GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel KVM: use-after-free from consuming an inactive gfn_to_pfn_cache via Xen shared-info race

CVSS 7.8CVE-2022-49882Kernel, userspace & hypervisorcurated

Impact

kvm_gpc_check() and kvm_gpc_refresh() did not reject an inactive gfn_to_pfn_cache, so a cache could end up valid but not on the mmu_notifier invalidation list - a classic use-after-free setup, reachable by racing KVM_XEN_ATTR_TYPE_SHARED_INFO against KVM_XEN_HVM_EVTCHN_SEND until the event-channel path dereferences a NULL or stale khva. A host-local attacker with /dev/kvm can crash the hypervisor, and a kernel UAF of this shape is the sort that gets driven toward privilege escalation. On a GPU host this means every co-tenant job on the node goes with it. Only hosts running KVM with Xen emulation enabled are affected; a plain KVM/QEMU fleet without KVM_XEN does not reach this code.

Who can reach it

Local user with access to /dev/kvm on a host built with KVM Xen emulation; the attacker drives it through VM ioctls. No remote path and no unauthenticated reach.

What to do

Update to a stable kernel carrying the fix (commits bfa9672f8fc9 / ecbcf030b456) and reboot the host, or reload the kvm module with all guests stopped. If KVM Xen emulation is not needed - it is not, on a normal GPU cloud - build or configure without it, which closes the path without a maintenance window for the kernel itself.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.