Hi
An otherwise fine running kernel-2.6.0-test2 repeatably gives this when
running the NPTL testsuite.
ksymoops output attached.
- kernel compiled with gcc-2.95.4 (s'pose I should try 3.2.3)
- recent binutils
- board is Tyan S2466N-4M with pair of Athlon 2200's
This is a UP kernel (trying to narrow down the cause).
Thanks
Greg
(not subscribed)
Greg Schafer <[email protected]> wrote:
>
> An otherwise fine running kernel-2.6.0-test2 repeatably gives this when
> running the NPTL testsuite.
>From where does one obtain the NPTL testsuite?
Greg Schafer wrote:
>
> An otherwise fine running kernel-2.6.0-test2 repeatably gives this when
> running the NPTL testsuite.
>
> ksymoops output attached.
>
> - kernel compiled with gcc-2.95.4 (s'pose I should try 3.2.3)
> - recent binutils
> - board is Tyan S2466N-4M with pair of Athlon 2200's
>
> This is a UP kernel (trying to narrow down the cause).
It looks like the list poisoning triggers:
ecx: 00200200 edx: 00100100
those are the poison values for the prev/next fields of lists (see
<linux/list.h>).
So it looks like switch_exec_pids() is removing a list entry that was
already removed.
Linus
On Tue, Aug 05, 2003 at 10:17:39PM -0700, Andrew Morton wrote:
> Greg Schafer <[email protected]> wrote:
> >
> > An otherwise fine running kernel-2.6.0-test2 repeatably gives this when
> > running the NPTL testsuite.
>
> From where does one obtain the NPTL testsuite?
It's part of the NPTL package which is provided as an addon to glibc:-
https://listman.redhat.com/archives/phil-list/2003-August/msg00000.html
Greg
On Tue, Aug 05, 2003 at 10:23:18PM -0700, OSDL wrote:
> It looks like the list poisoning triggers:
>
> ecx: 00200200 edx: 00100100
>
> those are the poison values for the prev/next fields of lists (see
> <linux/list.h>).
>
> So it looks like switch_exec_pids() is removing a list entry that was
> already removed.
Indeed. FWIW, a newer compiler gave the same results. This is way beyond my
debugging skills but I can reliably reproduce and am willing to test
anything you can suggest.
Greg