Inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage in XFS

Inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage in XFS

[Vladislav Bolkhovitin]

Hello,

After upgrading to 2.6.31 with SCST testings I started seeing the 
following lockdep messages. I have not seen them in 2.6.29. Looks like 
it's something recently introduced.

During the tests there was only one file used on XFS. It's a 5GB virtual 
device image. No SCST patches touch any I/O or memory management code.

Vlad

[ 4030.120972] =================================
[ 4030.121815] [ INFO: inconsistent lock state ]
[ 4030.121815] 2.6.31-scst-dbg #3
[ 4030.121815] ---------------------------------
[ 4030.121815] inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage.
[ 4030.121815] kswapd0/292 [HC0[0]:SC0[0]:HE1:SE1] takes:
[ 4030.121815]  (&(&ip->i_iolock)->mr_lock){++++-+}, at: [<f8fbef54>] 
xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815] {RECLAIM_FS-ON-R} state was registered at:
[ 4030.121815]   [<781707fb>] mark_held_locks+0x6a/0x9a
[ 4030.121815]   [<781708f5>] lockdep_trace_alloc+0xca/0xda
[ 4030.121815]   [<781c779e>] __alloc_pages_nodemask+0x8c/0x5db
[ 4030.121815]   [<781ca8db>] __do_page_cache_readahead+0xf8/0x1df
[ 4030.121815]   [<781ca9f7>] ra_submit+0x35/0x53
[ 4030.121815]   [<781cabcd>] ondemand_readahead+0x99/0x209
[ 4030.121815]   [<781cae50>] page_cache_sync_readahead+0x43/0x5c
[ 4030.121815]   [<781c2285>] generic_file_aio_read+0x3ed/0x67a
[ 4030.121815]   [<f8fec082>] xfs_read+0x13b/0x2b9 [xfs]
[ 4030.121815]   [<f8fe75c4>] xfs_file_aio_read+0x74/0xa0 [xfs]
[ 4030.121815]   [<f878074d>] do_sync_readv_writev+0xe8/0x153 [scst_vdisk]
[ 4030.121815]   [<f8784da4>] vdisk_do_job+0xd3e/0x1acf [scst_vdisk]
[ 4030.121815]   [<f8ee9939>] scst_do_real_exec+0xfe/0x62c [scst]
[ 4030.121815]   [<f8eea059>] scst_send_for_exec+0x1f2/0x710 [scst]
[ 4030.121815]   [<f8eec1d5>] scst_process_active_cmd+0x35d/0x1907 [scst]
[ 4030.121815]   [<f8eed809>] scst_do_job_active+0x8a/0x15d [scst]
[ 4030.121815]   [<f8eedaa3>] scst_cmd_thread+0xe8/0x29d [scst]
[ 4030.121815]   [<7815ac79>] kthread+0x84/0x8d
[ 4030.121815]   [<78103f57>] kernel_thread_helper+0x7/0x10
[ 4030.121815]   [<ffffffff>] 0xffffffff
[ 4030.121815] irq event stamp: 534933
[ 4030.121815] hardirqs last  enabled at (534933): [<7849b83d>] 
_spin_unlock_irqrestore+0x73/0x90
[ 4030.121815] hardirqs last disabled at (534932): [<781927c8>] 
call_rcu+0x2c/0x7d
[ 4030.121815] softirqs last  enabled at (534926): [<78147b28>] 
__do_softirq+0x1cf/0x20c
[ 4030.121815] softirqs last disabled at (534921): [<78105c5c>] 
do_softirq+0xaa/0xee
[ 4030.121815]
[ 4030.121815] other info that might help us debug this:
[ 4030.121815] 2 locks held by kswapd0/292:
[ 4030.121815]  #0:  (shrinker_rwsem){++++..}, at: [<781cd42e>] 
shrink_slab+0x31/0x1b8
[ 4030.121815]  #1:  (iprune_mutex){+.+.-.}, at: [<7820a351>] 
shrink_icache_memory+0x7a/0x274
[ 4030.121815]
[ 4030.121815] stack backtrace:
[ 4030.121815] Pid: 292, comm: kswapd0 Not tainted 2.6.31-scst-dbg #3
[ 4030.121815] Call Trace:
[ 4030.121815]  [<78497768>] ? printk+0x28/0x40
[ 4030.121815]  [<7816f8b9>] print_usage_bug+0x169/0x16e
[ 4030.121815]  [<781705d4>] mark_lock+0x1f6/0x3b3
[ 4030.121815]  [<7816fcbb>] ? check_usage_forwards+0x0/0xb0
[ 4030.121815]  [<78171876>] __lock_acquire+0x37d/0x102d
[ 4030.121815]  [<781707fb>] ? mark_held_locks+0x6a/0x9a
[ 4030.121815]  [<781033a7>] ? restore_all_notrace+0x0/0x18
[ 4030.121815]  [<f8fbef54>] ? xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815]  [<7849b83d>] ? _spin_unlock_irqrestore+0x73/0x90
[ 4030.121815]  [<78170af9>] ? trace_hardirqs_on_caller+0x13a/0x188
[ 4030.121815]  [<f8fbef54>] ? xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815]  [<7817260e>] lock_acquire+0xe8/0x127
[ 4030.121815]  [<f8fbef54>] ? xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815]  [<7815fd01>] down_write_nested+0x58/0xa5
[ 4030.121815]  [<f8fbef54>] ? xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815]  [<f8fbef54>] xfs_ilock+0x98/0x9f [xfs]
[ 4030.121815]  [<f8fbf12f>] xfs_ireclaim+0xa2/0xd5 [xfs]
[ 4030.121815]  [<f8fee9bc>] xfs_reclaim_inode+0xe6/0x150 [xfs]
[ 4030.121815]  [<f8fdd4cf>] xfs_reclaim+0xb2/0xb9 [xfs]
[ 4030.121815]  [<f8fecc4e>] xfs_fs_destroy_inode+0x3e/0x76 [xfs]
[ 4030.121815]  [<7820a100>] ? __destroy_inode+0x2b/0xa5
[ 4030.121815]  [<7820a1ae>] destroy_inode+0x34/0x5a
[ 4030.121815]  [<7820a26b>] dispose_list+0x97/0x103
[ 4030.121815]  [<7820a4ef>] shrink_icache_memory+0x218/0x274
[ 4030.121815]  [<781cd543>] shrink_slab+0x146/0x1b8
[ 4030.121815]  [<781cf248>] kswapd+0x501/0x625
[ 4030.121815]  [<781cc7ed>] ? isolate_pages_global+0x0/0x1fa
[ 4030.121815]  [<7815b03d>] ? autoremove_wake_function+0x0/0x5b
[ 4030.121815]  [<781ced47>] ? kswapd+0x0/0x625
[ 4030.121815]  [<7815ac79>] kthread+0x84/0x8d
[ 4030.121815]  [<7815abf5>] ? kthread+0x0/0x8d
[ 4030.121815]  [<78103f57>] kernel_thread_helper+0x7/0x10

