Re: [PATCH] sched_clock: prevent scd->clock from moving backwards

[Ingo Molnar <mingo@elte.hu>]

* Dave Kleikamp <shaggy@linux.vnet.ibm.com> wrote:

> On Thu, 2008-10-09 at 17:17 +0200, Ingo Molnar wrote:
> 
> > hm, -tip testing found a sporadic hard lockup during bootup, and i've 
> > bisected it back to this patch. They happened on 64-bit test-systems. 
> > I've attached the .config that produced the problem.
> > 
> > i reverted the patch and the lockups went away. But i cannot see what's 
> > wrong with it ...
> 
> I could have sworn I ran with the patch, but maybe I got my patch queue
> messed up and never tested it right.
> 
> I think I see the problem.
> 
> --- a/kernel/sched_clock.c
> +++ b/kernel/sched_clock.c
> @@ -118,13 +118,13 @@ static u64 __update_sched_clock(struct
> sched_clock_data *scd, u64 now)
>  
>         /*
>          * scd->clock = clamp(scd->tick_gtod + delta,
> -        *                    max(scd->tick_gtod, scd->clock),
> -        *                    scd->tick_gtod + TICK_NSEC);
> +        *                    max(scd->tick_gtod, scd->clock),
> +        *                    min(scd->clock, scd->tick_gtod +
> TICK_NSEC));
>          */
>  
>         clock = scd->tick_gtod + delta;
>         min_clock = wrap_max(scd->tick_gtod, scd->clock);
> -       max_clock = scd->tick_gtod + TICK_NSEC;
> +       max_clock = wrap_min(scd->clock, scd->tick_gtod + TICK_NSEC);
>  
>         clock = wrap_max(clock, min_clock);
>         clock = wrap_min(clock, max_clock);
> 
> We want wrap_max(scd->clock, scd->tick_gtod + TICK_NSEC), not
> wrap_min(). [...]

ah, so the lockup bug was probably that sched_clock() was never going 
forwards properly so some task was scheduled forever and livelocked the 
system?

> [...] The problem I am trying to fix is that scd->tick_gtod + 
> TICK_NSEC may be too low.  The upper bound needs to be at LEAST 
> scd->clock.  Limiting it to scd->clock all the time is disastrous.
> :-)
> 
> I'll fix the patch and retest it before sending it again.
> 
> Sorry about my sloppiness.

no problem - and it's good that our bad-patch filters worked properly 
and efficiently :-)

	Ingo