{
  "schema_version": "1.6.1",
  "id": "RPI-CVE-2026-98257",
  "modified": "2026-10-09T11:39:44Z",
  "upstream": [
    "CVE-2026-98257"
  ],
  "published": "2026-10-06T08:47:10Z",
  "summary": "rds: ib: use rds_conn_drop() on protocol version mismatch",
  "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nrds: ib: use rds_conn_drop() on protocol version mismatch\n\nrds_ib_cm_connect_complete() runs from the RDMA-CM event handler with\nconn->c_cm_lock held.  When the peer negotiates a protocol version\nolder than RDS_PROTOCOL_COMPAT_VERSION, the handler calls\nrds_conn_destroy(), which is only safe in the rmmod path: it\nsynchronously tears the connection down and flush_work()es the\nshutdown work cp_down_w.\n\nThat shutdown work (rds_conn_shutdown()) needs cp_cm_lock, which is\nthe very lock the event handler still holds, so the flush never\ncompletes: the two workers wait on each other and the RDS connection\nworkqueues stall for good.\n\nAll other RDMA-CM failure paths (REJECTED, CONNECT_ERROR,\nDISCONNECTED) use rds_conn_drop(), which marks the connection\nRDS_CONN_ERROR and schedules the shutdown work asynchronously.  Use\nit here as well.",
  "severity": [
    {
      "type": "CVSS_V3",
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/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": "6.1.189",
            "compiled_filter": false
          }
        }
      }
    }
  ],
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://www.cve.org/CVERecord?id=CVE-2026-98257"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/424019be3f637da76b74f44a40034669acbb166f"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/14fb90067d13f4494cff0f03243fae3cf2911cbc"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/344c72a0bc2703a6de4918ed935850e07dae3af2"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/b5c9f2951d330d4a198a64d6e60f79b4bd6ef078"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/d392b16acd0908f077289aad8bba066ff16de336"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/65ca0a5037152a5a80b8fe6aac80eb80458be573"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/7d8c22cb2cb1efb5e830a4545c76abf7b35cc2e8"
    },
    {
      "type": "FIX",
      "url": "https://git.kernel.org/stable/c/f97d8c7bab7843631206a114986c9059da03efeb"
    }
  ],
  "database_specific": {
    "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git",
    "upstream_modified": "2026-10-07T06:50:27Z",
    "severity": {
      "source": "kernel CNA",
      "cvss_v3_score": 7.5,
      "cvss_v3_severity": "High",
      "scenario": "AV:N - The bad input is dp_protocol_major/minor in the peer's RDMA-CM REP private data, which rds_ib_cm_connect_complete() reads on RDMA_CM_EVENT_ESTABLISHED from rds_rdma_cm_event_handler_cmn(). RDS/IB runs over RoCEv2, whose CM messages travel over routable UDP/IP, so a remote peer supplies the version that hits the rds_conn_destroy() branch.\nAC:L - The attacker produces the active-side state: connect to the RDS listener, then disconnect. IB has no lower-address rule in rds_queue_reconnect(), so the hashed conn reconnects to the attacker, who replies with version 3.0. The deadlock then fires every time.\nPR:N - RDS/IB connection setup has no authentication. rds_ib_cm_handle_connect() only checks the protocol version from the request, and the REP private data the fix concerns is taken from the peer as-is.\nUI:N - No local user action is needed. The victim kernel's own reconnect worker (rds_connect_worker via rds_queue_reconnect) starts the outgoing connection after the attacker drops the first one.\nS:U - The impact is a deadlock inside the same kernel that processes the RDMA-CM event. No VM, IOMMU or sandbox boundary is crossed.\nC:N - rds_conn_destroy() blocks in flush_work(&cp->cp_down_w) before it reaches kmem_cache_free(), so the conn is never freed and then used. No memory is disclosed.\nI:N - No memory is corrupted. The fault is a lock-ordering deadlock between the CM handler holding c_cm_lock and rds_conn_shutdown() waiting for it, not a write to freed or out-of-bounds memory.\nA:H - The RDMA-CM event worker blocks forever in flush_work() while holding c_cm_lock, and the RDS shutdown work blocks on mutex_lock(&cp->cp_cm_lock). This stalls the RDS connection workqueues and CM event handling for good, and the attacker can repeat it against new connections."
    }
  }
}