Bug 2178990
| Summary: | Containers with custom policy fails to start on OCP 4.13 duo to access denial to devices | |||
|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 9 | Reporter: | lpivarc | |
| Component: | container-selinux | Assignee: | Jindrich Novy <jnovy> | |
| Status: | CLOSED ERRATA | QA Contact: | atomic-bugs <atomic-bugs> | |
| Severity: | high | Docs Contact: | ||
| Priority: | unspecified | |||
| Version: | 9.3 | CC: | ajia, dbasunag, ddarrah, dollierp, dornelas, dshchedr, dwalsh, fdeutsch, jnovy, jwboyer, lfriedma, lsm5, mboddu, plautrba, qzhang, sgott, travier, tsweeney, vmojzis, ypu | |
| Target Milestone: | rc | Keywords: | Triaged | |
| Target Release: | 9.2 | |||
| Hardware: | Unspecified | |||
| OS: | Linux | |||
| Whiteboard: | ||||
| Fixed In Version: | container-selinux-2.205.0-1.el9_2 | Doc Type: | If docs needed, set a value | |
| Doc Text: | Story Points: | --- | ||
| Clone Of: | ||||
| : | 2179465 2181174 (view as bug list) | Environment: | ||
| Last Closed: | 2023-05-09 07:33:44 UTC | Type: | Bug | |
| Regression: | --- | Mount Type: | --- | |
| Documentation: | --- | CRM: | ||
| Verified Versions: | Category: | --- | ||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | ||
| Cloudforms Team: | --- | Target Upstream Version: | ||
| Embargoed: | ||||
| Bug Depends On: | ||||
| Bug Blocks: | 2177924, 2179392, 2179465, 2181174 | |||
This looks like a bug in the policy of whatever package shipped system_u:system_r:virt_launcher.process:s0:c804,c995 This is definitely not a container-selinux issue Let me just mention this was previously working, for example on 8.6 we are happily running this. My suspicion is that any workload on OCP using custom policy will face this issue. Are all policies expected to adjust to RHEL 9.2/OCP 4.13 now? In other words isn't this regression? So this was intentional. Because we wanted to allow two podman instances to run on a system and be isolated from each other, with one supporting container_file_t and the other using a different file type. Both being container_domains. I guess I can create a new domain, which works just like container_domain, but does not have access to container_file_t. If I add +manage_dirs_pattern(svirt_sandbox_domain, container_file_t, container_file_t) +manage_files_pattern(svirt_sandbox_domain, container_file_t, container_file_t) +manage_lnk_files_pattern(svirt_sandbox_domain, container_file_t, container_file_t) +manage_chr_files_pattern(svirt_sandbox_domain, container_file_t, container_file_t) +manage_blk_files_pattern(svirt_sandbox_domain, container_file_t, container_file_t) +manage_sock_files_pattern(svirt_sandbox_domain, container_file_t, container_file_t) Then I think that will fix your problem, and still give me the isolation I want. Brilliant, thank you for the clarity. What you're up to is completely sensible. Thank Daniel, It seems good to me as well. Can we, therefore, assume this will be fixed in 9.2 which would then OCP 4.13 consume? @dwalsh where is the new domain going to live, and is this something you can get together in the near term? FYI @dornelas and @travier We need a new version of container-selinux packaged up for RHEL. container-selinux-2.205.0 @dwalsh I'm assuming that adding container-selinux-2.205.0 needs to be a ZeroDay fix for RHEL 8.8 and 9.2. If you disagree, please let me know. @jnovy heads up and assigning to you for any further packaging or BZ needs. @travier heads up too. Re-requesting z+ and fixing ITR. Also dev acking as in the worst case we can remove the user_namespace class to allow the build to pass. *** Bug 2179392 has been marked as a duplicate of this bug. *** Setting back to Blocker+ and ITR of 9.2 Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (container-selinux bug fix and enhancement update), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2023:2206 The needinfo request[s] on this closed bug have been removed as they have been unresolved for 120 days |
Description of problem: Trying to run a container with custom policy[0] fails as the starting process cannot access devices (such as /dev/null). The produced avc is type=AVC msg=audit(1678952803.346:303): avc: denied { open } for pid=1905336 comm="virt-launcher-m" path="/dev/null" dev="tmpfs" ino=6 scontext=system_u:system_r:virt_launcher.process:s0:c804,c995 tcontext=system_u:object_r:container_file_t:s0:c804,c995 tclass=chr_file permissive=0 This seems to originate from https://github.com/containers/container-selinux/commit/24e57848527bcddad025316fa57493926ff1dfbf#diff-1cdb378311a88c884861f2d5996bca97ec20a49b1d66211dced87b2619c17ba9L826 [0] https://github.com/kubevirt/kubevirt/blob/main/cmd/virt-handler/virt_launcher.cil Version-Release number of selected component (if applicable): Running Openshift 4.13.0-0.nightly-2023-03-14-053612 which uses CentOS Stream CoreOS 413.92.202303061740-0 (Plow) Additionally: Name : container-selinux Epoch : 3 Version : 2.199.0 Release : 1.el9 Name : selinux-policy Version : 38.1.8 Release : 1.el9 How reproducible: Run container/pod with custom SELinux policy Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info: