fsnotify_recalc_mask() fails to handle the return value of
__fsnotify_recalc_mask(), which may return an inode pointer that needs
to be released via fsnotify_drop_object() when the connector's HAS_IREF
flag transitions from set to cleared.
This manifests as a hung task with the following call trace:
INFO: task umount:1234 blocked for more than 120 seconds.
Call Trace:
__schedule
schedule
fsnotify_sb_delete
generic_shutdown_super
kill_anon_super
cleanup_mnt
task_work_run
do_exit
do_group_exit
The race window that triggers the iref leak:
Thread A (adding mark) Thread B (removing mark)
────────────────────── ────────────────────────
fsnotify_add_mark_locked():
fsnotify_add_mark_list():
spin_lock(conn->lock)
add mark_B(evictable) to list
spin_unlock(conn->lock)
return
/* ---- gap: no lock held ---- */
fsnotify_detach_mark(mark_A):
spin_lock(mark_A->lock)
clear ATTACHED flag on mark_A
spin_unlock(mark_A->lock)
fsnotify_put_mark(mark_A)
fsnotify_recalc_mask():
spin_lock(conn->lock)
__fsnotify_recalc_mask():
/* mark_A skipped: ATTACHED cleared */
/* only mark_B(evictable) remains */
want_iref = false
has_iref = true /* not yet cleared */
-> HAS_IREF transitions true -> false
-> returns inode pointer
spin_unlock(conn->lock)
/* BUG: return value discarded!
* iput() and fsnotify_put_sb_watched_objects()
* are never called */
Fix this by deferring the transition true -> false of HAS_IREF flag from
fsnotify_recalc_mask() (Thread A) to fsnotify_put_mark() (thread B).
Fixes: c3638b5b13 ("fsnotify: allow adding an inode mark without pinning inode")
Signed-off-by: Xin Yin <yinxin.x@bytedance.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Link: https://patch.msgid.link/CAOQ4uxiPsbHb0o5voUKyPFMvBsDkG914FYDcs4C5UpBMNm0Vcg@mail.gmail.com
Signed-off-by: Jan Kara <jack@suse.cz>
This was done entirely with mindless brute force, using
git grep -l '\<k[vmz]*alloc_objs*(.*, GFP_KERNEL)' |
xargs sed -i 's/\(alloc_objs*(.*\), GFP_KERNEL)/\1)/'
to convert the new alloc_obj() users that had a simple GFP_KERNEL
argument to just drop that argument.
Note that due to the extreme simplicity of the scripting, any slightly
more complex cases spread over multiple lines would not be triggered:
they definitely exist, but this covers the vast bulk of the cases, and
the resulting diff is also then easier to check automatically.
For the same reason the 'flex' versions will be done as a separate
conversion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
Instead of iterating all inodes belonging to a superblock to find inode
marks and remove them on umount, iterate all inode connectors for the
superblock. This may be substantially faster since there are generally
much less inodes with fsnotify marks than all inodes. It also removes
one use of sb->s_inodes list which we strive to ultimately remove.
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Introduce a linked list tracking all inode connectors for a superblock.
We will use this list when the superblock is getting shutdown to
properly clean up all the inode marks instead of relying on scanning all
inodes in the superblock which can get rather slow.
Suggested-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
system_unbound_wq should be the default workqueue so as not to enforce
locality constraints for random work whenever it's not required.
Adding system_dfl_wq to encourage its use when unbound work should be used.
The old system_unbound_wq will be kept for a few release cycles.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://lore.kernel.org/20250916082906.77439-2-marco.crivellari@suse.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Ensure the superblock is kept alive until we're done with iput().
Holding a reference to an inode is not allowed unless we ensure the
superblock stays alive, which fsnotify does by keeping the
watched_objects count elevated, so iput() must happen before the
watched_objects decrement.
This can lead to a UAF of something like sb->s_fs_info in tmpfs, but the
UAF is hard to hit because race orderings that oops are more likely, thanks
to the CHECK_DATA_CORRUPTION() block in generic_shutdown_super().
