qed*: Utilize Firmware 8.15.3.0

This patch advances the qed* drivers into using the newer firmware -
This solves several firmware bugs, mostly related [but not limited to]
various init/deinit issues in various offloaded protocols.

It also introduces a major 4-Cached SGE change in firmware, which can be
seen in the storage drivers' changes.

In addition, this firmware is required for supporting the new QL41xxx
series of adapters; While this patch doesn't add the actual support,
the firmware contains the necessary initialization & firmware logic to
operate such adapters [actual support would be added later on].

Changes from Previous versions:
-------------------------------
 - V2 - fix kbuild-test robot warnings

Signed-off-by: Tomer Tayar <Tomer.Tayar@cavium.com>
Signed-off-by: Ram Amrani <Ram.Amrani@cavium.com>
Signed-off-by: Manish Rangankar <Manish.Rangankar@cavium.com>
Signed-off-by: Chad Dupuis <Chad.Dupuis@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

16 files changed, 2104 insertions, 1069 deletions

diff --git a/drivers/scsi/qedf/Makefile b/drivers/scsi/qedf/Makefile
index 64e9f507ce32..414f2a772a5f 100644
--- a/drivers/scsi/qedf/Makefile
+++ b/drivers/scsi/qedf/Makefile
@@ -1,5 +1,5 @@
 obj-$(CONFIG_QEDF) := qedf.o
 qedf-y = qedf_dbg.o qedf_main.o qedf_io.o qedf_fip.o \
-	 qedf_attr.o qedf_els.o
+	 qedf_attr.o qedf_els.o drv_scsi_fw_funcs.o drv_fcoe_fw_funcs.o
 
 qedf-$(CONFIG_DEBUG_FS) += qedf_debugfs.o
diff --git a/drivers/scsi/qedf/drv_fcoe_fw_funcs.c b/drivers/scsi/qedf/drv_fcoe_fw_funcs.c
new file mode 100644
index 000000000000..bb812db48da6
--- /dev/null
+++ b/drivers/scsi/qedf/drv_fcoe_fw_funcs.c
@@ -0,0 +1,190 @@
+/* QLogic FCoE Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+#include "drv_fcoe_fw_funcs.h"
+#include "drv_scsi_fw_funcs.h"
+
+#define FCOE_RX_ID ((u32)0x0000FFFF)
+
+static inline void init_common_sqe(struct fcoe_task_params *task_params,
+				   enum fcoe_sqe_request_type request_type)
+{
+	memset(task_params->sqe, 0, sizeof(*(task_params->sqe)));
+	SET_FIELD(task_params->sqe->flags, FCOE_WQE_REQ_TYPE,
+		  request_type);
+	task_params->sqe->task_id = task_params->itid;
+}
+
+int init_initiator_rw_fcoe_task(struct fcoe_task_params *task_params,
+				struct scsi_sgl_task_params *sgl_task_params,
+				struct regpair sense_data_buffer_phys_addr,
+				u32 task_retry_id,
+				u8 fcp_cmd_payload[32])
+{
+	struct fcoe_task_context *ctx = task_params->context;
+	struct ystorm_fcoe_task_st_ctx *y_st_ctx;
+	struct tstorm_fcoe_task_st_ctx *t_st_ctx;
+	struct ustorm_fcoe_task_ag_ctx *u_ag_ctx;
+	struct mstorm_fcoe_task_st_ctx *m_st_ctx;
+	u32 io_size, val;
+	bool slow_sgl;
+
+	memset(ctx, 0, sizeof(*(ctx)));
+	slow_sgl = scsi_is_slow_sgl(sgl_task_params->num_sges,
+				    sgl_task_params->small_mid_sge);
+	io_size = (task_params->task_type == FCOE_TASK_TYPE_WRITE_INITIATOR ?
+		   task_params->tx_io_size : task_params->rx_io_size);
+
+	/* Ystorm ctx */
+	y_st_ctx = &ctx->ystorm_st_context;
+	y_st_ctx->data_2_trns_rem = cpu_to_le32(io_size);
+	y_st_ctx->task_rety_identifier = cpu_to_le32(task_retry_id);
+	y_st_ctx->task_type = task_params->task_type;
+	memcpy(&y_st_ctx->tx_info_union.fcp_cmd_payload,
+	       fcp_cmd_payload, sizeof(struct fcoe_fcp_cmd_payload));
+
+	/* Tstorm ctx */
+	t_st_ctx = &ctx->tstorm_st_context;
+	t_st_ctx->read_only.dev_type = (task_params->is_tape_device == 1 ?
+					FCOE_TASK_DEV_TYPE_TAPE :
+					FCOE_TASK_DEV_TYPE_DISK);
+	t_st_ctx->read_only.cid = cpu_to_le32(task_params->conn_cid);
+	val = cpu_to_le32(task_params->cq_rss_number);
+	t_st_ctx->read_only.glbl_q_num = val;
+	t_st_ctx->read_only.fcp_cmd_trns_size = cpu_to_le32(io_size);
+	t_st_ctx->read_only.task_type = task_params->task_type;
+	SET_FIELD(t_st_ctx->read_write.flags,
+		  FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_EXP_FIRST_FRAME, 1);
+	t_st_ctx->read_write.rx_id = cpu_to_le32(FCOE_RX_ID);
+
+	/* Ustorm ctx */
+	u_ag_ctx = &ctx->ustorm_ag_context;
+	u_ag_ctx->global_cq_num = cpu_to_le32(task_params->cq_rss_number);
+
+	/* Mstorm buffer for sense/rsp data placement */
+	m_st_ctx = &ctx->mstorm_st_context;
+	val = cpu_to_le32(sense_data_buffer_phys_addr.hi);
+	m_st_ctx->rsp_buf_addr.hi = val;
+	val = cpu_to_le32(sense_data_buffer_phys_addr.lo);
+	m_st_ctx->rsp_buf_addr.lo = val;
+
+	if (task_params->task_type == FCOE_TASK_TYPE_WRITE_INITIATOR) {
+		/* Ystorm ctx */
+		y_st_ctx->expect_first_xfer = 1;
+
+		/* Set the amount of super SGEs. Can be up to 4. */
+		SET_FIELD(y_st_ctx->sgl_mode,
+			  YSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE,
+			  (slow_sgl ? SCSI_TX_SLOW_SGL : SCSI_FAST_SGL));
+		init_scsi_sgl_context(&y_st_ctx->sgl_params,
+				      &y_st_ctx->data_desc,
+				      sgl_task_params);
+
+		/* Mstorm ctx */
+		SET_FIELD(m_st_ctx->flags,
+			  MSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE,
+			  (slow_sgl ? SCSI_TX_SLOW_SGL : SCSI_FAST_SGL));
+	} else {
+		/* Tstorm ctx */
+		SET_FIELD(t_st_ctx->read_write.flags,
+			  FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_RX_SGL_MODE,
+			  (slow_sgl ? SCSI_TX_SLOW_SGL : SCSI_FAST_SGL));
+
+		/* Mstorm ctx */
+		m_st_ctx->data_2_trns_rem = cpu_to_le32(io_size);
+		init_scsi_sgl_context(&m_st_ctx->sgl_params,
+				      &m_st_ctx->data_desc,
+				      sgl_task_params);
+	}
+
+	init_common_sqe(task_params, SEND_FCOE_CMD);
+	return 0;
+}
+
+int init_initiator_midpath_unsolicited_fcoe_task(
+	struct fcoe_task_params *task_params,
+	struct fcoe_tx_mid_path_params *mid_path_fc_header,
+	struct scsi_sgl_task_params *tx_sgl_task_params,
+	struct scsi_sgl_task_params *rx_sgl_task_params,
+	u8 fw_to_place_fc_header)
+{
+	struct fcoe_task_context *ctx = task_params->context;
+	struct ystorm_fcoe_task_st_ctx *y_st_ctx;
+	struct tstorm_fcoe_task_st_ctx *t_st_ctx;
+	struct ustorm_fcoe_task_ag_ctx *u_ag_ctx;
+	struct mstorm_fcoe_task_st_ctx *m_st_ctx;
+	u32 val;
+
+	memset(ctx, 0, sizeof(*(ctx)));
+
+	/* Init Ystorm */
+	y_st_ctx = &ctx->ystorm_st_context;
+	init_scsi_sgl_context(&y_st_ctx->sgl_params,
+			      &y_st_ctx->data_desc,
+			      tx_sgl_task_params);
+	SET_FIELD(y_st_ctx->sgl_mode,
+		  YSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE, SCSI_FAST_SGL);
+	y_st_ctx->data_2_trns_rem = cpu_to_le32(task_params->tx_io_size);
+	y_st_ctx->task_type = task_params->task_type;
+	memcpy(&y_st_ctx->tx_info_union.tx_params.mid_path,
+	       mid_path_fc_header, sizeof(struct fcoe_tx_mid_path_params));
+
+	/* Init Mstorm */
+	m_st_ctx = &ctx->mstorm_st_context;
+	init_scsi_sgl_context(&m_st_ctx->sgl_params,
+			      &m_st_ctx->data_desc,
+			      rx_sgl_task_params);
+	SET_FIELD(m_st_ctx->flags,
+		  MSTORM_FCOE_TASK_ST_CTX_MP_INCLUDE_FC_HEADER,
+		  fw_to_place_fc_header);
+	m_st_ctx->data_2_trns_rem = cpu_to_le32(task_params->rx_io_size);
+
+	/* Init Tstorm */
+	t_st_ctx = &ctx->tstorm_st_context;
+	t_st_ctx->read_only.cid = cpu_to_le32(task_params->conn_cid);
+	val = cpu_to_le32(task_params->cq_rss_number);
+	t_st_ctx->read_only.glbl_q_num = val;
+	t_st_ctx->read_only.task_type = task_params->task_type;
+	SET_FIELD(t_st_ctx->read_write.flags,
+		  FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_EXP_FIRST_FRAME, 1);
+	t_st_ctx->read_write.rx_id = cpu_to_le32(FCOE_RX_ID);
+
+	/* Init Ustorm */
+	u_ag_ctx = &ctx->ustorm_ag_context;
+	u_ag_ctx->global_cq_num = cpu_to_le32(task_params->cq_rss_number);
+
+	/* Init SQE */
+	init_common_sqe(task_params, SEND_FCOE_MIDPATH);
+	task_params->sqe->additional_info_union.burst_length =
+				    tx_sgl_task_params->total_buffer_size;
+	SET_FIELD(task_params->sqe->flags,
+		  FCOE_WQE_NUM_SGES, tx_sgl_task_params->num_sges);
+	SET_FIELD(task_params->sqe->flags, FCOE_WQE_SGL_MODE,
+		  SCSI_FAST_SGL);
+
+	return 0;
+}
+
+int init_initiator_abort_fcoe_task(struct fcoe_task_params *task_params)
+{
+	init_common_sqe(task_params, SEND_FCOE_ABTS_REQUEST);
+	return 0;
+}
+
+int init_initiator_cleanup_fcoe_task(struct fcoe_task_params *task_params)
+{
+	init_common_sqe(task_params, FCOE_EXCHANGE_CLEANUP);
+	return 0;
+}
+
+int init_initiator_sequence_recovery_fcoe_task(
+	struct fcoe_task_params *task_params, u32 off)
+{
+	init_common_sqe(task_params, FCOE_SEQUENCE_RECOVERY);
+	task_params->sqe->additional_info_union.seq_rec_updated_offset = off;
+	return 0;
+}
diff --git a/drivers/scsi/qedf/drv_fcoe_fw_funcs.h b/drivers/scsi/qedf/drv_fcoe_fw_funcs.h
new file mode 100644
index 000000000000..617529b058f4
--- /dev/null
+++ b/drivers/scsi/qedf/drv_fcoe_fw_funcs.h
@@ -0,0 +1,93 @@
+/* QLogic FCoE Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+#ifndef _FCOE_FW_FUNCS_H
+#define _FCOE_FW_FUNCS_H
+#include "drv_scsi_fw_funcs.h"
+#include "qedf_hsi.h"
+#include <linux/qed/qed_if.h>
+
+struct fcoe_task_params {
+	/* Output parameter [set/filled by the HSI function] */
+	struct fcoe_task_context *context;
+
+	/* Output parameter [set/filled by the HSI function] */
+	struct fcoe_wqe *sqe;
+	enum fcoe_task_type task_type;
+	u32 tx_io_size; /* in bytes */
+	u32 rx_io_size; /* in bytes */
+	u32 conn_cid;
+	u16 itid;
+	u8 cq_rss_number;
+
+	 /* Whether it's Tape device or not (0=Disk, 1=Tape) */
+	u8 is_tape_device;
+};
+
+/**
+ * @brief init_initiator_rw_fcoe_task - Initializes FCoE task context for
+ * read/write task types and init fcoe_sqe
+ *
+ * @param task_params - Pointer to task parameters struct
+ * @param sgl_task_params - Pointer to SGL task params
+ * @param sense_data_buffer_phys_addr - Pointer to sense data buffer
+ * @param task_retry_id - retry identification - Used only for Tape device
+ * @param fcp_cmnd_payload - FCP CMD Payload
+ */
+int init_initiator_rw_fcoe_task(struct fcoe_task_params *task_params,
+	struct scsi_sgl_task_params *sgl_task_params,
+	struct regpair sense_data_buffer_phys_addr,
+	u32 task_retry_id,
+	u8 fcp_cmd_payload[32]);
+
+/**
+ * @brief init_initiator_midpath_fcoe_task - Initializes FCoE task context for
+ * midpath/unsolicited task types and init fcoe_sqe
+ *
+ * @param task_params - Pointer to task parameters struct
+ * @param mid_path_fc_header - FC header
+ * @param tx_sgl_task_params - Pointer to Tx SGL task params
+ * @param rx_sgl_task_params - Pointer to Rx SGL task params
+ * @param fw_to_place_fc_header	- Indication if the FW will place the FC header
+ * in addition to the data arrives.
+ */
+int init_initiator_midpath_unsolicited_fcoe_task(
+	struct fcoe_task_params *task_params,
+	struct fcoe_tx_mid_path_params *mid_path_fc_header,
+	struct scsi_sgl_task_params *tx_sgl_task_params,
+	struct scsi_sgl_task_params *rx_sgl_task_params,
+	u8 fw_to_place_fc_header);
+
+/**
+ * @brief init_initiator_abort_fcoe_task - Initializes FCoE task context for
+ * abort task types and init fcoe_sqe
+ *
+ * @param task_params - Pointer to task parameters struct
+ */
+int init_initiator_abort_fcoe_task(struct fcoe_task_params *task_params);
+
+/**
+ * @brief init_initiator_cleanup_fcoe_task - Initializes FCoE task context for
+ * cleanup task types and init fcoe_sqe
+ *
+ *
+ * @param task_params - Pointer to task parameters struct
+ */
+int init_initiator_cleanup_fcoe_task(struct fcoe_task_params *task_params);
+
+/**
+ * @brief init_initiator_cleanup_fcoe_task - Initializes FCoE task context for
+ * sequence recovery task types and init fcoe_sqe
+ *
+ *
+ * @param task_params - Pointer to task parameters struct
+ * @param desired_offset - The desired offest the task will be re-sent from
+ */
+int init_initiator_sequence_recovery_fcoe_task(
+	struct fcoe_task_params *task_params,
+	u32 desired_offset);
+#endif
diff --git a/drivers/scsi/qedf/drv_scsi_fw_funcs.c b/drivers/scsi/qedf/drv_scsi_fw_funcs.c
new file mode 100644
index 000000000000..11e0cc082ec0
--- /dev/null
+++ b/drivers/scsi/qedf/drv_scsi_fw_funcs.c
@@ -0,0 +1,44 @@
+/* QLogic FCoE Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+#include "drv_scsi_fw_funcs.h"
+
+#define SCSI_NUM_SGES_IN_CACHE 0x4
+
+bool scsi_is_slow_sgl(u16 num_sges, bool small_mid_sge)
+{
+	return (num_sges > SCSI_NUM_SGES_SLOW_SGL_THR && small_mid_sge);
+}
+
+void init_scsi_sgl_context(struct scsi_sgl_params *ctx_sgl_params,
+			   struct scsi_cached_sges *ctx_data_desc,
+			   struct scsi_sgl_task_params *sgl_task_params)
+{
+	/* no need to check for sgl_task_params->sgl validity */
+	u8 num_sges_to_init = sgl_task_params->num_sges >
+			      SCSI_NUM_SGES_IN_CACHE ? SCSI_NUM_SGES_IN_CACHE :
+			      sgl_task_params->num_sges;
+	u8 sge_index;
+	u32 val;
+
+	val = cpu_to_le32(sgl_task_params->sgl_phys_addr.lo);
+	ctx_sgl_params->sgl_addr.lo = val;
+	val = cpu_to_le32(sgl_task_params->sgl_phys_addr.hi);
+	ctx_sgl_params->sgl_addr.hi = val;
+	val = cpu_to_le32(sgl_task_params->total_buffer_size);
+	ctx_sgl_params->sgl_total_length = val;
+	ctx_sgl_params->sgl_num_sges = cpu_to_le16(sgl_task_params->num_sges);
+
+	for (sge_index = 0; sge_index < num_sges_to_init; sge_index++) {
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_addr.lo);
+		ctx_data_desc->sge[sge_index].sge_addr.lo = val;
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_addr.hi);
+		ctx_data_desc->sge[sge_index].sge_addr.hi = val;
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_len);
+		ctx_data_desc->sge[sge_index].sge_len = val;
+	}
+}
diff --git a/drivers/scsi/qedf/drv_scsi_fw_funcs.h b/drivers/scsi/qedf/drv_scsi_fw_funcs.h
new file mode 100644
index 000000000000..9cb45410bc45
--- /dev/null
+++ b/drivers/scsi/qedf/drv_scsi_fw_funcs.h
@@ -0,0 +1,85 @@
+/* QLogic FCoE Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+#ifndef _SCSI_FW_FUNCS_H
+#define _SCSI_FW_FUNCS_H
+#include <linux/qed/common_hsi.h>
+#include <linux/qed/storage_common.h>
+#include <linux/qed/fcoe_common.h>
+
+struct scsi_sgl_task_params {
+	struct scsi_sge *sgl;
+	struct regpair sgl_phys_addr;
+	u32 total_buffer_size;
+	u16 num_sges;
+
+	 /* true if SGL contains a small (< 4KB) SGE in middle(not 1st or last)
+	  * -> relevant for tx only
+	  */
+	bool small_mid_sge;
+};
+
+struct scsi_dif_task_params {
+	u32 initial_ref_tag;
+	bool initial_ref_tag_is_valid;
+	u16 application_tag;
+	u16 application_tag_mask;
+	u16 dif_block_size_log;
+	bool dif_on_network;
+	bool dif_on_host;
+	u8 host_guard_type;
+	u8 protection_type;
+	u8 ref_tag_mask;
+	bool crc_seed;
+
+	 /* Enable Connection error upon DIF error (segments with DIF errors are
+	  * dropped)
+	  */
+	bool tx_dif_conn_err_en;
+	bool ignore_app_tag;
+	bool keep_ref_tag_const;
+	bool validate_guard;
+	bool validate_app_tag;
+	bool validate_ref_tag;
+	bool forward_guard;
+	bool forward_app_tag;
+	bool forward_ref_tag;
+	bool forward_app_tag_with_mask;
+	bool forward_ref_tag_with_mask;
+};
+
+struct scsi_initiator_cmd_params {
+	 /* for cdb_size > default CDB size (extended CDB > 16 bytes) ->
+	  * pointer to the CDB buffer SGE
+	  */
+	struct scsi_sge extended_cdb_sge;
+
+	/* Physical address of sense data buffer for sense data - 256B buffer */
+	struct regpair sense_data_buffer_phys_addr;
+};
+
+/**
+ * @brief scsi_is_slow_sgl - checks for slow SGL
+ *
+ * @param num_sges - number of sges in SGL
+ * @param small_mid_sge - True is the SGL contains an SGE which is smaller than
+ * 4KB and its not the 1st or last SGE in the SGL
+ */
+bool scsi_is_slow_sgl(u16 num_sges, bool small_mid_sge);
+
+/**
+ * @brief init_scsi_sgl_context - initializes SGL task context
+ *
+ * @param sgl_params - SGL context parameters to initialize (output parameter)
+ * @param data_desc - context struct containing SGEs array to set (output
+ * parameter)
+ * @param sgl_task_params - SGL parameters (input)
+ */
+void init_scsi_sgl_context(struct scsi_sgl_params *sgl_params,
+	struct scsi_cached_sges *ctx_data_desc,
+	struct scsi_sgl_task_params *sgl_task_params);
+#endif
diff --git a/drivers/scsi/qedf/qedf.h b/drivers/scsi/qedf/qedf.h
index 96346a1b1515..40aeb6bb96a2 100644
--- a/drivers/scsi/qedf/qedf.h
+++ b/drivers/scsi/qedf/qedf.h
@@ -26,6 +26,7 @@
 #include <linux/qed/qed_ll2_if.h>
 #include "qedf_version.h"
 #include "qedf_dbg.h"
+#include "drv_fcoe_fw_funcs.h"
 
