Re: [BUGFIX][PATCH] memcg: fix wrong pointer initialization at pagemigration when memcg is disabled.

From: Daisuke Nishimura
Date: Wed Oct 28 2009 - 20:59:22 EST


On Thu, 29 Oct 2009 09:30:13 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
>
> Lee Schermerhorn reported that he saw bad pointer dereference
> in mem_cgroup_end_migration() when he disabled memcg by boot option.
>
> memcg's page migration logic works as
>
> mem_cgroup_prepare_migration(page, &ptr);
> do page migration
> mem_cgroup_end_migration(page, ptr);
>
> Now, ptr is not initialized in prepare_migration when memcg is disabled
> by boot option. This causes panic in end_migration. This patch fixes it.
>
> Reported-by: Lee Schermerhorn <Lee.Schermerhorn@xxxxxx>
> Cc: Balbir Singh <balbir@xxxxxxxxxx>
> Cc: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> ---
> mm/memcontrol.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> Index: linux-2.6.32-rc5/mm/memcontrol.c
> ===================================================================
> --- linux-2.6.32-rc5.orig/mm/memcontrol.c
> +++ linux-2.6.32-rc5/mm/memcontrol.c
> @@ -1990,7 +1990,8 @@ int mem_cgroup_prepare_migration(struct
> struct page_cgroup *pc;
> struct mem_cgroup *mem = NULL;
> int ret = 0;
> -
> + /* this pointer will be checked at end_migration */
> + *ptr = NULL;
> if (mem_cgroup_disabled())
> return 0;
>
>
I thought unmap_and_move() itself initializes "mem" to NULL, but it doesn't...
I personaly prefer initializing "mem" to NULL in unmap_and_move(), but anyway
I think this patch is also correct.

Reviewed-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>

And I think we should send a fix for this bug to -stable too.


Thanks,
Daisuke Nishimura.
--
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/