From: Dave Chinner <david@fromorbit.com>
To: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-xfs@vger.kernel.org, linux-mm@kvack.org,
"Darrick J. Wong" <djwong@kernel.org>,
Thomas Gleixner <tglx@linutronix.de>,
Andrew Morton <akpm@linux-foundation.org>,
Dave Chinner <dchinner@redhat.com>
Subject: [issue] kvmalloc doesn't repsect __GFP_NOLOCKDEP (was Re: [LOCKDEP] Circular locking dependency detected, fs_reclaim vs xfs_nondir_ilock_class)
Date: Sat, 20 Nov 2021 09:54:35 +1100 [thread overview]
Message-ID: <20211119225435.GZ449541@dread.disaster.area> (raw)
In-Reply-To: <20211119084444.ykmkdogsjdj3vtbi@linutronix.de>
[cc linux-mm@kvack.org]
On Fri, Nov 19, 2021 at 09:44:44AM +0100, Sebastian Andrzej Siewior wrote:
> Hi,
>
> left my box unattended for a while and lockdep reported this:
>
> | ======================================================
> | WARNING: possible circular locking dependency detected
> | 5.16.0-rc1+ #12 Not tainted
> | ------------------------------------------------------
> | kswapd1/510 is trying to acquire lock:
> | ffff88800c98ac70 (&xfs_nondir_ilock_class){++++}-{3:3}, at: xfs_icwalk_ag+0x365/0x810
> |
> | but task is already holding lock:
> | ffffffff82a76d60 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x4c4/0x5b0
> |
> | which lock already depends on the new lock.
> |
> |
> | the existing dependency chain (in reverse order) is:
> |
> | -> #1 (fs_reclaim){+.+.}-{0:0}:
> | fs_reclaim_acquire+0xa1/0xd0
> | __alloc_pages+0xed/0x380
> | new_slab+0x277/0x430
> | ___slab_alloc.constprop.0+0xb6a/0xfc0
> | __slab_alloc.constprop.0+0x42/0x80
> | __kmalloc_node+0xcc/0x200
> | xfs_attr_copy_value+0x6e/0x90
> | xfs_attr_get+0xa0/0xc0
> | xfs_get_acl+0xe4/0x210
> | get_acl.part.0+0x55/0x110
> | posix_acl_xattr_get+0x6a/0x120
> | vfs_getxattr+0x172/0x1a0
> | getxattr+0xb5/0x240
> | __x64_sys_fgetxattr+0x66/0xb0
> | do_syscall_64+0x59/0x80
> | entry_SYSCALL_64_after_hwframe+0x44/0xae
> |
> | -> #0 (&xfs_nondir_ilock_class){++++}-{3:3}:
> | __lock_acquire+0x12cb/0x2320
> | lock_acquire+0xc9/0x2e0
> | down_write_nested+0x42/0x110
> | xfs_icwalk_ag+0x365/0x810
> | xfs_icwalk+0x38/0xa0
> | xfs_reclaim_inodes_nr+0x90/0xc0
> | super_cache_scan+0x18e/0x1f0
> | shrink_slab.constprop.0+0x1cf/0x4d0
> | shrink_node+0x1e2/0x470
> | balance_pgdat+0x26d/0x5b0
> | kswapd+0x224/0x4e0
> | kthread+0x17a/0x1a0
> | ret_from_fork+0x1f/0x30
> |
> | other info that might help us debug this:
> |
> | Possible unsafe locking scenario:
> | CPU0 CPU1
> | ---- ----
> | lock(fs_reclaim);
> | lock(&xfs_nondir_ilock_class);
> | lock(fs_reclaim);
> | lock(&xfs_nondir_ilock_class);
> |
> | *** DEADLOCK ***
> |
> | 3 locks held by kswapd1/510:
> | #0: ffffffff82a76d60 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x4c4/0x5b0
> | #1: ffffffff82a69d58 (shrinker_rwsem){++++}-{3:3}, at: shrink_slab.constprop.0+0x3b/0x4d0
> | #2: ffff888553ccd0e8 (&type->s_umount_key#32){++++}-{3:3}, at: super_cache_scan+0x38/0x1f0
> |
> | stack backtrace:
> | CPU: 12 PID: 510 Comm: kswapd1 Not tainted 5.16.0-rc1+ #12
> | Hardware name: Intel Corporation S2600CP/S2600CP, BIOS SE5C600.86B.02.03.0003.041920141333 04/19/2014
> | Call Trace:
> | <TASK>
> | dump_stack_lvl+0x45/0x59
> | check_noncircular+0xff/0x110
> | __lock_acquire+0x12cb/0x2320
> | lock_acquire+0xc9/0x2e0
> | ? xfs_icwalk_ag+0x365/0x810
> | ? lock_is_held_type+0xd6/0x130
> | down_write_nested+0x42/0x110
> | ? xfs_icwalk_ag+0x365/0x810
> | xfs_icwalk_ag+0x365/0x810
> | xfs_icwalk+0x38/0xa0
> | xfs_reclaim_inodes_nr+0x90/0xc0
> | super_cache_scan+0x18e/0x1f0
> | shrink_slab.constprop.0+0x1cf/0x4d0
> | shrink_node+0x1e2/0x470
> | balance_pgdat+0x26d/0x5b0
> | ? lock_is_held_type+0xd6/0x130
> | kswapd+0x224/0x4e0
> | ? wait_woken+0x90/0x90
> | ? balance_pgdat+0x5b0/0x5b0
> | kthread+0x17a/0x1a0
> | ? set_kthread_struct+0x40/0x40
> | ret_from_fork+0x1f/0x30
> | </TASK>
>
> It appears to be related to commit
> d634525db63e9 ("xfs: replace kmem_alloc_large() with kvmalloc()")
>
> as SLUB's new_slab() has
> | return allocate_slab(s,
> | flags & (GFP_RECLAIM_MASK | GFP_CONSTRAINT_MASK), node);
>
> which drops the __GFP_NOLOCKDEP as used in xfs_attr_copy_value():
> | args->value = kvmalloc(valuelen, GFP_KERNEL | __GFP_NOLOCKDEP);
>
> However I can't tell if this should happen or not I'm just pointing out
> that it is.
Yup, that's a bug in the kvmalloc() implementation, not an XFS
issue. We're telling the memory allocator not to issue lockdep
warnings for this allocation because it's a known false positive
(impossible to be reclaiming an actively referenced inode, so cannot
deadlock like this).
Added linux-mm@kvack.org to the cc list so that they can add this to
the set of flags that kvmalloc and friends need to support
correctly.
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
prev parent reply other threads:[~2021-11-19 22:55 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-11-19 8:44 [LOCKDEP] Circular locking dependency detected, fs_reclaim vs xfs_nondir_ilock_class Sebastian Andrzej Siewior
2021-11-19 22:54 ` Dave Chinner [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20211119225435.GZ449541@dread.disaster.area \
--to=david@fromorbit.com \
--cc=akpm@linux-foundation.org \
--cc=bigeasy@linutronix.de \
--cc=dchinner@redhat.com \
--cc=djwong@kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-xfs@vger.kernel.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).