RDMA/rxe: Cleanup reset state handling in rxe_resp.c

Cleanup the handling of qp in the error state, reset state and during
rxe_qp_do_cleanup. The error state does about the same thing as the others
but has code spread all over.

This patch combines them in a cleaner way.

Link: https://lore.kernel.org/r/20230304174533.11296-4-rpearsonhpe@gmail.com
Signed-off-by: Ian Ziemba <ian.ziemba@hpe.com>
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

2 files changed, 57 insertions, 51 deletions

diff --git a/drivers/infiniband/sw/rxe/rxe.h b/drivers/infiniband/sw/rxe/rxe.h
index bd8a8ea4ea8f..d33dd6cf83d3 100644
--- a/drivers/infiniband/sw/rxe/rxe.h
+++ b/drivers/infiniband/sw/rxe/rxe.h
@@ -133,7 +133,6 @@ enum resp_states {
 	RESPST_ERR_LENGTH,
 	RESPST_ERR_CQ_OVERFLOW,
 	RESPST_ERROR,
-	RESPST_RESET,
 	RESPST_DONE,
 	RESPST_EXIT,
 };
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index 7cb1b962d665..8f9bbb14fa7a 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -42,7 +42,6 @@ static char *resp_state_name[] = {
 	[RESPST_ERR_LENGTH]			= "ERR_LENGTH",
 	[RESPST_ERR_CQ_OVERFLOW]		= "ERR_CQ_OVERFLOW",
 	[RESPST_ERROR]				= "ERROR",
-	[RESPST_RESET]				= "RESET",
 	[RESPST_DONE]				= "DONE",
 	[RESPST_EXIT]				= "EXIT",
 };
@@ -69,17 +68,6 @@ static inline enum resp_states get_req(struct rxe_qp *qp,
 {
 	struct sk_buff *skb;
 
-	if (qp->resp.state == QP_STATE_ERROR) {
-		while ((skb = skb_dequeue(&qp->req_pkts))) {
-			rxe_put(qp);
-			kfree_skb(skb);
-			ib_device_put(qp->ibqp.device);
-		}
-
-		/* go drain recv wr queue */
-		return RESPST_CHK_RESOURCE;
-	}
-
 	skb = skb_peek(&qp->req_pkts);
 	if (!skb)
 		return RESPST_EXIT;
@@ -334,24 +322,6 @@ static enum resp_states check_resource(struct rxe_qp *qp,
 {
 	struct rxe_srq *srq = qp->srq;
 
-	if (qp->resp.state == QP_STATE_ERROR) {
-		if (qp->resp.wqe) {
-			qp->resp.status = IB_WC_WR_FLUSH_ERR;
-			return RESPST_COMPLETE;
-		} else if (!srq) {
-			qp->resp.wqe = queue_head(qp->rq.queue,
-					QUEUE_TYPE_FROM_CLIENT);
-			if (qp->resp.wqe) {
-				qp->resp.status = IB_WC_WR_FLUSH_ERR;
-				return RESPST_COMPLETE;
-			} else {
-				return RESPST_EXIT;
-			}
-		} else {
-			return RESPST_EXIT;
-		}
-	}
-
 	if (pkt->mask & (RXE_READ_OR_ATOMIC_MASK | RXE_ATOMIC_WRITE_MASK)) {
 		/* it is the requesters job to not send
 		 * too many read/atomic ops, we just
@@ -1425,22 +1395,66 @@ static enum resp_states do_class_d1e_error(struct rxe_qp *qp)
 	}
 }
 
-static void rxe_drain_req_pkts(struct rxe_qp *qp, bool notify)
+/* drain incoming request packet queue */
+static void rxe_drain_req_pkts(struct rxe_qp *qp)
 {
 	struct sk_buff *skb;
-	struct rxe_queue *q = qp->rq.queue;
 
 	while ((skb = skb_dequeue(&qp->req_pkts))) {
 		rxe_put(qp);
 		kfree_skb(skb);
 		ib_device_put(qp->ibqp.device);
 	}
+}
+
+/* complete receive wqe with flush error */
+static int complete_flush(struct rxe_qp *qp, struct rxe_recv_wqe *wqe)
+{
+	struct rxe_cqe cqe = {};
+	struct ib_wc *wc = &cqe.ibwc;
+	struct ib_uverbs_wc *uwc = &cqe.uibwc;
+
+	if (qp->rcq->is_user) {
+		uwc->status = IB_WC_WR_FLUSH_ERR;
+		uwc->qp_num = qp_num(qp);
+		uwc->wr_id = wqe->wr_id;
+	} else {
+		wc->status = IB_WC_WR_FLUSH_ERR;
+		wc->qp = &qp->ibqp;
+		wc->wr_id = wqe->wr_id;
+	}
+
+	if (rxe_cq_post(qp->rcq, &cqe, 0))
+		return -ENOMEM;
+
+	return 0;
+}
+
+/* drain and optionally complete the recive queue
+ * if unable to complete a wqe stop completing and
+ * just flush the remaining wqes
+ */
+static void rxe_drain_recv_queue(struct rxe_qp *qp, bool notify)
+{
+	struct rxe_queue *q = qp->rq.queue;
+	struct rxe_recv_wqe *wqe;
+	int err;
 
-	if (notify)
+	if (qp->srq)
 		return;
 
-	while (!qp->srq && q && queue_head(q, q->type))
+	while ((wqe = queue_head(q, q->type))) {
+		if (notify) {
+			err = complete_flush(qp, wqe);
+			if (err) {
+				rxe_dbg_qp(qp, "complete failed for recv wqe");
+				notify = 0;
+			}
+		}
 		queue_advance_consumer(q, q->type);
+	}
+
+	qp->resp.wqe = NULL;
 }
 
 int rxe_responder(struct rxe_qp *qp)
@@ -1453,20 +1467,18 @@ int rxe_responder(struct rxe_qp *qp)
 	if (!rxe_get(qp))
 		return -EAGAIN;
 
-	qp->resp.aeth_syndrome = AETH_ACK_UNLIMITED;
-
-	if (!qp->valid)
+	if (!qp->valid || qp->resp.state == QP_STATE_ERROR ||
+	    qp->resp.state == QP_STATE_RESET) {
+		bool notify = qp->valid &&
+				(qp->resp.state == QP_STATE_ERROR);
+		rxe_drain_req_pkts(qp);
+		rxe_drain_recv_queue(qp, notify);
 		goto exit;
+	}
 
-	switch (qp->resp.state) {
-	case QP_STATE_RESET:
-		state = RESPST_RESET;
-		break;
+	qp->resp.aeth_syndrome = AETH_ACK_UNLIMITED;
 
-	default:
-		state = RESPST_GET_REQ;
-		break;
-	}
+	state = RESPST_GET_REQ;
 
 	while (1) {
 		rxe_dbg_qp(qp, "state = %s\n", resp_state_name[state]);
@@ -1625,11 +1637,6 @@ int rxe_responder(struct rxe_qp *qp)
 
 			goto exit;
 
-		case RESPST_RESET:
-			rxe_drain_req_pkts(qp, false);
-			qp->resp.wqe = NULL;
-			goto exit;
-
 		case RESPST_ERROR:
 			qp->resp.goto_error = 0;
 			rxe_dbg_qp(qp, "moved to error state\n");