IB/{rdmavt, hfi1, qib}: Remove AH refcount for UD QPs

Historically rdmavt destroy_ah() has returned an -EBUSY when the AH has a
non-zero reference count.  IBTA 11.2.2 notes no such return value or error
case:

	Output Modifiers:
	- Verb results:
	- Operation completed successfully.
	- Invalid HCA handle.
	- Invalid address handle.

ULPs never test for this error and this will leak memory.

The reference count exists to allow for driver independent progress
mechanisms to process UD SWQEs in parallel with post sends.  The SWQE will
hold a reference count until the UD SWQE completes and then drops the
reference.

Fix by removing need to reference count the AH.  Add a UD specific
allocation to each SWQE entry to cache the necessary information for
independent progress.  Copy the information during the post send
processing.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

4 files changed, 30 insertions, 29 deletions

diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c
index 41261e72c429..a84b44af7b97 100644
--- a/drivers/infiniband/hw/hfi1/qp.c
+++ b/drivers/infiniband/hw/hfi1/qp.c
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015 - 2018 Intel Corporation.
+ * Copyright(c) 2015 - 2019 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -348,7 +348,7 @@ int hfi1_setup_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe, bool *call_send)
 		break;
 	case IB_QPT_GSI:
 	case IB_QPT_UD:
-		ah = ibah_to_rvtah(wqe->ud_wr.ah);
+		ah = ibah_to_rvtah(wqe->ud_wr.wr.ah);
 		if (wqe->length > (1 << ah->log_pmtu))
 			return -EINVAL;
 		if (ibp->sl_to_sc[rdma_ah_get_sl(&ah->attr)] == 0xf)
diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c
index e16d499cfd1e..f8e796e45517 100644
--- a/drivers/infiniband/hw/hfi1/ud.c
+++ b/drivers/infiniband/hw/hfi1/ud.c
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015 - 2018 Intel Corporation.
+ * Copyright(c) 2015 - 2019 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -87,7 +87,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	rcu_read_lock();
 
 	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), &ibp->rvp,
-			    swqe->ud_wr.remote_qpn);
+			    swqe->ud_wr.wr.remote_qpn);
 	if (!qp) {
 		ibp->rvp.n_pkt_drops++;
 		rcu_read_unlock();
@@ -105,7 +105,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		goto drop;
 	}
 
-	ah_attr = &ibah_to_rvtah(swqe->ud_wr.ah)->attr;
+	ah_attr = swqe->ud_wr.attr;
 	ppd = ppd_from_ibp(ibp);
 
 	if (qp->ibqp.qp_num > 1) {
@@ -135,8 +135,8 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	if (qp->ibqp.qp_num) {
 		u32 qkey;
 
-		qkey = (int)swqe->ud_wr.remote_qkey < 0 ?
-			sqp->qkey : swqe->ud_wr.remote_qkey;
+		qkey = (int)swqe->ud_wr.wr.remote_qkey < 0 ?
+			sqp->qkey : swqe->ud_wr.wr.remote_qkey;
 		if (unlikely(qkey != qp->qkey))
 			goto drop; /* silently drop per IBTA spec */
 	}
@@ -240,7 +240,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	if (qp->ibqp.qp_type == IB_QPT_GSI || qp->ibqp.qp_type == IB_QPT_SMI) {
 		if (sqp->ibqp.qp_type == IB_QPT_GSI ||
 		    sqp->ibqp.qp_type == IB_QPT_SMI)
-			wc.pkey_index = swqe->ud_wr.pkey_index;
+			wc.pkey_index = swqe->ud_wr.wr.pkey_index;
 		else
 			wc.pkey_index = sqp->s_pkey_index;
 	} else {
@@ -282,20 +282,20 @@ static void hfi1_make_bth_deth(struct rvt_qp *qp, struct rvt_swqe *wqe,
 		bth0 |= IB_BTH_SOLICITED;
 	bth0 |= extra_bytes << 20;
 	if (qp->ibqp.qp_type == IB_QPT_GSI || qp->ibqp.qp_type == IB_QPT_SMI)
-		*pkey = hfi1_get_pkey(ibp, wqe->ud_wr.pkey_index);
+		*pkey = hfi1_get_pkey(ibp, wqe->ud_wr.wr.pkey_index);
 	else
 		*pkey = hfi1_get_pkey(ibp, qp->s_pkey_index);
 	if (!bypass)
 		bth0 |= *pkey;
 	ohdr->bth[0] = cpu_to_be32(bth0);
-	ohdr->bth[1] = cpu_to_be32(wqe->ud_wr.remote_qpn);
+	ohdr->bth[1] = cpu_to_be32(wqe->ud_wr.wr.remote_qpn);
 	ohdr->bth[2] = cpu_to_be32(mask_psn(wqe->psn));
 	/*
 	 * Qkeys with the high order bit set mean use the
 	 * qkey from the QP context instead of the WR (see 10.2.5).
 	 */
-	ohdr->u.ud.deth[0] = cpu_to_be32((int)wqe->ud_wr.remote_qkey < 0 ?
-					 qp->qkey : wqe->ud_wr.remote_qkey);
+	ohdr->u.ud.deth[0] = cpu_to_be32((int)wqe->ud_wr.wr.remote_qkey < 0 ?
+					 qp->qkey : wqe->ud_wr.wr.remote_qkey);
 	ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
 }
 
