[PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[Michal Kalderon]

This patch series adds iWARP support to our QL4xxxx networking adapters.
The code changes span across qed and qedr drivers, but this series contains
changes to qed only. Once the series is accepted, the qedr series will
be submitted to the rdma tree.
There is one additional qed patch which enables the iWARP, this patch is
delayed until the qedr series will be accepted. 

The patches were previously sent as an RFC, and these are the first 12
patches in the RFC series:
https://www.spinics.net/lists/linux-rdma/msg51416.html

This series was tested and built against net-next.

MAINTAINERS file is not updated in this PATCH as there is a pending patch
for qedr driver update https://patchwork.kernel.org/patch/9752761.

Michal Kalderon (12):
  qed: Introduce iWARP personality
  qed: Implement iWARP initialization, teardown and qp operations
  qed: Rename some ll2 related defines
  qed: Add iWARP support in ll2 connections
  qed: iWARP CM - setup a ll2 connection for handling SYN packets
  qed: iWARP CM add listener functions and initial SYN processing
  qed: iWARP CM add passive side connect
  qed: iWARP CM add active side connect
  qed: iWARP implement disconnect flows
  qed: iWARP CM add error handling
  qed: Add iWARP protocol support in context allocation
  qed: Add iWARP support for physical queue allocation

 drivers/net/ethernet/qlogic/qed/Makefile    |    2 +-
 drivers/net/ethernet/qlogic/qed/qed.h       |   30 +-
 drivers/net/ethernet/qlogic/qed/qed_cxt.c   |   21 +-
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |   36 +-
 drivers/net/ethernet/qlogic/qed/qed_hsi.h   |    1 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 2409 +++++++++++++++++++++++++++
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  189 +++
 drivers/net/ethernet/qlogic/qed/qed_l2.c    |   16 +-
 drivers/net/ethernet/qlogic/qed/qed_ll2.c   |   42 +-
 drivers/net/ethernet/qlogic/qed/qed_main.c  |   17 +-
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |  139 +-
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |    5 +
 drivers/net/ethernet/qlogic/qed/qed_roce.c  |   20 +
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |    7 +-
 include/linux/qed/common_hsi.h              |    4 +-
 include/linux/qed/iwarp_common.h            |   53 +
 include/linux/qed/qed_ll2_if.h              |    3 +-
 include/linux/qed/qed_rdma_if.h             |  114 ++
 18 files changed, 3008 insertions(+), 100 deletions(-)
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.c
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.h
 create mode 100644 include/linux/qed/iwarp_common.h

-- 
1.8.3.1

[PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[Michal Kalderon]

This patch series adds iWARP support to our QL4xxxx networking adapters.
The code changes span across qed and qedr drivers, but this series contains
changes to qed only. Once the series is accepted, the qedr series will
be submitted to the rdma tree.
There is one additional qed patch which enables the iWARP, this patch is
delayed until the qedr series will be accepted. 

The patches were previously sent as an RFC, and these are the first 12
patches in the RFC series:
https://www.spinics.net/lists/linux-rdma/msg51416.html

This series was tested and built against net-next.

MAINTAINERS file is not updated in this PATCH as there is a pending patch
for qedr driver update https://patchwork.kernel.org/patch/9752761.

Michal Kalderon (12):
  qed: Introduce iWARP personality
  qed: Implement iWARP initialization, teardown and qp operations
  qed: Rename some ll2 related defines
  qed: Add iWARP support in ll2 connections
  qed: iWARP CM - setup a ll2 connection for handling SYN packets
  qed: iWARP CM add listener functions and initial SYN processing
  qed: iWARP CM add passive side connect
  qed: iWARP CM add active side connect
  qed: iWARP implement disconnect flows
  qed: iWARP CM add error handling
  qed: Add iWARP protocol support in context allocation
  qed: Add iWARP support for physical queue allocation

 drivers/net/ethernet/qlogic/qed/Makefile    |    2 +-
 drivers/net/ethernet/qlogic/qed/qed.h       |   30 +-
 drivers/net/ethernet/qlogic/qed/qed_cxt.c   |   21 +-
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |   36 +-
 drivers/net/ethernet/qlogic/qed/qed_hsi.h   |    1 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 2409 +++++++++++++++++++++++++++
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  189 +++
 drivers/net/ethernet/qlogic/qed/qed_l2.c    |   16 +-
 drivers/net/ethernet/qlogic/qed/qed_ll2.c   |   42 +-
 drivers/net/ethernet/qlogic/qed/qed_main.c  |   17 +-
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |  139 +-
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |    5 +
 drivers/net/ethernet/qlogic/qed/qed_roce.c  |   20 +
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |    7 +-
 include/linux/qed/common_hsi.h              |    4 +-
 include/linux/qed/iwarp_common.h            |   53 +
 include/linux/qed/qed_ll2_if.h              |    3 +-
 include/linux/qed/qed_rdma_if.h             |  114 ++
 18 files changed, 3008 insertions(+), 100 deletions(-)
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.c
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.h
 create mode 100644 include/linux/qed/iwarp_common.h

-- 
1.8.3.1

[PATCH net-next 03/12] qed: Rename some ll2 related defines

[Michal Kalderon]

Make some names more generic as they will be used by iWARP too.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed.h     |  2 +-
 drivers/net/ethernet/qlogic/qed/qed_ll2.c | 29 ++++++++++++++---------------
 include/linux/qed/qed_ll2_if.h            |  2 +-
 3 files changed, 16 insertions(+), 17 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 22e1171..fd8cd5e 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -779,7 +779,7 @@ static inline u8 qed_concrete_to_sw_fid(struct qed_dev *cdev,
 }
 
 #define PURE_LB_TC 8
-#define OOO_LB_TC 9
+#define PKT_LB_TC 9
 
 int qed_configure_vport_wfq(struct qed_dev *cdev, u16 vp_id, u32 rate);
 void qed_configure_vp_wfq_on_link_change(struct qed_dev *cdev,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index be66f19..e235fb2 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -309,7 +309,7 @@ static void qed_ll2_txq_flush(struct qed_hwfn *p_hwfn, u8 connection_handle)
 		list_del(&p_pkt->list_entry);
 		b_last_packet = list_empty(&p_tx->active_descq);
 		list_add_tail(&p_pkt->list_entry, &p_tx->free_descq);
-		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO) {
 			struct qed_ooo_buffer *p_buffer;
 
 			p_buffer = (struct qed_ooo_buffer *)p_pkt->cookie;
@@ -532,7 +532,7 @@ static void qed_ll2_rxq_flush(struct qed_hwfn *p_hwfn, u8 connection_handle)
 
 		list_move_tail(&p_pkt->list_entry, &p_rx->free_descq);
 
-		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO) {
 			struct qed_ooo_buffer *p_buffer;
 
 			p_buffer = (struct qed_ooo_buffer *)p_pkt->cookie;
@@ -893,8 +893,7 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->drop_ttl0_flg = p_ll2_conn->input.rx_drop_ttl0_flg;
 	p_ramrod->inner_vlan_removal_en = p_ll2_conn->input.rx_vlan_removal_en;
 	p_ramrod->queue_id = p_ll2_conn->queue_id;
-	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_ISCSI_OOO) ? 0
-									  : 1;
+	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_OOO) ? 0 : 1;
 
 	if ((IS_MF_DEFAULT(p_hwfn) || IS_MF_SI(p_hwfn)) &&
 	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE)) {
@@ -924,7 +923,7 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 	if (!QED_LL2_TX_REGISTERED(p_ll2_conn))
 		return 0;
 
-	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO)
 		p_ll2_conn->tx_stats_en = 0;
 	else
 		p_ll2_conn->tx_stats_en = 1;
@@ -955,10 +954,10 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->pbl_size = cpu_to_le16(pbl_size);
 
 	switch (p_ll2_conn->input.tx_tc) {
-	case LB_TC:
+	case PURE_LB_TC:
 		pq_id = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_LB);
 		break;
-	case OOO_LB_TC:
+	case PKT_LB_TC:
 		pq_id = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OOO);
 		break;
 	default:
@@ -973,7 +972,7 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 		p_ramrod->conn_type = PROTOCOLID_FCOE;
 		break;
 	case QED_LL2_TYPE_ISCSI:
-	case QED_LL2_TYPE_ISCSI_OOO:
+	case QED_LL2_TYPE_OOO:
 		p_ramrod->conn_type = PROTOCOLID_ISCSI;
 		break;
 	case QED_LL2_TYPE_ROCE:
@@ -1142,7 +1141,7 @@ static int qed_ll2_acquire_connection_tx(struct qed_hwfn *p_hwfn,
 	u16 buf_idx;
 	int rc = 0;
 
-	if (p_ll2_info->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_info->input.conn_type != QED_LL2_TYPE_OOO)
 		return rc;
 
 	/* Correct number of requested OOO buffers if needed */
@@ -1280,7 +1279,7 @@ int qed_ll2_acquire_connection(void *cxt, struct qed_ll2_acquire_data *data)
 		goto q_allocate_fail;
 
 	/* Register callbacks for the Rx/Tx queues */
-	if (data->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+	if (data->input.conn_type == QED_LL2_TYPE_OOO) {
 		comp_rx_cb = qed_ll2_lb_rxq_completion;
 		comp_tx_cb = qed_ll2_lb_txq_completion;
 	} else {
@@ -1339,7 +1338,7 @@ static int qed_ll2_establish_connection_rx(struct qed_hwfn *p_hwfn,
 qed_ll2_establish_connection_ooo(struct qed_hwfn *p_hwfn,
 				 struct qed_ll2_info *p_ll2_conn)
 {
-	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_OOO)
 		return;
 
 	qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
@@ -1794,7 +1793,7 @@ int qed_ll2_terminate_connection(void *cxt, u8 connection_handle)
 		qed_ll2_rxq_flush(p_hwfn, connection_handle);
 	}
 
-	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO)
 		qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
 
 	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_FCOE) {
@@ -1816,7 +1815,7 @@ static void qed_ll2_release_connection_ooo(struct qed_hwfn *p_hwfn,
 {
 	struct qed_ooo_buffer *p_buffer;
 
-	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_OOO)
 		return;
 
 	qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
@@ -2063,7 +2062,7 @@ static void qed_ll2_set_conn_data(struct qed_dev *cdev,
 	ll2_cbs.cookie = QED_LEADING_HWFN(cdev);
 
 	if (lb) {
-		data->input.tx_tc = OOO_LB_TC;
+		data->input.tx_tc = PKT_LB_TC;
 		data->input.tx_dest = QED_LL2_TX_DEST_LB;
 	} else {
 		data->input.tx_tc = 0;
@@ -2080,7 +2079,7 @@ static int qed_ll2_start_ooo(struct qed_dev *cdev,
 	int rc;
 
 	qed_ll2_set_conn_data(cdev, &data, params,
-			      QED_LL2_TYPE_ISCSI_OOO, handle, true);
+			      QED_LL2_TYPE_OOO, handle, true);
 
 	rc = qed_ll2_acquire_connection(hwfn, &data);
 	if (rc) {
diff --git a/include/linux/qed/qed_ll2_if.h b/include/linux/qed/qed_ll2_if.h
index 5958b45e..c9c56bc 100644
--- a/include/linux/qed/qed_ll2_if.h
+++ b/include/linux/qed/qed_ll2_if.h
@@ -47,7 +47,7 @@ enum qed_ll2_conn_type {
 	QED_LL2_TYPE_FCOE,
 	QED_LL2_TYPE_ISCSI,
 	QED_LL2_TYPE_TEST,
-	QED_LL2_TYPE_ISCSI_OOO,
+	QED_LL2_TYPE_OOO,
 	QED_LL2_TYPE_RESERVED2,
 	QED_LL2_TYPE_ROCE,
 	QED_LL2_TYPE_RESERVED3,
-- 
1.8.3.1

[PATCH net-next 03/12] qed: Rename some ll2 related defines

[Michal Kalderon]

Make some names more generic as they will be used by iWARP too.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed.h     |  2 +-
 drivers/net/ethernet/qlogic/qed/qed_ll2.c | 29 ++++++++++++++---------------
 include/linux/qed/qed_ll2_if.h            |  2 +-
 3 files changed, 16 insertions(+), 17 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 22e1171..fd8cd5e 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -779,7 +779,7 @@ static inline u8 qed_concrete_to_sw_fid(struct qed_dev *cdev,
 }
 
 #define PURE_LB_TC 8
-#define OOO_LB_TC 9
+#define PKT_LB_TC 9
 
 int qed_configure_vport_wfq(struct qed_dev *cdev, u16 vp_id, u32 rate);
 void qed_configure_vp_wfq_on_link_change(struct qed_dev *cdev,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index be66f19..e235fb2 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -309,7 +309,7 @@ static void qed_ll2_txq_flush(struct qed_hwfn *p_hwfn, u8 connection_handle)
 		list_del(&p_pkt->list_entry);
 		b_last_packet = list_empty(&p_tx->active_descq);
 		list_add_tail(&p_pkt->list_entry, &p_tx->free_descq);
-		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO) {
 			struct qed_ooo_buffer *p_buffer;
 
 			p_buffer = (struct qed_ooo_buffer *)p_pkt->cookie;
@@ -532,7 +532,7 @@ static void qed_ll2_rxq_flush(struct qed_hwfn *p_hwfn, u8 connection_handle)
 
 		list_move_tail(&p_pkt->list_entry, &p_rx->free_descq);
 
-		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+		if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO) {
 			struct qed_ooo_buffer *p_buffer;
 
 			p_buffer = (struct qed_ooo_buffer *)p_pkt->cookie;
@@ -893,8 +893,7 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->drop_ttl0_flg = p_ll2_conn->input.rx_drop_ttl0_flg;
 	p_ramrod->inner_vlan_removal_en = p_ll2_conn->input.rx_vlan_removal_en;
 	p_ramrod->queue_id = p_ll2_conn->queue_id;
-	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_ISCSI_OOO) ? 0
-									  : 1;
+	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_OOO) ? 0 : 1;
 
 	if ((IS_MF_DEFAULT(p_hwfn) || IS_MF_SI(p_hwfn)) &&
 	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE)) {
@@ -924,7 +923,7 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 	if (!QED_LL2_TX_REGISTERED(p_ll2_conn))
 		return 0;
 
-	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO)
 		p_ll2_conn->tx_stats_en = 0;
 	else
 		p_ll2_conn->tx_stats_en = 1;
@@ -955,10 +954,10 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->pbl_size = cpu_to_le16(pbl_size);
 
 	switch (p_ll2_conn->input.tx_tc) {
-	case LB_TC:
+	case PURE_LB_TC:
 		pq_id = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_LB);
 		break;
-	case OOO_LB_TC:
+	case PKT_LB_TC:
 		pq_id = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OOO);
 		break;
 	default:
@@ -973,7 +972,7 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 		p_ramrod->conn_type = PROTOCOLID_FCOE;
 		break;
 	case QED_LL2_TYPE_ISCSI:
-	case QED_LL2_TYPE_ISCSI_OOO:
+	case QED_LL2_TYPE_OOO:
 		p_ramrod->conn_type = PROTOCOLID_ISCSI;
 		break;
 	case QED_LL2_TYPE_ROCE:
@@ -1142,7 +1141,7 @@ static int qed_ll2_acquire_connection_tx(struct qed_hwfn *p_hwfn,
 	u16 buf_idx;
 	int rc = 0;
 
-	if (p_ll2_info->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_info->input.conn_type != QED_LL2_TYPE_OOO)
 		return rc;
 
 	/* Correct number of requested OOO buffers if needed */
@@ -1280,7 +1279,7 @@ int qed_ll2_acquire_connection(void *cxt, struct qed_ll2_acquire_data *data)
 		goto q_allocate_fail;
 
 	/* Register callbacks for the Rx/Tx queues */
-	if (data->input.conn_type == QED_LL2_TYPE_ISCSI_OOO) {
+	if (data->input.conn_type == QED_LL2_TYPE_OOO) {
 		comp_rx_cb = qed_ll2_lb_rxq_completion;
 		comp_tx_cb = qed_ll2_lb_txq_completion;
 	} else {
@@ -1339,7 +1338,7 @@ static int qed_ll2_establish_connection_rx(struct qed_hwfn *p_hwfn,
 qed_ll2_establish_connection_ooo(struct qed_hwfn *p_hwfn,
 				 struct qed_ll2_info *p_ll2_conn)
 {
-	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_OOO)
 		return;
 
 	qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
@@ -1794,7 +1793,7 @@ int qed_ll2_terminate_connection(void *cxt, u8 connection_handle)
 		qed_ll2_rxq_flush(p_hwfn, connection_handle);
 	}
 
-	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_OOO)
 		qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
 
 	if (p_ll2_conn->input.conn_type == QED_LL2_TYPE_FCOE) {
@@ -1816,7 +1815,7 @@ static void qed_ll2_release_connection_ooo(struct qed_hwfn *p_hwfn,
 {
 	struct qed_ooo_buffer *p_buffer;
 
-	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_ISCSI_OOO)
+	if (p_ll2_conn->input.conn_type != QED_LL2_TYPE_OOO)
 		return;
 
 	qed_ooo_release_all_isles(p_hwfn, p_hwfn->p_ooo_info);
@@ -2063,7 +2062,7 @@ static void qed_ll2_set_conn_data(struct qed_dev *cdev,
 	ll2_cbs.cookie = QED_LEADING_HWFN(cdev);
 
 	if (lb) {
-		data->input.tx_tc = OOO_LB_TC;
+		data->input.tx_tc = PKT_LB_TC;
 		data->input.tx_dest = QED_LL2_TX_DEST_LB;
 	} else {
 		data->input.tx_tc = 0;
@@ -2080,7 +2079,7 @@ static int qed_ll2_start_ooo(struct qed_dev *cdev,
 	int rc;
 
 	qed_ll2_set_conn_data(cdev, &data, params,
-			      QED_LL2_TYPE_ISCSI_OOO, handle, true);
+			      QED_LL2_TYPE_OOO, handle, true);
 
 	rc = qed_ll2_acquire_connection(hwfn, &data);
 	if (rc) {
diff --git a/include/linux/qed/qed_ll2_if.h b/include/linux/qed/qed_ll2_if.h
index 5958b45e..c9c56bc 100644
--- a/include/linux/qed/qed_ll2_if.h
+++ b/include/linux/qed/qed_ll2_if.h
@@ -47,7 +47,7 @@ enum qed_ll2_conn_type {
 	QED_LL2_TYPE_FCOE,
 	QED_LL2_TYPE_ISCSI,
 	QED_LL2_TYPE_TEST,
-	QED_LL2_TYPE_ISCSI_OOO,
+	QED_LL2_TYPE_OOO,
 	QED_LL2_TYPE_RESERVED2,
 	QED_LL2_TYPE_ROCE,
 	QED_LL2_TYPE_RESERVED3,
-- 
1.8.3.1

[PATCH net-next 06/12] qed: iWARP CM add listener functions and initial SYN processing

[Michal Kalderon]

This patch adds the ability to add and remove listeners and identify
whether the SYN packet received is intended for iWARP or not. If
a listener is not found the SYN packet is posted back to the chip.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 269 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  23 +++
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |   2 +
 include/linux/qed/qed_rdma_if.h             |  52 ++++++
 4 files changed, 343 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index f3b4b32..2bab57c 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -31,6 +31,10 @@
  */
 #include <linux/if_ether.h>
 #include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/spinlock.h>
+#include <linux/tcp.h>
 #include "qed_cxt.h"
 #include "qed_hw.h"
 #include "qed_ll2.h"
@@ -477,6 +481,31 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
 }
 
+static void
+qed_iwarp_print_cm_info(struct qed_hwfn *p_hwfn,
+			struct qed_iwarp_cm_info *cm_info)
+{
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "ip_version = %d\n",
+		   cm_info->ip_version);
+
+	if (cm_info->ip_version == QED_TCP_IPV4)
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "remote_ip %pI4h:%x, local_ip %pI4h:%x vlan=%x\n",
+			   cm_info->remote_ip, cm_info->remote_port,
+			   cm_info->local_ip, cm_info->local_port,
+			   cm_info->vlan);
+	else
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "remote_ip %pI6h:%x, local_ip %pI6h:%x vlan=%x\n",
+			   cm_info->remote_ip, cm_info->remote_port,
+			   cm_info->local_ip, cm_info->local_port,
+			   cm_info->vlan);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "private_data_len = %x ord = %d, ird = %d\n",
+		   cm_info->private_data_len, cm_info->ord, cm_info->ird);
+}
+
 static int
 qed_iwarp_ll2_post_rx(struct qed_hwfn *p_hwfn,
 		      struct qed_iwarp_ll2_buff *buf, u8 handle)
@@ -497,11 +526,147 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	return rc;
 }
 
+static struct qed_iwarp_listener *
+qed_iwarp_get_listener(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_cm_info *cm_info)
+{
+	struct qed_iwarp_listener *listener = NULL;
+	static const u32 ip_zero[4] = { 0, 0, 0, 0 };
+	bool found = false;
+
+	qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+	list_for_each_entry(listener,
+			    &p_hwfn->p_rdma_info->iwarp.listen_list,
+			    list_entry) {
+		if (listener->port == cm_info->local_port) {
+			if (!memcmp(listener->ip_addr,
+				    ip_zero, sizeof(ip_zero))) {
+				found = true;
+				break;
+			}
+
+			if (!memcmp(listener->ip_addr,
+				    cm_info->local_ip,
+				    sizeof(cm_info->local_ip)) &&
+			    (listener->vlan == cm_info->vlan)) {
+				found = true;
+				break;
+			}
+		}
+	}
+
+	if (found) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener found = %p\n",
+			   listener);
+		return listener;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener not found\n");
+	return NULL;
+}
+
+static int
+qed_iwarp_parse_rx_pkt(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_cm_info *cm_info,
+		       void *buf,
+		       u8 *remote_mac_addr,
+		       u8 *local_mac_addr,
+		       int *payload_len, int *tcp_start_offset)
+{
+	struct vlan_ethhdr *vethh;
+	bool vlan_valid = false;
+	struct ipv6hdr *ip6h;
+	struct ethhdr *ethh;
+	struct tcphdr *tcph;
+	struct iphdr *iph;
+	int eth_hlen;
+	int ip_hlen;
+	int eth_type;
+	int i;
+
+	ethh = buf;
+	eth_type = ntohs(ethh->h_proto);
+	if (eth_type == ETH_P_8021Q) {
+		vlan_valid = true;
+		vethh = (struct vlan_ethhdr *)ethh;
+		cm_info->vlan = ntohs(vethh->h_vlan_TCI) & VLAN_VID_MASK;
+		eth_type = ntohs(vethh->h_vlan_encapsulated_proto);
+	}
+
+	eth_hlen = ETH_HLEN + (vlan_valid ? sizeof(u32) : 0);
+
+	memcpy(remote_mac_addr, ethh->h_source, ETH_ALEN);
+
+	memcpy(local_mac_addr, ethh->h_dest, ETH_ALEN);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_type =%d source mac: %pM\n",
+		   eth_type, ethh->h_source);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_hlen=%d destination mac: %pM\n",
+		   eth_hlen, ethh->h_dest);
+
+	iph = (struct iphdr *)((u8 *)(ethh) + eth_hlen);
+
+	if (eth_type == ETH_P_IP) {
+		cm_info->local_ip[0] = ntohl(iph->daddr);
+		cm_info->remote_ip[0] = ntohl(iph->saddr);
+		cm_info->ip_version = TCP_IPV4;
+
+		ip_hlen = (iph->ihl) * sizeof(u32);
+		*payload_len = ntohs(iph->tot_len) - ip_hlen;
+	} else if (eth_type == ETH_P_IPV6) {
+		ip6h = (struct ipv6hdr *)iph;
+		for (i = 0; i < 4; i++) {
+			cm_info->local_ip[i] =
+			    ntohl(ip6h->daddr.in6_u.u6_addr32[i]);
+			cm_info->remote_ip[i] =
+			    ntohl(ip6h->saddr.in6_u.u6_addr32[i]);
+		}
+		cm_info->ip_version = TCP_IPV6;
+
+		ip_hlen = sizeof(*ip6h);
+		*payload_len = ntohs(ip6h->payload_len);
+	} else {
+		DP_NOTICE(p_hwfn, "Unexpected ethertype on ll2 %x\n", eth_type);
+		return -EINVAL;
+	}
+
+	tcph = (struct tcphdr *)((u8 *)iph + ip_hlen);
+
+	if (!tcph->syn) {
+		DP_NOTICE(p_hwfn,
+			  "Only SYN type packet expected on this ll2 conn, iph->ihl=%d source=%d dest=%d\n",
+			  iph->ihl, tcph->source, tcph->dest);
+		return -EINVAL;
+	}
+
+	cm_info->local_port = ntohs(tcph->dest);
+	cm_info->remote_port = ntohs(tcph->source);
+
+	qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+	*tcp_start_offset = eth_hlen + ip_hlen;
+
+	return 0;
+}
+
 static void
 qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
 {
 	struct qed_iwarp_ll2_buff *buf = data->cookie;
+	struct qed_iwarp_listener *listener;
+	struct qed_ll2_tx_pkt_info tx_pkt;
+	struct qed_iwarp_cm_info cm_info;
 	struct qed_hwfn *p_hwfn = cxt;
+	u8 remote_mac_addr[ETH_ALEN];
+	u8 local_mac_addr[ETH_ALEN];
+	int tcp_start_offset;
+	u8 ll2_syn_handle;
+	int payload_len;
+	int rc;
+
+	memset(&cm_info, 0, sizeof(cm_info));
 
 	if (GET_FIELD(data->parse_flags,
 		      PARSING_AND_ERR_FLAGS_L4CHKSMWASCALCULATED) &&
@@ -510,11 +675,52 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 		goto err;
 	}
 
-	/* Process SYN packet - added later on in series */
+	rc = qed_iwarp_parse_rx_pkt(p_hwfn, &cm_info, (u8 *)(buf->data) +
+				    data->u.placement_offset, remote_mac_addr,
+				    local_mac_addr, &payload_len,
+				    &tcp_start_offset);
+	if (rc)
+		goto err;
 
+	/* Check if there is a listener for this 4-tuple+vlan */
+	ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle;
+	listener = qed_iwarp_get_listener(p_hwfn, &cm_info);
+	if (!listener) {
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "SYN received on tuple not listened on parse_flags=%d packet len=%d\n",
+			   data->parse_flags, data->length.packet_length);
+
+		memset(&tx_pkt, 0, sizeof(tx_pkt));
+		tx_pkt.num_of_bds = 1;
+		tx_pkt.vlan = data->vlan;
+
+		if (GET_FIELD(data->parse_flags,
+			      PARSING_AND_ERR_FLAGS_TAG8021QEXIST))
+			SET_FIELD(tx_pkt.bd_flags,
+				  CORE_TX_BD_DATA_VLAN_INSERTION, 1);
+
+		tx_pkt.l4_hdr_offset_w = (data->length.packet_length) >> 2;
+		tx_pkt.tx_dest = QED_LL2_TX_DEST_LB;
+		tx_pkt.first_frag = buf->data_phys_addr +
+				    data->u.placement_offset;
+		tx_pkt.first_frag_len = data->length.packet_length;
+		tx_pkt.cookie = buf;
+
+		rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_syn_handle,
+					       &tx_pkt, true);
+
+		if (rc) {
+			DP_NOTICE(p_hwfn,
+				  "Can't post SYN back to chip rc=%d\n", rc);
+			goto err;
+		}
+		return;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Received syn on listening port\n");
 err:
-	qed_iwarp_ll2_post_rx(p_hwfn, buf,
-			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+	qed_iwarp_ll2_post_rx(p_hwfn, buf, ll2_syn_handle);
 }
 
 static void qed_iwarp_ll2_rel_rx_pkt(void *cxt, u8 connection_handle,
@@ -700,6 +906,7 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
 
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.listen_list);
 
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
 				  qed_iwarp_async_event);
@@ -728,6 +935,62 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 	return 0;
 }
 
+int
+qed_iwarp_create_listen(void *rdma_cxt,
+			struct qed_iwarp_listen_in *iparams,
+			struct qed_iwarp_listen_out *oparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_listener *listener;
+
+	listener = kzalloc(sizeof(*listener), GFP_KERNEL);
+	if (!listener)
+		return -ENOMEM;
+
+	listener->ip_version = iparams->ip_version;
+	memcpy(listener->ip_addr, iparams->ip_addr, sizeof(listener->ip_addr));
+	listener->port = iparams->port;
+	listener->vlan = iparams->vlan;
+
+	listener->event_cb = iparams->event_cb;
+	listener->cb_context = iparams->cb_context;
+	listener->max_backlog = iparams->max_backlog;
+	oparams->handle = listener;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&listener->list_entry,
+		      &p_hwfn->p_rdma_info->iwarp.listen_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	DP_VERBOSE(p_hwfn,
+		   QED_MSG_RDMA,
+		   "callback=%p handle=%p ip=%x:%x:%x:%x port=0x%x vlan=0x%x\n",
+		   listener->event_cb,
+		   listener,
+		   listener->ip_addr[0],
+		   listener->ip_addr[1],
+		   listener->ip_addr[2],
+		   listener->ip_addr[3], listener->port, listener->vlan);
+
+	return 0;
+}
+
+int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle)
+{
+	struct qed_iwarp_listener *listener = handle;
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "handle=%p\n", handle);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_del(&listener->list_entry);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	kfree(listener);
+
+	return 0;
+}
+
 void
 qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 		   struct qed_rdma_query_qp_out_params *out_params)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 068b859..29005ac 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -54,6 +54,7 @@ struct qed_iwarp_ll2_buff {
 };
 
 struct qed_iwarp_info {
+	struct list_head listen_list;	/* qed_iwarp_listener */
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
 	u32 rcv_wnd_scale;
@@ -67,6 +68,21 @@ struct qed_iwarp_info {
 	enum mpa_rtr_type rtr_type;
 };
 
+struct qed_iwarp_listener {
+	struct list_head list_entry;
+
+	/* The event_cb function is called for connection requests.
+	 * The cb_context is passed to the event_cb function.
+	 */
+	iwarp_event_handler event_cb;
+	void *cb_context;
+	u32 max_backlog;
+	u32 ip_addr[4];
+	u16 port;
+	u16 vlan;
+	u8 ip_version;
+};
+
 int qed_iwarp_alloc(struct qed_hwfn *p_hwfn);
 
 int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
@@ -94,4 +110,11 @@ int qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp,
 void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_rdma_query_qp_out_params *out_params);
 
+int
+qed_iwarp_create_listen(void *rdma_cxt,
+			struct qed_iwarp_listen_in *iparams,
+			struct qed_iwarp_listen_out *oparams);
+
+int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
+
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index ee6887f..29de915 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -1772,6 +1772,8 @@ static int qed_roce_ll2_set_mac_filter(struct qed_dev *cdev,
 	.ll2_set_fragment_of_tx_packet = &qed_ll2_set_fragment_of_tx_packet,
 	.ll2_set_mac_filter = &qed_roce_ll2_set_mac_filter,
 	.ll2_get_stats = &qed_ll2_get_stats,
+	.iwarp_create_listen = &qed_iwarp_create_listen,
+	.iwarp_destroy_listen = &qed_iwarp_destroy_listen,
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void)
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 5b4bb09..28df568 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -470,6 +470,52 @@ struct qed_rdma_counters_out_params {
 #define QED_ROCE_TX_HEAD_FAILURE        (1)
 #define QED_ROCE_TX_FRAG_FAILURE        (2)
 
+enum qed_iwarp_event_type {
+	QED_IWARP_EVENT_MPA_REQUEST,	/* Passive side request received */
+};
+
+enum qed_tcp_ip_version {
+	QED_TCP_IPV4,
+	QED_TCP_IPV6,
+};
+
+struct qed_iwarp_cm_info {
+	enum qed_tcp_ip_version ip_version;
+	u32 remote_ip[4];
+	u32 local_ip[4];
+	u16 remote_port;
+	u16 local_port;
+	u16 vlan;
+	u8 ord;
+	u8 ird;
+	u16 private_data_len;
+	const void *private_data;
+};
+
+struct qed_iwarp_cm_event_params {
+	enum qed_iwarp_event_type event;
+	const struct qed_iwarp_cm_info *cm_info;
+	void *ep_context;	/* To be passed to accept call */
+	int status;
+};
+
+typedef int (*iwarp_event_handler) (void *context,
+				    struct qed_iwarp_cm_event_params *event);
+
+struct qed_iwarp_listen_in {
+	iwarp_event_handler event_cb;
+	void *cb_context;	/* passed to event_cb */
+	u32 max_backlog;
+	enum qed_tcp_ip_version ip_version;
+	u32 ip_addr[4];
+	u16 port;
+	u16 vlan;
+};
+
+struct qed_iwarp_listen_out {
+	void *handle;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -576,6 +622,12 @@ struct qed_rdma_ops {
 	int (*ll2_set_mac_filter)(struct qed_dev *cdev,
 				  u8 *old_mac_address, u8 *new_mac_address);
 
+	int (*iwarp_create_listen)(void *rdma_cxt,
+				   struct qed_iwarp_listen_in *iparams,
+				   struct qed_iwarp_listen_out *oparams);
+
+	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
+
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void);
-- 
1.8.3.1

[PATCH net-next 06/12] qed: iWARP CM add listener functions and initial SYN processing

[Michal Kalderon]

This patch adds the ability to add and remove listeners and identify
whether the SYN packet received is intended for iWARP or not. If
a listener is not found the SYN packet is posted back to the chip.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 269 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  23 +++
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |   2 +
 include/linux/qed/qed_rdma_if.h             |  52 ++++++
 4 files changed, 343 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index f3b4b32..2bab57c 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -31,6 +31,10 @@
  */
 #include <linux/if_ether.h>
 #include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/spinlock.h>
+#include <linux/tcp.h>
 #include "qed_cxt.h"
 #include "qed_hw.h"
 #include "qed_ll2.h"
@@ -477,6 +481,31 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
 }
 
+static void
+qed_iwarp_print_cm_info(struct qed_hwfn *p_hwfn,
+			struct qed_iwarp_cm_info *cm_info)
+{
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "ip_version = %d\n",
+		   cm_info->ip_version);
+
+	if (cm_info->ip_version == QED_TCP_IPV4)
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "remote_ip %pI4h:%x, local_ip %pI4h:%x vlan=%x\n",
+			   cm_info->remote_ip, cm_info->remote_port,
+			   cm_info->local_ip, cm_info->local_port,
+			   cm_info->vlan);
+	else
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "remote_ip %pI6h:%x, local_ip %pI6h:%x vlan=%x\n",
+			   cm_info->remote_ip, cm_info->remote_port,
+			   cm_info->local_ip, cm_info->local_port,
+			   cm_info->vlan);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "private_data_len = %x ord = %d, ird = %d\n",
+		   cm_info->private_data_len, cm_info->ord, cm_info->ird);
+}
+
 static int
 qed_iwarp_ll2_post_rx(struct qed_hwfn *p_hwfn,
 		      struct qed_iwarp_ll2_buff *buf, u8 handle)
@@ -497,11 +526,147 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	return rc;
 }
 
+static struct qed_iwarp_listener *
+qed_iwarp_get_listener(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_cm_info *cm_info)
+{
+	struct qed_iwarp_listener *listener = NULL;
+	static const u32 ip_zero[4] = { 0, 0, 0, 0 };
+	bool found = false;
+
+	qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+	list_for_each_entry(listener,
+			    &p_hwfn->p_rdma_info->iwarp.listen_list,
+			    list_entry) {
+		if (listener->port == cm_info->local_port) {
+			if (!memcmp(listener->ip_addr,
+				    ip_zero, sizeof(ip_zero))) {
+				found = true;
+				break;
+			}
+
+			if (!memcmp(listener->ip_addr,
+				    cm_info->local_ip,
+				    sizeof(cm_info->local_ip)) &&
+			    (listener->vlan == cm_info->vlan)) {
+				found = true;
+				break;
+			}
+		}
+	}
+
+	if (found) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener found = %p\n",
+			   listener);
+		return listener;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "listener not found\n");
+	return NULL;
+}
+
+static int
+qed_iwarp_parse_rx_pkt(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_cm_info *cm_info,
+		       void *buf,
+		       u8 *remote_mac_addr,
+		       u8 *local_mac_addr,
+		       int *payload_len, int *tcp_start_offset)
+{
+	struct vlan_ethhdr *vethh;
+	bool vlan_valid = false;
+	struct ipv6hdr *ip6h;
+	struct ethhdr *ethh;
+	struct tcphdr *tcph;
+	struct iphdr *iph;
+	int eth_hlen;
+	int ip_hlen;
+	int eth_type;
+	int i;
+
+	ethh = buf;
+	eth_type = ntohs(ethh->h_proto);
+	if (eth_type == ETH_P_8021Q) {
+		vlan_valid = true;
+		vethh = (struct vlan_ethhdr *)ethh;
+		cm_info->vlan = ntohs(vethh->h_vlan_TCI) & VLAN_VID_MASK;
+		eth_type = ntohs(vethh->h_vlan_encapsulated_proto);
+	}
+
+	eth_hlen = ETH_HLEN + (vlan_valid ? sizeof(u32) : 0);
+
+	memcpy(remote_mac_addr, ethh->h_source, ETH_ALEN);
+
+	memcpy(local_mac_addr, ethh->h_dest, ETH_ALEN);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_type =%d source mac: %pM\n",
+		   eth_type, ethh->h_source);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_hlen=%d destination mac: %pM\n",
+		   eth_hlen, ethh->h_dest);
+
+	iph = (struct iphdr *)((u8 *)(ethh) + eth_hlen);
+
+	if (eth_type == ETH_P_IP) {
+		cm_info->local_ip[0] = ntohl(iph->daddr);
+		cm_info->remote_ip[0] = ntohl(iph->saddr);
+		cm_info->ip_version = TCP_IPV4;
+
+		ip_hlen = (iph->ihl) * sizeof(u32);
+		*payload_len = ntohs(iph->tot_len) - ip_hlen;
+	} else if (eth_type == ETH_P_IPV6) {
+		ip6h = (struct ipv6hdr *)iph;
+		for (i = 0; i < 4; i++) {
+			cm_info->local_ip[i] =
+			    ntohl(ip6h->daddr.in6_u.u6_addr32[i]);
+			cm_info->remote_ip[i] =
+			    ntohl(ip6h->saddr.in6_u.u6_addr32[i]);
+		}
+		cm_info->ip_version = TCP_IPV6;
+
+		ip_hlen = sizeof(*ip6h);
+		*payload_len = ntohs(ip6h->payload_len);
+	} else {
+		DP_NOTICE(p_hwfn, "Unexpected ethertype on ll2 %x\n", eth_type);
+		return -EINVAL;
+	}
+
+	tcph = (struct tcphdr *)((u8 *)iph + ip_hlen);
+
+	if (!tcph->syn) {
+		DP_NOTICE(p_hwfn,
+			  "Only SYN type packet expected on this ll2 conn, iph->ihl=%d source=%d dest=%d\n",
+			  iph->ihl, tcph->source, tcph->dest);
+		return -EINVAL;
+	}
+
+	cm_info->local_port = ntohs(tcph->dest);
+	cm_info->remote_port = ntohs(tcph->source);
+
+	qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+	*tcp_start_offset = eth_hlen + ip_hlen;
+
+	return 0;
+}
+
 static void
 qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
 {
 	struct qed_iwarp_ll2_buff *buf = data->cookie;
+	struct qed_iwarp_listener *listener;
+	struct qed_ll2_tx_pkt_info tx_pkt;
+	struct qed_iwarp_cm_info cm_info;
 	struct qed_hwfn *p_hwfn = cxt;
+	u8 remote_mac_addr[ETH_ALEN];
+	u8 local_mac_addr[ETH_ALEN];
+	int tcp_start_offset;
+	u8 ll2_syn_handle;
+	int payload_len;
+	int rc;
+
+	memset(&cm_info, 0, sizeof(cm_info));
 
 	if (GET_FIELD(data->parse_flags,
 		      PARSING_AND_ERR_FLAGS_L4CHKSMWASCALCULATED) &&
@@ -510,11 +675,52 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 		goto err;
 	}
 
