Btrfs: fix btrfs page_mkwrite to return locked page
authorChris Mason <chris.mason@oracle.com>
Fri, 11 Sep 2009 16:33:12 +0000 (12:33 -0400)
committerChris Mason <chris.mason@oracle.com>
Fri, 11 Sep 2009 17:31:08 +0000 (13:31 -0400)
This closes a whole where the page may be written before
the page_mkwrite caller has a chance to dirty it

(thanks to Nick Piggin)

Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/inode.c

index 233fe6f..c846482 100644 (file)
@@ -4552,11 +4552,14 @@ again:
        }
        ClearPageChecked(page);
        set_page_dirty(page);
+       SetPageUptodate(page);
 
        BTRFS_I(inode)->last_trans = root->fs_info->generation + 1;
        unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
 
 out_unlock:
+       if (!ret)
+               return VM_FAULT_LOCKED;
        unlock_page(page);
 out:
        return ret;