This is an automated email from the ASF dual-hosted git repository.
chenBright pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/brpc.git
The following commit(s) were added to refs/heads/master by this push:
new 159c8550 Fix RDMA PollCq missing recv CQEs after re-arming the CQs
(#3425)
159c8550 is described below
commit 159c85505c84d78c030838d402536260805f4a26
Author: Bright Chen <[email protected]>
AuthorDate: Fri Aug 7 17:38:00 2026 +0800
Fix RDMA PollCq missing recv CQEs after re-arming the CQs (#3425)
PollCq only re-polled send_cq after arming both CQs, so a recv CQE
arriving in the one-shot notification race window of recv_cq was left
in the CQ and the RPC timed out. Restart the re-poll from recv_cq so
that both CQs are covered.
---
src/brpc/rdma/rdma_endpoint.cpp | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/src/brpc/rdma/rdma_endpoint.cpp b/src/brpc/rdma/rdma_endpoint.cpp
index 6c20ea39..8660d8d9 100644
--- a/src/brpc/rdma/rdma_endpoint.cpp
+++ b/src/brpc/rdma/rdma_endpoint.cpp
@@ -1492,6 +1492,17 @@ void RdmaEndpoint::PollCq(Socket* m) {
return;
}
notified = true;
+ // Both CQs have just been re-armed, thus both of them must be
+ // re-polled. Note that `cq' is `send_cq' here, so we have to
+ // switch back to `recv_cq' explicitly. Otherwise only
+ // `send_cq' would be re-polled, and a recv CQE arriving in
+ // the window between the poll and the notify of `recv_cq'
+ // would be left in the CQ without any following event
+ // (one shot notification is not triggered by the CQE which
+ // is already in the CQ before the arming), which stalls the
+ // connection until the next CQE happens to come.
+ send = false;
+ cq = ep->_resource->recv_cq;
continue;
}
if (!m->MoreReadEvents(&progress)) {
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
