CRITICAL Introduced in 6.16
kvm VNCR TLB Race
CVE-2026-89915
CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
KernelScan AI7.5HIGH
01Description
In the Linux kernel, the following vulnerability has been resolved: KVM: arm64: Remove VM-wide VNCR mapping counter The global VNCR mapping counter is used to decide whether an L1 provided VNCR page is mapped in L0 on any CPU at the point of dealing with a TLB invalidation. It is incremented when a mapping is made in the fixmap, and decremented when unmapped. As it turns out, this tracking has several flaws: - we are trying to invalidate TLBs, and the mapping is only an opportunistic consequence of the TLB. Checking this counter to decide whether a TLB needs to be invalidated may result in missed invalidations. - an L1 vcpu invalidating its own TLB (a very likely case) will not succeed in invalidating the VNCR pseudo TLB because that page is not mapped in L0 at this stage. Given that this tracking fails at delivering the minimum guarantees that are required and is only a performance optimisation, remove it completely.
02KernelScan AI Analysis
Risk summary
A guest VM running nested virtualization on arm64 can exploit a flawed VNCR mapping counter to trigger missed TLB invalidations, potentially leading to stale memory mappings and memory corruption in the host kernel. This requires a guest with nested virtualization enabled and can result in confidentiality, integrity, and availability impact on the host.
Vulnerability analysis
The VNCR mapping counter in KVM arm64 nested virtualization is used to decide whether TLB invalidation is needed for VNCR pages, but it fails to provide correct tracking: TLB invalidations can be missed when the counter is zero but stale mappings exist, and a guest vcpu invalidating its own TLB may not invalidate the VNCR pseudo TLB because the page is not yet mapped in L0. This can lead to stale memory mappings being used, potentially allowing a malicious L1 guest to access memory it should not, resulting in memory corruption. The fix removes the counter entirely and always takes the slow path for TLB invalidation when VNCR is in use, ensuring correctness at the cost of performance. This vulnerability is reachable from a guest VM using nested virtualization on arm64.
Lifecycle
03Fix Versions
| Branch | Introduced | Fixed in | Patch commit |
|---|---|---|---|
| 6.18 | 6.16 | 6.18.52 | 5453b85c7ebb |
| mainline | 6.16 | 7.3-rc1 | c55bc773b6e8 |
| 7.2 | 6.16 | 7.2.5 | 87c2bbf18982 |