| Age | Commit message (Collapse) | Author |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux
Pull module updates from Luis Chamberlain:
"There are a few fixes / cleanups from Vincent, Chunhui, and Petr, but
the most important part of this pull request is the Rust community
stepping up to help maintain both C / Rust code for future Rust module
support. We grow the set of modules maintainers by three now, and with
this hope to scale to help address what's needed to properly support
future Rust module support.
A lot of exciting stuff coming in future kernel releases.
This has been on linux-next for ~ 3 weeks now with no issues"
* tag 'modules-6.12-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux:
module: Refine kmemleak scanned areas
module: abort module loading when sysfs setup suffer errors
MAINTAINERS: scale modules with more reviewers
module: Clean up the description of MODULE_SIG_<type>
module: Split modules_install compression and in-kernel decompression
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev
Pull fbdev fixes from Helge Deller:
- crash fix in fbcon_putcs
- avoid a possible string memory overflow in sisfb
- minor code optimizations in omapfb and fbcon
* tag 'fbdev-for-6.12-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev:
fbdev: sisfb: Fix strbuf array overflow
fbcon: break earlier in search_fb_in_map and search_for_mapped_con
fbdev: omapfb: Call of_node_put(ep) only once in omapdss_of_find_source_for_first_ep()
fbcon: Fix a NULL pointer dereference issue in fbcon_putcs
|
|
Pull drm fixes from Dave Airlie:
"Regular fixes for the week to end the merge window, i915 and xe have a
few each, amdgpu makes up most of it with a bunch of SR-IOV related
fixes amongst others.
i915:
- Fix BMG support to UHBR13.5
- Two PSR fixes
- Fix colorimetry detection for DP
xe:
- Fix macro for checking minimum GuC version
- Fix CCS offset calculation for some BMG SKUs
- Fix locking on memory usage reporting via fdinfo and BO destroy
- Fix GPU page fault handler on a closed VM
- Fix overflow in oa batch buffer
amdgpu:
- MES 12 fix
- KFD fence sync fix
- SR-IOV fixes
- VCN 4.0.6 fix
- SDMA 7.x fix
- Bump driver version to note cleared VRAM support
- SWSMU fix
- CU occupancy logic fix
- SDMA queue fix"
* tag 'drm-next-2024-09-28' of https://gitlab.freedesktop.org/drm/kernel: (79 commits)
drm/amd/pm: update workload mask after the setting
drm/amdgpu: bump driver version for cleared VRAM
drm/amdgpu: fix vbios fetching for SR-IOV
drm/amdgpu: fix PTE copy corruption for sdma 7
drm/amdkfd: Add SDMA queue quantum support for GFX12
drm/amdgpu/vcn: enable AV1 on both instances
drm/amdkfd: Fix CU occupancy for GFX 9.4.3
drm/amdkfd: Update logic for CU occupancy calculations
drm/amdgpu: skip coredump after job timeout in SRIOV
drm/amdgpu: sync to KFD fences before clearing PTEs
drm/amdgpu/mes12: set enable_level_process_quantum_check
drm/i915/dp: Fix colorimetry detection
drm/amdgpu/mes12: reduce timeout
drm/amdgpu/mes11: reduce timeout
drm/amdgpu: use GEM references instead of TTMs v2
drm/amd/display: Allow backlight to go below `AMDGPU_DM_DEFAULT_MIN_BACKLIGHT`
drm/amd/display: Fix kdoc entry for 'tps' in 'dc_process_dmub_dpia_set_tps_notification'
drm/amdgpu: update golden regs for gfx12
drm/amdgpu: clean up vbios fetching code
drm/amd/display: handle nulled pipe context in DCE110's set_drr()
...
