Re: [PATCH RT] mm/slub: close possible memory-leak in kmem_cache_alloc_bulk()

From: Jesper Dangaard Brouer
Date: Wed Dec 13 2017 - 09:47:10 EST


On Wed, 13 Dec 2017 15:05:55 +0100
Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> wrote:

> Under certain circumstances we could leak elements which were moved to
> the local "to_free" list. The damage is limited since I can't find any
> users here.
>
> Cc: stable-rt@xxxxxxxxxxxxxxx
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> ---
> Jesper: There are no users of kmem_cache_alloc_bulk() and kfree_bulk().
> Only kmem_cache_free_bulk() is used since it was introduced. Do you
> think that it would make sense to remove those?

I would like to keep them.

Rao Shoaib (Cc'ed) is/was working on a patchset for RCU-bulk-free that
used the kfree_bulk() API.

I plan to use kmem_cache_alloc_bulk() in the bpf-map "cpumap", for bulk
allocating SKBs during dequeue of XDP frames. (My original bulk alloc
SKBs use-case during NAPI/softirq was never merged).


> mm/slub.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index ffd2fa0f415e..9053e929ce9d 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -3240,6 +3240,7 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
> return i;
> error:
> local_irq_enable();
> + free_delayed(&to_free);
> slab_post_alloc_hook(s, flags, i, p);
> __kmem_cache_free_bulk(s, i, p);
> return 0;

I've not seen free_delayed() before... and my cscope cannot find it...

--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat
LinkedIn: http://www.linkedin.com/in/brouer