 /* Helpers to extract upper and lower 32-bits of pointer */
 #define U64_HI(val) ((u32)(((u64)(val)) >> 32))
@@ -59,19 +60,17 @@
 #define UPSTREAM_KEEP		1
 
 struct qedf_mp_req {
-	uint8_t tm_flags;
-
 	uint32_t req_len;
 	void *req_buf;
 	dma_addr_t req_buf_dma;
-	struct fcoe_sge *mp_req_bd;
+	struct scsi_sge *mp_req_bd;
 	dma_addr_t mp_req_bd_dma;
 	struct fc_frame_header req_fc_hdr;
 
 	uint32_t resp_len;
 	void *resp_buf;
 	dma_addr_t resp_buf_dma;
-	struct fcoe_sge *mp_resp_bd;
+	struct scsi_sge *mp_resp_bd;
 	dma_addr_t mp_resp_bd_dma;
 	struct fc_frame_header resp_fc_hdr;
 };
@@ -119,6 +118,7 @@ struct qedf_ioreq {
 #define QEDF_CMD_IN_CLEANUP		0x2
 #define QEDF_CMD_SRR_SENT		0x3
 	u8 io_req_flags;
+	uint8_t tm_flags;
 	struct qedf_rport *fcport;
 	unsigned long flags;
 	enum qedf_ioreq_event event;
@@ -130,6 +130,8 @@ struct qedf_ioreq {
 	struct completion tm_done;
 	struct completion abts_done;
 	struct fcoe_task_context *task;
+	struct fcoe_task_params *task_params;
+	struct scsi_sgl_task_params *sgl_task_params;
 	int idx;
 /*
  * Need to allocate enough room for both sense data and FCP response data
@@ -199,8 +201,8 @@ struct qedf_rport {
 	dma_addr_t sq_pbl_dma;
 	u32 sq_pbl_size;
 	u32 sid;
-#define	QEDF_RPORT_TYPE_DISK		1
-#define	QEDF_RPORT_TYPE_TAPE		2
+#define	QEDF_RPORT_TYPE_DISK		0
+#define	QEDF_RPORT_TYPE_TAPE		1
 	uint dev_type; /* Disk or tape */
 	struct list_head peers;
 };
@@ -391,7 +393,7 @@ struct qedf_ctx {
 
 struct io_bdt {
 	struct qedf_ioreq *io_req;
-	struct fcoe_sge *bd_tbl;
+	struct scsi_sge *bd_tbl;
 	dma_addr_t bd_tbl_dma;
 	u16 bd_valid;
 };
@@ -400,7 +402,7 @@ struct qedf_cmd_mgr {
 	struct qedf_ctx *qedf;
 	u16 idx;
 	struct io_bdt **io_bdt_pool;
-#define FCOE_PARAMS_NUM_TASKS		4096
+#define FCOE_PARAMS_NUM_TASKS		2048
 	struct qedf_ioreq cmds[FCOE_PARAMS_NUM_TASKS];
 	spinlock_t lock;
 	atomic_t free_list_cnt;
@@ -465,9 +467,8 @@ extern void qedf_cmd_timer_set(struct qedf_ctx *qedf, struct qedf_ioreq *io_req,
 	unsigned int timer_msec);
 extern int qedf_init_mp_req(struct qedf_ioreq *io_req);
 extern void qedf_init_mp_task(struct qedf_ioreq *io_req,
-	struct fcoe_task_context *task_ctx);
-extern void qedf_add_to_sq(struct qedf_rport *fcport, u16 xid,
-	u32 ptu_invalidate, enum fcoe_task_type req_type, u32 offset);
+	struct fcoe_task_context *task_ctx, struct fcoe_wqe *wqe);
+extern u16 qedf_get_sqe_idx(struct qedf_rport *fcport);
 extern void qedf_ring_doorbell(struct qedf_rport *fcport);
 extern void qedf_process_els_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
 	struct qedf_ioreq *els_req);
diff --git a/drivers/scsi/qedf/qedf_els.c b/drivers/scsi/qedf/qedf_els.c
index 59f3e5c73a13..c505d41f6dc8 100644
--- a/drivers/scsi/qedf/qedf_els.c
+++ b/drivers/scsi/qedf/qedf_els.c
@@ -25,6 +25,9 @@ static int qedf_initiate_els(struct qedf_rport *fcport, unsigned int op,
 	uint16_t xid;
 	uint32_t start_time = jiffies / HZ;
 	uint32_t current_time;
+	struct fcoe_wqe *sqe;
+	unsigned long flags;
+	u16 sqe_idx;
 
 	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_ELS, "Sending ELS\n");
 
@@ -113,20 +116,25 @@ retry_els:
 	/* Obtain exchange id */
 	xid = els_req->xid;
 
+	spin_lock_irqsave(&fcport->rport_lock, flags);
+
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+
 	/* Initialize task context for this IO request */
 	task = qedf_get_task_mem(&qedf->tasks, xid);
-	qedf_init_mp_task(els_req, task);
+	qedf_init_mp_task(els_req, task, sqe);
 
 	/* Put timer on original I/O request */
 	if (timer_msec)
 		qedf_cmd_timer_set(qedf, els_req, timer_msec);
 
-	qedf_add_to_sq(fcport, xid, 0, FCOE_TASK_TYPE_MIDPATH, 0);
-
 	/* Ring doorbell */
 	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_ELS, "Ringing doorbell for ELS "
 		   "req\n");
 	qedf_ring_doorbell(fcport);
+	spin_unlock_irqrestore(&fcport->rport_lock, flags);
 els_err:
 	return rc;
 }
@@ -604,6 +612,8 @@ static void qedf_initiate_seq_cleanup(struct qedf_ioreq *orig_io_req,
 	struct qedf_rport *fcport;
 	unsigned long flags;
 	struct qedf_els_cb_arg *cb_arg;
+	struct fcoe_wqe *sqe;
+	u16 sqe_idx;
 
 	fcport = orig_io_req->fcport;
 
@@ -631,8 +641,13 @@ static void qedf_initiate_seq_cleanup(struct qedf_ioreq *orig_io_req,
 
 	spin_lock_irqsave(&fcport->rport_lock, flags);
 
-	qedf_add_to_sq(fcport, orig_io_req->xid, 0,
-	    FCOE_TASK_TYPE_SEQUENCE_CLEANUP, offset);
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+	orig_io_req->task_params->sqe = sqe;
+
+	init_initiator_sequence_recovery_fcoe_task(orig_io_req->task_params,
+						   offset);
 	qedf_ring_doorbell(fcport);
 
 	spin_unlock_irqrestore(&fcport->rport_lock, flags);
diff --git a/drivers/scsi/qedf/qedf_io.c b/drivers/scsi/qedf/qedf_io.c
index ee0dcf9d3aba..af2294635ab2 100644
--- a/drivers/scsi/qedf/qedf_io.c
+++ b/drivers/scsi/qedf/qedf_io.c
@@ -96,7 +96,7 @@ void qedf_cmd_mgr_free(struct qedf_cmd_mgr *cmgr)
 	if (!cmgr->io_bdt_pool)
 		goto free_cmd_pool;
 
-	bd_tbl_sz = QEDF_MAX_BDS_PER_CMD * sizeof(struct fcoe_sge);
+	bd_tbl_sz = QEDF_MAX_BDS_PER_CMD * sizeof(struct scsi_sge);
 	for (i = 0; i < num_ios; i++) {
 		bdt_info = cmgr->io_bdt_pool[i];
 		if (bdt_info->bd_tbl) {
@@ -119,6 +119,8 @@ free_cmd_pool:
 
 	for (i = 0; i < num_ios; i++) {
 		io_req = &cmgr->cmds[i];
+		kfree(io_req->sgl_task_params);
+		kfree(io_req->task_params);
 		/* Make sure we free per command sense buffer */
 		if (io_req->sense_buffer)
 			dma_free_coherent(&qedf->pdev->dev,
@@ -178,7 +180,7 @@ struct qedf_cmd_mgr *qedf_cmd_mgr_alloc(struct qedf_ctx *qedf)
 	spin_lock_init(&cmgr->lock);
 
 	/*
-	 * Initialize list of qedf_ioreq.
+	 * Initialize I/O request fields.
 	 */
 	xid = QEDF_MIN_XID;
 
@@ -196,6 +198,29 @@ struct qedf_cmd_mgr *qedf_cmd_mgr_alloc(struct qedf_ctx *qedf)
 		    GFP_KERNEL);
 		if (!io_req->sense_buffer)
 			goto mem_err;
+
+		/* Allocate task parameters to pass to f/w init funcions */
+		io_req->task_params = kzalloc(sizeof(*io_req->task_params),
+					      GFP_KERNEL);
+		if (!io_req->task_params) {
+			QEDF_ERR(&(qedf->dbg_ctx),
+				 "Failed to allocate task_params for xid=0x%x\n",
+				 i);
+			goto mem_err;
+		}
+
+		/*
+		 * Allocate scatter/gather list info to pass to f/w init
+		 * functions.
+		 */
+		io_req->sgl_task_params = kzalloc(
+		    sizeof(struct scsi_sgl_task_params), GFP_KERNEL);
+		if (!io_req->sgl_task_params) {
+			QEDF_ERR(&(qedf->dbg_ctx),
+				 "Failed to allocate sgl_task_params for xid=0x%x\n",
+				 i);
+			goto mem_err;
+		}
 	}
 
 	/* Allocate pool of io_bdts - one for each qedf_ioreq */
@@ -211,8 +236,8 @@ struct qedf_cmd_mgr *qedf_cmd_mgr_alloc(struct qedf_ctx *qedf)
 		cmgr->io_bdt_pool[i] = kmalloc(sizeof(struct io_bdt),
 		    GFP_KERNEL);
 		if (!cmgr->io_bdt_pool[i]) {
-			QEDF_WARN(&(qedf->dbg_ctx), "Failed to alloc "
-				   "io_bdt_pool[%d].\n", i);
+			QEDF_WARN(&(qedf->dbg_ctx),
+				  "Failed to alloc io_bdt_pool[%d].\n", i);
 			goto mem_err;
 		}
 	}
@@ -220,11 +245,11 @@ struct qedf_cmd_mgr *qedf_cmd_mgr_alloc(struct qedf_ctx *qedf)
 	for (i = 0; i < num_ios; i++) {
 		bdt_info = cmgr->io_bdt_pool[i];
 		bdt_info->bd_tbl = dma_alloc_coherent(&qedf->pdev->dev,
-		    QEDF_MAX_BDS_PER_CMD * sizeof(struct fcoe_sge),
+		    QEDF_MAX_BDS_PER_CMD * sizeof(struct scsi_sge),
 		    &bdt_info->bd_tbl_dma, GFP_KERNEL);
 		if (!bdt_info->bd_tbl) {
-			QEDF_WARN(&(qedf->dbg_ctx), "Failed to alloc "
-				   "bdt_tbl[%d].\n", i);
+			QEDF_WARN(&(qedf->dbg_ctx),
+				  "Failed to alloc bdt_tbl[%d].\n", i);
 			goto mem_err;
 		}
 	}
@@ -318,6 +343,7 @@ struct qedf_ioreq *qedf_alloc_cmd(struct qedf_rport *fcport, u8 cmd_type)
 	}
 	bd_tbl->io_req = io_req;
 	io_req->cmd_type = cmd_type;
+	io_req->tm_flags = 0;
 
 	/* Reset sequence offset data */
 	io_req->rx_buf_off = 0;
@@ -336,10 +362,9 @@ static void qedf_free_mp_resc(struct qedf_ioreq *io_req)
 {
 	struct qedf_mp_req *mp_req = &(io_req->mp_req);
 	struct qedf_ctx *qedf = io_req->fcport->qedf;
-	uint64_t sz = sizeof(struct fcoe_sge);
+	uint64_t sz = sizeof(struct scsi_sge);
 
 	/* clear tm flags */
-	mp_req->tm_flags = 0;
 	if (mp_req->mp_req_bd) {
 		dma_free_coherent(&qedf->pdev->dev, sz,
 		    mp_req->mp_req_bd, mp_req->mp_req_bd_dma);
@@ -387,7 +412,7 @@ void qedf_release_cmd(struct kref *ref)
 static int qedf_split_bd(struct qedf_ioreq *io_req, u64 addr, int sg_len,
 	int bd_index)
 {
-	struct fcoe_sge *bd = io_req->bd_tbl->bd_tbl;
+	struct scsi_sge *bd = io_req->bd_tbl->bd_tbl;
 	int frag_size, sg_frags;
 
 	sg_frags = 0;
@@ -398,7 +423,7 @@ static int qedf_split_bd(struct qedf_ioreq *io_req, u64 addr, int sg_len,
 			frag_size = sg_len;
 		bd[bd_index + sg_frags].sge_addr.lo = U64_LO(addr);
 		bd[bd_index + sg_frags].sge_addr.hi = U64_HI(addr);
-		bd[bd_index + sg_frags].size = (uint16_t)frag_size;
+		bd[bd_index + sg_frags].sge_len = (uint16_t)frag_size;
 
 		addr += (u64)frag_size;
 		sg_frags++;
@@ -413,7 +438,7 @@ static int qedf_map_sg(struct qedf_ioreq *io_req)
 	struct Scsi_Host *host = sc->device->host;
 	struct fc_lport *lport = shost_priv(host);
 	struct qedf_ctx *qedf = lport_priv(lport);
-	struct fcoe_sge *bd = io_req->bd_tbl->bd_tbl;
+	struct scsi_sge *bd = io_req->bd_tbl->bd_tbl;
 	struct scatterlist *sg;
 	int byte_count = 0;
 	int sg_count = 0;
@@ -439,7 +464,7 @@ static int qedf_map_sg(struct qedf_ioreq *io_req)
 
 		bd[bd_count].sge_addr.lo = (addr & 0xffffffff);
 		bd[bd_count].sge_addr.hi = (addr >> 32);
-		bd[bd_count].size = (u16)sg_len;
+		bd[bd_count].sge_len = (u16)sg_len;
 
 		return ++bd_count;
 	}
@@ -480,7 +505,7 @@ static int qedf_map_sg(struct qedf_ioreq *io_req)
 			sg_frags = 1;
 			bd[bd_count].sge_addr.lo = U64_LO(addr);
 			bd[bd_count].sge_addr.hi  = U64_HI(addr);
-			bd[bd_count].size = (uint16_t)sg_len;
+			bd[bd_count].sge_len = (uint16_t)sg_len;
 		}
 
 		bd_count += sg_frags;
@@ -498,7 +523,7 @@ static int qedf_map_sg(struct qedf_ioreq *io_req)
 static int qedf_build_bd_list_from_sg(struct qedf_ioreq *io_req)
 {
 	struct scsi_cmnd *sc = io_req->sc_cmd;
-	struct fcoe_sge *bd = io_req->bd_tbl->bd_tbl;
+	struct scsi_sge *bd = io_req->bd_tbl->bd_tbl;
 	int bd_count;
 
 	if (scsi_sg_count(sc)) {
@@ -508,7 +533,7 @@ static int qedf_build_bd_list_from_sg(struct qedf_ioreq *io_req)
 	} else {
 		bd_count = 0;
 		bd[0].sge_addr.lo = bd[0].sge_addr.hi = 0;
-		bd[0].size = 0;
+		bd[0].sge_len = 0;
 	}
 	io_req->bd_tbl->bd_valid = bd_count;
 
@@ -529,430 +554,223 @@ static void qedf_build_fcp_cmnd(struct qedf_ioreq *io_req,
 
 	/* 4 bytes: flag info */
 	fcp_cmnd->fc_pri_ta = 0;
-	fcp_cmnd->fc_tm_flags = io_req->mp_req.tm_flags;
+	fcp_cmnd->fc_tm_flags = io_req->tm_flags;
 	fcp_cmnd->fc_flags = io_req->io_req_flags;
 	fcp_cmnd->fc_cmdref = 0;
 
 	/* Populate data direction */
-	if (sc_cmd->sc_data_direction == DMA_TO_DEVICE)
-		fcp_cmnd->fc_flags |= FCP_CFL_WRDATA;
-	else if (sc_cmd->sc_data_direction == DMA_FROM_DEVICE)
+	if (io_req->cmd_type == QEDF_TASK_MGMT_CMD) {
 		fcp_cmnd->fc_flags |= FCP_CFL_RDDATA;
+	} else {
+		if (sc_cmd->sc_data_direction == DMA_TO_DEVICE)
+			fcp_cmnd->fc_flags |= FCP_CFL_WRDATA;
+		else if (sc_cmd->sc_data_direction == DMA_FROM_DEVICE)
+			fcp_cmnd->fc_flags |= FCP_CFL_RDDATA;
+	}
 
 	fcp_cmnd->fc_pri_ta = FCP_PTA_SIMPLE;
 
 	/* 16 bytes: CDB information */
-	memcpy(fcp_cmnd->fc_cdb, sc_cmd->cmnd, sc_cmd->cmd_len);
+	if (io_req->cmd_type != QEDF_TASK_MGMT_CMD)
+		memcpy(fcp_cmnd->fc_cdb, sc_cmd->cmnd, sc_cmd->cmd_len);
 
 	/* 4 bytes: FCP data length */
 	fcp_cmnd->fc_dl = htonl(io_req->data_xfer_len);
-
 }
 
 static void  qedf_init_task(struct qedf_rport *fcport, struct fc_lport *lport,
-	struct qedf_ioreq *io_req, u32 *ptu_invalidate,
-	struct fcoe_task_context *task_ctx)
+	struct qedf_ioreq *io_req, struct fcoe_task_context *task_ctx,
+	struct fcoe_wqe *sqe)
 {
 	enum fcoe_task_type task_type;
 	struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
 	struct io_bdt *bd_tbl = io_req->bd_tbl;
-	union fcoe_data_desc_ctx *data_desc;
-	u32 *fcp_cmnd;
+	u8 fcp_cmnd[32];
 	u32 tmp_fcp_cmnd[8];
-	int cnt, i;
-	int bd_count;
+	int bd_count = 0;
 	struct qedf_ctx *qedf = fcport->qedf;
 	uint16_t cq_idx = smp_processor_id() % qedf->num_queues;
-	u8 tmp_sgl_mode = 0;
-	u8 mst_sgl_mode = 0;
+	struct regpair sense_data_buffer_phys_addr;
+	u32 tx_io_size = 0;
+	u32 rx_io_size = 0;
+	int i, cnt;
 
-	memset(task_ctx, 0, sizeof(struct fcoe_task_context));
+	/* Note init_initiator_rw_fcoe_task memsets the task context */
 	io_req->task = task_ctx;
+	memset(task_ctx, 0, sizeof(struct fcoe_task_context));
+	memset(io_req->task_params, 0, sizeof(struct fcoe_task_params));
+	memset(io_req->sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
 
-	if (sc_cmd->sc_data_direction == DMA_TO_DEVICE)
-		task_type = FCOE_TASK_TYPE_WRITE_INITIATOR;
-	else
+	/* Set task type bassed on DMA directio of command */
+	if (io_req->cmd_type == QEDF_TASK_MGMT_CMD) {
 		task_type = FCOE_TASK_TYPE_READ_INITIATOR;
-
-	/* Y Storm context */
-	task_ctx->ystorm_st_context.expect_first_xfer = 1;
-	task_ctx->ystorm_st_context.data_2_trns_rem = io_req->data_xfer_len;
-	/* Check if this is required */
-	task_ctx->ystorm_st_context.ox_id = io_req->xid;
-	task_ctx->ystorm_st_context.task_rety_identifier =
-	    io_req->task_retry_identifier;
-
-	/* T Storm ag context */
-	SET_FIELD(task_ctx->tstorm_ag_context.flags0,
-	    TSTORM_FCOE_TASK_AG_CTX_CONNECTION_TYPE, PROTOCOLID_FCOE);
-	task_ctx->tstorm_ag_context.icid = (u16)fcport->fw_cid;
-
-	/* T Storm st context */
-	SET_FIELD(task_ctx->tstorm_st_context.read_write.flags,
-	    FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_EXP_FIRST_FRAME,
-	    1);
-	task_ctx->tstorm_st_context.read_write.rx_id = 0xffff;
-
-	task_ctx->tstorm_st_context.read_only.dev_type =
-	    FCOE_TASK_DEV_TYPE_DISK;
-	task_ctx->tstorm_st_context.read_only.conf_supported = 0;
-	task_ctx->tstorm_st_context.read_only.cid = fcport->fw_cid;
-
-	/* Completion queue for response. */
-	task_ctx->tstorm_st_context.read_only.glbl_q_num = cq_idx;
-	task_ctx->tstorm_st_context.read_only.fcp_cmd_trns_size =
-	    io_req->data_xfer_len;
-	task_ctx->tstorm_st_context.read_write.e_d_tov_exp_timeout_val =
-	    lport->e_d_tov;
-
-	task_ctx->ustorm_ag_context.global_cq_num = cq_idx;
-	io_req->fp_idx = cq_idx;
-
-	bd_count = bd_tbl->bd_valid;
-	if (task_type == FCOE_TASK_TYPE_WRITE_INITIATOR) {
-		/* Setup WRITE task */
-		struct fcoe_sge *fcoe_bd_tbl = bd_tbl->bd_tbl;
-
-		task_ctx->ystorm_st_context.task_type =
-		    FCOE_TASK_TYPE_WRITE_INITIATOR;
-		data_desc = &task_ctx->ystorm_st_context.data_desc;
-
-		if (io_req->use_slowpath) {
-			SET_FIELD(task_ctx->ystorm_st_context.sgl_mode,
-			    YSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE,
-			    FCOE_SLOW_SGL);
-			data_desc->slow.base_sgl_addr.lo =
-			    U64_LO(bd_tbl->bd_tbl_dma);
-			data_desc->slow.base_sgl_addr.hi =
-			    U64_HI(bd_tbl->bd_tbl_dma);
-			data_desc->slow.remainder_num_sges = bd_count;
-			data_desc->slow.curr_sge_off = 0;
-			data_desc->slow.curr_sgl_index = 0;
-			qedf->slow_sge_ios++;
-			io_req->sge_type = QEDF_IOREQ_SLOW_SGE;
-		} else {
-			SET_FIELD(task_ctx->ystorm_st_context.sgl_mode,
-			    YSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE,
-			    (bd_count <= 4) ? (enum fcoe_sgl_mode)bd_count :
-			    FCOE_MUL_FAST_SGES);
-
-			if (bd_count == 1) {
-				data_desc->single_sge.sge_addr.lo =
-				    fcoe_bd_tbl->sge_addr.lo;
-				data_desc->single_sge.sge_addr.hi =
-				    fcoe_bd_tbl->sge_addr.hi;
-				data_desc->single_sge.size =
-				    fcoe_bd_tbl->size;
-				data_desc->single_sge.is_valid_sge = 0;
-				qedf->single_sge_ios++;
-				io_req->sge_type = QEDF_IOREQ_SINGLE_SGE;
-			} else {
-				data_desc->fast.sgl_start_addr.lo =
-				    U64_LO(bd_tbl->bd_tbl_dma);
-				data_desc->fast.sgl_start_addr.hi =
-				    U64_HI(bd_tbl->bd_tbl_dma);
-				data_desc->fast.sgl_byte_offset =
-				    data_desc->fast.sgl_start_addr.lo &
-				    (QEDF_PAGE_SIZE - 1);
-				if (data_desc->fast.sgl_byte_offset > 0)
-					QEDF_ERR(&(qedf->dbg_ctx),
-					    "byte_offset=%u for xid=0x%x.\n",
-					    io_req->xid,
-					    data_desc->fast.sgl_byte_offset);
-				data_desc->fast.task_reuse_cnt =
-				    io_req->reuse_count;
-				io_req->reuse_count++;
-				if (io_req->reuse_count == QEDF_MAX_REUSE) {
-					*ptu_invalidate = 1;
-					io_req->reuse_count = 0;
-				}
-				qedf->fast_sge_ios++;
-				io_req->sge_type = QEDF_IOREQ_FAST_SGE;
-			}
-		}
-
-		/* T Storm context */
-		task_ctx->tstorm_st_context.read_only.task_type =
-		    FCOE_TASK_TYPE_WRITE_INITIATOR;
-
-		/* M Storm context */
-		tmp_sgl_mode = GET_FIELD(task_ctx->ystorm_st_context.sgl_mode,
-		    YSTORM_FCOE_TASK_ST_CTX_TX_SGL_MODE);
-		SET_FIELD(task_ctx->mstorm_st_context.non_fp.tx_rx_sgl_mode,
-		    FCOE_MSTORM_FCOE_TASK_ST_CTX_NON_FP_TX_SGL_MODE,
-		    tmp_sgl_mode);
-
 	} else {
-		/* Setup READ task */
-
-		/* M Storm context */
-		struct fcoe_sge *fcoe_bd_tbl = bd_tbl->bd_tbl;
-
-		data_desc = &task_ctx->mstorm_st_context.fp.data_desc;
-		task_ctx->mstorm_st_context.fp.data_2_trns_rem =
-		    io_req->data_xfer_len;
-
-		if (io_req->use_slowpath) {
-			SET_FIELD(
-			    task_ctx->mstorm_st_context.non_fp.tx_rx_sgl_mode,
-			    FCOE_MSTORM_FCOE_TASK_ST_CTX_NON_FP_RX_SGL_MODE,
-			    FCOE_SLOW_SGL);
-			data_desc->slow.base_sgl_addr.lo =
-			    U64_LO(bd_tbl->bd_tbl_dma);
-			data_desc->slow.base_sgl_addr.hi =
-			    U64_HI(bd_tbl->bd_tbl_dma);
-			data_desc->slow.remainder_num_sges =
-			    bd_count;
-			data_desc->slow.curr_sge_off = 0;
-			data_desc->slow.curr_sgl_index = 0;
-			qedf->slow_sge_ios++;
-			io_req->sge_type = QEDF_IOREQ_SLOW_SGE;
+		if (sc_cmd->sc_data_direction == DMA_TO_DEVICE) {
+			task_type = FCOE_TASK_TYPE_WRITE_INITIATOR;
+			tx_io_size = io_req->data_xfer_len;
 		} else {
-			SET_FIELD(
-			    task_ctx->mstorm_st_context.non_fp.tx_rx_sgl_mode,
-			    FCOE_MSTORM_FCOE_TASK_ST_CTX_NON_FP_RX_SGL_MODE,
-			    (bd_count <= 4) ? (enum fcoe_sgl_mode)bd_count :
-			    FCOE_MUL_FAST_SGES);
-
-			if (bd_count == 1) {
-				data_desc->single_sge.sge_addr.lo =
-				    fcoe_bd_tbl->sge_addr.lo;
-				data_desc->single_sge.sge_addr.hi =
-				    fcoe_bd_tbl->sge_addr.hi;
-				data_desc->single_sge.size =
-				    fcoe_bd_tbl->size;
-				data_desc->single_sge.is_valid_sge = 0;
-				qedf->single_sge_ios++;
-				io_req->sge_type = QEDF_IOREQ_SINGLE_SGE;
-			} else {
-				data_desc->fast.sgl_start_addr.lo =
-				    U64_LO(bd_tbl->bd_tbl_dma);
-				data_desc->fast.sgl_start_addr.hi =
-				    U64_HI(bd_tbl->bd_tbl_dma);
-				data_desc->fast.sgl_byte_offset = 0;
-				data_desc->fast.task_reuse_cnt =
-				    io_req->reuse_count;
-				io_req->reuse_count++;
-				if (io_req->reuse_count == QEDF_MAX_REUSE) {
-					*ptu_invalidate = 1;
-					io_req->reuse_count = 0;
-				}
-				qedf->fast_sge_ios++;
-				io_req->sge_type = QEDF_IOREQ_FAST_SGE;
-			}
+			task_type = FCOE_TASK_TYPE_READ_INITIATOR;
+			rx_io_size = io_req->data_xfer_len;
 		}
-
-		/* Y Storm context */
-		task_ctx->ystorm_st_context.expect_first_xfer = 0;
-		task_ctx->ystorm_st_context.task_type =
-		    FCOE_TASK_TYPE_READ_INITIATOR;
-
-		/* T Storm context */
-		task_ctx->tstorm_st_context.read_only.task_type =
-		    FCOE_TASK_TYPE_READ_INITIATOR;
-		mst_sgl_mode = GET_FIELD(
-		    task_ctx->mstorm_st_context.non_fp.tx_rx_sgl_mode,
-		    FCOE_MSTORM_FCOE_TASK_ST_CTX_NON_FP_RX_SGL_MODE);
-		SET_FIELD(task_ctx->tstorm_st_context.read_write.flags,
-		    FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_RX_SGL_MODE,
-		    mst_sgl_mode);
 	}
 