|
|
Pull ceph updates from Ilya Dryomov:
"Three CephFS fixes from Xiubo and Luis and a bunch of assorted
cleanups"
* tag 'ceph-for-6.12-rc1' of https://github.com/ceph/ceph-client:
ceph: remove the incorrect Fw reference check when dirtying pages
ceph: Remove empty definition in header file
ceph: Fix typo in the comment
ceph: fix a memory leak on cap_auths in MDS client
ceph: flush all caps releases when syncing the whole filesystem
ceph: rename ceph_flush_cap_releases() to ceph_flush_session_cap_releases()
libceph: use min() to simplify code in ceph_dns_resolve_name()
ceph: Convert to use jiffies macro
ceph: Remove unused declarations
|
|
Pull smb server fixes from Steve French:
- fix querying dentry for char/block special files
- small cleanup patches
* tag 'v6.12-rc-ksmbd-server-fixes' of git://git.samba.org/ksmbd:
ksmbd: Correct typos in multiple comments across various files
ksmbd: fix open failure from block and char device file
ksmbd: remove unsafe_memcpy use in session setup
ksmbd: Replace one-element arrays with flexible-array members
ksmbd: fix warning: comparison of distinct pointer types lacks a cast
|
|
git://git.samba.org/sfrench/cifs-2.6
Pull xmb client fixes from Steve French:
- Noisy log message cleanup
- Important netfs fix for cifs crash in generic/074
- Three minor improvements to use of hashing (multichannel and mount
improvements)
- Fix decryption crash for large read with small esize
* tag '6.12rc-more-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
smb: client: make SHA-512 TFM ephemeral
smb: client: make HMAC-MD5 TFM ephemeral
smb: client: stop flooding dmesg in smb2_calc_signature()
smb: client: allocate crypto only for primary server
smb: client: fix UAF in async decryption
netfs: Fix write oops in generic/346 (9p) and generic/074 (cifs)
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
if an inode backpointer points to a dirent that doesn't point back,
that's an error we should warn about.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
If the reader acquires the read lock and then the writer enters the slow
path, while the reader proceeds to the unlock path, the following scenario
can occur without the change:
writer: pcpu_read_count(lock) return 1 (so __do_six_trylock will return 0)
reader: this_cpu_dec(*lock->readers)
reader: smp_mb()
reader: state = atomic_read(&lock->state) (there is no waiting flag set)
writer: six_set_bitmask()
then the writer will sleep forever.
Signed-off-by: Alan Huang <mmpgouride@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
If we shut down successfully, there shouldn't be any logged ops to
resume.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Add a filesystem flag to indicate whether we did a clean recovery -
using c->sb.clean after we've got rw is incorrect, since c->sb is
updated whenever we write the superblock.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
We had a bug where disk accounting keys didn't always have their version
field set in journal replay; change the BUG_ON() to a WARN(), and
exclude this case since it's now checked for elsewhere (in the bkey
validate function).
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
This was added to avoid double-counting accounting keys in journal
replay. But applied incorrectly (easily done since it applies to the
transaction commit, not a particular update), it leads to skipping
in-mem accounting for real accounting updates, and failure to give them
a version number - which leads to journal replay becoming very confused
the next time around.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
give bversions a more distinct name, to aid in grepping
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Previously, check_inode() would delete unlinked inodes if they weren't
on the deleted list - this code dating from before there was a deleted
list.
But, if we crash during a logged op (truncate or finsert/fcollapse) of
an unlinked file, logged op resume will get confused if the inode has
already been deleted - instead, just add it to the deleted list if it
needs to be there; delete_dead_inodes runs after logged op resume.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
BCH_SB_ERRS() has a field for the actual enum val so that we can reorder
to reorganize, but the way BCH_SB_ERR_MAX was defined didn't allow for
this.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
__bch2_fsck_err() warns if the current task has a btree_trans object and
it wasn't passed in, because if it has to prompt for user input it has
to be able to unlock it.
