[CRIU] [PATCH 10/12] mounts: create a temporary directory for restoring non-root mntns

[Andrey Vagin]

All non-root namespaces will be restored as sub-trees of the root tree.

This patch creates a temporary directory and mount tmpfs in it, then
create directories for each non-root mount namespace.

tmpfs is quite useful here to simplify destroying this construction,
we don't need to unmount each namespace separately.

Signed-off-by: Andrey Vagin <avagin at openvz.org>
---
 cr-restore.c    | 22 ++++++++-----
 include/mount.h |  2 ++
 mount.c         | 98 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 115 insertions(+), 7 deletions(-)

diff --git a/cr-restore.c b/cr-restore.c
index 003e0f0..5825849 100644
--- a/cr-restore.c
+++ b/cr-restore.c
@@ -1281,10 +1281,10 @@ static int restore_task_with_children(void *_arg)
 		 * Thus -- mount proc at custom location for any new namespace
 		 */
 		if (mount_proc())
-			exit(1);
+			goto err;
 
 		if (root_prepare_shared())
-			exit(1);
+			goto err;
 	}
 
 	/*
@@ -1297,31 +1297,39 @@ static int restore_task_with_children(void *_arg)
 	ret = sigprocmask(SIG_BLOCK, &blockmask, NULL);
 	if (ret) {
 		pr_perror("%d: Can't block signals", current->pid.virt);
-		exit(1);
+		goto err;
 	}
 
 	if (prepare_mappings(pid))
-		exit(1);
+		goto err;
 
 	if ((!(ca->clone_flags & CLONE_FILES)) &&
 	    current->parent && current->parent->rst->fdt)
 		close_old_servie_fd(current->parent->rst->fdt->nr);
 
 	if (create_children_and_session())
-		exit(1);
+		goto err;
 
 	if (unmap_guard_pages())
-		exit(1);
+		goto err;
 
 	restore_pgid();
 
 	if (restore_finish_stage(CR_STATE_FORKING) < 0)
-		exit(1);
+		goto err;
+
+	if (current->parent == NULL && fini_mnt_ns())
+		exit (1);
 
 	if (current->state == TASK_HELPER)
 		return 0;
 
 	return restore_one_task(current->pid.virt, ca->core);
+err:
+	if (current->parent == NULL)
+		fini_mnt_ns();
+
+	exit(1);
 }
 
 static inline int stage_participants(int next_stage)
diff --git a/include/mount.h b/include/mount.h
index a76c027..c71f6e0 100644
--- a/include/mount.h
+++ b/include/mount.h
@@ -25,4 +25,6 @@ extern struct ns_desc mnt_ns_desc;
 extern dev_t phys_stat_resolve_dev(dev_t st_dev, const char *path);
 extern bool phys_stat_dev_match(dev_t st_dev, dev_t phys_dev, const char *path);
 
+extern int fini_mnt_ns(void);
+
 #endif /* __CR_MOUNT_H__ */
diff --git a/mount.c b/mount.c
index 1eed6c9..e6d54d0 100644
--- a/mount.c
+++ b/mount.c
@@ -1373,6 +1373,34 @@ static void free_mounts(void)
 	}
 }
 
+static char *get_mnt_roots(bool create)
+{
+	static char *mnt_roots = NULL;
+
+	if (!create || mnt_roots)
+		return mnt_roots;
+
+	if (chdir(opts.root ? : "/")) {
+		pr_perror("Unable to change working directory on %s", opts.root);
+		return NULL;
+	}
+
+	mnt_roots = strdup(".criu.mntns.XXXXXX");
+	if (mnt_roots == NULL) {
+		pr_perror("Can't allocate memory");
+		return NULL;
+	}
+
+	if (mkdtemp(mnt_roots) == NULL) {
+		pr_perror("Unable to create a temporary directory");
+		mnt_roots = NULL;
+		return NULL;
+	}
+
+	return mnt_roots;
+
+}
+
 static struct mount_info *read_mnt_ns_img(int ns_pid)
 {
 	MntEntry *me = NULL;
@@ -1452,6 +1480,45 @@ err:
 	return NULL;
 }
 
+/*
+ * All nested mount namespaces are restore as sub-trees of the root namespace.
+ */
+static int prepare_temporary_roots()
+{
+	char path[PATH_MAX];
+	struct ns_id *nsid;
+	char *mnt_roots = get_mnt_roots(false);
+
+	if (mnt_roots == NULL)
+		return 0;
+
+	if (mount("none", mnt_roots, "tmpfs", 0, NULL)) {
+		pr_perror("Unable to mount tmpfs in %s", mnt_roots);
+		return -1;
+	}
+	if (mount("none", mnt_roots, NULL, MS_PRIVATE, NULL))
+		return -1;
+
+	nsid = ns_ids;
+	while (nsid) {
+		if (nsid->nd != &mnt_ns_desc) {
+			nsid = nsid->next;
+			continue;
+		}
+
+		snprintf(path, sizeof(path), "%s/%d",
+				mnt_roots, nsid->id);
+
+		if (mkdir(path, 0600)) {
+			pr_perror("Unable to create %s", path);
+			return -1;
+		}
+		nsid = nsid->next;
+	}
+
+	return 0;
+}
+
 static int populate_mnt_ns(int ns_pid, struct mount_info *mis)
 {
 	struct mount_info *pms;
@@ -1459,6 +1526,9 @@ static int populate_mnt_ns(int ns_pid, struct mount_info *mis)
 	mntinfo_tree = NULL;
 	mntinfo = mis;
 
+	if (prepare_temporary_roots())
+		return -1;
+
 	pms = mnt_build_tree(mntinfo);
 	if (!pms)
 		return -1;
@@ -1470,6 +1540,34 @@ static int populate_mnt_ns(int ns_pid, struct mount_info *mis)
 	return mnt_tree_for_each(pms, do_mount_one);
 }
 
+int fini_mnt_ns()
+{
+	char *mnt_roots = get_mnt_roots(false);
+	int ret = 0;
+
+	if (mnt_roots == NULL)
+		return 0;
+
+	if (mount("none", mnt_roots, "none", MS_REC|MS_PRIVATE, NULL)) {
+		pr_perror("Can't remount root with MS_PRIVATE");
+		ret = 1;
+	}
+	/*
+	 * Don't exit after a first error, becuase this function
+	 * can be used to rollback in a error case
+	 */
+	if (umount2(mnt_roots, MNT_DETACH)) {
+		pr_perror("Can't unmount %s", mnt_roots);
+		ret = 1;
+	}
+	if (rmdir(mnt_roots)) {
+		pr_perror("Can't remove the directory %s", mnt_roots);
+		ret = 1;
+	}
+
+	return ret;
+}
+
 int prepare_mnt_ns(int ns_pid)
 {
 	int ret = -1;
-- 
1.8.3.1