GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux KVM pfncache: missed MMU-notifier invalidation maps a freed page into Xen-emulation caches

UnscoredCVE-2026-99472Kernel, userspace & hypervisorcurated

Impact

KVM's gfn_to_pfn_cache refresh can miss an MMU-notifier invalidation when the cached host virtual address is not backed by a memslot. It then publishes a mapping of a page that has already been freed. Later reads in kvm_xen_shared_info_init() or kvm_setup_guest_pvclock() turn this into a use-after-free in the host kernel. Public reproducers exist. It only matters on hosts that run KVM guests with Xen HVM emulation and HVA-based shared_info or vcpu_info attributes.

Who can reach it

Local. Needs access to /dev/kvm and the ability to set the KVM_XEN_ATTR_TYPE_SHARED_INFO_HVA or KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO_HVA attributes, which in practice means the VMM process or a user who can create VMs. The record does not describe a path for a guest to trigger it.

What to do

Install a kernel with the stable fix (a dedicated gpc_invalidate_seq), then drain and reboot each hypervisor host. Hosts that do not use KVM Xen emulation are not exposed through this path. The record gives no fixed version numbers.

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.