@@ -315,7 +315,7 @@ void hfi1_make_ud_req_9B(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	ah_attr = wqe->ud_wr.attr;
 
 	extra_bytes = -wqe->length & 3;
 	nwords = ((wqe->length + extra_bytes) >> 2) + SIZE_OF_CRC;
@@ -379,7 +379,7 @@ void hfi1_make_ud_req_16B(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 	struct hfi1_pportdata *ppd;
 	struct hfi1_ibport *ibp;
 	u32 dlid, slid, nwords, extra_bytes;
-	u32 dest_qp = wqe->ud_wr.remote_qpn;
+	u32 dest_qp = wqe->ud_wr.wr.remote_qpn;
 	u32 src_qp = qp->ibqp.qp_num;
 	u16 len, pkey;
 	u8 l4, sc5;
@@ -387,7 +387,7 @@ void hfi1_make_ud_req_16B(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	ah_attr = wqe->ud_wr.attr;
 
 	/*
 	 * Build 16B Management Packet if either the destination
@@ -449,7 +449,7 @@ void hfi1_make_ud_req_16B(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 
 	if (is_mgmt) {
 		l4 = OPA_16B_L4_FM;
-		pkey = hfi1_get_pkey(ibp, wqe->ud_wr.pkey_index);
+		pkey = hfi1_get_pkey(ibp, wqe->ud_wr.wr.pkey_index);
 		hfi1_16B_set_qpn(&ps->s_txreq->phdr.hdr.opah.u.mgmt,
 				 dest_qp, src_qp);
 	} else {
@@ -514,7 +514,7 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 	/* Construct the header. */
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	ah_attr = wqe->ud_wr.attr;
 	priv->hdr_type = hfi1_get_hdr_type(ppd->lid, ah_attr);
 	if ((!hfi1_check_mcast(rdma_ah_get_dlid(ah_attr))) ||
 	    (rdma_ah_get_dlid(ah_attr) == be32_to_cpu(OPA_LID_PERMISSIVE))) {
diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c
index a81905df2d0f..0e1d0d692891 100644
--- a/drivers/infiniband/hw/qib/qib_qp.c
+++ b/drivers/infiniband/hw/qib/qib_qp.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 - 2017 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2012 - 2019 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation.  * All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -398,7 +398,7 @@ int qib_check_send_wqe(struct rvt_qp *qp,
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:
 	case IB_QPT_UD:
-		ah = ibah_to_rvtah(wqe->ud_wr.ah);
+		ah = ibah_to_rvtah(wqe->ud_wr.wr.ah);
 		if (wqe->length > (1 << ah->log_pmtu))
 			return -EINVAL;
 		/* progress hint */
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index 32ad0b635fc6..d8c2c968909f 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2012 - 2019 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -63,7 +64,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	enum ib_qp_type sqptype, dqptype;
 
 	rcu_read_lock();
-	qp = rvt_lookup_qpn(rdi, &ibp->rvp, swqe->ud_wr.remote_qpn);
+	qp = rvt_lookup_qpn(rdi, &ibp->rvp, swqe->ud_wr.wr.remote_qpn);
 	if (!qp) {
 		ibp->rvp.n_pkt_drops++;
 		goto drop;
@@ -80,7 +81,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		goto drop;
 	}
 
