summaryrefslogtreecommitdiff
path: root/lib
diff options
context:
space:
mode:
Diffstat (limited to 'lib')
-rw-r--r--lib/Kconfig.debug7
-rw-r--r--lib/dynamic_debug.c8
-rw-r--r--lib/genalloc.c4
-rw-r--r--lib/iov_iter.c1
-rw-r--r--lib/kobject.c62
-rw-r--r--lib/math/Makefile2
-rw-r--r--lib/math/int_log.c133
-rw-r--r--lib/percpu_counter.c62
-rw-r--r--lib/raid6/Makefile1
-rw-r--r--lib/raid6/algos.c16
-rw-r--r--lib/raid6/loongarch.h38
-rw-r--r--lib/raid6/loongarch_simd.c422
-rw-r--r--lib/raid6/recov_loongarch_simd.c513
-rw-r--r--lib/raid6/test/Makefile12
-rw-r--r--lib/string_helpers.c15
-rw-r--r--lib/test_scanf.c2
16 files changed, 1242 insertions, 56 deletions
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index a8a1b0ac8b22..319cfbeb0738 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -355,6 +355,11 @@ endchoice # "Compressed Debug information"
config DEBUG_INFO_SPLIT
bool "Produce split debuginfo in .dwo files"
depends on $(cc-option,-gsplit-dwarf)
+ # RISC-V linker relaxation + -gsplit-dwarf has issues with LLVM and GCC
+ # prior to 12.x:
+ # https://github.com/llvm/llvm-project/issues/56642
+ # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99090
+ depends on !RISCV || GCC_VERSION >= 120000
help
Generate debug info into separate .dwo files. This significantly
reduces the build directory size for builds with DEBUG_INFO,
@@ -502,7 +507,7 @@ config SECTION_MISMATCH_WARN_ONLY
config DEBUG_FORCE_FUNCTION_ALIGN_64B
bool "Force all function address 64B aligned"
- depends on EXPERT && (X86_64 || ARM64 || PPC32 || PPC64 || ARC || S390)
+ depends on EXPERT && (X86_64 || ARM64 || PPC32 || PPC64 || ARC || RISCV || S390)
select FUNCTION_ALIGNMENT_64B
help
There are cases that a commit from one domain changes the function
diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
index fdd6d9800a70..6fba6423cc10 100644
--- a/lib/dynamic_debug.c
+++ b/lib/dynamic_debug.c
@@ -88,10 +88,11 @@ static inline const char *trim_prefix(const char *path)
return path + skip;
}
-static struct { unsigned flag:8; char opt_char; } opt_array[] = {
+static const struct { unsigned flag:8; char opt_char; } opt_array[] = {
{ _DPRINTK_FLAGS_PRINT, 'p' },
{ _DPRINTK_FLAGS_INCL_MODNAME, 'm' },
{ _DPRINTK_FLAGS_INCL_FUNCNAME, 'f' },
+ { _DPRINTK_FLAGS_INCL_SOURCENAME, 's' },
{ _DPRINTK_FLAGS_INCL_LINENO, 'l' },
{ _DPRINTK_FLAGS_INCL_TID, 't' },
{ _DPRINTK_FLAGS_NONE, '_' },
@@ -808,7 +809,7 @@ const struct kernel_param_ops param_ops_dyndbg_classes = {
};
EXPORT_SYMBOL(param_ops_dyndbg_classes);
-#define PREFIX_SIZE 64
+#define PREFIX_SIZE 128
static int remaining(int wrote)
{
@@ -836,6 +837,9 @@ static char *__dynamic_emit_prefix(const struct _ddebug *desc, char *buf)
if (desc->flags & _DPRINTK_FLAGS_INCL_FUNCNAME)
pos += snprintf(buf + pos, remaining(pos), "%s:",
desc->function);
+ if (desc->flags & _DPRINTK_FLAGS_INCL_SOURCENAME)
+ pos += snprintf(buf + pos, remaining(pos), "%s:",
+ trim_prefix(desc->filename));
if (desc->flags & _DPRINTK_FLAGS_INCL_LINENO)
pos += snprintf(buf + pos, remaining(pos), "%d:",
desc->lineno);
diff --git a/lib/genalloc.c b/lib/genalloc.c
index 6c644f954bc5..4fa5635bf81b 100644
--- a/lib/genalloc.c
+++ b/lib/genalloc.c
@@ -32,7 +32,9 @@
#include <linux/rculist.h>
#include <linux/interrupt.h>
#include <linux/genalloc.h>
-#include <linux/of_device.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
#include <linux/vmalloc.h>
static inline size_t chunk_size(const struct gen_pool_chunk *chunk)
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 424737045b97..b31597b0ca20 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -1557,6 +1557,7 @@ int import_ubuf(int rw, void __user *buf, size_t len, struct iov_iter *i)
iov_iter_ubuf(i, rw, buf, len);
return 0;
}
+EXPORT_SYMBOL_GPL(import_ubuf);
/**
* iov_iter_restore() - Restore a &struct iov_iter to the same state as when
diff --git a/lib/kobject.c b/lib/kobject.c
index 16d530f9c174..59dbcbdb1c91 100644
--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -56,6 +56,14 @@ void kobject_get_ownership(const struct kobject *kobj, kuid_t *uid, kgid_t *gid)
kobj->ktype->get_ownership(kobj, uid, gid);
}
+static bool kobj_ns_type_is_valid(enum kobj_ns_type type)
+{
+ if ((type <= KOBJ_NS_TYPE_NONE) || (type >= KOBJ_NS_TYPES))
+ return false;
+
+ return true;
+}
+
static int create_dir(struct kobject *kobj)
{
const struct kobj_type *ktype = get_ktype(kobj);
@@ -66,12 +74,10 @@ static int create_dir(struct kobject *kobj)
if (error)
return error;
- if (ktype) {
- error = sysfs_create_groups(kobj, ktype->default_groups);
- if (error) {
- sysfs_remove_dir(kobj);
- return error;
- }
+ error = sysfs_create_groups(kobj, ktype->default_groups);
+ if (error) {
+ sysfs_remove_dir(kobj);
+ return error;
}
/*
@@ -86,8 +92,7 @@ static int create_dir(struct kobject *kobj)
*/
ops = kobj_child_ns_ops(kobj);
if (ops) {
- BUG_ON(ops->type <= KOBJ_NS_TYPE_NONE);
- BUG_ON(ops->type >= KOBJ_NS_TYPES);
+ BUG_ON(!kobj_ns_type_is_valid(ops->type));
BUG_ON(!kobj_ns_type_registered(ops->type));
sysfs_enable_ns(kobj->sd);
@@ -584,8 +589,7 @@ static void __kobject_del(struct kobject *kobj)
sd = kobj->sd;
ktype = get_ktype(kobj);
- if (ktype)
- sysfs_remove_groups(kobj, ktype->default_groups);
+ sysfs_remove_groups(kobj, ktype->default_groups);
/* send "remove" if the caller did not do it but sent "add" */
if (kobj->state_add_uevent_sent && !kobj->state_remove_uevent_sent) {
@@ -662,10 +666,6 @@ static void kobject_cleanup(struct kobject *kobj)
pr_debug("'%s' (%p): %s, parent %p\n",
kobject_name(kobj), kobj, __func__, kobj->parent);
- if (t && !t->release)
- pr_debug("'%s' (%p): does not have a release() function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",
- kobject_name(kobj), kobj);
-
/* remove from sysfs if the caller did not do it */
if (kobj->state_in_sysfs) {
pr_debug("'%s' (%p): auto cleanup kobject_del\n",
@@ -676,10 +676,13 @@ static void kobject_cleanup(struct kobject *kobj)
parent = NULL;
}
- if (t && t->release) {
+ if (t->release) {
pr_debug("'%s' (%p): calling ktype release\n",
kobject_name(kobj), kobj);
t->release(kobj);
+ } else {
+ pr_debug("'%s' (%p): does not have a release() function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",
+ kobject_name(kobj), kobj);
}
/* free name if we allocated it */
@@ -854,6 +857,11 @@ int kset_register(struct kset *k)
if (!k)
return -EINVAL;
+ if (!k->kobj.ktype) {
+ pr_err("must have a ktype to be initialized properly!\n");
+ return -EINVAL;
+ }
+
kset_init(k);
err = kobject_add_internal(&k->kobj);
if (err) {
@@ -1017,11 +1025,7 @@ int kobj_ns_type_register(const struct kobj_ns_type_operations *ops)
spin_lock(&kobj_ns_type_lock);
error = -EINVAL;
- if (type >= KOBJ_NS_TYPES)
- goto out;
-
- error = -EINVAL;
- if (type <= KOBJ_NS_TYPE_NONE)
+ if (!kobj_ns_type_is_valid(type))
goto out;
error = -EBUSY;
