diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index 9088e26d3ac8b0aef86424d244c0465c4f33b87e..444028a3582a3898f2ddcedcdb69614904ab6f0d 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -230,7 +230,7 @@ static int qib_make_rc_ack(struct qib_ibdev *dev, struct rvt_qp *qp,
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_rc_req(struct rvt_qp *qp)
+int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index a5f07a64b228d2e8f5e1796f8892ddeda6aa37fd..b6777925629704d3a2c3c7bc51a5475a725ca2c3 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -739,7 +739,7 @@ void qib_do_send(struct rvt_qp *qp)
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-	int (*make_req)(struct rvt_qp *qp);
+	int (*make_req)(struct rvt_qp *qp, unsigned long *flags);
 	unsigned long flags;
 
 	if ((qp->ibqp.qp_type == IB_QPT_RC ||
@@ -781,7 +781,7 @@ void qib_do_send(struct rvt_qp *qp)
 			qp->s_hdrwords = 0;
 			spin_lock_irqsave(&qp->s_lock, flags);
 		}
-	} while (make_req(qp));
+	} while (make_req(qp, &flags));
 
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
index 7bdbc79ceaa3bbd7a25a87ea107b06f1460dd7ad..1d61bd04f4494bf12a6c020bad53a6fe1c6f8e81 100644
--- a/drivers/infiniband/hw/qib/qib_uc.c
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -45,7 +45,7 @@
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_uc_req(struct rvt_qp *qp)
+int qib_make_uc_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index d9502137de626b80966cfd200e3806c65c36b21a..846e6c726df7c67204f1dc1d5f1bfb066c36e149 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -238,7 +238,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
  *
  * Return 1 if constructed; otherwise, return 0.
  */
-int qib_make_ud_req(struct rvt_qp *qp)
+int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 {
 	struct qib_qp_priv *priv = qp->priv;
 	struct qib_other_headers *ohdr;
@@ -294,7 +294,7 @@ int qib_make_ud_req(struct rvt_qp *qp)
 		this_cpu_inc(ibp->pmastats->n_unicast_xmit);
 		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
 		if (unlikely(lid == ppd->lid)) {
-			unsigned long flags;
+			unsigned long tflags = *flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -307,10 +307,10 @@ int qib_make_ud_req(struct rvt_qp *qp)
 				goto bail;
 			}
 			qp->s_cur = next_cur;
-			local_irq_save(flags);
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, tflags);
 			qib_ud_loopback(qp, wqe);
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, tflags);
+			*flags = tflags;
 			qib_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done;
 		}
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index 4b76a8d593371b3929faf86437890ff8191ecf1c..6888f03c6d61007b6f4d86c1f73b403908e08279 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -430,11 +430,11 @@ void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 
 void qib_send_rc_ack(struct rvt_qp *qp);
 
-int qib_make_rc_req(struct rvt_qp *qp);
+int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags);
 
-int qib_make_uc_req(struct rvt_qp *qp);
+int qib_make_uc_req(struct rvt_qp *qp, unsigned long *flags);
 
-int qib_make_ud_req(struct rvt_qp *qp);
+int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags);
 
 int qib_register_ib_device(struct qib_devdata *);
 
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 08813cdbd475773b73bd9ee38ee9ff9cd40f0100..a659aec3c3c6b95823650c7b6bb3ac800649967a 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -831,7 +831,6 @@ void hfi1_do_send(struct rvt_qp *qp)
 	struct hfi1_pkt_state ps;
 	struct hfi1_qp_priv *priv = qp->priv;
 	int (*make_req)(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
-	unsigned long flags;
 	unsigned long timeout;
 	unsigned long timeout_int;
 	int cpu;
@@ -866,11 +865,11 @@ void hfi1_do_send(struct rvt_qp *qp)
 		timeout_int = SEND_RESCHED_TIMEOUT;
 	}
 
-	spin_lock_irqsave(&qp->s_lock, flags);
+	spin_lock_irqsave(&qp->s_lock, ps.flags);
 
 	/* Return if we are already busy processing a work request. */
 	if (!hfi1_send_ok(qp)) {
-		spin_unlock_irqrestore(&qp->s_lock, flags);
+		spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 		return;
 	}
 
@@ -884,7 +883,7 @@ void hfi1_do_send(struct rvt_qp *qp)
 	do {
 		/* Check for a constructed packet to be sent. */
 		if (qp->s_hdrwords != 0) {
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 			/*
 			 * If the packet cannot be sent now, return and
 			 * the send tasklet will be woken up later.
@@ -897,11 +896,14 @@ void hfi1_do_send(struct rvt_qp *qp)
 			if (unlikely(time_after(jiffies, timeout))) {
 				if (workqueue_congested(cpu,
 							ps.ppd->hfi1_wq)) {
-					spin_lock_irqsave(&qp->s_lock, flags);
+					spin_lock_irqsave(
+						&qp->s_lock,
+						ps.flags);
 					qp->s_flags &= ~RVT_S_BUSY;
 					hfi1_schedule_send(qp);
-					spin_unlock_irqrestore(&qp->s_lock,
-							       flags);
+					spin_unlock_irqrestore(
+						&qp->s_lock,
+						ps.flags);
 					this_cpu_inc(
 						*ps.ppd->dd->send_schedule);
 					return;
@@ -913,11 +915,11 @@ void hfi1_do_send(struct rvt_qp *qp)
 				}
 				timeout = jiffies + (timeout_int) / 8;
 			}
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, ps.flags);
 		}
 	} while (make_req(qp, &ps));
 
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+	spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 }
 
 /*
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index ae8a70f703ebcbb877abf630f501158688c6fb4a..1e503ad0bebb764ee1e05462604538e6324002f8 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -322,7 +322,7 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 			     (lid == ppd->lid ||
 			      (lid == be16_to_cpu(IB_LID_PERMISSIVE) &&
 			      qp->ibqp.qp_type == IB_QPT_GSI)))) {
-			unsigned long flags;
+			unsigned long tflags = ps->flags;
 			/*
 			 * If DMAs are in progress, we can't generate
 			 * a completion for the loopback packet since
@@ -335,10 +335,10 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 				goto bail;
 			}
 			qp->s_cur = next_cur;
-			local_irq_save(flags);
-			spin_unlock_irqrestore(&qp->s_lock, flags);
+			spin_unlock_irqrestore(&qp->s_lock, tflags);
 			ud_loopback(qp, wqe);
-			spin_lock_irqsave(&qp->s_lock, flags);
+			spin_lock_irqsave(&qp->s_lock, tflags);
+			ps->flags = tflags;
 			hfi1_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done_free_tx;
 		}
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 6c4670fffdbb8bf852a1ecaa6243339807bffd50..2ba1373f4fb4ca7e79d8bdaca4a79150e32212a8 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -215,6 +215,7 @@ struct hfi1_pkt_state {
 	struct hfi1_ibport *ibp;
 	struct hfi1_pportdata *ppd;
 	struct verbs_txreq *s_txreq;
+	unsigned long flags;
 };
 
 #define HFI1_PSN_CREDIT  16