Also, ensure that fsnotify_put_sb_watched_objects() doesn't call
fsnotify_sb_watched_objects() on a superblock that may have already been
freed, which would cause a UAF read of sb->s_fsnotify_info.
Cc: stable@kernel.org
Fixes: d2f277e26f ("fsnotify: rename fsnotify_{get,put}_sb_connectors()")
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Jan Kara <jack@suse.cz>
In some setups directories can have many (usually negative) dentries.
Hence __fsnotify_update_child_dentry_flags() function can take a
significant amount of time. Since the bulk of this function happens
under inode->i_lock this causes a significant contention on the lock
when we remove the watch from the directory as the
__fsnotify_update_child_dentry_flags() call from fsnotify_recalc_mask()
races with __fsnotify_update_child_dentry_flags() calls from
__fsnotify_parent() happening on children. This can lead upto softlockup
reports reported by users.
Fix the problem by calling fsnotify_update_children_dentry_flags() to
set PARENT_WATCHED flags only when parent starts watching children.
When parent stops watching children, clear false positive PARENT_WATCHED
flags lazily in __fsnotify_parent() for each accessed child.
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Commit e43de7f086 ("fsnotify: optimize the case of no marks of any type")
optimized the case where there are no fsnotify watchers on any of the
filesystem's objects.
It is quite common for a system to have a single local filesystem and
it is quite common for the system to have some inotify watches on some
config files or directories, so the optimization of no marks at all is
often not in effect.
Permission event watchers, which require high priority group are more
rare, so optimizing the case of no marks og high priority groups can
improve performance for more systems, especially for performance
sensitive io workloads.
Count per-sb watched objects by high priority groups and use that the
optimize out the call to __fsnotify_parent() and fsnotify() in fsnotify
permission hooks.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20240317184154.1200192-11-amir73il@gmail.com>
Define a container struct fsnotify_sb_info to hold per-sb state,
including the reference to sb marks connector.
Allocate the fsnotify_sb_info state before attaching connector to any
object on the sb and free it only when killing sb.
This state is going to be used for storing per priority watched objects
counters.
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20240317184154.1200192-8-amir73il@gmail.com>
We would like to count watched objects by priority group, so we will need
to update the watched object counter after adding/removing marks.
Create a helper fsnotify_update_sb_watchers() and call it after
attaching/detaching a mark, instead of fsnotify_{get,put}_sb_watchers()
only after attaching/detaching a connector.
Soon, we will use this helper to count watched objects by the highest
watching priority group.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20240317184154.1200192-7-amir73il@gmail.com>
Instead of passing fsnotify_connp_t, pass the pointer to the marked
object.
Store the object pointer in the connector and move the definition of
fsnotify_connp_t to internal fsnotify subsystem API, so it is no longer
used by fsnotify backends.
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20240317184154.1200192-6-amir73il@gmail.com>
Instead of counting the number of connectors in an sb, we would like
to count the number of watched objects per priority group.
As a start, create an accessor fsnotify_sb_watched_objects() to
s_fsnotify_connectors and rename the fsnotify_{get,put}_sb_connectors()
helpers to fsnotify_{get,put}_sb_watchers() to better describes the
counter.
Increment the counter at the end of fsnotify_attach_connector_to_object()
if connector was attached instead of decrementing it on race to connect.
This is fine, because fsnotify_delete_sb() cannot be running in parallel
to fsnotify_attach_connector_to_object() which requires a reference to
a filesystem object.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20240317184154.1200192-2-amir73il@gmail.com>
Some filesystems like fuse and nfs have zero or non-unique fsid.
We would like to avoid reporting ambiguous fsid in events, so we need
to avoid marking objects with same fsid and different sb.
To make this easier to enforce, store the fsid in the marks of the group
instead of in the shared conenctor.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20231130165619.3386452-2-amir73il@gmail.com>
fsnotify_add_mark() and variants implicitly take a reference on inode
when attaching a mark to an inode.
Make that behavior opt-out with the mark flag FSNOTIFY_MARK_FLAG_NO_IREF.
Instead of taking the inode reference when attaching connector to inode
and dropping the inode reference when detaching connector from inode,
take the inode reference on attach of the first mark that wants to hold
an inode reference and drop the inode reference on detach of the last
mark that wants to hold an inode reference.
Backends can "upgrade" an existing mark to take an inode reference, but
cannot "downgrade" a mark with inode reference to release the refernce.
This leaves the choice to the backend whether or not to pin the inode
when adding an inode mark.
This is intended to be used when adding a mark with ignored mask that is
used for optimization in cases where group can afford getting unneeded
events and reinstate the mark with ignored mask when inode is accessed
again after being evicted.
Link: https://lore.kernel.org/r/20220422120327.3459282-12-amir73il@gmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Create helpers to take and release the group mark_mutex lock.
Define a flag FSNOTIFY_GROUP_NOFS in fsnotify_group that determines
if the mark_mutex lock is fs reclaim safe or not. If not safe, the
lock helpers take the lock and disable direct fs reclaim.
In that case we annotate the mutex with a different lockdep class to
express to lockdep that an allocation of mark of an fs reclaim safe group
may take the group lock of another "NOFS" group to evict inodes.
For now, converted only the callers in common code and no backend
defines the NOFS flag. It is intended to be set by fanotify for
evictable marks support.
Link: https://lore.kernel.org/r/20220422120327.3459282-7-amir73il@gmail.com
Suggested-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220321112310.vpr7oxro2xkz5llh@quack3.lan/
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Instead of passing the allow_dups argument to fsnotify_add_mark()
as an argument, define the group flag FSNOTIFY_GROUP_DUPS to express
the allow_dups behavior and set this behavior at group creation time
for all calls of fsnotify_add_mark().
Rename the allow_dups argument to generic add_flags argument for future
use.
Link: https://lore.kernel.org/r/20220422120327.3459282-6-amir73il@gmail.com
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Commit 6960b0d909 ("fsnotify: change locking order") changed some
of the mark_mutex locks in direct reclaim path to use:
mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
This change is explained:
"...It uses nested locking to avoid deadlock in case we do the final
iput() on an inode which still holds marks and thus would take the
mutex again when calling fsnotify_inode_delete() in destroy_inode()."
The problem is that the mutex_lock_nested() is not a nested lock at
all. In fact, it has the opposite effect of preventing lockdep from
warning about a very possible deadlock.
Due to these wrong annotations, a deadlock that was introduced with
nfsd filecache in kernel v5.4 went unnoticed in v5.4.y for over two
years until it was reported recently by Khazhismel Kumykov, only to
find out that the deadlock was already fixed in kernel v5.5.
Fix the wrong lockdep annotations.
Cc: Khazhismel Kumykov <khazhy@google.com>
Fixes: 6960b0d909 ("fsnotify: change locking order")
Link: https://lore.kernel.org/r/20220321112310.vpr7oxro2xkz5llh@quack3.lan/
Link: https://lore.kernel.org/r/20220422120327.3459282-4-amir73il@gmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
fsnotify_parent() does not consider the parent's mark at all unless
the parent inode shows interest in events on children and in the
specific event.
So unless parent added an event to both its mark mask and ignored mask,
the event will not be ignored.
Fix this by declaring the interest of an object in an event when the
event is in either a mark mask or ignored mask.
Link: https://lore.kernel.org/r/20220223151438.790268-2-amir73il@gmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
They are two different types that use the same enum, so this confusing.
Use the object type to indicate the type of object mark is attached to
and the iter type to indicate the type of watch.
A group can have two different watches of the same object type (parent
and child watches) that match the same event.
Link: https://lore.kernel.org/r/20211129201537.1932819-3-amir73il@gmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Amir GoldsteinandJan Kara
Miklos SzerediandJan Kara
Linus Torvalds
Kees Cook
Marco CrivellariandChristian Brauner
Jann HornandJan Kara