[openib-general] ib_sa oops

[Hal Rosenstock]

When I did a modprobe -r ib_ipoib, I got the following oops when the
SA's send_handler is called on it's deregistering it's MAD client with
pending MADs.

I first bringup and configure IPoIB:
/sbin/modprobe ib_ipoib
/sbin/ifconfig ib0 192.168.0.20

I then do:
ping -b 192.168.0.255
and ctl-C before it cycles around the list a second time and then:

/sbin/modprobe -r ib_ipoib
Segmentation fault

/var/log/messages showed:
Nov  2 10:54:17 hpc-1 kernel: Unable to handle kernel paging request at
virtual
address f8a50407
Nov  2 10:54:17 hpc-1 kernel:  printing eip:
Nov  2 10:54:17 hpc-1 kernel: f8a50407
Nov  2 10:54:17 hpc-1 kernel: *pde = 019a5067
Nov  2 10:54:17 hpc-1 kernel: *pte = 00000000
Nov  2 10:54:17 hpc-1 kernel: Oops: 0000 [#1]
Nov  2 10:54:17 hpc-1 kernel: SMP
Nov  2 10:54:17 hpc-1 kernel: Modules linked in: ib_sa ib_mad
ib_services ib_mthca ib_core loop autofs e1000 ohci1394 ieee1394
parport_pc parport usbcore
Nov  2 10:54:17 hpc-1 kernel: CPU:    0
Nov  2 10:54:17 hpc-1 kernel: EIP:    0060:[<f8a50407>]    Not tainted
VLI
Nov  2 10:54:17 hpc-1 kernel: EFLAGS: 00010246   (2.6.9)
Nov  2 10:54:17 hpc-1 kernel: EIP is at 0xf8a50407
Nov  2 10:54:17 hpc-1 kernel: eax: e2f05280   ebx: 00000286   ecx:
00000000   edx: fffffffb
Nov  2 10:54:17 hpc-1 kernel: esi: c6ba3340   edi: c6ba3348   ebp:
fffffffb   esp: e6eebdfc
Nov  2 10:54:17 hpc-1 kernel: ds: 007b   es: 007b   ss: 0068
Nov  2 10:54:17 hpc-1 kernel: Process modprobe (pid: 12680,
threadinfo=e6eea000
task=f5f30230)
Nov  2 10:54:17 hpc-1 kernel: Stack: f8a217d8 fffffffb 00000000 e2f05280
e6eebe60 c02a1e5e 00000000 f5f30230
Nov  2 10:54:17 hpc-1 kernel:        c0117d96 00000000 00000000 00000003
c170b060 c6ff3a70 c6ff3830 c011685a
Nov  2 10:54:17 hpc-1 kernel:        f5f30230 e74b5800 f5f30230 00000000
e6eebe98 c02a1a92 c6ff3830 c170e4d0
Nov  2 10:54:17 hpc-1 kernel: Call Trace:
Nov  2 10:54:17 hpc-1 kernel:  [<f8a217d8>]
ib_sa_mcmember_rec_callback+0x5a/0x7f [ib_sa]
Nov  2 10:54:17 hpc-1 kernel:  [<c02a1e5e>]
wait_for_completion+0xc4/0xcc
Nov  2 10:54:17 hpc-1 kernel:  [<c0117d96>]
default_wake_function+0x0/0x12
Nov  2 10:54:17 hpc-1 kernel:  [<c011685a>] finish_task_switch+0x3a/0x83
Nov  2 10:54:17 hpc-1 kernel:  [<c02a1a92>] schedule+0x326/0x62e
Nov  2 10:54:17 hpc-1 kernel:  [<f8a21a24>] send_handler+0xaa/0xbc
[ib_sa]
Nov  2 10:54:17 hpc-1 kernel:  [<f89e8642>] cancel_mads+0xe5/0x127
[ib_mad]
Nov  2 10:54:17 hpc-1 kernel:  [<f89e737a>]
ib_unregister_mad_agent+0x16/0x135 [ib_mad]
Nov  2 10:54:17 hpc-1 kernel:  [<c0117d96>]
default_wake_function+0x0/0x12
Nov  2 10:54:17 hpc-1 kernel:  [<c0117d96>]
default_wake_function+0x0/0x12
Nov  2 10:54:17 hpc-1 kernel:  [<f89f2928>] ib_get_client_data+0x42/0x4e
[ib_core]
Nov  2 10:54:17 hpc-1 kernel:  [<f8a21d87>] ib_sa_remove_one+0x44/0x7d
[ib_sa]
Nov  2 10:54:17 hpc-1 kernel:  [<f89f28e1>]
ib_unregister_client+0xee/0xf3 [ib_core]
Nov  2 10:54:17 hpc-1 kernel:  [<c0130ecb>] try_stop_module+0x37/0x3b
Nov  2 10:54:17 hpc-1 kernel:  [<c0133941>] __try_stop_module+0x0/0x41
Nov  2 10:54:17 hpc-1 kernel:  [<f8a21dcf>] ib_sa_cleanup+0xf/0x13
[ib_sa]
Nov  2 10:54:17 hpc-1 kernel:  [<c01310c1>]
sys_delete_module+0x16d/0x19b
Nov  2 10:54:17 hpc-1 kernel:  [<c01474de>] sys_munmap+0x51/0x76
Nov  2 10:54:17 hpc-1 kernel:  [<c0105cf5>] sysenter_past_esp+0x52/0x71
Nov  2 10:54:17 hpc-1 kernel: Code:  Bad EIP value.

-- Hal