But plumbing the btree_trans through bkey_validate(), as well as
transaction restarts, is problematic - so instead make bkey fsck errors
FSCK_AUTOFIX, which doesn't need to warn.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
In order to check for accounting keys with version=0, we need to run
validation after they've been assigned version numbers.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
This fixes the following bug, where a disk accounting key has an invalid
replicas entry, and we attempt to add it to the superblock:
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): starting version 1.12: rebalance_work_acct_fix opts=metadata_replicas=2,data_replicas=2,foreground_target=ssd,background_target=hdd,nopromote_whole_extents,verbose,fsck,fix_errors=yes
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): recovering from clean shutdown, journal seq 15211644
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): accounting_read...
accounting not marked in superblock replicas
replicas cached: 1/1 [0], fixing
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): sb invalid before write: Invalid superblock section replicas_v0: invalid device 0 in entry cached: 1/1 [0]
replicas_v0 (size 88):
user: 2 [3 5] user: 2 [1 4] cached: 1 [2] btree: 2 [1 2] user: 2 [2 5] cached: 1 [0] cached: 1 [4] journal: 2 [1 5] user: 2 [1 2] user: 2 [2 3] user: 2 [3 4] user: 2 [4 5] cached: 1 [1] cached: 1 [3] cached: 1 [5] journal: 2 [1 2] journal: 2 [2 5] btree: 2 [2 5] user: 2 [1 3] user: 2 [1 5] user: 2 [2 4]
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): inconsistency detected - emergency read only at journal seq 15211644
accounting not marked in superblock replicas
replicas user: 1/1 [3], fixing
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): sb invalid before write: Invalid superblock section replicas_v0: invalid device 0 in entry cached: 1/1 [0]
replicas_v0 (size 96):
user: 2 [3 5] user: 2 [1 3] cached: 1 [2] btree: 2 [1 2] user: 2 [2 4] cached: 1 [0] cached: 1 [4] journal: 2 [1 5] user: 1 [3] user: 2 [1 5] user: 2 [3 4] user: 2 [4 5] cached: 1 [1] cached: 1 [3] cached: 1 [5] journal: 2 [1 2] journal: 2 [2 5] btree: 2 [2 5] user: 2 [1 2] user: 2 [1 4] user: 2 [2 3] user: 2 [2 5]
accounting not marked in superblock replicas
replicas user: 1/2 [3 7], fixing
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): sb invalid before write: Invalid superblock section replicas_v0: invalid device 7 in entry user: 1/2 [3 7]
replicas_v0 (size 96):
user: 2 [3 7] user: 2 [1 3] cached: 1 [2] btree: 2 [1 2] user: 2 [2 4] cached: 1 [0] cached: 1 [4] journal: 2 [1 5] user: 1 [3] user: 2 [1 5] user: 2 [3 4] user: 2 [4 5] cached: 1 [1] cached: 1 [3] cached: 1 [5] journal: 2 [1 2] journal: 2 [2 5] btree: 2 [2 5] user: 2 [1 2] user: 2 [1 4] user: 2 [2 3] user: 2 [2 5] user: 2 [3 5]
done
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): alloc_read... done
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): stripes_read... done
bcachefs (3c0860e8-07ca-4276-8954-11c1774be868): snapshots_read... done
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
accounting read was checking if accounting replicas entries were marked
in the superblock prior to applying accounting from the journal,
which meant that a recently removed device could spuriously trigger a
"not marked in superblocked" error (when journal entries zero out the
offending counter).
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Minor refactoring - replace multiple bool arguments with an enum; prep
work for fixing a bug in accounting read.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Dealing with outside state within a btree transaction is always tricky.
check_extents() and check_dirents() have to accumulate counters for
i_sectors and i_nlink (for subdirectories). There were two bugs:
- transaction commit may return a restart; therefore we have to commit
before accumulating to those counters
- get_inode_all_snapshots() may return a transaction restart, before
updating w->last_pos; then, on the restart,
check_i_sectors()/check_subdir_count() would see inodes that were not
for w->last_pos
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
dead code
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Returning a positive integer instead of an error code causes error paths
to become very confused.
Closes: syzbot+c0360e8367d6d8d04a66@syzkaller.appspotmail.com
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
The pointer clean points the memory allocated by kmemdup, when the
return value of bch2_sb_clean_validate_late is not zero. The memory
pointed by clean is leaked. So we should free it in this case.