@@ -1041,7 +1045,7 @@ int kobj_ns_type_registered(enum kobj_ns_type type)
int registered = 0;
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES))
+ if (kobj_ns_type_is_valid(type))
registered = kobj_ns_ops_tbl[type] != NULL;
spin_unlock(&kobj_ns_type_lock);
@@ -1052,7 +1056,7 @@ const struct kobj_ns_type_operations *kobj_child_ns_ops(const struct kobject *pa
{
const struct kobj_ns_type_operations *ops = NULL;
- if (parent && parent->ktype && parent->ktype->child_ns_type)
+ if (parent && parent->ktype->child_ns_type)
ops = parent->ktype->child_ns_type(parent);
return ops;
@@ -1068,8 +1072,7 @@ bool kobj_ns_current_may_mount(enum kobj_ns_type type)
bool may_mount = true;
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES) &&
- kobj_ns_ops_tbl[type])
+ if (kobj_ns_type_is_valid(type) && kobj_ns_ops_tbl[type])
may_mount = kobj_ns_ops_tbl[type]->current_may_mount();
spin_unlock(&kobj_ns_type_lock);
@@ -1081,8 +1084,7 @@ void *kobj_ns_grab_current(enum kobj_ns_type type)
void *ns = NULL;
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES) &&
- kobj_ns_ops_tbl[type])
+ if (kobj_ns_type_is_valid(type) && kobj_ns_ops_tbl[type])
ns = kobj_ns_ops_tbl[type]->grab_current_ns();
spin_unlock(&kobj_ns_type_lock);
@@ -1095,8 +1097,7 @@ const void *kobj_ns_netlink(enum kobj_ns_type type, struct sock *sk)
const void *ns = NULL;
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES) &&
- kobj_ns_ops_tbl[type])
+ if (kobj_ns_type_is_valid(type) && kobj_ns_ops_tbl[type])
ns = kobj_ns_ops_tbl[type]->netlink_ns(sk);
spin_unlock(&kobj_ns_type_lock);
@@ -1108,8 +1109,7 @@ const void *kobj_ns_initial(enum kobj_ns_type type)
const void *ns = NULL;
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES) &&
- kobj_ns_ops_tbl[type])
+ if (kobj_ns_type_is_valid(type) && kobj_ns_ops_tbl[type])
ns = kobj_ns_ops_tbl[type]->initial_ns();
spin_unlock(&kobj_ns_type_lock);
@@ -1119,7 +1119,7 @@ const void *kobj_ns_initial(enum kobj_ns_type type)
void kobj_ns_drop(enum kobj_ns_type type, void *ns)
{
spin_lock(&kobj_ns_type_lock);
- if ((type > KOBJ_NS_TYPE_NONE) && (type < KOBJ_NS_TYPES) &&
+ if (kobj_ns_type_is_valid(type) &&
kobj_ns_ops_tbl[type] && kobj_ns_ops_tbl[type]->drop_ns)
kobj_ns_ops_tbl[type]->drop_ns(ns);
spin_unlock(&kobj_ns_type_lock);
diff --git a/lib/math/Makefile b/lib/math/Makefile
index bfac26ddfc22..91fcdb0c9efe 100644
--- a/lib/math/Makefile
+++ b/lib/math/Makefile
@@ -1,5 +1,5 @@
# SPDX-License-Identifier: GPL-2.0-only
-obj-y += div64.o gcd.o lcm.o int_pow.o int_sqrt.o reciprocal_div.o
+obj-y += div64.o gcd.o lcm.o int_log.o int_pow.o int_sqrt.o reciprocal_div.o
obj-$(CONFIG_CORDIC) += cordic.o
obj-$(CONFIG_PRIME_NUMBERS) += prime_numbers.o
diff --git a/lib/math/int_log.c b/lib/math/int_log.c
new file mode 100644
index 000000000000..8f9da3a2ad39
--- /dev/null
+++ b/lib/math/int_log.c
@@ -0,0 +1,133 @@
+// SPDX-License-Identifier: LGPL-2.1-or-later
+/*
+ * Provides fixed-point logarithm operations.
+ *
+ * Copyright (C) 2006 Christoph Pfister (christophpfister@gmail.com)
+ */
+
+#include <linux/bitops.h>
+#include <linux/export.h>
+#include <linux/int_log.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+#include <asm/bug.h>
+
+static const unsigned short logtable[256] = {
+ 0x0000, 0x0171, 0x02e0, 0x044e, 0x05ba, 0x0725, 0x088e, 0x09f7,
+ 0x0b5d, 0x0cc3, 0x0e27, 0x0f8a, 0x10eb, 0x124b, 0x13aa, 0x1508,
+ 0x1664, 0x17bf, 0x1919, 0x1a71, 0x1bc8, 0x1d1e, 0x1e73, 0x1fc6,
+ 0x2119, 0x226a, 0x23ba, 0x2508, 0x2656, 0x27a2, 0x28ed, 0x2a37,
+ 0x2b80, 0x2cc8, 0x2e0f, 0x2f54, 0x3098, 0x31dc, 0x331e, 0x345f,
+ 0x359f, 0x36de, 0x381b, 0x3958, 0x3a94, 0x3bce, 0x3d08, 0x3e41,
+ 0x3f78, 0x40af, 0x41e4, 0x4319, 0x444c, 0x457f, 0x46b0, 0x47e1,
+ 0x4910, 0x4a3f, 0x4b6c, 0x4c99, 0x4dc5, 0x4eef, 0x5019, 0x5142,
+ 0x526a, 0x5391, 0x54b7, 0x55dc, 0x5700, 0x5824, 0x5946, 0x5a68,
+ 0x5b89, 0x5ca8, 0x5dc7, 0x5ee5, 0x6003, 0x611f, 0x623a, 0x6355,
+ 0x646f, 0x6588, 0x66a0, 0x67b7, 0x68ce, 0x69e4, 0x6af8, 0x6c0c,
+ 0x6d20, 0x6e32, 0x6f44, 0x7055, 0x7165, 0x7274, 0x7383, 0x7490,
+ 0x759d, 0x76aa, 0x77b5, 0x78c0, 0x79ca, 0x7ad3, 0x7bdb, 0x7ce3,
+ 0x7dea, 0x7ef0, 0x7ff6, 0x80fb, 0x81ff, 0x8302, 0x8405, 0x8507,
+ 0x8608, 0x8709, 0x8809, 0x8908, 0x8a06, 0x8b04, 0x8c01, 0x8cfe,
+ 0x8dfa, 0x8ef5, 0x8fef, 0x90e9, 0x91e2, 0x92db, 0x93d2, 0x94ca,
+ 0x95c0, 0x96b6, 0x97ab, 0x98a0, 0x9994, 0x9a87, 0x9b7a, 0x9c6c,
+ 0x9d5e, 0x9e4f, 0x9f3f, 0xa02e, 0xa11e, 0xa20c, 0xa2fa, 0xa3e7,
+ 0xa4d4, 0xa5c0, 0xa6ab, 0xa796, 0xa881, 0xa96a, 0xaa53, 0xab3c,
+ 0xac24, 0xad0c, 0xadf2, 0xaed9, 0xafbe, 0xb0a4, 0xb188, 0xb26c,
+ 0xb350, 0xb433, 0xb515, 0xb5f7, 0xb6d9, 0xb7ba, 0xb89a, 0xb97a,
+ 0xba59, 0xbb38, 0xbc16, 0xbcf4, 0xbdd1, 0xbead, 0xbf8a, 0xc065,
+ 0xc140, 0xc21b, 0xc2f5, 0xc3cf, 0xc4a8, 0xc580, 0xc658, 0xc730,
+ 0xc807, 0xc8de, 0xc9b4, 0xca8a, 0xcb5f, 0xcc34, 0xcd08, 0xcddc,
+ 0xceaf, 0xcf82, 0xd054, 0xd126, 0xd1f7, 0xd2c8, 0xd399, 0xd469,
+ 0xd538, 0xd607, 0xd6d6, 0xd7a4, 0xd872, 0xd93f, 0xda0c, 0xdad9,
+ 0xdba5, 0xdc70, 0xdd3b, 0xde06, 0xded0, 0xdf9a, 0xe063, 0xe12c,
+ 0xe1f5, 0xe2bd, 0xe385, 0xe44c, 0xe513, 0xe5d9, 0xe69f, 0xe765,
+ 0xe82a, 0xe8ef, 0xe9b3, 0xea77, 0xeb3b, 0xebfe, 0xecc1, 0xed83,
+ 0xee45, 0xef06, 0xefc8, 0xf088, 0xf149, 0xf209, 0xf2c8, 0xf387,
+ 0xf446, 0xf505, 0xf5c3, 0xf680, 0xf73e, 0xf7fb, 0xf8b7, 0xf973,
+ 0xfa2f, 0xfaea, 0xfba5, 0xfc60, 0xfd1a, 0xfdd4, 0xfe8e, 0xff47,
+};
+
+unsigned int intlog2(u32 value)
+{
+ /**
+ * returns: log2(value) * 2^24
+ * wrong result if value = 0 (log2(0) is undefined)
+ */
+ unsigned int msb;
+ unsigned int logentry;
+ unsigned int significand;
+ unsigned int interpolation;
+
+ if (unlikely(value == 0)) {
+ WARN_ON(1);
+ return 0;
+ }
+
+ /* first detect the msb (count begins at 0) */
+ msb = fls(value) - 1;
+
+ /**
+ * now we use a logtable after the following method:
+ *
+ * log2(2^x * y) * 2^24 = x * 2^24 + log2(y) * 2^24
+ * where x = msb and therefore 1 <= y < 2
+ * first y is determined by shifting the value left
+ * so that msb is bit 31
+ * 0x00231f56 -> 0x8C7D5800
+ * the result is y * 2^31 -> "significand"
+ * then the highest 9 bits are used for a table lookup
+ * the highest bit is discarded because it's always set
+ * the highest nine bits in our example are 100011000
+ * so we would use the entry 0x18
+ */
+ significand = value << (31 - msb);
+ logentry = (significand >> 23) % ARRAY_SIZE(logtable);
+
+ /**
+ * last step we do is interpolation because of the
+ * limitations of the log table the error is that part of
+ * the significand which isn't used for lookup then we
+ * compute the ratio between the error and the next table entry
+ * and interpolate it between the log table entry used and the
+ * next one the biggest error possible is 0x7fffff
+ * (in our example it's 0x7D5800)
+ * needed value for next table entry is 0x800000
+ * so the interpolation is
