For what it's worth: my nightly NFS test run, on upstream b2b436ec "Merge
tag 'trace-v5.16-rc8", spat this out last night.
The night before, on 75acfdb6 "Merge tag 'net-5.16-final' of
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net", I didn't see
it.
--b.
On Fri, Jan 07, 2022 at 12:35:10AM -0500, J. Bruce Fields wrote:
> [ 7814.324263] ======================================================
> [ 7814.324765] WARNING: possible circular locking dependency detected
> [ 7814.325285] 5.16.0-rc8-00052-gb2b436ec0205 #1221 Not tainted
> [ 7814.325802] ------------------------------------------------------
> [ 7814.326383] kswapd0/74 is trying to acquire lock:
> [ 7814.326730] ffff88801a06b828 (&xfs_dir_ilock_class){++++}-{3:3}, at: xfs_icwalk_ag+0x7af/0x1440
> [ 7814.327456]
> but task is already holding lock:
> [ 7814.327780] ffffffff85176b40 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x91e/0xad0
> [ 7814.328411]
> which lock already depends on the new lock.
>
> [ 7814.328857]
> the existing dependency chain (in reverse order) is:
> [ 7814.329367]
> -> #1 (fs_reclaim){+.+.}-{0:0}:
> [ 7814.329736] fs_reclaim_acquire+0xc0/0x100
> [ 7814.330092] __alloc_pages+0x14f/0x560
> [ 7814.330389] cache_alloc_refill+0x58c/0x780
> [ 7814.330762] __kmalloc+0x1f2/0x230
> [ 7814.331026] xfs_attr_copy_value+0xfb/0x170
> [ 7814.331365] xfs_attr_get+0x365/0x4b0
> [ 7814.331653] xfs_get_acl+0x1be/0x410
> [ 7814.331957] get_acl.part.0+0xb3/0x1e0
> [ 7814.332256] posix_acl_xattr_get+0x128/0x220
> [ 7814.332625] vfs_getxattr+0x1db/0x220
> [ 7814.332958] getxattr+0x100/0x2b0
> [ 7814.333248] path_getxattr+0xba/0x120
> [ 7814.333576] do_syscall_64+0x43/0x90
> [ 7814.333887] entry_SYSCALL_64_after_hwframe+0x44/0xae
> [ 7814.334329]
> -> #0 (&xfs_dir_ilock_class){++++}-{3:3}:
> [ 7814.334744] __lock_acquire+0x29f8/0x5b80
> [ 7814.335075] lock_acquire+0x1a6/0x4b0
> [ 7814.335419] down_write_nested+0x86/0x130
> [ 7814.335766] xfs_icwalk_ag+0x7af/0x1440
> [ 7814.336071] xfs_icwalk+0x4f/0xd0
> [ 7814.336327] xfs_reclaim_inodes_nr+0x130/0x1c0
> [ 7814.336691] super_cache_scan+0x3ae/0x520
> [ 7814.337034] shrink_slab.constprop.0+0x359/0x830
> [ 7814.337531] shrink_node+0x546/0xf20
> [ 7814.337867] balance_pgdat+0x41a/0xad0
> [ 7814.338183] kswapd+0x54c/0xa20
> [ 7814.338422] kthread+0x3b1/0x490
> [ 7814.338670] ret_from_fork+0x22/0x30
> [ 7814.338974]
> other info that might help us debug this:
>
> [ 7814.339375] Possible unsafe locking scenario:
>
> [ 7814.339730] CPU0 CPU1
> [ 7814.340049] ---- ----
> [ 7814.340368] lock(fs_reclaim);
> [ 7814.340558] lock(&xfs_dir_ilock_class);
> [ 7814.341114] lock(fs_reclaim);
> [ 7814.341610] lock(&xfs_dir_ilock_class);
> [ 7814.341919]
> *** DEADLOCK ***
>
> [ 7814.342134] 3 locks held by kswapd0/74:
> [ 7814.342410] #0: ffffffff85176b40 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x91e/0xad0
