[PATCH for-next] RDMA/rxe: Fix missing acks from responder

[PATCH for-next] RDMA/rxe: Fix missing acks from responder

[Bob Pearson]

All responder errors from request packets that do not
consume a receive WQE fail to generate acks for RC QPs.
This patch corrects this behavior by making the flow
follow the same path as request packets that do consume
a WQE after the completion.

Link: https://lore.kernel.org/linux-rdma/1a7286ac-bcea-40fb-2267-480134dd301b@gmail.com/
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
 drivers/infiniband/sw/rxe/rxe_comp.c |  1 -
 drivers/infiniband/sw/rxe/rxe_resp.c | 18 ++++++++----------
 2 files changed, 8 insertions(+), 11 deletions(-)

diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c
index a612b335baa0..2af26737d32d 100644
--- a/drivers/infiniband/sw/rxe/rxe_comp.c
+++ b/drivers/infiniband/sw/rxe/rxe_comp.c
@@ -676,7 +676,6 @@ int rxe_completer(void *arg)
 
 			/* there is nothing to retry in this case */
 			if (!wqe || (wqe->state == wqe_state_posted)) {
-				pr_warn("Retry attempted without a valid wqe\n");
 				ret = -EAGAIN;
 				goto done;
 			}
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index e0e9984a3cce..6c43df7bf9bc 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -816,8 +816,8 @@ static enum resp_states do_complete(struct rxe_qp *qp,
 	struct rxe_recv_wqe *wqe = qp->resp.wqe;
 	struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
 
-	if (unlikely(!wqe))
-		return RESPST_CLEANUP;
+	if (!wqe)
+		goto finish;
 
 	memset(&cqe, 0, sizeof(cqe));
 
@@ -917,12 +917,12 @@ static enum resp_states do_complete(struct rxe_qp *qp,
 	if (rxe_cq_post(qp->rcq, &cqe, pkt ? bth_se(pkt) : 1))
 		return RESPST_ERR_CQ_OVERFLOW;
 
-	if (qp->resp.state == QP_STATE_ERROR)
+finish:
+	if (unlikely(qp->resp.state == QP_STATE_ERROR))
 		return RESPST_CHK_RESOURCE;
-
-	if (!pkt)
+	if (unlikely(!pkt))
 		return RESPST_DONE;
-	else if (qp_type(qp) == IB_QPT_RC)
+	if (qp_type(qp) == IB_QPT_RC)
 		return RESPST_ACKNOWLEDGE;
 	else
 		return RESPST_CLEANUP;
@@ -1056,10 +1056,8 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
 	if (pkt->mask & RXE_SEND_MASK ||
 	    pkt->mask & RXE_WRITE_MASK) {
 		/* SEND. Ack again and cleanup. C9-105. */
-		if (bth_ack(pkt))
-			send_ack(qp, pkt, AETH_ACK_UNLIMITED, prev_psn);
-		rc = RESPST_CLEANUP;
-		goto out;
+		send_ack(qp, pkt, AETH_ACK_UNLIMITED, prev_psn);
+		return RESPST_CLEANUP;
 	} else if (pkt->mask & RXE_READ_MASK) {
 		struct resp_res *res;
 
-- 
2.27.0

Re: [PATCH for-next] RDMA/rxe: Fix missing acks from responder

[Jason Gunthorpe]

On Thu, Apr 01, 2021 at 07:10:17PM -0500, Bob Pearson wrote:
> All responder errors from request packets that do not
> consume a receive WQE fail to generate acks for RC QPs.
> This patch corrects this behavior by making the flow
> follow the same path as request packets that do consume
> a WQE after the completion.
> 
> Link: https://lore.kernel.org/linux-rdma/1a7286ac-bcea-40fb-2267-480134dd301b@gmail.com/
> Signed-off-by: Bob Pearson <rpearson@hpe.com>
> ---
>  drivers/infiniband/sw/rxe/rxe_comp.c |  1 -
>  drivers/infiniband/sw/rxe/rxe_resp.c | 18 ++++++++----------
>  2 files changed, 8 insertions(+), 11 deletions(-)

Applied to for-next, thanks

Jason