+	/* Setup the fields for fcoe_task_params */
+	io_req->task_params->context = task_ctx;
+	io_req->task_params->sqe = sqe;
+	io_req->task_params->task_type = task_type;
+	io_req->task_params->tx_io_size = tx_io_size;
+	io_req->task_params->rx_io_size = rx_io_size;
+	io_req->task_params->conn_cid = fcport->fw_cid;
+	io_req->task_params->itid = io_req->xid;
+	io_req->task_params->cq_rss_number = cq_idx;
+	io_req->task_params->is_tape_device = fcport->dev_type;
+
+	/* Fill in information for scatter/gather list */
+	if (io_req->cmd_type != QEDF_TASK_MGMT_CMD) {
+		bd_count = bd_tbl->bd_valid;
+		io_req->sgl_task_params->sgl = bd_tbl->bd_tbl;
+		io_req->sgl_task_params->sgl_phys_addr.lo =
+			U64_LO(bd_tbl->bd_tbl_dma);
+		io_req->sgl_task_params->sgl_phys_addr.hi =
+			U64_HI(bd_tbl->bd_tbl_dma);
+		io_req->sgl_task_params->num_sges = bd_count;
+		io_req->sgl_task_params->total_buffer_size =
+		    scsi_bufflen(io_req->sc_cmd);
+		io_req->sgl_task_params->small_mid_sge =
+			io_req->use_slowpath;
+	}
+
+	/* Fill in physical address of sense buffer */
+	sense_data_buffer_phys_addr.lo = U64_LO(io_req->sense_buffer_dma);
+	sense_data_buffer_phys_addr.hi = U64_HI(io_req->sense_buffer_dma);
+
 	/* fill FCP_CMND IU */
-	fcp_cmnd = (u32 *)task_ctx->ystorm_st_context.tx_info_union.fcp_cmd_payload.opaque;
-	qedf_build_fcp_cmnd(io_req, (struct fcp_cmnd *)&tmp_fcp_cmnd);
+	qedf_build_fcp_cmnd(io_req, (struct fcp_cmnd *)tmp_fcp_cmnd);
 
 	/* Swap fcp_cmnd since FC is big endian */
 	cnt = sizeof(struct fcp_cmnd) / sizeof(u32);
-
 	for (i = 0; i < cnt; i++) {
-		*fcp_cmnd = cpu_to_be32(tmp_fcp_cmnd[i]);
-		fcp_cmnd++;
+		tmp_fcp_cmnd[i] = cpu_to_be32(tmp_fcp_cmnd[i]);
+	}
+	memcpy(fcp_cmnd, tmp_fcp_cmnd, sizeof(struct fcp_cmnd));
+
+	init_initiator_rw_fcoe_task(io_req->task_params,
+				    io_req->sgl_task_params,
+				    sense_data_buffer_phys_addr,
+				    io_req->task_retry_identifier, fcp_cmnd);
+
+	/* Increment SGL type counters */
+	if (bd_count == 1) {
+		qedf->single_sge_ios++;
+		io_req->sge_type = QEDF_IOREQ_SINGLE_SGE;
+	} else if (io_req->use_slowpath) {
+		qedf->slow_sge_ios++;
+		io_req->sge_type = QEDF_IOREQ_SLOW_SGE;
+	} else {
+		qedf->fast_sge_ios++;
+		io_req->sge_type = QEDF_IOREQ_FAST_SGE;
 	}
-
-	/* M Storm context - Sense buffer */
-	task_ctx->mstorm_st_context.non_fp.rsp_buf_addr.lo =
-		U64_LO(io_req->sense_buffer_dma);
-	task_ctx->mstorm_st_context.non_fp.rsp_buf_addr.hi =
-		U64_HI(io_req->sense_buffer_dma);
 }
 
 void qedf_init_mp_task(struct qedf_ioreq *io_req,
-	struct fcoe_task_context *task_ctx)
+	struct fcoe_task_context *task_ctx, struct fcoe_wqe *sqe)
 {
 	struct qedf_mp_req *mp_req = &(io_req->mp_req);
 	struct qedf_rport *fcport = io_req->fcport;
 	struct qedf_ctx *qedf = io_req->fcport->qedf;
 	struct fc_frame_header *fc_hdr;
-	enum fcoe_task_type task_type = 0;
-	union fcoe_data_desc_ctx *data_desc;
+	struct fcoe_tx_mid_path_params task_fc_hdr;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
 
-	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "Initializing MP task "
-		   "for cmd_type = %d\n", io_req->cmd_type);
+	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC,
+		  "Initializing MP task for cmd_type=%d\n",
+		  io_req->cmd_type);
 
 	qedf->control_requests++;
 
-	/* Obtain task_type */
-	if ((io_req->cmd_type == QEDF_TASK_MGMT_CMD) ||
-	    (io_req->cmd_type == QEDF_ELS)) {
-		task_type = FCOE_TASK_TYPE_MIDPATH;
-	} else if (io_req->cmd_type == QEDF_ABTS) {
-		task_type = FCOE_TASK_TYPE_ABTS;
-	}
-
+	memset(&tx_sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
 	memset(task_ctx, 0, sizeof(struct fcoe_task_context));
+	memset(&task_fc_hdr, 0, sizeof(struct fcoe_tx_mid_path_params));
 
 	/* Setup the task from io_req for easy reference */
 	io_req->task = task_ctx;
 
-	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "task type = %d\n",
-		   task_type);
-
-	/* YSTORM only */
-	{
-		/* Initialize YSTORM task context */
-		struct fcoe_tx_mid_path_params *task_fc_hdr =
-		    &task_ctx->ystorm_st_context.tx_info_union.tx_params.mid_path;
-		memset(task_fc_hdr, 0, sizeof(struct fcoe_tx_mid_path_params));
-		task_ctx->ystorm_st_context.task_rety_identifier =
-		    io_req->task_retry_identifier;
-
-		/* Init SGL parameters */
-		if ((task_type == FCOE_TASK_TYPE_MIDPATH) ||
-		    (task_type == FCOE_TASK_TYPE_UNSOLICITED)) {
-			data_desc = &task_ctx->ystorm_st_context.data_desc;
-			data_desc->slow.base_sgl_addr.lo =
-			    U64_LO(mp_req->mp_req_bd_dma);
-			data_desc->slow.base_sgl_addr.hi =
-			    U64_HI(mp_req->mp_req_bd_dma);
-			data_desc->slow.remainder_num_sges = 1;
-			data_desc->slow.curr_sge_off = 0;
-			data_desc->slow.curr_sgl_index = 0;
-		}
-
-		fc_hdr = &(mp_req->req_fc_hdr);
-		if (task_type == FCOE_TASK_TYPE_MIDPATH) {
-			fc_hdr->fh_ox_id = io_req->xid;
-			fc_hdr->fh_rx_id = htons(0xffff);
-		} else if (task_type == FCOE_TASK_TYPE_UNSOLICITED) {
-			fc_hdr->fh_rx_id = io_req->xid;
-		}
+	/* Setup the fields for fcoe_task_params */
+	io_req->task_params->context = task_ctx;
+	io_req->task_params->sqe = sqe;
+	io_req->task_params->task_type = FCOE_TASK_TYPE_MIDPATH;
+	io_req->task_params->tx_io_size = io_req->data_xfer_len;
+	/* rx_io_size tells the f/w how large a response buffer we have */
+	io_req->task_params->rx_io_size = PAGE_SIZE;
+	io_req->task_params->conn_cid = fcport->fw_cid;
+	io_req->task_params->itid = io_req->xid;
+	/* Return middle path commands on CQ 0 */
+	io_req->task_params->cq_rss_number = 0;
+	io_req->task_params->is_tape_device = fcport->dev_type;
+
+	fc_hdr = &(mp_req->req_fc_hdr);
+	/* Set OX_ID and RX_ID based on driver task id */
+	fc_hdr->fh_ox_id = io_req->xid;
+	fc_hdr->fh_rx_id = htons(0xffff);
+
+	/* Set up FC header information */
+	task_fc_hdr.parameter = fc_hdr->fh_parm_offset;
+	task_fc_hdr.r_ctl = fc_hdr->fh_r_ctl;
+	task_fc_hdr.type = fc_hdr->fh_type;
+	task_fc_hdr.cs_ctl = fc_hdr->fh_cs_ctl;
+	task_fc_hdr.df_ctl = fc_hdr->fh_df_ctl;
+	task_fc_hdr.rx_id = fc_hdr->fh_rx_id;
+	task_fc_hdr.ox_id = fc_hdr->fh_ox_id;
+
+	/* Set up s/g list parameters for request buffer */
+	tx_sgl_task_params.sgl = mp_req->mp_req_bd;
+	tx_sgl_task_params.sgl_phys_addr.lo = U64_LO(mp_req->mp_req_bd_dma);
+	tx_sgl_task_params.sgl_phys_addr.hi = U64_HI(mp_req->mp_req_bd_dma);
+	tx_sgl_task_params.num_sges = 1;
+	/* Set PAGE_SIZE for now since sg element is that size ??? */
+	tx_sgl_task_params.total_buffer_size = io_req->data_xfer_len;
+	tx_sgl_task_params.small_mid_sge = 0;
+
+	/* Set up s/g list parameters for request buffer */
+	rx_sgl_task_params.sgl = mp_req->mp_resp_bd;
+	rx_sgl_task_params.sgl_phys_addr.lo = U64_LO(mp_req->mp_resp_bd_dma);
+	rx_sgl_task_params.sgl_phys_addr.hi = U64_HI(mp_req->mp_resp_bd_dma);
+	rx_sgl_task_params.num_sges = 1;
+	/* Set PAGE_SIZE for now since sg element is that size ??? */
+	rx_sgl_task_params.total_buffer_size = PAGE_SIZE;
+	rx_sgl_task_params.small_mid_sge = 0;
 
-		/* Fill FC Header into middle path buffer */
-		task_fc_hdr->parameter = fc_hdr->fh_parm_offset;
-		task_fc_hdr->r_ctl = fc_hdr->fh_r_ctl;
-		task_fc_hdr->type = fc_hdr->fh_type;
-		task_fc_hdr->cs_ctl = fc_hdr->fh_cs_ctl;
-		task_fc_hdr->df_ctl = fc_hdr->fh_df_ctl;
-		task_fc_hdr->rx_id = fc_hdr->fh_rx_id;
-		task_fc_hdr->ox_id = fc_hdr->fh_ox_id;
-
-		task_ctx->ystorm_st_context.data_2_trns_rem =
-		    io_req->data_xfer_len;
-		task_ctx->ystorm_st_context.task_type = task_type;
-	}
-
-	/* TSTORM ONLY */
-	{
-		task_ctx->tstorm_ag_context.icid = (u16)fcport->fw_cid;
-		task_ctx->tstorm_st_context.read_only.cid = fcport->fw_cid;
-		/* Always send middle-path repsonses on CQ #0 */
-		task_ctx->tstorm_st_context.read_only.glbl_q_num = 0;
-		io_req->fp_idx = 0;
-		SET_FIELD(task_ctx->tstorm_ag_context.flags0,
-		    TSTORM_FCOE_TASK_AG_CTX_CONNECTION_TYPE,
-		    PROTOCOLID_FCOE);
-		task_ctx->tstorm_st_context.read_only.task_type = task_type;
-		SET_FIELD(task_ctx->tstorm_st_context.read_write.flags,
-		    FCOE_TSTORM_FCOE_TASK_ST_CTX_READ_WRITE_EXP_FIRST_FRAME,
-		    1);
-		task_ctx->tstorm_st_context.read_write.rx_id = 0xffff;
-	}
-
-	/* MSTORM only */
-	{
-		if (task_type == FCOE_TASK_TYPE_MIDPATH) {
-			/* Initialize task context */
-			data_desc = &task_ctx->mstorm_st_context.fp.data_desc;
-
-			/* Set cache sges address and length */
-			data_desc->slow.base_sgl_addr.lo =
-			    U64_LO(mp_req->mp_resp_bd_dma);
-			data_desc->slow.base_sgl_addr.hi =
-			    U64_HI(mp_req->mp_resp_bd_dma);
-			data_desc->slow.remainder_num_sges = 1;
-			data_desc->slow.curr_sge_off = 0;
-			data_desc->slow.curr_sgl_index = 0;
 
-			/*
-			 * Also need to fil in non-fastpath response address
-			 * for middle path commands.
-			 */
-			task_ctx->mstorm_st_context.non_fp.rsp_buf_addr.lo =
-			    U64_LO(mp_req->mp_resp_bd_dma);
-			task_ctx->mstorm_st_context.non_fp.rsp_buf_addr.hi =
-			    U64_HI(mp_req->mp_resp_bd_dma);
-		}
-	}
-
-	/* USTORM ONLY */
-	{
-		task_ctx->ustorm_ag_context.global_cq_num = 0;
-	}
+	/*
+	 * Last arg is 0 as previous code did not set that we wanted the
+	 * fc header information.
+	 */
+	init_initiator_midpath_unsolicited_fcoe_task(io_req->task_params,
+						     &task_fc_hdr,
+						     &tx_sgl_task_params,
+						     &rx_sgl_task_params, 0);
 
-	/* I/O stats. Middle path commands always use slow SGEs */
-	qedf->slow_sge_ios++;
-	io_req->sge_type = QEDF_IOREQ_SLOW_SGE;
+	/* Midpath requests always consume 1 SGE */
+	qedf->single_sge_ios++;
 }
 
-void qedf_add_to_sq(struct qedf_rport *fcport, u16 xid, u32 ptu_invalidate,
-	enum fcoe_task_type req_type, u32 offset)
+/* Presumed that fcport->rport_lock is held */
+u16 qedf_get_sqe_idx(struct qedf_rport *fcport)
 {
-	struct fcoe_wqe *sqe;
 	uint16_t total_sqe = (fcport->sq_mem_size)/(sizeof(struct fcoe_wqe));
+	u16 rval;
 
-	sqe = &fcport->sq[fcport->sq_prod_idx];
+	rval = fcport->sq_prod_idx;
 
+	/* Adjust ring index */
 	fcport->sq_prod_idx++;
 	fcport->fw_sq_prod_idx++;
 	if (fcport->sq_prod_idx == total_sqe)
 		fcport->sq_prod_idx = 0;
 
-	switch (req_type) {
-	case FCOE_TASK_TYPE_WRITE_INITIATOR:
-	case FCOE_TASK_TYPE_READ_INITIATOR:
-		SET_FIELD(sqe->flags, FCOE_WQE_REQ_TYPE, SEND_FCOE_CMD);
-		if (ptu_invalidate)
-			SET_FIELD(sqe->flags, FCOE_WQE_INVALIDATE_PTU, 1);
-		break;
-	case FCOE_TASK_TYPE_MIDPATH:
-		SET_FIELD(sqe->flags, FCOE_WQE_REQ_TYPE, SEND_FCOE_MIDPATH);
-		break;
-	case FCOE_TASK_TYPE_ABTS:
-		SET_FIELD(sqe->flags, FCOE_WQE_REQ_TYPE,
-		    SEND_FCOE_ABTS_REQUEST);
-		break;
-	case FCOE_TASK_TYPE_EXCHANGE_CLEANUP:
-		SET_FIELD(sqe->flags, FCOE_WQE_REQ_TYPE,
-		     FCOE_EXCHANGE_CLEANUP);
-		break;
-	case FCOE_TASK_TYPE_SEQUENCE_CLEANUP:
-		SET_FIELD(sqe->flags, FCOE_WQE_REQ_TYPE,
-		    FCOE_SEQUENCE_RECOVERY);
-		/* NOTE: offset param only used for sequence recovery */
-		sqe->additional_info_union.seq_rec_updated_offset = offset;
-		break;
-	case FCOE_TASK_TYPE_UNSOLICITED:
-		break;
-	default:
-		break;
-	}
-
-	sqe->task_id = xid;
-
-	/* Make sure SQ data is coherent */
-	wmb();
-
+	return rval;
 }
 
 void qedf_ring_doorbell(struct qedf_rport *fcport)
@@ -1029,7 +847,8 @@ int qedf_post_io_req(struct qedf_rport *fcport, struct qedf_ioreq *io_req)
 	struct fcoe_task_context *task_ctx;
 	u16 xid;
 	enum fcoe_task_type req_type = 0;
-	u32 ptu_invalidate = 0;
+	struct fcoe_wqe *sqe;
+	u16 sqe_idx;
 
 	/* Initialize rest of io_req fileds */
 	io_req->data_xfer_len = scsi_bufflen(sc_cmd);
@@ -1061,6 +880,16 @@ int qedf_post_io_req(struct qedf_rport *fcport, struct qedf_ioreq *io_req)
 		return -EAGAIN;
 	}
 
+	if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
+		QEDF_ERR(&(qedf->dbg_ctx), "Session not offloaded yet.\n");
+		kref_put(&io_req->refcount, qedf_release_cmd);
+	}
+
+	/* Obtain free SQE */
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+
 	/* Get the task context */
 	task_ctx = qedf_get_task_mem(&qedf->tasks, xid);
 	if (!task_ctx) {
@@ -1070,15 +899,7 @@ int qedf_post_io_req(struct qedf_rport *fcport, struct qedf_ioreq *io_req)
 		return -EINVAL;
 	}
 
-	qedf_init_task(fcport, lport, io_req, &ptu_invalidate, task_ctx);
-
-	if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
-		QEDF_ERR(&(qedf->dbg_ctx), "Session not offloaded yet.\n");
-		kref_put(&io_req->refcount, qedf_release_cmd);
-	}
-
-	/* Obtain free SQ entry */
-	qedf_add_to_sq(fcport, xid, ptu_invalidate, req_type, 0);
+	qedf_init_task(fcport, lport, io_req, task_ctx, sqe);
 
 	/* Ring doorbell */
 	qedf_ring_doorbell(fcport);
@@ -1661,6 +1482,8 @@ int qedf_initiate_abts(struct qedf_ioreq *io_req, bool return_scsi_cmd_on_abts)
 	u32 r_a_tov = 0;
 	int rc = 0;
 	unsigned long flags;
+	struct fcoe_wqe *sqe;
+	u16 sqe_idx;
 
 	r_a_tov = rdata->r_a_tov;
 	lport = qedf->lport;
@@ -1712,10 +1535,12 @@ int qedf_initiate_abts(struct qedf_ioreq *io_req, bool return_scsi_cmd_on_abts)
 
 	spin_lock_irqsave(&fcport->rport_lock, flags);
 
-	/* Add ABTS to send queue */
-	qedf_add_to_sq(fcport, xid, 0, FCOE_TASK_TYPE_ABTS, 0);
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+	io_req->task_params->sqe = sqe;
 
-	/* Ring doorbell */
+	init_initiator_abort_fcoe_task(io_req->task_params);
 	qedf_ring_doorbell(fcport);
 
 	spin_unlock_irqrestore(&fcport->rport_lock, flags);
@@ -1784,8 +1609,8 @@ void qedf_process_abts_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
 int qedf_init_mp_req(struct qedf_ioreq *io_req)
 {
 	struct qedf_mp_req *mp_req;
-	struct fcoe_sge *mp_req_bd;
-	struct fcoe_sge *mp_resp_bd;
+	struct scsi_sge *mp_req_bd;
+	struct scsi_sge *mp_resp_bd;
 	struct qedf_ctx *qedf = io_req->fcport->qedf;
 	dma_addr_t addr;
 	uint64_t sz;
@@ -1819,7 +1644,7 @@ int qedf_init_mp_req(struct qedf_ioreq *io_req)
 	}
 
