Feature #11032

selinux policy for ceph-mon

Added by Sage Weil about 8 years ago. Updated over 7 years ago.

Target version:
% Done:


Affected Versions:
Pull request ID:


From an internal red hat discussion:

There are probably three distinct things we need to do to get cephs and
SELinux to work together.

1. It should just work out of the box. If there is any issue that
prevents cephs from working with SELinux in enforcing mode either on the
client or server, this needs to be fixed by the SELinux team ASAP. But
we need to know what is happening. Have your QE team do their testing
in enforcing mode and if things break have them work in permissive and
report bugs including the AVC messages. You can collect these via

ausearch -m avc -i -ts recent
Attach these to the bugzilla and work with SELinux engineers to get the
policy updated.

2. Start to build policy to confine Cephs Daemons. Anything that is
running on a cephs server as unconfined_service_t or initrc_t or init_t
will need to have policy written for it.

ps -eZ | grep unconfined_service

Will tells us the daemons that need policy written. Open a bugzilla for
any daemons that do not have policy written for them and the SELinux
team will work with you and QE to write policy.

3. Does cephs support XAttrs? SELinux relies on XATTR support to label
content differently. SELinux is a labeling system, we label every
process and every file system object and then write rules controlling
access between process labels and file labels. If a file system does
not support XAttrs then we can label the entire file system with the
same label via a mount "context" option, but this is not ideal since we
do not get the fine grained control. For example if we wanted to have
two different VM images on a cephs file system, we would want to have
them labeled differently.


#1 Updated by Sage Weil over 7 years ago

  • Target version set to v9.0.7

#2 Updated by Sage Weil over 7 years ago

  • Status changed from New to Resolved

Also available in: Atom PDF