{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2026-92488",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2026-92488"
  ],
  "published": "2026-09-17T16:10:40Z",
  "summary": "RDMA/erdma: complete object teardown when the destroy command fails",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nRDMA/erdma: complete object teardown when the destroy command fails\n\nerdma_destroy_qp(), erdma_destroy_cq(), erdma_dereg_mr(), and\nerdma_destroy_ah() returned early when erdma_post_cmd_wait() failed,\nleaking the queue buffers, MTTs, doorbells and the STAG, QPN, CQN and AHN\nidentifiers. A command timeout clears ERDMA_CMDQ_STATE_OK_BIT and\npermanently disables the command queue, so no retry can succeed; the RDMA\ncore keeps the object after a failed destructor and forced uverbs cleanup\nthen nulls the pointers, making the resources unreachable.\n\nWarn on failure but release every software-owned resource and return\nsuccess, since during terminal destruction the hardware command result is\nonly diagnostic.",
  "severity": [
    {
      "type": "CVSS_V3",
      "score": "CVSS:3.1/AV:L/AC:H/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": null,
            "compiled_filter": false
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2026-92488"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/5fcfc988ecf3e2bbe308b95c2c0d2f011d9afabe"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/ce7d205c264665517c25e8a3231cf2d0c2443e3c"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/334145d683ad3e31d052247f10807f2ebc950351"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/652befcba956ef357f480525ccbe25c59bc81d4d"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-09-18T17:55:42Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 7.0,
      "cvss_v3_severity": "High",
      "scenario": "AV:L - The bug is reached from ib_uverbs_write(IB_USER_VERBS_CMD_DESTROY_QP/DESTROY_CQ/DEREG_MR/DESTROY_AH) or ib_uverbs_ioctl(UVERBS_METHOD_*_DESTROY) on /dev/infiniband/uverbs*, which call uobj_destroy -> ib_destroy_qp_user/ib_destroy_cq_user/ib_dereg_mr_user -> erdma_destroy_*; a remote RoCE/iWARP peer never supplies that destroy command.\nAC:H - erdma_destroy_qp/cq/ah and erdma_dereg_mr only skip teardown when erdma_post_cmd_wait() fails. That wait returns -ETIME after ERDMA_CMDQ_TIMEOUT_MS and then clears ERDMA_CMDQ_STATE_OK_BIT, or -EIO on a hardware syndrome; the SQE is built by the driver from the kernel-owned QPN/CQN/lkey/AHN, so the attacker cannot force that failure.\nPR:L - ib_uverbs_open() only checks rdma_dev_access_netns() and uverbs_devnode() sets mode 0666. erdma_alloc_ucontext() requires CAP_SYS_RAWIO only when ERDMA_DEV_CAP_FLAGS_EXTEND_DB is unset; on extend-doorbell hardware an unprivileged local user can GET_CONTEXT, CREATE_QP/CQ and then hit the failing destructors.\nUI:N - The attacker issues DESTROY_* on their own uverbs fd or closes it so ib_uverbs_close() runs uverbs_destroy_ufile_hw(); no other user must mount a filesystem, open a file, or otherwise cooperate.\nS:U - A leftover erdma_cq/erdma_qp in cq_xa/qp_xa and a UAF of the associated ib_ucq_object stay inside the host kernel; this is not a KVM/Xen guest-to-host escape or an IOMMU/DMA boundary bypass.\nC:H - On command failure erdma_destroy_cq() returns before xa_erase(&dev->cq_xa). __uverbs_cleanup_ufile(RDMA_REMOVE_DRIVER_FAILURE) then nulls uobj->object without ib_destroy_cq_user and frees the ib_ucq_object, so erdma_ceq_completion_handler() -> ib_uverbs_comp_handler() reads the freed cq->uobject.\nI:H - The same ib_uverbs_comp_handler() then list_add_tail()s onto the freed ev_queue->event_list and ib_ucq_object.comp_list after DRIVER_FAILURE dropped the uobject, a use-after-free write into slab memory that can corrupt a reused object or freelist metadata.\nA:H - The CEQ tasklet UAF in ib_uverbs_comp_handler() can oops; independently, returning before erdma_free_idx()/put_mtt_entries()/dma_free_coherent() leaks STAG/QPN/CQN/AHN IDs and DMA queue buffers, and a cmdq timeout permanently disables further commands."
    }
  }
}