 	/* Allocate and map mp_req_bd and mp_resp_bd */
-	sz = sizeof(struct fcoe_sge);
+	sz = sizeof(struct scsi_sge);
 	mp_req->mp_req_bd = dma_alloc_coherent(&qedf->pdev->dev, sz,
 	    &mp_req->mp_req_bd_dma, GFP_KERNEL);
 	if (!mp_req->mp_req_bd) {
@@ -1841,7 +1666,7 @@ int qedf_init_mp_req(struct qedf_ioreq *io_req)
 	mp_req_bd = mp_req->mp_req_bd;
 	mp_req_bd->sge_addr.lo = U64_LO(addr);
 	mp_req_bd->sge_addr.hi = U64_HI(addr);
-	mp_req_bd->size = QEDF_PAGE_SIZE;
+	mp_req_bd->sge_len = QEDF_PAGE_SIZE;
 
 	/*
 	 * MP buffer is either a task mgmt command or an ELS.
@@ -1852,7 +1677,7 @@ int qedf_init_mp_req(struct qedf_ioreq *io_req)
 	addr = mp_req->resp_buf_dma;
 	mp_resp_bd->sge_addr.lo = U64_LO(addr);
 	mp_resp_bd->sge_addr.hi = U64_HI(addr);
-	mp_resp_bd->size = QEDF_PAGE_SIZE;
+	mp_resp_bd->sge_len = QEDF_PAGE_SIZE;
 
 	return 0;
 }
@@ -1895,6 +1720,8 @@ int qedf_initiate_cleanup(struct qedf_ioreq *io_req,
 	int tmo = 0;
 	int rc = SUCCESS;
 	unsigned long flags;
+	struct fcoe_wqe *sqe;
+	u16 sqe_idx;
 
 	fcport = io_req->fcport;
 	if (!fcport) {
@@ -1940,12 +1767,16 @@ int qedf_initiate_cleanup(struct qedf_ioreq *io_req,
 
 	init_completion(&io_req->tm_done);
 
-	/* Obtain free SQ entry */
 	spin_lock_irqsave(&fcport->rport_lock, flags);
-	qedf_add_to_sq(fcport, xid, 0, FCOE_TASK_TYPE_EXCHANGE_CLEANUP, 0);
 
-	/* Ring doorbell */
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+	io_req->task_params->sqe = sqe;
+
+	init_initiator_cleanup_fcoe_task(io_req->task_params);
 	qedf_ring_doorbell(fcport);
+
 	spin_unlock_irqrestore(&fcport->rport_lock, flags);
 
 	tmo = wait_for_completion_timeout(&io_req->tm_done,
@@ -1991,16 +1822,15 @@ static int qedf_execute_tmf(struct qedf_rport *fcport, struct scsi_cmnd *sc_cmd,
 	uint8_t tm_flags)
 {
 	struct qedf_ioreq *io_req;
-	struct qedf_mp_req *tm_req;
 	struct fcoe_task_context *task;
-	struct fc_frame_header *fc_hdr;
-	struct fcp_cmnd *fcp_cmnd;
 	struct qedf_ctx *qedf = fcport->qedf;
+	struct fc_lport *lport = qedf->lport;
 	int rc = 0;
 	uint16_t xid;
-	uint32_t sid, did;
 	int tmo = 0;
 	unsigned long flags;
+	struct fcoe_wqe *sqe;
+	u16 sqe_idx;
 
 	if (!sc_cmd) {
 		QEDF_ERR(&(qedf->dbg_ctx), "invalid arg\n");
@@ -2031,36 +1861,14 @@ static int qedf_execute_tmf(struct qedf_rport *fcport, struct scsi_cmnd *sc_cmd,
 	/* Set the return CPU to be the same as the request one */
 	io_req->cpu = smp_processor_id();
 
-	tm_req = (struct qedf_mp_req *)&(io_req->mp_req);
-
-	rc = qedf_init_mp_req(io_req);
-	if (rc == FAILED) {
-		QEDF_ERR(&(qedf->dbg_ctx), "Task mgmt MP request init "
-			  "failed\n");
-		kref_put(&io_req->refcount, qedf_release_cmd);
-		goto reset_tmf_err;
-	}
-
 	/* Set TM flags */
-	io_req->io_req_flags = 0;
-	tm_req->tm_flags = tm_flags;
+	io_req->io_req_flags = QEDF_READ;
+	io_req->data_xfer_len = 0;
+	io_req->tm_flags = tm_flags;
 
 	/* Default is to return a SCSI command when an error occurs */
 	io_req->return_scsi_cmd_on_abts = true;
 
-	/* Fill FCP_CMND */
-	qedf_build_fcp_cmnd(io_req, (struct fcp_cmnd *)tm_req->req_buf);
-	fcp_cmnd = (struct fcp_cmnd *)tm_req->req_buf;
-	memset(fcp_cmnd->fc_cdb, 0, FCP_CMND_LEN);
-	fcp_cmnd->fc_dl = 0;
-
-	/* Fill FC header */
-	fc_hdr = &(tm_req->req_fc_hdr);
-	sid = fcport->sid;
-	did = fcport->rdata->ids.port_id;
-	__fc_fill_fc_hdr(fc_hdr, FC_RCTL_DD_UNSOL_CMD, sid, did,
-			   FC_TYPE_FCP, FC_FC_FIRST_SEQ | FC_FC_END_SEQ |
-			   FC_FC_SEQ_INIT, 0);
 	/* Obtain exchange id */
 	xid = io_req->xid;
 
@@ -2069,16 +1877,18 @@ static int qedf_execute_tmf(struct qedf_rport *fcport, struct scsi_cmnd *sc_cmd,
 
 	/* Initialize task context for this IO request */
 	task = qedf_get_task_mem(&qedf->tasks, xid);
-	qedf_init_mp_task(io_req, task);
 
 	init_completion(&io_req->tm_done);
 
-	/* Obtain free SQ entry */
 	spin_lock_irqsave(&fcport->rport_lock, flags);
-	qedf_add_to_sq(fcport, xid, 0, FCOE_TASK_TYPE_MIDPATH, 0);
 
-	/* Ring doorbell */
+	sqe_idx = qedf_get_sqe_idx(fcport);
+	sqe = &fcport->sq[sqe_idx];
+	memset(sqe, 0, sizeof(struct fcoe_wqe));
+
+	qedf_init_task(fcport, lport, io_req, task, sqe);
 	qedf_ring_doorbell(fcport);
+
 	spin_unlock_irqrestore(&fcport->rport_lock, flags);
 
 	tmo = wait_for_completion_timeout(&io_req->tm_done,
@@ -2162,14 +1972,6 @@ void qedf_process_tmf_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
 	struct qedf_ioreq *io_req)
 {
 	struct fcoe_cqe_rsp_info *fcp_rsp;
-	struct fcoe_cqe_midpath_info *mp_info;
-
-
-	/* Get TMF response length from CQE */
-	mp_info = &cqe->cqe_info.midpath_info;
-	io_req->mp_req.resp_len = mp_info->data_placement_size;
-	QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_SCSI_TM,
-	    "Response len is %d.\n", io_req->mp_req.resp_len);
 
 	fcp_rsp = &cqe->cqe_info.rsp_info;
 	qedf_parse_fcp_rsp(io_req, fcp_rsp);
diff --git a/drivers/scsi/qedi/Makefile b/drivers/scsi/qedi/Makefile
index 2b3e16b24299..90a6925577cc 100644
--- a/drivers/scsi/qedi/Makefile
+++ b/drivers/scsi/qedi/Makefile
@@ -1,5 +1,5 @@
 obj-$(CONFIG_QEDI) := qedi.o
 qedi-y := qedi_main.o qedi_iscsi.o qedi_fw.o qedi_sysfs.o \
-	    qedi_dbg.o
+	    qedi_dbg.o qedi_fw_api.o
 
 qedi-$(CONFIG_DEBUG_FS) += qedi_debugfs.o
diff --git a/drivers/scsi/qedi/qedi_fw.c b/drivers/scsi/qedi/qedi_fw.c
index c9f0ef4e11b3..eca40b0513a3 100644
--- a/drivers/scsi/qedi/qedi_fw.c
+++ b/drivers/scsi/qedi/qedi_fw.c
@@ -14,6 +14,8 @@
 #include "qedi.h"
 #include "qedi_iscsi.h"
 #include "qedi_gbl.h"
+#include "qedi_fw_iscsi.h"
+#include "qedi_fw_scsi.h"
 
 static int qedi_send_iscsi_tmf(struct qedi_conn *qedi_conn,
 			       struct iscsi_task *mtask);
@@ -53,8 +55,8 @@ static void qedi_process_logout_resp(struct qedi_ctx *qedi,
 	resp_hdr->exp_cmdsn = cpu_to_be32(cqe_logout_response->exp_cmd_sn);
 	resp_hdr->max_cmdsn = cpu_to_be32(cqe_logout_response->max_cmd_sn);
 
-	resp_hdr->t2wait = cpu_to_be32(cqe_logout_response->time2wait);
-	resp_hdr->t2retain = cpu_to_be32(cqe_logout_response->time2retain);
+	resp_hdr->t2wait = cpu_to_be32(cqe_logout_response->time_2_wait);
+	resp_hdr->t2retain = cpu_to_be32(cqe_logout_response->time_2_retain);
 
 	QEDI_INFO(&qedi->dbg_ctx, QEDI_LOG_TID,
 		  "Freeing tid=0x%x for cid=0x%x\n",
@@ -975,81 +977,6 @@ exit_fp_process:
 	return;
 }
 
-static void qedi_add_to_sq(struct qedi_conn *qedi_conn, struct iscsi_task *task,
-			   u16 tid, uint16_t ptu_invalidate, int is_cleanup)
-{
-	struct iscsi_wqe *wqe;
-	struct iscsi_wqe_field *cont_field;
-	struct qedi_endpoint *ep;
-	struct scsi_cmnd *sc = task->sc;
-	struct iscsi_login_req *login_hdr;
-	struct qedi_cmd *cmd = task->dd_data;
-
-	login_hdr = (struct iscsi_login_req *)task->hdr;
-	ep = qedi_conn->ep;
-	wqe = &ep->sq[ep->sq_prod_idx];
-
-	memset(wqe, 0, sizeof(*wqe));
-
-	ep->sq_prod_idx++;
-	ep->fw_sq_prod_idx++;
-	if (ep->sq_prod_idx == QEDI_SQ_SIZE)
-		ep->sq_prod_idx = 0;
-
-	if (is_cleanup) {
-		SET_FIELD(wqe->flags, ISCSI_WQE_WQE_TYPE,
-			  ISCSI_WQE_TYPE_TASK_CLEANUP);
-		wqe->task_id = tid;
-		return;
-	}
-
-	if (ptu_invalidate) {
-		SET_FIELD(wqe->flags, ISCSI_WQE_PTU_INVALIDATE,
-			  ISCSI_WQE_SET_PTU_INVALIDATE);
-	}
-
-	cont_field = &wqe->cont_prevtid_union.cont_field;
-
-	switch (task->hdr->opcode & ISCSI_OPCODE_MASK) {
-	case ISCSI_OP_LOGIN:
-	case ISCSI_OP_TEXT:
-		SET_FIELD(wqe->flags, ISCSI_WQE_WQE_TYPE,
-			  ISCSI_WQE_TYPE_MIDDLE_PATH);
-		SET_FIELD(wqe->flags, ISCSI_WQE_NUM_FAST_SGES,
-			  1);
-		cont_field->contlen_cdbsize_field = ntoh24(login_hdr->dlength);
-		break;
-	case ISCSI_OP_LOGOUT:
-	case ISCSI_OP_NOOP_OUT:
-	case ISCSI_OP_SCSI_TMFUNC:
-		 SET_FIELD(wqe->flags, ISCSI_WQE_WQE_TYPE,
-			   ISCSI_WQE_TYPE_NORMAL);
-		break;
-	default:
-		if (!sc)
-			break;
-
-		SET_FIELD(wqe->flags, ISCSI_WQE_WQE_TYPE,
-			  ISCSI_WQE_TYPE_NORMAL);
-		cont_field->contlen_cdbsize_field =
-				(sc->sc_data_direction == DMA_TO_DEVICE) ?
-				scsi_bufflen(sc) : 0;
-		if (cmd->use_slowpath)
-			SET_FIELD(wqe->flags, ISCSI_WQE_NUM_FAST_SGES, 0);
-		else
-			SET_FIELD(wqe->flags, ISCSI_WQE_NUM_FAST_SGES,
-				  (sc->sc_data_direction ==
-				   DMA_TO_DEVICE) ?
-				  min((u16)QEDI_FAST_SGE_COUNT,
-				      (u16)cmd->io_tbl.sge_valid) : 0);
-		break;
-	}
-
-	wqe->task_id = tid;
-	/* Make sure SQ data is coherent */
-	wmb();
-}
-
 static void qedi_ring_doorbell(struct qedi_conn *qedi_conn)
 {
 	struct iscsi_db_data dbell = { 0 };
@@ -1076,96 +1003,116 @@ static void qedi_ring_doorbell(struct qedi_conn *qedi_conn)
 		  qedi_conn->iscsi_conn_id);
 }
 
+static u16 qedi_get_wqe_idx(struct qedi_conn *qedi_conn)
+{
+	struct qedi_endpoint *ep;
+	u16 rval;
+
+	ep = qedi_conn->ep;
+	rval = ep->sq_prod_idx;
+
+	/* Increament SQ index */
+	ep->sq_prod_idx++;
+	ep->fw_sq_prod_idx++;
+	if (ep->sq_prod_idx == QEDI_SQ_SIZE)
+		ep->sq_prod_idx = 0;
+
+	return rval;
+}
+
 int qedi_send_iscsi_login(struct qedi_conn *qedi_conn,
 			  struct iscsi_task *task)
 {
-	struct qedi_ctx *qedi = qedi_conn->qedi;
+	struct iscsi_login_req_hdr login_req_pdu_header;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
+	struct iscsi_task_params task_params;
 	struct iscsi_task_context *fw_task_ctx;
+	struct qedi_ctx *qedi = qedi_conn->qedi;
 	struct iscsi_login_req *login_hdr;
-	struct iscsi_login_req_hdr *fw_login_req = NULL;
-	struct iscsi_cached_sge_ctx *cached_sge = NULL;
-	struct iscsi_sge *single_sge = NULL;
-	struct iscsi_sge *req_sge = NULL;
-	struct iscsi_sge *resp_sge = NULL;
+	struct scsi_sge *req_sge = NULL;
+	struct scsi_sge *resp_sge = NULL;
 	struct qedi_cmd *qedi_cmd;
-	s16 ptu_invalidate = 0;
+	struct qedi_endpoint *ep;
 	s16 tid = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
-	req_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
-	resp_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+	req_sge = (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	resp_sge = (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
 	qedi_cmd = (struct qedi_cmd *)task->dd_data;
+	ep = qedi_conn->ep;
 	login_hdr = (struct iscsi_login_req *)task->hdr;
 
 	tid = qedi_get_task_idx(qedi);
 	if (tid == -1)
 		return -ENOMEM;
 
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
 
 	qedi_cmd->task_id = tid;
 
-	/* Ystorm context */
-	fw_login_req = &fw_task_ctx->ystorm_st_context.pdu_hdr.login_req;
-	fw_login_req->opcode = login_hdr->opcode;
-	fw_login_req->version_min = login_hdr->min_version;
-	fw_login_req->version_max = login_hdr->max_version;
-	fw_login_req->flags_attr = login_hdr->flags;
-	fw_login_req->isid_tabc = *((u16 *)login_hdr->isid + 2);
-	fw_login_req->isid_d = *((u32 *)login_hdr->isid);
-	fw_login_req->tsih = login_hdr->tsih;
-	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
-	fw_login_req->itt = qedi_set_itt(tid, get_itt(task->itt));
-	fw_login_req->cid = qedi_conn->iscsi_conn_id;
-	fw_login_req->cmd_sn = be32_to_cpu(login_hdr->cmdsn);
-	fw_login_req->exp_stat_sn = be32_to_cpu(login_hdr->exp_statsn);
-	fw_login_req->exp_stat_sn = 0;
-
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&login_req_pdu_header, 0, sizeof(login_req_pdu_header));
+	memset(&tx_sgl_task_params, 0, sizeof(tx_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(rx_sgl_task_params));
+	/* Update header info */
+	login_req_pdu_header.opcode = login_hdr->opcode;
+	login_req_pdu_header.version_min = login_hdr->min_version;
+	login_req_pdu_header.version_max = login_hdr->max_version;
+	login_req_pdu_header.flags_attr = login_hdr->flags;
+	login_req_pdu_header.isid_tabc = swab32p((u32 *)login_hdr->isid);
+	login_req_pdu_header.isid_d = swab16p((u16 *)&login_hdr->isid[4]);
+
+	login_req_pdu_header.tsih = login_hdr->tsih;
+	login_req_pdu_header.hdr_second_dword = ntoh24(login_hdr->dlength);
 
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						qedi->tid_reuse_count[tid]++;
-	cached_sge =
-	       &fw_task_ctx->ystorm_st_context.state.sgl_ctx_union.cached_sge;
-	cached_sge->sge.sge_len = req_sge->sge_len;
-	cached_sge->sge.sge_addr.lo = (u32)(qedi_conn->gen_pdu.req_dma_addr);
-	cached_sge->sge.sge_addr.hi =
-			     (u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
-
-	/* Mstorm context */
-	single_sge = &fw_task_ctx->mstorm_st_context.sgl_union.single_sge;
-	fw_task_ctx->mstorm_st_context.task_type = 0x2;
-	fw_task_ctx->mstorm_ag_context.task_cid = (u16)qedi_conn->iscsi_conn_id;
-	single_sge->sge_addr.lo = resp_sge->sge_addr.lo;
-	single_sge->sge_addr.hi = resp_sge->sge_addr.hi;
-	single_sge->sge_len = resp_sge->sge_len;
-
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SINGLE_SGE, 1);
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SLOW_IO, 0);
-	fw_task_ctx->mstorm_st_context.sgl_size = 1;
-	fw_task_ctx->mstorm_st_context.rem_task_size = resp_sge->sge_len;
-
-	/* Ustorm context */
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = resp_sge->sge_len;
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len =
-						ntoh24(login_hdr->dlength);
-	fw_task_ctx->ustorm_st_context.exp_data_sn = 0;
-	fw_task_ctx->ustorm_st_context.cq_rss_number = 0;
-	fw_task_ctx->ustorm_st_context.task_type = 0x2;
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-	fw_task_ctx->ustorm_ag_context.exp_data_acked =
-						 ntoh24(login_hdr->dlength);
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
-	SET_FIELD(fw_task_ctx->ustorm_st_context.flags,
-		  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 0);
+	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
+	login_req_pdu_header.itt = qedi_set_itt(tid, get_itt(task->itt));
+	login_req_pdu_header.cid = qedi_conn->iscsi_conn_id;
+	login_req_pdu_header.cmd_sn = be32_to_cpu(login_hdr->cmdsn);
+	login_req_pdu_header.exp_stat_sn = be32_to_cpu(login_hdr->exp_statsn);
+	login_req_pdu_header.exp_stat_sn = 0;
+
+	/* Fill tx AHS and rx buffer */
+	tx_sgl_task_params.sgl =
+			       (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	tx_sgl_task_params.sgl_phys_addr.lo =
+					 (u32)(qedi_conn->gen_pdu.req_dma_addr);
+	tx_sgl_task_params.sgl_phys_addr.hi =
+			      (u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
+	tx_sgl_task_params.total_buffer_size = ntoh24(login_hdr->dlength);
+	tx_sgl_task_params.num_sges = 1;
+
+	rx_sgl_task_params.sgl =
+			      (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+	rx_sgl_task_params.sgl_phys_addr.lo =
+					(u32)(qedi_conn->gen_pdu.resp_dma_addr);
+	rx_sgl_task_params.sgl_phys_addr.hi =
+			     (u32)((u64)qedi_conn->gen_pdu.resp_dma_addr >> 32);
+	rx_sgl_task_params.total_buffer_size = resp_sge->sge_len;
+	rx_sgl_task_params.num_sges = 1;
+
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = 0;
+	task_params.tx_io_size = ntoh24(login_hdr->dlength);
+	task_params.rx_io_size = resp_sge->sge_len;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+	rval = init_initiator_login_request_task(&task_params,
+						 &login_req_pdu_header,
+						 &tx_sgl_task_params,
+						 &rx_sgl_task_params);
+	if (rval)
+		return -1;
 
 	spin_lock(&qedi_conn->list_lock);
 	list_add_tail(&qedi_cmd->io_cmd, &qedi_conn->active_cmd_list);
@@ -1173,7 +1120,6 @@ int qedi_send_iscsi_login(struct qedi_conn *qedi_conn,
 	qedi_conn->active_cmd_count++;
 	spin_unlock(&qedi_conn->list_lock);
 
-	qedi_add_to_sq(qedi_conn, task, tid, ptu_invalidate, false);
 	qedi_ring_doorbell(qedi_conn);
 	return 0;
 }
@@ -1181,65 +1127,64 @@ int qedi_send_iscsi_login(struct qedi_conn *qedi_conn,
 int qedi_send_iscsi_logout(struct qedi_conn *qedi_conn,
 			   struct iscsi_task *task)
 {
-	struct qedi_ctx *qedi = qedi_conn->qedi;
-	struct iscsi_logout_req_hdr *fw_logout_req = NULL;
-	struct iscsi_task_context *fw_task_ctx = NULL;
+	struct iscsi_logout_req_hdr logout_pdu_header;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
+	struct iscsi_task_params task_params;
+	struct iscsi_task_context *fw_task_ctx;
 	struct iscsi_logout *logout_hdr = NULL;
-	struct qedi_cmd *qedi_cmd = NULL;
-	s16  tid = 0;
-	s16 ptu_invalidate = 0;
+	struct qedi_ctx *qedi = qedi_conn->qedi;
+	struct qedi_cmd *qedi_cmd;
+	struct qedi_endpoint *ep;
+	s16 tid = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
 	qedi_cmd = (struct qedi_cmd *)task->dd_data;
 	logout_hdr = (struct iscsi_logout *)task->hdr;
+	ep = qedi_conn->ep;
 
 	tid = qedi_get_task_idx(qedi);
 	if (tid == -1)
 		return -ENOMEM;
 
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
-
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
+
 	qedi_cmd->task_id = tid;
 
-	/* Ystorm context */
-	fw_logout_req = &fw_task_ctx->ystorm_st_context.pdu_hdr.logout_req;
-	fw_logout_req->opcode = ISCSI_OPCODE_LOGOUT_REQUEST;
-	fw_logout_req->reason_code = 0x80 | logout_hdr->flags;
-	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
-	fw_logout_req->itt = qedi_set_itt(tid, get_itt(task->itt));
-	fw_logout_req->exp_stat_sn = be32_to_cpu(logout_hdr->exp_statsn);
-	fw_logout_req->cmd_sn = be32_to_cpu(logout_hdr->cmdsn);
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&logout_pdu_header, 0, sizeof(logout_pdu_header));
+	memset(&tx_sgl_task_params, 0, sizeof(tx_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(rx_sgl_task_params));
 
