{ "schema_version": "1.6.1", "id": "RPI-CVE-2026-72469", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2026-72469" ], "published": "2026-08-15T06:04:33Z", "summary": "xprtrdma: Fix ep kref imbalance on ADDR_CHANGE", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nxprtrdma: Fix ep kref imbalance on ADDR_CHANGE\n\nrpcrdma_cm_event_handler() falls through to the disconnected: label\non RDMA_CM_EVENT_ADDR_CHANGE and calls rpcrdma_ep_put() with no\nmatching get when the event arrives before RDMA_CM_EVENT_ESTABLISHED.\nThe kref then underflows during connect teardown and\nrpcrdma_xprt_disconnect() operates on a freed ep.\n\nReference counts across a normal connection lifecycle:\n\n rpcrdma_ep_create() kref_init ->1\n rpcrdma_xprt_connect() ep_get ->2 (before post_recvs)\n RDMA_CM_EVENT_ESTABLISHED ep_get ->3\n RDMA_CM_EVENT_DISCONNECTED ep_put ->2\n rpcrdma_xprt_drain() ep_put ->1\n rpcrdma_xprt_disconnect() tail ep_put ->0 (ep_destroy)\n\nThe connect-time get in rpcrdma_xprt_connect(), taken just before\nrpcrdma_post_recvs() \"while there are outstanding Receives,\" is\nbalanced by rpcrdma_xprt_drain. ADDR_CHANGE before ESTABLISHED has\nno get to consume, so its put drops the count to 1 and the drain\nput then frees the ep while rpcrdma_xprt_disconnect() still holds a\npointer to it.\n\nFix by dispatching on the prior re_connect_status via xchg(): for\nprev == 0 (pre-ESTABLISHED) wake the connect waiter and return with\nno put; for prev == 1 call rpcrdma_force_disconnect() and return.\nThe case-1 arm relies on the subsequent RDMA_CM_EVENT_DISCONNECTED\nevent -- reliably delivered when rdma_disconnect() is called on a\nstill-connected cm_id -- to balance the ESTABLISHED get;\nrpcrdma_xprt_drain() continues to balance only that connect-time\nget. Any other prior value means teardown is already in flight.", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:A/AC:L/PR:N/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-2026-72469" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/cfd1bab66b042da7a778786685125656c695931b" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/d0479c2b12974aa188b10d221a5770126b118b6d" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/ffc07790539736a5d029f6a3c966b46c529f93a8" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/af9b65b29af341932625c4283dc7a23cdb62688a" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-08-17T05:50:24Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 8.8, "cvss_v3_severity": "High", "scenario": "AV:A - The bug is in the NFS/RPC-over-RDMA client reached during RDMA CM connect/reconnect to a remote server; ADDR_CHANGE is delivered on bonded RoCE/IB netdev failover on the storage network segment, a plausible adjacent-network trigger in HPC and cloud NFS-over-RDMA deployments.\nAC:L - Once NFS-over-RDMA is in use, an attacker can reliably drive repeated reconnect attempts and time or induce bonding failover so ADDR_CHANGE arrives pre-ESTABLISHED; the resulting kref underflow and UAF are deterministic, not a probabilistic memory-layout race.\nPR:N - Exploitation requires no privileges on the victim host\u2014only that NFS-over-RDMA is configured and reconnects occur\u2014while an adjacent attacker or malicious storage peer can provoke link failover and connection churn without authenticating to the victim system.\nUI:N - No victim user interaction is required beyond normal automated NFS-over-RDMA client reconnect activity already performed by the kernel when links fail or sessions drop.\nS:U - Impact is kernel memory corruption and potential local privilege escalation within the same kernel security domain, not a cross-boundary escape such as VM guest-to-host or IOMMU bypass.\nC:H - Premature free of rpcrdma_ep leaves dangling pointers used in rpcrdma_xprt_disconnect (rdma_disconnect, ib_drain_rq/sq), a classic kernel heap UAF that can be leveraged for arbitrary memory disclosure via controlled reuse of the freed endpoint object.\nI:H - Teardown continues to operate on the freed endpoint and associated RDMA/IB state, enabling heap corruption and potential arbitrary kernel write or control-flow hijack, not merely a benign NULL dereference crash.\nA:H - The kref imbalance causes premature endpoint destruction during disconnect teardown, which can trigger kernel oops/panic from use-after-free even when exploitation for code execution is not attempted." } } }