CVE-2026-46059: Linux Kernel Privilege Escalation Flaw

CVE-2026-46059 Overview

CVE-2026-46059 is a Linux kernel vulnerability in the nested SVM (nSVM) implementation of the Kernel-based Virtual Machine (KVM) subsystem. The flaw affects how KVM handles the NextRIP field in vmcb02 when emulating CPUs without the NRIPS (Next RIP Save) feature for nested guests. After the first L2 VMRUN, KVM incorrectly continues using the current RIP instead of the updated NextRIP, leading to incorrect instruction pointer state during nested virtualization with injected soft interrupts.

Critical Impact
Incorrect NextRIP handling in nested SVM virtualization can cause guest state corruption and incorrect instruction execution flow when L1 hypervisors run L2 guests with NRIPS disabled and soft interrupts injected.

Affected Products

Linux kernel — KVM nested SVM (nSVM) subsystem
Systems using AMD SVM virtualization with NRIPS feature disabled
Nested virtualization deployments with L1 hypervisors running L2 guests

Discovery Timeline

2026-05-27 - CVE-2026-46059 published to NVD
2026-05-27 - Last updated in NVD database

Technical Details for CVE-2026-46059

Vulnerability Analysis

The vulnerability resides in the Linux KVM nested SVM code path that emulates CPUs lacking the NRIPS feature. When NRIPS is disabled, L1 hypervisors do not provide NextRIP when running an L2 guest with an injected soft interrupt. Instead, L1 advances the current RIP before executing the instruction. KVM compensates by using the current RIP as the NextRIP value in vmcb02.

The flaw emerges after the first L2 VMRUN completes. At that point, NextRIP is updated either by the CPU or by KVM itself, meaning the current RIP is no longer the correct value to populate vmcb02. Reusing the stale current RIP after save/restore causes incorrect instruction pointer state in subsequent nested guest execution.

Root Cause

The root cause is incorrect state tracking after the first nested VMRUN. KVM unconditionally used the current RIP as NextRIP in vmcb02 without distinguishing between the initial nested run (where the current RIP is correct) and subsequent runs (where NextRIP has been updated and must be used instead). The same logical flaw applied to soft_int_next_rip, a narrower variant of the same condition.

Attack Vector

The vulnerability requires a nested virtualization configuration where an L1 hypervisor runs L2 guests on hardware without NRIPS, or with NRIPS disabled. A privileged L1 hypervisor injecting soft interrupts into an L2 guest can trigger the incorrect state condition across save/restore boundaries. Exploitation requires control over the L1 guest workload and is constrained to the nested SVM code path. See the upstream commits for technical details on the fix logic.

The vulnerability is described in prose only; refer to the kernel commit 3428ed1529a1 for the patch implementation.

Detection Methods for CVE-2026-46059

Indicators of Compromise

Unexpected guest crashes or instruction pointer corruption inside L2 virtual machines running under KVM nested SVM
Anomalous behavior in nested guests after soft interrupt injection, particularly on AMD systems with NRIPS unavailable or disabled
Kernel log messages indicating nested SVM state restore inconsistencies

Detection Strategies

Inventory Linux hosts running KVM with AMD SVM virtualization and identify kernel versions predating the upstream fix commits
Audit nested virtualization configurations to determine whether NRIPS is enabled on host CPUs and exposed to L1 guests
Correlate L2 guest stability issues with host kernel versions to identify systems running vulnerable nSVM code

Monitoring Recommendations

Monitor dmesg and kernel logs on virtualization hosts for KVM warnings related to nested SVM and VMCB state
Track kernel package versions across the virtualization fleet and alert on hosts that have not received the relevant stable kernel update
Log L2 guest reboot and crash events to surface patterns consistent with nested instruction pointer corruption

How to Mitigate CVE-2026-46059

Immediate Actions Required

Apply the upstream stable kernel updates referenced in the kernel.org commits to all affected Linux virtualization hosts
Identify hosts running nested virtualization with AMD SVM and prioritize them for patching
Coordinate maintenance windows with workload owners since updating the host kernel requires a reboot

Patch Information

The fix is published across the following upstream commits: 3428ed1529a1, 69fe1411a5ce, and 8d397582f6b5. The patch updates the nSVM logic to use the current RIP as vmcb02's NextRIP only when a nested run is pending, and to use NextRIP otherwise. The same treatment is applied to soft_int_next_rip. Apply the corresponding distribution kernel update once your Linux vendor backports these commits.

Workarounds

Disable nested virtualization on affected hosts by setting the kvm_amd module parameter nested=0 until the host kernel is patched
Avoid running L2 guests that rely on soft interrupt injection on hosts where NRIPS is unavailable or disabled
Where possible, enable NRIPS on host CPUs that support the feature to avoid the vulnerable emulation code path

bash

# Disable nested virtualization on AMD SVM hosts as a temporary workaround
sudo modprobe -r kvm_amd
sudo modprobe kvm_amd nested=0

# Persist the setting across reboots
echo "options kvm_amd nested=0" | sudo tee /etc/modprobe.d/kvm_amd.conf

# Verify nested virtualization is disabled
cat /sys/module/kvm_amd/parameters/nested