Merge branch 'qed-Add-iWARP-support-for-QL4xxxx'

Michal Kalderon says:

====================
qed: Add iWARP support for QL4xxxx

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.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

18 files changed, 3008 insertions, 100 deletions

diff --git a/drivers/net/ethernet/qlogic/qed/Makefile b/drivers/net/ethernet/qlogic/qed/Makefile
index 67452380b60e..82dd47068e18 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.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 14b08ee9e3ad..91003bc6f00b 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];
@@ -759,7 +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,
@@ -769,6 +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_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index e201214764db..af106be8cc08 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)
@@ -853,7 +855,7 @@ 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 +1035,7 @@ 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 +1835,7 @@ 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);
@@ -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,
@@ -2344,7 +2353,7 @@ qed_cxt_dynamic_ilt_alloc(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 49667ad9042d..6c87bed13bd2 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);
@@ -936,9 +940,16 @@ 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)) {
+			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) {
@@ -2057,7 +2068,7 @@ 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 +2079,7 @@ 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 +2092,12 @@ 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));
@@ -4122,3 +4131,14 @@ 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_hsi.h b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
index 3bf3614b3084..31fb0bffa098 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 000000000000..5cd20da2d4e0
--- /dev/null
+++ b/drivers/net/ethernet/qlogic/qed/qed_iwarp.c
@@ -0,0 +1,2409 @@
+/* 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 <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"
+#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_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)
+
+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) -
+		      QED_IWARP_PREALLOC_CNT;
+
+	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;
+}
+
+/* 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);
+
+	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);
+}
+
+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;
+}
+
+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)
+{
+	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;
+}
+
+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)
+{
+	struct qed_iwarp_ep *ep = qp->ep;
+	int wait_count = 0;
+	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;
+	}
+
+	/* 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)
+		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_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;
+	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,
+				 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));
+