diff mbox

[dpdk-dev,3/4] bcm: new poll mode driver

Message ID 1423247795-22399-4-git-send-email-stephen@networkplumber.org (mailing list archive)
State Superseded, archived
Headers show

Commit Message

Stephen Hemminger Feb. 6, 2015, 6:36 p.m. UTC
From: Stephen Hemminger <shemming@brocade.com>

Add driver for the Broadcom NetXtremeII 10 gigabit devices.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 lib/Makefile                        |     1 +
 lib/librte_pmd_bcm/Makefile         |    28 +
 lib/librte_pmd_bcm/bcm.c            | 11817 ++++++++++++++++++++++++++++++
 lib/librte_pmd_bcm/bcm.h            |  1998 +++++
 lib/librte_pmd_bcm/bcm_ethdev.c     |   544 ++
 lib/librte_pmd_bcm/bcm_ethdev.h     |    79 +
 lib/librte_pmd_bcm/bcm_logs.h       |    51 +
 lib/librte_pmd_bcm/bcm_rxtx.c       |   487 ++
 lib/librte_pmd_bcm/bcm_rxtx.h       |    85 +
 lib/librte_pmd_bcm/bcm_stats.c      |  1619 ++++
 lib/librte_pmd_bcm/bcm_stats.h      |   633 ++
 lib/librte_pmd_bcm/bcm_vfpf.c       |   597 ++
 lib/librte_pmd_bcm/bcm_vfpf.h       |   315 +
 lib/librte_pmd_bcm/debug.c          |   113 +
 lib/librte_pmd_bcm/ecore_fw_defs.h  |   423 ++
 lib/librte_pmd_bcm/ecore_hsi.h      |  6349 ++++++++++++++++
 lib/librte_pmd_bcm/ecore_init.h     |   842 +++
 lib/librte_pmd_bcm/ecore_init_ops.h |   886 +++
 lib/librte_pmd_bcm/ecore_mfw_req.h  |   207 +
 lib/librte_pmd_bcm/ecore_reg.h      |  3664 ++++++++++
 lib/librte_pmd_bcm/ecore_sp.c       |  5455 ++++++++++++++
 lib/librte_pmd_bcm/ecore_sp.h       |  1796 +++++
 lib/librte_pmd_bcm/elink.c          | 13378 ++++++++++++++++++++++++++++++++++
 lib/librte_pmd_bcm/elink.h          |   610 ++
 24 files changed, 51977 insertions(+)
 create mode 100644 lib/librte_pmd_bcm/Makefile
 create mode 100644 lib/librte_pmd_bcm/bcm.c
 create mode 100644 lib/librte_pmd_bcm/bcm.h
 create mode 100644 lib/librte_pmd_bcm/bcm_ethdev.c
 create mode 100644 lib/librte_pmd_bcm/bcm_ethdev.h
 create mode 100644 lib/librte_pmd_bcm/bcm_logs.h
 create mode 100644 lib/librte_pmd_bcm/bcm_rxtx.c
 create mode 100644 lib/librte_pmd_bcm/bcm_rxtx.h
 create mode 100644 lib/librte_pmd_bcm/bcm_stats.c
 create mode 100644 lib/librte_pmd_bcm/bcm_stats.h
 create mode 100644 lib/librte_pmd_bcm/bcm_vfpf.c
 create mode 100644 lib/librte_pmd_bcm/bcm_vfpf.h
 create mode 100644 lib/librte_pmd_bcm/debug.c
 create mode 100644 lib/librte_pmd_bcm/ecore_fw_defs.h
 create mode 100644 lib/librte_pmd_bcm/ecore_hsi.h
 create mode 100644 lib/librte_pmd_bcm/ecore_init.h
 create mode 100644 lib/librte_pmd_bcm/ecore_init_ops.h
 create mode 100644 lib/librte_pmd_bcm/ecore_mfw_req.h
 create mode 100644 lib/librte_pmd_bcm/ecore_reg.h
 create mode 100644 lib/librte_pmd_bcm/ecore_sp.c
 create mode 100644 lib/librte_pmd_bcm/ecore_sp.h
 create mode 100644 lib/librte_pmd_bcm/elink.c
 create mode 100644 lib/librte_pmd_bcm/elink.h
diff mbox

Patch

diff --git a/lib/Makefile b/lib/Makefile
index d617d81..023fc9c 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -45,6 +45,7 @@  DIRS-$(CONFIG_RTE_LIBRTE_E1000_PMD) += librte_pmd_e1000
 DIRS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += librte_pmd_ixgbe
 DIRS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += librte_pmd_i40e
 DIRS-$(CONFIG_RTE_LIBRTE_ENIC_PMD) += librte_pmd_enic
+DIRS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += librte_pmd_bcm
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_BOND) += librte_pmd_bond
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_RING) += librte_pmd_ring
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += librte_pmd_pcap
diff --git a/lib/librte_pmd_bcm/Makefile b/lib/librte_pmd_bcm/Makefile
new file mode 100644
index 0000000..2f3df65
--- /dev/null
+++ b/lib/librte_pmd_bcm/Makefile
@@ -0,0 +1,28 @@ 
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_bcm.a
+
+CFLAGS += -O3 -g
+CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -DZLIB_CONST
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += bcm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += bcm_rxtx.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += bcm_stats.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += bcm_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += ecore_sp.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += elink.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += bcm_vfpf.c
+SRCS-$(CONFIG_RTE_LIBRTE_BCM_DEBUG) += debug.c
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += lib/librte_eal lib/librte_ether lib/librte_hash
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BCM_PMD) += lib/librte_mempool lib/librte_mbuf
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/lib/librte_pmd_bcm/bcm.c b/lib/librte_pmd_bcm/bcm.c
new file mode 100644
index 0000000..26f190e
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm.c
@@ -0,0 +1,11817 @@ 
+/*-
+ * Copyright (c) 2007-2013 Broadcom Corporation. All rights reserved.
+ *
+ * Eric Davis        <edavis@broadcom.com>
+ * David Christensen <davidch@broadcom.com>
+ * Gary Zambrano     <zambrano@broadcom.com>
+ *
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ * All rights reserved.
+ *
+ * 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 Broadcom Corporation nor the name of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written consent.
+ *
+ * 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 OWNER 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.
+ */
+
+#define BCM_DRIVER_VERSION "1.78.18"
+
+#include "bcm.h"
+#include "bcm_vfpf.h"
+#include "ecore_sp.h"
+#include "ecore_init.h"
+#include "ecore_init_ops.h"
+
+#include "rte_pci_dev_ids.h"
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <zlib.h>
+
+static z_stream zlib_stream;
+
+#define EVL_VLID_MASK 0x0FFF
+
+#define BCM_DEF_SB_ATT_IDX 0x0001
+#define BCM_DEF_SB_IDX     0x0002
+
+/*
+ * FLR Support - bcm_pf_flr_clnup() is called during nic_load in the per
+ * function HW initialization.
