On Sun, 2007-09-30 at 09:20 +0200, Ingo Molnar wrote:
quoted text
> * Jan Luebbe <jluebbe@lasnet.de> wrote:
> > From: Jan Lübbe <jluebbe@lasnet.de>
> > 
> > The new behaviour of CFS exposes a race which occurs if a switch is
> > requested when vt_mode.mode is VT_PROCESS.
> > 
> > The process with vc->vt_pid is signaled before vc->vt_newvt is set. This
> > causes the switch to fail when triggered by the monitoing process
> > because the target is still -1.
> > 
> > Signed-off-by: Jan Lübbe <jluebbe@lasnet.de>
> 
> nifty! I'm wondering what the symptoms of this bug were - how did you 
> notice it? Also, i'm wondering how you managed to debug this - looks 
> quite hard to trigger this race, right?
> 
> Acked-by: Ingo Molnar <mingo@elte.hu>
> 
> 	Ingo

I'm currenly porting the OpenMoko patchset to 2.6.23-rc8. We are using
psplash and kdrive. When booting, I got an error from psplash about
being unable to switch consoles. Strangely, it was possible to start
kdrive manually. So i set up qemu for kernel debugging, set some
interesting breakpoints and then saw that it was scheduling away
directly after the kill_pid. Then it was quite obvious.

It was triggering every time, on the device and in qemu identically.
Most people testing the RCs probably don't run psplash (or any other
program which sets vt_mode.mode to VT_PROCESS), so it wasn't noticed
earlier.

Jan

-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/