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage in XFS

[Christoph Hellwig]

On Tue, Oct 06, 2009 at 04:40:37PM +0400, Vladislav Bolkhovitin wrote:
> Hello,
>
> After upgrading to 2.6.31 with SCST testings I started seeing the  
> following lockdep messages. I have not seen them in 2.6.29. Looks like  
> it's something recently introduced.

The lockdep code to track this has been recently introduced.

I've sent some patches to fix it ages ago, but so far no one has
bothered to review them.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage in XFS

[Eric Sandeen]

Christoph Hellwig wrote:
> On Tue, Oct 06, 2009 at 04:40:37PM +0400, Vladislav Bolkhovitin wrote:
>> Hello,
>>
>> After upgrading to 2.6.31 with SCST testings I started seeing the  
>> following lockdep messages. I have not seen them in 2.6.29. Looks like  
>> it's something recently introduced.
> 
> The lockdep code to track this has been recently introduced.
> 
> I've sent some patches to fix it ages ago, but so far no one has
> bothered to review them.

Christoph, remind me again which patch and I'll try to find time now
that I'm getting Fedora bug reports.  :)

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [xfs-masters] Inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage in XFS

[Christoph Hellwig]

On Wed, Oct 21, 2009 at 11:29:23AM -0500, Eric Sandeen wrote:
> > The lockdep code to track this has been recently introduced.
> > 
> > I've sent some patches to fix it ages ago, but so far no one has
> > bothered to review them.
> 
> Christoph, remind me again which patch and I'll try to find time now
> that I'm getting Fedora bug reports.  :)

I've recentlhy sent out some even simpler patches to sort this out, see
here:

	http://oss.sgi.com/pipermail/xfs/2009-October/042857.html
	http://oss.sgi.com/pipermail/xfs/2009-October/042858.html

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs