]> git.ipfire.org Git - thirdparty/linux.git/commit
kernfs: Relax constraint in draining guard
authorMichal Koutný <mkoutny@suse.com>
Mon, 5 May 2025 12:12:00 +0000 (14:12 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 21 May 2025 12:23:13 +0000 (14:23 +0200)
commit071d8e4c2a3b0999a9b822e2eb8854784a350f8a
treea8afbc6f47b0fc196a7063c8dbea5956cb0d5ef3
parent282582820ed7f60addd15223674fd6b99251e233
kernfs: Relax constraint in draining guard

The active reference lifecycle provides the break/unbreak mechanism but
the active reference is not truly active after unbreak -- callers don't
use it afterwards but it's important for proper pairing of kn->active
counting. Assuming this mechanism is in place, the WARN check in
kernfs_should_drain_open_files() is too sensitive -- it may transiently
catch those (rightful) callers between
kernfs_unbreak_active_protection() and kernfs_put_active() as found out by Chen
Ridong:

kernfs_remove_by_name_ns kernfs_get_active // active=1
__kernfs_remove   // active=0x80000002
kernfs_drain ...
wait_event
//waiting (active == 0x80000001)
kernfs_break_active_protection
// active = 0x80000001
// continue
kernfs_unbreak_active_protection
// active = 0x80000002
...
kernfs_should_drain_open_files
// warning occurs
kernfs_put_active

To avoid the false positives (mind panic_on_warn) remove the check altogether.
(This is meant as quick fix, I think active reference break/unbreak may be
simplified with larger rework.)

Fixes: bdb2fd7fc56e1 ("kernfs: Skip kernfs_drain_open_files() more aggressively")
Link: https://lore.kernel.org/r/kmmrseckjctb4gxcx2rdminrjnq2b4ipf7562nvfd432ld5v5m@2byj5eedkb2o/
Cc: Chen Ridong <chenridong@huawei.com>
Signed-off-by: Michal Koutný <mkoutny@suse.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20250505121201.879823-1-mkoutny@suse.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/kernfs/dir.c
fs/kernfs/file.c