mm/zswap: bugfix: memory leak when re-swapon
authorWeijie Yang <weijie.yang@samsung.com>
Wed, 16 Oct 2013 20:46:54 +0000 (13:46 -0700)
committerMarek Szyprowski <m.szyprowski@samsung.com>
Thu, 15 May 2014 05:28:10 +0000 (07:28 +0200)
zswap_tree is not freed when swapoff, and it got re-kmalloced in swapon,
so a memory leak occurs.

Free the memory of zswap_tree in zswap_frontswap_invalidate_area().

Signed-off-by: Weijie Yang <weijie.yang@samsung.com>
Reviewed-by: Bob Liu <bob.liu@oracle.com>
Cc: Minchan Kim <minchan@kernel.org>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: <stable@vger.kernel.org>
From: Weijie Yang <weijie.yang@samsung.com>
Subject: mm/zswap: bugfix: memory leak when invalidate and reclaim occur concurrently

Consider the following scenario:
thread 0: reclaim entry x (get refcount, but not call zswap_get_swap_cache_page)
thread 1: call zswap_frontswap_invalidate_page to invalidate entry x.
finished, entry x and its zbud is not freed as its refcount != 0
now, the swap_map[x] = 0
thread 0: now call zswap_get_swap_cache_page
swapcache_prepare return -ENOENT because entry x is not used any more
zswap_get_swap_cache_page return ZSWAP_SWAPCACHE_NOMEM
zswap_writeback_entry do nothing except put refcount
Now, the memory of zswap_entry x and its zpage leak.

Modify:
 - check the refcount in fail path, free memory if it is not referenced.

 - use ZSWAP_SWAPCACHE_FAIL instead of ZSWAP_SWAPCACHE_NOMEM as the fail path
   can be not only caused by nomem but also by invalidate.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Weijie Yang <weijie.yang@samsung.com>
Reviewed-by: Bob Liu <bob.liu@oracle.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: <stable@vger.kernel.org>
Acked-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Change-Id: I4a875e48714d73bf2c1f75b60d90776365c047de
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/zswap.c

index 841e35f..d93510c 100644 (file)
@@ -804,6 +804,10 @@ static void zswap_frontswap_invalidate_area(unsigned type)
        }
        tree->rbroot = RB_ROOT;
        spin_unlock(&tree->lock);
+
+       zbud_destroy_pool(tree->pool);
+       kfree(tree);
+       zswap_trees[type] = NULL;
 }
 
 static struct zbud_ops zswap_zbud_ops = {