[Serdev] kernel panic on ppc->solved upgrading the kernel

Wed Jun 8 16:52:00 UTC 2005

Hi all,

Obviously it was not a SER's failure ;) , I managed to upgrade to a newer kernel and now I can successfully run SER in both forking and non forking mode on a ppc platform.
By now no stress tests have been done so I can not say anything about stability or performance but at least it does not crash de system.
I am currently using the SYSV semaphores and if I have time I would try the new ppc optimizations.

Thanks Andrei,

Samuel

Unclassified.
>>> Samuel Osorio Calvo 06/03/05 11:31AM >>>
Unfortunately it did not work with the standard semaphores, the kernel crashes after the first call attemp.
When SER is starting the next couple of lines appear:

 0(0) ERROR: lock_set_init (SYSV): semget (..., 251, 0700) failed: Invalid argument
 0(0) INFO: semaphore arrays of size 250 allocated

However, I think it is not an "error" but just SER trying to determine the max lenght of the semaphores array, isn't it??

In fact, the first call attemp succeed but when I try to call any bash command, such as ls, the next appears (I don't know if it can help you to guess the problem.):

Oops: kernel access of bad area, sig: 11
NIP: C0011584 XER: 00000000 LR: C00114B4 SP: C4DF3EF0 REGS: c4df3e40 TRAP: 0300
   Not tainted
MSR: 00003032 EE: 0 PR: 0 FP: 1 ME: 1 IR/DR: 11
DAR: 047B5236, DSISR: 22000000
TASK = c4df2000[270] 'sh' Last syscall: 2
last math 00000000 last altivec 00000000
GPR00: C456E0B0 C4DF3EF0 C4DF2000 00000001 C4DF2384 00000000 C456E384 00000000
GPR08: C01BA78C 047B5182 C01BA2F4 00000498 24422482 100C3710 00000000 100C0000
GPR16: 00000000 00000000 100C1358 100C1408 00009032 04DF3F40 00000000 00000000
GPR24: C4DF3F50 7FFFF980 00000127 00000011 C7E1B1A0 C456E0A8 C7E1B5E0 C456E000
Call backtrace:
C0011440 C0006A2C C0003DDC 00000000 10033338 10027CE0 10026AC4
100235B8 10022FE0 10015DF8 100137C0 0FE74FC8 0000000

Do you think it can be an erroneous configuration of the kernel? lack of MMU? wrong library versions?? Might be a failure in the ssh connection between the host and the target??? I know it is really difficult to guess from the distance but any hint would be really useful.

BTW, do any of you know how to check if the kernel was effectively compiled with MMU?? I don't have currently the kernel sources or the configuration file.....so I have to guess from the installation... :(

Thanks again,

Samuel.

Unclassified.
>>> "Samuel Osorio Calvo" <samuel.osorio at nl.thalesgroup.com> 06/03/05 10:37AM >>>
Thanks, responses inline....

Samuel.

Unclassified.
>>> Andrei Pelinescu-Onciul <andrei at iptel.org> 06/03/05 02:10AM >>>
>On Jun 01, 2005 at 11:09, Samuel Osorio Calvo <samuel.osorio at nl.thalesgroup.com> wrote:
>> Hi!
>> 
>> Recently I had the opportunity to experiment porting SER to a power PC architecture (motorola 8260) and I successfully compiled it with the fastlocking changes for power PC commited by Andrei recently.
>> 
>> The issue is that I can make SER work in non fork mode but when I set to to forking mode, an access violation occurs, and a kernel panic pops up:
>> 
>> INIT: PANIC: segmentation violation! sleeping for 30 seconds
>
>Did it work with the older code?

In fact I first tried with the old code and then I upgrade the file, and with both SER gave the same error.

>You could also try to compile it without the assembly locks:
>make proper; make use_fast_lock=no all
I'll give it a try and report it...

>The code seems to be ok on normal powerpcs. I've tried it on an emulator
>(pearpc) running darwin.

>Andrei

_______________________________________________
Serdev mailing list
serdev at lists.iptel.org 
http://lists.iptel.org/mailman/listinfo/serdev