smb3: directory sync should not return an error 4.17-rc4-SMB3-Fixes
authorSteve French <smfrench@gmail.com>
Thu, 10 May 2018 15:59:37 +0000 (10:59 -0500)
committerSteve French <smfrench@gmail.com>
Fri, 11 May 2018 00:21:14 +0000 (19:21 -0500)
As with NFS, which ignores sync on directory handles,
fsync on a directory handle is a noop for CIFS/SMB3.
Do not return an error on it.  It breaks some database
apps otherwise.

Signed-off-by: Steve French <smfrench@gmail.com>
CC: Stable <stable@vger.kernel.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
fs/cifs/cifsfs.c

index f715609b13f34a412de9f0486750a05cf73340b6..5a5a0158cc8f5085cda312d25c10b21e49bd8d1d 100644 (file)
@@ -1047,6 +1047,18 @@ out:
        return rc;
 }
 
+/*
+ * Directory operations under CIFS/SMB2/SMB3 are synchronous, so fsync()
+ * is a dummy operation.
+ */
+static int cifs_dir_fsync(struct file *file, loff_t start, loff_t end, int datasync)
+{
+       cifs_dbg(FYI, "Sync directory - name: %pD datasync: 0x%x\n",
+                file, datasync);
+
+       return 0;
+}
+
 static ssize_t cifs_copy_file_range(struct file *src_file, loff_t off,
                                struct file *dst_file, loff_t destoff,
                                size_t len, unsigned int flags)
@@ -1181,6 +1193,7 @@ const struct file_operations cifs_dir_ops = {
        .copy_file_range = cifs_copy_file_range,
        .clone_file_range = cifs_clone_file_range,
        .llseek = generic_file_llseek,
+       .fsync = cifs_dir_fsync,
 };
 
 static void