Zygo Blaxell [Wed, 21 Oct 2026 02:15:41 +0000 (22:15 -0400)]
btrfs: write dirty device stats when no transaction is running
Device stats are counted in memory and written to their dev stats items
only by btrfs_run_dev_stats(), which runs inside a transaction commit.
Nothing starts a transaction for them: the transaction kthread commits
only a running transaction, and at unmount close_ctree() calls
btrfs_commit_super(), whose btrfs_commit_current_transaction() only
attaches to a running transaction and returns 0 when there is none.
Read errors, checksum failures and read repair (which rewrites the bad
copy with bios outside any transaction) count stats without starting a
transaction. So on a read-write mount whose workload only reads, every
count since the last commit is lost at unmount, and the next mount
shows the old values. For example, a cold read of a file with one
corrupted raid1 copy counts corruption_errs and repairs the copy, and
after unmount and mount the count is gone.
Start and commit a transaction to write the stats when any device has
dirty stats, the filesystem was mounted read-write and is not read-only,
shut down or in an error state, and no transaction is running:
- the transaction kthread does it at its commit_interval wakeup, so the
counts reach the disk within one commit interval;
- btrfs_commit_super() does it after committing the running transaction,
which covers unmount and remount read-only, including counts made by
the scrub, replace and balance that close_ctree() and
btrfs_remount_ro() stop just before.
btrfs_remount_ro() takes and releases transaction_kthread_mutex after it
sets SB_RDONLY, so a kthread pass that saw the filesystem writable can
not start a transaction after the final commit. A frozen filesystem is
skipped (sb_start_write_trylock()) and its counts are written after
thaw. Log replay at mount happens before BTRFS_FS_OPEN is set and does
not start one. The check reuses the lockless RCU pass of
btrfs_run_dev_stats(), now btrfs_dev_stats_dirty(), so a healthy idle
filesystem pays one list walk per commit interval and gets no new
transactions. The one exception is a filesystem whose devices have no
dev stats item yet: mkfs does not create them, and loading the stats at
mount marks such a device dirty, so the first read-write mount writes
the items in one commit even when idle.
This changes btrfs behaviour. An otherwise read-only workload on a
read-write mount now causes transaction commits and metadata writes
whenever errors were counted: up to one commit per commit interval while
errors keep coming, and one more at unmount. Until now such a workload
wrote no metadata at all, and someone who relies on that, or who prefers
losing the counts to writing to a filesystem that is returning errors,
may reasonably reject this patch.
Read-only mounts are unchanged and still lose their counts at unmount,
as do a crash and a shutdown or aborted filesystem.
Cheng Yongkang [Sun, 1 Nov 2026 07:14:44 +0000 (00:14 -0700)]
btrfs: handle zero num_stripes in btrfs_chunk_item_size()
validate_sys_chunk_array() calls btrfs_chunk_item_size() with the
on-disk num_stripes before btrfs_check_chunk_valid() has checked the
chunk. With CONFIG_BTRFS_ASSERT=y, a zero-stripe system chunk trips
ASSERT(num_stripes) instead of being rejected:
assertion failed: num_stripes, in fs/btrfs/volumes.h:714
kernel BUG at fs/btrfs/volumes.h:714!
RIP: 0010:btrfs_validate_super+0x3ce7/0x4210
Call Trace:
open_ctree
btrfs_get_tree
vfs_get_tree
do_new_mount
Drop the assertion and compute the size as the chunk header plus
num_stripes stripes. The size is unchanged for num_stripes > 0, and a
zero-stripe chunk now reaches btrfs_check_chunk_valid(), which rejects
it, so the mount fails with -EUCLEAN.
Suggested-by: Qu Wenruo <wqu@suse.com> Link: https://lore.kernel.org/linux-btrfs/d2ee0cc1-674e-4447-88da-d346122a43a4@suse.com/ Fixes: 2a9bb78cfd36 ("btrfs: validate system chunk array at btrfs_validate_super()") Signed-off-by: Cheng Yongkang <teel4res@gmail.com>
[ Slightly rework the commit message. ] Signed-off-by: Qu Wenruo <wqu@suse.com>
Cheng Yongkang [Sun, 1 Nov 2026 06:25:39 +0000 (23:25 -0700)]
btrfs: reject refs=0 subvolume without DEAD in SUBVOL_SYNC_WAIT
btrfs_ioctl_subvol_sync() assumes that a cached root with zero
references has BTRFS_ROOT_SUBVOL_DEAD set, and asserts that before
waiting for the cleaner. Two states break the assumption:
- A corrupt ROOT_ITEM with zero references and a matching orphan item,
on a writable filesystem.
- A failed subvolume deletion. btrfs_delete_subvolume() zeroes the
root references, aborts the transaction, which forces the filesystem
read-only, and then clears DEAD again. No on-disk metadata is
corrupted in this case.
With CONFIG_BTRFS_ASSERT=y, both reach a BUG().
Change the ASSERT() into a proper rejection to avoid triggering a BUG().
Fixes: 6c83d153ed86 ("btrfs: add new ioctl to wait for cleaned subvolumes") Cc: stable@vger.kernel.org Suggested-by: Qu Wenruo <quwenruo.btrfs@gmx.com> Link: https://lore.kernel.org/linux-btrfs/dd97efe9-91ec-495b-b698-acff1447421c@gmx.com/ Signed-off-by: Cheng Yongkang <teel4res@gmail.com>
[ Rework the commit message, drop the ad-hoc RW check. ] Signed-off-by: Qu Wenruo <wqu@suse.com>
Daniel Vacek [Fri, 30 Oct 2026 14:28:03 +0000 (15:28 +0100)]
btrfs: fix a livelock in wait_for_commit() on PREEMPT_RT
Peter reported [1] that running multiple RR tasks doing sync()
on fully preemptible kernel may starve other/normal sync() tasks
due to fs_info::trans_lock turned into rt_mutex.
This is caused by RR tasks contending over the mutex with priority over
other tasks, which in turn lack progress and they are no longer able to
remove TRANS_STATE_COMPLETED transactions from the fs_info::trans_list.
The offenders start looping in wait_for_commit() expecting the completed
transactions to get off the trans_list but since they're bouncing the
trans_lock the tasks with other/normal priority which are about to delist
such transactions never get a chance to do so due to lower priority.
A simple reproducer can be found in the mentioned email thread [1].
The situation can be resolved by boosting the priority of the starved
tasks or dropping priority of the offenders. Thought this fix prevents
the situation from happening in the first place. We simply delist the
transactions before changing them to completed state. That breaks the
loop condition.
Binbin Deng [Wed, 14 Oct 2026 02:37:33 +0000 (10:37 +0800)]
btrfs: wait for a grace period before freeing a new device
KASAN reports a slab-use-after-free in btrfs_statfs() walking
fs_devices->devices. btrfs_init_new_device() publishes the new device
with list_add_rcu() and can fail afterwards, for example with -ENOSPC
in init_first_rw_device() when a user with CAP_SYS_ADMIN adds a device
that is too small to a seeding filesystem mounted read-only. The error
path removes the entry with list_del_rcu() and frees the object in
btrfs_free_device() without waiting for a grace period, while
btrfs_calc_avail_data_space() and btrfs_ioctl_fs_info() keep reading
the device list under rcu_read_lock().
BUG: KASAN: slab-use-after-free in btrfs_statfs+0x11af/0x14a0
Read of size 8 at addr ffff8880096be858 by task poc/252
Call Trace:
<TASK>
dump_stack_lvl+0x53/0x70
print_report+0xd0/0x630
? __pfx__raw_spin_lock_irqsave+0x10/0x10
? filename_lookup+0x1a9/0x540
? btrfs_statfs+0x11af/0x14a0
kasan_report+0xce/0x100
? btrfs_statfs+0x11af/0x14a0
btrfs_statfs+0x11af/0x14a0
statfs_by_dentry+0x117/0x1e0
user_statfs+0xac/0x130
? __pfx_user_statfs+0x10/0x10
__do_sys_statfs+0x80/0xe0
? __pfx___do_sys_statfs+0x10/0x10
do_syscall_64+0xf9/0x540
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Fix by waiting for an RCU grace period after the device is removed
from the list and before it is freed.Place the synchronize_rcu() after
dropping the mutexes and before the `error_trans` label.
Fixes: 1f78160ce1b1 ("Btrfs: using rcu lock in the reader side of devices list") Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Binbin Deng <18983559317@163.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Sat, 24 Oct 2026 23:39:50 +0000 (10:09 +1030)]
btrfs: validate but do not set bdev blocksize
[BACKGROUND]
In the very early days, commit 325cd4bafeb6 ("Btrfs: properly set
blocksize when adding new device.") set the bdev blocksize as at that
time btrfs is using __bread() for super block IO, which went through
buffer heads thus requires the strict block size limit.
Then btrfs super block IO moved away from buffered heads, by commit 8f32380d3f29 ("btrfs: use the page cache for super block reading") and
commit 314b6dd0eebf ("btrfs: use bios instead of buffer_heads from super
block writeout").
So after that there is already no strict blocksize requirement for
btrfs, except the bdev's block size should not exceed 4K, which is also
the super block size of btrfs.
Btrfs data read/writes all follow sectorsize, which can be larger than
4K.
Btrfs metadata read/writes all follow nodesize, which is already 16K by
default.
Furthermore, there is already a bug report that set_blocksize() can
trigger a lockdep warning from btrfs, as set_blocksize() will acquire
inode lock, and under certain races can lead to lockdep warnings.
[ENHANCEMENT]
Remove the set_blocksize() calls, instead call bdev_validate_blocksize()
inside btrfs_open_device_deny_freeze() and btrfs_get_bdev_and_sb().
Btrfs doesn't use buffer heads, thus no strict set_blocksize()
requirement. But we still want to call bdev_validate_blocksize(), just in
case where a future block device with larger blocksize (e.g. 16K) shows
up.
For such devices btrfs super block bios will be rejected, so we want to call
bdev_validate_blocksize() early to reject such devices.
Furthermore, the bdev_validate_blocksize() itself is lockless, thus it
should resolve the lockdep report from syzbot.
Filipe Manana [Wed, 28 Oct 2026 15:37:59 +0000 (15:37 +0000)]
btrfs: zstd: pass fs_info to btrfs_warn() on failure to allocate workspace manager
The workspace manager is per filesystem and we have access to the fs_info
structure of the filesystem in zstd_alloc_workspace_manager(), so pass it
to the btrfs_warn() call so that we get a better log message (including
the device, state, etc).
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 28 Oct 2026 16:06:27 +0000 (16:06 +0000)]
btrfs: zstd: prevent leak of workspace below maximum level
When putting a workspace we always protect the workspace for the maximum
level from reclaim, since it can be used not just for that level but as
well for any level below it. However the logic is flawed because if we do
the put for a workspace at level 10 for example (maximum is level 15) and
there is currently no idle workspace for the maximum level because all
are currently being used, we don't add the workspace to the LRU list and
that can result in leaking it if we don't get more compression or
decompression requests for level 10 or any level below 10.
Fix this by skipping the addition to the LRU list only if the level of
the workspace matches the maximum level too (besides the list of idle
workspaces for the maximum level being empty).
Fixes: 3f93aef535c8 ("btrfs: add zstd compression level support") Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 28 Oct 2026 06:30:47 +0000 (17:00 +1030)]
btrfs: don't cache rbio if an error is hit
If rmw_rbio() hit an error after cache_rbio_pages() call, the rbio still
have RBIO_CACHE_READY_BIT set, thus can be utilized for future page
stealing.
But that rbio didn't really reach the disk, thus using the pages for
future steal can cause data inconsistency between in-memory cache and
on-disk data.
Fix the problem by clearing the RBIO_CACHE_READY_BIT on the error path
of rmw_rbio().
Fixes: 5eb30ee26fa4 ("btrfs: raid56: introduce the main entrance for RMW path") Assisted-by: LLM (find the bug) Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 28 Oct 2026 06:30:46 +0000 (17:00 +1030)]
btrfs: fix racy bio_list changes for failed RMW rbio
If an RMW work failed early, e.g. alloc_rbio_parity_pages() failed, it
won't get RBIO_RMW_LOCKED_BIT set, and get rbio_orig_end_io() called.
In that case, there can be a race between bio_list_get() on the rbio,
and new incoming bio merging into the rbio:
A is the rbio that rbio_orig_end_io() is running on.
a is the higher layer bio that generated the rbio A.
B is the new rbio that will be merged into A.
b is the higher layer bio that generated the rbio B.
rbio_orig_end_io(A) worker | lock_stripe_add(B)
--------------------------------------+---------------------------------
rbio_orig_end_io(A) | lock_stripe_add(B)
|- cur = bio_list_get(A) | |- rbio_can_merge(A, B)
| |- bio = A->bio_list.head; | | which returned true
| | which bio a is returned. | |
| | | |
| |- A->bio_list.tail = NULL; | |
| | | |
| | | |- merge_rbio(A, B)
| | | | |- bio_list_merge_init()
| | | | |- if (A->bio_list.tail)
| | | | | A->bio_list.tail is
| | | | | still NULL
| | | | |
| | | | |- A->bio_list.head = b
| | | | |- A->bio_list.tail = b
| |- A->bio_list.head = NULL; | |- return 1;
| |- return bio; | Caller expect rbio A to
| which is bio a. | complete bio b.
|
|- extra = bio_list_get()
| Which is NULL
|
|- rbio_endio_bio_list(cur);
Which only ends bio a.
In above case, the bio_list race can get bio b lost, and never ended.
Thankfully this race window is very tiny, so not practical in the real
world.
Fix it by setting RBIO_RMW_LOCKED_BIT for error path before calling
rbio_orig_end_io().
And since we're ensuring all RMW rbios have RBIO_RMW_LOCKED_BIT bit set
before calling rbio_orig_end_io(), add two ASSERT()s for that and remove
the old comment.
Fixes: 5eb30ee26fa4 ("btrfs: raid56: introduce the main entrance for RMW path") Assisted-by: LLM (find and analyze the bug) Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Thomas Fourier [Wed, 28 Oct 2026 09:50:51 +0000 (10:50 +0100)]
btrfs: zoned: fix double free of bitmaps on clone failure
btrfs_clone_dev_zone_info() duplicates the original orig_dev->zone_info
with kmemdup(), which also copies the seq_zones, empty_zones and
active_zones pointers. The clone then replaces them one by one with
fresh bitmaps. If one of the allocations fails, the error path frees all
three pointers, including those not yet replaced, which still point to
the bitmaps owned by the original device.