-	/* Process SYN packet - added later on in series */
+	rc = qed_iwarp_parse_rx_pkt(p_hwfn, &cm_info, (u8 *)(buf->data) +
+				    data->u.placement_offset, remote_mac_addr,
+				    local_mac_addr, &payload_len,
+				    &tcp_start_offset);
+	if (rc)
+		goto err;
 
+	/* Check if there is a listener for this 4-tuple+vlan */
+	ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle;
+	listener = qed_iwarp_get_listener(p_hwfn, &cm_info);
+	if (!listener) {
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "SYN received on tuple not listened on parse_flags=%d packet len=%d\n",
+			   data->parse_flags, data->length.packet_length);
+
+		memset(&tx_pkt, 0, sizeof(tx_pkt));
+		tx_pkt.num_of_bds = 1;
+		tx_pkt.vlan = data->vlan;
+
+		if (GET_FIELD(data->parse_flags,
+			      PARSING_AND_ERR_FLAGS_TAG8021QEXIST))
+			SET_FIELD(tx_pkt.bd_flags,
+				  CORE_TX_BD_DATA_VLAN_INSERTION, 1);
+
+		tx_pkt.l4_hdr_offset_w = (data->length.packet_length) >> 2;
+		tx_pkt.tx_dest = QED_LL2_TX_DEST_LB;
+		tx_pkt.first_frag = buf->data_phys_addr +
+				    data->u.placement_offset;
+		tx_pkt.first_frag_len = data->length.packet_length;
+		tx_pkt.cookie = buf;
+
+		rc = qed_ll2_prepare_tx_packet(p_hwfn, ll2_syn_handle,
+					       &tx_pkt, true);
+
+		if (rc) {
+			DP_NOTICE(p_hwfn,
+				  "Can't post SYN back to chip rc=%d\n", rc);
+			goto err;
+		}
+		return;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Received syn on listening port\n");
 err:
-	qed_iwarp_ll2_post_rx(p_hwfn, buf,
-			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+	qed_iwarp_ll2_post_rx(p_hwfn, buf, ll2_syn_handle);
 }
 
 static void qed_iwarp_ll2_rel_rx_pkt(void *cxt, u8 connection_handle,
@@ -700,6 +906,7 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
 
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.listen_list);
 
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
 				  qed_iwarp_async_event);
@@ -728,6 +935,62 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 	return 0;
 }
 
+int
+qed_iwarp_create_listen(void *rdma_cxt,
+			struct qed_iwarp_listen_in *iparams,
+			struct qed_iwarp_listen_out *oparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_listener *listener;
+
+	listener = kzalloc(sizeof(*listener), GFP_KERNEL);
+	if (!listener)
+		return -ENOMEM;
+
+	listener->ip_version = iparams->ip_version;
+	memcpy(listener->ip_addr, iparams->ip_addr, sizeof(listener->ip_addr));
+	listener->port = iparams->port;
+	listener->vlan = iparams->vlan;
+
+	listener->event_cb = iparams->event_cb;
+	listener->cb_context = iparams->cb_context;
+	listener->max_backlog = iparams->max_backlog;
+	oparams->handle = listener;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&listener->list_entry,
+		      &p_hwfn->p_rdma_info->iwarp.listen_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	DP_VERBOSE(p_hwfn,
+		   QED_MSG_RDMA,
+		   "callback=%p handle=%p ip=%x:%x:%x:%x port=0x%x vlan=0x%x\n",
+		   listener->event_cb,
+		   listener,
+		   listener->ip_addr[0],
+		   listener->ip_addr[1],
+		   listener->ip_addr[2],
+		   listener->ip_addr[3], listener->port, listener->vlan);
+
+	return 0;
+}
+
+int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle)
+{
+	struct qed_iwarp_listener *listener = handle;
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "handle=%p\n", handle);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_del(&listener->list_entry);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	kfree(listener);
+
+	return 0;
+}
+
 void
 qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 		   struct qed_rdma_query_qp_out_params *out_params)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 068b859..29005ac 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -54,6 +54,7 @@ struct qed_iwarp_ll2_buff {
 };
 
 struct qed_iwarp_info {
+	struct list_head listen_list;	/* qed_iwarp_listener */
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
 	u32 rcv_wnd_scale;
@@ -67,6 +68,21 @@ struct qed_iwarp_info {
 	enum mpa_rtr_type rtr_type;
 };
 
+struct qed_iwarp_listener {
+	struct list_head list_entry;
+
+	/* The event_cb function is called for connection requests.
+	 * The cb_context is passed to the event_cb function.
+	 */
+	iwarp_event_handler event_cb;
+	void *cb_context;
+	u32 max_backlog;
+	u32 ip_addr[4];
+	u16 port;
+	u16 vlan;
+	u8 ip_version;
+};
+
 int qed_iwarp_alloc(struct qed_hwfn *p_hwfn);
 
 int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
@@ -94,4 +110,11 @@ int qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp,
 void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_rdma_query_qp_out_params *out_params);
 
+int
+qed_iwarp_create_listen(void *rdma_cxt,
+			struct qed_iwarp_listen_in *iparams,
+			struct qed_iwarp_listen_out *oparams);
+
+int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
+
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index ee6887f..29de915 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -1772,6 +1772,8 @@ static int qed_roce_ll2_set_mac_filter(struct qed_dev *cdev,
 	.ll2_set_fragment_of_tx_packet = &qed_ll2_set_fragment_of_tx_packet,
 	.ll2_set_mac_filter = &qed_roce_ll2_set_mac_filter,
 	.ll2_get_stats = &qed_ll2_get_stats,
+	.iwarp_create_listen = &qed_iwarp_create_listen,
+	.iwarp_destroy_listen = &qed_iwarp_destroy_listen,
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void)
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 5b4bb09..28df568 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -470,6 +470,52 @@ struct qed_rdma_counters_out_params {
 #define QED_ROCE_TX_HEAD_FAILURE        (1)
 #define QED_ROCE_TX_FRAG_FAILURE        (2)
 
+enum qed_iwarp_event_type {
+	QED_IWARP_EVENT_MPA_REQUEST,	/* Passive side request received */
+};
+
+enum qed_tcp_ip_version {
+	QED_TCP_IPV4,
+	QED_TCP_IPV6,
+};
+
+struct qed_iwarp_cm_info {
+	enum qed_tcp_ip_version ip_version;
+	u32 remote_ip[4];
+	u32 local_ip[4];
+	u16 remote_port;
+	u16 local_port;
+	u16 vlan;
+	u8 ord;
+	u8 ird;
+	u16 private_data_len;
+	const void *private_data;
+};
+
+struct qed_iwarp_cm_event_params {
+	enum qed_iwarp_event_type event;
+	const struct qed_iwarp_cm_info *cm_info;
+	void *ep_context;	/* To be passed to accept call */
+	int status;
+};
+
+typedef int (*iwarp_event_handler) (void *context,
+				    struct qed_iwarp_cm_event_params *event);
+
+struct qed_iwarp_listen_in {
+	iwarp_event_handler event_cb;
+	void *cb_context;	/* passed to event_cb */
+	u32 max_backlog;
+	enum qed_tcp_ip_version ip_version;
+	u32 ip_addr[4];
+	u16 port;
+	u16 vlan;
+};
+
+struct qed_iwarp_listen_out {
+	void *handle;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -576,6 +622,12 @@ struct qed_rdma_ops {
 	int (*ll2_set_mac_filter)(struct qed_dev *cdev,
 				  u8 *old_mac_address, u8 *new_mac_address);
 
+	int (*iwarp_create_listen)(void *rdma_cxt,
+				   struct qed_iwarp_listen_in *iparams,
+				   struct qed_iwarp_listen_out *oparams);
+
+	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
+
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void);
-- 
1.8.3.1

[PATCH net-next 07/12] qed: iWARP CM add passive side connect

[Michal Kalderon]

This patch implements the passive side connect.
It addresses pre-allocating resources, creating a connection
element upon valid SYN packet received. Calling upper layer and
implementation of the accept/reject calls.

Error handling is not part of this patch.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed.h       |   2 +
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |  11 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 939 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  62 ++
 drivers/net/ethernet/qlogic/qed/qed_l2.c    |  13 -
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |   2 +
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |   2 +
 include/linux/qed/common_hsi.h              |   2 +
 include/linux/qed/qed_rdma_if.h             |  26 +-
 9 files changed, 1039 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index fd8cd5e..91003bc 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -789,6 +789,8 @@ void qed_configure_vp_wfq_on_link_change(struct qed_dev *cdev,
 void qed_clean_wfq_db(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
 int qed_device_num_engines(struct qed_dev *cdev);
 int qed_device_get_port_id(struct qed_dev *cdev);
+void qed_set_fw_mac_addr(__le16 *fw_msb,
+			 __le16 *fw_mid, __le16 *fw_lsb, u8 *mac);
 
 #define QED_LEADING_HWFN(dev)   (&dev->hwfns[0])
 
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 6c8505d..4060a6a 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -4127,3 +4127,14 @@ int qed_device_get_port_id(struct qed_dev *cdev)
 {
 	return (QED_LEADING_HWFN(cdev)->abs_pf_id) % qed_device_num_ports(cdev);
 }
+
+void qed_set_fw_mac_addr(__le16 *fw_msb,
+			 __le16 *fw_mid, __le16 *fw_lsb, u8 *mac)
+{
+	((u8 *)fw_msb)[0] = mac[1];
+	((u8 *)fw_msb)[1] = mac[0];
+	((u8 *)fw_mid)[0] = mac[3];
+	((u8 *)fw_mid)[1] = mac[2];
+	((u8 *)fw_lsb)[0] = mac[5];
+	((u8 *)fw_lsb)[1] = mac[4];
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index 2bab57c..a6dadae 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -44,9 +44,30 @@
 
 #define QED_IWARP_ORD_DEFAULT		32
 #define QED_IWARP_IRD_DEFAULT		32
+#define QED_IWARP_MAX_FW_MSS		4120
+
+#define QED_EP_SIG 0xecabcdef
+
+struct mpa_v2_hdr {
+	__be16 ird;
+	__be16 ord;
+};
+
+#define MPA_V2_PEER2PEER_MODEL  0x8000
+#define MPA_V2_SEND_RTR         0x4000	/* on ird */
+#define MPA_V2_READ_RTR         0x4000	/* on ord */
+#define MPA_V2_WRITE_RTR        0x8000
+#define MPA_V2_IRD_ORD_MASK     0x3FFF
+
+#define MPA_REV2(_mpa_rev) ((_mpa_rev) == MPA_NEGOTIATION_TYPE_ENHANCED)
+
+#define QED_IWARP_INVALID_TCP_CID	0xffffffff
 #define QED_IWARP_RCV_WND_SIZE_DEF	(256 * 1024)
 #define QED_IWARP_RCV_WND_SIZE_MIN	(64 * 1024)
+#define TIMESTAMP_HEADER_SIZE		(12)
+
 #define QED_IWARP_TS_EN			BIT(0)
+#define QED_IWARP_DA_EN			BIT(1)
 #define QED_IWARP_PARAM_CRC_NEEDED	(1)
 #define QED_IWARP_PARAM_P2P		(1)
 
@@ -63,7 +84,8 @@ void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn)
 	dev->max_inline = IWARP_REQ_MAX_INLINE_DATA_SIZE;
 	dev->max_qp = min_t(u32,
 			    IWARP_MAX_QPS,
-			    p_hwfn->p_rdma_info->num_qps);
+			    p_hwfn->p_rdma_info->num_qps) -
+		      QED_IWARP_PREALLOC_CNT;
 
 	dev->max_cq = dev->max_qp;
 
@@ -78,12 +100,22 @@ void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	p_hwfn->b_rdma_enabled_in_prs = true;
 }
 
+/* We have two cid maps, one for tcp which should be used only from passive
+ * syn processing and replacing a pre-allocated ep in the list. The second
+ * for active tcp and for QPs.
+ */
 static void qed_iwarp_cid_cleaned(struct qed_hwfn *p_hwfn, u32 cid)
 {
 	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
 
 	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
-	qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+
+	if (cid < QED_IWARP_PREALLOC_CNT)
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map,
+				    cid);
+	else
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+
 	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
 }
 
@@ -107,6 +139,45 @@ static int qed_iwarp_alloc_cid(struct qed_hwfn *p_hwfn, u32 *cid)
 	return rc;
 }
 
+static void qed_iwarp_set_tcp_cid(struct qed_hwfn *p_hwfn, u32 cid)
+{
+	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	qed_bmap_set_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+}
+
+/* This function allocates a cid for passive tcp (called from syn receive)
+ * the reason it's separate from the regular cid allocation is because it
+ * is assured that these cids already have ilt allocated. They are preallocated
+ * to ensure that we won't need to allocate memory during syn processing
+ */
+static int qed_iwarp_alloc_tcp_cid(struct qed_hwfn *p_hwfn, u32 *cid)
+{
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+
+	rc = qed_rdma_bmap_alloc_id(p_hwfn,
+				    &p_hwfn->p_rdma_info->tcp_cid_map, cid);
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+
+	if (rc) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "can't allocate iwarp tcp cid max-count=%d\n",
+			   p_hwfn->p_rdma_info->tcp_cid_map.max_count);
+
+		*cid = QED_IWARP_INVALID_TCP_CID;
+		return rc;
+	}
+
+	*cid += qed_cxt_get_proto_cid_start(p_hwfn,
+					    p_hwfn->p_rdma_info->proto);
+	return 0;
+}
+
 int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
 			struct qed_rdma_qp *qp,
 			struct qed_rdma_create_qp_out_params *out_params)
@@ -403,6 +474,26 @@ int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	return rc;
 }
 
+static void qed_iwarp_destroy_ep(struct qed_hwfn *p_hwfn,
+				 struct qed_iwarp_ep *ep,
+				 bool remove_from_active_list)
+{
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+			  sizeof(*ep->ep_buffer_virt),
+			  ep->ep_buffer_virt, ep->ep_buffer_phys);
+
+	if (remove_from_active_list) {
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_del(&ep->list_entry);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+
+	if (ep->qp)
+		ep->qp->ep = NULL;
+
+	kfree(ep);
+}
+
 int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
 	int rc = 0;
@@ -424,6 +515,507 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	return rc;
 }
 
+static int
+qed_iwarp_create_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep **ep_out)
+{
+	struct qed_iwarp_ep *ep;
+	int rc;
+
+	ep = kzalloc(sizeof(*ep), GFP_KERNEL);
+	if (!ep)
+		return -ENOMEM;
+
+	ep->state = QED_IWARP_EP_INIT;
+
+	ep->ep_buffer_virt = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+						sizeof(*ep->ep_buffer_virt),
+						&ep->ep_buffer_phys,
+						GFP_KERNEL);
+	if (!ep->ep_buffer_virt) {
+		rc = -ENOMEM;
+		goto err;
+	}
+
+	ep->sig = QED_EP_SIG;
+
+	*ep_out = ep;
+
+	return 0;
+
+err:
+	kfree(ep);
+	return rc;
+}
+
+static void
+qed_iwarp_print_tcp_ramrod(struct qed_hwfn *p_hwfn,
+			   struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod)
+{
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "local_mac=%x %x %x, remote_mac=%x %x %x\n",
+		   p_tcp_ramrod->tcp.local_mac_addr_lo,
+		   p_tcp_ramrod->tcp.local_mac_addr_mid,
+		   p_tcp_ramrod->tcp.local_mac_addr_hi,
+		   p_tcp_ramrod->tcp.remote_mac_addr_lo,
+		   p_tcp_ramrod->tcp.remote_mac_addr_mid,
+		   p_tcp_ramrod->tcp.remote_mac_addr_hi);
+
+	if (p_tcp_ramrod->tcp.ip_version == TCP_IPV4) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "local_ip=%pI4h:%x, remote_ip=%pI4h%x, vlan=%x\n",
+			   p_tcp_ramrod->tcp.local_ip,
+			   p_tcp_ramrod->tcp.local_port,
+			   p_tcp_ramrod->tcp.remote_ip,
+			   p_tcp_ramrod->tcp.remote_port,
+			   p_tcp_ramrod->tcp.vlan_id);
+	} else {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "local_ip=%pI6h:%x, remote_ip=%pI6h:%x, vlan=%x\n",
+			   p_tcp_ramrod->tcp.local_ip,
+			   p_tcp_ramrod->tcp.local_port,
+			   p_tcp_ramrod->tcp.remote_ip,
+			   p_tcp_ramrod->tcp.remote_port,
+			   p_tcp_ramrod->tcp.vlan_id);
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "flow_label=%x, ttl=%x, tos_or_tc=%x, mss=%x, rcv_wnd_scale=%x, connect_mode=%x, flags=%x\n",
+		   p_tcp_ramrod->tcp.flow_label,
+		   p_tcp_ramrod->tcp.ttl,
+		   p_tcp_ramrod->tcp.tos_or_tc,
+		   p_tcp_ramrod->tcp.mss,
+		   p_tcp_ramrod->tcp.rcv_wnd_scale,
+		   p_tcp_ramrod->tcp.connect_mode,
+		   p_tcp_ramrod->tcp.flags);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "syn_ip_payload_length=%x, lo=%x, hi=%x\n",
+		   p_tcp_ramrod->tcp.syn_ip_payload_length,
+		   p_tcp_ramrod->tcp.syn_phy_addr_lo,
+		   p_tcp_ramrod->tcp.syn_phy_addr_hi);
+}
+
+static int
+qed_iwarp_tcp_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod;
+	struct tcp_offload_params_opt2 *tcp;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	dma_addr_t async_output_phys;
+	dma_addr_t in_pdata_phys;
+	u16 physical_q;
+	u8 tcp_flags;
+	int rc;
+	int i;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = ep->tcp_cid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_CB;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_TCP_OFFLOAD,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		return rc;
+
+	p_tcp_ramrod = &p_ent->ramrod.iwarp_tcp_offload;
+
+	in_pdata_phys = ep->ep_buffer_phys +
+			offsetof(struct qed_iwarp_ep_memory, in_pdata);
+	DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.incoming_ulp_buffer.addr,
+		       in_pdata_phys);
+
+	p_tcp_ramrod->iwarp.incoming_ulp_buffer.len =
+	    cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata));
+
+	async_output_phys = ep->ep_buffer_phys +
+			    offsetof(struct qed_iwarp_ep_memory, async_output);
+	DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.async_eqe_output_buf,
+		       async_output_phys);
+
+	p_tcp_ramrod->iwarp.handle_for_async.hi = cpu_to_le32(PTR_HI(ep));
+	p_tcp_ramrod->iwarp.handle_for_async.lo = cpu_to_le32(PTR_LO(ep));
+
+	physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD);
+	p_tcp_ramrod->iwarp.physical_q0 = cpu_to_le16(physical_q);
+	physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK);
+	p_tcp_ramrod->iwarp.physical_q1 = cpu_to_le16(physical_q);
+	p_tcp_ramrod->iwarp.mpa_mode = iwarp_info->mpa_rev;
+
+	tcp = &p_tcp_ramrod->tcp;
+	qed_set_fw_mac_addr(&tcp->remote_mac_addr_hi,
+			    &tcp->remote_mac_addr_mid,
+			    &tcp->remote_mac_addr_lo, ep->remote_mac_addr);
+	qed_set_fw_mac_addr(&tcp->local_mac_addr_hi, &tcp->local_mac_addr_mid,
+			    &tcp->local_mac_addr_lo, ep->local_mac_addr);
+
+	tcp->vlan_id = cpu_to_le16(ep->cm_info.vlan);
+
+	tcp_flags = p_hwfn->p_rdma_info->iwarp.tcp_flags;
+	tcp->flags = 0;
+	SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_TS_EN,
+		  !!(tcp_flags & QED_IWARP_TS_EN));
+
+	SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_DA_EN,
+		  !!(tcp_flags & QED_IWARP_DA_EN));
+
+	tcp->ip_version = ep->cm_info.ip_version;
+
+	for (i = 0; i < 4; i++) {
+		tcp->remote_ip[i] = cpu_to_le32(ep->cm_info.remote_ip[i]);
+		tcp->local_ip[i] = cpu_to_le32(ep->cm_info.local_ip[i]);
+	}
+
+	tcp->remote_port = cpu_to_le16(ep->cm_info.remote_port);
+	tcp->local_port = cpu_to_le16(ep->cm_info.local_port);
+	tcp->mss = cpu_to_le16(ep->mss);
+	tcp->flow_label = 0;
+	tcp->ttl = 0x40;
+	tcp->tos_or_tc = 0;
+
+	tcp->rcv_wnd_scale = (u8)p_hwfn->p_rdma_info->iwarp.rcv_wnd_scale;
+	tcp->connect_mode = ep->connect_mode;
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		tcp->syn_ip_payload_length =
+			cpu_to_le16(ep->syn_ip_payload_length);
+		tcp->syn_phy_addr_hi = DMA_HI_LE(ep->syn_phy_addr);
+		tcp->syn_phy_addr_lo = DMA_LO_LE(ep->syn_phy_addr);
+	}
+
+	qed_iwarp_print_tcp_ramrod(p_hwfn, p_tcp_ramrod);
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "EP(0x%x) Offload completed rc=%d\n", ep->tcp_cid, rc);
+
+	return rc;
+}
+
+static void
+qed_iwarp_mpa_received(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	struct qed_iwarp_cm_event_params params;
+	struct mpa_v2_hdr *mpa_v2;
+	union async_output *async_data;
+	u16 mpa_ord, mpa_ird;
+	u8 mpa_hdr_size = 0;
+	u8 mpa_rev;
+
+	async_data = &ep->ep_buffer_virt->async_output;
+
+	mpa_rev = async_data->mpa_request.mpa_handshake_mode;
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "private_data_len=%x handshake_mode=%x private_data=(%x)\n",
+		   async_data->mpa_request.ulp_data_len,
+		   mpa_rev, *((u32 *)((u8 *)ep->ep_buffer_virt->in_pdata)));
+
+	if (mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
+		/* Read ord/ird values from private data buffer */
+		mpa_v2 = (struct mpa_v2_hdr *)ep->ep_buffer_virt->in_pdata;
+		mpa_hdr_size = sizeof(*mpa_v2);
+
+		mpa_ord = ntohs(mpa_v2->ord);
+		mpa_ird = ntohs(mpa_v2->ird);
+
+		/* Temprary store in cm_info incoming ord/ird requested, later
+		 * replace with negotiated value during accept
+		 */
+		ep->cm_info.ord = (u8)min_t(u16,
+					    (mpa_ord & MPA_V2_IRD_ORD_MASK),
+					    QED_IWARP_ORD_DEFAULT);
+
+		ep->cm_info.ird = (u8)min_t(u16,
+					    (mpa_ird & MPA_V2_IRD_ORD_MASK),
+					    QED_IWARP_IRD_DEFAULT);
+
+		/* Peer2Peer negotiation */
+		ep->rtr_type = MPA_RTR_TYPE_NONE;
+		if (mpa_ird & MPA_V2_PEER2PEER_MODEL) {
+			if (mpa_ord & MPA_V2_WRITE_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_WRITE;
+
+			if (mpa_ord & MPA_V2_READ_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_READ;
+
+			if (mpa_ird & MPA_V2_SEND_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_SEND;
+
+			ep->rtr_type &= iwarp_info->rtr_type;
+
+			/* if we're left with no match send our capabilities */
+			if (ep->rtr_type == MPA_RTR_TYPE_NONE)
+				ep->rtr_type = iwarp_info->rtr_type;
+		}
+
+		ep->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED;
+	} else {
+		ep->cm_info.ord = QED_IWARP_ORD_DEFAULT;
+		ep->cm_info.ird = QED_IWARP_IRD_DEFAULT;
+		ep->mpa_rev = MPA_NEGOTIATION_TYPE_BASIC;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x rtr:0x%x ulp_data_len = %x mpa_hdr_size = %x\n",
+		   mpa_rev, ep->cm_info.ord, ep->cm_info.ird, ep->rtr_type,
+		   async_data->mpa_request.ulp_data_len, mpa_hdr_size);
+
+	/* Strip mpa v2 hdr from private data before sending to upper layer */
+	ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_hdr_size;
+
+	ep->cm_info.private_data_len = async_data->mpa_request.ulp_data_len -
+				       mpa_hdr_size;
+
+	params.event = QED_IWARP_EVENT_MPA_REQUEST;
+	params.cm_info = &ep->cm_info;
+	params.ep_context = ep;
+	params.status = 0;
+
+	ep->state = QED_IWARP_EP_MPA_REQ_RCVD;
+	ep->event_cb(ep->cb_context, &params);
+}
+
+static int
+qed_iwarp_mpa_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct iwarp_mpa_offload_ramrod_data *p_mpa_ramrod;
+	struct qed_sp_init_data init_data;
+	dma_addr_t async_output_phys;
+	struct qed_spq_entry *p_ent;
+	dma_addr_t out_pdata_phys;
+	dma_addr_t in_pdata_phys;
+	struct qed_rdma_qp *qp;
+	bool reject;
+	int rc;
+
+	if (!ep)
+		return -EINVAL;
+
+	qp = ep->qp;
+	reject = !qp;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = reject ? ep->tcp_cid : qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		return rc;
+
+	p_mpa_ramrod = &p_ent->ramrod.iwarp_mpa_offload;
+	out_pdata_phys = ep->ep_buffer_phys +
+			 offsetof(struct qed_iwarp_ep_memory, out_pdata);
+	DMA_REGPAIR_LE(p_mpa_ramrod->common.outgoing_ulp_buffer.addr,
+		       out_pdata_phys);
+	p_mpa_ramrod->common.outgoing_ulp_buffer.len =
+	    ep->cm_info.private_data_len;
+	p_mpa_ramrod->common.crc_needed = p_hwfn->p_rdma_info->iwarp.crc_needed;
+
+	p_mpa_ramrod->common.out_rq.ord = ep->cm_info.ord;
+	p_mpa_ramrod->common.out_rq.ird = ep->cm_info.ird;
+
+	p_mpa_ramrod->tcp_cid = p_hwfn->hw_info.opaque_fid << 16 | ep->tcp_cid;
+
+	in_pdata_phys = ep->ep_buffer_phys +
+			offsetof(struct qed_iwarp_ep_memory, in_pdata);
+	p_mpa_ramrod->tcp_connect_side = ep->connect_mode;
+	DMA_REGPAIR_LE(p_mpa_ramrod->incoming_ulp_buffer.addr,
+		       in_pdata_phys);
+	p_mpa_ramrod->incoming_ulp_buffer.len =
+	    cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata));
+	async_output_phys = ep->ep_buffer_phys +
+			    offsetof(struct qed_iwarp_ep_memory, async_output);
+	DMA_REGPAIR_LE(p_mpa_ramrod->async_eqe_output_buf,
+		       async_output_phys);
+	p_mpa_ramrod->handle_for_async.hi = cpu_to_le32(PTR_HI(ep));
+	p_mpa_ramrod->handle_for_async.lo = cpu_to_le32(PTR_LO(ep));
+
+	if (!reject) {
+		DMA_REGPAIR_LE(p_mpa_ramrod->shared_queue_addr,
+			       qp->shared_queue_phys_addr);
+		p_mpa_ramrod->stats_counter_id =
+		    RESC_START(p_hwfn, QED_RDMA_STATS_QUEUE) + qp->stats_queue;
+	} else {
+		p_mpa_ramrod->common.reject = 1;
+	}
+
+	p_mpa_ramrod->mode = ep->mpa_rev;
+	SET_FIELD(p_mpa_ramrod->rtr_pref,
+		  IWARP_MPA_OFFLOAD_RAMROD_DATA_RTR_SUPPORTED, ep->rtr_type);
+
+	ep->state = QED_IWARP_EP_MPA_OFFLOADED;
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+	if (!reject)
+		ep->cid = qp->icid;	/* Now they're migrated. */
+
+	DP_VERBOSE(p_hwfn,
+		   QED_MSG_RDMA,
+		   "QP(0x%x) EP(0x%x) MPA Offload rc = %d IRD=0x%x ORD=0x%x rtr_type=%d mpa_rev=%d reject=%d\n",
+		   reject ? 0xffff : qp->icid,
+		   ep->tcp_cid,
+		   rc,
+		   ep->cm_info.ird,
+		   ep->cm_info.ord, ep->rtr_type, ep->mpa_rev, reject);
+	return rc;
+}
+
+static void
+qed_iwarp_return_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	ep->state = QED_IWARP_EP_INIT;
+	if (ep->qp)
+		ep->qp->ep = NULL;
+	ep->qp = NULL;
+	memset(&ep->cm_info, 0, sizeof(ep->cm_info));
+
+	if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) {
+		/* We don't care about the return code, it's ok if tcp_cid
+		 * remains invalid...in this case we'll defer allocation
+		 */
+		qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid);
+	}
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	list_del(&ep->list_entry);
+	list_add_tail(&ep->list_entry,
+		      &p_hwfn->p_rdma_info->iwarp.ep_free_list);
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+}
+
+#define QED_IWARP_CONNECT_MODE_STRING(ep) \
+	((ep)->connect_mode == TCP_CONNECT_PASSIVE) ? "Passive" : "Active"
+
+/* Called as a result of the event:
+ * IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE
+ */
+static void
+qed_iwarp_mpa_complete(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
+		   ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird);
+
+	params.cm_info = &ep->cm_info;
+
+	params.ep_context = ep;
+
+	ep->state = QED_IWARP_EP_CLOSED;
+
+	switch (fw_return_code) {
+	case RDMA_RETURN_OK:
+		ep->qp->max_rd_atomic_req = ep->cm_info.ord;
+		ep->qp->max_rd_atomic_resp = ep->cm_info.ird;
+		qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_RTS, 1);
+		ep->state = QED_IWARP_EP_ESTABLISHED;
+		params.status = 0;
+		break;
+	default:
+		params.status = -ECONNRESET;
+		break;
+	}
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
+static void
+qed_iwarp_mpa_v2_set_private(struct qed_hwfn *p_hwfn,
+			     struct qed_iwarp_ep *ep, u8 *mpa_data_size)
+{
+	struct mpa_v2_hdr *mpa_v2_params;
+	u16 mpa_ird, mpa_ord;
+
+	*mpa_data_size = 0;
+	if (MPA_REV2(ep->mpa_rev)) {
+		mpa_v2_params =
+		    (struct mpa_v2_hdr *)ep->ep_buffer_virt->out_pdata;
+		*mpa_data_size = sizeof(*mpa_v2_params);
+
+		mpa_ird = (u16)ep->cm_info.ird;
+		mpa_ord = (u16)ep->cm_info.ord;
+
+		if (ep->rtr_type != MPA_RTR_TYPE_NONE) {
+			mpa_ird |= MPA_V2_PEER2PEER_MODEL;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_SEND)
+				mpa_ird |= MPA_V2_SEND_RTR;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_WRITE)
+				mpa_ord |= MPA_V2_WRITE_RTR;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_READ)
+				mpa_ord |= MPA_V2_READ_RTR;
+		}
+
+		mpa_v2_params->ird = htons(mpa_ird);
+		mpa_v2_params->ord = htons(mpa_ord);
+
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "MPA_NEGOTIATE Header: [%x ord:%x ird] %x ord:%x ird:%x peer2peer:%x rtr_send:%x rtr_write:%x rtr_read:%x\n",
+			   mpa_v2_params->ird,
+			   mpa_v2_params->ord,
+			   *((u32 *)mpa_v2_params),
+			   mpa_ord & MPA_V2_IRD_ORD_MASK,
+			   mpa_ird & MPA_V2_IRD_ORD_MASK,
+			   !!(mpa_ird & MPA_V2_PEER2PEER_MODEL),
+			   !!(mpa_ird & MPA_V2_SEND_RTR),
+			   !!(mpa_ord & MPA_V2_WRITE_RTR),
+			   !!(mpa_ord & MPA_V2_READ_RTR));
+	}
+}
+
+static struct qed_iwarp_ep *qed_iwarp_get_free_ep(struct qed_hwfn *p_hwfn)
+{
+	struct qed_iwarp_ep *ep = NULL;
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	if (list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) {
+		DP_ERR(p_hwfn, "Ep list is empty\n");
+		goto out;
+	}
+
+	ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list,
+			      struct qed_iwarp_ep, list_entry);
+
+	/* in some cases we could have failed allocating a tcp cid when added
+	 * from accept / failure... retry now..this is not the common case.
+	 */
+	if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) {
+		rc = qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid);
+
+		/* if we fail we could look for another entry with a valid
+		 * tcp_cid, but since we don't expect to reach this anyway
+		 * it's not worth the handling
+		 */
+		if (rc) {
+			ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+			ep = NULL;
+			goto out;
+		}
+	}
+
+	list_del(&ep->list_entry);
+
+out:
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	return ep;
+}
+
 #define QED_IWARP_MAX_CID_CLEAN_TIME  100
 #define QED_IWARP_MAX_NO_PROGRESS_CNT 5
 
@@ -465,20 +1057,213 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 
 static int qed_iwarp_wait_for_all_cids(struct qed_hwfn *p_hwfn)
 {
+	int rc;
+	int i;
+
+	rc = qed_iwarp_wait_cid_map_cleared(p_hwfn,
+					    &p_hwfn->p_rdma_info->tcp_cid_map);
+	if (rc)
+		return rc;
+
+	/* Now free the tcp cids from the main cid map */
+	for (i = 0; i < QED_IWARP_PREALLOC_CNT; i++)
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, i);
+
 	/* Now wait for all cids to be completed */
 	return qed_iwarp_wait_cid_map_cleared(p_hwfn,
 					      &p_hwfn->p_rdma_info->cid_map);
 }
 
+static void qed_iwarp_free_prealloc_ep(struct qed_hwfn *p_hwfn)
+{
+	struct qed_iwarp_ep *ep;
+
+	while (!list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) {
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+		ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list,
+				      struct qed_iwarp_ep, list_entry);
+
+		if (!ep) {
+			spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+			break;
+		}
+		list_del(&ep->list_entry);
+
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+		if (ep->tcp_cid != QED_IWARP_INVALID_TCP_CID)
+			qed_iwarp_cid_cleaned(p_hwfn, ep->tcp_cid);
+
+		qed_iwarp_destroy_ep(p_hwfn, ep, false);
+	}
+}
+
+static int qed_iwarp_prealloc_ep(struct qed_hwfn *p_hwfn, bool init)
+{
+	struct qed_iwarp_ep *ep;
+	int rc = 0;
+	int count;
+	u32 cid;
+	int i;
+
+	count = init ? QED_IWARP_PREALLOC_CNT : 1;
+	for (i = 0; i < count; i++) {
+		rc = qed_iwarp_create_ep(p_hwfn, &ep);
+		if (rc)
+			return rc;
+
+		/* During initialization we allocate from the main pool,
+		 * afterwards we allocate only from the tcp_cid.
+		 */
+		if (init) {
+			rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+			if (rc)
+				goto err;
+			qed_iwarp_set_tcp_cid(p_hwfn, cid);
+		} else {
+			/* We don't care about the return code, it's ok if
+			 * tcp_cid remains invalid...in this case we'll
+			 * defer allocation
+			 */
+			qed_iwarp_alloc_tcp_cid(p_hwfn, &cid);
+		}
+
+		ep->tcp_cid = cid;
+
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_add_tail(&ep->list_entry,
+			      &p_hwfn->p_rdma_info->iwarp.ep_free_list);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+
+	return rc;
+
+err:
+	qed_iwarp_destroy_ep(p_hwfn, ep, false);
+
+	return rc;
+}
+
 int qed_iwarp_alloc(struct qed_hwfn *p_hwfn)
 {
+	int rc;
+
+	/* Allocate bitmap for tcp cid. These are used by passive side
+	 * to ensure it can allocate a tcp cid during dpc that was
+	 * pre-acquired and doesn't require dynamic allocation of ilt
+	 */
+	rc = qed_rdma_bmap_alloc(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map,
+				 QED_IWARP_PREALLOC_CNT, "TCP_CID");
+	if (rc) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "Failed to allocate tcp cid, rc = %d\n", rc);
+		return rc;
+	}
+
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_free_list);
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.iw_lock);
 
-	return 0;
+	return qed_iwarp_prealloc_ep(p_hwfn, true);
 }
 
 void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
+	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, 1);
+}
+
+int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+	int rc;
+
+	ep = (struct qed_iwarp_ep *)iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in accept is NULL\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n",
+		   iparams->qp->icid, ep->tcp_cid);
+
+	if ((iparams->ord > QED_IWARP_ORD_DEFAULT) ||
+	    (iparams->ird > QED_IWARP_IRD_DEFAULT)) {
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) EP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n",
+			   iparams->qp->icid,
+			   ep->tcp_cid, iparams->ord, iparams->ord);
+		return -EINVAL;
+	}
+
+	qed_iwarp_prealloc_ep(p_hwfn, false);
+
+	ep->cb_context = iparams->cb_context;
+	ep->qp = iparams->qp;
+	ep->qp->ep = ep;
+
+	if (ep->mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
+		/* Negotiate ord/ird: if upperlayer requested ord larger than
+		 * ird advertised by remote, we need to decrease our ord
+		 */
+		if (iparams->ord > ep->cm_info.ird)
+			iparams->ord = ep->cm_info.ird;
+
+		if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) &&
+		    (iparams->ird == 0))
+			iparams->ird = 1;
+	}
+
+	/* Update cm_info ord/ird to be negotiated values */
+	ep->cm_info.ord = iparams->ord;
+	ep->cm_info.ird = iparams->ird;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->private_data, iparams->private_data_len);
+
+	rc = qed_iwarp_mpa_offload(p_hwfn, ep);
+	if (rc)
+		qed_iwarp_modify_qp(p_hwfn,
+				    iparams->qp, QED_IWARP_QP_STATE_ERROR, 1);
+
+	return rc;
+}
+
+int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+
+	ep = iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in reject is NULL\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x)\n", ep->tcp_cid);
+
+	ep->cb_context = iparams->cb_context;
+	ep->qp = NULL;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->private_data, iparams->private_data_len);
+
+	return qed_iwarp_mpa_offload(p_hwfn, ep);
 }
 
 static void
@@ -526,6 +1311,38 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	return rc;
 }
 
