On Tue, 2017-05-23 at 22:08 +0200, Greg Kroah-Hartman wrote:
> 4.4-stable review patch. If anyone has any objections, please let me know.
>
> ------------------
>
> From: Dennis Yang <dennisyang@xxxxxxxx>
>
> commit 583da48e388f472e8818d9bb60ef6a1d40ee9f9d upstream.
>
> When growing raid5 device on machine with small memory, there is chance that
> mdadm will be killed and the following bug report can be observed. The same
> bug could also be reproduced in linux-4.10.6.
[...]
> The problem is that resize_stripes() releases new stripe_heads before assigning new
> slab cache to conf->slab_cache. If the shrinker function raid5_cache_scan() gets called
> after resize_stripes() starting releasing new stripes but right before new slab cache
> being assigned, it is possible that these new stripe_heads will be freed with the old
> slab_cache which was already been destoryed and that triggers this bug.
[...]
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -2232,6 +2232,10 @@ static int resize_stripes(struct r5conf
> err = -ENOMEM;
>
> mutex_unlock(&conf->cache_size_mutex);
> +
> + conf->slab_cache = sc;
> + conf->active_name = 1-conf->active_name;
> +
> /* Step 4, return new stripes to service */
> while(!list_empty(&newstripes)) {
> nsh = list_entry(newstripes.next, struct stripe_head, lru);
[...]
The assignments are still being done after conf->cache_size_mutex is
unlocked, so there still seems to be a race with raid5_cache_scan().
Shouldn't they be moved above the mutex_unlock()?
Ben.
--
Ben Hutchings
Software Developer, Codethink Ltd.