xref: /freebsd-12-stable/sys/dev/ocs_fc/ocs_hw.c

/*-
 * Copyright (c) 2017 Broadcom. All rights reserved.
 * The term "Broadcom" refers to Broadcom Limited and/or its subsidiaries.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice,
 *    this list of conditions and the following disclaimer.
 *
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *
 * 3. Neither the name of the copyright holder nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 *
 * $FreeBSD: stable/12/sys/dev/ocs_fc/ocs_hw.c 372539 2022-09-13 05:28:59Z git2svn $
 */

/**
 * @file
 * Defines and implements the Hardware Abstraction Layer (HW).
 * All interaction with the hardware is performed through the HW, which abstracts
 * the details of the underlying SLI-4 implementation.
 */

/**
 * @defgroup devInitShutdown Device Initialization and Shutdown
 * @defgroup domain Domain Functions
 * @defgroup port Port Functions
 * @defgroup node Remote Node Functions
 * @defgroup io IO Functions
 * @defgroup interrupt Interrupt handling
 * @defgroup os OS Required Functions
 */

#include "ocs.h"
#include "ocs_os.h"
#include "ocs_hw.h"
#include "ocs_hw_queues.h"

#define OCS_HW_MQ_DEPTH	128
#define OCS_HW_READ_FCF_SIZE	4096
#define OCS_HW_DEFAULT_AUTO_XFER_RDY_IOS	256
#define OCS_HW_WQ_TIMER_PERIOD_MS	500

/* values used for setting the auto xfer rdy parameters */
#define OCS_HW_AUTO_XFER_RDY_BLK_SIZE_DEFAULT		0 /* 512 bytes */
#define OCS_HW_AUTO_XFER_RDY_REF_TAG_IS_LBA_DEFAULT	TRUE
#define OCS_HW_AUTO_XFER_RDY_APP_TAG_VALID_DEFAULT	FALSE
#define OCS_HW_AUTO_XFER_RDY_APP_TAG_VALUE_DEFAULT	0
#define OCS_HW_REQUE_XRI_REGTAG			65534
/* max command and response buffer lengths -- arbitrary at the moment */
#define OCS_HW_DMTF_CLP_CMD_MAX	256
#define OCS_HW_DMTF_CLP_RSP_MAX	256

/* HW global data */
ocs_hw_global_t hw_global;

static void ocs_hw_queue_hash_add(ocs_queue_hash_t *, uint16_t, uint16_t);
static void ocs_hw_adjust_wqs(ocs_hw_t *hw);
static uint32_t ocs_hw_get_num_chutes(ocs_hw_t *hw);
static int32_t ocs_hw_cb_link(void *, void *);
static int32_t ocs_hw_cb_fip(void *, void *);
static int32_t ocs_hw_command_process(ocs_hw_t *, int32_t, uint8_t *, size_t);
static int32_t ocs_hw_mq_process(ocs_hw_t *, int32_t, sli4_queue_t *);
static int32_t ocs_hw_cb_read_fcf(ocs_hw_t *, int32_t, uint8_t *, void *);
static int32_t ocs_hw_cb_node_attach(ocs_hw_t *, int32_t, uint8_t *, void *);
static int32_t ocs_hw_cb_node_free(ocs_hw_t *, int32_t, uint8_t *, void *);
static int32_t ocs_hw_cb_node_free_all(ocs_hw_t *, int32_t, uint8_t *, void *);
static ocs_hw_rtn_e ocs_hw_setup_io(ocs_hw_t *);
static ocs_hw_rtn_e ocs_hw_init_io(ocs_hw_t *);
static int32_t ocs_hw_flush(ocs_hw_t *);
static int32_t ocs_hw_command_cancel(ocs_hw_t *);
static int32_t ocs_hw_io_cancel(ocs_hw_t *);
static void ocs_hw_io_quarantine(ocs_hw_t *hw, hw_wq_t *wq, ocs_hw_io_t *io);
static void ocs_hw_io_restore_sgl(ocs_hw_t *, ocs_hw_io_t *);
static int32_t ocs_hw_io_ini_sge(ocs_hw_t *, ocs_hw_io_t *, ocs_dma_t *, uint32_t, ocs_dma_t *);
static ocs_hw_rtn_e ocs_hw_firmware_write_lancer(ocs_hw_t *hw, ocs_dma_t *dma, uint32_t size, uint32_t offset, int last, ocs_hw_fw_cb_t cb, void *arg);
static int32_t ocs_hw_cb_fw_write(ocs_hw_t *, int32_t, uint8_t *, void  *);
static int32_t ocs_hw_cb_sfp(ocs_hw_t *, int32_t, uint8_t *, void  *);
static int32_t ocs_hw_cb_temp(ocs_hw_t *, int32_t, uint8_t *, void  *);
static int32_t ocs_hw_cb_link_stat(ocs_hw_t *, int32_t, uint8_t *, void  *);
static int32_t ocs_hw_cb_host_stat(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg);
static void ocs_hw_dmtf_clp_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg);
static int32_t ocs_hw_clp_resp_get_value(ocs_hw_t *hw, const char *keyword, char *value, uint32_t value_len, const char *resp, uint32_t resp_len);
typedef void (*ocs_hw_dmtf_clp_cb_t)(ocs_hw_t *hw, int32_t status, uint32_t result_len, void *arg);
static ocs_hw_rtn_e ocs_hw_exec_dmtf_clp_cmd(ocs_hw_t *hw, ocs_dma_t *dma_cmd, ocs_dma_t *dma_resp, uint32_t opts, ocs_hw_dmtf_clp_cb_t cb, void *arg);
static void ocs_hw_linkcfg_dmtf_clp_cb(ocs_hw_t *hw, int32_t status, uint32_t result_len, void *arg);

static int32_t __ocs_read_topology_cb(ocs_hw_t *, int32_t, uint8_t *, void *);
static ocs_hw_rtn_e ocs_hw_get_linkcfg(ocs_hw_t *, uint32_t, ocs_hw_port_control_cb_t, void *);
static ocs_hw_rtn_e ocs_hw_get_linkcfg_lancer(ocs_hw_t *, uint32_t, ocs_hw_port_control_cb_t, void *);
static ocs_hw_rtn_e ocs_hw_get_linkcfg_skyhawk(ocs_hw_t *, uint32_t, ocs_hw_port_control_cb_t, void *);
static ocs_hw_rtn_e ocs_hw_set_linkcfg(ocs_hw_t *, ocs_hw_linkcfg_e, uint32_t, ocs_hw_port_control_cb_t, void *);
static ocs_hw_rtn_e ocs_hw_set_linkcfg_lancer(ocs_hw_t *, ocs_hw_linkcfg_e, uint32_t, ocs_hw_port_control_cb_t, void *);
static ocs_hw_rtn_e ocs_hw_set_linkcfg_skyhawk(ocs_hw_t *, ocs_hw_linkcfg_e, uint32_t, ocs_hw_port_control_cb_t, void *);
static void ocs_hw_init_linkcfg_cb(int32_t status, uintptr_t value, void *arg);
static ocs_hw_rtn_e ocs_hw_set_eth_license(ocs_hw_t *hw, uint32_t license);
static ocs_hw_rtn_e ocs_hw_set_dif_seed(ocs_hw_t *hw);
static ocs_hw_rtn_e ocs_hw_set_dif_mode(ocs_hw_t *hw);
static void ocs_hw_io_free_internal(void *arg);
static void ocs_hw_io_free_port_owned(void *arg);
static ocs_hw_rtn_e ocs_hw_config_auto_xfer_rdy_t10pi(ocs_hw_t *hw, uint8_t *buf);
static ocs_hw_rtn_e ocs_hw_config_set_fdt_xfer_hint(ocs_hw_t *hw, uint32_t fdt_xfer_hint);
static void ocs_hw_wq_process_abort(void *arg, uint8_t *cqe, int32_t status);
static int32_t ocs_hw_config_mrq(ocs_hw_t *hw, uint8_t, uint16_t, uint16_t);
static ocs_hw_rtn_e ocs_hw_config_watchdog_timer(ocs_hw_t *hw);
static ocs_hw_rtn_e ocs_hw_config_sli_port_health_check(ocs_hw_t *hw, uint8_t query, uint8_t enable);

/* HW domain database operations */
static int32_t ocs_hw_domain_add(ocs_hw_t *, ocs_domain_t *);
static int32_t ocs_hw_domain_del(ocs_hw_t *, ocs_domain_t *);


/* Port state machine */
static void *__ocs_hw_port_alloc_init(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_port_alloc_read_sparm64(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_port_alloc_init_vpi(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_port_done(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_port_free_unreg_vpi(ocs_sm_ctx_t *, ocs_sm_event_t, void *);

/* Domain state machine */
static void *__ocs_hw_domain_init(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_domain_alloc_reg_fcfi(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void * __ocs_hw_domain_alloc_init_vfi(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_domain_free_unreg_vfi(ocs_sm_ctx_t *, ocs_sm_event_t, void *);
static void *__ocs_hw_domain_free_unreg_fcfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data);
static int32_t __ocs_hw_domain_cb(ocs_hw_t *, int32_t, uint8_t *, void *);
static int32_t __ocs_hw_port_cb(ocs_hw_t *, int32_t, uint8_t *, void *);
static int32_t __ocs_hw_port_realloc_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg);

/* BZ 161832 */
static void ocs_hw_check_sec_hio_list(ocs_hw_t *hw);

/* WQE timeouts */
static void target_wqe_timer_cb(void *arg);
static void shutdown_target_wqe_timer(ocs_hw_t *hw);

static inline void
ocs_hw_add_io_timed_wqe(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	if (hw->config.emulate_tgt_wqe_timeout && io->tgt_wqe_timeout) {
		/*
		 * Active WQE list currently only used for
		 * target WQE timeouts.
		 */
		ocs_lock(&hw->io_lock);
			ocs_list_add_tail(&hw->io_timed_wqe, io);
			io->submit_ticks = ocs_get_os_ticks();
		ocs_unlock(&hw->io_lock);
	}
}

static inline void
ocs_hw_remove_io_timed_wqe(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	if (hw->config.emulate_tgt_wqe_timeout) {
		/*
		 * If target wqe timeouts are enabled,
		 * remove from active wqe list.
		 */
		ocs_lock(&hw->io_lock);
			if (ocs_list_on_list(&io->wqe_link)) {
				ocs_list_remove(&hw->io_timed_wqe, io);
			}
		ocs_unlock(&hw->io_lock);
	}
}

static uint8_t ocs_hw_iotype_is_originator(uint16_t io_type)
{
	switch (io_type) {
	case OCS_HW_IO_INITIATOR_READ:
	case OCS_HW_IO_INITIATOR_WRITE:
	case OCS_HW_IO_INITIATOR_NODATA:
	case OCS_HW_FC_CT:
	case OCS_HW_ELS_REQ:
		return 1;
	default:
		return 0;
	}
}

static uint8_t ocs_hw_wcqe_abort_needed(uint16_t status, uint8_t ext, uint8_t xb)
{
	/* if exchange not active, nothing to abort */
	if (!xb) {
		return FALSE;
	}
	if (status == SLI4_FC_WCQE_STATUS_LOCAL_REJECT) {
		switch (ext) {
		/* exceptions where abort is not needed */
		case SLI4_FC_LOCAL_REJECT_INVALID_RPI: /* lancer returns this after unreg_rpi */
		case SLI4_FC_LOCAL_REJECT_ABORT_REQUESTED: /* abort already in progress */
			return FALSE;
		default:
			break;
		}
	}
	return TRUE;
}

/**
 * @brief Determine the number of chutes on the device.
 *
 * @par Description
 * Some devices require queue resources allocated per protocol processor
 * (chute). This function returns the number of chutes on this device.
 *
 * @param hw Hardware context allocated by the caller.
 *
 * @return Returns the number of chutes on the device for protocol.
 */
static uint32_t
ocs_hw_get_num_chutes(ocs_hw_t *hw)
{
	uint32_t num_chutes = 1;

	if (sli_get_is_dual_ulp_capable(&hw->sli) &&
	    sli_get_is_ulp_enabled(&hw->sli, 0) &&
	    sli_get_is_ulp_enabled(&hw->sli, 1)) {
		num_chutes = 2;
	}
	return num_chutes;
}

static ocs_hw_rtn_e
ocs_hw_link_event_init(ocs_hw_t *hw)
{
	if (hw == NULL) {
		ocs_log_err(hw->os, "bad parameter hw=%p\n", hw);
		return OCS_HW_RTN_ERROR;
	}

	hw->link.status = SLI_LINK_STATUS_MAX;
	hw->link.topology = SLI_LINK_TOPO_NONE;
	hw->link.medium = SLI_LINK_MEDIUM_MAX;
	hw->link.speed = 0;
	hw->link.loop_map = NULL;
	hw->link.fc_id = UINT32_MAX;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup devInitShutdown
 * @brief If this is physical port 0, then read the max dump size.
 *
 * @par Description
 * Queries the FW for the maximum dump size
 *
 * @param hw Hardware context allocated by the caller.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static ocs_hw_rtn_e
ocs_hw_read_max_dump_size(ocs_hw_t *hw)
{
	uint8_t	buf[SLI4_BMBX_SIZE];
	uint8_t bus, dev, func;
	int 	rc;

	/* lancer only */
	if ((SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) &&
	    (SLI4_IF_TYPE_LANCER_G7 != sli_get_if_type(&hw->sli))) {
		ocs_log_debug(hw->os, "Function only supported for I/F type 2\n");
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Make sure the FW is new enough to support this command. If the FW
	 * is too old, the FW will UE.
	 */
	if (hw->workaround.disable_dump_loc) {
		ocs_log_test(hw->os, "FW version is too old for this feature\n");
		return OCS_HW_RTN_ERROR;
	}

	/* attempt to detemine the dump size for function 0 only. */
	ocs_get_bus_dev_func(hw->os, &bus, &dev, &func);
	if (func == 0) {
		if (sli_cmd_common_set_dump_location(&hw->sli, buf,
							SLI4_BMBX_SIZE, 1, 0, NULL, 0)) {
			sli4_res_common_set_dump_location_t *rsp =
				(sli4_res_common_set_dump_location_t *)
				(buf + offsetof(sli4_cmd_sli_config_t,
						payload.embed));

			rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_test(hw->os, "set dump location command failed\n");
				return rc;
			} else {
				hw->dump_size = rsp->buffer_length;
				ocs_log_debug(hw->os, "Dump size %x\n", rsp->buffer_length);
			}
		}
	}
	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup devInitShutdown
 * @brief Set up the Hardware Abstraction Layer module.
 *
 * @par Description
 * Calls set up to configure the hardware.
 *
 * @param hw Hardware context allocated by the caller.
 * @param os Device abstraction.
 * @param port_type Protocol type of port, such as FC and NIC.
 *
 * @todo Why is port_type a parameter?
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_setup(ocs_hw_t *hw, ocs_os_handle_t os, sli4_port_type_e port_type)
{
	uint32_t i;
	char prop_buf[32];

	if (hw == NULL) {
		ocs_log_err(os, "bad parameter(s) hw=%p\n", hw);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->hw_setup_called) {
		/* Setup run-time workarounds.
		 * Call for each setup, to allow for hw_war_version
		 */
		ocs_hw_workaround_setup(hw);
		return OCS_HW_RTN_SUCCESS;
	}

	/*
	 * ocs_hw_init() relies on NULL pointers indicating that a structure
	 * needs allocation. If a structure is non-NULL, ocs_hw_init() won't
	 * free/realloc that memory
	 */
	ocs_memset(hw, 0, sizeof(ocs_hw_t));

	hw->hw_setup_called = TRUE;

	hw->os = os;

	ocs_lock_init(hw->os, &hw->cmd_lock, "HW_cmd_lock[%d]", ocs_instance(hw->os));
	ocs_list_init(&hw->cmd_head, ocs_command_ctx_t, link);
	ocs_list_init(&hw->cmd_pending, ocs_command_ctx_t, link);
	hw->cmd_head_count = 0;

	ocs_lock_init(hw->os, &hw->io_lock, "HW_io_lock[%d]", ocs_instance(hw->os));
	ocs_lock_init(hw->os, &hw->io_abort_lock, "HW_io_abort_lock[%d]", ocs_instance(hw->os));

	ocs_atomic_init(&hw->io_alloc_failed_count, 0);

	hw->config.speed = FC_LINK_SPEED_AUTO_16_8_4;
	hw->config.dif_seed = 0;
	hw->config.auto_xfer_rdy_blk_size_chip = OCS_HW_AUTO_XFER_RDY_BLK_SIZE_DEFAULT;
	hw->config.auto_xfer_rdy_ref_tag_is_lba = OCS_HW_AUTO_XFER_RDY_REF_TAG_IS_LBA_DEFAULT;
	hw->config.auto_xfer_rdy_app_tag_valid =  OCS_HW_AUTO_XFER_RDY_APP_TAG_VALID_DEFAULT;
	hw->config.auto_xfer_rdy_app_tag_value = OCS_HW_AUTO_XFER_RDY_APP_TAG_VALUE_DEFAULT;


	if (sli_setup(&hw->sli, hw->os, port_type)) {
		ocs_log_err(hw->os, "SLI setup failed\n");
		return OCS_HW_RTN_ERROR;
	}

	ocs_memset(hw->domains, 0, sizeof(hw->domains));

	ocs_memset(hw->fcf_index_fcfi, 0, sizeof(hw->fcf_index_fcfi));

	ocs_hw_link_event_init(hw);

	sli_callback(&hw->sli, SLI4_CB_LINK, ocs_hw_cb_link, hw);
	sli_callback(&hw->sli, SLI4_CB_FIP, ocs_hw_cb_fip, hw);

	/*
	 * Set all the queue sizes to the maximum allowed. These values may
	 * be changes later by the adjust and workaround functions.
	 */
	for (i = 0; i < ARRAY_SIZE(hw->num_qentries); i++) {
		hw->num_qentries[i] = sli_get_max_qentries(&hw->sli, i);
	}

	/*
	 * The RQ assignment for RQ pair mode.
	 */
	hw->config.rq_default_buffer_size = OCS_HW_RQ_SIZE_PAYLOAD;
	hw->config.n_io = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_XRI);
	if (ocs_get_property("auto_xfer_rdy_xri_cnt", prop_buf, sizeof(prop_buf)) == 0) {
		hw->config.auto_xfer_rdy_xri_cnt = ocs_strtoul(prop_buf, 0, 0);
	}

	/* by default, enable initiator-only auto-ABTS emulation */
	hw->config.i_only_aab = TRUE;

	/* Setup run-time workarounds */
	ocs_hw_workaround_setup(hw);

	/* HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB */
	if (hw->workaround.override_fcfi) {
		hw->first_domain_idx = -1;
	}

	/* Must be done after the workaround setup */
	if ((SLI4_IF_TYPE_LANCER_FC_ETH == sli_get_if_type(&hw->sli)) ||
	    (SLI4_IF_TYPE_LANCER_G7 == sli_get_if_type(&hw->sli))) {

		(void)ocs_hw_read_max_dump_size(hw);
	}

	/* calculate the number of WQs required. */
	ocs_hw_adjust_wqs(hw);

	/* Set the default dif mode */
	if (! sli_is_dif_inline_capable(&hw->sli)) {
		ocs_log_test(hw->os, "not inline capable, setting mode to separate\n");
		hw->config.dif_mode = OCS_HW_DIF_MODE_SEPARATE;
	}
	/* Workaround: BZ 161832 */
	if (hw->workaround.use_dif_sec_xri) {
		ocs_list_init(&hw->sec_hio_wait_list, ocs_hw_io_t, link);
	}

	/*
	 * Figure out the starting and max ULP to spread the WQs across the
	 * ULPs.
	 */
	if (sli_get_is_dual_ulp_capable(&hw->sli)) {
		if (sli_get_is_ulp_enabled(&hw->sli, 0) &&
		    sli_get_is_ulp_enabled(&hw->sli, 1)) {
			hw->ulp_start = 0;
			hw->ulp_max   = 1;
		} else if (sli_get_is_ulp_enabled(&hw->sli, 0)) {
			hw->ulp_start = 0;
			hw->ulp_max   = 0;
		} else {
			hw->ulp_start = 1;
			hw->ulp_max   = 1;
		}
	} else {
		if (sli_get_is_ulp_enabled(&hw->sli, 0)) {
			hw->ulp_start = 0;
			hw->ulp_max   = 0;
		} else {
			hw->ulp_start = 1;
			hw->ulp_max   = 1;
		}
	}
	ocs_log_debug(hw->os, "ulp_start %d, ulp_max %d\n",
		hw->ulp_start, hw->ulp_max);
	hw->config.queue_topology = hw_global.queue_topology_string;

	hw->qtop = ocs_hw_qtop_parse(hw, hw->config.queue_topology);

	hw->config.n_eq = hw->qtop->entry_counts[QTOP_EQ];
	hw->config.n_cq = hw->qtop->entry_counts[QTOP_CQ];
	hw->config.n_rq = hw->qtop->entry_counts[QTOP_RQ];
	hw->config.n_wq = hw->qtop->entry_counts[QTOP_WQ];
	hw->config.n_mq = hw->qtop->entry_counts[QTOP_MQ];

	/* Verify qtop configuration against driver supported configuration */
	if (hw->config.n_rq > OCE_HW_MAX_NUM_MRQ_PAIRS) {
		ocs_log_crit(hw->os, "Max supported MRQ pairs = %d\n",
				OCE_HW_MAX_NUM_MRQ_PAIRS);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->config.n_eq > OCS_HW_MAX_NUM_EQ) {
		ocs_log_crit(hw->os, "Max supported EQs = %d\n",
				OCS_HW_MAX_NUM_EQ);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->config.n_cq > OCS_HW_MAX_NUM_CQ) {
		ocs_log_crit(hw->os, "Max supported CQs = %d\n",
				OCS_HW_MAX_NUM_CQ);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->config.n_wq > OCS_HW_MAX_NUM_WQ) {
		ocs_log_crit(hw->os, "Max supported WQs = %d\n",
				OCS_HW_MAX_NUM_WQ);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->config.n_mq > OCS_HW_MAX_NUM_MQ) {
		ocs_log_crit(hw->os, "Max supported MQs = %d\n",
				OCS_HW_MAX_NUM_MQ);
		return OCS_HW_RTN_ERROR;
	}

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup devInitShutdown
 * @brief Allocate memory structures to prepare for the device operation.
 *
 * @par Description
 * Allocates memory structures needed by the device and prepares the device
 * for operation.
 * @n @n @b Note: This function may be called more than once (for example, at
 * initialization and then after a reset), but the size of the internal resources
 * may not be changed without tearing down the HW (ocs_hw_teardown()).
 *
 * @param hw Hardware context allocated by the caller.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_init(ocs_hw_t *hw)
{
	ocs_hw_rtn_e	rc;
	uint32_t	i = 0;
	uint8_t		buf[SLI4_BMBX_SIZE];
	uint32_t	max_rpi;
	int		rem_count;
	int	        written_size = 0;
	uint32_t	count;
	char		prop_buf[32];
	uint32_t ramdisc_blocksize = 512;
	uint32_t q_count = 0;
	/*
	 * Make sure the command lists are empty. If this is start-of-day,
	 * they'll be empty since they were just initialized in ocs_hw_setup.
	 * If we've just gone through a reset, the command and command pending
	 * lists should have been cleaned up as part of the reset (ocs_hw_reset()).
	 */
	ocs_lock(&hw->cmd_lock);
		if (!ocs_list_empty(&hw->cmd_head)) {
			ocs_log_test(hw->os, "command found on cmd list\n");
			ocs_unlock(&hw->cmd_lock);
			return OCS_HW_RTN_ERROR;
		}
		if (!ocs_list_empty(&hw->cmd_pending)) {
			ocs_log_test(hw->os, "command found on pending list\n");
			ocs_unlock(&hw->cmd_lock);
			return OCS_HW_RTN_ERROR;
		}
	ocs_unlock(&hw->cmd_lock);

	/* Free RQ buffers if prevously allocated */
	ocs_hw_rx_free(hw);

	/*
	 * The IO queues must be initialized here for the reset case. The
	 * ocs_hw_init_io() function will re-add the IOs to the free list.
	 * The cmd_head list should be OK since we free all entries in
	 * ocs_hw_command_cancel() that is called in the ocs_hw_reset().
	 */

	/* If we are in this function due to a reset, there may be stale items
	 * on lists that need to be removed.  Clean them up.
	 */
	rem_count=0;
	if (ocs_list_valid(&hw->io_wait_free)) {
		while ((!ocs_list_empty(&hw->io_wait_free))) {
			rem_count++;
			ocs_list_remove_head(&hw->io_wait_free);
		}
		if (rem_count > 0) {
			ocs_log_debug(hw->os, "removed %d items from io_wait_free list\n", rem_count);
		}
	}
	rem_count=0;
	if (ocs_list_valid(&hw->io_inuse)) {
		while ((!ocs_list_empty(&hw->io_inuse))) {
			rem_count++;
			ocs_list_remove_head(&hw->io_inuse);
		}
		if (rem_count > 0) {
			ocs_log_debug(hw->os, "removed %d items from io_inuse list\n", rem_count);
		}
	}
	rem_count=0;
	if (ocs_list_valid(&hw->io_free)) {
		while ((!ocs_list_empty(&hw->io_free))) {
			rem_count++;
			ocs_list_remove_head(&hw->io_free);
		}
		if (rem_count > 0) {
			ocs_log_debug(hw->os, "removed %d items from io_free list\n", rem_count);
		}
	}
	if (ocs_list_valid(&hw->io_port_owned)) {
		while ((!ocs_list_empty(&hw->io_port_owned))) {
			ocs_list_remove_head(&hw->io_port_owned);
		}
	}
	ocs_list_init(&hw->io_inuse, ocs_hw_io_t, link);
	ocs_list_init(&hw->io_free, ocs_hw_io_t, link);
	ocs_list_init(&hw->io_port_owned, ocs_hw_io_t, link);
	ocs_list_init(&hw->io_wait_free, ocs_hw_io_t, link);
	ocs_list_init(&hw->io_timed_wqe, ocs_hw_io_t, wqe_link);
	ocs_list_init(&hw->io_port_dnrx, ocs_hw_io_t, dnrx_link);

	/* If MRQ not required, Make sure we dont request feature. */
	if (hw->config.n_rq == 1) {
		hw->sli.config.features.flag.mrqp = FALSE;
	}

	if (sli_init(&hw->sli)) {
		ocs_log_err(hw->os, "SLI failed to initialize\n");
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Enable the auto xfer rdy feature if requested.
	 */
	hw->auto_xfer_rdy_enabled = FALSE;
	if (sli_get_auto_xfer_rdy_capable(&hw->sli) &&
	    hw->config.auto_xfer_rdy_size > 0) {
		if (hw->config.esoc){
			if (ocs_get_property("ramdisc_blocksize", prop_buf, sizeof(prop_buf)) == 0) {
				ramdisc_blocksize = ocs_strtoul(prop_buf, 0, 0);
			}
			written_size = sli_cmd_config_auto_xfer_rdy_hp(&hw->sli, buf, SLI4_BMBX_SIZE, hw->config.auto_xfer_rdy_size, 1, ramdisc_blocksize);
		} else {
			written_size = sli_cmd_config_auto_xfer_rdy(&hw->sli, buf, SLI4_BMBX_SIZE, hw->config.auto_xfer_rdy_size);
		}
		if (written_size) {
			rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_err(hw->os, "config auto xfer rdy failed\n");
				return rc;
			}
		}
		hw->auto_xfer_rdy_enabled = TRUE;

		if (hw->config.auto_xfer_rdy_t10_enable) {
			rc = ocs_hw_config_auto_xfer_rdy_t10pi(hw, buf);
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_err(hw->os, "set parameters auto xfer rdy T10 PI failed\n");
				return rc;
			}
		}
	}

	if(hw->sliport_healthcheck) {
		rc = ocs_hw_config_sli_port_health_check(hw, 0, 1);
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "Enabling Sliport Health check failed \n");
			return rc;
		}
	}

	/*
	 * Set FDT transfer hint, only works on Lancer
	 */
	if ((hw->sli.if_type == SLI4_IF_TYPE_LANCER_FC_ETH) && (OCS_HW_FDT_XFER_HINT != 0)) {
		/*
		 * Non-fatal error. In particular, we can disregard failure to set OCS_HW_FDT_XFER_HINT on
		 * devices with legacy firmware that do not support OCS_HW_FDT_XFER_HINT feature.
		 */
		ocs_hw_config_set_fdt_xfer_hint(hw, OCS_HW_FDT_XFER_HINT);
	}

	/*
	 * Verify that we have not exceeded any queue sizes
	 */
	q_count = MIN(sli_get_max_queue(&hw->sli, SLI_QTYPE_EQ),
					OCS_HW_MAX_NUM_EQ);
	if (hw->config.n_eq > q_count) {
		ocs_log_err(hw->os, "requested %d EQ but %d allowed\n",
			    hw->config.n_eq, q_count);
		return OCS_HW_RTN_ERROR;
	}

	q_count = MIN(sli_get_max_queue(&hw->sli, SLI_QTYPE_CQ),
					OCS_HW_MAX_NUM_CQ);
	if (hw->config.n_cq > q_count) {
		ocs_log_err(hw->os, "requested %d CQ but %d allowed\n",
			    hw->config.n_cq, q_count);
		return OCS_HW_RTN_ERROR;
	}

	q_count = MIN(sli_get_max_queue(&hw->sli, SLI_QTYPE_MQ),
					OCS_HW_MAX_NUM_MQ);
	if (hw->config.n_mq > q_count) {
		ocs_log_err(hw->os, "requested %d MQ but %d allowed\n",
			    hw->config.n_mq, q_count);
		return OCS_HW_RTN_ERROR;
	}

	q_count = MIN(sli_get_max_queue(&hw->sli, SLI_QTYPE_RQ),
					OCS_HW_MAX_NUM_RQ);
	if (hw->config.n_rq > q_count) {
		ocs_log_err(hw->os, "requested %d RQ but %d allowed\n",
			    hw->config.n_rq, q_count);
		return OCS_HW_RTN_ERROR;
	}

	q_count = MIN(sli_get_max_queue(&hw->sli, SLI_QTYPE_WQ),
					OCS_HW_MAX_NUM_WQ);
	if (hw->config.n_wq > q_count) {
		ocs_log_err(hw->os, "requested %d WQ but %d allowed\n",
			    hw->config.n_wq, q_count);
		return OCS_HW_RTN_ERROR;
	}

	/* zero the hashes */
	ocs_memset(hw->cq_hash, 0, sizeof(hw->cq_hash));
	ocs_log_debug(hw->os, "Max CQs %d, hash size = %d\n",
			OCS_HW_MAX_NUM_CQ, OCS_HW_Q_HASH_SIZE);

	ocs_memset(hw->rq_hash, 0, sizeof(hw->rq_hash));
	ocs_log_debug(hw->os, "Max RQs %d, hash size = %d\n",
			OCS_HW_MAX_NUM_RQ, OCS_HW_Q_HASH_SIZE);

	ocs_memset(hw->wq_hash, 0, sizeof(hw->wq_hash));
	ocs_log_debug(hw->os, "Max WQs %d, hash size = %d\n",
			OCS_HW_MAX_NUM_WQ, OCS_HW_Q_HASH_SIZE);


	rc = ocs_hw_init_queues(hw, hw->qtop);
	if (rc != OCS_HW_RTN_SUCCESS) {
		return rc;
	}

	max_rpi = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_RPI);
	i = sli_fc_get_rpi_requirements(&hw->sli, max_rpi);
	if (i) {
		ocs_dma_t payload_memory;

		rc = OCS_HW_RTN_ERROR;

		if (hw->rnode_mem.size) {
			ocs_dma_free(hw->os, &hw->rnode_mem);
		}

		if (ocs_dma_alloc(hw->os, &hw->rnode_mem, i, 4096)) {
			ocs_log_err(hw->os, "remote node memory allocation fail\n");
			return OCS_HW_RTN_NO_MEMORY;
		}

		payload_memory.size = 0;
		if (sli_cmd_fcoe_post_hdr_templates(&hw->sli, buf, SLI4_BMBX_SIZE,
					&hw->rnode_mem, UINT16_MAX, &payload_memory)) {
			rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);

			if (payload_memory.size != 0) {
				/* The command was non-embedded - need to free the dma buffer */
				ocs_dma_free(hw->os, &payload_memory);
			}
		}

		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "header template registration failed\n");
			return rc;
		}
	}

	/* Allocate and post RQ buffers */
	rc = ocs_hw_rx_allocate(hw);
	if (rc) {
		ocs_log_err(hw->os, "rx_allocate failed\n");
		return rc;
	}

	/* Populate hw->seq_free_list */
	if (hw->seq_pool == NULL) {
		uint32_t count = 0;
		uint32_t i;

		/* Sum up the total number of RQ entries, to use to allocate the sequence object pool */
		for (i = 0; i < hw->hw_rq_count; i++) {
			count += hw->hw_rq[i]->entry_count;
		}

		hw->seq_pool = ocs_array_alloc(hw->os, sizeof(ocs_hw_sequence_t), count);
		if (hw->seq_pool == NULL) {
			ocs_log_err(hw->os, "malloc seq_pool failed\n");
			return OCS_HW_RTN_NO_MEMORY;
		}
	}

	if(ocs_hw_rx_post(hw)) {
		ocs_log_err(hw->os, "WARNING - error posting RQ buffers\n");
	}

	/* Allocate rpi_ref if not previously allocated */
	if (hw->rpi_ref == NULL) {
		hw->rpi_ref = ocs_malloc(hw->os, max_rpi * sizeof(*hw->rpi_ref),
					  OCS_M_ZERO | OCS_M_NOWAIT);
		if (hw->rpi_ref == NULL) {
			ocs_log_err(hw->os, "rpi_ref allocation failure (%d)\n", i);
			return OCS_HW_RTN_NO_MEMORY;
		}
	}

	for (i = 0; i < max_rpi; i ++) {
		ocs_atomic_init(&hw->rpi_ref[i].rpi_count, 0);
		ocs_atomic_init(&hw->rpi_ref[i].rpi_attached, 0);
	}

	ocs_memset(hw->domains, 0, sizeof(hw->domains));

	/* HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB */
	if (hw->workaround.override_fcfi) {
		hw->first_domain_idx = -1;
	}

	ocs_memset(hw->fcf_index_fcfi, 0, sizeof(hw->fcf_index_fcfi));

	/* Register a FCFI to allow unsolicited frames to be routed to the driver */
	if (sli_get_medium(&hw->sli) == SLI_LINK_MEDIUM_FC) {

		if (hw->hw_mrq_count) {
			ocs_log_debug(hw->os, "using REG_FCFI MRQ\n");

			rc = ocs_hw_config_mrq(hw, SLI4_CMD_REG_FCFI_SET_FCFI_MODE, 0, 0);
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_err(hw->os, "REG_FCFI_MRQ FCFI registration failed\n");
				return rc;
			}

			rc = ocs_hw_config_mrq(hw, SLI4_CMD_REG_FCFI_SET_MRQ_MODE, 0, 0);
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_err(hw->os, "REG_FCFI_MRQ MRQ registration failed\n");
				return rc;
			}
		} else {
			sli4_cmd_rq_cfg_t rq_cfg[SLI4_CMD_REG_FCFI_NUM_RQ_CFG];

			ocs_log_debug(hw->os, "using REG_FCFI standard\n");

			/* Set the filter match/mask values from hw's filter_def values */
			for (i = 0; i < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; i++) {
				rq_cfg[i].rq_id = 0xffff;
				rq_cfg[i].r_ctl_mask =	(uint8_t)  hw->config.filter_def[i];
				rq_cfg[i].r_ctl_match = (uint8_t) (hw->config.filter_def[i] >> 8);
				rq_cfg[i].type_mask =	(uint8_t) (hw->config.filter_def[i] >> 16);
				rq_cfg[i].type_match =	(uint8_t) (hw->config.filter_def[i] >> 24);
			}

			/*
			 * Update the rq_id's of the FCF configuration (don't update more than the number
			 * of rq_cfg elements)
			 */
			for (i = 0; i < OCS_MIN(hw->hw_rq_count, SLI4_CMD_REG_FCFI_NUM_RQ_CFG); i++) {
				hw_rq_t *rq = hw->hw_rq[i];
				uint32_t j;
				for (j = 0; j < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; j++) {
					uint32_t mask = (rq->filter_mask != 0) ? rq->filter_mask : 1;
					if (mask & (1U << j)) {
						rq_cfg[j].rq_id = rq->hdr->id;
						ocs_log_debug(hw->os, "REG_FCFI: filter[%d] %08X -> RQ[%d] id=%d\n",
							j, hw->config.filter_def[j], i, rq->hdr->id);
					}
				}
			}

			rc = OCS_HW_RTN_ERROR;

			if (sli_cmd_reg_fcfi(&hw->sli, buf, SLI4_BMBX_SIZE, 0, rq_cfg, 0)) {
				rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
			}

			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_err(hw->os, "FCFI registration failed\n");
				return rc;
			}
			hw->fcf_indicator = ((sli4_cmd_reg_fcfi_t *)buf)->fcfi;
		}

	}

	/*
	 * Allocate the WQ request tag pool, if not previously allocated (the request tag value is 16 bits,
	 * thus the pool allocation size of 64k)
	 */
	rc = ocs_hw_reqtag_init(hw);
	if (rc) {
		ocs_log_err(hw->os, "ocs_pool_alloc hw_wq_callback_t failed: %d\n", rc);
		return rc;
	}

	rc = ocs_hw_setup_io(hw);
	if (rc) {
		ocs_log_err(hw->os, "IO allocation failure\n");
		return rc;
	}

	rc = ocs_hw_init_io(hw);
	if (rc) {
		ocs_log_err(hw->os, "IO initialization failure\n");
		return rc;
	}

	ocs_queue_history_init(hw->os, &hw->q_hist);

	/* get hw link config; polling, so callback will be called immediately */
	hw->linkcfg = OCS_HW_LINKCFG_NA;
	ocs_hw_get_linkcfg(hw, OCS_CMD_POLL, ocs_hw_init_linkcfg_cb, hw);

	/* if lancer ethernet, ethernet ports need to be enabled */
	if ((hw->sli.if_type == SLI4_IF_TYPE_LANCER_FC_ETH) &&
	    (sli_get_medium(&hw->sli) == SLI_LINK_MEDIUM_ETHERNET)) {
		if (ocs_hw_set_eth_license(hw, hw->eth_license)) {
			/* log warning but continue */
			ocs_log_err(hw->os, "Failed to set ethernet license\n");
		}
	}

	/* Set the DIF seed - only for lancer right now */
	if (SLI4_IF_TYPE_LANCER_FC_ETH == sli_get_if_type(&hw->sli) &&
	    ocs_hw_set_dif_seed(hw) != OCS_HW_RTN_SUCCESS) {
		ocs_log_err(hw->os, "Failed to set DIF seed value\n");
		return rc;
	}

	/* Set the DIF mode - skyhawk only */
	if (SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli) &&
	    sli_get_dif_capable(&hw->sli)) {
		rc = ocs_hw_set_dif_mode(hw);
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "Failed to set DIF mode value\n");
			return rc;
		}
	}

	/*
	 * Arming the EQ allows (e.g.) interrupts when CQ completions write EQ entries
	 */
	for (i = 0; i < hw->eq_count; i++) {
		sli_queue_arm(&hw->sli, &hw->eq[i], TRUE);
	}

	/*
	 * Initialize RQ hash
	 */
	for (i = 0; i < hw->rq_count; i++) {
		ocs_hw_queue_hash_add(hw->rq_hash, hw->rq[i].id, i);
	}

	/*
	 * Initialize WQ hash
	 */
	for (i = 0; i < hw->wq_count; i++) {
		ocs_hw_queue_hash_add(hw->wq_hash, hw->wq[i].id, i);
	}

	/*
	 * Arming the CQ allows (e.g.) MQ completions to write CQ entries
	 */
	for (i = 0; i < hw->cq_count; i++) {
		ocs_hw_queue_hash_add(hw->cq_hash, hw->cq[i].id, i);
		sli_queue_arm(&hw->sli, &hw->cq[i], TRUE);
	}

	/* record the fact that the queues are functional */
	hw->state = OCS_HW_STATE_ACTIVE;

	/* Note: Must be after the IOs are setup and the state is active*/
	if (ocs_hw_rqpair_init(hw)) {
		ocs_log_err(hw->os, "WARNING - error initializing RQ pair\n");
	}

	/* finally kick off periodic timer to check for timed out target WQEs */
	if (hw->config.emulate_tgt_wqe_timeout) {
		ocs_setup_timer(hw->os, &hw->wqe_timer, target_wqe_timer_cb, hw,
				OCS_HW_WQ_TIMER_PERIOD_MS);
	}

	/*
	 * Allocate a HW IOs for send frame.  Allocate one for each Class 1 WQ, or if there
	 * are none of those, allocate one for WQ[0]
	 */
	if ((count = ocs_varray_get_count(hw->wq_class_array[1])) > 0) {
		for (i = 0; i < count; i++) {
			hw_wq_t *wq = ocs_varray_iter_next(hw->wq_class_array[1]);
			wq->send_frame_io = ocs_hw_io_alloc(hw);
			if (wq->send_frame_io == NULL) {
				ocs_log_err(hw->os, "ocs_hw_io_alloc for send_frame_io failed\n");
			}
		}
	} else {
		hw->hw_wq[0]->send_frame_io = ocs_hw_io_alloc(hw);
		if (hw->hw_wq[0]->send_frame_io == NULL) {
			ocs_log_err(hw->os, "ocs_hw_io_alloc for send_frame_io failed\n");
		}
	}

	/* Initialize send frame frame sequence id */
	ocs_atomic_init(&hw->send_frame_seq_id, 0);

	/* Initialize watchdog timer if enabled by user */
	hw->expiration_logged = 0;
	if(hw->watchdog_timeout) {
		if((hw->watchdog_timeout < 1) || (hw->watchdog_timeout > 65534)) {
			ocs_log_err(hw->os, "watchdog_timeout out of range: Valid range is 1 - 65534\n");
		}else if(!ocs_hw_config_watchdog_timer(hw)) {
			ocs_log_info(hw->os, "watchdog timer configured with timeout = %d seconds \n", hw->watchdog_timeout);
		}
	}

	if (ocs_dma_alloc(hw->os, &hw->domain_dmem, 112, 4)) {
	   ocs_log_err(hw->os, "domain node memory allocation fail\n");
	   return OCS_HW_RTN_NO_MEMORY;
	}

	if (ocs_dma_alloc(hw->os, &hw->fcf_dmem, OCS_HW_READ_FCF_SIZE, OCS_HW_READ_FCF_SIZE)) {
	   ocs_log_err(hw->os, "domain fcf memory allocation fail\n");
	   return OCS_HW_RTN_NO_MEMORY;
	}

	if ((0 == hw->loop_map.size) &&	ocs_dma_alloc(hw->os, &hw->loop_map,
				SLI4_MIN_LOOP_MAP_BYTES, 4)) {
		ocs_log_err(hw->os, "Loop dma alloc failed size:%d \n", hw->loop_map.size);
	}

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @brief Configure Multi-RQ
 *
 * @param hw	Hardware context allocated by the caller.
 * @param mode	1 to set MRQ filters and 0 to set FCFI index
 * @param vlanid    valid in mode 0
 * @param fcf_index valid in mode 0
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_config_mrq(ocs_hw_t *hw, uint8_t mode, uint16_t vlanid, uint16_t fcf_index)
{
	uint8_t buf[SLI4_BMBX_SIZE], mrq_bitmask = 0;
	hw_rq_t *rq;
	sli4_cmd_reg_fcfi_mrq_t *rsp = NULL;
	uint32_t i, j;
	sli4_cmd_rq_cfg_t rq_filter[SLI4_CMD_REG_FCFI_MRQ_NUM_RQ_CFG];
	int32_t rc;

	if (mode == SLI4_CMD_REG_FCFI_SET_FCFI_MODE) {
		goto issue_cmd;
	}

	/* Set the filter match/mask values from hw's filter_def values */
	for (i = 0; i < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; i++) {
		rq_filter[i].rq_id = 0xffff;
		rq_filter[i].r_ctl_mask  = (uint8_t)  hw->config.filter_def[i];
		rq_filter[i].r_ctl_match = (uint8_t) (hw->config.filter_def[i] >> 8);
		rq_filter[i].type_mask   = (uint8_t) (hw->config.filter_def[i] >> 16);
		rq_filter[i].type_match  = (uint8_t) (hw->config.filter_def[i] >> 24);
	}

	/* Accumulate counts for each filter type used, build rq_ids[] list */
	for (i = 0; i < hw->hw_rq_count; i++) {
		rq = hw->hw_rq[i];
		for (j = 0; j < SLI4_CMD_REG_FCFI_MRQ_NUM_RQ_CFG; j++) {
			if (rq->filter_mask & (1U << j)) {
				if (rq_filter[j].rq_id != 0xffff) {
					/* Already used. Bailout ifts not RQset case */
					if (!rq->is_mrq || (rq_filter[j].rq_id != rq->base_mrq_id)) {
						ocs_log_err(hw->os, "Wrong queue topology.\n");
						return OCS_HW_RTN_ERROR;
					}
					continue;
				}

				if (rq->is_mrq) {
					rq_filter[j].rq_id = rq->base_mrq_id;
					mrq_bitmask |= (1U << j);
				} else {
					rq_filter[j].rq_id = rq->hdr->id;
				}
			}
		}
	}

issue_cmd:
	/* Invoke REG_FCFI_MRQ */
	rc = sli_cmd_reg_fcfi_mrq(&hw->sli,
				 buf,					/* buf */
				 SLI4_BMBX_SIZE,			/* size */
				 mode,					/* mode 1 */
				 fcf_index,				/* fcf_index */
				 vlanid,				/* vlan_id */
				 hw->config.rq_selection_policy,	/* RQ selection policy*/
				 mrq_bitmask,				/* MRQ bitmask */
				 hw->hw_mrq_count,			/* num_mrqs */
				 rq_filter);				/* RQ filter */
	if (rc == 0) {
		ocs_log_err(hw->os, "sli_cmd_reg_fcfi_mrq() failed: %d\n", rc);
		return OCS_HW_RTN_ERROR;
	}

	rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);

	rsp = (sli4_cmd_reg_fcfi_mrq_t *)buf;

	if ((rc != OCS_HW_RTN_SUCCESS) || (rsp->hdr.status)) {
		ocs_log_err(hw->os, "FCFI MRQ registration failed. cmd = %x status = %x\n",
			    rsp->hdr.command, rsp->hdr.status);
		return OCS_HW_RTN_ERROR;
	}

	if (mode == SLI4_CMD_REG_FCFI_SET_FCFI_MODE) {
		hw->fcf_indicator = rsp->fcfi;
	}
	return 0;
}

/**
 * @brief Callback function for getting linkcfg during HW initialization.
 *
 * @param status Status of the linkcfg get operation.
 * @param value Link configuration enum to which the link configuration is set.
 * @param arg Callback argument (ocs_hw_t *).
 *
 * @return None.
 */
static void
ocs_hw_init_linkcfg_cb(int32_t status, uintptr_t value, void *arg)
{
	ocs_hw_t *hw = (ocs_hw_t *)arg;
	if (status == 0) {
		hw->linkcfg = (ocs_hw_linkcfg_e)value;
	} else {
		hw->linkcfg = OCS_HW_LINKCFG_NA;
	}
	ocs_log_debug(hw->os, "linkcfg=%d\n", hw->linkcfg);
}

/**
 * @ingroup devInitShutdown
 * @brief Tear down the Hardware Abstraction Layer module.
 *
 * @par Description
 * Frees memory structures needed by the device, and shuts down the device. Does
 * not free the HW context memory (which is done by the caller).
 *
 * @param hw Hardware context allocated by the caller.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_teardown(ocs_hw_t *hw)
{
	uint32_t	i = 0;
	uint32_t	iters = 10;/*XXX*/
	uint32_t	max_rpi;
	uint32_t destroy_queues;
	uint32_t free_memory;

	if (!hw) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p\n", hw);
		return OCS_HW_RTN_ERROR;
	}

	destroy_queues = (hw->state == OCS_HW_STATE_ACTIVE);
	free_memory = (hw->state != OCS_HW_STATE_UNINITIALIZED);

	/* shutdown target wqe timer */
	shutdown_target_wqe_timer(hw);

	/* Cancel watchdog timer if enabled */
	if(hw->watchdog_timeout) {
		hw->watchdog_timeout = 0;
		ocs_hw_config_watchdog_timer(hw);
	}

	/* Cancel Sliport Healthcheck */
	if(hw->sliport_healthcheck) {
		hw->sliport_healthcheck = 0;
		ocs_hw_config_sli_port_health_check(hw, 0, 0);
	}

	if (hw->state != OCS_HW_STATE_QUEUES_ALLOCATED) {

		hw->state = OCS_HW_STATE_TEARDOWN_IN_PROGRESS;

		ocs_hw_flush(hw);

		/* If there are outstanding commands, wait for them to complete */
		while (!ocs_list_empty(&hw->cmd_head) && iters) {
			ocs_udelay(10000);
			ocs_hw_flush(hw);
			iters--;
		}

		if (ocs_list_empty(&hw->cmd_head)) {
			ocs_log_debug(hw->os, "All commands completed on MQ queue\n");
		} else {
			ocs_log_debug(hw->os, "Some commands still pending on MQ queue\n");
		}

		/* Cancel any remaining commands */
		ocs_hw_command_cancel(hw);
	} else {
		hw->state = OCS_HW_STATE_TEARDOWN_IN_PROGRESS;
	}

	ocs_lock_free(&hw->cmd_lock);