+static bool
+qed_iwarp_ep_exists(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info)
+{
+	struct qed_iwarp_ep *ep = NULL;
+	bool found = false;
+
+	list_for_each_entry(ep,
+			    &p_hwfn->p_rdma_info->iwarp.ep_list,
+			    list_entry) {
+		if ((ep->cm_info.local_port == cm_info->local_port) &&
+		    (ep->cm_info.remote_port == cm_info->remote_port) &&
+		    (ep->cm_info.vlan == cm_info->vlan) &&
+		    !memcmp(&ep->cm_info.local_ip, cm_info->local_ip,
+			    sizeof(cm_info->local_ip)) &&
+		    !memcmp(&ep->cm_info.remote_ip, cm_info->remote_ip,
+			    sizeof(cm_info->remote_ip))) {
+			found = true;
+			break;
+		}
+	}
+
+	if (found) {
+		DP_NOTICE(p_hwfn,
+			  "SYN received on active connection - dropping\n");
+		qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+		return true;
+	}
+
+	return false;
+}
+
 static struct qed_iwarp_listener *
 qed_iwarp_get_listener(struct qed_hwfn *p_hwfn,
 		       struct qed_iwarp_cm_info *cm_info)
@@ -596,9 +1413,8 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 
 	eth_hlen = ETH_HLEN + (vlan_valid ? sizeof(u32) : 0);
 
-	memcpy(remote_mac_addr, ethh->h_source, ETH_ALEN);
-
-	memcpy(local_mac_addr, ethh->h_dest, ETH_ALEN);
+	ether_addr_copy(remote_mac_addr, ethh->h_source);
+	ether_addr_copy(local_mac_addr, ethh->h_dest);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_type =%d source mac: %pM\n",
 		   eth_type, ethh->h_source);
@@ -661,9 +1477,12 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	struct qed_hwfn *p_hwfn = cxt;
 	u8 remote_mac_addr[ETH_ALEN];
 	u8 local_mac_addr[ETH_ALEN];
+	struct qed_iwarp_ep *ep;
 	int tcp_start_offset;
+	u8 ts_hdr_size = 0;
 	u8 ll2_syn_handle;
 	int payload_len;
+	u32 hdr_size;
 	int rc;
 
 	memset(&cm_info, 0, sizeof(cm_info));
@@ -719,6 +1538,49 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	}
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Received syn on listening port\n");
+	/* There may be an open ep on this connection if this is a syn
+	 * retrasnmit... need to make sure there isn't...
+	 */
+	if (qed_iwarp_ep_exists(p_hwfn, &cm_info))
+		goto err;
+
+	ep = qed_iwarp_get_free_ep(p_hwfn);
+	if (!ep)
+		goto err;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ether_addr_copy(ep->remote_mac_addr, remote_mac_addr);
+	ether_addr_copy(ep->local_mac_addr, local_mac_addr);
+
+	memcpy(&ep->cm_info, &cm_info, sizeof(ep->cm_info));
+
+	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
+		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
+
+	hdr_size = ((cm_info.ip_version == QED_TCP_IPV4) ? 40 : 60) +
+		   ts_hdr_size;
+	ep->mss = p_hwfn->p_rdma_info->iwarp.max_mtu - hdr_size;
+	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
+
+	ep->event_cb = listener->event_cb;
+	ep->cb_context = listener->cb_context;
+	ep->connect_mode = TCP_CONNECT_PASSIVE;
+
+	ep->syn = buf;
+	ep->syn_ip_payload_length = (u16)payload_len;
+	ep->syn_phy_addr = buf->data_phys_addr + data->u.placement_offset +
+			   tcp_start_offset;
+
+	rc = qed_iwarp_tcp_offload(p_hwfn, ep);
+	if (rc) {
+		qed_iwarp_return_ep(p_hwfn, ep);
+		goto err;
+	}
+
+	return;
 err:
 	qed_iwarp_ll2_post_rx(p_hwfn, buf, ll2_syn_handle);
 }
@@ -905,7 +1767,12 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 
 	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
 
+	iwarp_info->rtr_type =  MPA_RTR_TYPE_ZERO_SEND |
+				MPA_RTR_TYPE_ZERO_WRITE |
+				MPA_RTR_TYPE_ZERO_READ;
+
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_list);
 	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.listen_list);
 
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
@@ -918,6 +1785,7 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 {
 	int rc;
 
+	qed_iwarp_free_prealloc_ep(p_hwfn);
 	rc = qed_iwarp_wait_for_all_cids(p_hwfn);
 	if (rc)
 		return rc;
@@ -927,11 +1795,70 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
+void
+qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
+			   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	/* Done with the SYN packet, post back to ll2 rx */
+	qed_iwarp_ll2_post_rx(p_hwfn, ep->syn,
+			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+	ep->syn = NULL;
+
+	/* If connect failed - upper layer doesn't know about it */
+	qed_iwarp_mpa_received(p_hwfn, ep);
+}
+
+static inline bool
+qed_iwarp_check_ep_ok(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	if (!ep || (ep->sig != QED_EP_SIG)) {
+		DP_ERR(p_hwfn, "ERROR ON ASYNC ep=%p\n", ep);
+		return false;
+	}
+
+	return true;
+}
+
 static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 				 u8 fw_event_code, u16 echo,
 				 union event_ring_data *data,
 				 u8 fw_return_code)
 {
+	struct regpair *fw_handle = &data->rdma_data.async_handle;
+	struct qed_iwarp_ep *ep = NULL;
+	u16 cid;
+
+	ep = (struct qed_iwarp_ep *)(uintptr_t)HILO_64(fw_handle->hi,
+						       fw_handle->lo);
+
+	switch (fw_event_code) {
+	case IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE:
+		/* Async completion after TCP 3-way handshake */
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "EP(0x%x) IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE fw_ret_code=%d\n",
+			   ep->tcp_cid, fw_return_code);
+		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_mpa_complete(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_CID_CLEANED:
+		cid = (u16)le32_to_cpu(fw_handle->lo);
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "(0x%x)IWARP_EVENT_TYPE_ASYNC_CID_CLEANED\n", cid);
+		qed_iwarp_cid_cleaned(p_hwfn, cid);
+
+		break;
+	}
 	return 0;
 }
 
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 29005ac..bedac98 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -42,6 +42,8 @@ enum qed_iwarp_qp_state {
 
 enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
 
+#define QED_IWARP_PREALLOC_CNT  (256)
+
 #define QED_IWARP_LL2_SYN_TX_SIZE       (128)
 #define QED_IWARP_LL2_SYN_RX_SIZE       (256)
 #define QED_IWARP_MAX_SYN_PKT_SIZE      (128)
@@ -55,6 +57,8 @@ struct qed_iwarp_ll2_buff {
 
 struct qed_iwarp_info {
 	struct list_head listen_list;	/* qed_iwarp_listener */
+	struct list_head ep_list;	/* qed_iwarp_ep */
+	struct list_head ep_free_list;	/* pre-allocated ep's */
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
 	u32 rcv_wnd_scale;
@@ -68,6 +72,61 @@ struct qed_iwarp_info {
 	enum mpa_rtr_type rtr_type;
 };
 
+enum qed_iwarp_ep_state {
+	QED_IWARP_EP_INIT,
+	QED_IWARP_EP_MPA_REQ_RCVD,
+	QED_IWARP_EP_MPA_OFFLOADED,
+	QED_IWARP_EP_ESTABLISHED,
+	QED_IWARP_EP_CLOSED
+};
+
+union async_output {
+	struct iwarp_eqe_data_mpa_async_completion mpa_response;
+	struct iwarp_eqe_data_tcp_async_completion mpa_request;
+};
+
+#define QED_MAX_PRIV_DATA_LEN (512)
+struct qed_iwarp_ep_memory {
+	u8 in_pdata[QED_MAX_PRIV_DATA_LEN];
+	u8 out_pdata[QED_MAX_PRIV_DATA_LEN];
+	union async_output async_output;
+};
+
+/* Endpoint structure represents a TCP connection. This connection can be
+ * associated with a QP or not (in which case QP==NULL)
+ */
+struct qed_iwarp_ep {
+	struct list_head list_entry;
+	struct qed_rdma_qp *qp;
+	struct qed_iwarp_ep_memory *ep_buffer_virt;
+	dma_addr_t ep_buffer_phys;
+	enum qed_iwarp_ep_state state;
+	int sig;
+	struct qed_iwarp_cm_info cm_info;
+	enum tcp_connect_mode connect_mode;
+	enum mpa_rtr_type rtr_type;
+	enum mpa_negotiation_mode mpa_rev;
+	u32 tcp_cid;
+	u32 cid;
+	u16 mss;
+	u8 remote_mac_addr[6];
+	u8 local_mac_addr[6];
+	bool mpa_reply_processed;
+
+	/* For Passive side - syn packet related data */
+	u16 syn_ip_payload_length;
+	struct qed_iwarp_ll2_buff *syn;
+	dma_addr_t syn_phy_addr;
+
+	/* The event_cb function is called for asynchrounous events associated
+	 * with the ep. It is initialized at different entry points depending
+	 * on whether the ep is the tcp connection active side or passive side
+	 * The cb_context is passed to the event_cb function.
+	 */
+	iwarp_event_handler event_cb;
+	void *cb_context;
+};
+
 struct qed_iwarp_listener {
 	struct list_head list_entry;
 
@@ -115,6 +174,9 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_iwarp_listen_in *iparams,
 			struct qed_iwarp_listen_out *oparams);
 
+int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams);
+
+int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams);
 int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
 
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_l2.c b/drivers/net/ethernet/qlogic/qed/qed_l2.c
index 27ea54b..0ba5ec8 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_l2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_l2.c
@@ -1227,19 +1227,6 @@ static enum eth_filter_action qed_filter_action(enum qed_filter_opcode opcode)
 	return action;
 }
 
-static void qed_set_fw_mac_addr(__le16 *fw_msb,
-				__le16 *fw_mid,
-				__le16 *fw_lsb,
-				u8 *mac)
-{
-	((u8 *)fw_msb)[0] = mac[1];
-	((u8 *)fw_msb)[1] = mac[0];
-	((u8 *)fw_mid)[0] = mac[3];
-	((u8 *)fw_mid)[1] = mac[2];
-	((u8 *)fw_lsb)[0] = mac[5];
-	((u8 *)fw_lsb)[1] = mac[4];
-}
-
 static int
 qed_filter_ucast_common(struct qed_hwfn *p_hwfn,
 			u16 opaque_fid,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.h b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
index 90e4e0f..18ec9cb 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
@@ -85,6 +85,7 @@ struct qed_rdma_info {
 	struct qed_bmap qp_map;
 	struct qed_bmap srq_map;
 	struct qed_bmap cid_map;
+	struct qed_bmap tcp_cid_map;
 	struct qed_bmap real_cid_map;
 	struct qed_bmap dpi_map;
 	struct qed_bmap toggle_bits;
@@ -167,6 +168,7 @@ struct qed_rdma_qp {
 
 	void *shared_queue;
 	dma_addr_t shared_queue_phys_addr;
+	struct qed_iwarp_ep *ep;
 };
 
 #if IS_ENABLED(CONFIG_QED_RDMA)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sp.h b/drivers/net/ethernet/qlogic/qed/qed_sp.h
index c3752c5..ab4ad8a 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_sp.h
@@ -111,6 +111,8 @@ int qed_eth_cqe_completion(struct qed_hwfn *p_hwfn,
 	struct rdma_srq_destroy_ramrod_data rdma_destroy_srq;
 	struct rdma_srq_modify_ramrod_data rdma_modify_srq;
 	struct iwarp_create_qp_ramrod_data iwarp_create_qp;
+	struct iwarp_tcp_offload_ramrod_data iwarp_tcp_offload;
+	struct iwarp_mpa_offload_ramrod_data iwarp_mpa_offload;
 	struct iwarp_modify_qp_ramrod_data iwarp_modify_qp;
 	struct iwarp_init_func_ramrod_data iwarp_init_func;
 	struct fcoe_init_ramrod_params fcoe_init;
diff --git a/include/linux/qed/common_hsi.h b/include/linux/qed/common_hsi.h
index 885ae13..39e2a2a 100644
--- a/include/linux/qed/common_hsi.h
+++ b/include/linux/qed/common_hsi.h
@@ -38,6 +38,8 @@
 #include <linux/slab.h>
 
 /* dma_addr_t manip */
+#define PTR_LO(x)               ((u32)(((uintptr_t)(x)) & 0xffffffff))
+#define PTR_HI(x)               ((u32)((((uintptr_t)(x)) >> 16) >> 16))
 #define DMA_LO_LE(x)		cpu_to_le32(lower_32_bits(x))
 #define DMA_HI_LE(x)		cpu_to_le32(upper_32_bits(x))
 #define DMA_REGPAIR_LE(x, val)	do { \
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 28df568..c4c241f 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -471,7 +471,8 @@ struct qed_rdma_counters_out_params {
 #define QED_ROCE_TX_FRAG_FAILURE        (2)
 
 enum qed_iwarp_event_type {
-	QED_IWARP_EVENT_MPA_REQUEST,	/* Passive side request received */
+	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
+	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
 };
 
 enum qed_tcp_ip_version {
@@ -516,6 +517,23 @@ struct qed_iwarp_listen_out {
 	void *handle;
 };
 
+struct qed_iwarp_accept_in {
+	void *ep_context;
+	void *cb_context;
+	struct qed_rdma_qp *qp;
+	const void *private_data;
+	u16 private_data_len;
+	u8 ord;
+	u8 ird;
+};
+
+struct qed_iwarp_reject_in {
+	void *ep_context;
+	void *cb_context;
+	const void *private_data;
+	u16 private_data_len;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -626,6 +644,12 @@ struct qed_rdma_ops {
 				   struct qed_iwarp_listen_in *iparams,
 				   struct qed_iwarp_listen_out *oparams);
 
+	int (*iwarp_accept)(void *rdma_cxt,
+			    struct qed_iwarp_accept_in *iparams);
+
+	int (*iwarp_reject)(void *rdma_cxt,
+			    struct qed_iwarp_reject_in *iparams);
+
 	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
 
 };
-- 
1.8.3.1

[PATCH net-next 07/12] qed: iWARP CM add passive side connect

[Michal Kalderon]

This patch implements the passive side connect.
It addresses pre-allocating resources, creating a connection
element upon valid SYN packet received. Calling upper layer and
implementation of the accept/reject calls.

Error handling is not part of this patch.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed.h       |   2 +
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |  11 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 939 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  62 ++
 drivers/net/ethernet/qlogic/qed/qed_l2.c    |  13 -
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |   2 +
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |   2 +
 include/linux/qed/common_hsi.h              |   2 +
 include/linux/qed/qed_rdma_if.h             |  26 +-
 9 files changed, 1039 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index fd8cd5e..91003bc 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -789,6 +789,8 @@ void qed_configure_vp_wfq_on_link_change(struct qed_dev *cdev,
 void qed_clean_wfq_db(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
 int qed_device_num_engines(struct qed_dev *cdev);
 int qed_device_get_port_id(struct qed_dev *cdev);
+void qed_set_fw_mac_addr(__le16 *fw_msb,
+			 __le16 *fw_mid, __le16 *fw_lsb, u8 *mac);
 
 #define QED_LEADING_HWFN(dev)   (&dev->hwfns[0])
 
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 6c8505d..4060a6a 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -4127,3 +4127,14 @@ int qed_device_get_port_id(struct qed_dev *cdev)
 {
 	return (QED_LEADING_HWFN(cdev)->abs_pf_id) % qed_device_num_ports(cdev);
 }
+
+void qed_set_fw_mac_addr(__le16 *fw_msb,
+			 __le16 *fw_mid, __le16 *fw_lsb, u8 *mac)
+{
+	((u8 *)fw_msb)[0] = mac[1];
+	((u8 *)fw_msb)[1] = mac[0];
+	((u8 *)fw_mid)[0] = mac[3];
+	((u8 *)fw_mid)[1] = mac[2];
+	((u8 *)fw_lsb)[0] = mac[5];
+	((u8 *)fw_lsb)[1] = mac[4];
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index 2bab57c..a6dadae 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -44,9 +44,30 @@
 
 #define QED_IWARP_ORD_DEFAULT		32
 #define QED_IWARP_IRD_DEFAULT		32
+#define QED_IWARP_MAX_FW_MSS		4120
+
+#define QED_EP_SIG 0xecabcdef
+
+struct mpa_v2_hdr {
+	__be16 ird;
+	__be16 ord;
+};
+
+#define MPA_V2_PEER2PEER_MODEL  0x8000
+#define MPA_V2_SEND_RTR         0x4000	/* on ird */
+#define MPA_V2_READ_RTR         0x4000	/* on ord */
+#define MPA_V2_WRITE_RTR        0x8000
+#define MPA_V2_IRD_ORD_MASK     0x3FFF
+
+#define MPA_REV2(_mpa_rev) ((_mpa_rev) == MPA_NEGOTIATION_TYPE_ENHANCED)
+
+#define QED_IWARP_INVALID_TCP_CID	0xffffffff
 #define QED_IWARP_RCV_WND_SIZE_DEF	(256 * 1024)
 #define QED_IWARP_RCV_WND_SIZE_MIN	(64 * 1024)
+#define TIMESTAMP_HEADER_SIZE		(12)
+
 #define QED_IWARP_TS_EN			BIT(0)
+#define QED_IWARP_DA_EN			BIT(1)
 #define QED_IWARP_PARAM_CRC_NEEDED	(1)
 #define QED_IWARP_PARAM_P2P		(1)
 
@@ -63,7 +84,8 @@ void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn)
 	dev->max_inline = IWARP_REQ_MAX_INLINE_DATA_SIZE;
 	dev->max_qp = min_t(u32,
 			    IWARP_MAX_QPS,
-			    p_hwfn->p_rdma_info->num_qps);
+			    p_hwfn->p_rdma_info->num_qps) -
+		      QED_IWARP_PREALLOC_CNT;
 
 	dev->max_cq = dev->max_qp;
 
@@ -78,12 +100,22 @@ void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	p_hwfn->b_rdma_enabled_in_prs = true;
 }
 
+/* We have two cid maps, one for tcp which should be used only from passive
+ * syn processing and replacing a pre-allocated ep in the list. The second
+ * for active tcp and for QPs.
+ */
 static void qed_iwarp_cid_cleaned(struct qed_hwfn *p_hwfn, u32 cid)
 {
 	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
 
 	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
-	qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+
+	if (cid < QED_IWARP_PREALLOC_CNT)
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map,
+				    cid);
+	else
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+
 	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
 }
 
@@ -107,6 +139,45 @@ static int qed_iwarp_alloc_cid(struct qed_hwfn *p_hwfn, u32 *cid)
 	return rc;
 }
 
+static void qed_iwarp_set_tcp_cid(struct qed_hwfn *p_hwfn, u32 cid)
+{
+	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	qed_bmap_set_id(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+}
+
+/* This function allocates a cid for passive tcp (called from syn receive)
+ * the reason it's separate from the regular cid allocation is because it
+ * is assured that these cids already have ilt allocated. They are preallocated
+ * to ensure that we won't need to allocate memory during syn processing
+ */
+static int qed_iwarp_alloc_tcp_cid(struct qed_hwfn *p_hwfn, u32 *cid)
+{
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+
+	rc = qed_rdma_bmap_alloc_id(p_hwfn,
+				    &p_hwfn->p_rdma_info->tcp_cid_map, cid);
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+
+	if (rc) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "can't allocate iwarp tcp cid max-count=%d\n",
+			   p_hwfn->p_rdma_info->tcp_cid_map.max_count);
+
+		*cid = QED_IWARP_INVALID_TCP_CID;
+		return rc;
+	}
+
+	*cid += qed_cxt_get_proto_cid_start(p_hwfn,
+					    p_hwfn->p_rdma_info->proto);
+	return 0;
+}
+
 int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
 			struct qed_rdma_qp *qp,
 			struct qed_rdma_create_qp_out_params *out_params)
@@ -403,6 +474,26 @@ int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	return rc;
 }
 
+static void qed_iwarp_destroy_ep(struct qed_hwfn *p_hwfn,
+				 struct qed_iwarp_ep *ep,
+				 bool remove_from_active_list)
+{
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+			  sizeof(*ep->ep_buffer_virt),
+			  ep->ep_buffer_virt, ep->ep_buffer_phys);
+
+	if (remove_from_active_list) {
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_del(&ep->list_entry);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+
+	if (ep->qp)
+		ep->qp->ep = NULL;
+
+	kfree(ep);
+}
+
 int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
 	int rc = 0;
@@ -424,6 +515,507 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	return rc;
 }
 
+static int
+qed_iwarp_create_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep **ep_out)
+{
+	struct qed_iwarp_ep *ep;
+	int rc;
+
+	ep = kzalloc(sizeof(*ep), GFP_KERNEL);
+	if (!ep)
+		return -ENOMEM;
+
+	ep->state = QED_IWARP_EP_INIT;
+
+	ep->ep_buffer_virt = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+						sizeof(*ep->ep_buffer_virt),
+						&ep->ep_buffer_phys,
+						GFP_KERNEL);
+	if (!ep->ep_buffer_virt) {
+		rc = -ENOMEM;
+		goto err;
+	}
+
+	ep->sig = QED_EP_SIG;
+
+	*ep_out = ep;
+
+	return 0;
+
+err:
+	kfree(ep);
+	return rc;
+}
+
+static void
+qed_iwarp_print_tcp_ramrod(struct qed_hwfn *p_hwfn,
+			   struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod)
+{
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "local_mac=%x %x %x, remote_mac=%x %x %x\n",
+		   p_tcp_ramrod->tcp.local_mac_addr_lo,
+		   p_tcp_ramrod->tcp.local_mac_addr_mid,
+		   p_tcp_ramrod->tcp.local_mac_addr_hi,
+		   p_tcp_ramrod->tcp.remote_mac_addr_lo,
+		   p_tcp_ramrod->tcp.remote_mac_addr_mid,
+		   p_tcp_ramrod->tcp.remote_mac_addr_hi);
+
+	if (p_tcp_ramrod->tcp.ip_version == TCP_IPV4) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "local_ip=%pI4h:%x, remote_ip=%pI4h%x, vlan=%x\n",
+			   p_tcp_ramrod->tcp.local_ip,
+			   p_tcp_ramrod->tcp.local_port,
+			   p_tcp_ramrod->tcp.remote_ip,
+			   p_tcp_ramrod->tcp.remote_port,
+			   p_tcp_ramrod->tcp.vlan_id);
+	} else {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "local_ip=%pI6h:%x, remote_ip=%pI6h:%x, vlan=%x\n",
+			   p_tcp_ramrod->tcp.local_ip,
+			   p_tcp_ramrod->tcp.local_port,
+			   p_tcp_ramrod->tcp.remote_ip,
+			   p_tcp_ramrod->tcp.remote_port,
+			   p_tcp_ramrod->tcp.vlan_id);
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "flow_label=%x, ttl=%x, tos_or_tc=%x, mss=%x, rcv_wnd_scale=%x, connect_mode=%x, flags=%x\n",
+		   p_tcp_ramrod->tcp.flow_label,
+		   p_tcp_ramrod->tcp.ttl,
+		   p_tcp_ramrod->tcp.tos_or_tc,
+		   p_tcp_ramrod->tcp.mss,
+		   p_tcp_ramrod->tcp.rcv_wnd_scale,
+		   p_tcp_ramrod->tcp.connect_mode,
+		   p_tcp_ramrod->tcp.flags);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "syn_ip_payload_length=%x, lo=%x, hi=%x\n",
+		   p_tcp_ramrod->tcp.syn_ip_payload_length,
+		   p_tcp_ramrod->tcp.syn_phy_addr_lo,
+		   p_tcp_ramrod->tcp.syn_phy_addr_hi);
+}
+
+static int
+qed_iwarp_tcp_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	struct iwarp_tcp_offload_ramrod_data *p_tcp_ramrod;
+	struct tcp_offload_params_opt2 *tcp;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	dma_addr_t async_output_phys;
+	dma_addr_t in_pdata_phys;
+	u16 physical_q;
+	u8 tcp_flags;
+	int rc;
+	int i;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = ep->tcp_cid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_CB;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_TCP_OFFLOAD,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		return rc;
+
+	p_tcp_ramrod = &p_ent->ramrod.iwarp_tcp_offload;
+
+	in_pdata_phys = ep->ep_buffer_phys +
+			offsetof(struct qed_iwarp_ep_memory, in_pdata);
+	DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.incoming_ulp_buffer.addr,
+		       in_pdata_phys);
+
+	p_tcp_ramrod->iwarp.incoming_ulp_buffer.len =
+	    cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata));
+
+	async_output_phys = ep->ep_buffer_phys +
+			    offsetof(struct qed_iwarp_ep_memory, async_output);
+	DMA_REGPAIR_LE(p_tcp_ramrod->iwarp.async_eqe_output_buf,
+		       async_output_phys);
+
+	p_tcp_ramrod->iwarp.handle_for_async.hi = cpu_to_le32(PTR_HI(ep));
+	p_tcp_ramrod->iwarp.handle_for_async.lo = cpu_to_le32(PTR_LO(ep));
+
+	physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD);
+	p_tcp_ramrod->iwarp.physical_q0 = cpu_to_le16(physical_q);
+	physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK);
+	p_tcp_ramrod->iwarp.physical_q1 = cpu_to_le16(physical_q);
+	p_tcp_ramrod->iwarp.mpa_mode = iwarp_info->mpa_rev;
+
+	tcp = &p_tcp_ramrod->tcp;
+	qed_set_fw_mac_addr(&tcp->remote_mac_addr_hi,
+			    &tcp->remote_mac_addr_mid,
+			    &tcp->remote_mac_addr_lo, ep->remote_mac_addr);
+	qed_set_fw_mac_addr(&tcp->local_mac_addr_hi, &tcp->local_mac_addr_mid,
+			    &tcp->local_mac_addr_lo, ep->local_mac_addr);
+
+	tcp->vlan_id = cpu_to_le16(ep->cm_info.vlan);
+
+	tcp_flags = p_hwfn->p_rdma_info->iwarp.tcp_flags;
+	tcp->flags = 0;
+	SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_TS_EN,
+		  !!(tcp_flags & QED_IWARP_TS_EN));
+
+	SET_FIELD(tcp->flags, TCP_OFFLOAD_PARAMS_OPT2_DA_EN,
+		  !!(tcp_flags & QED_IWARP_DA_EN));
+
+	tcp->ip_version = ep->cm_info.ip_version;
+
+	for (i = 0; i < 4; i++) {
+		tcp->remote_ip[i] = cpu_to_le32(ep->cm_info.remote_ip[i]);
+		tcp->local_ip[i] = cpu_to_le32(ep->cm_info.local_ip[i]);
+	}
+
+	tcp->remote_port = cpu_to_le16(ep->cm_info.remote_port);
+	tcp->local_port = cpu_to_le16(ep->cm_info.local_port);
+	tcp->mss = cpu_to_le16(ep->mss);
+	tcp->flow_label = 0;
+	tcp->ttl = 0x40;
+	tcp->tos_or_tc = 0;
+
+	tcp->rcv_wnd_scale = (u8)p_hwfn->p_rdma_info->iwarp.rcv_wnd_scale;
+	tcp->connect_mode = ep->connect_mode;
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		tcp->syn_ip_payload_length =
+			cpu_to_le16(ep->syn_ip_payload_length);
+		tcp->syn_phy_addr_hi = DMA_HI_LE(ep->syn_phy_addr);
+		tcp->syn_phy_addr_lo = DMA_LO_LE(ep->syn_phy_addr);
+	}
+
+	qed_iwarp_print_tcp_ramrod(p_hwfn, p_tcp_ramrod);
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "EP(0x%x) Offload completed rc=%d\n", ep->tcp_cid, rc);
+
+	return rc;
+}
+
+static void
+qed_iwarp_mpa_received(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	struct qed_iwarp_cm_event_params params;
+	struct mpa_v2_hdr *mpa_v2;
+	union async_output *async_data;
+	u16 mpa_ord, mpa_ird;
+	u8 mpa_hdr_size = 0;
+	u8 mpa_rev;
+
+	async_data = &ep->ep_buffer_virt->async_output;
+
+	mpa_rev = async_data->mpa_request.mpa_handshake_mode;
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "private_data_len=%x handshake_mode=%x private_data=(%x)\n",
+		   async_data->mpa_request.ulp_data_len,
+		   mpa_rev, *((u32 *)((u8 *)ep->ep_buffer_virt->in_pdata)));
+
+	if (mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
+		/* Read ord/ird values from private data buffer */
+		mpa_v2 = (struct mpa_v2_hdr *)ep->ep_buffer_virt->in_pdata;
+		mpa_hdr_size = sizeof(*mpa_v2);
+
+		mpa_ord = ntohs(mpa_v2->ord);
+		mpa_ird = ntohs(mpa_v2->ird);
+
+		/* Temprary store in cm_info incoming ord/ird requested, later
+		 * replace with negotiated value during accept
+		 */
+		ep->cm_info.ord = (u8)min_t(u16,
+					    (mpa_ord & MPA_V2_IRD_ORD_MASK),
+					    QED_IWARP_ORD_DEFAULT);
+
+		ep->cm_info.ird = (u8)min_t(u16,
+					    (mpa_ird & MPA_V2_IRD_ORD_MASK),
+					    QED_IWARP_IRD_DEFAULT);
+
+		/* Peer2Peer negotiation */
+		ep->rtr_type = MPA_RTR_TYPE_NONE;
+		if (mpa_ird & MPA_V2_PEER2PEER_MODEL) {
+			if (mpa_ord & MPA_V2_WRITE_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_WRITE;
+
+			if (mpa_ord & MPA_V2_READ_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_READ;
+
+			if (mpa_ird & MPA_V2_SEND_RTR)
+				ep->rtr_type |= MPA_RTR_TYPE_ZERO_SEND;
+
+			ep->rtr_type &= iwarp_info->rtr_type;
+
+			/* if we're left with no match send our capabilities */
+			if (ep->rtr_type == MPA_RTR_TYPE_NONE)
+				ep->rtr_type = iwarp_info->rtr_type;
+		}
+
+		ep->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED;
+	} else {
+		ep->cm_info.ord = QED_IWARP_ORD_DEFAULT;
+		ep->cm_info.ird = QED_IWARP_IRD_DEFAULT;
+		ep->mpa_rev = MPA_NEGOTIATION_TYPE_BASIC;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x rtr:0x%x ulp_data_len = %x mpa_hdr_size = %x\n",
+		   mpa_rev, ep->cm_info.ord, ep->cm_info.ird, ep->rtr_type,
+		   async_data->mpa_request.ulp_data_len, mpa_hdr_size);
+
+	/* Strip mpa v2 hdr from private data before sending to upper layer */
+	ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_hdr_size;
+
+	ep->cm_info.private_data_len = async_data->mpa_request.ulp_data_len -
+				       mpa_hdr_size;
+
+	params.event = QED_IWARP_EVENT_MPA_REQUEST;
+	params.cm_info = &ep->cm_info;
+	params.ep_context = ep;
+	params.status = 0;
+
+	ep->state = QED_IWARP_EP_MPA_REQ_RCVD;
+	ep->event_cb(ep->cb_context, &params);
+}
+
+static int
+qed_iwarp_mpa_offload(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct iwarp_mpa_offload_ramrod_data *p_mpa_ramrod;
+	struct qed_sp_init_data init_data;
+	dma_addr_t async_output_phys;
+	struct qed_spq_entry *p_ent;
+	dma_addr_t out_pdata_phys;
+	dma_addr_t in_pdata_phys;
+	struct qed_rdma_qp *qp;
+	bool reject;
+	int rc;
+
+	if (!ep)
+		return -EINVAL;
+
+	qp = ep->qp;
+	reject = !qp;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = reject ? ep->tcp_cid : qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		return rc;
+
+	p_mpa_ramrod = &p_ent->ramrod.iwarp_mpa_offload;
+	out_pdata_phys = ep->ep_buffer_phys +
+			 offsetof(struct qed_iwarp_ep_memory, out_pdata);
+	DMA_REGPAIR_LE(p_mpa_ramrod->common.outgoing_ulp_buffer.addr,
+		       out_pdata_phys);
+	p_mpa_ramrod->common.outgoing_ulp_buffer.len =
+	    ep->cm_info.private_data_len;
+	p_mpa_ramrod->common.crc_needed = p_hwfn->p_rdma_info->iwarp.crc_needed;
+
+	p_mpa_ramrod->common.out_rq.ord = ep->cm_info.ord;
+	p_mpa_ramrod->common.out_rq.ird = ep->cm_info.ird;
+
+	p_mpa_ramrod->tcp_cid = p_hwfn->hw_info.opaque_fid << 16 | ep->tcp_cid;
+
+	in_pdata_phys = ep->ep_buffer_phys +
+			offsetof(struct qed_iwarp_ep_memory, in_pdata);
+	p_mpa_ramrod->tcp_connect_side = ep->connect_mode;
+	DMA_REGPAIR_LE(p_mpa_ramrod->incoming_ulp_buffer.addr,
+		       in_pdata_phys);
+	p_mpa_ramrod->incoming_ulp_buffer.len =
+	    cpu_to_le16(sizeof(ep->ep_buffer_virt->in_pdata));
+	async_output_phys = ep->ep_buffer_phys +
+			    offsetof(struct qed_iwarp_ep_memory, async_output);
+	DMA_REGPAIR_LE(p_mpa_ramrod->async_eqe_output_buf,
+		       async_output_phys);
+	p_mpa_ramrod->handle_for_async.hi = cpu_to_le32(PTR_HI(ep));
+	p_mpa_ramrod->handle_for_async.lo = cpu_to_le32(PTR_LO(ep));
+
+	if (!reject) {
+		DMA_REGPAIR_LE(p_mpa_ramrod->shared_queue_addr,
+			       qp->shared_queue_phys_addr);
+		p_mpa_ramrod->stats_counter_id =
+		    RESC_START(p_hwfn, QED_RDMA_STATS_QUEUE) + qp->stats_queue;
+	} else {
+		p_mpa_ramrod->common.reject = 1;
+	}
+
+	p_mpa_ramrod->mode = ep->mpa_rev;
+	SET_FIELD(p_mpa_ramrod->rtr_pref,
+		  IWARP_MPA_OFFLOAD_RAMROD_DATA_RTR_SUPPORTED, ep->rtr_type);
+
+	ep->state = QED_IWARP_EP_MPA_OFFLOADED;
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+	if (!reject)
+		ep->cid = qp->icid;	/* Now they're migrated. */
+
+	DP_VERBOSE(p_hwfn,
+		   QED_MSG_RDMA,
+		   "QP(0x%x) EP(0x%x) MPA Offload rc = %d IRD=0x%x ORD=0x%x rtr_type=%d mpa_rev=%d reject=%d\n",
+		   reject ? 0xffff : qp->icid,
+		   ep->tcp_cid,
+		   rc,
+		   ep->cm_info.ird,
+		   ep->cm_info.ord, ep->rtr_type, ep->mpa_rev, reject);
+	return rc;
+}
+
+static void
+qed_iwarp_return_ep(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	ep->state = QED_IWARP_EP_INIT;
+	if (ep->qp)
+		ep->qp->ep = NULL;
+	ep->qp = NULL;
+	memset(&ep->cm_info, 0, sizeof(ep->cm_info));
+
+	if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) {
+		/* We don't care about the return code, it's ok if tcp_cid
+		 * remains invalid...in this case we'll defer allocation
+		 */
+		qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid);
+	}
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	list_del(&ep->list_entry);
+	list_add_tail(&ep->list_entry,
+		      &p_hwfn->p_rdma_info->iwarp.ep_free_list);
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+}
+
+#define QED_IWARP_CONNECT_MODE_STRING(ep) \
+	((ep)->connect_mode == TCP_CONNECT_PASSIVE) ? "Passive" : "Active"
+
+/* Called as a result of the event:
+ * IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE
+ */
+static void
+qed_iwarp_mpa_complete(struct qed_hwfn *p_hwfn,
+		       struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
+		   ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird);
+
+	params.cm_info = &ep->cm_info;
+
+	params.ep_context = ep;
+
+	ep->state = QED_IWARP_EP_CLOSED;
+
+	switch (fw_return_code) {
+	case RDMA_RETURN_OK:
+		ep->qp->max_rd_atomic_req = ep->cm_info.ord;
+		ep->qp->max_rd_atomic_resp = ep->cm_info.ird;
+		qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_RTS, 1);
+		ep->state = QED_IWARP_EP_ESTABLISHED;
+		params.status = 0;
+		break;
+	default:
+		params.status = -ECONNRESET;
+		break;
+	}
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
+static void
+qed_iwarp_mpa_v2_set_private(struct qed_hwfn *p_hwfn,
+			     struct qed_iwarp_ep *ep, u8 *mpa_data_size)
+{
+	struct mpa_v2_hdr *mpa_v2_params;
+	u16 mpa_ird, mpa_ord;
+
+	*mpa_data_size = 0;
+	if (MPA_REV2(ep->mpa_rev)) {
+		mpa_v2_params =
+		    (struct mpa_v2_hdr *)ep->ep_buffer_virt->out_pdata;
+		*mpa_data_size = sizeof(*mpa_v2_params);
+
+		mpa_ird = (u16)ep->cm_info.ird;
+		mpa_ord = (u16)ep->cm_info.ord;
+
+		if (ep->rtr_type != MPA_RTR_TYPE_NONE) {
+			mpa_ird |= MPA_V2_PEER2PEER_MODEL;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_SEND)
+				mpa_ird |= MPA_V2_SEND_RTR;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_WRITE)
+				mpa_ord |= MPA_V2_WRITE_RTR;
+
+			if (ep->rtr_type & MPA_RTR_TYPE_ZERO_READ)
+				mpa_ord |= MPA_V2_READ_RTR;
+		}
+
+		mpa_v2_params->ird = htons(mpa_ird);
+		mpa_v2_params->ord = htons(mpa_ord);
+
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "MPA_NEGOTIATE Header: [%x ord:%x ird] %x ord:%x ird:%x peer2peer:%x rtr_send:%x rtr_write:%x rtr_read:%x\n",
+			   mpa_v2_params->ird,
+			   mpa_v2_params->ord,
+			   *((u32 *)mpa_v2_params),
+			   mpa_ord & MPA_V2_IRD_ORD_MASK,
+			   mpa_ird & MPA_V2_IRD_ORD_MASK,
+			   !!(mpa_ird & MPA_V2_PEER2PEER_MODEL),
+			   !!(mpa_ird & MPA_V2_SEND_RTR),
+			   !!(mpa_ord & MPA_V2_WRITE_RTR),
+			   !!(mpa_ord & MPA_V2_READ_RTR));
+	}
+}
+
+static struct qed_iwarp_ep *qed_iwarp_get_free_ep(struct qed_hwfn *p_hwfn)
+{
+	struct qed_iwarp_ep *ep = NULL;
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	if (list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) {
+		DP_ERR(p_hwfn, "Ep list is empty\n");
+		goto out;
+	}
+
+	ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list,
+			      struct qed_iwarp_ep, list_entry);
+
+	/* in some cases we could have failed allocating a tcp cid when added
+	 * from accept / failure... retry now..this is not the common case.
+	 */
+	if (ep->tcp_cid == QED_IWARP_INVALID_TCP_CID) {
+		rc = qed_iwarp_alloc_tcp_cid(p_hwfn, &ep->tcp_cid);
+
+		/* if we fail we could look for another entry with a valid
+		 * tcp_cid, but since we don't expect to reach this anyway
+		 * it's not worth the handling
+		 */
+		if (rc) {
+			ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+			ep = NULL;
+			goto out;
+		}
+	}
+
+	list_del(&ep->list_entry);
+
+out:
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	return ep;
+}
+
 #define QED_IWARP_MAX_CID_CLEAN_TIME  100
 #define QED_IWARP_MAX_NO_PROGRESS_CNT 5
 
