Linux kernel flaw CVE-2026-89775 allows ARM64 KVM host memory access
CVE-2026-89775 can let nested ARM64 KVM guests read and write freed host memory; Linux patches exist and no attacks are reported.
Researcher Hyunwoo Kim disclosed CVE-2026-89775, a type-truncation flaw in Linux KVM/arm64 nested virtualization that can leave a freed host page writable so a guest reads and writes it, described as 64 bits at a time, at a fixed address without a VM exit or hardware trap. The issue is local and requires nested virtualization, which is off by default and needs Armv8.4 FEAT_NV2; reports also describe a second path in which local users on hosts with open or world-writable /dev/kvm (mode 0666 on some distributions) could gain root. Vendor scores range from 7.8 to 9.3. The code was introduced on May 14, 2025 and fixed upstream on August 6, 2026 in commit 8053393680d4, with The Hacker News also listing fixes in Linux 6.18.51, 7.2.5, and 7.3-rc1 affecting RHEL 10 and Ubuntu 26.04 kernels. Sources disagree on mitigations: Red Hat is said to offer no workaround, while another report recommends turning off nested virtualization and limiting /dev/kvm until patched. No proof of concept has been published, no attacks have been observed, the flaw is not in CISA's KEV list, and predicted exploitation is below 1%; it is the fourth KVM guest-to-host escape Kim disclosed this year, after Januscape, Zapscape, and ITScape.
- CVE-2026-89775, disclosed by Hyunwoo Kim, is a Linux KVM/arm64 type-truncation flaw that can let a nested guest read and write a freed host page (64 bits at a time) at a fixed address without a VM exit.
- The guest-to-host path is local-only and needs nested virtualization, which is off by default and requires Armv8.4 FEAT_NV2 hardware.
- A second path is reported: local users on systems with open or world-writable /dev/kvm (mode 0666 on some distributions) could gain root.
- Vulnerable code was introduced on May 14, 2025, and fixed upstream on August 6, 2026 in commit 8053393680d4; The Hacker News also lists fixes in Linux 6.18.51, 7.2.5, and 7.3-rc1.
- RHEL 10 and Ubuntu 26.04 kernels are affected; vendor scores range from 7.8 to 9.3.
- Sources disagree on mitigations: Red Hat is reported to offer no workaround, while GBHackers advises disabling nested virtualization and limiting /dev/kvm until patched.
- No proof of concept has been published, no attacks have been observed, the flaw is not in CISA's KEV catalog, and predicted exploitation is below 1%.
- It is the fourth KVM guest-to-host escape Kim disclosed this year, after Januscape, Zapscape, and ITScape.
Coverage timelineoldest first · each row is one article
- · 4d agoNew Linux Kernel Flaw Gives ARM64 KVM Guests Read-Write Access to Host Memory
The Hacker News· 52
Linux kernel flaw CVE-2026-89775 lets ARM64 KVM guests read and write freed host memory, enabling guest-to-host escape.
- · 4d agoCritical Linux KVM Flaw Enables Guest-to-Host Escape on ARM64 Systems
GBHackers· 61
A Linux KVM bug on ARM64 with nested virtualization can let a guest read and write host kernel memory.
- · 4d agoLinux KVM/arm64 Vulnerability Lets Attackers Escape Virtual Machines and Gain Host Access
Cyber Security News· 68
Vulnerabilities in this storyAll →
- CVE-2026-897759.3—Improper TLB invalidation in Linux kernel KVM/arm64 risks cross-VM memory exposurepublished · Linux kernel - KVM arm64 nested virtualization (VNCR) TLB invalidation path
| CVE | Vulnerability | CVSS | EPSS | Flags | Affected | Exposure | Published |
|---|---|---|---|---|---|---|---|
| CVE-2026-89775 | Improper TLB invalidation in Linux kernel KVM/arm64 risks cross-VM memory exposure CVE-2026-89775 is a flaw in the Linux kernel's KVM implementation for arm64, where the helper pgshift_level_to_ttl() mishandles negative walk levels when computing the size covered by a stage-1 TLB invalidation on the VNCR (nested virtualization) path. When a guest's stage-1 MMU is off, the level is encoded as the special value -127 (S1_MMU_DISABLED), but this is cast to a u8 and truncated to its bottom two bits, producing an invalidation size of 0 so stale TLB entries survive an operation that should have removed them. On a hypervisor, stale translations can allow memory mappings to outlive the pages they refer to, potentially letting a local attacker or guest read or corrupt memory across the trust boundary, consistent with the critical 9.3 CVSS (local attack, scope change, high confidentiality/integrity/availability impact). Only arm64 Linux systems running KVM with the nested-virtualization (VNCR/FEAT_NV2) code paths are affected; the advisory does not specify concrete affected or fixed version numbers. No public proof-of-concept, CISA KEV listing, or known exploitation has been reported. |