---
id: "CVE-2026-99119"
kind: "vulnerability"
aliases: ["CVE-2026-99119"]
title: "Linux kernel AMD GPU driver: user queue buffer validation flaw on eviction"
severity: "unknown"
cvss: null
epss: null
exploited: false
status: "open"
category: "infrastructure"
products: []
ecosystems: []
tags: ["amd-gpu", "amdgpu", "infrastructure", "linux-kernel"]
not_affected_if: []
published_at: "2026-10-28T11:17:31+00:00"
first_seen_at: "2026-10-28T11:32:34+00:00"
updated_at: "2026-10-28T11:37:34+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-99119", "first_seen_at": "2026-10-28T11:32:34+00:00"}]
---

# Linux kernel AMD GPU driver: user queue buffer validation flaw on eviction

In the Linux kernel's AMD GPU (amdgpu) user queue code, buffer objects that were evicted before a user queue was created are not properly validated, so they never get an eviction fence re-armed. The sources do not state what an attacker gains from this beyond the fix description.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- unknown

## Fix

- Update to a Linux kernel release that includes upstream commit 18decfc9e8cc0fa70b6de9944cc27a37729edfbd (the fix is cherry-picked into stable branches; check the distribution's advisory for the fixed kernel version)

## Timeline

- 2026-10-28T11:32:34+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-99119

## From the sources

In the Linux kernel, the following vulnerability has been resolved:

drm/amdgpu/userq: actually validate userq BOs

This was supposed to loop over the individual evicted list, as the
kernel and always_valid evicted lists were already handled in
amdgpu_vm_validate.

Without this, BOs that were created and evicted before a userq was made
would never end up in an eviction fence rearm.

Also add the missing ref to the BO before drm_exec locks and refs it.

(cherry picked from commit 18decfc9e8cc0fa70b6de9944cc27a37729edfbd)

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_