	/* Free unregistered RPI if workaround is in force */
	if (hw->workaround.use_unregistered_rpi) {
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_RPI, hw->workaround.unregistered_rid);
	}

	max_rpi = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_RPI);
	if (hw->rpi_ref) {
		for (i = 0; i < max_rpi; i++) {
			if (ocs_atomic_read(&hw->rpi_ref[i].rpi_count)) {
				ocs_log_debug(hw->os, "non-zero ref [%d]=%d\n",
						i, ocs_atomic_read(&hw->rpi_ref[i].rpi_count));
			}
		}
		ocs_free(hw->os, hw->rpi_ref, max_rpi * sizeof(*hw->rpi_ref));
		hw->rpi_ref = NULL;
	}

	ocs_dma_free(hw->os, &hw->rnode_mem);

	if (hw->io) {
		for (i = 0; i < hw->config.n_io; i++) {
			if (hw->io[i] && (hw->io[i]->sgl != NULL) &&
			    (hw->io[i]->sgl->virt != NULL)) {
				if(hw->io[i]->is_port_owned) {
					ocs_lock_free(&hw->io[i]->axr_lock);
				}
				ocs_dma_free(hw->os, hw->io[i]->sgl);
			}
			ocs_free(hw->os, hw->io[i], sizeof(ocs_hw_io_t));
			hw->io[i] = NULL;
		}
		ocs_free(hw->os, hw->wqe_buffs, hw->config.n_io * hw->sli.config.wqe_size);
		hw->wqe_buffs = NULL;
		ocs_free(hw->os, hw->io, hw->config.n_io * sizeof(ocs_hw_io_t *));
		hw->io = NULL;
	}

	ocs_dma_free(hw->os, &hw->xfer_rdy);
	ocs_dma_free(hw->os, &hw->dump_sges);
	ocs_dma_free(hw->os, &hw->loop_map);

	ocs_lock_free(&hw->io_lock);
	ocs_lock_free(&hw->io_abort_lock);


	for (i = 0; i < hw->wq_count; i++) {
		sli_queue_free(&hw->sli, &hw->wq[i], destroy_queues, free_memory);
	}


	for (i = 0; i < hw->rq_count; i++) {
		sli_queue_free(&hw->sli, &hw->rq[i], destroy_queues, free_memory);
	}

	for (i = 0; i < hw->mq_count; i++) {
		sli_queue_free(&hw->sli, &hw->mq[i], destroy_queues, free_memory);
	}

	for (i = 0; i < hw->cq_count; i++) {
		sli_queue_free(&hw->sli, &hw->cq[i], destroy_queues, free_memory);
	}

	for (i = 0; i < hw->eq_count; i++) {
		sli_queue_free(&hw->sli, &hw->eq[i], destroy_queues, free_memory);
	}

	ocs_hw_qtop_free(hw->qtop);

	/* Free rq buffers */
	ocs_hw_rx_free(hw);

	hw_queue_teardown(hw);

	ocs_hw_rqpair_teardown(hw);

	if (sli_teardown(&hw->sli)) {
		ocs_log_err(hw->os, "SLI teardown failed\n");
	}

	ocs_queue_history_free(&hw->q_hist);

	/* record the fact that the queues are non-functional */
	hw->state = OCS_HW_STATE_UNINITIALIZED;

	/* free sequence free pool */
	ocs_array_free(hw->seq_pool);
	hw->seq_pool = NULL;

	/* free hw_wq_callback pool */
	ocs_pool_free(hw->wq_reqtag_pool);

	ocs_dma_free(hw->os, &hw->domain_dmem);
	ocs_dma_free(hw->os, &hw->fcf_dmem);
	/* Mark HW setup as not having been called */
	hw->hw_setup_called = FALSE;

	return OCS_HW_RTN_SUCCESS;
}

ocs_hw_rtn_e
ocs_hw_reset(ocs_hw_t *hw, ocs_hw_reset_e reset)
{
	uint32_t	i;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint32_t	iters;
	ocs_hw_state_e prev_state = hw->state;

	if (hw->state != OCS_HW_STATE_ACTIVE) {
		ocs_log_test(hw->os, "HW state %d is not active\n", hw->state);
	}

	hw->state = OCS_HW_STATE_RESET_IN_PROGRESS;

	/* shutdown target wqe timer */
	shutdown_target_wqe_timer(hw);

	ocs_hw_flush(hw);

	/*
	 * If an mailbox command requiring a DMA is outstanding (i.e. SFP/DDM),
	 * then the FW will UE when the reset is issued. So attempt to complete
	 * all mailbox commands.
	 */
	iters = 10;
	while (!ocs_list_empty(&hw->cmd_head) && iters) {
		ocs_udelay(10000);
		ocs_hw_flush(hw);
		iters--;
	}

	if (ocs_list_empty(&hw->cmd_head)) {
		ocs_log_debug(hw->os, "All commands completed on MQ queue\n");
	} else {
		ocs_log_debug(hw->os, "Some commands still pending on MQ queue\n");
	}

	/* Reset the chip */
	switch(reset) {
	case OCS_HW_RESET_FUNCTION:
		ocs_log_debug(hw->os, "issuing function level reset\n");
		if (sli_reset(&hw->sli)) {
			ocs_log_err(hw->os, "sli_reset failed\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_RESET_FIRMWARE:
		ocs_log_debug(hw->os, "issuing firmware reset\n");
		if (sli_fw_reset(&hw->sli)) {
			ocs_log_err(hw->os, "sli_soft_reset failed\n");
			rc = OCS_HW_RTN_ERROR;
		}
		/*
		 * Because the FW reset leaves the FW in a non-running state,
		 * follow that with a regular reset.
		 */
		ocs_log_debug(hw->os, "issuing function level reset\n");
		if (sli_reset(&hw->sli)) {
			ocs_log_err(hw->os, "sli_reset failed\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	default:
		ocs_log_test(hw->os, "unknown reset type - no reset performed\n");
		hw->state = prev_state;
		return OCS_HW_RTN_ERROR;
	}

	/* Not safe to walk command/io lists unless they've been initialized */
	if (prev_state != OCS_HW_STATE_UNINITIALIZED) {
		ocs_hw_command_cancel(hw);

		/* Clean up the inuse list, the free list and the wait free list */
		ocs_hw_io_cancel(hw);

		ocs_memset(hw->domains, 0, sizeof(hw->domains));
		ocs_memset(hw->fcf_index_fcfi, 0, sizeof(hw->fcf_index_fcfi));

		ocs_hw_link_event_init(hw);

		ocs_lock(&hw->io_lock);
			/* The io lists should be empty, but remove any that didn't get cleaned up. */
			while (!ocs_list_empty(&hw->io_timed_wqe)) {
				ocs_list_remove_head(&hw->io_timed_wqe);
			}
			/* Don't clean up the io_inuse list, the backend will do that when it finishes the IO */

			while (!ocs_list_empty(&hw->io_free)) {
				ocs_list_remove_head(&hw->io_free);
			}
			while (!ocs_list_empty(&hw->io_wait_free)) {
				ocs_list_remove_head(&hw->io_wait_free);
			}

			/* Reset the request tag pool, the HW IO request tags are reassigned in ocs_hw_setup_io() */
			ocs_hw_reqtag_reset(hw);

		ocs_unlock(&hw->io_lock);
	}

	if (prev_state != OCS_HW_STATE_UNINITIALIZED) {
		for (i = 0; i < hw->wq_count; i++) {
			sli_queue_reset(&hw->sli, &hw->wq[i]);
		}

		for (i = 0; i < hw->rq_count; i++) {
			sli_queue_reset(&hw->sli, &hw->rq[i]);
		}

		for (i = 0; i < hw->hw_rq_count; i++) {
			hw_rq_t *rq = hw->hw_rq[i];
			if (rq->rq_tracker != NULL) {
				uint32_t j;

				for (j = 0; j < rq->entry_count; j++) {
					rq->rq_tracker[j] = NULL;
				}
			}
		}

		for (i = 0; i < hw->mq_count; i++) {
			sli_queue_reset(&hw->sli, &hw->mq[i]);
		}

		for (i = 0; i < hw->cq_count; i++) {
			sli_queue_reset(&hw->sli, &hw->cq[i]);
		}

		for (i = 0; i < hw->eq_count; i++) {
			sli_queue_reset(&hw->sli, &hw->eq[i]);
		}

		/* Free rq buffers */
		ocs_hw_rx_free(hw);

		/* Teardown the HW queue topology */
		hw_queue_teardown(hw);
	} else {

		/* Free rq buffers */
		ocs_hw_rx_free(hw);
	}

	/*
	 * Re-apply the run-time workarounds after clearing the SLI config
	 * fields in sli_reset.
	 */
	ocs_hw_workaround_setup(hw);
	hw->state = OCS_HW_STATE_QUEUES_ALLOCATED;

	return rc;
}

int32_t
ocs_hw_get_num_eq(ocs_hw_t *hw)
{
	return hw->eq_count;
}

static int32_t
ocs_hw_get_fw_timed_out(ocs_hw_t *hw)
{
	/* The error values below are taken from LOWLEVEL_SET_WATCHDOG_TIMER_rev1.pdf
	* No further explanation is given in the document.
	* */
	return (sli_reg_read(&hw->sli, SLI4_REG_SLIPORT_ERROR1) == 0x2 &&
		sli_reg_read(&hw->sli, SLI4_REG_SLIPORT_ERROR2) == 0x10);
}


ocs_hw_rtn_e
ocs_hw_get(ocs_hw_t *hw, ocs_hw_property_e prop, uint32_t *value)
{
	ocs_hw_rtn_e		rc = OCS_HW_RTN_SUCCESS;
	int32_t			tmp;

	if (!value) {
		return OCS_HW_RTN_ERROR;
	}

	*value = 0;

	switch (prop) {
	case OCS_HW_N_IO:
		*value = hw->config.n_io;
		break;
	case OCS_HW_N_SGL:
		*value = (hw->config.n_sgl - SLI4_SGE_MAX_RESERVED);
		break;
	case OCS_HW_MAX_IO:
		*value = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_XRI);
		break;
	case OCS_HW_MAX_NODES:
		*value = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_RPI);
		break;
	case OCS_HW_MAX_RQ_ENTRIES:
		*value = hw->num_qentries[SLI_QTYPE_RQ];
		break;
	case OCS_HW_RQ_DEFAULT_BUFFER_SIZE:
		*value = hw->config.rq_default_buffer_size;
		break;
	case OCS_HW_AUTO_XFER_RDY_CAPABLE:
		*value = sli_get_auto_xfer_rdy_capable(&hw->sli);
		break;
	case OCS_HW_AUTO_XFER_RDY_XRI_CNT:
		*value = hw->config.auto_xfer_rdy_xri_cnt;
		break;
	case OCS_HW_AUTO_XFER_RDY_SIZE:
		*value = hw->config.auto_xfer_rdy_size;
		break;
	case OCS_HW_AUTO_XFER_RDY_BLK_SIZE:
		switch (hw->config.auto_xfer_rdy_blk_size_chip) {
		case 0:
			*value = 512;
			break;
		case 1:
			*value = 1024;
			break;
		case 2:
			*value = 2048;
			break;
		case 3:
			*value = 4096;
			break;
		case 4:
			*value = 520;
			break;
		default:
			*value = 0;
			rc = OCS_HW_RTN_ERROR;
			break;
		}
		break;
	case OCS_HW_AUTO_XFER_RDY_T10_ENABLE:
		*value = hw->config.auto_xfer_rdy_t10_enable;
		break;
	case OCS_HW_AUTO_XFER_RDY_P_TYPE:
		*value = hw->config.auto_xfer_rdy_p_type;
		break;
	case OCS_HW_AUTO_XFER_RDY_REF_TAG_IS_LBA:
		*value = hw->config.auto_xfer_rdy_ref_tag_is_lba;
		break;
	case OCS_HW_AUTO_XFER_RDY_APP_TAG_VALID:
		*value = hw->config.auto_xfer_rdy_app_tag_valid;
		break;
	case OCS_HW_AUTO_XFER_RDY_APP_TAG_VALUE:
		*value = hw->config.auto_xfer_rdy_app_tag_value;
		break;
	case OCS_HW_MAX_SGE:
		*value = sli_get_max_sge(&hw->sli);
		break;
	case OCS_HW_MAX_SGL:
		*value = sli_get_max_sgl(&hw->sli);
		break;
	case OCS_HW_TOPOLOGY:
		/*
		 * Infer link.status based on link.speed.
		 * Report OCS_HW_TOPOLOGY_NONE if the link is down.
		 */
		if (hw->link.speed == 0) {
			*value = OCS_HW_TOPOLOGY_NONE;
			break;
		}
		switch (hw->link.topology) {
		case SLI_LINK_TOPO_NPORT:
			*value = OCS_HW_TOPOLOGY_NPORT;
			break;
		case SLI_LINK_TOPO_LOOP:
			*value = OCS_HW_TOPOLOGY_LOOP;
			break;
		case SLI_LINK_TOPO_NONE:
			*value = OCS_HW_TOPOLOGY_NONE;
			break;
		default:
			ocs_log_test(hw->os, "unsupported topology %#x\n", hw->link.topology);
			rc = OCS_HW_RTN_ERROR;
			break;
		}
		break;
	case OCS_HW_CONFIG_TOPOLOGY:
		*value = hw->config.topology;
		break;
	case OCS_HW_LINK_SPEED:
		*value = hw->link.speed;
		break;
	case OCS_HW_LINK_CONFIG_SPEED:
		switch (hw->config.speed) {
		case FC_LINK_SPEED_10G:
			*value = 10000;
			break;
		case FC_LINK_SPEED_AUTO_16_8_4:
			*value = 0;
			break;
		case FC_LINK_SPEED_2G:
			*value = 2000;
			break;
		case FC_LINK_SPEED_4G:
			*value = 4000;
			break;
		case FC_LINK_SPEED_8G:
			*value = 8000;
			break;
		case FC_LINK_SPEED_16G:
			*value = 16000;
			break;
		case FC_LINK_SPEED_32G:
			*value = 32000;
			break;
		default:
			ocs_log_test(hw->os, "unsupported speed %#x\n", hw->config.speed);
			rc = OCS_HW_RTN_ERROR;
			break;
		}
		break;
	case OCS_HW_IF_TYPE:
		*value = sli_get_if_type(&hw->sli);
		break;
	case OCS_HW_SLI_REV:
		*value = sli_get_sli_rev(&hw->sli);
		break;
	case OCS_HW_SLI_FAMILY:
		*value = sli_get_sli_family(&hw->sli);
		break;
	case OCS_HW_DIF_CAPABLE:
		*value = sli_get_dif_capable(&hw->sli);
		break;
	case OCS_HW_DIF_SEED:
		*value = hw->config.dif_seed;
		break;
	case OCS_HW_DIF_MODE:
		*value = hw->config.dif_mode;
		break;
	case OCS_HW_DIF_MULTI_SEPARATE:
		/* Lancer supports multiple DIF separates */
		if (hw->sli.if_type == SLI4_IF_TYPE_LANCER_FC_ETH) {
			*value = TRUE;
		} else {
			*value = FALSE;
		}
		break;
	case OCS_HW_DUMP_MAX_SIZE:
		*value = hw->dump_size;
		break;
	case OCS_HW_DUMP_READY:
		*value = sli_dump_is_ready(&hw->sli);
		break;
	case OCS_HW_DUMP_PRESENT:
		*value = sli_dump_is_present(&hw->sli);
		break;
	case OCS_HW_RESET_REQUIRED:
		tmp = sli_reset_required(&hw->sli);
		if(tmp < 0) {
			rc = OCS_HW_RTN_ERROR;
		} else {
			*value = tmp;
		}
		break;
	case OCS_HW_FW_ERROR:
		*value = sli_fw_error_status(&hw->sli);
		break;
	case OCS_HW_FW_READY:
		*value = sli_fw_ready(&hw->sli);
		break;
	case OCS_HW_FW_TIMED_OUT:
		*value = ocs_hw_get_fw_timed_out(hw);
		break;
	case OCS_HW_HIGH_LOGIN_MODE:
		*value = sli_get_hlm_capable(&hw->sli);
		break;
	case OCS_HW_PREREGISTER_SGL:
		*value = sli_get_sgl_preregister_required(&hw->sli);
		break;
	case OCS_HW_HW_REV1:
		*value = sli_get_hw_revision(&hw->sli, 0);
		break;
	case OCS_HW_HW_REV2:
		*value = sli_get_hw_revision(&hw->sli, 1);
		break;
	case OCS_HW_HW_REV3:
		*value = sli_get_hw_revision(&hw->sli, 2);
		break;
	case OCS_HW_LINKCFG:
		*value = hw->linkcfg;
		break;
	case OCS_HW_ETH_LICENSE:
		*value = hw->eth_license;
		break;
	case OCS_HW_LINK_MODULE_TYPE:
		*value = sli_get_link_module_type(&hw->sli);
		break;
	case OCS_HW_NUM_CHUTES:
		*value = ocs_hw_get_num_chutes(hw);
		break;
	case OCS_HW_DISABLE_AR_TGT_DIF:
		*value = hw->workaround.disable_ar_tgt_dif;
		break;
	case OCS_HW_EMULATE_I_ONLY_AAB:
		*value = hw->config.i_only_aab;
		break;
	case OCS_HW_EMULATE_TARGET_WQE_TIMEOUT:
		*value = hw->config.emulate_tgt_wqe_timeout;
		break;
	case OCS_HW_VPD_LEN:
		*value = sli_get_vpd_len(&hw->sli);
		break;
	case OCS_HW_SGL_CHAINING_CAPABLE:
		*value = sli_get_is_sgl_chaining_capable(&hw->sli) || hw->workaround.sglc_misreported;
		break;
	case OCS_HW_SGL_CHAINING_ALLOWED:
		/*
		 * SGL Chaining is allowed in the following cases:
		 *   1. Lancer with host SGL Lists
		 *   2. Skyhawk with pre-registered SGL Lists
		 */
		*value = FALSE;
		if ((sli_get_is_sgl_chaining_capable(&hw->sli) || hw->workaround.sglc_misreported) &&
		    !sli_get_sgl_preregister(&hw->sli) &&
		    SLI4_IF_TYPE_LANCER_FC_ETH  == sli_get_if_type(&hw->sli)) {
			*value = TRUE;
		}

		if ((sli_get_is_sgl_chaining_capable(&hw->sli) || hw->workaround.sglc_misreported) &&
		    sli_get_sgl_preregister(&hw->sli) &&
		    ((SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) ||
			(SLI4_IF_TYPE_BE3_SKH_VF == sli_get_if_type(&hw->sli)))) {
			*value = TRUE;
		}
		break;
	case OCS_HW_SGL_CHAINING_HOST_ALLOCATED:
		/* Only lancer supports host allocated SGL Chaining buffers. */
		*value = ((sli_get_is_sgl_chaining_capable(&hw->sli) || hw->workaround.sglc_misreported) &&
			  (SLI4_IF_TYPE_LANCER_FC_ETH  == sli_get_if_type(&hw->sli)));
		break;
	case OCS_HW_SEND_FRAME_CAPABLE:
		if (hw->workaround.ignore_send_frame) {
			*value = 0;
		} else {
			/* Only lancer is capable */
			*value = sli_get_if_type(&hw->sli) == SLI4_IF_TYPE_LANCER_FC_ETH;
		}
		break;
	case OCS_HW_RQ_SELECTION_POLICY:
		*value = hw->config.rq_selection_policy;
		break;
	case OCS_HW_RR_QUANTA:
		*value = hw->config.rr_quanta;
		break;
	case OCS_HW_MAX_VPORTS:
		*value = sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_VPI);
	default:
		ocs_log_test(hw->os, "unsupported property %#x\n", prop);
		rc = OCS_HW_RTN_ERROR;
	}

	return rc;
}

void *
ocs_hw_get_ptr(ocs_hw_t *hw, ocs_hw_property_e prop)
{
	void	*rc = NULL;

	switch (prop) {
	case OCS_HW_WWN_NODE:
		rc = sli_get_wwn_node(&hw->sli);
		break;
	case OCS_HW_WWN_PORT:
		rc = sli_get_wwn_port(&hw->sli);
		break;
	case OCS_HW_VPD:
		/* make sure VPD length is non-zero */
		if (sli_get_vpd_len(&hw->sli)) {
			rc = sli_get_vpd(&hw->sli);
		}
		break;
	case OCS_HW_FW_REV:
		rc = sli_get_fw_name(&hw->sli, 0);
		break;
	case OCS_HW_FW_REV2:
		rc = sli_get_fw_name(&hw->sli, 1);
		break;
	case OCS_HW_IPL:
		rc = sli_get_ipl_name(&hw->sli);
		break;
	case OCS_HW_PORTNUM:
		rc = sli_get_portnum(&hw->sli);
		break;
	case OCS_HW_BIOS_VERSION_STRING:
		rc = sli_get_bios_version_string(&hw->sli);
		break;
	default:
		ocs_log_test(hw->os, "unsupported property %#x\n", prop);
	}

	return rc;
}



ocs_hw_rtn_e
ocs_hw_set(ocs_hw_t *hw, ocs_hw_property_e prop, uint32_t value)
{
	ocs_hw_rtn_e		rc = OCS_HW_RTN_SUCCESS;

	switch (prop) {
	case OCS_HW_N_IO:
		if (value > sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_XRI) ||
		    value == 0) {
			ocs_log_test(hw->os, "IO value out of range %d vs %d\n",
					value, sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_XRI));
			rc = OCS_HW_RTN_ERROR;
		} else {
			hw->config.n_io = value;
		}
		break;
	case OCS_HW_N_SGL:
		value += SLI4_SGE_MAX_RESERVED;
		if (value > sli_get_max_sgl(&hw->sli)) {
			ocs_log_test(hw->os, "SGL value out of range %d vs %d\n",
					value, sli_get_max_sgl(&hw->sli));
			rc = OCS_HW_RTN_ERROR;
		} else {
			hw->config.n_sgl = value;
		}
		break;
	case OCS_HW_TOPOLOGY:
		if ((sli_get_medium(&hw->sli) != SLI_LINK_MEDIUM_FC) &&
				(value != OCS_HW_TOPOLOGY_AUTO)) {
			ocs_log_test(hw->os, "unsupported topology=%#x medium=%#x\n",
					value, sli_get_medium(&hw->sli));
			rc = OCS_HW_RTN_ERROR;
			break;
		}

		switch (value) {
		case OCS_HW_TOPOLOGY_AUTO:
			if (sli_get_medium(&hw->sli) == SLI_LINK_MEDIUM_FC) {
				sli_set_topology(&hw->sli, SLI4_READ_CFG_TOPO_FC);
			} else {
				sli_set_topology(&hw->sli, SLI4_READ_CFG_TOPO_FCOE);
			}
			break;
		case OCS_HW_TOPOLOGY_NPORT:
			sli_set_topology(&hw->sli, SLI4_READ_CFG_TOPO_FC_DA);
			break;
		case OCS_HW_TOPOLOGY_LOOP:
			sli_set_topology(&hw->sli, SLI4_READ_CFG_TOPO_FC_AL);
			break;
		default:
			ocs_log_test(hw->os, "unsupported topology %#x\n", value);
			rc = OCS_HW_RTN_ERROR;
		}
		hw->config.topology = value;
		break;
	case OCS_HW_LINK_SPEED:
		if (sli_get_medium(&hw->sli) != SLI_LINK_MEDIUM_FC) {
			switch (value) {
			case 0: 	/* Auto-speed negotiation */
			case 10000:	/* FCoE speed */
				hw->config.speed = FC_LINK_SPEED_10G;
				break;
			default:
				ocs_log_test(hw->os, "unsupported speed=%#x medium=%#x\n",
						value, sli_get_medium(&hw->sli));
				rc = OCS_HW_RTN_ERROR;
			}
			break;
		}

		switch (value) {
		case 0:		/* Auto-speed negotiation */
			hw->config.speed = FC_LINK_SPEED_AUTO_16_8_4;
			break;
		case 2000:	/* FC speeds */
			hw->config.speed = FC_LINK_SPEED_2G;
			break;
		case 4000:
			hw->config.speed = FC_LINK_SPEED_4G;
			break;
		case 8000:
			hw->config.speed = FC_LINK_SPEED_8G;
			break;
		case 16000:
			hw->config.speed = FC_LINK_SPEED_16G;
			break;
		case 32000:
			hw->config.speed = FC_LINK_SPEED_32G;
			break;
		default:
			ocs_log_test(hw->os, "unsupported speed %d\n", value);
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_DIF_SEED:
		/* Set the DIF seed - only for lancer right now */
		if (SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) {
			ocs_log_test(hw->os, "DIF seed not supported for this device\n");
			rc = OCS_HW_RTN_ERROR;
		} else {
			hw->config.dif_seed = value;
		}
		break;
	case OCS_HW_DIF_MODE:
		switch (value) {
		case OCS_HW_DIF_MODE_INLINE:
			/*
			 *  Make sure we support inline DIF.
			 *
			 * Note: Having both bits clear means that we have old
			 *	FW that doesn't set the bits.
			 */
			if (sli_is_dif_inline_capable(&hw->sli)) {
				hw->config.dif_mode = value;
			} else {
				ocs_log_test(hw->os, "chip does not support DIF inline\n");
				rc = OCS_HW_RTN_ERROR;
			}
			break;
		case OCS_HW_DIF_MODE_SEPARATE:
			/* Make sure we support DIF separates. */
			if (sli_is_dif_separate_capable(&hw->sli)) {
				hw->config.dif_mode = value;
			} else {
				ocs_log_test(hw->os, "chip does not support DIF separate\n");
				rc = OCS_HW_RTN_ERROR;
			}
		}
		break;
	case OCS_HW_RQ_PROCESS_LIMIT: {
		hw_rq_t *rq;
		uint32_t i;

		/* For each hw_rq object, set its parent CQ limit value */
		for (i = 0; i < hw->hw_rq_count; i++) {
			rq = hw->hw_rq[i];
			hw->cq[rq->cq->instance].proc_limit = value;
		}
		break;
	}
	case OCS_HW_RQ_DEFAULT_BUFFER_SIZE:
		hw->config.rq_default_buffer_size = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_XRI_CNT:
		hw->config.auto_xfer_rdy_xri_cnt = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_SIZE:
		hw->config.auto_xfer_rdy_size = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_BLK_SIZE:
		switch (value) {
		case 512:
			hw->config.auto_xfer_rdy_blk_size_chip = 0;
			break;
		case 1024:
			hw->config.auto_xfer_rdy_blk_size_chip = 1;
			break;
		case 2048:
			hw->config.auto_xfer_rdy_blk_size_chip = 2;
			break;
		case 4096:
			hw->config.auto_xfer_rdy_blk_size_chip = 3;
			break;
		case 520:
			hw->config.auto_xfer_rdy_blk_size_chip = 4;
			break;
		default:
			ocs_log_err(hw->os, "Invalid block size %d\n",
				    value);
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_AUTO_XFER_RDY_T10_ENABLE:
		hw->config.auto_xfer_rdy_t10_enable = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_P_TYPE:
		hw->config.auto_xfer_rdy_p_type = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_REF_TAG_IS_LBA:
		hw->config.auto_xfer_rdy_ref_tag_is_lba = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_APP_TAG_VALID:
		hw->config.auto_xfer_rdy_app_tag_valid = value;
		break;
	case OCS_HW_AUTO_XFER_RDY_APP_TAG_VALUE:
		hw->config.auto_xfer_rdy_app_tag_value = value;
		break;
	case OCS_ESOC:
		hw->config.esoc = value;
	case OCS_HW_HIGH_LOGIN_MODE:
		rc = sli_set_hlm(&hw->sli, value);
		break;
	case OCS_HW_PREREGISTER_SGL:
		rc = sli_set_sgl_preregister(&hw->sli, value);
		break;
	case OCS_HW_ETH_LICENSE:
		hw->eth_license = value;
		break;
	case OCS_HW_EMULATE_I_ONLY_AAB:
		hw->config.i_only_aab = value;
		break;
	case OCS_HW_EMULATE_TARGET_WQE_TIMEOUT:
		hw->config.emulate_tgt_wqe_timeout = value;
		break;
	case OCS_HW_BOUNCE:
		hw->config.bounce = value;
		break;
	case OCS_HW_RQ_SELECTION_POLICY:
		hw->config.rq_selection_policy = value;
		break;
	case OCS_HW_RR_QUANTA:
		hw->config.rr_quanta = value;
		break;
	default:
		ocs_log_test(hw->os, "unsupported property %#x\n", prop);
		rc = OCS_HW_RTN_ERROR;
	}

	return rc;
}


ocs_hw_rtn_e
ocs_hw_set_ptr(ocs_hw_t *hw, ocs_hw_property_e prop, void *value)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	switch (prop) {
	case OCS_HW_WAR_VERSION:
		hw->hw_war_version = value;
		break;
	case OCS_HW_FILTER_DEF: {
		char *p = value;
		uint32_t idx = 0;

		for (idx = 0; idx < ARRAY_SIZE(hw->config.filter_def); idx++) {
			hw->config.filter_def[idx] = 0;
		}

		for (idx = 0; (idx < ARRAY_SIZE(hw->config.filter_def)) && (p != NULL) && *p; ) {
			hw->config.filter_def[idx++] = ocs_strtoul(p, 0, 0);
			p = ocs_strchr(p, ',');
			if (p != NULL) {
				p++;
			}
		}

		break;
	}
	default:
		ocs_log_test(hw->os, "unsupported property %#x\n", prop);
		rc = OCS_HW_RTN_ERROR;
		break;
	}
	return rc;
}
/**
 * @ingroup interrupt
 * @brief Check for the events associated with the interrupt vector.
 *
 * @param hw Hardware context.
 * @param vector Zero-based interrupt vector number.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
int32_t
ocs_hw_event_check(ocs_hw_t *hw, uint32_t vector)
{
	int32_t rc = 0;

	if (!hw) {
		ocs_log_err(NULL, "HW context NULL?!?\n");
		return -1;
	}

	if (vector > hw->eq_count) {
		ocs_log_err(hw->os, "vector %d. max %d\n",
				vector, hw->eq_count);
		return -1;
	}

	/*
	 * The caller should disable interrupts if they wish to prevent us
	 * from processing during a shutdown. The following states are defined:
	 *   OCS_HW_STATE_UNINITIALIZED - No queues allocated
	 *   OCS_HW_STATE_QUEUES_ALLOCATED - The state after a chip reset,
	 *                                    queues are cleared.
	 *   OCS_HW_STATE_ACTIVE - Chip and queues are operational
	 *   OCS_HW_STATE_RESET_IN_PROGRESS - reset, we still want completions
	 *   OCS_HW_STATE_TEARDOWN_IN_PROGRESS - We still want mailbox
	 *                                        completions.
	 */
	if (hw->state != OCS_HW_STATE_UNINITIALIZED) {
		rc = sli_queue_is_empty(&hw->sli, &hw->eq[vector]);

		/* Re-arm queue if there are no entries */
		if (rc != 0) {
			sli_queue_arm(&hw->sli, &hw->eq[vector], TRUE);
		}
	}
	return rc;
}

void
ocs_hw_unsol_process_bounce(void *arg)
{
	ocs_hw_sequence_t *seq = arg;
	ocs_hw_t *hw = seq->hw;

	ocs_hw_assert(hw != NULL);
	ocs_hw_assert(hw->callback.unsolicited != NULL);

	hw->callback.unsolicited(hw->args.unsolicited, seq);
}

int32_t
ocs_hw_process(ocs_hw_t *hw, uint32_t vector, uint32_t max_isr_time_msec)
{
	hw_eq_t *eq;
	int32_t rc = 0;

	CPUTRACE("");

	/*
	 * The caller should disable interrupts if they wish to prevent us
	 * from processing during a shutdown. The following states are defined:
	 *   OCS_HW_STATE_UNINITIALIZED - No queues allocated
	 *   OCS_HW_STATE_QUEUES_ALLOCATED - The state after a chip reset,
	 *                                    queues are cleared.
	 *   OCS_HW_STATE_ACTIVE - Chip and queues are operational
	 *   OCS_HW_STATE_RESET_IN_PROGRESS - reset, we still want completions
	 *   OCS_HW_STATE_TEARDOWN_IN_PROGRESS - We still want mailbox
	 *                                        completions.
	 */
	if (hw->state == OCS_HW_STATE_UNINITIALIZED) {
		return 0;
	}

	/* Get pointer to hw_eq_t */
	eq = hw->hw_eq[vector];

	OCS_STAT(eq->use_count++);

	rc = ocs_hw_eq_process(hw, eq, max_isr_time_msec);

	return rc;
}

/**
 * @ingroup interrupt
 * @brief Process events associated with an EQ.
 *
 * @par Description
 * Loop termination:
 * @n @n Without a mechanism to terminate the completion processing loop, it
 * is possible under some workload conditions for the loop to never terminate
 * (or at least take longer than the OS is happy to have an interrupt handler
 * or kernel thread context hold a CPU without yielding).
 * @n @n The approach taken here is to periodically check how much time
 * we have been in this
 * processing loop, and if we exceed a predetermined time (multiple seconds), the
 * loop is terminated, and ocs_hw_process() returns.
 *
 * @param hw Hardware context.
 * @param eq Pointer to HW EQ object.
 * @param max_isr_time_msec Maximum time in msec to stay in this function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
int32_t
ocs_hw_eq_process(ocs_hw_t *hw, hw_eq_t *eq, uint32_t max_isr_time_msec)
{
	uint8_t		eqe[sizeof(sli4_eqe_t)] = { 0 };
	uint32_t	done = FALSE;
	uint32_t	tcheck_count;
	time_t		tstart;
	time_t		telapsed;

	tcheck_count = OCS_HW_TIMECHECK_ITERATIONS;
	tstart = ocs_msectime();

	CPUTRACE("");

	while (!done && !sli_queue_read(&hw->sli, eq->queue, eqe)) {
		uint16_t	cq_id = 0;
		int32_t		rc;

		rc = sli_eq_parse(&hw->sli, eqe, &cq_id);
		if (unlikely(rc)) {
			if (rc > 0) {
				uint32_t i;

				/*
				 * Received a sentinel EQE indicating the EQ is full.
				 * Process all CQs
				 */
				for (i = 0; i < hw->cq_count; i++) {
					ocs_hw_cq_process(hw, hw->hw_cq[i]);
				}
				continue;
			} else {
				return rc;
			}
		} else {
			int32_t index = ocs_hw_queue_hash_find(hw->cq_hash, cq_id);
			if (likely(index >= 0)) {
				ocs_hw_cq_process(hw, hw->hw_cq[index]);
			} else {
				ocs_log_err(hw->os, "bad CQ_ID %#06x\n", cq_id);
			}
		}


		if (eq->queue->n_posted > (eq->queue->posted_limit)) {
			sli_queue_arm(&hw->sli, eq->queue, FALSE);
		}

		if (tcheck_count && (--tcheck_count == 0)) {
			tcheck_count = OCS_HW_TIMECHECK_ITERATIONS;
			telapsed = ocs_msectime() - tstart;
			if (telapsed >= max_isr_time_msec) {
				done = TRUE;
			}
		}
	}
	sli_queue_eq_arm(&hw->sli, eq->queue, TRUE);

	return 0;
}

/**
 * @brief Submit queued (pending) mbx commands.
 *
 * @par Description
 * Submit queued mailbox commands.
 * --- Assumes that hw->cmd_lock is held ---
 *
 * @param hw Hardware context.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
static int32_t
ocs_hw_cmd_submit_pending(ocs_hw_t *hw)
{
	ocs_command_ctx_t *ctx;
	int32_t rc = 0;

	/* Assumes lock held */

	/* Only submit MQE if there's room */
	while (hw->cmd_head_count < (OCS_HW_MQ_DEPTH - 1)) {
		ctx = ocs_list_remove_head(&hw->cmd_pending);
		if (ctx == NULL) {
			break;
		}
		ocs_list_add_tail(&hw->cmd_head, ctx);
		hw->cmd_head_count++;
		if (sli_queue_write(&hw->sli, hw->mq, ctx->buf) < 0) {
			ocs_log_test(hw->os, "sli_queue_write failed: %d\n", rc);
			rc = -1;
			break;
		}
	}
	return rc;
}

/**
 * @ingroup io
 * @brief Issue a SLI command.
 *
 * @par Description
 * Send a mailbox command to the hardware, and either wait for a completion
 * (OCS_CMD_POLL) or get an optional asynchronous completion (OCS_CMD_NOWAIT).
 *
 * @param hw Hardware context.
 * @param cmd Buffer containing a formatted command and results.
 * @param opts Command options:
 *  - OCS_CMD_POLL - Command executes synchronously and busy-waits for the completion.
 *  - OCS_CMD_NOWAIT - Command executes asynchronously. Uses callback.
 * @param cb Function callback used for asynchronous mode. May be NULL.
 * @n Prototype is <tt>(*cb)(void *arg, uint8_t *cmd)</tt>.
 * @n @n @b Note: If the
 * callback function pointer is NULL, the results of the command are silently
 * discarded, allowing this pointer to exist solely on the stack.
 * @param arg Argument passed to an asynchronous callback.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_command(ocs_hw_t *hw, uint8_t *cmd, uint32_t opts, void *cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;

	/*
	 * If the chip is in an error state (UE'd) then reject this mailbox
	 *  command.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		uint32_t err1 = sli_reg_read(&hw->sli, SLI4_REG_SLIPORT_ERROR1);
		uint32_t err2 = sli_reg_read(&hw->sli, SLI4_REG_SLIPORT_ERROR2);
		if (hw->expiration_logged == 0 && err1 == 0x2 && err2 == 0x10) {
			hw->expiration_logged = 1;
			ocs_log_crit(hw->os,"Emulex: Heartbeat expired after %d seconds\n",
					hw->watchdog_timeout);
		}
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		ocs_log_crit(hw->os, "status=%#x error1=%#x error2=%#x\n",
			sli_reg_read(&hw->sli, SLI4_REG_SLIPORT_STATUS),
			err1, err2);

		return OCS_HW_RTN_ERROR;
	}

	if (OCS_CMD_POLL == opts) {

		ocs_lock(&hw->cmd_lock);
		if (hw->mq->length && !sli_queue_is_empty(&hw->sli, hw->mq)) {
			/*
			 * Can't issue Boot-strap mailbox command with other
			 * mail-queue commands pending as this interaction is
			 * undefined
			 */
			rc = OCS_HW_RTN_ERROR;
		} else {
			void *bmbx = hw->sli.bmbx.virt;

			ocs_memset(bmbx, 0, SLI4_BMBX_SIZE);
			ocs_memcpy(bmbx, cmd, SLI4_BMBX_SIZE);

			if (sli_bmbx_command(&hw->sli) == 0) {
				rc = OCS_HW_RTN_SUCCESS;
				ocs_memcpy(cmd, bmbx, SLI4_BMBX_SIZE);
			}
		}
		ocs_unlock(&hw->cmd_lock);
	} else if (OCS_CMD_NOWAIT == opts) {
		ocs_command_ctx_t	*ctx = NULL;

		ctx = ocs_malloc(hw->os, sizeof(ocs_command_ctx_t), OCS_M_ZERO | OCS_M_NOWAIT);
		if (!ctx) {
			ocs_log_err(hw->os, "can't allocate command context\n");
			return OCS_HW_RTN_NO_RESOURCES;
		}

		if (hw->state != OCS_HW_STATE_ACTIVE) {
			ocs_log_err(hw->os, "Can't send command, HW state=%d\n", hw->state);
			ocs_free(hw->os, ctx, sizeof(*ctx));
			return OCS_HW_RTN_ERROR;
		}

		if (cb) {
			ctx->cb = cb;
			ctx->arg = arg;
		}
		ctx->buf = cmd;
		ctx->ctx = hw;

		ocs_lock(&hw->cmd_lock);

			/* Add to pending list */
			ocs_list_add_tail(&hw->cmd_pending, ctx);

			/* Submit as much of the pending list as we can */
			if (ocs_hw_cmd_submit_pending(hw) == 0) {
				rc = OCS_HW_RTN_SUCCESS;
			}

		ocs_unlock(&hw->cmd_lock);
	}

	return rc;
}

/**
 * @ingroup devInitShutdown
 * @brief Register a callback for the given event.
 *
 * @param hw Hardware context.
 * @param which Event of interest.
 * @param func Function to call when the event occurs.
 * @param arg Argument passed to the callback function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_callback(ocs_hw_t *hw, ocs_hw_callback_e which, void *func, void *arg)
{

	if (!hw || !func || (which >= OCS_HW_CB_MAX)) {
		ocs_log_err(NULL, "bad parameter hw=%p which=%#x func=%p\n",
			    hw, which, func);
		return OCS_HW_RTN_ERROR;
	}

	switch (which) {
	case OCS_HW_CB_DOMAIN:
		hw->callback.domain = func;
		hw->args.domain = arg;
		break;
	case OCS_HW_CB_PORT:
		hw->callback.port = func;
		hw->args.port = arg;
		break;
	case OCS_HW_CB_UNSOLICITED:
		hw->callback.unsolicited = func;
		hw->args.unsolicited = arg;
		break;
	case OCS_HW_CB_REMOTE_NODE:
		hw->callback.rnode = func;
		hw->args.rnode = arg;
		break;
	case OCS_HW_CB_BOUNCE:
		hw->callback.bounce = func;
		hw->args.bounce = arg;
		break;
	default:
		ocs_log_test(hw->os, "unknown callback %#x\n", which);
		return OCS_HW_RTN_ERROR;
	}

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup port
 * @brief Allocate a port object.
 *
 * @par Description
 * This function allocates a VPI object for the port and stores it in the
 * indicator field of the port object.
 *
 * @param hw Hardware context.
 * @param sport SLI port object used to connect to the domain.
 * @param domain Domain object associated with this port (may be NULL).
 * @param wwpn Port's WWPN in big-endian order, or NULL to use default.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_port_alloc(ocs_hw_t *hw, ocs_sli_port_t *sport, ocs_domain_t *domain,
		uint8_t *wwpn)
{
	uint8_t	*cmd = NULL;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint32_t index;

	sport->indicator = UINT32_MAX;
	sport->hw = hw;
	sport->ctx.app = sport;
	sport->sm_free_req_pending = 0;

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	if (wwpn) {
		ocs_memcpy(&sport->sli_wwpn, wwpn, sizeof(sport->sli_wwpn));
	}

	if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_VPI, &sport->indicator, &index)) {
		ocs_log_err(hw->os, "FCOE_VPI allocation failure\n");
		return OCS_HW_RTN_ERROR;
	}

	if (domain != NULL) {
		ocs_sm_function_t	next = NULL;

		cmd = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
		if (!cmd) {
			ocs_log_err(hw->os, "command memory allocation failed\n");
			rc = OCS_HW_RTN_NO_MEMORY;
			goto ocs_hw_port_alloc_out;
		}

		/* If the WWPN is NULL, fetch the default WWPN and WWNN before
		 * initializing the VPI
		 */
		if (!wwpn) {
			next = __ocs_hw_port_alloc_read_sparm64;
		} else {
			next = __ocs_hw_port_alloc_init_vpi;
		}

		ocs_sm_transition(&sport->ctx, next, cmd);
	} else if (!wwpn) {
		/* This is the convention for the HW, not SLI */
		ocs_log_test(hw->os, "need WWN for physical port\n");
		rc = OCS_HW_RTN_ERROR;
	} else {
		/* domain NULL and wwpn non-NULL */
		ocs_sm_transition(&sport->ctx, __ocs_hw_port_alloc_init, NULL);
	}

ocs_hw_port_alloc_out:
	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_free(hw->os, cmd, SLI4_BMBX_SIZE);

		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VPI, sport->indicator);
	}

	return rc;
}

/**
 * @ingroup port
 * @brief Attach a physical/virtual SLI port to a domain.
 *
 * @par Description
 * This function registers a previously-allocated VPI with the
 * device.
 *
 * @param hw Hardware context.
 * @param sport Pointer to the SLI port object.
 * @param fc_id Fibre Channel ID to associate with this port.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success, or an error code on failure.
 */
ocs_hw_rtn_e
ocs_hw_port_attach(ocs_hw_t *hw, ocs_sli_port_t *sport, uint32_t fc_id)
{
	uint8_t	*buf = NULL;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	if (!hw || !sport) {
		ocs_log_err(hw ? hw->os : NULL,
			"bad parameter(s) hw=%p sport=%p\n", hw,
			sport);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (!buf) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	sport->fc_id = fc_id;
	ocs_sm_post_event(&sport->ctx, OCS_EVT_HW_PORT_REQ_ATTACH, buf);
	return rc;
}

/**
 * @brief Called when the port control command completes.
 *
 * @par Description
 * We only need to free the mailbox command buffer.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_port_control(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	return 0;
}

/**
 * @ingroup port
 * @brief Control a port (initialize, shutdown, or set link configuration).
 *
 * @par Description
 * This function controls a port depending on the @c ctrl parameter:
 * - @b OCS_HW_PORT_INIT -
 * Issues the CONFIG_LINK and INIT_LINK commands for the specified port.
 * The HW generates an OCS_HW_DOMAIN_FOUND event when the link comes up.
 * .
 * - @b OCS_HW_PORT_SHUTDOWN -
 * Issues the DOWN_LINK command for the specified port.
 * The HW generates an OCS_HW_DOMAIN_LOST event when the link is down.
 * .
 * - @b OCS_HW_PORT_SET_LINK_CONFIG -
 * Sets the link configuration.
 *
 * @param hw Hardware context.
 * @param ctrl Specifies the operation:
 * - OCS_HW_PORT_INIT
 * - OCS_HW_PORT_SHUTDOWN
 * - OCS_HW_PORT_SET_LINK_CONFIG
 *
 * @param value Operation-specific value.
 * - OCS_HW_PORT_INIT - Selective reset AL_PA
 * - OCS_HW_PORT_SHUTDOWN - N/A
 * - OCS_HW_PORT_SET_LINK_CONFIG - An enum #ocs_hw_linkcfg_e value.
 *
 * @param cb Callback function to invoke the following operation.
 * - OCS_HW_PORT_INIT/OCS_HW_PORT_SHUTDOWN - NULL (link events
 * are handled by the OCS_HW_CB_DOMAIN callbacks).
 * - OCS_HW_PORT_SET_LINK_CONFIG - Invoked after linkcfg mailbox command
 * completes.
 *
 * @param arg Callback argument invoked after the command completes.
 * - OCS_HW_PORT_INIT/OCS_HW_PORT_SHUTDOWN - NULL (link events
 * are handled by the OCS_HW_CB_DOMAIN callbacks).
 * - OCS_HW_PORT_SET_LINK_CONFIG - Invoked after linkcfg mailbox command
 * completes.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_port_control(ocs_hw_t *hw, ocs_hw_port_e ctrl, uintptr_t value, ocs_hw_port_control_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;

	switch (ctrl) {
	case OCS_HW_PORT_INIT:
	{
		uint8_t	*init_link;
		uint32_t speed = 0;
		uint8_t reset_alpa = 0;

		if (SLI_LINK_MEDIUM_FC == sli_get_medium(&hw->sli)) {
			uint8_t	*cfg_link;

			cfg_link = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
			if (cfg_link == NULL) {
				ocs_log_err(hw->os, "no buffer for command\n");
				return OCS_HW_RTN_NO_MEMORY;
			}

			if (sli_cmd_config_link(&hw->sli, cfg_link, SLI4_BMBX_SIZE)) {
				rc = ocs_hw_command(hw, cfg_link, OCS_CMD_NOWAIT,
							ocs_hw_cb_port_control, NULL);
			}

			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_free(hw->os, cfg_link, SLI4_BMBX_SIZE);
				ocs_log_err(hw->os, "CONFIG_LINK failed\n");
				break;
			}
			speed = hw->config.speed;
			reset_alpa = (uint8_t)(value & 0xff);
		} else {
			speed = FC_LINK_SPEED_10G;
		}

		/*
		 * Bring link up, unless FW version is not supported
		 */
		if (hw->workaround.fw_version_too_low) {
			if (SLI4_IF_TYPE_LANCER_FC_ETH == hw->sli.if_type) {
				ocs_log_err(hw->os, "Cannot bring up link.  Please update firmware to %s or later (current version is %s)\n",
					OCS_FW_VER_STR(OCS_MIN_FW_VER_LANCER), (char *) sli_get_fw_name(&hw->sli,0));
			} else {
				ocs_log_err(hw->os, "Cannot bring up link.  Please update firmware to %s or later (current version is %s)\n",
					OCS_FW_VER_STR(OCS_MIN_FW_VER_SKYHAWK), (char *) sli_get_fw_name(&hw->sli, 0));
			}

			return OCS_HW_RTN_ERROR;
		}

		rc = OCS_HW_RTN_ERROR;

		/* Allocate a new buffer for the init_link command */
		init_link = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
		if (init_link == NULL) {
			ocs_log_err(hw->os, "no buffer for command\n");
			return OCS_HW_RTN_NO_MEMORY;
		}


		if (sli_cmd_init_link(&hw->sli, init_link, SLI4_BMBX_SIZE, speed, reset_alpa)) {
			rc = ocs_hw_command(hw, init_link, OCS_CMD_NOWAIT,
						ocs_hw_cb_port_control, NULL);
		}
		/* Free buffer on error, since no callback is coming */
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_free(hw->os, init_link, SLI4_BMBX_SIZE);
			ocs_log_err(hw->os, "INIT_LINK failed\n");
		}
		break;
	}
	case OCS_HW_PORT_SHUTDOWN:
	{
		uint8_t	*down_link;

		down_link = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
		if (down_link == NULL) {
			ocs_log_err(hw->os, "no buffer for command\n");
			return OCS_HW_RTN_NO_MEMORY;
		}
		if (sli_cmd_down_link(&hw->sli, down_link, SLI4_BMBX_SIZE)) {
			rc = ocs_hw_command(hw, down_link, OCS_CMD_NOWAIT,
						ocs_hw_cb_port_control, NULL);
		}
		/* Free buffer on error, since no callback is coming */
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_free(hw->os, down_link, SLI4_BMBX_SIZE);
			ocs_log_err(hw->os, "DOWN_LINK failed\n");
		}
		break;
	}
	case OCS_HW_PORT_SET_LINK_CONFIG:
		rc = ocs_hw_set_linkcfg(hw, (ocs_hw_linkcfg_e)value, OCS_CMD_NOWAIT, cb, arg);
		break;
	default:
		ocs_log_test(hw->os, "unhandled control %#x\n", ctrl);
		break;
	}

	return rc;
}


/**
 * @ingroup port
 * @brief Free port resources.
 *
 * @par Description
 * Issue the UNREG_VPI command to free the assigned VPI context.
 *
 * @param hw Hardware context.
 * @param sport SLI port object used to connect to the domain.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_port_free(ocs_hw_t *hw, ocs_sli_port_t *sport)
{
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;

	if (!hw || !sport) {
		ocs_log_err(hw ? hw->os : NULL,
			"bad parameter(s) hw=%p sport=%p\n", hw,
			sport);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	ocs_sm_post_event(&sport->ctx, OCS_EVT_HW_PORT_REQ_FREE, NULL);
	return rc;
}

