Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessGain host-kernel execution on arm64 pKVM host
-
ExploitReach hyp stack pages left in host stage-2
-
ExecutionRead or corrupt per-CPU hypervisor stack memory
-
ImpactBreak host/hypervisor isolation boundary
Vulnerability Assessment
| Exploitation Conditions | Requires the target to be running KVM in protected mode (pKVM) on arm64 - the vulnerable fix_host_ownership() ownership-transfer path only exists in that configuration. The attacker must already hold host-kernel-level privilege (CVSS PR:H, AV:L), i.e. local code execution within the untrusted host that pKVM isolates; it is not reachable by an unprivileged user or over a network. The concrete defect: each CPU's hyp stack is mapped in the private VA range (for its guard page) and has no linear-map alias, so the host-stage-2 ownership walker never removed those pages, leaving them accessible to the host. Limiting factors: arm64 only, pKVM must be enabled, requires high host privilege, and exploitation is local with no user interaction needed. |
| Impact | Signals are consistent and point to a real-but-bounded priority rather than an urgent widespread threat. The CVSS base score of 8.2 (CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:H) is driven mostly by the scope change (S:C) and full C/I/A impact, reflecting that the flaw crosses the host→hypervisor trust boundary; however, AV:L and PR:H mean the attacker must already be executing with high privilege locally (effectively host-kernel level), which is a significant barrier. EPSS is low at 0.20% (9th percentile), the CVE is NOT on the CISA KEV list, and no exploit or PoC reference is present - only vendor patch commits. The vuln.today decision label is 'this_week', consistent with a serious isolation bug that warrants prompt patching but is not an emergency. The signals do not conflict: low EPSS plus no KEV plus a local/high-privilege vector together indicate this is a severe correctness/isolation defect in a narrow configuration (arm64 protected KVM / pKVM), not a mass-exploitation candidate. Its importance is specific to the pKVM threat model, where the host is deliberately untrusted and is supposed to be unable to reach hypervisor memory; this bug defeats exactly that guarantee. |
| Exploit ScenarioDerived | On an arm64 system running KVM in protected mode (pKVM), an attacker who has already achieved host-kernel execution reads or corrupts the per-CPU hypervisor stack pages, which were left mapped in the host's stage-2 translation because fix_host_ownership() only walked the linear-map alias and missed the stacks' private VA range. This lets a compromised host violate the host/hypervisor isolation pKVM is designed to enforce, exposing or tampering with hypervisor state. No public exploit or PoC was identified at time of analysis; the only references are the upstream kernel fix commits. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 24, 2026
Linux kernel arm64 KVM running in protected mode (pKVM) fails to remove each CPU's hypervisor stack pages from the host stage-2 mapping, leaving hypervisor-private memory reachable by the host. The flaw sits in the fix_host_ownership() path, which only walks the linear-map alias of memblock regions and therefore misses stacks that are mapped solely in the private VA range for their guard page. Only arm64 systems with pKVM enabled are affected, and the attacker must already hold high privilege in the host kernel (CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:H, 8.2), so this is a local isolation-bypass primitive rather than a network-reachable flaw; no public exploit has been identified at time of analysis, it is not in CISA KEV, and EPSS is 0.20% (9th percentile).
RemediationAI
Vendor-released patch: Linux 6.18.55, 7.2.9 and 7.3-rc5 (stable commits e5dd2a91a7ae3a276e6d79806127e923936e241b, 22df195bff84300a3f04813d6148765c77ac565a, 3a8c562892b96f35bba1e00d5e455a15963bbb92); upgrade arm64 KVM hosts to the fixed build for their branch, or cherry-pick the corresponding stable commit if staying on an older series, and reboot so the hypervisor and host stage-2 setup are rebuilt. Given the this_week priority, KEV status of not listed and EPSS of 0.20%, schedule this within the current weekly patch cycle rather than as an emergency. Where patching must be deferred, the exposure is confined to arm64 hosts running pKVM in protected mode, so not enabling (or temporarily disabling) protected mode removes the vulnerable ownership-transfer path - at the cost of losing the confidential-computing isolation that pKVM provides for that workload, which may be unacceptable for multi-tenant confidential guests. Note that the flaw only yields hypervisor memory access to code that already holds host-kernel privilege, so tightening which principals can obtain host-kernel code execution is a useful but not complete compensating control; no public exploit has been identified at time of analysis.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
pKVM (protected KVM) on arm64 runs a hypervisor that is meant to be isolated from the untrusted host kernel: the host's own stage-2 translation tables are used to deny the host access to hypervisor memory. During ownership transfer, fix_host_ownership() iterates memblock regions using their linear-map alias to strip those pages from the host stage-2 mapping. Per-CPU hypervisor stacks are an exception: because they need a guard page, they are mapped in a private VA range and have no linear-map alias, so the walker never encountered them and their pages stayed mapped into host stage-2. The fix teaches the ownership walker to walk each stack's VA range with the same walker. The weakness is an access-control/isolation defect (no CWE was assigned in the supplied data) whose consequence is that host-side code retains read and write reachability to hypervisor stack memory, which is why the vector carries Scope:Changed and High confidentiality, integrity and availability impact on the protected component. The trigger is specific to arm64 protected-mode deployments; non-pKVM arm64 and other architectures do not exercise this path.
Know when the next one hits your stack
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:H
Attacker needs local host-kernel privilege (AV:L, PR:H) with no user interaction; the flaw crosses the host→hypervisor boundary (S:C) exposing and tampering hyp memory (C:H/I:H/A:H).
Primary rating from Vendor (Linux).
CVSS VectorVendor: Linux
Lifecycle Timeline
4DescriptionCVE.org
In the Linux kernel, the following vulnerability has been resolved:
KVM: arm64: Transfer the hyp stack pages out of the host stage-2
fix_host_ownership() walks only the linear-map alias of each memblock region, and the per-CPU hyp stack, mapped in the private VA range for its guard page, has none.
Walk each stack's VA range with the same walker.
Executive ViewAI
A high-severity isolation defect in the Linux kernel's arm64 protected KVM (pKVM) hypervisor means hypervisor-private memory on affected systems can be reached by the host, undermining the boundary pKVM is deployed to enforce. Exposure is confined to arm64 systems running pKVM, and triggering it requires an attacker who already holds high privilege in the host kernel, so this is a privilege-escalation and containment-risk issue rather than a remotely reachable threat. Leadership should care because the affected boundary is typically the core security control protecting confidential-computing and multi-tenant workloads on arm64 infrastructure.
Apply vendor patch.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Information disclosure occurs when an application unintentionally exposes sensitive data that aids attackers in reconnaissance or directly compromises security.
EUVD-2026-103706
GHSA-7374-3pfm-f9q8
NVD
CVE.org
EPSS
Exploit-DB