+ */
+#define FLR_WAIT_USEC     10000	/* 10 msecs */
+#define FLR_WAIT_INTERVAL 50	/* usecs */
+#define FLR_POLL_CNT      (FLR_WAIT_USEC / FLR_WAIT_INTERVAL)	/* 200 */
+
+struct pbf_pN_buf_regs {
+	int pN;
+	uint32_t init_crd;
+	uint32_t crd;
+	uint32_t crd_freed;
+};
+
+struct pbf_pN_cmd_regs {
+	int pN;
+	uint32_t lines_occup;
+	uint32_t lines_freed;
+};
+
+/* resources needed for unloading a previously loaded device */
+
+#define BCM_PREV_WAIT_NEEDED 1
+rte_spinlock_t bcm_prev_mtx;
+struct bcm_prev_list_node {
+	LIST_ENTRY(bcm_prev_list_node) node;
+	uint8_t bus;
+	uint8_t slot;
+	uint8_t path;
+	uint8_t aer;
+	uint8_t undi;
+};
+static LIST_HEAD(, bcm_prev_list_node) bcm_prev_list =
+LIST_HEAD_INITIALIZER(bcm_prev_list);
+
+static int load_count[2][3] = { {0} };	/* per-path: 0-common, 1-port0, 2-port1 */
+
+static void bcm_cmng_fns_init(struct bcm_softc *sc, uint8_t read_cfg,
+			      uint8_t cmng_type);
+static int bcm_get_cmng_fns_mode(struct bcm_softc *sc);
+static void storm_memset_cmng(struct bcm_softc *sc, struct cmng_init *cmng,
+			      uint8_t port);
+static void bcm_set_reset_global(struct bcm_softc *sc);
+static void bcm_set_reset_in_progress(struct bcm_softc *sc);
+static uint8_t bcm_reset_is_done(struct bcm_softc *sc, int engine);
+static uint8_t bcm_clear_pf_load(struct bcm_softc *sc);
+static uint8_t bcm_chk_parity_attn(struct bcm_softc *sc, uint8_t * global,
+				   uint8_t print);
+static void bcm_int_disable(struct bcm_softc *sc);
+static int bcm_release_leader_lock(struct bcm_softc *sc);
+static void bcm_pf_disable(struct bcm_softc *sc);
+static void bcm_update_rx_prod(struct bcm_softc *sc, struct bcm_fastpath *fp,
+			       uint16_t rx_bd_prod, uint16_t rx_cq_prod);
+static void bcm_link_report(struct bcm_softc *sc);
+void bcm_link_status_update(struct bcm_softc *sc);
+static int bcm_alloc_mem(struct bcm_softc *sc);
+static void bcm_free_mem(struct bcm_softc *sc);
+static int bcm_alloc_fw_stats_mem(struct bcm_softc *sc);
+static void bcm_free_fw_stats_mem(struct bcm_softc *sc);
+static __attribute__ ((noinline))
+int bcm_nic_load(struct bcm_softc *sc);
+
+static int bcm_handle_sp_tq(struct bcm_softc *sc);
+static void bcm_handle_fp_tq(struct bcm_fastpath *fp, int scan_fp);
+static void bcm_periodic_stop(struct bcm_softc *sc);
+static void bcm_ack_sb(struct bcm_softc *sc, uint8_t igu_sb_id, uint8_t storm,
+		       uint16_t index, uint8_t op, uint8_t update);
+
+int bcm_test_bit(int nr, volatile unsigned long *addr)
+{
+	int res;
+	mb();
+	res = ((*addr) & (1UL << nr)) != 0;
+	mb();
+	return res;
+}
+
+void bcm_set_bit(unsigned int nr, volatile unsigned long *addr)
+{
+	__sync_fetch_and_or(addr, (1UL << nr));
+}
+
+void bcm_clear_bit(int nr, volatile unsigned long *addr)
+{
+	__sync_fetch_and_and(addr, ~(1UL << nr));
+}
+
+int bcm_test_and_clear_bit(int nr, volatile unsigned long *addr)
+{
+	unsigned long mask = (1UL << nr);
+	return __sync_fetch_and_and(addr, ~mask) & mask;
+}
+
+int bcm_cmpxchg(volatile int *addr, int old, int new)
+{
+	return __sync_val_compare_and_swap(addr, old, new);
+}
+
+int
+bcm_dma_alloc(struct bcm_softc *sc, size_t size, struct bcm_dma *dma,
+	      const char *msg, uint32_t align)
+{
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	const struct rte_memzone *z;
+
+	dma->sc = sc;
+	if (IS_PF(sc))
+		sprintf(mz_name, "bcm%d_%s_%lx", SC_ABS_FUNC(sc), msg,
+			rte_get_timer_cycles());
+	else
+		sprintf(mz_name, "bcm%d_%s_%lx", sc->pcie_device, msg,
+			rte_get_timer_cycles());
+
+	/* Caller must take care that strlen(mz_name) < RTE_MEMZONE_NAMESIZE */
+	z = rte_memzone_reserve_aligned(mz_name, (uint64_t) (size),
+					rte_lcore_to_socket_id(rte_lcore_id()),
+					0, align);
+	if (z == NULL) {
+		PMD_DRV_LOG(ERR, "DMA alloc failed for %s", msg);
+		return -ENOMEM;
+	}
+	dma->paddr = (uint64_t) z->phys_addr;
+	dma->vaddr = z->addr;
+
+	PMD_DRV_LOG(DEBUG, "%s: virt=%p phys=%lx", msg, dma->vaddr, dma->paddr);
+
+	return 0;
+}
+
+static int bcm_acquire_hw_lock(struct bcm_softc *sc, uint32_t resource)
+{
+	uint32_t lock_status;
+	uint32_t resource_bit = (1 << resource);
+	int func = SC_FUNC(sc);
+	uint32_t hw_lock_control_reg;
+	int cnt;
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* validate the resource is within range */
+	if (resource > HW_LOCK_MAX_RESOURCE_VALUE) {
+		PMD_DRV_LOG(NOTICE,
+			    "resource 0x%x > HW_LOCK_MAX_RESOURCE_VALUE",
+			    resource);
+		return -1;
+	}
+
+	if (func <= 5) {
+		hw_lock_control_reg = (MISC_REG_DRIVER_CONTROL_1 + (func * 8));
+	} else {
+		hw_lock_control_reg =
+		    (MISC_REG_DRIVER_CONTROL_7 + ((func - 6) * 8));
+	}
+
+	/* validate the resource is not already taken */
+	lock_status = REG_RD(sc, hw_lock_control_reg);
+	if (lock_status & resource_bit) {
+		PMD_DRV_LOG(NOTICE,
+			    "resource in use (status 0x%x bit 0x%x)",
+			    lock_status, resource_bit);
+		return -1;
+	}
+
+	/* try every 5ms for 5 seconds */
+	for (cnt = 0; cnt < 1000; cnt++) {
+		REG_WR(sc, (hw_lock_control_reg + 4), resource_bit);
+		lock_status = REG_RD(sc, hw_lock_control_reg);
+		if (lock_status & resource_bit) {
+			return 0;
+		}
+		DELAY(5000);
+	}
+
+	PMD_DRV_LOG(NOTICE, "Resource lock timeout!");
+	return -1;
+}
+
+static int bcm_release_hw_lock(struct bcm_softc *sc, uint32_t resource)
+{
+	uint32_t lock_status;
+	uint32_t resource_bit = (1 << resource);
+	int func = SC_FUNC(sc);
+	uint32_t hw_lock_control_reg;
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* validate the resource is within range */
+	if (resource > HW_LOCK_MAX_RESOURCE_VALUE) {
+		PMD_DRV_LOG(NOTICE,
+			    "resource 0x%x > HW_LOCK_MAX_RESOURCE_VALUE",
+			    resource);
+		return -1;
+	}
+
+	if (func <= 5) {
+		hw_lock_control_reg = (MISC_REG_DRIVER_CONTROL_1 + (func * 8));
+	} else {
+		hw_lock_control_reg =
+		    (MISC_REG_DRIVER_CONTROL_7 + ((func - 6) * 8));
+	}
+
+	/* validate the resource is currently taken */
+	lock_status = REG_RD(sc, hw_lock_control_reg);
+	if (!(lock_status & resource_bit)) {
+		PMD_DRV_LOG(NOTICE,
+			    "resource not in use (status 0x%x bit 0x%x)",
+			    lock_status, resource_bit);
+		return -1;
+	}
+
+	REG_WR(sc, hw_lock_control_reg, resource_bit);
+	return 0;
+}
+
+/* copy command into DMAE command memory and set DMAE command Go */
+void bcm_post_dmae(struct bcm_softc *sc, struct dmae_command *dmae, int idx)
+{
+	uint32_t cmd_offset;
+	uint32_t i;
+
+	cmd_offset = (DMAE_REG_CMD_MEM + (sizeof(struct dmae_command) * idx));
+	for (i = 0; i < ((sizeof(struct dmae_command) / 4)); i++) {
+		REG_WR(sc, (cmd_offset + (i * 4)), *(((uint32_t *) dmae) + i));
+	}
+
+	REG_WR(sc, dmae_reg_go_c[idx], 1);
+}
+
+uint32_t bcm_dmae_opcode_add_comp(uint32_t opcode, uint8_t comp_type)
+{
+	return (opcode | ((comp_type << DMAE_COMMAND_C_DST_SHIFT) |
+			  DMAE_COMMAND_C_TYPE_ENABLE));
+}
+
+uint32_t bcm_dmae_opcode_clr_src_reset(uint32_t opcode)
+{
+	return (opcode & ~DMAE_COMMAND_SRC_RESET);
+}
+
+uint32_t
+bcm_dmae_opcode(struct bcm_softc * sc, uint8_t src_type, uint8_t dst_type,
+		uint8_t with_comp, uint8_t comp_type)
+{
+	uint32_t opcode = 0;
+
+	opcode |= ((src_type << DMAE_COMMAND_SRC_SHIFT) |
+		   (dst_type << DMAE_COMMAND_DST_SHIFT));
+
+	opcode |= (DMAE_COMMAND_SRC_RESET | DMAE_COMMAND_DST_RESET);
+
+	opcode |= (SC_PORT(sc) ? DMAE_CMD_PORT_1 : DMAE_CMD_PORT_0);
+
+	opcode |= ((SC_VN(sc) << DMAE_COMMAND_E1HVN_SHIFT) |
+		   (SC_VN(sc) << DMAE_COMMAND_DST_VN_SHIFT));
+
+	opcode |= (DMAE_COM_SET_ERR << DMAE_COMMAND_ERR_POLICY_SHIFT);
+
+#ifdef __BIG_ENDIAN
+	opcode |= DMAE_CMD_ENDIANITY_B_DW_SWAP;
+#else
+	opcode |= DMAE_CMD_ENDIANITY_DW_SWAP;
