splice: fix bad unlock_page() in error case
authorJens Axboe <jens.axboe@oracle.com>
Fri, 20 Jul 2007 13:18:12 +0000 (15:18 +0200)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Fri, 20 Jul 2007 16:07:01 +0000 (09:07 -0700)
If add_to_page_cache_lru() fails, the page will not be locked. But
splice jumps to an error path that does a page release and unlock,
causing a BUG() in unlock_page().

Fix this by adding one more label that just releases the page. This bug
was actually triggered on EL5 by gurudas pai <gurudas.pai@oracle.com>
using fio.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/splice.c

index 22496d2a73fa1ab5efe6f663cfc1a92a9a316543..0a09732180847ce6cb7d67f91477c2314296003a 100644 (file)
@@ -594,7 +594,7 @@ find_page:
                ret = add_to_page_cache_lru(page, mapping, index,
                                            GFP_KERNEL);
                if (unlikely(ret))
-                       goto out;
+                       goto out_release;
        }
 
        ret = mapping->a_ops->prepare_write(file, page, offset, offset+this_len);
@@ -650,8 +650,9 @@ find_page:
         */
        mark_page_accessed(page);
 out:
-       page_cache_release(page);
        unlock_page(page);
+out_release:
+       page_cache_release(page);
 out_ret:
        return ret;
 }