[lkml]   [2021]   [May]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
Subject[PATCH 2/3] init/do_cmounts.c: introduce 'user_root' for initramfs
From: Menglong Dong <>

During the kernel initialization, the root of the mount tree is
created with file system type of ramfs or tmpfs.

While using initramfs as the root file system, cpio file is unpacked
into the rootfs. Thus, this rootfs is exactly what users see in user
space, and some problems arose: this rootfs has no parent mount,
which make it can't be umounted or pivot_root.

'pivot_root' is used to change the rootfs and clean the old mounts,
and it is essential for some users, such as docker. Docker use
'pivot_root' to change the root fs of a process if the current root
fs is a block device of initrd. However, when it comes to initramfs,
things is different: docker has to use 'chroot()' to change the root
fs, as 'pivot_root()' is not supported in initramfs.

The usage of 'chroot()' to create root fs for a container introduced
a lot problems.

First, 'chroot()' can't clean the old mountpoints which inherited
from the host. It means that the mountpoints in host will have a
duplicate in every container. Users may remove a USB after it
is umounted successfully in the host. However, the USB may still
be mounted in containers, although it can't be seen by the 'mount'
commond. This means the USB is not released yet, and data may not
write back. Therefore, data lose arise.

Second, net-namespace leak is another problem. The net-namespace
of containers will be mounted in /var/run/docker/netns/ in host
by dockerd. It means that the net-namespace of a container will
be mounted in containers which are created after it. Things
become worse now, as the net-namespace can't be remove after
the destroy of that container, as it is still mounted in other
containers. If users want to recreate that container, he will
fail if a certain mac address is to be binded with the container,
as it is not release yet.

In this patch, a second mount, which is called 'user root', is
created before 'cpio' unpacking. The file system of 'user root'
is exactly the same as rootfs, and both ramfs and tmpfs are
supported. Then, the 'cpio' is unpacked into the 'user root'.
Now, the rootfs has a parent mount, and pivot_root() will be
supported for initramfs.

What's more, after this patch, 'rootflags' in boot cmd is supported
by initramfs. Therefore, users can set the size of tmpfs with

Signed-off-by: Menglong Dong <>
init/do_mounts.c | 72 ++++++++++++++++++++++++++++++++++++++++++++++++
init/do_mounts.h | 7 ++++-
init/initramfs.c | 10 +++++++
3 files changed, 88 insertions(+), 1 deletion(-)

diff --git a/init/do_mounts.c b/init/do_mounts.c
index a78e44ee6adb..943cb58846fb 100644
--- a/init/do_mounts.c
+++ b/init/do_mounts.c
@@ -617,6 +617,78 @@ void __init prepare_namespace(void)

+static __init bool is_tmpfs_enabled(void)
+ return (!root_fs_names || strstr(root_fs_names, "tmpfs")) &&
+ !saved_root_name[0];
+static __init bool is_ramfs_enabled(void)
+ return true;
+struct fs_user_root {
+ bool (*enabled)(void);
+ char *dev_name;
+ char *fs_name;
+static struct fs_user_root user_roots[] __initdata = {
+ {.fs_name = "tmpfs", .enabled = is_tmpfs_enabled },
+ {.fs_name = "ramfs", .enabled = is_ramfs_enabled }
+static struct fs_user_root * __initdata user_root;
+/* Mount the user_root on '/'. */
+int __init mount_user_root(void)
+ return do_mount_root(user_root->dev_name,
+ user_root->fs_name,
+ root_mountflags & ~MS_RDONLY,
+ root_mount_data);
+ * This function is used to chroot to new initramfs root that
+ * we unpacked on success. It will chdir to '/' and umount
+ * the secound mount on failure.
+ */
+void __init end_mount_user_root(bool succeed)
+ if (!succeed)
+ goto on_failed;
+ if (!ramdisk_exec_exist(false))
+ goto on_failed;
+ init_mount(".", "/", NULL, MS_MOVE, NULL);
+ init_chroot(".");
+ return;
+ init_chdir("/");
+ init_umount("/..", 0);
+void __init init_user_rootfs(void)
+ struct fs_user_root *root;
+ int i;
+ for (i = 0; i < ARRAY_SIZE(user_roots); i++) {
+ root = &user_roots[i];
+ if (root->enabled()) {
+ user_root = root;
+ break;
+ }
+ }
static bool is_tmpfs;
static int rootfs_init_fs_context(struct fs_context *fc)
diff --git a/init/do_mounts.h b/init/do_mounts.h
index 7a29ac3e427b..92c004bdd320 100644
--- a/init/do_mounts.h
+++ b/init/do_mounts.h
@@ -10,9 +10,14 @@
#include <linux/root_dev.h>
#include <linux/init_syscalls.h>

+extern int root_mountflags;
void mount_block_root(char *name, int flags);
void mount_root(void);
-extern int root_mountflags;
+int mount_user_root(void);
+void end_mount_user_root(bool succeed);
+void init_user_rootfs(void);
+bool ramdisk_exec_exist(bool abs);

static inline __init int create_dev(char *name, dev_t dev)
diff --git a/init/initramfs.c b/init/initramfs.c
index af27abc59643..ffa78932ae65 100644
--- a/init/initramfs.c
+++ b/init/initramfs.c
@@ -16,6 +16,8 @@
#include <linux/namei.h>
#include <linux/init_syscalls.h>

+#include "do_mounts.h"
static ssize_t __init xwrite(struct file *file, const char *p, size_t count,
loff_t *pos)
@@ -682,15 +684,23 @@ static void __init do_populate_rootfs(void *unused, async_cookie_t cookie)
printk(KERN_INFO "Unpacking initramfs...\n");

+ init_user_rootfs();
+ if (mount_user_root())
+ panic("Failed to create user root");
err = unpack_to_rootfs((char *)initrd_start, initrd_end - initrd_start);
if (err) {
+ end_mount_user_root(false);
printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
+ goto done;

+ end_mount_user_root(true);
* If the initrd region is overlapped with crashkernel reserved region,
 \ /
  Last update: 2021-05-22 13:33    [W:0.123 / U:0.484 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site