+#endif
+
+	if (with_comp) {
+		opcode = bcm_dmae_opcode_add_comp(opcode, comp_type);
+	}
+
+	return opcode;
+}
+
+static void
+bcm_prep_dmae_with_comp(struct bcm_softc *sc, struct dmae_command *dmae,
+			uint8_t src_type, uint8_t dst_type)
+{
+	memset(dmae, 0, sizeof(struct dmae_command));
+
+	/* set the opcode */
+	dmae->opcode = bcm_dmae_opcode(sc, src_type, dst_type,
+				       TRUE, DMAE_COMP_PCI);
+
+	/* fill in the completion parameters */
+	dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, wb_comp));
+	dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, wb_comp));
+	dmae->comp_val = DMAE_COMP_VAL;
+}
+
+/* issue a DMAE command over the init channel and wait for completion */
+static int
+bcm_issue_dmae_with_comp(struct bcm_softc *sc, struct dmae_command *dmae)
+{
+	uint32_t *wb_comp = BCM_SP(sc, wb_comp);
+	int timeout = CHIP_REV_IS_SLOW(sc) ? 400000 : 4000;
+
+	/* reset completion */
+	*wb_comp = 0;
+
+	/* post the command on the channel used for initializations */
+	bcm_post_dmae(sc, dmae, INIT_DMAE_C(sc));
+
+	/* wait for completion */
+	DELAY(500);
+
+	while ((*wb_comp & ~DMAE_PCI_ERR_FLAG) != DMAE_COMP_VAL) {
+		if (!timeout ||
+		    (sc->recovery_state != BCM_RECOVERY_DONE &&
+		     sc->recovery_state != BCM_RECOVERY_NIC_LOADING)) {
+			PMD_DRV_LOG(INFO, "DMAE timeout!");
+			return DMAE_TIMEOUT;
+		}
+
+		timeout--;
+		DELAY(50);
+	}
+
+	if (*wb_comp & DMAE_PCI_ERR_FLAG) {
+		PMD_DRV_LOG(INFO, "DMAE PCI error!");
+		return DMAE_PCI_ERROR;
+	}
+
+	return 0;
+}
+
+void bcm_read_dmae(struct bcm_softc *sc, uint32_t src_addr, uint32_t len32)
+{
+	struct dmae_command dmae;
+	uint32_t *data;
+	uint32_t i;
+	int rc;
+
+	if (!sc->dmae_ready) {
+		data = BCM_SP(sc, wb_data[0]);
+
+		for (i = 0; i < len32; i++) {
+			data[i] = REG_RD(sc, (src_addr + (i * 4)));
+		}
+
+		return;
+	}
+
+	/* set opcode and fixed command fields */
+	bcm_prep_dmae_with_comp(sc, &dmae, DMAE_SRC_GRC, DMAE_DST_PCI);
+
+	/* fill in addresses and len */
+	dmae.src_addr_lo = (src_addr >> 2);	/* GRC addr has dword resolution */
+	dmae.src_addr_hi = 0;
+	dmae.dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, wb_data));
+	dmae.dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, wb_data));
+	dmae.len = len32;
+
+	/* issue the command and wait for completion */
+	if ((rc = bcm_issue_dmae_with_comp(sc, &dmae)) != 0) {
+		rte_panic("DMAE failed (%d)", rc);
+	};
+}
+
+void
+bcm_write_dmae(struct bcm_softc *sc, phys_addr_t dma_addr, uint32_t dst_addr,
+	       uint32_t len32)
+{
+	struct dmae_command dmae;
+	int rc;
+
+	if (!sc->dmae_ready) {
+		ecore_init_str_wr(sc, dst_addr, BCM_SP(sc, wb_data[0]), len32);
+		return;
+	}
+
+	/* set opcode and fixed command fields */
+	bcm_prep_dmae_with_comp(sc, &dmae, DMAE_SRC_PCI, DMAE_DST_GRC);
+
+	/* fill in addresses and len */
+	dmae.src_addr_lo = U64_LO(dma_addr);
+	dmae.src_addr_hi = U64_HI(dma_addr);
+	dmae.dst_addr_lo = (dst_addr >> 2);	/* GRC addr has dword resolution */
+	dmae.dst_addr_hi = 0;
+	dmae.len = len32;
+
+	/* issue the command and wait for completion */
+	if ((rc = bcm_issue_dmae_with_comp(sc, &dmae)) != 0) {
+		rte_panic("DMAE failed (%d)", rc);
+	}
+}
+
+static void
+bcm_write_dmae_phys_len(struct bcm_softc *sc, phys_addr_t phys_addr,
+			uint32_t addr, uint32_t len)
+{
+	uint32_t dmae_wr_max = DMAE_LEN32_WR_MAX(sc);
+	uint32_t offset = 0;
+
+	while (len > dmae_wr_max) {
+		bcm_write_dmae(sc, (phys_addr + offset),	/* src DMA address */
+			       (addr + offset),	/* dst GRC address */
+			       dmae_wr_max);
+		offset += (dmae_wr_max * 4);
+		len -= dmae_wr_max;
+	}
+
+	bcm_write_dmae(sc, (phys_addr + offset),	/* src DMA address */
+		       (addr + offset),	/* dst GRC address */
+		       len);
+}
+
+void
+bcm_set_ctx_validation(struct bcm_softc *sc, struct eth_context *cxt,
+		       uint32_t cid)
+{
+	/* ustorm cxt validation */
+	cxt->ustorm_ag_context.cdu_usage =
+	    CDU_RSRVD_VALUE_TYPE_A(HW_CID(sc, cid),
+				   CDU_REGION_NUMBER_UCM_AG,
+				   ETH_CONNECTION_TYPE);
+	/* xcontext validation */
+	cxt->xstorm_ag_context.cdu_reserved =
+	    CDU_RSRVD_VALUE_TYPE_A(HW_CID(sc, cid),
+				   CDU_REGION_NUMBER_XCM_AG,
+				   ETH_CONNECTION_TYPE);
+}
+
+static void
+bcm_storm_memset_hc_timeout(struct bcm_softc *sc, uint8_t fw_sb_id,
+			    uint8_t sb_index, uint8_t ticks)
+{
+	uint32_t addr =
+	    (BAR_CSTRORM_INTMEM +
+	     CSTORM_STATUS_BLOCK_DATA_TIMEOUT_OFFSET(fw_sb_id, sb_index));
+
+	REG_WR8(sc, addr, ticks);
+}
+
+static void
+bcm_storm_memset_hc_disable(struct bcm_softc *sc, uint16_t fw_sb_id,
+			    uint8_t sb_index, uint8_t disable)
+{
+	uint32_t enable_flag =
+	    (disable) ? 0 : (1 << HC_INDEX_DATA_HC_ENABLED_SHIFT);
+	uint32_t addr =
+	    (BAR_CSTRORM_INTMEM +
+	     CSTORM_STATUS_BLOCK_DATA_FLAGS_OFFSET(fw_sb_id, sb_index));
+	uint8_t flags;
+
+	/* clear and set */
+	flags = REG_RD8(sc, addr);
+	flags &= ~HC_INDEX_DATA_HC_ENABLED;
+	flags |= enable_flag;
+	REG_WR8(sc, addr, flags);
+}
+
+void
+bcm_update_coalesce_sb_index(struct bcm_softc *sc, uint8_t fw_sb_id,
+			     uint8_t sb_index, uint8_t disable, uint16_t usec)
+{
+	uint8_t ticks = (usec / 4);
+
+	bcm_storm_memset_hc_timeout(sc, fw_sb_id, sb_index, ticks);
+
+	disable = (disable) ? 1 : ((usec) ? 0 : 1);
+	bcm_storm_memset_hc_disable(sc, fw_sb_id, sb_index, disable);
+}
+
+uint32_t elink_cb_reg_read(struct bcm_softc *sc, uint32_t reg_addr)
+{
+	return REG_RD(sc, reg_addr);
+}
+
+void elink_cb_reg_write(struct bcm_softc *sc, uint32_t reg_addr, uint32_t val)
+{
+	REG_WR(sc, reg_addr, val);
+}
+
+void
+elink_cb_event_log(__rte_unused struct bcm_softc *sc,
+		   __rte_unused const elink_log_id_t elink_log_id, ...)
+{
+	PMD_DRV_LOG(DEBUG, "ELINK EVENT LOG (%d)", elink_log_id);
+}
+
+static int bcm_set_spio(struct bcm_softc *sc, int spio, uint32_t mode)
+{
+	uint32_t spio_reg;
+
+	/* Only 2 SPIOs are configurable */
+	if ((spio != MISC_SPIO_SPIO4) && (spio != MISC_SPIO_SPIO5)) {
+		PMD_DRV_LOG(NOTICE, "Invalid SPIO 0x%x", spio);
+		return -1;
+	}
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_SPIO);
+
+	/* read SPIO and mask except the float bits */
+	spio_reg = (REG_RD(sc, MISC_REG_SPIO) & MISC_SPIO_FLOAT);
+
+	switch (mode) {
+	case MISC_SPIO_OUTPUT_LOW:
+		/* clear FLOAT and set CLR */
+		spio_reg &= ~(spio << MISC_SPIO_FLOAT_POS);
+		spio_reg |= (spio << MISC_SPIO_CLR_POS);
+		break;
+
+	case MISC_SPIO_OUTPUT_HIGH:
+		/* clear FLOAT and set SET */
+		spio_reg &= ~(spio << MISC_SPIO_FLOAT_POS);
+		spio_reg |= (spio << MISC_SPIO_SET_POS);
+		break;
+
+	case MISC_SPIO_INPUT_HI_Z:
+		/* set FLOAT */
+		spio_reg |= (spio << MISC_SPIO_FLOAT_POS);
+		break;
+
+	default:
+		break;
+	}
+
+	REG_WR(sc, MISC_REG_SPIO, spio_reg);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_SPIO);
+
+	return 0;
+}
+
+static int bcm_gpio_read(struct bcm_softc *sc, int gpio_num, uint8_t port)
+{
+	/* The GPIO should be swapped if swap register is set and active */
+	int gpio_port = ((REG_RD(sc, NIG_REG_PORT_SWAP) &&
+			  REG_RD(sc, NIG_REG_STRAP_OVERRIDE)) ^ port);
+	int gpio_shift = gpio_num;
+	if (gpio_port)
+		gpio_shift += MISC_REGISTERS_GPIO_PORT_SHIFT;
+
+	uint32_t gpio_mask = (1 << gpio_shift);
+	uint32_t gpio_reg;
+
+	if (gpio_num > MISC_REGISTERS_GPIO_3) {
+		PMD_DRV_LOG(NOTICE, "Invalid GPIO %d", gpio_num);
+		return -1;
+	}
+
+	/* read GPIO value */
+	gpio_reg = REG_RD(sc, MISC_REG_GPIO);
+
+	/* get the requested pin value */
+	return ((gpio_reg & gpio_mask) == gpio_mask) ? 1 : 0;
+}
+
+static int
+bcm_gpio_write(struct bcm_softc *sc, int gpio_num, uint32_t mode, uint8_t port)
