Re: [PATCH -rt] memcg: use migrate_disable()/migrate_enable( ) inmemcg_check_events()

[Yong Zhang]

On Thu, Nov 17, 2011 at 11:25:56AM +0100, Thomas Gleixner wrote:
> On Thu, 17 Nov 2011, Yong Zhang wrote:
> 
> > On Wed, Nov 16, 2011 at 06:02:42PM +0100, Thomas Gleixner wrote:
> > > On Wed, 16 Nov 2011, Steven Rostedt wrote:
> > > > On Wed, 2011-11-16 at 17:16 +0800, Yong Zhang wrote:
> > > > > Looking at commit 4799401f [memcg: Fix race condition in
> > > > > memcg_check_events() with this_cpu usage], we just want
> > > > > to disable migration. So use the right API in -rt. This
> > > > > will cure below warning.
> > > > No this won't work. Not even for -rt. If we disable migration but not
> > > > preemption, then two tasks can take this path. And the checks in
> > > > __memcg_event_check() will be corrupted because nothing is protecting
> > > > the updates from two tasks going into the same path.
> > > > 
> > > > Perhaps a local_lock would work.
> > > 
> > > Yes, that's the only sensible option for now. Untested patch below.
> > 
> > Works for me.
> 
> Johannes came up with a different solution. Could you please give it a try?

Works too :)

Thanks,
Yong

> 
> Thanks,
> 
> 	tglx
> 
> ------------->
> Subject: [patch] mm: memcg: shorten preempt-disabled section around event checks
> 
> Only the ratelimit checks themselves have to run with preemption
> disabled, the resulting actions - checking for usage thresholds,
> updating the soft limit tree - can and should run with preemption
> enabled.
> 
> Signed-off-by: Johannes Weiner <jweiner@xxxxxxxxxx>
> ---
>  mm/memcontrol.c |   73 ++++++++++++++++++++++++++----------------------------
>  1 files changed, 35 insertions(+), 38 deletions(-)
> 
> Thomas, HTH and it is probably interesting for upstream as well.
> Unfortunately, I'm in the middle of moving right now, so this is
> untested except for compiling.
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 6aff93c..8e62d3e 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -683,37 +683,32 @@ static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
>  	return total;
>  }
>  
> -static bool __memcg_event_check(struct mem_cgroup *memcg, int target)
> +static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
> +				       enum mem_cgroup_events_target target)
>  {
>  	unsigned long val, next;
>  
>  	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
>  	next = __this_cpu_read(memcg->stat->targets[target]);
>  	/* from time_after() in jiffies.h */
> -	return ((long)next - (long)val < 0);
> -}
> -
> -static void __mem_cgroup_target_update(struct mem_cgroup *memcg, int target)
> -{
> -	unsigned long val, next;
> -
> -	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
> -
> -	switch (target) {
> -	case MEM_CGROUP_TARGET_THRESH:
> -		next = val + THRESHOLDS_EVENTS_TARGET;
> -		break;
> -	case MEM_CGROUP_TARGET_SOFTLIMIT:
> -		next = val + SOFTLIMIT_EVENTS_TARGET;
> -		break;
> -	case MEM_CGROUP_TARGET_NUMAINFO:
> -		next = val + NUMAINFO_EVENTS_TARGET;
> -		break;
> -	default:
> -		return;
> +	if ((long)next - (long)val < 0) {
> +		switch (target) {
> +		case MEM_CGROUP_TARGET_THRESH:
> +			next = val + THRESHOLDS_EVENTS_TARGET;
> +			break;
> +		case MEM_CGROUP_TARGET_SOFTLIMIT:
> +			next = val + SOFTLIMIT_EVENTS_TARGET;
> +			break;
> +		case MEM_CGROUP_TARGET_NUMAINFO:
> +			next = val + NUMAINFO_EVENTS_TARGET;
> +			break;
> +		default:
> +			break;
> +		}
> +		__this_cpu_write(memcg->stat->targets[target], next);
> +		return true;
>  	}
> -
> -	__this_cpu_write(memcg->stat->targets[target], next);
> +	return false;
>  }
>  
>  /*
> @@ -724,25 +719,27 @@ static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
>  {
>  	preempt_disable();
>  	/* threshold event is triggered in finer grain than soft limit */
> -	if (unlikely(__memcg_event_check(memcg, MEM_CGROUP_TARGET_THRESH))) {
> +	if (unlikely(mem_cgroup_event_ratelimit(memcg,
> +						MEM_CGROUP_TARGET_THRESH))) {
> +		bool do_softlimit, do_numainfo;
> +
> +		do_softlimit = mem_cgroup_event_ratelimit(memcg,
> +						MEM_CGROUP_TARGET_SOFTLIMIT);
> +#if MAX_NUMNODES > 1
> +		do_numainfo = mem_cgroup_event_ratelimit(memcg,
> +						MEM_CGROUP_TARGET_NUMAINFO);
> +#endif
> +		preempt_enable();
> +
>  		mem_cgroup_threshold(memcg);
> -		__mem_cgroup_target_update(memcg, MEM_CGROUP_TARGET_THRESH);
> -		if (unlikely(__memcg_event_check(memcg,
> -			     MEM_CGROUP_TARGET_SOFTLIMIT))) {
> +		if (unlikely(do_softlimit))
>  			mem_cgroup_update_tree(memcg, page);
> -			__mem_cgroup_target_update(memcg,
> -						   MEM_CGROUP_TARGET_SOFTLIMIT);
> -		}
>  #if MAX_NUMNODES > 1
> -		if (unlikely(__memcg_event_check(memcg,
> -			MEM_CGROUP_TARGET_NUMAINFO))) {
> +		if (unlikely(do_numainfo))
>  			atomic_inc(&memcg->numainfo_events);
> -			__mem_cgroup_target_update(memcg,
> -				MEM_CGROUP_TARGET_NUMAINFO);
> -		}
>  #endif
> -	}
> -	preempt_enable();
> +	} else
> +		preempt_enable();
>  }
>  
>  static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
> -- 
> 1.7.6.4
> 
> --
> 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/

-- 
Only stand for myself
--
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/