Re: [PATCH v6 2/2] init/do_mounts.c: create second mount for initramfs

[Christian Brauner]

On Sat, Jun 05, 2021 at 10:47:07PM +0800, Menglong Dong wrote:
> Hello,
> 
> On Sat, Jun 5, 2021 at 7:50 PM Christian Brauner
> <christian.brauner@xxxxxxxxxx> wrote:
> >
> > On Sat, Jun 05, 2021 at 11:44:47AM +0800, menglong8.dong@xxxxxxxxx wrote:
> > > From: Menglong Dong <dong.menglong@xxxxxxxxxx>
> > >
> > > If using container platforms such as Docker, upon initialization it
> > > wants to use pivot_root() so that currently mounted devices do not
> > > propagate to containers. An example of value in this is that
> > > a USB device connected prior to the creation of a containers on the
> > > host gets disconnected after a container is created; if the
> > > USB device was mounted on containers, but already removed and
> > > umounted on the host, the mount point will not go away until all
> > > containers unmount the USB device.
> > >
> > > Another reason for container platforms such as Docker to use pivot_root
> > > is that upon initialization the net-namspace is mounted under
> > > /var/run/docker/netns/ on the host by dockerd. Without pivot_root
> > > Docker must either wait to create the network namespace prior to
> > > the creation of containers or simply deal with leaking this to each
> > > container.
> > >
> > > pivot_root is supported if the rootfs is a initrd or block device, but
> > > it's not supported if the rootfs uses an initramfs (tmpfs). This means
> > > container platforms today must resort to using block devices if
> > > they want to pivot_root from the rootfs. A workaround to use chroot()
> > > is not a clean viable option given every container will have a
> > > duplicate of every mount point on the host.
> > >
> > > In order to support using container platforms such as Docker on
> > > all the supported rootfs types we must extend Linux to support
> > > pivot_root on initramfs as well. This patch does the work to do
> > > just that.
> > >
> > > pivot_root will unmount the mount of the rootfs from its parent mount
> > > and mount the new root to it. However, when it comes to initramfs, it
> > > donesn't work, because the root filesystem has not parent mount, which
> > > makes initramfs not supported by pivot_root.
> > >
> > > In order to make pivot_root supported on initramfs, we create a second
> > > mount with type of rootfs before unpacking cpio, and change root to
> > > this mount after unpacking.
> > >
> > > While mounting the second rootfs, 'rootflags' is passed, and it means
> > > that we can set options for the mount of rootfs in boot cmd now.
> > > For example, the size of tmpfs can be set with 'rootflags=size=1024M'.
> > >
> > > Signed-off-by: Menglong Dong <dong.menglong@xxxxxxxxxx>
> > > ---
> > >  init/do_mounts.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
> > >  init/do_mounts.h | 17 ++++++++++++++++-
> > >  init/initramfs.c |  8 ++++++++
> > >  usr/Kconfig      | 10 ++++++++++
> > >  4 files changed, 78 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/init/do_mounts.c b/init/do_mounts.c
> > > index a78e44ee6adb..715bdaa89b81 100644
> > > --- a/init/do_mounts.c
> > > +++ b/init/do_mounts.c
> > > @@ -618,6 +618,49 @@ void __init prepare_namespace(void)
> > >  }
> > >
> > >  static bool is_tmpfs;
> > > +#ifdef CONFIG_INITRAMFS_MOUNT
> > > +
> > > +/*
> > > + * Give systems running from the initramfs and making use of pivot_root a
> > > + * proper mount so it can be umounted during pivot_root.
> > > + */
> > > +int __init prepare_mount_rootfs(void)
> > > +{
> > > +     char *rootfs = "ramfs";
> > > +
> > > +     if (is_tmpfs)
> > > +             rootfs = "tmpfs";
> > > +
> > > +     return do_mount_root(rootfs, rootfs,
> > > +                          root_mountflags & ~MS_RDONLY,
> > > +                          root_mount_data);
> > > +}
> > > +
> > > +/*
> > > + * Revert to previous mount by chdir to '/' and unmounting the second
> > > + * mount.
> > > + */
> > > +void __init revert_mount_rootfs(void)
> > > +{
> > > +     init_chdir("/");
> > > +     init_umount(".", MNT_DETACH);
> > > +}
> > > +
> > > +/*
> > > + * Change root to the new rootfs that mounted in prepare_mount_rootfs()
> > > + * if cpio is unpacked successfully and 'ramdisk_execute_command' exist.
> > > + */
> > > +void __init finish_mount_rootfs(void)
> > > +{
> > > +     init_mount(".", "/", NULL, MS_MOVE, NULL);
> > > +     if (likely(ramdisk_exec_exist()))
> > > +             init_chroot(".");
> > > +     else
> > > +             revert_mount_rootfs();
> > > +}
> > > +
> > > +#define rootfs_init_fs_context ramfs_init_fs_context
> >
> > Sorry, I think we're nearly there. What's the rationale for using ramfs
> > when unconditionally when a separate mount for initramfs is requested?
> > Meaning, why do we need this define at all?
> 
> I think it's necessary, as I explained in the third patch. When the rootfs
> is a block device, ramfs is used in init_mount_tree() unconditionally,
> which can be seen from the enable of is_tmpfs.
> 
> That makes sense, because rootfs will not become the root if a block
> device is specified by 'root' in boot cmd, so it makes no sense to use
> tmpfs, because ramfs is more simple.
> 
> Here, I make rootfs as ramfs for the same reason: the first mount is not
> used as the root, so make it ramfs which is more simple.

Ok. If you don't mind I'd like to pull and test this before moving
further. (Btw, I talked about this at Plumbers before btw.)
What did you use for testing this? Any way you can share it?