Re: [PATCH 2/2] SLUB: Disable debugging if it increases the minimumpage order

[Christoph Lameter]

We have had that with SLAB. NO! This leads to the situation that some
slabs have debug on and some have not. You just do not know which.

Note that CONFIG_SLUB_DEBUG only enables the code to debug a slab. It does
not enable debugging for each slab. CONFIG_SLUB_DEBUG_ON does that.

On Thu, 11 Jun 2009, Pekka J Enberg wrote:

> From: Pekka Enberg <penberg@xxxxxxxxxxxxxx>
>
> As CONFIG_SLUB_DEBUG is almost always enabled, we need to disable debugging on
> per-cache basis if the added debug metadata increases minimum page order. This
> is a problem for large caches such as kmalloc-4096 as seen in the following
> bugzilla report:
>
>   http://bugzilla.kernel.org/show_bug.cgi?id=13319
>
> Cc: Christoph Lameter <cl@xxxxxxxxxxxxxxxxxxxx>
> Cc: Larry Finger <Larry.Finger@xxxxxxxxxxxx>
> Cc: Mel Gorman <mel@xxxxxxxxx>
> Signed-off-by: Pekka Enberg <penberg@xxxxxxxxxxxxxx>
> ---
>  mm/slub.c |   13 +++++++++++++
>  1 files changed, 13 insertions(+), 0 deletions(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 2bbacfc..45080d3 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -2385,6 +2385,16 @@ static int calculate_sizes(struct kmem_cache *s, int forced_order)
>
>  }
>
> +#define MAX_DEBUG_SIZE (3 * sizeof(void *) + 2 * sizeof(struct track))
> +
> +static bool must_disable_debug(size_t size)
> +{
> +	/*
> +	 * Disable debugging if it increases the minimum page order.
> +	 */
> +	return get_order(size + MAX_DEBUG_SIZE) > get_order(size);
> +}
> +
>  static int kmem_cache_open(struct kmem_cache *s, gfp_t gfpflags,
>  		const char *name, size_t size,
>  		size_t align, unsigned long flags,
> @@ -2397,6 +2407,9 @@ static int kmem_cache_open(struct kmem_cache *s, gfp_t gfpflags,
>  	s->align = align;
>  	s->flags = kmem_cache_flags(size, flags, name, ctor);
>
> +	if (must_disable_debug(size))
> +		s->flags &= ~(SLAB_POISON|SLAB_RED_ZONE|SLAB_STORE_USER);
> +
>  	if (!calculate_sizes(s, -1))
>  		goto error;
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/