[PATCH] - Allocate larger cache_cache if order 0 fails

[PATCH] - Allocate larger cache_cache if order 0 fails

[Jack Steiner]

kmem_cache_init() incorrectly assumes that the cache_cache object will
fit in an order 0 allocation. On very large systems, this is not
true. Change the code to try larger order allocations if order 0 fails.

	Signed-off-by: Jack Steiner <steiner@sgi.com>




Index: linux/mm/slab.c
===================================================================
--- linux.orig/mm/slab.c	2006-03-01 16:07:31.000000000 -0600
+++ linux/mm/slab.c	2006-03-03 13:21:52.000000000 -0600
@@ -1124,6 +1124,7 @@ void __init kmem_cache_init(void)
 	struct cache_sizes *sizes;
 	struct cache_names *names;
 	int i;
+	int order;
 
 	for (i = 0; i < NUM_INIT_LISTS; i++) {
 		kmem_list3_init(&initkmem_list3[i]);
@@ -1167,11 +1168,15 @@ void __init kmem_cache_init(void)
 
 	cache_cache.buffer_size = ALIGN(cache_cache.buffer_size, cache_line_size());
 
-	cache_estimate(0, cache_cache.buffer_size, cache_line_size(), 0,
-		       &left_over, &cache_cache.num);
+	for (order = 0; order < MAX_ORDER; order++) {
+		cache_estimate(order, cache_cache.buffer_size, cache_line_size(), 0,
+		       	&left_over, &cache_cache.num);
+		if (cache_cache.num)
+			break;
+	}
 	if (!cache_cache.num)
 		BUG();
-
+	cache_cache.gfporder = order;
 	cache_cache.colour = left_over / cache_cache.colour_off;
 	cache_cache.slab_size = ALIGN(cache_cache.num * sizeof(kmem_bufctl_t) +
 				      sizeof(struct slab), cache_line_size());

Re: [PATCH] - Allocate larger cache_cache if order 0 fails

[Pekka Enberg]

Hi Jack,

On 3/7/06, Jack Steiner <steiner@sgi.com> wrote:
> -       cache_estimate(0, cache_cache.buffer_size, cache_line_size(), 0,
> -                      &left_over, &cache_cache.num);
> +       for (order = 0; order < MAX_ORDER; order++) {
> +               cache_estimate(order, cache_cache.buffer_size, cache_line_size(), 0,
> +                       &left_over, &cache_cache.num);
> +               if (cache_cache.num)
> +                       break;
> +       }

Any reason why you can't use calculate_slab_order() here?

                                   Pekka

Re: [PATCH] - Allocate larger cache_cache if order 0 fails

[Jack Steiner]

On Tue, Mar 07, 2006 at 09:36:04PM +0200, Pekka Enberg wrote:
> Hi Jack,
> 
> On 3/7/06, Jack Steiner <steiner@sgi.com> wrote:
> > -       cache_estimate(0, cache_cache.buffer_size, cache_line_size(), 0,
> > -                      &left_over, &cache_cache.num);
> > +       for (order = 0; order < MAX_ORDER; order++) {
> > +               cache_estimate(order, cache_cache.buffer_size, cache_line_size(), 0,
> > +                       &left_over, &cache_cache.num);
> > +               if (cache_cache.num)
> > +                       break;
> > +       }
> 
> Any reason why you can't use calculate_slab_order() here?
> 
>                                    Pekka

I think either will work & the amount of code is about the same. 

I chose the above because it was easier to see that change had no effect 
on existing platforms.

Does anyone see a compelling reason for a different but equivalent implementation??

---
Jack

Re: [PATCH] - Allocate larger cache_cache if order 0 fails

[Pekka J Enberg]

On Tue, 7 Mar 2006, Jack Steiner wrote:
> Does anyone see a compelling reason for a different but equivalent implementation??

Less duplicate code?

			Pekka