Clear the three pointers right after kmemdup() so the error path only frees
what this function allocated.
Fixes: 21e61ec6d0bb ("btrfs: zoned: clone zoned device info when cloning a device") Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Thomas Fourier <fourier.thomas@gmail.com> Signed-off-by: David Sterba <dsterba@suse.com>
When mounting a filesystem on a zoned drive passed into a VM via a
virtio-blk disk lockdep sporadically complains about a possible deadlock
between BTRFS' dev_replace::rwsem and virtio-blk's vdev_mutex:
======================================================
WARNING: possible circular locking dependency detected
7.3.0-rc6+ #317 Not tainted
------------------------------------------------------
mount/7418 is trying to acquire lock: ffff888102c07860 (&vblk->vdev_mutex){+.+.}-{4:4}, at: virtblk_report_zones+0xda/0x440
but task is already holding lock: ffff8881fd725ea8 (&fs_info->dev_replace.rwsem){++++}-{4:4}, at: btrfs_load_block_group_zone_info+0x186/0xe50
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
locks held by mount/7418: 3, last CPU#0:
#0: ffff888110aa6868 (&fc->uapi_mutex){+.+.}-{4:4}, at: __do_sys_fsconfig+0x2ae/0x690
#1: ffff888112ba20e8 (&type->s_umount_key#31/1){+.+.}-{4:4}, at: sget_fc+0x148/0x8d0
#2: ffff8881fd725ea8 (&fs_info->dev_replace.rwsem){++++}-{4:4}, at: btrfs_load_block_group_zone_info+0x186/0xe50
Holding the rwsem was widened in 0090d6e1b210 ("btrfs: zoned: fix
use-after-free due to race with dev replace"), but it now also covers
calling into blkdev_report_zones_cached(). The device replace ioctl()
only works on a mounted filesystem and the only caller with setting
'new' to false is read_one_block_group() which is only called during
mount, so it is safe to drop the rwsem before calling into
blkdev_report_zones_cached().
Fixes: 0090d6e1b210 ("btrfs: zoned: fix use-after-free due to race with dev replace") Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: zoned: don't wait for ordered extents in do_zone_finish()
do_zone_finish() calls btrfs_wait_ordered_roots(), which holds
fs_info->ordered_operations_mutex while waiting for writeback. It can be
reached from writeback itself:
If another task holds the mutex and waits for that writeback, this
deadlocks.
In general, we want to ensure all writes in the block group to
complete before finishing the zone, as relocation does. However, the
current callers do not need the wait (or cannot wait) in do_zone_finish():
- Relocation already waits for reservations and ordered extents itself
before calling btrfs_zone_finish().
- btrfs_delete_unused_bgs() checks that ->reserved is zero and sets the
block group read-only while holding groups_sem, so there can be no
in-flight data write.
- btrfs_zone_finish_one_bg() is called from the allocation path and
cannot wait.
Dropping the wait is still safe. Once the block group is read-only,
->reserved can only drop, and for data it only drops after the ordered
extent's I/O has completed. So if ->reserved is zero, no data write is
left in flight. If it is non-zero, do_zone_finish() returns -EAGAIN,
which all callers handle.
This makes btrfs_zone_finish_one_bg() fail with -EAGAIN when it races
with an allocation in the selected block group.
Fixes: 393f646e34c1 ("btrfs: zoned: finish least available block group on data bg allocation") Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 26 Oct 2026 17:34:10 +0000 (17:34 +0000)]
btrfs: lzo: use memcpy_to_folio() in copy_compressed_data_to_bio()
We are manually mapping the folio, copying to it with memcpy() and then
unmapping, but this can all simply be done with memcpy_to_folio() that
is provided by the MM layer. So use memcpy_to_folio() instead.
Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Johannes Thumshirn <johannes.thumshrin@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 26 Oct 2026 17:31:37 +0000 (17:31 +0000)]
btrfs: zlib: use memcpy_from_folio() in copy_data_into_buffer()
We are manually mapping the folio, copying from it with memcpy() and then
unmapping, but this can all simply be done with memcpy_from_folio() that
is provided by the MM layer. So use memcpy_from_folio() instead.
Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Johannes Thumshirn <johannes.thumshrin@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Thu, 22 Oct 2026 09:39:29 +0000 (20:09 +1030)]
btrfs: add stripe_len < PAGE_SIZE support for alloc_rbio_*_pages()
The functions alloc_rbio_parity_pages() and alloc_rbio_data_pages()
didn't support stripe_len < PAGE_SIZE cases, and fallback to
alloc_rbio_pages() instead.
That's mostly to avoid the case where both data and P/Q blocks can exist
in the same page.
However even in that case, it's not a big deal, the existing
infrastructures like index_stripe_sectors() can handle it correctly.
Update those functions to support stripe_len < PAGE_SIZE cases, with two
points to note:
- Extra attention needed for the round direction
For alloc_rbio_data_pages() we need to round up, so the last data page
gets allocated.
For alloc_rbio_parity_pages() we need to round down, so the first
P/Q page get allocated.
- Mixed page allocation
We can call alloc_rbio_parity_pages() first, then
alloc_rbio_data_pages().
In that case, it's alloc_pages_bulk() saving us, which always skips
allocated slot. Thus we're still safe.
Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Thu, 22 Oct 2026 09:39:28 +0000 (20:09 +1030)]
btrfs: add stripe_len < PAGE_SIZE support for steal_rbio()
The function steal_rbio() is utilized to use the pages that belongs to
uptodate data blocks, without doing a page copy.
That is done by just updating the dest->stripe_pages[] pointer to use
the point inside @src directly.
That function previously will reject stripe_len < PAGE_SIZE cases, as a
page can contain both data and P/Q blocks.
But that shouldn't be a big deal, as long as the corresponding P/Q
blocks are not marked uptodate, we can still steal the page that also
contains P/Q blocks.
So rework steal_rbio() to support stripe_len < PAGE_SIZE by:
- Calculate the last page index correctly
Instead of iterating all pages but skip P/Q pages, calculate the last
page index properly.
The last page may contain both data and P/Q blocks, but we can still
steal the page.
- Update the bitmap in one go
Instead of updating the stripe_uptodate_bitmap for every page, we can
do the update in just one go, out of the main loop.
* full_page_sectors_uptodate()
This one is very annoying, because of its __maybe_unused
attribution.
They are based on the old BTRFS_STRIPE_LEN >= PAGE_SIZE assumption,
and no longer utilized.
- Make index_stripe_sectors() update the stripe_paddrs[] without a page
This doesn't cause any functional change for now, as there is no one
accessing the source stripe_paddrs[] after a steal.
But this should still make the function a little more robust.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 23 Oct 2026 09:26:08 +0000 (10:26 +0100)]
btrfs: remove fs_info argument from btrfs_add_bg_to_space_info()
It's not used, so remove it.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 23 Oct 2026 09:16:54 +0000 (10:16 +0100)]
btrfs: remove redundant space_info assignment in btrfs_add_bg_to_space_info()
There's no point in setting the block group's space_info member as the
space_info was extracted from the block group early when the variable was
declared and did not change.
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 21 Oct 2026 19:34:59 +0000 (20:34 +0100)]
btrfs: send: fix missing hole punching during incremental send
When using the no-holes feature (enabled by default since btrfs-progs
v5.15) and doing an incremental send, we can miss issuing a hole punching
command when the following conditions happen:
1) A hole is punched in a tail section of a file extent item that happens
to be the last item in a leaf (for example the extent covers the file
range [1M, 3M) and a hole is punched in the range [2M, 3M)). Between
the end of this file extent item's range and the i_size, there are
other file extent items;
2) The next leaf was not modified in the send snapshot, remaining shared
with the parent snapshot.
In this scenario we are unable to detect a hole was punched and never
issue a hole punch command because with no-holes there we don't have an
explicit file extent item to mark a hole.
Fix this by teaching maybe_send_hole() to detect such case.
A test case for fstests will follow soon.
btrfs: force a transaction commit when fsyncing stripe-tree extents
An extent in a block-group that uses the RAID stripe-tree has its logical
to physical mapping stored in that tree, which only becomes persistent
when the stripe-tree is written out at transaction commit. But the RAID
stripe-tree is not logged in the log tree.
A fsync therefore logs a file-extent item whose logical range cannot be
resolved after a crash. Replay restores the file-extent item and the
file has the correct size, but reading it fails with EIO and nothing is
reported.
Force a full commit when logging such an extent. Extents reach the log
both from log_one_extent() and from copy_items(), so check in both.
copy_items() skips extents from a past transaction, their mapping is
already committed, the same reason their checksums are not logged.
Fixes: 02c372e1f016 ("btrfs: add support for inserting raid stripe extents") Reviewed-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 22:48:47 +0000 (09:18 +1030)]
btrfs: remove BUG*() from raid56.c
There are several BUG()/BUG_ON() inside raid56.c, which cause
checkpatch warning if code is touching them. This is especially annoying
if using b4, with btrfs-workflow git hooks applied, which makes b4 to
wait for user input but without any output.
Address them by removing one very impossible BUG(), and using ASSERT() for
the remaining ones:
- __remove_rbio_from_cache()
- rbio_is_full()
- unlock_stripe()
Replace the BUG_ON() with ASSERT().
- finish_parity_scrub()
Remove the BUG(), as alloc_rbio() has already properly calculated
real_stripes and nr_data using the number of parity.
Furthermore, initialize @has_qstripe at the beginning of the function
using the bioc->map_type directly.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:43 +0000 (14:09 +1030)]
btrfs: introduce extra limits on btrfs_chunk::stripe_len
The new limits are:
- Stripe_len must be no smaller than sectorsize
As sectorsize is still the minimal unit.
- Stripe_len must be no larger than BTRFS_STRIPE_LEN
We still have a lot of things requiring on BTRFS_STRIPE_LEN alignment,
e.g. dev extent and chunk allocation.
A stripe_len larger than BTRFS_STRIPE_LEN will cause problems for the
existing chunk allocator.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:41 +0000 (14:09 +1030)]
btrfs: convert raid56.c to use rbio->bioc->stripe_len
So that raid56 can support stripe length other than BTRFS_STRIPE_LEN.
There are several extra things to note:
- btrfs_raid_bio::stripe_npages is removed
With stripe_len < BTRFS_STRIPE_LEN, there is no guarantee that a page
won't cross multiple stripes.
E.g. with 4K stripe_len and sectorsize, 2 data stripes, with 64K page
size, one page will cover the full stripe.
So stripe_npages no longer makes sense, just remove it.
- alloc_rbio_data_pages()/alloc_rbio_parity_pages() fallback
For the same above reason, a page can contain both data and parity
blocks.
In that case we can no longer only allocate pages for data/parity.
Add a fallback that if stripe_len is smaller than page size, always
allocate all the pages.
This is not the best solution, but should be good enough for now until
the variable stripe_len is properly tested.
- steal_rbio() fallback
If stripe_len is smaller than PAGE_SIZE, it's hard to only steal the
data pages.
Again skip the stealing if stripe_len is smaller than PAGE_SIZE.
Reviewed-by: Johannes Thusmhirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:39 +0000 (14:09 +1030)]
btrfs: add a new 'stripe_len' parameter to btrfs_stripe_nr_to_offset()
That helper is currently using a fixed BTRFS_STRIPE_LEN, change it to
accept a 'u32 stripe_len'.
Also add an ASSERT() to make sure the @stripe_len passed in is power of
2, and no larger than BTRFS_STRIPE_LEN.
There is a special case where we still pass BTRFS_STRIPE_LEN into
btrfs_stripe_nr_to_offset(), that's for chunk allocation to determine
the minimal device extent size.
In that case, since our future planned map->stripe_len is always power
of 2 and no larger than BTRFS_STRIPE_LEN, we are totally fine to use the
old BTRFS_STRIPE_LEN.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:38 +0000 (14:09 +1030)]
btrfs: change scrub to use stripe_len
Several points to note:
- scrub_ctx::stripes[] stripe_len updates
We always pre-initialize scrub_ctx::stripes[] at the very beginning of
scrub, without holding any btrfs_chunk_map.
So init_scrub_stripe() still uses BTRFS_STRIPE_LEN, and all allocated
memory are based on that stripe length.
But before scrubbing each chunk, update_scrub_stripe_len() will be
called to update the scrub_stripe::stripe_len.
- Stripe_len should not exceed BTRFS_STRIPE_LEN
Since scrub_stripe's heap memory are all based on BTRFS_STRIPE_LEN,
any stripe_len larger than BTRFS_STRIPE_LEN will cause
access-beyond-boundary.
Thankfully there is no plan to support such stripe_len >
BTRFS_STRIPE_LEN any time soon.
- alloc_scrub_bbio() changes
Previously we ensure BTRFS_STRIPE_LEN >= PAGE_SIZE, so
BTRFS_STRIPE_LEN >> PAGE_SHIFT can never be zero.
But now stripe_len can be smaller than PAGE_SIZE. E.g. stripe_len 4K,
page size 64K.
So we have to use "round_up(stripe_len, PAGE_SIZE) >> PAGE_SHIFT"
instead.
Finally alloc_scrub_bbio() is always called with a scrub_stripe
pointer as the @private parameter, change the type from (void *) to
(struct scrub_stripe *), so that we can grab the stripe_len without
a new parameter.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:36 +0000 (14:09 +1030)]
btrfs: load the on-disk btrfs_chunk::stripe_len
Currently btrfs doesn't load btrfs_chunk::stripe_len into
btrfs_chunk_map, as tree-checker will reject any chunk whose stripe_len
is not 64K.
But for the future we will want stripe length other than 64K.
As a preparation for the variable stripe length in the future, load
btrfs_chunk::stripe_len into btrfs_chunk_map::stripe_len and
stripe_shift.
For now stripe_len is still 64K, no matter if it's loaded from the
on-disk metadata or newly allocated chunk.
As tree-checker still rejects any chunk whose stripe_len is not 64K.