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						  qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						qedi->tid_reuse_count[tid]++;
-	fw_logout_req->cid = qedi_conn->iscsi_conn_id;
-	fw_task_ctx->ystorm_st_context.state.buffer_offset[0] = 0;
-
-	/* Mstorm context */
-	fw_task_ctx->mstorm_st_context.task_type = ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->mstorm_ag_context.task_cid = (u16)qedi_conn->iscsi_conn_id;
-
-	/* Ustorm context */
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = 0;
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len = 0;
-	fw_task_ctx->ustorm_st_context.exp_data_sn = 0;
-	fw_task_ctx->ustorm_st_context.task_type =  ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->ustorm_st_context.cq_rss_number = 0;
-
-	SET_FIELD(fw_task_ctx->ustorm_st_context.flags,
-		  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 0);
-	SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-		  ISCSI_REG1_NUM_FAST_SGES, 0);
-
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
+	/* Update header info */
+	logout_pdu_header.opcode = logout_hdr->opcode;
+	logout_pdu_header.reason_code = 0x80 | logout_hdr->flags;
+	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
+	logout_pdu_header.itt = qedi_set_itt(tid, get_itt(task->itt));
+	logout_pdu_header.exp_stat_sn = be32_to_cpu(logout_hdr->exp_statsn);
+	logout_pdu_header.cmd_sn = be32_to_cpu(logout_hdr->cmdsn);
+	logout_pdu_header.cid = qedi_conn->iscsi_conn_id;
+
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = 0;
+	task_params.tx_io_size = 0;
+	task_params.rx_io_size = 0;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+
+	rval = init_initiator_logout_request_task(&task_params,
+						  &logout_pdu_header,
+						  NULL, NULL);
+	if (rval)
+		return -1;
 
 	spin_lock(&qedi_conn->list_lock);
 	list_add_tail(&qedi_cmd->io_cmd, &qedi_conn->active_cmd_list);
@@ -1247,9 +1192,7 @@ int qedi_send_iscsi_logout(struct qedi_conn *qedi_conn,
 	qedi_conn->active_cmd_count++;
 	spin_unlock(&qedi_conn->list_lock);
 
-	qedi_add_to_sq(qedi_conn, task, tid, ptu_invalidate, false);
 	qedi_ring_doorbell(qedi_conn);
-
 	return 0;
 }
 
@@ -1533,47 +1476,46 @@ ldel_exit:
 static int qedi_send_iscsi_tmf(struct qedi_conn *qedi_conn,
 			       struct iscsi_task *mtask)
 {
-	struct iscsi_conn *conn = qedi_conn->cls_conn->dd_data;
+	struct iscsi_tmf_request_hdr tmf_pdu_header;
+	struct iscsi_task_params task_params;
 	struct qedi_ctx *qedi = qedi_conn->qedi;
 	struct iscsi_task_context *fw_task_ctx;
-	struct iscsi_tmf_request_hdr *fw_tmf_request;
-	struct iscsi_sge *single_sge;
-	struct qedi_cmd *qedi_cmd;
-	struct qedi_cmd *cmd;
+	struct iscsi_conn *conn = qedi_conn->cls_conn->dd_data;
 	struct iscsi_task *ctask;
 	struct iscsi_tm *tmf_hdr;
-	struct iscsi_sge *req_sge;
-	struct iscsi_sge *resp_sge;
-	u32 lun[2];
-	s16 tid = 0, ptu_invalidate = 0;
+	struct qedi_cmd *qedi_cmd;
+	struct qedi_cmd *cmd;
+	struct qedi_endpoint *ep;
+	u32 scsi_lun[2];
+	s16 tid = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
-	req_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
-	resp_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
-	qedi_cmd = (struct qedi_cmd *)mtask->dd_data;
 	tmf_hdr = (struct iscsi_tm *)mtask->hdr;
+	qedi_cmd = (struct qedi_cmd *)mtask->dd_data;
+	ep = qedi_conn->ep;
 
-	tid = qedi_cmd->task_id;
-	qedi_update_itt_map(qedi, tid, mtask->itt, qedi_cmd);
+	tid = qedi_get_task_idx(qedi);
+	if (tid == -1)
+		return -ENOMEM;
 
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
 
-	fw_tmf_request = &fw_task_ctx->ystorm_st_context.pdu_hdr.tmf_request;
-	fw_tmf_request->itt = qedi_set_itt(tid, get_itt(mtask->itt));
-	fw_tmf_request->cmd_sn = be32_to_cpu(tmf_hdr->cmdsn);
+	qedi_cmd->task_id = tid;
 
-	memcpy(lun, &tmf_hdr->lun, sizeof(struct scsi_lun));
-	fw_tmf_request->lun.lo = be32_to_cpu(lun[0]);
-	fw_tmf_request->lun.hi = be32_to_cpu(lun[1]);
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&tmf_pdu_header, 0, sizeof(tmf_pdu_header));
 
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						qedi->tid_reuse_count[tid]++;
+	/* Update header info */
+	qedi_update_itt_map(qedi, tid, mtask->itt, qedi_cmd);
+	tmf_pdu_header.itt = qedi_set_itt(tid, get_itt(mtask->itt));
+	tmf_pdu_header.cmd_sn = be32_to_cpu(tmf_hdr->cmdsn);
+
+	memcpy(scsi_lun, &tmf_hdr->lun, sizeof(struct scsi_lun));
+	tmf_pdu_header.lun.lo = be32_to_cpu(scsi_lun[0]);
+	tmf_pdu_header.lun.hi = be32_to_cpu(scsi_lun[1]);
 
 	if ((tmf_hdr->flags & ISCSI_FLAG_TM_FUNC_MASK) ==
 	     ISCSI_TM_FUNC_ABORT_TASK) {
@@ -1584,53 +1526,34 @@ static int qedi_send_iscsi_tmf(struct qedi_conn *qedi_conn,
 			return 0;
 		}
 		cmd = (struct qedi_cmd *)ctask->dd_data;
-		fw_tmf_request->rtt =
+		tmf_pdu_header.rtt =
 				qedi_set_itt(cmd->task_id,
 					     get_itt(tmf_hdr->rtt));
 	} else {
-		fw_tmf_request->rtt = ISCSI_RESERVED_TAG;
+		tmf_pdu_header.rtt = ISCSI_RESERVED_TAG;
 	}
 
-	fw_tmf_request->opcode = tmf_hdr->opcode;
-	fw_tmf_request->function = tmf_hdr->flags;
-	fw_tmf_request->hdr_second_dword = ntoh24(tmf_hdr->dlength);
-	fw_tmf_request->ref_cmd_sn = be32_to_cpu(tmf_hdr->refcmdsn);
-
-	single_sge = &fw_task_ctx->mstorm_st_context.sgl_union.single_sge;
-	fw_task_ctx->mstorm_st_context.task_type = ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->mstorm_ag_context.task_cid = (u16)qedi_conn->iscsi_conn_id;
-	single_sge->sge_addr.lo = resp_sge->sge_addr.lo;
-	single_sge->sge_addr.hi = resp_sge->sge_addr.hi;
-	single_sge->sge_len = resp_sge->sge_len;
-
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SINGLE_SGE, 1);
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SLOW_IO, 0);
-	fw_task_ctx->mstorm_st_context.sgl_size = 1;
-	fw_task_ctx->mstorm_st_context.rem_task_size = resp_sge->sge_len;
-
-	/* Ustorm context */
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = 0;
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len = 0;
-	fw_task_ctx->ustorm_st_context.exp_data_sn = 0;
-	fw_task_ctx->ustorm_st_context.task_type =  ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->ustorm_st_context.cq_rss_number = 0;
-
-	SET_FIELD(fw_task_ctx->ustorm_st_context.flags,
-		  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 0);
-	SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-		  ISCSI_REG1_NUM_FAST_SGES, 0);
-
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
-	fw_task_ctx->ustorm_st_context.lun.lo = be32_to_cpu(lun[0]);
-	fw_task_ctx->ustorm_st_context.lun.hi = be32_to_cpu(lun[1]);
+	tmf_pdu_header.opcode = tmf_hdr->opcode;
+	tmf_pdu_header.function = tmf_hdr->flags;
+	tmf_pdu_header.hdr_second_dword = ntoh24(tmf_hdr->dlength);
+	tmf_pdu_header.ref_cmd_sn = be32_to_cpu(tmf_hdr->refcmdsn);
 
-	QEDI_INFO(&qedi->dbg_ctx, QEDI_LOG_SCSI_TM,
-		  "Add TMF to SQ, tmf tid=0x%x, itt=0x%x, cid=0x%x\n",
-		  tid,  mtask->itt, qedi_conn->iscsi_conn_id);
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = 0;
+	task_params.tx_io_size = 0;
+	task_params.rx_io_size = 0;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+	rval = init_initiator_tmf_request_task(&task_params,
+					       &tmf_pdu_header);
+	if (rval)
+		return -1;
 
 	spin_lock(&qedi_conn->list_lock);
 	list_add_tail(&qedi_cmd->io_cmd, &qedi_conn->active_cmd_list);
@@ -1638,7 +1561,6 @@ static int qedi_send_iscsi_tmf(struct qedi_conn *qedi_conn,
 	qedi_conn->active_cmd_count++;
 	spin_unlock(&qedi_conn->list_lock);
 
-	qedi_add_to_sq(qedi_conn, mtask, tid, ptu_invalidate, false);
 	qedi_ring_doorbell(qedi_conn);
 	return 0;
 }
@@ -1689,101 +1611,98 @@ int qedi_iscsi_abort_work(struct qedi_conn *qedi_conn,
 int qedi_send_iscsi_text(struct qedi_conn *qedi_conn,
 			 struct iscsi_task *task)
 {
-	struct qedi_ctx *qedi = qedi_conn->qedi;
+	struct iscsi_text_request_hdr text_request_pdu_header;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
+	struct iscsi_task_params task_params;
 	struct iscsi_task_context *fw_task_ctx;
-	struct iscsi_text_request_hdr *fw_text_request;
-	struct iscsi_cached_sge_ctx *cached_sge;
-	struct iscsi_sge *single_sge;
-	struct qedi_cmd *qedi_cmd;
-	/* For 6.5 hdr iscsi_hdr */
+	struct qedi_ctx *qedi = qedi_conn->qedi;
 	struct iscsi_text *text_hdr;
-	struct iscsi_sge *req_sge;
-	struct iscsi_sge *resp_sge;
-	s16 ptu_invalidate = 0;
+	struct scsi_sge *req_sge = NULL;
+	struct scsi_sge *resp_sge = NULL;
+	struct qedi_cmd *qedi_cmd;
+	struct qedi_endpoint *ep;
 	s16 tid = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
-	req_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
-	resp_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+	req_sge = (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	resp_sge = (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
 	qedi_cmd = (struct qedi_cmd *)task->dd_data;
 	text_hdr = (struct iscsi_text *)task->hdr;
+	ep = qedi_conn->ep;
 
 	tid = qedi_get_task_idx(qedi);
 	if (tid == -1)
 		return -ENOMEM;
 
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
 
 	qedi_cmd->task_id = tid;
 
-	/* Ystorm context */
-	fw_text_request =
-			&fw_task_ctx->ystorm_st_context.pdu_hdr.text_request;
-	fw_text_request->opcode = text_hdr->opcode;
-	fw_text_request->flags_attr = text_hdr->flags;
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&text_request_pdu_header, 0, sizeof(text_request_pdu_header));
+	memset(&tx_sgl_task_params, 0, sizeof(tx_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(rx_sgl_task_params));
+
+	/* Update header info */
+	text_request_pdu_header.opcode = text_hdr->opcode;
+	text_request_pdu_header.flags_attr = text_hdr->flags;
 
 	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
-	fw_text_request->itt = qedi_set_itt(tid, get_itt(task->itt));
-	fw_text_request->ttt = text_hdr->ttt;
-	fw_text_request->cmd_sn = be32_to_cpu(text_hdr->cmdsn);
-	fw_text_request->exp_stat_sn = be32_to_cpu(text_hdr->exp_statsn);
-	fw_text_request->hdr_second_dword = ntoh24(text_hdr->dlength);
-
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						     qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						   qedi->tid_reuse_count[tid]++;
-
-	cached_sge =
-	       &fw_task_ctx->ystorm_st_context.state.sgl_ctx_union.cached_sge;
-	cached_sge->sge.sge_len = req_sge->sge_len;
-	cached_sge->sge.sge_addr.lo = (u32)(qedi_conn->gen_pdu.req_dma_addr);
-	cached_sge->sge.sge_addr.hi =
+	text_request_pdu_header.itt = qedi_set_itt(tid, get_itt(task->itt));
+	text_request_pdu_header.ttt = text_hdr->ttt;
+	text_request_pdu_header.cmd_sn = be32_to_cpu(text_hdr->cmdsn);
+	text_request_pdu_header.exp_stat_sn = be32_to_cpu(text_hdr->exp_statsn);
+	text_request_pdu_header.hdr_second_dword = ntoh24(text_hdr->dlength);
+
+	/* Fill tx AHS and rx buffer */
+	tx_sgl_task_params.sgl =
+			       (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	tx_sgl_task_params.sgl_phys_addr.lo =
+					 (u32)(qedi_conn->gen_pdu.req_dma_addr);
+	tx_sgl_task_params.sgl_phys_addr.hi =
 			      (u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
+	tx_sgl_task_params.total_buffer_size = req_sge->sge_len;
+	tx_sgl_task_params.num_sges = 1;
+
+	rx_sgl_task_params.sgl =
+			      (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+	rx_sgl_task_params.sgl_phys_addr.lo =
+					(u32)(qedi_conn->gen_pdu.resp_dma_addr);
+	rx_sgl_task_params.sgl_phys_addr.hi =
+			     (u32)((u64)qedi_conn->gen_pdu.resp_dma_addr >> 32);
+	rx_sgl_task_params.total_buffer_size = resp_sge->sge_len;
+	rx_sgl_task_params.num_sges = 1;
+
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = 0;
+	task_params.tx_io_size = ntoh24(text_hdr->dlength);
+	task_params.rx_io_size = resp_sge->sge_len;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+	rval = init_initiator_text_request_task(&task_params,
+						&text_request_pdu_header,
+						&tx_sgl_task_params,
+						&rx_sgl_task_params);
+	if (rval)
+		return -1;
 
-	/* Mstorm context */
-	single_sge = &fw_task_ctx->mstorm_st_context.sgl_union.single_sge;
-	fw_task_ctx->mstorm_st_context.task_type = 0x2;
-	fw_task_ctx->mstorm_ag_context.task_cid = (u16)qedi_conn->iscsi_conn_id;
-	single_sge->sge_addr.lo = resp_sge->sge_addr.lo;
-	single_sge->sge_addr.hi = resp_sge->sge_addr.hi;
-	single_sge->sge_len = resp_sge->sge_len;
-
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SINGLE_SGE, 1);
-	SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-		  ISCSI_MFLAGS_SLOW_IO, 0);
-	fw_task_ctx->mstorm_st_context.sgl_size = 1;
-	fw_task_ctx->mstorm_st_context.rem_task_size = resp_sge->sge_len;
-
-	/* Ustorm context */
-	fw_task_ctx->ustorm_ag_context.exp_data_acked =
-						      ntoh24(text_hdr->dlength);
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = resp_sge->sge_len;
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len =
-						      ntoh24(text_hdr->dlength);
-	fw_task_ctx->ustorm_st_context.exp_data_sn =
-					      be32_to_cpu(text_hdr->exp_statsn);
-	fw_task_ctx->ustorm_st_context.cq_rss_number = 0;
-	fw_task_ctx->ustorm_st_context.task_type = 0x2;
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
-
-	/*  Add command in active command list */
 	spin_lock(&qedi_conn->list_lock);
 	list_add_tail(&qedi_cmd->io_cmd, &qedi_conn->active_cmd_list);
 	qedi_cmd->io_cmd_in_list = true;
 	qedi_conn->active_cmd_count++;
 	spin_unlock(&qedi_conn->list_lock);
 
-	qedi_add_to_sq(qedi_conn, task, tid, ptu_invalidate, false);
 	qedi_ring_doorbell(qedi_conn);
-
 	return 0;
 }
 
@@ -1791,58 +1710,62 @@ int qedi_send_iscsi_nopout(struct qedi_conn *qedi_conn,
 			   struct iscsi_task *task,
 			   char *datap, int data_len, int unsol)
 {
+	struct iscsi_nop_out_hdr nop_out_pdu_header;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
+	struct iscsi_task_params task_params;
 	struct qedi_ctx *qedi = qedi_conn->qedi;
 	struct iscsi_task_context *fw_task_ctx;
-	struct iscsi_nop_out_hdr *fw_nop_out;
-	struct qedi_cmd *qedi_cmd;
-	/* For 6.5 hdr iscsi_hdr */
 	struct iscsi_nopout *nopout_hdr;
-	struct iscsi_cached_sge_ctx *cached_sge;
-	struct iscsi_sge *single_sge;
-	struct iscsi_sge *req_sge;
-	struct iscsi_sge *resp_sge;
-	u32 lun[2];
-	s16 ptu_invalidate = 0;
+	struct scsi_sge *req_sge = NULL;
+	struct scsi_sge *resp_sge = NULL;
+	struct qedi_cmd *qedi_cmd;
+	struct qedi_endpoint *ep;
+	u32 scsi_lun[2];
 	s16 tid = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
-	req_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
-	resp_sge = (struct iscsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+	req_sge = (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	resp_sge = (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
 	qedi_cmd = (struct qedi_cmd *)task->dd_data;
 	nopout_hdr = (struct iscsi_nopout *)task->hdr;
+	ep = qedi_conn->ep;
 
 	tid = qedi_get_task_idx(qedi);
-	if (tid == -1) {
-		QEDI_WARN(&qedi->dbg_ctx, "Invalid tid\n");
+	if (tid == -1)
 		return -ENOMEM;
-	}
-
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
 
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
+
 	qedi_cmd->task_id = tid;
 
-	/* Ystorm context */
-	fw_nop_out = &fw_task_ctx->ystorm_st_context.pdu_hdr.nop_out;
-	SET_FIELD(fw_nop_out->flags_attr, ISCSI_NOP_OUT_HDR_CONST1, 1);
-	SET_FIELD(fw_nop_out->flags_attr, ISCSI_NOP_OUT_HDR_RSRV, 0);
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&nop_out_pdu_header, 0, sizeof(nop_out_pdu_header));
+	memset(&tx_sgl_task_params, 0, sizeof(tx_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(rx_sgl_task_params));
+
+	/* Update header info */
+	nop_out_pdu_header.opcode = nopout_hdr->opcode;
+	SET_FIELD(nop_out_pdu_header.flags_attr, ISCSI_NOP_OUT_HDR_CONST1, 1);
+	SET_FIELD(nop_out_pdu_header.flags_attr, ISCSI_NOP_OUT_HDR_RSRV, 0);
 
-	memcpy(lun, &nopout_hdr->lun, sizeof(struct scsi_lun));
-	fw_nop_out->lun.lo = be32_to_cpu(lun[0]);
-	fw_nop_out->lun.hi = be32_to_cpu(lun[1]);
+	memcpy(scsi_lun, &nopout_hdr->lun, sizeof(struct scsi_lun));
+	nop_out_pdu_header.lun.lo = be32_to_cpu(scsi_lun[0]);
+	nop_out_pdu_header.lun.hi = be32_to_cpu(scsi_lun[1]);
+	nop_out_pdu_header.cmd_sn = be32_to_cpu(nopout_hdr->cmdsn);
+	nop_out_pdu_header.exp_stat_sn = be32_to_cpu(nopout_hdr->exp_statsn);
 
 	qedi_update_itt_map(qedi, tid, task->itt, qedi_cmd);
 
 	if (nopout_hdr->ttt != ISCSI_TTT_ALL_ONES) {
-		fw_nop_out->itt = be32_to_cpu(nopout_hdr->itt);
-		fw_nop_out->ttt = be32_to_cpu(nopout_hdr->ttt);
-		fw_task_ctx->ystorm_st_context.state.buffer_offset[0] = 0;
-		fw_task_ctx->ystorm_st_context.state.local_comp = 1;
-		SET_FIELD(fw_task_ctx->ustorm_st_context.flags,
-			  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 1);
+		nop_out_pdu_header.itt = be32_to_cpu(nopout_hdr->itt);
+		nop_out_pdu_header.ttt = be32_to_cpu(nopout_hdr->ttt);
 	} else {
-		fw_nop_out->itt = qedi_set_itt(tid, get_itt(task->itt));
-		fw_nop_out->ttt = ISCSI_TTT_ALL_ONES;
-		fw_task_ctx->ystorm_st_context.state.buffer_offset[0] = 0;
+		nop_out_pdu_header.itt = qedi_set_itt(tid, get_itt(task->itt));
+		nop_out_pdu_header.ttt = ISCSI_TTT_ALL_ONES;
 
 		spin_lock(&qedi_conn->list_lock);
 		list_add_tail(&qedi_cmd->io_cmd, &qedi_conn->active_cmd_list);
@@ -1851,53 +1774,46 @@ int qedi_send_iscsi_nopout(struct qedi_conn *qedi_conn,
 		spin_unlock(&qedi_conn->list_lock);
 	}
 