+{
+	/* The GPIO should be swapped if swap register is set and active */
+	int gpio_port = ((REG_RD(sc, NIG_REG_PORT_SWAP) &&
+			  REG_RD(sc, NIG_REG_STRAP_OVERRIDE)) ^ port);
+	int gpio_shift = gpio_num;
+	if (gpio_port)
+		gpio_shift += MISC_REGISTERS_GPIO_PORT_SHIFT;
+
+	uint32_t gpio_mask = (1 << gpio_shift);
+	uint32_t gpio_reg;
+
+	if (gpio_num > MISC_REGISTERS_GPIO_3) {
+		PMD_DRV_LOG(NOTICE, "Invalid GPIO %d", gpio_num);
+		return -1;
+	}
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	/* read GPIO and mask except the float bits */
+	gpio_reg = (REG_RD(sc, MISC_REG_GPIO) & MISC_REGISTERS_GPIO_FLOAT);
+
+	switch (mode) {
+	case MISC_REGISTERS_GPIO_OUTPUT_LOW:
+		/* clear FLOAT and set CLR */
+		gpio_reg &= ~(gpio_mask << MISC_REGISTERS_GPIO_FLOAT_POS);
+		gpio_reg |= (gpio_mask << MISC_REGISTERS_GPIO_CLR_POS);
+		break;
+
+	case MISC_REGISTERS_GPIO_OUTPUT_HIGH:
+		/* clear FLOAT and set SET */
+		gpio_reg &= ~(gpio_mask << MISC_REGISTERS_GPIO_FLOAT_POS);
+		gpio_reg |= (gpio_mask << MISC_REGISTERS_GPIO_SET_POS);
+		break;
+
+	case MISC_REGISTERS_GPIO_INPUT_HI_Z:
+		/* set FLOAT */
+		gpio_reg |= (gpio_mask << MISC_REGISTERS_GPIO_FLOAT_POS);
+		break;
+
+	default:
+		break;
+	}
+
+	REG_WR(sc, MISC_REG_GPIO, gpio_reg);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	return 0;
+}
+
+static int
+bcm_gpio_mult_write(struct bcm_softc *sc, uint8_t pins, uint32_t mode)
+{
+	uint32_t gpio_reg;
+
+	/* any port swapping should be handled by caller */
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	/* read GPIO and mask except the float bits */
+	gpio_reg = REG_RD(sc, MISC_REG_GPIO);
+	gpio_reg &= ~(pins << MISC_REGISTERS_GPIO_FLOAT_POS);
+	gpio_reg &= ~(pins << MISC_REGISTERS_GPIO_CLR_POS);
+	gpio_reg &= ~(pins << MISC_REGISTERS_GPIO_SET_POS);
+
+	switch (mode) {
+	case MISC_REGISTERS_GPIO_OUTPUT_LOW:
+		/* set CLR */
+		gpio_reg |= (pins << MISC_REGISTERS_GPIO_CLR_POS);
+		break;
+
+	case MISC_REGISTERS_GPIO_OUTPUT_HIGH:
+		/* set SET */
+		gpio_reg |= (pins << MISC_REGISTERS_GPIO_SET_POS);
+		break;
+
+	case MISC_REGISTERS_GPIO_INPUT_HI_Z:
+		/* set FLOAT */
+		gpio_reg |= (pins << MISC_REGISTERS_GPIO_FLOAT_POS);
+		break;
+
+	default:
+		PMD_DRV_LOG(NOTICE, "Invalid GPIO mode assignment %d", mode);
+		bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+		return -1;
+	}
+
+	REG_WR(sc, MISC_REG_GPIO, gpio_reg);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	return 0;
+}
+
+static int
+bcm_gpio_int_write(struct bcm_softc *sc, int gpio_num, uint32_t mode,
+		   uint8_t port)
+{
+	/* The GPIO should be swapped if swap register is set and active */
+	int gpio_port = ((REG_RD(sc, NIG_REG_PORT_SWAP) &&
+			  REG_RD(sc, NIG_REG_STRAP_OVERRIDE)) ^ port);
+	int gpio_shift = gpio_num;
+	if (gpio_port)
+		gpio_shift += MISC_REGISTERS_GPIO_PORT_SHIFT;
+
+	uint32_t gpio_mask = (1 << gpio_shift);
+	uint32_t gpio_reg;
+
+	if (gpio_num > MISC_REGISTERS_GPIO_3) {
+		PMD_DRV_LOG(NOTICE, "Invalid GPIO %d", gpio_num);
+		return -1;
+	}
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	/* read GPIO int */
+	gpio_reg = REG_RD(sc, MISC_REG_GPIO_INT);
+
+	switch (mode) {
+	case MISC_REGISTERS_GPIO_INT_OUTPUT_CLR:
+		/* clear SET and set CLR */
+		gpio_reg &= ~(gpio_mask << MISC_REGISTERS_GPIO_INT_SET_POS);
+		gpio_reg |= (gpio_mask << MISC_REGISTERS_GPIO_INT_CLR_POS);
+		break;
+
+	case MISC_REGISTERS_GPIO_INT_OUTPUT_SET:
+		/* clear CLR and set SET */
+		gpio_reg &= ~(gpio_mask << MISC_REGISTERS_GPIO_INT_CLR_POS);
+		gpio_reg |= (gpio_mask << MISC_REGISTERS_GPIO_INT_SET_POS);
+		break;
+
+	default:
+		break;
+	}
+
+	REG_WR(sc, MISC_REG_GPIO_INT, gpio_reg);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_GPIO);
+
+	return 0;
+}
+
+uint32_t
+elink_cb_gpio_read(struct bcm_softc * sc, uint16_t gpio_num, uint8_t port)
+{
+	return bcm_gpio_read(sc, gpio_num, port);
+}
+
+uint8_t elink_cb_gpio_write(struct bcm_softc * sc, uint16_t gpio_num, uint8_t mode,	/* 0=low 1=high */
+			    uint8_t port)
+{
+	return bcm_gpio_write(sc, gpio_num, mode, port);
+}
+
+uint8_t
+elink_cb_gpio_mult_write(struct bcm_softc * sc, uint8_t pins,
+			 uint8_t mode /* 0=low 1=high */ )
+{
+	return bcm_gpio_mult_write(sc, pins, mode);
+}
+
+uint8_t elink_cb_gpio_int_write(struct bcm_softc * sc, uint16_t gpio_num, uint8_t mode,	/* 0=low 1=high */
+				uint8_t port)
+{
+	return bcm_gpio_int_write(sc, gpio_num, mode, port);
+}
+
+void elink_cb_notify_link_changed(struct bcm_softc *sc)
+{
+	REG_WR(sc, (MISC_REG_AEU_GENERAL_ATTN_12 +
+		    (SC_FUNC(sc) * sizeof(uint32_t))), 1);
+}
+
+/* send the MCP a request, block until there is a reply */
+uint32_t
+elink_cb_fw_command(struct bcm_softc *sc, uint32_t command, uint32_t param)
+{
+	int mb_idx = SC_FW_MB_IDX(sc);
+	uint32_t seq;
+	uint32_t rc = 0;
+	uint32_t cnt = 1;
+	uint8_t delay = CHIP_REV_IS_SLOW(sc) ? 100 : 10;
+
+	seq = ++sc->fw_seq;
+	SHMEM_WR(sc, func_mb[mb_idx].drv_mb_param, param);
+	SHMEM_WR(sc, func_mb[mb_idx].drv_mb_header, (command | seq));
+
+	PMD_DRV_LOG(DEBUG,
+		    "wrote command 0x%08x to FW MB param 0x%08x",
+		    (command | seq), param);
+
+	/* Let the FW do it's magic. GIve it up to 5 seconds... */
+	do {
+		DELAY(delay * 1000);
+		rc = SHMEM_RD(sc, func_mb[mb_idx].fw_mb_header);
+	} while ((seq != (rc & FW_MSG_SEQ_NUMBER_MASK)) && (cnt++ < 500));
+
+	/* is this a reply to our command? */
+	if (seq == (rc & FW_MSG_SEQ_NUMBER_MASK)) {
+		rc &= FW_MSG_CODE_MASK;
+	} else {
+		/* Ruh-roh! */
+		PMD_DRV_LOG(NOTICE, "FW failed to respond!");
+		rc = 0;
+	}
+
+	return rc;
+}
+
+static uint32_t
+bcm_fw_command(struct bcm_softc *sc, uint32_t command, uint32_t param)
+{
+	return elink_cb_fw_command(sc, command, param);
+}
+
+static void
+__storm_memset_dma_mapping(struct bcm_softc *sc, uint32_t addr,
+			   phys_addr_t mapping)
+{
+	REG_WR(sc, addr, U64_LO(mapping));
+	REG_WR(sc, (addr + 4), U64_HI(mapping));
+}
+
+static void
+storm_memset_spq_addr(struct bcm_softc *sc, phys_addr_t mapping,
+		      uint16_t abs_fid)
+{
+	uint32_t addr = (XSEM_REG_FAST_MEMORY +
+			 XSTORM_SPQ_PAGE_BASE_OFFSET(abs_fid));
+	__storm_memset_dma_mapping(sc, addr, mapping);
+}
+
+static void
+storm_memset_vf_to_pf(struct bcm_softc *sc, uint16_t abs_fid, uint16_t pf_id)
+{
+	REG_WR8(sc, (BAR_XSTRORM_INTMEM + XSTORM_VF_TO_PF_OFFSET(abs_fid)),
+		pf_id);
+	REG_WR8(sc, (BAR_CSTRORM_INTMEM + CSTORM_VF_TO_PF_OFFSET(abs_fid)),
+		pf_id);
+	REG_WR8(sc, (BAR_TSTRORM_INTMEM + TSTORM_VF_TO_PF_OFFSET(abs_fid)),
+		pf_id);
+	REG_WR8(sc, (BAR_USTRORM_INTMEM + USTORM_VF_TO_PF_OFFSET(abs_fid)),
+		pf_id);
+}
+
+static void
+storm_memset_func_en(struct bcm_softc *sc, uint16_t abs_fid, uint8_t enable)
+{
+	REG_WR8(sc, (BAR_XSTRORM_INTMEM + XSTORM_FUNC_EN_OFFSET(abs_fid)),
+		enable);
+	REG_WR8(sc, (BAR_CSTRORM_INTMEM + CSTORM_FUNC_EN_OFFSET(abs_fid)),
+		enable);
+	REG_WR8(sc, (BAR_TSTRORM_INTMEM + TSTORM_FUNC_EN_OFFSET(abs_fid)),
+		enable);
+	REG_WR8(sc, (BAR_USTRORM_INTMEM + USTORM_FUNC_EN_OFFSET(abs_fid)),
+		enable);
+}
+
+static void
+storm_memset_eq_data(struct bcm_softc *sc, struct event_ring_data *eq_data,
+		     uint16_t pfid)
+{
+	uint32_t addr;
+	size_t size;
+
+	addr = (BAR_CSTRORM_INTMEM + CSTORM_EVENT_RING_DATA_OFFSET(pfid));
+	size = sizeof(struct event_ring_data);
+	ecore_storm_memset_struct(sc, addr, size, (uint32_t *) eq_data);
+}
+
+static void
+storm_memset_eq_prod(struct bcm_softc *sc, uint16_t eq_prod, uint16_t pfid)
+{
+	uint32_t addr = (BAR_CSTRORM_INTMEM +
+			 CSTORM_EVENT_RING_PROD_OFFSET(pfid));
+	REG_WR16(sc, addr, eq_prod);
+}
+
+/*
+ * Post a slowpath command.