+ * (error / 0x800000) * (logtable_next - logtable_current)
+ * in the implementation the division is moved to the end for
+ * better accuracy there is also an overflow correction if
+ * logtable_next is 256
+ */
+ interpolation = ((significand & 0x7fffff) *
+ ((logtable[(logentry + 1) % ARRAY_SIZE(logtable)] -
+ logtable[logentry]) & 0xffff)) >> 15;
+
+ /* now we return the result */
+ return ((msb << 24) + (logtable[logentry] << 8) + interpolation);
+}
+EXPORT_SYMBOL(intlog2);
+
+unsigned int intlog10(u32 value)
+{
+ /**
+ * returns: log10(value) * 2^24
+ * wrong result if value = 0 (log10(0) is undefined)
+ */
+ u64 log;
+
+ if (unlikely(value == 0)) {
+ WARN_ON(1);
+ return 0;
+ }
+
+ log = intlog2(value);
+
+ /**
+ * we use the following method:
+ * log10(x) = log2(x) * log10(2)
+ */
+
+ return (log * 646456993) >> 31;
+}
+EXPORT_SYMBOL(intlog10);
diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c
index 5004463c4f9f..9073430dc865 100644
--- a/lib/percpu_counter.c
+++ b/lib/percpu_counter.c
@@ -151,48 +151,72 @@ s64 __percpu_counter_sum(struct percpu_counter *fbc)
}
EXPORT_SYMBOL(__percpu_counter_sum);
-int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
- struct lock_class_key *key)
+int __percpu_counter_init_many(struct percpu_counter *fbc, s64 amount,
+ gfp_t gfp, u32 nr_counters,
+ struct lock_class_key *key)
{
unsigned long flags __maybe_unused;
-
- raw_spin_lock_init(&fbc->lock);
- lockdep_set_class(&fbc->lock, key);
- fbc->count = amount;
- fbc->counters = alloc_percpu_gfp(s32, gfp);
- if (!fbc->counters)
+ size_t counter_size;
+ s32 __percpu *counters;
+ u32 i;
+
+ counter_size = ALIGN(sizeof(*counters), __alignof__(*counters));
+ counters = __alloc_percpu_gfp(nr_counters * counter_size,
+ __alignof__(*counters), gfp);
+ if (!counters) {
+ fbc[0].counters = NULL;
return -ENOMEM;
+ }
- debug_percpu_counter_activate(fbc);
+ for (i = 0; i < nr_counters; i++) {
+ raw_spin_lock_init(&fbc[i].lock);
+ lockdep_set_class(&fbc[i].lock, key);
+#ifdef CONFIG_HOTPLUG_CPU
+ INIT_LIST_HEAD(&fbc[i].list);
+#endif
+ fbc[i].count = amount;
+ fbc[i].counters = (void *)counters + (i * counter_size);
+
+ debug_percpu_counter_activate(&fbc[i]);
+ }
#ifdef CONFIG_HOTPLUG_CPU
- INIT_LIST_HEAD(&fbc->list);
spin_lock_irqsave(&percpu_counters_lock, flags);
- list_add(&fbc->list, &percpu_counters);
+ for (i = 0; i < nr_counters; i++)
+ list_add(&fbc[i].list, &percpu_counters);
spin_unlock_irqrestore(&percpu_counters_lock, flags);
#endif
return 0;
}
-EXPORT_SYMBOL(__percpu_counter_init);
+EXPORT_SYMBOL(__percpu_counter_init_many);
-void percpu_counter_destroy(struct percpu_counter *fbc)
+void percpu_counter_destroy_many(struct percpu_counter *fbc, u32 nr_counters)
{
unsigned long flags __maybe_unused;
+ u32 i;
+
+ if (WARN_ON_ONCE(!fbc))
+ return;
- if (!fbc->counters)
+ if (!fbc[0].counters)
return;
- debug_percpu_counter_deactivate(fbc);
+ for (i = 0; i < nr_counters; i++)
+ debug_percpu_counter_deactivate(&fbc[i]);
#ifdef CONFIG_HOTPLUG_CPU
spin_lock_irqsave(&percpu_counters_lock, flags);
- list_del(&fbc->list);
+ for (i = 0; i < nr_counters; i++)
+ list_del(&fbc[i].list);
spin_unlock_irqrestore(&percpu_counters_lock, flags);
#endif
- free_percpu(fbc->counters);
- fbc->counters = NULL;
+
+ free_percpu(fbc[0].counters);
+
+ for (i = 0; i < nr_counters; i++)
+ fbc[i].counters = NULL;
}
-EXPORT_SYMBOL(percpu_counter_destroy);
+EXPORT_SYMBOL(percpu_counter_destroy_many);
int percpu_counter_batch __read_mostly = 32;
EXPORT_SYMBOL(percpu_counter_batch);
diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile
index 45e17619422b..035b0a4db476 100644
--- a/lib/raid6/Makefile
+++ b/lib/raid6/Makefile
@@ -9,6 +9,7 @@ raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o \
vpermxor1.o vpermxor2.o vpermxor4.o vpermxor8.o
raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o recov_neon.o recov_neon_inner.o
raid6_pq-$(CONFIG_S390) += s390vx8.o recov_s390xc.o
+raid6_pq-$(CONFIG_LOONGARCH) += loongarch_simd.o recov_loongarch_simd.o
hostprogs += mktables
diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c
index a22a05c9af8a..0ec534faf019 100644
--- a/lib/raid6/algos.c
+++ b/lib/raid6/algos.c
@@ -73,6 +73,14 @@ const struct raid6_calls * const raid6_algos[] = {
&raid6_neonx2,
&raid6_neonx1,
#endif
+#ifdef CONFIG_LOONGARCH
+#ifdef CONFIG_CPU_HAS_LASX
+ &raid6_lasx,
+#endif
+#ifdef CONFIG_CPU_HAS_LSX
+ &raid6_lsx,
+#endif
+#endif
#if defined(__ia64__)
&raid6_intx32,
&raid6_intx16,
@@ -104,6 +112,14 @@ const struct raid6_recov_calls *const raid6_recov_algos[] = {
#if defined(CONFIG_KERNEL_MODE_NEON)
&raid6_recov_neon,
#endif
+#ifdef CONFIG_LOONGARCH
+#ifdef CONFIG_CPU_HAS_LASX
+ &raid6_recov_lasx,
+#endif
+#ifdef CONFIG_CPU_HAS_LSX
+ &raid6_recov_lsx,
+#endif
+#endif
&raid6_recov_intx1,
NULL
};
diff --git a/lib/raid6/loongarch.h b/lib/raid6/loongarch.h
new file mode 100644
index 000000000000..acfc33ce7056
--- /dev/null
+++ b/lib/raid6/loongarch.h
@@ -0,0 +1,38 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) 2023 WANG Xuerui <git@xen0n.name>
+ *
+ * raid6/loongarch.h
+ *
+ * Definitions common to LoongArch RAID-6 code only
+ */
+
+#ifndef _LIB_RAID6_LOONGARCH_H
+#define _LIB_RAID6_LOONGARCH_H
+
+#ifdef __KERNEL__
+
+#include <asm/cpu-features.h>
+#include <asm/fpu.h>
+
+#else /* for user-space testing */
+
+#include <sys/auxv.h>
+
+/* have to supply these defines for glibc 2.37- and musl */
+#ifndef HWCAP_LOONGARCH_LSX
+#define HWCAP_LOONGARCH_LSX (1 << 4)
+#endif
+#ifndef HWCAP_LOONGARCH_LASX
+#define HWCAP_LOONGARCH_LASX (1 << 5)
+#endif
+
+#define kernel_fpu_begin()
+#define kernel_fpu_end()
+
+#define cpu_has_lsx (getauxval(AT_HWCAP) & HWCAP_LOONGARCH_LSX)
+#define cpu_has_lasx (getauxval(AT_HWCAP) & HWCAP_LOONGARCH_LASX)
+
+#endif /* __KERNEL__ */
+
+#endif /* _LIB_RAID6_LOONGARCH_H */
diff --git a/lib/raid6/loongarch_simd.c b/lib/raid6/loongarch_simd.c
new file mode 100644
index 000000000000..aa5d9f924ca3
--- /dev/null
+++ b/lib/raid6/loongarch_simd.c
@@ -0,0 +1,422 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * RAID6 syndrome calculations in LoongArch SIMD (LSX & LASX)
+ *
+ * Copyright 2023 WANG Xuerui <git@xen0n.name>
+ *
+ * Based on the generic RAID-6 code (int.uc):
+ *
+ * Copyright 2002-2004 H. Peter Anvin
+ */
+
+#include <linux/raid/pq.h>
+#include "loongarch.h"
+
+/*
+ * The vector algorithms are currently priority 0, which means the generic
+ * scalar algorithms are not being disabled if vector support is present.
+ * This is like the similar LoongArch RAID5 XOR code, with the main reason
+ * repeated here: it cannot be ruled out at this point of time, that some
+ * future (maybe reduced) models could run the vector algorithms slower than
+ * the scalar ones, maybe for errata or micro-op reasons. It may be
+ * appropriate to revisit this after one or two more uarch generations.