The two new members fit into the existing hole of btrfs_chunk_map, so
the size of btrfs_chunk_map stay the same.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Wed, 21 Oct 2026 03:39:35 +0000 (14:09 +1030)]
btrfs: reduce the width of num_stripes and sub_stripes
The on-disk btrfs_chunk structure uses u16 for both num_stripes and
sub_stripes, so there is no need to use "int" to represent them in
btrfs_chunk_map.
Follow the on-disk format to use u16 for those members, including
verified_stripes too.
And since verified_stripes is u16, it no longer completely fill the
hole between @refs and @start.
Instead use @num_stripes and @sub_stripes to fill the hole, and leave a 6
bytes hole between @verified_stripes and @stripes.
This doesn't save any space for btrfs_chunk_map, but leaves enough space
for future expansion.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: zoned: avoid underflow of space info bytes_zone_unusable
The two functions btrfs_free_reserved_bytes() and
btrfs_reset_unused_block_groups() both manipulate
space_info->bytes_zone_unusable without using the accessor function
btrfs_space_info_update_bytes_zone_unusable() potentially risking an
underflow (in case of btrfs_reset_unused_block_groups()).
Use btrfs_space_info_update_bytes_zone_unusable() in both function, in
the case of btrfs_free_reserved_bytes() it is mostly for documenting
purposes.
Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: zoned: requeue block group if zone reset bails out
btrfs_reset_unused_block_groups() drops the block-group from
fs_info->unused_bgs before performing the work that can actually fail,
the actual submission of a REQ_OP_ZONE_RESET.
Every failed operation after this leaves the block-group dangling,
physically reset, but the in-memory structures aren't and it is not
requeued to be reset on fs_info->unused_bgs.
Instead of bailing out, re-add this block-group to fs_info->unused_bgs
list.
Fixes: 453a73c3069a ("btrfs: zoned: reclaim unused zone by zone resetting") Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: zoned: fix possible UAF in wait_eb_writebacks()
wait_eb_writebacks() iterates the fs_info->buffer_tree xarray and waits
for the writeback of these extent-buffers. Waiting for writeback needs
to be done without the rcu_read_lock() held (because it can sleep) and
thus the loop drops the rcu_read_lock() before calling into
wait_on_extent_buffer_writeback(). But extent_buffers are freed through
RCU, so if btrfs_release_extent_buffer_rcu() is scheduled while we're
still waiting on the writeback, the extent_buffer will be freed causing
a use-after-free.
Similar to what is done in find_extent_buffer_nolock(), get a reference
to the extent_buffer before calling into
wait_on_extent_buffer_writeback() so a successful writeback does not
free the extent_buffer while we still have a reference to it.
Fixes: 2dd7e7bc0282 ("btrfs: zoned: wait for extent buffer IOs before finishing a zone") Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: zoned: only change active zone counter on successful (de)activation
btrfs_zone_activate() unconditionally decrements the reserved_active_zones
counter regardless if btrfs_dev_set_active_zone() fails to set the zone
active, i.e. because it raced with another call that already set the bit
in the bitmask, or not.
This can lead to a double decrement of the counter in case the bit has
already been set.
Only decrement the counter *iff* btrfs_dev_set_active_zone() successfully
marked the zone in the bitmap.
Mirror this behaviour when clearing the bit again.
Fixes: a7e1ac7bdc5a ("btrfs: zoned: reserve zones for an active metadata/system block group") Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
Tao Cui [Fri, 9 Oct 2026 07:53:25 +0000 (15:53 +0800)]
btrfs: allow idmapped DEFRAG ioctls
btrfs_ioctl_defrag() checks MAY_WRITE against nop_mnt_idmap, so on an
idmapped mount the owner comparison uses the caller's fsuid against the
raw on-disk uid and the ioctl fails with -EPERM even for the file's
owner. Pass the mount idmap down so defrag works on idmapped mounts.
The check, added in 616d374efa23 ("btrfs: allow defrag on a file opened
read-only that has rw permissions"), is not a privilege gate. It only
tests whether the file could have been opened for writing, and a user
who owns the file on the idmapped mount can already open it O_RDWR,
write() to it or fallocate() it. Defragmenting a regular file only
rearranges the caller's own extents; there is nothing it can rewrite
that the caller could not rewrite anyway. Whole-subvolume defrag on a
directory keeps its capable(CAP_SYS_ADMIN) requirement, and the
!capable() guard around this check is left as is.
FIDEDUPERANGE already works this way for unprivileged callers:
may_dedupe_file() in fs/remap_range.c compares the inode owner through
file_mnt_idmap(file) and falls back to inode_permission() with the same
idmap, and dedupe can rewrite one file's extents from another file's
contents, which is a stronger operation than defrag.
On regular mounts file_mnt_idmap() is nop_mnt_idmap and nothing
changes.
Reviewed-by: Seth Forshee <sforshee@kernel.org> Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Christian Brauner (Amutable) <brauner@kernel.org> Signed-off-by: Tao Cui <cuitao@kylinos.cn> Signed-off-by: David Sterba <dsterba@suse.com>
guolingxing [Sun, 18 Oct 2026 15:38:10 +0000 (15:38 +0000)]
btrfs: prevent deleting the wrong subvolume by ID
When deleting a subvolume by ID, btrfs_ioctl_snap_destroy() first
resolves the requested ID to a name, then looks up that name again
under the parent directory lock.
A concurrent rename can change which subvolume is associated with
the name between these operations. Since the deletion path does not
verify the ID of the resolved subvolume, it can delete a different
subvolume while leaving the requested one intact.
This was reproduced by repeatedly exchanging the names of two
subvolumes using renameat2(RENAME_EXCHANGE) while deleting one of
them by ID. The wrong subvolume was deleted in 9 out of 14 independent
tests.
Fix this by checking that the resolved subvolume ID matches the
requested ID while holding the parent directory lock. Return -EAGAIN
on a mismatch so that userspace can retry the operation instead of
being told that the requested subvolume does not exist.
Fixes: 949964c92843 ("btrfs: add new BTRFS_IOC_SNAP_DESTROY_V2 ioctl") CC: stable@vger.kernel.org Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: guolingxing <guolingxing@tju.edu.cn> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Sun, 18 Oct 2026 04:17:02 +0000 (14:47 +1030)]
btrfs: make chunk size validation to be simpler
Commit 6ded22c1bfe6 ("btrfs: reduce div64 calls by limiting the number
of stripes of a chunk to u32") introduced a chunk size limit check, but
it is using a wrong result in the comment:
U32_MAX << BTRFS_STRIPE_LEN_SHIFT results 256T, not 256G.
Instead of calculating the size limit, using an immediate number
(SZ_256G).
This has the following benefit:
- Still large enough for future expansion
Although the check is 1024 times more strict, I don't think we will
allow 256G chunk anytime soon.
And we have higher chance (10 more bits) to catch a bitflip in the
high bits.
- Much harder to diverge from the comment
We don't need to calculate the number anymore.
Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 16 Oct 2026 17:02:23 +0000 (18:02 +0100)]
btrfs: scrub: fix race between scrub and scrub progress ioctl
The scrub progress ioctl is copying the scrub progress stats without
acquiring the spinlock that protects it and a concurrent scrub can be
updating the stats at the same time which can lead to load tearing
issues. So take the spinlock when copying the progress stats.
Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 16 Oct 2026 16:53:00 +0000 (17:53 +0100)]
btrfs: scrub: fix race leading to use-after-free in the scrub progress ioctl
If the scrub progress ioctl is called when replace is about to finish, we
have a race that can result in a use-after-free of the scrub context:
1) A scrub is finishing and is at btrfs_scrub_dev();
2) A concurrent task called the scrub progress ioctl and is in
btrfs_scrub_progress(), where it takes the device_list_mutex and grabs
the device. Then it starts using the device's scrub context (for a
memcpy operation);
3) While the scrub progress ioctl task is running, the scrub task sets
the device's scrub_ctx to NULL (acquiring fs_info->scrub_lock) and
then calls scrub_put_ctx() against the scrub context, dropping the
last reference and freeing the memory - while the scrub progress ioctl
task is still using it, resulting in a use-after-free.
Fix this by changing btrfs_scrub_progress() to take fs_info->scrub_lock
before using the scrub context.
Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 16 Oct 2026 16:14:49 +0000 (17:14 +0100)]
btrfs: scrub: fix lost wakeup and hang when cancelling a scrub that is finishing
When scrub finishes, it does a wakeup on the fs_info->scrub_pause_wait
wait queue and then sets dev->scrub_sctx to NULL. On the other hand the
scrub cancel code (btrfs_scrub_cancel_dev()) waits on that wait queue
for dev->scrub_sctx to be NULL. So this means if btrfs_scrub_cancel_dev()
is called right when scrub is finishing and sees the wakeup before the
NULL is assigned to dev->scrub_sctx by btrfs_scrub_dev(), it hangs.
Fix this by doing the wakeup in btrfs_scrub_dev() after setting
dev->scrub_sctx to NULL.
Fixes: a2de733c78fa ("btrfs: scrub") Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 16 Oct 2026 11:55:55 +0000 (12:55 +0100)]
btrfs: scrub: allocate stripe buffer with GFP_KERNEL
In init_scrub_stripe() we are allocating a stripe's buffer with GFP_NOFS
but we can use GFP_KERNEL instead since we are not holding a transaction
handle or any locks that could cause a deadlock if GFP_KERNEL triggers a
transaction commit.
So switch to GFP_KERNEL. We are also doing GFP_KERNEL for the allocation
of stripe->csums just a few lines below.
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Thu, 15 Oct 2026 08:59:50 +0000 (19:29 +1030)]
btrfs: lzo: remove the special handling inside one fs block
Since commit 3eaf5f082c4c ("btrfs: extract inlined creation into a
dedicated delalloc helper"), even for inlined compressed extents the cb
range will be exact one block, aka [0, sector).
And several attempts to bring back the old compression range [0, i_size]
all failed with extra races against i_size.
So I think the behavior will stay. With that said, we can further
simplify the lzo compression path.
Since the range is always aligned to block size, and lzo compression is
already done block by block, there is no reason to handle sub-block
sized range.
We can safely remove the @sectorsize_mask, @sector_off and @in_len
variables, and always use @sectorsize as input length for each
compression iteration.
Finally add two ASSERT()s to make sure the range passed into
btrfs_compress_bio() is always block aligned.
Reviewed-by: Daniel Vacek <neelx@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Fri, 16 Oct 2026 00:48:17 +0000 (11:18 +1030)]
btrfs: rename btrfs_chunk_map::stripe_size
The member btrfs_chunk_map::stripe_size sounds very similar to
btrfs_chunk::stripe_len, but they have very different meanings.
btrfs_chunk_map::stripe_size is in fact the device extent size for the
chunk. E.g. a 2GiB RAID0 chunk with 2 stripes,
btrfs_chunk_map::stripe_size will be 1GiB (2GiB / 2).
But btrfs_chunk::stripe_len is the stripe length that striping profiles
should follow, and the value is now fixed to 64KiB.
E.g. for a 2GiB RAID0 chunk with 2 stripes, every write will be split
into 64KiB stripes, and then distributed across the 2 devices.
To avoid confusion, rename btrfs_chunk_map::stripe_size to
dev_extent_size.
And since we're here, also rename the helper btrfs_calc_stripe_length()
to btrfs_calc_dev_extent_size().
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 15 Oct 2026 17:45:45 +0000 (18:45 +0100)]
btrfs: scrub: fix lost error return value when scrubbing raid56 parity stripe
If we find errors we are returning the value of 'ret', which has a value
of 0 from some previous function call that succeeded, instead of returning
an -EIO error. Fix this by returning -EIO instead of 'ret'.
Fixes: fab0c0f03cfd ("btrfs: introduce BTRFS_PATH_AUTO_RELEASE() helper") Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 15 Oct 2026 14:55:55 +0000 (15:55 +0100)]
btrfs: fix fiemap merging compressed extents
Compressed extents are never supposed to be merged by the fiemap code and
reported as a single extent, as each extent's content comes from a
different compression stream and has its own header and context.
We currently rely on checking the logical and physical address adjacency
to prevent merging compressed extents but that is broken. Consider the
following reproducer which explains why the merging happens:
$ cat test.sh
#!/bin/bash
DEV=/dev/sdi
MNT=/mnt/sdi
mkfs.btrfs -s 4K -f $DEV
mount -o compress $DEV $MNT
# Do a write that compresses to a single 4K extent and represents the
# maximum compressed chunk size (128K).
xfs_io -f -c "pwrite 0 128K" $MNT/foo
# Sync to flush delalloc and allocate the 4K extent.
sync -f $MNT
# Now allocate a 124K extent in some other file, which will be adjacent
# (to the right) of the 4K extent previously allocated.
xfs_io -f -c "falloc 0 124K" $MNT/filler
# Do another write to the main file at its current EOF that is
# compressible and fits in a 4K extent.
xfs_io -c "pwrite 128K 128K" $MNT/foo
# Sync to flush delalloc and allocate the 4K extent.
sync -f $MNT
# Run fiemap which should report 2 encoded extents, each with a
# logical size of 128K.
xfs_io -c "fiemap -v" $MNT/foo
Fix this by never attempting to merge extents if the flags contain
FIEMAP_EXTENT_ENCODED.
Fixes: 4751832da990 ("btrfs: fiemap: Cache and merge fiemap extent before submit it to user") Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
The read itself eventually succeeds, and no read error is returned to
user space.
[CAUSE]
Since v7.2 btrfs has enabled large data folio support, now a folio can
contain multiple blocks even if the page size matches the fs block size.
And even before that, btrfs already has block size < page size support,
thus all involved code should assume a folio can contain multiple
blocks.
However the function btrfs_verify_folio() calls fsverity_verify_folio(),
which verifies the whole folio, even if the read range only covers part
of the folio.
This forces fsverity to verify garbage data that is not yet read from
disk, and causes verification failure.
This is a long-standing bug, but previously it only affected bs < ps
cases which is not that common.
Now with the large folio support, it's much easier to trigger with bs ==
ps cases.
[FIX]
Instead of calling fsverity_verify_folio(), call
fsverity_verify_blocks() to verify only the read range.
