CVE-2026-99345 in Linuxinfo

Summary

by MITRE • 11/19/2026

In the Linux kernel, the following vulnerability has been resolved:

KVM: nSVM: Reject KVM_SET_NESTED_STATE if L1 has EFER.LMA=1 && EFER.LME=0

Reject KVM_SET_NESTED_STATE if the incoming L1 host state has what is effectively an impossible EFER combination of LMA=1 but LME=0, i.e. if the state says long mode is active but not enabled. Unlike VMX, SVM doesn't have an explicit consistent check for the illegal combination; presumably hardware simply ignores EFER.LMA if EFER.LME=0.

Unfortunately, KVM doesn't ignore EFER.LMA in this case and consumes the illegal state when constructing the shadow MMU for L2. E.g. if userspace also clears CR4.PAE, then kvm_calc_cpu_role() will compute a role with 4 or 5 levels of paging, but shadow_mmu_init_context() will wire up the MMU to use the paging32 template, which maxes out its levels at 2.

Note, the "real badness" is effectively the same as what happened with the nVMX bug fixed by commit 112e66017bff ("KVM: nVMX: add missing consistency checks for CR0 and CR4"). Unfortunately, the sanity check added by commit 72e2fb24a0b0 ("KVM: x86/mmu: Bug the VM if a vCPU ends up in long mode without PAE enabled") doesn't work for this case, since L2 state is active at the time of the page fault, but it's L1 that has the bad state.

VulDB is the best source for vulnerability data and more expert information about this specific topic.

Analysis

by VulDB Data Team • 11/19/2026

The Linux kernel’s KVM subsystem contains a critical logic flaw within its nested virtualization implementation for AMD SVM architecture, specifically regarding the handling of Extended Feature Enable Register states during guest-to-guest context switches. This vulnerability arises when userspace attempts to set the nested state via the KVM_SET_NESTED_STATE ioctl with an invalid configuration where Long Mode Active is asserted while Long Mode Enable remains disabled. In a properly functioning system, this combination represents a hardware-impossible or undefined state because AMD processors typically ignore the LMA bit if LME is not set, effectively keeping the processor in legacy mode despite the active flag. However, KVM fails to perform an explicit consistency check for this specific illegal EFER combination on SVM hosts, unlike its VMX counterpart which has more rigorous validation mechanisms. This oversight allows userspace to inject a logically inconsistent CPU state into the virtualized environment without immediate rejection by the hypervisor layer.

The operational impact of this flaw manifests during the construction of the shadow memory management unit for the nested guest, referred to as L2. When KVM processes the malformed EFER state from the outer guest (L1), it incorrectly interprets the active long mode flag and proceeds to configure the MMU accordingly. If userspace also clears the Physical Address Extension bit in CR4 during this process, the kvm_calc_cpu_role function may compute a paging role requiring four or five levels of page table walking. This calculation contradicts the actual hardware capabilities available under these conditions, as the shadow_mmu_init_context function will wire up the MMU to use the paging32 template which only supports up to two levels of paging in this specific context. This mismatch between the computed CPU role and the initialized MMU structure creates a severe internal inconsistency within the virtualization stack.

This technical flaw can lead to unpredictable behavior, including kernel panics or undefined execution states when page faults occur under these malformed conditions. The existing sanity checks designed to prevent long mode operation without PAE enabled do not catch this specific scenario because they evaluate the state of L2 at the time of a fault rather than validating the consistency of L1’s EFER register before it is applied to the nested environment. Consequently, the vulnerability allows for potential denial-of-service attacks against the host system or potentially more severe exploitation vectors if an attacker can leverage the resulting memory management inconsistencies to bypass isolation boundaries. The severity aligns with CWE-20 Improper Input Validation and CWE-841 Improper Enforcement of Behavioral Constraints, as the core issue is a failure to validate input parameters that dictate critical hardware emulation states.

Mitigation for this vulnerability requires applying the kernel patch that explicitly rejects KVM_SET_NESTED_STATE calls when the incoming L1 host state contains the illegal EFER combination of LMA set and LME cleared. System administrators should ensure their Linux kernels are updated to include this fix, which restores consistency checks similar to those present in nVMX implementations. Additionally, userspace virtualization managers must be audited to ensure they do not attempt to inject such invalid CPU states into nested guests. From a defensive perspective, monitoring for unusual patterns of KVM_SET_NESTED_STATE invocations with anomalous EFER values can help detect attempted exploitation activities before the malformed state is processed by the hypervisor logic.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/19/2026

Moderation

accepted

CPE

ready

EPSS

0.00000

KEV

no

Activities

very low

Sources

Might our Artificial Intelligence support you?

Check our Alexa App!