Fixes: a37ad1a3aba9 ("bcachefs: sb-clean.c")
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
In downgrade_table_extra, the return value is needed. When it
return failed, we should exit immediately.
Fixes: 7773df19c35f ("bcachefs: metadata version bucket_stripe_sectors")
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
A couple small error handling fixes
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
this allows for various cleanups in fsck
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
syzbot reported a null ptr deref in __copy_user [0]
In __bch2_read_super, when a corrupt backup superblock matches the
default opts offset, no error is assigned to ret and the freed superblock
gets through, possibly being assigned as the best sb in bch2_fs_open and
being later dereferenced, causing a fault. Assign EINVALID to ret when
iterating through layout.
[0]: https://syzkaller.appspot.com/bug?extid=18a5c5e8a9c856944876
Reported-by: syzbot+18a5c5e8a9c856944876@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=18a5c5e8a9c856944876
Signed-off-by: Diogo Jahchan Koike <djahchankoike@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
check_topology doesn't need the srcu lock and doesn't use normal btree
transactions - we can just drop the srcu lock.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
fsck_err() jumps to the fsck_err label when bailing out; need to make
sure bp_iter was initialized...
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Zero-initialize part of allocated bounce buffer which wasn't touched by
subsequent bch2_key_sort_fix_overlapping to mitigate later uinit-value
use KMSAN bug[1].
After applying the patch reproducer still triggers stack overflow[2] but
it seems unrelated to the uninit-value use warning. After further
investigation it was found that stack overflow occurs because KMSAN adds
too many function calls[3]. Backtrace of where the stack magic number gets
smashed was added as a reply to syzkaller thread[3].
It was confirmed that task's stack magic number gets smashed after the code
path where KSMAN detects uninit-value use is executed, so it can be assumed
that it doesn't contribute in any way to uninit-value use detection.
[1] https://syzkaller.appspot.com/bug?extid=6f655a60d3244d0c6718
[2] https://lore.kernel.org/lkml/66e57e46.050a0220.115905.0002.GAE@google.com
[3] https://lore.kernel.org/all/rVaWgPULej8K7HqMPNIu8kVNyXNjjCiTB-QBtItLFBmk0alH6fV2tk4joVPk97Evnuv4ZRDd8HB5uDCkiFG6u81xKdzDj-KrtIMJSlF6Kt8=@proton.me
Reported-by: syzbot+6f655a60d3244d0c6718@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=6f655a60d3244d0c6718
Fixes: ec4edd7b9d20 ("bcachefs: Prep work for variable size btree node buffers")
Suggested-by: Kent Overstreet <kent.overstreet@linux.dev>
Signed-off-by: Piotr Zalewski <pZ010001011111@proton.me>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
This fixes a kasan splat in propagate_key_to_snapshot_leaves() -
varint_decode_fast() does reads (that it never uses) up to 7 bytes past
the end of the integer.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Most or all errors will be autofix in the future, we're currently just
doing the ones that we know are well tested.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
https://gitlab.freedesktop.org/agd5f/linux into drm-next
amd-drm-fixes-6.12-2024-09-27:
amdgpu:
- MES 12 fix
- KFD fence sync fix
- SR-IOV fixes
- VCN 4.0.6 fix
- SDMA 7.x fix
- Bump driver version to note cleared VRAM support
- SWSMU fix
amdgpu:
- CU occupancy logic fix
- SDMA queue fix
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240927202819.2978109-1-alexander.deucher@amd.com
|
|
The values of the variables xres and yres are placed in strbuf.
These variables are obtained from strbuf1.
The strbuf1 array contains digit characters
and a space if the array contains non-digit characters.
Then, when executing sprintf(strbuf, "%ux%ux8", xres, yres);
more than 16 bytes will be written to strbuf.
