Re: [v7 7/8] mm: only IPI CPUs to drain local pages if they exist

[Mel Gorman]

On Mon, Jan 30, 2012 at 05:14:37PM +0200, Gilad Ben-Yossef wrote:
> >> +     for_each_online_cpu(cpu) {
> >> +             bool has_pcps = false;
> >> +             for_each_populated_zone(zone) {
> >> +                     pcp = per_cpu_ptr(zone->pageset, cpu);
> >> +                     if (pcp->pcp.count) {
> >> +                             has_pcps = true;
> >> +                             break;
> >> +                     }
> >> +             }
> >> +             if (has_pcps)
> >> +                     cpumask_set_cpu(cpu, &cpus_with_pcps);
> >> +             else
> >> +                     cpumask_clear_cpu(cpu, &cpus_with_pcps);
> >> +     }
> >
> > Lets take two CPUs running this code at the same time. CPU 1 has per-cpu
> > pages in all zones. CPU 2 has no per-cpu pages in any zone. If both run
> > at the same time, CPU 2 can be clearing the mask for CPU 1 before it has
> > had a chance to send the IPI. This means we'll miss sending IPIs to CPUs
> > that we intended to.
> 
> I'm confused. You seem to be assuming that each CPU is looking at its own pcps
> only (per zone).

/me slaps self

I was assuming exactly this.

> Assuming no change in the state of the pcps when both CPUs
> run this code at the same time, both of them should mark the bit for
> their respective
> CPUs the same, unless one of them raced and managed to send the IPI to clear
> pcps from the other, at which point you might see one of them send a
> spurious IPI
> to drains pcps that have been drained - but that isn't bad.
> 

Indeed, the race is tiny and the consequences are not important.

> At least, that is what I meant the code to do and what I believe it
> does. What have I
> missed?
> 

Nothing, the problem was on my side. Sorry for the noise.

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