/**
 * @ingroup domain
 * @brief Allocate a fabric domain object.
 *
 * @par Description
 * This function starts a series of commands needed to connect to the domain, including
 *   - REG_FCFI
 *   - INIT_VFI
 *   - READ_SPARMS
 *   .
 * @b Note: Not all SLI interface types use all of the above commands.
 * @n @n Upon successful allocation, the HW generates a OCS_HW_DOMAIN_ALLOC_OK
 * event. On failure, it generates a OCS_HW_DOMAIN_ALLOC_FAIL event.
 *
 * @param hw Hardware context.
 * @param domain Pointer to the domain object.
 * @param fcf FCF index.
 * @param vlan VLAN ID.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_domain_alloc(ocs_hw_t *hw, ocs_domain_t *domain, uint32_t fcf, uint32_t vlan)
{
	uint8_t		*cmd = NULL;
	uint32_t	index;

	if (!hw || !domain || !domain->sport) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p domain=%p sport=%p\n",
				hw, domain, domain ? domain->sport : NULL);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	cmd = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (!cmd) {
		ocs_log_err(hw->os, "command memory allocation failed\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	domain->dma = hw->domain_dmem;

	domain->hw = hw;
	domain->sm.app = domain;
	domain->fcf = fcf;
	domain->fcf_indicator = UINT32_MAX;
	domain->vlan_id = vlan;
	domain->indicator = UINT32_MAX;

	if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_VFI, &domain->indicator, &index)) {
		ocs_log_err(hw->os, "FCOE_VFI allocation failure\n");

		ocs_free(hw->os, cmd, SLI4_BMBX_SIZE);

		return OCS_HW_RTN_ERROR;
	}

	ocs_sm_transition(&domain->sm, __ocs_hw_domain_init, cmd);
	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup domain
 * @brief Attach a SLI port to a domain.
 *
 * @param hw Hardware context.
 * @param domain Pointer to the domain object.
 * @param fc_id Fibre Channel ID to associate with this port.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_domain_attach(ocs_hw_t *hw, ocs_domain_t *domain, uint32_t fc_id)
{
	uint8_t	*buf = NULL;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	if (!hw || !domain) {
		ocs_log_err(hw ? hw->os : NULL,
			"bad parameter(s) hw=%p domain=%p\n",
			hw, domain);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (!buf) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	domain->sport->fc_id = fc_id;
	ocs_sm_post_event(&domain->sm, OCS_EVT_HW_DOMAIN_REQ_ATTACH, buf);
	return rc;
}

/**
 * @ingroup domain
 * @brief Free a fabric domain object.
 *
 * @par Description
 * Free both the driver and SLI port resources associated with the domain.
 *
 * @param hw Hardware context.
 * @param domain Pointer to the domain object.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_domain_free(ocs_hw_t *hw, ocs_domain_t *domain)
{
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;

	if (!hw || !domain) {
		ocs_log_err(hw ? hw->os : NULL,
			"bad parameter(s) hw=%p domain=%p\n",
			hw, domain);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	ocs_sm_post_event(&domain->sm, OCS_EVT_HW_DOMAIN_REQ_FREE, NULL);
	return rc;
}

/**
 * @ingroup domain
 * @brief Free a fabric domain object.
 *
 * @par Description
 * Free the driver resources associated with the domain. The difference between
 * this call and ocs_hw_domain_free() is that this call assumes resources no longer
 * exist on the SLI port, due to a reset or after some error conditions.
 *
 * @param hw Hardware context.
 * @param domain Pointer to the domain object.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_domain_force_free(ocs_hw_t *hw, ocs_domain_t *domain)
{
	if (!hw || !domain) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p domain=%p\n", hw, domain);
		return OCS_HW_RTN_ERROR;
	}

	sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VFI, domain->indicator);

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup node
 * @brief Allocate a remote node object.
 *
 * @param hw Hardware context.
 * @param rnode Allocated remote node object to initialize.
 * @param fc_addr FC address of the remote node.
 * @param sport SLI port used to connect to remote node.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_node_alloc(ocs_hw_t *hw, ocs_remote_node_t *rnode, uint32_t fc_addr,
		ocs_sli_port_t *sport)
{
	/* Check for invalid indicator */
	if (UINT32_MAX != rnode->indicator) {
		ocs_log_err(hw->os, "FCOE_RPI allocation failure addr=%#x rpi=%#x\n",
				fc_addr, rnode->indicator);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	/* NULL SLI port indicates an unallocated remote node */
	rnode->sport = NULL;

	if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_RPI, &rnode->indicator, &rnode->index)) {
		ocs_log_err(hw->os, "FCOE_RPI allocation failure addr=%#x\n",
				fc_addr);
		return OCS_HW_RTN_ERROR;
	}

	rnode->fc_id = fc_addr;
	rnode->sport = sport;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup node
 * @brief Update a remote node object with the remote port's service parameters.
 *
 * @param hw Hardware context.
 * @param rnode Allocated remote node object to initialize.
 * @param sparms DMA buffer containing the remote port's service parameters.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_node_attach(ocs_hw_t *hw, ocs_remote_node_t *rnode, ocs_dma_t *sparms)
{
	ocs_hw_rtn_e	rc = OCS_HW_RTN_ERROR;
	uint8_t		*buf = NULL;
	uint32_t	count = 0;

	if (!hw || !rnode || !sparms) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p rnode=%p sparms=%p\n",
			    hw, rnode, sparms);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (!buf) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/*
	 * If the attach count is non-zero, this RPI has already been registered.
	 * Otherwise, register the RPI
	 */
	if (rnode->index == UINT32_MAX) {
		ocs_log_err(NULL, "bad parameter rnode->index invalid\n");
		ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_ERROR;
	}
	count = ocs_atomic_add_return(&hw->rpi_ref[rnode->index].rpi_count, 1);
	if (count) {
		/*
		 * Can't attach multiple FC_ID's to a node unless High Login
		 * Mode is enabled
		 */
		if (sli_get_hlm(&hw->sli) == FALSE) {
			ocs_log_test(hw->os, "attach to already attached node HLM=%d count=%d\n",
					sli_get_hlm(&hw->sli), count);
			rc = OCS_HW_RTN_SUCCESS;
		} else {
			rnode->node_group = TRUE;
			rnode->attached = ocs_atomic_read(&hw->rpi_ref[rnode->index].rpi_attached);
			rc = rnode->attached  ? OCS_HW_RTN_SUCCESS_SYNC : OCS_HW_RTN_SUCCESS;
		}
	} else {
		rnode->node_group = FALSE;

		ocs_display_sparams("", "reg rpi", 0, NULL, sparms->virt);
		if (sli_cmd_reg_rpi(&hw->sli, buf, SLI4_BMBX_SIZE, rnode->fc_id,
					rnode->indicator, rnode->sport->indicator,
					sparms, 0, (hw->auto_xfer_rdy_enabled && hw->config.auto_xfer_rdy_t10_enable))) {
			rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT,
					ocs_hw_cb_node_attach, rnode);
		}
	}

	if (count || rc) {
		if (rc < OCS_HW_RTN_SUCCESS) {
			ocs_atomic_sub_return(&hw->rpi_ref[rnode->index].rpi_count, 1);
			ocs_log_err(hw->os, "%s error\n", count ? "HLM" : "REG_RPI");
		}
		ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
	}

	return rc;
}

/**
 * @ingroup node
 * @brief Free a remote node resource.
 *
 * @param hw Hardware context.
 * @param rnode Remote node object to free.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_node_free_resources(ocs_hw_t *hw, ocs_remote_node_t *rnode)
{
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;

	if (!hw || !rnode) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p rnode=%p\n",
			    hw, rnode);
		return OCS_HW_RTN_ERROR;
	}

	if (rnode->sport) {
		if (!rnode->attached) {
			if (rnode->indicator != UINT32_MAX) {
				if (sli_resource_free(&hw->sli, SLI_RSRC_FCOE_RPI, rnode->indicator)) {
					ocs_log_err(hw->os, "FCOE_RPI free failure RPI %d addr=%#x\n",
						    rnode->indicator, rnode->fc_id);
					rc = OCS_HW_RTN_ERROR;
				} else {
					rnode->node_group = FALSE;
					rnode->indicator = UINT32_MAX;
					rnode->index = UINT32_MAX;
					rnode->free_group = FALSE;
				}
			}
		} else {
			ocs_log_err(hw->os, "Error: rnode is still attached\n");
			rc = OCS_HW_RTN_ERROR;
		}
	}

	return rc;
}


/**
 * @ingroup node
 * @brief Free a remote node object.
 *
 * @param hw Hardware context.
 * @param rnode Remote node object to free.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_node_detach(ocs_hw_t *hw, ocs_remote_node_t *rnode)
{
	uint8_t	*buf = NULL;
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS_SYNC;
	uint32_t	index = UINT32_MAX;

	if (!hw || !rnode) {
		ocs_log_err(NULL, "bad parameter(s) hw=%p rnode=%p\n",
			    hw, rnode);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	index = rnode->index;

	if (rnode->sport) {
		uint32_t	count = 0;
		uint32_t	fc_id;

		if (!rnode->attached) {
			return OCS_HW_RTN_SUCCESS_SYNC;
		}

		buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
		if (!buf) {
			ocs_log_err(hw->os, "no buffer for command\n");
			return OCS_HW_RTN_NO_MEMORY;
		}

		count = ocs_atomic_sub_return(&hw->rpi_ref[index].rpi_count, 1);

		if (count <= 1) {
			/* There are no other references to this RPI
			 * so unregister it and free the resource. */
			fc_id = UINT32_MAX;
			rnode->node_group = FALSE;
			rnode->free_group = TRUE;
		} else {
			if (sli_get_hlm(&hw->sli) == FALSE) {
				ocs_log_test(hw->os, "Invalid count with HLM disabled, count=%d\n",
						count);
			}
			fc_id = rnode->fc_id & 0x00ffffff;
		}

		rc = OCS_HW_RTN_ERROR;

		if (sli_cmd_unreg_rpi(&hw->sli, buf, SLI4_BMBX_SIZE, rnode->indicator,
					SLI_RSRC_FCOE_RPI, fc_id)) {
			rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, ocs_hw_cb_node_free, rnode);
		}

		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "UNREG_RPI failed\n");
			ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
			rc = OCS_HW_RTN_ERROR;
		}
	}

	return rc;
}

/**
 * @ingroup node
 * @brief Free all remote node objects.
 *
 * @param hw Hardware context.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_node_free_all(ocs_hw_t *hw)
{
	uint8_t	*buf = NULL;
	ocs_hw_rtn_e	rc = OCS_HW_RTN_ERROR;

	if (!hw) {
		ocs_log_err(NULL, "bad parameter hw=%p\n", hw);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Check if the chip is in an error state (UE'd) before proceeding.
	 */
	if (sli_fw_error_status(&hw->sli) > 0) {
		ocs_log_crit(hw->os, "Chip is in an error state - reset needed\n");
		return OCS_HW_RTN_ERROR;
	}

	buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (!buf) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_unreg_rpi(&hw->sli, buf, SLI4_BMBX_SIZE, 0xffff,
				SLI_RSRC_FCOE_FCFI, UINT32_MAX)) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, ocs_hw_cb_node_free_all,
				NULL);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_err(hw->os, "UNREG_RPI failed\n");
		ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
		rc = OCS_HW_RTN_ERROR;
	}

	return rc;
}

ocs_hw_rtn_e
ocs_hw_node_group_alloc(ocs_hw_t *hw, ocs_remote_node_group_t *ngroup)
{

	if (!hw || !ngroup) {
		ocs_log_err(NULL, "bad parameter hw=%p ngroup=%p\n",
				hw, ngroup);
		return OCS_HW_RTN_ERROR;
	}

	if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_RPI, &ngroup->indicator,
				&ngroup->index)) {
		ocs_log_err(hw->os, "FCOE_RPI allocation failure addr=%#x\n",
				ngroup->indicator);
		return OCS_HW_RTN_ERROR;
	}

	return OCS_HW_RTN_SUCCESS;
}

ocs_hw_rtn_e
ocs_hw_node_group_attach(ocs_hw_t *hw, ocs_remote_node_group_t *ngroup, ocs_remote_node_t *rnode)
{

	if (!hw || !ngroup || !rnode) {
		ocs_log_err(NULL, "bad parameter hw=%p ngroup=%p rnode=%p\n",
			    hw, ngroup, rnode);
		return OCS_HW_RTN_ERROR;
	}

	if (rnode->attached) {
		ocs_log_err(hw->os, "node already attached RPI=%#x addr=%#x\n",
			    rnode->indicator, rnode->fc_id);
		return OCS_HW_RTN_ERROR;
	}

	if (sli_resource_free(&hw->sli, SLI_RSRC_FCOE_RPI, rnode->indicator)) {
		ocs_log_err(hw->os, "FCOE_RPI free failure RPI=%#x\n",
				rnode->indicator);
		return OCS_HW_RTN_ERROR;
	}

	rnode->indicator = ngroup->indicator;
	rnode->index = ngroup->index;

	return OCS_HW_RTN_SUCCESS;
}

ocs_hw_rtn_e
ocs_hw_node_group_free(ocs_hw_t *hw, ocs_remote_node_group_t *ngroup)
{
	int	ref;

	if (!hw || !ngroup) {
		ocs_log_err(NULL, "bad parameter hw=%p ngroup=%p\n",
				hw, ngroup);
		return OCS_HW_RTN_ERROR;
	}

	ref = ocs_atomic_read(&hw->rpi_ref[ngroup->index].rpi_count);
	if (ref) {
		/* Hmmm, the reference count is non-zero */
		ocs_log_debug(hw->os, "node group reference=%d (RPI=%#x)\n",
				ref, ngroup->indicator);

		if (sli_resource_free(&hw->sli, SLI_RSRC_FCOE_RPI, ngroup->indicator)) {
			ocs_log_err(hw->os, "FCOE_RPI free failure RPI=%#x\n",
				    ngroup->indicator);
			return OCS_HW_RTN_ERROR;
		}

		ocs_atomic_set(&hw->rpi_ref[ngroup->index].rpi_count, 0);
	}

	ngroup->indicator = UINT32_MAX;
	ngroup->index = UINT32_MAX;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @brief Initialize IO fields on each free call.
 *
 * @n @b Note: This is done on each free call (as opposed to each
 * alloc call) because port-owned XRIs are not
 * allocated with ocs_hw_io_alloc() but are freed with this
 * function.
 *
 * @param io Pointer to HW IO.
 */
static inline void
ocs_hw_init_free_io(ocs_hw_io_t *io)
{
	/*
	 * Set io->done to NULL, to avoid any callbacks, should
	 * a completion be received for one of these IOs
	 */
	io->done = NULL;
	io->abort_done = NULL;
	io->status_saved = 0;
	io->abort_in_progress = FALSE;
	io->port_owned_abort_count = 0;
	io->rnode = NULL;
	io->type = 0xFFFF;
	io->wq = NULL;
	io->ul_io = NULL;
	io->tgt_wqe_timeout = 0;
}

/**
 * @ingroup io
 * @brief Lockless allocate a HW IO object.
 *
 * @par Description
 * Assume that hw->ocs_lock is held. This function is only used if
 * use_dif_sec_xri workaround is being used.
 *
 * @param hw Hardware context.
 *
 * @return Returns a pointer to an object on success, or NULL on failure.
 */
static inline ocs_hw_io_t *
_ocs_hw_io_alloc(ocs_hw_t *hw)
{
	ocs_hw_io_t	*io = NULL;

	if (NULL != (io = ocs_list_remove_head(&hw->io_free))) {
		ocs_list_add_tail(&hw->io_inuse, io);
		io->state = OCS_HW_IO_STATE_INUSE;
		io->quarantine = FALSE;
		io->quarantine_first_phase = TRUE;
		io->abort_reqtag = UINT32_MAX;
		ocs_ref_init(&io->ref, ocs_hw_io_free_internal, io);
	} else {
		ocs_atomic_add_return(&hw->io_alloc_failed_count, 1);
	}

	return io;
}
/**
 * @ingroup io
 * @brief Allocate a HW IO object.
 *
 * @par Description
 * @n @b Note: This function applies to non-port owned XRIs
 * only.
 *
 * @param hw Hardware context.
 *
 * @return Returns a pointer to an object on success, or NULL on failure.
 */
ocs_hw_io_t *
ocs_hw_io_alloc(ocs_hw_t *hw)
{
	ocs_hw_io_t	*io = NULL;

	ocs_lock(&hw->io_lock);
		io = _ocs_hw_io_alloc(hw);
	ocs_unlock(&hw->io_lock);

	return io;
}

/**
 * @ingroup io
 * @brief Allocate/Activate a port owned HW IO object.
 *
 * @par Description
 * This function is called by the transport layer when an XRI is
 * allocated by the SLI-Port. This will "activate" the HW IO
 * associated with the XRI received from the SLI-Port to mirror
 * the state of the XRI.
 * @n @n @b Note: This function applies to port owned XRIs only.
 *
 * @param hw Hardware context.
 * @param io Pointer HW IO to activate/allocate.
 *
 * @return Returns a pointer to an object on success, or NULL on failure.
 */
ocs_hw_io_t *
ocs_hw_io_activate_port_owned(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	if (ocs_ref_read_count(&io->ref) > 0) {
		ocs_log_err(hw->os, "Bad parameter: refcount > 0\n");
		return NULL;
	}

	if (io->wq != NULL) {
		ocs_log_err(hw->os, "XRI %x already in use\n", io->indicator);
		return NULL;
	}

	ocs_ref_init(&io->ref, ocs_hw_io_free_port_owned, io);
	io->xbusy = TRUE;

	return io;
}

/**
 * @ingroup io
 * @brief When an IO is freed, depending on the exchange busy flag, and other
 * workarounds, move it to the correct list.
 *
 * @par Description
 * @n @b Note: Assumes that the hw->io_lock is held and the item has been removed
 * from the busy or wait_free list.
 *
 * @param hw Hardware context.
 * @param io Pointer to the IO object to move.
 */
static void
ocs_hw_io_free_move_correct_list(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	if (io->xbusy) {
		/* add to wait_free list and wait for XRI_ABORTED CQEs to clean up */
		ocs_list_add_tail(&hw->io_wait_free, io);
		io->state = OCS_HW_IO_STATE_WAIT_FREE;
	} else {
		/* IO not busy, add to free list */
		ocs_list_add_tail(&hw->io_free, io);
		io->state = OCS_HW_IO_STATE_FREE;
	}

	/* BZ 161832 workaround */
	if (hw->workaround.use_dif_sec_xri) {
		ocs_hw_check_sec_hio_list(hw);
	}
}

/**
 * @ingroup io
 * @brief Free a HW IO object. Perform cleanup common to
 * port and host-owned IOs.
 *
 * @param hw Hardware context.
 * @param io Pointer to the HW IO object.
 */
static inline void
ocs_hw_io_free_common(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	/* initialize IO fields */
	ocs_hw_init_free_io(io);

	/* Restore default SGL */
	ocs_hw_io_restore_sgl(hw, io);
}

/**
 * @ingroup io
 * @brief Free a HW IO object associated with a port-owned XRI.
 *
 * @param arg Pointer to the HW IO object.
 */
static void
ocs_hw_io_free_port_owned(void *arg)
{
	ocs_hw_io_t *io = (ocs_hw_io_t *)arg;
	ocs_hw_t *hw = io->hw;

	/*
	 * For auto xfer rdy, if the dnrx bit is set, then add it to the list of XRIs
	 * waiting for buffers.
	 */
	if (io->auto_xfer_rdy_dnrx) {
		ocs_lock(&hw->io_lock);
			/* take a reference count because we still own the IO until the buffer is posted */
			ocs_ref_init(&io->ref, ocs_hw_io_free_port_owned, io);
			ocs_list_add_tail(&hw->io_port_dnrx, io);
		ocs_unlock(&hw->io_lock);
	}

	/* perform common cleanup */
	ocs_hw_io_free_common(hw, io);
}

/**
 * @ingroup io
 * @brief Free a previously-allocated HW IO object. Called when
 * IO refcount goes to zero (host-owned IOs only).
 *
 * @param arg Pointer to the HW IO object.
 */
static void
ocs_hw_io_free_internal(void *arg)
{
	ocs_hw_io_t *io = (ocs_hw_io_t *)arg;
	ocs_hw_t *hw = io->hw;

	/* perform common cleanup */
	ocs_hw_io_free_common(hw, io);

	ocs_lock(&hw->io_lock);
		/* remove from in-use list */
		ocs_list_remove(&hw->io_inuse, io);
		ocs_hw_io_free_move_correct_list(hw, io);
	ocs_unlock(&hw->io_lock);
}

/**
 * @ingroup io
 * @brief Free a previously-allocated HW IO object.
 *
 * @par Description
 * @n @b Note: This function applies to port and host owned XRIs.
 *
 * @param hw Hardware context.
 * @param io Pointer to the HW IO object.
 *
 * @return Returns a non-zero value if HW IO was freed, 0 if references
 * on the IO still exist, or a negative value if an error occurred.
 */
int32_t
ocs_hw_io_free(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	/* just put refcount */
	if (ocs_ref_read_count(&io->ref) <= 0) {
		ocs_log_err(hw->os, "Bad parameter: refcount <= 0 xri=%x tag=%x\n",
			    io->indicator, io->reqtag);
		return -1;
	}

	return ocs_ref_put(&io->ref); /* ocs_ref_get(): ocs_hw_io_alloc() */
}

/**
 * @ingroup io
 * @brief Check if given HW IO is in-use
 *
 * @par Description
 * This function returns TRUE if the given HW IO has been
 * allocated and is in-use, and FALSE otherwise. It applies to
 * port and host owned XRIs.
 *
 * @param hw Hardware context.
 * @param io Pointer to the HW IO object.
 *
 * @return TRUE if an IO is in use, or FALSE otherwise.
 */
uint8_t
ocs_hw_io_inuse(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	return (ocs_ref_read_count(&io->ref) > 0);
}

/**
 * @brief Write a HW IO to a work queue.
 *
 * @par Description
 * A HW IO is written to a work queue.
 *
 * @param wq Pointer to work queue.
 * @param wqe Pointer to WQ entry.
 *
 * @n @b Note: Assumes the SLI-4 queue lock is held.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
static int32_t
_hw_wq_write(hw_wq_t *wq, ocs_hw_wqe_t *wqe)
{
	int32_t rc;
	int32_t queue_rc;

	/* Every so often, set the wqec bit to generate comsummed completions */
	if (wq->wqec_count) {
		wq->wqec_count--;
	}
	if (wq->wqec_count == 0) {
		sli4_generic_wqe_t *genwqe = (void*)wqe->wqebuf;
		genwqe->wqec = 1;
		wq->wqec_count = wq->wqec_set_count;
	}

	/* Decrement WQ free count */
	wq->free_count--;

	queue_rc = _sli_queue_write(&wq->hw->sli, wq->queue, wqe->wqebuf);

	if (queue_rc < 0) {
		rc = -1;
	} else {
		rc = 0;
		ocs_queue_history_wq(&wq->hw->q_hist, (void *) wqe->wqebuf, wq->queue->id, queue_rc);
	}

	return rc;
}

/**
 * @brief Write a HW IO to a work queue.
 *
 * @par Description
 * A HW IO is written to a work queue.
 *
 * @param wq Pointer to work queue.
 * @param wqe Pointer to WQE entry.
 *
 * @n @b Note: Takes the SLI-4 queue lock.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
int32_t
hw_wq_write(hw_wq_t *wq, ocs_hw_wqe_t *wqe)
{
	int32_t rc = 0;

	sli_queue_lock(wq->queue);
		if ( ! ocs_list_empty(&wq->pending_list)) {
			ocs_list_add_tail(&wq->pending_list, wqe);
			OCS_STAT(wq->wq_pending_count++;)
			while ((wq->free_count > 0) && ((wqe = ocs_list_remove_head(&wq->pending_list)) != NULL)) {
				rc = _hw_wq_write(wq, wqe);
				if (rc < 0) {
					break;
				}
				if (wqe->abort_wqe_submit_needed) {
					wqe->abort_wqe_submit_needed = 0;
					sli_abort_wqe(&wq->hw->sli, wqe->wqebuf, wq->hw->sli.config.wqe_size, SLI_ABORT_XRI,
							wqe->send_abts, wqe->id, 0, wqe->abort_reqtag, SLI4_CQ_DEFAULT );
					ocs_list_add_tail(&wq->pending_list, wqe);
					OCS_STAT(wq->wq_pending_count++;)
				}
			}
		} else {
			if (wq->free_count > 0) {
				rc = _hw_wq_write(wq, wqe);
			} else {
				ocs_list_add_tail(&wq->pending_list, wqe);
				OCS_STAT(wq->wq_pending_count++;)
			}
		}

	sli_queue_unlock(wq->queue);

	return rc;

}

/**
 * @brief Update free count and submit any pending HW IOs
 *
 * @par Description
 * The WQ free count is updated, and any pending HW IOs are submitted that
 * will fit in the queue.
 *
 * @param wq Pointer to work queue.
 * @param update_free_count Value added to WQs free count.
 *
 * @return None.
 */
static void
hw_wq_submit_pending(hw_wq_t *wq, uint32_t update_free_count)
{
	ocs_hw_wqe_t *wqe;

	sli_queue_lock(wq->queue);

		/* Update free count with value passed in */
		wq->free_count += update_free_count;

		while ((wq->free_count > 0) && ((wqe = ocs_list_remove_head(&wq->pending_list)) != NULL)) {
			_hw_wq_write(wq, wqe);

			if (wqe->abort_wqe_submit_needed) {
				wqe->abort_wqe_submit_needed = 0;
				sli_abort_wqe(&wq->hw->sli, wqe->wqebuf, wq->hw->sli.config.wqe_size, SLI_ABORT_XRI,
						wqe->send_abts, wqe->id, 0, wqe->abort_reqtag, SLI4_CQ_DEFAULT);
				ocs_list_add_tail(&wq->pending_list, wqe);
				OCS_STAT(wq->wq_pending_count++;)
			}
		}

	sli_queue_unlock(wq->queue);
}

/**
 * @brief Check to see if there are any BZ 161832 workaround waiting IOs
 *
 * @par Description
 * Checks hw->sec_hio_wait_list, if an IO is waiting for a HW IO, then try
 * to allocate a secondary HW io, and dispatch it.
 *
 * @n @b Note: hw->io_lock MUST be taken when called.
 *
 * @param hw pointer to HW object
 *
 * @return none
 */
static void
ocs_hw_check_sec_hio_list(ocs_hw_t *hw)
{
	ocs_hw_io_t *io;
	ocs_hw_io_t *sec_io;
	int rc = 0;

	while (!ocs_list_empty(&hw->sec_hio_wait_list)) {
		uint16_t flags;

		sec_io = _ocs_hw_io_alloc(hw);
		if (sec_io == NULL) {
			break;
		}

		io = ocs_list_remove_head(&hw->sec_hio_wait_list);
		ocs_list_add_tail(&hw->io_inuse, io);
		io->state = OCS_HW_IO_STATE_INUSE;
		io->sec_hio = sec_io;

		/* mark secondary XRI for second and subsequent data phase as quarantine */
		if (io->xbusy) {
			sec_io->quarantine = TRUE;
		}

		flags = io->sec_iparam.fcp_tgt.flags;
		if (io->xbusy) {
			flags |= SLI4_IO_CONTINUATION;
		} else {
			flags &= ~SLI4_IO_CONTINUATION;
		}

		io->tgt_wqe_timeout = io->sec_iparam.fcp_tgt.timeout;

		/* Complete (continue) TRECV IO */
		if (io->xbusy) {
			if (sli_fcp_cont_treceive64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl,
				io->first_data_sge,
				io->sec_iparam.fcp_tgt.offset, io->sec_len, io->indicator, io->sec_hio->indicator,
				io->reqtag, SLI4_CQ_DEFAULT,
				io->sec_iparam.fcp_tgt.ox_id, io->rnode->indicator, io->rnode,
				flags,
				io->sec_iparam.fcp_tgt.dif_oper, io->sec_iparam.fcp_tgt.blk_size, io->sec_iparam.fcp_tgt.cs_ctl, io->sec_iparam.fcp_tgt.app_id)) {
					ocs_log_test(hw->os, "TRECEIVE WQE error\n");
					break;
			}
		} else {
			if (sli_fcp_treceive64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl,
				io->first_data_sge,
				io->sec_iparam.fcp_tgt.offset, io->sec_len, io->indicator,
				io->reqtag, SLI4_CQ_DEFAULT,
				io->sec_iparam.fcp_tgt.ox_id, io->rnode->indicator, io->rnode,
				flags,
				io->sec_iparam.fcp_tgt.dif_oper, io->sec_iparam.fcp_tgt.blk_size,
				io->sec_iparam.fcp_tgt.cs_ctl, io->sec_iparam.fcp_tgt.app_id)) {
					ocs_log_test(hw->os, "TRECEIVE WQE error\n");
					break;
			}
		}

		if (io->wq == NULL) {
			io->wq = ocs_hw_queue_next_wq(hw, io);
			ocs_hw_assert(io->wq != NULL);
		}
		io->xbusy = TRUE;

		/*
		 * Add IO to active io wqe list before submitting, in case the
		 * wcqe processing preempts this thread.
		 */
		ocs_hw_add_io_timed_wqe(hw, io);
		rc = hw_wq_write(io->wq, &io->wqe);
		if (rc >= 0) {
			/* non-negative return is success */
			rc = 0;
		} else {
			/* failed to write wqe, remove from active wqe list */
			ocs_log_err(hw->os, "sli_queue_write failed: %d\n", rc);
			io->xbusy = FALSE;
			ocs_hw_remove_io_timed_wqe(hw, io);
		}
	}
}

/**
 * @ingroup io
 * @brief Send a Single Request/Response Sequence (SRRS).
 *
 * @par Description
 * This routine supports communication sequences consisting of a single
 * request and single response between two endpoints. Examples include:
 *  - Sending an ELS request.
 *  - Sending an ELS response - To send an ELS reponse, the caller must provide
 * the OX_ID from the received request.
 *  - Sending a FC Common Transport (FC-CT) request - To send a FC-CT request,
 * the caller must provide the R_CTL, TYPE, and DF_CTL
 * values to place in the FC frame header.
 *  .
 * @n @b Note: The caller is expected to provide both send and receive
 * buffers for requests. In the case of sending a response, no receive buffer
 * is necessary and the caller may pass in a NULL pointer.
 *
 * @param hw Hardware context.
 * @param type Type of sequence (ELS request/response, FC-CT).
 * @param io Previously-allocated HW IO object.
 * @param send DMA memory holding data to send (for example, ELS request, BLS response).
 * @param len Length, in bytes, of data to send.
 * @param receive Optional DMA memory to hold a response.
 * @param rnode Destination of data (that is, a remote node).
 * @param iparam IO parameters (ELS response and FC-CT).
 * @param cb Function call upon completion of sending the data (may be NULL).
 * @param arg Argument to pass to IO completion function.
 *
 * @return Returns 0 on success, or a non-zero on failure.
 */
ocs_hw_rtn_e
ocs_hw_srrs_send(ocs_hw_t *hw, ocs_hw_io_type_e type, ocs_hw_io_t *io,
		  ocs_dma_t *send, uint32_t len, ocs_dma_t *receive,
		  ocs_remote_node_t *rnode, ocs_hw_io_param_t *iparam,
		  ocs_hw_srrs_cb_t cb, void *arg)
{
	sli4_sge_t	*sge = NULL;
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;
	uint16_t	local_flags = 0;

	if (!hw || !io || !rnode || !iparam) {
		ocs_log_err(NULL, "bad parm hw=%p io=%p send=%p receive=%p rnode=%p iparam=%p\n",
			    hw, io, send, receive, rnode, iparam);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->state != OCS_HW_STATE_ACTIVE) {
		ocs_log_test(hw->os, "cannot send SRRS, HW state=%d\n", hw->state);
		return OCS_HW_RTN_ERROR;
	}

	if (ocs_hw_is_xri_port_owned(hw, io->indicator)) {
		/* We must set the XC bit for port owned XRIs */
		local_flags |= SLI4_IO_CONTINUATION;
	}
	io->rnode = rnode;
	io->type  = type;
	io->done = cb;
	io->arg  = arg;

	sge = io->sgl->virt;

	/* clear both SGE */
	ocs_memset(io->sgl->virt, 0, 2 * sizeof(sli4_sge_t));

	if (send) {
		sge[0].buffer_address_high = ocs_addr32_hi(send->phys);
		sge[0].buffer_address_low  = ocs_addr32_lo(send->phys);
		sge[0].sge_type = SLI4_SGE_TYPE_DATA;
		sge[0].buffer_length = len;
	}

	if ((OCS_HW_ELS_REQ == type) || (OCS_HW_FC_CT == type)) {
		sge[1].buffer_address_high = ocs_addr32_hi(receive->phys);
		sge[1].buffer_address_low  = ocs_addr32_lo(receive->phys);
		sge[1].sge_type = SLI4_SGE_TYPE_DATA;
		sge[1].buffer_length = receive->size;
		sge[1].last = TRUE;
	} else {
		sge[0].last = TRUE;
	}

	switch (type) {
	case OCS_HW_ELS_REQ:
		if ( (!send) || sli_els_request64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, io->sgl,
							*((uint8_t *)(send->virt)), /* req_type */
							len, receive->size,
							iparam->els.timeout, io->indicator, io->reqtag, SLI4_CQ_DEFAULT, rnode)) {
			ocs_log_err(hw->os, "REQ WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_ELS_RSP:
		if ( (!send) || sli_xmit_els_rsp64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, send, len,
					   io->indicator, io->reqtag, SLI4_CQ_DEFAULT,
					   iparam->els.ox_id,
							rnode, local_flags, UINT32_MAX)) {
			ocs_log_err(hw->os, "RSP WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_ELS_RSP_SID:
		if ( (!send) || sli_xmit_els_rsp64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, send, len,
					   io->indicator, io->reqtag, SLI4_CQ_DEFAULT,
					   iparam->els_sid.ox_id,
							rnode, local_flags, iparam->els_sid.s_id)) {
			ocs_log_err(hw->os, "RSP (SID) WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_FC_CT:
		if ( (!send) || sli_gen_request64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, io->sgl, len,
					  receive->size, iparam->fc_ct.timeout, io->indicator,
					  io->reqtag, SLI4_CQ_DEFAULT, rnode, iparam->fc_ct.r_ctl,
					  iparam->fc_ct.type, iparam->fc_ct.df_ctl)) {
			ocs_log_err(hw->os, "GEN WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_FC_CT_RSP:
		if ( (!send) || sli_xmit_sequence64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, io->sgl, len,
					  iparam->fc_ct_rsp.timeout, iparam->fc_ct_rsp.ox_id, io->indicator,
					  io->reqtag, rnode, iparam->fc_ct_rsp.r_ctl,
					  iparam->fc_ct_rsp.type, iparam->fc_ct_rsp.df_ctl)) {
			ocs_log_err(hw->os, "XMIT SEQ WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_BLS_ACC:
	case OCS_HW_BLS_RJT:
	{
		sli_bls_payload_t	bls;

		if (OCS_HW_BLS_ACC == type) {
			bls.type = SLI_BLS_ACC;
			ocs_memcpy(&bls.u.acc, iparam->bls.payload, sizeof(bls.u.acc));
		} else {
			bls.type = SLI_BLS_RJT;
			ocs_memcpy(&bls.u.rjt, iparam->bls.payload, sizeof(bls.u.rjt));
		}

		bls.ox_id = iparam->bls.ox_id;
		bls.rx_id = iparam->bls.rx_id;

		if (sli_xmit_bls_rsp64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &bls,
					   io->indicator, io->reqtag,
					   SLI4_CQ_DEFAULT,
					   rnode, UINT32_MAX)) {
			ocs_log_err(hw->os, "XMIT_BLS_RSP64 WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	}
	case OCS_HW_BLS_ACC_SID:
	{
		sli_bls_payload_t	bls;

		bls.type = SLI_BLS_ACC;
		ocs_memcpy(&bls.u.acc, iparam->bls_sid.payload, sizeof(bls.u.acc));

		bls.ox_id = iparam->bls_sid.ox_id;
		bls.rx_id = iparam->bls_sid.rx_id;

		if (sli_xmit_bls_rsp64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &bls,
					   io->indicator, io->reqtag,
					   SLI4_CQ_DEFAULT,
					   rnode, iparam->bls_sid.s_id)) {
			ocs_log_err(hw->os, "XMIT_BLS_RSP64 WQE SID error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	}
	case OCS_HW_BCAST:
		if ( (!send) || sli_xmit_bcast64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, send, len,
					iparam->bcast.timeout, io->indicator, io->reqtag,
					SLI4_CQ_DEFAULT, rnode,
					iparam->bcast.r_ctl, iparam->bcast.type, iparam->bcast.df_ctl)) {
			ocs_log_err(hw->os, "XMIT_BCAST64 WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	default:
		ocs_log_err(hw->os, "bad SRRS type %#x\n", type);
		rc = OCS_HW_RTN_ERROR;
	}

	if (OCS_HW_RTN_SUCCESS == rc) {
		if (io->wq == NULL) {
			io->wq = ocs_hw_queue_next_wq(hw, io);
			ocs_hw_assert(io->wq != NULL);
		}
		io->xbusy = TRUE;

		/*
		 * Add IO to active io wqe list before submitting, in case the
		 * wcqe processing preempts this thread.
		 */
		OCS_STAT(io->wq->use_count++);
		ocs_hw_add_io_timed_wqe(hw, io);
		rc = hw_wq_write(io->wq, &io->wqe);
		if (rc >= 0) {
			/* non-negative return is success */
			rc = 0;
		} else {
			/* failed to write wqe, remove from active wqe list */
			ocs_log_err(hw->os, "sli_queue_write failed: %d\n", rc);
			io->xbusy = FALSE;
			ocs_hw_remove_io_timed_wqe(hw, io);
		}
	}

	return rc;
}

/**
 * @ingroup io
 * @brief Send a read, write, or response IO.
 *
 * @par Description
 * This routine supports sending a higher-level IO (for example, FCP) between two endpoints
 * as a target or initiator. Examples include:
 *  - Sending read data and good response (target).
 *  - Sending a response (target with no data or after receiving write data).
 *  .
 * This routine assumes all IOs use the SGL associated with the HW IO. Prior to
 * calling this routine, the data should be loaded using ocs_hw_io_add_sge().
 *
 * @param hw Hardware context.
 * @param type Type of IO (target read, target response, and so on).
 * @param io Previously-allocated HW IO object.
 * @param len Length, in bytes, of data to send.
 * @param iparam IO parameters.
 * @param rnode Destination of data (that is, a remote node).
 * @param cb Function call upon completion of sending data (may be NULL).
 * @param arg Argument to pass to IO completion function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 *
 * @todo
 *  - Support specifiying relative offset.
 *  - Use a WQ other than 0.
 */
ocs_hw_rtn_e
ocs_hw_io_send(ocs_hw_t *hw, ocs_hw_io_type_e type, ocs_hw_io_t *io,
		uint32_t len, ocs_hw_io_param_t *iparam, ocs_remote_node_t *rnode,
		void *cb, void *arg)
{
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;
	uint32_t	rpi;
	uint8_t		send_wqe = TRUE;

	CPUTRACE("");

	if (!hw || !io || !rnode || !iparam) {
		ocs_log_err(NULL, "bad parm hw=%p io=%p iparam=%p rnode=%p\n",
			    hw, io, iparam, rnode);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->state != OCS_HW_STATE_ACTIVE) {
		ocs_log_err(hw->os, "cannot send IO, HW state=%d\n", hw->state);
		return OCS_HW_RTN_ERROR;
	}

	rpi = rnode->indicator;

	if (hw->workaround.use_unregistered_rpi && (rpi == UINT32_MAX)) {
		rpi = hw->workaround.unregistered_rid;
		ocs_log_test(hw->os, "using unregistered RPI: %d\n", rpi);
	}

	/*
	 * Save state needed during later stages
	 */
	io->rnode = rnode;
	io->type  = type;
	io->done  = cb;
	io->arg   = arg;

	/*
	 * Format the work queue entry used to send the IO
	 */
	switch (type) {
	case OCS_HW_IO_INITIATOR_READ:
		/*
		 * If use_dif_quarantine workaround is in effect, and dif_separates then mark the
		 * initiator read IO for quarantine
		 */
		if (hw->workaround.use_dif_quarantine && (hw->config.dif_mode == OCS_HW_DIF_MODE_SEPARATE) &&
		    (iparam->fcp_tgt.dif_oper != OCS_HW_DIF_OPER_DISABLED)) {
			io->quarantine = TRUE;
		}

		ocs_hw_io_ini_sge(hw, io, iparam->fcp_ini.cmnd, iparam->fcp_ini.cmnd_size,
				iparam->fcp_ini.rsp);

		if (sli_fcp_iread64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl, io->first_data_sge, len,
					io->indicator, io->reqtag, SLI4_CQ_DEFAULT, rpi, rnode,
					iparam->fcp_ini.dif_oper, iparam->fcp_ini.blk_size,
					iparam->fcp_ini.timeout)) {
			ocs_log_err(hw->os, "IREAD WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_IO_INITIATOR_WRITE:
		ocs_hw_io_ini_sge(hw, io, iparam->fcp_ini.cmnd, iparam->fcp_ini.cmnd_size,
				iparam->fcp_ini.rsp);

		if (sli_fcp_iwrite64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl, io->first_data_sge,
					 len, iparam->fcp_ini.first_burst,
					 io->indicator, io->reqtag,
					SLI4_CQ_DEFAULT, rpi, rnode,
					iparam->fcp_ini.dif_oper, iparam->fcp_ini.blk_size,
					iparam->fcp_ini.timeout)) {
			ocs_log_err(hw->os, "IWRITE WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_IO_INITIATOR_NODATA:
		ocs_hw_io_ini_sge(hw, io, iparam->fcp_ini.cmnd, iparam->fcp_ini.cmnd_size,
				iparam->fcp_ini.rsp);

		if (sli_fcp_icmnd64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl,
					io->indicator, io->reqtag, SLI4_CQ_DEFAULT,
					rpi, rnode, iparam->fcp_ini.timeout)) {
			ocs_log_err(hw->os, "ICMND WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}
		break;
	case OCS_HW_IO_TARGET_WRITE: {
		uint16_t flags = iparam->fcp_tgt.flags;
		fcp_xfer_rdy_iu_t *xfer = io->xfer_rdy.virt;

		/*
		 * Fill in the XFER_RDY for IF_TYPE 0 devices
		 */
		*((uint32_t *)xfer->fcp_data_ro) = ocs_htobe32(iparam->fcp_tgt.offset);
		*((uint32_t *)xfer->fcp_burst_len) = ocs_htobe32(len);
		*((uint32_t *)xfer->rsvd) = 0;

		if (io->xbusy) {
			flags |= SLI4_IO_CONTINUATION;
		} else {
			flags &= ~SLI4_IO_CONTINUATION;
		}

		io->tgt_wqe_timeout = iparam->fcp_tgt.timeout;

		/*
		 * If use_dif_quarantine workaround is in effect, and this is a DIF enabled IO
		 * then mark the target write IO for quarantine
		 */
		if (hw->workaround.use_dif_quarantine && (hw->config.dif_mode == OCS_HW_DIF_MODE_SEPARATE) &&
		    (iparam->fcp_tgt.dif_oper != OCS_HW_DIF_OPER_DISABLED)) {
			io->quarantine = TRUE;
		}

		/*
		 * BZ 161832 Workaround:
		 * Check for use_dif_sec_xri workaround.  Note, even though the first dataphase
		 * doesn't really need a secondary XRI, we allocate one anyway, as this avoids the
		 * potential for deadlock where all XRI's are allocated as primaries to IOs that
		 * are on hw->sec_hio_wait_list.   If this secondary XRI is not for the first
		 * data phase, it is marked for quarantine.
		 */
		if (hw->workaround.use_dif_sec_xri && (iparam->fcp_tgt.dif_oper != OCS_HW_DIF_OPER_DISABLED)) {

			/*
			 * If we have allocated a chained SGL for skyhawk, then
			 * we can re-use this for the sec_hio.
			 */
			if (io->ovfl_io != NULL) {
				io->sec_hio = io->ovfl_io;
				io->sec_hio->quarantine = TRUE;
			} else {
				io->sec_hio = ocs_hw_io_alloc(hw);
			}
			if (io->sec_hio == NULL) {
				/* Failed to allocate, so save full request context and put
				 * this IO on the wait list
				 */
				io->sec_iparam = *iparam;
				io->sec_len = len;
				ocs_lock(&hw->io_lock);
					ocs_list_remove(&hw->io_inuse,  io);
					ocs_list_add_tail(&hw->sec_hio_wait_list, io);
					io->state = OCS_HW_IO_STATE_WAIT_SEC_HIO;
					hw->sec_hio_wait_count++;
				ocs_unlock(&hw->io_lock);
				send_wqe = FALSE;
				/* Done */
				break;
			}
			/* We quarantine the secondary IO if this is the second or subsequent data phase */
			if (io->xbusy) {
				io->sec_hio->quarantine = TRUE;
			}
		}

		/*
		 * If not the first data phase, and io->sec_hio has been allocated, then issue
		 * FCP_CONT_TRECEIVE64 WQE, otherwise use the usual FCP_TRECEIVE64 WQE
		 */
		if (io->xbusy && (io->sec_hio != NULL)) {
			if (sli_fcp_cont_treceive64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl, io->first_data_sge,
						   iparam->fcp_tgt.offset, len, io->indicator, io->sec_hio->indicator,
						   io->reqtag, SLI4_CQ_DEFAULT,
						   iparam->fcp_tgt.ox_id, rpi, rnode,
						   flags,
						   iparam->fcp_tgt.dif_oper, iparam->fcp_tgt.blk_size,
						   iparam->fcp_tgt.cs_ctl, iparam->fcp_tgt.app_id)) {
				ocs_log_err(hw->os, "TRECEIVE WQE error\n");
				rc = OCS_HW_RTN_ERROR;
			}
		} else {
			if (sli_fcp_treceive64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl, io->first_data_sge,
						   iparam->fcp_tgt.offset, len, io->indicator, io->reqtag,
						   SLI4_CQ_DEFAULT,
						   iparam->fcp_tgt.ox_id, rpi, rnode,
						   flags,
						   iparam->fcp_tgt.dif_oper, iparam->fcp_tgt.blk_size,
						   iparam->fcp_tgt.cs_ctl, iparam->fcp_tgt.app_id)) {
				ocs_log_err(hw->os, "TRECEIVE WQE error\n");
				rc = OCS_HW_RTN_ERROR;
			}
		}
		break;
	}
	case OCS_HW_IO_TARGET_READ: {
		uint16_t flags = iparam->fcp_tgt.flags;

		if (io->xbusy) {
			flags |= SLI4_IO_CONTINUATION;
		} else {
			flags &= ~SLI4_IO_CONTINUATION;
		}

		io->tgt_wqe_timeout = iparam->fcp_tgt.timeout;
		if (sli_fcp_tsend64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, &io->def_sgl, io->first_data_sge,
					iparam->fcp_tgt.offset, len, io->indicator, io->reqtag,
					SLI4_CQ_DEFAULT,
					iparam->fcp_tgt.ox_id, rpi, rnode,
					flags,
					iparam->fcp_tgt.dif_oper,
					iparam->fcp_tgt.blk_size,
					iparam->fcp_tgt.cs_ctl,
					iparam->fcp_tgt.app_id)) {
			ocs_log_err(hw->os, "TSEND WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		} else if (hw->workaround.retain_tsend_io_length) {
			io->length = len;
		}
		break;
	}
	case OCS_HW_IO_TARGET_RSP: {
		uint16_t flags = iparam->fcp_tgt.flags;

		if (io->xbusy) {
			flags |= SLI4_IO_CONTINUATION;
		} else {
			flags &= ~SLI4_IO_CONTINUATION;
		}

		/* post a new auto xfer ready buffer */
		if (hw->auto_xfer_rdy_enabled && io->is_port_owned) {
			if ((io->auto_xfer_rdy_dnrx = ocs_hw_rqpair_auto_xfer_rdy_buffer_post(hw, io, 1))) {
				flags |= SLI4_IO_DNRX;
			}
		}

		io->tgt_wqe_timeout = iparam->fcp_tgt.timeout;
		if (sli_fcp_trsp64_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size,
					&io->def_sgl,
					len,
					io->indicator, io->reqtag,
					SLI4_CQ_DEFAULT,
					iparam->fcp_tgt.ox_id,
					rpi, rnode,
					flags, iparam->fcp_tgt.cs_ctl,
					io->is_port_owned,
					iparam->fcp_tgt.app_id)) {
			ocs_log_err(hw->os, "TRSP WQE error\n");
			rc = OCS_HW_RTN_ERROR;
		}

		break;
	}
	default:
		ocs_log_err(hw->os, "unsupported IO type %#x\n", type);
		rc = OCS_HW_RTN_ERROR;
	}

	if (send_wqe && (OCS_HW_RTN_SUCCESS == rc)) {
		if (io->wq == NULL) {
			io->wq = ocs_hw_queue_next_wq(hw, io);
			ocs_hw_assert(io->wq != NULL);
		}

		io->xbusy = TRUE;

		/*
		 * Add IO to active io wqe list before submitting, in case the
		 * wcqe processing preempts this thread.
		 */
		OCS_STAT(hw->tcmd_wq_submit[io->wq->instance]++);
		OCS_STAT(io->wq->use_count++);
		ocs_hw_add_io_timed_wqe(hw, io);
		rc = hw_wq_write(io->wq, &io->wqe);
		if (rc >= 0) {
			/* non-negative return is success */
			rc = 0;
		} else {
			/* failed to write wqe, remove from active wqe list */
			ocs_log_err(hw->os, "sli_queue_write failed: %d\n", rc);
			io->xbusy = FALSE;
			ocs_hw_remove_io_timed_wqe(hw, io);
		}
	}

	return rc;
}

/**
 * @brief Send a raw frame
 *
 * @par Description
 * Using the SEND_FRAME_WQE, a frame consisting of header and payload is sent.
 *
 * @param hw Pointer to HW object.
 * @param hdr Pointer to a little endian formatted FC header.
 * @param sof Value to use as the frame SOF.
 * @param eof Value to use as the frame EOF.
 * @param payload Pointer to payload DMA buffer.
 * @param ctx Pointer to caller provided send frame context.
 * @param callback Callback function.
 * @param arg Callback function argument.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
ocs_hw_rtn_e
ocs_hw_send_frame(ocs_hw_t *hw, fc_header_le_t *hdr, uint8_t sof, uint8_t eof, ocs_dma_t *payload,
		   ocs_hw_send_frame_context_t *ctx, void (*callback)(void *arg, uint8_t *cqe, int32_t status), void *arg)
{
	int32_t rc;
	ocs_hw_wqe_t *wqe;
	uint32_t xri;
	hw_wq_t *wq;

	wqe = &ctx->wqe;

