Vulnerability-Lookup

GHSA-XPW2-C5CH-25QM

Vulnerability from github – Published: 2026-10-28 12:31 – Updated: 2026-10-31 15:30

Details

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

drm/amdgpu: keep the eviction fence until the last handle is closed

amdgpu_gem_object_close() detaches the user queue eviction fence from the BO before it checks bo_va->ref_count, so closing any handle of a BO removes the fence, even when the file still has another handle to the BO and the bo_va with its GPU mappings stays. The BO can then be evicted without suspending the user queues of the file, which keep accessing it through mappings that are only marked for update.

Seen on navi31 with a SDMA user queue: a BO opened twice in the same file (GEM_FLINK + GEM_OPEN) loses its eviction fence when the second handle is closed (traced attach/detach of the fence):

attach bo=ffff8c78a6c84c00 GEM_CLOSE handle=7 detach bo=ffff8c78a6c84c00

while handle 6 and the mapping are still in use.

Only detach the fence when the last reference to the bo_va goes away, matching amdgpu_gem_object_open(), which attaches it only when the bo_va is created. With this the fence stays until the file is closed.

(cherry picked from commit 8387982f03d4026b7dc11b0b5e3118298eee4d6a)

Severity

7.0 (High)


                  
                    CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H

Show details on source website

JSON

To clipboard

{
  "affected": [],
  "aliases": [
    "CVE-2026-99148"
  ],
  "database_specific": {
    "cwe_ids": [],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-10-28T11:17:34Z",
    "severity": "HIGH"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: keep the eviction fence until the last handle is closed\n\namdgpu_gem_object_close() detaches the user queue eviction fence from\nthe BO before it checks bo_va-\u003eref_count, so closing any handle of a BO\nremoves the fence, even when the file still has another handle to the\nBO and the bo_va with its GPU mappings stays. The BO can then be evicted\nwithout suspending the user queues of the file, which keep accessing it\nthrough mappings that are only marked for update.\n\nSeen on navi31 with a SDMA user queue: a BO opened twice in the same\nfile (GEM_FLINK + GEM_OPEN) loses its eviction fence when the second\nhandle is closed (traced attach/detach of the fence):\n\n  attach bo=ffff8c78a6c84c00\n  GEM_CLOSE handle=7\n  detach bo=ffff8c78a6c84c00\n\nwhile handle 6 and the mapping are still in use.\n\nOnly detach the fence when the last reference to the bo_va goes away,\nmatching amdgpu_gem_object_open(), which attaches it only when the\nbo_va is created. With this the fence stays until the file is closed.\n\n(cherry picked from commit 8387982f03d4026b7dc11b0b5e3118298eee4d6a)",
  "id": "GHSA-xpw2-c5ch-25qm",
  "modified": "2026-10-31T15:30:22Z",
  "published": "2026-10-28T12:31:05Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-99148"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/3c44cbcec5eb370713bbd2c315ed7c34d8ccec2b"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/6467c4cc83b94850bfea614e9f98276edf9a7c3f"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/dce8f7f03a47b1744dccf506268f9229b67fab47"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ]
}

CVE-2026-99148 (GCVE-0-2026-99148)

Vulnerability from cvelistv5 – Published: 2026-10-28 10:35 – Updated: 2026-10-31 12:43

Title

drm/amdgpu: keep the eviction fence until the last handle is closed

Summary

In the Linux kernel, the following vulnerability has been resolved: drm/amdgpu: keep the eviction fence until the last handle is closed amdgpu_gem_object_close() detaches the user queue eviction fence from the BO before it checks bo_va->ref_count, so closing any handle of a BO removes the fence, even when the file still has another handle to the BO and the bo_va with its GPU mappings stays. The BO can then be evicted without suspending the user queues of the file, which keep accessing it through mappings that are only marked for update. Seen on navi31 with a SDMA user queue: a BO opened twice in the same file (GEM_FLINK + GEM_OPEN) loses its eviction fence when the second handle is closed (traced attach/detach of the fence): attach bo=ffff8c78a6c84c00 GEM_CLOSE handle=7 detach bo=ffff8c78a6c84c00 while handle 6 and the mapping are still in use. Only detach the fence when the last reference to the bo_va goes away, matching amdgpu_gem_object_open(), which attaches it only when the bo_va is created. With this the fence stays until the file is closed. (cherry picked from commit 8387982f03d4026b7dc11b0b5e3118298eee4d6a)