Reported-by: Roger Alasà <vonbloom@gmail.com> Link: https://lore.kernel.org/linux-btrfs/CAELqY9==Ter2XQPRp=gjW5T75T+7kcZCXpn5roF2L6tY6mQNxw@mail.gmail.com/ Fixes: 146054090b08 ("btrfs: initial fsverity support") Tested-by: Roger Alasà <vonbloom@gmail.com> Reviewed-by: Eric Biggers <ebiggers@kernel.org> Tested-by: Tito Duarte <duartito@gmail.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
guolingxing [Thu, 15 Oct 2026 09:59:01 +0000 (09:59 +0000)]
btrfs: verity: propagate orphan insertion errors when enabling verity
btrfs_begin_enable_verity() ignores the return value of
btrfs_orphan_add() and unconditionally returns 0. If orphan insertion
fails, btrfs_orphan_add() aborts the transaction, but the generic
fs-verity code still proceeds with Merkle tree construction because
the begin_enable_verity callback reported success.
Return the error so that fs-verity stops immediately instead of
continuing after the initialization failure.
The transaction is already aborted by btrfs_orphan_add() and the
handle is released by btrfs_end_transaction(), so no additional
cleanup is needed.
guolingxing [Thu, 15 Oct 2026 06:53:19 +0000 (06:53 +0000)]
btrfs: sysfs: restore NOFS context on early return in btrfs_sysfs_add_block_group_type()
btrfs_sysfs_add_block_group_type() sets up a NOFS context with
memalloc_nofs_save() because kobject_add() performs GFP_KERNEL
allocations deep in its call chain, and the function may be called
while holding a transaction handle or a lock required for a
transaction commit, where reclaim could deadlock.
Every other return path of this function calls memalloc_nofs_restore()
except the early return taken when another thread has already created
the kobject for the same raid index (this can happen when restriping).
Taking that path leaves PF_MEMALLOC_NOFS set on the calling task, so
all subsequent memory allocations performed by that task remain
restricted to the NOFS scope, which may lead to unexpected allocation
failures or deadlocks later on.
Restore the NOFS context before returning from that path.
Fixes: 49ea112da0e6 ("btrfs: do not create raid sysfs entries under any locks") CC: stable@vger.kernel.org Signed-off-by: guolingxing <guolingxing@tju.edu.cn> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Tue, 13 Oct 2026 16:04:00 +0000 (17:04 +0100)]
btrfs: defrag: prevent inode truncated to zero from not being defragged again
If we add an inode to autodefrag and the inode is truncated to an i_size
of zero before it's picked by autodefrag (run by the cleaner kthread),
we exit btrfs_run_defrag_inode() leaving the BTRFS_INODE_IN_DEFRAG flag
set in the inode, which prevents the inode from being auto defragged in
the future until it's evicted, loaded again and small writes happen to
it again.
Fix this by clearing BTRFS_INODE_IN_DEFRAG before we check if the current
offset is at or beyond the inode's i_size.
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Tue, 13 Oct 2026 05:41:20 +0000 (16:11 +1030)]
btrfs: cleanup dirty_cowonly_roots during btrfs_free_fs_roots()
[EXISTING ERROR HANDLING QUIRK]
Sashiko mentioned an existing problem that, if btrfs_abort_transaction()
is triggered during the following call site, some roots will be attached
to dirty_cowonly_roots without being cleaned up:
btrfs_commit_transaction()
|- commit_cowonly_roots()
|- list_move_tail();
| One root's dirty_list is moved from dirty_cowonly_roots to
| switch_commits.
| But there are still other dirty roots in dirty_cowonly_roots.
|
|- update_cowonly_root()
| |- btrfs_update_root()
| |- btrfs_abort_transaction()
|
|- return ret;
In that case, those dirty roots linked into dirty_cowonly_roots will
never be removed from that list.
As normally we require a transaction commit to move them, but since the
fs is already marked error, no new transaction can be started.
For roots that are already moved to switch_commits list, they are
cleaned up by btrfs_cleanup_one_transaction().
This is not going to cause memory leak on the btrfs_root structure
itself, as after a transaction is aborted, no one will read
dirty_cowonly_roots.
Furthermore dirty_cowonly_roots only tracks trees that are not
subvolume or log trees. They should not have root->dirty_log_pages.
And qgroup swapped blocks are going to be released during
btrfs_put_root().
Thus no memory leak from those two members either.
But we're still relying on a hidden behavior to keep us from leaking
memory.
[ENHANCEMENT]
Unlink all roots that are still linked to dirty_cowonly_roots during
btrfs_free_fs_roots(), and do one extra ASSERT() to make sure that their
dirty_log_pages are empty.
At that timing there is no other thread thus we're safe to unlink roots
without holding trans_lock.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Tue, 13 Oct 2026 05:41:19 +0000 (16:11 +1030)]
btrfs: cleanup qgroup swapped blocks at root release
Normally btrfs_qgroup_clean_swapped_blocks() is called when switching
the commit root of a subvolume tree.
However if a transaction is aborted before all dirty subvolumes are
linked to switch_commits, those left-over dirty subvolumes won't get
their qgroup swapped blocks cleaned up.
And since a transaction is already aborted, no new transaction is
allowed to be started, thus no way to cleanup the qgroup swapped blocks
for those subvolume trees.
Fix the problem by calling btrfs_qgroup_clean_swapped_blocks() inside
btrfs_put_root() when the last ref is dropped, so that at unmount time
those left-over qgroup swapped blocks are properly released.
This fix is inspired by the recent Sashiko reports about similar cases
on cowonly roots.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Tue, 13 Oct 2026 05:41:17 +0000 (16:11 +1030)]
btrfs: always remove quota_root from dirty_cowonly_roots
[ERROR HANDLING QUIRK]
When btrfs_quota_disable() failed at btrfs_clean_quota_tree() or
btrfs_del_root(), the quota_root->dirty_list is not removed from
fs_info->dirty_cowonly_roots, but the quota_root itself will be freed.
This leaves a dangling pointer in the dirty_cowonly_roots.
This is not a huge problem, as the transaction is aborted thus no one
will read the dirty_cowonly_roots for now.
Furthermore, quota_root should never be logged nor have qgroup swapped
tree blocks, thus it won't cause leaks from extent io tree nor qgroup
swapped blocks.
[ENHANCEMENT]
For the incoming cleanup of dirty_cowonly_roots during
btrfs_free_fs_roots(), any root that is still linked to
dirty_cowonly_roots will be read again, and this existing dangling
pointer is going to cause problem.
To prepare for that cleanup, move the list_del() call into the out
label, so that even if the transaction is aborted, we still unlink
quota_root from dirty_cowonly_roots before freeing it.
Assisted-by: LLM (exposing the bug during review) Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Adarsh Das [Wed, 7 Oct 2026 16:03:49 +0000 (21:33 +0530)]
btrfs: unwind device add when sysfs registration fails
btrfs_init_new_device() ignored the return value of
btrfs_sysfs_add_device(). When sysfs link creation failed (e.g. -EEXIST),
the add path continued, aborted the transaction with the same errno, and
tripped WARN_ON(btrfs_abort_should_print_stack()) while leaving the device
half-registered in memory.
Check the return value and unwind when registration fails, without
aborting the filesystem for errnos like -EEXIST. If add failed before
anything was published in sysfs, tear down the in-memory device state only
and do not call btrfs_sysfs_remove_device(). When the block-device link
was created but devid kobject registration fails, drop the link inside
btrfs_sysfs_add_device() before returning the error.
Reported-by: syzbot+3bf3e110b2d406b8166c@syzkaller.appspotmail.com Link: https://lore.kernel.org/all/6a791c94.01d0871a.3a0d52.0099.GAE@google.com Tested-by: syzbot+3bf3e110b2d406b8166c@syzkaller.appspotmail.com Assisted-by: LLM Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Adarsh Das <adarshdas950@gmail.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: add LOGICAL_INO_V2 flag for avoiding the tree modification log
At Meta, we use the LOGICAL_INO_V2 ioctl to estimate sharing-aware space
usage. We sample the logical address space and resolve each sample back
to the inodes which reference it. The measurement is statistical, so
with enough samples it gives good bounds on real usage.
Currently in the LOGICAL_INO_V2 ioctl path, iterate_extent_inodes()
(which can walk backrefs either against the current trees or against the
commit roots) acquires a tree modification log sequence number when
walking the current trees, since those trees are being modified while
the walk is in progress. The log is then enabled filesystem wide for as
long as any such walk is running, and is only disabled again once the
last walk has finished.
While it is enabled the cost impacts the whole filesystem, not only the
caller. Modifications of interior nodes of the extent tree and of
subvolume trees have to be recorded, serialized on a single filesystem
wide lock. Delayed refs with a sequence number at or above the oldest
active one can not be run (the run path bails out with -EAGAIN in
btrfs_run_delayed_refs_for_head()), and metadata refs above it can not
be merged, so they accumulate and that work is deferred to transaction
commit.
Currently, userspace has no way to ask for anything cheaper. Whenever a
transaction is running, the walk uses the tree modification log, even if
the caller would tolerate results that could be up to one commit interval
stale.
Give userspace this option by adding BTRFS_LOGICAL_INO_ARGS_COMMIT_ROOT,
which makes the backref walk use the commit roots and therefore never
enables the tree modification log. Behavior is unchanged unless the flag
gets explicitly passed by userspace. iterate_extent_inodes() already
takes a search_commit_root argument for this, which scrub, send and the
data relocation warning path all pass as true. The flag simply plumbs it
through to userspace.
Hengyu Liang [Sun, 11 Oct 2026 15:58:15 +0000 (11:58 -0400)]
btrfs: fix chattr failing on nodatasum files that have the compress flag
Commit cf1afec09e9f ("btrfs: validate properties before setting them")
made btrfs_set_prop() validate a property before setting it. For the
compression property, the validation fails with -EINVAL when the inode
has NODATASUM.
However, chattr writes back all the flags it read from the file, and
btrfs_fileattr_set() sets the compression property whenever the new
flags include FS_COMPR_FL, even when the compression flags do not
change. A file created on a filesystem mounted with nodatasum has
NODATASUM and still inherits the compress flag from its directory. As
of now, chattr fails on such a file when it changes any other flag.
Before commit cf1afec09e9f ("btrfs: validate properties before setting
them"), the last command succeeds.
After that commit, it fails with EINVAL.
This patch will make btrfs_fileattr_set() leave the compression flags
and the compression property alone when the call changes neither
FS_COMPR_FL nor FS_NOCOMP_FL.
Fixes: cf1afec09e9f ("btrfs: validate properties before setting them") CC: stable@vger.kernel.org # 7.2+ Suggested-by: Qu Wenruo <wqu@suse.com> Assisted-by: LLM Signed-off-by: Hengyu Liang <hengyul@cs.unc.edu> Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Linus Torvalds [Fri, 30 Oct 2026 18:27:10 +0000 (08:27 -1000)]
Merge tag 'spi-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Pull spi fixes from Mark Brown:
"Two patches here, one a straightforward device quirk and the other a
fix for a potentially nasty data corruption bug in the zynq-qspi
driver"
* tag 'spi-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
spi: cs42l43: Add speaker ID quirk for Dell XPS 13 DX13260 SKU 0E54
spi: zynq-qspi: fix address buffer corruption and support 4-byte address
Linus Torvalds [Fri, 30 Oct 2026 18:24:02 +0000 (08:24 -1000)]
Merge tag 'regulator-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator
Pull regulator fix from Mark Brown:
"One fix here, we had an issue with deferred hardware disables where
if we were deferring while suspend was happening we would let the
workqueue run which could result in us trying to access hardware after
it or the buses used to access it were suspended.
We fix this by completing any work as we suspend the individual
regulators"
* tag 'regulator-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: irq: freeze deferred hardware access during suspend
Linus Torvalds [Fri, 30 Oct 2026 18:20:58 +0000 (08:20 -1000)]
Merge tag 'regmap-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap
Pull regmap fix from Mark Brown:
"One fix for a real issue that fortunately doesn't affect any in tree
configurations"
* tag 'regmap-fix-v7.4-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
regmap: irq: fix incorrect index calculation for read_sub_irq_data()
Linus Torvalds [Fri, 30 Oct 2026 17:45:41 +0000 (07:45 -1000)]
Merge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull more SCSI updates from James Bottomley:
"A couple of late arriving driver updates (lpfc, ufs) with the main
addition being the lock context analysis update where clang static
analysis coupled with annotations are used to verify driver guarantees
about locks and warn on compile if they're inconsistent. This is
mostly annotations with only a few code changes (except in scsi_debug
which needed a fairly big rework to validate with locking context
analysis)"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (103 commits)
scsi: ufs: core: Do not release a reserved request in ufshcd_abort()
scsi: ufs: core: Remove completed-without-interrupt path in ufshcd_abort()
scsi: ufs: core: Fix double release in ufshcd_eh_timed_out()
scsi: ufs: core: Fix double completion in ufshcd_clear_lu_cmds()
scsi: ufs: core: Add a helper for clearing hba->outstanding_reqs
scsi: ufs: core: Remove an always-false WARN_ONCE()
scsi: fnic: Update module authors
scsi: ufs: qcom: Power down UFS during system suspend on Milos
scsi: MAINTAINERS: Update Marvell Ethernet storage maintainers
scsi: handle out-of-range SG_SET_TIMEOUT values
scsi: core: Update cdrom.h coments
scsi: st: MAINTAINERS: Add Laurence Oberman and John Meneghini
scsi: sd: Mark the disk dead if runtime resume fails in sd_remove()
scsi: libiscsi_tcp: Prevent R2T range check overflow
scsi: ufs: core: Fix stale comment in ufshcd_mcq_config_nr_queues()
scsi: fnic: Annotate fnic_nvme functions with __must_hold()
scsi: ufs: core: Fix a deadlock between clock scaling and device commands
scsi: ufs: core: Clear stale UIC result byte before peer DME retries
scsi: ufs: bsg: Reject UIC commands with a non-zero result byte
scsi: ufs: bsg: Validate request_len before accessing request
...
