Linux Kernel CVE-2026-46059

Exploitation requires all of the following: (1) AMD x86 hardware with SVM support on a host running Linux kernel 5.8 through the affected commit range; (2) KVM nested virtualization explicitly enabled via the 'nested=1' module parameter for kvm_amd; (3) an L1 hypervisor guest that disables the NRIPS CPU feature when launching L2 guests - this is a non-default hypervisor configuration; (4) soft interrupt injection into the L2 guest by the L1 hypervisor prior to VMRUN; (5) a save/restore cycle (e.g., VM migration, suspend/resume) occurring after the first L2 VMRUN. Intel VMX-based hosts are entirely unaffected. Standard single-level KVM guests with no nested virtualization are unaffected. The PR:L requirement means the attacker must have a local account with sufficient privileges to interact with the KVM subsystem or influence L1 hypervisor configuration. The combination of AMD-only, nested-VM-required, and NRIPS-disabled conditions makes this a highly targeted rather than broadly exploitable flaw.

The CVSS 3.1 base score of 5.5 (AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H) reflects a local, low-complexity attack yielding high availability impact with no confidentiality or integrity consequences. The EPSS score of 0.02% (4th percentile) places this well below exploitation thresholds and aligns with the highly specific triggering conditions. The vulnerability is not listed in the CISA KEV catalog, and no public exploit code has been identified at time of analysis. Real-world priority is LOW: exploitation is confined to AMD hardware running KVM with nested virtualization explicitly enabled, an L1 hypervisor that disables NRIPS, and a soft interrupt injection followed by a save/restore cycle - a combination absent from typical production cloud or virtualization deployments. The official AC:L metric may underweight the specificity of required conditions; an independent assessment of AC:H is warranted given the multi-step, configuration-dependent trigger. The 'Code Injection' tag in the input is inconsistent with the CVSS C:N/I:N impact profile and the patch description - this appears to be a misclassification and should not elevate risk estimates.

A local user with standard access to an AMD KVM host configures an L1 hypervisor VM with NRIPS disabled and injects a soft interrupt into an L2 guest before issuing VMRUN. After the first L2 VMRUN completes and KVM updates the NextRIP, the attacker (or the natural VM lifecycle) triggers a save/restore cycle - for example via VM migration or suspend/resume - causing KVM to overwrite NextRIP with the stale current RIP, corrupting vmcb02 state and crashing the nested guest or the host KVM subsystem. No public exploit code exists, and the multi-step, configuration-dependent nature of this scenario substantially limits opportunistic exploitation.