Re: [PATCH] Re: AMD Mobile Semprons (3500+, 3600+,...) break with nohz and highres enabled

[Thomas Renninger <trenn@suse.de>]

On Wednesday 16 July 2008 09:03:12 pm you wrote:
> On Wed, 16 Jul 2008, Thomas Renninger wrote:
> > This is against Andi's (ak) release branch,
> > hope that is ok...
>
> No idea what that is.
The branch Andi will merge his ACPI stuff into 2.6.27.
>
> > Can this one be reviewed/added, pls.
>
> Reviewed yes. Added no.
>
> >  kernel/time/tick-sched.c |   31 +++++++++++++++++++++++++++++++
> >  1 files changed, 31 insertions(+), 0 deletions(-)
> >
> > diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> > index b854a89..bda0164 100644
> > --- a/kernel/time/tick-sched.c
> > +++ b/kernel/time/tick-sched.c
> > @@ -589,6 +589,32 @@ static enum hrtimer_restart tick_sched_timer(struct
> > hrtimer *timer)
> >  	return HRTIMER_RESTART;
> >  }
> >
> > +#if defined(CONFIG_X86) && defined(CONFIG_NO_HZ)
> > +static __init int is_nohz_broken(void)
> > +{
>
> We definitely do not add x86 quirks into the generic code.
So the right solution (if nobody cares about this one) is to set 
tick_no_hz_enabled global and disable it through:
arch/x86/kernel/quirks.c
is that correct?
>
> Is there really no other way to fix this problem ?
I have no idea.
Richard's last message:
"I can help out as needed, just a work deadline I have to meet so I 
haven't been able to dig further.  Just point me in the direction and 
I'll gladly be someones hands on the beastly machine"

has silently be ignored.

> Which chipset is 
> involved here ? SB400/600 perhaps ?
No, it is probably much older and I am pretty sure that it is not chipset 
related, but related to the "Mobile" in the Mobile Sempron as this exactly 
seem to hit 3500 and 3600 Mobile Sempron CPUs.

I can't help much further because:
  - I have not such a machine
  - I don't know these parts. I hoped to get some hints to help these guys

       Thomas