In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
RDMA\/rxe: Avoid reprocessing the current packet after the QP enters the error state<\/p>\n
When do_complete() finds the QP in the error state it returns
\nRESPST_CHK_RESOURCE. Before commit 49dc9c1f0c7e (“RDMA\/rxe: Cleanup
\nreset state handling in rxe_resp.c”) this was the flush loop:
\ncheck_resource() had an error-state branch that fetched each remaining
\nrecv WQE and completed it with IB_WC_WR_FLUSH_ERR, without touching
\nthe current packet. That commit removed the error-state branch from
\ncheck_resource() (draining is now done at rxe_receiver() entry) but
\nkept the do_complete() error-state return.<\/p>\n
As a result, when a QP moves to the error state while a packet is
\nbeing completed – e.g. an rdma_cm disconnect racing with receive
\nprocessing – the responder state machine loops back into the request
\nprocessing chain with the already-completed packet still in hand:
\ncheck_resource() fetches a fresh recv WQE, execute()\/send_data_in()
\ncopies the same packet payload again, do_complete() posts another
\nIB_WC_SUCCESS CQE (qp->resp.status is still 0), and control returns
\nto the error-state check. The loop re-executes the same packet once
\nper posted recv WQE (observed: ~1000 duplicate IB_WC_SUCCESS
\ncompletions of one SEND, one per ~8us, matching the RQ occupancy)
\nuntil the RQ is exhausted, after which qp->resp.wqe is NULL and
\nsend_data_in() dereferences it:<\/p>\n
BUG: kernel NULL pointer dereference, address: 0000000000000014
\n Workqueue: rxe_wq do_work
\n RIP: copy_data+0x29\/0x1f0
\n Call Trace:
\n send_data_in+0x25\/0x50
\n rxe_receiver+0xf36\/0x1dd0<\/p>\n
The duplicate completions are indistinguishable from real receives to
\nthe ULP. During an rds stress test, the message was accepted as new and
\ndelivered the same datagram to user space hundreds of times, corrupting
\nthe stream; any ULP that relies on RC exactly-once delivery is affected.<\/p>\n
A live packet reaching the error-state check in do_complete() has
\nbeen executed and completed exactly once and must be consumed, not
\nre-processed. Return RESPST_CLEANUP for it (dequeue and free); keep
\nreturning RESPST_CHK_RESOURCE for the pkt == NULL case.<\/p>\n","protected":false},"excerpt":{"rendered":"
In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
RDMA\/rxe: Avoid reprocessing the current packet after the QP enters the error state<\/p>\n
When do_complete() finds the QP in the error s<\/p>\n","protected":false},"template":"","meta":[],"categories":[],"tags":[],"class_list":["post-25964","vuln","type-vuln","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln\/25964","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln"}],"about":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/types\/vuln"}],"wp:attachment":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=25964"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=25964"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=25964"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}