It is suggested to increase the size of the strbuf array to 24.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Signed-off-by: Andrey Shumilin <shum.sdl@nppct.ru>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fix from Rafael Wysocki:
"Fix idle states enumeration in the intel_idle driver on platforms
supporting multiple flavors of the C6 idle state (Artem Bityutskiy)"
* tag 'pm-6.12-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
intel_idle: fix ACPI _CST matching for newer Xeon platforms
|
|
select_rq_task() already checked that 'p->nr_cpus_allowed > 1',
'p->nr_cpus_allowed == 1' checker in scx_select_cpu_dfl() is redundant.
Signed-off-by: Zhang Qiao <zhangqiao22@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
The enable path uses three big locks - scx_fork_rwsem, scx_cgroup_rwsem and
cpus_read_lock. Currently, the locks are grabbed together which is prone to
locking order problems.
For example, currently, there is a possible deadlock involving
scx_fork_rwsem and cpus_read_lock. cpus_read_lock has to nest inside
scx_fork_rwsem due to locking order existing in other subsystems. However,
there exists a dependency in the other direction during hotplug if hotplug
needs to fork a new task, which happens in some cases. This leads to the
following deadlock:
scx_ops_enable() hotplug
percpu_down_write(&cpu_hotplug_lock)
percpu_down_write(&scx_fork_rwsem)
block on cpu_hotplug_lock
kthread_create() waits for kthreadd
kthreadd blocks on scx_fork_rwsem
Note that this doesn't trigger lockdep because the hotplug side dependency
bounces through kthreadd.
With the preceding scx_cgroup_enabled change, this can be solved by
decoupling cpus_read_lock, which is needed for static_key manipulations,
from the other two locks.
- Move the first block of static_key manipulations outside of scx_fork_rwsem
and scx_cgroup_rwsem. This is now safe with the preceding
scx_cgroup_enabled change.
- Drop scx_cgroup_rwsem and scx_fork_rwsem between the two task iteration
blocks so that __scx_ops_enabled static_key enabling is outside the two
rwsems.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Aboorva Devarajan <aboorvad@linux.ibm.com>
Link: http://lkml.kernel.org/r/8cd0ec0c4c7c1bc0119e61fbef0bee9d5e24022d.camel@linux.ibm.com
|
|
The disable path uses three big locks - scx_fork_rwsem, scx_cgroup_rwsem and
cpus_read_lock. Currently, the locks are grabbed together which is prone to
locking order problems. With the preceding scx_cgroup_enabled change, we can
decouple them:
- As cgroup disabling no longer requires modifying a static_key which
requires cpus_read_lock(), no need to grab cpus_read_lock() before
grabbing scx_cgroup_rwsem.
- cgroup can now be independently disabled before tasks are moved back to
the fair class.
Relocate scx_cgroup_exit() invocation before scx_fork_rwsem is grabbed, drop
now unnecessary cpus_read_lock() and move static_key operations out of
scx_fork_rwsem. This decouples all three locks in the disable path.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Aboorva Devarajan <aboorvad@linux.ibm.com>
Link: http://lkml.kernel.org/r/8cd0ec0c4c7c1bc0119e61fbef0bee9d5e24022d.camel@linux.ibm.com
|
|
scx_tg_online()
If the BPF scheduler does not implement ops.cgroup_init(), scx_tg_online()
didn't set SCX_TG_INITED which meant that ops.cgroup_exit(), even if
implemented, won't be called from scx_tg_offline(). This is because
SCX_HAS_OP(cgroupt_init) is used to test both whether SCX cgroup operations
are enabled and ops.cgroup_init() exists.
Fix it by introducing a separate bool scx_cgroup_enabled to gate cgroup
operations and use SCX_HAS_OP(cgroup_init) only to test whether
ops.cgroup_init() exists. Make all cgroup operations consistently use
scx_cgroup_enabled to test whether cgroup operations are enabled.
scx_cgroup_enabled is added instead of using scx_enabled() to ease planned
locking updates.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
scx_ops_init_task() and the follow-up scx_ops_enable_task() in the fork path
were gated by scx_enabled() test and thus __scx_ops_enabled had to be turned
on before the first scx_ops_init_task() loop in scx_ops_enable(). However,
if an external entity causes sched_class switch before the loop is complete,
tasks which are not initialized could be switched to SCX.