	/* populate the callback object */
	ctx->hw = hw;

	/* Fetch and populate request tag */
	ctx->wqcb = ocs_hw_reqtag_alloc(hw, callback, arg);
	if (ctx->wqcb == NULL) {
		ocs_log_err(hw->os, "can't allocate request tag\n");
		return OCS_HW_RTN_NO_RESOURCES;
	}

	/* Choose a work queue, first look for a class[1] wq, otherwise just use wq[0] */
	wq = ocs_varray_iter_next(hw->wq_class_array[1]);
	if (wq == NULL) {
		wq = hw->hw_wq[0];
	}

	/* Set XRI and RX_ID in the header based on which WQ, and which send_frame_io we are using */
	xri = wq->send_frame_io->indicator;

	/* Build the send frame WQE */
	rc = sli_send_frame_wqe(&hw->sli, wqe->wqebuf, hw->sli.config.wqe_size, sof, eof, (uint32_t*) hdr, payload,
				payload->len, OCS_HW_SEND_FRAME_TIMEOUT, xri, ctx->wqcb->instance_index);
	if (rc) {
		ocs_log_err(hw->os, "sli_send_frame_wqe failed: %d\n", rc);
		return OCS_HW_RTN_ERROR;
	}

	/* Write to WQ */
	rc = hw_wq_write(wq, wqe);
	if (rc) {
		ocs_log_err(hw->os, "hw_wq_write failed: %d\n", rc);
		return OCS_HW_RTN_ERROR;
	}

	OCS_STAT(wq->use_count++);

	return rc ? OCS_HW_RTN_ERROR : OCS_HW_RTN_SUCCESS;
}

ocs_hw_rtn_e
ocs_hw_io_register_sgl(ocs_hw_t *hw, ocs_hw_io_t *io, ocs_dma_t *sgl, uint32_t sgl_count)
{
	if (sli_get_sgl_preregister(&hw->sli)) {
		ocs_log_err(hw->os, "can't use temporary SGL with pre-registered SGLs\n");
		return OCS_HW_RTN_ERROR;
	}
	io->ovfl_sgl = sgl;
	io->ovfl_sgl_count = sgl_count;
	io->ovfl_io = NULL;

	return OCS_HW_RTN_SUCCESS;
}

static void
ocs_hw_io_restore_sgl(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	/* Restore the default */
	io->sgl = &io->def_sgl;
	io->sgl_count = io->def_sgl_count;

	/*
	 * For skyhawk, we need to free the IO allocated for the chained
	 * SGL. For all devices, clear the overflow fields on the IO.
	 *
	 * Note: For DIF IOs, we may be using the same XRI for the sec_hio and
	 *       the chained SGLs. If so, then we clear the ovfl_io field
	 *       when the sec_hio is freed.
	 */
	if (io->ovfl_io != NULL) {
		ocs_hw_io_free(hw, io->ovfl_io);
		io->ovfl_io = NULL;
	}

	/* Clear the overflow SGL */
	io->ovfl_sgl = NULL;
	io->ovfl_sgl_count = 0;
	io->ovfl_lsp = NULL;
}

/**
 * @ingroup io
 * @brief Initialize the scatter gather list entries of an IO.
 *
 * @param hw Hardware context.
 * @param io Previously-allocated HW IO object.
 * @param type Type of IO (target read, target response, and so on).
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_io_init_sges(ocs_hw_t *hw, ocs_hw_io_t *io, ocs_hw_io_type_e type)
{
	sli4_sge_t	*data = NULL;
	uint32_t	i = 0;
	uint32_t	skips = 0;

	if (!hw || !io) {
		ocs_log_err(hw ? hw->os : NULL, "bad parameter hw=%p io=%p\n",
			    hw, io);
		return OCS_HW_RTN_ERROR;
	}

	/* Clear / reset the scatter-gather list */
	io->sgl = &io->def_sgl;
	io->sgl_count = io->def_sgl_count;
	io->first_data_sge = 0;

	ocs_memset(io->sgl->virt, 0, 2 * sizeof(sli4_sge_t));
	io->n_sge = 0;
	io->sge_offset = 0;

	io->type = type;

	data = io->sgl->virt;

	/*
	 * Some IO types have underlying hardware requirements on the order
	 * of SGEs. Process all special entries here.
	 */
	switch (type) {
	case OCS_HW_IO_INITIATOR_READ:
	case OCS_HW_IO_INITIATOR_WRITE:
	case OCS_HW_IO_INITIATOR_NODATA:
		/*
		 * No skips, 2 special for initiator I/Os
		 * The addresses and length are written later
		 */
		/* setup command pointer */
		data->sge_type = SLI4_SGE_TYPE_DATA;
		data++;

		/* setup response pointer */
		data->sge_type = SLI4_SGE_TYPE_DATA;

		if (OCS_HW_IO_INITIATOR_NODATA == type) {
			data->last = TRUE;
		}
		data++;

		io->n_sge = 2;
		break;
	case OCS_HW_IO_TARGET_WRITE:
#define OCS_TARGET_WRITE_SKIPS	2
		skips = OCS_TARGET_WRITE_SKIPS;

		/* populate host resident XFER_RDY buffer */
		data->sge_type = SLI4_SGE_TYPE_DATA;
		data->buffer_address_high = ocs_addr32_hi(io->xfer_rdy.phys);
		data->buffer_address_low  = ocs_addr32_lo(io->xfer_rdy.phys);
		data->buffer_length = io->xfer_rdy.size;
		data++;

		skips--;

		io->n_sge = 1;
		break;
	case OCS_HW_IO_TARGET_READ:
		/*
		 * For FCP_TSEND64, the first 2 entries are SKIP SGE's
		 */
#define OCS_TARGET_READ_SKIPS	2
		skips = OCS_TARGET_READ_SKIPS;
		break;
	case OCS_HW_IO_TARGET_RSP:
		/*
		 * No skips, etc. for FCP_TRSP64
		 */
		break;
	default:
		ocs_log_err(hw->os, "unsupported IO type %#x\n", type);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Write skip entries
	 */
	for (i = 0; i < skips; i++) {
		data->sge_type = SLI4_SGE_TYPE_SKIP;
		data++;
	}

	io->n_sge += skips;

	/*
	 * Set last
	 */
	data->last = TRUE;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup io
 * @brief Add a T10 PI seed scatter gather list entry.
 *
 * @param hw Hardware context.
 * @param io Previously-allocated HW IO object.
 * @param dif_info Pointer to T10 DIF fields, or NULL if no DIF.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_io_add_seed_sge(ocs_hw_t *hw, ocs_hw_io_t *io, ocs_hw_dif_info_t *dif_info)
{
	sli4_sge_t	*data = NULL;
	sli4_diseed_sge_t *dif_seed;

	/* If no dif_info, or dif_oper is disabled, then just return success */
	if ((dif_info == NULL) || (dif_info->dif_oper == OCS_HW_DIF_OPER_DISABLED)) {
		return OCS_HW_RTN_SUCCESS;
	}

	if (!hw || !io) {
		ocs_log_err(hw ? hw->os : NULL, "bad parameter hw=%p io=%p dif_info=%p\n",
			    hw, io, dif_info);
		return OCS_HW_RTN_ERROR;
	}

	data = io->sgl->virt;
	data += io->n_sge;

	/* If we are doing T10 DIF add the DIF Seed SGE */
	ocs_memset(data, 0, sizeof(sli4_diseed_sge_t));
	dif_seed = (sli4_diseed_sge_t *)data;
	dif_seed->ref_tag_cmp = dif_info->ref_tag_cmp;
	dif_seed->ref_tag_repl = dif_info->ref_tag_repl;
	dif_seed->app_tag_repl = dif_info->app_tag_repl;
	dif_seed->repl_app_tag = dif_info->repl_app_tag;
	if (SLI4_IF_TYPE_LANCER_FC_ETH != hw->sli.if_type) {
		dif_seed->atrt = dif_info->disable_app_ref_ffff;
		dif_seed->at = dif_info->disable_app_ffff;
	}
	dif_seed->sge_type = SLI4_SGE_TYPE_DISEED;
	/* Workaround for SKH (BZ157233) */
	if (((io->type == OCS_HW_IO_TARGET_WRITE) || (io->type == OCS_HW_IO_INITIATOR_READ)) &&
		(SLI4_IF_TYPE_LANCER_FC_ETH != hw->sli.if_type) && dif_info->dif_separate) {
		dif_seed->sge_type = SLI4_SGE_TYPE_SKIP;
	}

	dif_seed->app_tag_cmp = dif_info->app_tag_cmp;
	dif_seed->dif_blk_size = dif_info->blk_size;
	dif_seed->auto_incr_ref_tag = dif_info->auto_incr_ref_tag;
	dif_seed->check_app_tag = dif_info->check_app_tag;
	dif_seed->check_ref_tag = dif_info->check_ref_tag;
	dif_seed->check_crc = dif_info->check_guard;
	dif_seed->new_ref_tag = dif_info->repl_ref_tag;

	switch(dif_info->dif_oper) {
	case OCS_HW_SGE_DIF_OP_IN_NODIF_OUT_CRC:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_NODIF_OUT_CRC;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_NODIF_OUT_CRC;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CRC_OUT_NODIF:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CRC_OUT_NODIF;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CRC_OUT_NODIF;
		break;
	case OCS_HW_SGE_DIF_OP_IN_NODIF_OUT_CHKSUM:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_NODIF_OUT_CHKSUM;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_NODIF_OUT_CHKSUM;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CHKSUM_OUT_NODIF:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_NODIF;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_NODIF;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CRC_OUT_CRC:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CRC_OUT_CRC;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CRC_OUT_CRC;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CHKSUM_OUT_CHKSUM:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_CHKSUM;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_CHKSUM;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CRC_OUT_CHKSUM:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CRC_OUT_CHKSUM;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CRC_OUT_CHKSUM;
		break;
	case OCS_HW_SGE_DIF_OP_IN_CHKSUM_OUT_CRC:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_CRC;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_CHKSUM_OUT_CRC;
		break;
	case OCS_HW_SGE_DIF_OP_IN_RAW_OUT_RAW:
		dif_seed->dif_op_rx = SLI4_SGE_DIF_OP_IN_RAW_OUT_RAW;
		dif_seed->dif_op_tx = SLI4_SGE_DIF_OP_IN_RAW_OUT_RAW;
		break;
	default:
		ocs_log_err(hw->os, "unsupported DIF operation %#x\n",
			    dif_info->dif_oper);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Set last, clear previous last
	 */
	data->last = TRUE;
	if (io->n_sge) {
		data[-1].last = FALSE;
	}

	io->n_sge++;

	return OCS_HW_RTN_SUCCESS;
}

static ocs_hw_rtn_e
ocs_hw_io_overflow_sgl(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	sli4_lsp_sge_t *lsp;

	/* fail if we're already pointing to the overflow SGL */
	if (io->sgl == io->ovfl_sgl) {
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * For skyhawk, we can use another SGL to extend the SGL list. The
	 * Chained entry must not be in the first 4 entries.
	 *
	 * Note: For DIF enabled IOs, we will use the ovfl_io for the sec_hio.
	 */
	if (sli_get_sgl_preregister(&hw->sli) &&
	    io->def_sgl_count > 4 &&
	    io->ovfl_io == NULL &&
	    ((SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) ||
		(SLI4_IF_TYPE_BE3_SKH_VF == sli_get_if_type(&hw->sli)))) {
		io->ovfl_io = ocs_hw_io_alloc(hw);
		if (io->ovfl_io != NULL) {
			/*
			 * Note: We can't call ocs_hw_io_register_sgl() here
			 * because it checks that SGLs are not pre-registered
			 * and for shyhawk, preregistered SGLs are required.
			 */
			io->ovfl_sgl = &io->ovfl_io->def_sgl;
			io->ovfl_sgl_count = io->ovfl_io->def_sgl_count;
		}
	}

	/* fail if we don't have an overflow SGL registered */
	if (io->ovfl_sgl == NULL) {
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * Overflow, we need to put a link SGE in the last location of the current SGL, after
	 * copying the the last SGE to the overflow SGL
	 */

	((sli4_sge_t*)io->ovfl_sgl->virt)[0] = ((sli4_sge_t*)io->sgl->virt)[io->n_sge - 1];

	lsp = &((sli4_lsp_sge_t*)io->sgl->virt)[io->n_sge - 1];
	ocs_memset(lsp, 0, sizeof(*lsp));

	if ((SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) ||
	    (SLI4_IF_TYPE_BE3_SKH_VF == sli_get_if_type(&hw->sli))) {
		sli_skh_chain_sge_build(&hw->sli,
					(sli4_sge_t*)lsp,
					io->ovfl_io->indicator,
					0, /* frag_num */
					0); /* offset */
	} else {
		lsp->buffer_address_high = ocs_addr32_hi(io->ovfl_sgl->phys);
		lsp->buffer_address_low  = ocs_addr32_lo(io->ovfl_sgl->phys);
		lsp->sge_type = SLI4_SGE_TYPE_LSP;
		lsp->last = 0;
		io->ovfl_lsp = lsp;
		io->ovfl_lsp->segment_length = sizeof(sli4_sge_t);
	}

	/* Update the current SGL pointer, and n_sgl */
	io->sgl = io->ovfl_sgl;
	io->sgl_count = io->ovfl_sgl_count;
	io->n_sge = 1;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup io
 * @brief Add a scatter gather list entry to an IO.
 *
 * @param hw Hardware context.
 * @param io Previously-allocated HW IO object.
 * @param addr Physical address.
 * @param length Length of memory pointed to by @c addr.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_io_add_sge(ocs_hw_t *hw, ocs_hw_io_t *io, uintptr_t addr, uint32_t length)
{
	sli4_sge_t	*data = NULL;

	if (!hw || !io || !addr || !length) {
		ocs_log_err(hw ? hw->os : NULL,
			    "bad parameter hw=%p io=%p addr=%lx length=%u\n",
			    hw, io, addr, length);
		return OCS_HW_RTN_ERROR;
	}

	if ((length != 0) && (io->n_sge + 1) > io->sgl_count) {
		if (ocs_hw_io_overflow_sgl(hw, io) != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "SGL full (%d)\n", io->n_sge);
			return OCS_HW_RTN_ERROR;
		}
	}

	if (length > sli_get_max_sge(&hw->sli)) {
		ocs_log_err(hw->os, "length of SGE %d bigger than allowed %d\n",
			    length, sli_get_max_sge(&hw->sli));
		return OCS_HW_RTN_ERROR;
	}

	data = io->sgl->virt;
	data += io->n_sge;

	data->sge_type = SLI4_SGE_TYPE_DATA;
	data->buffer_address_high = ocs_addr32_hi(addr);
	data->buffer_address_low  = ocs_addr32_lo(addr);
	data->buffer_length = length;
	data->data_offset = io->sge_offset;
	/*
	 * Always assume this is the last entry and mark as such.
	 * If this is not the first entry unset the "last SGE"
	 * indication for the previous entry
	 */
	data->last = TRUE;
	if (io->n_sge) {
		data[-1].last = FALSE;
	}

	/* Set first_data_bde if not previously set */
	if (io->first_data_sge == 0) {
		io->first_data_sge = io->n_sge;
	}

	io->sge_offset += length;
	io->n_sge++;

	/* Update the linked segment length (only executed after overflow has begun) */
	if (io->ovfl_lsp != NULL) {
		io->ovfl_lsp->segment_length = io->n_sge * sizeof(sli4_sge_t);
	}

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup io
 * @brief Add a T10 DIF scatter gather list entry to an IO.
 *
 * @param hw Hardware context.
 * @param io Previously-allocated HW IO object.
 * @param addr DIF physical address.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_io_add_dif_sge(ocs_hw_t *hw, ocs_hw_io_t *io, uintptr_t addr)
{
	sli4_dif_sge_t	*data = NULL;

	if (!hw || !io || !addr) {
		ocs_log_err(hw ? hw->os : NULL,
			    "bad parameter hw=%p io=%p addr=%lx\n",
			    hw, io, addr);
		return OCS_HW_RTN_ERROR;
	}

	if ((io->n_sge + 1) > hw->config.n_sgl) {
		if (ocs_hw_io_overflow_sgl(hw, io) != OCS_HW_RTN_ERROR) {
			ocs_log_err(hw->os, "SGL full (%d)\n", io->n_sge);
			return OCS_HW_RTN_ERROR;
		}
	}

	data = io->sgl->virt;
	data += io->n_sge;

	data->sge_type = SLI4_SGE_TYPE_DIF;
	/* Workaround for SKH (BZ157233) */
	if (((io->type == OCS_HW_IO_TARGET_WRITE) || (io->type == OCS_HW_IO_INITIATOR_READ)) &&
		(SLI4_IF_TYPE_LANCER_FC_ETH != hw->sli.if_type)) {
		data->sge_type = SLI4_SGE_TYPE_SKIP;
	}

	data->buffer_address_high = ocs_addr32_hi(addr);
	data->buffer_address_low  = ocs_addr32_lo(addr);

	/*
	 * Always assume this is the last entry and mark as such.
	 * If this is not the first entry unset the "last SGE"
	 * indication for the previous entry
	 */
	data->last = TRUE;
	if (io->n_sge) {
		data[-1].last = FALSE;
	}

	io->n_sge++;

	return OCS_HW_RTN_SUCCESS;
}

/**
 * @ingroup io
 * @brief Abort a previously-started IO.
 *
 * @param hw Hardware context.
 * @param io_to_abort The IO to abort.
 * @param send_abts Boolean to have the hardware automatically
 * generate an ABTS.
 * @param cb Function call upon completion of the abort (may be NULL).
 * @param arg Argument to pass to abort completion function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_io_abort(ocs_hw_t *hw, ocs_hw_io_t *io_to_abort, uint32_t send_abts, void *cb, void *arg)
{
	sli4_abort_type_e atype = SLI_ABORT_MAX;
	uint32_t	id = 0, mask = 0;
	ocs_hw_rtn_e	rc = OCS_HW_RTN_SUCCESS;
	hw_wq_callback_t *wqcb;

	if (!hw || !io_to_abort) {
		ocs_log_err(hw ? hw->os : NULL,
			    "bad parameter hw=%p io=%p\n",
			    hw, io_to_abort);
		return OCS_HW_RTN_ERROR;
	}

	if (hw->state != OCS_HW_STATE_ACTIVE) {
		ocs_log_err(hw->os, "cannot send IO abort, HW state=%d\n",
			    hw->state);
		return OCS_HW_RTN_ERROR;
	}

	/* take a reference on IO being aborted */
	if (ocs_ref_get_unless_zero(&io_to_abort->ref) == 0) {
		/* command no longer active */
		ocs_log_test(hw ? hw->os : NULL,
				"io not active xri=0x%x tag=0x%x\n",
				io_to_abort->indicator, io_to_abort->reqtag);
		return OCS_HW_RTN_IO_NOT_ACTIVE;
	}

	/* non-port owned XRI checks */
	/* Must have a valid WQ reference */
	if (io_to_abort->wq == NULL) {
		ocs_log_test(hw->os, "io_to_abort xri=0x%x not active on WQ\n",
				io_to_abort->indicator);
		ocs_ref_put(&io_to_abort->ref); /* ocs_ref_get(): same function */
		return OCS_HW_RTN_IO_NOT_ACTIVE;
	}

	/* Validation checks complete; now check to see if already being aborted */
	ocs_lock(&hw->io_abort_lock);
		if (io_to_abort->abort_in_progress) {
			ocs_unlock(&hw->io_abort_lock);
			ocs_ref_put(&io_to_abort->ref); /* ocs_ref_get(): same function */
			ocs_log_debug(hw ? hw->os : NULL,
				"io already being aborted xri=0x%x tag=0x%x\n",
				io_to_abort->indicator, io_to_abort->reqtag);
			return OCS_HW_RTN_IO_ABORT_IN_PROGRESS;
		}

		/*
		 * This IO is not already being aborted. Set flag so we won't try to
		 * abort it again. After all, we only have one abort_done callback.
		 */
		io_to_abort->abort_in_progress = 1;
	ocs_unlock(&hw->io_abort_lock);

	/*
	 * If we got here, the possibilities are:
	 * - host owned xri
	 *	- io_to_abort->wq_index != UINT32_MAX
	 *		- submit ABORT_WQE to same WQ
	 * - port owned xri:
	 *	- rxri: io_to_abort->wq_index == UINT32_MAX
	 *		- submit ABORT_WQE to any WQ
	 *	- non-rxri
	 *		- io_to_abort->index != UINT32_MAX
	 *			- submit ABORT_WQE to same WQ
	 *		- io_to_abort->index == UINT32_MAX
	 *			- submit ABORT_WQE to any WQ
	 */
	io_to_abort->abort_done = cb;
	io_to_abort->abort_arg  = arg;

	atype = SLI_ABORT_XRI;
	id = io_to_abort->indicator;

	/* Allocate a request tag for the abort portion of this IO */
	wqcb = ocs_hw_reqtag_alloc(hw, ocs_hw_wq_process_abort, io_to_abort);
	if (wqcb == NULL) {
		ocs_log_err(hw->os, "can't allocate request tag\n");
		return OCS_HW_RTN_NO_RESOURCES;
	}
	io_to_abort->abort_reqtag = wqcb->instance_index;

	/*
	 * If the wqe is on the pending list, then set this wqe to be
	 * aborted when the IO's wqe is removed from the list.
	 */
	if (io_to_abort->wq != NULL) {
		sli_queue_lock(io_to_abort->wq->queue);
			if (ocs_list_on_list(&io_to_abort->wqe.link)) {
				io_to_abort->wqe.abort_wqe_submit_needed = 1;
				io_to_abort->wqe.send_abts = send_abts;
				io_to_abort->wqe.id = id;
				io_to_abort->wqe.abort_reqtag = io_to_abort->abort_reqtag;
				sli_queue_unlock(io_to_abort->wq->queue);
				return 0;
		}
		sli_queue_unlock(io_to_abort->wq->queue);
	}

	if (sli_abort_wqe(&hw->sli, io_to_abort->wqe.wqebuf, hw->sli.config.wqe_size, atype, send_abts, id, mask,
			  io_to_abort->abort_reqtag, SLI4_CQ_DEFAULT)) {
		ocs_log_err(hw->os, "ABORT WQE error\n");
		io_to_abort->abort_reqtag = UINT32_MAX;
		ocs_hw_reqtag_free(hw, wqcb);
		rc = OCS_HW_RTN_ERROR;
	}

	if (OCS_HW_RTN_SUCCESS == rc) {
		if (io_to_abort->wq == NULL) {
			io_to_abort->wq = ocs_hw_queue_next_wq(hw, io_to_abort);
			ocs_hw_assert(io_to_abort->wq != NULL);
		}
		/* ABORT_WQE does not actually utilize an XRI on the Port,
		 * therefore, keep xbusy as-is to track the exchange's state,
		 * not the ABORT_WQE's state
		 */
		rc = hw_wq_write(io_to_abort->wq, &io_to_abort->wqe);
		if (rc > 0) {
			/* non-negative return is success */
			rc = 0;
			/* can't abort an abort so skip adding to timed wqe list */
		}
	}

	if (OCS_HW_RTN_SUCCESS != rc) {
		ocs_lock(&hw->io_abort_lock);
			io_to_abort->abort_in_progress = 0;
		ocs_unlock(&hw->io_abort_lock);
		ocs_ref_put(&io_to_abort->ref); /* ocs_ref_get(): same function */
	}
	return rc;
}

/**
 * @ingroup io
 * @brief Return the OX_ID/RX_ID of the IO.
 *
 * @param hw Hardware context.
 * @param io HW IO object.
 *
 * @return Returns X_ID on success, or -1 on failure.
 */
int32_t
ocs_hw_io_get_xid(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	if (!hw || !io) {
		ocs_log_err(hw ? hw->os : NULL,
			    "bad parameter hw=%p io=%p\n", hw, io);
		return -1;
	}

	return io->indicator;
}


typedef struct ocs_hw_fw_write_cb_arg {
	ocs_hw_fw_cb_t cb;
	void *arg;
} ocs_hw_fw_write_cb_arg_t;

typedef struct ocs_hw_sfp_cb_arg {
	ocs_hw_sfp_cb_t cb;
	void *arg;
	ocs_dma_t payload;
} ocs_hw_sfp_cb_arg_t;

typedef struct ocs_hw_temp_cb_arg {
	ocs_hw_temp_cb_t cb;
	void *arg;
} ocs_hw_temp_cb_arg_t;

typedef struct ocs_hw_link_stat_cb_arg {
	ocs_hw_link_stat_cb_t cb;
	void *arg;
} ocs_hw_link_stat_cb_arg_t;

typedef struct ocs_hw_host_stat_cb_arg {
	ocs_hw_host_stat_cb_t cb;
	void *arg;
} ocs_hw_host_stat_cb_arg_t;

typedef struct ocs_hw_dump_get_cb_arg {
	ocs_hw_dump_get_cb_t cb;
	void *arg;
	void *mbox_cmd;
} ocs_hw_dump_get_cb_arg_t;

typedef struct ocs_hw_dump_clear_cb_arg {
	ocs_hw_dump_clear_cb_t cb;
	void *arg;
	void *mbox_cmd;
} ocs_hw_dump_clear_cb_arg_t;

/**
 * @brief Write a portion of a firmware image to the device.
 *
 * @par Description
 * Calls the correct firmware write function based on the device type.
 *
 * @param hw Hardware context.
 * @param dma DMA structure containing the firmware image chunk.
 * @param size Size of the firmware image chunk.
 * @param offset Offset, in bytes, from the beginning of the firmware image.
 * @param last True if this is the last chunk of the image.
 * Causes the image to be committed to flash.
 * @param cb Pointer to a callback function that is called when the command completes.
 * The callback function prototype is
 * <tt>void cb(int32_t status, uint32_t bytes_written, void *arg)</tt>.
 * @param arg Pointer to be passed to the callback function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_firmware_write(ocs_hw_t *hw, ocs_dma_t *dma, uint32_t size, uint32_t offset, int last, ocs_hw_fw_cb_t cb, void *arg)
{
	if (hw->sli.if_type == SLI4_IF_TYPE_LANCER_FC_ETH) {
		return ocs_hw_firmware_write_lancer(hw, dma, size, offset, last, cb, arg);
	} else {
		/* Write firmware_write for BE3/Skyhawk not supported */
		return -1;
	}
}

/**
 * @brief Write a portion of a firmware image to the Emulex XE201 ASIC (Lancer).
 *
 * @par Description
 * Creates a SLI_CONFIG mailbox command, fills it with the correct values to write a
 * firmware image chunk, and then sends the command with ocs_hw_command(). On completion,
 * the callback function ocs_hw_fw_write_cb() gets called to free the mailbox
 * and to signal the caller that the write has completed.
 *
 * @param hw Hardware context.
 * @param dma DMA structure containing the firmware image chunk.
 * @param size Size of the firmware image chunk.
 * @param offset Offset, in bytes, from the beginning of the firmware image.
 * @param last True if this is the last chunk of the image. Causes the image to be committed to flash.
 * @param cb Pointer to a callback function that is called when the command completes.
 * The callback function prototype is
 * <tt>void cb(int32_t status, uint32_t bytes_written, void *arg)</tt>.
 * @param arg Pointer to be passed to the callback function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_firmware_write_lancer(ocs_hw_t *hw, ocs_dma_t *dma, uint32_t size, uint32_t offset, int last, ocs_hw_fw_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	uint8_t *mbxdata;
	ocs_hw_fw_write_cb_arg_t *cb_arg;
	int noc=0;	/* No Commit bit - set to 1 for testing */

	if (SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) {
		ocs_log_test(hw->os, "Function only supported for I/F type 2\n");
		return OCS_HW_RTN_ERROR;
	}

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_fw_write_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;

	if (sli_cmd_common_write_object(&hw->sli, mbxdata, SLI4_BMBX_SIZE, noc, last,
			size, offset, "/prg/", dma)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_cb_fw_write, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "COMMON_WRITE_OBJECT failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_fw_write_cb_arg_t));
	}

	return rc;

}

/**
 * @brief Called when the WRITE OBJECT command completes.
 *
 * @par Description
 * Get the number of bytes actually written out of the response, free the mailbox
 * that was malloc'd by ocs_hw_firmware_write(),
 * then call the callback and pass the status and bytes written.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is <tt>void cb(int32_t status, uint32_t bytes_written)</tt>.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_fw_write(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{

	sli4_cmd_sli_config_t* mbox_rsp = (sli4_cmd_sli_config_t*) mqe;
	sli4_res_common_write_object_t* wr_obj_rsp = (sli4_res_common_write_object_t*) &(mbox_rsp->payload.embed);
	ocs_hw_fw_write_cb_arg_t *cb_arg = arg;
	uint32_t bytes_written;
	uint16_t mbox_status;
	uint32_t change_status;

	bytes_written = wr_obj_rsp->actual_write_length;
	mbox_status = mbox_rsp->hdr.status;
	change_status = wr_obj_rsp->change_status;

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_status) {
				status = mbox_status;
			}
			cb_arg->cb(status, bytes_written, change_status, cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_fw_write_cb_arg_t));
	}

	return 0;

}

/**
 * @brief Called when the READ_TRANSCEIVER_DATA command completes.
 *
 * @par Description
 * Get the number of bytes read out of the response, free the mailbox that was malloc'd
 * by ocs_hw_get_sfp(), then call the callback and pass the status and bytes written.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is
 * <tt>void cb(int32_t status, uint32_t bytes_written, uint32_t *data, void *arg)</tt>.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_sfp(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{

	ocs_hw_sfp_cb_arg_t *cb_arg = arg;
	ocs_dma_t *payload = NULL;
	sli4_res_common_read_transceiver_data_t* mbox_rsp = NULL;
	uint32_t bytes_written;

	if (cb_arg) {
		payload = &(cb_arg->payload);
		if (cb_arg->cb) {
			mbox_rsp = (sli4_res_common_read_transceiver_data_t*) payload->virt;
			bytes_written = mbox_rsp->hdr.response_length;
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(hw->os, status, bytes_written, mbox_rsp->page_data, cb_arg->arg);
		}

		ocs_dma_free(hw->os, &cb_arg->payload);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_sfp_cb_arg_t));
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	return 0;
}

/**
 * @ingroup io
 * @brief Function to retrieve the SFP information.
 *
 * @param hw Hardware context.
 * @param page The page of SFP data to retrieve (0xa0 or 0xa2).
 * @param cb Function call upon completion of sending the data (may be NULL).
 * @param arg Argument to pass to IO completion function.
 *
 * @return Returns OCS_HW_RTN_SUCCESS, OCS_HW_RTN_ERROR, or OCS_HW_RTN_NO_MEMORY.
 */
ocs_hw_rtn_e
ocs_hw_get_sfp(ocs_hw_t *hw, uint16_t page, ocs_hw_sfp_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	ocs_hw_sfp_cb_arg_t *cb_arg;
	uint8_t *mbxdata;

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_sfp_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;

	/* payload holds the non-embedded portion */
	if (ocs_dma_alloc(hw->os, &cb_arg->payload, sizeof(sli4_res_common_read_transceiver_data_t),
			  OCS_MIN_DMA_ALIGNMENT)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_sfp_cb_arg_t));
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* Send the HW command */
	if (sli_cmd_common_read_transceiver_data(&hw->sli, mbxdata, SLI4_BMBX_SIZE, page,
	    &cb_arg->payload)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_cb_sfp, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "READ_TRANSCEIVER_DATA failed with status %d\n",
				rc);
		ocs_dma_free(hw->os, &cb_arg->payload);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_sfp_cb_arg_t));
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
	}

	return rc;
}

/**
 * @brief Function to retrieve the temperature information.
 *
 * @param hw Hardware context.
 * @param cb Function call upon completion of sending the data (may be NULL).
 * @param arg Argument to pass to IO completion function.
 *
 * @return Returns OCS_HW_RTN_SUCCESS, OCS_HW_RTN_ERROR, or OCS_HW_RTN_NO_MEMORY.
 */
ocs_hw_rtn_e
ocs_hw_get_temperature(ocs_hw_t *hw, ocs_hw_temp_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	ocs_hw_temp_cb_arg_t *cb_arg;
	uint8_t *mbxdata;

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_temp_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;

	if (sli_cmd_dump_type4(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
				SLI4_WKI_TAG_SAT_TEM)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_cb_temp, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "DUMP_TYPE4 failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_temp_cb_arg_t));
	}

	return rc;
}

/**
 * @brief Called when the DUMP command completes.
 *
 * @par Description
 * Get the temperature data out of the response, free the mailbox that was malloc'd
 * by ocs_hw_get_temperature(), then call the callback and pass the status and data.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is defined by ocs_hw_temp_cb_t.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_temp(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{

	sli4_cmd_dump4_t* mbox_rsp = (sli4_cmd_dump4_t*) mqe;
	ocs_hw_temp_cb_arg_t *cb_arg = arg;
	uint32_t curr_temp = mbox_rsp->resp_data[0]; /* word 5 */
	uint32_t crit_temp_thrshld = mbox_rsp->resp_data[1]; /* word 6*/
	uint32_t warn_temp_thrshld = mbox_rsp->resp_data[2]; /* word 7 */
	uint32_t norm_temp_thrshld = mbox_rsp->resp_data[3]; /* word 8 */
	uint32_t fan_off_thrshld = mbox_rsp->resp_data[4];   /* word 9 */
	uint32_t fan_on_thrshld = mbox_rsp->resp_data[5];    /* word 10 */

	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(status,
				   curr_temp,
				   crit_temp_thrshld,
				   warn_temp_thrshld,
				   norm_temp_thrshld,
				   fan_off_thrshld,
				   fan_on_thrshld,
				   cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_temp_cb_arg_t));
	}
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return 0;
}

/**
 * @brief Function to retrieve the link statistics.
 *
 * @param hw Hardware context.
 * @param req_ext_counters If TRUE, then the extended counters will be requested.
 * @param clear_overflow_flags If TRUE, then overflow flags will be cleared.
 * @param clear_all_counters If TRUE, the counters will be cleared.
 * @param cb Function call upon completion of sending the data (may be NULL).
 * @param arg Argument to pass to IO completion function.
 *
 * @return Returns OCS_HW_RTN_SUCCESS, OCS_HW_RTN_ERROR, or OCS_HW_RTN_NO_MEMORY.
 */
ocs_hw_rtn_e
ocs_hw_get_link_stats(ocs_hw_t *hw,
			uint8_t req_ext_counters,
			uint8_t clear_overflow_flags,
			uint8_t clear_all_counters,
			ocs_hw_link_stat_cb_t cb,
			void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	ocs_hw_link_stat_cb_arg_t *cb_arg;
	uint8_t *mbxdata;

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_link_stat_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;

	if (sli_cmd_read_link_stats(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
				    req_ext_counters,
				    clear_overflow_flags,
				    clear_all_counters)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_cb_link_stat, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "READ_LINK_STATS failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_link_stat_cb_arg_t));
	}

	return rc;
}

/**
 * @brief Called when the READ_LINK_STAT command completes.
 *
 * @par Description
 * Get the counters out of the response, free the mailbox that was malloc'd
 * by ocs_hw_get_link_stats(), then call the callback and pass the status and data.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is defined by ocs_hw_link_stat_cb_t.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_link_stat(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{

	sli4_cmd_read_link_stats_t* mbox_rsp = (sli4_cmd_read_link_stats_t*) mqe;
	ocs_hw_link_stat_cb_arg_t *cb_arg = arg;
	ocs_hw_link_stat_counts_t counts[OCS_HW_LINK_STAT_MAX];
	uint32_t num_counters = (mbox_rsp->gec ? 20 : 13);

	ocs_memset(counts, 0, sizeof(ocs_hw_link_stat_counts_t) *
		   OCS_HW_LINK_STAT_MAX);

	counts[OCS_HW_LINK_STAT_LINK_FAILURE_COUNT].overflow = mbox_rsp->w02of;
	counts[OCS_HW_LINK_STAT_LOSS_OF_SYNC_COUNT].overflow = mbox_rsp->w03of;
	counts[OCS_HW_LINK_STAT_LOSS_OF_SIGNAL_COUNT].overflow = mbox_rsp->w04of;
	counts[OCS_HW_LINK_STAT_PRIMITIVE_SEQ_COUNT].overflow = mbox_rsp->w05of;
	counts[OCS_HW_LINK_STAT_INVALID_XMIT_WORD_COUNT].overflow = mbox_rsp->w06of;
	counts[OCS_HW_LINK_STAT_CRC_COUNT].overflow = mbox_rsp->w07of;
	counts[OCS_HW_LINK_STAT_PRIMITIVE_SEQ_TIMEOUT_COUNT].overflow = mbox_rsp->w08of;
	counts[OCS_HW_LINK_STAT_ELASTIC_BUFFER_OVERRUN_COUNT].overflow = mbox_rsp->w09of;
	counts[OCS_HW_LINK_STAT_ARB_TIMEOUT_COUNT].overflow = mbox_rsp->w10of;
	counts[OCS_HW_LINK_STAT_ADVERTISED_RCV_B2B_CREDIT].overflow = mbox_rsp->w11of;
	counts[OCS_HW_LINK_STAT_CURR_RCV_B2B_CREDIT].overflow = mbox_rsp->w12of;
	counts[OCS_HW_LINK_STAT_ADVERTISED_XMIT_B2B_CREDIT].overflow = mbox_rsp->w13of;
	counts[OCS_HW_LINK_STAT_CURR_XMIT_B2B_CREDIT].overflow = mbox_rsp->w14of;
	counts[OCS_HW_LINK_STAT_RCV_EOFA_COUNT].overflow = mbox_rsp->w15of;
	counts[OCS_HW_LINK_STAT_RCV_EOFDTI_COUNT].overflow = mbox_rsp->w16of;
	counts[OCS_HW_LINK_STAT_RCV_EOFNI_COUNT].overflow = mbox_rsp->w17of;
	counts[OCS_HW_LINK_STAT_RCV_SOFF_COUNT].overflow = mbox_rsp->w18of;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_AER_COUNT].overflow = mbox_rsp->w19of;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_RPI_COUNT].overflow = mbox_rsp->w20of;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_XRI_COUNT].overflow = mbox_rsp->w21of;

	counts[OCS_HW_LINK_STAT_LINK_FAILURE_COUNT].counter = mbox_rsp->link_failure_error_count;
	counts[OCS_HW_LINK_STAT_LOSS_OF_SYNC_COUNT].counter = mbox_rsp->loss_of_sync_error_count;
	counts[OCS_HW_LINK_STAT_LOSS_OF_SIGNAL_COUNT].counter = mbox_rsp->loss_of_signal_error_count;
	counts[OCS_HW_LINK_STAT_PRIMITIVE_SEQ_COUNT].counter = mbox_rsp->primitive_sequence_error_count;
	counts[OCS_HW_LINK_STAT_INVALID_XMIT_WORD_COUNT].counter = mbox_rsp->invalid_transmission_word_error_count;
	counts[OCS_HW_LINK_STAT_CRC_COUNT].counter = mbox_rsp->crc_error_count;
	counts[OCS_HW_LINK_STAT_PRIMITIVE_SEQ_TIMEOUT_COUNT].counter = mbox_rsp->primitive_sequence_event_timeout_count;
	counts[OCS_HW_LINK_STAT_ELASTIC_BUFFER_OVERRUN_COUNT].counter = mbox_rsp->elastic_buffer_overrun_error_count;
	counts[OCS_HW_LINK_STAT_ARB_TIMEOUT_COUNT].counter = mbox_rsp->arbitration_fc_al_timout_count;
	counts[OCS_HW_LINK_STAT_ADVERTISED_RCV_B2B_CREDIT].counter = mbox_rsp->advertised_receive_bufftor_to_buffer_credit;
	counts[OCS_HW_LINK_STAT_CURR_RCV_B2B_CREDIT].counter = mbox_rsp->current_receive_buffer_to_buffer_credit;
	counts[OCS_HW_LINK_STAT_ADVERTISED_XMIT_B2B_CREDIT].counter = mbox_rsp->advertised_transmit_buffer_to_buffer_credit;
	counts[OCS_HW_LINK_STAT_CURR_XMIT_B2B_CREDIT].counter = mbox_rsp->current_transmit_buffer_to_buffer_credit;
	counts[OCS_HW_LINK_STAT_RCV_EOFA_COUNT].counter = mbox_rsp->received_eofa_count;
	counts[OCS_HW_LINK_STAT_RCV_EOFDTI_COUNT].counter = mbox_rsp->received_eofdti_count;
	counts[OCS_HW_LINK_STAT_RCV_EOFNI_COUNT].counter = mbox_rsp->received_eofni_count;
	counts[OCS_HW_LINK_STAT_RCV_SOFF_COUNT].counter = mbox_rsp->received_soff_count;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_AER_COUNT].counter = mbox_rsp->received_dropped_no_aer_count;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_RPI_COUNT].counter = mbox_rsp->received_dropped_no_available_rpi_resources_count;
	counts[OCS_HW_LINK_STAT_RCV_DROPPED_NO_XRI_COUNT].counter = mbox_rsp->received_dropped_no_available_xri_resources_count;

	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(status,
				   num_counters,
				   counts,
				   cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_link_stat_cb_arg_t));
	}
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return 0;
}

/**
 * @brief Function to retrieve the link and host statistics.
 *
 * @param hw Hardware context.
 * @param cc clear counters, if TRUE all counters will be cleared.
 * @param cb Function call upon completion of receiving the data.
 * @param arg Argument to pass to pointer fc hosts statistics structure.
 *
 * @return Returns OCS_HW_RTN_SUCCESS, OCS_HW_RTN_ERROR, or OCS_HW_RTN_NO_MEMORY.
 */
ocs_hw_rtn_e
ocs_hw_get_host_stats(ocs_hw_t *hw, uint8_t cc, ocs_hw_host_stat_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	ocs_hw_host_stat_cb_arg_t *cb_arg;
	uint8_t *mbxdata;

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_host_stat_cb_arg_t), 0);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	 }

	 cb_arg->cb = cb;
	 cb_arg->arg = arg;

	 /* Send the HW command to get the host stats */
	if (sli_cmd_read_status(&hw->sli, mbxdata, SLI4_BMBX_SIZE, cc)) {
		 rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_cb_host_stat, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "READ_HOST_STATS failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_host_stat_cb_arg_t));
	}

	return rc;
}


/**
 * @brief Called when the READ_STATUS command completes.
 *
 * @par Description
 * Get the counters out of the response, free the mailbox that was malloc'd
 * by ocs_hw_get_host_stats(), then call the callback and pass
 * the status and data.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is defined by
 * ocs_hw_host_stat_cb_t.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_host_stat(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{

	sli4_cmd_read_status_t* mbox_rsp = (sli4_cmd_read_status_t*) mqe;
	ocs_hw_host_stat_cb_arg_t *cb_arg = arg;
	ocs_hw_host_stat_counts_t counts[OCS_HW_HOST_STAT_MAX];
	uint32_t num_counters = OCS_HW_HOST_STAT_MAX;

	ocs_memset(counts, 0, sizeof(ocs_hw_host_stat_counts_t) *
		   OCS_HW_HOST_STAT_MAX);

	counts[OCS_HW_HOST_STAT_TX_KBYTE_COUNT].counter = mbox_rsp->transmit_kbyte_count;
	counts[OCS_HW_HOST_STAT_RX_KBYTE_COUNT].counter = mbox_rsp->receive_kbyte_count;
	counts[OCS_HW_HOST_STAT_TX_FRAME_COUNT].counter = mbox_rsp->transmit_frame_count;
	counts[OCS_HW_HOST_STAT_RX_FRAME_COUNT].counter = mbox_rsp->receive_frame_count;
	counts[OCS_HW_HOST_STAT_TX_SEQ_COUNT].counter = mbox_rsp->transmit_sequence_count;
	counts[OCS_HW_HOST_STAT_RX_SEQ_COUNT].counter = mbox_rsp->receive_sequence_count;
	counts[OCS_HW_HOST_STAT_TOTAL_EXCH_ORIG].counter = mbox_rsp->total_exchanges_originator;
	counts[OCS_HW_HOST_STAT_TOTAL_EXCH_RESP].counter = mbox_rsp->total_exchanges_responder;
	counts[OCS_HW_HOSY_STAT_RX_P_BSY_COUNT].counter = mbox_rsp->receive_p_bsy_count;
	counts[OCS_HW_HOST_STAT_RX_F_BSY_COUNT].counter = mbox_rsp->receive_f_bsy_count;
	counts[OCS_HW_HOST_STAT_DROP_FRM_DUE_TO_NO_RQ_BUF_COUNT].counter = mbox_rsp->dropped_frames_due_to_no_rq_buffer_count;
	counts[OCS_HW_HOST_STAT_EMPTY_RQ_TIMEOUT_COUNT].counter = mbox_rsp->empty_rq_timeout_count;
	counts[OCS_HW_HOST_STAT_DROP_FRM_DUE_TO_NO_XRI_COUNT].counter = mbox_rsp->dropped_frames_due_to_no_xri_count;
	counts[OCS_HW_HOST_STAT_EMPTY_XRI_POOL_COUNT].counter = mbox_rsp->empty_xri_pool_count;


	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(status,
				   num_counters,
				   counts,
				   cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_host_stat_cb_arg_t));
	}
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return 0;
}

/**
 * @brief HW link configuration enum to the CLP string value mapping.
 *
 * This structure provides a mapping from the ocs_hw_linkcfg_e
 * enum (enum exposed for the OCS_HW_PORT_SET_LINK_CONFIG port
 * control) to the CLP string that is used
 * in the DMTF_CLP_CMD mailbox command.
 */
typedef struct ocs_hw_linkcfg_map_s {
	ocs_hw_linkcfg_e linkcfg;
	const char *clp_str;
} ocs_hw_linkcfg_map_t;

/**
 * @brief Mapping from the HW linkcfg enum to the CLP command value
 * string.
 */
static ocs_hw_linkcfg_map_t linkcfg_map[] = {
	{OCS_HW_LINKCFG_4X10G, "ELX_4x10G"},
	{OCS_HW_LINKCFG_1X40G, "ELX_1x40G"},
	{OCS_HW_LINKCFG_2X16G, "ELX_2x16G"},
	{OCS_HW_LINKCFG_4X8G, "ELX_4x8G"},
	{OCS_HW_LINKCFG_4X1G, "ELX_4x1G"},
	{OCS_HW_LINKCFG_2X10G, "ELX_2x10G"},
	{OCS_HW_LINKCFG_2X10G_2X8G, "ELX_2x10G_2x8G"}};

/**
 * @brief HW link configuration enum to Skyhawk link config ID mapping.
 *
 * This structure provides a mapping from the ocs_hw_linkcfg_e
 * enum (enum exposed for the OCS_HW_PORT_SET_LINK_CONFIG port
 * control) to the link config ID numbers used by Skyhawk
 */
typedef struct ocs_hw_skyhawk_linkcfg_map_s {
	ocs_hw_linkcfg_e linkcfg;
	uint32_t	config_id;
} ocs_hw_skyhawk_linkcfg_map_t;

/**
 * @brief Mapping from the HW linkcfg enum to the Skyhawk link config IDs
 */
static ocs_hw_skyhawk_linkcfg_map_t skyhawk_linkcfg_map[] = {
	{OCS_HW_LINKCFG_4X10G, 0x0a},
	{OCS_HW_LINKCFG_1X40G, 0x09},
};

/**
 * @brief Helper function for getting the HW linkcfg enum from the CLP
 * string value
 *
 * @param clp_str CLP string value from OEMELX_LinkConfig.
 *
 * @return Returns the HW linkcfg enum corresponding to clp_str.
 */
static ocs_hw_linkcfg_e
ocs_hw_linkcfg_from_clp(const char *clp_str)
{
	uint32_t i;
	for (i = 0; i < ARRAY_SIZE(linkcfg_map); i++) {
		if (ocs_strncmp(linkcfg_map[i].clp_str, clp_str, ocs_strlen(clp_str)) == 0) {
			return linkcfg_map[i].linkcfg;
		}
	}
	return OCS_HW_LINKCFG_NA;
}

/**
 * @brief Helper function for getting the CLP string value from the HW
 * linkcfg enum.
 *
 * @param linkcfg HW linkcfg enum.
 *
 * @return Returns the OEMELX_LinkConfig CLP string value corresponding to
 * given linkcfg.
 */
static const char *
ocs_hw_clp_from_linkcfg(ocs_hw_linkcfg_e linkcfg)
{
	uint32_t i;
	for (i = 0; i < ARRAY_SIZE(linkcfg_map); i++) {
		if (linkcfg_map[i].linkcfg == linkcfg) {
			return linkcfg_map[i].clp_str;
		}
	}
	return NULL;
}

/**
 * @brief Helper function for getting a Skyhawk link config ID from the HW
 * linkcfg enum.
 *
 * @param linkcfg HW linkcfg enum.
 *
 * @return Returns the Skyhawk link config ID corresponding to
 * given linkcfg.
 */
static uint32_t
ocs_hw_config_id_from_linkcfg(ocs_hw_linkcfg_e linkcfg)
{
	uint32_t i;
	for (i = 0; i < ARRAY_SIZE(skyhawk_linkcfg_map); i++) {
		if (skyhawk_linkcfg_map[i].linkcfg == linkcfg) {
			return skyhawk_linkcfg_map[i].config_id;
		}
	}
	return 0;
}

/**
 * @brief Helper function for getting the HW linkcfg enum from a
 * Skyhawk config ID.
 *
 * @param config_id Skyhawk link config ID.
 *
 * @return Returns the HW linkcfg enum corresponding to config_id.
 */
static ocs_hw_linkcfg_e
ocs_hw_linkcfg_from_config_id(const uint32_t config_id)
{
	uint32_t i;
	for (i = 0; i < ARRAY_SIZE(skyhawk_linkcfg_map); i++) {
		if (skyhawk_linkcfg_map[i].config_id == config_id) {
			return skyhawk_linkcfg_map[i].linkcfg;
		}
	}
	return OCS_HW_LINKCFG_NA;
}

/**
 * @brief Link configuration callback argument.
 */
typedef struct ocs_hw_linkcfg_cb_arg_s {
	ocs_hw_port_control_cb_t cb;
	void *arg;
	uint32_t opts;
	int32_t status;
	ocs_dma_t dma_cmd;
	ocs_dma_t dma_resp;
	uint32_t result_len;
} ocs_hw_linkcfg_cb_arg_t;