+ */
+
+#ifdef CONFIG_CPU_HAS_LSX
+#define NSIZE 16
+
+static int raid6_has_lsx(void)
+{
+ return cpu_has_lsx;
+}
+
+static void raid6_lsx_gen_syndrome(int disks, size_t bytes, void **ptrs)
+{
+ u8 **dptr = (u8 **)ptrs;
+ u8 *p, *q;
+ int d, z, z0;
+
+ z0 = disks - 3; /* Highest data disk */
+ p = dptr[z0+1]; /* XOR parity */
+ q = dptr[z0+2]; /* RS syndrome */
+
+ kernel_fpu_begin();
+
+ /*
+ * $vr0, $vr1, $vr2, $vr3: wp
+ * $vr4, $vr5, $vr6, $vr7: wq
+ * $vr8, $vr9, $vr10, $vr11: wd
+ * $vr12, $vr13, $vr14, $vr15: w2
+ * $vr16, $vr17, $vr18, $vr19: w1
+ */
+ for (d = 0; d < bytes; d += NSIZE*4) {
+ /* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
+ asm volatile("vld $vr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
+ asm volatile("vld $vr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
+ asm volatile("vld $vr2, %0" : : "m"(dptr[z0][d+2*NSIZE]));
+ asm volatile("vld $vr3, %0" : : "m"(dptr[z0][d+3*NSIZE]));
+ asm volatile("vori.b $vr4, $vr0, 0");
+ asm volatile("vori.b $vr5, $vr1, 0");
+ asm volatile("vori.b $vr6, $vr2, 0");
+ asm volatile("vori.b $vr7, $vr3, 0");
+ for (z = z0-1; z >= 0; z--) {
+ /* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
+ asm volatile("vld $vr8, %0" : : "m"(dptr[z][d+0*NSIZE]));
+ asm volatile("vld $vr9, %0" : : "m"(dptr[z][d+1*NSIZE]));
+ asm volatile("vld $vr10, %0" : : "m"(dptr[z][d+2*NSIZE]));
+ asm volatile("vld $vr11, %0" : : "m"(dptr[z][d+3*NSIZE]));
+ /* wp$$ ^= wd$$; */
+ asm volatile("vxor.v $vr0, $vr0, $vr8");
+ asm volatile("vxor.v $vr1, $vr1, $vr9");
+ asm volatile("vxor.v $vr2, $vr2, $vr10");
+ asm volatile("vxor.v $vr3, $vr3, $vr11");
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("vslti.b $vr12, $vr4, 0");
+ asm volatile("vslti.b $vr13, $vr5, 0");
+ asm volatile("vslti.b $vr14, $vr6, 0");
+ asm volatile("vslti.b $vr15, $vr7, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("vslli.b $vr16, $vr4, 1");
+ asm volatile("vslli.b $vr17, $vr5, 1");
+ asm volatile("vslli.b $vr18, $vr6, 1");
+ asm volatile("vslli.b $vr19, $vr7, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("vandi.b $vr12, $vr12, 0x1d");
+ asm volatile("vandi.b $vr13, $vr13, 0x1d");
+ asm volatile("vandi.b $vr14, $vr14, 0x1d");
+ asm volatile("vandi.b $vr15, $vr15, 0x1d");
+ /* w1$$ ^= w2$$; */
+ asm volatile("vxor.v $vr16, $vr16, $vr12");
+ asm volatile("vxor.v $vr17, $vr17, $vr13");
+ asm volatile("vxor.v $vr18, $vr18, $vr14");
+ asm volatile("vxor.v $vr19, $vr19, $vr15");
+ /* wq$$ = w1$$ ^ wd$$; */
+ asm volatile("vxor.v $vr4, $vr16, $vr8");
+ asm volatile("vxor.v $vr5, $vr17, $vr9");
+ asm volatile("vxor.v $vr6, $vr18, $vr10");
+ asm volatile("vxor.v $vr7, $vr19, $vr11");
+ }
+ /* *(unative_t *)&p[d+NSIZE*$$] = wp$$; */
+ asm volatile("vst $vr0, %0" : "=m"(p[d+NSIZE*0]));
+ asm volatile("vst $vr1, %0" : "=m"(p[d+NSIZE*1]));
+ asm volatile("vst $vr2, %0" : "=m"(p[d+NSIZE*2]));
+ asm volatile("vst $vr3, %0" : "=m"(p[d+NSIZE*3]));
+ /* *(unative_t *)&q[d+NSIZE*$$] = wq$$; */
+ asm volatile("vst $vr4, %0" : "=m"(q[d+NSIZE*0]));
+ asm volatile("vst $vr5, %0" : "=m"(q[d+NSIZE*1]));
+ asm volatile("vst $vr6, %0" : "=m"(q[d+NSIZE*2]));
+ asm volatile("vst $vr7, %0" : "=m"(q[d+NSIZE*3]));
+ }
+
+ kernel_fpu_end();
+}
+
+static void raid6_lsx_xor_syndrome(int disks, int start, int stop,
+ size_t bytes, void **ptrs)
+{
+ u8 **dptr = (u8 **)ptrs;
+ u8 *p, *q;
+ int d, z, z0;
+
+ z0 = stop; /* P/Q right side optimization */
+ p = dptr[disks-2]; /* XOR parity */
+ q = dptr[disks-1]; /* RS syndrome */
+
+ kernel_fpu_begin();
+
+ /*
+ * $vr0, $vr1, $vr2, $vr3: wp
+ * $vr4, $vr5, $vr6, $vr7: wq
+ * $vr8, $vr9, $vr10, $vr11: wd
+ * $vr12, $vr13, $vr14, $vr15: w2
+ * $vr16, $vr17, $vr18, $vr19: w1
+ */
+ for (d = 0; d < bytes; d += NSIZE*4) {
+ /* P/Q data pages */
+ /* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
+ asm volatile("vld $vr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
+ asm volatile("vld $vr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
+ asm volatile("vld $vr2, %0" : : "m"(dptr[z0][d+2*NSIZE]));
+ asm volatile("vld $vr3, %0" : : "m"(dptr[z0][d+3*NSIZE]));
+ asm volatile("vori.b $vr4, $vr0, 0");
+ asm volatile("vori.b $vr5, $vr1, 0");
+ asm volatile("vori.b $vr6, $vr2, 0");
+ asm volatile("vori.b $vr7, $vr3, 0");
+ for (z = z0-1; z >= start; z--) {
+ /* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
+ asm volatile("vld $vr8, %0" : : "m"(dptr[z][d+0*NSIZE]));
+ asm volatile("vld $vr9, %0" : : "m"(dptr[z][d+1*NSIZE]));
+ asm volatile("vld $vr10, %0" : : "m"(dptr[z][d+2*NSIZE]));
+ asm volatile("vld $vr11, %0" : : "m"(dptr[z][d+3*NSIZE]));
+ /* wp$$ ^= wd$$; */
+ asm volatile("vxor.v $vr0, $vr0, $vr8");
+ asm volatile("vxor.v $vr1, $vr1, $vr9");
+ asm volatile("vxor.v $vr2, $vr2, $vr10");
+ asm volatile("vxor.v $vr3, $vr3, $vr11");
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("vslti.b $vr12, $vr4, 0");
+ asm volatile("vslti.b $vr13, $vr5, 0");
+ asm volatile("vslti.b $vr14, $vr6, 0");
+ asm volatile("vslti.b $vr15, $vr7, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("vslli.b $vr16, $vr4, 1");
+ asm volatile("vslli.b $vr17, $vr5, 1");
+ asm volatile("vslli.b $vr18, $vr6, 1");
+ asm volatile("vslli.b $vr19, $vr7, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("vandi.b $vr12, $vr12, 0x1d");
+ asm volatile("vandi.b $vr13, $vr13, 0x1d");
+ asm volatile("vandi.b $vr14, $vr14, 0x1d");
+ asm volatile("vandi.b $vr15, $vr15, 0x1d");
+ /* w1$$ ^= w2$$; */
+ asm volatile("vxor.v $vr16, $vr16, $vr12");
+ asm volatile("vxor.v $vr17, $vr17, $vr13");
+ asm volatile("vxor.v $vr18, $vr18, $vr14");
+ asm volatile("vxor.v $vr19, $vr19, $vr15");
+ /* wq$$ = w1$$ ^ wd$$; */
+ asm volatile("vxor.v $vr4, $vr16, $vr8");
+ asm volatile("vxor.v $vr5, $vr17, $vr9");
+ asm volatile("vxor.v $vr6, $vr18, $vr10");
+ asm volatile("vxor.v $vr7, $vr19, $vr11");
+ }
+
+ /* P/Q left side optimization */
+ for (z = start-1; z >= 0; z--) {
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("vslti.b $vr12, $vr4, 0");
+ asm volatile("vslti.b $vr13, $vr5, 0");
+ asm volatile("vslti.b $vr14, $vr6, 0");
+ asm volatile("vslti.b $vr15, $vr7, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("vslli.b $vr16, $vr4, 1");
+ asm volatile("vslli.b $vr17, $vr5, 1");
+ asm volatile("vslli.b $vr18, $vr6, 1");
+ asm volatile("vslli.b $vr19, $vr7, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("vandi.b $vr12, $vr12, 0x1d");
+ asm volatile("vandi.b $vr13, $vr13, 0x1d");
+ asm volatile("vandi.b $vr14, $vr14, 0x1d");
+ asm volatile("vandi.b $vr15, $vr15, 0x1d");
+ /* wq$$ = w1$$ ^ w2$$; */
+ asm volatile("vxor.v $vr4, $vr16, $vr12");
+ asm volatile("vxor.v $vr5, $vr17, $vr13");
+ asm volatile("vxor.v $vr6, $vr18, $vr14");
+ asm volatile("vxor.v $vr7, $vr19, $vr15");
+ }
+ /*
+ * *(unative_t *)&p[d+NSIZE*$$] ^= wp$$;
+ * *(unative_t *)&q[d+NSIZE*$$] ^= wq$$;
+ */
+ asm volatile(
+ "vld $vr20, %0\n\t"
+ "vld $vr21, %1\n\t"
+ "vld $vr22, %2\n\t"
+ "vld $vr23, %3\n\t"
+ "vld $vr24, %4\n\t"