+ *
+ * A slowpath command is used to propogate a configuration change through
+ * the controller in a controlled manner, allowing each STORM processor and
+ * other H/W blocks to phase in the change.  The commands sent on the
+ * slowpath are referred to as ramrods.  Depending on the ramrod used the
+ * completion of the ramrod will occur in different ways.  Here's a
+ * breakdown of ramrods and how they complete:
+ *
+ * RAMROD_CMD_ID_ETH_PORT_SETUP
+ *   Used to setup the leading connection on a port.  Completes on the
+ *   Receive Completion Queue (RCQ) of that port (typically fp[0]).
+ *
+ * RAMROD_CMD_ID_ETH_CLIENT_SETUP
+ *   Used to setup an additional connection on a port.  Completes on the
+ *   RCQ of the multi-queue/RSS connection being initialized.
+ *
+ * RAMROD_CMD_ID_ETH_STAT_QUERY
+ *   Used to force the storm processors to update the statistics database
+ *   in host memory.  This ramrod is send on the leading connection CID and
+ *   completes as an index increment of the CSTORM on the default status
+ *   block.
+ *
+ * RAMROD_CMD_ID_ETH_UPDATE
+ *   Used to update the state of the leading connection, usually to udpate
+ *   the RSS indirection table.  Completes on the RCQ of the leading
+ *   connection. (Not currently used under FreeBSD until OS support becomes
+ *   available.)
+ *
+ * RAMROD_CMD_ID_ETH_HALT
+ *   Used when tearing down a connection prior to driver unload.  Completes
+ *   on the RCQ of the multi-queue/RSS connection being torn down.  Don't
+ *   use this on the leading connection.
+ *
+ * RAMROD_CMD_ID_ETH_SET_MAC
+ *   Sets the Unicast/Broadcast/Multicast used by the port.  Completes on
+ *   the RCQ of the leading connection.
+ *
+ * RAMROD_CMD_ID_ETH_CFC_DEL
+ *   Used when tearing down a conneciton prior to driver unload.  Completes
+ *   on the RCQ of the leading connection (since the current connection
+ *   has been completely removed from controller memory).
+ *
+ * RAMROD_CMD_ID_ETH_PORT_DEL
+ *   Used to tear down the leading connection prior to driver unload,
+ *   typically fp[0].  Completes as an index increment of the CSTORM on the
+ *   default status block.
+ *
+ * RAMROD_CMD_ID_ETH_FORWARD_SETUP
+ *   Used for connection offload.  Completes on the RCQ of the multi-queue
+ *   RSS connection that is being offloaded.  (Not currently used under
+ *   FreeBSD.)
+ *
+ * There can only be one command pending per function.
+ *
+ * Returns:
+ *   0 = Success, !0 = Failure.
+ */
+
+/* must be called under the spq lock */
+static inline struct eth_spe *bcm_sp_get_next(struct bcm_softc *sc)
+{
+	struct eth_spe *next_spe = sc->spq_prod_bd;
+
+	if (sc->spq_prod_bd == sc->spq_last_bd) {
+		/* wrap back to the first eth_spq */
+		sc->spq_prod_bd = sc->spq;
+		sc->spq_prod_idx = 0;
+	} else {
+		sc->spq_prod_bd++;
+		sc->spq_prod_idx++;
+	}
+
+	return next_spe;
+}
+
+/* must be called under the spq lock */
+static void bcm_sp_prod_update(struct bcm_softc *sc)
+{
+	int func = SC_FUNC(sc);
+
+	/*
+	 * Make sure that BD data is updated before writing the producer.
+	 * BD data is written to the memory, the producer is read from the
+	 * memory, thus we need a full memory barrier to ensure the ordering.
+	 */
+	mb();
+
+	REG_WR16(sc, (BAR_XSTRORM_INTMEM + XSTORM_SPQ_PROD_OFFSET(func)),
+		 sc->spq_prod_idx);
+
+	mb();
+}
+
+/**
+ * bcm_is_contextless_ramrod - check if the current command ends on EQ
+ *
+ * @cmd:      command to check
+ * @cmd_type: command type
+ */
+static int bcm_is_contextless_ramrod(int cmd, int cmd_type)
+{
+	if ((cmd_type == NONE_CONNECTION_TYPE) ||
+	    (cmd == RAMROD_CMD_ID_ETH_FORWARD_SETUP) ||
+	    (cmd == RAMROD_CMD_ID_ETH_CLASSIFICATION_RULES) ||
+	    (cmd == RAMROD_CMD_ID_ETH_FILTER_RULES) ||
+	    (cmd == RAMROD_CMD_ID_ETH_MULTICAST_RULES) ||
+	    (cmd == RAMROD_CMD_ID_ETH_SET_MAC) ||
+	    (cmd == RAMROD_CMD_ID_ETH_RSS_UPDATE)) {
+		return TRUE;
+	} else {
+		return FALSE;
+	}
+}
+
+/**
+ * bcm_sp_post - place a single command on an SP ring
+ *
+ * @sc:         driver handle
+ * @command:    command to place (e.g. SETUP, FILTER_RULES, etc.)
+ * @cid:        SW CID the command is related to
+ * @data_hi:    command private data address (high 32 bits)
+ * @data_lo:    command private data address (low 32 bits)
+ * @cmd_type:   command type (e.g. NONE, ETH)
+ *
+ * SP data is handled as if it's always an address pair, thus data fields are
+ * not swapped to little endian in upper functions. Instead this function swaps
+ * data as if it's two uint32 fields.
+ */
+int
+bcm_sp_post(struct bcm_softc *sc, int command, int cid, uint32_t data_hi,
+	    uint32_t data_lo, int cmd_type)
+{
+	struct eth_spe *spe;
+	uint16_t type;
+	int common;
+
+	common = bcm_is_contextless_ramrod(command, cmd_type);
+
+	if (common) {
+		if (!atomic_load_acq_long(&sc->eq_spq_left)) {
+			PMD_DRV_LOG(INFO, "EQ ring is full!");
+			return -1;
+		}
+	} else {
+		if (!atomic_load_acq_long(&sc->cq_spq_left)) {
+			PMD_DRV_LOG(INFO, "SPQ ring is full!");
+			return -1;
+		}
+	}
+
+	spe = bcm_sp_get_next(sc);
+
+	/* CID needs port number to be encoded int it */
+	spe->hdr.conn_and_cmd_data =
+	    htole32((command << SPE_HDR_CMD_ID_SHIFT) | HW_CID(sc, cid));
+
+	type = (cmd_type << SPE_HDR_CONN_TYPE_SHIFT) & SPE_HDR_CONN_TYPE;
+
+	/* TBD: Check if it works for VFs */
+	type |= ((SC_FUNC(sc) << SPE_HDR_FUNCTION_ID_SHIFT) &
+		 SPE_HDR_FUNCTION_ID);
+
+	spe->hdr.type = htole16(type);
+
+	spe->data.update_data_addr.hi = htole32(data_hi);
+	spe->data.update_data_addr.lo = htole32(data_lo);
+
+	/*
+	 * It's ok if the actual decrement is issued towards the memory
+	 * somewhere between the lock and unlock. Thus no more explict
+	 * memory barrier is needed.
