LU-17202 target: set SB_KERNMOUNT

set SB_KERNMOUNT on server's mountpoint so umount process block
until the server (MDS/OST/MGS) is really umounted and released
the corresponding block device.

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I03ac2dc6bb7cd4d93f5a1729fccb976aa1ebd5d4

diff --git a/lustre/target/tgt_mount.c b/lustre/target/tgt_mount.c
index eeab3c1..692d266 100644 (file)
--- a/lustre/target/tgt_mount.c
+++ b/lustre/target/tgt_mount.c
@@ -2065,7 +2065,8 @@ static int server_fill_super_common(struct super_block *sb)
        sb->s_blocksize_bits = log2(sb->s_blocksize);
        sb->s_magic = LUSTRE_SUPER_MAGIC;
        sb->s_maxbytes = 0; /* we don't allow file IO on server mountpoints */
-       sb->s_flags |= SB_RDONLY;
+       /* set SB_KERNMOUNT to make umount synchronous */
+       sb->s_flags |= SB_RDONLY | SB_KERNMOUNT;
        sb->s_op = &server_ops;
 
        root = new_inode(sb);