@@ -465,20 +1057,213 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 
 static int qed_iwarp_wait_for_all_cids(struct qed_hwfn *p_hwfn)
 {
+	int rc;
+	int i;
+
+	rc = qed_iwarp_wait_cid_map_cleared(p_hwfn,
+					    &p_hwfn->p_rdma_info->tcp_cid_map);
+	if (rc)
+		return rc;
+
+	/* Now free the tcp cids from the main cid map */
+	for (i = 0; i < QED_IWARP_PREALLOC_CNT; i++)
+		qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, i);
+
 	/* Now wait for all cids to be completed */
 	return qed_iwarp_wait_cid_map_cleared(p_hwfn,
 					      &p_hwfn->p_rdma_info->cid_map);
 }
 
+static void qed_iwarp_free_prealloc_ep(struct qed_hwfn *p_hwfn)
+{
+	struct qed_iwarp_ep *ep;
+
+	while (!list_empty(&p_hwfn->p_rdma_info->iwarp.ep_free_list)) {
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+		ep = list_first_entry(&p_hwfn->p_rdma_info->iwarp.ep_free_list,
+				      struct qed_iwarp_ep, list_entry);
+
+		if (!ep) {
+			spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+			break;
+		}
+		list_del(&ep->list_entry);
+
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+		if (ep->tcp_cid != QED_IWARP_INVALID_TCP_CID)
+			qed_iwarp_cid_cleaned(p_hwfn, ep->tcp_cid);
+
+		qed_iwarp_destroy_ep(p_hwfn, ep, false);
+	}
+}
+
+static int qed_iwarp_prealloc_ep(struct qed_hwfn *p_hwfn, bool init)
+{
+	struct qed_iwarp_ep *ep;
+	int rc = 0;
+	int count;
+	u32 cid;
+	int i;
+
+	count = init ? QED_IWARP_PREALLOC_CNT : 1;
+	for (i = 0; i < count; i++) {
+		rc = qed_iwarp_create_ep(p_hwfn, &ep);
+		if (rc)
+			return rc;
+
+		/* During initialization we allocate from the main pool,
+		 * afterwards we allocate only from the tcp_cid.
+		 */
+		if (init) {
+			rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+			if (rc)
+				goto err;
+			qed_iwarp_set_tcp_cid(p_hwfn, cid);
+		} else {
+			/* We don't care about the return code, it's ok if
+			 * tcp_cid remains invalid...in this case we'll
+			 * defer allocation
+			 */
+			qed_iwarp_alloc_tcp_cid(p_hwfn, &cid);
+		}
+
+		ep->tcp_cid = cid;
+
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_add_tail(&ep->list_entry,
+			      &p_hwfn->p_rdma_info->iwarp.ep_free_list);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+
+	return rc;
+
+err:
+	qed_iwarp_destroy_ep(p_hwfn, ep, false);
+
+	return rc;
+}
+
 int qed_iwarp_alloc(struct qed_hwfn *p_hwfn)
 {
+	int rc;
+
+	/* Allocate bitmap for tcp cid. These are used by passive side
+	 * to ensure it can allocate a tcp cid during dpc that was
+	 * pre-acquired and doesn't require dynamic allocation of ilt
+	 */
+	rc = qed_rdma_bmap_alloc(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map,
+				 QED_IWARP_PREALLOC_CNT, "TCP_CID");
+	if (rc) {
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "Failed to allocate tcp cid, rc = %d\n", rc);
+		return rc;
+	}
+
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_free_list);
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.iw_lock);
 
-	return 0;
+	return qed_iwarp_prealloc_ep(p_hwfn, true);
 }
 
 void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
+	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->tcp_cid_map, 1);
+}
+
+int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+	int rc;
+
+	ep = (struct qed_iwarp_ep *)iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in accept is NULL\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n",
+		   iparams->qp->icid, ep->tcp_cid);
+
+	if ((iparams->ord > QED_IWARP_ORD_DEFAULT) ||
+	    (iparams->ird > QED_IWARP_IRD_DEFAULT)) {
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) EP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n",
+			   iparams->qp->icid,
+			   ep->tcp_cid, iparams->ord, iparams->ord);
+		return -EINVAL;
+	}
+
+	qed_iwarp_prealloc_ep(p_hwfn, false);
+
+	ep->cb_context = iparams->cb_context;
+	ep->qp = iparams->qp;
+	ep->qp->ep = ep;
+
+	if (ep->mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
+		/* Negotiate ord/ird: if upperlayer requested ord larger than
+		 * ird advertised by remote, we need to decrease our ord
+		 */
+		if (iparams->ord > ep->cm_info.ird)
+			iparams->ord = ep->cm_info.ird;
+
+		if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) &&
+		    (iparams->ird == 0))
+			iparams->ird = 1;
+	}
+
+	/* Update cm_info ord/ird to be negotiated values */
+	ep->cm_info.ord = iparams->ord;
+	ep->cm_info.ird = iparams->ird;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->private_data, iparams->private_data_len);
+
+	rc = qed_iwarp_mpa_offload(p_hwfn, ep);
+	if (rc)
+		qed_iwarp_modify_qp(p_hwfn,
+				    iparams->qp, QED_IWARP_QP_STATE_ERROR, 1);
+
+	return rc;
+}
+
+int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+
+	ep = iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in reject is NULL\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x)\n", ep->tcp_cid);
+
+	ep->cb_context = iparams->cb_context;
+	ep->qp = NULL;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->private_data, iparams->private_data_len);
+
+	return qed_iwarp_mpa_offload(p_hwfn, ep);
 }
 
 static void
@@ -526,6 +1311,38 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	return rc;
 }
 
+static bool
+qed_iwarp_ep_exists(struct qed_hwfn *p_hwfn, struct qed_iwarp_cm_info *cm_info)
+{
+	struct qed_iwarp_ep *ep = NULL;
+	bool found = false;
+
+	list_for_each_entry(ep,
+			    &p_hwfn->p_rdma_info->iwarp.ep_list,
+			    list_entry) {
+		if ((ep->cm_info.local_port == cm_info->local_port) &&
+		    (ep->cm_info.remote_port == cm_info->remote_port) &&
+		    (ep->cm_info.vlan == cm_info->vlan) &&
+		    !memcmp(&ep->cm_info.local_ip, cm_info->local_ip,
+			    sizeof(cm_info->local_ip)) &&
+		    !memcmp(&ep->cm_info.remote_ip, cm_info->remote_ip,
+			    sizeof(cm_info->remote_ip))) {
+			found = true;
+			break;
+		}
+	}
+
+	if (found) {
+		DP_NOTICE(p_hwfn,
+			  "SYN received on active connection - dropping\n");
+		qed_iwarp_print_cm_info(p_hwfn, cm_info);
+
+		return true;
+	}
+
+	return false;
+}
+
 static struct qed_iwarp_listener *
 qed_iwarp_get_listener(struct qed_hwfn *p_hwfn,
 		       struct qed_iwarp_cm_info *cm_info)
@@ -596,9 +1413,8 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 
 	eth_hlen = ETH_HLEN + (vlan_valid ? sizeof(u32) : 0);
 
-	memcpy(remote_mac_addr, ethh->h_source, ETH_ALEN);
-
-	memcpy(local_mac_addr, ethh->h_dest, ETH_ALEN);
+	ether_addr_copy(remote_mac_addr, ethh->h_source);
+	ether_addr_copy(local_mac_addr, ethh->h_dest);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "eth_type =%d source mac: %pM\n",
 		   eth_type, ethh->h_source);
@@ -661,9 +1477,12 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	struct qed_hwfn *p_hwfn = cxt;
 	u8 remote_mac_addr[ETH_ALEN];
 	u8 local_mac_addr[ETH_ALEN];
+	struct qed_iwarp_ep *ep;
 	int tcp_start_offset;
+	u8 ts_hdr_size = 0;
 	u8 ll2_syn_handle;
 	int payload_len;
+	u32 hdr_size;
 	int rc;
 
 	memset(&cm_info, 0, sizeof(cm_info));
@@ -719,6 +1538,49 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 	}
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Received syn on listening port\n");
+	/* There may be an open ep on this connection if this is a syn
+	 * retrasnmit... need to make sure there isn't...
+	 */
+	if (qed_iwarp_ep_exists(p_hwfn, &cm_info))
+		goto err;
+
+	ep = qed_iwarp_get_free_ep(p_hwfn);
+	if (!ep)
+		goto err;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ether_addr_copy(ep->remote_mac_addr, remote_mac_addr);
+	ether_addr_copy(ep->local_mac_addr, local_mac_addr);
+
+	memcpy(&ep->cm_info, &cm_info, sizeof(ep->cm_info));
+
+	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
+		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
+
+	hdr_size = ((cm_info.ip_version == QED_TCP_IPV4) ? 40 : 60) +
+		   ts_hdr_size;
+	ep->mss = p_hwfn->p_rdma_info->iwarp.max_mtu - hdr_size;
+	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
+
+	ep->event_cb = listener->event_cb;
+	ep->cb_context = listener->cb_context;
+	ep->connect_mode = TCP_CONNECT_PASSIVE;
+
+	ep->syn = buf;
+	ep->syn_ip_payload_length = (u16)payload_len;
+	ep->syn_phy_addr = buf->data_phys_addr + data->u.placement_offset +
+			   tcp_start_offset;
+
+	rc = qed_iwarp_tcp_offload(p_hwfn, ep);
+	if (rc) {
+		qed_iwarp_return_ep(p_hwfn, ep);
+		goto err;
+	}
+
+	return;
 err:
 	qed_iwarp_ll2_post_rx(p_hwfn, buf, ll2_syn_handle);
 }
@@ -905,7 +1767,12 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 
 	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
 
+	iwarp_info->rtr_type =  MPA_RTR_TYPE_ZERO_SEND |
+				MPA_RTR_TYPE_ZERO_WRITE |
+				MPA_RTR_TYPE_ZERO_READ;
+
 	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.ep_list);
 	INIT_LIST_HEAD(&p_hwfn->p_rdma_info->iwarp.listen_list);
 
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
@@ -918,6 +1785,7 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 {
 	int rc;
 
+	qed_iwarp_free_prealloc_ep(p_hwfn);
 	rc = qed_iwarp_wait_for_all_cids(p_hwfn);
 	if (rc)
 		return rc;
@@ -927,11 +1795,70 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
+void
+qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
+			   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	/* Done with the SYN packet, post back to ll2 rx */
+	qed_iwarp_ll2_post_rx(p_hwfn, ep->syn,
+			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+	ep->syn = NULL;
+
+	/* If connect failed - upper layer doesn't know about it */
+	qed_iwarp_mpa_received(p_hwfn, ep);
+}
+
+static inline bool
+qed_iwarp_check_ep_ok(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	if (!ep || (ep->sig != QED_EP_SIG)) {
+		DP_ERR(p_hwfn, "ERROR ON ASYNC ep=%p\n", ep);
+		return false;
+	}
+
+	return true;
+}
+
 static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 				 u8 fw_event_code, u16 echo,
 				 union event_ring_data *data,
 				 u8 fw_return_code)
 {
+	struct regpair *fw_handle = &data->rdma_data.async_handle;
+	struct qed_iwarp_ep *ep = NULL;
+	u16 cid;
+
+	ep = (struct qed_iwarp_ep *)(uintptr_t)HILO_64(fw_handle->hi,
+						       fw_handle->lo);
+
+	switch (fw_event_code) {
+	case IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE:
+		/* Async completion after TCP 3-way handshake */
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "EP(0x%x) IWARP_EVENT_TYPE_ASYNC_CONNECT_COMPLETE fw_ret_code=%d\n",
+			   ep->tcp_cid, fw_return_code);
+		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_mpa_complete(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_CID_CLEANED:
+		cid = (u16)le32_to_cpu(fw_handle->lo);
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "(0x%x)IWARP_EVENT_TYPE_ASYNC_CID_CLEANED\n", cid);
+		qed_iwarp_cid_cleaned(p_hwfn, cid);
+
+		break;
+	}
 	return 0;
 }
 
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 29005ac..bedac98 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -42,6 +42,8 @@ enum qed_iwarp_qp_state {
 
 enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
 
+#define QED_IWARP_PREALLOC_CNT  (256)
+
 #define QED_IWARP_LL2_SYN_TX_SIZE       (128)
 #define QED_IWARP_LL2_SYN_RX_SIZE       (256)
 #define QED_IWARP_MAX_SYN_PKT_SIZE      (128)
@@ -55,6 +57,8 @@ struct qed_iwarp_ll2_buff {
 
 struct qed_iwarp_info {
 	struct list_head listen_list;	/* qed_iwarp_listener */
+	struct list_head ep_list;	/* qed_iwarp_ep */
+	struct list_head ep_free_list;	/* pre-allocated ep's */
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
 	u32 rcv_wnd_scale;
@@ -68,6 +72,61 @@ struct qed_iwarp_info {
 	enum mpa_rtr_type rtr_type;
 };
 
+enum qed_iwarp_ep_state {
+	QED_IWARP_EP_INIT,
+	QED_IWARP_EP_MPA_REQ_RCVD,
+	QED_IWARP_EP_MPA_OFFLOADED,
+	QED_IWARP_EP_ESTABLISHED,
+	QED_IWARP_EP_CLOSED
+};
+
+union async_output {
+	struct iwarp_eqe_data_mpa_async_completion mpa_response;
+	struct iwarp_eqe_data_tcp_async_completion mpa_request;
+};
+
+#define QED_MAX_PRIV_DATA_LEN (512)
+struct qed_iwarp_ep_memory {
+	u8 in_pdata[QED_MAX_PRIV_DATA_LEN];
+	u8 out_pdata[QED_MAX_PRIV_DATA_LEN];
+	union async_output async_output;
+};
+
+/* Endpoint structure represents a TCP connection. This connection can be
+ * associated with a QP or not (in which case QP==NULL)
+ */
+struct qed_iwarp_ep {
+	struct list_head list_entry;
+	struct qed_rdma_qp *qp;
+	struct qed_iwarp_ep_memory *ep_buffer_virt;
+	dma_addr_t ep_buffer_phys;
+	enum qed_iwarp_ep_state state;
+	int sig;
+	struct qed_iwarp_cm_info cm_info;
+	enum tcp_connect_mode connect_mode;
+	enum mpa_rtr_type rtr_type;
+	enum mpa_negotiation_mode mpa_rev;
+	u32 tcp_cid;
+	u32 cid;
+	u16 mss;
+	u8 remote_mac_addr[6];
+	u8 local_mac_addr[6];
+	bool mpa_reply_processed;
+
+	/* For Passive side - syn packet related data */
+	u16 syn_ip_payload_length;
+	struct qed_iwarp_ll2_buff *syn;
+	dma_addr_t syn_phy_addr;
+
+	/* The event_cb function is called for asynchrounous events associated
+	 * with the ep. It is initialized at different entry points depending
+	 * on whether the ep is the tcp connection active side or passive side
+	 * The cb_context is passed to the event_cb function.
+	 */
+	iwarp_event_handler event_cb;
+	void *cb_context;
+};
+
 struct qed_iwarp_listener {
 	struct list_head list_entry;
 
@@ -115,6 +174,9 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_iwarp_listen_in *iparams,
 			struct qed_iwarp_listen_out *oparams);
 
+int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams);
+
+int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams);
 int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
 
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_l2.c b/drivers/net/ethernet/qlogic/qed/qed_l2.c
index 27ea54b..0ba5ec8 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_l2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_l2.c
@@ -1227,19 +1227,6 @@ static enum eth_filter_action qed_filter_action(enum qed_filter_opcode opcode)
 	return action;
 }
 
-static void qed_set_fw_mac_addr(__le16 *fw_msb,
-				__le16 *fw_mid,
-				__le16 *fw_lsb,
-				u8 *mac)
-{
-	((u8 *)fw_msb)[0] = mac[1];
-	((u8 *)fw_msb)[1] = mac[0];
-	((u8 *)fw_mid)[0] = mac[3];
-	((u8 *)fw_mid)[1] = mac[2];
-	((u8 *)fw_lsb)[0] = mac[5];
-	((u8 *)fw_lsb)[1] = mac[4];
-}
-
 static int
 qed_filter_ucast_common(struct qed_hwfn *p_hwfn,
 			u16 opaque_fid,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.h b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
index 90e4e0f..18ec9cb 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
@@ -85,6 +85,7 @@ struct qed_rdma_info {
 	struct qed_bmap qp_map;
 	struct qed_bmap srq_map;
 	struct qed_bmap cid_map;
+	struct qed_bmap tcp_cid_map;
 	struct qed_bmap real_cid_map;
 	struct qed_bmap dpi_map;
 	struct qed_bmap toggle_bits;
@@ -167,6 +168,7 @@ struct qed_rdma_qp {
 
 	void *shared_queue;
 	dma_addr_t shared_queue_phys_addr;
+	struct qed_iwarp_ep *ep;
 };
 
 #if IS_ENABLED(CONFIG_QED_RDMA)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sp.h b/drivers/net/ethernet/qlogic/qed/qed_sp.h
index c3752c5..ab4ad8a 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_sp.h
@@ -111,6 +111,8 @@ int qed_eth_cqe_completion(struct qed_hwfn *p_hwfn,
 	struct rdma_srq_destroy_ramrod_data rdma_destroy_srq;
 	struct rdma_srq_modify_ramrod_data rdma_modify_srq;
 	struct iwarp_create_qp_ramrod_data iwarp_create_qp;
+	struct iwarp_tcp_offload_ramrod_data iwarp_tcp_offload;
+	struct iwarp_mpa_offload_ramrod_data iwarp_mpa_offload;
 	struct iwarp_modify_qp_ramrod_data iwarp_modify_qp;
 	struct iwarp_init_func_ramrod_data iwarp_init_func;
 	struct fcoe_init_ramrod_params fcoe_init;
diff --git a/include/linux/qed/common_hsi.h b/include/linux/qed/common_hsi.h
index 885ae13..39e2a2a 100644
--- a/include/linux/qed/common_hsi.h
+++ b/include/linux/qed/common_hsi.h
@@ -38,6 +38,8 @@
 #include <linux/slab.h>
 
 /* dma_addr_t manip */
+#define PTR_LO(x)               ((u32)(((uintptr_t)(x)) & 0xffffffff))
+#define PTR_HI(x)               ((u32)((((uintptr_t)(x)) >> 16) >> 16))
 #define DMA_LO_LE(x)		cpu_to_le32(lower_32_bits(x))
 #define DMA_HI_LE(x)		cpu_to_le32(upper_32_bits(x))
 #define DMA_REGPAIR_LE(x, val)	do { \
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 28df568..c4c241f 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -471,7 +471,8 @@ struct qed_rdma_counters_out_params {
 #define QED_ROCE_TX_FRAG_FAILURE        (2)
 
 enum qed_iwarp_event_type {
-	QED_IWARP_EVENT_MPA_REQUEST,	/* Passive side request received */
+	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
+	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
 };
 
 enum qed_tcp_ip_version {
@@ -516,6 +517,23 @@ struct qed_iwarp_listen_out {
 	void *handle;
 };
 
+struct qed_iwarp_accept_in {
+	void *ep_context;
+	void *cb_context;
+	struct qed_rdma_qp *qp;
+	const void *private_data;
+	u16 private_data_len;
+	u8 ord;
+	u8 ird;
+};
+
+struct qed_iwarp_reject_in {
+	void *ep_context;
+	void *cb_context;
+	const void *private_data;
+	u16 private_data_len;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -626,6 +644,12 @@ struct qed_rdma_ops {
 				   struct qed_iwarp_listen_in *iparams,
 				   struct qed_iwarp_listen_out *oparams);
 
+	int (*iwarp_accept)(void *rdma_cxt,
+			    struct qed_iwarp_accept_in *iparams);
+
+	int (*iwarp_reject)(void *rdma_cxt,
+			    struct qed_iwarp_reject_in *iparams);
+
 	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
 
 };
-- 
1.8.3.1

[PATCH net-next 10/12] qed: iWARP CM add error handling

[Michal Kalderon]

This patch introduces error handling for errors that occurred during
connection establishment.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 183 +++++++++++++++++++++++++++-
 include/linux/qed/qed_rdma_if.h             |   9 ++
 2 files changed, 190 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index 84bcda3..5cd20da 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -1001,12 +1001,75 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 		ep->state = QED_IWARP_EP_ESTABLISHED;
 		params.status = 0;
 		break;
+	case IWARP_CONN_ERROR_MPA_TIMEOUT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA timeout\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -EBUSY;
+		break;
+	case IWARP_CONN_ERROR_MPA_ERROR_REJECT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Reject\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_RST:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA reset(tcp cid: 0x%x)\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid,
+			  ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_MPA_FIN:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA received FIN\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INSUF_IRD:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA insufficient ird\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_RTR_MISMATCH:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA RTR MISMATCH\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INVALID_PACKET:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_LOCAL_ERROR:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Local Error\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_TERMINATE:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA TERMINATE\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
 	default:
 		params.status = -ECONNRESET;
 		break;
 	}
 
 	ep->event_cb(ep->cb_context, &params);
+
+	/* on passive side, if there is no associated QP (REJECT) we need to
+	 * return the ep to the pool, (in the regular case we add an element
+	 * in accept instead of this one.
+	 * In both cases we need to remove it from the ep_list.
+	 */
+	if (fw_return_code != RDMA_RETURN_OK) {
+		ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+		if ((ep->connect_mode == TCP_CONNECT_PASSIVE) &&
+		    (!ep->qp)) {	/* Rejected */
+			qed_iwarp_return_ep(p_hwfn, ep);
+		} else {
+			spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+			list_del(&ep->list_entry);
+			spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		}
+	}
 }
 
 static void
@@ -2011,6 +2074,42 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 		params.event = QED_IWARP_EVENT_DISCONNECT;
 		event_cb = true;
 		break;
+	case IWARP_EXCEPTION_DETECTED_RQ_EMPTY:
+		params.event = QED_IWARP_EVENT_RQ_EMPTY;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_IRQ_FULL:
+		params.event = QED_IWARP_EVENT_IRQ_FULL;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LLP_TIMEOUT:
+		params.event = QED_IWARP_EVENT_LLP_TIMEOUT;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_REMOTE_PROTECTION_ERROR:
+		params.event = QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_CQ_OVERFLOW:
+		params.event = QED_IWARP_EVENT_CQ_OVERFLOW;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LOCAL_CATASTROPHIC:
+		params.event = QED_IWARP_EVENT_QP_CATASTROPHIC;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LOCAL_ACCESS_ERROR:
+		params.event = QED_IWARP_EVENT_LOCAL_ACCESS_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_REMOTE_OPERATION_ERROR:
+		params.event = QED_IWARP_EVENT_REMOTE_OPERATION_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_TERMINATE_RECEIVED:
+		params.event = QED_IWARP_EVENT_TERMINATE_RECEIVED;
+		event_cb = true;
+		break;
 	default:
 		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 			   "Unhandled exception received...fw_ret_code=%d\n",
@@ -2025,6 +2124,66 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 	}
 }
 
+static void
+qed_iwarp_tcp_connect_unsuccessful(struct qed_hwfn *p_hwfn,
+				   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	memset(&params, 0, sizeof(params));
+	params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE;
+	params.ep_context = ep;
+	params.cm_info = &ep->cm_info;
+	ep->state = QED_IWARP_EP_CLOSED;
+
+	switch (fw_return_code) {
+	case IWARP_CONN_ERROR_TCP_CONNECT_INVALID_PACKET:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "%s(0x%x) TCP connect got invalid packet\n",
+			   QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_TCP_CONNECTION_RST:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "%s(0x%x) TCP Connection Reset\n",
+			   QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_TCP_CONNECT_TIMEOUT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) TCP timeout\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -EBUSY;
+		break;
+	case IWARP_CONN_ERROR_MPA_NOT_SUPPORTED_VER:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA not supported VER\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INVALID_PACKET:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	default:
+		DP_ERR(p_hwfn,
+		       "%s(0x%x) Unexpected return code tcp connect: %d\n",
+		       QED_IWARP_CONNECT_MODE_STRING(ep),
+		       ep->tcp_cid, fw_return_code);
+		params.status = -ECONNRESET;
+		break;
+	}
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+		qed_iwarp_return_ep(p_hwfn, ep);
+	} else {
+		ep->event_cb(ep->cb_context, &params);
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_del(&ep->list_entry);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+}
+
 void
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
@@ -2038,9 +2197,17 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 		ep->syn = NULL;
 
 		/* If connect failed - upper layer doesn't know about it */
-		qed_iwarp_mpa_received(p_hwfn, ep);
+		if (fw_return_code == RDMA_RETURN_OK)
+			qed_iwarp_mpa_received(p_hwfn, ep);
+		else
+			qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep,
+							   fw_return_code);
 	} else {
-		qed_iwarp_mpa_offload(p_hwfn, ep);
+		if (fw_return_code == RDMA_RETURN_OK)
+			qed_iwarp_mpa_offload(p_hwfn, ep);
+		else
+			qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep,
+							   fw_return_code);
 	}
 }
 
@@ -2123,6 +2290,18 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 		qed_iwarp_cid_cleaned(p_hwfn, cid);
 
 		break;
+	case IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW:
+		DP_NOTICE(p_hwfn, "IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW\n");
+
+		p_hwfn->p_rdma_info->events.affiliated_event(
+			p_hwfn->p_rdma_info->events.context,
+			QED_IWARP_EVENT_CQ_OVERFLOW,
+			(void *)fw_handle);
+		break;
+	default:
+		DP_ERR(p_hwfn, "Received unexpected async iwarp event %d\n",
+		       fw_event_code);
+		return -EINVAL;
 	}
 	return 0;
 }
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 01966c3..4dd72ba 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -476,7 +476,16 @@ enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
 	QED_IWARP_EVENT_DISCONNECT,
 	QED_IWARP_EVENT_CLOSE,
+	QED_IWARP_EVENT_IRQ_FULL,
+	QED_IWARP_EVENT_RQ_EMPTY,
+	QED_IWARP_EVENT_LLP_TIMEOUT,
+	QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR,
+	QED_IWARP_EVENT_CQ_OVERFLOW,
+	QED_IWARP_EVENT_QP_CATASTROPHIC,
 	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
+	QED_IWARP_EVENT_LOCAL_ACCESS_ERROR,
+	QED_IWARP_EVENT_REMOTE_OPERATION_ERROR,
+	QED_IWARP_EVENT_TERMINATE_RECEIVED
 };
 
 enum qed_tcp_ip_version {
-- 
1.8.3.1

[PATCH net-next 10/12] qed: iWARP CM add error handling

[Michal Kalderon]

This patch introduces error handling for errors that occurred during
connection establishment.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 183 +++++++++++++++++++++++++++-
 include/linux/qed/qed_rdma_if.h             |   9 ++
 2 files changed, 190 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index 84bcda3..5cd20da 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -1001,12 +1001,75 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 		ep->state = QED_IWARP_EP_ESTABLISHED;
 		params.status = 0;
 		break;
+	case IWARP_CONN_ERROR_MPA_TIMEOUT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA timeout\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -EBUSY;
+		break;
+	case IWARP_CONN_ERROR_MPA_ERROR_REJECT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Reject\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_RST:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA reset(tcp cid: 0x%x)\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid,
+			  ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_MPA_FIN:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA received FIN\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INSUF_IRD:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA insufficient ird\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_RTR_MISMATCH:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA RTR MISMATCH\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INVALID_PACKET:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_LOCAL_ERROR:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Local Error\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_TERMINATE:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA TERMINATE\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->cid);
+		params.status = -ECONNREFUSED;
+		break;
 	default:
 		params.status = -ECONNRESET;
 		break;
 	}
 
 	ep->event_cb(ep->cb_context, &params);
+
+	/* on passive side, if there is no associated QP (REJECT) we need to
+	 * return the ep to the pool, (in the regular case we add an element
+	 * in accept instead of this one.
+	 * In both cases we need to remove it from the ep_list.
+	 */
+	if (fw_return_code != RDMA_RETURN_OK) {
+		ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+		if ((ep->connect_mode == TCP_CONNECT_PASSIVE) &&
+		    (!ep->qp)) {	/* Rejected */
+			qed_iwarp_return_ep(p_hwfn, ep);
+		} else {
+			spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+			list_del(&ep->list_entry);
+			spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		}
+	}
 }
 
 static void
@@ -2011,6 +2074,42 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 		params.event = QED_IWARP_EVENT_DISCONNECT;
 		event_cb = true;
 		break;
+	case IWARP_EXCEPTION_DETECTED_RQ_EMPTY:
+		params.event = QED_IWARP_EVENT_RQ_EMPTY;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_IRQ_FULL:
+		params.event = QED_IWARP_EVENT_IRQ_FULL;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LLP_TIMEOUT:
+		params.event = QED_IWARP_EVENT_LLP_TIMEOUT;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_REMOTE_PROTECTION_ERROR:
+		params.event = QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_CQ_OVERFLOW:
+		params.event = QED_IWARP_EVENT_CQ_OVERFLOW;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LOCAL_CATASTROPHIC:
+		params.event = QED_IWARP_EVENT_QP_CATASTROPHIC;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LOCAL_ACCESS_ERROR:
+		params.event = QED_IWARP_EVENT_LOCAL_ACCESS_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_REMOTE_OPERATION_ERROR:
+		params.event = QED_IWARP_EVENT_REMOTE_OPERATION_ERROR;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_TERMINATE_RECEIVED:
+		params.event = QED_IWARP_EVENT_TERMINATE_RECEIVED;
+		event_cb = true;
+		break;
 	default:
 		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 			   "Unhandled exception received...fw_ret_code=%d\n",
@@ -2025,6 +2124,66 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 	}
 }
 
+static void
+qed_iwarp_tcp_connect_unsuccessful(struct qed_hwfn *p_hwfn,
+				   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	memset(&params, 0, sizeof(params));
+	params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE;
+	params.ep_context = ep;
+	params.cm_info = &ep->cm_info;
+	ep->state = QED_IWARP_EP_CLOSED;
+
+	switch (fw_return_code) {
+	case IWARP_CONN_ERROR_TCP_CONNECT_INVALID_PACKET:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "%s(0x%x) TCP connect got invalid packet\n",
+			   QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_TCP_CONNECTION_RST:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "%s(0x%x) TCP Connection Reset\n",
+			   QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	case IWARP_CONN_ERROR_TCP_CONNECT_TIMEOUT:
+		DP_NOTICE(p_hwfn, "%s(0x%x) TCP timeout\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -EBUSY;
+		break;
+	case IWARP_CONN_ERROR_MPA_NOT_SUPPORTED_VER:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA not supported VER\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNREFUSED;
+		break;
+	case IWARP_CONN_ERROR_MPA_INVALID_PACKET:
+		DP_NOTICE(p_hwfn, "%s(0x%x) MPA Invalid Packet\n",
+			  QED_IWARP_CONNECT_MODE_STRING(ep), ep->tcp_cid);
+		params.status = -ECONNRESET;
+		break;
+	default:
+		DP_ERR(p_hwfn,
+		       "%s(0x%x) Unexpected return code tcp connect: %d\n",
+		       QED_IWARP_CONNECT_MODE_STRING(ep),
+		       ep->tcp_cid, fw_return_code);
+		params.status = -ECONNRESET;
+		break;
+	}
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		ep->tcp_cid = QED_IWARP_INVALID_TCP_CID;
+		qed_iwarp_return_ep(p_hwfn, ep);
+	} else {
+		ep->event_cb(ep->cb_context, &params);
+		spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+		list_del(&ep->list_entry);
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	}
+}
+
 void
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
@@ -2038,9 +2197,17 @@ void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
 		ep->syn = NULL;
 
 		/* If connect failed - upper layer doesn't know about it */
-		qed_iwarp_mpa_received(p_hwfn, ep);
+		if (fw_return_code == RDMA_RETURN_OK)
+			qed_iwarp_mpa_received(p_hwfn, ep);
+		else
+			qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep,
+							   fw_return_code);
 	} else {
-		qed_iwarp_mpa_offload(p_hwfn, ep);
+		if (fw_return_code == RDMA_RETURN_OK)
+			qed_iwarp_mpa_offload(p_hwfn, ep);
+		else
+			qed_iwarp_tcp_connect_unsuccessful(p_hwfn, ep,
+							   fw_return_code);
 	}
 }
 
@@ -2123,6 +2290,18 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 		qed_iwarp_cid_cleaned(p_hwfn, cid);
 
 		break;
+	case IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW:
+		DP_NOTICE(p_hwfn, "IWARP_EVENT_TYPE_ASYNC_CQ_OVERFLOW\n");
+
+		p_hwfn->p_rdma_info->events.affiliated_event(
+			p_hwfn->p_rdma_info->events.context,
+			QED_IWARP_EVENT_CQ_OVERFLOW,
+			(void *)fw_handle);
+		break;
+	default:
+		DP_ERR(p_hwfn, "Received unexpected async iwarp event %d\n",
+		       fw_event_code);
+		return -EINVAL;
 	}
 	return 0;
 }
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index 01966c3..4dd72ba 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -476,7 +476,16 @@ enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
 	QED_IWARP_EVENT_DISCONNECT,
 	QED_IWARP_EVENT_CLOSE,
+	QED_IWARP_EVENT_IRQ_FULL,
+	QED_IWARP_EVENT_RQ_EMPTY,
+	QED_IWARP_EVENT_LLP_TIMEOUT,
+	QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR,
+	QED_IWARP_EVENT_CQ_OVERFLOW,
+	QED_IWARP_EVENT_QP_CATASTROPHIC,
 	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
+	QED_IWARP_EVENT_LOCAL_ACCESS_ERROR,
+	QED_IWARP_EVENT_REMOTE_OPERATION_ERROR,
+	QED_IWARP_EVENT_TERMINATE_RECEIVED
 };
 
 enum qed_tcp_ip_version {
-- 
1.8.3.1

[PATCH net-next 01/12] qed: Introduce iWARP personality

[Michal Kalderon]

iWARP personality introduced the need for differentiating in several
places in the code whether we are RoCE, iWARP or either. This
leads to introducing new macros for querying the personality.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed.h      | 26 +++++++++++++++++++++++---
 drivers/net/ethernet/qlogic/qed/qed_cxt.c  |  8 ++++----
 drivers/net/ethernet/qlogic/qed/qed_dev.c  | 12 +++++-------
 drivers/net/ethernet/qlogic/qed/qed_l2.c   |  3 +--
 drivers/net/ethernet/qlogic/qed/qed_ll2.c  |  2 +-
 drivers/net/ethernet/qlogic/qed/qed_main.c | 17 ++++++++---------
 include/linux/qed/common_hsi.h             |  2 +-
 7 files changed, 43 insertions(+), 27 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 14b08ee..22e1171 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -210,14 +210,16 @@ struct qed_tunn_update_params {
 
 /* The PCI personality is not quite synonymous to protocol ID:
  * 1. All personalities need CORE connections
- * 2. The Ethernet personality may support also the RoCE protocol
+ * 2. The Ethernet personality may support also the RoCE/iWARP protocol
  */
 enum qed_pci_personality {
 	QED_PCI_ETH,
 	QED_PCI_FCOE,
 	QED_PCI_ISCSI,
 	QED_PCI_ETH_ROCE,
-	QED_PCI_DEFAULT /* default in shmem */
+	QED_PCI_ETH_IWARP,
+	QED_PCI_ETH_RDMA,
+	QED_PCI_DEFAULT, /* default in shmem */
 };
 
 /* All VFs are symmetric, all counters are PF + all VFs */
@@ -277,6 +279,7 @@ enum qed_dev_cap {
 	QED_DEV_CAP_FCOE,
 	QED_DEV_CAP_ISCSI,
 	QED_DEV_CAP_ROCE,
+	QED_DEV_CAP_IWARP,
 };
 
 enum qed_wol_support {
@@ -286,7 +289,24 @@ enum qed_wol_support {
 
 struct qed_hw_info {
 	/* PCI personality */
-	enum qed_pci_personality	personality;
+	enum qed_pci_personality personality;
+#define QED_IS_RDMA_PERSONALITY(dev)			    \
+	((dev)->hw_info.personality == QED_PCI_ETH_ROCE ||  \
+	 (dev)->hw_info.personality == QED_PCI_ETH_IWARP || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_ROCE_PERSONALITY(dev)			   \
+	((dev)->hw_info.personality == QED_PCI_ETH_ROCE || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_IWARP_PERSONALITY(dev)			    \
+	((dev)->hw_info.personality == QED_PCI_ETH_IWARP || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_L2_PERSONALITY(dev)		      \
+	((dev)->hw_info.personality == QED_PCI_ETH || \
+	 QED_IS_RDMA_PERSONALITY(dev))
+#define QED_IS_FCOE_PERSONALITY(dev) \
+	((dev)->hw_info.personality == QED_PCI_FCOE)
+#define QED_IS_ISCSI_PERSONALITY(dev) \
+	((dev)->hw_info.personality == QED_PCI_ISCSI)
 
 	/* Resource Allocation scheme results */
 	u32				resc_start[QED_MAX_RESC];
diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index e201214..38716f7 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
@@ -853,7 +853,7 @@ u32 qed_cxt_cfg_ilt_compute_excess(struct qed_hwfn *p_hwfn, u32 used_lines)
 	if (!excess_lines)
 		return 0;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_RDMA_PERSONALITY(p_hwfn))
 		return 0;
 
 	p_mngr = p_hwfn->p_cxt_mngr;
@@ -1033,7 +1033,7 @@ static int qed_ilt_blk_alloc(struct qed_hwfn *p_hwfn,
 	u32 lines, line, sz_left, lines_to_skip = 0;
 
 	/* Special handling for RoCE that supports dynamic allocation */
-	if ((p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) &&
+	if (QED_IS_RDMA_PERSONALITY(p_hwfn) &&
 	    ((ilt_client == ILT_CLI_CDUT) || ilt_client == ILT_CLI_TSDM))
 		return 0;
 
@@ -1833,7 +1833,7 @@ static void qed_tm_init_pf(struct qed_hwfn *p_hwfn)
 		tm_offset += tm_iids.pf_tids[i];
 	}
 
-	if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE)
+	if (QED_IS_RDMA_PERSONALITY(p_hwfn))
 		active_seg_mask = 0;
 
 	STORE_RT_REG(p_hwfn, TM_REG_PF_ENABLE_TASK_RT_OFFSET, active_seg_mask);
@@ -2344,7 +2344,7 @@ int qed_cxt_get_tid_mem_info(struct qed_hwfn *p_hwfn,
 		       last_cid_allocated - 1);
 
 		if (!p_hwfn->b_rdma_enabled_in_prs) {
-			/* Enable RoCE search */
+			/* Enable RDMA search */
 			qed_wr(p_hwfn, p_ptt, p_hwfn->rdma_prs_search_reg, 1);
 			p_hwfn->b_rdma_enabled_in_prs = true;
 		}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 49667ad..68e6182 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -936,7 +936,7 @@ int qed_resc_alloc(struct qed_dev *cdev)
 