Linus Torvalds [Fri, 30 Oct 2026 16:33:23 +0000 (06:33 -1000)]
Merge tag 'block-7.4-20261029' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux
Pull more block updates from Jens Axboe:
"Bigger than usual for an end-of-merge-window pull request, main reason
being that the MD side I had missed a pull on, and both NVMe and block
got a good pile of LLM driven fixes where some of them may as well
just go in now rather than wait a full cycle. This contains:
- NVMe fixes via Keith:
- nvme-tcp fallout from the parallel queue allocation: initialize
pf_cache_lock and the header digest before requests are set up,
and don't dereference a NULL queue->ctrl on reconnect
- nvme-fabrics queue mapping fixes: reset io_queues on reconnect,
limit each queue type to the online CPUs, and clear a poll map
with no poll queues
- nvme-tcp io_work priority inversion with send_mutex held
(Seiji), error recovery on read_sock failure and skipping TLS
control records while waiting for the ICResp (Martin B), TLS
handshake timeout validation, and an address attribute race
- nvme-rdma NULL metadata_sgl dereference with T10-PI
- nvme-pci early boot hang with the quirks parameter
- nvme-multipath ANA log bounds underflow, and an nvme-fc
discovery reference leak
- nvmet hardening: rdma inline SGL out-of-bounds write and short
connect requests, fc use-after-free in LS create connection and
teardown on admin queue failure, pci-epf transport error
reporting and SGL length checks, zns NUMD wrap, fc stale read
data, auth state and status fixes, ports list and qid_max
locking, rdma double puts and irq-safe locking, and a double
free on metadata SGL allocation failure
- nvmet fixes from Josef: big discard length truncation, NULL
derefs for PCI endpoint targets and debugfs host_traddr, stop
listening when a tcp port is removed, and don't BUG() in
nvme-loop when a scan runs beside a reset
- MD fixes via Yu Kuai:
- raid5 validation: reject too few disks, invalid reshape disks, a
per-device size below one chunk, and an inconsistent reshape
state from the superblock instead of BUG(), plus journal payload
size checks
- raid5 error path and lifetime fixes: extra_page leak and NULL
derefs in free_conf/setup_conf, draining released_stripes before
destroying the cache, RCU for the big_stripe_tree lookup, and
per-array io_unit slab names
- md core: reject v0.90 arrays with too many devices, validate
bblog_size, take reconfig_mutex when enabling PPL, fix the
mddev->private NULL deref when stopping an array under IO, and
don't hand out the array before mddev->kobj is added
- raid10 missing barrier wakeup, raid0 device count validation, a
bogus mddev_unlock() in lbs_store(), and shared helpers
- blk-mq fixes from Josef:
- don't trust the bio cookie in blk_mq_poll()
- keep tag sharing fair after a queue update
- don't bring back flags of a completed request from the timeout
handler
- honor a restarted timer
- initialize the dma state for P2P
- quiesce a queue joining a quiesced tag set
- set up and exit hctxs under tag_list_lock
- free the tag bitmaps in the SRCU callback
- blk-cgroup and friends from Josef:
- keep the old blkg while reassociating a bio
- free struct ioc after a grace period
- don't truncate v1 iops limits
- fix root io.stat on 32-bit
- don't free a blkg with its punt work queued
- flush iolatency enable_work after going offline
- pass dead group I/O counts on and make reset_stats reset them
- don't enter reclaim under rq_qos_mutex
- Other core fixes from Josef:
- a disk use-after-free in sync_bdevs()
- don't losing the status of a failed flush request
- 64-bit sector math for mac partitions
- apply the segment size to the integrity buffer
- blk-iocost: abort a weight transfer on invalid donation weights,
and publish hweights as one atomic value
- Fix a NULL deref and use-after-free in __blk_crypto_evict_key(),
reachable via dm-inlinecrypt
- Fix recursive i_rwsem locking in blkdev_write_iter(), from two
fixes for the same race landing in the window
- Advance the iterator after copy-in for SG_DXFER_TO_FROM_DEV, fixing
an infinite loop in blk_rq_map_user_iov()
- Check the open mode for LOOP_SET_STATUS in the compat ioctl
- rnbd-srv use-after-free and kobject fixes
- drbd timer and req_lock fixes
- zloop cleanups and error path fixes
- cdrom header and kernel-doc cleanups"
* tag 'block-7.4-20261029' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux: (109 commits)
blk-crypto: fix NULL deref and UAF in __blk_crypto_evict_key()
zloop: check that the zone backing file support ->{read,write}_iter
zloop: fput file after failed zone init
zloop: share more code for opening a zone backing file
zloop: remove a duplicate call to zloop_get_block_size
block: fix kernel-doc warnings in blk-map.c and blk-mq-dma.c
blk-iocost: publish hweights as one atomic value
blk-iocost: abort weight transfer on invalid donation weights
rnbd-srv: publish the sess_dev at the end of OPEN
rnbd-srv: fix use-after-free on two CLOSE messages
block/rnbd-srv: drop the sess_dev kobject unlocked
block/rnbd-srv: don't free the device on a sysfs error
drbd: take req_lock to skip the initial sync as peer
drbd: stop the start_resync timer of a dying device
block: apply the segment size to the integrity buffer
blk-mq: free the bitmaps of tags in the SRCU callback
block: partitions/mac: compute the sectors in 64 bits
blk-mq: set up a queue under tag_list_lock
blk-mq: exit the hctxs of a queue under tag_list_lock
blk-mq: quiesce a queue that joins a quiesced tag set
...
Linus Torvalds [Fri, 30 Oct 2026 16:29:43 +0000 (06:29 -1000)]
Merge tag 'io_uring-7.4-20261029' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux
Pull io_uring fixes from Jens Axboe:
- A batch of fixes from Chris:
- Charge registered buffer arrays, rsrc nodes, tctx nodes and
personalities to memcg, and free the XArray nodes a failed store
leaves behind.
- Clear bpf_ctx while submitting SQEs from a bpf-ops program, as the
socket code otherwise skips both the capability check and
lock_sock() for a setsockopt from the SQ.
- Free the task context after an RCU grace period, closing a UAF
when a foreign task frees the last request.
- Stop io_ring_exit_work() from waiting on an SQPOLL thread held in
get_signal(), and drop sqd->lock while the thread waits for
in-flight requests, as either one could pin every iou_exit slot
- io-wq fixes:
- Don't create new workers once the exit has begun, which could
leave an exiting worker in the thread group under exec
- save the worker type before dropping its reference (KASAN UAF)
- Defer io-wq enqueues from a foreign task to task_work, as an
IOPOLL reap by another task could race io_wq_put_and_exit()
- Keep the drain active when the deferred request being started is
itself a drain request, and don't attach a new ring to an SQPOLL
thread that is already canceling
- Fix the multishot recv length limit accounting, and count cancelled
timeouts in cq_timeouts so an armed off timeout doesn't fire early
- Advance the buffer ring head when a pickup empties it, and keep the
recvmsg buffer select length within MAX_RW_COUNT
- Mask the socket type in the bpf filter population so SOCK_CLOEXEC or
SOCK_NONBLOCK can't bypass a type filter, and enforce the per-task
register restrictions for blind registration
- Complete a non-blocking connect() that fails with -EAGAIN rather than
retrying it through poll and io-wq, matching connect(2)
- napi list traversal cleanup, and remove a useless min_not_zero()
* tag 'io_uring-7.4-20261029' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux: (21 commits)
io_uring/sqpoll: don't attach to a thread that is canceling
io_uring/net: fix multishot recv length limit accounting
io_uring/timeout: count cancelled timeouts in cq_timeouts
io_uring/napi: clean up list traversal with list_for_each_entry_safe()
io_uring: defer foreign io-wq enqueues to task_work
io_uring/io-wq: save worker type before dropping its reference
io_uring/cancel: drop sqd->lock while SQPOLL waits
io_uring/sqpoll: don't wait for the thread to exit
io_uring: free the task context after a grace period
io_uring/bpf-ops: clear bpf_ctx while submitting SQEs
io_uring: free XArray nodes left by a failed store
io_uring: charge tctx nodes and personalities
io_uring/rsrc: charge buffer arrays and nodes to memcg
io_uring/io-wq: no new workers once the exit has begun
io_uring: keep the drain active when a deferred drain request is started
io_uring/net: don't retry a non-blocking connect() that fails with -EAGAIN
io_uring/register: enforce per-task restrictions in io_uring_register_blind()
io_uring/net: mask socket type with SOCK_TYPE_MASK in io_socket_bpf_populate()
io_uring/net: Remove useless min_not_zero() usage
io_uring/net: keep recvmsg buffer select length within MAX_RW_COUNT
...
Linus Torvalds [Fri, 30 Oct 2026 16:18:43 +0000 (06:18 -1000)]
Merge tag 'clk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux
Pull clk updates from Stephen Boyd:
"Here's the clk changes for the merge window which is dominated in the
diffstat by Qualcomm churning out new SoCs that need clk drivers.
Looking beyond the sheer amount of lines of mostly data there we have
a clk nexus DT binding change in the core clk framework and a driver
core patch to add a subsys driver macro.
The clock nexus support is useful for devices that map clks internally
or for connectors on a board that route clock signals to another
board. I suspect the use of this will expand in the future, so we have
unit tests to make sure it doesn't break.
The driver core change adds subsys_driver() and
subsys_platform_driver() so we can register early drivers without
open-coding that all (module_platform_driver() is at device initcall
level so it isn't be used).
Beyond the core framework bits we have lots of changes all over the
tree to initialize 'struct clk_init_data'. The potential risk here is
that someone forgets to initialize all the members of that structure
and either the clk flags cause weird behavior later on or the parents
are pointing at junk and the clk is orphaned forever. The patches just
initialize the thing with zeroes to be safe. Longer term we should
probably move this struct out of 'struct clk_hw' and pass it during
clk registration time. That's a much bigger change and would require
restructuring entire drivers that rely on the pointer being there.
Finally, we have the usual fixes like calculation updates for
dividers, handling error paths during clk registration, or fixing clk
data now that drivers are using them. Those patches are dwarfed by the
amount of new SoC support and only affect those SoCs and the consumers
of those clks. Overall I'm more concerned the clk_init_data patches
are going to expose some latent bugs that we didn't notice before
because something was relying on clks being orphaned and now they're
not.
Core:
- subsys_driver() / subsys_platform_driver() macros
- Clock nexus DT binding support (with KUnit tests)
- Spread-spectrum configuration validation
- Fully initialize struct clk_init_data across many drivers
New Drivers:
- Anlogic DR1V90 CRU
- ZTE zx297520v3 top, matrix, and LSP clocks
- Samsung Exynos5515 and ACPM driver extended to Exynos850.
- Qualcomm Eliza camera, camera-BIST, GPU, and video clks
- Qualcomm Hawi / Maili camera and GPU controllers
- Qualcomm Kuno global, RPMh
- Qualcomm MSM8952 global, RPM
- Qualcomm Nord camera/video clocks, EMAC interconnect, PHY resets
- Qualcomm IPQ5210 / IPQ9650 network clocks
- i.MX93 350 MHz fractional-N GP-PLL for display
- i.MX pll14xx spread-spectrum clock generation
- Amlogic A9 SCMI clock controller
- Allwinner single-divider "update bit" support
- EN7523 Serdes port selection"
* tag 'clk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux: (316 commits)
clk: imx: pll14xx: support spread spectrum clock generation
clk: cdce925: Size the PLL child name buffer for any int
clk: stm32: fix clock array size for STM32MP21
dt-bindings: clock: Convert HiSilicon clocks to DT schema
clk: lmk04832: Handle reset GPIO lookup errors
dt-bindings: clock: anlogic,dr1v90-cru: make external clocks optional
clk: tegra: bpmp: remove kcalloc
clk: mmp2-audio: add missing PM_CLK dependency
clk: zte: Introduce a driver for zx297520v3 LSP clocks
clk: zte: Introduce a driver for zx297520v3 matrix clocks
clk: zte: Introduce a driver for zx297520v3 top clocks
clk: zte: Add zx PLL support infrastructure
clk: zte: Add regmap-based clocks
clk: zte: Add Clock registration infrastructure
dt-bindings: clk: zte: Add zx297520v3 LSP clock and reset controller
dt-bindings: clk: zte: Add zx297520v3 matrix clock and reset controller
dt-bindings: clk: zte: Add zx297520v3 top clock and reset controller
clk: nuvoton: ma35d1: Add missing WDT/WWDT parent clocks
dt-bindings: clock: ma35d1: Add missing WDT/WWDT parent clocks
dt-bindings: clock: ma35d1: Drop CLK_MAX_IDX define
...
Jens Axboe [Fri, 30 Oct 2026 13:27:28 +0000 (07:27 -0600)]
io_uring/sqpoll: don't attach to a thread that is canceling
io_sq_offload_create() refuses to attach a ring to an SQPOLL thread
that has already gone, by checking sqd->thread under sqd->lock. Since
the thread drops the lock while io_uring_cancel_generic() waits for its
requests, that check can run while the thread is on its way out and
still see it. The new ring is then added to ctx_list, io_uring_setup()
succeeds, the thread finishes canceling and exits without ever having
looked at the ring, and io_uring_enter() fails with -EOWNERDEAD.
Only a thread that exits on a fatal signal can be caught like this, as
the stop bit is set once the last ring has let go of the sqd, when no fd
is left to attach through. A fatal signal is group wide and attaching
needs the same thread group, so the attacher is on its way out as well.
Still, make the check see it: set the stop bit before canceling,
whatever the reason for the exit, and test it when attaching.
Reported-by: Karl Mehltretter <kmehltretter@gmail.com> Fixes: cb0306c52ffa ("io_uring/cancel: drop sqd->lock while SQPOLL waits") Signed-off-by: Jens Axboe <axboe@kernel.dk>
Linus Torvalds [Fri, 30 Oct 2026 00:00:23 +0000 (14:00 -1000)]
Merge tag 'fscrypt-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/linux
Pull fscrypt fixes from Eric Biggers:
- Add more checks to prevent encryption on "large block size" ext4
filesystems (s_blocksize > PAGE_SIZE), which isn't supported yet.
There was already a mount-time check, but it wasn't enough.
