Commit b659997e authored Jun 27, 2023 by Shuhei Matsumoto Committed by Tomasz Zawadzki Mar 05, 2024

nvme_rdma: Move rqpair->num_completions++ prior to request_ready()



We will want to call nvme_rdma_qpair_process_submits() in
nvme_rdma_req_complete() if rqpair->num_completions is non-zero even if
rqpair is removed from group->active_qpairs.

As a preparation, move rqpair->num_comlpletions++ prior to
nvme_rdma_request_ready() in nvme_rdma_process_recv_completion() and
nvme_rdma_process_send_completion().

Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I1f22ac1e8236f9a5f526fe2e67cd0c051398c9bd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/18887


Reviewed-by: Jim Harris <jim.harris@samsung.com>
Reviewed-by: Ben Walker <ben@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>

parent d1fd0d13

lib/nvme/nvme_rdma.c

+4 −2

Original line number	Diff line number	Diff line
		@@ -2540,6 +2540,8 @@ nvme_rdma_process_recv_completion(struct nvme_rdma_poller *poller, struct ibv_wc
		return 0;
		}

		rqpair->num_completions++;

		nvme_rdma_request_ready(rqpair, rdma_req);

		if (!rqpair->delay_cmd_submit) {
		@@ -2550,7 +2552,6 @@ nvme_rdma_process_recv_completion(struct nvme_rdma_poller *poller, struct ibv_wc
		}
		}

		rqpair->num_completions++;
		return 1;

		err_wc:
		@@ -2619,6 +2620,8 @@ nvme_rdma_process_send_completion(struct nvme_rdma_poller *poller,
		return 0;
		}

		rqpair->num_completions++;

		nvme_rdma_request_ready(rqpair, rdma_req);

		if (!rqpair->delay_cmd_submit) {
		@@ -2629,7 +2632,6 @@ nvme_rdma_process_send_completion(struct nvme_rdma_poller *poller,
		}
		}

		rqpair->num_completions++;
		return 1;
		}