Btrfs: fix filemap_flush call in btrfs_file_release

commit f6dc45c7a93a011dff6eb9b2ffda59c390c7705a upstream.

We should only be flushing on close if the file was flagged as needing
it during truncate.  I broke this with my ordered data vs transaction
commit deadlock fix.

Thanks to Miao Xie for catching this.

Signed-off-by: Chris Mason <clm@fb.com>
Reported-by: Miao Xie <miaox@cn.fujitsu.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index d3afac292d677e0ed492255ca7cb3890cde717b1..ab1fd668020d88538e3c1b9ed1bc79fff2aae3e6 100644 (file)
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1840,7 +1840,15 @@ int btrfs_release_file(struct inode *inode, struct file *filp)
 {
        if (filp->private_data)
                btrfs_ioctl_trans_end(filp);
-       filemap_flush(inode->i_mapping);
+       /*
+        * ordered_data_close is set by settattr when we are about to truncate
+        * a file from a non-zero size to a zero size.  This tries to
+        * flush down new bytes that may have been written if the
+        * application were using truncate to replace a file in place.
+        */
+       if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
+                              &BTRFS_I(inode)->runtime_flags))
+                       filemap_flush(inode->i_mapping);
        return 0;
 }