{
  "affected": [
    {
      "ranges": [
        {
          "events": [
            {
              "introduced": "98f37634b12b17ad5c56db8fb63cf9d7dc55d74c"
            },
            {
              "fixed": "0639ea767fe04c288a8d6cb826100fe3d95d4936"
            },
            {
              "fixed": "3afc64c61ce906a04f073ca350b46de10e8302f9"
            }
          ],
          "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git",
          "type": "GIT"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Linux",
        "name": "Kernel"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "7.1.0"
            },
            {
              "fixed": "7.1.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "database_specific": {
    "cna_assigner": "Linux",
    "osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/72xxx/CVE-2026-72112.json"
  },
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nio_uring/bpf-ops: reject re-registration of an already-bound ops\n\nio_install_bpf() only rejects a second registration on the ctx side\n(ctx-\u003ebpf_ops) and sets the per-map back-pointer ops-\u003epriv\nunconditionally. The struct_ops link path never advances a map past\nBPF_STRUCT_OPS_STATE_READY, so the same io_uring_bpf_ops map can be\nregistered more than once, and bpf_io_reg() re-resolves the target ring\nvia fget(ops-\u003ering_fd) on every call. A caller can therefore point the\nsame ring_fd at a different io_ring_ctx between two BPF_LINK_CREATE\ncalls.\n\nThe second registration passes the ctx-\u003ebpf_ops check (the new ctx has\nnone) and overwrites ops-\u003epriv, orphaning the first ctx. Teardown\n(io_eject_bpf()/bpf_io_unreg()) only reaches a ctx through ops-\u003epriv, so\nthe orphaned ctx is never torn down: its ctx-\u003eloop_step keeps pointing\ninto the struct_ops trampoline, which is freed once the map is gone. A\nlater io_uring_enter() on the orphaned ring then calls the dangling\nctx-\u003eloop_step from io_run_loop() -- a use-after-free of freed\nexecutable memory, reachable by a task with CAP_BPF + CAP_PERFMON.\n\nReject registration when ops-\u003epriv is already set, as hid_bpf_reg()\ndoes for its struct_ops.",
  "id": "CVE-2026-72112",
  "modified": "2026-08-18T03:31:01.492554617Z",
  "published": "2026-08-15T05:52:53.275Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/0639ea767fe04c288a8d6cb826100fe3d95d4936"
    },
    {
      "type": "WEB",
      "url": "https://git.kernel.org/stable/c/3afc64c61ce906a04f073ca350b46de10e8302f9"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/72xxx/CVE-2026-72112.json"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-72112"
    },
    {
      "type": "PACKAGE",
      "url": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git"
    }
  ],
  "schema_version": "1.9.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "io_uring/bpf-ops: reject re-registration of an already-bound ops"
}