 		/* EQ */
 		n_eqes = qed_chain_get_capacity(&p_hwfn->p_spq->chain);
-		if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) {
+		if (QED_IS_RDMA_PERSONALITY(p_hwfn)) {
 			num_cons = qed_cxt_get_proto_cid_count(p_hwfn,
 							       PROTOCOLID_ROCE,
 							       NULL) * 2;
@@ -2057,7 +2057,7 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 	qed_int_get_num_sbs(p_hwfn, &sb_cnt);
 
 	if (IS_ENABLED(CONFIG_QED_RDMA) &&
-	    p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) {
+	    QED_IS_RDMA_PERSONALITY(p_hwfn)) {
 		/* Roce CNQ each requires: 1 status block + 1 CNQ. We divide
 		 * the status blocks equally between L2 / RoCE but with
 		 * consideration as to how many l2 queues / cnqs we have.
@@ -2068,9 +2068,7 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 
 		non_l2_sbs = feat_num[QED_RDMA_CNQ];
 	}
-
-	if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE ||
-	    p_hwfn->hw_info.personality == QED_PCI_ETH) {
+	if (QED_IS_L2_PERSONALITY(p_hwfn)) {
 		/* Start by allocating VF queues, then PF's */
 		feat_num[QED_VF_L2_QUE] = min_t(u32,
 						RESC_NUM(p_hwfn, QED_L2_QUEUE),
@@ -2083,12 +2081,12 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 							 QED_VF_L2_QUE));
 	}
 
-	if (p_hwfn->hw_info.personality == QED_PCI_FCOE)
+	if (QED_IS_FCOE_PERSONALITY(p_hwfn))
 		feat_num[QED_FCOE_CQ] =  min_t(u32, sb_cnt.cnt,
 					       RESC_NUM(p_hwfn,
 							QED_CMDQS_CQS));
 
-	if (p_hwfn->hw_info.personality == QED_PCI_ISCSI)
+	if (QED_IS_ISCSI_PERSONALITY(p_hwfn))
 		feat_num[QED_ISCSI_CQ] = min_t(u32, sb_cnt.cnt,
 					       RESC_NUM(p_hwfn,
 							QED_CMDQS_CQS));
diff --git a/drivers/net/ethernet/qlogic/qed/qed_l2.c b/drivers/net/ethernet/qlogic/qed/qed_l2.c
index e57699b..27ea54b 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_l2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_l2.c
@@ -79,8 +79,7 @@ int qed_l2_alloc(struct qed_hwfn *p_hwfn)
 	unsigned long **pp_qids;
 	u32 i;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH &&
-	    p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_L2_PERSONALITY(p_hwfn))
 		return 0;
 
 	p_l2_info = kzalloc(sizeof(*p_l2_info), GFP_KERNEL);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index 17f9b0a..be66f19 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -1421,7 +1421,7 @@ int qed_ll2_establish_connection(void *cxt, u8 connection_handle)
 	if (rc)
 		goto out;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_RDMA_PERSONALITY(p_hwfn))
 		qed_wr(p_hwfn, p_ptt, PRS_REG_USE_LIGHT_L2, 1);
 
 	qed_ll2_establish_connection_ooo(p_hwfn, p_ll2_conn);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_main.c b/drivers/net/ethernet/qlogic/qed/qed_main.c
index 16cc30b..b113996 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_main.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_main.c
@@ -237,6 +237,8 @@ static int qed_init_pci(struct qed_dev *cdev, struct pci_dev *pdev)
 int qed_fill_dev_info(struct qed_dev *cdev,
 		      struct qed_dev_info *dev_info)
 {
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_hw_info *hw_info = &p_hwfn->hw_info;
 	struct qed_tunnel_info *tun = &cdev->tunnel;
 	struct qed_ptt  *ptt;
 
@@ -260,11 +262,10 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 	dev_info->pci_mem_start = cdev->pci_params.mem_start;
 	dev_info->pci_mem_end = cdev->pci_params.mem_end;
 	dev_info->pci_irq = cdev->pci_params.irq;
-	dev_info->rdma_supported = (cdev->hwfns[0].hw_info.personality ==
-				    QED_PCI_ETH_ROCE);
+	dev_info->rdma_supported = QED_IS_RDMA_PERSONALITY(p_hwfn);
 	dev_info->is_mf_default = IS_MF_DEFAULT(&cdev->hwfns[0]);
 	dev_info->dev_type = cdev->type;
-	ether_addr_copy(dev_info->hw_mac, cdev->hwfns[0].hw_info.hw_mac_addr);
+	ether_addr_copy(dev_info->hw_mac, hw_info->hw_mac_addr);
 
 	if (IS_PF(cdev)) {
 		dev_info->fw_major = FW_MAJOR_VERSION;
@@ -274,8 +275,7 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 		dev_info->mf_mode = cdev->mf_mode;
 		dev_info->tx_switching = true;
 
-		if (QED_LEADING_HWFN(cdev)->hw_info.b_wol_support ==
-		    QED_WOL_SUPPORT_PME)
+		if (hw_info->b_wol_support == QED_WOL_SUPPORT_PME)
 			dev_info->wol_support = true;
 
 		dev_info->abs_pf_id = QED_LEADING_HWFN(cdev)->abs_pf_id;
@@ -304,7 +304,7 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 				    &dev_info->mfw_rev, NULL);
 	}
 
-	dev_info->mtu = QED_LEADING_HWFN(cdev)->hw_info.mtu;
+	dev_info->mtu = hw_info->mtu;
 
 	return 0;
 }
@@ -790,7 +790,7 @@ static int qed_slowpath_setup_int(struct qed_dev *cdev,
 				       cdev->num_hwfns;
 
 	if (!IS_ENABLED(CONFIG_QED_RDMA) ||
-	    QED_LEADING_HWFN(cdev)->hw_info.personality != QED_PCI_ETH_ROCE)
+	    !QED_IS_RDMA_PERSONALITY(QED_LEADING_HWFN(cdev)))
 		return 0;
 
 	for_each_hwfn(cdev, i)
@@ -931,8 +931,7 @@ static void qed_update_pf_params(struct qed_dev *cdev,
 	/* In case we might support RDMA, don't allow qede to be greedy
 	 * with the L2 contexts. Allow for 64 queues [rx, tx, xdp] per hwfn.
 	 */
-	if (QED_LEADING_HWFN(cdev)->hw_info.personality ==
-	    QED_PCI_ETH_ROCE) {
+	if (QED_IS_RDMA_PERSONALITY(QED_LEADING_HWFN(cdev))) {
 		u16 *num_cons;
 
 		num_cons = &params->eth_pf_params.num_cons;
diff --git a/include/linux/qed/common_hsi.h b/include/linux/qed/common_hsi.h
index a567cbf..885ae13 100644
--- a/include/linux/qed/common_hsi.h
+++ b/include/linux/qed/common_hsi.h
@@ -778,7 +778,7 @@ enum protocol_type {
 	PROTOCOLID_ROCE,
 	PROTOCOLID_CORE,
 	PROTOCOLID_ETH,
-	PROTOCOLID_RESERVED4,
+	PROTOCOLID_IWARP,
 	PROTOCOLID_RESERVED5,
 	PROTOCOLID_PREROCE,
 	PROTOCOLID_COMMON,
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 01/12] qed: Introduce iWARP personality

[Michal Kalderon]

iWARP personality introduced the need for differentiating in several
places in the code whether we are RoCE, iWARP or either. This
leads to introducing new macros for querying the personality.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed.h      | 26 +++++++++++++++++++++++---
 drivers/net/ethernet/qlogic/qed/qed_cxt.c  |  8 ++++----
 drivers/net/ethernet/qlogic/qed/qed_dev.c  | 12 +++++-------
 drivers/net/ethernet/qlogic/qed/qed_l2.c   |  3 +--
 drivers/net/ethernet/qlogic/qed/qed_ll2.c  |  2 +-
 drivers/net/ethernet/qlogic/qed/qed_main.c | 17 ++++++++---------
 include/linux/qed/common_hsi.h             |  2 +-
 7 files changed, 43 insertions(+), 27 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 14b08ee..22e1171 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -210,14 +210,16 @@ struct qed_tunn_update_params {
 
 /* The PCI personality is not quite synonymous to protocol ID:
  * 1. All personalities need CORE connections
- * 2. The Ethernet personality may support also the RoCE protocol
+ * 2. The Ethernet personality may support also the RoCE/iWARP protocol
  */
 enum qed_pci_personality {
 	QED_PCI_ETH,
 	QED_PCI_FCOE,
 	QED_PCI_ISCSI,
 	QED_PCI_ETH_ROCE,
-	QED_PCI_DEFAULT /* default in shmem */
+	QED_PCI_ETH_IWARP,
+	QED_PCI_ETH_RDMA,
+	QED_PCI_DEFAULT, /* default in shmem */
 };
 
 /* All VFs are symmetric, all counters are PF + all VFs */
@@ -277,6 +279,7 @@ enum qed_dev_cap {
 	QED_DEV_CAP_FCOE,
 	QED_DEV_CAP_ISCSI,
 	QED_DEV_CAP_ROCE,
+	QED_DEV_CAP_IWARP,
 };
 
 enum qed_wol_support {
@@ -286,7 +289,24 @@ enum qed_wol_support {
 
 struct qed_hw_info {
 	/* PCI personality */
-	enum qed_pci_personality	personality;
+	enum qed_pci_personality personality;
+#define QED_IS_RDMA_PERSONALITY(dev)			    \
+	((dev)->hw_info.personality == QED_PCI_ETH_ROCE ||  \
+	 (dev)->hw_info.personality == QED_PCI_ETH_IWARP || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_ROCE_PERSONALITY(dev)			   \
+	((dev)->hw_info.personality == QED_PCI_ETH_ROCE || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_IWARP_PERSONALITY(dev)			    \
+	((dev)->hw_info.personality == QED_PCI_ETH_IWARP || \
+	 (dev)->hw_info.personality == QED_PCI_ETH_RDMA)
+#define QED_IS_L2_PERSONALITY(dev)		      \
+	((dev)->hw_info.personality == QED_PCI_ETH || \
+	 QED_IS_RDMA_PERSONALITY(dev))
+#define QED_IS_FCOE_PERSONALITY(dev) \
+	((dev)->hw_info.personality == QED_PCI_FCOE)
+#define QED_IS_ISCSI_PERSONALITY(dev) \
+	((dev)->hw_info.personality == QED_PCI_ISCSI)
 
 	/* Resource Allocation scheme results */
 	u32				resc_start[QED_MAX_RESC];
diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index e201214..38716f7 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
@@ -853,7 +853,7 @@ u32 qed_cxt_cfg_ilt_compute_excess(struct qed_hwfn *p_hwfn, u32 used_lines)
 	if (!excess_lines)
 		return 0;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_RDMA_PERSONALITY(p_hwfn))
 		return 0;
 
 	p_mngr = p_hwfn->p_cxt_mngr;
@@ -1033,7 +1033,7 @@ static int qed_ilt_blk_alloc(struct qed_hwfn *p_hwfn,
 	u32 lines, line, sz_left, lines_to_skip = 0;
 
 	/* Special handling for RoCE that supports dynamic allocation */
-	if ((p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) &&
+	if (QED_IS_RDMA_PERSONALITY(p_hwfn) &&
 	    ((ilt_client == ILT_CLI_CDUT) || ilt_client == ILT_CLI_TSDM))
 		return 0;
 
@@ -1833,7 +1833,7 @@ static void qed_tm_init_pf(struct qed_hwfn *p_hwfn)
 		tm_offset += tm_iids.pf_tids[i];
 	}
 
-	if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE)
+	if (QED_IS_RDMA_PERSONALITY(p_hwfn))
 		active_seg_mask = 0;
 
 	STORE_RT_REG(p_hwfn, TM_REG_PF_ENABLE_TASK_RT_OFFSET, active_seg_mask);
@@ -2344,7 +2344,7 @@ int qed_cxt_get_tid_mem_info(struct qed_hwfn *p_hwfn,
 		       last_cid_allocated - 1);
 
 		if (!p_hwfn->b_rdma_enabled_in_prs) {
-			/* Enable RoCE search */
+			/* Enable RDMA search */
 			qed_wr(p_hwfn, p_ptt, p_hwfn->rdma_prs_search_reg, 1);
 			p_hwfn->b_rdma_enabled_in_prs = true;
 		}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 49667ad..68e6182 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -936,7 +936,7 @@ int qed_resc_alloc(struct qed_dev *cdev)
 
 		/* EQ */
 		n_eqes = qed_chain_get_capacity(&p_hwfn->p_spq->chain);
-		if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) {
+		if (QED_IS_RDMA_PERSONALITY(p_hwfn)) {
 			num_cons = qed_cxt_get_proto_cid_count(p_hwfn,
 							       PROTOCOLID_ROCE,
 							       NULL) * 2;
@@ -2057,7 +2057,7 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 	qed_int_get_num_sbs(p_hwfn, &sb_cnt);
 
 	if (IS_ENABLED(CONFIG_QED_RDMA) &&
-	    p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE) {
+	    QED_IS_RDMA_PERSONALITY(p_hwfn)) {
 		/* Roce CNQ each requires: 1 status block + 1 CNQ. We divide
 		 * the status blocks equally between L2 / RoCE but with
 		 * consideration as to how many l2 queues / cnqs we have.
@@ -2068,9 +2068,7 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 
 		non_l2_sbs = feat_num[QED_RDMA_CNQ];
 	}
-
-	if (p_hwfn->hw_info.personality == QED_PCI_ETH_ROCE ||
-	    p_hwfn->hw_info.personality == QED_PCI_ETH) {
+	if (QED_IS_L2_PERSONALITY(p_hwfn)) {
 		/* Start by allocating VF queues, then PF's */
 		feat_num[QED_VF_L2_QUE] = min_t(u32,
 						RESC_NUM(p_hwfn, QED_L2_QUEUE),
@@ -2083,12 +2081,12 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 							 QED_VF_L2_QUE));
 	}
 
-	if (p_hwfn->hw_info.personality == QED_PCI_FCOE)
+	if (QED_IS_FCOE_PERSONALITY(p_hwfn))
 		feat_num[QED_FCOE_CQ] =  min_t(u32, sb_cnt.cnt,
 					       RESC_NUM(p_hwfn,
 							QED_CMDQS_CQS));
 
-	if (p_hwfn->hw_info.personality == QED_PCI_ISCSI)
+	if (QED_IS_ISCSI_PERSONALITY(p_hwfn))
 		feat_num[QED_ISCSI_CQ] = min_t(u32, sb_cnt.cnt,
 					       RESC_NUM(p_hwfn,
 							QED_CMDQS_CQS));
diff --git a/drivers/net/ethernet/qlogic/qed/qed_l2.c b/drivers/net/ethernet/qlogic/qed/qed_l2.c
index e57699b..27ea54b 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_l2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_l2.c
@@ -79,8 +79,7 @@ int qed_l2_alloc(struct qed_hwfn *p_hwfn)
 	unsigned long **pp_qids;
 	u32 i;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH &&
-	    p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_L2_PERSONALITY(p_hwfn))
 		return 0;
 
 	p_l2_info = kzalloc(sizeof(*p_l2_info), GFP_KERNEL);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index 17f9b0a..be66f19 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -1421,7 +1421,7 @@ int qed_ll2_establish_connection(void *cxt, u8 connection_handle)
 	if (rc)
 		goto out;
 
-	if (p_hwfn->hw_info.personality != QED_PCI_ETH_ROCE)
+	if (!QED_IS_RDMA_PERSONALITY(p_hwfn))
 		qed_wr(p_hwfn, p_ptt, PRS_REG_USE_LIGHT_L2, 1);
 
 	qed_ll2_establish_connection_ooo(p_hwfn, p_ll2_conn);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_main.c b/drivers/net/ethernet/qlogic/qed/qed_main.c
index 16cc30b..b113996 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_main.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_main.c
@@ -237,6 +237,8 @@ static int qed_init_pci(struct qed_dev *cdev, struct pci_dev *pdev)
 int qed_fill_dev_info(struct qed_dev *cdev,
 		      struct qed_dev_info *dev_info)
 {
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_hw_info *hw_info = &p_hwfn->hw_info;
 	struct qed_tunnel_info *tun = &cdev->tunnel;
 	struct qed_ptt  *ptt;
 
@@ -260,11 +262,10 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 	dev_info->pci_mem_start = cdev->pci_params.mem_start;
 	dev_info->pci_mem_end = cdev->pci_params.mem_end;
 	dev_info->pci_irq = cdev->pci_params.irq;
-	dev_info->rdma_supported = (cdev->hwfns[0].hw_info.personality ==
-				    QED_PCI_ETH_ROCE);
+	dev_info->rdma_supported = QED_IS_RDMA_PERSONALITY(p_hwfn);
 	dev_info->is_mf_default = IS_MF_DEFAULT(&cdev->hwfns[0]);
 	dev_info->dev_type = cdev->type;
-	ether_addr_copy(dev_info->hw_mac, cdev->hwfns[0].hw_info.hw_mac_addr);
+	ether_addr_copy(dev_info->hw_mac, hw_info->hw_mac_addr);
 
 	if (IS_PF(cdev)) {
 		dev_info->fw_major = FW_MAJOR_VERSION;
@@ -274,8 +275,7 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 		dev_info->mf_mode = cdev->mf_mode;
 		dev_info->tx_switching = true;
 
-		if (QED_LEADING_HWFN(cdev)->hw_info.b_wol_support ==
-		    QED_WOL_SUPPORT_PME)
+		if (hw_info->b_wol_support == QED_WOL_SUPPORT_PME)
 			dev_info->wol_support = true;
 
 		dev_info->abs_pf_id = QED_LEADING_HWFN(cdev)->abs_pf_id;
@@ -304,7 +304,7 @@ int qed_fill_dev_info(struct qed_dev *cdev,
 				    &dev_info->mfw_rev, NULL);
 	}
 
-	dev_info->mtu = QED_LEADING_HWFN(cdev)->hw_info.mtu;
+	dev_info->mtu = hw_info->mtu;
 
 	return 0;
 }
@@ -790,7 +790,7 @@ static int qed_slowpath_setup_int(struct qed_dev *cdev,
 				       cdev->num_hwfns;
 
 	if (!IS_ENABLED(CONFIG_QED_RDMA) ||
-	    QED_LEADING_HWFN(cdev)->hw_info.personality != QED_PCI_ETH_ROCE)
+	    !QED_IS_RDMA_PERSONALITY(QED_LEADING_HWFN(cdev)))
 		return 0;
 
 	for_each_hwfn(cdev, i)
@@ -931,8 +931,7 @@ static void qed_update_pf_params(struct qed_dev *cdev,
 	/* In case we might support RDMA, don't allow qede to be greedy
 	 * with the L2 contexts. Allow for 64 queues [rx, tx, xdp] per hwfn.
 	 */
-	if (QED_LEADING_HWFN(cdev)->hw_info.personality ==
-	    QED_PCI_ETH_ROCE) {
+	if (QED_IS_RDMA_PERSONALITY(QED_LEADING_HWFN(cdev))) {
 		u16 *num_cons;
 
 		num_cons = &params->eth_pf_params.num_cons;
diff --git a/include/linux/qed/common_hsi.h b/include/linux/qed/common_hsi.h
index a567cbf..885ae13 100644
--- a/include/linux/qed/common_hsi.h
+++ b/include/linux/qed/common_hsi.h
@@ -778,7 +778,7 @@ enum protocol_type {
 	PROTOCOLID_ROCE,
 	PROTOCOLID_CORE,
 	PROTOCOLID_ETH,
-	PROTOCOLID_RESERVED4,
+	PROTOCOLID_IWARP,
 	PROTOCOLID_RESERVED5,
 	PROTOCOLID_PREROCE,
 	PROTOCOLID_COMMON,
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 02/12] qed: Implement iWARP initialization, teardown and qp operations

This patch adds iWARP support for flows that have common code
between RoCE and iWARP, such as initialization, teardown and
qp setup verbs: create, destroy, modify, query.
It introduces the iWARP specific files qed_iwarp.[ch] and
iwarp_common.h

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/Makefile    |   2 +-
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |   9 +-
 drivers/net/ethernet/qlogic/qed/qed_hsi.h   |   1 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 531 ++++++++++++++++++++++++++++
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  85 +++++
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  | 133 +++++--
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |   3 +
 drivers/net/ethernet/qlogic/qed/qed_roce.c  |  20 ++
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |   5 +-
 include/linux/qed/iwarp_common.h            |  53 +++
 include/linux/qed/qed_rdma_if.h             |   1 +
 11 files changed, 803 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.c
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.h
 create mode 100644 include/linux/qed/iwarp_common.h

diff --git a/drivers/net/ethernet/qlogic/qed/Makefile b/drivers/net/ethernet/qlogic/qed/Makefile
index 6745238..82dd470 100644
--- a/drivers/net/ethernet/qlogic/qed/Makefile
+++ b/drivers/net/ethernet/qlogic/qed/Makefile
@@ -5,6 +5,6 @@ qed-y := qed_cxt.o qed_dev.o qed_hw.o qed_init_fw_funcs.o qed_init_ops.o \
 	 qed_selftest.o qed_dcbx.o qed_debug.o qed_ptp.o
 qed-$(CONFIG_QED_SRIOV) += qed_sriov.o qed_vf.o
 qed-$(CONFIG_QED_LL2) += qed_ll2.o
-qed-$(CONFIG_QED_RDMA) += qed_roce.o qed_rdma.o
+qed-$(CONFIG_QED_RDMA) += qed_roce.o qed_rdma.o qed_iwarp.o
 qed-$(CONFIG_QED_ISCSI) += qed_iscsi.o qed_ooo.o
 qed-$(CONFIG_QED_FCOE) += qed_fcoe.o
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 68e6182..6c8505d 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -937,8 +937,15 @@ int qed_resc_alloc(struct qed_dev *cdev)
 		/* EQ */
 		n_eqes = qed_chain_get_capacity(&p_hwfn->p_spq->chain);
 		if (QED_IS_RDMA_PERSONALITY(p_hwfn)) {
+			enum protocol_type rdma_proto;
+
+			if (QED_IS_ROCE_PERSONALITY(p_hwfn))
+				rdma_proto = PROTOCOLID_ROCE;
+			else
+				rdma_proto = PROTOCOLID_IWARP;
+
 			num_cons = qed_cxt_get_proto_cid_count(p_hwfn,
-							       PROTOCOLID_ROCE,
+							       rdma_proto,
 							       NULL) * 2;
 			n_eqes += num_cons + 2 * MAX_NUM_VFS_BB;
 		} else if (p_hwfn->hw_info.personality == QED_PCI_ISCSI) {
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hsi.h b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
index 3bf3614..31fb0bf 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hsi.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
@@ -46,6 +46,7 @@
 #include <linux/qed/fcoe_common.h>
 #include <linux/qed/eth_common.h>
 #include <linux/qed/iscsi_common.h>
+#include <linux/qed/iwarp_common.h>
 #include <linux/qed/rdma_common.h>
 #include <linux/qed/roce_common.h>
 #include <linux/qed/qed_fcoe_if.h>
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
new file mode 100644
index 0000000..a8bd5f8
--- /dev/null
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -0,0 +1,531 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "qed_cxt.h"
+#include "qed_hw.h"
+#include "qed_rdma.h"
+#include "qed_reg_addr.h"
+#include "qed_sp.h"
+
+#define QED_IWARP_ORD_DEFAULT		32
+#define QED_IWARP_IRD_DEFAULT		32
+#define QED_IWARP_RCV_WND_SIZE_DEF	(256 * 1024)
+#define QED_IWARP_RCV_WND_SIZE_MIN	(64 * 1024)
+#define QED_IWARP_TS_EN			BIT(0)
+#define QED_IWARP_PARAM_CRC_NEEDED	(1)
+#define QED_IWARP_PARAM_P2P		(1)
+
+static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
+				 u8 fw_event_code, u16 echo,
+				 union event_ring_data *data,
+				 u8 fw_return_code);
+
+/* Override devinfo with iWARP specific values */
+void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn)
+{
+	struct qed_rdma_device *dev = p_hwfn->p_rdma_info->dev;
+
+	dev->max_inline = IWARP_REQ_MAX_INLINE_DATA_SIZE;
+	dev->max_qp = min_t(u32,
+			    IWARP_MAX_QPS,
+			    p_hwfn->p_rdma_info->num_qps);
+
+	dev->max_cq = dev->max_qp;
+
+	dev->max_qp_resp_rd_atomic_resc = QED_IWARP_IRD_DEFAULT;
+	dev->max_qp_req_rd_atomic_resc = QED_IWARP_ORD_DEFAULT;
+}
+
+void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_TCP;
+	qed_wr(p_hwfn, p_ptt, p_hwfn->rdma_prs_search_reg, 1);
+	p_hwfn->b_rdma_enabled_in_prs = true;
+}
+
+static void qed_iwarp_cid_cleaned(struct qed_hwfn *p_hwfn, u32 cid)
+{
+	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+}
+
+static int qed_iwarp_alloc_cid(struct qed_hwfn *p_hwfn, u32 *cid)
+{
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	rc = qed_rdma_bmap_alloc_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed in allocating iwarp cid\n");
+		return rc;
+	}
+	*cid += qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	rc = qed_cxt_dynamic_ilt_alloc(p_hwfn, QED_ELEM_CXT, *cid);
+	if (rc)
+		qed_iwarp_cid_cleaned(p_hwfn, *cid);
+
+	return rc;
+}
+
+int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
+			struct qed_rdma_qp *qp,
+			struct qed_rdma_create_qp_out_params *out_params)
+{
+	struct iwarp_create_qp_ramrod_data *p_ramrod;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	u16 physical_queue;
+	u32 cid;
+	int rc;
+
+	qp->shared_queue = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+					      IWARP_SHARED_QUEUE_PAGE_SIZE,
+					      &qp->shared_queue_phys_addr,
+					      GFP_KERNEL);
+	if (!qp->shared_queue)
+		return -ENOMEM;
+
+	out_params->sq_pbl_virt = (u8 *)qp->shared_queue +
+	    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET;
+	out_params->sq_pbl_phys = qp->shared_queue_phys_addr +
+	    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET;
+	out_params->rq_pbl_virt = (u8 *)qp->shared_queue +
+	    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET;
+	out_params->rq_pbl_phys = qp->shared_queue_phys_addr +
+	    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET;
+
+	rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+	if (rc)
+		goto err1;
+
+	qp->icid = (u16)cid;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.cid = qp->icid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_CREATE_QP,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		goto err2;
+
+	p_ramrod = &p_ent->ramrod.iwarp_create_qp;
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_FMR_AND_RESERVED_EN,
+		  qp->fmr_and_reserved_lkey);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_SIGNALED_COMP, qp->signal_all);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_RDMA_RD_EN,
+		  qp->incoming_rdma_read_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_RDMA_WR_EN,
+		  qp->incoming_rdma_write_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_ATOMIC_EN,
+		  qp->incoming_atomic_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_SRQ_FLG, qp->use_srq);
+
+	p_ramrod->pd = qp->pd;
+	p_ramrod->sq_num_pages = qp->sq_num_pages;
+	p_ramrod->rq_num_pages = qp->rq_num_pages;
+
+	p_ramrod->qp_handle_for_cqe.hi = cpu_to_le32(qp->qp_handle.hi);
+	p_ramrod->qp_handle_for_cqe.lo = cpu_to_le32(qp->qp_handle.lo);
+
+	p_ramrod->cq_cid_for_sq =
+	    cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->sq_cq_id);
+	p_ramrod->cq_cid_for_rq =
+	    cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->rq_cq_id);
+
+	p_ramrod->dpi = cpu_to_le16(qp->dpi);
+
+	physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD);
+	p_ramrod->physical_q0 = cpu_to_le16(physical_queue);
+	physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK);
+	p_ramrod->physical_q1 = cpu_to_le16(physical_queue);
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+	if (rc)
+		goto err2;
+
+	return rc;
+
+err2:
+	qed_iwarp_cid_cleaned(p_hwfn, cid);
+err1:
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+			  IWARP_SHARED_QUEUE_PAGE_SIZE,
+			  qp->shared_queue, qp->shared_queue_phys_addr);
+
+	return rc;
+}
+
+static int qed_iwarp_modify_fw(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	struct iwarp_modify_qp_ramrod_data *p_ramrod;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	int rc;
+
+	/* Get SPQ entry */
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MODIFY_QP,
+				 p_hwfn->p_rdma_info->proto, &init_data);
+	if (rc)
+		return rc;
+
+	p_ramrod = &p_ent->ramrod.iwarp_modify_qp;
+	SET_FIELD(p_ramrod->flags, IWARP_MODIFY_QP_RAMROD_DATA_STATE_TRANS_EN,
+		  0x1);
+	if (qp->iwarp_state == QED_IWARP_QP_STATE_CLOSING)
+		p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_CLOSING;
+	else
+		p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_ERROR;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x)rc=%d\n", qp->icid, rc);
+
+	return rc;
+}
+
+enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state)
+{
+	switch (state) {
+	case QED_ROCE_QP_STATE_RESET:
+	case QED_ROCE_QP_STATE_INIT:
+	case QED_ROCE_QP_STATE_RTR:
+		return QED_IWARP_QP_STATE_IDLE;
+	case QED_ROCE_QP_STATE_RTS:
+		return QED_IWARP_QP_STATE_RTS;
+	case QED_ROCE_QP_STATE_SQD:
+		return QED_IWARP_QP_STATE_CLOSING;
+	case QED_ROCE_QP_STATE_ERR:
+		return QED_IWARP_QP_STATE_ERROR;
+	case QED_ROCE_QP_STATE_SQE:
+		return QED_IWARP_QP_STATE_TERMINATE;
+	default:
+		return QED_IWARP_QP_STATE_ERROR;
+	}
+}
+
+static enum qed_roce_qp_state
+qed_iwarp2roce_state(enum qed_iwarp_qp_state state)
+{
+	switch (state) {
+	case QED_IWARP_QP_STATE_IDLE:
+		return QED_ROCE_QP_STATE_INIT;
+	case QED_IWARP_QP_STATE_RTS:
+		return QED_ROCE_QP_STATE_RTS;
+	case QED_IWARP_QP_STATE_TERMINATE:
+		return QED_ROCE_QP_STATE_SQE;
+	case QED_IWARP_QP_STATE_CLOSING:
+		return QED_ROCE_QP_STATE_SQD;
+	case QED_IWARP_QP_STATE_ERROR:
+		return QED_ROCE_QP_STATE_ERR;
+	default:
+		return QED_ROCE_QP_STATE_ERR;
+	}
+}
+
+const char *iwarp_state_names[] = {
+	"IDLE",
+	"RTS",
+	"TERMINATE",
+	"CLOSING",
+	"ERROR",
+};
+
+int
+qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn,
+		    struct qed_rdma_qp *qp,
+		    enum qed_iwarp_qp_state new_state, bool internal)
+{
+	enum qed_iwarp_qp_state prev_iw_state;
+	bool modify_fw = false;
+	int rc = 0;
+
+	/* modify QP can be called from upper-layer or as a result of async
+	 * RST/FIN... therefore need to protect
+	 */
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	prev_iw_state = qp->iwarp_state;
+
+	if (prev_iw_state == new_state) {
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+		return 0;
+	}
+
+	switch (prev_iw_state) {
+	case QED_IWARP_QP_STATE_IDLE:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_RTS:
+			qp->iwarp_state = QED_IWARP_QP_STATE_RTS;
+			break;
+		case QED_IWARP_QP_STATE_ERROR:
+			qp->iwarp_state = QED_IWARP_QP_STATE_ERROR;
+			if (!internal)
+				modify_fw = true;
+			break;
+		default:
+			break;
+		}
+		break;
+	case QED_IWARP_QP_STATE_RTS:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_CLOSING:
+			if (!internal)
+				modify_fw = true;
+
+			qp->iwarp_state = QED_IWARP_QP_STATE_CLOSING;
+			break;
+		case QED_IWARP_QP_STATE_ERROR:
+			if (!internal)
+				modify_fw = true;
+			qp->iwarp_state = QED_IWARP_QP_STATE_ERROR;
+			break;
+		default:
+			break;
+		}
+		break;
+	case QED_IWARP_QP_STATE_ERROR:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_IDLE:
+
+			qp->iwarp_state = new_state;
+			break;
+		case QED_IWARP_QP_STATE_CLOSING:
+			/* could happen due to race... do nothing.... */
+			break;
+		default:
+			rc = -EINVAL;
+		}
+		break;
+	case QED_IWARP_QP_STATE_TERMINATE:
+	case QED_IWARP_QP_STATE_CLOSING:
+		qp->iwarp_state = new_state;
+		break;
+	default:
+		break;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) %s --> %s%s\n",
+		   qp->icid,
+		   iwarp_state_names[prev_iw_state],
+		   iwarp_state_names[qp->iwarp_state],
+		   internal ? "internal" : "");
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+
+	if (modify_fw)
+		rc = qed_iwarp_modify_fw(p_hwfn, qp);
+
+	return rc;
+}
+
+int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	int rc;
+
+	/* Get SPQ entry */
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_DESTROY_QP,
+				 p_hwfn->p_rdma_info->proto, &init_data);
+	if (rc)
+		return rc;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) rc = %d\n", qp->icid, rc);
+
+	return rc;
+}
+
+int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	int rc = 0;
+
+	if (qp->iwarp_state != QED_IWARP_QP_STATE_ERROR) {
+		rc = qed_iwarp_modify_qp(p_hwfn, qp,
+					 QED_IWARP_QP_STATE_ERROR, false);
+		if (rc)
+			return rc;
+	}
+
+	rc = qed_iwarp_fw_destroy(p_hwfn, qp);
+
+	if (qp->shared_queue)
+		dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+				  IWARP_SHARED_QUEUE_PAGE_SIZE,
+				  qp->shared_queue, qp->shared_queue_phys_addr);
+
+	return rc;
+}
+
+#define QED_IWARP_MAX_CID_CLEAN_TIME  100
+#define QED_IWARP_MAX_NO_PROGRESS_CNT 5
+
+/* This function waits for all the bits of a bmap to be cleared, as long as
+ * there is progress ( i.e. the number of bits left to be cleared decreases )
+ * the function continues.
+ */
+static int
+qed_iwarp_wait_cid_map_cleared(struct qed_hwfn *p_hwfn, struct qed_bmap *bmap)
+{
+	int prev_weight = 0;
+	int wait_count = 0;
+	int weight = 0;
+
+	weight = bitmap_weight(bmap->bitmap, bmap->max_count);
+	prev_weight = weight;
+
+	while (weight) {
+		msleep(QED_IWARP_MAX_CID_CLEAN_TIME);
+
+		weight = bitmap_weight(bmap->bitmap, bmap->max_count);
+
+		if (prev_weight == weight) {
+			wait_count++;
+		} else {
+			prev_weight = weight;
+			wait_count = 0;
+		}
+
+		if (wait_count > QED_IWARP_MAX_NO_PROGRESS_CNT) {
+			DP_NOTICE(p_hwfn,
+				  "%s bitmap wait timed out (%d cids pending)\n",
+				  bmap->name, weight);
+			return -EBUSY;
+		}
+	}
+	return 0;
+}
+
+static int qed_iwarp_wait_for_all_cids(struct qed_hwfn *p_hwfn)
+{
+	/* Now wait for all cids to be completed */
+	return qed_iwarp_wait_cid_map_cleared(p_hwfn,
+					      &p_hwfn->p_rdma_info->cid_map);
+}
+
+int qed_iwarp_alloc(struct qed_hwfn *p_hwfn)
+{
+	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	return 0;
+}
+
+void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
+{
+}
+
+int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+		    struct qed_rdma_start_in_params *params)
+{
+	struct qed_iwarp_info *iwarp_info;
+	u32 rcv_wnd_size;
+	int rc = 0;
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+
+	iwarp_info->tcp_flags = QED_IWARP_TS_EN;
+	rcv_wnd_size = QED_IWARP_RCV_WND_SIZE_DEF;
+
+	/* value 0 is used for ilog2(QED_IWARP_RCV_WND_SIZE_MIN) */
+	iwarp_info->rcv_wnd_scale = ilog2(rcv_wnd_size) -
+	    ilog2(QED_IWARP_RCV_WND_SIZE_MIN);
+	iwarp_info->crc_needed = QED_IWARP_PARAM_CRC_NEEDED;
+	iwarp_info->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED;
+
+	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
+
+	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+
+	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
+				  qed_iwarp_async_event);
+
+	return rc;
+}
+
+int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	int rc;
+
+	rc = qed_iwarp_wait_for_all_cids(p_hwfn);
+	if (rc)
+		return rc;
+
+	qed_spq_unregister_async_cb(p_hwfn, PROTOCOLID_IWARP);
+
+	return 0;
+}
+
+static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
+				 u8 fw_event_code, u16 echo,
+				 union event_ring_data *data,
+				 u8 fw_return_code)
+{
+	return 0;
+}
+
+void
+qed_iwarp_query_qp(struct qed_rdma_qp *qp,
+		   struct qed_rdma_query_qp_out_params *out_params)
+{
+	out_params->state = qed_iwarp2roce_state(qp->iwarp_state);
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
new file mode 100644
index 0000000..05e5e45
--- /dev/null
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -0,0 +1,85 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef _QED_IWARP_H
+#define _QED_IWARP_H
+
+enum qed_iwarp_qp_state {
+	QED_IWARP_QP_STATE_IDLE,
+	QED_IWARP_QP_STATE_RTS,
+	QED_IWARP_QP_STATE_TERMINATE,
+	QED_IWARP_QP_STATE_CLOSING,
+	QED_IWARP_QP_STATE_ERROR,
+};
+
+enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
+
+struct qed_iwarp_info {
+	spinlock_t iw_lock;	/* for iwarp resources */
+	spinlock_t qp_lock;	/* for teardown races */
+	u32 rcv_wnd_scale;
+	u16 max_mtu;
+	u8 mac_addr[ETH_ALEN];
+	u8 crc_needed;
+	u8 tcp_flags;
+	u8 peer2peer;
+	enum mpa_negotiation_mode mpa_rev;
+	enum mpa_rtr_type rtr_type;
+};
+
+int qed_iwarp_alloc(struct qed_hwfn *p_hwfn);
+
+int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+		    struct qed_rdma_start_in_params *params);
+
+int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
+void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn);
+
+void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn);
+
+void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
+int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
+			struct qed_rdma_qp *qp,
+			struct qed_rdma_create_qp_out_params *out_params);
+
+int qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp,
+			enum qed_iwarp_qp_state new_state, bool internal);
+
+int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp);
+
+int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp);
+
+void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
+			struct qed_rdma_query_qp_out_params *out_params);
+
+#endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index df76e21..ee6887f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -161,7 +161,10 @@ static int qed_rdma_alloc(struct qed_hwfn *p_hwfn,
 	num_cons = qed_cxt_get_proto_cid_count(p_hwfn, p_rdma_info->proto,
 					       NULL);
 
-	p_rdma_info->num_qps = num_cons / 2;
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		p_rdma_info->num_qps = num_cons;
+	else
+		p_rdma_info->num_qps = num_cons / 2; /* 2 cids per qp */
 
 	num_tasks = qed_cxt_get_proto_tid_count(p_hwfn, PROTOCOLID_ROCE);
 
@@ -252,6 +255,13 @@ static int qed_rdma_alloc(struct qed_hwfn *p_hwfn,
 			   "Failed to allocate real cid bitmap, rc = %d\n", rc);
 		goto free_cid_map;
 	}
+
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		rc = qed_iwarp_alloc(p_hwfn);
+
+	if (rc)
+		goto free_cid_map;
+
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Allocation successful\n");
 	return 0;
 
