Fix parallel mount failure caused by ZoL specific behavior

The main reason parallel mount behaves differently on ZoL is that
pthreads fork/execs another process for /bin/mount, whereas illumos
and FreeBSD calls mount(2) and nmount(2) respectively.

By taking a global lock around /bin/mount (and also /bin/umount)
call, ZoL will behave similarly to other implementations.

See openzfs#8833 for details.

Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>

lib/libzfs/libzfs_mount.c

@@ -352,6 +352,8 @@ zfs_is_mountable(zfs_handle_t *zhp, char *buf, size_t buflen,
  * http://www.kernel.org/pub/linux/utils/util-linux/libmount-docs/index.html
  */
 
+static pthread_mutex_t mount_lock = PTHREAD_MUTEX_INITIALIZER;
+
 static int
 do_mount(const char *src, const char *mntpt, char *opts)
 {
@@ -366,7 +368,9 @@ do_mount(const char *src, const char *mntpt, char *opts)
 	int rc;
 
 	/* Return only the most critical mount error */
+	pthread_mutex_lock(&mount_lock);
 	rc = libzfs_run_process(argv[0], argv, STDOUT_VERBOSE|STDERR_VERBOSE);
+	pthread_mutex_unlock(&mount_lock);
 	if (rc) {
 		if (rc & MOUNT_FILEIO)
 			return (EIO);
@@ -409,7 +413,10 @@ do_unmount(const char *mntpt, int flags)
 	}
 
 	argv[count] = (char *)mntpt;
+
+	pthread_mutex_lock(&mount_lock);
 	rc = libzfs_run_process(argv[0], argv, STDOUT_VERBOSE|STDERR_VERBOSE);
+	pthread_mutex_unlock(&mount_lock);
 
 	return (rc ? EINVAL : 0);
 }