- Add missing memory barrier to fscrypt_find_mode_key()
- Add missing zeroization in an error path
* tag 'fscrypt-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/linux:
ext4: reject encrypted inodes when block size > page size
ext4: refuse encryption when block size > page size
fscrypt: Fix missing barrier in fscrypt_find_mode_key()
fscrypt: Explicitly zeroize sw_secret in error path
Linus Torvalds [Thu, 29 Oct 2026 23:31:48 +0000 (13:31 -1000)]
Merge tag 'ntfs3_for_7.4' of https://github.com/Paragon-Software-Group/linux-ntfs3
Pull ntfs3 updates from Konstantin Komarov:
"The bulk of this cycle is hardening against corrupted or maliciously
crafted volumes. Most of these fixes come from fuzzing and KASAN
reports. The main areas are:
- $LogFile replay, where several unchecked on-disk sizes could make
mount write past the MFT record buffer
- MFT record and index parsing
- extended attributes, where a bad $EA entry could leak up to 64K of
kernel heap to userspace through getxattr()
Reading a damaged compressed file could also return uninitialized page
cache contents to userspace; that is fixed too.
A few fixes affect healthy volumes as well: a heap overflow in
fallocate(INSERT_RANGE) on small files, a use-after-free when a
WOF-compressed file is opened for writing during a concurrent read, a
lock-order deadlock in readdir, and stale disk data showing up in
newly allocated clusters of compressed files.
On the feature side there are new tracepoints covering mount, log
replay, lookup, directory index, allocation, iomap and file I/O. There
is also a new 'ugm' mount option (on by default): with ugm=0, newly
created files don't get the WSL uid/gid/mode extended attributes,
which Windows doesn't use.
The rest is smaller correctness fixes and cleanups"
* tag 'ntfs3_for_7.4' of https://github.com/Paragon-Software-Group/linux-ntfs3: (52 commits)
ntfs3: add file I/O tracepoints
ntfs3: add iomap tracepoints
ntfs3: add allocation tracepoints
ntfs3: add directory index tracepoints
ntfs3: add create inode tracepoint
ntfs3: add namei tracepoints
ntfs3: add mount and log replay tracepoints
fs/ntfs3: validate the EA value length in ntfs_read_ea()
fs/ntfs3: handle partial allocation in attr_data_get_block_locked()
ntfs3: fix mapping pairs bounds check for alignment
fs/ntfs3: bound DeleteAttribute asize against rec->used in do_action
ntfs3: release ni_lock before dir_emit in ntfs_readdir to fix deadlocks
fs/ntfs3: bound CreateAttribute asize against record_size in do_action
fs/ntfs3: do not mark frame pages uptodate on read error
fs/ntfs3: prevent positive E_NTFS_NONRESIDENT from reaching writeback error handling
fs/ntfs3: fix ntfs_fnd leak when compare function is missing
fs/ntfs3: fix stale attr_b in attr_allocate_frame()
fs/ntfs3: reject unaligned hole punching on compressed files
fs/ntfs3: take ni_lock() once in ntfs_readdir()
fs/ntfs3: reserve a compression unit in maxbytes_sparse
...
Linus Torvalds [Thu, 29 Oct 2026 23:24:49 +0000 (13:24 -1000)]
Merge tag 'kgdb-7.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux
Pull kgdb updates from Daniel Thompson:
"Varied updates this cycle.
Naveen Kumar Chaudhary fixed a memory leak when the environment is
already full, eliminated to weird historical cruft (CMD_BUFLEN having
different values in different files) and removed some dead code.
Fang Xieyan tackled a family of related buffer overflows in the gdb
remote protocol parsing (great fixes from a defensive coding point of
view but not security relavent; gdb can already do anything it wants
to)"
* tag 'kgdb-7.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux:
kgdb: Fix buffer overflow in the qRcmd packet handler
kgdb: Fix buffer overflow in the 'M' and 'X' packet handlers
kgdb: Fix buffer overflow in the 'm' packet handler
kgdb: remove unreachable return in kgdb_reenter_check()
kdb: unify CMD_BUFLEN definition into kdb_private.h
kdb: fix memory leak in kdb_setenv() when environment is full
* tag 'hyperv-next-signed-20261028' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
Drivers: hv: Remove leftover support for old Hyper-V versions
mshv: Remove unused *_old memory deposit functions
mshv: Redesign hypervisor memory deposit logic
mshv: Rename memory deposit memory functions to _old
mshv_vtl: reject poll files with multiple waitqueues
mshv_vtl: detach poll waiter on POLLFREE
mshv_vtl: initialize poll file mutex before device registration
Documentation: hyperv: Update information on guest CPU offlining
Documentation: hyperv: Remove support for guests on Hyper-V 2012/2012R2
Drivers: hv: Run VMBus kexec cleanup from syscore shutdown
kexec: Add helper to report an active kexec reboot
tools: hv: remove unused rte_smp_mb()
Documentation: hyperv: Update documentation for sending VMBus packets
Drivers: hv: vmbus: Check ACPI companion before use
Documentation: hyperv: Describe validation of the ring buffer indices
Drivers: hv: vmbus: Keep the ring byte counts sane for a bad index
Drivers: hv: vmbus: Compute ring byte counts from a caller-held snapshot
Drivers: hv: vmbus: Use *_ONCE() to access shared ring buffer indices
Drivers: hv: vmbus: Bounds check the shared ring buffer indices
Linus Torvalds [Thu, 29 Oct 2026 22:33:25 +0000 (12:33 -1000)]
Merge tag 'riscv-for-linus-7.4-mw1' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux
Pull RISC-V updates from Paul Walmsley:
- Add support for crash hotplug
- Improve user copy performance on cores with fast misaligned accesses
- Optimize code paths involving pgtable_l{4,5}_enabled() by patching in
the return values during early boot (since these don't change during
runtime), or by specifying the return values directly in Kconfig
files (for purpose-built kernels)
- Change the PCIe bus reassignment behavior to depend on the
PCI_REASSIGN_ALL_BUS PCI flag (as ARM64 does)
- Avoid an unnecessary sstatus CSR read during context switch
- Make the sbi_ecall() macro variadic, improving code clarity
- Refactor __switch_to_vector() to prepare for upcoming vector context
switch optimizations
- Add a kselftest for vectorized user copy
- Add bounds checking on the shadow stack restore token address to
avoid it being used as an attack vector
- Fix the shadow stack and indirect branch landing pad control
functions to only update the CPU's envcfg CSR when operating on the
current task
- For MIPS RISC-V systems, add support for early boot-applied errata;
and for P8700 RISC-V systems, avoid WFI usage in the kernel (since it
is buggy)
- Add the "B" RISC-V ISA extension to hwcap and hwprobe reporting
- Add support for specifying and reporting the "Zic64b" RISC-V ISA
extension, and add it to the SpacemiT K1, K3, and Sophgo SG2044 DT
files
- Require the block size of cache management operations to be specified
in DT schema validation for the RISC-V ISA extensions Zicbom, Zicbop,
and Zicboz
- Some miscellaneous fixes: fix an incorrect bitfield width in the
RVG_SYSTEM_CSR_MASK bitmask macro; make RISCV_ALTERNATIVE mandatory,
now that XIP kernel support has been removed; and remove some
unnecessary includes in kernel/smp.c
* tag 'riscv-for-linus-7.4-mw1' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: (28 commits)
riscv: mm: unexport _pgtable_l4_enabled and _pgtable_l5_enabled
riscv: introduce RISCV_ISA_SV48 and RISCV_ISA_SV57
riscv: support early isa ext and use it to optimize pgtable_l4|l5_enabled
riscv: convert pgtable_l4|l5_enabled to inline function
riscv: remove RISCV_ALTERNATIVE Kconfig option
riscv: Add a getter for user PMLEN support
riscv: dts: sophgo: sg2044: Add Zic64b ISA extension
riscv: dts: spacemit: k1: Add Zic64b ISA extension
riscv: dts: spacemit: k3: Add Zic64b ISA extension
riscv: Add Zic64b to cpufeature and hwprobe
dt-bindings: riscv: Add Zic64b extension description
dt-bindings: riscv: Require block-size for Zicbom, Zicbop, and Zicboz
riscv: Add B to hwcap and hwprobe
riscv: remove unused profile.h includes
riscv: cfi: Only write envcfg CSR when task is current
riscv: cfi: Bounds check restore token before swap
riscv: Add ERRATA_MIPS_P8700_WFI to replace WFI with mips.pause
selftest: riscv: test vectorized user copy
riscv: vector: refactor context switch for kernel-mode vector
riscv: do not read csr in vector context switch
...
Linus Torvalds [Thu, 29 Oct 2026 19:40:12 +0000 (09:40 -1000)]
Merge tag 'rust-fixes-7.4' of git://git.kernel.org/pub/scm/linux/kernel/git/ojeda/linux
Pull Rust fixes from Miguel Ojeda:
"Toolchain and infrastructure:
- Fix 'objtool' warnings for the upcoming Rust 1.101.0 (expected
2026-12-24) by adding one more 'noreturn' function
'kernel' crate:
- Fix KCFI failures on every call from C to Rust IRQ handlers and
'hrtimer' callbacks by giving the C enum bindings the correct
'cfi_encoding'. Also prevent the same issue for DMA callbacks
taking 'dma_data_direction'
- 'workqueue' module: fix a runtime warning when queueing work on the
deprecated 'system_wq' by switching to 'system_{percpu,dfl}_wq'"
* tag 'rust-fixes-7.4' of git://git.kernel.org/pub/scm/linux/kernel/git/ojeda/linux:
rust: dma: declare cfi_encoding for dma_data_direction
rust: hrtimer: declare cfi_encoding for hrtimer_restart
rust: irq: declare cfi_encoding for irqreturn
objtool/rust: add one more `noreturn` Rust function for Rust 1.101.0
rust: workqueue: replace deprecated system_wq with system_{percpu,dfl}_wq
- Revert advertising the TCP MSS from the configured MTU
- netlink: check sk_rcvbuf before charging the dump skb
- eth:
- ti: am65-cpts: fix support for PTP_EXTTS_REQUEST2 ioctl
- mlx4: always create the eth and ib auxiliary devices
- tg3: invert restart quirk logic to restrict SNP power-down bypass
- bnxt: recycle agg buffers when Rx head buffer allocation fails
Previous releases - always broken:
- net: fix frag_list leak in skb_segment_list() error path
- sctp: reject forged cookie peer_addr with unknown address family
- vxlan: keep reserved_bits across changelink
- net: limit priority VLAN untag depth
- inet: frags: prevent acceptance of partially authenticated reassembly
- gro: do not GRO GSO packets with inconsistent payload
- tcp: avoid double destroy in tcp_abort() of an unaccepted child"
* tag 'net-7.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (137 commits)
tun: drop batched packets while the device is down
selftests: net: netlink-dumps: start a dump with a notification queued
netlink: check sk_rcvbuf before charging the dump skb
net/sched: act_ife: do not consume the inner header on egress
openvswitch: use eth_type_trans_careful() in internal_dev_recv()
net: use eth_type_trans_careful() in __dev_forward_skb2()
loopback: use eth_type_trans_careful()
net: add eth_type_trans_careful()
tcp: initialize mapped children before route lookup
tcp: fix reordering addition overflow in MTU probing
tcp: avoid double destroy in tcp_abort() of an unaccepted child
net: bgmac: fix TX DMA unmapping and slot cleanup
gro: do not GRO GSO packets with inconsistent payload
net/packet: check TX_RING fragments before memory accounting
eth: bnxt: hand XDP frags over to the copybreak skb
eth: bnxt: free XDP frags when xdp_do_redirect() fails
eth: bnxt: don't leak buffers when XDP agg buffer allocation fails
eth: bnxt: resize Rx head buffers when rx-copybreak changes
eth: bnxt: recycle agg buffers when Rx head buffer allocation fails
netdev-genl: ignore NETDEV_UNREGISTER rebroadcasts
...
Linus Torvalds [Thu, 29 Oct 2026 18:10:23 +0000 (08:10 -1000)]
Merge tag 'bpf-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Pull bpf fixes from Alexei Starovoitov:
- Remove BPF_F_ARENA_SCALAR flag and .arena.data section in libbpf that
were added for rust-bpf in this merge window. We found a cleaner
approach that does it on compiler side instead of kernel/libbpf. No
released kernel has them (Alexei Starovoitov)
- Fix bpf_send_signal_task() returning -EPERM when it is called from a
kernel thread (Aditya Sharma)
- Various fixes (Alexei Starovoitov):
- Use emit_rsb_call() to call bpf_unwind_target() in x86 JIT to fix
objtool warning during the build
- Fix scalar evolution: clear delta of a widened register, don't
widen a pointer to a buffer, don't clamp a register whose type
changed in the loop, and reset the range of a clamped packet
pointer
- Finish the insn in the state that maybe_fork_scalars() splits
off, so that pruning doesn't drop it at the head of a counted
loop
- Preserve the registers that carry stack arguments across
arch_bpf_timed_may_goto() on x86-64, arm64 and riscv64
- Don't re-sort subprog_info[] for every new subprogram
- Fix flaky map_kptr and rcu_reentry selftests
- Use memmove() in bpf_skb_load_bytes(), bpf_skb_load_bytes_relative()
and bpf_xdp_load_bytes(), since the destination can be a slice of the
same packet (Ayushman Rout)
- Fixes and follow-ups for scalar evolution: reject pointer operands in
32-bit recurrences and latches, don't intersect with the current
register range on clamp to avoid a false 'verifier bug', avoid false
infinite loop reports, and recognize a 32-bit copy of the induction
variable (Eduard Zingerman)
- Fix out-of-bounds write to btf->types[] for split BTF and propagate
btf_add_type() errors in btf_check_all_metas() (Feng Yang)
- Fix use-after-free in bpf_timer callback that fires after its inner
map was freed (Mani Tofigh)
- Fix x86 JIT stopping before jumps converge: size backward jumps by
the insns they jump over (Mykyta Yatsenko)
- Fix use-after-free of struct_ops trampoline: dissociate struct_ops
progs before the grace period of the map (Qiliang Yuan)
- Validate stack argument reads against the caller state to fix
out-of-bounds read in the verifier, and reject callbacks with stack
arguments (Weiming Shi)
- Tie borrowed iterator results and pointers loaded from trusted fields
to the lifetime of their source, so that iterator next, destroy and
reference release invalidate them (Xu Yunxiang)
* tag 'bpf-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf: (48 commits)
bpf: Don't re-sort subprog_info[] for every new subprogram
bpf: Fix out-of-bounds btf->types write for split BTF
selftests/bpf: Test stack arguments across may_goto
bpf: Keep stack arguments in arch_bpf_timed_may_goto()
selftests/bpf: Test maybe_fork_scalars() at the head of a counted loop
bpf: Finish the insn in the state that maybe_fork_scalars() splits off
selftests/bpf: Test the clamp of pointers in scalar evolution
bpf: Reset the range of a packet pointer that scalar evolution clamps
bpf: Don't clamp a register whose type changed in the loop
selftests/bpf: Test a loop that moves a pointer to a buffer
bpf: Don't widen a pointer to a buffer in scalar evolution
selftests/bpf: Avoid a page fault in rcu_reentry while the handler is armed
selftests/bpf: Wait for the last maps of map_kptr to be freed
selftests/bpf: Test that a widened register is not linked
bpf: Clear delta of a register that scalar evolution widens
bpf, x86: Remove r0_on_stack from fetching atomics in arena
bpf: Remove BPF_F_ARENA_SCALAR
selftests/bpf: Remove tests for arena access through numbers
libbpf: Keep format strings of bpf_printk() in .rodata
libbpf: Remove .arena.data
...