+ "vld $vr25, %5\n\t"
+ "vld $vr26, %6\n\t"
+ "vld $vr27, %7\n\t"
+ "vxor.v $vr20, $vr20, $vr0\n\t"
+ "vxor.v $vr21, $vr21, $vr1\n\t"
+ "vxor.v $vr22, $vr22, $vr2\n\t"
+ "vxor.v $vr23, $vr23, $vr3\n\t"
+ "vxor.v $vr24, $vr24, $vr4\n\t"
+ "vxor.v $vr25, $vr25, $vr5\n\t"
+ "vxor.v $vr26, $vr26, $vr6\n\t"
+ "vxor.v $vr27, $vr27, $vr7\n\t"
+ "vst $vr20, %0\n\t"
+ "vst $vr21, %1\n\t"
+ "vst $vr22, %2\n\t"
+ "vst $vr23, %3\n\t"
+ "vst $vr24, %4\n\t"
+ "vst $vr25, %5\n\t"
+ "vst $vr26, %6\n\t"
+ "vst $vr27, %7\n\t"
+ : "+m"(p[d+NSIZE*0]), "+m"(p[d+NSIZE*1]),
+ "+m"(p[d+NSIZE*2]), "+m"(p[d+NSIZE*3]),
+ "+m"(q[d+NSIZE*0]), "+m"(q[d+NSIZE*1]),
+ "+m"(q[d+NSIZE*2]), "+m"(q[d+NSIZE*3])
+ );
+ }
+
+ kernel_fpu_end();
+}
+
+const struct raid6_calls raid6_lsx = {
+ raid6_lsx_gen_syndrome,
+ raid6_lsx_xor_syndrome,
+ raid6_has_lsx,
+ "lsx",
+ .priority = 0 /* see the comment near the top of the file for reason */
+};
+
+#undef NSIZE
+#endif /* CONFIG_CPU_HAS_LSX */
+
+#ifdef CONFIG_CPU_HAS_LASX
+#define NSIZE 32
+
+static int raid6_has_lasx(void)
+{
+ return cpu_has_lasx;
+}
+
+static void raid6_lasx_gen_syndrome(int disks, size_t bytes, void **ptrs)
+{
+ u8 **dptr = (u8 **)ptrs;
+ u8 *p, *q;
+ int d, z, z0;
+
+ z0 = disks - 3; /* Highest data disk */
+ p = dptr[z0+1]; /* XOR parity */
+ q = dptr[z0+2]; /* RS syndrome */
+
+ kernel_fpu_begin();
+
+ /*
+ * $xr0, $xr1: wp
+ * $xr2, $xr3: wq
+ * $xr4, $xr5: wd
+ * $xr6, $xr7: w2
+ * $xr8, $xr9: w1
+ */
+ for (d = 0; d < bytes; d += NSIZE*2) {
+ /* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
+ asm volatile("xvld $xr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
+ asm volatile("xvld $xr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
+ asm volatile("xvori.b $xr2, $xr0, 0");
+ asm volatile("xvori.b $xr3, $xr1, 0");
+ for (z = z0-1; z >= 0; z--) {
+ /* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
+ asm volatile("xvld $xr4, %0" : : "m"(dptr[z][d+0*NSIZE]));
+ asm volatile("xvld $xr5, %0" : : "m"(dptr[z][d+1*NSIZE]));
+ /* wp$$ ^= wd$$; */
+ asm volatile("xvxor.v $xr0, $xr0, $xr4");
+ asm volatile("xvxor.v $xr1, $xr1, $xr5");
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("xvslti.b $xr6, $xr2, 0");
+ asm volatile("xvslti.b $xr7, $xr3, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("xvslli.b $xr8, $xr2, 1");
+ asm volatile("xvslli.b $xr9, $xr3, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("xvandi.b $xr6, $xr6, 0x1d");
+ asm volatile("xvandi.b $xr7, $xr7, 0x1d");
+ /* w1$$ ^= w2$$; */
+ asm volatile("xvxor.v $xr8, $xr8, $xr6");
+ asm volatile("xvxor.v $xr9, $xr9, $xr7");
+ /* wq$$ = w1$$ ^ wd$$; */
+ asm volatile("xvxor.v $xr2, $xr8, $xr4");
+ asm volatile("xvxor.v $xr3, $xr9, $xr5");
+ }
+ /* *(unative_t *)&p[d+NSIZE*$$] = wp$$; */
+ asm volatile("xvst $xr0, %0" : "=m"(p[d+NSIZE*0]));
+ asm volatile("xvst $xr1, %0" : "=m"(p[d+NSIZE*1]));
+ /* *(unative_t *)&q[d+NSIZE*$$] = wq$$; */
+ asm volatile("xvst $xr2, %0" : "=m"(q[d+NSIZE*0]));
+ asm volatile("xvst $xr3, %0" : "=m"(q[d+NSIZE*1]));
+ }
+
+ kernel_fpu_end();
+}
+
+static void raid6_lasx_xor_syndrome(int disks, int start, int stop,
+ size_t bytes, void **ptrs)
+{
+ u8 **dptr = (u8 **)ptrs;
+ u8 *p, *q;
+ int d, z, z0;
+
+ z0 = stop; /* P/Q right side optimization */
+ p = dptr[disks-2]; /* XOR parity */
+ q = dptr[disks-1]; /* RS syndrome */
+
+ kernel_fpu_begin();
+
+ /*
+ * $xr0, $xr1: wp
+ * $xr2, $xr3: wq
+ * $xr4, $xr5: wd
+ * $xr6, $xr7: w2
+ * $xr8, $xr9: w1
+ */
+ for (d = 0; d < bytes; d += NSIZE*2) {
+ /* P/Q data pages */
+ /* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
+ asm volatile("xvld $xr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
+ asm volatile("xvld $xr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
+ asm volatile("xvori.b $xr2, $xr0, 0");
+ asm volatile("xvori.b $xr3, $xr1, 0");
+ for (z = z0-1; z >= start; z--) {
+ /* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
+ asm volatile("xvld $xr4, %0" : : "m"(dptr[z][d+0*NSIZE]));
+ asm volatile("xvld $xr5, %0" : : "m"(dptr[z][d+1*NSIZE]));
+ /* wp$$ ^= wd$$; */
+ asm volatile("xvxor.v $xr0, $xr0, $xr4");
+ asm volatile("xvxor.v $xr1, $xr1, $xr5");
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("xvslti.b $xr6, $xr2, 0");
+ asm volatile("xvslti.b $xr7, $xr3, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("xvslli.b $xr8, $xr2, 1");
+ asm volatile("xvslli.b $xr9, $xr3, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("xvandi.b $xr6, $xr6, 0x1d");
+ asm volatile("xvandi.b $xr7, $xr7, 0x1d");
+ /* w1$$ ^= w2$$; */
+ asm volatile("xvxor.v $xr8, $xr8, $xr6");
+ asm volatile("xvxor.v $xr9, $xr9, $xr7");
+ /* wq$$ = w1$$ ^ wd$$; */
+ asm volatile("xvxor.v $xr2, $xr8, $xr4");
+ asm volatile("xvxor.v $xr3, $xr9, $xr5");
+ }
+
+ /* P/Q left side optimization */
+ for (z = start-1; z >= 0; z--) {
+ /* w2$$ = MASK(wq$$); */
+ asm volatile("xvslti.b $xr6, $xr2, 0");
+ asm volatile("xvslti.b $xr7, $xr3, 0");
+ /* w1$$ = SHLBYTE(wq$$); */
+ asm volatile("xvslli.b $xr8, $xr2, 1");
+ asm volatile("xvslli.b $xr9, $xr3, 1");
+ /* w2$$ &= NBYTES(0x1d); */
+ asm volatile("xvandi.b $xr6, $xr6, 0x1d");
+ asm volatile("xvandi.b $xr7, $xr7, 0x1d");
+ /* wq$$ = w1$$ ^ w2$$; */
+ asm volatile("xvxor.v $xr2, $xr8, $xr6");
+ asm volatile("xvxor.v $xr3, $xr9, $xr7");
+ }
+ /*
+ * *(unative_t *)&p[d+NSIZE*$$] ^= wp$$;
+ * *(unative_t *)&q[d+NSIZE*$$] ^= wq$$;
+ */
+ asm volatile(
+ "xvld $xr10, %0\n\t"
+ "xvld $xr11, %1\n\t"
+ "xvld $xr12, %2\n\t"
+ "xvld $xr13, %3\n\t"
+ "xvxor.v $xr10, $xr10, $xr0\n\t"
+ "xvxor.v $xr11, $xr11, $xr1\n\t"
+ "xvxor.v $xr12, $xr12, $xr2\n\t"
+ "xvxor.v $xr13, $xr13, $xr3\n\t"
+ "xvst $xr10, %0\n\t"
+ "xvst $xr11, %1\n\t"
+ "xvst $xr12, %2\n\t"
+ "xvst $xr13, %3\n\t"
+ : "+m"(p[d+NSIZE*0]), "+m"(p[d+NSIZE*1]),
+ "+m"(q[d+NSIZE*0]), "+m"(q[d+NSIZE*1])
+ );
+ }
+
+ kernel_fpu_end();
+}
+
+const struct raid6_calls raid6_lasx = {
+ raid6_lasx_gen_syndrome,
+ raid6_lasx_xor_syndrome,
+ raid6_has_lasx,
+ "lasx",
+ .priority = 0 /* see the comment near the top of the file for reason */
+};
+#undef NSIZE
+#endif /* CONFIG_CPU_HAS_LASX */
diff --git a/lib/raid6/recov_loongarch_simd.c b/lib/raid6/recov_loongarch_simd.c
new file mode 100644
index 000000000000..94aeac85e6f7
--- /dev/null
+++ b/lib/raid6/recov_loongarch_simd.c
@@ -0,0 +1,513 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * RAID6 recovery algorithms in LoongArch SIMD (LSX & LASX)
+ *
+ * Copyright (C) 2023 WANG Xuerui <git@xen0n.name>
+ *
+ * Originally based on recov_avx2.c and recov_ssse3.c:
+ *
+ * Copyright (C) 2012 Intel Corporation
+ * Author: Jim Kukunas <james.t.kukunas@linux.intel.com>
+ */
+
+#include <linux/raid/pq.h>
+#include "loongarch.h"
+
+/*
+ * Unlike with the syndrome calculation algorithms, there's no boot-time
+ * selection of recovery algorithms by benchmarking, so we have to specify
+ * the priorities and hope the future cores will all have decent vector
+ * support (i.e. no LASX slower than LSX, or even scalar code).
