ceph backend, secret key leak

Bug #1849624 reported by raphael.glon on 2019-10-24
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Brian Rosmaita
OpenStack Security Advisory
OpenStack Security Notes
Brian Rosmaita

Bug Description

Cinder + ceph backend, secret key leak

Conditions: cinder + ceph backend + rbd_keyring_conf set in cinder config files

As an authenticated simple user create a cinder volume that ends up on a ceph backend,
Then reuse the os.initialize_connection api call
(used by nova-compute/cinder-backup to attach volumes locally to the host running the services):

curl -g -i -X POST https://<cinder_controller>/v3/c495530af57611e9bc14bbaa251e1e96/volumes/7e59b91e-d426-4294-bfc5-dfdebcb21879/action \
    -H "Accept: application/json" \
    -H "Content-Type: application/json" \
    -H "OpenStack-API-Version: volume 3.15" \
    -H "X-Auth-Token: $TOKEN" \
    -d '{"os-initialize_connection": {"connector":{}}}'

If you do not want to forge the http request, openstack clients and extensions may prove helpful.

As root:

apt-get install python3-oslo.privsep virtualenv python3-dev python3-os-brick gcc ceph-common
virtualenv -p python3 venv_openstack
source venv_openstack/bin/activate
pip install python-openstackclient
pip install python-cinderclient
pip install os-brick
pip install python-brick-cinderclient-ext
cinder create vol 1
cinder --debug local-attach 7e59b91e-d426-4294-bfc5-dfdebcb21879

This leaks the ceph credentials for the whole ceph cluster, leaving anyone able to go through ceph acls to get access
to all the volumes within the cluster.

   "connection_info" : {
      "data" : {
         "access_mode" : "rw",
         "secret_uuid" : "SECRET_UUID",
         "cluster_name" : "ceph",
         "encrypted" : false,
         "auth_enabled" : true,
         "discard" : true,
         "qos_specs" : {
            "write_iops_sec" : "3050",
            "read_iops_sec" : "3050"
         "keyring" : "SECRETFILETOHIDE",
         "ports" : [
         "name" : "volumes/volume-7e59b91e-d426-4294-bfc5-dfdebcb21879",
         "secret_type" : "ceph",
         "hosts" : [
         "volume_id" : "7e59b91e-d426-4294-bfc5-dfdebcb21879",
         "auth_username" : "cinder"
      "driver_volume_type" : "rbd"

Quick workaround:
1. Remove rbd_keyring_conf param from any cinder config file, this will mitigate the information disclosure.
2. For cinder backups to still work, providers should instead deploy their ceph keyring secrets directly on cinder-backup hosts
(/etc/cinder/<backend_name>.keyring.conf, to be confirmed).

Note that nova-compute hosts should not be impacted by the change, because ceph secrets are expected to be stored in
libvirt secrets already, thus making this keyring disclose useless to it.
(to be confirmed, there may be other compute drivers that might be impacted by this)

Quick code fix:
Mandatory: revert this commit https://review.opendev.org/#/c/456672/
Optional: revert this one https://review.opendev.org/#/c/465044/, harmless in itself, but pointless once the first one has been reverted

Long term code fix proposals:
What the os.initialize_connection api call is meant to: allow simple users to use cinder as block storage as a service
in order to attach volumes outside the scope of any virtual machines/nova.
Thus, information returned by this call should give enough information for a volume attach to be possible for the caller but they should not disclose
anything that would allow him to do more than that.
Since it is not possible at all with ceph to do so (no tenant isolation within ceph cluster),
the related cinder backend for ceph should not implement this route at all
There is indeed no reason why cinder should disclose anything here about ceph cluster, including hosts, cluster-ids,
if the attach is doomed to fail for users missing secret informations anyway.
Then, any 'admin' service using this call to locally attach the volumes (nova-compute, cinder-backup...) should be modified to:
- check caller rw permissions on requested volumes
- escalate the request
- go through a new admin api route, not this 'user' one

description: updated
Sean McGinnis (sean-mcginnis) wrote :

Eric, is this something you can comment on? Or know someone on the team that can?

Changed in cinder:
assignee: nobody → Eric Harney (eharney)
Jeremy Stanley (fungi) wrote :

Since this report concerns a possible security risk, an incomplete security advisory task has been added while the core security reviewers for the affected project or projects confirm the bug and discuss the scope of any vulnerability along with potential solutions.

Changed in ossa:
status: New → Incomplete
description: updated
Jeremy Stanley (fungi) wrote :

Note that it's possible for Nova to leak Ceph credentials under certain situations if the fix for bug 1837877 (OSSA-2019-003) is not applied.

Brian Rosmaita (brian-rosmaita) wrote :

First, we need to be clear that if the rbd_keyring_conf option is NOT used (which is the standard deployment scenario), this vulnerability does not arise.

Second, my proposal for fixing this is:
(1) Deprecate the option now (in Ussuri) for removal in V.
(2) Issue the OSSA explaining the problem and explaining that the mitigation is to not use the rbd_keyring_conf option.

