RDMA/cxgb4: Fix QP flush logic

This patch makes following fixes in QP flush logic:

- correctly flushes unsignaled WRs followed by a signaled WR
- supports for flushing a CQ bound to multiple QPs
- resets cidx_flush if a active queue starts getting HW CQEs again
- marks WQ in error when we leave RTS. This was only being done for
  user queues, but we need it for kernel queues too so that
  post_send/post_recv will start returning the appropriate error
  synchronously
- eats unsignaled read resp CQEs. HW always inserts CQEs so we must
  silently discard them if the read work request was unsignaled.
- handles QP flushes with pending SW CQEs. The flush and out of order
  completion logic has a bug where if out of order completions are
  flushed but not yet polled by the consumer and the qp is then
  flushed then we end up inserting duplicate completions.
- c4iw_flush_sq() should only flush wrs that have not already been
  flushed.  Since we already track where in the SQ we've flushed via
  sq.cidx_flush, just start at that point and flush any remaining.
  This bug only caused a problem in the presence of unsignaled work
  requests.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Vipul Pandya <vipul@chelsio.com>

[ Fixed sparse warning due to htonl/ntohl confusion.  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>

drivers/infiniband/hw/cxgb4/qp.c

@@ -737,6 +737,7 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		swsqe->idx = qhp->wq.sq.pidx;
 		swsqe->complete = 0;
 		swsqe->signaled = (wr->send_flags & IB_SEND_SIGNALED);
+		swsqe->flushed = 0;
 		swsqe->wr_id = wr->wr_id;
 
 		init_wr_hdr(wqe, qhp->wq.sq.pidx, fw_opcode, fw_flags, len16);
@@ -1006,7 +1007,15 @@ static void __flush_qp(struct c4iw_qp *qhp, struct c4iw_cq *rchp,
 	/* locking hierarchy: cq lock first, then qp lock. */
 	spin_lock_irqsave(&rchp->lock, flag);
 	spin_lock(&qhp->lock);
-	c4iw_flush_hw_cq(&rchp->cq);
+
+	if (qhp->wq.flushed) {
+		spin_unlock(&qhp->lock);
+		spin_unlock_irqrestore(&rchp->lock, flag);
+		return;
+	}
+	qhp->wq.flushed = 1;
+
+	c4iw_flush_hw_cq(rchp);
 	c4iw_count_rcqes(&rchp->cq, &qhp->wq, &count);
 	flushed = c4iw_flush_rq(&qhp->wq, &rchp->cq, count);
 	spin_unlock(&qhp->lock);
@@ -1020,9 +1029,9 @@ static void __flush_qp(struct c4iw_qp *qhp, struct c4iw_cq *rchp,
 	/* locking hierarchy: cq lock first, then qp lock. */
 	spin_lock_irqsave(&schp->lock, flag);
 	spin_lock(&qhp->lock);
-	c4iw_flush_hw_cq(&schp->cq);
-	c4iw_count_scqes(&schp->cq, &qhp->wq, &count);
-	flushed = c4iw_flush_sq(&qhp->wq, &schp->cq, count);
+	if (schp != rchp)
+		c4iw_flush_hw_cq(schp);
+	flushed = c4iw_flush_sq(qhp);
 	spin_unlock(&qhp->lock);
 	spin_unlock_irqrestore(&schp->lock, flag);
 	if (flushed) {
@@ -1037,11 +1046,11 @@ static void flush_qp(struct c4iw_qp *qhp)
 	struct c4iw_cq *rchp, *schp;
 	unsigned long flag;
 
-	rchp = get_chp(qhp->rhp, qhp->attr.rcq);
-	schp = get_chp(qhp->rhp, qhp->attr.scq);
+	rchp = to_c4iw_cq(qhp->ibqp.recv_cq);
+	schp = to_c4iw_cq(qhp->ibqp.send_cq);
 
+	t4_set_wq_in_error(&qhp->wq);
 	if (qhp->ibqp.uobject) {
-		t4_set_wq_in_error(&qhp->wq);
 		t4_set_cq_in_error(&rchp->cq);
 		spin_lock_irqsave(&rchp->comp_handler_lock, flag);
 		(*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context);
@@ -1330,8 +1339,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 				disconnect = 1;
 				c4iw_get_ep(&qhp->ep->com);
 			}
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			ret = rdma_fini(rhp, qhp, ep);
 			if (ret)
 				goto err;
@@ -1340,8 +1348,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 			set_state(qhp, C4IW_QP_STATE_TERMINATE);
 			qhp->attr.layer_etype = attrs->layer_etype;
 			qhp->attr.ecode = attrs->ecode;
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			ep = qhp->ep;
 			if (!internal)
 				terminate = 1;
@@ -1350,8 +1357,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 			break;
 		case C4IW_QP_STATE_ERROR:
 			set_state(qhp, C4IW_QP_STATE_ERROR);
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			if (!internal) {
 				abort = 1;
 				disconnect = 1;
@@ -1552,12 +1558,12 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 
 	ucontext = pd->uobject ? to_c4iw_ucontext(pd->uobject->context) : NULL;
 
-
 	qhp = kzalloc(sizeof(*qhp), GFP_KERNEL);
 	if (!qhp)
 		return ERR_PTR(-ENOMEM);
 	qhp->wq.sq.size = sqsize;
 	qhp->wq.sq.memsize = (sqsize + 1) * sizeof *qhp->wq.sq.queue;
+	qhp->wq.sq.flush_cidx = -1;
 	qhp->wq.rq.size = rqsize;
 	qhp->wq.rq.memsize = (rqsize + 1) * sizeof *qhp->wq.rq.queue;