Severity

7 (High)


                        
                          CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H

Assigner

Linux CNA under the mitre root

CNA scorecard D 53/100 over 6639 records in the last 180 days details

References

3 references

URL	Tags
https://git.kernel.org/stable/c/dce8f7f03a47b1744…
https://git.kernel.org/stable/c/6467c4cc83b94850b…
https://git.kernel.org/stable/c/3c44cbcec5eb37071…

Impacted products

2 products

Vendor	Product	Version
Linux	Linux	Affected: 31f7efcdca4d1caaa3a0babc33377e27e6f9b593 , < dce8f7f03a47b1744dccf506268f9229b67fab47 (git) Affected: 31f7efcdca4d1caaa3a0babc33377e27e6f9b593 , < 6467c4cc83b94850bfea614e9f98276edf9a7c3f (git) Affected: 31f7efcdca4d1caaa3a0babc33377e27e6f9b593 , < 3c44cbcec5eb370713bbd2c315ed7c34d8ccec2b (git)
Linux	Linux	Affected: 6.16 Unaffected: 0 , < 6.16 (semver) Unaffected: 6.18.59 , ≤ 6.18.* (semver) Unaffected: 7.2.13 , ≤ 7.2.* (semver) Unaffected: 7.3 , ≤ * (original_commit_for_fix)

