{
  "affected": [
    {
      "ranges": [
        {
          "events": [
            {
              "introduced": "fd1fa48b935f22c7d99713bf33846e14a6bb6ab9"
            },
            {
              "fixed": "1ae594845217ce740b27ec5a202e9e8cb144e0aa"
            },
            {
              "fixed": "e0a1404b34e0002f03f7979cdb791a9b5c3963f5"
            }
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "type": "GIT"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Linux",
        "name": "Kernel"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "7.0.0"
            },
            {
              "fixed": "7.3.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "database_specific": {
    "cna_assigner": "Linux",
    "osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/99xxx/CVE-2026-99621.json"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: unlock the VM and BO when attaching a KFD BO fails\n\nkfd_mem_attach() locks the page directory and the BO with a drm_exec\ntransaction on its stack, but leaves through the unwind label without\ndrm_exec_fini() when locking fails, e.g. when MAP_MEMORY_TO_GPU gets a\nsignal while the page directory or the BO is busy. Reservations already\nacquired then remain locked with their ww_acquire_ctx pointing into the\nexpired stack frame.\n\nWith one thread doing GEM_VA map/unmap on the KFD VM and\nanother one mapping KFD BOs while getting SIGUSR1 every 20 us:\n\n  BUG: KASAN: stack-out-of-bounds in __ww_mutex_lock.constprop.0+0x34f5/0x3da0\n   ww_mutex_lock_interruptible+0x2f/0x150\n   drm_exec_lock_obj+0x107/0x2d0 [drm_exec]\n   kfd_mem_attach+0x4dd/0x1b60 [amdgpu]\n\nand the process ends up in D state.\n\nFinish the transaction before unwinding.\n\n(cherry picked from commit 7b885fa27e4b9fa5cbdc64e66fc35872aee296f5)",
  "id": "CVE-2026-99621",
  "modified": "2026-11-20T06:45:26.982356026Z",
  "published": "2026-11-20T05:49:41.407Z",
  "references": [
    {
      "type": "PACKAGE",
      "url": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/1ae594845217ce740b27ec5a202e9e8cb144e0aa"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/e0a1404b34e0002f03f7979cdb791a9b5c3963f5"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/99xxx/CVE-2026-99621.json"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-99621"
    }
  ],
  "schema_version": "1.9.2",
  "summary": "drm/amdgpu: unlock the VM and BO when attaching a KFD BO fails"
}