Btrfs: fix file clone ioctl for bookend extents
authorChris Mason <chris.mason@oracle.com>
Fri, 9 Oct 2009 15:29:53 +0000 (11:29 -0400)
committerChris Mason <chris.mason@oracle.com>
Fri, 9 Oct 2009 15:29:53 +0000 (11:29 -0400)
The file clone ioctl was incorrectly taking the offset into the
extent on disk into account when calculating the length of the
cloned extent.

The length never changes based on the offset into the physical extent.

Test case:

fallocate -l 1g image
mke2fs image
bcp image image2
e2fsck -f image2

(errors on image2)

The math bug ends up wrapping the length of the extent, and things
go wrong from there.

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

index e8795becad4cda9ce6e97ce87c77edc62f73c9a0..cdbb054102b9f860ee4d119dfff39e891ac29438 100644 (file)
@@ -1123,8 +1123,10 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
                                        datao += off - key.offset;
                                        datal -= off - key.offset;
                                }
-                               if (key.offset + datao + datal > off + len)
-                                       datal = off + len - key.offset - datao;
+
+                               if (key.offset + datal > off + len)
+                                       datal = off + len - key.offset;
+
                                /* disko == 0 means it's a hole */
                                if (!disko)
                                        datao = 0;