{"id":"CVE-2026-99212","published":"2026-10-28T11:17:42.650","lastModified":"2026-10-28T11:17:42.650","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nnet/rds: don't let the rdma_cm destroy an id RDS still owns on route failure\n\nrds_rdma_cm_event_handler_cmn() hands the return value of\nrdma_resolve_route() straight back to the rdma_cm on\nRDMA_CM_EVENT_ADDR_RESOLVED.  A synchronous failure there - -ENOMEM\nfrom the route work allocation, an SA query that cannot be set up, a\nRoCE route with no usable device - makes addr_handler() destroy the\nid the event was delivered on.  That id is ic->i_cm_id, and nothing\nclears the pointer: the connection sits in RDS_CONN_CONNECTING with a\nfreed id until its shutdown calls rdma_disconnect() and\nrdma_destroy_id() on it.\n\nThe ROUTE_RESOLVED case already avoids this: rds_ib_cm_initiate_connect()\nforces a zero return while ic->i_cm_id == cm_id, as the comment there\nexplains.  Do the same here - drop the connection and return 0 - so\nthe id stays RDS's to destroy from the shutdown, and the reconnect\ngets a fresh one.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/91ece3efa064f0b3b3c67b9b045d9eeb4a034e43","tags":[]},{"url":"https://git.kernel.org/stable/c/98d446b7c013254a8a7b49720557676a3a851fde","tags":[]},{"url":"https://git.kernel.org/stable/c/a7c7d0ff876377fea7ae876a28d6df8d9349df57","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}