Btrfs: walk compressed pages based on the nr_pages count instead of bytes

The byte walk counting was awkward and error prone.  This uses the
number of pages sent the higher layer to build bios.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

1 files changed, 1 insertions, 1 deletions

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 9797592dc86b..6739424c0fe6 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -338,7 +338,7 @@ again:
         if (!btrfs_test_flag(inode, NOCOMPRESS) &&
             btrfs_test_opt(root, COMPRESS)) {
                 WARN_ON(pages);
-                pages = kmalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
+                pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
 
                 /* we want to make sure the amount of IO required to satisfy
                  * a random read is reasonably small, so we limit the size