/**
 * @brief Set link configuration.
 *
 * @param hw Hardware context.
 * @param value Link configuration enum to which the link configuration is
 * set.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_linkcfg(ocs_hw_t *hw, ocs_hw_linkcfg_e value, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	if (!sli_link_is_configurable(&hw->sli)) {
		ocs_log_debug(hw->os, "Function not supported\n");
		return OCS_HW_RTN_ERROR;
	}

	if (SLI4_IF_TYPE_LANCER_FC_ETH == sli_get_if_type(&hw->sli)) {
		return ocs_hw_set_linkcfg_lancer(hw, value, opts, cb, arg);
	} else if ((SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) ||
		   (SLI4_IF_TYPE_BE3_SKH_VF == sli_get_if_type(&hw->sli))) {
		return ocs_hw_set_linkcfg_skyhawk(hw, value, opts, cb, arg);
	} else {
		ocs_log_test(hw->os, "Function not supported for this IF_TYPE\n");
		return OCS_HW_RTN_ERROR;
	}
}

/**
 * @brief Set link configuration for Lancer
 *
 * @param hw Hardware context.
 * @param value Link configuration enum to which the link configuration is
 * set.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_linkcfg_lancer(ocs_hw_t *hw, ocs_hw_linkcfg_e value, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	char cmd[OCS_HW_DMTF_CLP_CMD_MAX];
	ocs_hw_linkcfg_cb_arg_t *cb_arg;
	const char *value_str = NULL;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* translate ocs_hw_linkcfg_e to CLP string */
	value_str = ocs_hw_clp_from_linkcfg(value);

	/* allocate memory for callback argument */
	cb_arg = ocs_malloc(hw->os, sizeof(*cb_arg), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		return OCS_HW_RTN_NO_MEMORY;
	}

	ocs_snprintf(cmd, OCS_HW_DMTF_CLP_CMD_MAX, "set / OEMELX_LinkConfig=%s", value_str);
	/* allocate DMA for command  */
	if (ocs_dma_alloc(hw->os, &cb_arg->dma_cmd, ocs_strlen(cmd)+1, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		return OCS_HW_RTN_NO_MEMORY;
	}
	ocs_memset(cb_arg->dma_cmd.virt, 0, ocs_strlen(cmd)+1);
	ocs_memcpy(cb_arg->dma_cmd.virt, cmd, ocs_strlen(cmd));

	/* allocate DMA for response */
	if (ocs_dma_alloc(hw->os, &cb_arg->dma_resp, OCS_HW_DMTF_CLP_RSP_MAX, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		return OCS_HW_RTN_NO_MEMORY;
	}
	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->opts = opts;

	rc = ocs_hw_exec_dmtf_clp_cmd(hw, &cb_arg->dma_cmd, &cb_arg->dma_resp,
					opts, ocs_hw_linkcfg_dmtf_clp_cb, cb_arg);

	if (opts == OCS_CMD_POLL || rc != OCS_HW_RTN_SUCCESS) {
		/* if failed, or polling, free memory here; if success and not
		 * polling, will free in callback function
		 */
		if (rc) {
			ocs_log_test(hw->os, "CLP cmd=\"%s\" failed\n",
					(char *)cb_arg->dma_cmd.virt);
		}
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_dma_free(hw->os, &cb_arg->dma_resp);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}
	return rc;
}

/**
 * @brief Callback for ocs_hw_set_linkcfg_skyhawk
 *
 * @param hw Hardware context.
 * @param status Status from the RECONFIG_GET_LINK_INFO command.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback argument.
 *
 * @return none
 */
static void
ocs_hw_set_active_link_config_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	ocs_hw_linkcfg_cb_arg_t *cb_arg = (ocs_hw_linkcfg_cb_arg_t *)arg;

	if (status) {
		ocs_log_test(hw->os, "SET_RECONFIG_LINK_ID failed, status=%d\n", status);
	}

	/* invoke callback */
	if (cb_arg->cb) {
		cb_arg->cb(status, 0, cb_arg->arg);
	}

	/* if polling, will free memory in calling function */
	if (cb_arg->opts != OCS_CMD_POLL) {
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}
}

/**
 * @brief Set link configuration for a Skyhawk
 *
 * @param hw Hardware context.
 * @param value Link configuration enum to which the link configuration is
 * set.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_linkcfg_skyhawk(ocs_hw_t *hw, ocs_hw_linkcfg_e value, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	uint8_t *mbxdata;
	ocs_hw_linkcfg_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint32_t config_id;

	config_id = ocs_hw_config_id_from_linkcfg(value);

	if (config_id == 0) {
		ocs_log_test(hw->os, "Link config %d not supported by Skyhawk\n", value);
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_linkcfg_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;

	if (sli_cmd_common_set_reconfig_link_id(&hw->sli, mbxdata, SLI4_BMBX_SIZE, NULL, 0, config_id)) {
		rc = ocs_hw_command(hw, mbxdata, opts, ocs_hw_set_active_link_config_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_err(hw->os, "SET_RECONFIG_LINK_ID failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_linkcfg_cb_arg_t));
	} else if (opts == OCS_CMD_POLL) {
		/* if we're polling we have to call the callback here. */
		ocs_hw_set_active_link_config_cb(hw, 0, mbxdata, cb_arg);
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_linkcfg_cb_arg_t));
	} else {
		/* We weren't poling, so the callback got called */
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
	}

	return rc;
}

/**
 * @brief Get link configuration.
 *
 * @param hw Hardware context.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_get_linkcfg(ocs_hw_t *hw, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	if (!sli_link_is_configurable(&hw->sli)) {
		ocs_log_debug(hw->os, "Function not supported\n");
		return OCS_HW_RTN_ERROR;
	}

	if ((SLI4_IF_TYPE_LANCER_FC_ETH == sli_get_if_type(&hw->sli)) ||
	    (SLI4_IF_TYPE_LANCER_G7 == sli_get_if_type(&hw->sli))){
		return ocs_hw_get_linkcfg_lancer(hw, opts, cb, arg);
	} else if ((SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) ||
		   (SLI4_IF_TYPE_BE3_SKH_VF == sli_get_if_type(&hw->sli))) {
		return ocs_hw_get_linkcfg_skyhawk(hw, opts, cb, arg);
	} else {
		ocs_log_test(hw->os, "Function not supported for this IF_TYPE\n");
		return OCS_HW_RTN_ERROR;
	}
}

/**
 * @brief Get link configuration for a Lancer
 *
 * @param hw Hardware context.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_get_linkcfg_lancer(ocs_hw_t *hw, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	char cmd[OCS_HW_DMTF_CLP_CMD_MAX];
	ocs_hw_linkcfg_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* allocate memory for callback argument */
	cb_arg = ocs_malloc(hw->os, sizeof(*cb_arg), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		return OCS_HW_RTN_NO_MEMORY;
	}

	ocs_snprintf(cmd, OCS_HW_DMTF_CLP_CMD_MAX, "show / OEMELX_LinkConfig");

	/* allocate DMA for command  */
	if (ocs_dma_alloc(hw->os, &cb_arg->dma_cmd, ocs_strlen(cmd)+1, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* copy CLP command to DMA command */
	ocs_memset(cb_arg->dma_cmd.virt, 0, ocs_strlen(cmd)+1);
	ocs_memcpy(cb_arg->dma_cmd.virt, cmd, ocs_strlen(cmd));

	/* allocate DMA for response */
	if (ocs_dma_alloc(hw->os, &cb_arg->dma_resp, OCS_HW_DMTF_CLP_RSP_MAX, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		return OCS_HW_RTN_NO_MEMORY;
	}
	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->opts = opts;

	rc = ocs_hw_exec_dmtf_clp_cmd(hw, &cb_arg->dma_cmd, &cb_arg->dma_resp,
					opts, ocs_hw_linkcfg_dmtf_clp_cb, cb_arg);

	if (opts == OCS_CMD_POLL || rc != OCS_HW_RTN_SUCCESS) {
		/* if failed or polling, free memory here; if not polling and success,
		 * will free in callback function
		 */
		if (rc) {
			ocs_log_test(hw->os, "CLP cmd=\"%s\" failed\n",
					(char *)cb_arg->dma_cmd.virt);
		}
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_dma_free(hw->os, &cb_arg->dma_resp);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}
	return rc;
}


/**
 * @brief Get the link configuration callback.
 *
 * @param hw Hardware context.
 * @param status Status from the RECONFIG_GET_LINK_INFO command.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback argument.
 *
 * @return none
 */
static void
ocs_hw_get_active_link_config_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	ocs_hw_linkcfg_cb_arg_t *cb_arg = (ocs_hw_linkcfg_cb_arg_t *)arg;
	sli4_res_common_get_reconfig_link_info_t *rsp = cb_arg->dma_cmd.virt;
	ocs_hw_linkcfg_e value = OCS_HW_LINKCFG_NA;

	if (status) {
		ocs_log_test(hw->os, "GET_RECONFIG_LINK_INFO failed, status=%d\n", status);
	} else {
		/* Call was successful */
		value = ocs_hw_linkcfg_from_config_id(rsp->active_link_config_id);
	}

	/* invoke callback */
	if (cb_arg->cb) {
		cb_arg->cb(status, value, cb_arg->arg);
	}

	/* if polling, will free memory in calling function */
	if (cb_arg->opts != OCS_CMD_POLL) {
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}
}

/**
 * @brief Get link configuration for a Skyhawk.
 *
 * @param hw Hardware context.
 * @param opts Mailbox command options (OCS_CMD_NOWAIT/POLL).
 * @param cb Callback function to invoke following mbx command.
 * @param arg Callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_get_linkcfg_skyhawk(ocs_hw_t *hw, uint32_t opts, ocs_hw_port_control_cb_t cb, void *arg)
{
	uint8_t *mbxdata;
	ocs_hw_linkcfg_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_linkcfg_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->opts = opts;

	/* dma_mem holds the non-embedded portion */
	if (ocs_dma_alloc(hw->os, &cb_arg->dma_cmd, sizeof(sli4_res_common_get_reconfig_link_info_t), 4)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_linkcfg_cb_arg_t));
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_common_get_reconfig_link_info(&hw->sli, mbxdata, SLI4_BMBX_SIZE, &cb_arg->dma_cmd)) {
		rc = ocs_hw_command(hw, mbxdata, opts, ocs_hw_get_active_link_config_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_err(hw->os, "GET_RECONFIG_LINK_INFO failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_linkcfg_cb_arg_t));
	} else if (opts == OCS_CMD_POLL) {
		/* if we're polling we have to call the callback here. */
		ocs_hw_get_active_link_config_cb(hw, 0, mbxdata, cb_arg);
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_linkcfg_cb_arg_t));
	} else {
		/* We weren't poling, so the callback got called */
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
	}

	return rc;
}

/**
 * @brief Sets the DIF seed value.
 *
 * @param hw Hardware context.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_dif_seed(ocs_hw_t *hw)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t buf[SLI4_BMBX_SIZE];
	sli4_req_common_set_features_dif_seed_t seed_param;

	ocs_memset(&seed_param, 0, sizeof(seed_param));
	seed_param.seed = hw->config.dif_seed;

	/* send set_features command */
	if (sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
					SLI4_SET_FEATURES_DIF_SEED,
					4,
					(uint32_t*)&seed_param)) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
		if (rc) {
			ocs_log_err(hw->os, "ocs_hw_command returns %d\n", rc);
		} else {
			ocs_log_debug(hw->os, "DIF seed set to 0x%x\n",
					hw->config.dif_seed);
		}
	} else {
		ocs_log_err(hw->os, "sli_cmd_common_set_features failed\n");
		rc = OCS_HW_RTN_ERROR;
	}
	return rc;
}


/**
 * @brief Sets the DIF mode value.
 *
 * @param hw Hardware context.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_dif_mode(ocs_hw_t *hw)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t buf[SLI4_BMBX_SIZE];
	sli4_req_common_set_features_t10_pi_mem_model_t mode_param;

	ocs_memset(&mode_param, 0, sizeof(mode_param));
	mode_param.tmm = (hw->config.dif_mode == OCS_HW_DIF_MODE_INLINE ? 0 : 1);

	/* send set_features command */
	if (sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
					SLI4_SET_FEATURES_DIF_MEMORY_MODE,
					sizeof(mode_param),
					(uint32_t*)&mode_param)) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
		if (rc) {
			ocs_log_err(hw->os, "ocs_hw_command returns %d\n", rc);
		} else {
			ocs_log_test(hw->os, "DIF mode set to %s\n",
				(hw->config.dif_mode == OCS_HW_DIF_MODE_INLINE ? "inline" : "separate"));
		}
	} else {
		ocs_log_err(hw->os, "sli_cmd_common_set_features failed\n");
		rc = OCS_HW_RTN_ERROR;
	}
	return rc;
}

static void
ocs_hw_watchdog_timer_cb(void *arg)
{
	ocs_hw_t *hw = (ocs_hw_t *)arg;

	ocs_hw_config_watchdog_timer(hw);
	return;
}

static void
ocs_hw_cb_cfg_watchdog(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	uint16_t timeout = hw->watchdog_timeout;

	if (status != 0) {
		ocs_log_err(hw->os, "config watchdog timer failed, rc = %d\n", status);
	} else {
		if(timeout != 0) {
			/* keeping callback 500ms before timeout to keep heartbeat alive */
			ocs_setup_timer(hw->os, &hw->watchdog_timer, ocs_hw_watchdog_timer_cb, hw, (timeout*1000 - 500) );
		}else {
			ocs_del_timer(&hw->watchdog_timer);
		}
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	return;
}

/**
 * @brief Set configuration parameters for watchdog timer feature.
 *
 * @param hw Hardware context.
 * @param timeout Timeout for watchdog timer in seconds
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_config_watchdog_timer(ocs_hw_t *hw)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t *buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);

	sli4_cmd_lowlevel_set_watchdog(&hw->sli, buf, SLI4_BMBX_SIZE, hw->watchdog_timeout);
	rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, ocs_hw_cb_cfg_watchdog, NULL);
	if (rc) {
		ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
		ocs_log_err(hw->os, "config watchdog timer failed, rc = %d\n", rc);
	}
	return rc;
}

/**
 * @brief Set configuration parameters for auto-generate xfer_rdy T10 PI feature.
 *
 * @param hw Hardware context.
 * @param buf Pointer to a mailbox buffer area.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_config_auto_xfer_rdy_t10pi(ocs_hw_t *hw, uint8_t *buf)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	sli4_req_common_set_features_xfer_rdy_t10pi_t param;

	ocs_memset(&param, 0, sizeof(param));
	param.rtc = (hw->config.auto_xfer_rdy_ref_tag_is_lba ? 0 : 1);
	param.atv = (hw->config.auto_xfer_rdy_app_tag_valid ? 1 : 0);
	param.tmm = ((hw->config.dif_mode == OCS_HW_DIF_MODE_INLINE) ? 0 : 1);
	param.app_tag = hw->config.auto_xfer_rdy_app_tag_value;
	param.blk_size = hw->config.auto_xfer_rdy_blk_size_chip;

	switch (hw->config.auto_xfer_rdy_p_type) {
	case 1:
		param.p_type = 0;
		break;
	case 3:
		param.p_type = 2;
		break;
	default:
		ocs_log_err(hw->os, "unsupported p_type %d\n",
			hw->config.auto_xfer_rdy_p_type);
		return OCS_HW_RTN_ERROR;
	}

	/* build the set_features command */
	sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
				    SLI4_SET_FEATURES_SET_CONFIG_AUTO_XFER_RDY_T10PI,
				    sizeof(param),
				    &param);


	rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
	if (rc) {
		ocs_log_err(hw->os, "ocs_hw_command returns %d\n", rc);
	} else {
		ocs_log_test(hw->os, "Auto XFER RDY T10 PI configured rtc:%d atv:%d p_type:%d app_tag:%x blk_size:%d\n",
				param.rtc, param.atv, param.p_type,
				param.app_tag, param.blk_size);
	}

	return rc;
}


/**
 * @brief enable sli port health check
 *
 * @param hw Hardware context.
 * @param buf Pointer to a mailbox buffer area.
 * @param query current status of the health check feature enabled/disabled
 * @param enable if 1: enable 0: disable
 * @param buf Pointer to a mailbox buffer area.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_config_sli_port_health_check(ocs_hw_t *hw, uint8_t query, uint8_t enable)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t buf[SLI4_BMBX_SIZE];
	sli4_req_common_set_features_health_check_t param;

	ocs_memset(&param, 0, sizeof(param));
	param.hck = enable;
	param.qry = query;

	/* build the set_features command */
	sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
				    SLI4_SET_FEATURES_SLI_PORT_HEALTH_CHECK,
				    sizeof(param),
				    &param);

	rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
	if (rc) {
		ocs_log_err(hw->os, "ocs_hw_command returns %d\n", rc);
	} else {
		ocs_log_test(hw->os, "SLI Port Health Check is enabled \n");
	}

	return rc;
}

/**
 * @brief Set FTD transfer hint feature
 *
 * @param hw Hardware context.
 * @param fdt_xfer_hint size in bytes where read requests are segmented.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_config_set_fdt_xfer_hint(ocs_hw_t *hw, uint32_t fdt_xfer_hint)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t buf[SLI4_BMBX_SIZE];
	sli4_req_common_set_features_set_fdt_xfer_hint_t param;

	ocs_memset(&param, 0, sizeof(param));
	param.fdt_xfer_hint = fdt_xfer_hint;
	/* build the set_features command */
	sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
				    SLI4_SET_FEATURES_SET_FTD_XFER_HINT,
				    sizeof(param),
				    &param);


	rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
	if (rc) {
		ocs_log_warn(hw->os, "set FDT hint %d failed: %d\n", fdt_xfer_hint, rc);
	} else {
		ocs_log_debug(hw->os, "Set FTD transfer hint to %d\n", param.fdt_xfer_hint);
	}

	return rc;
}

/**
 * @brief Get the link configuration callback.
 *
 * @param hw Hardware context.
 * @param status Status from the DMTF CLP command.
 * @param result_len Length, in bytes, of the DMTF CLP result.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static void
ocs_hw_linkcfg_dmtf_clp_cb(ocs_hw_t *hw, int32_t status, uint32_t result_len, void *arg)
{
	int32_t rval;
	char retdata_str[64];
	ocs_hw_linkcfg_cb_arg_t *cb_arg = (ocs_hw_linkcfg_cb_arg_t *)arg;
	ocs_hw_linkcfg_e linkcfg = OCS_HW_LINKCFG_NA;

	if (status) {
		ocs_log_test(hw->os, "CLP cmd failed, status=%d\n", status);
	} else {
		/* parse CLP response to get return data */
		rval = ocs_hw_clp_resp_get_value(hw, "retdata", retdata_str,
						  sizeof(retdata_str),
						  cb_arg->dma_resp.virt,
						  result_len);

		if (rval <= 0) {
			ocs_log_err(hw->os, "failed to get retdata %d\n", result_len);
		} else {
			/* translate string into hw enum */
			linkcfg = ocs_hw_linkcfg_from_clp(retdata_str);
		}
	}

	/* invoke callback */
	if (cb_arg->cb) {
		cb_arg->cb(status, linkcfg, cb_arg->arg);
	}

	/* if polling, will free memory in calling function */
	if (cb_arg->opts != OCS_CMD_POLL) {
		ocs_dma_free(hw->os, &cb_arg->dma_cmd);
		ocs_dma_free(hw->os, &cb_arg->dma_resp);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}
}

/**
 * @brief Set the Lancer dump location
 * @par Description
 * This function tells a Lancer chip to use a specific DMA
 * buffer as a dump location rather than the internal flash.
 *
 * @param hw Hardware context.
 * @param num_buffers The number of DMA buffers to hold the dump (1..n).
 * @param dump_buffers DMA buffers to hold the dump.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
ocs_hw_rtn_e
ocs_hw_set_dump_location(ocs_hw_t *hw, uint32_t num_buffers, ocs_dma_t *dump_buffers, uint8_t fdb)
{
	uint8_t bus, dev, func;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t	buf[SLI4_BMBX_SIZE];

	/*
	 * Make sure the FW is new enough to support this command. If the FW
	 * is too old, the FW will UE.
	 */
	if (hw->workaround.disable_dump_loc) {
		ocs_log_test(hw->os, "FW version is too old for this feature\n");
		return OCS_HW_RTN_ERROR;
	}

	/* This command is only valid for physical port 0 */
	ocs_get_bus_dev_func(hw->os, &bus, &dev, &func);
	if (fdb == 0 && func != 0) {
		ocs_log_test(hw->os, "function only valid for pci function 0, %d passed\n",
			     func);
		return OCS_HW_RTN_ERROR;
	}

	/*
	 * If a single buffer is used, then it may be passed as is to the chip. For multiple buffers,
	 * We must allocate a SGL list and then pass the address of the list to the chip.
	 */
	if (num_buffers > 1) {
		uint32_t sge_size = num_buffers * sizeof(sli4_sge_t);
		sli4_sge_t *sge;
		uint32_t i;

		if (hw->dump_sges.size < sge_size) {
			ocs_dma_free(hw->os, &hw->dump_sges);
			if (ocs_dma_alloc(hw->os, &hw->dump_sges, sge_size, OCS_MIN_DMA_ALIGNMENT)) {
				ocs_log_err(hw->os, "SGE DMA allocation failed\n");
				return OCS_HW_RTN_NO_MEMORY;
			}
		}
		/* build the SGE list */
		ocs_memset(hw->dump_sges.virt, 0, hw->dump_sges.size);
		hw->dump_sges.len = sge_size;
		sge = hw->dump_sges.virt;
		for (i = 0; i < num_buffers; i++) {
			sge[i].buffer_address_high = ocs_addr32_hi(dump_buffers[i].phys);
			sge[i].buffer_address_low = ocs_addr32_lo(dump_buffers[i].phys);
			sge[i].last = (i == num_buffers - 1 ? 1 : 0);
			sge[i].buffer_length = dump_buffers[i].size;
		}
		rc = sli_cmd_common_set_dump_location(&hw->sli, (void *)buf,
						      SLI4_BMBX_SIZE, FALSE, TRUE,
						      &hw->dump_sges, fdb);
	} else {
		dump_buffers->len = dump_buffers->size;
		rc = sli_cmd_common_set_dump_location(&hw->sli, (void *)buf,
						      SLI4_BMBX_SIZE, FALSE, FALSE,
						      dump_buffers, fdb);
	}

	if (rc) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_POLL,
				     NULL, NULL);
		if (rc) {
			ocs_log_err(hw->os, "ocs_hw_command returns %d\n",
				rc);
		}
	} else {
		ocs_log_err(hw->os,
			"sli_cmd_common_set_dump_location failed\n");
		rc = OCS_HW_RTN_ERROR;
	}

	return rc;
}


/**
 * @brief Set the Ethernet license.
 *
 * @par Description
 * This function sends the appropriate mailbox command (DMTF
 * CLP) to set the Ethernet license to the given license value.
 * Since it is used during the time of ocs_hw_init(), the mailbox
 * command is sent via polling (the BMBX route).
 *
 * @param hw Hardware context.
 * @param license 32-bit license value.
 *
 * @return Returns OCS_HW_RTN_SUCCESS on success.
 */
static ocs_hw_rtn_e
ocs_hw_set_eth_license(ocs_hw_t *hw, uint32_t license)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	char cmd[OCS_HW_DMTF_CLP_CMD_MAX];
	ocs_dma_t dma_cmd;
	ocs_dma_t dma_resp;

	/* only for lancer right now */
	if (SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) {
		ocs_log_test(hw->os, "Function only supported for I/F type 2\n");
		return OCS_HW_RTN_ERROR;
	}

	ocs_snprintf(cmd, OCS_HW_DMTF_CLP_CMD_MAX, "set / OEMELX_Ethernet_License=%X", license);
	/* allocate DMA for command  */
	if (ocs_dma_alloc(hw->os, &dma_cmd, ocs_strlen(cmd)+1, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		return OCS_HW_RTN_NO_MEMORY;
	}
	ocs_memset(dma_cmd.virt, 0, ocs_strlen(cmd)+1);
	ocs_memcpy(dma_cmd.virt, cmd, ocs_strlen(cmd));

	/* allocate DMA for response */
	if (ocs_dma_alloc(hw->os, &dma_resp, OCS_HW_DMTF_CLP_RSP_MAX, 4096)) {
		ocs_log_err(hw->os, "malloc failed\n");
		ocs_dma_free(hw->os, &dma_cmd);
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* send DMTF CLP command mbx and poll */
	if (ocs_hw_exec_dmtf_clp_cmd(hw, &dma_cmd, &dma_resp, OCS_CMD_POLL, NULL, NULL)) {
		ocs_log_err(hw->os, "CLP cmd=\"%s\" failed\n", (char *)dma_cmd.virt);
		rc = OCS_HW_RTN_ERROR;
	}

	ocs_dma_free(hw->os, &dma_cmd);
	ocs_dma_free(hw->os, &dma_resp);
	return rc;
}

/**
 * @brief Callback argument structure for the DMTF CLP commands.
 */
typedef struct ocs_hw_clp_cb_arg_s {
	ocs_hw_dmtf_clp_cb_t cb;
	ocs_dma_t *dma_resp;
	int32_t status;
	uint32_t opts;
	void *arg;
} ocs_hw_clp_cb_arg_t;

/**
 * @brief Execute the DMTF CLP command.
 *
 * @param hw Hardware context.
 * @param dma_cmd DMA buffer containing the CLP command.
 * @param dma_resp DMA buffer that will contain the response (if successful).
 * @param opts Mailbox command options (such as OCS_CMD_NOWAIT and POLL).
 * @param cb Callback function.
 * @param arg Callback argument.
 *
 * @return Returns the number of bytes written to the response
 * buffer on success, or a negative value if failed.
 */
static ocs_hw_rtn_e
ocs_hw_exec_dmtf_clp_cmd(ocs_hw_t *hw, ocs_dma_t *dma_cmd, ocs_dma_t *dma_resp, uint32_t opts, ocs_hw_dmtf_clp_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	ocs_hw_clp_cb_arg_t *cb_arg;
	uint8_t *mbxdata;

	/* allocate DMA for mailbox */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* allocate memory for callback argument */
	cb_arg = ocs_malloc(hw->os, sizeof(*cb_arg), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->dma_resp = dma_resp;
	cb_arg->opts = opts;

	/* Send the HW command */
	if (sli_cmd_dmtf_exec_clp_cmd(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
				      dma_cmd, dma_resp)) {
		rc = ocs_hw_command(hw, mbxdata, opts, ocs_hw_dmtf_clp_cb, cb_arg);

		if (opts == OCS_CMD_POLL && rc == OCS_HW_RTN_SUCCESS) {
			/* if we're polling, copy response and invoke callback to
			 * parse result */
			ocs_memcpy(mbxdata, hw->sli.bmbx.virt, SLI4_BMBX_SIZE);
			ocs_hw_dmtf_clp_cb(hw, 0, mbxdata, cb_arg);

			/* set rc to resulting or "parsed" status */
			rc = cb_arg->status;
		}

		/* if failed, or polling, free memory here */
		if (opts == OCS_CMD_POLL || rc != OCS_HW_RTN_SUCCESS) {
			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_test(hw->os, "ocs_hw_command failed\n");
			}
			ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
			ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		}
	} else {
		ocs_log_test(hw->os, "sli_cmd_dmtf_exec_clp_cmd failed\n");
		rc = OCS_HW_RTN_ERROR;
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
	}

	return rc;
}


/**
 * @brief Called when the DMTF CLP command completes.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback argument.
 *
 * @return None.
 *
 */
static void
ocs_hw_dmtf_clp_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	int32_t cb_status = 0;
	sli4_cmd_sli_config_t* mbox_rsp = (sli4_cmd_sli_config_t*) mqe;
	sli4_res_dmtf_exec_clp_cmd_t *clp_rsp = (sli4_res_dmtf_exec_clp_cmd_t *) mbox_rsp->payload.embed;
	ocs_hw_clp_cb_arg_t *cb_arg = arg;
	uint32_t result_len = 0;
	int32_t stat_len;
	char stat_str[8];

	/* there are several status codes here, check them all and condense
	 * into a single callback status
	 */
	if (status || mbox_rsp->hdr.status || clp_rsp->clp_status) {
		ocs_log_debug(hw->os, "status=x%x/x%x/x%x  addl=x%x clp=x%x detail=x%x\n",
			status,
			mbox_rsp->hdr.status,
			clp_rsp->hdr.status,
			clp_rsp->hdr.additional_status,
			clp_rsp->clp_status,
			clp_rsp->clp_detailed_status);
		if (status) {
			cb_status = status;
		} else if (mbox_rsp->hdr.status) {
			cb_status = mbox_rsp->hdr.status;
		} else {
			cb_status = clp_rsp->clp_status;
		}
	} else {
		result_len = clp_rsp->resp_length;
	}

	if (cb_status) {
		goto ocs_hw_cb_dmtf_clp_done;
	}

	if ((result_len == 0) || (cb_arg->dma_resp->size < result_len)) {
		ocs_log_test(hw->os, "Invalid response length: resp_len=%zu result len=%d\n",
			     cb_arg->dma_resp->size, result_len);
		cb_status = -1;
		goto ocs_hw_cb_dmtf_clp_done;
	}

	/* parse CLP response to get status */
	stat_len = ocs_hw_clp_resp_get_value(hw, "status", stat_str,
					      sizeof(stat_str),
					      cb_arg->dma_resp->virt,
					      result_len);

	if (stat_len <= 0) {
		ocs_log_test(hw->os, "failed to get status %d\n", stat_len);
		cb_status = -1;
		goto ocs_hw_cb_dmtf_clp_done;
	}

	if (ocs_strcmp(stat_str, "0") != 0) {
		ocs_log_test(hw->os, "CLP status indicates failure=%s\n", stat_str);
		cb_status = -1;
		goto ocs_hw_cb_dmtf_clp_done;
	}

ocs_hw_cb_dmtf_clp_done:

	/* save status in cb_arg for callers with NULL cb's + polling */
	cb_arg->status = cb_status;
	if (cb_arg->cb) {
		cb_arg->cb(hw, cb_status, result_len, cb_arg->arg);
	}
	/* if polling, caller will free memory */
	if (cb_arg->opts != OCS_CMD_POLL) {
		ocs_free(hw->os, cb_arg, sizeof(*cb_arg));
		ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	}
}

/**
 * @brief Parse the CLP result and get the value corresponding to the given
 * keyword.
 *
 * @param hw Hardware context.
 * @param keyword CLP keyword for which the value is returned.
 * @param value Location to which the resulting value is copied.
 * @param value_len Length of the value parameter.
 * @param resp Pointer to the response buffer that is searched
 * for the keyword and value.
 * @param resp_len Length of response buffer passed in.
 *
 * @return Returns the number of bytes written to the value
 * buffer on success, or a negative vaue on failure.
 */
static int32_t
ocs_hw_clp_resp_get_value(ocs_hw_t *hw, const char *keyword, char *value, uint32_t value_len, const char *resp, uint32_t resp_len)
{
	char *start = NULL;
	char *end = NULL;

	/* look for specified keyword in string */
	start = ocs_strstr(resp, keyword);
	if (start == NULL) {
		ocs_log_test(hw->os, "could not find keyword=%s in CLP response\n",
			     keyword);
		return -1;
	}

	/* now look for '=' and go one past */
	start = ocs_strchr(start, '=');
	if (start == NULL) {
		ocs_log_test(hw->os, "could not find \'=\' in CLP response for keyword=%s\n",
			     keyword);
		return -1;
	}
	start++;

	/* \r\n terminates value */
	end = ocs_strstr(start, "\r\n");
	if (end == NULL) {
		ocs_log_test(hw->os, "could not find \\r\\n for keyword=%s in CLP response\n",
			     keyword);
		return -1;
	}

	/* make sure given result array is big enough */
	if ((end - start + 1) > value_len) {
		ocs_log_test(hw->os, "value len=%d not large enough for actual=%ld\n",
			     value_len, (end-start));
		return -1;
	}

	ocs_strncpy(value, start, (end - start));
	value[end-start] = '\0';
	return (end-start+1);
}

/**
 * @brief Cause chip to enter an unrecoverable error state.
 *
 * @par Description
 * Cause chip to enter an unrecoverable error state. This is
 * used when detecting unexpected FW behavior so that the FW can be
 * hwted from the driver as soon as the error is detected.
 *
 * @param hw Hardware context.
 * @param dump Generate dump as part of reset.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 *
 */
ocs_hw_rtn_e
ocs_hw_raise_ue(ocs_hw_t *hw, uint8_t dump)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	if (sli_raise_ue(&hw->sli, dump) != 0) {
		rc = OCS_HW_RTN_ERROR;
	} else {
		if (hw->state != OCS_HW_STATE_UNINITIALIZED) {
			hw->state = OCS_HW_STATE_QUEUES_ALLOCATED;
		}
	}

	return rc;
}

/**
 * @brief Called when the OBJECT_GET command completes.
 *
 * @par Description
 * Get the number of bytes actually written out of the response, free the mailbox
 * that was malloc'd by ocs_hw_dump_get(), then call the callback
 * and pass the status and bytes read.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is <tt>void cb(int32_t status, uint32_t bytes_read)</tt>.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_dump_get(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	sli4_cmd_sli_config_t* mbox_rsp = (sli4_cmd_sli_config_t*) mqe;
	sli4_res_common_read_object_t* rd_obj_rsp = (sli4_res_common_read_object_t*) mbox_rsp->payload.embed;
	ocs_hw_dump_get_cb_arg_t *cb_arg = arg;
	uint32_t bytes_read;
	uint8_t eof;

	bytes_read = rd_obj_rsp->actual_read_length;
	eof = rd_obj_rsp->eof;

	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(status, bytes_read, eof, cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg->mbox_cmd, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_dump_get_cb_arg_t));
	}

	return 0;
}


/**
 * @brief Read a dump image to the host.
 *
 * @par Description
 * Creates a SLI_CONFIG mailbox command, fills in the correct values to read a
 * dump image chunk, then sends the command with the ocs_hw_command(). On completion,
 * the callback function ocs_hw_cb_dump_get() gets called to free the mailbox
 * and signal the caller that the read has completed.
 *
 * @param hw Hardware context.
 * @param dma DMA structure to transfer the dump chunk into.
 * @param size Size of the dump chunk.
 * @param offset Offset, in bytes, from the beginning of the dump.
 * @param cb Pointer to a callback function that is called when the command completes.
 * The callback function prototype is
 * <tt>void cb(int32_t status, uint32_t bytes_read, uint8_t eof, void *arg)</tt>.
 * @param arg Pointer to be passed to the callback function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_dump_get(ocs_hw_t *hw, ocs_dma_t *dma, uint32_t size, uint32_t offset, ocs_hw_dump_get_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	uint8_t *mbxdata;
	ocs_hw_dump_get_cb_arg_t *cb_arg;
	uint32_t opts = (hw->state == OCS_HW_STATE_ACTIVE ? OCS_CMD_NOWAIT : OCS_CMD_POLL);

	if (SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) {
		ocs_log_test(hw->os, "Function only supported for I/F type 2\n");
		return OCS_HW_RTN_ERROR;
	}

	if (1 != sli_dump_is_present(&hw->sli)) {
		ocs_log_test(hw->os, "No dump is present\n");
		return OCS_HW_RTN_ERROR;
	}

	if (1 == sli_reset_required(&hw->sli)) {
		ocs_log_test(hw->os, "device reset required\n");
		return OCS_HW_RTN_ERROR;
	}

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_dump_get_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->mbox_cmd = mbxdata;

	if (sli_cmd_common_read_object(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
			size, offset, "/dbg/dump.bin", dma)) {
		rc = ocs_hw_command(hw, mbxdata, opts, ocs_hw_cb_dump_get, cb_arg);
		if (rc == 0 && opts == OCS_CMD_POLL) {
			ocs_memcpy(mbxdata, hw->sli.bmbx.virt, SLI4_BMBX_SIZE);
			rc = ocs_hw_cb_dump_get(hw, 0, mbxdata, cb_arg);
		}
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "COMMON_READ_OBJECT failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_dump_get_cb_arg_t));
	}

	return rc;
}

/**
 * @brief Called when the OBJECT_DELETE command completes.
 *
 * @par Description
 * Free the mailbox that was malloc'd
 * by ocs_hw_dump_clear(), then call the callback and pass the status.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 * The callback function prototype is <tt>void cb(int32_t status, void *arg)</tt>.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_dump_clear(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	ocs_hw_dump_clear_cb_arg_t *cb_arg = arg;
	sli4_cmd_sli_config_t* mbox_rsp = (sli4_cmd_sli_config_t*) mqe;

	if (cb_arg) {
		if (cb_arg->cb) {
			if ((status == 0) && mbox_rsp->hdr.status) {
				status = mbox_rsp->hdr.status;
			}
			cb_arg->cb(status, cb_arg->arg);
		}

		ocs_free(hw->os, cb_arg->mbox_cmd, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_dump_clear_cb_arg_t));
	}

	return 0;
}

/**
 * @brief Clear a dump image from the device.
 *
 * @par Description
 * Creates a SLI_CONFIG mailbox command, fills it with the correct values to clear
 * the dump, then sends the command with ocs_hw_command(). On completion,
 * the callback function ocs_hw_cb_dump_clear() gets called to free the mailbox
 * and to signal the caller that the write has completed.
 *
 * @param hw Hardware context.
 * @param cb Pointer to a callback function that is called when the command completes.
 * The callback function prototype is
 * <tt>void cb(int32_t status, uint32_t bytes_written, void *arg)</tt>.
 * @param arg Pointer to be passed to the callback function.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_dump_clear(ocs_hw_t *hw, ocs_hw_dump_clear_cb_t cb, void *arg)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;
	uint8_t *mbxdata;
	ocs_hw_dump_clear_cb_arg_t *cb_arg;
	uint32_t opts = (hw->state == OCS_HW_STATE_ACTIVE ? OCS_CMD_NOWAIT : OCS_CMD_POLL);

	if (SLI4_IF_TYPE_LANCER_FC_ETH != sli_get_if_type(&hw->sli)) {
		ocs_log_test(hw->os, "Function only supported for I/F type 2\n");
		return OCS_HW_RTN_ERROR;
	}

	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_dump_clear_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = arg;
	cb_arg->mbox_cmd = mbxdata;

	if (sli_cmd_common_delete_object(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
			"/dbg/dump.bin")) {
		rc = ocs_hw_command(hw, mbxdata, opts, ocs_hw_cb_dump_clear, cb_arg);
		if (rc == 0 && opts == OCS_CMD_POLL) {
			ocs_memcpy(mbxdata, hw->sli.bmbx.virt, SLI4_BMBX_SIZE);
			rc = ocs_hw_cb_dump_clear(hw, 0, mbxdata, cb_arg);
		}
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "COMMON_DELETE_OBJECT failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_dump_clear_cb_arg_t));
	}

	return rc;
}

typedef struct ocs_hw_get_port_protocol_cb_arg_s {
	ocs_get_port_protocol_cb_t cb;
	void *arg;
	uint32_t pci_func;
	ocs_dma_t payload;
} ocs_hw_get_port_protocol_cb_arg_t;

/**
 * @brief Called for the completion of get_port_profile for a
 *        user request.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE.
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_get_port_protocol_cb(ocs_hw_t *hw, int32_t status,
			    uint8_t *mqe, void *arg)
{
	ocs_hw_get_port_protocol_cb_arg_t *cb_arg = arg;
	ocs_dma_t *payload = &(cb_arg->payload);
	sli4_res_common_get_profile_config_t* response = (sli4_res_common_get_profile_config_t*) payload->virt;
	ocs_hw_port_protocol_e port_protocol;
	int num_descriptors;
	sli4_resource_descriptor_v1_t *desc_p;
	sli4_pcie_resource_descriptor_v1_t *pcie_desc_p;
	int i;

	port_protocol = OCS_HW_PORT_PROTOCOL_OTHER;

	num_descriptors = response->desc_count;
	desc_p = (sli4_resource_descriptor_v1_t *)response->desc;
	for (i=0; i<num_descriptors; i++) {
		if (desc_p->descriptor_type == SLI4_RESOURCE_DESCRIPTOR_TYPE_PCIE) {
			pcie_desc_p = (sli4_pcie_resource_descriptor_v1_t*) desc_p;
			if (pcie_desc_p->pf_number == cb_arg->pci_func) {
				switch(pcie_desc_p->pf_type) {
				case 0x02:
					port_protocol = OCS_HW_PORT_PROTOCOL_ISCSI;
					break;
				case 0x04:
					port_protocol = OCS_HW_PORT_PROTOCOL_FCOE;
					break;
				case 0x10:
					port_protocol = OCS_HW_PORT_PROTOCOL_FC;
					break;
				default:
					port_protocol = OCS_HW_PORT_PROTOCOL_OTHER;
					break;
				}
			}
		}

		desc_p = (sli4_resource_descriptor_v1_t *) ((uint8_t *)desc_p + desc_p->descriptor_length);
	}

	if (cb_arg->cb) {
		cb_arg->cb(status, port_protocol, cb_arg->arg);

	}

	ocs_dma_free(hw->os, &cb_arg->payload);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_port_protocol_cb_arg_t));
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return 0;
}

/**
 * @ingroup io
 * @brief  Get the current port protocol.
 * @par Description
 * Issues a SLI4 COMMON_GET_PROFILE_CONFIG mailbox.  When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param pci_func PCI function to query for current protocol.
 * @param cb Callback function to be called when the command completes.
 * @param ul_arg An argument that is passed to the callback function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
ocs_hw_rtn_e
ocs_hw_get_port_protocol(ocs_hw_t *hw, uint32_t pci_func,
	ocs_get_port_protocol_cb_t cb, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_get_port_protocol_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* Only supported on Skyhawk */
	if (sli_get_if_type(&hw->sli) != SLI4_IF_TYPE_BE3_SKH_PF) {
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}


	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_get_port_protocol_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;
	cb_arg->pci_func = pci_func;

	/* dma_mem holds the non-embedded portion */
	if (ocs_dma_alloc(hw->os, &cb_arg->payload, 4096, 4)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_port_protocol_cb_arg_t));
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_common_get_profile_config(&hw->sli, mbxdata, SLI4_BMBX_SIZE, &cb_arg->payload)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_get_port_protocol_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "GET_PROFILE_CONFIG failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_fw_write_cb_arg_t));
		ocs_dma_free(hw->os, &cb_arg->payload);
	}

	return rc;

}

typedef struct ocs_hw_set_port_protocol_cb_arg_s {
	ocs_set_port_protocol_cb_t cb;
	void *arg;
	ocs_dma_t payload;
	uint32_t new_protocol;
	uint32_t pci_func;
} ocs_hw_set_port_protocol_cb_arg_t;

/**
 * @brief Called for the completion of set_port_profile for a
 *        user request.
 *
 * @par Description
 * This is the second of two callbacks for the set_port_protocol
 * function. The set operation is a read-modify-write. This
 * callback is called when the write (SET_PROFILE_CONFIG)
 * completes.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE.
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return 0 on success, non-zero otherwise
 */
static int32_t
ocs_hw_set_port_protocol_cb2(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_set_port_protocol_cb_arg_t *cb_arg = arg;

	if (cb_arg->cb) {
		cb_arg->cb( status, cb_arg->arg);
	}

	ocs_dma_free(hw->os, &(cb_arg->payload));
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, arg, sizeof(ocs_hw_set_port_protocol_cb_arg_t));

	return 0;
}

/**
 * @brief Called for the completion of set_port_profile for a
 *        user request.
 *
 * @par Description
 * This is the first of two callbacks for the set_port_protocol
 * function.  The set operation is a read-modify-write.  This
 * callback is called when the read completes
 * (GET_PROFILE_CONFG).  It will updated the resource
 * descriptors, then queue the write (SET_PROFILE_CONFIG).
 *
 * On entry there are three memory areas that were allocated by
 * ocs_hw_set_port_protocol.  If a failure is detected in this
 * function those need to be freed.  If this function succeeds
 * it allocates three more areas.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value otherwise.
 */
static int32_t
ocs_hw_set_port_protocol_cb1(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_set_port_protocol_cb_arg_t *cb_arg = arg;
	ocs_dma_t *payload = &(cb_arg->payload);
	sli4_res_common_get_profile_config_t* response = (sli4_res_common_get_profile_config_t*) payload->virt;
	int num_descriptors;
	sli4_resource_descriptor_v1_t *desc_p;
	sli4_pcie_resource_descriptor_v1_t *pcie_desc_p;
	int i;
	ocs_hw_set_port_protocol_cb_arg_t *new_cb_arg;
	ocs_hw_port_protocol_e new_protocol;
	uint8_t *dst;
	sli4_isap_resouce_descriptor_v1_t *isap_desc_p;
	uint8_t *mbxdata;
	int pci_descriptor_count;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	int num_fcoe_ports = 0;
	int num_iscsi_ports = 0;

	new_protocol = (ocs_hw_port_protocol_e)cb_arg->new_protocol;

	num_descriptors = response->desc_count;

	/* Count PCI descriptors */
	pci_descriptor_count = 0;
	desc_p = (sli4_resource_descriptor_v1_t *)response->desc;
	for (i=0; i<num_descriptors; i++) {
		if (desc_p->descriptor_type == SLI4_RESOURCE_DESCRIPTOR_TYPE_PCIE) {
			++pci_descriptor_count;
		}
		desc_p = (sli4_resource_descriptor_v1_t *) ((uint8_t *)desc_p + desc_p->descriptor_length);
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}


	/* cb_arg holds the data that will be passed to the callback on completion */
	new_cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_set_port_protocol_cb_arg_t), OCS_M_NOWAIT);
	if (new_cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	new_cb_arg->cb = cb_arg->cb;
	new_cb_arg->arg = cb_arg->arg;

	/* Allocate memory for the descriptors we're going to send.  This is
	 * one for each PCI descriptor plus one ISAP descriptor. */
	if (ocs_dma_alloc(hw->os, &new_cb_arg->payload, sizeof(sli4_req_common_set_profile_config_t) +
			  (pci_descriptor_count * sizeof(sli4_pcie_resource_descriptor_v1_t)) +
			  sizeof(sli4_isap_resouce_descriptor_v1_t), 4)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, new_cb_arg, sizeof(ocs_hw_set_port_protocol_cb_arg_t));
		return OCS_HW_RTN_NO_MEMORY;
	}

	sli_cmd_common_set_profile_config(&hw->sli, mbxdata, SLI4_BMBX_SIZE,
						   &new_cb_arg->payload,
						   0, pci_descriptor_count+1, 1);

	/* Point dst to the first descriptor entry in the SET_PROFILE_CONFIG command */
	dst = (uint8_t *)&(((sli4_req_common_set_profile_config_t *) new_cb_arg->payload.virt)->desc);

	/* Loop over all descriptors.  If the descriptor is a PCIe descriptor, copy it
	 * to the SET_PROFILE_CONFIG command to be written back.  If it's the descriptor
	 * that we're trying to change also set its pf_type.
	 */
	desc_p = (sli4_resource_descriptor_v1_t *)response->desc;
	for (i=0; i<num_descriptors; i++) {
		if (desc_p->descriptor_type == SLI4_RESOURCE_DESCRIPTOR_TYPE_PCIE) {
			pcie_desc_p = (sli4_pcie_resource_descriptor_v1_t*) desc_p;
			if (pcie_desc_p->pf_number == cb_arg->pci_func) {
				/* This is the PCIe descriptor for this OCS instance.
				 * Update it with the new pf_type */
				switch(new_protocol) {
				case OCS_HW_PORT_PROTOCOL_FC:
					pcie_desc_p->pf_type = SLI4_PROTOCOL_FC;
					break;
				case OCS_HW_PORT_PROTOCOL_FCOE:
					pcie_desc_p->pf_type = SLI4_PROTOCOL_FCOE;
					break;
				case OCS_HW_PORT_PROTOCOL_ISCSI:
					pcie_desc_p->pf_type = SLI4_PROTOCOL_ISCSI;
					break;
				default:
					pcie_desc_p->pf_type = SLI4_PROTOCOL_DEFAULT;
					break;
				}

			}

			if (pcie_desc_p->pf_type == SLI4_PROTOCOL_FCOE) {
				++num_fcoe_ports;
			}
			if (pcie_desc_p->pf_type == SLI4_PROTOCOL_ISCSI) {
				++num_iscsi_ports;
			}
			ocs_memcpy(dst, pcie_desc_p, sizeof(sli4_pcie_resource_descriptor_v1_t));
			dst += sizeof(sli4_pcie_resource_descriptor_v1_t);
		}

		desc_p = (sli4_resource_descriptor_v1_t *) ((uint8_t *)desc_p + desc_p->descriptor_length);
	}

	/* Create an ISAP resource descriptor */
	isap_desc_p = (sli4_isap_resouce_descriptor_v1_t*)dst;
	isap_desc_p->descriptor_type = SLI4_RESOURCE_DESCRIPTOR_TYPE_ISAP;
	isap_desc_p->descriptor_length = sizeof(sli4_isap_resouce_descriptor_v1_t);
	if (num_iscsi_ports > 0) {
		isap_desc_p->iscsi_tgt = 1;
		isap_desc_p->iscsi_ini = 1;
		isap_desc_p->iscsi_dif = 1;
	}
	if (num_fcoe_ports > 0) {
		isap_desc_p->fcoe_tgt = 1;
		isap_desc_p->fcoe_ini = 1;
		isap_desc_p->fcoe_dif = 1;
	}

	/* At this point we're done with the memory allocated by ocs_port_set_protocol */
	ocs_dma_free(hw->os, &cb_arg->payload);
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_set_port_protocol_cb_arg_t));


	/* Send a SET_PROFILE_CONFIG mailbox command with the new descriptors */
	rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_set_port_protocol_cb2, new_cb_arg);
	if (rc) {
		ocs_log_err(hw->os, "Error posting COMMON_SET_PROFILE_CONFIG\n");
		/* Call the upper level callback to report a failure */
		if (new_cb_arg->cb) {
			new_cb_arg->cb( rc, new_cb_arg->arg);
		}

		/* Free the memory allocated by this function */
		ocs_dma_free(hw->os, &new_cb_arg->payload);
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, new_cb_arg, sizeof(ocs_hw_set_port_protocol_cb_arg_t));
	}


	return rc;
}

/**
 * @ingroup io
 * @brief  Set the port protocol.
 * @par Description
 * Setting the port protocol is a read-modify-write operation.
 * This function submits a GET_PROFILE_CONFIG command to read
 * the current settings.  The callback function will modify the
 * settings and issue the write.
 *
 * On successful completion this function will have allocated
 * two regular memory areas and one dma area which will need to
 * get freed later in the callbacks.
 *
 * @param hw Hardware context.
 * @param new_protocol New protocol to use.
 * @param pci_func PCI function to configure.
 * @param cb Callback function to be called when the command completes.
 * @param ul_arg An argument that is passed to the callback function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
ocs_hw_rtn_e
ocs_hw_set_port_protocol(ocs_hw_t *hw, ocs_hw_port_protocol_e new_protocol,
		uint32_t pci_func, ocs_set_port_protocol_cb_t cb, void *ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_set_port_protocol_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;

