remount to private subtree after unshare

systemd will makes the VFS tree 'shared' on boot, so mount
operations would propagate between the namespace and the
parent system.
make sure the whole VFS tree is private to the namespace.

lib/linux_ns.c

@@ -66,7 +66,10 @@ void linux_namespace_start(void *argv) {
 			uwsgi_error("clone()");
 			exit(1);
 		}
-
+		if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL)) {
+			uwsgi_error("mount()");
+			exit(1);
+		}
 		// run the post-jail scripts
 		if (setenv("UWSGI_JAIL_PID", uwsgi_num2str((int) pid), 1)) {
 			uwsgi_error("setenv()");
@@ -173,6 +176,7 @@ void linux_namespace_jail() {
 	uwsgi_log("remounting /proc\n");
 	if (mount("proc", "/proc", "proc", 0, NULL)) {
 		uwsgi_error("mount()");
+		exit(1);
 	}
 
 	struct uwsgi_string_list *usl = uwsgi.ns_keep_mount;