Skip to contentSkip to conversation

Oops in 2.6.19.1

DISCUSSION Chuck Ebbert 4 participants 5 messages Dec 20, 2006 – Dec 30, 2006

lore
Share & tools
lore
#0 lore raw

In-Reply-To: <[email protected]>

Quoted 4 lines · Alistair John Strachan

Any ideas?

BUG: unable to handle kernel NULL pointer dereference at virtual address 

00000009

    83 ca 10                  or     $0x10,%edx
    3b                        .byte 0x3b
    87 68 01                  xchg   %ebp,0x1(%eax)   <=====
    00 00                     add    %al,(%eax)

Somehow it is trying to execute code in the middle of an instruction. That almost never works, even when the resulting fragment is a legal opcode. :)

The real instruction is:

    3b 87 68 01 00 00 00        cmp    0x168(%edi),%eax

I'd guess you have some kind of hardware problem. It could also be a kernel problem where the saved address was corrupted during an interrupt, but that's not likely.

Signature · 1 line
MBTI: IXTP
#1 lore raw

On Wednesday 20 December 2006 20:48, Chuck Ebbert wrote:
[snip]

Quoted 3 lines

I'd guess you have some kind of hardware problem. It could also be a kernel problem where the saved address was corrupted during an interrupt, but that's not likely.

Seems pretty unlikely on a 4 year old Via Epia. Never had any problems with it before now.

Maybe a cosmic ray event? ;-)

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.
#2 lore raw
Quoted 4 lines · Alistair John Strachan

Seems pretty unlikely on a 4 year old Via Epia. Never had any problems with it before now.

Maybe a cosmic ray event? ;-)

More likely a stray alpha particle from a radioactive decay in the actual chip casing - I saw some research a while back that said that the average commodity system should *expect* to see 1 or 2 alpha-induced single-bit errors per year, and the chance that *you* saw the event was directly related to whether the memory had ECC, and how much of the other circuitry had ECC on it....

#3 lore raw
In reply to Chuck Ebbert
Quoted 17 lines · Chuck Ebbert

In-Reply-To: <[email protected]>

Quoted 3 lines · Alistair John Strachan

Any ideas?

BUG: unable to handle kernel NULL pointer dereference at virtual address 00000009
    83 ca 10                  or     $0x10,%edx
    3b                        .byte 0x3b
    87 68 01                  xchg   %ebp,0x1(%eax)   <=====
    00 00                     add    %al,(%eax)

Somehow it is trying to execute code in the middle of an instruction. That almost never works, even when the resulting fragment is a legal opcode. :)

Quoted 7 lines

The real instruction is:

    3b 87 68 01 00 00 00        cmp    0x168(%edi),%eax

I'd guess you have some kind of hardware problem. It could also be a kernel problem where the saved address was corrupted during an interrupt, but that's not likely.

This looks rather strange.
The times I have seen this sort of problem is:
1) when one bit of the kernel is corrupting another part of it.
2) Kernel modules compiled with different gcc than rest of kernel.
3) kernel headers do not match the kernel being used.

One way to start tracking this down would be to run it with the fewest amount of kernel modules loaded as one can, but still reproduce the problem.

James

#4 lore raw
Quoted 5 lines · James Courtier-Dutton
Quoted 3 lines

I'd guess you have some kind of hardware problem. It could also be a kernel problem where the saved address was corrupted during an interrupt, but that's not likely.

This looks rather strange.

[snip]

Quoted 1 line

2) Kernel modules compiled with different gcc than rest of kernel.

Previously there was only one GCC version (4.1.1 totally replaced 3.4.3, and is the system wide GCC), now I have installed 3.4.6 into /opt/gcc-3.4.6 and it is only PATH'ed explicitly by me when I wish to compile a kernel using it:

export PATH=/opt/gcc-3.4.6/bin:$PATH
cp /boot/config-2.6.19-test .config
make oldconfig
make

Quoted 1 line

3) kernel headers do not match the kernel being used.

The tree is a pristine 2.6.19.

Quoted 3 lines

One way to start tracking this down would be to run it with the fewest amount of kernel modules loaded as one can, but still reproduce the problem.

Crippling the machine, though. Impractical for something that isn't immediately reproducible.

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.