	/* Only supported on Skyhawk */
	if (sli_get_if_type(&hw->sli) != SLI4_IF_TYPE_BE3_SKH_PF) {
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}


	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_set_port_protocol_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;
	cb_arg->new_protocol = new_protocol;
	cb_arg->pci_func = pci_func;

	/* dma_mem holds the non-embedded portion */
	if (ocs_dma_alloc(hw->os, &cb_arg->payload, 4096, 4)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_port_protocol_cb_arg_t));
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_common_get_profile_config(&hw->sli, mbxdata, SLI4_BMBX_SIZE, &cb_arg->payload)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_set_port_protocol_cb1, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "GET_PROFILE_CONFIG failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_fw_write_cb_arg_t));
		ocs_dma_free(hw->os, &cb_arg->payload);
	}

	return rc;
}

typedef struct ocs_hw_get_profile_list_cb_arg_s {
	ocs_get_profile_list_cb_t cb;
	void *arg;
	ocs_dma_t payload;
} ocs_hw_get_profile_list_cb_arg_t;

/**
 * @brief Called for the completion of get_profile_list for a
 *        user request.
 * @par Description
 * This function is called when the COMMMON_GET_PROFILE_LIST
 * mailbox completes.  The response will be in
 * ctx->non_embedded_mem.virt.  This function parses the
 * response and creates a ocs_hw_profile_list, then calls the
 * mgmt_cb callback function and passes that list to it.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_get_profile_list_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_profile_list_t *list;
	ocs_hw_get_profile_list_cb_arg_t *cb_arg = arg;
	ocs_dma_t *payload = &(cb_arg->payload);
	sli4_res_common_get_profile_list_t *response = (sli4_res_common_get_profile_list_t *)payload->virt;
	int i;
	int num_descriptors;

	list = ocs_malloc(hw->os, sizeof(ocs_hw_profile_list_t), OCS_M_ZERO);
	if (list == NULL) {
		ocs_log_err(hw->os, "failed to malloc list\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	list->num_descriptors = response->profile_descriptor_count;

	num_descriptors = list->num_descriptors;
	if (num_descriptors > OCS_HW_MAX_PROFILES) {
		num_descriptors = OCS_HW_MAX_PROFILES;
	}

	for (i=0; i<num_descriptors; i++) {
		list->descriptors[i].profile_id = response->profile_descriptor[i].profile_id;
		list->descriptors[i].profile_index = response->profile_descriptor[i].profile_index;
		ocs_strcpy(list->descriptors[i].profile_description, (char *)response->profile_descriptor[i].profile_description);
	}

	if (cb_arg->cb) {
		cb_arg->cb(status, list, cb_arg->arg);
	} else {
		ocs_free(hw->os, list, sizeof(*list));
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_dma_free(hw->os, &cb_arg->payload);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_profile_list_cb_arg_t));

	return 0;
}

/**
 * @ingroup io
 * @brief  Get a list of available profiles.
 * @par Description
 * Issues a SLI-4 COMMON_GET_PROFILE_LIST mailbox.  When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param cb Callback function to be called when the
 *      	  command completes.
 * @param ul_arg An argument that is passed to the callback
 *      	 function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
ocs_hw_rtn_e
ocs_hw_get_profile_list(ocs_hw_t *hw, ocs_get_profile_list_cb_t cb, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_get_profile_list_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* Only supported on Skyhawk */
	if (sli_get_if_type(&hw->sli) != SLI4_IF_TYPE_BE3_SKH_PF) {
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}


	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_get_profile_list_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;

	/* dma_mem holds the non-embedded portion */
	if (ocs_dma_alloc(hw->os, &cb_arg->payload, sizeof(sli4_res_common_get_profile_list_t), 4)) {
		ocs_log_err(hw->os, "Failed to allocate DMA buffer\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_profile_list_cb_arg_t));
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_common_get_profile_list(&hw->sli, mbxdata, SLI4_BMBX_SIZE, 0, &cb_arg->payload)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_get_profile_list_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "GET_PROFILE_LIST failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_dma_free(hw->os, &cb_arg->payload);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_profile_list_cb_arg_t));
	}

	return rc;
}

typedef struct ocs_hw_get_active_profile_cb_arg_s {
	ocs_get_active_profile_cb_t cb;
	void *arg;
} ocs_hw_get_active_profile_cb_arg_t;

/**
 * @brief Called for the completion of get_active_profile for a
 *        user request.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_get_active_profile_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_get_active_profile_cb_arg_t *cb_arg = arg;
	sli4_cmd_sli_config_t* mbox_rsp = (sli4_cmd_sli_config_t*) mqe;
	sli4_res_common_get_active_profile_t* response = (sli4_res_common_get_active_profile_t*) mbox_rsp->payload.embed;
	uint32_t active_profile;

	active_profile = response->active_profile_id;

	if (cb_arg->cb) {
		cb_arg->cb(status, active_profile, cb_arg->arg);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_active_profile_cb_arg_t));

	return 0;
}

/**
 * @ingroup io
 * @brief  Get the currently active profile.
 * @par Description
 * Issues a SLI-4 COMMON_GET_ACTIVE_PROFILE mailbox. When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param cb Callback function to be called when the
 *	     command completes.
 * @param ul_arg An argument that is passed to the callback
 *      	 function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
int32_t
ocs_hw_get_active_profile(ocs_hw_t *hw, ocs_get_active_profile_cb_t cb, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_get_active_profile_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* Only supported on Skyhawk */
	if (sli_get_if_type(&hw->sli) != SLI4_IF_TYPE_BE3_SKH_PF) {
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_get_active_profile_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;

	if (sli_cmd_common_get_active_profile(&hw->sli, mbxdata, SLI4_BMBX_SIZE)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_get_active_profile_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "GET_ACTIVE_PROFILE failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_active_profile_cb_arg_t));
	}

	return rc;
}

typedef struct ocs_hw_get_nvparms_cb_arg_s {
	ocs_get_nvparms_cb_t cb;
	void *arg;
} ocs_hw_get_nvparms_cb_arg_t;

/**
 * @brief Called for the completion of get_nvparms for a
 *        user request.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE.
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return 0 on success, non-zero otherwise
 */
static int32_t
ocs_hw_get_nvparms_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_get_nvparms_cb_arg_t *cb_arg = arg;
	sli4_cmd_read_nvparms_t* mbox_rsp = (sli4_cmd_read_nvparms_t*) mqe;

	if (cb_arg->cb) {
		cb_arg->cb(status, mbox_rsp->wwpn, mbox_rsp->wwnn, mbox_rsp->hard_alpa,
				mbox_rsp->preferred_d_id, cb_arg->arg);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_nvparms_cb_arg_t));

	return 0;
}

/**
 * @ingroup io
 * @brief  Read non-volatile parms.
 * @par Description
 * Issues a SLI-4 READ_NVPARMS mailbox. When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param cb Callback function to be called when the
 *	  command completes.
 * @param ul_arg An argument that is passed to the callback
 *	  function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
int32_t
ocs_hw_get_nvparms(ocs_hw_t *hw, ocs_get_nvparms_cb_t cb, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_get_nvparms_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_get_nvparms_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;

	if (sli_cmd_read_nvparms(&hw->sli, mbxdata, SLI4_BMBX_SIZE)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_get_nvparms_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "READ_NVPARMS failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_nvparms_cb_arg_t));
	}

	return rc;
}

typedef struct ocs_hw_set_nvparms_cb_arg_s {
	ocs_set_nvparms_cb_t cb;
	void *arg;
} ocs_hw_set_nvparms_cb_arg_t;

/**
 * @brief Called for the completion of set_nvparms for a
 *        user request.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE.
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_set_nvparms_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_set_nvparms_cb_arg_t *cb_arg = arg;

	if (cb_arg->cb) {
		cb_arg->cb(status, cb_arg->arg);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_set_nvparms_cb_arg_t));

	return 0;
}

/**
 * @ingroup io
 * @brief  Write non-volatile parms.
 * @par Description
 * Issues a SLI-4 WRITE_NVPARMS mailbox. When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param cb Callback function to be called when the
 *	  command completes.
 * @param wwpn Port's WWPN in big-endian order, or NULL to use default.
 * @param wwnn Port's WWNN in big-endian order, or NULL to use default.
 * @param hard_alpa A hard AL_PA address setting used during loop
 * initialization. If no hard AL_PA is required, set to 0.
 * @param preferred_d_id A preferred D_ID address setting
 * that may be overridden with the CONFIG_LINK mailbox command.
 * If there is no preference, set to 0.
 * @param ul_arg An argument that is passed to the callback
 *	  function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
int32_t
ocs_hw_set_nvparms(ocs_hw_t *hw, ocs_set_nvparms_cb_t cb, uint8_t *wwpn,
		uint8_t *wwnn, uint8_t hard_alpa, uint32_t preferred_d_id, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_set_nvparms_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_set_nvparms_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;

	if (sli_cmd_write_nvparms(&hw->sli, mbxdata, SLI4_BMBX_SIZE, wwpn, wwnn, hard_alpa, preferred_d_id)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_set_nvparms_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "SET_NVPARMS failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_set_nvparms_cb_arg_t));
	}

	return rc;
}



/**
 * @brief Called to obtain the count for the specified type.
 *
 * @param hw Hardware context.
 * @param io_count_type IO count type (inuse, free, wait_free).
 *
 * @return Returns the number of IOs on the specified list type.
 */
uint32_t
ocs_hw_io_get_count(ocs_hw_t *hw, ocs_hw_io_count_type_e io_count_type)
{
	ocs_hw_io_t *io = NULL;
	uint32_t count = 0;

	ocs_lock(&hw->io_lock);

	switch (io_count_type) {
	case OCS_HW_IO_INUSE_COUNT :
		ocs_list_foreach(&hw->io_inuse, io) {
			count++;
		}
		break;
	case OCS_HW_IO_FREE_COUNT :
		 ocs_list_foreach(&hw->io_free, io) {
			 count++;
		 }
		 break;
	case OCS_HW_IO_WAIT_FREE_COUNT :
		 ocs_list_foreach(&hw->io_wait_free, io) {
			 count++;
		 }
		 break;
	case OCS_HW_IO_PORT_OWNED_COUNT:
		 ocs_list_foreach(&hw->io_port_owned, io) {
			 count++;
		 }
		 break;
	case OCS_HW_IO_N_TOTAL_IO_COUNT :
		count = hw->config.n_io;
		break;
	}

	ocs_unlock(&hw->io_lock);

	return count;
}

/**
 * @brief Called to obtain the count of produced RQs.
 *
 * @param hw Hardware context.
 *
 * @return Returns the number of RQs produced.
 */
uint32_t
ocs_hw_get_rqes_produced_count(ocs_hw_t *hw)
{
	uint32_t count = 0;
	uint32_t i;
	uint32_t j;

	for (i = 0; i < hw->hw_rq_count; i++) {
		hw_rq_t *rq = hw->hw_rq[i];
		if (rq->rq_tracker != NULL) {
			for (j = 0; j < rq->entry_count; j++) {
				if (rq->rq_tracker[j] != NULL) {
					count++;
				}
			}
		}
	}

	return count;
}

typedef struct ocs_hw_set_active_profile_cb_arg_s {
	ocs_set_active_profile_cb_t cb;
	void *arg;
} ocs_hw_set_active_profile_cb_arg_t;

/**
 * @brief Called for the completion of set_active_profile for a
 *        user request.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_set_active_profile_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_set_active_profile_cb_arg_t *cb_arg = arg;

	if (cb_arg->cb) {
		cb_arg->cb(status, cb_arg->arg);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	ocs_free(hw->os, cb_arg, sizeof(ocs_hw_get_active_profile_cb_arg_t));

	return 0;
}

/**
 * @ingroup io
 * @brief  Set the currently active profile.
 * @par Description
 * Issues a SLI4 COMMON_GET_ACTIVE_PROFILE mailbox. When the
 * command completes the provided mgmt callback function is
 * called.
 *
 * @param hw Hardware context.
 * @param profile_id Profile ID to activate.
 * @param cb Callback function to be called when the command completes.
 * @param ul_arg An argument that is passed to the callback function.
 *
 * @return
 * - OCS_HW_RTN_SUCCESS on success.
 * - OCS_HW_RTN_NO_MEMORY if a malloc fails.
 * - OCS_HW_RTN_NO_RESOURCES if unable to get a command
 *   context.
 * - OCS_HW_RTN_ERROR on any other error.
 */
int32_t
ocs_hw_set_active_profile(ocs_hw_t *hw, ocs_set_active_profile_cb_t cb, uint32_t profile_id, void* ul_arg)
{
	uint8_t *mbxdata;
	ocs_hw_set_active_profile_cb_arg_t *cb_arg;
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;

	/* Only supported on Skyhawk */
	if (sli_get_if_type(&hw->sli) != SLI4_IF_TYPE_BE3_SKH_PF) {
		return OCS_HW_RTN_ERROR;
	}

	/* mbxdata holds the header of the command */
	mbxdata = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mbxdata == NULL) {
		ocs_log_err(hw->os, "failed to malloc mbox\n");
		return OCS_HW_RTN_NO_MEMORY;
	}


	/* cb_arg holds the data that will be passed to the callback on completion */
	cb_arg = ocs_malloc(hw->os, sizeof(ocs_hw_set_active_profile_cb_arg_t), OCS_M_NOWAIT);
	if (cb_arg == NULL) {
		ocs_log_err(hw->os, "failed to malloc cb_arg\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		return OCS_HW_RTN_NO_MEMORY;
	}

	cb_arg->cb = cb;
	cb_arg->arg = ul_arg;

	if (sli_cmd_common_set_active_profile(&hw->sli, mbxdata, SLI4_BMBX_SIZE, 0, profile_id)) {
		rc = ocs_hw_command(hw, mbxdata, OCS_CMD_NOWAIT, ocs_hw_set_active_profile_cb, cb_arg);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "SET_ACTIVE_PROFILE failed\n");
		ocs_free(hw->os, mbxdata, SLI4_BMBX_SIZE);
		ocs_free(hw->os, cb_arg, sizeof(ocs_hw_set_active_profile_cb_arg_t));
	}

	return rc;
}



/*
 * Private functions
 */

/**
 * @brief Update the queue hash with the ID and index.
 *
 * @param hash Pointer to hash table.
 * @param id ID that was created.
 * @param index The index into the hash object.
 */
static void
ocs_hw_queue_hash_add(ocs_queue_hash_t *hash, uint16_t id, uint16_t index)
{
	uint32_t	hash_index = id & (OCS_HW_Q_HASH_SIZE - 1);

	/*
	 * Since the hash is always bigger than the number of queues, then we
	 * never have to worry about an infinite loop.
	 */
	while(hash[hash_index].in_use) {
		hash_index = (hash_index + 1) & (OCS_HW_Q_HASH_SIZE - 1);
	}

	/* not used, claim the entry */
	hash[hash_index].id = id;
	hash[hash_index].in_use = 1;
	hash[hash_index].index = index;
}

/**
 * @brief Find index given queue ID.
 *
 * @param hash Pointer to hash table.
 * @param id ID to find.
 *
 * @return Returns the index into the HW cq array or -1 if not found.
 */
int32_t
ocs_hw_queue_hash_find(ocs_queue_hash_t *hash, uint16_t id)
{
	int32_t	rc = -1;
	int32_t	index = id & (OCS_HW_Q_HASH_SIZE - 1);

	/*
	 * Since the hash is always bigger than the maximum number of Qs, then we
	 * never have to worry about an infinite loop. We will always find an
	 * unused entry.
	 */
	do {
		if (hash[index].in_use &&
		    hash[index].id == id) {
			rc = hash[index].index;
		} else {
			index = (index + 1) & (OCS_HW_Q_HASH_SIZE - 1);
		}
	} while(rc == -1 && hash[index].in_use);

	return rc;
}

static int32_t
ocs_hw_domain_add(ocs_hw_t *hw, ocs_domain_t *domain)
{
	int32_t		rc = OCS_HW_RTN_ERROR;
	uint16_t	fcfi = UINT16_MAX;

	if ((hw == NULL) || (domain == NULL)) {
		ocs_log_err(NULL, "bad parameter hw=%p domain=%p\n",
				hw, domain);
		return OCS_HW_RTN_ERROR;
	}

	fcfi = domain->fcf_indicator;

	if (fcfi < SLI4_MAX_FCFI) {
		uint16_t	fcf_index = UINT16_MAX;

		ocs_log_debug(hw->os, "adding domain %p @ %#x\n",
				domain, fcfi);
		hw->domains[fcfi] = domain;

		/* HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB */
		if (hw->workaround.override_fcfi) {
			if (hw->first_domain_idx < 0) {
				hw->first_domain_idx = fcfi;
			}
		}

		fcf_index = domain->fcf;

		if (fcf_index < SLI4_MAX_FCF_INDEX) {
			ocs_log_debug(hw->os, "adding map of FCF index %d to FCFI %d\n",
				      fcf_index, fcfi);
			hw->fcf_index_fcfi[fcf_index] = fcfi;
			rc = OCS_HW_RTN_SUCCESS;
		} else {
			ocs_log_test(hw->os, "FCF index %d out of range (max %d)\n",
				     fcf_index, SLI4_MAX_FCF_INDEX);
			hw->domains[fcfi] = NULL;
		}
	} else {
		ocs_log_test(hw->os, "FCFI %#x out of range (max %#x)\n",
				fcfi, SLI4_MAX_FCFI);
	}

	return rc;
}

static int32_t
ocs_hw_domain_del(ocs_hw_t *hw, ocs_domain_t *domain)
{
	int32_t		rc = OCS_HW_RTN_ERROR;
	uint16_t	fcfi = UINT16_MAX;

	if ((hw == NULL) || (domain == NULL)) {
		ocs_log_err(NULL, "bad parameter hw=%p domain=%p\n",
				hw, domain);
		return OCS_HW_RTN_ERROR;
	}

	fcfi = domain->fcf_indicator;

	if (fcfi < SLI4_MAX_FCFI) {
		uint16_t	fcf_index = UINT16_MAX;

		ocs_log_debug(hw->os, "deleting domain %p @ %#x\n",
				domain, fcfi);

		if (domain != hw->domains[fcfi]) {
			ocs_log_test(hw->os, "provided domain %p does not match stored domain %p\n",
				     domain, hw->domains[fcfi]);
			return OCS_HW_RTN_ERROR;
		}

		hw->domains[fcfi] = NULL;

		/* HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB */
		if (hw->workaround.override_fcfi) {
			if (hw->first_domain_idx == fcfi) {
				hw->first_domain_idx = -1;
			}
		}

		fcf_index = domain->fcf;

		if (fcf_index < SLI4_MAX_FCF_INDEX) {
			if (hw->fcf_index_fcfi[fcf_index] == fcfi) {
				hw->fcf_index_fcfi[fcf_index] = 0;
				rc = OCS_HW_RTN_SUCCESS;
			} else {
				ocs_log_test(hw->os, "indexed FCFI %#x doesn't match provided %#x @ %d\n",
					     hw->fcf_index_fcfi[fcf_index], fcfi, fcf_index);
			}
		} else {
			ocs_log_test(hw->os, "FCF index %d out of range (max %d)\n",
				     fcf_index, SLI4_MAX_FCF_INDEX);
		}
	} else {
		ocs_log_test(hw->os, "FCFI %#x out of range (max %#x)\n",
				fcfi, SLI4_MAX_FCFI);
	}

	return rc;
}

ocs_domain_t *
ocs_hw_domain_get(ocs_hw_t *hw, uint16_t fcfi)
{

	if (hw == NULL) {
		ocs_log_err(NULL, "bad parameter hw=%p\n", hw);
		return NULL;
	}

	if (fcfi < SLI4_MAX_FCFI) {
		return hw->domains[fcfi];
	} else {
		ocs_log_test(hw->os, "FCFI %#x out of range (max %#x)\n",
				fcfi, SLI4_MAX_FCFI);
		return NULL;
	}
}

static ocs_domain_t *
ocs_hw_domain_get_indexed(ocs_hw_t *hw, uint16_t fcf_index)
{

	if (hw == NULL) {
		ocs_log_err(NULL, "bad parameter hw=%p\n", hw);
		return NULL;
	}

	if (fcf_index < SLI4_MAX_FCF_INDEX) {
		return ocs_hw_domain_get(hw, hw->fcf_index_fcfi[fcf_index]);
	} else {
		ocs_log_test(hw->os, "FCF index %d out of range (max %d)\n",
			     fcf_index, SLI4_MAX_FCF_INDEX);
		return NULL;
	}
}

/**
 * @brief Quaratine an IO by taking a reference count and adding it to the
 *        quarantine list. When the IO is popped from the list then the
 *        count is released and the IO MAY be freed depending on whether
 *        it is still referenced by the IO.
 *
 *        @n @b Note: BZ 160124 - If this is a target write or an initiator read using
 *        DIF, then we must add the XRI to a quarantine list until we receive
 *        4 more completions of this same type.
 *
 * @param hw Hardware context.
 * @param wq Pointer to the WQ associated with the IO object to quarantine.
 * @param io Pointer to the io object to quarantine.
 */
static void
ocs_hw_io_quarantine(ocs_hw_t *hw, hw_wq_t *wq, ocs_hw_io_t *io)
{
	ocs_quarantine_info_t *q_info = &wq->quarantine_info;
	uint32_t	index;
	ocs_hw_io_t	*free_io = NULL;

	/* return if the QX bit was clear */
	if (!io->quarantine) {
		return;
	}

	/* increment the IO refcount to prevent it from being freed before the quarantine is over */
	if (ocs_ref_get_unless_zero(&io->ref) == 0) {
		/* command no longer active */
		ocs_log_debug(hw ? hw->os : NULL,
			      "io not active xri=0x%x tag=0x%x\n",
			      io->indicator, io->reqtag);
		return;
	}

	sli_queue_lock(wq->queue);
		index = q_info->quarantine_index;
		free_io = q_info->quarantine_ios[index];
		q_info->quarantine_ios[index] = io;
		q_info->quarantine_index = (index + 1) % OCS_HW_QUARANTINE_QUEUE_DEPTH;
	sli_queue_unlock(wq->queue);

	if (free_io != NULL) {
		ocs_ref_put(&free_io->ref); /* ocs_ref_get(): same function */
	}
}

/**
 * @brief Process entries on the given completion queue.
 *
 * @param hw Hardware context.
 * @param cq Pointer to the HW completion queue object.
 *
 * @return None.
 */
void
ocs_hw_cq_process(ocs_hw_t *hw, hw_cq_t *cq)
{
	uint8_t		cqe[sizeof(sli4_mcqe_t)];
	uint16_t	rid = UINT16_MAX;
	sli4_qentry_e	ctype;		/* completion type */
	int32_t		status;
	uint32_t	n_processed = 0;
	time_t		tstart;
	time_t		telapsed;

	tstart = ocs_msectime();

	while (!sli_queue_read(&hw->sli, cq->queue, cqe)) {
		status = sli_cq_parse(&hw->sli, cq->queue, cqe, &ctype, &rid);
		/*
		 * The sign of status is significant. If status is:
		 * == 0 : call completed correctly and the CQE indicated success
		 *  > 0 : call completed correctly and the CQE indicated an error
		 *  < 0 : call failed and no information is available about the CQE
		 */
		if (status < 0) {
			if (status == -2) {
				/* Notification that an entry was consumed, but not completed */
				continue;
			}

			break;
		}

		switch (ctype) {
		case SLI_QENTRY_ASYNC:
			CPUTRACE("async");
			sli_cqe_async(&hw->sli, cqe);
			break;
		case SLI_QENTRY_MQ:
			/*
			 * Process MQ entry. Note there is no way to determine
			 * the MQ_ID from the completion entry.
			 */
			CPUTRACE("mq");
			ocs_hw_mq_process(hw, status, hw->mq);
			break;
		case SLI_QENTRY_OPT_WRITE_CMD:
			ocs_hw_rqpair_process_auto_xfr_rdy_cmd(hw, cq, cqe);
			break;
		case SLI_QENTRY_OPT_WRITE_DATA:
			ocs_hw_rqpair_process_auto_xfr_rdy_data(hw, cq, cqe);
			break;
		case SLI_QENTRY_WQ:
			CPUTRACE("wq");
			ocs_hw_wq_process(hw, cq, cqe, status, rid);
			break;
		case SLI_QENTRY_WQ_RELEASE: {
			uint32_t wq_id = rid;
			uint32_t index = ocs_hw_queue_hash_find(hw->wq_hash, wq_id);
			hw_wq_t *wq = hw->hw_wq[index];

			/* Submit any HW IOs that are on the WQ pending list */
			hw_wq_submit_pending(wq, wq->wqec_set_count);

			break;
		}

		case SLI_QENTRY_RQ:
			CPUTRACE("rq");
			ocs_hw_rqpair_process_rq(hw, cq, cqe);
			break;
		case SLI_QENTRY_XABT: {
			CPUTRACE("xabt");
			ocs_hw_xabt_process(hw, cq, cqe, rid);
			break;

		}
		default:
			ocs_log_test(hw->os, "unhandled ctype=%#x rid=%#x\n", ctype, rid);
			break;
		}

		n_processed++;
		if (n_processed == cq->queue->proc_limit) {
			break;
		}

		if (cq->queue->n_posted >= (cq->queue->posted_limit)) {
			sli_queue_arm(&hw->sli, cq->queue, FALSE);
		}
	}

	sli_queue_arm(&hw->sli, cq->queue, TRUE);

	if (n_processed > cq->queue->max_num_processed) {
		cq->queue->max_num_processed = n_processed;
	}
	telapsed = ocs_msectime() - tstart;
	if (telapsed > cq->queue->max_process_time) {
		cq->queue->max_process_time = telapsed;
	}
}

/**
 * @brief Process WQ completion queue entries.
 *
 * @param hw Hardware context.
 * @param cq Pointer to the HW completion queue object.
 * @param cqe Pointer to WQ completion queue.
 * @param status Completion status.
 * @param rid Resource ID (IO tag).
 *
 * @return none
 */
void
ocs_hw_wq_process(ocs_hw_t *hw, hw_cq_t *cq, uint8_t *cqe, int32_t status, uint16_t rid)
{
	hw_wq_callback_t *wqcb;

	ocs_queue_history_cqe(&hw->q_hist, SLI_QENTRY_WQ, (void *)cqe, ((sli4_fc_wcqe_t *)cqe)->status, cq->queue->id,
			      ((cq->queue->index - 1) & (cq->queue->length - 1)));

	if(rid == OCS_HW_REQUE_XRI_REGTAG) {
		if(status) {
			ocs_log_err(hw->os, "reque xri failed, status = %d \n", status);
		}
		return;
	}

	wqcb = ocs_hw_reqtag_get_instance(hw, rid);
	if (wqcb == NULL) {
		ocs_log_err(hw->os, "invalid request tag: x%x\n", rid);
		return;
	}

	if (wqcb->callback == NULL) {
		ocs_log_err(hw->os, "wqcb callback is NULL\n");
		return;
	}

	(*wqcb->callback)(wqcb->arg, cqe, status);
}

/**
 * @brief Process WQ completions for IO requests
 *
 * @param arg Generic callback argument
 * @param cqe Pointer to completion queue entry
 * @param status Completion status
 *
 * @par Description
 * @n @b Note:  Regarding io->reqtag, the reqtag is assigned once when HW IOs are initialized
 * in ocs_hw_setup_io(), and don't need to be returned to the hw->wq_reqtag_pool.
 *
 * @return None.
 */
static void
ocs_hw_wq_process_io(void *arg, uint8_t *cqe, int32_t status)
{
	ocs_hw_io_t *io = arg;
	ocs_hw_t *hw = io->hw;
	sli4_fc_wcqe_t *wcqe = (void *)cqe;
	uint32_t	len = 0;
	uint32_t ext = 0;
	uint8_t out_of_order_axr_cmd = 0;
	uint8_t out_of_order_axr_data = 0;
	uint8_t lock_taken = 0;
#if defined(OCS_DISC_SPIN_DELAY)
	uint32_t delay = 0;
	char prop_buf[32];
#endif

	/*
	 * For the primary IO, this will also be used for the
	 * response. So it is important to only set/clear this
	 * flag on the first data phase of the IO because
	 * subsequent phases will be done on the secondary XRI.
	 */
	if (io->quarantine && io->quarantine_first_phase) {
		io->quarantine = (wcqe->qx == 1);
		ocs_hw_io_quarantine(hw, io->wq, io);
	}
	io->quarantine_first_phase = FALSE;

	/* BZ 161832 - free secondary HW IO */
	if (io->sec_hio != NULL &&
	    io->sec_hio->quarantine) {
		/*
		 * If the quarantine flag is set on the
		 * IO, then set it on the secondary IO
		 * based on the quarantine XRI (QX) bit
		 * sent by the FW.
		 */
		io->sec_hio->quarantine = (wcqe->qx == 1);
		/* use the primary io->wq because it is not set on the secondary IO. */
		ocs_hw_io_quarantine(hw, io->wq, io->sec_hio);
	}

	ocs_hw_remove_io_timed_wqe(hw, io);

	/* clear xbusy flag if WCQE[XB] is clear */
	if (io->xbusy && wcqe->xb == 0) {
		io->xbusy = FALSE;
	}

	/* get extended CQE status */
	switch (io->type) {
	case OCS_HW_BLS_ACC:
	case OCS_HW_BLS_ACC_SID:
		break;
	case OCS_HW_ELS_REQ:
		sli_fc_els_did(&hw->sli, cqe, &ext);
		len = sli_fc_response_length(&hw->sli, cqe);
		break;
	case OCS_HW_ELS_RSP:
	case OCS_HW_ELS_RSP_SID:
	case OCS_HW_FC_CT_RSP:
		break;
	case OCS_HW_FC_CT:
		len = sli_fc_response_length(&hw->sli, cqe);
		break;
	case OCS_HW_IO_TARGET_WRITE:
		len = sli_fc_io_length(&hw->sli, cqe);
#if defined(OCS_DISC_SPIN_DELAY)
		if (ocs_get_property("disk_spin_delay", prop_buf, sizeof(prop_buf)) == 0) {
			delay = ocs_strtoul(prop_buf, 0, 0);
			ocs_udelay(delay);
		}
#endif
		break;
	case OCS_HW_IO_TARGET_READ:
		len = sli_fc_io_length(&hw->sli, cqe);
		/*
		 * if_type == 2 seems to return 0 "total length placed" on
		 * FCP_TSEND64_WQE completions. If this appears to happen,
		 * use the CTIO data transfer length instead.
		 */
		if (hw->workaround.retain_tsend_io_length && !len && !status) {
			len = io->length;
		}

		break;
	case OCS_HW_IO_TARGET_RSP:
		if(io->is_port_owned) {
			ocs_lock(&io->axr_lock);
			lock_taken = 1;
			if(io->axr_buf->call_axr_cmd) {
				out_of_order_axr_cmd = 1;
			}
			if(io->axr_buf->call_axr_data) {
				out_of_order_axr_data = 1;
			}
		}
		break;
	case OCS_HW_IO_INITIATOR_READ:
		len = sli_fc_io_length(&hw->sli, cqe);
		break;
	case OCS_HW_IO_INITIATOR_WRITE:
		len = sli_fc_io_length(&hw->sli, cqe);
		break;
	case OCS_HW_IO_INITIATOR_NODATA:
		break;
	case OCS_HW_IO_DNRX_REQUEUE:
		/* release the count for re-posting the buffer */
		//ocs_hw_io_free(hw, io);
		break;
	default:
		ocs_log_test(hw->os, "XXX unhandled io type %#x for XRI 0x%x\n",
			     io->type, io->indicator);
		break;
	}
	if (status) {
		ext = sli_fc_ext_status(&hw->sli, cqe);
		/* Emulate IAAB=0 for initiator WQEs only; i.e. automatically
		 * abort exchange if an error occurred and exchange is still busy.
		 */
		if (hw->config.i_only_aab &&
		    (ocs_hw_iotype_is_originator(io->type)) &&
		    (ocs_hw_wcqe_abort_needed(status, ext, wcqe->xb))) {
			ocs_hw_rtn_e rc;

			ocs_log_debug(hw->os, "aborting xri=%#x tag=%#x\n",
				      io->indicator, io->reqtag);
			/*
			 * Because the initiator will not issue another IO phase, then it is OK to issue the
			 * callback on the abort completion, but for consistency with the target, wait for the
			 * XRI_ABORTED CQE to issue the IO callback.
			 */
			rc = ocs_hw_io_abort(hw, io, TRUE, NULL, NULL);

			if (rc == OCS_HW_RTN_SUCCESS) {
				/* latch status to return after abort is complete */
				io->status_saved = 1;
				io->saved_status = status;
				io->saved_ext = ext;
				io->saved_len = len;
				goto exit_ocs_hw_wq_process_io;
			} else if (rc == OCS_HW_RTN_IO_ABORT_IN_PROGRESS) {
				/*
				 * Already being aborted by someone else (ABTS
				 * perhaps). Just fall through and return original
				 * error.
				 */
				ocs_log_debug(hw->os, "abort in progress xri=%#x tag=%#x\n",
					      io->indicator, io->reqtag);

			} else {
				/* Failed to abort for some other reason, log error */
				ocs_log_test(hw->os, "Failed to abort xri=%#x tag=%#x rc=%d\n",
					     io->indicator, io->reqtag, rc);
			}
		}

		/*
		 * If we're not an originator IO, and XB is set, then issue abort for the IO from within the HW
		 */
		if ( (! ocs_hw_iotype_is_originator(io->type)) && wcqe->xb) {
			ocs_hw_rtn_e rc;

			ocs_log_debug(hw->os, "aborting xri=%#x tag=%#x\n", io->indicator, io->reqtag);

			/*
			 * Because targets may send a response when the IO completes using the same XRI, we must
			 * wait for the XRI_ABORTED CQE to issue the IO callback
			 */
			rc = ocs_hw_io_abort(hw, io, FALSE, NULL, NULL);
			if (rc == OCS_HW_RTN_SUCCESS) {
				/* latch status to return after abort is complete */
				io->status_saved = 1;
				io->saved_status = status;
				io->saved_ext = ext;
				io->saved_len = len;
				goto exit_ocs_hw_wq_process_io;
			} else if (rc == OCS_HW_RTN_IO_ABORT_IN_PROGRESS) {
				/*
				 * Already being aborted by someone else (ABTS
				 * perhaps). Just fall through and return original
				 * error.
				 */
				ocs_log_debug(hw->os, "abort in progress xri=%#x tag=%#x\n",
					      io->indicator, io->reqtag);

			} else {
				/* Failed to abort for some other reason, log error */
				ocs_log_test(hw->os, "Failed to abort xri=%#x tag=%#x rc=%d\n",
					     io->indicator, io->reqtag, rc);
			}
		}
	}
	/* BZ 161832 - free secondary HW IO */
	if (io->sec_hio != NULL) {
		ocs_hw_io_free(hw, io->sec_hio);
		io->sec_hio = NULL;
	}

	if (io->done != NULL) {
		ocs_hw_done_t  done = io->done;
		void		*arg = io->arg;

		io->done = NULL;

		if (io->status_saved) {
			/* use latched status if exists */
			status = io->saved_status;
			len = io->saved_len;
			ext = io->saved_ext;
			io->status_saved = 0;
		}

		/* Restore default SGL */
		ocs_hw_io_restore_sgl(hw, io);
		done(io, io->rnode, len, status, ext, arg);
	}

	if(out_of_order_axr_cmd) {
		/* bounce enabled, single RQ, we snoop the ox_id to choose the cpuidx */
		if (hw->config.bounce) {
			fc_header_t *hdr = io->axr_buf->cmd_seq->header->dma.virt;
			uint32_t s_id = fc_be24toh(hdr->s_id);
			uint32_t d_id = fc_be24toh(hdr->d_id);
			uint32_t ox_id =  ocs_be16toh(hdr->ox_id);
			if (hw->callback.bounce != NULL) {
				(*hw->callback.bounce)(ocs_hw_unsol_process_bounce, io->axr_buf->cmd_seq, s_id, d_id, ox_id);
			}
		}else {
			hw->callback.unsolicited(hw->args.unsolicited, io->axr_buf->cmd_seq);
		}

		if(out_of_order_axr_data) {
			/* bounce enabled, single RQ, we snoop the ox_id to choose the cpuidx */
			if (hw->config.bounce) {
				fc_header_t *hdr = io->axr_buf->seq.header->dma.virt;
				uint32_t s_id = fc_be24toh(hdr->s_id);
				uint32_t d_id = fc_be24toh(hdr->d_id);
				uint32_t ox_id =  ocs_be16toh(hdr->ox_id);
				if (hw->callback.bounce != NULL) {
					(*hw->callback.bounce)(ocs_hw_unsol_process_bounce, &io->axr_buf->seq, s_id, d_id, ox_id);
				}
			}else {
				hw->callback.unsolicited(hw->args.unsolicited, &io->axr_buf->seq);
			}
		}
	}

exit_ocs_hw_wq_process_io:
	if(lock_taken) {
		ocs_unlock(&io->axr_lock);
	}
}

/**
 * @brief Process WQ completions for abort requests.
 *
 * @param arg Generic callback argument.
 * @param cqe Pointer to completion queue entry.
 * @param status Completion status.
 *
 * @return None.
 */
static void
ocs_hw_wq_process_abort(void *arg, uint8_t *cqe, int32_t status)
{
	ocs_hw_io_t *io = arg;
	ocs_hw_t *hw = io->hw;
	uint32_t ext = 0;
	uint32_t len = 0;
	hw_wq_callback_t *wqcb;

	/*
	 * For IOs that were aborted internally, we may need to issue the callback here depending
	 * on whether a XRI_ABORTED CQE is expected ot not. If the status is Local Reject/No XRI, then
	 * issue the callback now.
	*/
	ext = sli_fc_ext_status(&hw->sli, cqe);
	if (status == SLI4_FC_WCQE_STATUS_LOCAL_REJECT &&
	    ext == SLI4_FC_LOCAL_REJECT_NO_XRI &&
		io->done != NULL) {
		ocs_hw_done_t  done = io->done;
		void		*arg = io->arg;

		io->done = NULL;

		/*
		 * Use latched status as this is always saved for an internal abort
		 *
		 * Note: We wont have both a done and abort_done function, so don't worry about
		 *       clobbering the len, status and ext fields.
		 */
		status = io->saved_status;
		len = io->saved_len;
		ext = io->saved_ext;
		io->status_saved = 0;
		done(io, io->rnode, len, status, ext, arg);
	}

	if (io->abort_done != NULL) {
		ocs_hw_done_t  done = io->abort_done;
		void		*arg = io->abort_arg;

		io->abort_done = NULL;

		done(io, io->rnode, len, status, ext, arg);
	}
	ocs_lock(&hw->io_abort_lock);
		/* clear abort bit to indicate abort is complete */
		io->abort_in_progress = 0;
	ocs_unlock(&hw->io_abort_lock);

	/* Free the WQ callback */
	ocs_hw_assert(io->abort_reqtag != UINT32_MAX);
	wqcb = ocs_hw_reqtag_get_instance(hw, io->abort_reqtag);
	ocs_hw_reqtag_free(hw, wqcb);

	/*
	 * Call ocs_hw_io_free() because this releases the WQ reservation as
	 * well as doing the refcount put. Don't duplicate the code here.
	 */
	(void)ocs_hw_io_free(hw, io);
}

/**
 * @brief Process XABT completions
 *
 * @param hw Hardware context.
 * @param cq Pointer to the HW completion queue object.
 * @param cqe Pointer to WQ completion queue.
 * @param rid Resource ID (IO tag).
 *
 *
 * @return None.
 */
void
ocs_hw_xabt_process(ocs_hw_t *hw, hw_cq_t *cq, uint8_t *cqe, uint16_t rid)
{
	/* search IOs wait free list */
	ocs_hw_io_t *io = NULL;

	io = ocs_hw_io_lookup(hw, rid);

	ocs_queue_history_cqe(&hw->q_hist, SLI_QENTRY_XABT, (void *)cqe, 0, cq->queue->id,
			      ((cq->queue->index - 1) & (cq->queue->length - 1)));
	if (io == NULL) {
		/* IO lookup failure should never happen */
		ocs_log_err(hw->os, "Error: xabt io lookup failed rid=%#x\n", rid);
		return;
	}

	if (!io->xbusy) {
		ocs_log_debug(hw->os, "xabt io not busy rid=%#x\n", rid);
	} else {
		/* mark IO as no longer busy */
		io->xbusy = FALSE;
	}

       if (io->is_port_owned) {
               ocs_lock(&hw->io_lock);
               /* Take reference so that below callback will not free io before reque */
               ocs_ref_get(&io->ref);
               ocs_unlock(&hw->io_lock);
       }



	/* For IOs that were aborted internally, we need to issue any pending callback here. */
	if (io->done != NULL) {
		ocs_hw_done_t  done = io->done;
		void		*arg = io->arg;

		/* Use latched status as this is always saved for an internal abort */
		int32_t status = io->saved_status;
		uint32_t len = io->saved_len;
		uint32_t ext = io->saved_ext;

		io->done = NULL;
		io->status_saved = 0;

		done(io, io->rnode, len, status, ext, arg);
	}

	/* Check to see if this is a port owned XRI */
	if (io->is_port_owned) {
		ocs_lock(&hw->io_lock);
		ocs_hw_reque_xri(hw, io);
		ocs_unlock(&hw->io_lock);
		/* Not hanlding reque xri completion, free io */
		ocs_hw_io_free(hw, io);
		return;
	}

	ocs_lock(&hw->io_lock);
		if ((io->state == OCS_HW_IO_STATE_INUSE) || (io->state == OCS_HW_IO_STATE_WAIT_FREE)) {
			/* if on wait_free list, caller has already freed IO;
			 * remove from wait_free list and add to free list.
			 * if on in-use list, already marked as no longer busy;
			 * just leave there and wait for caller to free.
			 */
			if (io->state == OCS_HW_IO_STATE_WAIT_FREE) {
				io->state = OCS_HW_IO_STATE_FREE;
				ocs_list_remove(&hw->io_wait_free, io);
				ocs_hw_io_free_move_correct_list(hw, io);
			}
		}
	ocs_unlock(&hw->io_lock);
}

/**
 * @brief Adjust the number of WQs and CQs within the HW.
 *
 * @par Description
 * Calculates the number of WQs and associated CQs needed in the HW based on
 * the number of IOs. Calculates the starting CQ index for each WQ, RQ and
 * MQ.
 *
 * @param hw Hardware context allocated by the caller.
 */
static void
ocs_hw_adjust_wqs(ocs_hw_t *hw)
{
	uint32_t max_wq_num = sli_get_max_queue(&hw->sli, SLI_QTYPE_WQ);
	uint32_t max_wq_entries = hw->num_qentries[SLI_QTYPE_WQ];
	uint32_t max_cq_entries = hw->num_qentries[SLI_QTYPE_CQ];

	/*
	 * possibly adjust the the size of the WQs so that the CQ is twice as
	 * big as the WQ to allow for 2 completions per IO. This allows us to
	 * handle multi-phase as well as aborts.
	 */
	if (max_cq_entries < max_wq_entries * 2) {
		max_wq_entries = hw->num_qentries[SLI_QTYPE_WQ] = max_cq_entries / 2;
	}

	/*
	 * Calculate the number of WQs to use base on the number of IOs.
	 *
	 * Note: We need to reserve room for aborts which must be sent down
	 *       the same WQ as the IO. So we allocate enough WQ space to
	 *       handle 2 times the number of IOs. Half of the space will be
	 *       used for normal IOs and the other hwf is reserved for aborts.
	 */
	hw->config.n_wq = ((hw->config.n_io * 2) + (max_wq_entries - 1)) / max_wq_entries;

	/*
	 * For performance reasons, it is best to use use a minimum of 4 WQs
	 * for BE3 and Skyhawk.
	 */
	if (hw->config.n_wq < 4 &&
	    SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) {
		hw->config.n_wq = 4;
	}

	/*
	 * For dual-chute support, we need to have at least one WQ per chute.
	 */
	if (hw->config.n_wq < 2 &&
	    ocs_hw_get_num_chutes(hw) > 1) {
		hw->config.n_wq = 2;
	}

	/* make sure we haven't exceeded the max supported in the HW */
	if (hw->config.n_wq > OCS_HW_MAX_NUM_WQ) {
		hw->config.n_wq = OCS_HW_MAX_NUM_WQ;
	}

	/* make sure we haven't exceeded the chip maximum */
	if (hw->config.n_wq > max_wq_num) {
		hw->config.n_wq = max_wq_num;
	}

	/*
	 * Using Queue Topology string, we divide by number of chutes
	 */
	hw->config.n_wq /= ocs_hw_get_num_chutes(hw);
}

static int32_t
ocs_hw_command_process(ocs_hw_t *hw, int32_t status, uint8_t *mqe, size_t size)
{
	ocs_command_ctx_t *ctx = NULL;

	ocs_lock(&hw->cmd_lock);
		if (NULL == (ctx = ocs_list_remove_head(&hw->cmd_head))) {
			ocs_log_err(hw->os, "XXX no command context?!?\n");
			ocs_unlock(&hw->cmd_lock);
			return -1;
		}

		hw->cmd_head_count--;

		/* Post any pending requests */
		ocs_hw_cmd_submit_pending(hw);

	ocs_unlock(&hw->cmd_lock);

	if (ctx->cb) {
		if (ctx->buf) {
			ocs_memcpy(ctx->buf, mqe, size);
		}
		ctx->cb(hw, status, ctx->buf, ctx->arg);
	}

	ocs_memset(ctx, 0, sizeof(ocs_command_ctx_t));
	ocs_free(hw->os, ctx, sizeof(ocs_command_ctx_t));

	return 0;
}




/**
 * @brief Process entries on the given mailbox queue.
 *
 * @param hw Hardware context.
 * @param status CQE status.
 * @param mq Pointer to the mailbox queue object.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_mq_process(ocs_hw_t *hw, int32_t status, sli4_queue_t *mq)
{
	uint8_t		mqe[SLI4_BMBX_SIZE];

	if (!sli_queue_read(&hw->sli, mq, mqe)) {
		ocs_hw_command_process(hw, status, mqe, mq->size);
	}

	return 0;
}

/**
 * @brief Read a FCF table entry.
 *
 * @param hw Hardware context.
 * @param index Table index to read. Use SLI4_FCOE_FCF_TABLE_FIRST for the first
 * read and the next_index field from the FCOE_READ_FCF_TABLE command
 * for subsequent reads.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static ocs_hw_rtn_e
ocs_hw_read_fcf(ocs_hw_t *hw, uint32_t index)
{
	uint8_t		*buf = NULL;
	int32_t		rc = OCS_HW_RTN_ERROR;

	buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (!buf) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	if (sli_cmd_fcoe_read_fcf_table(&hw->sli, buf, SLI4_BMBX_SIZE, &hw->fcf_dmem,
			index)) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, ocs_hw_cb_read_fcf, &hw->fcf_dmem);
	}

	if (rc != OCS_HW_RTN_SUCCESS) {
		ocs_log_test(hw->os, "FCOE_READ_FCF_TABLE failed\n");
		ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
	}

	return rc;
}

/**
 * @brief Callback function for the FCOE_READ_FCF_TABLE command.
 *
 * @par Description
 * Note that the caller has allocated:
 *  - DMA memory to hold the table contents
 *  - DMA memory structure
 *  - Command/results buffer
 *  .
 * Each of these must be freed here.
 *
 * @param hw Hardware context.
 * @param status Hardware status.
 * @param mqe Pointer to the mailbox command/results buffer.
 * @param arg Pointer to the DMA memory structure.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_cb_read_fcf(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_dma_t	*dma = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;

	if (status || hdr->status) {
		ocs_log_test(hw->os, "bad status cqe=%#x mqe=%#x\n",
				status, hdr->status);
	} else if (dma->virt) {
		sli4_res_fcoe_read_fcf_table_t *read_fcf = dma->virt;

		/* if FC or FCOE and FCF entry valid, process it */
		if (read_fcf->fcf_entry.fc ||
				(read_fcf->fcf_entry.val && !read_fcf->fcf_entry.sol)) {
			if (hw->callback.domain != NULL) {
				ocs_domain_record_t drec = {0};

				if (read_fcf->fcf_entry.fc) {
					/*
					 * This is a pseudo FCF entry. Create a domain
					 * record based on the read topology information
					 */
					drec.speed = hw->link.speed;
					drec.fc_id = hw->link.fc_id;
					drec.is_fc = TRUE;
					if (SLI_LINK_TOPO_LOOP == hw->link.topology) {
						drec.is_loop = TRUE;
						ocs_memcpy(drec.map.loop, hw->link.loop_map,
							   sizeof(drec.map.loop));
					} else if (SLI_LINK_TOPO_NPORT == hw->link.topology) {
						drec.is_nport = TRUE;
					}
				} else {
					drec.index = read_fcf->fcf_entry.fcf_index;
					drec.priority = read_fcf->fcf_entry.fip_priority;

					/* copy address, wwn and vlan_bitmap */
					ocs_memcpy(drec.address, read_fcf->fcf_entry.fcf_mac_address,
						   sizeof(drec.address));
					ocs_memcpy(drec.wwn, read_fcf->fcf_entry.fabric_name_id,
						   sizeof(drec.wwn));
					ocs_memcpy(drec.map.vlan, read_fcf->fcf_entry.vlan_bitmap,
						   sizeof(drec.map.vlan));

					drec.is_ethernet = TRUE;
					drec.is_nport = TRUE;
				}

				hw->callback.domain(hw->args.domain,
						OCS_HW_DOMAIN_FOUND,
						&drec);
			}
		} else {
			/* if FCOE and FCF is not valid, ignore it */
			ocs_log_test(hw->os, "ignore invalid FCF entry\n");
		}

		if (SLI4_FCOE_FCF_TABLE_LAST != read_fcf->next_index) {
			ocs_hw_read_fcf(hw, read_fcf->next_index);
		}
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	//ocs_dma_free(hw->os, dma);
	//ocs_free(hw->os, dma, sizeof(ocs_dma_t));

	return 0;
}

/**
 * @brief Callback function for the SLI link events.
 *
 * @par Description
 * This function allocates memory which must be freed in its callback.
 *
 * @param ctx Hardware context pointer (that is, ocs_hw_t *).
 * @param e Event structure pointer (that is, sli4_link_event_t *).
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static int32_t
ocs_hw_cb_link(void *ctx, void *e)
{
	ocs_hw_t	*hw = ctx;
	sli4_link_event_t *event = e;
	ocs_domain_t	*d = NULL;
	uint32_t	i = 0;
	int32_t		rc = OCS_HW_RTN_ERROR;
	ocs_t 		*ocs = hw->os;

	ocs_hw_link_event_init(hw);

	switch (event->status) {
	case SLI_LINK_STATUS_UP:

		hw->link = *event;

		if (SLI_LINK_TOPO_NPORT == event->topology) {
			device_printf(ocs->dev, "Link Up, NPORT, speed is %d\n", event->speed);
			ocs_hw_read_fcf(hw, SLI4_FCOE_FCF_TABLE_FIRST);
		} else if (SLI_LINK_TOPO_LOOP == event->topology) {
			uint8_t	*buf = NULL;
			device_printf(ocs->dev, "Link Up, LOOP, speed is %d\n", event->speed);

			buf = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
			if (!buf) {
				ocs_log_err(hw->os, "no buffer for command\n");
				break;
			}

			if (sli_cmd_read_topology(&hw->sli, buf, SLI4_BMBX_SIZE, &hw->loop_map)) {
				rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, __ocs_read_topology_cb, NULL);
			}

			if (rc != OCS_HW_RTN_SUCCESS) {
				ocs_log_test(hw->os, "READ_TOPOLOGY failed\n");
				ocs_free(hw->os, buf, SLI4_BMBX_SIZE);
			}
		} else {
			device_printf(ocs->dev, "Link Up, unsupported topology (%#x), speed is %d\n",
					event->topology, event->speed);
		}
		break;
	case SLI_LINK_STATUS_DOWN:
		device_printf(ocs->dev, "Link Down\n");

		hw->link.status = event->status;

		for (i = 0; d = hw->domains[i], i < SLI4_MAX_FCFI; i++) {
			if (d != NULL &&
			    hw->callback.domain != NULL) {
				hw->callback.domain(hw->args.domain, OCS_HW_DOMAIN_LOST, d);
			}
		}
		break;
	default:
		ocs_log_test(hw->os, "unhandled link status %#x\n", event->status);
		break;
	}

	return 0;
}

static int32_t
ocs_hw_cb_fip(void *ctx, void *e)
{
	ocs_hw_t	*hw = ctx;
	ocs_domain_t	*domain = NULL;
	sli4_fip_event_t *event = e;

	/* Find the associated domain object */
	if (event->type == SLI4_FCOE_FIP_FCF_CLEAR_VLINK) {
		ocs_domain_t *d = NULL;
		uint32_t	i = 0;

		/* Clear VLINK is different from the other FIP events as it passes back
		 * a VPI instead of a FCF index. Check all attached SLI ports for a
		 * matching VPI */
		for (i = 0; d = hw->domains[i], i < SLI4_MAX_FCFI; i++) {
			if (d != NULL) {
				ocs_sport_t	*sport = NULL;

				ocs_list_foreach(&d->sport_list, sport) {
					if (sport->indicator == event->index) {
						domain = d;
						break;
					}
				}

				if (domain != NULL) {
					break;
				}
			}
		}
	} else {
		domain = ocs_hw_domain_get_indexed(hw, event->index);
	}

	switch (event->type) {
	case SLI4_FCOE_FIP_FCF_DISCOVERED:
		ocs_hw_read_fcf(hw, event->index);
		break;
	case SLI4_FCOE_FIP_FCF_DEAD:
		if (domain != NULL &&
		    hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain, OCS_HW_DOMAIN_LOST, domain);
		}
		break;
	case SLI4_FCOE_FIP_FCF_CLEAR_VLINK:
		if (domain != NULL &&
		    hw->callback.domain != NULL) {
			/*
			 * We will want to issue rediscover FCF when this domain is free'd  in order
			 * to invalidate the FCF table
			 */
			domain->req_rediscover_fcf = TRUE;
			hw->callback.domain(hw->args.domain, OCS_HW_DOMAIN_LOST, domain);
		}
		break;
	case SLI4_FCOE_FIP_FCF_MODIFIED:
		if (domain != NULL &&
		    hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain, OCS_HW_DOMAIN_LOST, domain);
		}

		ocs_hw_read_fcf(hw, event->index);
		break;
	default:
		ocs_log_test(hw->os, "unsupported event %#x\n", event->type);
	}

	return 0;
}

static int32_t
ocs_hw_cb_node_attach(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_remote_node_t *rnode = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_hw_remote_node_event_e	evt = 0;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status cqe=%#x mqe=%#x\n", status,
				hdr->status);
		ocs_atomic_sub_return(&hw->rpi_ref[rnode->index].rpi_count, 1);
		rnode->attached = FALSE;
		ocs_atomic_set(&hw->rpi_ref[rnode->index].rpi_attached, 0);
		evt = OCS_HW_NODE_ATTACH_FAIL;
	} else {
		rnode->attached = TRUE;
		ocs_atomic_set(&hw->rpi_ref[rnode->index].rpi_attached, 1);
		evt = OCS_HW_NODE_ATTACH_OK;
	}

	if (hw->callback.rnode != NULL) {
		hw->callback.rnode(hw->args.rnode, evt, rnode);
	}
	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return 0;
}

static int32_t
ocs_hw_cb_node_free(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_remote_node_t *rnode = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_hw_remote_node_event_e	evt = OCS_HW_NODE_FREE_FAIL;
	int32_t		rc = 0;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status cqe=%#x mqe=%#x\n", status,
				hdr->status);

		/*
		 * In certain cases, a non-zero MQE status is OK (all must be true):
		 *   - node is attached
		 *   - if High Login Mode is enabled, node is part of a node group
		 *   - status is 0x1400
		 */
		if (!rnode->attached || ((sli_get_hlm(&hw->sli) == TRUE) && !rnode->node_group) ||
				(hdr->status != SLI4_MBOX_STATUS_RPI_NOT_REG)) {
			rc = -1;
		}
	}

	if (rc == 0) {
		rnode->node_group = FALSE;
		rnode->attached = FALSE;

		if (ocs_atomic_read(&hw->rpi_ref[rnode->index].rpi_count) == 0) {
			ocs_atomic_set(&hw->rpi_ref[rnode->index].rpi_attached, 0);
		}

		evt = OCS_HW_NODE_FREE_OK;
	}

	if (hw->callback.rnode != NULL) {
		hw->callback.rnode(hw->args.rnode, evt, rnode);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return rc;
}

static int32_t
ocs_hw_cb_node_free_all(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_hw_remote_node_event_e	evt = OCS_HW_NODE_FREE_FAIL;
	int32_t		rc = 0;
	uint32_t	i;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status cqe=%#x mqe=%#x\n", status,
				hdr->status);
	} else {
		evt = OCS_HW_NODE_FREE_ALL_OK;
	}

	if (evt == OCS_HW_NODE_FREE_ALL_OK) {
		for (i = 0; i < sli_get_max_rsrc(&hw->sli, SLI_RSRC_FCOE_RPI); i++) {
			ocs_atomic_set(&hw->rpi_ref[i].rpi_count, 0);
		}

		if (sli_resource_reset(&hw->sli, SLI_RSRC_FCOE_RPI)) {
			ocs_log_test(hw->os, "FCOE_RPI free all failure\n");
			rc = -1;
		}
	}

	if (hw->callback.rnode != NULL) {
		hw->callback.rnode(hw->args.rnode, evt, NULL);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	return rc;
}

/**
 * @brief Initialize the pool of HW IO objects.
 *
 * @param hw Hardware context.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
static ocs_hw_rtn_e
ocs_hw_setup_io(ocs_hw_t *hw)
{
	uint32_t	i = 0;
	ocs_hw_io_t	*io = NULL;
	uintptr_t	xfer_virt = 0;
	uintptr_t	xfer_phys = 0;
	uint32_t	index;
	uint8_t		new_alloc = TRUE;

	if (NULL == hw->io) {
		hw->io = ocs_malloc(hw->os, hw->config.n_io * sizeof(ocs_hw_io_t *), OCS_M_ZERO | OCS_M_NOWAIT);

		if (NULL == hw->io) {
			ocs_log_err(hw->os, "IO pointer memory allocation failed, %d Ios at size %zu\n",
				    hw->config.n_io,
				    sizeof(ocs_hw_io_t *));
			return OCS_HW_RTN_NO_MEMORY;
		}
		for (i = 0; i < hw->config.n_io; i++) {
			hw->io[i] = ocs_malloc(hw->os, sizeof(ocs_hw_io_t),
						OCS_M_ZERO | OCS_M_NOWAIT);
			if (hw->io[i] == NULL) {
				ocs_log_err(hw->os, "IO(%d) memory allocation failed\n", i);
				goto error;
			}
		}

		/* Create WQE buffs for IO */
		hw->wqe_buffs = ocs_malloc(hw->os, hw->config.n_io * hw->sli.config.wqe_size,
				OCS_M_ZERO | OCS_M_NOWAIT);
		if (NULL == hw->wqe_buffs) {
			ocs_free(hw->os, hw->io, hw->config.n_io * sizeof(ocs_hw_io_t));
			ocs_log_err(hw->os, "%s: IO WQE buff allocation failed, %d Ios at size %zu\n",
					__func__, hw->config.n_io, hw->sli.config.wqe_size);
			return OCS_HW_RTN_NO_MEMORY;
		}

	} else {
		/* re-use existing IOs, including SGLs */
		new_alloc = FALSE;
	}

	if (new_alloc) {
		if (ocs_dma_alloc(hw->os, &hw->xfer_rdy,
					sizeof(fcp_xfer_rdy_iu_t) * hw->config.n_io,
					4/*XXX what does this need to be? */)) {
			ocs_log_err(hw->os, "XFER_RDY buffer allocation failed\n");
			return OCS_HW_RTN_NO_MEMORY;
		}
	}
	xfer_virt = (uintptr_t)hw->xfer_rdy.virt;
	xfer_phys = hw->xfer_rdy.phys;

	for (i = 0; i < hw->config.n_io; i++) {
		hw_wq_callback_t *wqcb;

		io = hw->io[i];

		/* initialize IO fields */
		io->hw = hw;

		/* Assign a WQE buff */
		io->wqe.wqebuf = &hw->wqe_buffs[i * hw->sli.config.wqe_size];

		/* Allocate the request tag for this IO */
		wqcb = ocs_hw_reqtag_alloc(hw, ocs_hw_wq_process_io, io);
		if (wqcb == NULL) {
			ocs_log_err(hw->os, "can't allocate request tag\n");
			return OCS_HW_RTN_NO_RESOURCES;
		}
		io->reqtag = wqcb->instance_index;

		/* Now for the fields that are initialized on each free */
		ocs_hw_init_free_io(io);

		/* The XB flag isn't cleared on IO free, so initialize it to zero here */
		io->xbusy = 0;

		if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_XRI, &io->indicator, &index)) {
			ocs_log_err(hw->os, "sli_resource_alloc failed @ %d\n", i);
			return OCS_HW_RTN_NO_MEMORY;
		}

		if (new_alloc && ocs_dma_alloc(hw->os, &io->def_sgl, hw->config.n_sgl * sizeof(sli4_sge_t), 64)) {
			ocs_log_err(hw->os, "ocs_dma_alloc failed @ %d\n", i);
			ocs_memset(&io->def_sgl, 0, sizeof(ocs_dma_t));
			return OCS_HW_RTN_NO_MEMORY;
		}
		io->def_sgl_count = hw->config.n_sgl;
		io->sgl = &io->def_sgl;
		io->sgl_count = io->def_sgl_count;

		if (hw->xfer_rdy.size) {
			io->xfer_rdy.virt = (void *)xfer_virt;
			io->xfer_rdy.phys = xfer_phys;
			io->xfer_rdy.size = sizeof(fcp_xfer_rdy_iu_t);

			xfer_virt += sizeof(fcp_xfer_rdy_iu_t);
			xfer_phys += sizeof(fcp_xfer_rdy_iu_t);
		}
	}

	return OCS_HW_RTN_SUCCESS;
error:
	for (i = 0; i < hw->config.n_io && hw->io[i]; i++) {
		ocs_free(hw->os, hw->io[i], sizeof(ocs_hw_io_t));
		hw->io[i] = NULL;
	}

	return OCS_HW_RTN_NO_MEMORY;
}

static ocs_hw_rtn_e
ocs_hw_init_io(ocs_hw_t *hw)
{
	uint32_t        i = 0, io_index = 0;
	uint32_t        prereg = 0;
	ocs_hw_io_t	*io = NULL;
	uint8_t		cmd[SLI4_BMBX_SIZE];
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint32_t	nremaining;
	uint32_t	n = 0;
	uint32_t	sgls_per_request = 256;
	ocs_dma_t	**sgls = NULL;
	ocs_dma_t	reqbuf = { 0 };

	prereg = sli_get_sgl_preregister(&hw->sli);

	if (prereg) {
		sgls = ocs_malloc(hw->os, sizeof(*sgls) * sgls_per_request, OCS_M_NOWAIT);
		if (sgls == NULL) {
			ocs_log_err(hw->os, "ocs_malloc sgls failed\n");
			return OCS_HW_RTN_NO_MEMORY;
		}

		rc = ocs_dma_alloc(hw->os, &reqbuf, 32 + sgls_per_request*16, OCS_MIN_DMA_ALIGNMENT);
		if (rc) {
			ocs_log_err(hw->os, "ocs_dma_alloc reqbuf failed\n");
			ocs_free(hw->os, sgls, sizeof(*sgls) * sgls_per_request);
			return OCS_HW_RTN_NO_MEMORY;
		}
	}

	io = hw->io[io_index];
	for (nremaining = hw->config.n_io; nremaining; nremaining -= n) {
		if (prereg) {
			/* Copy address of SGL's into local sgls[] array, break out if the xri
			 * is not contiguous.
			 */
			for (n = 0; n < MIN(sgls_per_request, nremaining); n++) {
				/* Check that we have contiguous xri values */
				if (n > 0) {
					if (hw->io[io_index + n]->indicator != (hw->io[io_index + n-1]->indicator+1)) {
						break;
					}
				}
				sgls[n] = hw->io[io_index + n]->sgl;
			}

			if (sli_cmd_fcoe_post_sgl_pages(&hw->sli, cmd, sizeof(cmd),
						io->indicator, n, sgls, NULL, &reqbuf)) {
				if (ocs_hw_command(hw, cmd, OCS_CMD_POLL, NULL, NULL)) {
					rc = OCS_HW_RTN_ERROR;
					ocs_log_err(hw->os, "SGL post failed\n");
					break;
				}
			}
		} else {
			n = nremaining;
		}

		/* Add to tail if successful */
		for (i = 0; i < n; i ++) {
			io->is_port_owned = 0;
			io->state = OCS_HW_IO_STATE_FREE;
			ocs_list_add_tail(&hw->io_free, io);
			io = hw->io[io_index+1];
			io_index++;
		}
	}

	if (prereg) {
		ocs_dma_free(hw->os, &reqbuf);
		ocs_free(hw->os, sgls, sizeof(*sgls) * sgls_per_request);
	}

	return rc;
}

static int32_t
ocs_hw_flush(ocs_hw_t *hw)
{
	uint32_t	i = 0;

	/* Process any remaining completions */
	for (i = 0; i < hw->eq_count; i++) {
		ocs_hw_process(hw, i, ~0);
	}

	return 0;
}

static int32_t
ocs_hw_command_cancel(ocs_hw_t *hw)
{

	ocs_lock(&hw->cmd_lock);

	/*
	 * Manually clean up remaining commands. Note: since this calls
	 * ocs_hw_command_process(), we'll also process the cmd_pending
	 * list, so no need to manually clean that out.
	 */
	while (!ocs_list_empty(&hw->cmd_head)) {
		uint8_t		mqe[SLI4_BMBX_SIZE] = { 0 };
		ocs_command_ctx_t *ctx = ocs_list_get_head(&hw->cmd_head);

		ocs_log_test(hw->os, "hung command %08x\n",
				NULL == ctx ? UINT32_MAX :
				(NULL == ctx->buf ? UINT32_MAX : *((uint32_t *)ctx->buf)));
		ocs_unlock(&hw->cmd_lock);
		ocs_hw_command_process(hw, -1/*Bad status*/, mqe, SLI4_BMBX_SIZE);
		ocs_lock(&hw->cmd_lock);
	}

	ocs_unlock(&hw->cmd_lock);

	return 0;
}

/**
 * @brief Find IO given indicator (xri).
 *
 * @param hw Hal context.
 * @param indicator Indicator (xri) to look for.
 *
 * @return Returns io if found, NULL otherwise.
 */
ocs_hw_io_t *
ocs_hw_io_lookup(ocs_hw_t *hw, uint32_t xri)
{
	uint32_t ioindex;
	ioindex = xri - hw->sli.config.extent[SLI_RSRC_FCOE_XRI].base[0];
	return hw->io[ioindex];
}

/**
 * @brief Issue any pending callbacks for an IO and remove off the timer and pending lists.
 *
 * @param hw Hal context.
 * @param io Pointer to the IO to cleanup.
 */
static void
ocs_hw_io_cancel_cleanup(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	ocs_hw_done_t  done = io->done;
	ocs_hw_done_t  abort_done = io->abort_done;

	/* first check active_wqe list and remove if there */
	if (ocs_list_on_list(&io->wqe_link)) {
		ocs_list_remove(&hw->io_timed_wqe, io);
	}

	/* Remove from WQ pending list */
	if ((io->wq != NULL) && ocs_list_on_list(&io->wq->pending_list)) {
		ocs_list_remove(&io->wq->pending_list, io);
	}

	if (io->done) {
		void		*arg = io->arg;

		io->done = NULL;
		ocs_unlock(&hw->io_lock);
		done(io, io->rnode, 0, SLI4_FC_WCQE_STATUS_SHUTDOWN, 0, arg);
		ocs_lock(&hw->io_lock);
	}

	if (io->abort_done != NULL) {
		void		*abort_arg = io->abort_arg;

		io->abort_done = NULL;
		ocs_unlock(&hw->io_lock);
		abort_done(io, io->rnode, 0, SLI4_FC_WCQE_STATUS_SHUTDOWN, 0, abort_arg);
		ocs_lock(&hw->io_lock);
	}
}

static int32_t
ocs_hw_io_cancel(ocs_hw_t *hw)
{
	ocs_hw_io_t	*io = NULL;
	ocs_hw_io_t	*tmp_io = NULL;
	uint32_t	iters = 100; /* One second limit */

	/*
	 * Manually clean up outstanding IO.
	 * Only walk through list once: the backend will cleanup any IOs when done/abort_done is called.
	 */
	ocs_lock(&hw->io_lock);
	ocs_list_foreach_safe(&hw->io_inuse, io, tmp_io) {
		ocs_hw_done_t  done = io->done;
		ocs_hw_done_t  abort_done = io->abort_done;

		ocs_hw_io_cancel_cleanup(hw, io);

		/*
		 * Since this is called in a reset/shutdown
		 * case, If there is no callback, then just
		 * free the IO.
		 *
		 * Note: A port owned XRI cannot be on
		 *       the in use list. We cannot call
		 *       ocs_hw_io_free() because we already
		 *       hold the io_lock.
		 */
		if (done == NULL &&
		    abort_done == NULL) {
			/*
			 * Since this is called in a reset/shutdown
			 * case, If there is no callback, then just
			 * free the IO.
			 */
			ocs_hw_io_free_common(hw, io);
			ocs_list_remove(&hw->io_inuse, io);
			ocs_hw_io_free_move_correct_list(hw, io);
		}
	}

	/*
	 * For port owned XRIs, they are not on the in use list, so
	 * walk though XRIs and issue any callbacks.
	 */
	ocs_list_foreach_safe(&hw->io_port_owned, io, tmp_io) {
		/* check  list and remove if there */
		if (ocs_list_on_list(&io->dnrx_link)) {
			ocs_list_remove(&hw->io_port_dnrx, io);
			ocs_ref_put(&io->ref); /* ocs_ref_get(): same function */
		}
		ocs_hw_io_cancel_cleanup(hw, io);
		ocs_list_remove(&hw->io_port_owned, io);
		ocs_hw_io_free_common(hw, io);
	}
	ocs_unlock(&hw->io_lock);

	/* Give time for the callbacks to complete */
	do {
		ocs_udelay(10000);
		iters--;
	} while (!ocs_list_empty(&hw->io_inuse) && iters);

	/* Leave a breadcrumb that cleanup is not yet complete. */
	if (!ocs_list_empty(&hw->io_inuse)) {
		ocs_log_test(hw->os, "io_inuse list is not empty\n");
	}

	return 0;
}

static int32_t
ocs_hw_io_ini_sge(ocs_hw_t *hw, ocs_hw_io_t *io, ocs_dma_t *cmnd, uint32_t cmnd_size,
		ocs_dma_t *rsp)
{
	sli4_sge_t	*data = NULL;

	if (!hw || !io) {
		ocs_log_err(NULL, "bad parm hw=%p io=%p\n", hw, io);
		return OCS_HW_RTN_ERROR;
	}

	data = io->def_sgl.virt;

	/* setup command pointer */
	data->buffer_address_high = ocs_addr32_hi(cmnd->phys);
	data->buffer_address_low  = ocs_addr32_lo(cmnd->phys);
	data->buffer_length = cmnd_size;
	data++;

	/* setup response pointer */
	data->buffer_address_high = ocs_addr32_hi(rsp->phys);
	data->buffer_address_low  = ocs_addr32_lo(rsp->phys);
	data->buffer_length = rsp->size;

	return 0;
}

static int32_t
__ocs_read_topology_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	sli4_cmd_read_topology_t *read_topo = (sli4_cmd_read_topology_t *)mqe;

	if (status || read_topo->hdr.status) {
		ocs_log_debug(hw->os, "bad status cqe=%#x mqe=%#x\n",
				status, read_topo->hdr.status);
		ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
		return -1;
	}

	switch (read_topo->attention_type) {
	case SLI4_READ_TOPOLOGY_LINK_UP:
		hw->link.status = SLI_LINK_STATUS_UP;
		break;
	case SLI4_READ_TOPOLOGY_LINK_DOWN:
		hw->link.status = SLI_LINK_STATUS_DOWN;
		break;
	case SLI4_READ_TOPOLOGY_LINK_NO_ALPA:
		hw->link.status = SLI_LINK_STATUS_NO_ALPA;
		break;
	default:
		hw->link.status = SLI_LINK_STATUS_MAX;
		break;
	}

	switch (read_topo->topology) {
	case SLI4_READ_TOPOLOGY_NPORT:
		hw->link.topology = SLI_LINK_TOPO_NPORT;
		break;
	case SLI4_READ_TOPOLOGY_FC_AL:
		hw->link.topology = SLI_LINK_TOPO_LOOP;
		if (SLI_LINK_STATUS_UP == hw->link.status) {
			hw->link.loop_map = hw->loop_map.virt;
		}
		hw->link.fc_id = read_topo->acquired_al_pa;
		break;
	default:
		hw->link.topology = SLI_LINK_TOPO_MAX;
		break;
	}

	hw->link.medium = SLI_LINK_MEDIUM_FC;

	switch (read_topo->link_current.link_speed) {
	case SLI4_READ_TOPOLOGY_SPEED_1G:
		hw->link.speed =  1 * 1000;
		break;
	case SLI4_READ_TOPOLOGY_SPEED_2G:
		hw->link.speed =  2 * 1000;
		break;
	case SLI4_READ_TOPOLOGY_SPEED_4G:
		hw->link.speed =  4 * 1000;
		break;
	case SLI4_READ_TOPOLOGY_SPEED_8G:
		hw->link.speed =  8 * 1000;
		break;
	case SLI4_READ_TOPOLOGY_SPEED_16G:
		hw->link.speed = 16 * 1000;
		hw->link.loop_map = NULL;
		break;
	case SLI4_READ_TOPOLOGY_SPEED_32G:
		hw->link.speed = 32 * 1000;
		hw->link.loop_map = NULL;
		break;
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);

	ocs_hw_read_fcf(hw, SLI4_FCOE_FCF_TABLE_FIRST);

	return 0;
}

static int32_t
__ocs_hw_port_common(const char *funcname, ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_EXIT:
		/* ignore */
		break;

	case OCS_EVT_HW_PORT_REQ_FREE:
	case OCS_EVT_HW_PORT_REQ_ATTACH:
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		/* fall through */
	default:
		ocs_log_test(hw->os, "%s %-20s not handled\n", funcname, ocs_sm_event_name(evt));
		break;
	}

	return 0;
}

static void *
__ocs_hw_port_free_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_FREE_FAIL, sport);
		}
		break;
	default:
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_freed(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free SLI resource */
		if (sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VPI, sport->indicator)) {
			ocs_log_err(hw->os, "FCOE_VPI free failure addr=%#x\n", sport->fc_id);
		}

		/* free mailbox buffer */
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_FREE_OK, sport);
		}
		break;
	default:
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_attach_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free SLI resource */
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VPI, sport->indicator);

		/* free mailbox buffer */
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}

		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_ATTACH_FAIL, sport);
		}
		if (sport->sm_free_req_pending) {
			ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		}
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_free_unreg_vpi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;
	uint8_t		*cmd = NULL;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* allocate memory and send unreg_vpi */
		cmd = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
		if (!cmd) {
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (0 == sli_cmd_unreg_vpi(&hw->sli, cmd, SLI4_BMBX_SIZE, sport->indicator,
					   SLI4_UNREG_TYPE_PORT)) {
			ocs_log_err(hw->os, "UNREG_VPI format failure\n");
			ocs_free(hw->os, cmd, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, cmd, OCS_CMD_NOWAIT, __ocs_hw_port_cb, sport)) {
			ocs_log_err(hw->os, "UNREG_VPI command failure\n");
			ocs_free(hw->os, cmd, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_port_freed, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_port_free_report_fail, data);
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_free_nop(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* Forward to execute in mailbox completion processing context */
		if (ocs_hw_async_call(hw, __ocs_hw_port_realloc_cb, sport)) {
			ocs_log_err(hw->os, "ocs_hw_async_call failed\n");
		}
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_port_freed, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_port_free_report_fail, data);
		break;
	default:
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_attached(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_ATTACH_OK, sport);
		}
		if (sport->sm_free_req_pending) {
			ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		}
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* virtual/physical port request free */
		ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_attach_reg_vpi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (0 == sli_cmd_reg_vpi(&hw->sli, data, SLI4_BMBX_SIZE, sport, FALSE)) {
			ocs_log_err(hw->os, "REG_VPI format failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_port_cb, sport)) {
			ocs_log_err(hw->os, "REG_VPI command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_port_attached, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_port_attach_report_fail, data);
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* Wait for attach response and then free */
		sport->sm_free_req_pending = 1;
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_done(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free SLI resource */
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VPI, sport->indicator);

		/* free mailbox buffer */
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_allocated(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_ALLOC_OK, sport);
		}
		/* If there is a pending free request, then handle it now */
		if (sport->sm_free_req_pending) {
			ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		}
		break;
	case OCS_EVT_HW_PORT_REQ_ATTACH:
		/* virtual port requests attach */
		ocs_sm_transition(ctx, __ocs_hw_port_attach_reg_vpi, data);
		break;
	case OCS_EVT_HW_PORT_ATTACH_OK:
		/* physical port attached (as part of attaching domain) */
		ocs_sm_transition(ctx, __ocs_hw_port_attached, data);
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* virtual port request free */
		if (SLI4_IF_TYPE_LANCER_FC_ETH == sli_get_if_type(&hw->sli)) {
			ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		} else {
			/*
			 * Note: BE3/Skyhawk will respond with a status of 0x20
			 *       unless the reg_vpi has been issued, so we can
			 *       skip the unreg_vpi for these adapters.
			 *
			 * Send a nop to make sure that free doesn't occur in
			 * same context
			 */
			ocs_sm_transition(ctx, __ocs_hw_port_free_nop, NULL);
		}
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_alloc_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free SLI resource */
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VPI, sport->indicator);

		/* free mailbox buffer */
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}

		if (hw->callback.port != NULL) {
			hw->callback.port(hw->args.port,
					OCS_HW_PORT_ALLOC_FAIL, sport);
		}

		/* If there is a pending free request, then handle it now */
		if (sport->sm_free_req_pending) {
			ocs_sm_transition(ctx, __ocs_hw_port_free_unreg_vpi, NULL);
		}
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_alloc_read_sparm64(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;
	uint8_t		*payload = NULL;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* allocate memory for the service parameters */
		if (ocs_dma_alloc(hw->os, &sport->dma, 112, 4)) {
			ocs_log_err(hw->os, "Failed to allocate DMA memory\n");
			ocs_sm_transition(ctx, __ocs_hw_port_done, data);
			break;
		}

		if (0 == sli_cmd_read_sparm64(&hw->sli, data, SLI4_BMBX_SIZE,
					&sport->dma, sport->indicator)) {
			ocs_log_err(hw->os, "READ_SPARM64 allocation failure\n");
			ocs_dma_free(hw->os, &sport->dma);
			ocs_sm_transition(ctx, __ocs_hw_port_done, data);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_port_cb, sport)) {
			ocs_log_err(hw->os, "READ_SPARM64 command failure\n");
			ocs_dma_free(hw->os, &sport->dma);
			ocs_sm_transition(ctx, __ocs_hw_port_done, data);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		payload = sport->dma.virt;

		ocs_display_sparams(sport->display_name, "sport sparm64", 0, NULL, payload);

		ocs_memcpy(&sport->sli_wwpn, payload + SLI4_READ_SPARM64_WWPN_OFFSET,
				sizeof(sport->sli_wwpn));
		ocs_memcpy(&sport->sli_wwnn, payload + SLI4_READ_SPARM64_WWNN_OFFSET,
				sizeof(sport->sli_wwnn));

		ocs_dma_free(hw->os, &sport->dma);
		ocs_sm_transition(ctx, __ocs_hw_port_alloc_init_vpi, data);
		break;
	case OCS_EVT_ERROR:
		ocs_dma_free(hw->os, &sport->dma);
		ocs_sm_transition(ctx, __ocs_hw_port_alloc_report_fail, data);
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* Wait for attach response and then free */
		sport->sm_free_req_pending = 1;
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_alloc_init(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* no-op */
		break;
	case OCS_EVT_HW_PORT_ALLOC_OK:
		ocs_sm_transition(ctx, __ocs_hw_port_allocated, NULL);
		break;
	case OCS_EVT_HW_PORT_ALLOC_FAIL:
		ocs_sm_transition(ctx, __ocs_hw_port_alloc_report_fail, NULL);
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* Wait for attach response and then free */
		sport->sm_free_req_pending = 1;
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_port_alloc_init_vpi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_sli_port_t	*sport = ctx->app;
	ocs_hw_t	*hw = sport->hw;

	smtrace("port");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* If there is a pending free request, then handle it now */
		if (sport->sm_free_req_pending) {
			ocs_sm_transition(ctx, __ocs_hw_port_freed, NULL);
			return NULL;
		}

		/* TODO XXX transitioning to done only works if this is called
		 * directly from ocs_hw_port_alloc BUT not if called from
		 * read_sparm64. In the later case, we actually want to go
		 * through report_ok/fail
		 */
		if (0 == sli_cmd_init_vpi(&hw->sli, data, SLI4_BMBX_SIZE,
					sport->indicator, sport->domain->indicator)) {
			ocs_log_err(hw->os, "INIT_VPI allocation failure\n");
			ocs_sm_transition(ctx, __ocs_hw_port_done, data);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_port_cb, sport)) {
			ocs_log_err(hw->os, "INIT_VPI command failure\n");
			ocs_sm_transition(ctx, __ocs_hw_port_done, data);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_port_allocated, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_port_alloc_report_fail, data);
		break;
	case OCS_EVT_HW_PORT_REQ_FREE:
		/* Wait for attach response and then free */
		sport->sm_free_req_pending = 1;
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_port_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static int32_t
__ocs_hw_port_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_sli_port_t *sport = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_sm_event_t	evt;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status vpi=%#x st=%x hdr=%x\n",
			      sport->indicator, status, hdr->status);
		evt = OCS_EVT_ERROR;
	} else {
		evt = OCS_EVT_RESPONSE;
	}

	ocs_sm_post_event(&sport->ctx, evt, mqe);

	return 0;
}

static int32_t
__ocs_hw_port_realloc_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_sli_port_t *sport = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_sm_event_t	evt;
	uint8_t *mqecpy;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status vpi=%#x st=%x hdr=%x\n",
			      sport->indicator, status, hdr->status);
		evt = OCS_EVT_ERROR;
	} else {
		evt = OCS_EVT_RESPONSE;
	}

	/*
	 * In this case we have to malloc a mailbox command buffer, as it is reused
	 * in the state machine post event call, and eventually freed
	 */
	mqecpy = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
	if (mqecpy == NULL) {
		ocs_log_err(hw->os, "malloc mqecpy failed\n");
		return -1;
	}
	ocs_memcpy(mqecpy, mqe, SLI4_BMBX_SIZE);

	ocs_sm_post_event(&sport->ctx, evt, mqecpy);

	return 0;
}

/***************************************************************************
 * Domain state machine
 */

static int32_t
__ocs_hw_domain_common(const char *funcname, ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_EXIT:
		/* ignore */
		break;

	default:
		ocs_log_test(hw->os, "%s %-20s not handled\n", funcname, ocs_sm_event_name(evt));
		break;
	}

	return 0;
}

static void *
__ocs_hw_domain_alloc_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free command buffer */
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		/* free SLI resources */
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VFI, domain->indicator);
		/* TODO how to free FCFI (or do we at all)? */

		if (hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain,
					OCS_HW_DOMAIN_ALLOC_FAIL,
					domain);
		}
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_attached(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free mailbox buffer and send alloc ok to physical sport */
		ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		ocs_sm_post_event(&domain->sport->ctx, OCS_EVT_HW_PORT_ATTACH_OK, NULL);

		/* now inform registered callbacks */
		if (hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain,
					OCS_HW_DOMAIN_ATTACH_OK,
					domain);
		}
		break;
	case OCS_EVT_HW_DOMAIN_REQ_FREE:
		ocs_sm_transition(ctx, __ocs_hw_domain_free_unreg_vfi, NULL);
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_attach_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data != NULL) {
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		}
		/* free SLI resources */
		sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VFI, domain->indicator);
		/* TODO how to free FCFI (or do we at all)? */

		if (hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain,
					OCS_HW_DOMAIN_ATTACH_FAIL,
					domain);
		}
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_attach_reg_vfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:

		ocs_display_sparams("", "reg vpi", 0, NULL, domain->dma.virt);

		if (0 == sli_cmd_reg_vfi(&hw->sli, data, SLI4_BMBX_SIZE, domain)) {
			ocs_log_err(hw->os, "REG_VFI format failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "REG_VFI command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_domain_attached, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_domain_attach_report_fail, data);
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_allocated(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* free mailbox buffer and send alloc ok to physical sport */
		ocs_free(hw->os, data, SLI4_BMBX_SIZE);
		ocs_sm_post_event(&domain->sport->ctx, OCS_EVT_HW_PORT_ALLOC_OK, NULL);

		ocs_hw_domain_add(hw, domain);

		/* now inform registered callbacks */
		if (hw->callback.domain != NULL) {
			hw->callback.domain(hw->args.domain,
					OCS_HW_DOMAIN_ALLOC_OK,
					domain);
		}
		break;
	case OCS_EVT_HW_DOMAIN_REQ_ATTACH:
		ocs_sm_transition(ctx, __ocs_hw_domain_attach_reg_vfi, data);
		break;
	case OCS_EVT_HW_DOMAIN_REQ_FREE:
		/* unreg_fcfi/vfi */
		if (SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) {
			ocs_sm_transition(ctx, __ocs_hw_domain_free_unreg_fcfi, NULL);
		} else {
			ocs_sm_transition(ctx, __ocs_hw_domain_free_unreg_vfi, NULL);
		}
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_alloc_read_sparm64(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (0 == sli_cmd_read_sparm64(&hw->sli, data, SLI4_BMBX_SIZE,
					&domain->dma, SLI4_READ_SPARM64_VPI_DEFAULT)) {
			ocs_log_err(hw->os, "READ_SPARM64 format failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "READ_SPARM64 command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_EXIT:
		break;
	case OCS_EVT_RESPONSE:
		ocs_display_sparams(domain->display_name, "domain sparm64", 0, NULL, domain->dma.virt);

		ocs_sm_transition(ctx, __ocs_hw_domain_allocated, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_domain_alloc_report_fail, data);
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_alloc_init_vfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_sli_port_t	*sport = domain->sport;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (0 == sli_cmd_init_vfi(&hw->sli, data, SLI4_BMBX_SIZE, domain->indicator,
					domain->fcf_indicator, sport->indicator)) {
			ocs_log_err(hw->os, "INIT_VFI format failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "INIT_VFI command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_EXIT:
		break;
	case OCS_EVT_RESPONSE:
		ocs_sm_transition(ctx, __ocs_hw_domain_alloc_read_sparm64, data);
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_domain_alloc_report_fail, data);
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_alloc_reg_fcfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER: {
		sli4_cmd_rq_cfg_t rq_cfg[SLI4_CMD_REG_FCFI_NUM_RQ_CFG];
		uint32_t i;

		/* Set the filter match/mask values from hw's filter_def values */
		for (i = 0; i < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; i++) {
			rq_cfg[i].rq_id = 0xffff;
			rq_cfg[i].r_ctl_mask = (uint8_t) hw->config.filter_def[i];
			rq_cfg[i].r_ctl_match = (uint8_t) (hw->config.filter_def[i] >> 8);
			rq_cfg[i].type_mask = (uint8_t) (hw->config.filter_def[i] >> 16);
			rq_cfg[i].type_match = (uint8_t) (hw->config.filter_def[i] >> 24);
		}

		/* Set the rq_id for each, in order of RQ definition */
		for (i = 0; i < hw->hw_rq_count; i++) {
			if (i >= ARRAY_SIZE(rq_cfg)) {
				ocs_log_warn(hw->os, "more RQs than REG_FCFI filter entries\n");
				break;
			}
			rq_cfg[i].rq_id = hw->hw_rq[i]->hdr->id;
		}

		if (!data) {
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (hw->hw_mrq_count) {
			if (OCS_HW_RTN_SUCCESS != ocs_hw_config_mrq(hw, SLI4_CMD_REG_FCFI_SET_FCFI_MODE,
				 domain->vlan_id, domain->fcf)) {
				ocs_log_err(hw->os, "REG_FCFI_MRQ format failure\n");
				ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
				break;
			}

		} else {
			if (0 == sli_cmd_reg_fcfi(&hw->sli, data, SLI4_BMBX_SIZE, domain->fcf,
						rq_cfg, domain->vlan_id)) {
				ocs_log_err(hw->os, "REG_FCFI format failure\n");
				ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
				break;
			}
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "REG_FCFI command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	}
	case OCS_EVT_EXIT:
		break;
	case OCS_EVT_RESPONSE:
		if (!data) {
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		domain->fcf_indicator = ((sli4_cmd_reg_fcfi_t *)data)->fcfi;

		/*
		 * IF_TYPE 0 devices do not support explicit VFI and VPI initialization
		 * and instead rely on implicit initialization during VFI registration.
		 * Short circuit normal processing here for those devices.
		 */
		if (SLI4_IF_TYPE_BE3_SKH_PF == sli_get_if_type(&hw->sli)) {
			ocs_sm_transition(ctx, __ocs_hw_domain_alloc_read_sparm64, data);
		} else {
			ocs_sm_transition(ctx, __ocs_hw_domain_alloc_init_vfi, data);
		}
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_domain_alloc_report_fail, data);
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_init(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (sli_get_medium(&hw->sli) == SLI_LINK_MEDIUM_FC) {
			/*
			 * For FC, the HW alread registered a FCFI
			 * Copy FCF information into the domain and jump to INIT_VFI
			 */
			domain->fcf_indicator = hw->fcf_indicator;
			ocs_sm_transition(&domain->sm, __ocs_hw_domain_alloc_init_vfi, data);
		} else {
			ocs_sm_transition(&domain->sm, __ocs_hw_domain_alloc_reg_fcfi, data);
		}
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_free_report_fail(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (domain != NULL) {
			ocs_hw_t	*hw = domain->hw;

			ocs_hw_domain_del(hw, domain);

			if (hw->callback.domain != NULL) {
				hw->callback.domain(hw->args.domain,
						     OCS_HW_DOMAIN_FREE_FAIL,
						     domain);
			}
		}

		/* free command buffer */
		if (data != NULL) {
			ocs_free(domain != NULL ? domain->hw->os : NULL, data, SLI4_BMBX_SIZE);
		}
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_freed(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* Free DMA and mailbox buffer */
		if (domain != NULL) {
			ocs_hw_t *hw = domain->hw;

			/* free VFI resource */
			sli_resource_free(&hw->sli, SLI_RSRC_FCOE_VFI,
					  domain->indicator);

			ocs_hw_domain_del(hw, domain);

			/* inform registered callbacks */
			if (hw->callback.domain != NULL) {
				hw->callback.domain(hw->args.domain,
						     OCS_HW_DOMAIN_FREE_OK,
						     domain);
			}
		}
		if (data != NULL) {
			ocs_free(NULL, data, SLI4_BMBX_SIZE);
		}
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}


static void *
__ocs_hw_domain_free_redisc_fcf(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		/* if we're in the middle of a teardown, skip sending rediscover */
		if (hw->state == OCS_HW_STATE_TEARDOWN_IN_PROGRESS) {
			ocs_sm_transition(ctx, __ocs_hw_domain_freed, data);
			break;
		}
		if (0 == sli_cmd_fcoe_rediscover_fcf(&hw->sli, data, SLI4_BMBX_SIZE, domain->fcf)) {
			ocs_log_err(hw->os, "REDISCOVER_FCF format failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "REDISCOVER_FCF command failure\n");
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
		}
		break;
	case OCS_EVT_RESPONSE:
	case OCS_EVT_ERROR:
		/* REDISCOVER_FCF can fail if none exist */
		ocs_sm_transition(ctx, __ocs_hw_domain_freed, data);
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_free_unreg_fcfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;

	smtrace("domain");

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data == NULL) {
			data = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
			if (!data) {
				ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
				break;
			}
		}

		if (0 == sli_cmd_unreg_fcfi(&hw->sli, data, SLI4_BMBX_SIZE, domain->fcf_indicator)) {
			ocs_log_err(hw->os, "UNREG_FCFI format failure\n");
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "UNREG_FCFI command failure\n");
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_RESPONSE:
		if (domain->req_rediscover_fcf) {
			domain->req_rediscover_fcf = FALSE;
			ocs_sm_transition(ctx, __ocs_hw_domain_free_redisc_fcf, data);
		} else {
			ocs_sm_transition(ctx, __ocs_hw_domain_freed, data);
		}
		break;
	case OCS_EVT_ERROR:
		ocs_sm_transition(ctx, __ocs_hw_domain_free_report_fail, data);
		break;
	case OCS_EVT_EXIT:
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

static void *
__ocs_hw_domain_free_unreg_vfi(ocs_sm_ctx_t *ctx, ocs_sm_event_t evt, void *data)
{
	ocs_domain_t	*domain = ctx->app;
	ocs_hw_t	*hw = domain->hw;
	uint8_t		is_fc = FALSE;

	smtrace("domain");

	is_fc = (sli_get_medium(&hw->sli) == SLI_LINK_MEDIUM_FC);

	switch (evt) {
	case OCS_EVT_ENTER:
		if (data == NULL) {
			data = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_ZERO | OCS_M_NOWAIT);
			if (!data) {
				ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
				break;
			}
		}

		if (0 == sli_cmd_unreg_vfi(&hw->sli, data, SLI4_BMBX_SIZE, domain,
					SLI4_UNREG_TYPE_DOMAIN)) {
			ocs_log_err(hw->os, "UNREG_VFI format failure\n");
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}

		if (ocs_hw_command(hw, data, OCS_CMD_NOWAIT, __ocs_hw_domain_cb, domain)) {
			ocs_log_err(hw->os, "UNREG_VFI command failure\n");
			ocs_free(hw->os, data, SLI4_BMBX_SIZE);
			ocs_sm_post_event(ctx, OCS_EVT_ERROR, NULL);
			break;
		}
		break;
	case OCS_EVT_ERROR:
		if (is_fc) {
			ocs_sm_transition(ctx, __ocs_hw_domain_free_report_fail, data);
		} else {
			ocs_sm_transition(ctx, __ocs_hw_domain_free_unreg_fcfi, data);
		}
		break;
	case OCS_EVT_RESPONSE:
		if (is_fc) {
			ocs_sm_transition(ctx, __ocs_hw_domain_freed, data);
		} else {
			ocs_sm_transition(ctx, __ocs_hw_domain_free_unreg_fcfi, data);
		}
		break;
	default:
		__ocs_hw_domain_common(__func__, ctx, evt, data);
		break;
	}

	return NULL;
}

/* callback for domain alloc/attach/free */
static int32_t
__ocs_hw_domain_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_domain_t	*domain = arg;
	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;
	ocs_sm_event_t	evt;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status vfi=%#x st=%x hdr=%x\n",
			      domain->indicator, status, hdr->status);
		evt = OCS_EVT_ERROR;
	} else {
		evt = OCS_EVT_RESPONSE;
	}

	ocs_sm_post_event(&domain->sm, evt, mqe);

	return 0;
}

static int32_t
target_wqe_timer_nop_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_io_t *io = NULL;
	ocs_hw_io_t *io_next = NULL;
	uint64_t ticks_current = ocs_get_os_ticks();
	uint32_t sec_elapsed;

	sli4_mbox_command_header_t	*hdr = (sli4_mbox_command_header_t *)mqe;

	if (status || hdr->status) {
		ocs_log_debug(hw->os, "bad status st=%x hdr=%x\n",
			      status, hdr->status);
		/* go ahead and proceed with wqe timer checks... */
	}

	/* loop through active WQE list and check for timeouts */
	ocs_lock(&hw->io_lock);
		ocs_list_foreach_safe(&hw->io_timed_wqe, io, io_next) {
			sec_elapsed = ((ticks_current - io->submit_ticks) / ocs_get_os_tick_freq());

			/*
			 * If elapsed time > timeout, abort it. No need to check type since
			 * it wouldn't be on this list unless it was a target WQE
			 */
			if (sec_elapsed > io->tgt_wqe_timeout) {
				ocs_log_test(hw->os, "IO timeout xri=0x%x tag=0x%x type=%d\n",
					     io->indicator, io->reqtag, io->type);

				/* remove from active_wqe list so won't try to abort again */
				ocs_list_remove(&hw->io_timed_wqe, io);

				/* save status of "timed out" for when abort completes */
				io->status_saved = 1;
				io->saved_status = SLI4_FC_WCQE_STATUS_TARGET_WQE_TIMEOUT;
				io->saved_ext = 0;
				io->saved_len = 0;

				/* now abort outstanding IO */
				ocs_hw_io_abort(hw, io, FALSE, NULL, NULL);
			}
			/*
			 * need to go through entire list since each IO could have a
			 * different timeout value
			 */
		}
	ocs_unlock(&hw->io_lock);

	/* if we're not in the middle of shutting down, schedule next timer */
	if (!hw->active_wqe_timer_shutdown) {
		ocs_setup_timer(hw->os, &hw->wqe_timer, target_wqe_timer_cb, hw, OCS_HW_WQ_TIMER_PERIOD_MS);
	}
	hw->in_active_wqe_timer = FALSE;
	return 0;
}

static void
target_wqe_timer_cb(void *arg)
{
	ocs_hw_t *hw = (ocs_hw_t *)arg;

	/* delete existing timer; will kick off new timer after checking wqe timeouts */
	hw->in_active_wqe_timer = TRUE;
	ocs_del_timer(&hw->wqe_timer);

	/* Forward timer callback to execute in the mailbox completion processing context */
	if (ocs_hw_async_call(hw, target_wqe_timer_nop_cb, hw)) {
		ocs_log_test(hw->os, "ocs_hw_async_call failed\n");
	}
}

static void
shutdown_target_wqe_timer(ocs_hw_t *hw)
{
	uint32_t	iters = 100;

	if (hw->config.emulate_tgt_wqe_timeout) {
		/* request active wqe timer shutdown, then wait for it to complete */
		hw->active_wqe_timer_shutdown = TRUE;

		/* delete WQE timer and wait for timer handler to complete (if necessary) */
		ocs_del_timer(&hw->wqe_timer);

		/* now wait for timer handler to complete (if necessary) */
		while (hw->in_active_wqe_timer && iters) {
			/*
			 * if we happen to have just sent NOP mailbox command, make sure
			 * completions are being processed
			 */
			ocs_hw_flush(hw);
			iters--;
		}

		if (iters == 0) {
			ocs_log_test(hw->os, "Failed to shutdown active wqe timer\n");
		}
	}
}

/**
 * @brief Determine if HW IO is owned by the port.
 *
 * @par Description
 * Determines if the given HW IO has been posted to the chip.
 *
 * @param hw Hardware context allocated by the caller.
 * @param io HW IO.
 *
 * @return Returns TRUE if given HW IO is port-owned.
 */
uint8_t
ocs_hw_is_io_port_owned(ocs_hw_t *hw, ocs_hw_io_t *io)
{
	/* Check to see if this is a port owned XRI */
	return io->is_port_owned;
}

/**
 * @brief Return TRUE if exchange is port-owned.
 *
 * @par Description
 * Test to see if the xri is a port-owned xri.
 *
 * @param hw Hardware context.
 * @param xri Exchange indicator.
 *
 * @return Returns TRUE if XRI is a port owned XRI.
 */

uint8_t
ocs_hw_is_xri_port_owned(ocs_hw_t *hw, uint32_t xri)
{
	ocs_hw_io_t *io = ocs_hw_io_lookup(hw, xri);
	return (io == NULL ? FALSE : io->is_port_owned);
}

/**
 * @brief Returns an XRI from the port owned list to the host.
 *
 * @par Description
 * Used when the POST_XRI command fails as well as when the RELEASE_XRI completes.
 *
 * @param hw Hardware context.
 * @param xri_base The starting XRI number.
 * @param xri_count The number of XRIs to free from the base.
 */
static void
ocs_hw_reclaim_xri(ocs_hw_t *hw, uint16_t xri_base, uint16_t xri_count)
{
	ocs_hw_io_t	*io;
	uint32_t i;

	for (i = 0; i < xri_count; i++) {
		io = ocs_hw_io_lookup(hw, xri_base + i);