+ */
+
+#ifdef CONFIG_CPU_HAS_LSX
+static int raid6_has_lsx(void)
+{
+ return cpu_has_lsx;
+}
+
+static void raid6_2data_recov_lsx(int disks, size_t bytes, int faila,
+ int failb, void **ptrs)
+{
+ u8 *p, *q, *dp, *dq;
+ const u8 *pbmul; /* P multiplier table for B data */
+ const u8 *qmul; /* Q multiplier table (for both) */
+
+ p = (u8 *)ptrs[disks - 2];
+ q = (u8 *)ptrs[disks - 1];
+
+ /*
+ * Compute syndrome with zero for the missing data pages
+ * Use the dead data pages as temporary storage for
+ * delta p and delta q
+ */
+ dp = (u8 *)ptrs[faila];
+ ptrs[faila] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 2] = dp;
+ dq = (u8 *)ptrs[failb];
+ ptrs[failb] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 1] = dq;
+
+ raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+ /* Restore pointer table */
+ ptrs[faila] = dp;
+ ptrs[failb] = dq;
+ ptrs[disks - 2] = p;
+ ptrs[disks - 1] = q;
+
+ /* Now, pick the proper data tables */
+ pbmul = raid6_vgfmul[raid6_gfexi[failb - faila]];
+ qmul = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila] ^ raid6_gfexp[failb]]];
+
+ kernel_fpu_begin();
+
+ /*
+ * vr20, vr21: qmul
+ * vr22, vr23: pbmul
+ */
+ asm volatile("vld $vr20, %0" : : "m" (qmul[0]));
+ asm volatile("vld $vr21, %0" : : "m" (qmul[16]));
+ asm volatile("vld $vr22, %0" : : "m" (pbmul[0]));
+ asm volatile("vld $vr23, %0" : : "m" (pbmul[16]));
+
+ while (bytes) {
+ /* vr4 - vr7: Q */
+ asm volatile("vld $vr4, %0" : : "m" (q[0]));
+ asm volatile("vld $vr5, %0" : : "m" (q[16]));
+ asm volatile("vld $vr6, %0" : : "m" (q[32]));
+ asm volatile("vld $vr7, %0" : : "m" (q[48]));
+ /* vr4 - vr7: Q + Qxy */
+ asm volatile("vld $vr8, %0" : : "m" (dq[0]));
+ asm volatile("vld $vr9, %0" : : "m" (dq[16]));
+ asm volatile("vld $vr10, %0" : : "m" (dq[32]));
+ asm volatile("vld $vr11, %0" : : "m" (dq[48]));
+ asm volatile("vxor.v $vr4, $vr4, $vr8");
+ asm volatile("vxor.v $vr5, $vr5, $vr9");
+ asm volatile("vxor.v $vr6, $vr6, $vr10");
+ asm volatile("vxor.v $vr7, $vr7, $vr11");
+ /* vr0 - vr3: P */
+ asm volatile("vld $vr0, %0" : : "m" (p[0]));
+ asm volatile("vld $vr1, %0" : : "m" (p[16]));
+ asm volatile("vld $vr2, %0" : : "m" (p[32]));
+ asm volatile("vld $vr3, %0" : : "m" (p[48]));
+ /* vr0 - vr3: P + Pxy */
+ asm volatile("vld $vr8, %0" : : "m" (dp[0]));
+ asm volatile("vld $vr9, %0" : : "m" (dp[16]));
+ asm volatile("vld $vr10, %0" : : "m" (dp[32]));
+ asm volatile("vld $vr11, %0" : : "m" (dp[48]));
+ asm volatile("vxor.v $vr0, $vr0, $vr8");
+ asm volatile("vxor.v $vr1, $vr1, $vr9");
+ asm volatile("vxor.v $vr2, $vr2, $vr10");
+ asm volatile("vxor.v $vr3, $vr3, $vr11");
+
+ /* vr8 - vr11: higher 4 bits of each byte of (Q + Qxy) */
+ asm volatile("vsrli.b $vr8, $vr4, 4");
+ asm volatile("vsrli.b $vr9, $vr5, 4");
+ asm volatile("vsrli.b $vr10, $vr6, 4");
+ asm volatile("vsrli.b $vr11, $vr7, 4");
+ /* vr4 - vr7: lower 4 bits of each byte of (Q + Qxy) */
+ asm volatile("vandi.b $vr4, $vr4, 0x0f");
+ asm volatile("vandi.b $vr5, $vr5, 0x0f");
+ asm volatile("vandi.b $vr6, $vr6, 0x0f");
+ asm volatile("vandi.b $vr7, $vr7, 0x0f");
+ /* lookup from qmul[0] */
+ asm volatile("vshuf.b $vr4, $vr20, $vr20, $vr4");
+ asm volatile("vshuf.b $vr5, $vr20, $vr20, $vr5");
+ asm volatile("vshuf.b $vr6, $vr20, $vr20, $vr6");
+ asm volatile("vshuf.b $vr7, $vr20, $vr20, $vr7");
+ /* lookup from qmul[16] */
+ asm volatile("vshuf.b $vr8, $vr21, $vr21, $vr8");
+ asm volatile("vshuf.b $vr9, $vr21, $vr21, $vr9");
+ asm volatile("vshuf.b $vr10, $vr21, $vr21, $vr10");
+ asm volatile("vshuf.b $vr11, $vr21, $vr21, $vr11");
+ /* vr16 - vr19: B(Q + Qxy) */
+ asm volatile("vxor.v $vr16, $vr8, $vr4");
+ asm volatile("vxor.v $vr17, $vr9, $vr5");
+ asm volatile("vxor.v $vr18, $vr10, $vr6");
+ asm volatile("vxor.v $vr19, $vr11, $vr7");
+
+ /* vr4 - vr7: higher 4 bits of each byte of (P + Pxy) */
+ asm volatile("vsrli.b $vr4, $vr0, 4");
+ asm volatile("vsrli.b $vr5, $vr1, 4");
+ asm volatile("vsrli.b $vr6, $vr2, 4");
+ asm volatile("vsrli.b $vr7, $vr3, 4");
+ /* vr12 - vr15: lower 4 bits of each byte of (P + Pxy) */
+ asm volatile("vandi.b $vr12, $vr0, 0x0f");
+ asm volatile("vandi.b $vr13, $vr1, 0x0f");
+ asm volatile("vandi.b $vr14, $vr2, 0x0f");
+ asm volatile("vandi.b $vr15, $vr3, 0x0f");
+ /* lookup from pbmul[0] */
+ asm volatile("vshuf.b $vr12, $vr22, $vr22, $vr12");
+ asm volatile("vshuf.b $vr13, $vr22, $vr22, $vr13");
+ asm volatile("vshuf.b $vr14, $vr22, $vr22, $vr14");
+ asm volatile("vshuf.b $vr15, $vr22, $vr22, $vr15");
+ /* lookup from pbmul[16] */
+ asm volatile("vshuf.b $vr4, $vr23, $vr23, $vr4");
+ asm volatile("vshuf.b $vr5, $vr23, $vr23, $vr5");
+ asm volatile("vshuf.b $vr6, $vr23, $vr23, $vr6");
+ asm volatile("vshuf.b $vr7, $vr23, $vr23, $vr7");
+ /* vr4 - vr7: A(P + Pxy) */
+ asm volatile("vxor.v $vr4, $vr4, $vr12");
+ asm volatile("vxor.v $vr5, $vr5, $vr13");
+ asm volatile("vxor.v $vr6, $vr6, $vr14");
+ asm volatile("vxor.v $vr7, $vr7, $vr15");
+
+ /* vr4 - vr7: A(P + Pxy) + B(Q + Qxy) = Dx */
+ asm volatile("vxor.v $vr4, $vr4, $vr16");
+ asm volatile("vxor.v $vr5, $vr5, $vr17");
+ asm volatile("vxor.v $vr6, $vr6, $vr18");
+ asm volatile("vxor.v $vr7, $vr7, $vr19");
+ asm volatile("vst $vr4, %0" : "=m" (dq[0]));
+ asm volatile("vst $vr5, %0" : "=m" (dq[16]));
+ asm volatile("vst $vr6, %0" : "=m" (dq[32]));
+ asm volatile("vst $vr7, %0" : "=m" (dq[48]));
+
+ /* vr0 - vr3: P + Pxy + Dx = Dy */
+ asm volatile("vxor.v $vr0, $vr0, $vr4");
+ asm volatile("vxor.v $vr1, $vr1, $vr5");
+ asm volatile("vxor.v $vr2, $vr2, $vr6");
+ asm volatile("vxor.v $vr3, $vr3, $vr7");
+ asm volatile("vst $vr0, %0" : "=m" (dp[0]));
+ asm volatile("vst $vr1, %0" : "=m" (dp[16]));
+ asm volatile("vst $vr2, %0" : "=m" (dp[32]));
+ asm volatile("vst $vr3, %0" : "=m" (dp[48]));
+
+ bytes -= 64;
+ p += 64;
+ q += 64;
+ dp += 64;
+ dq += 64;
+ }
+
+ kernel_fpu_end();
+}
+
+static void raid6_datap_recov_lsx(int disks, size_t bytes, int faila,
+ void **ptrs)
+{
+ u8 *p, *q, *dq;
+ const u8 *qmul; /* Q multiplier table */
+
+ p = (u8 *)ptrs[disks - 2];
+ q = (u8 *)ptrs[disks - 1];
+
+ /*
+ * Compute syndrome with zero for the missing data page
+ * Use the dead data page as temporary storage for delta q
+ */
+ dq = (u8 *)ptrs[faila];
+ ptrs[faila] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 1] = dq;