@@ -329,6 +339,9 @@ static void qed_rdma_resc_free(struct qed_hwfn *p_hwfn)
 {
 	struct qed_rdma_info *p_rdma_info = p_hwfn->p_rdma_info;
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_resc_free(p_hwfn);
+
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->cid_map, 1);
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->pd_map, 1);
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->dpi_map, 1);
@@ -470,6 +483,9 @@ static void qed_rdma_init_devinfo(struct qed_hwfn *p_hwfn,
 
 	if (pci_status_control & PCI_EXP_DEVCTL2_LTR_EN)
 		SET_FIELD(dev->dev_caps, QED_RDMA_DEV_CAP_ATOMIC_OP, 1);
+
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_init_devinfo(p_hwfn);
 }
 
 static void qed_rdma_init_port(struct qed_hwfn *p_hwfn)
@@ -490,29 +506,17 @@ static void qed_rdma_init_port(struct qed_hwfn *p_hwfn)
 
 static int qed_rdma_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 {
-	u32 ll2_ethertype_en;
+	int rc = 0;
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW\n");
 	p_hwfn->b_rdma_enabled_in_prs = false;
 
-	qed_wr(p_hwfn, p_ptt, PRS_REG_ROCE_DEST_QP_MAX_PF, 0);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_init_hw(p_hwfn, p_ptt);
+	else
+		rc = qed_roce_init_hw(p_hwfn, p_ptt);
 
-	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_ROCE;
-
-	/* We delay writing to this reg until first cid is allocated. See
-	 * qed_cxt_dynamic_ilt_alloc function for more details
-	 */
-	ll2_ethertype_en = qed_rd(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN);
-	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
-	       (ll2_ethertype_en | 0x01));
-
-	if (qed_cxt_get_proto_cid_start(p_hwfn, PROTOCOLID_ROCE) % 2) {
-		DP_NOTICE(p_hwfn, "The first RoCE's cid should be even\n");
-		return -EINVAL;
-	}
-
-	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW - Done\n");
-	return 0;
+	return rc;
 }
 
 static int qed_rdma_start_fw(struct qed_hwfn *p_hwfn,
@@ -544,7 +548,10 @@ static int qed_rdma_start_fw(struct qed_hwfn *p_hwfn,
 	if (rc)
 		return rc;
 
-	p_ramrod = &p_ent->ramrod.roce_init_func.rdma;
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		p_ramrod = &p_ent->ramrod.iwarp_init_func.rdma;
+	else
+		p_ramrod = &p_ent->ramrod.roce_init_func.rdma;
 
 	p_params_header = &p_ramrod->params_header;
 	p_params_header->cnq_start_offset = (u8)RESC_START(p_hwfn,
@@ -641,7 +648,15 @@ static int qed_rdma_setup(struct qed_hwfn *p_hwfn,
 	if (rc)
 		return rc;
 
-	qed_roce_setup(p_hwfn);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_setup(p_hwfn, p_ptt, params);
+		if (rc)
+			return rc;
+	} else {
+		rc = qed_roce_setup(p_hwfn);
+		if (rc)
+			return rc;
+	}
 
 	return qed_rdma_start_fw(p_hwfn, params, p_ptt);
 }
@@ -675,7 +690,16 @@ int qed_rdma_stop(void *rdma_cxt)
 	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
 	       (ll2_ethertype_en & 0xFFFE));
 
-	qed_roce_stop(p_hwfn);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_stop(p_hwfn, p_ptt);
+		if (rc) {
+			qed_ptt_release(p_hwfn, p_ptt);
+			return rc;
+		}
+	} else {
+		qed_roce_stop(p_hwfn);
+	}
+
 	qed_ptt_release(p_hwfn, p_ptt);
 
 	/* Get SPQ entry */
@@ -810,7 +834,9 @@ static int qed_fill_rdma_dev_info(struct qed_dev *cdev,
 
 	memset(info, 0, sizeof(*info));
 
-	info->rdma_type = QED_RDMA_TYPE_ROCE;
+	info->rdma_type = QED_IS_ROCE_PERSONALITY(p_hwfn) ?
+	    QED_RDMA_TYPE_ROCE : QED_RDMA_TYPE_IWARP;
+
 	info->user_dpm_enabled = (p_hwfn->db_bar_no_edpm == 0);
 
 	qed_fill_dev_info(cdev, &info->common);
@@ -1112,7 +1138,7 @@ static int qed_rdma_query_qp(void *rdma_cxt,
 			     struct qed_rdma_query_qp_out_params *out_params)
 {
 	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
-	int rc;
+	int rc = 0;
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "icid = %08x\n", qp->icid);
 
@@ -1138,7 +1164,10 @@ static int qed_rdma_query_qp(void *rdma_cxt,
 	out_params->max_dest_rd_atomic = qp->max_rd_atomic_resp;
 	out_params->sqd_async = qp->sqd_async;
 
-	rc = qed_roce_query_qp(p_hwfn, qp, out_params);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_query_qp(qp, out_params);
+	else
+		rc = qed_roce_query_qp(p_hwfn, qp, out_params);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Query QP, rc = %d\n", rc);
 	return rc;
@@ -1151,7 +1180,10 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "icid = %08x\n", qp->icid);
 
-	rc = qed_roce_destroy_qp(p_hwfn, qp);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		rc = qed_iwarp_destroy_qp(p_hwfn, qp);
+	else
+		rc = qed_roce_destroy_qp(p_hwfn, qp);
 
 	/* free qp params struct */
 	kfree(qp);
@@ -1190,20 +1222,27 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 		return NULL;
 	}
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		if (in_params->sq_num_pages * sizeof(struct regpair) >
+		    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_MAX_SIZE) {
+			DP_NOTICE(p_hwfn->cdev,
+				  "Sq num pages: %d exceeds maximum\n",
+				  in_params->sq_num_pages);
+			return NULL;
+		}
+		if (in_params->rq_num_pages * sizeof(struct regpair) >
+		    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_MAX_SIZE) {
+			DP_NOTICE(p_hwfn->cdev,
+				  "Rq num pages: %d exceeds maximum\n",
+				  in_params->rq_num_pages);
+			return NULL;
+		}
+	}
+
 	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
 	if (!qp)
 		return NULL;
 
-	rc = qed_roce_alloc_cid(p_hwfn, &qp->icid);
-	qp->qpid = ((0xFF << 16) | qp->icid);
-
-	DP_INFO(p_hwfn, "ROCE qpid=%x\n", qp->qpid);
-
-	if (rc) {
-		kfree(qp);
-		return NULL;
-	}
-
 	qp->cur_state = QED_ROCE_QP_STATE_RESET;
 	qp->qp_handle.hi = cpu_to_le32(in_params->qp_handle_hi);
 	qp->qp_handle.lo = cpu_to_le32(in_params->qp_handle_lo);
@@ -1226,6 +1265,19 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 	qp->e2e_flow_control_en = qp->use_srq ? false : true;
 	qp->stats_queue = in_params->stats_queue;
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_create_qp(p_hwfn, qp, out_params);
+		qp->qpid = qp->icid;
+	} else {
+		rc = qed_roce_alloc_cid(p_hwfn, &qp->icid);
+		qp->qpid = ((0xFF << 16) | qp->icid);
+	}
+
+	if (rc) {
+		kfree(qp);
+		return NULL;
+	}
+
 	out_params->icid = qp->icid;
 	out_params->qp_id = qp->qpid;
 
@@ -1324,7 +1376,14 @@ static int qed_rdma_modify_qp(void *rdma_cxt,
 			   qp->cur_state);
 	}
 
-	rc = qed_roce_modify_qp(p_hwfn, qp, prev_state, params);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		enum qed_iwarp_qp_state new_state =
+		    qed_roce2iwarp_state(qp->cur_state);
+
+		rc = qed_iwarp_modify_qp(p_hwfn, qp, new_state, 0);
+	} else {
+		rc = qed_roce_modify_qp(p_hwfn, qp, prev_state, params);
+	}
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Modify QP, rc = %d\n", rc);
 	return rc;
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.h b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
index d91e5c4..90e4e0f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
@@ -42,6 +42,7 @@
 #include "qed.h"
 #include "qed_dev_api.h"
 #include "qed_hsi.h"
+#include "qed_iwarp.h"
 #include "qed_roce.h"
 
 #define QED_RDMA_MAX_FMR                    (RDMA_MAX_TIDS)
@@ -97,6 +98,7 @@ struct qed_rdma_info {
 	u16 queue_zone_base;
 	u16 max_queue_zones;
 	enum protocol_type proto;
+	struct qed_iwarp_info iwarp;
 };
 
 struct qed_rdma_qp {
@@ -105,6 +107,7 @@ struct qed_rdma_qp {
 	u32 qpid;
 	u16 icid;
 	enum qed_roce_qp_state cur_state;
+	enum qed_iwarp_qp_state iwarp_state;
 	bool use_srq;
 	bool signal_all;
 	bool fmr_and_reserved_lkey;
diff --git a/drivers/net/ethernet/qlogic/qed/qed_roce.c b/drivers/net/ethernet/qlogic/qed/qed_roce.c
index e53adc3..fb7c2d1 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_roce.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_roce.c
@@ -1149,3 +1149,23 @@ int qed_roce_setup(struct qed_hwfn *p_hwfn)
 					 qed_roce_async_event);
 }
 
+int qed_roce_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	u32 ll2_ethertype_en;
+
+	qed_wr(p_hwfn, p_ptt, PRS_REG_ROCE_DEST_QP_MAX_PF, 0);
+
+	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_ROCE;
+
+	ll2_ethertype_en = qed_rd(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN);
+	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
+	       (ll2_ethertype_en | 0x01));
+
+	if (qed_cxt_get_proto_cid_start(p_hwfn, PROTOCOLID_ROCE) % 2) {
+		DP_NOTICE(p_hwfn, "The first RoCE's cid should be even\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW - Done\n");
+	return 0;
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sp.h b/drivers/net/ethernet/qlogic/qed/qed_sp.h
index 56c95fb..c3752c5 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_sp.h
@@ -104,12 +104,15 @@ int qed_eth_cqe_completion(struct qed_hwfn *p_hwfn,
 	struct roce_query_qp_req_ramrod_data roce_query_qp_req;
 	struct roce_destroy_qp_resp_ramrod_data roce_destroy_qp_resp;
 	struct roce_destroy_qp_req_ramrod_data roce_destroy_qp_req;
+	struct roce_init_func_ramrod_data roce_init_func;
 	struct rdma_create_cq_ramrod_data rdma_create_cq;
 	struct rdma_destroy_cq_ramrod_data rdma_destroy_cq;
 	struct rdma_srq_create_ramrod_data rdma_create_srq;
 	struct rdma_srq_destroy_ramrod_data rdma_destroy_srq;
 	struct rdma_srq_modify_ramrod_data rdma_modify_srq;
-	struct roce_init_func_ramrod_data roce_init_func;
+	struct iwarp_create_qp_ramrod_data iwarp_create_qp;
+	struct iwarp_modify_qp_ramrod_data iwarp_modify_qp;
+	struct iwarp_init_func_ramrod_data iwarp_init_func;
 	struct fcoe_init_ramrod_params fcoe_init;
 	struct fcoe_conn_offload_ramrod_params fcoe_conn_ofld;
 	struct fcoe_conn_terminate_ramrod_params fcoe_conn_terminate;
diff --git a/include/linux/qed/iwarp_common.h b/include/linux/qed/iwarp_common.h
new file mode 100644
index 0000000..b8b3e1c
--- /dev/null
+++ b/include/linux/qed/iwarp_common.h
@@ -0,0 +1,53 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __IWARP_COMMON__
+#define __IWARP_COMMON__
+#include <linux/qed/rdma_common.h>
+/************************/
+/* IWARP FW CONSTANTS	*/
+/************************/
+
+#define IWARP_ACTIVE_MODE 0
+#define IWARP_PASSIVE_MODE 1
+
+#define IWARP_SHARED_QUEUE_PAGE_SIZE		(0x8000)
+#define IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET   (0x4000)
+#define IWARP_SHARED_QUEUE_PAGE_RQ_PBL_MAX_SIZE (0x1000)
+#define IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET   (0x5000)
+#define IWARP_SHARED_QUEUE_PAGE_SQ_PBL_MAX_SIZE (0x3000)
+
+#define IWARP_REQ_MAX_INLINE_DATA_SIZE          (128)
+#define IWARP_REQ_MAX_SINGLE_SQ_WQE_SIZE        (176)
+
+#define IWARP_MAX_QPS                           (64 * 1024)
+
+#endif /* __IWARP_COMMON__ */
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index ff9be01..5b4bb09 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -491,6 +491,7 @@ struct qed_roce_ll2_packet {
 
 enum qed_rdma_type {
 	QED_RDMA_TYPE_ROCE,
+	QED_RDMA_TYPE_IWARP
 };
 
 struct qed_dev_rdma_info {
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 02/12] qed: Implement iWARP initialization, teardown and qp operations

[Michal Kalderon]

This patch adds iWARP support for flows that have common code
between RoCE and iWARP, such as initialization, teardown and
qp setup verbs: create, destroy, modify, query.
It introduces the iWARP specific files qed_iwarp.[ch] and
iwarp_common.h

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/Makefile    |   2 +-
 drivers/net/ethernet/qlogic/qed/qed_dev.c   |   9 +-
 drivers/net/ethernet/qlogic/qed/qed_hsi.h   |   1 +
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 531 ++++++++++++++++++++++++++++
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  85 +++++
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  | 133 +++++--
 drivers/net/ethernet/qlogic/qed/qed_rdma.h  |   3 +
 drivers/net/ethernet/qlogic/qed/qed_roce.c  |  20 ++
 drivers/net/ethernet/qlogic/qed/qed_sp.h    |   5 +-
 include/linux/qed/iwarp_common.h            |  53 +++
 include/linux/qed/qed_rdma_if.h             |   1 +
 11 files changed, 803 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.c
 create mode 100644 drivers/net/ethernet/qlogic/qed/qed_iwarp.h
 create mode 100644 include/linux/qed/iwarp_common.h

diff --git a/drivers/net/ethernet/qlogic/qed/Makefile b/drivers/net/ethernet/qlogic/qed/Makefile
index 6745238..82dd470 100644
--- a/drivers/net/ethernet/qlogic/qed/Makefile
+++ b/drivers/net/ethernet/qlogic/qed/Makefile
@@ -5,6 +5,6 @@ qed-y := qed_cxt.o qed_dev.o qed_hw.o qed_init_fw_funcs.o qed_init_ops.o \
 	 qed_selftest.o qed_dcbx.o qed_debug.o qed_ptp.o
 qed-$(CONFIG_QED_SRIOV) += qed_sriov.o qed_vf.o
 qed-$(CONFIG_QED_LL2) += qed_ll2.o
-qed-$(CONFIG_QED_RDMA) += qed_roce.o qed_rdma.o
+qed-$(CONFIG_QED_RDMA) += qed_roce.o qed_rdma.o qed_iwarp.o
 qed-$(CONFIG_QED_ISCSI) += qed_iscsi.o qed_ooo.o
 qed-$(CONFIG_QED_FCOE) += qed_fcoe.o
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 68e6182..6c8505d 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -937,8 +937,15 @@ int qed_resc_alloc(struct qed_dev *cdev)
 		/* EQ */
 		n_eqes = qed_chain_get_capacity(&p_hwfn->p_spq->chain);
 		if (QED_IS_RDMA_PERSONALITY(p_hwfn)) {
+			enum protocol_type rdma_proto;
+
+			if (QED_IS_ROCE_PERSONALITY(p_hwfn))
+				rdma_proto = PROTOCOLID_ROCE;
+			else
+				rdma_proto = PROTOCOLID_IWARP;
+
 			num_cons = qed_cxt_get_proto_cid_count(p_hwfn,
-							       PROTOCOLID_ROCE,
+							       rdma_proto,
 							       NULL) * 2;
 			n_eqes += num_cons + 2 * MAX_NUM_VFS_BB;
 		} else if (p_hwfn->hw_info.personality == QED_PCI_ISCSI) {
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hsi.h b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
index 3bf3614..31fb0bf 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hsi.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
@@ -46,6 +46,7 @@
 #include <linux/qed/fcoe_common.h>
 #include <linux/qed/eth_common.h>
 #include <linux/qed/iscsi_common.h>
+#include <linux/qed/iwarp_common.h>
 #include <linux/qed/rdma_common.h>
 #include <linux/qed/roce_common.h>
 #include <linux/qed/qed_fcoe_if.h>
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
new file mode 100644
index 0000000..a8bd5f8
--- /dev/null
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -0,0 +1,531 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "qed_cxt.h"
+#include "qed_hw.h"
+#include "qed_rdma.h"
+#include "qed_reg_addr.h"
+#include "qed_sp.h"
+
+#define QED_IWARP_ORD_DEFAULT		32
+#define QED_IWARP_IRD_DEFAULT		32
+#define QED_IWARP_RCV_WND_SIZE_DEF	(256 * 1024)
+#define QED_IWARP_RCV_WND_SIZE_MIN	(64 * 1024)
+#define QED_IWARP_TS_EN			BIT(0)
+#define QED_IWARP_PARAM_CRC_NEEDED	(1)
+#define QED_IWARP_PARAM_P2P		(1)
+
+static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
+				 u8 fw_event_code, u16 echo,
+				 union event_ring_data *data,
+				 u8 fw_return_code);
+
+/* Override devinfo with iWARP specific values */
+void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn)
+{
+	struct qed_rdma_device *dev = p_hwfn->p_rdma_info->dev;
+
+	dev->max_inline = IWARP_REQ_MAX_INLINE_DATA_SIZE;
+	dev->max_qp = min_t(u32,
+			    IWARP_MAX_QPS,
+			    p_hwfn->p_rdma_info->num_qps);
+
+	dev->max_cq = dev->max_qp;
+
+	dev->max_qp_resp_rd_atomic_resc = QED_IWARP_IRD_DEFAULT;
+	dev->max_qp_req_rd_atomic_resc = QED_IWARP_ORD_DEFAULT;
+}
+
+void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_TCP;
+	qed_wr(p_hwfn, p_ptt, p_hwfn->rdma_prs_search_reg, 1);
+	p_hwfn->b_rdma_enabled_in_prs = true;
+}
+
+static void qed_iwarp_cid_cleaned(struct qed_hwfn *p_hwfn, u32 cid)
+{
+	cid -= qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	qed_bmap_release_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+}
+
+static int qed_iwarp_alloc_cid(struct qed_hwfn *p_hwfn, u32 *cid)
+{
+	int rc;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->lock);
+	rc = qed_rdma_bmap_alloc_id(p_hwfn, &p_hwfn->p_rdma_info->cid_map, cid);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->lock);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed in allocating iwarp cid\n");
+		return rc;
+	}
+	*cid += qed_cxt_get_proto_cid_start(p_hwfn, p_hwfn->p_rdma_info->proto);
+
+	rc = qed_cxt_dynamic_ilt_alloc(p_hwfn, QED_ELEM_CXT, *cid);
+	if (rc)
+		qed_iwarp_cid_cleaned(p_hwfn, *cid);
+
+	return rc;
+}
+
+int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
+			struct qed_rdma_qp *qp,
+			struct qed_rdma_create_qp_out_params *out_params)
+{
+	struct iwarp_create_qp_ramrod_data *p_ramrod;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	u16 physical_queue;
+	u32 cid;
+	int rc;
+
+	qp->shared_queue = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+					      IWARP_SHARED_QUEUE_PAGE_SIZE,
+					      &qp->shared_queue_phys_addr,
+					      GFP_KERNEL);
+	if (!qp->shared_queue)
+		return -ENOMEM;
+
+	out_params->sq_pbl_virt = (u8 *)qp->shared_queue +
+	    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET;
+	out_params->sq_pbl_phys = qp->shared_queue_phys_addr +
+	    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET;
+	out_params->rq_pbl_virt = (u8 *)qp->shared_queue +
+	    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET;
+	out_params->rq_pbl_phys = qp->shared_queue_phys_addr +
+	    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET;
+
+	rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+	if (rc)
+		goto err1;
+
+	qp->icid = (u16)cid;
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.cid = qp->icid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_CREATE_QP,
+				 PROTOCOLID_IWARP, &init_data);
+	if (rc)
+		goto err2;
+
+	p_ramrod = &p_ent->ramrod.iwarp_create_qp;
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_FMR_AND_RESERVED_EN,
+		  qp->fmr_and_reserved_lkey);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_SIGNALED_COMP, qp->signal_all);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_RDMA_RD_EN,
+		  qp->incoming_rdma_read_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_RDMA_WR_EN,
+		  qp->incoming_rdma_write_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_ATOMIC_EN,
+		  qp->incoming_atomic_en);
+
+	SET_FIELD(p_ramrod->flags,
+		  IWARP_CREATE_QP_RAMROD_DATA_SRQ_FLG, qp->use_srq);
+
+	p_ramrod->pd = qp->pd;
+	p_ramrod->sq_num_pages = qp->sq_num_pages;
+	p_ramrod->rq_num_pages = qp->rq_num_pages;
+
+	p_ramrod->qp_handle_for_cqe.hi = cpu_to_le32(qp->qp_handle.hi);
+	p_ramrod->qp_handle_for_cqe.lo = cpu_to_le32(qp->qp_handle.lo);
+
+	p_ramrod->cq_cid_for_sq =
+	    cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->sq_cq_id);
+	p_ramrod->cq_cid_for_rq =
+	    cpu_to_le32((p_hwfn->hw_info.opaque_fid << 16) | qp->rq_cq_id);
+
+	p_ramrod->dpi = cpu_to_le16(qp->dpi);
+
+	physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_OFLD);
+	p_ramrod->physical_q0 = cpu_to_le16(physical_queue);
+	physical_queue = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_ACK);
+	p_ramrod->physical_q1 = cpu_to_le16(physical_queue);
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+	if (rc)
+		goto err2;
+
+	return rc;
+
+err2:
+	qed_iwarp_cid_cleaned(p_hwfn, cid);
+err1:
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+			  IWARP_SHARED_QUEUE_PAGE_SIZE,
+			  qp->shared_queue, qp->shared_queue_phys_addr);
+
+	return rc;
+}
+
+static int qed_iwarp_modify_fw(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	struct iwarp_modify_qp_ramrod_data *p_ramrod;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	int rc;
+
+	/* Get SPQ entry */
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MODIFY_QP,
+				 p_hwfn->p_rdma_info->proto, &init_data);
+	if (rc)
+		return rc;
+
+	p_ramrod = &p_ent->ramrod.iwarp_modify_qp;
+	SET_FIELD(p_ramrod->flags, IWARP_MODIFY_QP_RAMROD_DATA_STATE_TRANS_EN,
+		  0x1);
+	if (qp->iwarp_state == QED_IWARP_QP_STATE_CLOSING)
+		p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_CLOSING;
+	else
+		p_ramrod->transition_to_state = IWARP_MODIFY_QP_STATE_ERROR;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x)rc=%d\n", qp->icid, rc);
+
+	return rc;
+}
+
+enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state)
+{
+	switch (state) {
+	case QED_ROCE_QP_STATE_RESET:
+	case QED_ROCE_QP_STATE_INIT:
+	case QED_ROCE_QP_STATE_RTR:
+		return QED_IWARP_QP_STATE_IDLE;
+	case QED_ROCE_QP_STATE_RTS:
+		return QED_IWARP_QP_STATE_RTS;
+	case QED_ROCE_QP_STATE_SQD:
+		return QED_IWARP_QP_STATE_CLOSING;
+	case QED_ROCE_QP_STATE_ERR:
+		return QED_IWARP_QP_STATE_ERROR;
+	case QED_ROCE_QP_STATE_SQE:
+		return QED_IWARP_QP_STATE_TERMINATE;
+	default:
+		return QED_IWARP_QP_STATE_ERROR;
+	}
+}
+
+static enum qed_roce_qp_state
+qed_iwarp2roce_state(enum qed_iwarp_qp_state state)
+{
+	switch (state) {
+	case QED_IWARP_QP_STATE_IDLE:
+		return QED_ROCE_QP_STATE_INIT;
+	case QED_IWARP_QP_STATE_RTS:
+		return QED_ROCE_QP_STATE_RTS;
+	case QED_IWARP_QP_STATE_TERMINATE:
+		return QED_ROCE_QP_STATE_SQE;
+	case QED_IWARP_QP_STATE_CLOSING:
+		return QED_ROCE_QP_STATE_SQD;
+	case QED_IWARP_QP_STATE_ERROR:
+		return QED_ROCE_QP_STATE_ERR;
+	default:
+		return QED_ROCE_QP_STATE_ERR;
+	}
+}
+
+const char *iwarp_state_names[] = {
+	"IDLE",
+	"RTS",
+	"TERMINATE",
+	"CLOSING",
+	"ERROR",
+};
+
+int
+qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn,
+		    struct qed_rdma_qp *qp,
+		    enum qed_iwarp_qp_state new_state, bool internal)
+{
+	enum qed_iwarp_qp_state prev_iw_state;
+	bool modify_fw = false;
+	int rc = 0;
+
+	/* modify QP can be called from upper-layer or as a result of async
+	 * RST/FIN... therefore need to protect
+	 */
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+	prev_iw_state = qp->iwarp_state;
+
+	if (prev_iw_state == new_state) {
+		spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+		return 0;
+	}
+
+	switch (prev_iw_state) {
+	case QED_IWARP_QP_STATE_IDLE:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_RTS:
+			qp->iwarp_state = QED_IWARP_QP_STATE_RTS;
+			break;
+		case QED_IWARP_QP_STATE_ERROR:
+			qp->iwarp_state = QED_IWARP_QP_STATE_ERROR;
+			if (!internal)
+				modify_fw = true;
+			break;
+		default:
+			break;
+		}
+		break;
+	case QED_IWARP_QP_STATE_RTS:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_CLOSING:
+			if (!internal)
+				modify_fw = true;
+
+			qp->iwarp_state = QED_IWARP_QP_STATE_CLOSING;
+			break;
+		case QED_IWARP_QP_STATE_ERROR:
+			if (!internal)
+				modify_fw = true;
+			qp->iwarp_state = QED_IWARP_QP_STATE_ERROR;
+			break;
+		default:
+			break;
+		}
+		break;
+	case QED_IWARP_QP_STATE_ERROR:
+		switch (new_state) {
+		case QED_IWARP_QP_STATE_IDLE:
+
+			qp->iwarp_state = new_state;
+			break;
+		case QED_IWARP_QP_STATE_CLOSING:
+			/* could happen due to race... do nothing.... */
+			break;
+		default:
+			rc = -EINVAL;
+		}
+		break;
+	case QED_IWARP_QP_STATE_TERMINATE:
+	case QED_IWARP_QP_STATE_CLOSING:
+		qp->iwarp_state = new_state;
+		break;
+	default:
+		break;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) %s --> %s%s\n",
+		   qp->icid,
+		   iwarp_state_names[prev_iw_state],
+		   iwarp_state_names[qp->iwarp_state],
+		   internal ? "internal" : "");
+
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+
+	if (modify_fw)
+		rc = qed_iwarp_modify_fw(p_hwfn, qp);
+
+	return rc;
+}
+
+int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	int rc;
+
+	/* Get SPQ entry */
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_DESTROY_QP,
+				 p_hwfn->p_rdma_info->proto, &init_data);
+	if (rc)
+		return rc;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) rc = %d\n", qp->icid, rc);
+
+	return rc;
+}
+
+int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
+{
+	int rc = 0;
+
+	if (qp->iwarp_state != QED_IWARP_QP_STATE_ERROR) {
+		rc = qed_iwarp_modify_qp(p_hwfn, qp,
+					 QED_IWARP_QP_STATE_ERROR, false);
+		if (rc)
+			return rc;
+	}
+
+	rc = qed_iwarp_fw_destroy(p_hwfn, qp);
+
+	if (qp->shared_queue)
+		dma_free_coherent(&p_hwfn->cdev->pdev->dev,
+				  IWARP_SHARED_QUEUE_PAGE_SIZE,
+				  qp->shared_queue, qp->shared_queue_phys_addr);
+
+	return rc;
+}
+
+#define QED_IWARP_MAX_CID_CLEAN_TIME  100
+#define QED_IWARP_MAX_NO_PROGRESS_CNT 5
+
+/* This function waits for all the bits of a bmap to be cleared, as long as
+ * there is progress ( i.e. the number of bits left to be cleared decreases )
+ * the function continues.
+ */
+static int
+qed_iwarp_wait_cid_map_cleared(struct qed_hwfn *p_hwfn, struct qed_bmap *bmap)
+{
+	int prev_weight = 0;
+	int wait_count = 0;
+	int weight = 0;
+
+	weight = bitmap_weight(bmap->bitmap, bmap->max_count);
+	prev_weight = weight;
+
+	while (weight) {
+		msleep(QED_IWARP_MAX_CID_CLEAN_TIME);
+
+		weight = bitmap_weight(bmap->bitmap, bmap->max_count);
+
+		if (prev_weight == weight) {
+			wait_count++;
+		} else {
+			prev_weight = weight;
+			wait_count = 0;
+		}
+
+		if (wait_count > QED_IWARP_MAX_NO_PROGRESS_CNT) {
+			DP_NOTICE(p_hwfn,
+				  "%s bitmap wait timed out (%d cids pending)\n",
+				  bmap->name, weight);
+			return -EBUSY;
+		}
+	}
+	return 0;
+}
+
+static int qed_iwarp_wait_for_all_cids(struct qed_hwfn *p_hwfn)
+{
+	/* Now wait for all cids to be completed */
+	return qed_iwarp_wait_cid_map_cleared(p_hwfn,
+					      &p_hwfn->p_rdma_info->cid_map);
+}
+
+int qed_iwarp_alloc(struct qed_hwfn *p_hwfn)
+{
+	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	return 0;
+}
+
+void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
+{
+}
+
+int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+		    struct qed_rdma_start_in_params *params)
+{
+	struct qed_iwarp_info *iwarp_info;
+	u32 rcv_wnd_size;
+	int rc = 0;
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+
+	iwarp_info->tcp_flags = QED_IWARP_TS_EN;
+	rcv_wnd_size = QED_IWARP_RCV_WND_SIZE_DEF;
+
+	/* value 0 is used for ilog2(QED_IWARP_RCV_WND_SIZE_MIN) */
+	iwarp_info->rcv_wnd_scale = ilog2(rcv_wnd_size) -
+	    ilog2(QED_IWARP_RCV_WND_SIZE_MIN);
+	iwarp_info->crc_needed = QED_IWARP_PARAM_CRC_NEEDED;
+	iwarp_info->mpa_rev = MPA_NEGOTIATION_TYPE_ENHANCED;
+
+	iwarp_info->peer2peer = QED_IWARP_PARAM_P2P;
+
+	spin_lock_init(&p_hwfn->p_rdma_info->iwarp.qp_lock);
+
+	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
+				  qed_iwarp_async_event);
+
+	return rc;
+}
+
+int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	int rc;
+
+	rc = qed_iwarp_wait_for_all_cids(p_hwfn);
+	if (rc)
+		return rc;
+
+	qed_spq_unregister_async_cb(p_hwfn, PROTOCOLID_IWARP);
+
+	return 0;
+}
+
+static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
+				 u8 fw_event_code, u16 echo,
+				 union event_ring_data *data,
+				 u8 fw_return_code)
+{
+	return 0;
+}
+
+void
+qed_iwarp_query_qp(struct qed_rdma_qp *qp,
+		   struct qed_rdma_query_qp_out_params *out_params)
+{
+	out_params->state = qed_iwarp2roce_state(qp->iwarp_state);
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
new file mode 100644
index 0000000..05e5e45
--- /dev/null
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -0,0 +1,85 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef _QED_IWARP_H
+#define _QED_IWARP_H
+
+enum qed_iwarp_qp_state {
+	QED_IWARP_QP_STATE_IDLE,
+	QED_IWARP_QP_STATE_RTS,
+	QED_IWARP_QP_STATE_TERMINATE,
+	QED_IWARP_QP_STATE_CLOSING,
+	QED_IWARP_QP_STATE_ERROR,
+};
+
+enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
+
+struct qed_iwarp_info {
+	spinlock_t iw_lock;	/* for iwarp resources */
+	spinlock_t qp_lock;	/* for teardown races */
+	u32 rcv_wnd_scale;
+	u16 max_mtu;
+	u8 mac_addr[ETH_ALEN];
+	u8 crc_needed;
+	u8 tcp_flags;
+	u8 peer2peer;
+	enum mpa_negotiation_mode mpa_rev;
+	enum mpa_rtr_type rtr_type;
+};
+
+int qed_iwarp_alloc(struct qed_hwfn *p_hwfn);
+
+int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+		    struct qed_rdma_start_in_params *params);
+
+int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
+void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn);
+
+void qed_iwarp_init_devinfo(struct qed_hwfn *p_hwfn);
+
+void qed_iwarp_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
+int qed_iwarp_create_qp(struct qed_hwfn *p_hwfn,
+			struct qed_rdma_qp *qp,
+			struct qed_rdma_create_qp_out_params *out_params);
+
+int qed_iwarp_modify_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp,
+			enum qed_iwarp_qp_state new_state, bool internal);
+
+int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp);
+
+int qed_iwarp_fw_destroy(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp);
+
+void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
+			struct qed_rdma_query_qp_out_params *out_params);
+
+#endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index df76e21..ee6887f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -161,7 +161,10 @@ static int qed_rdma_alloc(struct qed_hwfn *p_hwfn,
 	num_cons = qed_cxt_get_proto_cid_count(p_hwfn, p_rdma_info->proto,
 					       NULL);
 
-	p_rdma_info->num_qps = num_cons / 2;
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		p_rdma_info->num_qps = num_cons;
+	else
+		p_rdma_info->num_qps = num_cons / 2; /* 2 cids per qp */
 
 	num_tasks = qed_cxt_get_proto_tid_count(p_hwfn, PROTOCOLID_ROCE);
 
@@ -252,6 +255,13 @@ static int qed_rdma_alloc(struct qed_hwfn *p_hwfn,
 			   "Failed to allocate real cid bitmap, rc = %d\n", rc);
 		goto free_cid_map;
 	}
+
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		rc = qed_iwarp_alloc(p_hwfn);
+
+	if (rc)
+		goto free_cid_map;
+
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Allocation successful\n");
 	return 0;
 
@@ -329,6 +339,9 @@ static void qed_rdma_resc_free(struct qed_hwfn *p_hwfn)
 {
 	struct qed_rdma_info *p_rdma_info = p_hwfn->p_rdma_info;
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_resc_free(p_hwfn);
+
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->cid_map, 1);
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->pd_map, 1);
 	qed_rdma_bmap_free(p_hwfn, &p_hwfn->p_rdma_info->dpi_map, 1);
@@ -470,6 +483,9 @@ static void qed_rdma_init_devinfo(struct qed_hwfn *p_hwfn,
 
 	if (pci_status_control & PCI_EXP_DEVCTL2_LTR_EN)
 		SET_FIELD(dev->dev_caps, QED_RDMA_DEV_CAP_ATOMIC_OP, 1);
+
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_init_devinfo(p_hwfn);
 }
 
 static void qed_rdma_init_port(struct qed_hwfn *p_hwfn)
@@ -490,29 +506,17 @@ static void qed_rdma_init_port(struct qed_hwfn *p_hwfn)
 
 static int qed_rdma_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 {
-	u32 ll2_ethertype_en;
+	int rc = 0;
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW\n");
 	p_hwfn->b_rdma_enabled_in_prs = false;
 
-	qed_wr(p_hwfn, p_ptt, PRS_REG_ROCE_DEST_QP_MAX_PF, 0);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_init_hw(p_hwfn, p_ptt);
+	else
+		rc = qed_roce_init_hw(p_hwfn, p_ptt);
 
-	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_ROCE;
-
-	/* We delay writing to this reg until first cid is allocated. See
-	 * qed_cxt_dynamic_ilt_alloc function for more details
-	 */
-	ll2_ethertype_en = qed_rd(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN);
-	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
-	       (ll2_ethertype_en | 0x01));
-
-	if (qed_cxt_get_proto_cid_start(p_hwfn, PROTOCOLID_ROCE) % 2) {
-		DP_NOTICE(p_hwfn, "The first RoCE's cid should be even\n");
-		return -EINVAL;
-	}
-
-	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW - Done\n");
-	return 0;
+	return rc;
 }
 
 static int qed_rdma_start_fw(struct qed_hwfn *p_hwfn,
@@ -544,7 +548,10 @@ static int qed_rdma_start_fw(struct qed_hwfn *p_hwfn,
 	if (rc)
 		return rc;
 
-	p_ramrod = &p_ent->ramrod.roce_init_func.rdma;
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		p_ramrod = &p_ent->ramrod.iwarp_init_func.rdma;
+	else
+		p_ramrod = &p_ent->ramrod.roce_init_func.rdma;
 
 	p_params_header = &p_ramrod->params_header;
 	p_params_header->cnq_start_offset = (u8)RESC_START(p_hwfn,
@@ -641,7 +648,15 @@ static int qed_rdma_setup(struct qed_hwfn *p_hwfn,
 	if (rc)
 		return rc;
 
-	qed_roce_setup(p_hwfn);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_setup(p_hwfn, p_ptt, params);
+		if (rc)
+			return rc;
+	} else {
+		rc = qed_roce_setup(p_hwfn);
+		if (rc)
+			return rc;
+	}
 
 	return qed_rdma_start_fw(p_hwfn, params, p_ptt);
 }
@@ -675,7 +690,16 @@ int qed_rdma_stop(void *rdma_cxt)
 	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
 	       (ll2_ethertype_en & 0xFFFE));
 
-	qed_roce_stop(p_hwfn);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_stop(p_hwfn, p_ptt);
+		if (rc) {
+			qed_ptt_release(p_hwfn, p_ptt);
+			return rc;
+		}
+	} else {
+		qed_roce_stop(p_hwfn);
+	}
+
 	qed_ptt_release(p_hwfn, p_ptt);
 
 	/* Get SPQ entry */
@@ -810,7 +834,9 @@ static int qed_fill_rdma_dev_info(struct qed_dev *cdev,
 
 	memset(info, 0, sizeof(*info));
 
-	info->rdma_type = QED_RDMA_TYPE_ROCE;
+	info->rdma_type = QED_IS_ROCE_PERSONALITY(p_hwfn) ?
+	    QED_RDMA_TYPE_ROCE : QED_RDMA_TYPE_IWARP;
+
 	info->user_dpm_enabled = (p_hwfn->db_bar_no_edpm == 0);
 
 	qed_fill_dev_info(cdev, &info->common);
@@ -1112,7 +1138,7 @@ static int qed_rdma_query_qp(void *rdma_cxt,
 			     struct qed_rdma_query_qp_out_params *out_params)
 {
 	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
-	int rc;
+	int rc = 0;
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "icid = %08x\n", qp->icid);
 
@@ -1138,7 +1164,10 @@ static int qed_rdma_query_qp(void *rdma_cxt,
 	out_params->max_dest_rd_atomic = qp->max_rd_atomic_resp;
 	out_params->sqd_async = qp->sqd_async;
 
-	rc = qed_roce_query_qp(p_hwfn, qp, out_params);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		qed_iwarp_query_qp(qp, out_params);
+	else
+		rc = qed_roce_query_qp(p_hwfn, qp, out_params);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Query QP, rc = %d\n", rc);
 	return rc;
@@ -1151,7 +1180,10 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "icid = %08x\n", qp->icid);
 
-	rc = qed_roce_destroy_qp(p_hwfn, qp);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn))
+		rc = qed_iwarp_destroy_qp(p_hwfn, qp);
+	else
+		rc = qed_roce_destroy_qp(p_hwfn, qp);
 
