CVE-2026-89811

drm/amdkfd: Add TLB flush after MES queue eviction/suspension

Published Sep 16, 2026 CVSS 3.1 · 8.8 High EPSS < 1% KEV · Not listed
Description
In the Linux kernel, the following vulnerability has been resolved:

drm/amdkfd: Add TLB flush after MES queue eviction/suspension

MES (Micro Engine Scheduler) does not perform heavy-weight TLB
invalidation after unmapping queues, unlike HWS which does this
automatically. This causes a race condition where in-flight DMA
descriptors can access memory that has been unmapped, leading to page
faults and GPU queue hangs during SVM page migration.

The issue manifests as KFDSVMRangeTest.MultiThreadMigrationTest
failures on gfx1151 (Strix Point) with XNACK mode 1 enabled - the GPU
compute queue hangs with packets submitted but never consumed.

Add kfd_flush_tlb() calls after MES queue removal in two locations:
- evict_process_queues_cpsch(): after all queues removed during eviction
- suspend_queues(): after debug/criu queue suspension (with mem_fence barrier)

This ensures all in-flight memory accesses from unmapped queues are
flushed before memory is freed or migrated.

(cherry picked from commit f5c4f88e0f9c45a8fb9dfac0c1df726c95e41b77)

Analysis

AI-generated

Technical impact

A race condition in the Linux kernel’s AMD KFD driver allows in‑flight DMA descriptors to access memory that has already been unmapped during MES queue eviction or suspension, leading to page faults and GPU compute queue hangs. This interruption can suspend GPU workloads and effectively deny service to applications relying on the graphics subsystem.

Affected Systems

Linux kernel versions lacking the recent kernel Float‑based TLB flush after MES queue removal in the drm/amdkfd component are affected. The vulnerability applies to any system using this component prior to the inclusion of the patch in the kernel source tree.

Risk and Exploitability

The vulnerability scores a CVSS of 8.8, indicating high severity, but the EPSS score is below 1%, meaning the likelihood of exploitation is very low. It is not listed in CISA’s KEV catalog. The attack vector is likely local and requires the ability to drive GPU queues, so an attacker would need to execute privileged or high‑level code that interacts with the graphics driver to trigger the race.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Update the Linux kernel to a version that includes the kfd_flush_tlb patch (commit f5c4f88e0f9c45a8fb9dfac0c1df726c95e41b77).
  • Upgrade the AMD KFD driver to a release that implements TLB flushing after MES queue eviction and suspension.
  • If a kernel upgrade is not immediately possible, avoid using graphics queues that trigger MES eviction by disabling features such as SVM page migration or XNACK mode 1 until the patch is applied.

Generated by OpenCVE AI on September 18, 2026 at 09:16 UTC.

Tracking

Sign in to view the affected projects.

Advisories

Source ID Title
Debian DLA Debian DLA DLA-4904-1 linux-6.12 security update
Debian DLA Debian DLA DLA-4906-1 linux security update
Ubuntu USN Ubuntu USN USN-9002-1 Linux kernel (NVIDIA Tegra) vulnerabilities
Ubuntu USN Ubuntu USN USN-9019-1 Linux kernel (Azure) vulnerabilities
Ubuntu USN Ubuntu USN USN-9020-1 Linux kernel vulnerabilities
History
Type Values Removed Values Added
Weaknesses CWE-362
CWE-416
Type Values Removed Values Added
Metrics cvssV3_1

{'score': 8.8, 'vector': 'CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H'}

Type Values Removed Values Added
Description In the Linux kernel, the following vulnerability has been resolved: drm/amdkfd: Add TLB flush after MES queue eviction/suspension MES (Micro Engine Scheduler) does not perform heavy-weight TLB invalidation after unmapping queues, unlike HWS which does this automatically. This causes a race condition where in-flight DMA descriptors can access memory that has been unmapped, leading to page faults and GPU queue hangs during SVM page migration. The issue manifests as KFDSVMRangeTest.MultiThreadMigrationTest failures on gfx1151 (Strix Point) with XNACK mode 1 enabled - the GPU compute queue hangs with packets submitted but never consumed. Add kfd_flush_tlb() calls after MES queue removal in two locations: - evict_process_queues_cpsch(): after all queues removed during eviction - suspend_queues(): after debug/criu queue suspension (with mem_fence barrier) This ensures all in-flight memory accesses from unmapped queues are flushed before memory is freed or migrated. (cherry picked from commit f5c4f88e0f9c45a8fb9dfac0c1df726c95e41b77)
Title drm/amdkfd: Add TLB flush after MES queue eviction/suspension
First Time appeared Linux
Linux linux Kernel
CPEs cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*
Vendors & Products Linux
Linux linux Kernel
References
Weaknesses
  • CWE-362

    Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition')

  • CWE-416

    Use After Free

Sources

Status
PUBLISHED
Assigner
Linux
Published
Updated
Oct 30, 2026, 12:33 PM
Reserved
Sep 11, 2026, 7:38 PM
Status
Received
Published
Sep 16, 2026, 11:16 AM
Modified
Oct 30, 2026, 1:20 PM