NULL pointer dereference bug in 2.6.18-rc6-CITI_NFS4_ALL-1

[Bryce Harrington]

Jason noticed that in the latest CITI patch the kernel hits a bug during
the LTP run with a krb5p mount.  There's also some interesting looking
console output during the LTP runs with the other security flavors as
well.

BUG: unable to handle kernel NULL pointer dereference at virtual address
00000000
printing eip:
c040d211
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP 
Modules linked in:
CPU:    0
EIP:    0060:[<c040d211>]    Not tainted VLI
EFLAGS: 00010212   (2.6.18-rc6-CITI_NFS4_ALL-1 #1) 
EIP is at svc_process+0x40/0x5c5
eax: 00000000   ebx: f366c070   ecx: f7b50550   edx: f6bf8720
esi: 00000001   edi: f366c000   ebp: f3894fb0   esp: f3894f7c
ds: 007b   es: 007b   ss: 0068
Process nfsd (pid: 15671, ti=f3894000 task=f7b50550 task.ti=f3894000)
Stack: 00000000 00000213 f3894f98 c041c380 00000002 00000002 f6bf8720
f366c04c 
c041c6d1 f3894fb4 000ca365 00000001 000ca365 f3894fe4 c01e131b f366c000 
00000002 f3894fc8 00000000 fffffeff ffffffff fffffef8 ffffffff c01e11a2 
Call Trace:
[<c01037d6>] show_stack_log_lvl+0x8a/0x92
[<c0103937>] show_registers+0x11d/0x186
[<c0103b23>] die+0x10c/0x1c2
[<c0113f1d>] do_page_fault+0x3e0/0x4bc
[<c01034a9>] error_code+0x39/0x40
[<c01e131b>] nfsd+0x179/0x283


Is this a known issue?  I haven't narrowed to which testcase in LTP is
causing the failure but would guess it's likely to be either fsx or
fsstress.  We can work on isolating these, if it might be useful?

Thanks,
Bryce

----- Forwarded message from Jason Neighbors <jasonn at osdl.org> -----

Date: Thu, 14 Sep 2006 14:39:57 -0700
From: Jason Neighbors <jasonn at osdl.org>
To: bryce at osdl.org
Subject: 2.6.18-rc6-CITI_NFS4_ALL-1

Hey,
don't remember if you already looked at it, but the latest CITI kernel seems to have a NULL pointer dereference issue:

http://crucible.osdl.org/runs/2058/report/01-consoles.txt

http://crucible.osdl.org/runs/2004/report/01-consoles.txt
http://crucible.osdl.org/runs/2004/sysinfo/nfs03.console

Didn't see it mentioned in your report for 2004.
Guess it only appears on 32-bit runs, so maybe why the ffsb scores were low, etc.

-- 
Jason Neighbors
x1939

----- End forwarded message -----