+	 */
+	if (common) {
+		atomic_subtract_acq_long(&sc->eq_spq_left, 1);
+	} else {
+		atomic_subtract_acq_long(&sc->cq_spq_left, 1);
+	}
+
+	PMD_DRV_LOG(DEBUG,
+		    "SPQE[%x] (%x:%x) (cmd, common?) (%d,%d) hw_cid %x"
+		    "data (%x:%x) type(0x%x) left (CQ, EQ) (%lx,%lx)",
+		    sc->spq_prod_idx,
+		    (uint32_t) U64_HI(sc->spq_dma.paddr),
+		    (uint32_t) (U64_LO(sc->spq_dma.paddr) +
+				(uint8_t *) sc->spq_prod_bd -
+				(uint8_t *) sc->spq), command, common,
+		    HW_CID(sc, cid), data_hi, data_lo, type,
+		    atomic_load_acq_long(&sc->cq_spq_left),
+		    atomic_load_acq_long(&sc->eq_spq_left));
+
+	bcm_sp_prod_update(sc);
+
+	return 0;
+}
+
+static void bcm_drv_pulse(struct bcm_softc *sc)
+{
+	SHMEM_WR(sc, func_mb[SC_FW_MB_IDX(sc)].drv_pulse_mb,
+		 sc->fw_drv_pulse_wr_seq);
+}
+
+static int bcm_tx_queue_has_work(const struct bcm_fastpath *fp)
+{
+	uint16_t hw_cons;
+	struct bcm_tx_queue *txq = fp->sc->tx_queues[fp->index];
+
+	if (unlikely(!txq)) {
+		PMD_TX_LOG(ERR, "ERROR: TX queue is NULL");
+		return 0;
+	}
+
+	mb();			/* status block fields can change */
+	hw_cons = le16toh(*fp->tx_cons_sb);
+	return (hw_cons != txq->tx_pkt_head);
+}
+
+static uint8_t bcm_has_tx_work(struct bcm_fastpath *fp)
+{
+	/* expand this for multi-cos if ever supported */
+	return bcm_tx_queue_has_work(fp);
+}
+
+static int bcm_has_rx_work(struct bcm_fastpath *fp)
+{
+	uint16_t rx_cq_cons_sb;
+	struct bcm_rx_queue *rxq;
+	rxq = fp->sc->rx_queues[fp->index];
+	if (unlikely(!rxq)) {
+		PMD_RX_LOG(ERR, "ERROR: RX queue is NULL");
+		return 0;
+	}
+
+	mb();			/* status block fields can change */
+	rx_cq_cons_sb = le16toh(*fp->rx_cq_cons_sb);
+	if (unlikely((rx_cq_cons_sb & MAX_RCQ_ENTRIES(rxq)) ==
+		     MAX_RCQ_ENTRIES(rxq)))
+		rx_cq_cons_sb++;
+	return (rxq->rx_cq_head != rx_cq_cons_sb);
+}
+
+static void
+bcm_sp_event(struct bcm_softc *sc, struct bcm_fastpath *fp,
+	     union eth_rx_cqe *rr_cqe)
+{
+#ifdef RTE_LIBRTE_BCM_DEBUG
+	int cid = SW_CID(rr_cqe->ramrod_cqe.conn_and_cmd_data);
+#endif
+	int command = CQE_CMD(rr_cqe->ramrod_cqe.conn_and_cmd_data);
+	enum ecore_queue_cmd drv_cmd = ECORE_Q_CMD_MAX;
+	struct ecore_queue_sp_obj *q_obj = &BCM_SP_OBJ(sc, fp).q_obj;
+
+	PMD_DRV_LOG(DEBUG,
+		    "fp=%d cid=%d got ramrod #%d state is %x type is %d",
+		    fp->index, cid, command, sc->state,
+		    rr_cqe->ramrod_cqe.ramrod_type);
+
+	switch (command) {
+	case (RAMROD_CMD_ID_ETH_CLIENT_UPDATE):
+		PMD_DRV_LOG(DEBUG, "got UPDATE ramrod. CID %d", cid);
+		drv_cmd = ECORE_Q_CMD_UPDATE;
+		break;
+
+	case (RAMROD_CMD_ID_ETH_CLIENT_SETUP):
+		PMD_DRV_LOG(DEBUG, "got MULTI[%d] setup ramrod", cid);
+		drv_cmd = ECORE_Q_CMD_SETUP;
+		break;
+
+	case (RAMROD_CMD_ID_ETH_TX_QUEUE_SETUP):
+		PMD_DRV_LOG(DEBUG, "got MULTI[%d] tx-only setup ramrod", cid);
+		drv_cmd = ECORE_Q_CMD_SETUP_TX_ONLY;
+		break;
+
+	case (RAMROD_CMD_ID_ETH_HALT):
+		PMD_DRV_LOG(DEBUG, "got MULTI[%d] halt ramrod", cid);
+		drv_cmd = ECORE_Q_CMD_HALT;
+		break;
+
+	case (RAMROD_CMD_ID_ETH_TERMINATE):
+		PMD_DRV_LOG(DEBUG, "got MULTI[%d] teminate ramrod", cid);
+		drv_cmd = ECORE_Q_CMD_TERMINATE;
+		break;
+
+	case (RAMROD_CMD_ID_ETH_EMPTY):
+		PMD_DRV_LOG(DEBUG, "got MULTI[%d] empty ramrod", cid);
+		drv_cmd = ECORE_Q_CMD_EMPTY;
+		break;
+
+	default:
+		PMD_DRV_LOG(DEBUG,
+			    "ERROR: unexpected MC reply (%d)"
+			    "on fp[%d]", command, fp->index);
+		return;
+	}
+
+	if ((drv_cmd != ECORE_Q_CMD_MAX) &&
+	    q_obj->complete_cmd(sc, q_obj, drv_cmd)) {
+		/*
+		 * q_obj->complete_cmd() failure means that this was
+		 * an unexpected completion.
+		 *
+		 * In this case we don't want to increase the sc->spq_left
+		 * because apparently we haven't sent this command the first
+		 * place.
+		 */
+		// rte_panic("Unexpected SP completion");
+		return;
+	}
+
+	atomic_add_acq_long(&sc->cq_spq_left, 1);
+
+	PMD_DRV_LOG(DEBUG, "sc->cq_spq_left 0x%lx",
+		    atomic_load_acq_long(&sc->cq_spq_left));
+}
+
+static uint8_t bcm_rxeof(struct bcm_softc *sc, struct bcm_fastpath *fp)
+{
+	struct bcm_rx_queue *rxq;
+	uint16_t bd_cons, bd_prod, bd_prod_fw, comp_ring_cons;
+	uint16_t hw_cq_cons, sw_cq_cons, sw_cq_prod;
+
+	rxq = sc->rx_queues[fp->index];
+	if (!rxq) {
+		PMD_RX_LOG(ERR, "RX queue %d is NULL", fp->index);
+		return 0;
+	}
+
+	/* CQ "next element" is of the size of the regular element */
+	hw_cq_cons = le16toh(*fp->rx_cq_cons_sb);
+	if (unlikely((hw_cq_cons & USABLE_RCQ_ENTRIES_PER_PAGE) ==
+		     USABLE_RCQ_ENTRIES_PER_PAGE)) {
+		hw_cq_cons++;
+	}
+
+	bd_cons = rxq->rx_bd_head;
+	bd_prod = rxq->rx_bd_tail;
+	bd_prod_fw = bd_prod;
+	sw_cq_cons = rxq->rx_cq_head;
+	sw_cq_prod = rxq->rx_cq_tail;
+
+	/*
+	 * Memory barrier necessary as speculative reads of the rx
+	 * buffer can be ahead of the index in the status block
+	 */
+	rmb();
+
+	while (sw_cq_cons != hw_cq_cons) {
+		union eth_rx_cqe *cqe;
+		struct eth_fast_path_rx_cqe *cqe_fp;
+		uint8_t cqe_fp_flags;
+		enum eth_rx_cqe_type cqe_fp_type;
+
+		comp_ring_cons = RCQ_ENTRY(sw_cq_cons, rxq);
+		bd_prod = RX_BD(bd_prod, rxq);
+		bd_cons = RX_BD(bd_cons, rxq);
+
+		cqe = &rxq->cq_ring[comp_ring_cons];
+		cqe_fp = &cqe->fast_path_cqe;
+		cqe_fp_flags = cqe_fp->type_error_flags;
+		cqe_fp_type = cqe_fp_flags & ETH_FAST_PATH_RX_CQE_TYPE;
+
+		/* is this a slowpath msg? */
+		if (CQE_TYPE_SLOW(cqe_fp_type)) {
+			bcm_sp_event(sc, fp, cqe);
+			goto next_cqe;
+		}
+
+		/* is this an error packet? */
+		if (unlikely(cqe_fp_flags &
+			     ETH_FAST_PATH_RX_CQE_PHY_DECODE_ERR_FLG)) {
+			PMD_RX_LOG(DEBUG, "flags 0x%x rx packet %u",
+				   cqe_fp_flags, sw_cq_cons);
+			goto next_rx;
+		}
+
+		PMD_RX_LOG(DEBUG, "Dropping fastpath called from attn poller!");
+
+next_rx:
+		bd_cons = NEXT_RX_BD(bd_cons);
+		bd_prod = NEXT_RX_BD(bd_prod);
+		bd_prod_fw = NEXT_RX_BD(bd_prod_fw);
+
+next_cqe:
+		sw_cq_prod = NEXT_RCQ_IDX(sw_cq_prod);
+		sw_cq_cons = NEXT_RCQ_IDX(sw_cq_cons);
+
+	}			/* while work to do */
+
+	rxq->rx_bd_head = bd_cons;
+	rxq->rx_bd_tail = bd_prod_fw;
+	rxq->rx_cq_head = sw_cq_cons;
+	rxq->rx_cq_tail = sw_cq_prod;
+
+	/* Update producers */
+	bcm_update_rx_prod(sc, fp, bd_prod_fw, sw_cq_prod);
+
+	return (sw_cq_cons != hw_cq_cons);
+}
+
+static uint16_t
+bcm_free_tx_pkt(__rte_unused struct bcm_fastpath *fp, struct bcm_tx_queue *txq,
+		uint16_t pkt_idx, uint16_t bd_idx)
+{
+	struct eth_tx_start_bd *tx_start_bd =
+	    &txq->tx_ring[TX_BD(bd_idx, txq)].start_bd;
+	uint16_t nbd = rte_le_to_cpu_16(tx_start_bd->nbd);
+	struct rte_mbuf *tx_mbuf = txq->sw_ring[TX_BD(pkt_idx, txq)];
+
+	if (likely(tx_mbuf != NULL)) {
+		rte_pktmbuf_free(tx_mbuf);
+	} else {
+		PMD_RX_LOG(ERR, "fp[%02d] lost mbuf %lu",
+			   fp->index, TX_BD(pkt_idx, txq));
+	}
+
+	txq->sw_ring[TX_BD(pkt_idx, txq)] = NULL;
+	txq->nb_tx_avail += nbd;
+
+	while (nbd--)
+		bd_idx = NEXT_TX_BD(bd_idx);
+
+	return bd_idx;
+}
+