-	fw_nop_out->opcode = ISCSI_OPCODE_NOP_OUT;
-	fw_nop_out->cmd_sn = be32_to_cpu(nopout_hdr->cmdsn);
-	fw_nop_out->exp_stat_sn = be32_to_cpu(nopout_hdr->exp_statsn);
-
-	cached_sge =
-	       &fw_task_ctx->ystorm_st_context.state.sgl_ctx_union.cached_sge;
-	cached_sge->sge.sge_len = req_sge->sge_len;
-	cached_sge->sge.sge_addr.lo = (u32)(qedi_conn->gen_pdu.req_dma_addr);
-	cached_sge->sge.sge_addr.hi =
-			(u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
-
-	/* Mstorm context */
-	fw_task_ctx->mstorm_st_context.task_type = ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->mstorm_ag_context.task_cid = (u16)qedi_conn->iscsi_conn_id;
-
-	single_sge = &fw_task_ctx->mstorm_st_context.sgl_union.single_sge;
-	single_sge->sge_addr.lo = resp_sge->sge_addr.lo;
-	single_sge->sge_addr.hi = resp_sge->sge_addr.hi;
-	single_sge->sge_len = resp_sge->sge_len;
-	fw_task_ctx->mstorm_st_context.rem_task_size = resp_sge->sge_len;
-
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						qedi->tid_reuse_count[tid]++;
-	/* Ustorm context */
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = resp_sge->sge_len;
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len = data_len;
-	fw_task_ctx->ustorm_st_context.exp_data_sn = 0;
-	fw_task_ctx->ustorm_st_context.task_type =  ISCSI_TASK_TYPE_MIDPATH;
-	fw_task_ctx->ustorm_st_context.cq_rss_number = 0;
-
-	SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-		  ISCSI_REG1_NUM_FAST_SGES, 0);
-
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
-
-	fw_task_ctx->ustorm_st_context.lun.lo = be32_to_cpu(lun[0]);
-	fw_task_ctx->ustorm_st_context.lun.hi = be32_to_cpu(lun[1]);
-
-	qedi_add_to_sq(qedi_conn, task, tid, ptu_invalidate, false);
+	/* Fill tx AHS and rx buffer */
+	if (data_len) {
+		tx_sgl_task_params.sgl =
+			       (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+		tx_sgl_task_params.sgl_phys_addr.lo =
+					 (u32)(qedi_conn->gen_pdu.req_dma_addr);
+		tx_sgl_task_params.sgl_phys_addr.hi =
+			      (u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
+		tx_sgl_task_params.total_buffer_size = data_len;
+		tx_sgl_task_params.num_sges = 1;
+
+		rx_sgl_task_params.sgl =
+			      (struct scsi_sge *)qedi_conn->gen_pdu.resp_bd_tbl;
+		rx_sgl_task_params.sgl_phys_addr.lo =
+					(u32)(qedi_conn->gen_pdu.resp_dma_addr);
+		rx_sgl_task_params.sgl_phys_addr.hi =
+			     (u32)((u64)qedi_conn->gen_pdu.resp_dma_addr >> 32);
+		rx_sgl_task_params.total_buffer_size = resp_sge->sge_len;
+		rx_sgl_task_params.num_sges = 1;
+	}
+
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = 0;
+	task_params.tx_io_size = data_len;
+	task_params.rx_io_size = resp_sge->sge_len;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+	rval = init_initiator_nop_out_task(&task_params,
+					   &nop_out_pdu_header,
+					   &tx_sgl_task_params,
+					   &rx_sgl_task_params);
+	if (rval)
+		return -1;
+
 	qedi_ring_doorbell(qedi_conn);
 	return 0;
 }
@@ -1905,7 +1821,7 @@ int qedi_send_iscsi_nopout(struct qedi_conn *qedi_conn,
 static int qedi_split_bd(struct qedi_cmd *cmd, u64 addr, int sg_len,
 			 int bd_index)
 {
-	struct iscsi_sge *bd = cmd->io_tbl.sge_tbl;
+	struct scsi_sge *bd = cmd->io_tbl.sge_tbl;
 	int frag_size, sg_frags;
 
 	sg_frags = 0;
@@ -1938,7 +1854,7 @@ static int qedi_split_bd(struct qedi_cmd *cmd, u64 addr, int sg_len,
 static int qedi_map_scsi_sg(struct qedi_ctx *qedi, struct qedi_cmd *cmd)
 {
 	struct scsi_cmnd *sc = cmd->scsi_cmd;
-	struct iscsi_sge *bd = cmd->io_tbl.sge_tbl;
+	struct scsi_sge *bd = cmd->io_tbl.sge_tbl;
 	struct scatterlist *sg;
 	int byte_count = 0;
 	int bd_count = 0;
@@ -2040,7 +1956,7 @@ static void qedi_iscsi_map_sg_list(struct qedi_cmd *cmd)
 		if (bd_count == 0)
 			return;
 	} else {
-		struct iscsi_sge *bd = cmd->io_tbl.sge_tbl;
+		struct scsi_sge *bd = cmd->io_tbl.sge_tbl;
 
 		bd[0].sge_addr.lo = 0;
 		bd[0].sge_addr.hi = 0;
@@ -2136,244 +2052,182 @@ int qedi_iscsi_send_ioreq(struct iscsi_task *task)
 	struct qedi_conn *qedi_conn = conn->dd_data;
 	struct qedi_cmd *cmd = task->dd_data;
 	struct scsi_cmnd *sc = task->sc;
+	struct iscsi_cmd_hdr cmd_pdu_header;
+	struct scsi_sgl_task_params tx_sgl_task_params;
+	struct scsi_sgl_task_params rx_sgl_task_params;
+	struct scsi_sgl_task_params *prx_sgl = NULL;
+	struct scsi_sgl_task_params *ptx_sgl = NULL;
+	struct iscsi_task_params task_params;
+	struct iscsi_conn_params conn_params;
+	struct scsi_initiator_cmd_params cmd_params;
 	struct iscsi_task_context *fw_task_ctx;
-	struct iscsi_cached_sge_ctx *cached_sge;
-	struct iscsi_phys_sgl_ctx *phys_sgl;
-	struct iscsi_virt_sgl_ctx *virt_sgl;
-	struct ystorm_iscsi_task_st_ctx *yst_cxt;
-	struct mstorm_iscsi_task_st_ctx *mst_cxt;
-	struct iscsi_sgl *sgl_struct;
-	struct iscsi_sge *single_sge;
+	struct iscsi_cls_conn *cls_conn;
 	struct iscsi_scsi_req *hdr = (struct iscsi_scsi_req *)task->hdr;
-	struct iscsi_sge *bd = cmd->io_tbl.sge_tbl;
-	enum iscsi_task_type task_type;
-	struct iscsi_cmd_hdr *fw_cmd;
-	u32 lun[2];
-	u32 exp_data;
-	u16 cq_idx = smp_processor_id() % qedi->num_queues;
-	s16 ptu_invalidate = 0;
+	enum iscsi_task_type task_type = MAX_ISCSI_TASK_TYPE;
+	struct qedi_endpoint *ep;
+	u32 scsi_lun[2];
 	s16 tid = 0;
-	u8 num_fast_sgs;
+	u16 sq_idx = 0;
+	u16 cq_idx;
+	int rval = 0;
 
-	tid = qedi_get_task_idx(qedi);
-	if (tid == -1)
-		return -ENOMEM;
+	ep = qedi_conn->ep;
+	cls_conn = qedi_conn->cls_conn;
+	conn = cls_conn->dd_data;
 
 	qedi_iscsi_map_sg_list(cmd);
+	int_to_scsilun(sc->device->lun, (struct scsi_lun *)scsi_lun);
 
-	int_to_scsilun(sc->device->lun, (struct scsi_lun *)lun);
-	fw_task_ctx = qedi_get_task_mem(&qedi->tasks, tid);
+	tid = qedi_get_task_idx(qedi);
+	if (tid == -1)
+		return -ENOMEM;
 
+	fw_task_ctx =
+	     (struct iscsi_task_context *)qedi_get_task_mem(&qedi->tasks, tid);
 	memset(fw_task_ctx, 0, sizeof(struct iscsi_task_context));
-	cmd->task_id = tid;
 
-	/* Ystorm context */
-	fw_cmd = &fw_task_ctx->ystorm_st_context.pdu_hdr.cmd;
-	SET_FIELD(fw_cmd->flags_attr, ISCSI_CMD_HDR_ATTR, ISCSI_ATTR_SIMPLE);
+	cmd->task_id = tid;
 
+	memset(&task_params, 0, sizeof(task_params));
+	memset(&cmd_pdu_header, 0, sizeof(cmd_pdu_header));
+	memset(&tx_sgl_task_params, 0, sizeof(tx_sgl_task_params));
+	memset(&rx_sgl_task_params, 0, sizeof(rx_sgl_task_params));
+	memset(&conn_params, 0, sizeof(conn_params));
+	memset(&cmd_params, 0, sizeof(cmd_params));
+
+	cq_idx = smp_processor_id() % qedi->num_queues;
+	/* Update header info */
+	SET_FIELD(cmd_pdu_header.flags_attr, ISCSI_CMD_HDR_ATTR,
+		  ISCSI_ATTR_SIMPLE);
 	if (sc->sc_data_direction == DMA_TO_DEVICE) {
-		if (conn->session->initial_r2t_en) {
-			exp_data = min((conn->session->imm_data_en *
-					conn->max_xmit_dlength),
-				       conn->session->first_burst);
-			exp_data = min(exp_data, scsi_bufflen(sc));
-			fw_task_ctx->ustorm_ag_context.exp_data_acked =
-							  cpu_to_le32(exp_data);
-		} else {
-			fw_task_ctx->ustorm_ag_context.exp_data_acked =
-			      min(conn->session->first_burst, scsi_bufflen(sc));
-		}
-
-		SET_FIELD(fw_cmd->flags_attr, ISCSI_CMD_HDR_WRITE, 1);
+		SET_FIELD(cmd_pdu_header.flags_attr,
+			  ISCSI_CMD_HDR_WRITE, 1);
 		task_type = ISCSI_TASK_TYPE_INITIATOR_WRITE;
 	} else {
-		if (scsi_bufflen(sc))
-			SET_FIELD(fw_cmd->flags_attr, ISCSI_CMD_HDR_READ, 1);
+		SET_FIELD(cmd_pdu_header.flags_attr,
+			  ISCSI_CMD_HDR_READ, 1);
 		task_type = ISCSI_TASK_TYPE_INITIATOR_READ;
 	}
 
-	fw_cmd->lun.lo = be32_to_cpu(lun[0]);
-	fw_cmd->lun.hi = be32_to_cpu(lun[1]);
+	cmd_pdu_header.lun.lo = be32_to_cpu(scsi_lun[0]);
+	cmd_pdu_header.lun.hi = be32_to_cpu(scsi_lun[1]);
 
 	qedi_update_itt_map(qedi, tid, task->itt, cmd);
-	fw_cmd->itt = qedi_set_itt(tid, get_itt(task->itt));
-	fw_cmd->expected_transfer_length = scsi_bufflen(sc);
-	fw_cmd->cmd_sn = be32_to_cpu(hdr->cmdsn);
-	fw_cmd->opcode = hdr->opcode;
-	qedi_cpy_scsi_cdb(sc, (u32 *)fw_cmd->cdb);
-
-	/* Mstorm context */
-	fw_task_ctx->mstorm_st_context.sense_db.lo = (u32)cmd->sense_buffer_dma;
-	fw_task_ctx->mstorm_st_context.sense_db.hi =
-					(u32)((u64)cmd->sense_buffer_dma >> 32);
-	fw_task_ctx->mstorm_ag_context.task_cid = qedi_conn->iscsi_conn_id;
-	fw_task_ctx->mstorm_st_context.task_type = task_type;
-
-	if (qedi->tid_reuse_count[tid] == QEDI_MAX_TASK_NUM) {
-		ptu_invalidate = 1;
-		qedi->tid_reuse_count[tid] = 0;
-	}
-	fw_task_ctx->ystorm_st_context.state.reuse_count =
-						     qedi->tid_reuse_count[tid];
-	fw_task_ctx->mstorm_st_context.reuse_count =
-						   qedi->tid_reuse_count[tid]++;
-
-	/* Ustorm context */
-	fw_task_ctx->ustorm_st_context.rem_rcv_len = scsi_bufflen(sc);
-	fw_task_ctx->ustorm_st_context.exp_data_transfer_len = scsi_bufflen(sc);
-	fw_task_ctx->ustorm_st_context.exp_data_sn =
-						   be32_to_cpu(hdr->exp_statsn);
-	fw_task_ctx->ustorm_st_context.task_type = task_type;
-	fw_task_ctx->ustorm_st_context.cq_rss_number = cq_idx;
-	fw_task_ctx->ustorm_ag_context.icid = (u16)qedi_conn->iscsi_conn_id;
-
-	SET_FIELD(fw_task_ctx->ustorm_ag_context.flags1,
-		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
-	SET_FIELD(fw_task_ctx->ustorm_st_context.flags,
-		  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 0);
-
-	num_fast_sgs = (cmd->io_tbl.sge_valid ?
-			min((u16)QEDI_FAST_SGE_COUNT,
-			    (u16)cmd->io_tbl.sge_valid) : 0);
-	SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-		  ISCSI_REG1_NUM_FAST_SGES, num_fast_sgs);
-
-	fw_task_ctx->ustorm_st_context.lun.lo = be32_to_cpu(lun[0]);
-	fw_task_ctx->ustorm_st_context.lun.hi = be32_to_cpu(lun[1]);
-
-	QEDI_INFO(&qedi->dbg_ctx, QEDI_LOG_IO, "Total sge count [%d]\n",
-		  cmd->io_tbl.sge_valid);
-
-	yst_cxt = &fw_task_ctx->ystorm_st_context;
-	mst_cxt = &fw_task_ctx->mstorm_st_context;
-	/* Tx path */
+	cmd_pdu_header.itt = qedi_set_itt(tid, get_itt(task->itt));
+	cmd_pdu_header.expected_transfer_length = cpu_to_be32(hdr->data_length);
+	cmd_pdu_header.hdr_second_dword = ntoh24(hdr->dlength);
+	cmd_pdu_header.cmd_sn = be32_to_cpu(hdr->cmdsn);
+	cmd_pdu_header.opcode = hdr->opcode;
+	qedi_cpy_scsi_cdb(sc, (u32 *)cmd_pdu_header.cdb);
+
+	/* Fill tx AHS and rx buffer */
 	if (task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE) {
-		/* not considering  superIO or FastIO */
-		if (cmd->io_tbl.sge_valid == 1) {
-			cached_sge = &yst_cxt->state.sgl_ctx_union.cached_sge;
-			cached_sge->sge.sge_addr.lo = bd[0].sge_addr.lo;
-			cached_sge->sge.sge_addr.hi = bd[0].sge_addr.hi;
-			cached_sge->sge.sge_len = bd[0].sge_len;
-			qedi->cached_sgls++;
-		} else if ((cmd->io_tbl.sge_valid != 1) && cmd->use_slowpath) {
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SLOW_IO, 1);
-			SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-				  ISCSI_REG1_NUM_FAST_SGES, 0);
-			phys_sgl = &yst_cxt->state.sgl_ctx_union.phys_sgl;
-			phys_sgl->sgl_base.lo = (u32)(cmd->io_tbl.sge_tbl_dma);
-			phys_sgl->sgl_base.hi =
-				     (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32);
-			phys_sgl->sgl_size = cmd->io_tbl.sge_valid;
-			qedi->slow_sgls++;
-		} else if ((cmd->io_tbl.sge_valid != 1) && !cmd->use_slowpath) {
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SLOW_IO, 0);
-			SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-				  ISCSI_REG1_NUM_FAST_SGES,
-				  min((u16)QEDI_FAST_SGE_COUNT,
-				      (u16)cmd->io_tbl.sge_valid));
-			virt_sgl = &yst_cxt->state.sgl_ctx_union.virt_sgl;
-			virt_sgl->sgl_base.lo = (u32)(cmd->io_tbl.sge_tbl_dma);
-			virt_sgl->sgl_base.hi =
+		tx_sgl_task_params.sgl = cmd->io_tbl.sge_tbl;
+		tx_sgl_task_params.sgl_phys_addr.lo =
+						 (u32)(cmd->io_tbl.sge_tbl_dma);
+		tx_sgl_task_params.sgl_phys_addr.hi =
 				      (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32);
-			virt_sgl->sgl_initial_offset =
-				 (u32)bd[0].sge_addr.lo & (QEDI_PAGE_SIZE - 1);
-			qedi->fast_sgls++;
-		}
-		fw_task_ctx->mstorm_st_context.sgl_size = cmd->io_tbl.sge_valid;
-		fw_task_ctx->mstorm_st_context.rem_task_size = scsi_bufflen(sc);
-	} else {
-	/* Rx path */
-		if (cmd->io_tbl.sge_valid == 1) {
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SLOW_IO, 0);
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SINGLE_SGE, 1);
-			single_sge = &mst_cxt->sgl_union.single_sge;
-			single_sge->sge_addr.lo = bd[0].sge_addr.lo;
-			single_sge->sge_addr.hi = bd[0].sge_addr.hi;
-			single_sge->sge_len = bd[0].sge_len;
-			qedi->cached_sgls++;
-		} else if ((cmd->io_tbl.sge_valid != 1) && cmd->use_slowpath) {
-			sgl_struct = &mst_cxt->sgl_union.sgl_struct;
-			sgl_struct->sgl_addr.lo =
-						(u32)(cmd->io_tbl.sge_tbl_dma);
-			sgl_struct->sgl_addr.hi =
-				     (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32);
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SLOW_IO, 1);
-			SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-				  ISCSI_REG1_NUM_FAST_SGES, 0);
-			sgl_struct->updated_sge_size = 0;
-			sgl_struct->updated_sge_offset = 0;
-			qedi->slow_sgls++;
-		} else if ((cmd->io_tbl.sge_valid != 1) && !cmd->use_slowpath) {
-			sgl_struct = &mst_cxt->sgl_union.sgl_struct;
-			sgl_struct->sgl_addr.lo =
-						(u32)(cmd->io_tbl.sge_tbl_dma);
-			sgl_struct->sgl_addr.hi =
-				     (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32);
-			sgl_struct->byte_offset =
-				(u32)bd[0].sge_addr.lo & (QEDI_PAGE_SIZE - 1);
-			SET_FIELD(fw_task_ctx->mstorm_st_context.flags.mflags,
-				  ISCSI_MFLAGS_SLOW_IO, 0);
-			SET_FIELD(fw_task_ctx->ustorm_st_context.reg1.reg1_map,
-				  ISCSI_REG1_NUM_FAST_SGES, 0);
-			sgl_struct->updated_sge_size = 0;
-			sgl_struct->updated_sge_offset = 0;
-			qedi->fast_sgls++;
-		}
-		fw_task_ctx->mstorm_st_context.sgl_size = cmd->io_tbl.sge_valid;
-		fw_task_ctx->mstorm_st_context.rem_task_size = scsi_bufflen(sc);
-	}
-
-	if (cmd->io_tbl.sge_valid == 1)
-		/* Singel-SGL */
-		qedi->use_cached_sge = true;
-	else {
+		tx_sgl_task_params.total_buffer_size = scsi_bufflen(sc);
+		tx_sgl_task_params.num_sges = cmd->io_tbl.sge_valid;
 		if (cmd->use_slowpath)
-			qedi->use_slow_sge = true;
-		else
-			qedi->use_fast_sge = true;
-	}
+			tx_sgl_task_params.small_mid_sge = true;
+	} else if (task_type == ISCSI_TASK_TYPE_INITIATOR_READ) {
+		rx_sgl_task_params.sgl = cmd->io_tbl.sge_tbl;
+		rx_sgl_task_params.sgl_phys_addr.lo =
+						 (u32)(cmd->io_tbl.sge_tbl_dma);
+		rx_sgl_task_params.sgl_phys_addr.hi =
+				      (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32);
+		rx_sgl_task_params.total_buffer_size = scsi_bufflen(sc);
+		rx_sgl_task_params.num_sges = cmd->io_tbl.sge_valid;
+	}
+
+	/* Add conn param */
+	conn_params.first_burst_length = conn->session->first_burst;
+	conn_params.max_send_pdu_length = conn->max_xmit_dlength;
+	conn_params.max_burst_length = conn->session->max_burst;
+	if (conn->session->initial_r2t_en)
+		conn_params.initial_r2t = true;
+	if (conn->session->imm_data_en)
+		conn_params.immediate_data = true;
+
+	/* Add cmd params */
+	cmd_params.sense_data_buffer_phys_addr.lo = (u32)cmd->sense_buffer_dma;
+	cmd_params.sense_data_buffer_phys_addr.hi =
+					(u32)((u64)cmd->sense_buffer_dma >> 32);
+	/* Fill fw input params */
+	task_params.context = fw_task_ctx;
+	task_params.conn_icid = (u16)qedi_conn->iscsi_conn_id;
+	task_params.itid = tid;
+	task_params.cq_rss_number = cq_idx;
+	if (task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE)
+		task_params.tx_io_size = scsi_bufflen(sc);
+	else if (task_type == ISCSI_TASK_TYPE_INITIATOR_READ)
+		task_params.rx_io_size = scsi_bufflen(sc);
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+	task_params.sqe = &ep->sq[sq_idx];
+
 	QEDI_INFO(&qedi->dbg_ctx, QEDI_LOG_IO,
-		  "%s: %s-SGL: num_sges=0x%x first-sge-lo=0x%x first-sge-hi=0x%x",
+		  "%s: %s-SGL: sg_len=0x%x num_sges=0x%x first-sge-lo=0x%x first-sge-hi=0x%x\n",
 		  (task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE) ?
 		  "Write " : "Read ", (cmd->io_tbl.sge_valid == 1) ?
 		  "Single" : (cmd->use_slowpath ? "SLOW" : "FAST"),
-		  (u16)cmd->io_tbl.sge_valid, (u32)(cmd->io_tbl.sge_tbl_dma),
+		  (u16)cmd->io_tbl.sge_valid, scsi_bufflen(sc),
+		  (u32)(cmd->io_tbl.sge_tbl_dma),
 		  (u32)((u64)cmd->io_tbl.sge_tbl_dma >> 32));
 
-	/*  Add command in active command list */
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+
+	if (task_params.tx_io_size != 0)
+		ptx_sgl = &tx_sgl_task_params;
+	if (task_params.rx_io_size != 0)
+		prx_sgl = &rx_sgl_task_params;
+
+	rval = init_initiator_rw_iscsi_task(&task_params, &conn_params,
+					    &cmd_params, &cmd_pdu_header,
+					    ptx_sgl, prx_sgl,
+					    NULL);
+	if (rval)
+		return -1;
+
 	spin_lock(&qedi_conn->list_lock);
 	list_add_tail(&cmd->io_cmd, &qedi_conn->active_cmd_list);
 	cmd->io_cmd_in_list = true;
 	qedi_conn->active_cmd_count++;
 	spin_unlock(&qedi_conn->list_lock);
 
-	qedi_add_to_sq(qedi_conn, task, tid, ptu_invalidate, false);
 	qedi_ring_doorbell(qedi_conn);
-	if (qedi_io_tracing)
-		qedi_trace_io(qedi, task, tid, QEDI_IO_TRACE_REQ);
-
 	return 0;
 }
 
 int qedi_iscsi_cleanup_task(struct iscsi_task *task, bool mark_cmd_node_deleted)
 {
+	struct iscsi_task_params task_params;
+	struct qedi_endpoint *ep;
 	struct iscsi_conn *conn = task->conn;
 	struct qedi_conn *qedi_conn = conn->dd_data;
 	struct qedi_cmd *cmd = task->dd_data;
-	s16 ptu_invalidate = 0;
+	u16 sq_idx = 0;
+	int rval = 0;
 
 	QEDI_INFO(&qedi_conn->qedi->dbg_ctx, QEDI_LOG_SCSI_TM,
 		  "issue cleanup tid=0x%x itt=0x%x task_state=%d cmd_state=0%x cid=0x%x\n",
 		  cmd->task_id, get_itt(task->itt), task->state,
 		  cmd->state, qedi_conn->iscsi_conn_id);
 
-	qedi_add_to_sq(qedi_conn, task, cmd->task_id, ptu_invalidate, true);
-	qedi_ring_doorbell(qedi_conn);
+	memset(&task_params, 0, sizeof(task_params));
+	ep = qedi_conn->ep;
+
+	sq_idx = qedi_get_wqe_idx(qedi_conn);
+
+	task_params.sqe = &ep->sq[sq_idx];
+	memset(task_params.sqe, 0, sizeof(struct iscsi_wqe));
+	task_params.itid = cmd->task_id;
 
