FS#31252 - [irqbalance] segmentation fault caused by applied patch

Attached to Project: Arch Linux
Opened by Gerardo Exequiel Pozzi (djgera) - Wednesday, 22 August 2012, 20:44 GMT
Last edited by Dan McGee (toofishes) - Tuesday, 04 September 2012, 23:48 GMT
Task Type Bug Report
Category Upstream Bugs
Status Closed
Assigned To Dan McGee (toofishes)
Architecture All
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description: segmentation fault, recompiling without the patch fixes the issue.

irqbalance[27051]: segfault at 7d0 ip b761aa22 sp bfba3810 error 4 in libglib-2.0.so.0.3200.4[b75b9000+f7000]

Additional info:
irqbalance-1.0.3.20110809-1 (1.0.3-1 works fine)
Recompiled with -g and gdb full backtrace, and running in debug mode attached.

Steps to reproduce:
Run irqbalance then wait fews seconds
This task depends upon

Closed by  Dan McGee (toofishes)
Tuesday, 04 September 2012, 23:48 GMT
Reason for closing:  Fixed
Additional comments about closing:  1.0.4
Comment by Dan McGee (toofishes) - Wednesday, 22 August 2012, 21:03 GMT
Is there a possibility you can report this upstream? The patch was simply the diff between the last released version and git master as of 2011-08-09.

http://code.google.com/p/irqbalance/
You will probably want to get in touch with Neil Horman; if you clone the git repo you'll have his email address from the commit messages. Thanks!
Comment by Gerardo Exequiel Pozzi (djgera) - Wednesday, 22 August 2012, 21:25 GMT
  • Field changed: Category (Packages: Extra → Upstream Bugs)
OK, I bisected it, and found the bad commit, sending mail...

$ git bisect bad
fd24d8f30012a02f7d107174bb460eb641d8cbce is the first bad commit
commit fd24d8f30012a02f7d107174bb460eb641d8cbce
Author: Neil Horman <nhorman@tuxdriver.com>
Date: Thu Aug 9 14:18:23 2012 -0400

Improve rescan ability for newly allocated interrupts

When an interrupt is registered after irqbalance starts we just add it as a misc
interrupt, even if its something that we want to balance. Instead, lets add the
irq to a temp list, force a rescan, then make sure the newly discovered irqs get
added properly on the rescan.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>

:100644 100644 3c76cea3d4b7545f38e31a91960ed7efacd6dea0 1ae4ff99da976f7c3c4e21d8d20e65ae709c86c1 M classify.c
:100644 100644 636a69ea7001168b117cf03b992c8f75597eb059 e46f31fe9ec5ea332f40046d9950622082f99142 M irqbalance.h
:100644 100644 f1d67456f9e96f07edc655f539f06b156b03575a 431fffa0c6f3f694b270cd0d9fab0601d38f40c0 M procinterrupts.c
Comment by Gerardo Exequiel Pozzi (djgera) - Wednesday, 22 August 2012, 21:30 GMT
@Dan: I guess you made a typo 2011 should be 2012 ;)
Comment by Dan McGee (toofishes) - Wednesday, 22 August 2012, 21:31 GMT
Haha yes, thanks Gerardo.
Comment by Gerardo Exequiel Pozzi (djgera) - Thursday, 23 August 2012, 15:12 GMT
On 08/23/2012 08:09 AM, Neil Horman wrote:
> On Wed, Aug 22, 2012 at 06:29:20PM -0300, Gerardo Exequiel Pozzi wrote:
>> Hello Neil
>>
>> irqbalance crashes on my system, bad commit is
>> fd24d8f30012a02f7d107174bb460eb641d8cbce
>>
>> For more details please see here [#1] (gdb backtrace, and running in
>> debug mode)
>>
>> [#1] https://bugs.archlinux.org/task/31252
>>
>> Good Luck!
>>
> Can you please open a bug at code.google.com, as the arch linux bug indicates
> please?
>
> Also, I'm unable to reproduce your problem. Given that the crash happens in
> g_slice_alloc, I'm guessing that the problem may be a memory corruptor. Could
> you please run the following:
> G_SLICE=always-malloc valgrind --tool=memcheck /usr/sbin/irqbalance --debug
>
> You'll need to disable the irqbalance service when you do this of course.
>
> thanks
> Neil
>

Done: http://code.google.com/p/irqbalance/issues/detail?id=38
Comment by Gerardo Exequiel Pozzi (djgera) - Friday, 24 August 2012, 17:00 GMT

Loading...