-	ah_attr = &ibah_to_rvtah(swqe->ud_wr.ah)->attr;
+	ah_attr = swqe->ud_wr.attr;
 	ppd = ppd_from_ibp(ibp);
 
 	if (qp->ibqp.qp_num > 1) {
@@ -110,8 +111,8 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	if (qp->ibqp.qp_num) {
 		u32 qkey;
 
-		qkey = (int)swqe->ud_wr.remote_qkey < 0 ?
-			sqp->qkey : swqe->ud_wr.remote_qkey;
+		qkey = (int)swqe->ud_wr.wr.remote_qkey < 0 ?
+			sqp->qkey : swqe->ud_wr.wr.remote_qkey;
 		if (unlikely(qkey != qp->qkey))
 			goto drop;
 	}
@@ -203,7 +204,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	wc.qp = &qp->ibqp;
 	wc.src_qp = sqp->ibqp.qp_num;
 	wc.pkey_index = qp->ibqp.qp_type == IB_QPT_GSI ?
-		swqe->ud_wr.pkey_index : 0;
+		swqe->ud_wr.wr.pkey_index : 0;
 	wc.slid = ppd->lid | (rdma_ah_get_path_bits(ah_attr) &
 				((1 << ppd->lmc) - 1));
 	wc.sl = rdma_ah_get_sl(ah_attr);
@@ -270,7 +271,7 @@ int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 	/* Construct the header. */
 	ibp = to_iport(qp->ibqp.device, qp->port_num);
 	ppd = ppd_from_ibp(ibp);
-	ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
+	ah_attr = wqe->ud_wr.attr;
 	if (rdma_ah_get_dlid(ah_attr) >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
 		if (rdma_ah_get_dlid(ah_attr) !=
 				be16_to_cpu(IB_LID_PERMISSIVE))
@@ -362,7 +363,7 @@ int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 	bth0 |= extra_bytes << 20;
 	bth0 |= qp->ibqp.qp_type == IB_QPT_SMI ? QIB_DEFAULT_P_KEY :
 		qib_get_pkey(ibp, qp->ibqp.qp_type == IB_QPT_GSI ?
-			     wqe->ud_wr.pkey_index : qp->s_pkey_index);
+			     wqe->ud_wr.wr.pkey_index : qp->s_pkey_index);
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	/*
 	 * Use the multicast QP if the destination LID is a multicast LID.
@@ -371,14 +372,14 @@ int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 			be16_to_cpu(IB_MULTICAST_LID_BASE) &&
 		rdma_ah_get_dlid(ah_attr) != be16_to_cpu(IB_LID_PERMISSIVE) ?
 		cpu_to_be32(QIB_MULTICAST_QPN) :
-		cpu_to_be32(wqe->ud_wr.remote_qpn);
+		cpu_to_be32(wqe->ud_wr.wr.remote_qpn);
 	ohdr->bth[2] = cpu_to_be32(wqe->psn & QIB_PSN_MASK);
 	/*
 	 * Qkeys with the high order bit set mean use the
 	 * qkey from the QP context instead of the WR (see 10.2.5).
 	 */
-	ohdr->u.ud.deth[0] = cpu_to_be32((int)wqe->ud_wr.remote_qkey < 0 ?
-					 qp->qkey : wqe->ud_wr.remote_qkey);
+	ohdr->u.ud.deth[0] = cpu_to_be32((int)wqe->ud_wr.wr.remote_qkey < 0 ?
+					 qp->qkey : wqe->ud_wr.wr.remote_qkey);
 	ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
 
 done: