lkml.org 
[lkml]   [2011]   [Oct]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    SubjectRe: [PATCH v2 5/6] slub: Only IPI CPUs that have per cpu obj to flush
    From
    On Fri, Oct 28, 2011 at 6:06 AM, Christoph Lameter <cl@gentwo.org> wrote:
    > On Sun, 23 Oct 2011, Gilad Ben-Yossef wrote:
    >
    >> diff --git a/include/linux/slub_def.h b/include/linux/slub_def.h
    >> index f58d641..b130f61 100644
    >> --- a/include/linux/slub_def.h
    >> +++ b/include/linux/slub_def.h
    >> @@ -102,6 +102,9 @@ struct kmem_cache {
    >>        */
    >>       int remote_node_defrag_ratio;
    >>  #endif
    >> +
    >> +     /* Which CPUs hold local slabs for this cache. */
    >> +     cpumask_var_t cpus_with_slabs;
    >>       struct kmem_cache_node *node[MAX_NUMNODES];
    >>  };
    >
    > Please do not add fields to structures for passing parameters to
    > functions. This just increases the complexity of the patch and extends a
    > structures needlessly.

    The field was added to provide storage to cpus_with_slabs during
    flush_all, since otherwise cpus_with_slabs, being a cpumask, would
    require a kmem_cache allocation in the middle of flush_all for
    CONFIG_CPUMASK_OFF_STACK=y case, which Pekka E. objected to.

    The next patch in the series makes the field (and overhead) only added
    for CONFIG_CPUMASK_OFF_STACK=y case but I wanted to break out the
    addition to the patch core feature and the optimization of only adding
    the field for CONFIG_CPUMASK_OFF_STACK=y, so this patch as is without
    the next one is only good for bisect value.

    I should have probably have commented about it in the description of
    this patch and not only in the next one. Sorry about that. I will fix
    it for the next round.

    >
    >> diff --git a/mm/slub.c b/mm/slub.c
    >> index 7c54fe8..f8cbf2d 100644
    >> --- a/mm/slub.c
    >> +++ b/mm/slub.c
    >> @@ -1948,7 +1948,18 @@ static void flush_cpu_slab(void *d)
    >>
    >>  static void flush_all(struct kmem_cache *s)
    >>  {
    >> -     on_each_cpu(flush_cpu_slab, s, 1);
    >> +     struct kmem_cache_cpu *c;
    >> +     int cpu;
    >> +
    >> +     for_each_online_cpu(cpu) {
    >> +             c = per_cpu_ptr(s->cpu_slab, cpu);
    >> +             if (c && c->page)
    >> +                     cpumask_set_cpu(cpu, s->cpus_with_slabs);
    >> +             else
    >> +                     cpumask_clear_cpu(cpu, s->cpus_with_slabs);
    >> +     }
    >> +
    >> +     on_each_cpu_mask(s->cpus_with_slabs, flush_cpu_slab, s, 1);
    >>  }
    >
    >
    > You do not need s->cpus_with_slabs to be in kmem_cache. Make it a local
    > variable instead.

    That is what the next patch does - for CONFIG_CPUMASK_OFFSTACK=n, alt least.

    Thanks,
    Gilad


    --
    Gilad Ben-Yossef
    Chief Coffee Drinker
    gilad@benyossef.com
    Israel Cell: +972-52-8260388
    US Cell: +1-973-8260388
    http://benyossef.com

    "I've seen things you people wouldn't believe. Goto statements used to
    implement co-routines. I watched C structures being stored in
    registers. All those moments will be lost in time... like tears in
    rain... Time to die. "
    --
    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/

    \
     
     \ /
      Last update: 2011-10-28 10:53    [W:0.026 / U:0.412 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site