Re: [PATCH] patch-slab-split-08-reap

[Andrew Morton <akpm@digeo.com>]

Manfred Spraul wrote:
> 
>  int __init cpucache_init(void)
>  {
>         struct list_head* p;
> +       int i;
> 
>         down(&cache_chain_sem);
>         g_cpucache_up = FULL;
> 
> -       p = &cache_cache.next;
> -       do {
> +       list_for_each(p, &cache_chain) {
>                 kmem_cache_t* cachep = list_entry(p, kmem_cache_t, next);
>                 enable_cpucache(cachep);
>                 p = cachep->next.next;
> -       } while (p != &cache_cache.next);
> +       }

You're only visiting every second member in this list walk.
I added the below diff which gets me to a login prompt.  I need
to get another rollup out - I'll beat on the slab changes a bit
and include them if they look solid; otherwise I'll upload an
incremental diff for them.

--- 2.5.40/mm/slab.c~cpucache_init-fix	Sat Oct  5 17:06:28 2002
+++ 2.5.40-akpm/mm/slab.c	Sat Oct  5 17:06:28 2002
@@ -697,17 +697,14 @@ void __init kmem_cache_sizes_init(void)
 
 int __init cpucache_init(void)
 {
-	struct list_head* p;
+	kmem_cache_t *cachep;
 	int i;
 
 	down(&cache_chain_sem);
 	g_cpucache_up = FULL;
 
-	list_for_each(p, &cache_chain) {
-		kmem_cache_t* cachep = list_entry(p, kmem_cache_t, next);
+	list_for_each_entry(cachep, &cache_chain, next)
 		enable_cpucache(cachep);
-		p = cachep->next.next;
-	}
 	
 	/* 
 	 * Register the timers that return unneeded

.