Linus Torvalds [Thu, 29 Oct 2026 17:49:07 +0000 (07:49 -1000)]
Merge tag 'kthread-fix-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks
Pull kthread fix from Frederic Weisbecker:
"Fix a bug where dynamic CPU isolation through cpuset isolated
partitions can move a kthread out of the housekeeping range.
The fix makes sure the fallback affinity is always a set of online
housekeeper"
* tag 'kthread-fix-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks:
kthread: Fall back when no preferred housekeeping CPU is online
Linus Torvalds [Thu, 29 Oct 2026 17:38:53 +0000 (07:38 -1000)]
Merge tag 'sched_ext-for-7.4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext
Pull sched_ext fixes from Tejun Heo:
"This is a big pull for fixes, but much of it was in the works before
the merge window.
Most of these fixes are in sub-scheduler support, which nests BPF
schedulers under a parent that hands them CPUs. It was heavily
reworked over recent cycles and is now functionally complete. Porting
real schedulers on it drove the latest additions and keeps finding
corner cases.
For example, once CPUs are shared and can be taken back, the kernel
must keep every task moving when its scheduler can't run it or is
failing, and those fallbacks interact subtly with each other and with
other new core features.
- The fallback paths meant to keep a task moving when its scheduler
can't run it or is failing could themselves leave it stuck
- The stall watchdog could blame a scheduler for time it wasn't
responsible for, even taking down all of sched_ext
- Sub-schedulers were told late, or not at all, which CPUs they could
use
- Deferred callbacks could stay queued or be dropped under core
scheduling and proxy execution, with a sched/core fix to the
duplicate check
- Hardening: tracing programs can no longer call kfuncs that could
deadlock the CPU"
* tag 'sched_ext-for-7.4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext: (34 commits)
sched_ext: Reject scx_bpf_sub_grant/revoke() in tracing programs
sched_ext: Fix the rescue overload pick ejecting the wrong sub under bypass
sched_ext: Fix use-after-free of a NONE rescuee leaving its sched while running
sched_ext: Fix a lone IMMED task waiting behind a kept prev
sched_ext: Fix queued bypassed tasks stalling when the host loses SCX_CAP_BASE
sched_ext: Let the bypass load balancer move tasks homed on other cpus
sched_ext: Factor out bypass_lb_move() and insert with the donor DSQ unlocked
sched_ext: Fix the bypass load balancer stranding tasks on cpus where the host lacks SCX_CAP_BASE
sched_ext: Fix bypassed tasks stalling when the host lacks SCX_CAP_BASE
sched_ext: Expire prev's slice in dispatch_one() under bypass
sched_ext: Fix a cpu idling after put_prev_task_scx() queues its task on the local DSQ
sched_ext: Refuse extending a kernel slice
sched_ext: Mark the rescue grant as a kernel slice with SCX_TASK_KERNEL_SLICE
sched_ext: Fix a cpu idling with a runnable rescuee
sched_ext: Restore a rescuee's slice after ops.stopping()
sched_ext: Fix rescue served time disagreeing with the slice under proxy execution
sched_ext: Move refill_task_slice_dfl() below apply_slice_vtime()
sched_ext: Fix bypassed tasks looping on a host without SCX_CAP_ENQ
sched_ext: Fix bypassed tasks stranded by the reenqueue limit
sched_ext: Clear holding_cpu where the claim is confirmed
...
Linus Torvalds [Thu, 29 Oct 2026 17:35:19 +0000 (07:35 -1000)]
Merge tag 'cgroup-for-7.4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup fixes from Tejun Heo:
- cgroup.kill did not kill a process whose thread group leader had
exited while other threads still ran, nor those threads once moved to
another cgroup, letting a process survive its container being killed
- A cgroup.procs write under favordynmods could migrate a newly forked
process before fork linked it to its cgroup, corrupting cgroup task
lists
- cpuset partition fixes for nested hierarchies: invalidation, revival,
shrinking and switches between root and isolated could leak isolated
CPUs, skip or miscompute the housekeeping check, or leave wrong
effective CPUs:
- A root partition can no longer be created under an isolated
parent or stay valid under one, as disabling it could leave no
housekeeping CPUs
- Enabling a partition failed when a grandchild held tasks,
although the parent had none of its own
- Separately, an allocation racing with a cpuset.mems update could
drop its mempolicy nodemask and land on the wrong node, and a
memoryless node in cpuset.mems triggered a spurious "movable
nodes only" warning
- Two no-functional-change cleanups of the effective CPU
computation
* tag 'cgroup-for-7.4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (27 commits)
cgroup/cpuset: Integrate compute_partition_effective_cpumask() into compute_effective_cpumask()
cgroup/cpuset: Relocate compute_*effective_cpumask() and compute_effective_nodemask()
cgroup/cpuset: Release invalid isolated CPUs with a shrunk root partition parent
cgroup/cpuset: Release invalid isolated CPUs in compute_partition_effective_cpumask()
cgroup/cpuset: Make sure compute_partition_effective_cpumask() is only called for valid partition
cgroup/cpuset: Prevent recursive call of remote_cpus_update() via update_cpumasks_hier()
selftests/cgroup: Check for a single range nohz_full CPUs before running test matrix
cgroup/cpuset: Minor fixes to prstate_housekeeping_conflict()
cgroup/cpuset: Fix incorrect effective_cpus bug in update_cpumasks_hier()
cgroup/cpuset: Properly compute effective xcpus in update_parent_effective_cpumask()
cgroup/cpuset: Find furthest isolated ancestor for invalidated isolated partition
cgroup: don't kill migrated processes through their exiting threads
cgroup/cpuset: check effective mems for a movable-only configuration
cgroup/cpuset: Fix task checking when enabling a partition
cgroup: serialize migrations of newly forked processes
selftests/cgroup: test cgroup.kill with an exited thread group leader
cgroup: fix cgroup.kill for exited thread group leader
selftests/cgroup: Skip test_cpuset_prs.sh test if some required CPUs are offline
selftests/cgroup: Add tests for nohz_full housekeeping check
selftests/cgroup: Add tests for isolcpus housekeeping check
...
==================================================================
blk-crypto: loop0: no support for wrapped keys
Oops: general protection fault, probably for non-canonical address
0xdffffc000000000b: 0000 [#1] SMP KASAN PTI
KASAN: null-ptr-deref in range [0x0000000000000058-0x000000000000005f]
CPU: 5 UID: 0 PID: 901 Comm: dmsetup 7.3.0-rc3 #1 PREEMPT(lazy)
RIP: 0010:__blk_crypto_evict_key+0x2e/0x3b0
Call Trace:
blk_crypto_evict_key+0x24c/0x320
inlinecrypt_dtr+0xf6/0x160
inlinecrypt_ctr+0x2d3/0xbd0
dm_table_add_target+0x661/0xf40
table_load+0x1c4/0x490
ctl_ioctl+0x32f/0x5f0
dm_ctl_ioctl+0xe/0x20
==================================================================
dm-inlinecrypt is the only caller that can evict a key that was never
started, as fscrypt frees such a key without evicting it.
Fix this by returning early from blk_crypto_fallback_evict_key() when
tfms_inited[] is not set for the key's mode. tfms_inited[] is only set
after blk_crypto_fallback_init() has successfully initialized the
profile, so the crash above is no longer reachable.
The bug has been latent since commit c984ff1423ae ("blk-crypto:
dynamically allocate fallback profile") was merged in v6.5, and only
became actually triggerable when commit e7f57d2c47e2 ("dm-inlinecrypt:
add target for inline block device encryption") was merged in v7.2.
Xiao Liang [Tue, 27 Oct 2026 10:18:27 +0000 (18:18 +0800)]
tun: drop batched packets while the device is down
tun_sendmsg() handles TUN_MSG_PTR batches from vhost-net by calling
tun_xdp_one() instead of tun_get_user(). But this path does not check
IFF_UP, allowing packets from a guest to be injected into the host
receive path while the device is administratively down.
Check IFF_UP in tun_xdp_one() before running XDP or building an skb.
Drop the packet when the device is down, matching tun_get_user().
Fixes: 043d222f93ab ("tuntap: accept an array of XDP buffs through sendmsg()") Signed-off-by: Xiao Liang <shaw.leon@gmail.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20261027101830.85327-1-shaw.leon@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
netlink: fix a dump livelock with a notification queued
A netlink dump that starts while a notification is still charged to a
socket with a small receive buffer fails with ENOBUFS, and every retry
from recvmsg() fails the same way. A reader that peeks before reading,
as libnl does, never consumes the message at the head of the queue and
spins forever. I hit this with "iw dev wlan0 scan" on an arm64 board.
Patch 1 goes back to checking sk_rcvbuf before charging the dump skb,
as before commit ae8f160e7eb2. Patch 2 adds a selftest; it fails
without patch 1 and is skipped without the privileges to create a
netns.
====================
selftests: net: netlink-dumps: start a dump with a notification queued
Add a test which starts a link dump on a socket with SO_RCVBUF 2048
while a link notification is still queued, after a large read made the
kernel attempt large dump skbs. It reads like libnl does, peeking
before each read, and expects the dump to complete with a zero status
and at least one link. The notifications come from bringing lo up and
down, so the test runs in its own netns and is skipped without the
privileges to create one.
netlink: check sk_rcvbuf before charging the dump skb
Since commit ae8f160e7eb2 ("netlink: Fix wraparounds of
sk->sk_rmem_alloc."), netlink_dump() charges the dump skb first and
queues it only if sk_rmem_alloc is still below sk_rcvbuf. Commit a215b5723922 ("netlink: make sure we allow at least one dump skb")
added an exception for an empty socket.
A dump that starts while something else is still charged to a socket
with a small receive buffer fails with ENOBUFS and stays running.
Every recvmsg() retries it, fails again and sets sk_err, so a reader
that peeks before reading, as libnl does, gets ENOBUFS on the read
after each peek, never consumes the message at the head of the queue
and spins forever.
I hit this with "iw dev wlan0 scan" hanging at 100% CPU on an arm64
board with a 6.6 based kernel (iw 5.19 sets SO_RCVBUF to 8KiB). A
kprobe trace showed the scan dump starting while the NEW_SCAN_RESULTS
notification iw had just read was still charged to its socket,
because iw had preempted the sender before netlink_broadcast()
returned.
Check the limit before charging, as before ae8f160e7eb2, which also
covers the empty socket. netlink_recvmsg() continues a dump only at
or below half of sk_rcvbuf, so the skb is then always queued, and a
dump goes over the limit by at most one skb. Refuse a skb that would
take sk_rmem_alloc past INT_MAX, so the wraparound stays fixed.
Reproduced on net in a KVM guest with a userspace reproducer and the
selftest in the next patch; both pass with this fix.
====================
net: fix eth_type_trans() crashes on skbs coming from the stack
eth_type_trans() assumes ETH_HLEN bytes are present in skb->head.
NIC drivers always provide this, and they should keep using the fast
version, without an extra pskb_may_pull().
Devices looping skbs from their transmit path back to their receive
path get skbs built by the stack instead. tc actions (act_vlan pop_eth,
act_ife decode), BPF helpers (bpf_skb_change_head() before a redirect)
or OVS actions (pop_eth followed by pop_mpls/pop_nsh) can leave fewer
than ETH_HLEN bytes in the linear part, and eth_type_trans() then hits
the BUG() in __skb_pull().
Instead of fixing every producer, or adding a check to all
eth_type_trans() users, fix the few consumers exposed to this:
- Patch 1 adds eth_type_trans_careful(), which pulls the Ethernet
header first and returns a drop reason on failure.
- Patch 2 uses it in loopback_xmit().
- Patch 3 uses it in __dev_forward_skb2(), used by BPF redirects to
ingress and by veth, macvlan, ipvlan, netdevsim and l2tp_eth.
- Patch 4 uses it in OVS internal_dev_recv(), covering both the MPLS
(TEB) and NSH pop cases at once (only the MPLS case was tested).
- Patch 5 (from Jamal) fixes act_ife decode on egress, which was
consuming the inner Ethernet header. This is no longer a crash after
patch 2, but egress decode never delivered a correct frame.
Other eth_type_trans() users already pull the header or never see
such skbs:
- vrf_process_v{4,6}_outbound() already pull ETH_HLEN + IP header
before vrf_local_xmit().
- tun already uses pskb_may_pull_reason(skb, ETH_HLEN).
- XDP: bpf_xdp_adjust_head() and bpf_xdp_adjust_tail() keep at least
ETH_HLEN bytes in the linear part, and multi-buffer tail shrinking
only trims frags.
Tested under virtme-ng with AF_PACKET reproducers. Without the series,
all four cases crash in eth_type_trans():
- lo egress "action vlan pop_eth", eth/eth/8192 bytes frame
- lo egress "action ife decode pipe", IFE frame with 8192 bytes payload
- lo egress "action vlan pop_eth continue", then a BPF program doing
bpf_skb_change_head(skb, 1, 0) + bpf_redirect(veth1, BPF_F_INGRESS)
- OVS flow "pop_eth,pop_mpls(0x6558),0" to the internal port,
eth/mpls/eth/4096 bytes frame
With the series, no crash. An ICMP echo inside an IFE frame decoded
on lo egress is now received and answered. Normal traffic on lo,
veth (including 8000 byte pings) and vrf is unchanged.
