Btrfs: add missing inode version, ctime and mtime updates when punching hole
authorFilipe Manana <fdmanana@suse.com>
Wed, 19 Jun 2019 12:05:50 +0000 (13:05 +0100)
committerDavid Sterba <dsterba@suse.com>
Tue, 2 Jul 2019 10:30:50 +0000 (12:30 +0200)
If the range for which we are punching a hole covers only part of a page,
we end up updating the inode item but we skip the update of the inode's
iversion, mtime and ctime. Fix that by ensuring we update those properties
of the inode.

A patch for fstests test case generic/059 that tests this as been sent
along with this fix.

Fixes: 2aaa66558172b0 ("Btrfs: add hole punching")
Fixes: e8c1c76e804b18 ("Btrfs: add missing inode update when punching hole")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/file.c

index 5370152ea7e3affd8b587c99d3a70049b7a4c779..b455bdf46faa9e4ddc765f97e15a0c85366bf443 100644 (file)
@@ -2711,6 +2711,11 @@ out_only_mutex:
                 * for detecting, at fsync time, if the inode isn't yet in the
                 * log tree or it's there but not up to date.
                 */
+               struct timespec64 now = current_time(inode);
+
+               inode_inc_iversion(inode);
+               inode->i_mtime = now;
+               inode->i_ctime = now;
                trans = btrfs_start_transaction(root, 1);
                if (IS_ERR(trans)) {
                        err = PTR_ERR(trans);