CRI-O v1.23.3

The release notes have been generated for the commit range v1.23.2…724e932 on Wed, 19 Oct 2022 17:00:04 UTC.

Downloads

Download one of our static release bundles via our Google Cloud Bucket:

CRI-O now logs the stage of container or pod creation under system load. This allows users to find why their creation requests are stalling. (#6303, @haircommander)

Fix CVE-2022-27652 by dropping and refusing to add any inheritable capabilities (#5770, @haircommander)
Fix a bug where ExecSync requests (exec probes) could use an arbitrary amount of memory and disk. Output from ExecSync requests is now limited to 16MB (the amount that exec output was limited to in the dockershim). Disk limiting requires conmon 2.1.2 to work. See https://github.com/cri-o/cri-o/security/advisories/GHSA-fcm2-6c3h-pg6j and CVE-2022-1708 for more information. (#5937, @haircommander)
Fix a bug where child processes of containers in the host’s PID namespace appear to leak after the child exits (#5977, @haircommander)
Fix a panic caused by using the improper Result structure version when setting up CNI plugins (#6320, @juanfresia)
Revert fix for CVE-2022-27652 by re-adding inheritable capabilities. While there is a workaround, we believe this causes regression mid cycle with is contrary to CRI-O’s backporting policy. The risk of the CVE is low, and so there is little risk in reverting here. (#6168, @haircommander)
Update cni_default_network to not be commented by default in the output of crio config (#5566, @haircommander)

Disable systemd-mode cgroup detection if /sys/fs/cgroup is bind mounted from the host (#5815, @openshift-cherrypick-robot)
Fix a bug where CRI-O would leak a log file if a container failed to be created and the pod hadn’t yet been cleaned up. (#5817, @openshift-cherrypick-robot)
Fix a bug where a container is stuck in INIT because CRI-O believes it to be paused, and never updates the state (#6143, @openshift-cherrypick-robot)
Fix a bug where exit files were never cleaned up from /run/crio/exits (#6171, @openshift-cherrypick-robot)
Fix a rare deadlock while communicating to systemd (RHBZ 2082344) (#5961, @openshift-cherrypick-robot)
Fix a segfault when multiple container stops come in for the same container (#5782, @openshift-cherrypick-robot)
Fixed io.kubernetes.cri-o.TrySkipVolumeSELinuxLabel annotation usage with the OpenShift MCS. (#5792, @openshift-cherrypick-robot)
Internal pod and container creation timeouts now account for changes in runtime-request-timeout in the Kubelet (#5853, @openshift-cherrypick-robot)
Use default umask 0o022 if CRI-O runs under a different umask value. (#6053, @openshift-cherrypick-robot)

Nothing has changed.