Re: [syzbot] [mm?] possible deadlock in rhashtable_free_and_destroy

[Hillf Danton <hdanton@sina.com>]

On Tue, 21 Apr 2026 14:27:35 -0700 Shakeel Butt wrote:
>Ccing relevant folks as this seems related to recent change of moving
>simple_xattrs from rbtree to rhastable.
>
>On Tue, Apr 21, 2026 at 08:34:22AM -0700, syzbot wrote:
>> Hello,
>> 
>> syzbot found the following issue on:
>> 
>> HEAD commit:    8541d8f725c6 Merge tag 'mtd/for-7.1' of git://git.kernel.o..
>> git tree:       upstream
>> console output: https://syzkaller.appspot.com/x/log.txt?x=15380836580000
>> kernel config:  https://syzkaller.appspot.com/x/.config?x=7e54da1916e8d11f
>> dashboard link: https://syzkaller.appspot.com/bug?extid=5af806780f38a5fe691f
>> compiler:       gcc (Debian 14.2.0-19) 14.2.0, GNU ld (GNU Binutils for Debian) 2.44
>> 
>> Unfortunately, I don't have any reproducer for this issue yet.
>> 
>> Downloadable assets:
>> disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/d900f083ada3/non_bootable_disk-8541d8f7.raw.xz
>> vmlinux: https://storage.googleapis.com/syzbot-assets/22dfea2c37c2/vmlinux-8541d8f7.xz
>> kernel image: https://storage.googleapis.com/syzbot-assets/e2f93ad68fe3/bzImage-8541d8f7.xz
>> 
>> IMPORTANT: if you fix the issue, please add the following tag to the commit:
>> Reported-by: syzbot+5af806780f38a5fe691f@syzkaller.appspotmail.com
>> 
>> ======================================================
>> WARNING: possible circular locking dependency detected
>> syzkaller #0 Tainted: G             L     
>> ------------------------------------------------------
>> kswapd0/108 is trying to acquire lock:
>> ffff888056f3c4e8 (&ht->mutex){+.+.}-{4:4}, at: rhashtable_free_and_destroy+0x3d/0x9b0 lib/rhashtable.c:1154
>> 
>> but task is already holding lock:
>> ffffffff8e9b0800 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0xb5d/0x1ac0 mm/vmscan.c:7102
>> 
>> which lock already depends on the new lock.
>> 
>> 
>> the existing dependency chain (in reverse order) is:
>> 
>> -> #1 (fs_reclaim){+.+.}-{0:0}:
>>        __fs_reclaim_acquire mm/page_alloc.c:4327 [inline]
>>        fs_reclaim_acquire+0xc4/0x100 mm/page_alloc.c:4341
>>        might_alloc include/linux/sched/mm.h:317 [inline]
>>        slab_pre_alloc_hook mm/slub.c:4520 [inline]
>>        slab_alloc_node mm/slub.c:4875 [inline]
>>        __do_kmalloc_node mm/slub.c:5294 [inline]
>>        __kvmalloc_node_noprof+0xcc/0xa00 mm/slub.c:6828
>>        bucket_table_alloc.isra.0+0x88/0x460 lib/rhashtable.c:186
>>        rhashtable_rehash_alloc+0x68/0x110 lib/rhashtable.c:368
>>        rht_deferred_worker+0x1d9/0x1fd0 lib/rhashtable.c:429
>>        process_one_work+0xa0e/0x1980 kernel/workqueue.c:3302
>>        process_scheduled_works kernel/workqueue.c:3385 [inline]
>>        worker_thread+0x5ef/0xe50 kernel/workqueue.c:3466
>>        kthread+0x370/0x450 kernel/kthread.c:436
>>        ret_from_fork+0x72b/0xd50 arch/x86/kernel/process.c:158
>>        ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245
>> 
>> -> #0 (&ht->mutex){+.+.}-{4:4}:
>>        check_prev_add kernel/locking/lockdep.c:3165 [inline]
>>        check_prevs_add kernel/locking/lockdep.c:3284 [inline]
>>        validate_chain kernel/locking/lockdep.c:3908 [inline]
>>        __lock_acquire+0x14b8/0x2630 kernel/locking/lockdep.c:5237
>>        lock_acquire kernel/locking/lockdep.c:5868 [inline]
>>        lock_acquire+0x1b1/0x370 kernel/locking/lockdep.c:5825
>>        __mutex_lock_common kernel/locking/mutex.c:632 [inline]
>>        __mutex_lock+0x1a4/0x1b10 kernel/locking/mutex.c:806
>
> The cancel_work_sync(&ht->run_work) in rhashtable_free_and_destroy() should
> avoid concurrent mutex lockers from rht_deferred_worker(). Seems like false

Yes.

> positive as all rhastables share single static lockdep class for the ht->mutex.
>
Nope, given schedule_work() in the &ht->run_work callback, if the run_work is
scheduled once more, the race window between the destroy and work callback is
still open. And simple fix looks like disableing the work item before cancel.

> Simple fix would be to move rhashtable_rehash_alloc() to use NOFS allocation or
> we can introduce more fine grained lockdep classes for rhashtables. 
>
>>        rhashtable_free_and_destroy+0x3d/0x9b0 lib/rhashtable.c:1154
>>        shmem_evict_inode+0x1ae/0xc40 mm/shmem.c:1429
>>        evict+0x3c2/0xad0 fs/inode.c:841
>>        iput_final fs/inode.c:1960 [inline]
>>        iput.part.0+0x605/0xf50 fs/inode.c:2009
>>        iput+0x35/0x40 fs/inode.c:1975
>>        dentry_unlink_inode+0x2a1/0x490 fs/dcache.c:467
>>        __dentry_kill+0x1d0/0x600 fs/dcache.c:670
>>        finish_dput+0x76/0x480 fs/dcache.c:879
>>        dput.part.0+0x456/0x570 fs/dcache.c:928
>>        dput+0x1f/0x30 fs/dcache.c:920
>>        ovl_destroy_inode+0x3e/0x190 fs/overlayfs/super.c:217
>>        destroy_inode+0xcb/0x1c0 fs/inode.c:394
>>        evict+0x599/0xad0 fs/inode.c:865
>>        iput_final fs/inode.c:1960 [inline]
>>        iput.part.0+0x605/0xf50 fs/inode.c:2009
>>        iput+0x35/0x40 fs/inode.c:1975
>>        dentry_unlink_inode+0x2a1/0x490 fs/dcache.c:467
>>        __dentry_kill+0x1d0/0x600 fs/dcache.c:670
>>        shrink_kill fs/dcache.c:1147 [inline]
>>        shrink_dentry_list+0x180/0x5e0 fs/dcache.c:1174
>>        prune_dcache_sb+0xea/0x150 fs/dcache.c:1256
>>        super_cache_scan+0x328/0x550 fs/super.c:223
>>        do_shrink_slab+0x416/0x1240 mm/shrinker.c:440
>>        shrink_slab_memcg mm/shrinker.c:557 [inline]
>>        shrink_slab+0xa7d/0x12e0 mm/shrinker.c:635
>>        shrink_one+0x398/0x7f0 mm/vmscan.c:4932
>>        shrink_many mm/vmscan.c:4993 [inline]
>>        lru_gen_shrink_node mm/vmscan.c:5071 [inline]
>>        shrink_node+0x2673/0x3dc0 mm/vmscan.c:6059
>>        kswapd_shrink_node mm/vmscan.c:6913 [inline]
>>        balance_pgdat+0xaaf/0x1ac0 mm/vmscan.c:7089
>>        kswapd+0x557/0xb60 mm/vmscan.c:7362
>>        kthread+0x370/0x450 kernel/kthread.c:436
>>        ret_from_fork+0x72b/0xd50 arch/x86/kernel/process.c:158
>>        ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245
>> 
>> other info that might help us debug this:
>> 
>>  Possible unsafe locking scenario:
>> 
>>        CPU0                    CPU1
>>        ----                    ----
>>   lock(fs_reclaim);
>>                                lock(&ht->mutex);
>>                                lock(fs_reclaim);
>>   lock(&ht->mutex);
>> 
>>  *** DEADLOCK ***