+
+ raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+ /* Restore pointer table */
+ ptrs[faila] = dq;
+ ptrs[disks - 1] = q;
+
+ /* Now, pick the proper data tables */
+ qmul = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila]]];
+
+ kernel_fpu_begin();
+
+ /* vr22, vr23: qmul */
+ asm volatile("vld $vr22, %0" : : "m" (qmul[0]));
+ asm volatile("vld $vr23, %0" : : "m" (qmul[16]));
+
+ while (bytes) {
+ /* vr0 - vr3: P + Dx */
+ asm volatile("vld $vr0, %0" : : "m" (p[0]));
+ asm volatile("vld $vr1, %0" : : "m" (p[16]));
+ asm volatile("vld $vr2, %0" : : "m" (p[32]));
+ asm volatile("vld $vr3, %0" : : "m" (p[48]));
+ /* vr4 - vr7: Qx */
+ asm volatile("vld $vr4, %0" : : "m" (dq[0]));
+ asm volatile("vld $vr5, %0" : : "m" (dq[16]));
+ asm volatile("vld $vr6, %0" : : "m" (dq[32]));
+ asm volatile("vld $vr7, %0" : : "m" (dq[48]));
+ /* vr4 - vr7: Q + Qx */
+ asm volatile("vld $vr8, %0" : : "m" (q[0]));
+ asm volatile("vld $vr9, %0" : : "m" (q[16]));
+ asm volatile("vld $vr10, %0" : : "m" (q[32]));
+ asm volatile("vld $vr11, %0" : : "m" (q[48]));
+ asm volatile("vxor.v $vr4, $vr4, $vr8");
+ asm volatile("vxor.v $vr5, $vr5, $vr9");
+ asm volatile("vxor.v $vr6, $vr6, $vr10");
+ asm volatile("vxor.v $vr7, $vr7, $vr11");
+
+ /* vr8 - vr11: higher 4 bits of each byte of (Q + Qx) */
+ asm volatile("vsrli.b $vr8, $vr4, 4");
+ asm volatile("vsrli.b $vr9, $vr5, 4");
+ asm volatile("vsrli.b $vr10, $vr6, 4");
+ asm volatile("vsrli.b $vr11, $vr7, 4");
+ /* vr4 - vr7: lower 4 bits of each byte of (Q + Qx) */
+ asm volatile("vandi.b $vr4, $vr4, 0x0f");
+ asm volatile("vandi.b $vr5, $vr5, 0x0f");
+ asm volatile("vandi.b $vr6, $vr6, 0x0f");
+ asm volatile("vandi.b $vr7, $vr7, 0x0f");
+ /* lookup from qmul[0] */
+ asm volatile("vshuf.b $vr4, $vr22, $vr22, $vr4");
+ asm volatile("vshuf.b $vr5, $vr22, $vr22, $vr5");
+ asm volatile("vshuf.b $vr6, $vr22, $vr22, $vr6");
+ asm volatile("vshuf.b $vr7, $vr22, $vr22, $vr7");
+ /* lookup from qmul[16] */
+ asm volatile("vshuf.b $vr8, $vr23, $vr23, $vr8");
+ asm volatile("vshuf.b $vr9, $vr23, $vr23, $vr9");
+ asm volatile("vshuf.b $vr10, $vr23, $vr23, $vr10");
+ asm volatile("vshuf.b $vr11, $vr23, $vr23, $vr11");
+ /* vr4 - vr7: qmul(Q + Qx) = Dx */
+ asm volatile("vxor.v $vr4, $vr4, $vr8");
+ asm volatile("vxor.v $vr5, $vr5, $vr9");
+ asm volatile("vxor.v $vr6, $vr6, $vr10");
+ asm volatile("vxor.v $vr7, $vr7, $vr11");
+ asm volatile("vst $vr4, %0" : "=m" (dq[0]));
+ asm volatile("vst $vr5, %0" : "=m" (dq[16]));
+ asm volatile("vst $vr6, %0" : "=m" (dq[32]));
+ asm volatile("vst $vr7, %0" : "=m" (dq[48]));
+
+ /* vr0 - vr3: P + Dx + Dx = P */
+ asm volatile("vxor.v $vr0, $vr0, $vr4");
+ asm volatile("vxor.v $vr1, $vr1, $vr5");
+ asm volatile("vxor.v $vr2, $vr2, $vr6");
+ asm volatile("vxor.v $vr3, $vr3, $vr7");
+ asm volatile("vst $vr0, %0" : "=m" (p[0]));
+ asm volatile("vst $vr1, %0" : "=m" (p[16]));
+ asm volatile("vst $vr2, %0" : "=m" (p[32]));
+ asm volatile("vst $vr3, %0" : "=m" (p[48]));
+
+ bytes -= 64;
+ p += 64;
+ q += 64;
+ dq += 64;
+ }
+
+ kernel_fpu_end();
+}
+
+const struct raid6_recov_calls raid6_recov_lsx = {
+ .data2 = raid6_2data_recov_lsx,
+ .datap = raid6_datap_recov_lsx,
+ .valid = raid6_has_lsx,
+ .name = "lsx",
+ .priority = 1,
+};
+#endif /* CONFIG_CPU_HAS_LSX */
+
+#ifdef CONFIG_CPU_HAS_LASX
+static int raid6_has_lasx(void)
+{
+ return cpu_has_lasx;
+}
+
+static void raid6_2data_recov_lasx(int disks, size_t bytes, int faila,
+ int failb, void **ptrs)
+{
+ u8 *p, *q, *dp, *dq;
+ const u8 *pbmul; /* P multiplier table for B data */
+ const u8 *qmul; /* Q multiplier table (for both) */
+
+ p = (u8 *)ptrs[disks - 2];
+ q = (u8 *)ptrs[disks - 1];
+
+ /*
+ * Compute syndrome with zero for the missing data pages
+ * Use the dead data pages as temporary storage for
+ * delta p and delta q
+ */
+ dp = (u8 *)ptrs[faila];
+ ptrs[faila] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 2] = dp;
+ dq = (u8 *)ptrs[failb];
+ ptrs[failb] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 1] = dq;
+
+ raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+ /* Restore pointer table */
+ ptrs[faila] = dp;
+ ptrs[failb] = dq;
+ ptrs[disks - 2] = p;
+ ptrs[disks - 1] = q;
+
+ /* Now, pick the proper data tables */
+ pbmul = raid6_vgfmul[raid6_gfexi[failb - faila]];
+ qmul = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila] ^ raid6_gfexp[failb]]];
+
+ kernel_fpu_begin();
+
+ /*
+ * xr20, xr21: qmul
+ * xr22, xr23: pbmul
+ */
+ asm volatile("vld $vr20, %0" : : "m" (qmul[0]));
+ asm volatile("vld $vr21, %0" : : "m" (qmul[16]));
+ asm volatile("vld $vr22, %0" : : "m" (pbmul[0]));
+ asm volatile("vld $vr23, %0" : : "m" (pbmul[16]));
+ asm volatile("xvreplve0.q $xr20, $xr20");
+ asm volatile("xvreplve0.q $xr21, $xr21");
+ asm volatile("xvreplve0.q $xr22, $xr22");
+ asm volatile("xvreplve0.q $xr23, $xr23");
+
+ while (bytes) {
+ /* xr0, xr1: Q */
+ asm volatile("xvld $xr0, %0" : : "m" (q[0]));
+ asm volatile("xvld $xr1, %0" : : "m" (q[32]));
+ /* xr0, xr1: Q + Qxy */
+ asm volatile("xvld $xr4, %0" : : "m" (dq[0]));
+ asm volatile("xvld $xr5, %0" : : "m" (dq[32]));
+ asm volatile("xvxor.v $xr0, $xr0, $xr4");
+ asm volatile("xvxor.v $xr1, $xr1, $xr5");
+ /* xr2, xr3: P */
+ asm volatile("xvld $xr2, %0" : : "m" (p[0]));
+ asm volatile("xvld $xr3, %0" : : "m" (p[32]));
+ /* xr2, xr3: P + Pxy */
+ asm volatile("xvld $xr4, %0" : : "m" (dp[0]));
+ asm volatile("xvld $xr5, %0" : : "m" (dp[32]));
+ asm volatile("xvxor.v $xr2, $xr2, $xr4");
+ asm volatile("xvxor.v $xr3, $xr3, $xr5");
+
+ /* xr4, xr5: higher 4 bits of each byte of (Q + Qxy) */
+ asm volatile("xvsrli.b $xr4, $xr0, 4");
+ asm volatile("xvsrli.b $xr5, $xr1, 4");
+ /* xr0, xr1: lower 4 bits of each byte of (Q + Qxy) */
+ asm volatile("xvandi.b $xr0, $xr0, 0x0f");
+ asm volatile("xvandi.b $xr1, $xr1, 0x0f");
+ /* lookup from qmul[0] */
+ asm volatile("xvshuf.b $xr0, $xr20, $xr20, $xr0");
+ asm volatile("xvshuf.b $xr1, $xr20, $xr20, $xr1");
+ /* lookup from qmul[16] */
+ asm volatile("xvshuf.b $xr4, $xr21, $xr21, $xr4");
+ asm volatile("xvshuf.b $xr5, $xr21, $xr21, $xr5");
+ /* xr6, xr7: B(Q + Qxy) */
+ asm volatile("xvxor.v $xr6, $xr4, $xr0");
+ asm volatile("xvxor.v $xr7, $xr5, $xr1");
+
