cachefiles: do not yet allow on idmapped mounts
authorChristian Brauner <christian.brauner@ubuntu.com>
Wed, 24 Mar 2021 08:51:10 +0000 (08:51 +0000)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 24 Mar 2021 17:20:22 +0000 (10:20 -0700)
Based on discussions (e.g. in [1]) my understanding of cachefiles and
the cachefiles userspace daemon is that it creates a cache on a local
filesystem (e.g. ext4, xfs etc.) for a network filesystem. The way this
is done is by writing "bind" to /dev/cachefiles and pointing it to the
directory to use as the cache.

Currently this directory can technically also be an idmapped mount but
cachefiles aren't yet fully aware of such mounts and thus don't take the
idmapping into account when creating cache entries. This could leave
users confused as the ownership of the files wouldn't match to what they
expressed in the idmapping. Block cache files on idmapped mounts until
the fscache rework is done and we have ported it to support idmapped
mounts.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: linux-cachefs@redhat.com
Link: https://lore.kernel.org/lkml/20210303161528.n3jzg66ou2wa43qb@wittgenstein
Link: https://lore.kernel.org/r/20210316112257.2974212-1-christian.brauner@ubuntu.com/
Link: https://listman.redhat.com/archives/linux-cachefs/2021-March/msg00044.html
Link: https://lore.kernel.org/r/20210319114146.410329-1-christian.brauner@ubuntu.com/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/cachefiles/bind.c

index dfb14dbddf51d9d16f3361bd16b8ada627bffb20..38bb7764b454092ad65d1793b54b71b0db570c52 100644 (file)
@@ -118,6 +118,12 @@ static int cachefiles_daemon_add_cache(struct cachefiles_cache *cache)
        cache->mnt = path.mnt;
        root = path.dentry;
 
+       ret = -EINVAL;
+       if (mnt_user_ns(path.mnt) != &init_user_ns) {
+               pr_warn("File cache on idmapped mounts not supported");
+               goto error_unsupported;
+       }
+
        /* check parameters */
        ret = -EOPNOTSUPP;
        if (d_is_negative(root) ||