		/*
		 * if this is an auto xfer rdy XRI, then we need to release any
		 * buffer attached to the XRI before moving the XRI back to the free pool.
		 */
		if (hw->auto_xfer_rdy_enabled) {
			ocs_hw_rqpair_auto_xfer_rdy_move_to_host(hw, io);
		}

		ocs_lock(&hw->io_lock);
			ocs_list_remove(&hw->io_port_owned, io);
			io->is_port_owned = 0;
			ocs_list_add_tail(&hw->io_free, io);
		ocs_unlock(&hw->io_lock);
	}
}

/**
 * @brief Called when the POST_XRI command completes.
 *
 * @par Description
 * Free the mailbox command buffer and reclaim the XRIs on failure.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_post_xri(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	sli4_cmd_post_xri_t	*post_xri = (sli4_cmd_post_xri_t*)mqe;

	/* Reclaim the XRIs as host owned if the command fails */
	if (status != 0) {
		ocs_log_debug(hw->os, "Status 0x%x for XRI base 0x%x, cnt =x%x\n",
			      status, post_xri->xri_base, post_xri->xri_count);
		ocs_hw_reclaim_xri(hw, post_xri->xri_base, post_xri->xri_count);
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	return 0;
}

/**
 * @brief Issues a mailbox command to move XRIs from the host-controlled pool to the port.
 *
 * @param hw Hardware context.
 * @param xri_start The starting XRI to post.
 * @param num_to_post The number of XRIs to post.
 *
 * @return Returns OCS_HW_RTN_NO_MEMORY, OCS_HW_RTN_ERROR, or OCS_HW_RTN_SUCCESS.
 */

static ocs_hw_rtn_e
ocs_hw_post_xri(ocs_hw_t *hw, uint32_t xri_start, uint32_t num_to_post)
{
	uint8_t	*post_xri;
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;

	/* Since we need to allocate for mailbox queue, just always allocate */
	post_xri = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (post_xri == NULL) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* Register the XRIs */
	if (sli_cmd_post_xri(&hw->sli, post_xri, SLI4_BMBX_SIZE,
			     xri_start, num_to_post)) {
		rc = ocs_hw_command(hw, post_xri, OCS_CMD_NOWAIT, ocs_hw_cb_post_xri, NULL);
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_free(hw->os, post_xri, SLI4_BMBX_SIZE);
			ocs_log_err(hw->os, "post_xri failed\n");
		}
	}
	return rc;
}

/**
 * @brief Move XRIs from the host-controlled pool to the port.
 *
 * @par Description
 * Removes IOs from the free list and moves them to the port.
 *
 * @param hw Hardware context.
 * @param num_xri The number of XRIs being requested to move to the chip.
 *
 * @return Returns the number of XRIs that were moved.
 */

uint32_t
ocs_hw_xri_move_to_port_owned(ocs_hw_t *hw, uint32_t num_xri)
{
	ocs_hw_io_t	*io;
	uint32_t i;
	uint32_t num_posted = 0;

	/*
	 * Note: We cannot use ocs_hw_io_alloc() because that would place the
	 *       IO on the io_inuse list. We need to move from the io_free to
	 *       the io_port_owned list.
	 */
	ocs_lock(&hw->io_lock);

	for (i = 0; i < num_xri; i++) {

		if (NULL != (io = ocs_list_remove_head(&hw->io_free))) {
			ocs_hw_rtn_e rc;

			/*
			 * if this is an auto xfer rdy XRI, then we need to attach a
			 * buffer to the XRI before submitting it to the chip. If a
			 * buffer is unavailable, then we cannot post it, so return it
			 * to the free pool.
			 */
			if (hw->auto_xfer_rdy_enabled) {
				/* Note: uses the IO lock to get the auto xfer rdy buffer */
				ocs_unlock(&hw->io_lock);
				rc = ocs_hw_rqpair_auto_xfer_rdy_move_to_port(hw, io);
				ocs_lock(&hw->io_lock);
				if (rc != OCS_HW_RTN_SUCCESS) {
					ocs_list_add_head(&hw->io_free, io);
					break;
				}
			}
			ocs_lock_init(hw->os, &io->axr_lock, "HW_axr_lock[%d]", io->indicator);
			io->is_port_owned = 1;
			ocs_list_add_tail(&hw->io_port_owned, io);

			/* Post XRI */
			if (ocs_hw_post_xri(hw, io->indicator, 1) != OCS_HW_RTN_SUCCESS ) {
				ocs_hw_reclaim_xri(hw, io->indicator, i);
				break;
			}
			num_posted++;
		} else {
			/* no more free XRIs */
			break;
		}
	}
	ocs_unlock(&hw->io_lock);

	return num_posted;
}

/**
 * @brief Called when the RELEASE_XRI command completes.
 *
 * @par Description
 * Move the IOs back to the free pool on success.
 *
 * @param hw Hardware context.
 * @param status Status field from the mbox completion.
 * @param mqe Mailbox response structure.
 * @param arg Pointer to a callback function that signals the caller that the command is done.
 *
 * @return Returns 0.
 */
static int32_t
ocs_hw_cb_release_xri(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void  *arg)
{
	sli4_cmd_release_xri_t	*release_xri = (sli4_cmd_release_xri_t*)mqe;
	uint8_t i;

	/* Reclaim the XRIs as host owned if the command fails */
	if (status != 0) {
		ocs_log_err(hw->os, "Status 0x%x\n", status);
	} else {
		for (i = 0; i < release_xri->released_xri_count; i++) {
			uint16_t xri = ((i & 1) == 0 ? release_xri->xri_tbl[i/2].xri_tag0 :
					release_xri->xri_tbl[i/2].xri_tag1);
			ocs_hw_reclaim_xri(hw, xri, 1);
		}
	}

	ocs_free(hw->os, mqe, SLI4_BMBX_SIZE);
	return 0;
}

/**
 * @brief Move XRIs from the port-controlled pool to the host.
 *
 * Requests XRIs from the FW to return to the host-owned pool.
 *
 * @param hw Hardware context.
 * @param num_xri The number of XRIs being requested to moved from the chip.
 *
 * @return Returns 0 for success, or a negative error code value for failure.
 */

ocs_hw_rtn_e
ocs_hw_xri_move_to_host_owned(ocs_hw_t *hw, uint8_t num_xri)
{
	uint8_t	*release_xri;
	ocs_hw_rtn_e rc = OCS_HW_RTN_ERROR;

	/* non-local buffer required for mailbox queue */
	release_xri = ocs_malloc(hw->os, SLI4_BMBX_SIZE, OCS_M_NOWAIT);
	if (release_xri == NULL) {
		ocs_log_err(hw->os, "no buffer for command\n");
		return OCS_HW_RTN_NO_MEMORY;
	}

	/* release the XRIs */
	if (sli_cmd_release_xri(&hw->sli, release_xri, SLI4_BMBX_SIZE, num_xri)) {
		rc = ocs_hw_command(hw, release_xri, OCS_CMD_NOWAIT, ocs_hw_cb_release_xri, NULL);
		if (rc != OCS_HW_RTN_SUCCESS) {
			ocs_log_err(hw->os, "release_xri failed\n");
		}
	}
	/* If we are polling or an error occurred, then free the mailbox buffer */
	if (release_xri != NULL && rc != OCS_HW_RTN_SUCCESS) {
		ocs_free(hw->os, release_xri, SLI4_BMBX_SIZE);
	}
	return rc;
}


/**
 * @brief Allocate an ocs_hw_rx_buffer_t array.
 *
 * @par Description
 * An ocs_hw_rx_buffer_t array is allocated, along with the required DMA memory.
 *
 * @param hw Pointer to HW object.
 * @param rqindex RQ index for this buffer.
 * @param count Count of buffers in array.
 * @param size Size of buffer.
 *
 * @return Returns the pointer to the allocated ocs_hw_rq_buffer_t array.
 */
static ocs_hw_rq_buffer_t *
ocs_hw_rx_buffer_alloc(ocs_hw_t *hw, uint32_t rqindex, uint32_t count, uint32_t size)
{
	ocs_t *ocs = hw->os;
	ocs_hw_rq_buffer_t *rq_buf = NULL;
	ocs_hw_rq_buffer_t *prq;
	uint32_t i;

	if (count != 0) {
		rq_buf = ocs_malloc(hw->os, sizeof(*rq_buf) * count, OCS_M_NOWAIT | OCS_M_ZERO);
		if (rq_buf == NULL) {
			ocs_log_err(hw->os, "Failure to allocate unsolicited DMA trackers\n");
			return NULL;
		}

		for (i = 0, prq = rq_buf; i < count; i ++, prq++) {
			prq->rqindex = rqindex;
			if (ocs_dma_alloc(ocs, &prq->dma, size, OCS_MIN_DMA_ALIGNMENT)) {
				ocs_log_err(hw->os, "DMA allocation failed\n");
				ocs_free(hw->os, rq_buf, sizeof(*rq_buf) * count);
				rq_buf = NULL;
				break;
			}
		}
	}
	return rq_buf;
}

/**
 * @brief Free an ocs_hw_rx_buffer_t array.
 *
 * @par Description
 * The ocs_hw_rx_buffer_t array is freed, along with allocated DMA memory.
 *
 * @param hw Pointer to HW object.
 * @param rq_buf Pointer to ocs_hw_rx_buffer_t array.
 * @param count Count of buffers in array.
 *
 * @return None.
 */
static void
ocs_hw_rx_buffer_free(ocs_hw_t *hw, ocs_hw_rq_buffer_t *rq_buf, uint32_t count)
{
	ocs_t *ocs = hw->os;
	uint32_t i;
	ocs_hw_rq_buffer_t *prq;

	if (rq_buf != NULL) {
		for (i = 0, prq = rq_buf; i < count; i++, prq++) {
			ocs_dma_free(ocs, &prq->dma);
		}
		ocs_free(hw->os, rq_buf, sizeof(*rq_buf) * count);
	}
}

/**
 * @brief Allocate the RQ data buffers.
 *
 * @param hw Pointer to HW object.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_rx_allocate(ocs_hw_t *hw)
{
	ocs_t *ocs = hw->os;
	uint32_t i;
	int32_t rc = OCS_HW_RTN_SUCCESS;
	uint32_t rqindex = 0;
	hw_rq_t *rq;
	uint32_t hdr_size = OCS_HW_RQ_SIZE_HDR;
	uint32_t payload_size = hw->config.rq_default_buffer_size;

	rqindex = 0;

	for (i = 0; i < hw->hw_rq_count; i++) {
		rq = hw->hw_rq[i];

		/* Allocate header buffers */
		rq->hdr_buf = ocs_hw_rx_buffer_alloc(hw, rqindex, rq->entry_count, hdr_size);
		if (rq->hdr_buf == NULL) {
			ocs_log_err(ocs, "ocs_hw_rx_buffer_alloc hdr_buf failed\n");
			rc = OCS_HW_RTN_ERROR;
			break;
		}

		ocs_log_debug(hw->os, "rq[%2d] rq_id %02d header  %4d by %4d bytes\n", i, rq->hdr->id,
			      rq->entry_count, hdr_size);

		rqindex++;

		/* Allocate payload buffers */
		rq->payload_buf = ocs_hw_rx_buffer_alloc(hw, rqindex, rq->entry_count, payload_size);
		if (rq->payload_buf == NULL) {
			ocs_log_err(ocs, "ocs_hw_rx_buffer_alloc fb_buf failed\n");
			rc = OCS_HW_RTN_ERROR;
			break;
		}
		ocs_log_debug(hw->os, "rq[%2d] rq_id %02d default %4d by %4d bytes\n", i, rq->data->id,
			      rq->entry_count, payload_size);
		rqindex++;
	}

	return rc ? OCS_HW_RTN_ERROR : OCS_HW_RTN_SUCCESS;
}

/**
 * @brief Post the RQ data buffers to the chip.
 *
 * @param hw Pointer to HW object.
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_rx_post(ocs_hw_t *hw)
{
	uint32_t i;
	uint32_t idx;
	uint32_t rq_idx;
	int32_t rc = 0;

	/*
	 * In RQ pair mode, we MUST post the header and payload buffer at the
	 * same time.
	 */
	for (rq_idx = 0, idx = 0; rq_idx < hw->hw_rq_count; rq_idx++) {
		hw_rq_t *rq = hw->hw_rq[rq_idx];

		for (i = 0; i < rq->entry_count-1; i++) {
			ocs_hw_sequence_t *seq = ocs_array_get(hw->seq_pool, idx++);
			ocs_hw_assert(seq != NULL);

			seq->header = &rq->hdr_buf[i];

			seq->payload = &rq->payload_buf[i];

			rc = ocs_hw_sequence_free(hw, seq);
			if (rc) {
				break;
			}
		}
		if (rc) {
			break;
		}
	}

	return rc;
}

/**
 * @brief Free the RQ data buffers.
 *
 * @param hw Pointer to HW object.
 *
 */
void
ocs_hw_rx_free(ocs_hw_t *hw)
{
	hw_rq_t *rq;
	uint32_t i;

	/* Free hw_rq buffers */
	for (i = 0; i < hw->hw_rq_count; i++) {
		rq = hw->hw_rq[i];
		if (rq != NULL) {
			ocs_hw_rx_buffer_free(hw, rq->hdr_buf, rq->entry_count);
			rq->hdr_buf = NULL;
			ocs_hw_rx_buffer_free(hw, rq->payload_buf, rq->entry_count);
			rq->payload_buf = NULL;
		}
	}
}

/**
 * @brief HW async call context structure.
 */
typedef struct {
	ocs_hw_async_cb_t callback;
	void *arg;
	uint8_t cmd[SLI4_BMBX_SIZE];
} ocs_hw_async_call_ctx_t;

/**
 * @brief HW async callback handler
 *
 * @par Description
 * This function is called when the NOP mailbox command completes.  The callback stored
 * in the requesting context is invoked.
 *
 * @param hw Pointer to HW object.
 * @param status Completion status.
 * @param mqe Pointer to mailbox completion queue entry.
 * @param arg Caller-provided argument.
 *
 * @return None.
 */
static void
ocs_hw_async_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_async_call_ctx_t *ctx = arg;

	if (ctx != NULL) {
		if (ctx->callback != NULL) {
			(*ctx->callback)(hw, status, mqe, ctx->arg);
		}
		ocs_free(hw->os, ctx, sizeof(*ctx));
	}
}

/**
 * @brief Make an async callback using NOP mailbox command
 *
 * @par Description
 * Post a NOP mailbox command; the callback with argument is invoked upon completion
 * while in the event processing context.
 *
 * @param hw Pointer to HW object.
 * @param callback Pointer to callback function.
 * @param arg Caller-provided callback.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
int32_t
ocs_hw_async_call(ocs_hw_t *hw, ocs_hw_async_cb_t callback, void *arg)
{
	ocs_hw_async_call_ctx_t *ctx;

	/*
	 * Allocate a callback context (which includes the mailbox command buffer), we need
	 * this to be persistent as the mailbox command submission may be queued and executed later
	 * execution.
	 */
	ctx = ocs_malloc(hw->os, sizeof(*ctx), OCS_M_ZERO | OCS_M_NOWAIT);
	if (ctx == NULL) {
		ocs_log_err(hw->os, "failed to malloc async call context\n");
		return OCS_HW_RTN_NO_MEMORY;
	}
	ctx->callback = callback;
	ctx->arg = arg;

	/* Build and send a NOP mailbox command */
	if (sli_cmd_common_nop(&hw->sli, ctx->cmd, sizeof(ctx->cmd), 0) == 0) {
		ocs_log_err(hw->os, "COMMON_NOP format failure\n");
		ocs_free(hw->os, ctx, sizeof(*ctx));
		return OCS_HW_RTN_ERROR;
	}

	if (ocs_hw_command(hw, ctx->cmd, OCS_CMD_NOWAIT, ocs_hw_async_cb, ctx)) {
		ocs_log_err(hw->os, "COMMON_NOP command failure\n");
		ocs_free(hw->os, ctx, sizeof(*ctx));
		return OCS_HW_RTN_ERROR;
	}
	return OCS_HW_RTN_SUCCESS;
}

/**
 * @brief Initialize the reqtag pool.
 *
 * @par Description
 * The WQ request tag pool is initialized.
 *
 * @param hw Pointer to HW object.
 *
 * @return Returns 0 on success, or a negative error code value on failure.
 */
ocs_hw_rtn_e
ocs_hw_reqtag_init(ocs_hw_t *hw)
{
	if (hw->wq_reqtag_pool == NULL) {
		hw->wq_reqtag_pool = ocs_pool_alloc(hw->os, sizeof(hw_wq_callback_t), 65536, TRUE);
		if (hw->wq_reqtag_pool == NULL) {
			ocs_log_err(hw->os, "ocs_pool_alloc hw_wq_callback_t failed\n");
			return OCS_HW_RTN_NO_MEMORY;
		}
	}
	ocs_hw_reqtag_reset(hw);
	return OCS_HW_RTN_SUCCESS;
}

/**
 * @brief Allocate a WQ request tag.
 *
 * Allocate and populate a WQ request tag from the WQ request tag pool.
 *
 * @param hw Pointer to HW object.
 * @param callback Callback function.
 * @param arg Pointer to callback argument.
 *
 * @return Returns pointer to allocated WQ request tag, or NULL if object cannot be allocated.
 */
hw_wq_callback_t *
ocs_hw_reqtag_alloc(ocs_hw_t *hw, void (*callback)(void *arg, uint8_t *cqe, int32_t status), void *arg)
{
	hw_wq_callback_t *wqcb;

	ocs_hw_assert(callback != NULL);

	wqcb = ocs_pool_get(hw->wq_reqtag_pool);
	if (wqcb != NULL) {
		ocs_hw_assert(wqcb->callback == NULL);
		wqcb->callback = callback;
		wqcb->arg = arg;
	}
	return wqcb;
}

/**
 * @brief Free a WQ request tag.
 *
 * Free the passed in WQ request tag.
 *
 * @param hw Pointer to HW object.
 * @param wqcb Pointer to WQ request tag object to free.
 *
 * @return None.
 */
void
ocs_hw_reqtag_free(ocs_hw_t *hw, hw_wq_callback_t *wqcb)
{
	ocs_hw_assert(wqcb->callback != NULL);
	wqcb->callback = NULL;
	wqcb->arg = NULL;
	ocs_pool_put(hw->wq_reqtag_pool, wqcb);
}

/**
 * @brief Return WQ request tag by index.
 *
 * @par Description
 * Return pointer to WQ request tag object given an index.
 *
 * @param hw Pointer to HW object.
 * @param instance_index Index of WQ request tag to return.
 *
 * @return Pointer to WQ request tag, or NULL.
 */
hw_wq_callback_t *
ocs_hw_reqtag_get_instance(ocs_hw_t *hw, uint32_t instance_index)
{
	hw_wq_callback_t *wqcb;

	wqcb = ocs_pool_get_instance(hw->wq_reqtag_pool, instance_index);
	if (wqcb == NULL) {
		ocs_log_err(hw->os, "wqcb for instance %d is null\n", instance_index);
	}
	return wqcb;
}

/**
 * @brief Reset the WQ request tag pool.
 *
 * @par Description
 * Reset the WQ request tag pool, returning all to the free list.
 *
 * @param hw pointer to HW object.
 *
 * @return None.
 */
void
ocs_hw_reqtag_reset(ocs_hw_t *hw)
{
	hw_wq_callback_t *wqcb;
	uint32_t i;

	/* Remove all from freelist */
	while(ocs_pool_get(hw->wq_reqtag_pool) != NULL) {
		;
	}

	/* Put them all back */
	for (i = 0; ((wqcb = ocs_pool_get_instance(hw->wq_reqtag_pool, i)) != NULL); i++) {
		wqcb->instance_index = i;
		wqcb->callback = NULL;
		wqcb->arg = NULL;
		ocs_pool_put(hw->wq_reqtag_pool, wqcb);
	}
}

/**
 * @brief Handle HW assertion
 *
 * HW assert, display diagnostic message, and abort.
 *
 * @param cond string describing failing assertion condition
 * @param filename file name
 * @param linenum line number
 *
 * @return none
 */
void
_ocs_hw_assert(const char *cond, const char *filename, int linenum)
{
	ocs_printf("%s(%d): HW assertion (%s) failed\n", filename, linenum, cond);
	ocs_abort();
		/* no return */
}

/**
 * @brief Handle HW verify
 *
 * HW verify, display diagnostic message, dump stack and return.
 *
 * @param cond string describing failing verify condition
 * @param filename file name
 * @param linenum line number
 *
 * @return none
 */
void
_ocs_hw_verify(const char *cond, const char *filename, int linenum)
{
	ocs_printf("%s(%d): HW verify (%s) failed\n", filename, linenum, cond);
	ocs_print_stack();
}

/**
 * @brief Reque XRI
 *
 * @par Description
 * Reque XRI
 *
 * @param hw Pointer to HW object.
 * @param io Pointer to HW IO
 *
 * @return Return 0 if successful else returns -1
 */
int32_t
ocs_hw_reque_xri( ocs_hw_t *hw, ocs_hw_io_t *io )
{
	int32_t rc = 0;

	rc = ocs_hw_rqpair_auto_xfer_rdy_buffer_post(hw, io, 1);
	if (rc) {
		ocs_list_add_tail(&hw->io_port_dnrx, io);
		rc = -1;
		goto exit_ocs_hw_reque_xri;
	}

	io->auto_xfer_rdy_dnrx = 0;
	io->type = OCS_HW_IO_DNRX_REQUEUE;
	if (sli_requeue_xri_wqe(&hw->sli, io->wqe.wqebuf, hw->sli.config.wqe_size, io->indicator, OCS_HW_REQUE_XRI_REGTAG, SLI4_CQ_DEFAULT)) {
		/* Clear buffer from XRI */
		ocs_pool_put(hw->auto_xfer_rdy_buf_pool, io->axr_buf);
		io->axr_buf = NULL;

		ocs_log_err(hw->os, "requeue_xri WQE error\n");
		ocs_list_add_tail(&hw->io_port_dnrx, io);

		rc = -1;
		goto exit_ocs_hw_reque_xri;
	}

	if (io->wq == NULL) {
		io->wq = ocs_hw_queue_next_wq(hw, io);
		ocs_hw_assert(io->wq != NULL);
	}

	/*
	 * Add IO to active io wqe list before submitting, in case the
	 * wcqe processing preempts this thread.
	 */
	OCS_STAT(hw->tcmd_wq_submit[io->wq->instance]++);
	OCS_STAT(io->wq->use_count++);

	rc = hw_wq_write(io->wq, &io->wqe);
	if (rc < 0) {
		ocs_log_err(hw->os, "sli_queue_write reque xri failed: %d\n", rc);
		rc = -1;
	}

exit_ocs_hw_reque_xri:
	return 0;
}

uint32_t
ocs_hw_get_def_wwn(ocs_t *ocs, uint32_t chan, uint64_t *wwpn, uint64_t *wwnn)
{
	sli4_t *sli4 = &ocs->hw.sli;
	ocs_dma_t       dma;
	uint8_t		*payload = NULL;

	int indicator = sli4->config.extent[SLI_RSRC_FCOE_VPI].base[0] + chan;

	/* allocate memory for the service parameters */
	if (ocs_dma_alloc(ocs, &dma, 112, 4)) {
		ocs_log_err(ocs, "Failed to allocate DMA memory\n");
		return 1;
	}

	if (0 == sli_cmd_read_sparm64(sli4, sli4->bmbx.virt, SLI4_BMBX_SIZE,
				&dma, indicator)) {
		ocs_log_err(ocs, "READ_SPARM64 allocation failure\n");
		ocs_dma_free(ocs, &dma);
		return 1;
	}

	if (sli_bmbx_command(sli4)) {
		ocs_log_err(ocs, "READ_SPARM64 command failure\n");
		ocs_dma_free(ocs, &dma);
		return 1;
	}

	payload = dma.virt;
	ocs_memcpy(wwpn, payload + SLI4_READ_SPARM64_WWPN_OFFSET, sizeof(*wwpn));
	ocs_memcpy(wwnn, payload + SLI4_READ_SPARM64_WWNN_OFFSET, sizeof(*wwnn));
	ocs_dma_free(ocs, &dma);
	return 0;
}

uint32_t
ocs_hw_get_config_persistent_topology(ocs_hw_t *hw)
{
        uint32_t topology = OCS_HW_TOPOLOGY_AUTO;
	sli4_t *sli = &hw->sli;

        if (!sli_persist_topology_enabled(sli))
                return topology;

        switch (sli->config.pt) {
                case SLI4_INIT_LINK_F_P2P_ONLY:
                        topology = OCS_HW_TOPOLOGY_NPORT;
                        break;
                case SLI4_INIT_LINK_F_FCAL_ONLY:
                        topology = OCS_HW_TOPOLOGY_LOOP;
                        break;
                default:
                        break;
        }

        return topology;
}

/*
 * @brief Persistent topology configuration callback argument.
 */
typedef struct ocs_hw_persistent_topo_cb_arg {
	ocs_sem_t semaphore;
	int32_t status;
} ocs_hw_persistent_topo_cb_arg_t;

/*
 * @brief Called after the completion of set persistent topology request
 *
 * @par Description
 * This is callback fn for the set_persistent_topology
 * function. This callback is called when the common feature mbx cmd
 * completes.
 *
 * @param hw Hardware context.
 * @param status The status from the MQE.
 * @param mqe Pointer to mailbox command buffer.
 * @param arg Pointer to a callback argument.
 *
 * @return 0 on success, non-zero otherwise
 */
static int32_t
ocs_hw_set_persistent_topolgy_cb(ocs_hw_t *hw, int32_t status, uint8_t *mqe, void *arg)
{
	ocs_hw_persistent_topo_cb_arg_t *req = (ocs_hw_persistent_topo_cb_arg_t *)arg;

	req->status = status;

	ocs_sem_v(&req->semaphore);

	return 0;
}

/**
 * @brief Set persistent topology
 *
 * Sets the persistent topology(PT) feature using
 * COMMON_SET_FEATURES cmd. If mbx cmd succeeds, update the
 * topology into sli config. PT stores the value to be set into link_flags
 * of the cmd INIT_LINK, to bring up the link.
 *
 * SLI specs defines following for PT:
 *     When TF is set to 0:
 *       0 Reserved
 *       1 Attempt point-to-point initialization (direct attach or Fabric topology).
 *       2 Attempt FC-AL loop initialization.
 *       3 Reserved
 *
 *      When TF is set to 1:
 *       0 Attempt FC-AL loop initialization; if it fails, attempt point-to-point initialization.
 *       1 Attempt point-to-point initialization; if it fails, attempt FC-AL loop initialization.
 *       2 Reserved
 *      3 Reserved
 *
 *     Note: Topology failover is only available on Lancer G5. This command will fail
 *     if TF is set to 1 on any other ASICs
 *
 * @param hw Pointer to hw
 * @param topology topology value to be set, provided through
 *        elxsdkutil set-topology cmd
 *
 * @return Returns 0 on success, or a non-zero value on failure.
 */
ocs_hw_rtn_e
ocs_hw_set_persistent_topology(ocs_hw_t *hw, uint32_t topology, uint32_t opts)
{
	ocs_hw_rtn_e rc = OCS_HW_RTN_SUCCESS;
	uint8_t buf[SLI4_BMBX_SIZE];
	sli4_req_common_set_features_persistent_topo_param_t param;
	ocs_hw_persistent_topo_cb_arg_t request;

	ocs_memset(&param, 0, sizeof(param));
	param.persistent_topo = topology;

	switch (topology) {
	case OCS_HW_TOPOLOGY_AUTO:
		if (sli_get_asic_type(&hw->sli) == SLI4_ASIC_TYPE_LANCER) {
			param.persistent_topo = SLI4_INIT_LINK_F_P2P_FAIL_OVER;
			param.topo_failover = 1;
		} else {
			param.persistent_topo = SLI4_INIT_LINK_F_P2P_ONLY;;
			param.topo_failover = 0;
		}
		break;

	case OCS_HW_TOPOLOGY_NPORT:
		param.persistent_topo = SLI4_INIT_LINK_F_P2P_ONLY;
		param.topo_failover = 0;
		break;

	case OCS_HW_TOPOLOGY_LOOP:
		param.persistent_topo = SLI4_INIT_LINK_F_FCAL_ONLY;
		param.topo_failover = 0;
		break;

	default:
		ocs_log_err(hw->os, "unsupported topology %#x\n", topology);
		return -1;
	}

	ocs_sem_init(&request.semaphore, 0, "set_persistent_topo");

	/* build the set_features command */
	sli_cmd_common_set_features(&hw->sli, buf, SLI4_BMBX_SIZE,
		SLI4_SET_FEATURES_PERSISTENT_TOPOLOGY, sizeof(param), &param);

	if (opts == OCS_CMD_POLL) {
		rc = ocs_hw_command(hw, buf, OCS_CMD_POLL, NULL, NULL);
		if (rc) {
			ocs_log_err(hw->os, "Failed to set persistent topology, rc: %#x\n", rc);
			return rc;
		}
	} else {

		// there's no response for this feature command
		rc = ocs_hw_command(hw, buf, OCS_CMD_NOWAIT, ocs_hw_set_persistent_topolgy_cb, &request);
		if (rc) {
			ocs_log_err(hw->os, "Failed to set persistent topology, rc: %#x\n", rc);
			return rc;
		}

		if (ocs_sem_p(&request.semaphore, OCS_SEM_FOREVER)) {
			ocs_log_err(hw->os, "ocs_sem_p failed\n");
			return -ENXIO;
		}

		if (request.status) {
			ocs_log_err(hw->os, "set persistent topology failed; status: %d\n", request.status);
			return -EFAULT;
		}
	}

	sli_config_persistent_topology(&hw->sli, &param);

	return rc;
}

/**
 * @page fc_hw_api_overview HW APIs
 * - @ref devInitShutdown
 * - @ref domain
 * - @ref port
 * - @ref node
 * - @ref io
 * - @ref interrupt
 *
 * <div class="overview">
 * The Hardware Abstraction Layer (HW) insulates the higher-level code from the SLI-4
 * message details, but the higher level code must still manage domains, ports,
 * IT nexuses, and IOs. The HW API is designed to help the higher level manage
 * these objects.<br><br>
 *
 * The HW uses function callbacks to notify the higher-level code of events
 * that are received from the chip. There are currently three types of
 * functions that may be registered:
 *
 * <ul><li>domain – This function is called whenever a domain event is generated
 * within the HW. Examples include a new FCF is discovered, a connection
 * to a domain is disrupted, and allocation callbacks.</li>
 * <li>unsolicited – This function is called whenever new data is received in
 * the SLI-4 receive queue.</li>
 * <li>rnode – This function is called for remote node events, such as attach status
 * and  allocation callbacks.</li></ul>
 *
 * Upper layer functions may be registered by using the ocs_hw_callback() function.
 *
 * <img src="elx_fc_hw.jpg" alt="FC/FCoE HW" title="FC/FCoE HW" align="right"/>
 * <h2>FC/FCoE HW API</h2>
 * The FC/FCoE HW component builds upon the SLI-4 component to establish a flexible
 * interface for creating the necessary common objects and sending I/Os. It may be used
 * “as is” in customer implementations or it can serve as an example of typical interactions
 * between a driver and the SLI-4 hardware. The broad categories of functionality include:
 *
 * <ul><li>Setting-up and tearing-down of the HW.</li>
 * <li>Allocating and using the common objects (SLI Port, domain, remote node).</li>
 * <li>Sending and receiving I/Os.</li></ul>
 *
 * <h3>HW Setup</h3>
 * To set up the HW:
 *
 * <ol>
 * <li>Set up the HW object using ocs_hw_setup().<br>
 * This step performs a basic configuration of the SLI-4 component and the HW to
 * enable querying the hardware for its capabilities. At this stage, the HW is not
 * capable of general operations (such as, receiving events or sending I/Os).</li><br><br>
 * <li>Configure the HW according to the driver requirements.<br>
 * The HW provides functions to discover hardware capabilities (ocs_hw_get()), as
 * well as configures the amount of resources required (ocs_hw_set()). The driver
 * must also register callback functions (ocs_hw_callback()) to receive notification of
 * various asynchronous events.<br><br>
 * @b Note: Once configured, the driver must initialize the HW (ocs_hw_init()). This
 * step creates the underlying queues, commits resources to the hardware, and
 * prepares the hardware for operation. While the hardware is operational, the
 * port is not online, and cannot send or receive data.</li><br><br>
 * <br><br>
 * <li>Finally, the driver can bring the port online (ocs_hw_port_control()).<br>
 * When the link comes up, the HW determines if a domain is present and notifies the
 * driver using the domain callback function. This is the starting point of the driver's
 * interaction with the common objects.<br><br>
 * @b Note: For FCoE, there may be more than one domain available and, therefore,
 * more than one callback.</li>
 * </ol>
 *
 * <h3>Allocating and Using Common Objects</h3>
 * Common objects provide a mechanism through which the various OneCore Storage
 * driver components share and track information. These data structures are primarily
 * used to track SLI component information but can be extended by other components, if
 * needed. The main objects are:
 *
 * <ul><li>DMA – the ocs_dma_t object describes a memory region suitable for direct
 * memory access (DMA) transactions.</li>
 * <li>SCSI domain – the ocs_domain_t object represents the SCSI domain, including
 * any infrastructure devices such as FC switches and FC forwarders. The domain
 * object contains both an FCFI and a VFI.</li>
 * <li>SLI Port (sport) – the ocs_sli_port_t object represents the connection between
 * the driver and the SCSI domain. The SLI Port object contains a VPI.</li>
 * <li>Remote node – the ocs_remote_node_t represents a connection between the SLI
 * Port and another device in the SCSI domain. The node object contains an RPI.</li></ul>
 *
 * Before the driver can send I/Os, it must allocate the SCSI domain, SLI Port, and remote
 * node common objects and establish the connections between them. The goal is to
 * connect the driver to the SCSI domain to exchange I/Os with other devices. These
 * common object connections are shown in the following figure, FC Driver Common Objects:
 * <img src="elx_fc_common_objects.jpg"
 * alt="FC Driver Common Objects" title="FC Driver Common Objects" align="center"/>
 *
 * The first step is to create a connection to the domain by allocating an SLI Port object.
 * The SLI Port object represents a particular FC ID and must be initialized with one. With
 * the SLI Port object, the driver can discover the available SCSI domain(s). On identifying
 * a domain, the driver allocates a domain object and attaches to it using the previous SLI
 * port object.<br><br>
 *
 * @b Note: In some cases, the driver may need to negotiate service parameters (that is,
 * FLOGI) with the domain before attaching.<br><br>
 *
 * Once attached to the domain, the driver can discover and attach to other devices
 * (remote nodes). The exact discovery method depends on the driver, but it typically
 * includes using a position map, querying the fabric name server, or an out-of-band
 * method. In most cases, it is necessary to log in with devices before performing I/Os.
 * Prior to sending login-related ELS commands (ocs_hw_srrs_send()), the driver must
 * allocate a remote node object (ocs_hw_node_alloc()). If the login negotiation is
 * successful, the driver must attach the nodes (ocs_hw_node_attach()) to the SLI Port
 * before exchanging FCP I/O.<br><br>
 *
 * @b Note: The HW manages both the well known fabric address and the name server as
 * nodes in the domain. Therefore, the driver must allocate node objects prior to
 * communicating with either of these entities.
 *
 * <h3>Sending and Receiving I/Os</h3>
 * The HW provides separate interfaces for sending BLS/ ELS/ FC-CT and FCP, but the
 * commands are conceptually similar. Since the commands complete asynchronously,
 * the caller must provide a HW I/O object that maintains the I/O state, as well as
 * provide a callback function. The driver may use the same callback function for all I/O
 * operations, but each operation must use a unique HW I/O object. In the SLI-4
 * architecture, there is a direct association between the HW I/O object and the SGL used
 * to describe the data. Therefore, a driver typically performs the following operations:
 *
 * <ul><li>Allocates a HW I/O object (ocs_hw_io_alloc()).</li>
 * <li>Formats the SGL, specifying both the HW I/O object and the SGL.
 * (ocs_hw_io_init_sges() and ocs_hw_io_add_sge()).</li>
 * <li>Sends the HW I/O (ocs_hw_io_send()).</li></ul>
 *
 * <h3>HW Tear Down</h3>
 * To tear-down the HW:
 *
 * <ol><li>Take the port offline (ocs_hw_port_control()) to prevent receiving further
 * data andevents.</li>
 * <li>Destroy the HW object (ocs_hw_teardown()).</li>
 * <li>Free any memory used by the HW, such as buffers for unsolicited data.</li></ol>
 * <br>
 * </div><!-- overview -->
 *
 */




/**
 * This contains all hw runtime workaround code.  Based on the asic type,
 * asic revision, and range of fw revisions, a particular workaround may be enabled.
 *
 * A workaround may consist of overriding a particular HW/SLI4 value that was initialized
 * during ocs_hw_setup() (for example the MAX_QUEUE overrides for mis-reported queue
 * sizes). Or if required, elements of the ocs_hw_workaround_t structure may be set to
 * control specific runtime behavior.
 *
 * It is intended that the controls in ocs_hw_workaround_t be defined functionally.  So we
 * would have the driver look like:  "if (hw->workaround.enable_xxx) then ...", rather than
 * what we might previously see as "if this is a BE3, then do xxx"
 *
 */


#define HW_FWREV_ZERO		(0ull)
#define HW_FWREV_MAX		(~0ull)

#define SLI4_ASIC_TYPE_ANY	0
#define SLI4_ASIC_REV_ANY	0

/**
 * @brief Internal definition of workarounds
 */

typedef enum {
	HW_WORKAROUND_TEST = 1,
	HW_WORKAROUND_MAX_QUEUE,	/**< Limits all queues */
	HW_WORKAROUND_MAX_RQ,		/**< Limits only the RQ */
	HW_WORKAROUND_RETAIN_TSEND_IO_LENGTH,
	HW_WORKAROUND_WQE_COUNT_METHOD,
	HW_WORKAROUND_RQE_COUNT_METHOD,
	HW_WORKAROUND_USE_UNREGISTERD_RPI,
	HW_WORKAROUND_DISABLE_AR_TGT_DIF, /**< Disable of auto-response target DIF */
	HW_WORKAROUND_DISABLE_SET_DUMP_LOC,
	HW_WORKAROUND_USE_DIF_QUARANTINE,
	HW_WORKAROUND_USE_DIF_SEC_XRI,		/**< Use secondary xri for multiple data phases */
	HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB,	/**< FCFI reported in SRB not correct, use "first" registered domain */
	HW_WORKAROUND_FW_VERSION_TOO_LOW,	/**< The FW version is not the min version supported by this driver */
	HW_WORKAROUND_SGLC_MISREPORTED,	/**< Chip supports SGL Chaining but SGLC is not set in SLI4_PARAMS */
	HW_WORKAROUND_IGNORE_SEND_FRAME_CAPABLE,	/**< Don't use SEND_FRAME capable if FW version is too old */
} hw_workaround_e;

/**
 * @brief Internal workaround structure instance
 */

typedef struct {
	sli4_asic_type_e asic_type;
	sli4_asic_rev_e asic_rev;
	uint64_t fwrev_low;
	uint64_t fwrev_high;

	hw_workaround_e workaround;
	uint32_t value;
} hw_workaround_t;

static hw_workaround_t hw_workarounds[] = {
	{SLI4_ASIC_TYPE_ANY,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_TEST, 999},

	/* Bug: 127585: if_type == 2 returns 0 for total length placed on
	 * FCP_TSEND64_WQE completions.   Note, original driver code enables this
	 * workaround for all asic types
	 */
	{SLI4_ASIC_TYPE_ANY,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_RETAIN_TSEND_IO_LENGTH, 0},

	/* Bug: unknown, Lancer A0 has mis-reported max queue depth */
	{SLI4_ASIC_TYPE_LANCER,	SLI4_ASIC_REV_A0, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_MAX_QUEUE, 2048},

	/* Bug: 143399, BE3 has mis-reported max RQ queue depth */
	{SLI4_ASIC_TYPE_BE3,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV(4,6,293,0),
		HW_WORKAROUND_MAX_RQ, 2048},

	/* Bug: 143399, skyhawk has mis-reported max RQ queue depth */
	{SLI4_ASIC_TYPE_SKYHAWK, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV(10,0,594,0),
		HW_WORKAROUND_MAX_RQ, 2048},

	/* Bug: 103487, BE3 before f/w 4.2.314.0 has mis-reported WQE count method */
	{SLI4_ASIC_TYPE_BE3,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV(4,2,314,0),
		HW_WORKAROUND_WQE_COUNT_METHOD, 1},

	/* Bug: 103487, BE3 before f/w 4.2.314.0 has mis-reported RQE count method */
	{SLI4_ASIC_TYPE_BE3,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV(4,2,314,0),
		HW_WORKAROUND_RQE_COUNT_METHOD, 1},

	/* Bug: 142968, BE3 UE with RPI == 0xffff */
	{SLI4_ASIC_TYPE_BE3,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_USE_UNREGISTERD_RPI, 0},

	/* Bug: unknown, Skyhawk won't support auto-response on target T10-PI  */
	{SLI4_ASIC_TYPE_SKYHAWK, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_DISABLE_AR_TGT_DIF, 0},

	{SLI4_ASIC_TYPE_LANCER,	SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV(1,1,65,0),
		HW_WORKAROUND_DISABLE_SET_DUMP_LOC, 0},

	/* Bug: 160124, Skyhawk quarantine DIF XRIs  */
	{SLI4_ASIC_TYPE_SKYHAWK, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_USE_DIF_QUARANTINE, 0},

	/* Bug: 161832, Skyhawk use secondary XRI for multiple data phase TRECV */
	{SLI4_ASIC_TYPE_SKYHAWK, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_USE_DIF_SEC_XRI, 0},

	/* Bug: xxxxxx, FCFI reported in SRB not corrrect */
	{SLI4_ASIC_TYPE_LANCER, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB, 0},
#if 0
	/* Bug: 165642, FW version check for driver */
	{SLI4_ASIC_TYPE_LANCER, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_1(OCS_MIN_FW_VER_LANCER),
		HW_WORKAROUND_FW_VERSION_TOO_LOW, 0},
#endif
	{SLI4_ASIC_TYPE_SKYHAWK, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_1(OCS_MIN_FW_VER_SKYHAWK),
		HW_WORKAROUND_FW_VERSION_TOO_LOW, 0},

	/* Bug 177061, Lancer FW does not set the SGLC bit */
	{SLI4_ASIC_TYPE_LANCER, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_SGLC_MISREPORTED, 0},

	/* BZ 181208/183914, enable this workaround for ALL revisions */
	{SLI4_ASIC_TYPE_ANY, SLI4_ASIC_REV_ANY, HW_FWREV_ZERO, HW_FWREV_MAX,
		HW_WORKAROUND_IGNORE_SEND_FRAME_CAPABLE, 0},
};

/**
 * @brief Function prototypes
 */

static int32_t ocs_hw_workaround_match(ocs_hw_t *hw, hw_workaround_t *w);

/**
 * @brief Parse the firmware version (name)
 *
 * Parse a string of the form a.b.c.d, returning a uint64_t packed as defined
 * by the HW_FWREV() macro
 *
 * @param fwrev_string pointer to the firmware string
 *
 * @return packed firmware revision value
 */

static uint64_t
parse_fw_version(const char *fwrev_string)
{
	int v[4] = {0};
	const char *p;
	int i;

	for (p = fwrev_string, i = 0; *p && (i < 4); i ++) {
		v[i] = ocs_strtoul(p, 0, 0);
		while(*p && *p != '.') {
			p ++;
		}
		if (*p) {
			p ++;
		}
	}

	/* Special case for bootleg releases with f/w rev 0.0.9999.0, set to max value */
	if (v[2] == 9999) {
		return HW_FWREV_MAX;
	} else {
		return HW_FWREV(v[0], v[1], v[2], v[3]);
	}
}

/**
 * @brief Test for a workaround match
 *
 * Looks at the asic type, asic revision, and fw revision, and returns TRUE if match.
 *
 * @param hw Pointer to the HW structure
 * @param w Pointer to a workaround structure entry
 *
 * @return Return TRUE for a match
 */

static int32_t
ocs_hw_workaround_match(ocs_hw_t *hw, hw_workaround_t *w)
{
	return (((w->asic_type == SLI4_ASIC_TYPE_ANY) || (w->asic_type == hw->sli.asic_type)) &&
		    ((w->asic_rev == SLI4_ASIC_REV_ANY) || (w->asic_rev == hw->sli.asic_rev)) &&
		    (w->fwrev_low <= hw->workaround.fwrev) &&
		    ((w->fwrev_high == HW_FWREV_MAX) || (hw->workaround.fwrev < w->fwrev_high)));
}

/**
 * @brief Setup HW runtime workarounds
 *
 * The function is called at the end of ocs_hw_setup() to setup any runtime workarounds
 * based on the HW/SLI setup.
 *
 * @param hw Pointer to HW structure
 *
 * @return none
 */

void
ocs_hw_workaround_setup(struct ocs_hw_s *hw)
{
	hw_workaround_t *w;
	sli4_t *sli4 = &hw->sli;
	uint32_t i;

	/* Initialize the workaround settings */
	ocs_memset(&hw->workaround, 0, sizeof(hw->workaround));

	/* If hw_war_version is non-null, then its a value that was set by a module parameter
	 * (sorry for the break in abstraction, but workarounds are ... well, workarounds)
	 */

	if (hw->hw_war_version) {
		hw->workaround.fwrev = parse_fw_version(hw->hw_war_version);
	} else {
		hw->workaround.fwrev = parse_fw_version((char*) sli4->config.fw_name[0]);
	}

	/* Walk the workaround list, if a match is found, then handle it */
	for (i = 0, w = hw_workarounds; i < ARRAY_SIZE(hw_workarounds); i++, w++) {
		if (ocs_hw_workaround_match(hw, w)) {
			switch(w->workaround) {

			case HW_WORKAROUND_TEST: {
				ocs_log_debug(hw->os, "Override: test: %d\n", w->value);
				break;
			}

			case HW_WORKAROUND_RETAIN_TSEND_IO_LENGTH: {
				ocs_log_debug(hw->os, "HW Workaround: retain TSEND IO length\n");
				hw->workaround.retain_tsend_io_length = 1;
				break;
			}
			case HW_WORKAROUND_MAX_QUEUE: {
				sli4_qtype_e q;

				ocs_log_debug(hw->os, "HW Workaround: override max_qentries: %d\n", w->value);
				for (q = SLI_QTYPE_EQ; q < SLI_QTYPE_MAX; q++) {
					if (hw->num_qentries[q] > w->value) {
						hw->num_qentries[q] = w->value;
					}
				}
				break;
			}
			case HW_WORKAROUND_MAX_RQ: {
				ocs_log_debug(hw->os, "HW Workaround: override RQ max_qentries: %d\n", w->value);
				if (hw->num_qentries[SLI_QTYPE_RQ] > w->value) {
					hw->num_qentries[SLI_QTYPE_RQ] = w->value;
				}
				break;
			}
			case HW_WORKAROUND_WQE_COUNT_METHOD: {
				ocs_log_debug(hw->os, "HW Workaround: set WQE count method=%d\n", w->value);
				sli4->config.count_method[SLI_QTYPE_WQ] = w->value;
				sli_calc_max_qentries(sli4);
				break;
			}
			case HW_WORKAROUND_RQE_COUNT_METHOD: {
				ocs_log_debug(hw->os, "HW Workaround: set RQE count method=%d\n", w->value);
				sli4->config.count_method[SLI_QTYPE_RQ] = w->value;
				sli_calc_max_qentries(sli4);
				break;
			}
			case HW_WORKAROUND_USE_UNREGISTERD_RPI:
				ocs_log_debug(hw->os, "HW Workaround: use unreg'd RPI if rnode->indicator == 0xFFFF\n");
				hw->workaround.use_unregistered_rpi = TRUE;
				/*
				 * Allocate an RPI that is never registered, to be used in the case where
				 * a node has been unregistered, and its indicator (RPI) value is set to 0xFFFF
				 */
				if (sli_resource_alloc(&hw->sli, SLI_RSRC_FCOE_RPI, &hw->workaround.unregistered_rid,
					&hw->workaround.unregistered_index)) {
					ocs_log_err(hw->os, "sli_resource_alloc unregistered RPI failed\n");
					hw->workaround.use_unregistered_rpi = FALSE;
				}
				break;
			case HW_WORKAROUND_DISABLE_AR_TGT_DIF:
				ocs_log_debug(hw->os, "HW Workaround: disable AR on T10-PI TSEND\n");
				hw->workaround.disable_ar_tgt_dif = TRUE;
				break;
			case HW_WORKAROUND_DISABLE_SET_DUMP_LOC:
				ocs_log_debug(hw->os, "HW Workaround: disable set_dump_loc\n");
				hw->workaround.disable_dump_loc = TRUE;
				break;
			case HW_WORKAROUND_USE_DIF_QUARANTINE:
				ocs_log_debug(hw->os, "HW Workaround: use DIF quarantine\n");
				hw->workaround.use_dif_quarantine = TRUE;
				break;
			case HW_WORKAROUND_USE_DIF_SEC_XRI:
				ocs_log_debug(hw->os, "HW Workaround: use DIF secondary xri\n");
				hw->workaround.use_dif_sec_xri = TRUE;
				break;
			case HW_WORKAROUND_OVERRIDE_FCFI_IN_SRB:
				ocs_log_debug(hw->os, "HW Workaround: override FCFI in SRB\n");
				hw->workaround.override_fcfi = TRUE;
				break;

			case HW_WORKAROUND_FW_VERSION_TOO_LOW:
				ocs_log_debug(hw->os, "HW Workaround: fw version is below the minimum for this driver\n");
				hw->workaround.fw_version_too_low = TRUE;
				break;
			case HW_WORKAROUND_SGLC_MISREPORTED:
				ocs_log_debug(hw->os, "HW Workaround: SGLC misreported - chaining is enabled\n");
				hw->workaround.sglc_misreported = TRUE;
				break;
			case HW_WORKAROUND_IGNORE_SEND_FRAME_CAPABLE:
				ocs_log_debug(hw->os, "HW Workaround: not SEND_FRAME capable - disabled\n");
				hw->workaround.ignore_send_frame = TRUE;
				break;
			} /* switch(w->workaround) */
		}
	}
}