The following can be reproduced by running a program which keeps toggling a
process between SCHED_OTHER and SCHED_EXT using sched_setscheduler(2).
sched_ext: Invalid task state transition 0 -> 3 for fish[1623]
WARNING: CPU: 1 PID: 1650 at kernel/sched/ext.c:3392 scx_ops_enable_task+0x1a1/0x200
...
Sched_ext: simple (enabling)
RIP: 0010:scx_ops_enable_task+0x1a1/0x200
...
switching_to_scx+0x13/0xa0
__sched_setscheduler+0x850/0xa50
do_sched_setscheduler+0x104/0x1c0
__x64_sys_sched_setscheduler+0x18/0x30
do_syscall_64+0x7b/0x140
entry_SYSCALL_64_after_hwframe+0x76/0x7e
Fix it by gating scx_ops_init_task() separately using
scx_ops_init_task_enabled. __scx_ops_enabled is now set after all tasks are
finished with scx_ops_init_task().
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
scx_ops_enable() has two task iteration loops. The first one calls
scx_ops_init_task() on every task and the latter switches the eligible ones
into SCX. The first loop left the tasks in SCX_TASK_INIT state and then the
second loop switched it into READY before switching the task into SCX.
The distinction between INIT and READY is only meaningful in the fork path
where it's used to tell whether the task finished forking so that we can
tell ops.exit_task() accordingly. Leaving task in INIT state between the two
loops is incosistent with the fork path and incorrect. The following can be
triggered by running a program which keeps toggling a task between
SCHED_OTHER and SCHED_SCX while enabling a task:
sched_ext: Invalid task state transition 1 -> 3 for fish[1526]
WARNING: CPU: 2 PID: 1615 at kernel/sched/ext.c:3393 scx_ops_enable_task+0x1a1/0x200
...
Sched_ext: qmap (enabling+all)
RIP: 0010:scx_ops_enable_task+0x1a1/0x200
...
switching_to_scx+0x13/0xa0
__sched_setscheduler+0x850/0xa50
do_sched_setscheduler+0x104/0x1c0
__x64_sys_sched_setscheduler+0x18/0x30
do_syscall_64+0x7b/0x140
entry_SYSCALL_64_after_hwframe+0x76/0x7e
Fix it by transitioning to READY in the first loop right after
scx_ops_init_task() succeeds.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: David Vernet <void@manifault.com>
|
|
scx_ops_enable() used preempt_disable() around the task iteration loop to
switch tasks into SCX to guarantee forward progress of the task which is
running scx_ops_enable(). However, in the gap between setting
__scx_ops_enabled and preeempt_disable(), an external entity can put tasks
including the enabling one into SCX prematurely, which can lead to
malfunctions including stalls.
The bypass mode can wrap the entire enabling operation and guarantee forward
progress no matter what the BPF scheduler does. Use the bypass mode instead
to guarantee forward progress while enabling.
While at it, release and regrab scx_tasks_lock between the two task
iteration locks in scx_ops_enable() for clarity as there is no reason to
keep holding the lock between them.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
The distinction between SCX_OPS_PREPPING and SCX_OPS_ENABLING is not used
anywhere and only adds confusion. Drop SCX_OPS_PREPPING.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
check_hotplug_seq() is used to detect CPU hotplug event which occurred while
the BPF scheduler is being loaded so that initialization can be retried if
CPU hotplug events take place before the CPU hotplug callbacks are online.
As such, the best place to call it is in the same cpu_read_lock() section
that enables the CPU hotplug ops. Currently, it is called in the next
cpus_read_lock() block in scx_ops_enable(). The side effect of this
placement is a small window in which hotplug sequence detection can trigger
unnecessarily, which isn't critical.
Move check_hotplug_seq() invocation to the same cpus_read_lock() block as
the hotplug operation enablement to close the window and get the invocation
out of the way for planned locking updates.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: David Vernet <void@manifault.com>
|
