From: Lennart Poettering <lennart@poettering.net>
Date: Fri, 19 May 2023 12:22:14 +0000 (+0200)
Subject: switch-root: sync() before MNT_DETACH unmounting all old mounts
X-Git-Tag: v254-rc1~438^2
X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=5221b55bb3b78d6dafcc33a609513b594fb62dbf;p=thirdparty%2Fsystemd.git

switch-root: sync() before MNT_DETACH unmounting all old mounts

As discussed here:

https://github.com/systemd/systemd/pull/27670#discussion_r1196716794
---

diff --git a/src/shared/switch-root.c b/src/shared/switch-root.c
index 2431facba00..d5b3d169bf8 100644
--- a/src/shared/switch-root.c
+++ b/src/shared/switch-root.c
@@ -63,6 +63,13 @@ int switch_root(const char *new_root,
                         (void) mkdir_p_label(resolved_old_root_after, 0755);
         }
 
+        /* We are about to unmount various file systems with MNT_DETACH (either explicitly via umount() or
+         * indirectly via pivot_root()), and thus do not synchronously wait for them to be fully sync'ed —
+         * all while making them invisible/inaccessible in the file system tree for later code. That makes
+         * sync'ing them then difficult. Let's hence issue a manual sync() here, so that we at least can
+         * guarantee all file systems are an a good state before entering this state. */
+        sync();
+
         /* Work-around for kernel design: the kernel refuses MS_MOVE if any file systems are mounted
          * MS_SHARED. Hence remount them MS_PRIVATE here as a work-around.
          *