Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Pull networking updates from David Millar:
 "Here are some highlights from the 2065 networking commits that
  happened this development cycle:

   1) XDP support for IXGBE (John Fastabend) and thunderx (Sunil Kowuri)

   2) Add a generic XDP driver, so that anyone can test XDP even if they
      lack a networking device whose driver has explicit XDP support
      (me).

   3) Sparc64 now has an eBPF JIT too (me)

   4) Add a BPF program testing framework via BPF_PROG_TEST_RUN (Alexei
      Starovoitov)

   5) Make netfitler network namespace teardown less expensive (Florian
      Westphal)

   6) Add symmetric hashing support to nft_hash (Laura Garcia Liebana)

   7) Implement NAPI and GRO in netvsc driver (Stephen Hemminger)

   8) Support TC flower offload statistics in mlxsw (Arkadi Sharshevsky)

   9) Multiqueue support in stmmac driver (Joao Pinto)

  10) Remove TCP timewait recycling, it never really could possibly work
      well in the real world and timestamp randomization really zaps any
      hint of usability this feature had (Soheil Hassas Yeganeh)

  11) Support level3 vs level4 ECMP route hashing in ipv4 (Nikolay
      Aleksandrov)

  12) Add socket busy poll support to epoll (Sridhar Samudrala)

  13) Netlink extended ACK support (Johannes Berg, Pablo Neira Ayuso,
      and several others)

  14) IPSEC hw offload infrastructure (Steffen Klassert)"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (2065 commits)
  tipc: refactor function tipc_sk_recv_stream()
  tipc: refactor function tipc_sk_recvmsg()
  net: thunderx: Optimize page recycling for XDP
  net: thunderx: Support for XDP header adjustment
  net: thunderx: Add support for XDP_TX
  net: thunderx: Add support for XDP_DROP
  net: thunderx: Add basic XDP support
  net: thunderx: Cleanup receive buffer allocation
  net: thunderx: Optimize CQE_TX handling
  net: thunderx: Optimize RBDR descriptor handling
  net: thunderx: Support for page recycling
  ipx: call ipxitf_put() in ioctl error path
  net: sched: add helpers to handle extended actions
  qed*: Fix issues in the ptp filter config implementation.
  qede: Fix concurrency issue in PTP Tx path processing.
  stmmac: Add support for SIMATIC IOT2000 platform
  net: hns: fix ethtool_get_strings overflow in hns driver
  tcp: fix wraparound issue in tcp_lp
  bpf, arm64: fix jit branch offset related to ldimm64
  bpf, arm64: implement jiting of BPF_XADD
  ...

17 files changed, 2105 insertions, 1070 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..8c65e3b034dc
--- /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 (0xFFFFu)
+
+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_le16(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_le16(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 46debe5034af..1d7f90d0adc1 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;
-
-