Show details on NVD website

JSON

To clipboard

{
  "containers": {
    "cna": {
      "affected": [
        {
          "defaultStatus": "unaffected",
          "product": "Linux",
          "programFiles": [
            "drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c"
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "versions": [
            {
              "lessThan": "dce8f7f03a47b1744dccf506268f9229b67fab47",
              "status": "affected",
              "version": "31f7efcdca4d1caaa3a0babc33377e27e6f9b593",
              "versionType": "git"
            },
            {
              "lessThan": "6467c4cc83b94850bfea614e9f98276edf9a7c3f",
              "status": "affected",
              "version": "31f7efcdca4d1caaa3a0babc33377e27e6f9b593",
              "versionType": "git"
            },
            {
              "lessThan": "3c44cbcec5eb370713bbd2c315ed7c34d8ccec2b",
              "status": "affected",
              "version": "31f7efcdca4d1caaa3a0babc33377e27e6f9b593",
              "versionType": "git"
            }
          ]
        },
        {
          "defaultStatus": "affected",
          "product": "Linux",
          "programFiles": [
            "drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c"
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "vendor": "Linux",
          "versions": [
            {
              "status": "affected",
              "version": "6.16"
            },
            {
              "lessThan": "6.16",
              "status": "unaffected",
              "version": "0",
              "versionType": "semver"
            },
            {
              "lessThanOrEqual": "6.18.*",
              "status": "unaffected",
              "version": "6.18.59",
              "versionType": "semver"
            },
            {
              "lessThanOrEqual": "7.2.*",
              "status": "unaffected",
              "version": "7.2.13",
              "versionType": "semver"
            },
            {
              "lessThanOrEqual": "*",
              "status": "unaffected",
              "version": "7.3",
              "versionType": "original_commit_for_fix"
            }
          ]
        }
      ],
      "cpeApplicability": [
        {
          "nodes": [
            {
              "cpeMatch": [
                {
                  "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "6.18.59",
                  "versionStartIncluding": "6.16",
                  "vulnerable": true
                },
                {
                  "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "7.2.13",
                  "versionStartIncluding": "6.16",
                  "vulnerable": true
                },
                {
                  "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "7.3",
                  "versionStartIncluding": "6.16",
                  "vulnerable": true
                }
              ],
              "negate": false,
              "operator": "OR"
            }
          ]
        }
      ],
      "descriptions": [
        {
          "lang": "en",
          "value": "In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: keep the eviction fence until the last handle is closed\n\namdgpu_gem_object_close() detaches the user queue eviction fence from\nthe BO before it checks bo_va-\u003eref_count, so closing any handle of a BO\nremoves the fence, even when the file still has another handle to the\nBO and the bo_va with its GPU mappings stays. The BO can then be evicted\nwithout suspending the user queues of the file, which keep accessing it\nthrough mappings that are only marked for update.\n\nSeen on navi31 with a SDMA user queue: a BO opened twice in the same\nfile (GEM_FLINK + GEM_OPEN) loses its eviction fence when the second\nhandle is closed (traced attach/detach of the fence):\n\n  attach bo=ffff8c78a6c84c00\n  GEM_CLOSE handle=7\n  detach bo=ffff8c78a6c84c00\n\nwhile handle 6 and the mapping are still in use.\n\nOnly detach the fence when the last reference to the bo_va goes away,\nmatching amdgpu_gem_object_open(), which attaches it only when the\nbo_va is created. With this the fence stays until the file is closed.\n\n(cherry picked from commit 8387982f03d4026b7dc11b0b5e3118298eee4d6a)"
        }
      ],
      "metrics": [
        {
          "cvssV3_1": {
            "baseScore": 7,
            "baseSeverity": "HIGH",
            "vectorString": "CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H",
            "version": "3.1"
          },
          "scenarios": [
            {
              "lang": "en",
              "value": "AV:L - A local process reaches patched amdgpu_gem_object_close() through DRM_IOCTL_GEM_CLOSE -\u003e drm_gem_handle_delete() after duplicating the BO handle with DRM_IOCTL_GEM_FLINK/OPEN; no network or physical input reaches this path.\nAC:H - The handle-close and memory-pressure sequence is attacker-controlled, but affected GFX11/GFX12 and SDMA6/7 paths set disable_uq=true for the default amdgpu.user_queue=-1; exploitation requires an administrator-controlled non-default user-queue configuration.\nPR:L - FLINK/OPEN require an authenticated primary-node DRM client, but not root, and amdgpu_userq_priority_permit() allows normal-priority queues without capabilities; CAP_SYS_NICE or DRM master is required only for high priority.\nUI:N - The same process can issue GEM_FLINK, GEM_OPEN, GEM_CLOSE, create the user queue, ring its doorbell, and generate eviction pressure without another user\u0027s action.\nS:U - The stale per-file amdgpu_fpriv GPUVM mapping can affect BO memory managed by the same host DRM/kernel authority; it does not cross a guest-host or IOMMU security authority.\nC:H - After the fence is detached, amdgpu_bo_move_notify() only marks the bo_va moved, leaving its readable PTE stale until queue restoration; attacker-controlled SDMA copy packets can read BO backing that reuses the released physical VRAM.\nI:H - The same stale writable GPU PTE lets attacker-controlled user-queue packets write into a reallocated BO\u0027s backing memory, providing cross-allocation GPU-memory corruption rather than merely corrupting the attacker\u0027s moved BO.\nA:H - Accessing moved or reused ring/data backing can hang the user queue or GPU; amdgpu_userq_hang_detect_work() can escalate failed per-queue recovery to amdgpu_device_gpu_recover(), disrupting all GPU clients."
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2026-10-31T12:43:54.856Z",
        "orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
        "shortName": "Linux"
      },
      "references": [
        {
          "url": "https://git.kernel.org/stable/c/dce8f7f03a47b1744dccf506268f9229b67fab47"
        },
        {
          "url": "https://git.kernel.org/stable/c/6467c4cc83b94850bfea614e9f98276edf9a7c3f"
        },
        {
          "url": "https://git.kernel.org/stable/c/3c44cbcec5eb370713bbd2c315ed7c34d8ccec2b"
        }
      ],
      "title": "drm/amdgpu: keep the eviction fence until the last handle is closed",
      "x_generator": {
        "engine": "bippy-1.2.0"
      }
    }
  },
  "cveMetadata": {
    "assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67",
    "assignerShortName": "Linux",
    "cveId": "CVE-2026-99148",
    "datePublished": "2026-10-28T10:35:04.360Z",
    "dateReserved": "2026-09-25T10:25:14.396Z",
    "dateUpdated": "2026-10-31T12:43:54.856Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2"
}

Sightings

Author	Source	Type	Date	Other

Nomenclature

Seen: The vulnerability was mentioned, discussed, or observed by the user.
Confirmed: The vulnerability has been validated from an analyst's perspective.
Published Proof of Concept: A public proof of concept is available for this vulnerability.
Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
Not confirmed: The user expressed doubt about the validity of the vulnerability.
Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Vulnerability timeline

Sightings and recorded dates, grouped by day · UTC

Order JSON

How durations are calculated

Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.

Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.

Publication & reservation Sightings KEV catalogs Other timestamps

Loading timeline…

The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base model, served locally by ML-Gateway. They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.

Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.

Action not permitted