Bug 1231936

+++ This bug was initially created as a clone of Bug #1231134 +++

SELinux denies any access from inside a Docker container to a mounted GlusterFS volume, even though virt_use_fusefs boolean is set to 'on'.

AVC gathered with disabled dontaudit rules (why dontaudit this ????):
type=SYSCALL msg=audit(1434098028.965:515): arch=c000003e syscall=257 success=no exit=-13 a0=ffffffffffffff9c a1=194bc60 a2=90800 a3=0 items=0 ppid=4012 pid=4054 auid=4294967295 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts3 ses=4294967295 comm="ls" exe="/usr/bin/ls" subj=system_u:system_r:svirt_lxc_net_t:s0:c131,c679 key=(null)
type=AVC msg=audit(1434098028.965:515): avc:  denied  { read } for  pid=4054 comm="ls" name="/" dev="fuse" ino=1 scontext=system_u:system_r:svirt_lxc_net_t:s0:c131,c679 tcontext=system_u:object_r:fusefs_t:s0 tclass=dir permissive=0


Version-Release number of selected component (if applicable):
selinux-policy-targeted-3.13.1-105.13.fc21.noarch

How reproducible:
always

Steps to Reproduce:
1. mount a GlusterFS volume to /mnt/test (or use any other fuse filesystem)
mount -t glusterfs my-gluster-server:test_vol /mnt/test

2. use it as a container volume 
docker run -ti -v /mnt/test:/mnt/test fedora:21 bash

3. (inside the container) look at /mnt/test
ls /mnt/test

Actual results:
ls: cannot open directory /mnt/test: Permission denied

Expected results:
see content of the directory

Additional info:
As stated above, 'setsebool -P virt_use_fusefs on' does not help. On the other way, 'setsebool -P virt_use_nfs on' helps to get the same working for NFS.

Also, these dontaudit rules only complicate things, it's hard to guess who rejects access in a complex environment where Docker, Kubernetes, SElinux, kernel namespaces and Gluster are involved.

--- Additional comment from Jeff Vance on 2015-06-12 16:07:27 CDT ---

I've observed that, although you cannot ls the mounted dir (as if r access to the dir is denied), you can pass through it (x access) and cat files in the gluster volume.

--- Additional comment from Daniel Walsh on 2015-06-15 08:12:42 CDT ---

I just added a 

virt_sandbox_use_fusefs boolean to docker-selinux.

If we got proper labeling on gluster files via XAttrs we would not have this problem.

This boolean will allow docker containers to attack any fusefs files on the system, including all gluster mounted drives, if it can get access to them.