 	/* free qp params struct */
 	kfree(qp);
@@ -1190,20 +1222,27 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 		return NULL;
 	}
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		if (in_params->sq_num_pages * sizeof(struct regpair) >
+		    IWARP_SHARED_QUEUE_PAGE_SQ_PBL_MAX_SIZE) {
+			DP_NOTICE(p_hwfn->cdev,
+				  "Sq num pages: %d exceeds maximum\n",
+				  in_params->sq_num_pages);
+			return NULL;
+		}
+		if (in_params->rq_num_pages * sizeof(struct regpair) >
+		    IWARP_SHARED_QUEUE_PAGE_RQ_PBL_MAX_SIZE) {
+			DP_NOTICE(p_hwfn->cdev,
+				  "Rq num pages: %d exceeds maximum\n",
+				  in_params->rq_num_pages);
+			return NULL;
+		}
+	}
+
 	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
 	if (!qp)
 		return NULL;
 
-	rc = qed_roce_alloc_cid(p_hwfn, &qp->icid);
-	qp->qpid = ((0xFF << 16) | qp->icid);
-
-	DP_INFO(p_hwfn, "ROCE qpid=%x\n", qp->qpid);
-
-	if (rc) {
-		kfree(qp);
-		return NULL;
-	}
-
 	qp->cur_state = QED_ROCE_QP_STATE_RESET;
 	qp->qp_handle.hi = cpu_to_le32(in_params->qp_handle_hi);
 	qp->qp_handle.lo = cpu_to_le32(in_params->qp_handle_lo);
@@ -1226,6 +1265,19 @@ static int qed_rdma_destroy_qp(void *rdma_cxt, struct qed_rdma_qp *qp)
 	qp->e2e_flow_control_en = qp->use_srq ? false : true;
 	qp->stats_queue = in_params->stats_queue;
 
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		rc = qed_iwarp_create_qp(p_hwfn, qp, out_params);
+		qp->qpid = qp->icid;
+	} else {
+		rc = qed_roce_alloc_cid(p_hwfn, &qp->icid);
+		qp->qpid = ((0xFF << 16) | qp->icid);
+	}
+
+	if (rc) {
+		kfree(qp);
+		return NULL;
+	}
+
 	out_params->icid = qp->icid;
 	out_params->qp_id = qp->qpid;
 
@@ -1324,7 +1376,14 @@ static int qed_rdma_modify_qp(void *rdma_cxt,
 			   qp->cur_state);
 	}
 
-	rc = qed_roce_modify_qp(p_hwfn, qp, prev_state, params);
+	if (QED_IS_IWARP_PERSONALITY(p_hwfn)) {
+		enum qed_iwarp_qp_state new_state =
+		    qed_roce2iwarp_state(qp->cur_state);
+
+		rc = qed_iwarp_modify_qp(p_hwfn, qp, new_state, 0);
+	} else {
+		rc = qed_roce_modify_qp(p_hwfn, qp, prev_state, params);
+	}
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Modify QP, rc = %d\n", rc);
 	return rc;
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.h b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
index d91e5c4..90e4e0f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.h
@@ -42,6 +42,7 @@
 #include "qed.h"
 #include "qed_dev_api.h"
 #include "qed_hsi.h"
+#include "qed_iwarp.h"
 #include "qed_roce.h"
 
 #define QED_RDMA_MAX_FMR                    (RDMA_MAX_TIDS)
@@ -97,6 +98,7 @@ struct qed_rdma_info {
 	u16 queue_zone_base;
 	u16 max_queue_zones;
 	enum protocol_type proto;
+	struct qed_iwarp_info iwarp;
 };
 
 struct qed_rdma_qp {
@@ -105,6 +107,7 @@ struct qed_rdma_qp {
 	u32 qpid;
 	u16 icid;
 	enum qed_roce_qp_state cur_state;
+	enum qed_iwarp_qp_state iwarp_state;
 	bool use_srq;
 	bool signal_all;
 	bool fmr_and_reserved_lkey;
diff --git a/drivers/net/ethernet/qlogic/qed/qed_roce.c b/drivers/net/ethernet/qlogic/qed/qed_roce.c
index e53adc3..fb7c2d1 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_roce.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_roce.c
@@ -1149,3 +1149,23 @@ int qed_roce_setup(struct qed_hwfn *p_hwfn)
 					 qed_roce_async_event);
 }
 
+int qed_roce_init_hw(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	u32 ll2_ethertype_en;
+
+	qed_wr(p_hwfn, p_ptt, PRS_REG_ROCE_DEST_QP_MAX_PF, 0);
+
+	p_hwfn->rdma_prs_search_reg = PRS_REG_SEARCH_ROCE;
+
+	ll2_ethertype_en = qed_rd(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN);
+	qed_wr(p_hwfn, p_ptt, PRS_REG_LIGHT_L2_ETHERTYPE_EN,
+	       (ll2_ethertype_en | 0x01));
+
+	if (qed_cxt_get_proto_cid_start(p_hwfn, PROTOCOLID_ROCE) % 2) {
+		DP_NOTICE(p_hwfn, "The first RoCE's cid should be even\n");
+		return -EINVAL;
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "Initializing HW - Done\n");
+	return 0;
+}
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sp.h b/drivers/net/ethernet/qlogic/qed/qed_sp.h
index 56c95fb..c3752c5 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_sp.h
@@ -104,12 +104,15 @@ int qed_eth_cqe_completion(struct qed_hwfn *p_hwfn,
 	struct roce_query_qp_req_ramrod_data roce_query_qp_req;
 	struct roce_destroy_qp_resp_ramrod_data roce_destroy_qp_resp;
 	struct roce_destroy_qp_req_ramrod_data roce_destroy_qp_req;
+	struct roce_init_func_ramrod_data roce_init_func;
 	struct rdma_create_cq_ramrod_data rdma_create_cq;
 	struct rdma_destroy_cq_ramrod_data rdma_destroy_cq;
 	struct rdma_srq_create_ramrod_data rdma_create_srq;
 	struct rdma_srq_destroy_ramrod_data rdma_destroy_srq;
 	struct rdma_srq_modify_ramrod_data rdma_modify_srq;
-	struct roce_init_func_ramrod_data roce_init_func;
+	struct iwarp_create_qp_ramrod_data iwarp_create_qp;
+	struct iwarp_modify_qp_ramrod_data iwarp_modify_qp;
+	struct iwarp_init_func_ramrod_data iwarp_init_func;
 	struct fcoe_init_ramrod_params fcoe_init;
 	struct fcoe_conn_offload_ramrod_params fcoe_conn_ofld;
 	struct fcoe_conn_terminate_ramrod_params fcoe_conn_terminate;
diff --git a/include/linux/qed/iwarp_common.h b/include/linux/qed/iwarp_common.h
new file mode 100644
index 0000000..b8b3e1c
--- /dev/null
+++ b/include/linux/qed/iwarp_common.h
@@ -0,0 +1,53 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __IWARP_COMMON__
+#define __IWARP_COMMON__
+#include <linux/qed/rdma_common.h>
+/************************/
+/* IWARP FW CONSTANTS	*/
+/************************/
+
+#define IWARP_ACTIVE_MODE 0
+#define IWARP_PASSIVE_MODE 1
+
+#define IWARP_SHARED_QUEUE_PAGE_SIZE		(0x8000)
+#define IWARP_SHARED_QUEUE_PAGE_RQ_PBL_OFFSET   (0x4000)
+#define IWARP_SHARED_QUEUE_PAGE_RQ_PBL_MAX_SIZE (0x1000)
+#define IWARP_SHARED_QUEUE_PAGE_SQ_PBL_OFFSET   (0x5000)
+#define IWARP_SHARED_QUEUE_PAGE_SQ_PBL_MAX_SIZE (0x3000)
+
+#define IWARP_REQ_MAX_INLINE_DATA_SIZE          (128)
+#define IWARP_REQ_MAX_SINGLE_SQ_WQE_SIZE        (176)
+
+#define IWARP_MAX_QPS                           (64 * 1024)
+
+#endif /* __IWARP_COMMON__ */
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index ff9be01..5b4bb09 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -491,6 +491,7 @@ struct qed_roce_ll2_packet {
 
 enum qed_rdma_type {
 	QED_RDMA_TYPE_ROCE,
+	QED_RDMA_TYPE_IWARP
 };
 
 struct qed_dev_rdma_info {
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 04/12] qed: Add iWARP support in ll2 connections

[Michal Kalderon]

Add a new connection type for iWARP ll2 connections for setting
correct ll2 filters and connection type to FW.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_ll2.c | 13 +++++++++++--
 include/linux/qed/qed_ll2_if.h            |  1 +
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index e235fb2..c06ad4f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -896,7 +896,8 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_OOO) ? 0 : 1;
 
 	if ((IS_MF_DEFAULT(p_hwfn) || IS_MF_SI(p_hwfn)) &&
-	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE)) {
+	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE) &&
+	    (conn_type != QED_LL2_TYPE_IWARP)) {
 		p_ramrod->mf_si_bcast_accept_all = 1;
 		p_ramrod->mf_si_mcast_accept_all = 1;
 	} else {
@@ -972,12 +973,20 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 		p_ramrod->conn_type = PROTOCOLID_FCOE;
 		break;
 	case QED_LL2_TYPE_ISCSI:
-	case QED_LL2_TYPE_OOO:
 		p_ramrod->conn_type = PROTOCOLID_ISCSI;
 		break;
 	case QED_LL2_TYPE_ROCE:
 		p_ramrod->conn_type = PROTOCOLID_ROCE;
 		break;
+	case QED_LL2_TYPE_IWARP:
+		p_ramrod->conn_type = PROTOCOLID_IWARP;
+		break;
+	case QED_LL2_TYPE_OOO:
+		if (p_hwfn->hw_info.personality == QED_PCI_ISCSI)
+			p_ramrod->conn_type = PROTOCOLID_ISCSI;
+		else
+			p_ramrod->conn_type = PROTOCOLID_IWARP;
+		break;
 	default:
 		p_ramrod->conn_type = PROTOCOLID_ETH;
 		DP_NOTICE(p_hwfn, "Unknown connection type: %d\n", conn_type);
diff --git a/include/linux/qed/qed_ll2_if.h b/include/linux/qed/qed_ll2_if.h
index c9c56bc..dd7a3b8 100644
--- a/include/linux/qed/qed_ll2_if.h
+++ b/include/linux/qed/qed_ll2_if.h
@@ -50,6 +50,7 @@ enum qed_ll2_conn_type {
 	QED_LL2_TYPE_OOO,
 	QED_LL2_TYPE_RESERVED2,
 	QED_LL2_TYPE_ROCE,
+	QED_LL2_TYPE_IWARP,
 	QED_LL2_TYPE_RESERVED3,
 	MAX_QED_LL2_RX_CONN_TYPE
 };
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 04/12] qed: Add iWARP support in ll2 connections

[Michal Kalderon]

Add a new connection type for iWARP ll2 connections for setting
correct ll2 filters and connection type to FW.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_ll2.c | 13 +++++++++++--
 include/linux/qed/qed_ll2_if.h            |  1 +
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
index e235fb2..c06ad4f 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
@@ -896,7 +896,8 @@ static int qed_sp_ll2_rx_queue_start(struct qed_hwfn *p_hwfn,
 	p_ramrod->main_func_queue = (conn_type == QED_LL2_TYPE_OOO) ? 0 : 1;
 
 	if ((IS_MF_DEFAULT(p_hwfn) || IS_MF_SI(p_hwfn)) &&
-	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE)) {
+	    p_ramrod->main_func_queue && (conn_type != QED_LL2_TYPE_ROCE) &&
+	    (conn_type != QED_LL2_TYPE_IWARP)) {
 		p_ramrod->mf_si_bcast_accept_all = 1;
 		p_ramrod->mf_si_mcast_accept_all = 1;
 	} else {
@@ -972,12 +973,20 @@ static int qed_sp_ll2_tx_queue_start(struct qed_hwfn *p_hwfn,
 		p_ramrod->conn_type = PROTOCOLID_FCOE;
 		break;
 	case QED_LL2_TYPE_ISCSI:
-	case QED_LL2_TYPE_OOO:
 		p_ramrod->conn_type = PROTOCOLID_ISCSI;
 		break;
 	case QED_LL2_TYPE_ROCE:
 		p_ramrod->conn_type = PROTOCOLID_ROCE;
 		break;
+	case QED_LL2_TYPE_IWARP:
+		p_ramrod->conn_type = PROTOCOLID_IWARP;
+		break;
+	case QED_LL2_TYPE_OOO:
+		if (p_hwfn->hw_info.personality == QED_PCI_ISCSI)
+			p_ramrod->conn_type = PROTOCOLID_ISCSI;
+		else
+			p_ramrod->conn_type = PROTOCOLID_IWARP;
+		break;
 	default:
 		p_ramrod->conn_type = PROTOCOLID_ETH;
 		DP_NOTICE(p_hwfn, "Unknown connection type: %d\n", conn_type);
diff --git a/include/linux/qed/qed_ll2_if.h b/include/linux/qed/qed_ll2_if.h
index c9c56bc..dd7a3b8 100644
--- a/include/linux/qed/qed_ll2_if.h
+++ b/include/linux/qed/qed_ll2_if.h
@@ -50,6 +50,7 @@ enum qed_ll2_conn_type {
 	QED_LL2_TYPE_OOO,
 	QED_LL2_TYPE_RESERVED2,
 	QED_LL2_TYPE_ROCE,
+	QED_LL2_TYPE_IWARP,
 	QED_LL2_TYPE_RESERVED3,
 	MAX_QED_LL2_RX_CONN_TYPE
 };
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 05/12] qed: iWARP CM - setup a ll2 connection for handling SYN packets

[Michal Kalderon]

iWARP handles incoming SYN packets using the ll2 interface. This patch
implements ll2 setup and teardown. Additional ll2 connections will
be used in the future which are not part of this patch series.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 211 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  12 ++
 2 files changed, 220 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a8bd5f8..f3b4b32 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -29,8 +29,11 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
 #include "qed_cxt.h"
 #include "qed_hw.h"
+#include "qed_ll2.h"
 #include "qed_rdma.h"
 #include "qed_reg_addr.h"
 #include "qed_sp.h"
@@ -474,12 +477,214 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
 }
 
+static int
+qed_iwarp_ll2_post_rx(struct qed_hwfn *p_hwfn,
+		      struct qed_iwarp_ll2_buff *buf, u8 handle)
+{
+	int rc;
+
+	rc = qed_ll2_post_rx_buffer(p_hwfn, handle, buf->data_phys_addr,
+				    (u16)buf->buff_size, buf, 1);
+	if (rc) {
+		DP_NOTICE(p_hwfn,
+			  "Failed to repost rx buffer to ll2 rc = %d, handle=%d\n",
+			  rc, handle);
+		dma_free_coherent(&p_hwfn->cdev->pdev->dev, buf->buff_size,
+				  buf->data, buf->data_phys_addr);
+		kfree(buf);
+	}
+
+	return rc;
+}
+
+static void
+qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
+{
+	struct qed_iwarp_ll2_buff *buf = data->cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	if (GET_FIELD(data->parse_flags,
+		      PARSING_AND_ERR_FLAGS_L4CHKSMWASCALCULATED) &&
+	    GET_FIELD(data->parse_flags, PARSING_AND_ERR_FLAGS_L4CHKSMERROR)) {
+		DP_NOTICE(p_hwfn, "Syn packet received with checksum error\n");
+		goto err;
+	}
+
+	/* Process SYN packet - added later on in series */
+
+err:
+	qed_iwarp_ll2_post_rx(p_hwfn, buf,
+			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+}
+
+static void qed_iwarp_ll2_rel_rx_pkt(void *cxt, u8 connection_handle,
+				     void *cookie, dma_addr_t rx_buf_addr,
+				     bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size,
+			  buffer->data, buffer->data_phys_addr);
+	kfree(buffer);
+}
+
+static void qed_iwarp_ll2_comp_tx_pkt(void *cxt, u8 connection_handle,
+				      void *cookie, dma_addr_t first_frag_addr,
+				      bool b_last_fragment, bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	/* this was originally an rx packet, post it back */
+	qed_iwarp_ll2_post_rx(p_hwfn, buffer, connection_handle);
+}
+
+static void qed_iwarp_ll2_rel_tx_pkt(void *cxt, u8 connection_handle,
+				     void *cookie, dma_addr_t first_frag_addr,
+				     bool b_last_fragment, bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	if (!buffer)
+		return;
+
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size,
+			  buffer->data, buffer->data_phys_addr);
+
+	kfree(buffer);
+}
+
+static int qed_iwarp_ll2_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	int rc = 0;
+
+	if (iwarp_info->ll2_syn_handle != QED_IWARP_HANDLE_INVAL) {
+		rc = qed_ll2_terminate_connection(p_hwfn,
+						  iwarp_info->ll2_syn_handle);
+		if (rc)
+			DP_INFO(p_hwfn, "Failed to terminate syn connection\n");
+
+		qed_ll2_release_connection(p_hwfn, iwarp_info->ll2_syn_handle);
+		iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL;
+	}
+
+	qed_llh_remove_mac_filter(p_hwfn,
+				  p_ptt, p_hwfn->p_rdma_info->iwarp.mac_addr);
+	return rc;
+}
+
+static int
+qed_iwarp_ll2_alloc_buffers(struct qed_hwfn *p_hwfn,
+			    int num_rx_bufs, int buff_size, u8 ll2_handle)
+{
+	struct qed_iwarp_ll2_buff *buffer;
+	int rc = 0;
+	int i;
+
+	for (i = 0; i < num_rx_bufs; i++) {
+		buffer = kzalloc(sizeof(*buffer), GFP_KERNEL);
+		if (!buffer) {
+			rc = -ENOMEM;
+			break;
+		}
+
+		buffer->data = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+						  buff_size,
+						  &buffer->data_phys_addr,
+						  GFP_KERNEL);
+		if (!buffer->data) {
+			kfree(buffer);
+			rc = -ENOMEM;
+			break;
+		}
+
+		buffer->buff_size = buff_size;
+		rc = qed_iwarp_ll2_post_rx(p_hwfn, buffer, ll2_handle);
+		if (rc)
+			/* buffers will be deallocated by qed_ll2 */
+			break;
+	}
+	return rc;
+}
+
+#define QED_IWARP_MAX_BUF_SIZE(mtu)				     \
+	ALIGN((mtu) + ETH_HLEN + 2 * VLAN_HLEN + 2 + ETH_CACHE_LINE_SIZE, \
+		ETH_CACHE_LINE_SIZE)
+
+static int
+qed_iwarp_ll2_start(struct qed_hwfn *p_hwfn,
+		    struct qed_rdma_start_in_params *params,
+		    struct qed_ptt *p_ptt)
+{
+	struct qed_iwarp_info *iwarp_info;
+	struct qed_ll2_acquire_data data;
+	struct qed_ll2_cbs cbs;
+	int rc = 0;
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL;
+
+	iwarp_info->max_mtu = params->max_mtu;
+
+	ether_addr_copy(p_hwfn->p_rdma_info->iwarp.mac_addr, params->mac_addr);
+
+	rc = qed_llh_add_mac_filter(p_hwfn, p_ptt, params->mac_addr);
+	if (rc)
+		return rc;
+
+	/* Start SYN connection */
+	cbs.rx_comp_cb = qed_iwarp_ll2_comp_syn_pkt;
+	cbs.rx_release_cb = qed_iwarp_ll2_rel_rx_pkt;
+	cbs.tx_comp_cb = qed_iwarp_ll2_comp_tx_pkt;
+	cbs.tx_release_cb = qed_iwarp_ll2_rel_tx_pkt;
+	cbs.cookie = p_hwfn;
+
+	memset(&data, 0, sizeof(data));
+	data.input.conn_type = QED_LL2_TYPE_IWARP;
+	data.input.mtu = QED_IWARP_MAX_SYN_PKT_SIZE;
+	data.input.rx_num_desc = QED_IWARP_LL2_SYN_RX_SIZE;
+	data.input.tx_num_desc = QED_IWARP_LL2_SYN_TX_SIZE;
+	data.input.tx_max_bds_per_packet = 1;	/* will never be fragmented */
+	data.input.tx_tc = PKT_LB_TC;
+	data.input.tx_dest = QED_LL2_TX_DEST_LB;
+	data.p_connection_handle = &iwarp_info->ll2_syn_handle;
+	data.cbs = &cbs;
+
+	rc = qed_ll2_acquire_connection(p_hwfn, &data);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed to acquire LL2 connection\n");
+		qed_llh_remove_mac_filter(p_hwfn, p_ptt, params->mac_addr);
+		return rc;
+	}
+
+	rc = qed_ll2_establish_connection(p_hwfn, iwarp_info->ll2_syn_handle);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed to establish LL2 connection\n");
+		goto err;
+	}
+
+	rc = qed_iwarp_ll2_alloc_buffers(p_hwfn,
+					 QED_IWARP_LL2_SYN_RX_SIZE,
+					 QED_IWARP_MAX_SYN_PKT_SIZE,
+					 iwarp_info->ll2_syn_handle);
+	if (rc)
+		goto err;
+
+	return rc;
+err:
+	qed_iwarp_ll2_stop(p_hwfn, p_ptt);
+
+	return rc;
+}
+
 int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 		    struct qed_rdma_start_in_params *params)
 {
 	struct qed_iwarp_info *iwarp_info;
 	u32 rcv_wnd_size;
-	int rc = 0;
 
 	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
 
@@ -499,7 +704,7 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
 				  qed_iwarp_async_event);
 
-	return rc;
+	return qed_iwarp_ll2_start(p_hwfn, params, p_ptt);
 }
 
 int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
@@ -512,7 +717,7 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 
 	qed_spq_unregister_async_cb(p_hwfn, PROTOCOLID_IWARP);
 
-	return 0;
+	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
 static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 05e5e45..068b859 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -42,6 +42,17 @@ enum qed_iwarp_qp_state {
 
 enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
 
+#define QED_IWARP_LL2_SYN_TX_SIZE       (128)
+#define QED_IWARP_LL2_SYN_RX_SIZE       (256)
+#define QED_IWARP_MAX_SYN_PKT_SIZE      (128)
+#define QED_IWARP_HANDLE_INVAL			(0xff)
+
+struct qed_iwarp_ll2_buff {
+	void *data;
+	dma_addr_t data_phys_addr;
+	u32 buff_size;
+};
+
 struct qed_iwarp_info {
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
@@ -50,6 +61,7 @@ struct qed_iwarp_info {
 	u8 mac_addr[ETH_ALEN];
 	u8 crc_needed;
 	u8 tcp_flags;
+	u8 ll2_syn_handle;
 	u8 peer2peer;
 	enum mpa_negotiation_mode mpa_rev;
 	enum mpa_rtr_type rtr_type;
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 05/12] qed: iWARP CM - setup a ll2 connection for handling SYN packets

[Michal Kalderon]

iWARP handles incoming SYN packets using the ll2 interface. This patch
implements ll2 setup and teardown. Additional ll2 connections will
be used in the future which are not part of this patch series.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 211 +++++++++++++++++++++++++++-
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |  12 ++
 2 files changed, 220 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a8bd5f8..f3b4b32 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -29,8 +29,11 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
 #include "qed_cxt.h"
 #include "qed_hw.h"
+#include "qed_ll2.h"
 #include "qed_rdma.h"
 #include "qed_reg_addr.h"
 #include "qed_sp.h"
@@ -474,12 +477,214 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 {
 }
 
+static int
+qed_iwarp_ll2_post_rx(struct qed_hwfn *p_hwfn,
+		      struct qed_iwarp_ll2_buff *buf, u8 handle)
+{
+	int rc;
+
+	rc = qed_ll2_post_rx_buffer(p_hwfn, handle, buf->data_phys_addr,
+				    (u16)buf->buff_size, buf, 1);
+	if (rc) {
+		DP_NOTICE(p_hwfn,
+			  "Failed to repost rx buffer to ll2 rc = %d, handle=%d\n",
+			  rc, handle);
+		dma_free_coherent(&p_hwfn->cdev->pdev->dev, buf->buff_size,
+				  buf->data, buf->data_phys_addr);
+		kfree(buf);
+	}
+
+	return rc;
+}
+
+static void
+qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
+{
+	struct qed_iwarp_ll2_buff *buf = data->cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	if (GET_FIELD(data->parse_flags,
+		      PARSING_AND_ERR_FLAGS_L4CHKSMWASCALCULATED) &&
+	    GET_FIELD(data->parse_flags, PARSING_AND_ERR_FLAGS_L4CHKSMERROR)) {
+		DP_NOTICE(p_hwfn, "Syn packet received with checksum error\n");
+		goto err;
+	}
+
+	/* Process SYN packet - added later on in series */
+
+err:
+	qed_iwarp_ll2_post_rx(p_hwfn, buf,
+			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
+}
+
+static void qed_iwarp_ll2_rel_rx_pkt(void *cxt, u8 connection_handle,
+				     void *cookie, dma_addr_t rx_buf_addr,
+				     bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size,
+			  buffer->data, buffer->data_phys_addr);
+	kfree(buffer);
+}
+
+static void qed_iwarp_ll2_comp_tx_pkt(void *cxt, u8 connection_handle,
+				      void *cookie, dma_addr_t first_frag_addr,
+				      bool b_last_fragment, bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	/* this was originally an rx packet, post it back */
+	qed_iwarp_ll2_post_rx(p_hwfn, buffer, connection_handle);
+}
+
+static void qed_iwarp_ll2_rel_tx_pkt(void *cxt, u8 connection_handle,
+				     void *cookie, dma_addr_t first_frag_addr,
+				     bool b_last_fragment, bool b_last_packet)
+{
+	struct qed_iwarp_ll2_buff *buffer = cookie;
+	struct qed_hwfn *p_hwfn = cxt;
+
+	if (!buffer)
+		return;
+
+	dma_free_coherent(&p_hwfn->cdev->pdev->dev, buffer->buff_size,
+			  buffer->data, buffer->data_phys_addr);
+
+	kfree(buffer);
+}
+
+static int qed_iwarp_ll2_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	struct qed_iwarp_info *iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	int rc = 0;
+
+	if (iwarp_info->ll2_syn_handle != QED_IWARP_HANDLE_INVAL) {
+		rc = qed_ll2_terminate_connection(p_hwfn,
+						  iwarp_info->ll2_syn_handle);
+		if (rc)
+			DP_INFO(p_hwfn, "Failed to terminate syn connection\n");
+
+		qed_ll2_release_connection(p_hwfn, iwarp_info->ll2_syn_handle);
+		iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL;
+	}
+
+	qed_llh_remove_mac_filter(p_hwfn,
+				  p_ptt, p_hwfn->p_rdma_info->iwarp.mac_addr);
+	return rc;
+}
+
+static int
+qed_iwarp_ll2_alloc_buffers(struct qed_hwfn *p_hwfn,
+			    int num_rx_bufs, int buff_size, u8 ll2_handle)
+{
+	struct qed_iwarp_ll2_buff *buffer;
+	int rc = 0;
+	int i;
+
+	for (i = 0; i < num_rx_bufs; i++) {
+		buffer = kzalloc(sizeof(*buffer), GFP_KERNEL);
+		if (!buffer) {
+			rc = -ENOMEM;
+			break;
+		}
+
+		buffer->data = dma_alloc_coherent(&p_hwfn->cdev->pdev->dev,
+						  buff_size,
+						  &buffer->data_phys_addr,
+						  GFP_KERNEL);
+		if (!buffer->data) {
+			kfree(buffer);
+			rc = -ENOMEM;
+			break;
+		}
+
+		buffer->buff_size = buff_size;
+		rc = qed_iwarp_ll2_post_rx(p_hwfn, buffer, ll2_handle);
+		if (rc)
+			/* buffers will be deallocated by qed_ll2 */
+			break;
+	}
+	return rc;
+}
+
+#define QED_IWARP_MAX_BUF_SIZE(mtu)				     \
+	ALIGN((mtu) + ETH_HLEN + 2 * VLAN_HLEN + 2 + ETH_CACHE_LINE_SIZE, \
+		ETH_CACHE_LINE_SIZE)
+
+static int
+qed_iwarp_ll2_start(struct qed_hwfn *p_hwfn,
+		    struct qed_rdma_start_in_params *params,
+		    struct qed_ptt *p_ptt)
+{
+	struct qed_iwarp_info *iwarp_info;
+	struct qed_ll2_acquire_data data;
+	struct qed_ll2_cbs cbs;
+	int rc = 0;
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+	iwarp_info->ll2_syn_handle = QED_IWARP_HANDLE_INVAL;
+
+	iwarp_info->max_mtu = params->max_mtu;
+
+	ether_addr_copy(p_hwfn->p_rdma_info->iwarp.mac_addr, params->mac_addr);
+
+	rc = qed_llh_add_mac_filter(p_hwfn, p_ptt, params->mac_addr);
+	if (rc)
+		return rc;
+
+	/* Start SYN connection */
+	cbs.rx_comp_cb = qed_iwarp_ll2_comp_syn_pkt;
+	cbs.rx_release_cb = qed_iwarp_ll2_rel_rx_pkt;
+	cbs.tx_comp_cb = qed_iwarp_ll2_comp_tx_pkt;
+	cbs.tx_release_cb = qed_iwarp_ll2_rel_tx_pkt;
+	cbs.cookie = p_hwfn;
+
+	memset(&data, 0, sizeof(data));
+	data.input.conn_type = QED_LL2_TYPE_IWARP;
+	data.input.mtu = QED_IWARP_MAX_SYN_PKT_SIZE;
+	data.input.rx_num_desc = QED_IWARP_LL2_SYN_RX_SIZE;
+	data.input.tx_num_desc = QED_IWARP_LL2_SYN_TX_SIZE;
+	data.input.tx_max_bds_per_packet = 1;	/* will never be fragmented */
+	data.input.tx_tc = PKT_LB_TC;
+	data.input.tx_dest = QED_LL2_TX_DEST_LB;
+	data.p_connection_handle = &iwarp_info->ll2_syn_handle;
+	data.cbs = &cbs;
+
+	rc = qed_ll2_acquire_connection(p_hwfn, &data);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed to acquire LL2 connection\n");
+		qed_llh_remove_mac_filter(p_hwfn, p_ptt, params->mac_addr);
+		return rc;
+	}
+
+	rc = qed_ll2_establish_connection(p_hwfn, iwarp_info->ll2_syn_handle);
+	if (rc) {
+		DP_NOTICE(p_hwfn, "Failed to establish LL2 connection\n");
+		goto err;
+	}
+
+	rc = qed_iwarp_ll2_alloc_buffers(p_hwfn,
+					 QED_IWARP_LL2_SYN_RX_SIZE,
+					 QED_IWARP_MAX_SYN_PKT_SIZE,
+					 iwarp_info->ll2_syn_handle);
+	if (rc)
+		goto err;
+
+	return rc;
+err:
+	qed_iwarp_ll2_stop(p_hwfn, p_ptt);
+
+	return rc;
+}
+
 int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 		    struct qed_rdma_start_in_params *params)
 {
 	struct qed_iwarp_info *iwarp_info;
 	u32 rcv_wnd_size;
-	int rc = 0;
 
 	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
 
@@ -499,7 +704,7 @@ int qed_iwarp_setup(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
 	qed_spq_register_async_cb(p_hwfn, PROTOCOLID_IWARP,
 				  qed_iwarp_async_event);
 
-	return rc;
+	return qed_iwarp_ll2_start(p_hwfn, params, p_ptt);
 }
 
 int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
@@ -512,7 +717,7 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 
 	qed_spq_unregister_async_cb(p_hwfn, PROTOCOLID_IWARP);
 
-	return 0;
+	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
 static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index 05e5e45..068b859 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -42,6 +42,17 @@ enum qed_iwarp_qp_state {
 
 enum qed_iwarp_qp_state qed_roce2iwarp_state(enum qed_roce_qp_state state);
 
+#define QED_IWARP_LL2_SYN_TX_SIZE       (128)
+#define QED_IWARP_LL2_SYN_RX_SIZE       (256)
+#define QED_IWARP_MAX_SYN_PKT_SIZE      (128)
+#define QED_IWARP_HANDLE_INVAL			(0xff)
+
+struct qed_iwarp_ll2_buff {
+	void *data;
+	dma_addr_t data_phys_addr;
+	u32 buff_size;
+};
+
 struct qed_iwarp_info {
 	spinlock_t iw_lock;	/* for iwarp resources */
 	spinlock_t qp_lock;	/* for teardown races */
@@ -50,6 +61,7 @@ struct qed_iwarp_info {
 	u8 mac_addr[ETH_ALEN];
 	u8 crc_needed;
 	u8 tcp_flags;
+	u8 ll2_syn_handle;
 	u8 peer2peer;
 	enum mpa_negotiation_mode mpa_rev;
 	enum mpa_rtr_type rtr_type;
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 08/12] qed: iWARP CM add active side connect

[Michal Kalderon]

This patch implements the active side connect.
Offload a connection, process MPA reply and send RTR.
In some of the common passive/active functions, the active side
will work in blocking mode.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 240 ++++++++++++++++++++++++++--
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |   7 +
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |   4 +
 include/linux/qed/qed_rdma_if.h             |  26 +++
 4 files changed, 265 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a6dadae..a5da9fc 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -611,7 +611,10 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	memset(&init_data, 0, sizeof(init_data));
 	init_data.cid = ep->tcp_cid;
 	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
-	init_data.comp_mode = QED_SPQ_MODE_CB;
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE)
+		init_data.comp_mode = QED_SPQ_MODE_CB;
+	else
+		init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
 
 	rc = qed_sp_init_request(p_hwfn, &p_ent,
 				 IWARP_RAMROD_CMD_ID_TCP_OFFLOAD,
@@ -711,7 +714,7 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 		   "private_data_len=%x handshake_mode=%x private_data=(%x)\n",
 		   async_data->mpa_request.ulp_data_len,
-		   mpa_rev, *((u32 *)((u8 *)ep->ep_buffer_virt->in_pdata)));
+		   mpa_rev, *((u32 *)(ep->ep_buffer_virt->in_pdata)));
 
 	if (mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
 		/* Read ord/ird values from private data buffer */
@@ -801,7 +804,10 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	init_data.cid = reject ? ep->tcp_cid : qp->icid;
 	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
 
-	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE)
+		init_data.comp_mode = QED_SPQ_MODE_CB;
+	else
+		init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
 
 	rc = qed_sp_init_request(p_hwfn, &p_ent,
 				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD,
@@ -890,6 +896,59 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
 }
 
+void
+qed_iwarp_parse_private_data(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct mpa_v2_hdr *mpa_v2_params;
+	union async_output *async_data;
+	u16 mpa_ird, mpa_ord;
+	u8 mpa_data_size = 0;
+
+	if (MPA_REV2(p_hwfn->p_rdma_info->iwarp.mpa_rev)) {
+		mpa_v2_params =
+			(struct mpa_v2_hdr *)(ep->ep_buffer_virt->in_pdata);
+		mpa_data_size = sizeof(*mpa_v2_params);
+		mpa_ird = ntohs(mpa_v2_params->ird);
+		mpa_ord = ntohs(mpa_v2_params->ord);
+
+		ep->cm_info.ird = (u8)(mpa_ord & MPA_V2_IRD_ORD_MASK);
+		ep->cm_info.ord = (u8)(mpa_ird & MPA_V2_IRD_ORD_MASK);
+	}
+	async_data = &ep->ep_buffer_virt->async_output;
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_data_size;
+	ep->cm_info.private_data_len = async_data->mpa_response.ulp_data_len -
+				       mpa_data_size;
+}
+
+void
+qed_iwarp_mpa_reply_arrived(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		DP_NOTICE(p_hwfn,
+			  "MPA reply event not expected on passive side!\n");
+		return;
+	}
+
+	params.event = QED_IWARP_EVENT_ACTIVE_MPA_REPLY;
+
+	qed_iwarp_parse_private_data(p_hwfn, ep);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
+		   ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird);
+
+	params.cm_info = &ep->cm_info;
+	params.ep_context = ep;
+	params.status = 0;
+
+	ep->mpa_reply_processed = true;
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
 #define QED_IWARP_CONNECT_MODE_STRING(ep) \
 	((ep)->connect_mode == TCP_CONNECT_PASSIVE) ? "Passive" : "Active"
 
@@ -902,7 +961,13 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
 	struct qed_iwarp_cm_event_params params;
 
-	params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE)
+		params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE;
+	else
+		params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE && !ep->mpa_reply_processed)
+		qed_iwarp_parse_private_data(p_hwfn, ep);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
@@ -977,6 +1042,102 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	}
 }
 
+int qed_iwarp_connect(void *rdma_cxt,
+		      struct qed_iwarp_connect_in *iparams,
+		      struct qed_iwarp_connect_out *oparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_info *iwarp_info;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+	u8 ts_hdr_size = 0;
+	u32 cid;
+	int rc;
+
+	if ((iparams->cm_info.ord > QED_IWARP_ORD_DEFAULT) ||
+	    (iparams->cm_info.ird > QED_IWARP_IRD_DEFAULT)) {
+		DP_NOTICE(p_hwfn,
+			  "QP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n",
+			  iparams->qp->icid, iparams->cm_info.ord,
+			  iparams->cm_info.ird);
+
+		return -EINVAL;
+	}
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+
+	/* Allocate ep object */
+	rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+	if (rc)
+		return rc;
+
+	rc = qed_iwarp_create_ep(p_hwfn, &ep);
+	if (rc)
+		goto err;
+
+	ep->tcp_cid = cid;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ep->qp = iparams->qp;
+	ep->qp->ep = ep;
+	ether_addr_copy(ep->remote_mac_addr, iparams->remote_mac_addr);
+	ether_addr_copy(ep->local_mac_addr, iparams->local_mac_addr);
+	memcpy(&ep->cm_info, &iparams->cm_info, sizeof(ep->cm_info));
+
+	ep->cm_info.ord = iparams->cm_info.ord;
+	ep->cm_info.ird = iparams->cm_info.ird;
+
+	ep->rtr_type = iwarp_info->rtr_type;
+	if (!iwarp_info->peer2peer)
+		ep->rtr_type = MPA_RTR_TYPE_NONE;
+
+	if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) && (ep->cm_info.ord == 0))
+		ep->cm_info.ord = 1;
+
+	ep->mpa_rev = iwarp_info->mpa_rev;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->cm_info.private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->cm_info.private_data,
+	       iparams->cm_info.private_data_len);
+
+	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
+		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
+
+	ep->mss = iparams->mss - ts_hdr_size;
+	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
+
+	ep->event_cb = iparams->event_cb;
+	ep->cb_context = iparams->cb_context;
+	ep->connect_mode = TCP_CONNECT_ACTIVE;
+
+	oparams->ep_context = ep;
+
+	rc = qed_iwarp_tcp_offload(p_hwfn, ep);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x) rc = %d\n",
+		   iparams->qp->icid, ep->tcp_cid, rc);
+
+	if (rc) {
+		qed_iwarp_destroy_ep(p_hwfn, ep, true);
+		goto err;
+	}
+
+	return rc;
+err:
+	qed_iwarp_cid_cleaned(p_hwfn, cid);
+
+	return rc;
+}
+
 static struct qed_iwarp_ep *qed_iwarp_get_free_ep(struct qed_hwfn *p_hwfn)
 {
 	struct qed_iwarp_ep *ep = NULL;
@@ -1174,12 +1335,12 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 
 int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams)
 {
-	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
+	struct qed_hwfn *p_hwfn = rdma_cxt;
 	struct qed_iwarp_ep *ep;
 	u8 mpa_data_size = 0;
 	int rc;
 
