Re: [patch] preempt-trace.patch (mono preempt-trace)

From: Dominik Karall
Date: Fri Aug 05 2005 - 09:25:37 EST


On Friday 05 August 2005 12:48, Ingo Molnar wrote:
> * Andrew Morton <akpm@xxxxxxxx> wrote:
> > I think Ingo was planning on coming up with some infrastructure which
> > would allow us to debug this further.
>
> yeah. I've done this today and have split it out of the -RT tree, see
> the patch below. After some exposure in -mm i'd like this feature to go
> upstream too.
>
> the patch is against recent Linus trees, 2.6.13-rc4 or later should all
> work. Dominik, could you try it and send us the new kernel logs whenever
> you happen to hit that warning message again? (Please also enable
> CONFIG_KALLSYMS_ALL, so that we get as much symbolic data as possible.)

Here's a preempt trace output from mono. To compile preempt-trace.patch I
remove the traps.c patch and added u32 definition for out_count in handle.c.
After those changes, the kernel compiled fine.

Now here's the output, let me know if it is ok, or if you can make any reveals
where the bug is located.

BUG: mono[10011] exited with nonzero preempt_count 1!
---------------------------
| preempt count: 00000001 ]
| 1 level deep critical section nesting:
----------------------------------------
.. [<ffffffff803f791e>] .... _spin_lock+0xe/0x70
.....[<0000000000000000>] .. ( <= 0x0)

If there is anything I should test, let me know!

dominik

Attachment: pgp00000.pgp
Description: PGP signature