Skip to contentSkip to conversation

Kernel Error / Crash

DISCUSSION werner 2 participants 3 messages Jul 20, 2007

lore
Share & tools
lore

werner

author
#0 lore raw

To: [email protected]

copied from the screen:

...
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
Starting udevd: /sbin/udevd -- daemon
------------[cut here]-------------------

Kernel BUG at mm/slub.c:2401!
Invalid opcode: 0000[#1]
SMP
Modules linked in:
CPU: 0

EIP: 0060: [<c0174ec0>] Not tainted VLI
EFLAGS: 00010046 (2.6.22-git13-i486-1mn #1)
EIP is at ksize + 0x40/0x50
.....

Obs: With 2.6.22-git11 that problem didnt occur

Werner Landgraf
[email protected]

#1 lore raw

Hi Werner,

Quoted 11 lines · werner

To: [email protected]

copied from the screen:

...
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
Starting udevd: /sbin/udevd -- daemon
------------[cut here]-------------------

Kernel BUG at mm/slub.c:2401!

That's ksize(invalid_address) -- when the address belongs to an !PageSlab page.

Quoted 6 lines

Invalid opcode: 0000[#1]
SMP

Modules linked in:
CPU: 0

EIP: 0060: [<c0174ec0>] Not tainted VLI
EFLAGS: 00010046 (2.6.22-git13-i486-1mn #1)

Why is the kernel called -git13-i486-1mn (the -1mn part)?

Is this a pure vanilla kernel or do you have additional patches applied on top of that?

Quoted 1 line

EIP is at ksize + 0x40/0x50

Hmm. This could be from krealloc(), but that's just a wild guess.

Quoted 1 line

.....

Please send us the entire oops log (including the registers contents, stack backtrace, "Code:" thing, et al.) -- it's *impossible* to debug this case, otherwise.

Quoted 4 lines

Obs: With 2.6.22-git11 that problem didnt occur

Werner Landgraf
[email protected]

Thanks,

Satyam

#2 lore raw
Quoted 15 lines · Satyam Sharma

Hi Werner,

Quoted 11 lines · werner

To: [email protected]

copied from the screen:

...
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
Starting udevd: /sbin/udevd -- daemon
------------[cut here]-------------------

Kernel BUG at mm/slub.c:2401!
Quoted 2 lines

That's ksize(invalid_address) -- when the address belongs to an !PageSlab page.

Oh, ok so this is fixed in 9550b105b8646f916862aee3ab7b25020ca14159 in mainline -git for some time now.

[ turns out the bug was in ksize() itself, so there wasn't need for a stack trace after all :-p ]