> [ 7814.343086] #1: ffffffff85158fb0 (shrinker_rwsem){++++}-{3:3}, at: shrink_slab.constprop.0+0x71/0x830
> [ 7814.343867] #2: ffff88800e8fc0e0 (&type->s_umount_key#48){++++}-{3:3}, at: super_cache_scan+0x53/0x520
> [ 7814.344654]
> stack backtrace:
> [ 7814.344837] CPU: 0 PID: 74 Comm: kswapd0 Not tainted 5.16.0-rc8-00052-gb2b436ec0205 #1221
> [ 7814.345594] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-6.fc35 04/01/2014
> [ 7814.346354] Call Trace:
> [ 7814.346478] <TASK>
> [ 7814.346569] dump_stack_lvl+0x45/0x59
> [ 7814.346835] check_noncircular+0x23e/0x2e0
> [ 7814.347116] ? print_circular_bug+0x450/0x450
> [ 7814.347420] ? filter_irq_stacks+0x90/0x90
> [ 7814.347707] ? check_noncircular+0x132/0x2e0
> [ 7814.348004] ? add_lock_to_list.constprop.0+0x193/0x5b0
> [ 7814.348394] __lock_acquire+0x29f8/0x5b80
> [ 7814.348670] ? lockdep_hardirqs_on_prepare+0x400/0x400
> [ 7814.349054] lock_acquire+0x1a6/0x4b0
> [ 7814.349328] ? xfs_icwalk_ag+0x7af/0x1440
> [ 7814.349665] ? lock_release+0x6d0/0x6d0
> [ 7814.349977] ? lock_is_held_type+0xd7/0x130
> [ 7814.350290] down_write_nested+0x86/0x130
> [ 7814.350563] ? xfs_icwalk_ag+0x7af/0x1440
> [ 7814.350838] ? rwsem_down_write_slowpath+0x10f0/0x10f0
> [ 7814.351218] ? rcu_read_lock_sched_held+0x3f/0x70
> [ 7814.351599] ? xfs_ilock+0x1da/0x330
> [ 7814.351835] xfs_icwalk_ag+0x7af/0x1440
> [ 7814.352094] ? xfs_inode_free_cowblocks+0x1e0/0x1e0
> [ 7814.352449] ? lock_is_held_type+0xd7/0x130
> [ 7814.352741] ? find_held_lock+0x2c/0x110
> [ 7814.353007] ? lock_release+0x3b8/0x6d0
> [ 7814.353296] ? xfs_perag_get_tag+0xe2/0x430
> [ 7814.353646] ? lock_downgrade+0x690/0x690
> [ 7814.353956] ? lock_is_held_type+0xd7/0x130
> [ 7814.354247] ? rcu_read_lock_sched_held+0x3f/0x70
> [ 7814.354585] ? xfs_perag_get_tag+0x279/0x430
> [ 7814.354885] ? xfs_perag_get+0x310/0x310
> [ 7814.355149] ? xfs_icwalk+0x83/0xd0
> [ 7814.355372] ? rcu_read_lock_sched_held+0x3f/0x70
> [ 7814.355714] xfs_icwalk+0x4f/0xd0
> [ 7814.355922] xfs_reclaim_inodes_nr+0x130/0x1c0
> [ 7814.356236] ? xfs_reclaim_inodes+0x1d0/0x1d0
> [ 7814.356543] super_cache_scan+0x3ae/0x520
> [ 7814.356822] shrink_slab.constprop.0+0x359/0x830
> [ 7814.357177] ? move_pages_to_lru+0xd80/0xd80
> [ 7814.357534] shrink_node+0x546/0xf20
> [ 7814.357781] balance_pgdat+0x41a/0xad0
> [ 7814.358039] ? shrink_node+0xf20/0xf20
> [ 7814.358294] ? lock_is_held_type+0xd7/0x130
> [ 7814.358595] ? lock_is_held_type+0xd7/0x130
> [ 7814.358921] kswapd+0x54c/0xa20
> [ 7814.359115] ? balance_pgdat+0xad0/0xad0
> [ 7814.359379] ? _raw_spin_unlock_irqrestore+0x2d/0x50
> [ 7814.359845] ? lockdep_hardirqs_on+0x79/0x100
> [ 7814.360152] ? finish_wait+0x270/0x270
> [ 7814.360400] ? __kthread_parkme+0xcc/0x1f0
> [ 7814.360682] ? balance_pgdat+0xad0/0xad0
> [ 7814.360982] kthread+0x3b1/0x490
> [ 7814.361214] ? _raw_spin_unlock_irq+0x24/0x50
> [ 7814.361575] ? set_kthread_struct+0x100/0x100
> [ 7814.361914] ret_from_fork+0x22/0x30
> [ 7814.362159] </TASK>