+	rval = init_cleanup_task(&task_params);
+	if (rval)
+		return rval;
+
+	qedi_ring_doorbell(qedi_conn);
 	return 0;
 }
diff --git a/drivers/scsi/qedi/qedi_fw_api.c b/drivers/scsi/qedi/qedi_fw_api.c
new file mode 100644
index 000000000000..fd354d4e03eb
--- /dev/null
+++ b/drivers/scsi/qedi/qedi_fw_api.c
@@ -0,0 +1,781 @@
+/* QLogic iSCSI Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+
+#include <linux/types.h>
+#include <asm/byteorder.h>
+#include "qedi_hsi.h"
+#include <linux/qed/qed_if.h>
+
+#include "qedi_fw_iscsi.h"
+#include "qedi_fw_scsi.h"
+
+#define SCSI_NUM_SGES_IN_CACHE 0x4
+
+static bool scsi_is_slow_sgl(u16 num_sges, bool small_mid_sge)
+{
+	return (num_sges > SCSI_NUM_SGES_SLOW_SGL_THR && small_mid_sge);
+}
+
+static
+void init_scsi_sgl_context(struct scsi_sgl_params *ctx_sgl_params,
+			   struct scsi_cached_sges *ctx_data_desc,
+			   struct scsi_sgl_task_params *sgl_task_params)
+{
+	u8 sge_index;
+	u8 num_sges;
+	u32 val;
+
+	num_sges = (sgl_task_params->num_sges > SCSI_NUM_SGES_IN_CACHE) ?
+			     SCSI_NUM_SGES_IN_CACHE : sgl_task_params->num_sges;
+
+	/* sgl params */
+	val = cpu_to_le32(sgl_task_params->sgl_phys_addr.lo);
+	ctx_sgl_params->sgl_addr.lo = val;
+	val = cpu_to_le32(sgl_task_params->sgl_phys_addr.hi);
+	ctx_sgl_params->sgl_addr.hi = val;
+	val = cpu_to_le32(sgl_task_params->total_buffer_size);
+	ctx_sgl_params->sgl_total_length = val;
+	ctx_sgl_params->sgl_num_sges = cpu_to_le16(sgl_task_params->num_sges);
+
+	for (sge_index = 0; sge_index < num_sges; sge_index++) {
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_addr.lo);
+		ctx_data_desc->sge[sge_index].sge_addr.lo = val;
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_addr.hi);
+		ctx_data_desc->sge[sge_index].sge_addr.hi = val;
+		val = cpu_to_le32(sgl_task_params->sgl[sge_index].sge_len);
+		ctx_data_desc->sge[sge_index].sge_len = val;
+	}
+}
+
+static u32 calc_rw_task_size(struct iscsi_task_params *task_params,
+			     enum iscsi_task_type task_type,
+			     struct scsi_sgl_task_params *sgl_task_params,
+			     struct scsi_dif_task_params *dif_task_params)
+{
+	u32 io_size;
+
+	if (task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE ||
+	    task_type == ISCSI_TASK_TYPE_TARGET_READ)
+		io_size = task_params->tx_io_size;
+	else
+		io_size = task_params->rx_io_size;
+
+	if (!io_size)
+		return 0;
+
+	if (!dif_task_params)
+		return io_size;
+
+	return !dif_task_params->dif_on_network ?
+	       io_size : sgl_task_params->total_buffer_size;
+}
+
+static void
+init_dif_context_flags(struct iscsi_dif_flags *ctx_dif_flags,
+		       struct scsi_dif_task_params *dif_task_params)
+{
+	if (!dif_task_params)
+		return;
+
+	SET_FIELD(ctx_dif_flags->flags, ISCSI_DIF_FLAGS_PROT_INTERVAL_SIZE_LOG,
+		  dif_task_params->dif_block_size_log);
+	SET_FIELD(ctx_dif_flags->flags, ISCSI_DIF_FLAGS_DIF_TO_PEER,
+		  dif_task_params->dif_on_network ? 1 : 0);
+	SET_FIELD(ctx_dif_flags->flags, ISCSI_DIF_FLAGS_HOST_INTERFACE,
+		  dif_task_params->dif_on_host ? 1 : 0);
+}
+
+static void init_sqe(struct iscsi_task_params *task_params,
+		     struct scsi_sgl_task_params *sgl_task_params,
+		     struct scsi_dif_task_params *dif_task_params,
+		     struct iscsi_common_hdr *pdu_header,
+		     struct scsi_initiator_cmd_params *cmd_params,
+		     enum iscsi_task_type task_type,
+		     bool is_cleanup)
+{
+	if (!task_params->sqe)
+		return;
+
+	memset(task_params->sqe, 0, sizeof(*task_params->sqe));
+	task_params->sqe->task_id = cpu_to_le16(task_params->itid);
+	if (is_cleanup) {
+		SET_FIELD(task_params->sqe->flags, ISCSI_WQE_WQE_TYPE,
+			  ISCSI_WQE_TYPE_TASK_CLEANUP);
+		return;
+	}
+
+	switch (task_type) {
+	case ISCSI_TASK_TYPE_INITIATOR_WRITE:
+	{
+		u32 buf_size = 0;
+		u32 num_sges = 0;
+
+		init_dif_context_flags(&task_params->sqe->prot_flags,
+				       dif_task_params);
+
+		SET_FIELD(task_params->sqe->flags, ISCSI_WQE_WQE_TYPE,
+			  ISCSI_WQE_TYPE_NORMAL);
+
+		if (task_params->tx_io_size) {
+			buf_size = calc_rw_task_size(task_params, task_type,
+						     sgl_task_params,
+						     dif_task_params);
+
+		if (scsi_is_slow_sgl(sgl_task_params->num_sges,
+				     sgl_task_params->small_mid_sge))
+			num_sges = ISCSI_WQE_NUM_SGES_SLOWIO;
+		else
+			num_sges = min(sgl_task_params->num_sges,
+				       (u16)SCSI_NUM_SGES_SLOW_SGL_THR);
+	}
+
+	SET_FIELD(task_params->sqe->flags, ISCSI_WQE_NUM_SGES, num_sges);
+	SET_FIELD(task_params->sqe->contlen_cdbsize, ISCSI_WQE_CONT_LEN,
+		  buf_size);
+
+	if (GET_FIELD(pdu_header->hdr_second_dword,
+		      ISCSI_CMD_HDR_TOTAL_AHS_LEN))
+		SET_FIELD(task_params->sqe->contlen_cdbsize, ISCSI_WQE_CDB_SIZE,
+			  cmd_params->extended_cdb_sge.sge_len);
+	}
+		break;
+	case ISCSI_TASK_TYPE_INITIATOR_READ:
+		SET_FIELD(task_params->sqe->flags, ISCSI_WQE_WQE_TYPE,
+			  ISCSI_WQE_TYPE_NORMAL);
+
+		if (GET_FIELD(pdu_header->hdr_second_dword,
+			      ISCSI_CMD_HDR_TOTAL_AHS_LEN))
+			SET_FIELD(task_params->sqe->contlen_cdbsize,
+				  ISCSI_WQE_CDB_SIZE,
+				  cmd_params->extended_cdb_sge.sge_len);
+		break;
+	case ISCSI_TASK_TYPE_LOGIN_RESPONSE:
+	case ISCSI_TASK_TYPE_MIDPATH:
+	{
+		bool advance_statsn = true;
+
+		if (task_type == ISCSI_TASK_TYPE_LOGIN_RESPONSE)
+			SET_FIELD(task_params->sqe->flags, ISCSI_WQE_WQE_TYPE,
+				  ISCSI_WQE_TYPE_LOGIN);
+		else
+			SET_FIELD(task_params->sqe->flags, ISCSI_WQE_WQE_TYPE,
+				  ISCSI_WQE_TYPE_MIDDLE_PATH);
+
+		if (task_type == ISCSI_TASK_TYPE_MIDPATH) {
+			u8 opcode = GET_FIELD(pdu_header->hdr_first_byte,
+					      ISCSI_COMMON_HDR_OPCODE);
+
+			if (opcode != ISCSI_OPCODE_TEXT_RESPONSE &&
+			    (opcode != ISCSI_OPCODE_NOP_IN ||
+			    pdu_header->itt == ISCSI_TTT_ALL_ONES))
+				advance_statsn = false;
+		}
+
+		SET_FIELD(task_params->sqe->flags, ISCSI_WQE_RESPONSE,
+			  advance_statsn ? 1 : 0);
+
+		if (task_params->tx_io_size) {
+			SET_FIELD(task_params->sqe->contlen_cdbsize,
+				  ISCSI_WQE_CONT_LEN, task_params->tx_io_size);
+
+		if (scsi_is_slow_sgl(sgl_task_params->num_sges,
+				     sgl_task_params->small_mid_sge))
+			SET_FIELD(task_params->sqe->flags, ISCSI_WQE_NUM_SGES,
+				  ISCSI_WQE_NUM_SGES_SLOWIO);
+		else
+			SET_FIELD(task_params->sqe->flags, ISCSI_WQE_NUM_SGES,
+				  min(sgl_task_params->num_sges,
+				      (u16)SCSI_NUM_SGES_SLOW_SGL_THR));
+		}
+	}
+		break;
+	default:
+		break;
+	}
+}
+
+static void init_default_iscsi_task(struct iscsi_task_params *task_params,
+				    struct data_hdr *pdu_header,
+				    enum iscsi_task_type task_type)
+{
+	struct iscsi_task_context *context;
+	u16 index;
+	u32 val;
+
+	context = task_params->context;
+	memset(context, 0, sizeof(*context));
+
+	for (index = 0; index <
+	     ARRAY_SIZE(context->ystorm_st_context.pdu_hdr.data.data);
+	     index++) {
+		val = cpu_to_le32(pdu_header->data[index]);
+		context->ystorm_st_context.pdu_hdr.data.data[index] = val;
+	}
+
+	context->mstorm_st_context.task_type = task_type;
+	context->mstorm_ag_context.task_cid =
+					    cpu_to_le16(task_params->conn_icid);
+
+	SET_FIELD(context->ustorm_ag_context.flags1,
+		  USTORM_ISCSI_TASK_AG_CTX_R2T2RECV, 1);
+
+	context->ustorm_st_context.task_type = task_type;
+	context->ustorm_st_context.cq_rss_number = task_params->cq_rss_number;
+	context->ustorm_ag_context.icid = cpu_to_le16(task_params->conn_icid);
+}
+
+static
+void init_initiator_rw_cdb_ystorm_context(struct ystorm_iscsi_task_st_ctx *ystc,
+					  struct scsi_initiator_cmd_params *cmd)
+{
+	union iscsi_task_hdr *ctx_pdu_hdr = &ystc->pdu_hdr;
+	u32 val;
+
+	if (!cmd->extended_cdb_sge.sge_len)
+		return;
+
+	SET_FIELD(ctx_pdu_hdr->ext_cdb_cmd.hdr_second_dword,
+		  ISCSI_EXT_CDB_CMD_HDR_CDB_SIZE,
+		  cmd->extended_cdb_sge.sge_len);
+	val = cpu_to_le32(cmd->extended_cdb_sge.sge_addr.lo);
+	ctx_pdu_hdr->ext_cdb_cmd.cdb_sge.sge_addr.lo = val;
+	val = cpu_to_le32(cmd->extended_cdb_sge.sge_addr.hi);
+	ctx_pdu_hdr->ext_cdb_cmd.cdb_sge.sge_addr.hi = val;
+	val = cpu_to_le32(cmd->extended_cdb_sge.sge_len);
+	ctx_pdu_hdr->ext_cdb_cmd.cdb_sge.sge_len  = val;
+}
+
+static
+void init_ustorm_task_contexts(struct ustorm_iscsi_task_st_ctx *ustorm_st_cxt,
+			       struct ustorm_iscsi_task_ag_ctx *ustorm_ag_cxt,
+			       u32 remaining_recv_len,
+			       u32 expected_data_transfer_len,
+			       u8 num_sges, bool tx_dif_conn_err_en)
+{
+	u32 val;
+
+	ustorm_st_cxt->rem_rcv_len = cpu_to_le32(remaining_recv_len);
+	ustorm_ag_cxt->exp_data_acked = cpu_to_le32(expected_data_transfer_len);
+	val = cpu_to_le32(expected_data_transfer_len);
+	ustorm_st_cxt->exp_data_transfer_len = val;
+	SET_FIELD(ustorm_st_cxt->reg1.reg1_map, ISCSI_REG1_NUM_SGES, num_sges);
+	SET_FIELD(ustorm_ag_cxt->flags2,
+		  USTORM_ISCSI_TASK_AG_CTX_DIF_ERROR_CF_EN,
+		  tx_dif_conn_err_en ? 1 : 0);
+}
+
+static
+void set_rw_exp_data_acked_and_cont_len(struct iscsi_task_context *context,
+					struct iscsi_conn_params  *conn_params,
+					enum iscsi_task_type task_type,
+					u32 task_size,
+					u32 exp_data_transfer_len,
+					u8 total_ahs_length)
+{
+	u32 max_unsolicited_data = 0, val;
+
+	if (total_ahs_length &&
+	    (task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE ||
+	     task_type == ISCSI_TASK_TYPE_INITIATOR_READ))
+		SET_FIELD(context->ustorm_st_context.flags2,
+			  USTORM_ISCSI_TASK_ST_CTX_AHS_EXIST, 1);
+
+	switch (task_type) {
+	case ISCSI_TASK_TYPE_INITIATOR_WRITE:
+		if (!conn_params->initial_r2t)
+			max_unsolicited_data = conn_params->first_burst_length;
+		else if (conn_params->immediate_data)
+			max_unsolicited_data =
+					  min(conn_params->first_burst_length,
+					      conn_params->max_send_pdu_length);
+
+		context->ustorm_ag_context.exp_data_acked =
+				   cpu_to_le32(total_ahs_length == 0 ?
+						min(exp_data_transfer_len,
+						    max_unsolicited_data) :
+						((u32)(total_ahs_length +
+						       ISCSI_AHS_CNTL_SIZE)));
+		break;
+	case ISCSI_TASK_TYPE_TARGET_READ:
+		val = cpu_to_le32(exp_data_transfer_len);
+		context->ustorm_ag_context.exp_data_acked = val;
+		break;
+	case ISCSI_TASK_TYPE_INITIATOR_READ:
+		context->ustorm_ag_context.exp_data_acked =
+					cpu_to_le32((total_ahs_length == 0 ? 0 :
+						     total_ahs_length +
+						     ISCSI_AHS_CNTL_SIZE));
+		break;
+	case ISCSI_TASK_TYPE_TARGET_WRITE:
+		val = cpu_to_le32(task_size);
+		context->ustorm_ag_context.exp_cont_len = val;
+		break;
+	default:
+		break;
+	}
+}
+
+static
+void init_rtdif_task_context(struct rdif_task_context *rdif_context,
+			     struct tdif_task_context *tdif_context,
+			     struct scsi_dif_task_params *dif_task_params,
+			     enum iscsi_task_type task_type)
+{
+	u32 val;
+
+	if (!dif_task_params->dif_on_network || !dif_task_params->dif_on_host)
+		return;
+
+	if (task_type == ISCSI_TASK_TYPE_TARGET_WRITE ||
+	    task_type == ISCSI_TASK_TYPE_INITIATOR_READ) {
+		rdif_context->app_tag_value =
+				  cpu_to_le16(dif_task_params->application_tag);
+		rdif_context->partial_crc_value = cpu_to_le16(0xffff);
+		val = cpu_to_le32(dif_task_params->initial_ref_tag);
+		rdif_context->initial_ref_tag = val;
+		rdif_context->app_tag_mask =
+			     cpu_to_le16(dif_task_params->application_tag_mask);
+		SET_FIELD(rdif_context->flags0, RDIF_TASK_CONTEXT_CRC_SEED,
+			  dif_task_params->crc_seed ? 1 : 0);
+		SET_FIELD(rdif_context->flags0, RDIF_TASK_CONTEXT_HOSTGUARDTYPE,
+			  dif_task_params->host_guard_type);
+		SET_FIELD(rdif_context->flags0,
+			  RDIF_TASK_CONTEXT_PROTECTIONTYPE,
+			  dif_task_params->protection_type);
+		SET_FIELD(rdif_context->flags0,
+			  RDIF_TASK_CONTEXT_INITIALREFTAGVALID, 1);
+		SET_FIELD(rdif_context->flags0,
+			  RDIF_TASK_CONTEXT_KEEPREFTAGCONST,
+			  dif_task_params->keep_ref_tag_const ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_VALIDATEAPPTAG,
+			  (dif_task_params->validate_app_tag &&
+			  dif_task_params->dif_on_network) ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_VALIDATEGUARD,
+			  (dif_task_params->validate_guard &&
+			  dif_task_params->dif_on_network) ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_VALIDATEREFTAG,
+			  (dif_task_params->validate_ref_tag &&
+			  dif_task_params->dif_on_network) ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_HOSTINTERFACE,
+			  dif_task_params->dif_on_host ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_NETWORKINTERFACE,
+			  dif_task_params->dif_on_network ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_FORWARDGUARD,
+			  dif_task_params->forward_guard ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_FORWARDAPPTAG,
+			  dif_task_params->forward_app_tag ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_FORWARDREFTAG,
+			  dif_task_params->forward_ref_tag ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_FORWARDAPPTAGWITHMASK,
+			  dif_task_params->forward_app_tag_with_mask ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_FORWARDREFTAGWITHMASK,
+			  dif_task_params->forward_ref_tag_with_mask ? 1 : 0);
+		SET_FIELD(rdif_context->flags1,
+			  RDIF_TASK_CONTEXT_INTERVALSIZE,
+			  dif_task_params->dif_block_size_log - 9);
+		SET_FIELD(rdif_context->state,
+			  RDIF_TASK_CONTEXT_REFTAGMASK,
+			  dif_task_params->ref_tag_mask);
+		SET_FIELD(rdif_context->state, RDIF_TASK_CONTEXT_IGNOREAPPTAG,
+			  dif_task_params->ignore_app_tag);
+	}
+
+	if (task_type == ISCSI_TASK_TYPE_TARGET_READ ||
+	    task_type == ISCSI_TASK_TYPE_INITIATOR_WRITE) {
+		tdif_context->app_tag_value =
+				  cpu_to_le16(dif_task_params->application_tag);
+		tdif_context->partial_crc_valueB =
+		       cpu_to_le16(dif_task_params->crc_seed ? 0xffff : 0x0000);
+		tdif_context->partial_crc_value_a =
+		       cpu_to_le16(dif_task_params->crc_seed ? 0xffff : 0x0000);
+		SET_FIELD(tdif_context->flags0, TDIF_TASK_CONTEXT_CRC_SEED,
+			  dif_task_params->crc_seed ? 1 : 0);
+
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_SETERRORWITHEOP,
+			  dif_task_params->tx_dif_conn_err_en ? 1 : 0);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_FORWARDGUARD,
+			  dif_task_params->forward_guard   ? 1 : 0);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_FORWARDAPPTAG,
+			  dif_task_params->forward_app_tag ? 1 : 0);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_FORWARDREFTAG,
+			  dif_task_params->forward_ref_tag ? 1 : 0);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_INTERVALSIZE,
+			  dif_task_params->dif_block_size_log - 9);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_HOSTINTERFACE,
+			  dif_task_params->dif_on_host    ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_NETWORKINTERFACE,
+			  dif_task_params->dif_on_network ? 1 : 0);
+		val = cpu_to_le32(dif_task_params->initial_ref_tag);
+		tdif_context->initial_ref_tag = val;
+		tdif_context->app_tag_mask =
+			     cpu_to_le16(dif_task_params->application_tag_mask);
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_HOSTGUARDTYPE,
+			  dif_task_params->host_guard_type);
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_PROTECTIONTYPE,
+			  dif_task_params->protection_type);
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_INITIALREFTAGVALID,
+			  dif_task_params->initial_ref_tag_is_valid ? 1 : 0);
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_KEEPREFTAGCONST,
+			  dif_task_params->keep_ref_tag_const ? 1 : 0);
+		SET_FIELD(tdif_context->flags1, TDIF_TASK_CONTEXT_VALIDATEGUARD,
+			  (dif_task_params->validate_guard &&
+			   dif_task_params->dif_on_host) ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_VALIDATEAPPTAG,
+			  (dif_task_params->validate_app_tag &&
+			  dif_task_params->dif_on_host) ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_VALIDATEREFTAG,
+			  (dif_task_params->validate_ref_tag &&
+			   dif_task_params->dif_on_host) ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_FORWARDAPPTAGWITHMASK,
+			  dif_task_params->forward_app_tag_with_mask ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_FORWARDREFTAGWITHMASK,
+			  dif_task_params->forward_ref_tag_with_mask ? 1 : 0);
+		SET_FIELD(tdif_context->flags1,
+			  TDIF_TASK_CONTEXT_REFTAGMASK,
+			  dif_task_params->ref_tag_mask);
+		SET_FIELD(tdif_context->flags0,
+			  TDIF_TASK_CONTEXT_IGNOREAPPTAG,
+			  dif_task_params->ignore_app_tag ? 1 : 0);
+	}
+}
+
+static void set_local_completion_context(struct iscsi_task_context *context)
+{
+	SET_FIELD(context->ystorm_st_context.state.flags,
+		  YSTORM_ISCSI_TASK_STATE_LOCAL_COMP, 1);
+	SET_FIELD(context->ustorm_st_context.flags,
+		  USTORM_ISCSI_TASK_ST_CTX_LOCAL_COMP, 1);
+}
+
+static int init_rw_iscsi_task(struct iscsi_task_params *task_params,
+			      enum iscsi_task_type task_type,
+			      struct iscsi_conn_params *conn_params,
+			      struct iscsi_common_hdr *pdu_header,
+			      struct scsi_sgl_task_params *sgl_task_params,
+			      struct scsi_initiator_cmd_params *cmd_params,
+			      struct scsi_dif_task_params *dif_task_params)
+{
+	u32 exp_data_transfer_len = conn_params->max_burst_length;
+	struct iscsi_task_context *cxt;
+	bool slow_io = false;
+	u32 task_size, val;
+	u8 num_sges = 0;
+
+	task_size = calc_rw_task_size(task_params, task_type, sgl_task_params,
+				      dif_task_params);
+
+	init_default_iscsi_task(task_params, (struct data_hdr *)pdu_header,
+				task_type);
+
+	cxt = task_params->context;
+
+	val = cpu_to_le32(task_size);
+	cxt->ystorm_st_context.pdu_hdr.cmd.expected_transfer_length = val;
+	init_initiator_rw_cdb_ystorm_context(&cxt->ystorm_st_context,
+					     cmd_params);
+	val = cpu_to_le32(cmd_params->sense_data_buffer_phys_addr.lo);
+	cxt->mstorm_st_context.sense_db.lo = val;
+
+	val = cpu_to_le32(cmd_params->sense_data_buffer_phys_addr.hi);
+	cxt->mstorm_st_context.sense_db.hi = val;
+
+	if (task_params->tx_io_size) {
+		init_dif_context_flags(&cxt->ystorm_st_context.state.dif_flags,
+				       dif_task_params);
+		init_scsi_sgl_context(&cxt->ystorm_st_context.state.sgl_params,
+				      &cxt->ystorm_st_context.state.data_desc,
+				      sgl_task_params);
+
+		slow_io = scsi_is_slow_sgl(sgl_task_params->num_sges,
+					   sgl_task_params->small_mid_sge);
+
+		num_sges = !slow_io ? min_t(u16, sgl_task_params->num_sges,
+					    (u16)SCSI_NUM_SGES_SLOW_SGL_THR) :
+				      ISCSI_WQE_NUM_SGES_SLOWIO;
+
+		if (slow_io) {
+			SET_FIELD(cxt->ystorm_st_context.state.flags,
+				  YSTORM_ISCSI_TASK_STATE_SLOW_IO, 1);
+		}
+	} else if (task_params->rx_io_size) {
+		init_dif_context_flags(&cxt->mstorm_st_context.dif_flags,
+				       dif_task_params);
+		init_scsi_sgl_context(&cxt->mstorm_st_context.sgl_params,
+				      &cxt->mstorm_st_context.data_desc,
+				      sgl_task_params);
+		num_sges = !scsi_is_slow_sgl(sgl_task_params->num_sges,
+				sgl_task_params->small_mid_sge) ?
+				min_t(u16, sgl_task_params->num_sges,
+				      (u16)SCSI_NUM_SGES_SLOW_SGL_THR) :
+				ISCSI_WQE_NUM_SGES_SLOWIO;
+		cxt->mstorm_st_context.rem_task_size = cpu_to_le32(task_size);
+	}
+
+	if (exp_data_transfer_len > task_size  ||
+	    task_type != ISCSI_TASK_TYPE_TARGET_WRITE)
+		exp_data_transfer_len = task_size;
+
+	init_ustorm_task_contexts(&task_params->context->ustorm_st_context,
+				  &task_params->context->ustorm_ag_context,
+				  task_size, exp_data_transfer_len, num_sges,
+				  dif_task_params ?
+				  dif_task_params->tx_dif_conn_err_en : false);
+
+	set_rw_exp_data_acked_and_cont_len(task_params->context, conn_params,
+					   task_type, task_size,
+					   exp_data_transfer_len,
+					GET_FIELD(pdu_header->hdr_second_dword,
+						  ISCSI_CMD_HDR_TOTAL_AHS_LEN));
+
+	if (dif_task_params)
+		init_rtdif_task_context(&task_params->context->rdif_context,
+					&task_params->context->tdif_context,
+					dif_task_params, task_type);
+
+	init_sqe(task_params, sgl_task_params, dif_task_params, pdu_header,
+		 cmd_params, task_type, false);
+
+	return 0;
+}
+
+int init_initiator_rw_iscsi_task(struct iscsi_task_params *task_params,
+				 struct iscsi_conn_params *conn_params,
+				 struct scsi_initiator_cmd_params *cmd_params,
+				 struct iscsi_cmd_hdr *cmd_header,
+				 struct scsi_sgl_task_params *tx_sgl_params,
+				 struct scsi_sgl_task_params *rx_sgl_params,
+				 struct scsi_dif_task_params *dif_task_params)
+{
+	if (GET_FIELD(cmd_header->flags_attr, ISCSI_CMD_HDR_WRITE))
+		return init_rw_iscsi_task(task_params,
+					  ISCSI_TASK_TYPE_INITIATOR_WRITE,
+					  conn_params,
+					  (struct iscsi_common_hdr *)cmd_header,
+					  tx_sgl_params, cmd_params,
+					  dif_task_params);
+	else if (GET_FIELD(cmd_header->flags_attr, ISCSI_CMD_HDR_READ))
+		return init_rw_iscsi_task(task_params,
+					  ISCSI_TASK_TYPE_INITIATOR_READ,
+					  conn_params,
+					  (struct iscsi_common_hdr *)cmd_header,
+					  rx_sgl_params, cmd_params,
+					  dif_task_params);
+	else
+		return -1;
+}
+
+int init_initiator_login_request_task(struct iscsi_task_params *task_params,
+				      struct iscsi_login_req_hdr  *login_header,
+				      struct scsi_sgl_task_params *tx_params,
+				      struct scsi_sgl_task_params *rx_params)
+{
+	struct iscsi_task_context *cxt;
+
+	cxt = task_params->context;
+
+	init_default_iscsi_task(task_params,
+				(struct data_hdr *)login_header,
+				ISCSI_TASK_TYPE_MIDPATH);
+
+	init_ustorm_task_contexts(&cxt->ustorm_st_context,
+				  &cxt->ustorm_ag_context,
+				  task_params->rx_io_size ?
+				  rx_params->total_buffer_size : 0,
+				  task_params->tx_io_size ?
+				  tx_params->total_buffer_size : 0, 0,
+				  0);
+
+	if (task_params->tx_io_size)
+		init_scsi_sgl_context(&cxt->ystorm_st_context.state.sgl_params,
+				      &cxt->ystorm_st_context.state.data_desc,
+				      tx_params);
+
+	if (task_params->rx_io_size)
+		init_scsi_sgl_context(&cxt->mstorm_st_context.sgl_params,
+				      &cxt->mstorm_st_context.data_desc,
+				      rx_params);
+
+	cxt->mstorm_st_context.rem_task_size =
+			cpu_to_le32(task_params->rx_io_size ?
+				    rx_params->total_buffer_size : 0);
+
+	init_sqe(task_params, tx_params, NULL,
+		 (struct iscsi_common_hdr *)login_header, NULL,
+		 ISCSI_TASK_TYPE_MIDPATH, false);
+
+	return 0;
+}
+
+int init_initiator_nop_out_task(struct iscsi_task_params *task_params,
+				struct iscsi_nop_out_hdr *nop_out_pdu_header,
+				struct scsi_sgl_task_params *tx_sgl_task_params,
+				struct scsi_sgl_task_params *rx_sgl_task_params)
+{
+	struct iscsi_task_context *cxt;
+
+	cxt = task_params->context;
+
+	init_default_iscsi_task(task_params,
+				(struct data_hdr *)nop_out_pdu_header,
+				ISCSI_TASK_TYPE_MIDPATH);
+
+	if (nop_out_pdu_header->itt == ISCSI_ITT_ALL_ONES)
+		set_local_completion_context(task_params->context);
+
+	if (task_params->tx_io_size)
+		init_scsi_sgl_context(&cxt->ystorm_st_context.state.sgl_params,
+				      &cxt->ystorm_st_context.state.data_desc,
+				      tx_sgl_task_params);
+
+	if (task_params->rx_io_size)
+		init_scsi_sgl_context(&cxt->mstorm_st_context.sgl_params,
+				      &cxt->mstorm_st_context.data_desc,
+				      rx_sgl_task_params);
+
+	init_ustorm_task_contexts(&cxt->ustorm_st_context,
+				  &cxt->ustorm_ag_context,
+				  task_params->rx_io_size ?
+				  rx_sgl_task_params->total_buffer_size : 0,
+				  task_params->tx_io_size ?
+				  tx_sgl_task_params->total_buffer_size : 0,
+				  0, 0);
+
+	cxt->mstorm_st_context.rem_task_size =
+				cpu_to_le32(task_params->rx_io_size ?
+					rx_sgl_task_params->total_buffer_size :
+					0);
+
+	init_sqe(task_params, tx_sgl_task_params, NULL,
+		 (struct iscsi_common_hdr *)nop_out_pdu_header, NULL,
+		 ISCSI_TASK_TYPE_MIDPATH, false);
+
+	return 0;
+}
+
+int init_initiator_logout_request_task(struct iscsi_task_params *task_params,
+				       struct iscsi_logout_req_hdr *logout_hdr,
+				       struct scsi_sgl_task_params *tx_params,
+				       struct scsi_sgl_task_params *rx_params)
+{
+	struct iscsi_task_context *cxt;
+
+	cxt = task_params->context;
+
+	init_default_iscsi_task(task_params,
+				(struct data_hdr *)logout_hdr,
+				ISCSI_TASK_TYPE_MIDPATH);
+
+	if (task_params->tx_io_size)
+		init_scsi_sgl_context(&cxt->ystorm_st_context.state.sgl_params,
+				      &cxt->ystorm_st_context.state.data_desc,
+				      tx_params);
+
+	if (task_params->rx_io_size)
+		init_scsi_sgl_context(&cxt->mstorm_st_context.sgl_params,
+				      &cxt->mstorm_st_context.data_desc,
+				      rx_params);
+
+	init_ustorm_task_contexts(&cxt->ustorm_st_context,
+				  &cxt->ustorm_ag_context,
+				  task_params->rx_io_size ?
+				  rx_params->total_buffer_size : 0,
+				  task_params->tx_io_size ?
+				  tx_params->total_buffer_size : 0,
+				  0, 0);
+
+	cxt->mstorm_st_context.rem_task_size =
+					cpu_to_le32(task_params->rx_io_size ?
+					rx_params->total_buffer_size : 0);
+
+	init_sqe(task_params, tx_params, NULL,
+		 (struct iscsi_common_hdr *)logout_hdr, NULL,
+		 ISCSI_TASK_TYPE_MIDPATH, false);
+
+	return 0;
+}
+
+int init_initiator_tmf_request_task(struct iscsi_task_params *task_params,
+				    struct iscsi_tmf_request_hdr *tmf_header)
+{
+	init_default_iscsi_task(task_params, (struct data_hdr *)tmf_header,
+				ISCSI_TASK_TYPE_MIDPATH);
+
+	init_sqe(task_params, NULL, NULL,
+		 (struct iscsi_common_hdr *)tmf_header, NULL,
+		 ISCSI_TASK_TYPE_MIDPATH, false);
+
+	return 0;
+}
+
+int init_initiator_text_request_task(struct iscsi_task_params *task_params,
+				     struct iscsi_text_request_hdr *text_header,
+				     struct scsi_sgl_task_params *tx_params,
+				     struct scsi_sgl_task_params *rx_params)
+{
+	struct iscsi_task_context *cxt;
+
+	cxt = task_params->context;
+
+	init_default_iscsi_task(task_params,
+				(struct data_hdr *)text_header,
+				ISCSI_TASK_TYPE_MIDPATH);
+
+	if (task_params->tx_io_size)
+		init_scsi_sgl_context(&cxt->ystorm_st_context.state.sgl_params,
+				      &cxt->ystorm_st_context.state.data_desc,
+				      tx_params);
+
+	if (task_params->rx_io_size)
+		init_scsi_sgl_context(&cxt->mstorm_st_context.sgl_params,
+				      &cxt->mstorm_st_context.data_desc,
+				      rx_params);
+
+	cxt->mstorm_st_context.rem_task_size =
+				cpu_to_le32(task_params->rx_io_size ?
+					rx_params->total_buffer_size : 0);
+
+	init_ustorm_task_contexts(&cxt->ustorm_st_context,
+				  &cxt->ustorm_ag_context,
+				  task_params->rx_io_size ?
+				  rx_params->total_buffer_size : 0,
+				  task_params->tx_io_size ?
+				  tx_params->total_buffer_size : 0, 0, 0);
+
+	init_sqe(task_params, tx_params, NULL,
+		 (struct iscsi_common_hdr *)text_header, NULL,
+		 ISCSI_TASK_TYPE_MIDPATH, false);
+
+	return 0;
+}
+
+int init_cleanup_task(struct iscsi_task_params *task_params)
+{
+	init_sqe(task_params, NULL, NULL, NULL, NULL, ISCSI_TASK_TYPE_MIDPATH,
+		 true);
+	return 0;
+}
diff --git a/drivers/scsi/qedi/qedi_fw_iscsi.h b/drivers/scsi/qedi/qedi_fw_iscsi.h
new file mode 100644
index 000000000000..b6f24f91849d
--- /dev/null
+++ b/drivers/scsi/qedi/qedi_fw_iscsi.h
@@ -0,0 +1,117 @@
+/*
+ * QLogic iSCSI Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+
+#ifndef _QEDI_FW_ISCSI_H_
+#define _QEDI_FW_ISCSI_H_
+
+#include "qedi_fw_scsi.h"
+
+struct iscsi_task_params {
+	struct iscsi_task_context *context;
+	struct iscsi_wqe	  *sqe;
+	u32			  tx_io_size;
+	u32			  rx_io_size;
+	u16			  conn_icid;
+	u16			  itid;
+	u8			  cq_rss_number;
+};
+
+struct iscsi_conn_params {
+	u32	first_burst_length;
+	u32	max_send_pdu_length;
+	u32	max_burst_length;
+	bool	initial_r2t;
+	bool	immediate_data;
+};
+
+/* @brief init_initiator_read_iscsi_task - initializes iSCSI Initiator Read
+ * task context.
+ *
+ * @param task_params	  - Pointer to task parameters struct
+ * @param conn_params	  - Connection Parameters
+ * @param cmd_params	  - command specific parameters
+ * @param cmd_pdu_header  - PDU Header Parameters
+ * @param sgl_task_params - Pointer to SGL task params
+ * @param dif_task_params - Pointer to DIF parameters struct
+ */
+int init_initiator_rw_iscsi_task(struct iscsi_task_params *task_params,
+				 struct iscsi_conn_params *conn_params,
+				 struct scsi_initiator_cmd_params *cmd_params,
+				 struct iscsi_cmd_hdr *cmd_pdu_header,
+				 struct scsi_sgl_task_params *tx_sgl_params,
+				 struct scsi_sgl_task_params *rx_sgl_params,
+				 struct scsi_dif_task_params *dif_task_params);
+
+/* @brief init_initiator_login_request_task - initializes iSCSI Initiator Login
+ * Request task context.
+ *
+ * @param task_params		  - Pointer to task parameters struct
+ * @param login_req_pdu_header    - PDU Header Parameters
+ * @param tx_sgl_task_params	  - Pointer to SGL task params
+ * @param rx_sgl_task_params	  - Pointer to SGL task params
+ */
+int init_initiator_login_request_task(struct iscsi_task_params *task_params,
+				      struct iscsi_login_req_hdr *login_header,
+				      struct scsi_sgl_task_params *tx_params,
+				      struct scsi_sgl_task_params *rx_params);
+
+/* @brief init_initiator_nop_out_task - initializes iSCSI Initiator NOP Out
+ * task context.
+ *
+ * @param task_params		- Pointer to task parameters struct
+ * @param nop_out_pdu_header    - PDU Header Parameters
+ * @param tx_sgl_task_params	- Pointer to SGL task params
+ * @param rx_sgl_task_params	- Pointer to SGL task params
+ */
+int init_initiator_nop_out_task(struct iscsi_task_params *task_params,
+				struct iscsi_nop_out_hdr *nop_out_pdu_header,
+				struct scsi_sgl_task_params *tx_sgl_params,
+				struct scsi_sgl_task_params *rx_sgl_params);
+
+/* @brief init_initiator_logout_request_task - initializes iSCSI Initiator
+ * Logout Request task context.
+ *
+ * @param task_params		- Pointer to task parameters struct
+ * @param logout_pdu_header  - PDU Header Parameters
+ * @param tx_sgl_task_params	- Pointer to SGL task params
+ * @param rx_sgl_task_params	- Pointer to SGL task params
+ */
+int init_initiator_logout_request_task(struct iscsi_task_params *task_params,
+				       struct iscsi_logout_req_hdr *logout_hdr,
+				       struct scsi_sgl_task_params *tx_params,
+				       struct scsi_sgl_task_params *rx_params);
+
+/* @brief init_initiator_tmf_request_task - initializes iSCSI Initiator TMF
+ * task context.
+ *
+ * @param task_params	- Pointer to task parameters struct
+ * @param tmf_pdu_header - PDU Header Parameters
+ */
+int init_initiator_tmf_request_task(struct iscsi_task_params *task_params,
+				    struct iscsi_tmf_request_hdr *tmf_header);
+
+/* @brief init_initiator_text_request_task - initializes iSCSI Initiator Text
+ * Request task context.
+ *
+ * @param task_params		     - Pointer to task parameters struct
+ * @param text_request_pdu_header    - PDU Header Parameters
+ * @param tx_sgl_task_params	     - Pointer to Tx SGL task params
+ * @param rx_sgl_task_params	     - Pointer to Rx SGL task params
+ */
+int init_initiator_text_request_task(struct iscsi_task_params *task_params,
+				     struct iscsi_text_request_hdr *text_header,
+				     struct scsi_sgl_task_params *tx_params,
+				     struct scsi_sgl_task_params *rx_params);
+
+/* @brief init_cleanup_task - initializes Clean task (SQE)
+ *
+ * @param task_params - Pointer to task parameters struct
+ */
+int init_cleanup_task(struct iscsi_task_params *task_params);
+#endif
diff --git a/drivers/scsi/qedi/qedi_fw_scsi.h b/drivers/scsi/qedi/qedi_fw_scsi.h
new file mode 100644
index 000000000000..cdaf918f1019
--- /dev/null
+++ b/drivers/scsi/qedi/qedi_fw_scsi.h
@@ -0,0 +1,55 @@
+/*
+ * QLogic iSCSI Offload Driver
+ * Copyright (c) 2016 Cavium Inc.
+ *
+ * This software is available 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.
+ */
+
+#ifndef _QEDI_FW_SCSI_H_
+#define _QEDI_FW_SCSI_H_
+
+#include <linux/types.h>
+#include <asm/byteorder.h>
+#include "qedi_hsi.h"
+#include <linux/qed/qed_if.h>
+
+struct scsi_sgl_task_params {
+	struct scsi_sge	*sgl;
+	struct regpair	sgl_phys_addr;
+	u32		total_buffer_size;
+	u16		num_sges;
+	bool		small_mid_sge;
+};
+
+struct scsi_dif_task_params {
+	u32	initial_ref_tag;
+	bool	initial_ref_tag_is_valid;
+	u16	application_tag;
+	u16	application_tag_mask;
+	u16	dif_block_size_log;
+	bool	dif_on_network;
+	bool	dif_on_host;
+	u8	host_guard_type;
+	u8	protection_type;
+	u8	ref_tag_mask;
+	bool	crc_seed;
+	bool	tx_dif_conn_err_en;
+	bool	ignore_app_tag;
+	bool	keep_ref_tag_const;
+	bool	validate_guard;
+	bool	validate_app_tag;
+	bool	validate_ref_tag;
+	bool	forward_guard;
+	bool	forward_app_tag;
+	bool	forward_ref_tag;
+	bool	forward_app_tag_with_mask;
+	bool	forward_ref_tag_with_mask;
+};
+
+struct scsi_initiator_cmd_params {
+	struct scsi_sge	extended_cdb_sge;
+	struct regpair	sense_data_buffer_phys_addr;
+};
+#endif
diff --git a/drivers/scsi/qedi/qedi_iscsi.c b/drivers/scsi/qedi/qedi_iscsi.c
index b9f79d36142d..d5eff68507e5 100644
--- a/drivers/scsi/qedi/qedi_iscsi.c
+++ b/drivers/scsi/qedi/qedi_iscsi.c
@@ -175,7 +175,7 @@ static void qedi_destroy_cmd_pool(struct qedi_ctx *qedi,
 		if (cmd->io_tbl.sge_tbl)
 			dma_free_coherent(&qedi->pdev->dev,
 					  QEDI_ISCSI_MAX_BDS_PER_CMD *
-					  sizeof(struct iscsi_sge),
+					  sizeof(struct scsi_sge),
 					  cmd->io_tbl.sge_tbl,
 					  cmd->io_tbl.sge_tbl_dma);
 