-	ep = (struct qed_iwarp_ep *)iparams->ep_context;
+	ep = iparams->ep_context;
 	if (!ep) {
 		DP_ERR(p_hwfn, "Ep Context receive in accept is NULL\n");
 		return -EINVAL;
@@ -1799,13 +1960,19 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
 {
-	/* Done with the SYN packet, post back to ll2 rx */
-	qed_iwarp_ll2_post_rx(p_hwfn, ep->syn,
-			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
-	ep->syn = NULL;
+	u8 ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle;
 
-	/* If connect failed - upper layer doesn't know about it */
-	qed_iwarp_mpa_received(p_hwfn, ep);
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		/* Done with the SYN packet, post back to ll2 rx */
+		qed_iwarp_ll2_post_rx(p_hwfn, ep->syn, ll2_syn_handle);
+
+		ep->syn = NULL;
+
+		/* If connect failed - upper layer doesn't know about it */
+		qed_iwarp_mpa_received(p_hwfn, ep);
+	} else {
+		qed_iwarp_mpa_offload(p_hwfn, ep);
+	}
 }
 
 static inline bool
@@ -1842,6 +2009,16 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 			   ep->tcp_cid, fw_return_code);
 		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
 		break;
+		/* Async event for active side only */
+	case IWARP_EVENT_TYPE_ASYNC_ENHANCED_MPA_REPLY_ARRIVED:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_MPA_REPLY_ARRIVED fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_mpa_reply_arrived(p_hwfn, ep);
+		break;
 	case IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE:
 		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
 			return -EINVAL;
@@ -1918,6 +2095,45 @@ int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle)
 	return 0;
 }
 
+int qed_iwarp_send_rtr(void *rdma_cxt, struct qed_iwarp_send_rtr_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	struct qed_iwarp_ep *ep;
+	struct qed_rdma_qp *qp;
+	int rc;
+
+	ep = iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in send_rtr is NULL\n");
+		return -EINVAL;
+	}
+
+	qp = ep->qp;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n",
+		   qp->icid, ep->tcp_cid);
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_CB;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD_SEND_RTR,
+				 PROTOCOLID_IWARP, &init_data);
+
+	if (rc)
+		return rc;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "rc = 0x%x\n", rc);
+
+	return rc;
+}
+
 void
 qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 		   struct qed_rdma_query_qp_out_params *out_params)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index bedac98..148ef3c 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -170,6 +170,11 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_rdma_query_qp_out_params *out_params);
 
 int
+qed_iwarp_connect(void *rdma_cxt,
+		  struct qed_iwarp_connect_in *iparams,
+		  struct qed_iwarp_connect_out *oparams);
+
+int
 qed_iwarp_create_listen(void *rdma_cxt,
 			struct qed_iwarp_listen_in *iparams,
 			struct qed_iwarp_listen_out *oparams);
@@ -179,4 +184,6 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams);
 int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
 
+int qed_iwarp_send_rtr(void *rdma_cxt, struct qed_iwarp_send_rtr_in *iparams);
+
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index 29de915..6fb9951 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -1772,8 +1772,12 @@ static int qed_roce_ll2_set_mac_filter(struct qed_dev *cdev,
 	.ll2_set_fragment_of_tx_packet = &qed_ll2_set_fragment_of_tx_packet,
 	.ll2_set_mac_filter = &qed_roce_ll2_set_mac_filter,
 	.ll2_get_stats = &qed_ll2_get_stats,
+	.iwarp_connect = &qed_iwarp_connect,
 	.iwarp_create_listen = &qed_iwarp_create_listen,
 	.iwarp_destroy_listen = &qed_iwarp_destroy_listen,
+	.iwarp_accept = &qed_iwarp_accept,
+	.iwarp_reject = &qed_iwarp_reject,
+	.iwarp_send_rtr = &qed_iwarp_send_rtr,
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void)
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index c4c241f..e9514a6 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -473,6 +473,8 @@ struct qed_rdma_counters_out_params {
 enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
 	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
+	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
+	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
 };
 
 enum qed_tcp_ip_version {
@@ -503,6 +505,20 @@ struct qed_iwarp_cm_event_params {
 typedef int (*iwarp_event_handler) (void *context,
 				    struct qed_iwarp_cm_event_params *event);
 
+struct qed_iwarp_connect_in {
+	iwarp_event_handler event_cb;
+	void *cb_context;
+	struct qed_rdma_qp *qp;
+	struct qed_iwarp_cm_info cm_info;
+	u16 mss;
+	u8 remote_mac_addr[ETH_ALEN];
+	u8 local_mac_addr[ETH_ALEN];
+};
+
+struct qed_iwarp_connect_out {
+	void *ep_context;
+};
+
 struct qed_iwarp_listen_in {
 	iwarp_event_handler event_cb;
 	void *cb_context;	/* passed to event_cb */
@@ -534,6 +550,10 @@ struct qed_iwarp_reject_in {
 	u16 private_data_len;
 };
 
+struct qed_iwarp_send_rtr_in {
+	void *ep_context;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -640,6 +660,10 @@ struct qed_rdma_ops {
 	int (*ll2_set_mac_filter)(struct qed_dev *cdev,
 				  u8 *old_mac_address, u8 *new_mac_address);
 
+	int (*iwarp_connect)(void *rdma_cxt,
+			     struct qed_iwarp_connect_in *iparams,
+			     struct qed_iwarp_connect_out *oparams);
+
 	int (*iwarp_create_listen)(void *rdma_cxt,
 				   struct qed_iwarp_listen_in *iparams,
 				   struct qed_iwarp_listen_out *oparams);
@@ -652,6 +676,8 @@ struct qed_rdma_ops {
 
 	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
 
+	int (*iwarp_send_rtr)(void *rdma_cxt,
+			      struct qed_iwarp_send_rtr_in *iparams);
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void);
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 08/12] qed: iWARP CM add active side connect

[Michal Kalderon]

This patch implements the active side connect.
Offload a connection, process MPA reply and send RTR.
In some of the common passive/active functions, the active side
will work in blocking mode.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 240 ++++++++++++++++++++++++++--
 drivers/net/ethernet/qlogic/qed/qed_iwarp.h |   7 +
 drivers/net/ethernet/qlogic/qed/qed_rdma.c  |   4 +
 include/linux/qed/qed_rdma_if.h             |  26 +++
 4 files changed, 265 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a6dadae..a5da9fc 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -611,7 +611,10 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	memset(&init_data, 0, sizeof(init_data));
 	init_data.cid = ep->tcp_cid;
 	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
-	init_data.comp_mode = QED_SPQ_MODE_CB;
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE)
+		init_data.comp_mode = QED_SPQ_MODE_CB;
+	else
+		init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
 
 	rc = qed_sp_init_request(p_hwfn, &p_ent,
 				 IWARP_RAMROD_CMD_ID_TCP_OFFLOAD,
@@ -711,7 +714,7 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 		   "private_data_len=%x handshake_mode=%x private_data=(%x)\n",
 		   async_data->mpa_request.ulp_data_len,
-		   mpa_rev, *((u32 *)((u8 *)ep->ep_buffer_virt->in_pdata)));
+		   mpa_rev, *((u32 *)(ep->ep_buffer_virt->in_pdata)));
 
 	if (mpa_rev == MPA_NEGOTIATION_TYPE_ENHANCED) {
 		/* Read ord/ird values from private data buffer */
@@ -801,7 +804,10 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	init_data.cid = reject ? ep->tcp_cid : qp->icid;
 	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
 
-	init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE)
+		init_data.comp_mode = QED_SPQ_MODE_CB;
+	else
+		init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
 
 	rc = qed_sp_init_request(p_hwfn, &p_ent,
 				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD,
@@ -890,6 +896,59 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
 }
 
+void
+qed_iwarp_parse_private_data(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct mpa_v2_hdr *mpa_v2_params;
+	union async_output *async_data;
+	u16 mpa_ird, mpa_ord;
+	u8 mpa_data_size = 0;
+
+	if (MPA_REV2(p_hwfn->p_rdma_info->iwarp.mpa_rev)) {
+		mpa_v2_params =
+			(struct mpa_v2_hdr *)(ep->ep_buffer_virt->in_pdata);
+		mpa_data_size = sizeof(*mpa_v2_params);
+		mpa_ird = ntohs(mpa_v2_params->ird);
+		mpa_ord = ntohs(mpa_v2_params->ord);
+
+		ep->cm_info.ird = (u8)(mpa_ord & MPA_V2_IRD_ORD_MASK);
+		ep->cm_info.ord = (u8)(mpa_ird & MPA_V2_IRD_ORD_MASK);
+	}
+	async_data = &ep->ep_buffer_virt->async_output;
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->in_pdata + mpa_data_size;
+	ep->cm_info.private_data_len = async_data->mpa_response.ulp_data_len -
+				       mpa_data_size;
+}
+
+void
+qed_iwarp_mpa_reply_arrived(struct qed_hwfn *p_hwfn, struct qed_iwarp_ep *ep)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		DP_NOTICE(p_hwfn,
+			  "MPA reply event not expected on passive side!\n");
+		return;
+	}
+
+	params.event = QED_IWARP_EVENT_ACTIVE_MPA_REPLY;
+
+	qed_iwarp_parse_private_data(p_hwfn, ep);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
+		   ep->mpa_rev, ep->cm_info.ord, ep->cm_info.ird);
+
+	params.cm_info = &ep->cm_info;
+	params.ep_context = ep;
+	params.status = 0;
+
+	ep->mpa_reply_processed = true;
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
 #define QED_IWARP_CONNECT_MODE_STRING(ep) \
 	((ep)->connect_mode == TCP_CONNECT_PASSIVE) ? "Passive" : "Active"
 
@@ -902,7 +961,13 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
 	struct qed_iwarp_cm_event_params params;
 
-	params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE)
+		params.event = QED_IWARP_EVENT_ACTIVE_COMPLETE;
+	else
+		params.event = QED_IWARP_EVENT_PASSIVE_COMPLETE;
+
+	if (ep->connect_mode == TCP_CONNECT_ACTIVE && !ep->mpa_reply_processed)
+		qed_iwarp_parse_private_data(p_hwfn, ep);
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
 		   "MPA_NEGOTIATE (v%d): ORD: 0x%x IRD: 0x%x\n",
@@ -977,6 +1042,102 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 	}
 }
 
+int qed_iwarp_connect(void *rdma_cxt,
+		      struct qed_iwarp_connect_in *iparams,
+		      struct qed_iwarp_connect_out *oparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_iwarp_info *iwarp_info;
+	struct qed_iwarp_ep *ep;
+	u8 mpa_data_size = 0;
+	u8 ts_hdr_size = 0;
+	u32 cid;
+	int rc;
+
+	if ((iparams->cm_info.ord > QED_IWARP_ORD_DEFAULT) ||
+	    (iparams->cm_info.ird > QED_IWARP_IRD_DEFAULT)) {
+		DP_NOTICE(p_hwfn,
+			  "QP(0x%x) ERROR: Invalid ord(0x%x)/ird(0x%x)\n",
+			  iparams->qp->icid, iparams->cm_info.ord,
+			  iparams->cm_info.ird);
+
+		return -EINVAL;
+	}
+
+	iwarp_info = &p_hwfn->p_rdma_info->iwarp;
+
+	/* Allocate ep object */
+	rc = qed_iwarp_alloc_cid(p_hwfn, &cid);
+	if (rc)
+		return rc;
+
+	rc = qed_iwarp_create_ep(p_hwfn, &ep);
+	if (rc)
+		goto err;
+
+	ep->tcp_cid = cid;
+
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_add_tail(&ep->list_entry, &p_hwfn->p_rdma_info->iwarp.ep_list);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ep->qp = iparams->qp;
+	ep->qp->ep = ep;
+	ether_addr_copy(ep->remote_mac_addr, iparams->remote_mac_addr);
+	ether_addr_copy(ep->local_mac_addr, iparams->local_mac_addr);
+	memcpy(&ep->cm_info, &iparams->cm_info, sizeof(ep->cm_info));
+
+	ep->cm_info.ord = iparams->cm_info.ord;
+	ep->cm_info.ird = iparams->cm_info.ird;
+
+	ep->rtr_type = iwarp_info->rtr_type;
+	if (!iwarp_info->peer2peer)
+		ep->rtr_type = MPA_RTR_TYPE_NONE;
+
+	if ((ep->rtr_type & MPA_RTR_TYPE_ZERO_READ) && (ep->cm_info.ord == 0))
+		ep->cm_info.ord = 1;
+
+	ep->mpa_rev = iwarp_info->mpa_rev;
+
+	qed_iwarp_mpa_v2_set_private(p_hwfn, ep, &mpa_data_size);
+
+	ep->cm_info.private_data = ep->ep_buffer_virt->out_pdata;
+	ep->cm_info.private_data_len = iparams->cm_info.private_data_len +
+				       mpa_data_size;
+
+	memcpy((u8 *)ep->ep_buffer_virt->out_pdata + mpa_data_size,
+	       iparams->cm_info.private_data,
+	       iparams->cm_info.private_data_len);
+
+	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
+		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
+
+	ep->mss = iparams->mss - ts_hdr_size;
+	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
+
+	ep->event_cb = iparams->event_cb;
+	ep->cb_context = iparams->cb_context;
+	ep->connect_mode = TCP_CONNECT_ACTIVE;
+
+	oparams->ep_context = ep;
+
+	rc = qed_iwarp_tcp_offload(p_hwfn, ep);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x) rc = %d\n",
+		   iparams->qp->icid, ep->tcp_cid, rc);
+
+	if (rc) {
+		qed_iwarp_destroy_ep(p_hwfn, ep, true);
+		goto err;
+	}
+
+	return rc;
+err:
+	qed_iwarp_cid_cleaned(p_hwfn, cid);
+
+	return rc;
+}
+
 static struct qed_iwarp_ep *qed_iwarp_get_free_ep(struct qed_hwfn *p_hwfn)
 {
 	struct qed_iwarp_ep *ep = NULL;
@@ -1174,12 +1335,12 @@ void qed_iwarp_resc_free(struct qed_hwfn *p_hwfn)
 
 int qed_iwarp_accept(void *rdma_cxt, struct qed_iwarp_accept_in *iparams)
 {
-	struct qed_hwfn *p_hwfn = (struct qed_hwfn *)rdma_cxt;
+	struct qed_hwfn *p_hwfn = rdma_cxt;
 	struct qed_iwarp_ep *ep;
 	u8 mpa_data_size = 0;
 	int rc;
 
-	ep = (struct qed_iwarp_ep *)iparams->ep_context;
+	ep = iparams->ep_context;
 	if (!ep) {
 		DP_ERR(p_hwfn, "Ep Context receive in accept is NULL\n");
 		return -EINVAL;
@@ -1799,13 +1960,19 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
 {
-	/* Done with the SYN packet, post back to ll2 rx */
-	qed_iwarp_ll2_post_rx(p_hwfn, ep->syn,
-			      p_hwfn->p_rdma_info->iwarp.ll2_syn_handle);
-	ep->syn = NULL;
+	u8 ll2_syn_handle = p_hwfn->p_rdma_info->iwarp.ll2_syn_handle;
 
-	/* If connect failed - upper layer doesn't know about it */
-	qed_iwarp_mpa_received(p_hwfn, ep);
+	if (ep->connect_mode == TCP_CONNECT_PASSIVE) {
+		/* Done with the SYN packet, post back to ll2 rx */
+		qed_iwarp_ll2_post_rx(p_hwfn, ep->syn, ll2_syn_handle);
+
+		ep->syn = NULL;
+
+		/* If connect failed - upper layer doesn't know about it */
+		qed_iwarp_mpa_received(p_hwfn, ep);
+	} else {
+		qed_iwarp_mpa_offload(p_hwfn, ep);
+	}
 }
 
 static inline bool
@@ -1842,6 +2009,16 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 			   ep->tcp_cid, fw_return_code);
 		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
 		break;
+		/* Async event for active side only */
+	case IWARP_EVENT_TYPE_ASYNC_ENHANCED_MPA_REPLY_ARRIVED:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_MPA_REPLY_ARRIVED fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_mpa_reply_arrived(p_hwfn, ep);
+		break;
 	case IWARP_EVENT_TYPE_ASYNC_MPA_HANDSHAKE_COMPLETE:
 		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
 			return -EINVAL;
@@ -1918,6 +2095,45 @@ int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle)
 	return 0;
 }
 
+int qed_iwarp_send_rtr(void *rdma_cxt, struct qed_iwarp_send_rtr_in *iparams)
+{
+	struct qed_hwfn *p_hwfn = rdma_cxt;
+	struct qed_sp_init_data init_data;
+	struct qed_spq_entry *p_ent;
+	struct qed_iwarp_ep *ep;
+	struct qed_rdma_qp *qp;
+	int rc;
+
+	ep = iparams->ep_context;
+	if (!ep) {
+		DP_ERR(p_hwfn, "Ep Context receive in send_rtr is NULL\n");
+		return -EINVAL;
+	}
+
+	qp = ep->qp;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "QP(0x%x) EP(0x%x)\n",
+		   qp->icid, ep->tcp_cid);
+
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qp->icid;
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+	init_data.comp_mode = QED_SPQ_MODE_CB;
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 IWARP_RAMROD_CMD_ID_MPA_OFFLOAD_SEND_RTR,
+				 PROTOCOLID_IWARP, &init_data);
+
+	if (rc)
+		return rc;
+
+	rc = qed_spq_post(p_hwfn, p_ent, NULL);
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "rc = 0x%x\n", rc);
+
+	return rc;
+}
+
 void
 qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 		   struct qed_rdma_query_qp_out_params *out_params)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
index bedac98..148ef3c 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.h
@@ -170,6 +170,11 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 			struct qed_rdma_query_qp_out_params *out_params);
 
 int
+qed_iwarp_connect(void *rdma_cxt,
+		  struct qed_iwarp_connect_in *iparams,
+		  struct qed_iwarp_connect_out *oparams);
+
+int
 qed_iwarp_create_listen(void *rdma_cxt,
 			struct qed_iwarp_listen_in *iparams,
 			struct qed_iwarp_listen_out *oparams);
@@ -179,4 +184,6 @@ void qed_iwarp_query_qp(struct qed_rdma_qp *qp,
 int qed_iwarp_reject(void *rdma_cxt, struct qed_iwarp_reject_in *iparams);
 int qed_iwarp_destroy_listen(void *rdma_cxt, void *handle);
 
+int qed_iwarp_send_rtr(void *rdma_cxt, struct qed_iwarp_send_rtr_in *iparams);
+
 #endif
diff --git a/drivers/net/ethernet/qlogic/qed/qed_rdma.c b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
index 29de915..6fb9951 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_rdma.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_rdma.c
@@ -1772,8 +1772,12 @@ static int qed_roce_ll2_set_mac_filter(struct qed_dev *cdev,
 	.ll2_set_fragment_of_tx_packet = &qed_ll2_set_fragment_of_tx_packet,
 	.ll2_set_mac_filter = &qed_roce_ll2_set_mac_filter,
 	.ll2_get_stats = &qed_ll2_get_stats,
+	.iwarp_connect = &qed_iwarp_connect,
 	.iwarp_create_listen = &qed_iwarp_create_listen,
 	.iwarp_destroy_listen = &qed_iwarp_destroy_listen,
+	.iwarp_accept = &qed_iwarp_accept,
+	.iwarp_reject = &qed_iwarp_reject,
+	.iwarp_send_rtr = &qed_iwarp_send_rtr,
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void)
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index c4c241f..e9514a6 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -473,6 +473,8 @@ struct qed_rdma_counters_out_params {
 enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
 	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
+	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
+	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
 };
 
 enum qed_tcp_ip_version {
@@ -503,6 +505,20 @@ struct qed_iwarp_cm_event_params {
 typedef int (*iwarp_event_handler) (void *context,
 				    struct qed_iwarp_cm_event_params *event);
 
+struct qed_iwarp_connect_in {
+	iwarp_event_handler event_cb;
+	void *cb_context;
+	struct qed_rdma_qp *qp;
+	struct qed_iwarp_cm_info cm_info;
+	u16 mss;
+	u8 remote_mac_addr[ETH_ALEN];
+	u8 local_mac_addr[ETH_ALEN];
+};
+
+struct qed_iwarp_connect_out {
+	void *ep_context;
+};
+
 struct qed_iwarp_listen_in {
 	iwarp_event_handler event_cb;
 	void *cb_context;	/* passed to event_cb */
@@ -534,6 +550,10 @@ struct qed_iwarp_reject_in {
 	u16 private_data_len;
 };
 
+struct qed_iwarp_send_rtr_in {
+	void *ep_context;
+};
+
 struct qed_roce_ll2_header {
 	void *vaddr;
 	dma_addr_t baddr;
@@ -640,6 +660,10 @@ struct qed_rdma_ops {
 	int (*ll2_set_mac_filter)(struct qed_dev *cdev,
 				  u8 *old_mac_address, u8 *new_mac_address);
 
+	int (*iwarp_connect)(void *rdma_cxt,
+			     struct qed_iwarp_connect_in *iparams,
+			     struct qed_iwarp_connect_out *oparams);
+
 	int (*iwarp_create_listen)(void *rdma_cxt,
 				   struct qed_iwarp_listen_in *iparams,
 				   struct qed_iwarp_listen_out *oparams);
@@ -652,6 +676,8 @@ struct qed_rdma_ops {
 
 	int (*iwarp_destroy_listen)(void *rdma_cxt, void *handle);
 
+	int (*iwarp_send_rtr)(void *rdma_cxt,
+			      struct qed_iwarp_send_rtr_in *iparams);
 };
 
 const struct qed_rdma_ops *qed_get_rdma_ops(void);
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 09/12] qed: iWARP implement disconnect flows

[Michal Kalderon]

This patch takes care of active/passive disconnect flows.
Disconnect flows can be initiated remotely, in which case a async event
will arrive from peer and indicated to qedr driver. These
are referred to as exceptions. When a QP is destroyed, it needs to check
that it's associated ep has been closed.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 90 ++++++++++++++++++++++++++++-
 include/linux/qed/qed_rdma_if.h             |  2 +
 2 files changed, 91 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a5da9fc..84bcda3 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -496,6 +496,8 @@ static void qed_iwarp_destroy_ep(struct qed_hwfn *p_hwfn,
 
 int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
+	struct qed_iwarp_ep *ep = qp->ep;
+	int wait_count = 0;
 	int rc = 0;
 
 	if (qp->iwarp_state != QED_IWARP_QP_STATE_ERROR) {
@@ -505,6 +507,18 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 			return rc;
 	}
 
+	/* Make sure ep is closed before returning and freeing memory. */
+	if (ep) {
+		while (ep->state != QED_IWARP_EP_CLOSED && wait_count++ < 200)
+			msleep(100);
+
+		if (ep->state != QED_IWARP_EP_CLOSED)
+			DP_NOTICE(p_hwfn, "ep state close timeout state=%x\n",
+				  ep->state);
+
+		qed_iwarp_destroy_ep(p_hwfn, ep, false);
+	}
+
 	rc = qed_iwarp_fw_destroy(p_hwfn, qp);
 
 	if (qp->shared_queue)
@@ -1956,6 +1970,61 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
+void qed_iwarp_qp_in_error(struct qed_hwfn *p_hwfn,
+			   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_ERROR, true);
+
+	params.event = QED_IWARP_EVENT_CLOSE;
+	params.ep_context = ep;
+	params.cm_info = &ep->cm_info;
+	params.status = (fw_return_code == IWARP_QP_IN_ERROR_GOOD_CLOSE) ?
+			 0 : -ECONNRESET;
+
+	ep->state = QED_IWARP_EP_CLOSED;
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_del(&ep->list_entry);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
+void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
+				  struct qed_iwarp_ep *ep, int fw_ret_code)
+{
+	struct qed_iwarp_cm_event_params params;
+	bool event_cb = false;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x) fw_ret_code=%d\n",
+		   ep->cid, fw_ret_code);
+
+	switch (fw_ret_code) {
+	case IWARP_EXCEPTION_DETECTED_LLP_CLOSED:
+		params.status = 0;
+		params.event = QED_IWARP_EVENT_DISCONNECT;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LLP_RESET:
+		params.status = -ECONNRESET;
+		params.event = QED_IWARP_EVENT_DISCONNECT;
+		event_cb = true;
+		break;
+	default:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "Unhandled exception received...fw_ret_code=%d\n",
+			   fw_ret_code);
+		break;
+	}
+
+	if (event_cb) {
+		params.ep_context = ep;
+		params.cm_info = &ep->cm_info;
+		ep->event_cb(ep->cb_context, &params);
+	}
+}
+
 void
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
@@ -2009,8 +2078,27 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 			   ep->tcp_cid, fw_return_code);
 		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
 		break;
-		/* Async event for active side only */
+	case IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_exception_received(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE:
+		/* Async completion for Close Connection ramrod */
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_qp_in_error(p_hwfn, ep, fw_return_code);
+		break;
 	case IWARP_EVENT_TYPE_ASYNC_ENHANCED_MPA_REPLY_ARRIVED:
+		/* Async event for active side only */
 		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
 			return -EINVAL;
 		DP_VERBOSE(p_hwfn,
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index e9514a6..01966c3 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -474,6 +474,8 @@ enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
 	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
 	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
+	QED_IWARP_EVENT_DISCONNECT,
+	QED_IWARP_EVENT_CLOSE,
 	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
 };
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 09/12] qed: iWARP implement disconnect flows

[Michal Kalderon]

This patch takes care of active/passive disconnect flows.
Disconnect flows can be initiated remotely, in which case a async event
will arrive from peer and indicated to qedr driver. These
are referred to as exceptions. When a QP is destroyed, it needs to check
that it's associated ep has been closed.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_iwarp.c | 90 ++++++++++++++++++++++++++++-
 include/linux/qed/qed_rdma_if.h             |  2 +
 2 files changed, 91 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
index a5da9fc..84bcda3 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -496,6 +496,8 @@ static void qed_iwarp_destroy_ep(struct qed_hwfn *p_hwfn,
 
 int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
+	struct qed_iwarp_ep *ep = qp->ep;
+	int wait_count = 0;
 	int rc = 0;
 
 	if (qp->iwarp_state != QED_IWARP_QP_STATE_ERROR) {
@@ -505,6 +507,18 @@ int qed_iwarp_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 			return rc;
 	}
 
+	/* Make sure ep is closed before returning and freeing memory. */
+	if (ep) {
+		while (ep->state != QED_IWARP_EP_CLOSED && wait_count++ < 200)
+			msleep(100);
+
+		if (ep->state != QED_IWARP_EP_CLOSED)
+			DP_NOTICE(p_hwfn, "ep state close timeout state=%x\n",
+				  ep->state);
+
+		qed_iwarp_destroy_ep(p_hwfn, ep, false);
+	}
+
 	rc = qed_iwarp_fw_destroy(p_hwfn, qp);
 
 	if (qp->shared_queue)
@@ -1956,6 +1970,61 @@ int qed_iwarp_stop(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	return qed_iwarp_ll2_stop(p_hwfn, p_ptt);
 }
 
+void qed_iwarp_qp_in_error(struct qed_hwfn *p_hwfn,
+			   struct qed_iwarp_ep *ep, u8 fw_return_code)
+{
+	struct qed_iwarp_cm_event_params params;
+
+	qed_iwarp_modify_qp(p_hwfn, ep->qp, QED_IWARP_QP_STATE_ERROR, true);
+
+	params.event = QED_IWARP_EVENT_CLOSE;
+	params.ep_context = ep;
+	params.cm_info = &ep->cm_info;
+	params.status = (fw_return_code == IWARP_QP_IN_ERROR_GOOD_CLOSE) ?
+			 0 : -ECONNRESET;
+
+	ep->state = QED_IWARP_EP_CLOSED;
+	spin_lock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+	list_del(&ep->list_entry);
+	spin_unlock_bh(&p_hwfn->p_rdma_info->iwarp.iw_lock);
+
+	ep->event_cb(ep->cb_context, &params);
+}
+
+void qed_iwarp_exception_received(struct qed_hwfn *p_hwfn,
+				  struct qed_iwarp_ep *ep, int fw_ret_code)
+{
+	struct qed_iwarp_cm_event_params params;
+	bool event_cb = false;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "EP(0x%x) fw_ret_code=%d\n",
+		   ep->cid, fw_ret_code);
+
+	switch (fw_ret_code) {
+	case IWARP_EXCEPTION_DETECTED_LLP_CLOSED:
+		params.status = 0;
+		params.event = QED_IWARP_EVENT_DISCONNECT;
+		event_cb = true;
+		break;
+	case IWARP_EXCEPTION_DETECTED_LLP_RESET:
+		params.status = -ECONNRESET;
+		params.event = QED_IWARP_EVENT_DISCONNECT;
+		event_cb = true;
+		break;
+	default:
+		DP_VERBOSE(p_hwfn, QED_MSG_RDMA,
+			   "Unhandled exception received...fw_ret_code=%d\n",
+			   fw_ret_code);
+		break;
+	}
+
+	if (event_cb) {
+		params.ep_context = ep;
+		params.cm_info = &ep->cm_info;
+		ep->event_cb(ep->cb_context, &params);
+	}
+}
+
 void
 qed_iwarp_connect_complete(struct qed_hwfn *p_hwfn,
 			   struct qed_iwarp_ep *ep, u8 fw_return_code)
@@ -2009,8 +2078,27 @@ static int qed_iwarp_async_event(struct qed_hwfn *p_hwfn,
 			   ep->tcp_cid, fw_return_code);
 		qed_iwarp_connect_complete(p_hwfn, ep, fw_return_code);
 		break;
-		/* Async event for active side only */
+	case IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED:
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_EXCEPTION_DETECTED fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_exception_received(p_hwfn, ep, fw_return_code);
+		break;
+	case IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE:
+		/* Async completion for Close Connection ramrod */
+		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
+			return -EINVAL;
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_RDMA,
+			   "QP(0x%x) IWARP_EVENT_TYPE_ASYNC_QP_IN_ERROR_STATE fw_ret_code=%d\n",
+			   ep->cid, fw_return_code);
+		qed_iwarp_qp_in_error(p_hwfn, ep, fw_return_code);
+		break;
 	case IWARP_EVENT_TYPE_ASYNC_ENHANCED_MPA_REPLY_ARRIVED:
+		/* Async event for active side only */
 		if (!qed_iwarp_check_ep_ok(p_hwfn, ep))
 			return -EINVAL;
 		DP_VERBOSE(p_hwfn,
diff --git a/include/linux/qed/qed_rdma_if.h b/include/linux/qed/qed_rdma_if.h
index e9514a6..01966c3 100644
--- a/include/linux/qed/qed_rdma_if.h
+++ b/include/linux/qed/qed_rdma_if.h
@@ -474,6 +474,8 @@ enum qed_iwarp_event_type {
 	QED_IWARP_EVENT_MPA_REQUEST,	  /* Passive side request received */
 	QED_IWARP_EVENT_PASSIVE_COMPLETE, /* ack on mpa response */
 	QED_IWARP_EVENT_ACTIVE_COMPLETE,  /* Active side reply received */
+	QED_IWARP_EVENT_DISCONNECT,
+	QED_IWARP_EVENT_CLOSE,
 	QED_IWARP_EVENT_ACTIVE_MPA_REPLY,
 };
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 11/12] qed: Add iWARP protocol support in context allocation

[Michal Kalderon]

When computing how much memory is required for the different hw clients
iWARP protocol should be taken into account

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_cxt.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index 38716f7..af106be 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
@@ -246,14 +246,16 @@ struct qed_cxt_mngr {
 static bool src_proto(enum protocol_type type)
 {
 	return type == PROTOCOLID_ISCSI ||
-	       type == PROTOCOLID_FCOE;
+	       type == PROTOCOLID_FCOE ||
+	       type == PROTOCOLID_IWARP;
 }
 
 static bool tm_cid_proto(enum protocol_type type)
 {
 	return type == PROTOCOLID_ISCSI ||
 	       type == PROTOCOLID_FCOE ||
-	       type == PROTOCOLID_ROCE;
+	       type == PROTOCOLID_ROCE ||
+	       type == PROTOCOLID_IWARP;
 }
 
 static bool tm_tid_proto(enum protocol_type type)
@@ -2068,6 +2070,11 @@ static void qed_rdma_set_pf_params(struct qed_hwfn *p_hwfn,
 	num_srqs = min_t(u32, 32 * 1024, p_params->num_srqs);
 
 	switch (p_hwfn->hw_info.personality) {
+	case QED_PCI_ETH_IWARP:
+		/* Each QP requires one connection */
+		num_cons = min_t(u32, IWARP_MAX_QPS, p_params->num_qps);
+		proto = PROTOCOLID_IWARP;
+		break;
 	case QED_PCI_ETH_ROCE:
 		num_qps = min_t(u32, ROCE_MAX_QPS, p_params->num_qps);
 		num_cons = num_qps * 2;	/* each QP requires two connections */
@@ -2103,6 +2110,8 @@ int qed_cxt_set_pf_params(struct qed_hwfn *p_hwfn, u32 rdma_tasks)
 	qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_CORE, core_cids, 0);
 
 	switch (p_hwfn->hw_info.personality) {
+	case QED_PCI_ETH_RDMA:
+	case QED_PCI_ETH_IWARP:
 	case QED_PCI_ETH_ROCE:
 	{
 			qed_rdma_set_pf_params(p_hwfn,
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 11/12] qed: Add iWARP protocol support in context allocation

[Michal Kalderon]

When computing how much memory is required for the different hw clients
iWARP protocol should be taken into account

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_cxt.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index 38716f7..af106be 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
@@ -246,14 +246,16 @@ struct qed_cxt_mngr {
 static bool src_proto(enum protocol_type type)
 {
 	return type == PROTOCOLID_ISCSI ||
-	       type == PROTOCOLID_FCOE;
+	       type == PROTOCOLID_FCOE ||
+	       type == PROTOCOLID_IWARP;
 }
 
 static bool tm_cid_proto(enum protocol_type type)
 {
 	return type == PROTOCOLID_ISCSI ||
 	       type == PROTOCOLID_FCOE ||
-	       type == PROTOCOLID_ROCE;
+	       type == PROTOCOLID_ROCE ||
+	       type == PROTOCOLID_IWARP;
 }
 
 static bool tm_tid_proto(enum protocol_type type)
@@ -2068,6 +2070,11 @@ static void qed_rdma_set_pf_params(struct qed_hwfn *p_hwfn,
 	num_srqs = min_t(u32, 32 * 1024, p_params->num_srqs);
 
 	switch (p_hwfn->hw_info.personality) {
+	case QED_PCI_ETH_IWARP:
+		/* Each QP requires one connection */
+		num_cons = min_t(u32, IWARP_MAX_QPS, p_params->num_qps);
+		proto = PROTOCOLID_IWARP;
+		break;
 	case QED_PCI_ETH_ROCE:
 		num_qps = min_t(u32, ROCE_MAX_QPS, p_params->num_qps);
 		num_cons = num_qps * 2;	/* each QP requires two connections */
@@ -2103,6 +2110,8 @@ int qed_cxt_set_pf_params(struct qed_hwfn *p_hwfn, u32 rdma_tasks)
 	qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_CORE, core_cids, 0);
 
 	switch (p_hwfn->hw_info.personality) {
+	case QED_PCI_ETH_RDMA:
+	case QED_PCI_ETH_IWARP:
 	case QED_PCI_ETH_ROCE:
 	{
 			qed_rdma_set_pf_params(p_hwfn,
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 12/12] qed: Add iWARP support for physical queue allocation

[Michal Kalderon]

iWARP has different physical queue requirements than RoCE

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_dev.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 4060a6a..6c87bed 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -216,6 +216,10 @@ static u32 qed_get_pq_flags(struct qed_hwfn *p_hwfn)
 	case QED_PCI_ETH_ROCE:
 		flags |= PQ_FLAGS_MCOS | PQ_FLAGS_OFLD | PQ_FLAGS_LLT;
 		break;
+	case QED_PCI_ETH_IWARP:
+		flags |= PQ_FLAGS_MCOS | PQ_FLAGS_ACK | PQ_FLAGS_OOO |
+		    PQ_FLAGS_OFLD;
+		break;
 	default:
 		DP_ERR(p_hwfn,
 		       "unknown personality %d\n", p_hwfn->hw_info.personality);
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH net-next 12/12] qed: Add iWARP support for physical queue allocation

[Michal Kalderon]

iWARP has different physical queue requirements than RoCE

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Yuval Mintz <Yuval.Mintz-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/net/ethernet/qlogic/qed/qed_dev.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 4060a6a..6c87bed 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -216,6 +216,10 @@ static u32 qed_get_pq_flags(struct qed_hwfn *p_hwfn)
 	case QED_PCI_ETH_ROCE:
 		flags |= PQ_FLAGS_MCOS | PQ_FLAGS_OFLD | PQ_FLAGS_LLT;
 		break;
+	case QED_PCI_ETH_IWARP:
+		flags |= PQ_FLAGS_MCOS | PQ_FLAGS_ACK | PQ_FLAGS_OOO |
+		    PQ_FLAGS_OFLD;
+		break;
 	default:
 		DP_ERR(p_hwfn,
 		       "unknown personality %d\n", p_hwfn->hw_info.personality);
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[David Miller]

From: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Date: Sun, 2 Jul 2017 10:29:20 +0300

> This patch series adds iWARP support to our QL4xxxx networking adapters.
> The code changes span across qed and qedr drivers, but this series contains
> changes to qed only. Once the series is accepted, the qedr series will
> be submitted to the rdma tree.
> There is one additional qed patch which enables the iWARP, this patch is
> delayed until the qedr series will be accepted. 
> 
> The patches were previously sent as an RFC, and these are the first 12
> patches in the RFC series:
> https://www.spinics.net/lists/linux-rdma/msg51416.html
> 
> This series was tested and built against net-next.
> 
> MAINTAINERS file is not updated in this PATCH as there is a pending patch
> for qedr driver update https://patchwork.kernel.org/patch/9752761.

Series applied, thanks.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[David Miller]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: Text/Plain; charset=iso-8859-7, Size: 451 bytes --]


You really have to compile test your work and do something with
the warnings:

drivers/net/ethernet/qlogic/qed/qed_iwarp.c:1721:5: warning: ¡ll2_syn_handle¢ may be used uninitialized in this funct

This one is completely legitimate, you can goto "err" and use
the ll2_syn_handle without it being initialized.
N‹§²æìr¸›yúèšØb²X¬¶Ç§vØ^–)Þº{.nÇ+‰·¥Š{±­ÙšŠ{ayº\x1dʇڙë,j\a­¢f£¢·hš‹»öì\x17/oSc¾™Ú³9˜uÀ¦æå‰È&jw¨®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿïêäz¹Þ–Šàþf£¢·hšˆ§~ˆmš

Re: [PATCH net-next 00/12] qed: Add iWARP support for QL4xxxx

[Kalderon, Michal]

From: David Miller <davem@davemloft.net>
Sent: Monday, July 3, 2017 11:59 AM

> You really have to compile test your work and do something with
> the warnings:

> drivers/net/ethernet/qlogic/qed/qed_iwarp.c:1721:5: warning: ‘ll2_syn_handle’ may be used uninitialized in this funct

> This one is completely legitimate, you can goto "err" and use
> the ll2_syn_handle without it being initialized.

Sorry about that, this warning didn't appear locally (gcc 4.8.5).
Fix is on its way (after verifying with newer gcc).