[Xen-devel] 2.6.23.8 paravirt block hang

["Christopher S. Aker" <caker@xxxxxxxxxxxx>]

I have a test user reporting hangs during disk IO on paravirt_ops 
2.6.23.8, Xen 3.1.2 (64 bit Xen, pae host and guest).  Blocked processes 
continue to go up until the system is unusable.  No output on the 
guest's dmesg, nor anything unusual on the dom0 is logged.
root@dallas38:/xendev# xm block-list xen8
Vdev  BE handle state evt-ch ring-ref BE-path
51712    0    0     4      15     8 
/local/domain/0/backend/vbd/96/51712
51728    0    0     4      16     9 
/local/domain/0/backend/vbd/96/51728
blkback.96.xv S C037369D     0   851     19           852 16848 (L-TLB)
       cab1df24 00000246 cc3b6580 c037369d c0101227 cea24684 cea24684 
266fbeb5
       0001c834 00001868 00000000 cab1dfbc ca365030 267124fc 0001c834 
c1213920
       00000000 4ddd48df 00000001 c6f76040 cab1dfbc c9e9b954 c02c373b 
cfac4b84
Call Trace:
 [<c037369d>] scsi_dispatch_cmd+0x180/0x2a5
 [<c02c373b>] __generic_unplug_device+0x1f/0x25
 [<c02c3ebc>] generic_unplug_device+0x15/0x42
 [<c03e389b>] dm_table_unplug_all+0x21/0x2a
 [<c01304cb>] prepare_to_wait+0x12/0x4f
 [<c0353d55>] blkif_schedule+0x359/0x574
 [<c04918fb>] schedule+0x394/0x879
 [<c0130360>] autoremove_wake_function+0x0/0x37
 [<c03539fc>] blkif_schedule+0x0/0x574
 [<c013029a>] kthread+0xde/0xe2
 [<c01301bc>] kthread+0x0/0xe2
 [<c0102b75>] kernel_thread_helper+0x5/0xb
blkback.96.xv S C9037F10     0   852     19                 851 (L-TLB)
       c9037f24 00000246 00000002 c9037f10 00000001 cfa0050c cfac4b84 
dfbc6185
       0001c4c2 00000776 00000000 c9037fbc cfafa570 dfbcc1b2 0001c4c2 
c1213920
       c02c3662 0007088d 00000000 d04503c0 c9037fbc c9e9b3a8 c02c373b 
cfac4b84
Call Trace:
 [<c02c3662>] blk_remove_plug+0x38/0x6f
 [<c02c373b>] __generic_unplug_device+0x1f/0x25
 [<c02c3ebc>] generic_unplug_device+0x15/0x42
 [<c03e389b>] dm_table_unplug_all+0x21/0x2a
 [<c0353d55>] blkif_schedule+0x359/0x574
 [<c04918fb>] schedule+0x394/0x879
 [<c0130360>] autoremove_wake_function+0x0/0x37
 [<c03539fc>] blkif_schedule+0x0/0x574
 [<c013029a>] kthread+0xde/0xe2
 [<c01301bc>] kthread+0x0/0xe2
 [<c0102b75>] kernel_thread_helper+0x5/0xb

Do either of these stacks look unusual?

Have any other hints on how to further debug this?  He can reproduce it 
quite easily (un/tarring kernel ball, git --reset, etc).
-Chris

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel