Btrfs: copy everything if we've created an inline extent

commit bdc20e67e82cfc4901d3a5a0d79104b0e2296d83 upstream.

I noticed while looking into a tree logging bug that we aren't logging inline
extents properly.  Since this requires copying and it shouldn't happen too often
just force us to copy everything for the inode into the tree log when we have an
inline extent.  With this patch we have valid data after a crash when we write
an inline extent.  Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index cc93b23..8a338c2 100644 (file)
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -265,6 +265,7 @@ static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
                return 1;
        }
 
+       set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
        btrfs_delalloc_release_metadata(inode, end + 1 - start);
        btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
        return 0;