Commit f7b85051 authored Nov 28, 2023 by Shuhei Matsumoto Committed by Tomasz Zawadzki Dec 14, 2023

nvme_rdma: Clear rqpair->num_completions after it is used for resubmission



It is enough if we clear rqpair->num_completions after it is actually
used for resubmission. This is helpful to reduce connected qpair
traversal loop.

Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I99ab47b211d703d722768eac64239712c8142d7f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/18841


Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <ben@nvidia.com>
Community-CI: Mellanox Build Bot

parent 77967a60

lib/nvme/nvme_rdma.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -3086,6 +3086,7 @@ nvme_rdma_qpair_process_submits(struct spdk_nvme_qpair *qpair)
		}
		if (rqpair->num_completions > 0) {
		nvme_qpair_resubmit_requests(qpair, rqpair->num_completions);
		rqpair->num_completions = 0;
		}
		}

		@@ -3118,7 +3119,6 @@ nvme_rdma_poll_group_process_completions(struct spdk_nvme_transport_poll_group *

		STAILQ_FOREACH_SAFE(qpair, &tgroup->connected_qpairs, poll_group_stailq, tmp_qpair) {
		rqpair = nvme_rdma_qpair(qpair);
		rqpair->num_completions = 0;

		if (spdk_unlikely(nvme_qpair_get_state(qpair) == NVME_QPAIR_CONNECTING)) {
		rc = nvme_rdma_ctrlr_connect_qpair_poll(qpair->ctrlr, qpair);