@@ -191,7 +191,7 @@ static int qedi_alloc_sget(struct qedi_ctx *qedi, struct iscsi_session *session,
 			   struct qedi_cmd *cmd)
 {
 	struct qedi_io_bdt *io = &cmd->io_tbl;
-	struct iscsi_sge *sge;
+	struct scsi_sge *sge;
 
 	io->sge_tbl = dma_alloc_coherent(&qedi->pdev->dev,
 					 QEDI_ISCSI_MAX_BDS_PER_CMD *
@@ -708,22 +708,20 @@ static void qedi_conn_get_stats(struct iscsi_cls_conn *cls_conn,
 
 static void qedi_iscsi_prep_generic_pdu_bd(struct qedi_conn *qedi_conn)
 {
-	struct iscsi_sge *bd_tbl;
+	struct scsi_sge *bd_tbl;
 
-	bd_tbl = (struct iscsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
+	bd_tbl = (struct scsi_sge *)qedi_conn->gen_pdu.req_bd_tbl;
 
 	bd_tbl->sge_addr.hi =
 		(u32)((u64)qedi_conn->gen_pdu.req_dma_addr >> 32);
 	bd_tbl->sge_addr.lo = (u32)qedi_conn->gen_pdu.req_dma_addr;
 	bd_tbl->sge_len = qedi_conn->gen_pdu.req_wr_ptr -
 				qedi_conn->gen_pdu.req_buf;
-	bd_tbl->reserved0 = 0;
-	bd_tbl = (struct iscsi_sge  *)qedi_conn->gen_pdu.resp_bd_tbl;
+	bd_tbl = (struct scsi_sge  *)qedi_conn->gen_pdu.resp_bd_tbl;
 	bd_tbl->sge_addr.hi =
 			(u32)((u64)qedi_conn->gen_pdu.resp_dma_addr >> 32);
 	bd_tbl->sge_addr.lo = (u32)qedi_conn->gen_pdu.resp_dma_addr;
 	bd_tbl->sge_len = ISCSI_DEF_MAX_RECV_SEG_LEN;
-	bd_tbl->reserved0 = 0;
 }
 
 static int qedi_iscsi_send_generic_request(struct iscsi_task *task)
diff --git a/drivers/scsi/qedi/qedi_iscsi.h b/drivers/scsi/qedi/qedi_iscsi.h
index d3c06bbddb4e..3247287cb0e7 100644
--- a/drivers/scsi/qedi/qedi_iscsi.h
+++ b/drivers/scsi/qedi/qedi_iscsi.h
@@ -102,7 +102,7 @@ struct qedi_endpoint {
 #define QEDI_SQ_WQES_MIN	16
 
 struct qedi_io_bdt {
-	struct iscsi_sge *sge_tbl;
+	struct scsi_sge *sge_tbl;
 	dma_addr_t sge_tbl_dma;
 	u16 sge_valid;
 };
diff --git a/drivers/scsi/qedi/qedi_version.h b/drivers/scsi/qedi/qedi_version.h
index 9543a1b139d4..d61e3ac22e67 100644
--- a/drivers/scsi/qedi/qedi_version.h
+++ b/drivers/scsi/qedi/qedi_version.h
@@ -7,8 +7,8 @@
  * this source tree.
  */
 
-#define QEDI_MODULE_VERSION	"8.10.3.0"
+#define QEDI_MODULE_VERSION	"8.10.4.0"
 #define QEDI_DRIVER_MAJOR_VER		8
 #define QEDI_DRIVER_MINOR_VER		10
-#define QEDI_DRIVER_REV_VER		3
+#define QEDI_DRIVER_REV_VER		4
 #define QEDI_DRIVER_ENG_VER		0