XFS trouble: task xfssyncd blocked for more than 120 seconds.

[Julian Einwag]

Hi,

we're currently using XFS on our POP/IMAP server for storage of the  
user's mailboxes (Linux kernel version 2.6.27.10, x86_64).
Howevere, we're experencing some seemingly random filesystem lockups  
during which the filesystem can still be read, but not written. The  
following error messages show in in the system logs:

Jan 27 17:03:01 postman-64b kernel: "echo 0 > /proc/sys/kernel/ 
hung_task_timeout_secs" disables this message.
Jan 27 17:03:01 postman-64b kernel: xfssyncd      D  
0000000100921c06     0  3090      2
Jan 27 17:03:01 postman-64b kernel:  ffff88067b1cbd70 0000000000000046  
0000000000000000 0000000000000001
Jan 27 17:03:01 postman-64b kernel:  ffff88067a6124c0 ffff88067cc7c980  
ffff88067a612718 00000002a8421200
Jan 27 17:03:01 postman-64b kernel:  00000000fffffffd 0000000000001000  
0000000000000000 0000000000000001
Jan 27 17:03:01 postman-64b kernel: Call Trace:
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8041d432>]  
schedule_timeout+0x1e/0xad
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01ce88e>]  
xfs_buf_iorequest+0x5d/0x61 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01d2352>]  
xfs_bdstrat_cb+0x36/0x3a [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8041c9e4>]  
wait_for_common+0xc8/0x132
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8022b373>]  
default_wake_function+0x0/0xe
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01c70ba>]  
xfs_finish_reclaim+0x110/0x13b [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01c7169>]  
xfs_finish_reclaim_all+0x84/0xc9 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01c634b>] xfs_syncsub 
+0x4c/0x226 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01d3d54>]  
xfs_sync_worker+0x17/0x36 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01d4cb3>] xfssyncd 
+0x133/0x182 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffffa01d4b80>] xfssyncd 
+0x0/0x182 [xfs]
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff80244a68>] kthread 
+0x47/0x73
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8022f77f>] schedule_tail 
+0x27/0x5f
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8020ccf9>] child_rip 
+0xa/0x11
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff802f9395>]  
generic_unplug_device+0x0/0x2d
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff80244a21>] kthread 
+0x0/0x73
Jan 27 17:03:01 postman-64b kernel:  [<ffffffff8020ccef>] child_rip 
+0x0/0x11

Only rebooting helps.

The filesystem is hosted on a hardware-based RAID5 array.

We're currently out of ideas on what might be causing this, so any  
hints would be greately appreciated.

Best regards,
Julian Einwag