On Fri 27-09-24 11:31:50, Amir Goldstein wrote:
> On Fri, Sep 27, 2024 at 11:12 AM Lizhi Xu <lizhi.xu@xxxxxxxxxxxxx> wrote:
> >
> > [Syzbot reported]
> > WARNING: possible circular locking dependency detected
> > 6.11.0-rc4-syzkaller-00019-gb311c1b497e5 #0 Not tainted
> > ------------------------------------------------------
> > kswapd0/78 is trying to acquire lock:
> > ffff88801b8d8930 (&group->mark_mutex){+.+.}-{3:3}, at: fsnotify_group_lock include/linux/fsnotify_backend.h:270 [inline]
> > ffff88801b8d8930 (&group->mark_mutex){+.+.}-{3:3}, at: fsnotify_destroy_mark+0x38/0x3c0 fs/notify/mark.c:578
> >
> > but task is already holding lock:
> > ffffffff8ea2fd60 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat mm/vmscan.c:6841 [inline]
> > ffffffff8ea2fd60 (fs_reclaim){+.+.}-{0:0}, at: kswapd+0xbb4/0x35a0 mm/vmscan.c:7223
> >
> > which lock already depends on the new lock.
> >
> >
> > the existing dependency chain (in reverse order) is:
> >
> > -> #1 (fs_reclaim){+.+.}-{0:0}:
> > lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5759
> > __fs_reclaim_acquire mm/page_alloc.c:3818 [inline]
> > fs_reclaim_acquire+0x88/0x140 mm/page_alloc.c:3832
> > might_alloc include/linux/sched/mm.h:334 [inline]
> > slab_pre_alloc_hook mm/slub.c:3939 [inline]
> > slab_alloc_node mm/slub.c:4017 [inline]
> > kmem_cache_alloc_noprof+0x3d/0x2a0 mm/slub.c:4044
> > inotify_new_watch fs/notify/inotify/inotify_user.c:599 [inline]
> > inotify_update_watch fs/notify/inotify/inotify_user.c:647 [inline]
> > __do_sys_inotify_add_watch fs/notify/inotify/inotify_user.c:786 [inline]
> > __se_sys_inotify_add_watch+0x72e/0x1070 fs/notify/inotify/inotify_user.c:729
> > do_syscall_x64 arch/x86/entry/common.c:52 [inline]
> > do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
> > entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >
> > -> #0 (&group->mark_mutex){+.+.}-{3:3}:
> > check_prev_add kernel/locking/lockdep.c:3133 [inline]
> > check_prevs_add kernel/locking/lockdep.c:3252 [inline]
> > validate_chain+0x18e0/0x5900 kernel/locking/lockdep.c:3868
> > __lock_acquire+0x137a/0x2040 kernel/locking/lockdep.c:5142
> > lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5759
> > __mutex_lock_common kernel/locking/mutex.c:608 [inline]
> > __mutex_lock+0x136/0xd70 kernel/locking/mutex.c:752
> > fsnotify_group_lock include/linux/fsnotify_backend.h:270 [inline]
> > fsnotify_destroy_mark+0x38/0x3c0 fs/notify/mark.c:578
> > fsnotify_destroy_marks+0x14a/0x660 fs/notify/mark.c:934
> > fsnotify_inoderemove include/linux/fsnotify.h:264 [inline]
> > dentry_unlink_inode+0x2e0/0x430 fs/dcache.c:403
> > __dentry_kill+0x20d/0x630 fs/dcache.c:610
> > shrink_kill+0xa9/0x2c0 fs/dcache.c:1055
> > shrink_dentry_list+0x2c0/0x5b0 fs/dcache.c:1082
> > prune_dcache_sb+0x10f/0x180 fs/dcache.c:1163
> > super_cache_scan+0x34f/0x4b0 fs/super.c:221
> > do_shrink_slab+0x701/0x1160 mm/shrinker.c:435
> > shrink_slab+0x1093/0x14d0 mm/shrinker.c:662
> > shrink_one+0x43b/0x850 mm/vmscan.c:4815
> > shrink_many mm/vmscan.c:4876 [inline]
> > lru_gen_shrink_node mm/vmscan.c:4954 [inline]
> > shrink_node+0x3799/0x3de0 mm/vmscan.c:5934
> > kswapd_shrink_node mm/vmscan.c:6762 [inline]
> > balance_pgdat mm/vmscan.c:6954 [inline]
> > kswapd+0x1bcd/0x35a0 mm/vmscan.c:7223
> > kthread+0x2f0/0x390 kernel/kthread.c:389
> > ret_from_fork+0x4b/0x80 arch/x86/kernel/process.c:147
> > ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:244
> >
> > other info that might help us debug this:
> >
> > Possible unsafe locking scenario:
> >
> > CPU0 CPU1
> > ---- ----
> > lock(fs_reclaim);
> > lock(&group->mark_mutex);
> > lock(fs_reclaim);
> > lock(&group->mark_mutex);
> >
> > *** DEADLOCK ***
> >
> > [Analysis]
> > The inotify_new_watch() call passes through GFP_KERNEL, use memalloc_nofs_save/
> > memalloc_nofs_restore to make sure we don't end up with the fs reclaim dependency.
>
> I don't think this can actually happen, because an inode with
> an inotify mark cannot get evicted,
Well, in the trace above dentry reclaim apparently raced with unlink and so
ended up going to dentry_unlink_inode() -> fsnotify_inoderemove() which
does indeed end up grabbing group->mark_mutex.
> but I cannot think of a way to annotate
> this to lockdep, so if we need to silence lockdep, this is what
> FSNOTIFY_GROUP_NOFS was created for.
So yes, inotify needs FSNOTIFY_GROUP_NOFS as well. In fact this trace shows
that any notification group needs to use NOFS allocations to be safe
against this race so we can just remove FSNOTIFY_GROUP_NOFS and
unconditionally do memalloc_nofs_save() in fsnotify_group_lock(). Lizhi,
will you send a patch please?
Honza
> > Reported-and-tested-by: syzbot+c679f13773f295d2da53@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Closes: https://syzkaller.appspot.com/bug?extid=c679f13773f295d2da53
> > Signed-off-by: Lizhi Xu <lizhi.xu@xxxxxxxxxxxxx>
> > ---
> > fs/notify/inotify/inotify_user.c | 7 ++++++-
> > 1 file changed, 6 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c
> > index c7e451d5bd51..70b77b6186a6 100644
> > --- a/fs/notify/inotify/inotify_user.c
> > +++ b/fs/notify/inotify/inotify_user.c
> > @@ -643,8 +643,13 @@ static int inotify_update_watch(struct fsnotify_group *group, struct inode *inod
> > /* try to update and existing watch with the new arg */
> > ret = inotify_update_existing_watch(group, inode, arg);
> > /* no mark present, try to add a new one */
> > - if (ret == -ENOENT)
> > + if (ret == -ENOENT) {
> > + unsigned int nofs_flag;
> > +
> > + nofs_flag = memalloc_nofs_save();
> > ret = inotify_new_watch(group, inode, arg);
> > + memalloc_nofs_restore(nofs_flag);
> > + }
> > fsnotify_group_unlock(group);
> >
> > return ret;
> > --
> > 2.43.0
> >
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
