From 27034e7c64b00a1f2467afb5ebb1d5b9b1a06ce1 Mon Sep 17 00:00:00 2001
From: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Date: Wed, 11 Jul 2018 13:43:33 +0300
Subject: [PATCH] mount: fix regression where open_mountpoint failed on
 readonly fs

If we fail to create temporary directory for doing a clean mount we can
make mount clean reusing the code which enters new mountns to umount
overmounts. As when last process exits mntns all mounts are implicitly
cleaned from children, see in kernel source - sys_exit->do_exit
->exit_task_namespaces->switch_task_namespaces->free_nsproxy
->put_mnt_ns->umount_tree->drop_collected_mounts->umount_tree:

    /* Hide the mounts from mnt_mounts */
    list_for_each_entry(p, &tmp_list, mnt_list) {
            list_del_init(&p->mnt_child);
    }

Fixes commit b6cfb1ce2948 ("mount: make open_mountpoint handle overmouts
properly")

https://github.com/checkpoint-restore/criu/issues/520
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Acked-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
---
 criu/mount.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)

diff --git a/criu/mount.c b/criu/mount.c
index b56164e953..9cc8f6e940 100644
--- a/criu/mount.c
+++ b/criu/mount.c
@@ -1325,10 +1325,18 @@ int ns_open_mountpoint(void *arg)
 	if (umount_overmounts(mi))
 		goto err;
 
-	/* Save fd which we opened for parent due to CLONE_FILES flag */
-	*fd = get_clean_fd(mi);
-	if (*fd < 0)
+	/*
+	 * Save fd which we opened for parent due to CLONE_FILES flag
+	 *
+	 * Mount can still have children in it, but we don't need to clean it
+	 * explicitly as when last process exits mntns all mounts in it are
+	 * cleaned from their children, and we are exactly the last process.
+	 */
+	*fd = open(mi->mountpoint, O_DIRECTORY|O_RDONLY);
+	if (*fd < 0) {
+		pr_perror("Unable to open %s", mi->mountpoint);
 		goto err;
+	}
 
 	return 0;
 err:
@@ -1367,18 +1375,22 @@ int open_mountpoint(struct mount_info *pm)
 
 	if (!mnt_is_overmounted(pm)) {
 		pr_info("\tmount has children %s\n", pm->mountpoint);
-
 		fd = get_clean_fd(pm);
-		if (fd < 0)
-			goto err;
-	} else {
+	}
+
+	/*
+	 * Mount is overmounted or probably we can't create a temporary
+	 * direcotry for a cleaned mount
+	 */
+	if (fd < 0) {
 		int pid, status;
 		struct clone_arg ca = {
 			.mi = pm,
 			.fd = &fd
 		};
 
-		pr_info("\tmount is overmounted %s\n", pm->mountpoint);
+		pr_info("\tmount is overmounted or has children %s\n",
+				pm->mountpoint);
 
 		/*
 		 * We are overmounted - not accessible in a regular way. We