Re: [PATCH v2] mm/slub: Simplify __kmem_cache_alias()

[Hyeonggon Yoo]

On Fri, Jun 03, 2022 at 10:35:55PM +0800, sxwjean@xxxxxx wrote:
> From: Xiongwei Song <xiongwei.song@xxxxxxxxxxxxx>
> 
> There is no need to do anything if sysfs_slab_alias() return nonzero
> value after getting a mergeable cache.
> 
> Signed-off-by: Xiongwei Song <xiongwei.song@xxxxxxxxxxxxx>
> Reviewed-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> ---
> v2: Collect Reviewed-by tag from Muchun.
> ---
>  mm/slub.c | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
> 
> diff --git a/mm/slub.c b/mm/slub.c
> index d8d5abf49f5f..9444277d669a 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -4861,6 +4861,9 @@ __kmem_cache_alias(const char *name, unsigned int size, unsigned int align,
>  
>  	s = find_mergeable(size, align, flags, name, ctor);
>  	if (s) {
> +		if (sysfs_slab_alias(s, name))
> +			return NULL;
> +
>  		s->refcount++;
>

I think we should not expose sysfs attributes before initializing
what can be read via sysfs attribute (object_size).

>  		/*
> @@ -4869,11 +4872,6 @@ __kmem_cache_alias(const char *name, unsigned int size, unsigned int align,
>  		 */
>  		s->object_size = max(s->object_size, size);

this calculation should be done before sysfs_slab_alias().

Thanks,
Hyeonggon

>  		s->inuse = max(s->inuse, ALIGN(size, sizeof(void *)));
> -
> -		if (sysfs_slab_alias(s, name)) {
> -			s->refcount--;
> -			s = NULL;
> -		}
>  	}
>  
>  	return s;
> -- 
> 2.30.2
>