Re: PROBLEM: kmem_cache_create: duplicate cache fat_cache

[OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>]

Manfred Spraul <manfred@colorfullife.com> writes:

> OGAWA Hirofumi wrote:
>
>>Ummm... why is this normal situation? Didn't you run the
>>modules_install after changed .config?  Anyway, this patch returns
>>NULL instead of calling BUG().  This case seems to also happen with
>>user error.
>>
> Either return NULL or ignore the collision. I don't know what's the
> better solution.
>
> I'm open to either approach - it depends on what the
> kmem_cache_create() caller expects.

> --- 2.6/mm/slab.c	2005-06-05 17:29:01.000000000 +0200
> +++ build-2.6/mm/slab.c	2005-06-06 21:34:38.000000000 +0200
> @@ -1480,9 +1480,14 @@
>  			} 	
>  			if (!strcmp(pc->name,name)) { 
>  				printk("kmem_cache_create: duplicate cache %s\n",name); 
> -				up(&cache_chain_sem); 
> -				unlock_cpu_hotplug();
> -				BUG(); 
> +				/* This is a severe bug, because it breaks
> +				 * tuning by writing to /proc/slabinfo.
> +				 * But everything else works, and since
> +				 * duplicate caches typically happen if
> +				 * someone inserts a module twice, we'll
> +				 * continue.
> +				 */
> +				WARN_ON(1); 
>  			}	
>  		}
>  		set_fs(old_fs);

Ah, I see. I think this is friendly to developers and probably more
proper than my patch for this problem.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>