Re: [PATCH] Locking optimization for cache_reap

[Andrew Morton <akpm@osdl.org>]

Dimitri Sivanich <sivanich@sgi.com> wrote:
>
> Here is another cache_reap optimization that reduces latency when
> applied after the 'Move cache_reap out of timer context' patch I
> submitted on 7/14 (for inclusion in -mm next week).
> 
> This applies to 2.6.8-rc2 + the above mentioned patch.

How does it "reduce latency"?

It looks like a reasonable cleanup, but afaict it will result in the
per-cache spinlock actually being held for longer periods, thus increasing
latencies???

> 
> 
> Index: linux/mm/slab.c
> ===================================================================
> --- linux.orig/mm/slab.c
> +++ linux/mm/slab.c
> @@ -2619,27 +2619,6 @@ static void enable_cpucache (kmem_cache_
>  					cachep->name, -err);
>  }
>  
> -static void drain_array(kmem_cache_t *cachep, struct array_cache *ac)
> -{
> -	int tofree;
> -
> -	check_irq_off();
> -	if (ac->touched) {
> -		ac->touched = 0;
> -	} else if (ac->avail) {
> -		tofree = (ac->limit+4)/5;
> -		if (tofree > ac->avail) {
> -			tofree = (ac->avail+1)/2;
> -		}
> -		spin_lock(&cachep->spinlock);
> -		free_block(cachep, ac_entry(ac), tofree);
> -		spin_unlock(&cachep->spinlock);
> -		ac->avail -= tofree;
> -		memmove(&ac_entry(ac)[0], &ac_entry(ac)[tofree],
> -					sizeof(void*)*ac->avail);
> -	}
> -}
> -
>  static void drain_array_locked(kmem_cache_t *cachep,
>  				struct array_cache *ac, int force)
>  {
> @@ -2697,16 +2676,14 @@ static void cache_reap (void *unused)
>  			goto next;
>  
>  		check_irq_on();
> -		local_irq_disable();
> -		drain_array(searchp, ac_data(searchp));
>  
> -		if(time_after(searchp->lists.next_reap, jiffies))
> -			goto next_irqon;
> +		spin_lock_irq(&searchp->spinlock);
> +
> +		drain_array_locked(searchp, ac_data(searchp), 0);
>  
> -		spin_lock(&searchp->spinlock);
> -		if(time_after(searchp->lists.next_reap, jiffies)) {
> +		if(time_after(searchp->lists.next_reap, jiffies))
>  			goto next_unlock;
> -		}
> +
>  		searchp->lists.next_reap = jiffies + REAPTIMEOUT_LIST3;
>  
>  		if (searchp->lists.shared)
> @@ -2739,9 +2716,7 @@ static void cache_reap (void *unused)
>  			spin_lock_irq(&searchp->spinlock);
>  		} while(--tofree > 0);
>  next_unlock:
> -		spin_unlock(&searchp->spinlock);
> -next_irqon:
> -		local_irq_enable();
> +		spin_unlock_irq(&searchp->spinlock);
>  next:
>  		;
>  	}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>