block: check for proper length of iov entries earlier in blk_rq_map_user_iov()
authorXiaotian Feng <dfeng@redhat.com>
Mon, 29 Nov 2010 09:03:55 +0000 (10:03 +0100)
committerJens Axboe <jaxboe@fusionio.com>
Mon, 29 Nov 2010 09:04:50 +0000 (10:04 +0100)
commit 9284bcf checks for proper length of iov entries in
blk_rq_map_user_iov(). But if the map is unaligned, kernel
will break out the loop without checking for the proper length.
So we need to check the proper length before the unalign check.

Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
block/blk-map.c

index 5d5dbe47c2285ee7ccb5f3ca784f9cdb8cb6df3e..e663ac2d8e68f70ff17ce274f3cebec16c1dd18c 100644 (file)
@@ -201,12 +201,13 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq,
        for (i = 0; i < iov_count; i++) {
                unsigned long uaddr = (unsigned long)iov[i].iov_base;
 
+               if (!iov[i].iov_len)
+                       return -EINVAL;
+
                if (uaddr & queue_dma_alignment(q)) {
                        unaligned = 1;
                        break;
                }
-               if (!iov[i].iov_len)
-                       return -EINVAL;
        }
 
        if (unaligned || (q->dma_pad_mask & len) || map_data)