summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorChris Mason <chris.mason@oracle.com>2009-10-09 11:29:53 -0400
committerChris Mason <chris.mason@oracle.com>2009-10-09 11:29:53 -0400
commitac6889cbb254be1ffea376bea4a96ce9be0e0ed0 (patch)
tree8f0e073af940b1bdfb134b945a03a1b43b614352
parente9061e214810c9534381a705a1b46533e09f2676 (diff)
downloadop-kernel-dev-ac6889cbb254be1ffea376bea4a96ce9be0e0ed0.zip
op-kernel-dev-ac6889cbb254be1ffea376bea4a96ce9be0e0ed0.tar.gz
Btrfs: fix file clone ioctl for bookend extents
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>
-rw-r--r--fs/btrfs/ioctl.c6
1 files changed, 4 insertions, 2 deletions
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index e8795be..cdbb054 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -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;
OpenPOWER on IntegriCloud