Jamal Hadi Salim [Wed, 28 Oct 2026 12:27:03 +0000 (13:27 +0100)]
net/sched: act_ife: do not consume the inner header on egress
The issue:
tcf_ife_decode() unconditionally calls eth_type_trans(), which pulls
ETH_HLEN. On egress this consumes the inner Ethernet header, and the
device receives an skb whose data points to the L3 header:
- loopback_xmit() parses L3 data as Ethernet: the "ethertype" comes
from the inner IPv4 source address, and the decapsulated packet is
dropped or misparsed.
- a real NIC transmits the frame without its Ethernet header.
(The __skb_pull() BUG this used to trigger in loopback_xmit() when the
remaining linear area was short is handled by "loopback: use
eth_type_trans_careful()".)
The Fix:
Parse the inner Ethernet header with eth_type_trans() in both directions,
then restore the egress data position and MAC length. On ingress the
existing behavior is unchanged. On egress, push the inner Ethernet header
back after parsing so the xmit path sees the original frame and later
actions see a consistent network header.
Conditions to recreate the bug:
- egress tc filter on lo: action ife decode pipe;
- AF_PACKET SOCK_RAW frame: outer Ethernet with ETH_P_IFE, IFE metadata,
inner Ethernet + a valid IPv4 ICMP echo request to a local address;
- before this patch, no echo is received; after it, the echo is
delivered and answered.
Fixes: ef6980b6becb ("introduce IFE action") Cc: stable@vger.kernel.org Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
[edumazet: reworded changelog, the crash is now handled in loopback] Signed-off-by: Eric Dumazet <edumazet@kernel.org> Acked-by: Alexei Starovoitov <ast@kernel.org> Link: https://patch.msgid.link/20261028122703.2093730-6-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet [Wed, 28 Oct 2026 12:27:02 +0000 (13:27 +0100)]
openvswitch: use eth_type_trans_careful() in internal_dev_recv()
OVS actions like pop_eth followed by pop_mpls(ETH_P_TEB) or pop_nsh
can expose an inner Ethernet header which is not in the linear part
of the skb. Output to an internal port then crashes in eth_type_trans():
flow: in_port(0),eth(),eth_type(0x8847),mpls(lse=0x00010140)
actions: pop_eth,pop_mpls(0x6558),0
# AF_PACKET SOCK_RAW on dp0: eth / mpls / eth / 4096 bytes of payload
Use eth_type_trans_careful(), and drop the packet if the Ethernet
header can not be pulled. This covers all OVS actions at the consumer,
instead of fixing each producer.
Fixes: b2d0f5d5dc53 ("openvswitch: enable NSH support") Fixes: f66b53fdbb22 ("openvswitch: New MPLS actions for layer 2 tunnelling") Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Aaron Conole <aconole@redhat.com> Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com> Tested-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20261028122703.2093730-5-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet [Wed, 28 Oct 2026 12:27:01 +0000 (13:27 +0100)]
net: use eth_type_trans_careful() in __dev_forward_skb2()
A tc egress BPF program can push a tiny link-layer header with
bpf_skb_change_head() on an skb whose head is almost empty, then
redirect it to an Ethernet device ingress. __bpf_redirect_common()
only checks mac_header < network_header, and __dev_forward_skb2()
crashes in eth_type_trans():
tc filter add dev lo egress pref 1 matchall action vlan pop_eth continue
tc filter add dev lo egress pref 2 bpf da obj prog.o
# prog: bpf_skb_change_head(skb, 1, 0);
# return bpf_redirect(veth1_ifindex, BPF_F_INGRESS);
# AF_PACKET SOCK_RAW on lo: eth / eth / 8192 bytes of payload
Use eth_type_trans_careful(), and drop the packet if the Ethernet
header can not be pulled. This also covers other dev_forward_skb()
users (veth, macvlan, ipvlan, netdevsim, l2tp_eth).
Update the empty_skb BPF selftest: at lwt_xmit, BPF_PROG_TEST_RUN
strips the Ethernet header, so the "veth ETH_HLEN+1" packet is a 1-byte
skb. Redirecting it to veth ingress used to read the Ethernet header
past skb->tail and deliver a 1-byte frame; it is now dropped with
NET_RX_DROP, as veth_xmit() already does at egress.
Fixes: 3a0af8fd61f9 ("bpf: BPF for lightweight tunnel infrastructure") Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com> Tested-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20261028122703.2093730-4-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet [Wed, 28 Oct 2026 12:26:59 +0000 (13:26 +0100)]
net: add eth_type_trans_careful()
eth_type_trans() assumes ETH_HLEN bytes are present in skb->head.
NIC drivers always provide this, and they should keep using the fast
version.
Devices looping skbs from their transmit path back to their receive
path get skbs built by the stack instead. Their link-layer header
might have been popped by tc actions (act_vlan pop_eth, act_ife decode)
or OVS actions (pop_eth + pop_mpls / pop_nsh), leaving fewer than
ETH_HLEN bytes in the linear part. eth_type_trans() then hits
the BUG() in __skb_pull().
Add eth_type_trans_careful(), which first pulls the Ethernet header,
and returns a drop reason on failure, so that the few devices exposed
to this can drop such packets.
Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Aaron Conole <aconole@redhat.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com> Tested-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Link: https://patch.msgid.link/20261028122703.2093730-2-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kyle Zeng [Wed, 28 Oct 2026 15:41:01 +0000 (16:41 +0100)]
tcp: initialize mapped children before route lookup
IPv4 passive opens on a dual-stack listener go through
tcp_v4_syn_recv_sock() with tcp_v6_mapped_child_init() as the optional
initializer. The callback currently runs after the output route lookup.
If that lookup fails, tcp_done() destroys a clone whose pinet6 and
ipv6_fl_list still refer to the listener's state.
With the flowlabel list head in inet_sock, fl6_free_socklist() walks
the clone's list and RCU-frees its nodes without unlinking them from the
live listener. Subsequent flowlabel operations can use or free the nodes
again. Commit 858d2a4f67ff ("tcp: fix potential race in
tcp_v6_syn_recv_sock()") moved the initializer before ehash insertion,
but did not cover this error path.
Invoke the existing initializer after the IPv4 fields it consumes have
been set, but before route lookup can fail. It needs neither the
destination nor sk_setup_caps(), and all later disposal paths remain
covered.
Fixes: 1c17f4373d4d ("ipv6: Move ipv6_fl_list from ipv6_pinfo to inet_sock.") Signed-off-by: Kyle Zeng <kylebot@openai.com> Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com> Link: https://patch.msgid.link/20261028154101.2153360-1-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kyle Zeng [Wed, 28 Oct 2026 16:02:26 +0000 (17:02 +0100)]
tcp: fix reordering addition overflow in MTU probing
Commit efb8763d7bbb ("net: ipv4: bound TCP reordering sysctl writes and
MTU probe sizes") widened size_needed and the MSS multiplication, but
tp->reordering + 1 is still evaluated as u32. RTAX_REORDERING accepts
an unrestricted u32, and tcp_init_metrics() can install U32_MAX.
In that case the extra-data requirement becomes zero. A write queue
with probe_size - 1 data bytes and a FIN satisfies the sequence-space
check. The coalescing walk can pass the queue head to
tcp_skb_can_collapse(), and tcp_clone_payload() can produce an skb
whose advertised payload exceeds the data in its fragments. Local
delivery followed by TCP_ZEROCOPY_RECEIVE can then walk beyond the
fragment array. KASAN reports a slab-out-of-bounds read in
can_map_frag().
Cast reordering before incrementing it so the arithmetic handles the
entire u32 metric range. For U32_MAX, the required size exceeds the
32-bit queue and window limits, so the existing checks skip the probe.
The calculation is unchanged for all other reordering values.
Kyle Zeng [Wed, 28 Oct 2026 16:08:47 +0000 (17:08 +0100)]
tcp: avoid double destroy in tcp_abort() of an unaccepted child
tcp_abort() checks for TCP_CLOSE while only owning the socket lock,
then grabs the BH lock before calling tcp_done_with_error().
For a child still sitting in a listener accept queue (which can be
reached via SOCK_DESTROY or bpf_sock_destroy()), inet_csk_listen_stop()
can run in this window: it only holds the child BH lock, and calls
inet_child_forget(), which disconnects, orphans and destroys the child.
tcp_abort() then calls tcp_done_with_error() -> tcp_done(), which sees
SOCK_DEAD and calls inet_csk_destroy_sock() a second time, leading to
a double sock_put() and a use-after-free.
Recheck TCP_CLOSE after acquiring the BH lock, unless we just closed
a listener ourselves, in which case tcp_done_with_error() must still
run to report the error.
Note that the WARN_ON(sock_owned_by_user(child)) in
inet_csk_listen_stop() can still trigger in this window.
Fixes: bac76cf89816 ("tcp: fix forever orphan socket caused by tcp_abort") Signed-off-by: Kyle Zeng <kylebot@openai.com>
[edumazet: recheck TCP_CLOSE under the BH lock instead of moving the test, smaller diff] Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com> Link: https://patch.msgid.link/20261028160847.2170959-1-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rosen Penev [Wed, 28 Oct 2026 16:48:39 +0000 (09:48 -0700)]
net: bgmac: fix TX DMA unmapping and slot cleanup
When mapping a fragment fails, bgmac_dma_tx_add() unwinds with the
wrong slots. It passes the failed fragment's slot->dma_addr, which
holds the mapping error, to dma_unmap_single() for the head. The
fragment loop indexes from ring->end instead of ring->end + 1, so it
unmaps the head a second time as a page and leaves the last mapped
fragment mapped. None of the affected slots get their dma_addr cleared
either, which includes the head slot when mapping the head fails.
bgmac_dma_tx_ring_free() unmaps every slot with a non-zero dma_addr,
so tearing down the ring unmaps these addresses again. It also frees
the skbs of pending TX slots but leaves slot->skb and slot->dma_addr
set. The slots survive a down/up cycle, and bgmac_dma_tx_add() only
sets slot->skb on the last slot of a packet. If a slot with a stale
pointer is later used for the head or a fragment of a scatter/gather
packet, bgmac_dma_tx_free() frees the old skb a second time when that
slot completes.
bgmac_dma_tx_ring_free() also picks dma_unmap_single() or
dma_unmap_page() by looking at slot->skb, which is set on the last slot
of a packet, while the head was mapped with dma_map_single() and the
fragments with skb_frag_dma_map().
In the error path, unmap the head and each mapped fragment from their
own slots, and clear dma_addr on every slot the failed packet touched.
In bgmac_dma_tx_ring_free(), choose the unmap function by the SOF flag
of the descriptor like bgmac_dma_tx_free() does, unmap every slot
before freeing any skb, since a packet can wrap around the end of the
ring with its skb on the last slot, and clear both fields.
This was found through code inspection by an LLM while hitting other
bugs on an Asus RT-N16. It did not hit these ones specifically.
Kyle Zeng [Wed, 28 Oct 2026 17:13:50 +0000 (18:13 +0100)]
gro: do not GRO GSO packets with inconsistent payload
tcp_gro_receive() excludes empty non-GSO packets with the unsigned
len - 1 check, but its GSO branch only checks that gso_size matches
the held packet's MSS. A trusted TCP GSO skb can reach GRO again with
no payload after TC pedit increases its TCP data offset.
If the skb is fully linear, skb_gro_receive() then enters its fragment
transfer path with nr_frags == 0. The do-while loop copies skb_frag_t
entries backwards beyond both fragment arrays. An unprivileged process
using its own user and network namespaces can trigger a KASAN
slab-out-of-bounds write and panic through veth, pedit and GRE.
More generally, the payload of such a packet can be shrunk to any
value, making it inconsistent with its gso_segs, and a GSO packet
with gso_segs == 0 contributes no segment to the GRO packet.
Do not aggregate GSO packets with gso_segs == 0, and require the
payload of GSO packets to be consistent with gso_segs and gso_size,
the last segment being allowed to be short. This rejects the empty
skb before either TCP merge helper, and covers both IPv4 and IPv6.
Fixes: 5eddb24901ee ("gro: add support of (hw)gro packets to gro stack") Signed-off-by: Kyle Zeng <kylebot@openai.com>
[edumazet: check payload against gso_segs, refuse gso_segs == 0] Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20261028171350.2205175-1-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kyle Zeng [Wed, 28 Oct 2026 18:14:50 +0000 (19:14 +0100)]
net/packet: check TX_RING fragments before memory accounting
tpacket_fill_skb() adds the full non-linear payload length to
skb->truesize and sk_wmem_alloc before checking MAX_SKB_FRAGS. With
PACKET_VNET_HDR, userspace can submit a ring-sized frame even when it
greatly exceeds the device MTU.
A near-INT_MAX frame can saturate sk_wmem_alloc before the fragment
loop rejects it. A subsequent refcount_add(0x40000001) can temporarily
turn REFCOUNT_SATURATED into 1. A concurrent TX completion can then
drop that apparent last reference and free the still-live packet
socket before refcount_warn_saturate() restores the counter.
Check that the remaining payload fits in MAX_SKB_FRAGS pages before
accounting it. Include the initial page offset in the calculation,
and move the existing -EFAULT handling out of the fragment loop.
This preserves the accepted packet sizes and error code while
preventing an invalid frame from injecting an oversized byte charge.
Fixes: 69e3c75f4d54 ("net: TX_RING and packet mmap") Cc: stable@vger.kernel.org Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com> Signed-off-by: Kyle Zeng <kylebot@openai.com>
[edumazet: use net_warn_ratelimited() instead of pr_err(), add a Fixes: tag] Signed-off-by: Eric Dumazet <edumazet@kernel.org> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20261028181450.2221013-1-edumazet@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
eth: bnxt: fix Rx buffer leaks under OOM
Commit bd3beb5c8691 ("eth: bnxt: disable rx-copybreak by default")
changed the defaults putting users on buggy(ier) driver paths.
Fix various ways bnxt may leak buffers under OOM. The first patch
fixes the real bug that was hit in production. Second patch fixes
an issue I hit while testing the first one. The next 3 patches
fix other Rx buffer leaks pointed out by an LLM during review.
====================