{"id":"CVE-2026-64281","title":"In the Linux kernel, the following vulnerability has been resolved:\n\nsvcrdma: wake sq waiters when the transport closes\n\nThreads parked in svc_rdma_sq_wait() on sc_sq_ticket_wait or\nsc_send_wait can hang indefinitely in TASK_UNINTERRUPTI…","summary":"In the Linux kernel, the following vulnerability has been resolved:\n\nsvcrdma: wake sq waiters when the transport closes\n\nThreads parked in svc_rdma_sq_wait() on sc_sq_ticket_wait or\nsc_send_wait can hang indefinitely in TASK_UNINTERRUPTI…","severity":"high","cvss":7.5,"cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","published":"2026-07-25","updated":"2026-07-27","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-64281","references":[{"url":"https://git.kernel.org/stable/c/40eedc4253dbda0b29b7961200534dfcecb48ace","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"},{"url":"https://git.kernel.org/stable/c/e5248a7426030db1e126363f72afdb3b71339a5c","label":"416baaa9-dc9f-4396-8d5f-8c081fb06d67"}],"tags":["nvd"],"epss":0.00468,"epssPercentile":0.39542,"ingestedAt":"2026-07-27T06:16:50.040Z","slug":"CVE-2026-64281","body":"## Overview\n\nIn the Linux kernel, the following vulnerability has been resolved:\n\nsvcrdma: wake sq waiters when the transport closes\n\nThreads parked in svc_rdma_sq_wait() on sc_sq_ticket_wait or\nsc_send_wait can hang indefinitely in TASK_UNINTERRUPTIBLE state\nacross transport teardown, pinning svc_xprt references and\nblocking svc_rdma_free().\n\nThe close path sets XPT_CLOSE before invoking xpo_detach and both\nwait_event predicates include an XPT_CLOSE term, but the\npredicates are re-evaluated only on wakeup. sc_sq_ticket_wait has\nno completion-driven wake path; it is advanced solely by the\nchained ticket handoff inside svc_rdma_sq_wait() itself. Without\nan explicit wake at close, parked threads never observe\nXPT_CLOSE, hold their svc_xprt_get reference forever, and\nsvc_rdma_free() blocks on xpt_ref dropping to zero.\n\nTwo close entry points reach this transport. Local teardown runs\nsvc_rdma_detach() from svc_handle_xprt() -> svc_delete_xprt() ->\nxpo_detach() on a worker thread. A remote disconnect arrives at\nsvc_rdma_cma_handler(), which calls svc_xprt_deferred_close():\nthat sets XPT_CLOSE and enqueues the transport but does not\naccess either RDMA waitqueue, so a worker already parked in\nsvc_rdma_sq_wait() never re-evaluates its predicate. With every\nworker parked on this transport, no thread is available to run\nthe local teardown either, and the wake site there is\nunreachable.\n\nIntroduce svc_rdma_xprt_deferred_close(), a thin svcrdma wrapper\nthat calls svc_xprt_deferred_close() and then wakes both\nsc_sq_ticket_wait and sc_send_wait. Convert the svcrdma producers\nthat called svc_xprt_deferred_close() directly:\nsvc_rdma_cma_handler(), qp_event_handler(),\nsvc_rdma_post_send_err(), svc_rdma_wc_send(), the sendto drop\npath, the rw completion error paths, and the recvfrom flush and\nread-list error paths.\n\nWake both waitqueues from svc_rdma_detach() as well. The\nsynchronous svc_xprt_close() path (backchannel ENOTCONN, device\nremoval via svc_rdma_xprt_done) reaches detach without flowing\nthrough svc_xprt_deferred_close() and therefore does not invoke\nthe new helper.\n\n[ cel: add svc_rdma_xprt_deferred_close() to complete the fix ]\n\n## Remediation\n\nRefer to the linked advisories for vendor-supplied fixes and affected version ranges.","depth":"twilight","depthScore":41,"depthScoreParts":{"impact":41.3,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}