[lxc-devel] [lxc/master] autodev: adapt to changes in Linux 4.18

[brauner on Github]

A non-text attachment was scrubbed...
Name: not available
Type: text/x-mailbox
Size: 1465 bytes
Desc: not available
URL: <http://lists.linuxcontainers.org/pipermail/lxc-devel/attachments/20180629/8fcd53be/attachment.bin>
-------------- next part --------------
From cd541e9415fcb2bf5f05111d24aa9e1573327f8f Mon Sep 17 00:00:00 2001
From: Christian Brauner <christian.brauner at ubuntu.com>
Date: Fri, 29 Jun 2018 13:58:52 +0200
Subject: [PATCH] autodev: adapt to changes in Linux 4.18

Starting with commit
55956b59df33 ("vfs: Allow userns root to call mknod on owned filesystems.")
Linux will allow mknod() in user namespaces for userns root if CAP_MKNOD is
available.
However, these device nodes are useless since

static struct super_block *alloc_super(struct file_system_type *type, int flags,
                                       struct user_namespace *user_ns)
{
        /* <snip> */

        if (s->s_user_ns != &init_user_ns)
                s->s_iflags |= SB_I_NODEV;

        /* <snip> */
}

will set the SB_I_NODEV flag on the filesystem. When a device node created in
non-init userns is open()ed the call chain will hit:

bool may_open_dev(const struct path *path)
{
        return !(path->mnt->mnt_flags & MNT_NODEV) &&
                !(path->mnt->mnt_sb->s_iflags & SB_I_NODEV);
}

which will cause an EPERM because the device node is located on an fs
owned by non-init-userns and thus doesn't grant access to device nodes due to
SB_I_NODEV.

The solution is straightforward. Unless you're real root you should bind-mount
device nodes.

Signed-off-by: Christian Brauner <christian.brauner at ubuntu.com>
---
 src/lxc/conf.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/src/lxc/conf.c b/src/lxc/conf.c
index ee0e7bddd..cc37b8637 100644
--- a/src/lxc/conf.c
+++ b/src/lxc/conf.c
@@ -1251,7 +1251,7 @@ static int lxc_fill_autodev(const struct lxc_rootfs *rootfs)
 	int i, ret;
 	char path[MAXPATHLEN];
 	mode_t cmask;
-	bool can_mknod = true;
+	int can_mknod = 1;
 
 	ret = snprintf(path, MAXPATHLEN, "%s/dev",
 		       rootfs->path ? rootfs->mount : "");
@@ -1274,7 +1274,11 @@ static int lxc_fill_autodev(const struct lxc_rootfs *rootfs)
 		if (ret < 0 || ret >= MAXPATHLEN)
 			return -1;
 
-		if (can_mknod) {
+		/* See
+		 * - https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=55956b59df336f6738da916dbb520b6e37df9fbd
+		 * - https://lists.linuxfoundation.org/pipermail/containers/2018-June/039176.html
+		 */
+		if (can_mknod == 2 || (can_mknod == 1 && !am_host_unpriv())) {
 			ret = mknod(path, device->mode, makedev(device->maj, device->min));
 			if (ret == 0 || (ret < 0 && errno == EEXIST)) {
 				DEBUG("Created device node \"%s\"", path);
@@ -1289,7 +1293,9 @@ static int lxc_fill_autodev(const struct lxc_rootfs *rootfs)
 			/* This can e.g. happen when the container is
 			 * unprivileged or CAP_MKNOD has been dropped.
 			 */
-			can_mknod = false;
+			can_mknod = 2;
+		} else {
+			can_mknod = 0;
 		}
 
 		ret = mknod(path, S_IFREG, 0);