fs: require CAP_SYS_ADMIN in target namespace for idmapped mounts

Idmapped mounts should not allow a user to map file ownsership into a
range of ids which is not under the control of that user. However, we
currently don't check whether the mounter is privileged wrt to the
target user namespace.

Currently no FS_USERNS_MOUNT filesystems support idmapped mounts, thus
this is not a problem as only CAP_SYS_ADMIN in init_user_ns is allowed
to set up idmapped mounts. But this could change in the future, so add a
check to refuse to create idmapped mounts when the mounter does not have
CAP_SYS_ADMIN in the target user namespace.

Fixes: bd303368b776 ("fs: support mapped mounts of mapped filesystems")
Signed-off-by: Seth Forshee <sforshee@digitalocean.com>
Reviewed-by: Christian Brauner (Microsoft) <brauner@kernel.org>
Link: https://lore.kernel.org/r/20220816164752.2595240-1-sforshee@digitalocean.com
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>

diff --git a/fs/namespace.c b/fs/namespace.c
index 68789f896f0819e0abeb509e90aea20dbc0fc7c7..df137ba19d3756bd08d544f0d7e95eb52cb4869c 100644 (file)
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -4238,6 +4238,13 @@ static int build_mount_idmapped(const struct mount_attr *attr, size_t usize,
                err = -EPERM;
                goto out_fput;
        }
+
+       /* We're not controlling the target namespace. */
+       if (!ns_capable(mnt_userns, CAP_SYS_ADMIN)) {
+               err = -EPERM;
+               goto out_fput;
+       }
+
        kattr->mnt_userns = get_user_ns(mnt_userns);
 
 out_fput: