE); - grant = hyp_grant_give(domid, atop(kvtophys(addr)), 0); + grant = hyp_grant_give(domid, atop(addr), 0); /* and give it to backend. */ i = sprintf(port_name, "%u", grant); @@ -396,12 +471,12 @@ void hyp_net_init(void) { /* Get a page for packet reception */ for (i= 0; i<WINDOW; i++) { - if (kmem_alloc_wired(kernel_map, &addr, PAGE_SIZE) != KERN_SUCCESS) + if ((addr = vm_page_grab_phys_addr()) == -1) panic("eth: couldn't allocate space for store tx_ring"); - nd->rx_buf[i] = (void*)phystokv(kvtophys(addr)); - nd->rx_buf_pfn[i] = atop(kvtophys((vm_offset_t)nd->rx_buf[i])); + nd->rx_buf[i] = (void*)phystokv(addr); + nd->rx_buf_pfn[i] = atop(addr); if (!nd->rx_copy) { - if (hyp_do_update_va_mapping(kvtolin(addr), 0, UVMF_INVLPG|UVMF_ALL)) + if (hyp_do_update_va_mapping(kvtolin(nd->rx_buf[i]), 0, UVMF_INVLPG|UVMF_ALL)) panic("eth: couldn't clear rx kv buf %d at %p", i, addr); } /* and enqueue it to backend. */ @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by diff --git a/xen/store.c b/xen/store.c index 3c6baebf..739dc367 100644 --- a/xen/store.c +++ b/xen/store.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -20,6 +20,7 @@ #include <mach/mig_support.h> #include <machine/pmap.h> #include <machine/ipl.h> +#include <kern/kalloc.h> #include <stdarg.h> #include <string.h> #include <alloca.h> @@ -328,7 +329,9 @@ void hyp_store_init(void) return; simple_lock_init(&lock); store = (void*) mfn_to_kv(boot_info.store_mfn); +#ifdef MACH_PV_PAGETABLES pmap_set_page_readwrite(store); +#endif /* MACH_PV_PAGETABLES */ /* SPL sched */ hyp_evt_handler(boot_info.store_evtchn, hyp_store_handler, 0, SPL7); } diff --git a/xen/store.h b/xen/store.h index 4b3ee187..ae236eb6 100644 --- a/xen/store.h +++ b/xen/store.h @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -116,11 +116,6 @@ readtodc(tp) unsigned64_t t = hyp_get_time(); u_int n = t / 1000000000; -#ifndef MACH_KERNEL - n += tz.tz_minuteswest * 60; - if (tz.tz_dsttime) - n -= 3600; -#endif /* MACH_KERNEL */ *tp = n; return(0); @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -1,5 +1,5 @@ /* - * Copyright (C) 2007-2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2007-2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -30,13 +30,6 @@ #include "xen.h" #include "evt.h" -void hyp_invalidate_pte(pt_entry_t *pte) -{ - if (!hyp_mmu_update_pte(kv_to_ma(pte), (*pte) & ~INTEL_PTE_VALID)) - panic("%s:%d could not set pte %p(%p) to %p(%p)\n",__FILE__,__LINE__,pte,(vm_offset_t) kv_to_ma(pte),*pte,ma_to_pa(*pte)); - hyp_mmuext_op_void(MMUEXT_TLB_FLUSH_LOCAL); -} - void hyp_debug() { panic("debug"); @@ -1,5 +1,5 @@ /* - * Copyright (C) 2006-2010, 2011 Samuel Thibault <samuel.thibault@ens-lyon.org> + * Copyright (C) 2006-2009, 2011 Free Software Foundation * * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -21,7 +21,6 @@ void hyp_init(void); void hyp_dev_init(void); -void hyp_invalidate_pte(pt_entry_t *pte); void hyp_idle(void); void hyp_p2m_init(void); |