Entreprise française de cybersécurité depuis 2004
☎ 03 60 47 09 81 - info@securiteinfo.com
☎ 03 60 47 09 81 - info@securiteinfo.com
In the Linux kernel, the following vulnerability has been resolved:perf/x86/intel: KVM: Mask PEBS_ENABLE loaded for guest with vCPU's value.When generating the MSR_IA32_PEBS_ENABLE value that will be loaded onVM-Entry to a KVM guest, mask the value with the vCPU's desired PEBS_ENABLEvalue. Consulting only the host kernel's host vs. guest masks results inrunning the guest with PEBS enabled even when the guest doesn't want to usePEBS. Because KVM uses perf events to proxy the guest virtual PMU, simplylooking at exclude_host can't differentiate between events created by hostuserspace, and events created by KVM on behalf of the guest.Running the guest with PEBS unexpectedly enabled typically manifests ascrashes due to a near-infinite stream of #PFs. E.g. if the guest hasn'twritten MSR_IA32_DS_AREA, the CPU will hit page faults on address '0' whentrying to record PEBS events.The issue is most easily reproduced by running `perf kvm top` from beforecommit 7b100989b4f6 ("perf evlist: Remove __evlist__add_default") (afterwhich, `perf kvm top` effectively stopped using PEBS). The userspace sideof perf creates a guest-only PEBS event, which intel_guest_get_msrs()misconstrues a guest-*owned* PEBS event.Arguably, this is a userspace bug, as enabling PEBS on guest-only eventssimply cannot work, and userspace can kill VMs in many other ways (thereis no danger to the host). However, even if this is considered to be baduserspace behavior, there's zero downside to perf/KVM restricting PEBS toguest-owned events.Note, commit 854250329c02 ("KVM: x86/pmu: Disable guest PEBS temporarilyin two rare situations") fixed the case where host userspace is profilingKVM *and* userspace, but missed the case where userspace is profiling onlyKVM.
No PoCs from references.
- https://github.com/runwhen-contrib/helm-charts
- https://github.com/w4zu/Debian_security