Jeremy Stanley (fungi) wrote :

Based on this, I think that (2) will more correctly be an OSSN (security note rather than advisory) since it's recommending a change in configuration rather than providing a patch for deployers to apply. Probably closest classification in our report taxonomy is B2 "A vulnerability that can only be fixed in master, security note for stable branches, e.g., default config value is insecure" (though in this case it sounds like it's at least not the default value): https://security.openstack.org/vmt-process.html#incident-report-taxonomy

Jeremy Stanley (fungi) wrote :

Er, I meant class B1 there (which is the classification matching the description I quoted).

Brian Rosmaita (brian-rosmaita) wrote :

Thanks, Jeremy. An OSSN for class B1 sounds correct to me.

I'd like to get some feedback on my proposal to fix this by simply removing the option. It's not clear to me from the patch that introduced the rbd_keyring_conf (or the bug associated with it) why this config option is necessary, but I may be missing something.

Walt Boring (walter-boring) wrote :

Just so we are clear here. The keyring itself is not being passed in the call, what is being passed is the path and name of the keyring file on disk. So the credentials aren't being leaked here, just the name and path of the file that exists. You would still have to get access on the host where the keyring exists to even read it.

Walt Boring (walter-boring) wrote :

Also, as of note, the reason this feature was put in place to begin with is that a host can have access to multiple ceph clusters, and it's not always possible to force the filename to the 'standard' naming scheme. This would in fact make cinder not able to talk to a secondary cluster that requires a different keyring filename.

Walt Boring (walter-boring) wrote :

ok my bad, I was looking at the original patch that seemed like it was returning the file path. returning the key itself is not good.

OK, to move this along:

1. it looks like there is not a security fix we can issue at this time without redesigning the feature
2. the vulnerability is restricted to non-standard deployments
3. there is a workaround (the "Quick Workaround" outlined by Raphael in the Bug Description)
4. so, as of now, the mitigation is to use the workaround
5. since a patch isn't coming, we should issue an OSSN describing the vulnerability and the workaround
6. Cinder will deprecate the option referencing the OSSN
7. we can then make this bug public and continue a discussion of Walt's point that this is a legitimate use case and figure out how to address it

It would be great to have 1-6 done this week so that we can discuss this at the Forum/PTG next week so we can get an idea of how many users this affects (in the sense that they can't do the workaround, i.e., point #7).

Jeremy Stanley (fungi) wrote :

Brian's plan makes sense aside from #7. Instead I recommend we switch the bug to public while seeking volunteers to draft a note about it at https://wiki.openstack.org/wiki/Security_Notes for increased visibility. Keeping it under a private embargo when there's no patch forthcoming serves no purpose that I can see, since the advised workaround is already available and can be found in the discussion here.

Gage Hugo (gagehugo) wrote :

Agreed on switching this to public and having someone draft an OSSN for this.

I have no objection to Jeremy's recommendation to make the bug public first and then draft the OSSN.

OK, to continue moving this along, I will volunteer to draft the OSSN.

Jeremy Stanley (fungi) wrote :

Since there were no objections to the revised plan, I've switched the report to public, set the advisory task to won't fix and added a confirmed task for a security note. Thanks!

Changed in ossa:
status: Incomplete → Won't Fix
Changed in ossn:
status: New → Confirmed
assignee: nobody → Brian Rosmaita (brian-rosmaita)
description: updated
information type: Private Security → Public
tags: added: security

Using the Cinder part of the bug to track deprecating the option for removal in V; deprecation notice should reference the OSSN.

Changed in cinder:
assignee: Eric Harney (eharney) → Brian Rosmaita (brian-rosmaita)
importance: Undecided → Low
milestone: none → ussuri-1
status: New → In Progress

A note about step 2 in the Quick Workaround in the bug description: Gorka Eguileor noticed that the correct file location is actually:


See https://opendev.org/openstack/os-brick/src/commit/87171abef8bf2336f15ce3a7949f77d7999e11b7/os_brick/initiator/connectors/rbd.py#L76

Gage Hugo (gagehugo) on 2019-10-31
Changed in ossn:
status: Confirmed → In Progress
Changed in ossn:
status: In Progress → Fix Released
Changed in cinder:
status: In Progress → Fix Committed

This actually hasn't merged yet. It's been approved but has been stuck in the gate due to circumstances beyond Cinder's control.

Changed in cinder:
status: Fix Committed → In Progress

Reviewed: https://review.opendev.org/692428
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=b3c68b777a53e78fb77b7af98952bcc7d97bb06f
Submitter: Zuul
Branch: master

commit b3c68b777a53e78fb77b7af98952bcc7d97bb06f
Author: Brian Rosmaita <email address hidden>
Date: Thu Oct 31 14:17:19 2019 -0400

    Deprecate rbd_keyring_conf option

    This option presents a security risk; see OSSN-0085.

    Change-Id: I345a3b4bf3b328b0e547016f481518d252f734b9
    Partial-bug: #1849624

Changed in cinder:
status: In Progress → Fix Committed
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers