{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2026-31566",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2026-31566"
  ],
  "published": "2026-04-24T14:37:34Z",
  "summary": "drm/amdgpu: Fix fence put before wait in amdgpu_amdkfd_submit_ib",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: Fix fence put before wait in amdgpu_amdkfd_submit_ib\n\namdgpu_amdkfd_submit_ib() submits a GPU job and gets a fence\nfrom amdgpu_ib_schedule(). This fence is used to wait for job\ncompletion.\n\nCurrently, the code drops the fence reference using dma_fence_put()\nbefore calling dma_fence_wait().\n\nIf dma_fence_put() releases the last reference, the fence may be\nfreed before dma_fence_wait() is called. This can lead to a\nuse-after-free.\n\nFix this by waiting on the fence first and releasing the reference\nonly after dma_fence_wait() completes.\n\nFixes the below:\ndrivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c:697 amdgpu_amdkfd_submit_ib() warn: passing freed memory 'f' (line 696)\n\n(cherry picked from commit 8b9e5259adc385b61a6590a13b82ae0ac2bd3482)",
  "severity": [
    {
      "type": "CVSS_V3",
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H"
    }
  ],
  "affected": [
    {
      "package": {
        "ecosystem": "Raspberry Pi OS:12",
        "name": "linux"
      },
      "versions": [
        "1:6.1.47-1+rpt4"
      ],
      "ecosystem_specific": {
        "binaries": [
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-2712",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v6",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v7",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v7l",
            "binary_version": "1:6.1.47-1+rpt4"
          },
          {
            "binary_name": "linux-image-6.1.0-rpi3-rpi-v8",
            "binary_version": "1:6.1.47-1+rpt4"
          }
        ]
      },
      "database_specific": {
        "per_version": {
          "1:6.1.47-1+rpt4": {
            "linux_commit": "655fc658a15ae7a6f37103754adb39ba52a9a14e",
            "upstream_version": "6.1.47",
            "assessment": "commit ancestry (strak rule)",
            "upstream_fixed": "6.1.168",
            "compiled_filter": false
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2026-31566"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/bc7760c107dc08ef3e231d72c492e67b0a86848b"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/e23602eb0779760544314ed3905fa6a89a4e4070"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/138e42be35ff2ce6572ae744de851ea286cf3c69"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/39820864eacd886f1a6f817414fb8f9ea3e9a2b4"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/42d248726a0837640452b71c5a202ca3d35239ec"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/7150850146ebfa4ca998f653f264b8df6f7f85be"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-08-05T13:16:28Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 7.8,
      "cvss_v3_severity": "High",
      "scenario": "AV:L - The vulnerable `amdgpu_amdkfd_submit_ib()` path is reached from local userspace through `/dev/kfd` and AMDGPU/KFD queue ioctls, not through network, adjacent, or physical input.\nAC:L - On an affected AMDGPU KFD/NO_HWS Hawaii deployment, an attacker can drive the queue create/destroy path and repeatedly trigger the fence lifetime bug without relying on victim-controlled timing. When uncertain, the lower-complexity option is appropriate.\nPR:L - The path requires local access to the KFD/AMDGPU device interfaces and a render/DRM VM FD, but no kernel administrative capability is checked by the relevant KFD queue ioctls. This is ordinary low-privileged local GPU access rather than real root.\nUI:N - After obtaining local device access, the attacker can issue the ioctl sequence themselves. No separate victim action is required.\nS:U - The impact is within the same kernel security authority: corrupting or crashing kernel/driver state from a local kernel attack surface. It does not cross a VM, IOMMU, or comparable security boundary.\nC:H - The bug is a use-after-free of a `dma_fence`/AMDGPU fence object used by `dma_fence_wait()`. Per kernel scoring guidance, UAF memory corruption can be leveraged for high confidentiality impact.\nI:H - The same UAF exposes kernel object lifetime corruption in a wait path using function pointers, locks, callbacks, and list state. Per guidance, UAF memory corruption is scored as high integrity impact.\nA:H - Even without full exploitation, using a freed fence in `dma_fence_wait()` can cause kernel warnings, crashes, hangs, or panics. A kernel UAF therefore has high availability impact."
    }
  }
}