+ /* xr4, xr5: higher 4 bits of each byte of (P + Pxy) */
+ asm volatile("xvsrli.b $xr4, $xr2, 4");
+ asm volatile("xvsrli.b $xr5, $xr3, 4");
+ /* xr0, xr1: lower 4 bits of each byte of (P + Pxy) */
+ asm volatile("xvandi.b $xr0, $xr2, 0x0f");
+ asm volatile("xvandi.b $xr1, $xr3, 0x0f");
+ /* lookup from pbmul[0] */
+ asm volatile("xvshuf.b $xr0, $xr22, $xr22, $xr0");
+ asm volatile("xvshuf.b $xr1, $xr22, $xr22, $xr1");
+ /* lookup from pbmul[16] */
+ asm volatile("xvshuf.b $xr4, $xr23, $xr23, $xr4");
+ asm volatile("xvshuf.b $xr5, $xr23, $xr23, $xr5");
+ /* xr0, xr1: A(P + Pxy) */
+ asm volatile("xvxor.v $xr0, $xr0, $xr4");
+ asm volatile("xvxor.v $xr1, $xr1, $xr5");
+
+ /* xr0, xr1: A(P + Pxy) + B(Q + Qxy) = Dx */
+ asm volatile("xvxor.v $xr0, $xr0, $xr6");
+ asm volatile("xvxor.v $xr1, $xr1, $xr7");
+
+ /* xr2, xr3: P + Pxy + Dx = Dy */
+ asm volatile("xvxor.v $xr2, $xr2, $xr0");
+ asm volatile("xvxor.v $xr3, $xr3, $xr1");
+
+ asm volatile("xvst $xr0, %0" : "=m" (dq[0]));
+ asm volatile("xvst $xr1, %0" : "=m" (dq[32]));
+ asm volatile("xvst $xr2, %0" : "=m" (dp[0]));
+ asm volatile("xvst $xr3, %0" : "=m" (dp[32]));
+
+ bytes -= 64;
+ p += 64;
+ q += 64;
+ dp += 64;
+ dq += 64;
+ }
+
+ kernel_fpu_end();
+}
+
+static void raid6_datap_recov_lasx(int disks, size_t bytes, int faila,
+ void **ptrs)
+{
+ u8 *p, *q, *dq;
+ const u8 *qmul; /* Q multiplier table */
+
+ p = (u8 *)ptrs[disks - 2];
+ q = (u8 *)ptrs[disks - 1];
+
+ /*
+ * Compute syndrome with zero for the missing data page
+ * Use the dead data page as temporary storage for delta q
+ */
+ dq = (u8 *)ptrs[faila];
+ ptrs[faila] = (void *)raid6_empty_zero_page;
+ ptrs[disks - 1] = dq;
+
+ raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+ /* Restore pointer table */
+ ptrs[faila] = dq;
+ ptrs[disks - 1] = q;
+
+ /* Now, pick the proper data tables */
+ qmul = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila]]];
+
+ kernel_fpu_begin();
+
+ /* xr22, xr23: qmul */
+ asm volatile("vld $vr22, %0" : : "m" (qmul[0]));
+ asm volatile("xvreplve0.q $xr22, $xr22");
+ asm volatile("vld $vr23, %0" : : "m" (qmul[16]));
+ asm volatile("xvreplve0.q $xr23, $xr23");
+
+ while (bytes) {
+ /* xr0, xr1: P + Dx */
+ asm volatile("xvld $xr0, %0" : : "m" (p[0]));
+ asm volatile("xvld $xr1, %0" : : "m" (p[32]));
+ /* xr2, xr3: Qx */
+ asm volatile("xvld $xr2, %0" : : "m" (dq[0]));
+ asm volatile("xvld $xr3, %0" : : "m" (dq[32]));
+ /* xr2, xr3: Q + Qx */
+ asm volatile("xvld $xr4, %0" : : "m" (q[0]));
+ asm volatile("xvld $xr5, %0" : : "m" (q[32]));
+ asm volatile("xvxor.v $xr2, $xr2, $xr4");
+ asm volatile("xvxor.v $xr3, $xr3, $xr5");
+
+ /* xr4, xr5: higher 4 bits of each byte of (Q + Qx) */
+ asm volatile("xvsrli.b $xr4, $xr2, 4");
+ asm volatile("xvsrli.b $xr5, $xr3, 4");
+ /* xr2, xr3: lower 4 bits of each byte of (Q + Qx) */
+ asm volatile("xvandi.b $xr2, $xr2, 0x0f");
+ asm volatile("xvandi.b $xr3, $xr3, 0x0f");
+ /* lookup from qmul[0] */
+ asm volatile("xvshuf.b $xr2, $xr22, $xr22, $xr2");
+ asm volatile("xvshuf.b $xr3, $xr22, $xr22, $xr3");
+ /* lookup from qmul[16] */
+ asm volatile("xvshuf.b $xr4, $xr23, $xr23, $xr4");
+ asm volatile("xvshuf.b $xr5, $xr23, $xr23, $xr5");
+ /* xr2, xr3: qmul(Q + Qx) = Dx */
+ asm volatile("xvxor.v $xr2, $xr2, $xr4");
+ asm volatile("xvxor.v $xr3, $xr3, $xr5");
+
+ /* xr0, xr1: P + Dx + Dx = P */
+ asm volatile("xvxor.v $xr0, $xr0, $xr2");
+ asm volatile("xvxor.v $xr1, $xr1, $xr3");
+
+ asm volatile("xvst $xr2, %0" : "=m" (dq[0]));
+ asm volatile("xvst $xr3, %0" : "=m" (dq[32]));
+ asm volatile("xvst $xr0, %0" : "=m" (p[0]));
+ asm volatile("xvst $xr1, %0" : "=m" (p[32]));
+
+ bytes -= 64;
+ p += 64;
+ q += 64;
+ dq += 64;
+ }
+
+ kernel_fpu_end();
+}
+
+const struct raid6_recov_calls raid6_recov_lasx = {
+ .data2 = raid6_2data_recov_lasx,
+ .datap = raid6_datap_recov_lasx,
+ .valid = raid6_has_lasx,
+ .name = "lasx",
+ .priority = 2,
+};
+#endif /* CONFIG_CPU_HAS_LASX */
diff --git a/lib/raid6/test/Makefile b/lib/raid6/test/Makefile
index 1f693ea3b980..2abe0076a636 100644
--- a/lib/raid6/test/Makefile
+++ b/lib/raid6/test/Makefile
@@ -41,6 +41,16 @@ ifeq ($(findstring ppc,$(ARCH)),ppc)
gcc -c -x c - >/dev/null && rm ./-.o && echo yes)
endif
+ifeq ($(ARCH),loongarch64)
+ CFLAGS += -I../../../arch/loongarch/include -DCONFIG_LOONGARCH=1
+ CFLAGS += $(shell echo 'vld $$vr0, $$zero, 0' | \
+ gcc -c -x assembler - >/dev/null 2>&1 && \
+ rm ./-.o && echo -DCONFIG_CPU_HAS_LSX=1)
+ CFLAGS += $(shell echo 'xvld $$xr0, $$zero, 0' | \
+ gcc -c -x assembler - >/dev/null 2>&1 && \
+ rm ./-.o && echo -DCONFIG_CPU_HAS_LASX=1)
+endif
+
ifeq ($(IS_X86),yes)
OBJS += mmx.o sse1.o sse2.o avx2.o recov_ssse3.o recov_avx2.o avx512.o recov_avx512.o
CFLAGS += -DCONFIG_X86
@@ -54,6 +64,8 @@ else ifeq ($(HAS_ALTIVEC),yes)
CFLAGS += -DCONFIG_ALTIVEC
OBJS += altivec1.o altivec2.o altivec4.o altivec8.o \
vpermxor1.o vpermxor2.o vpermxor4.o vpermxor8.o
+else ifeq ($(ARCH),loongarch64)
+ OBJS += loongarch_simd.o recov_loongarch_simd.o
endif
.c.o:
diff --git a/lib/string_helpers.c b/lib/string_helpers.c
index d3b1dd718daf..9982344cca34 100644
--- a/lib/string_helpers.c
+++ b/lib/string_helpers.c
@@ -719,6 +719,21 @@ char *kstrdup_quotable_file(struct file *file, gfp_t gfp)
}
EXPORT_SYMBOL_GPL(kstrdup_quotable_file);
+/*
+ * Returns duplicate string in which the @old characters are replaced by @new.
+ */
+char *kstrdup_and_replace(const char *src, char old, char new, gfp_t gfp)
+{
+ char *dst;
+
+ dst = kstrdup(src, gfp);
+ if (!dst)
+ return NULL;
+
+ return strreplace(dst, old, new);
+}
+EXPORT_SYMBOL_GPL(kstrdup_and_replace);
+
/**
* kasprintf_strarray - allocate and fill array of sequential strings
* @gfp: flags for the slab allocator
diff --git a/lib/test_scanf.c b/lib/test_scanf.c
index b620cf7de503..a2707af2951a 100644
--- a/lib/test_scanf.c
+++ b/lib/test_scanf.c
@@ -606,7 +606,7 @@ static void __init numbers_slice(void)
#define test_number_prefix(T, str, scan_fmt, expect0, expect1, n_args, fn) \
do { \
const T expect[2] = { expect0, expect1 }; \
- T result[2] = {~expect[0], ~expect[1]}; \
+ T result[2] = { (T)~expect[0], (T)~expect[1] }; \
\
_test(fn, &expect, str, scan_fmt, n_args, &result[0], &result[1]); \
} while (0)
generated by cgit v1.2.3 (git 2.46.0) at 2025-10-23 15:33:40 +0000