Re: [BUG] local_softirq_pending storm

[Anant Nitya <kernel@prachanda.info>]

On Friday 18 May 2007 18:31:17 Thomas Gleixner wrote:
> On Thu, 2007-05-17 at 12:11 +0530, Anant Nitya wrote:
> > On Friday 11 May 2007 03:28:46 Thomas Gleixner wrote:
> > > Ok, that's consistent with earlier reports. The problem surfaces when
> > > one of the SMT-"cpus" goes idle. The problem goes away when you disable
> > > hyperthreading.
> >
> > Yes with HT disabled in BIOS there is no local_softirq_pending messages.
> > BTW why does this problem persist only with X ?
>
> No idea. I uploaded a debug patch against 2.6.22-rc1 to
>
> http://www.tglx.de/private/tglx/2.6.22-rc1-hrt-debug.patch
>
> Can you give it a try and report the output ?

I am compiling kernel with above patch applied and will post the results.

> > > When you apply the ratelimit patch, does the softlockup problem
> > > persist ?
> >
> > Yes, though softlockup is rare and mostly hit when system is under high
> > load. Apart of that I am also getting following messages consistently
> > across multiple boot cycles with NOHZ=y and ratelimit patch applied.
> >
> > May 15 11:51:22 rudra kernel: [ 2594.341068] Clocksource tsc unstable
> > (delta = 28111260302 ns)
> > May 15 11:51:22 rudra kernel: [ 2594.343194] Time: acpi_pm clocksource
> > has been installed.
>
> That's informal. The TSC is detected to be unstable and replaced by the
> pm timer. Nothing to worry about. It happens with NOHZ=n as well,
> right ?
Nope it don't appear with nohz=no till now across so many boot cycles, it 
always appears with nohz!=no.

Regards
Ananitya