[dm-devel] Last path failure causes oops

[Lars Marowsky-Bree]

Hi, I'm finally finding time to work and play with dm-mpath again. It works,
unless there's a failure ;-)

(I've added a couple more log messages in places so that it's a little
bit more obvious. Right now, dm/dm-mpath rather keeps its troubles to
itself; if this was a relationship, I'd be sending it to a relationship
counselor that talking is much better than silently breaking...)

Anyway:

device-mapper: dm-multipath: IO error on path 8:0, sector 256
device-mapper: dm-multipath: no valid paths left.
Buffer I/O error on device dm-4, logical block 31
Unable to handle kernel NULL pointer dereference at virtual address 00000010
 printing eip:
f89cc8e0
*pde = 00102001
Oops: 0000 [#1]
SMP 
CPU:    0
EIP:    0060:[<f89cc8e0>]    Tainted: G  U
EFLAGS: 00010282   (2.6.5-7.162-bigsmp SLES9_SP2_BRANCH-200504181325230000) 
EIP is at multipath_end_io+0x90/0x240 [dm_multipath]
eax: 00000000   ebx: f7f60b80   ecx: fffffffb   edx: 00000000
esi: f2dad1c0   edi: 00000000   ebp: f737c45c   esp: cde59eac
ds: 007b   es: 007b   ss: 0068
Process events/0 (pid: 18, threadinfo=cde58000 task=cde48000)
Stack: f2dadf00 00001000 00000008 00000000 00000001 f7f60b90 00000000 fffffffb 
       f6e153f0 f7dfb3f0 f2dad1c0 f89e1580 f6e153f8 f8914080 f89cc850 00000000 
       f2dad1c0 00001000 f89e1510 fffffffb c017f078 00000001 00000296 00000296 
Call Trace:
 [<f89e1580>] clone_endio+0x70/0x120 [dm_mod]
 [<f89cc850>] multipath_end_io+0x0/0x240 [dm_multipath]
 [<f89e1510>] clone_endio+0x0/0x120 [dm_mod]
 [<c017f078>] bio_endio+0x68/0xa0
 [<f89ccdac>] process_queued_ios+0x12c/0x140 [dm_multipath]
 [<c013c2a6>] worker_thread+0x186/0x230
 [<f89ccc80>] process_queued_ios+0x0/0x140 [dm_multipath]
 [<c0123910>] default_wake_function+0x0/0x10
 [<c0123910>] default_wake_function+0x0/0x10
 [<c013c120>] worker_thread+0x0/0x230
 [<c0140019>] kthread+0xf9/0x12d
 [<c013ff20>] kthread+0x0/0x12d
 [<c0107005>] kernel_thread_helper+0x5/0x10

Code: 8b 40 10 c7 04 24 e8 da 9c f8 83 c0 14 89 44 24 04 e8 ca f1 
 Dumping to block device (104,5) on CPU 0 ...

Any ideas?


Sincerely,
    Lars Marowsky-Brée <lmb at suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX Products GmbH - A Novell Business