fix(cozystack): stop disabling KVM nested virtualization in Talos skills#18
Conversation
The talos-bootstrap skill wrote kvm_intel.nested=0 / kvm_amd.nested=0 into every generated node config and flagged their absence during the pre-apply review; the cluster-upgrade skill re-checked the kernel command line afterwards. That turned a temporary workaround for CVE-2026-53359 into a permanent policy the skills enforced on the operator's behalf. Nested virtualization is a feature real users depend on, and the skills have no basis to take it away. CVE-2026-53359 and its sibling CVE-2026-46113 are use-after-free bugs in the KVM x86 shadow MMU, and both are fixed in the kernel. Linux 6.18.38 carries both upstream patches, and Talos v1.13.6 ships that kernel. Running a patched Talos release is the remedy, so the host no longer has to give up the feature. Signed-off-by: Aleksei Sviridkin <f@lex.la>
|
Warning Review limit reached
Next review available in: 31 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (3)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Warning Gemini encountered an error creating the review. You can try again by commenting |
1 participant
Summary
The
cozystackplugin's Talos skills no longer disable KVM nested virtualization. They stop writingkvm_intel.nested=0/kvm_amd.nested=0into generated node configs, and stop verifying those args after an upgrade.Why
Turning nested virtualization off was a stopgap for CVE-2026-53359 ("Januscape"), adopted while no released Talos carried the kernel fix. Encoding it in the skills made it worse than a stopgap:
talos-bootstrapinjected the args into every generatednodes/<name>.yamland flagged their absence during the pre-apply review, andcluster-upgradere-checked the kernel command line afterwards. An operator who legitimately runs nested guests would have had the capability stripped at bootstrap, and would have been flagged as non-compliant every time they put it back. Nested virtualization is a feature real users depend on; the skills have no business taking it away.It is also unnecessary now. CVE-2026-53359 and its sibling CVE-2026-46113 are use-after-free bugs in the KVM x86 shadow MMU (
arch/x86/kvm/mmu/mmu.c), and a complete fix needs both upstream patches:81ccda30b4e8for the unexpected-role path and0cb2af2ea66afor the unexpected-GFN path. Stable kernels are cumulative, so on the 6.18 branch both patches are present from 6.18.38 onwards. Talos v1.13.6 ships Linux 6.18.38 and therefore carries both fixes. Keeping nodes on a patched Talos release is the remedy; giving up the feature is not.Change
talos-bootstrapskill: the generated node overlay no longer setsmachine.install.grubUseUKICmdline/extraKernelArgs, and the pre-apply review no longer demands thenested=0args.talos-bootstrapreference (manual-steps.md): same block dropped from the node-config example.cluster-upgradereference (post-upgrade-checks.md): the post-upgrade/proc/cmdlinecheck is gone.The three files are restored byte-for-byte to their pre-mitigation state. Nothing else changes: the skills do not gain a Talos version gate, and no other guidance is touched.