+/* processes transmit completions */
+uint8_t bcm_txeof(__rte_unused struct bcm_softc * sc, struct bcm_fastpath * fp)
+{
+	uint16_t bd_cons, hw_cons, sw_cons;
+	__rte_unused uint16_t tx_bd_avail;
+
+	struct bcm_tx_queue *txq = fp->sc->tx_queues[fp->index];
+
+	if (unlikely(!txq)) {
+		PMD_TX_LOG(ERR, "ERROR: TX queue is NULL");
+		return 0;
+	}
+
+	bd_cons = txq->tx_bd_head;
+	hw_cons = rte_le_to_cpu_16(*fp->tx_cons_sb);
+	sw_cons = txq->tx_pkt_head;
+
+	while (sw_cons != hw_cons) {
+		bd_cons = bcm_free_tx_pkt(fp, txq, sw_cons, bd_cons);
+		sw_cons++;
+	}
+
+	txq->tx_pkt_head = sw_cons;
+	txq->tx_bd_head = bd_cons;
+
+	tx_bd_avail = txq->nb_tx_avail;
+
+	PMD_TX_LOG(DEBUG, "fp[%02d] avail=%u cons_sb=%u, "
+		   "pkt_head=%u pkt_tail=%u bd_head=%u bd_tail=%u",
+		   fp->index, tx_bd_avail, hw_cons,
+		   txq->tx_pkt_head, txq->tx_pkt_tail,
+		   txq->tx_bd_head, txq->tx_bd_tail);
+	return TRUE;
+}
+
+static void bcm_drain_tx_queues(struct bcm_softc *sc)
+{
+	struct bcm_fastpath *fp;
+	int i, count;
+
+	/* wait until all TX fastpath tasks have completed */
+	for (i = 0; i < sc->num_queues; i++) {
+		fp = &sc->fp[i];
+
+		count = 1000;
+
+		while (bcm_has_tx_work(fp)) {
+			bcm_txeof(sc, fp);
+
+			if (count == 0) {
+				PMD_TX_LOG(ERR,
+					   "Timeout waiting for fp[%d] "
+					   "transmits to complete!", i);
+				rte_panic("tx drain failure");
+				return;
+			}
+
+			count--;
+			DELAY(1000);
+			rmb();
+		}
+	}
+
+	return;
+}
+
+static int
+bcm_del_all_macs(struct bcm_softc *sc, struct ecore_vlan_mac_obj *mac_obj,
+		 int mac_type, uint8_t wait_for_comp)
+{
+	unsigned long ramrod_flags = 0, vlan_mac_flags = 0;
+	int rc;
+
+	/* wait for completion of requested */
+	if (wait_for_comp) {
+		bcm_set_bit(RAMROD_COMP_WAIT, &ramrod_flags);
+	}
+
+	/* Set the mac type of addresses we want to clear */
+	bcm_set_bit(mac_type, &vlan_mac_flags);
+
+	rc = mac_obj->delete_all(sc, mac_obj, &vlan_mac_flags, &ramrod_flags);
+	if (rc < 0)
+		PMD_DRV_LOG(ERR, "Failed to delete MACs (%d)", rc);
+
+	return rc;
+}
+
+int
+bcm_fill_accept_flags(struct bcm_softc *sc, uint32_t rx_mode,
+		      unsigned long *rx_accept_flags,
+		      unsigned long *tx_accept_flags)
+{
+	/* Clear the flags first */
+	*rx_accept_flags = 0;
+	*tx_accept_flags = 0;
+
+	switch (rx_mode) {
+	case BCM_RX_MODE_NONE:
+		/*
+		 * 'drop all' supersedes any accept flags that may have been
+		 * passed to the function.
+		 */
+		break;
+
+	case BCM_RX_MODE_NORMAL:
+		bcm_set_bit(ECORE_ACCEPT_UNICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_MULTICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, rx_accept_flags);
+
+		/* internal switching mode */
+		bcm_set_bit(ECORE_ACCEPT_UNICAST, tx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_MULTICAST, tx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, tx_accept_flags);
+
+		break;
+
+	case BCM_RX_MODE_ALLMULTI:
+		bcm_set_bit(ECORE_ACCEPT_UNICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_ALL_MULTICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, rx_accept_flags);
+
+		/* internal switching mode */
+		bcm_set_bit(ECORE_ACCEPT_UNICAST, tx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_ALL_MULTICAST, tx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, tx_accept_flags);
+
+		break;
+
+	case BCM_RX_MODE_PROMISC:
+		/*
+		 * According to deffinition of SI mode, iface in promisc mode
+		 * should receive matched and unmatched (in resolution of port)
+		 * unicast packets.
+		 */
+		bcm_set_bit(ECORE_ACCEPT_UNMATCHED, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_UNICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_ALL_MULTICAST, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, rx_accept_flags);
+
+		/* internal switching mode */
+		bcm_set_bit(ECORE_ACCEPT_ALL_MULTICAST, tx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_BROADCAST, tx_accept_flags);
+
+		if (IS_MF_SI(sc)) {
+			bcm_set_bit(ECORE_ACCEPT_ALL_UNICAST, tx_accept_flags);
+		} else {
+			bcm_set_bit(ECORE_ACCEPT_UNICAST, tx_accept_flags);
+		}
+
+		break;
+
+	default:
+		PMD_RX_LOG(ERR, "Unknown rx_mode (%d)", rx_mode);
+		return -1;
+	}
+
+	/* Set ACCEPT_ANY_VLAN as we do not enable filtering by VLAN */
+	if (rx_mode != BCM_RX_MODE_NONE) {
+		bcm_set_bit(ECORE_ACCEPT_ANY_VLAN, rx_accept_flags);
+		bcm_set_bit(ECORE_ACCEPT_ANY_VLAN, tx_accept_flags);
+	}
+
+	return 0;
+}
+
+static int
+bcm_set_q_rx_mode(struct bcm_softc *sc, uint8_t cl_id,
+		  unsigned long rx_mode_flags,
+		  unsigned long rx_accept_flags,
+		  unsigned long tx_accept_flags, unsigned long ramrod_flags)
+{
+	struct ecore_rx_mode_ramrod_params ramrod_param;
+	int rc;
+
+	memset(&ramrod_param, 0, sizeof(ramrod_param));
+
+	/* Prepare ramrod parameters */
+	ramrod_param.cid = 0;
+	ramrod_param.cl_id = cl_id;
+	ramrod_param.rx_mode_obj = &sc->rx_mode_obj;
+	ramrod_param.func_id = SC_FUNC(sc);
+
+	ramrod_param.pstate = &sc->sp_state;
+	ramrod_param.state = ECORE_FILTER_RX_MODE_PENDING;
+
+	ramrod_param.rdata = BCM_SP(sc, rx_mode_rdata);
+	ramrod_param.rdata_mapping =
+	    (phys_addr_t) ((void *)BCM_SP_MAPPING(sc, rx_mode_rdata)),
+	    bcm_set_bit(ECORE_FILTER_RX_MODE_PENDING, &sc->sp_state);
+
+	ramrod_param.ramrod_flags = ramrod_flags;
+	ramrod_param.rx_mode_flags = rx_mode_flags;
+
+	ramrod_param.rx_accept_flags = rx_accept_flags;
+	ramrod_param.tx_accept_flags = tx_accept_flags;
+
+	rc = ecore_config_rx_mode(sc, &ramrod_param);
+	if (rc < 0) {
+		PMD_RX_LOG(ERR, "Set rx_mode %d failed", sc->rx_mode);
+		return rc;
+	}
+
+	return 0;
+}
+
+int bcm_set_storm_rx_mode(struct bcm_softc *sc)
+{
+	unsigned long rx_mode_flags = 0, ramrod_flags = 0;
+	unsigned long rx_accept_flags = 0, tx_accept_flags = 0;
+	int rc;
+
+	rc = bcm_fill_accept_flags(sc, sc->rx_mode, &rx_accept_flags,
+				   &tx_accept_flags);
+	if (rc) {
+		return rc;
+	}
+
+	bcm_set_bit(RAMROD_RX, &ramrod_flags);
+	bcm_set_bit(RAMROD_TX, &ramrod_flags);
+	bcm_set_bit(RAMROD_COMP_WAIT, &ramrod_flags);
+
+	return bcm_set_q_rx_mode(sc, sc->fp[0].cl_id, rx_mode_flags,
+				 rx_accept_flags, tx_accept_flags,
+				 ramrod_flags);
+}
+
+/* returns the "mcp load_code" according to global load_count array */
+static int bcm_nic_load_no_mcp(struct bcm_softc *sc)
+{
+	int path = SC_PATH(sc);
+	int port = SC_PORT(sc);
+
+	PMD_DRV_LOG(INFO, "NO MCP - load counts[%d]      %d, %d, %d",
+		    path, load_count[path][0], load_count[path][1],
+		    load_count[path][2]);
+
+	load_count[path][0]++;
+	load_count[path][1 + port]++;
+	PMD_DRV_LOG(INFO, "NO MCP - new load counts[%d]  %d, %d, %d",
+		    path, load_count[path][0], load_count[path][1],
+		    load_count[path][2]);
+	if (load_count[path][0] == 1)
+		return FW_MSG_CODE_DRV_LOAD_COMMON;
+	else if (load_count[path][1 + port] == 1)
+		return FW_MSG_CODE_DRV_LOAD_PORT;
+	else
+		return FW_MSG_CODE_DRV_LOAD_FUNCTION;
+}
+
+/* returns the "mcp load_code" according to global load_count array */
+static int bcm_nic_unload_no_mcp(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	int path = SC_PATH(sc);
+
+	PMD_DRV_LOG(INFO, "NO MCP - load counts[%d]      %d, %d, %d",
+		    path, load_count[path][0], load_count[path][1],
+		    load_count[path][2]);
+	load_count[path][0]--;
+	load_count[path][1 + port]--;
+	PMD_DRV_LOG(INFO, "NO MCP - new load counts[%d]  %d, %d, %d",
+		    path, load_count[path][0], load_count[path][1],
+		    load_count[path][2]);
+	if (load_count[path][0] == 0) {
+		return FW_MSG_CODE_DRV_UNLOAD_COMMON;
+	} else if (load_count[path][1 + port] == 0) {
+		return FW_MSG_CODE_DRV_UNLOAD_PORT;
+	} else {
+		return FW_MSG_CODE_DRV_UNLOAD_FUNCTION;
+	}
+}
+
+/* request unload mode from the MCP: COMMON, PORT or FUNCTION */
+static uint32_t bcm_send_unload_req(struct bcm_softc *sc, int unload_mode)
+{
+	uint32_t reset_code = 0;
+
+	/* Select the UNLOAD request mode */
+	if (unload_mode == UNLOAD_NORMAL) {
+		reset_code = DRV_MSG_CODE_UNLOAD_REQ_WOL_DIS;
+	} else {
+		reset_code = DRV_MSG_CODE_UNLOAD_REQ_WOL_DIS;
+	}
+
+	/* Send the request to the MCP */
+	if (!BCM_NOMCP(sc)) {
+		reset_code = bcm_fw_command(sc, reset_code, 0);
+	} else {
+		reset_code = bcm_nic_unload_no_mcp(sc);
+	}
+
+	return reset_code;
+}
+
+/* send UNLOAD_DONE command to the MCP */
+static void bcm_send_unload_done(struct bcm_softc *sc, uint8_t keep_link)
+{
+	uint32_t reset_param =
+	    keep_link ? DRV_MSG_CODE_UNLOAD_SKIP_LINK_RESET : 0;
+
+	/* Report UNLOAD_DONE to MCP */
+	if (!BCM_NOMCP(sc)) {
+		bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_DONE, reset_param);
+	}
+}
+
+static int bcm_func_wait_started(struct bcm_softc *sc)
+{
+	int tout = 50;
+
+	if (!sc->port.pmf) {
+		return 0;
+	}
+
+	/*
+	 * (assumption: No Attention from MCP at this stage)
+	 * PMF probably in the middle of TX disable/enable transaction
+	 * 1. Sync IRS for default SB
+	 * 2. Sync SP queue - this guarantees us that attention handling started
+	 * 3. Wait, that TX disable/enable transaction completes
+	 *
+	 * 1+2 guarantee that if DCBX attention was scheduled it already changed
+	 * pending bit of transaction from STARTED-->TX_STOPPED, if we already
+	 * received completion for the transaction the state is TX_STOPPED.
+	 * State will return to STARTED after completion of TX_STOPPED-->STARTED
+	 * transaction.
+	 */
+
+	while (ecore_func_get_state(sc, &sc->func_obj) !=
+	       ECORE_F_STATE_STARTED && tout--) {
+		DELAY(20000);
+	}
+
+	if (ecore_func_get_state(sc, &sc->func_obj) != ECORE_F_STATE_STARTED) {
+		/*
+		 * Failed to complete the transaction in a "good way"
+		 * Force both transactions with CLR bit.
+		 */
+		struct ecore_func_state_params func_params = { NULL };
+
+		PMD_DRV_LOG(NOTICE, "Unexpected function state! "
+			    "Forcing STARTED-->TX_STOPPED-->STARTED");
+
+		func_params.f_obj = &sc->func_obj;
+		bcm_set_bit(RAMROD_DRV_CLR_ONLY, &func_params.ramrod_flags);
+
+		/* STARTED-->TX_STOPPED */
+		func_params.cmd = ECORE_F_CMD_TX_STOP;
+		ecore_func_state_change(sc, &func_params);
+
+		/* TX_STOPPED-->STARTED */
+		func_params.cmd = ECORE_F_CMD_TX_START;
+		return ecore_func_state_change(sc, &func_params);
+	}
+
+	return 0;
+}
+
+static int bcm_stop_queue(struct bcm_softc *sc, int index)
+{
+	struct bcm_fastpath *fp = &sc->fp[index];
+	struct ecore_queue_state_params q_params = { NULL };
+	int rc;
+
+	PMD_DRV_LOG(DEBUG, "stopping queue %d cid %d", index, fp->index);
+
+	q_params.q_obj = &sc->sp_objs[fp->index].q_obj;
+	/* We want to wait for completion in this context */
+	bcm_set_bit(RAMROD_COMP_WAIT, &q_params.ramrod_flags);
+
+	/* Stop the primary connection: */
+
+	/* ...halt the connection */
+	q_params.cmd = ECORE_Q_CMD_HALT;
+	rc = ecore_queue_state_change(sc, &q_params);
+	if (rc) {
+		return rc;
+	}
+
+	/* ...terminate the connection */
+	q_params.cmd = ECORE_Q_CMD_TERMINATE;
+	memset(&q_params.params.terminate, 0,
+	       sizeof(q_params.params.terminate));
+	q_params.params.terminate.cid_index = FIRST_TX_COS_INDEX;
+	rc = ecore_queue_state_change(sc, &q_params);
+	if (rc) {
+		return rc;
+	}
+
+	/* ...delete cfc entry */
+	q_params.cmd = ECORE_Q_CMD_CFC_DEL;
+	memset(&q_params.params.cfc_del, 0, sizeof(q_params.params.cfc_del));
+	q_params.params.cfc_del.cid_index = FIRST_TX_COS_INDEX;
+	return ecore_queue_state_change(sc, &q_params);
+}
+
+/* wait for the outstanding SP commands */
+static uint8_t bcm_wait_sp_comp(struct bcm_softc *sc, unsigned long mask)
+{
+	unsigned long tmp;
+	int tout = 5000;	/* wait for 5 secs tops */
+
+	while (tout--) {
+		mb();
+		if (!(atomic_load_acq_long(&sc->sp_state) & mask)) {
+			return TRUE;
+		}
+
+		DELAY(1000);
+	}
+
+	mb();
+
+	tmp = atomic_load_acq_long(&sc->sp_state);
+	if (tmp & mask) {
+		PMD_DRV_LOG(INFO, "Filtering completion timed out: "
+			    "sp_state 0x%lx, mask 0x%lx", tmp, mask);
+		return FALSE;
+	}
+
+	return FALSE;
+}
+
+static int bcm_func_stop(struct bcm_softc *sc)
+{
+	struct ecore_func_state_params func_params = { NULL };
+	int rc;
+
+	/* prepare parameters for function state transitions */
+	bcm_set_bit(RAMROD_COMP_WAIT, &func_params.ramrod_flags);
+	func_params.f_obj = &sc->func_obj;
+	func_params.cmd = ECORE_F_CMD_STOP;
+
+	/*
+	 * Try to stop the function the 'good way'. If it fails (in case
+	 * of a parity error during bcm_chip_cleanup()) and we are
+	 * not in a debug mode, perform a state transaction in order to
+	 * enable further HW_RESET transaction.
+	 */
+	rc = ecore_func_state_change(sc, &func_params);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "FUNC_STOP ramrod failed. "
+			    "Running a dry transaction");
+		bcm_set_bit(RAMROD_DRV_CLR_ONLY, &func_params.ramrod_flags);
+		return ecore_func_state_change(sc, &func_params);
+	}
+
+	return 0;
+}
+
+static int bcm_reset_hw(struct bcm_softc *sc, uint32_t load_code)
+{
+	struct ecore_func_state_params func_params = { NULL };
+
+	/* Prepare parameters for function state transitions */
+	bcm_set_bit(RAMROD_COMP_WAIT, &func_params.ramrod_flags);
+
+	func_params.f_obj = &sc->func_obj;
+	func_params.cmd = ECORE_F_CMD_HW_RESET;
+
+	func_params.params.hw_init.load_phase = load_code;
+
+	return ecore_func_state_change(sc, &func_params);
+}
+
+static void bcm_int_disable_sync(struct bcm_softc *sc, int disable_hw)
+{
+	if (disable_hw) {
+		/* prevent the HW from sending interrupts */
+		bcm_int_disable(sc);
+	}
+}
+
+static void
+bcm_chip_cleanup(struct bcm_softc *sc, uint32_t unload_mode, uint8_t keep_link)
+{
+	int port = SC_PORT(sc);
+	struct ecore_mcast_ramrod_params rparam = { NULL };
+	uint32_t reset_code;
+	int i, rc = 0;
+
+	bcm_drain_tx_queues(sc);
+
+	/* give HW time to discard old tx messages */
+	DELAY(1000);
+
+	/* Clean all ETH MACs */
+	rc = bcm_del_all_macs(sc, &sc->sp_objs[0].mac_obj, ECORE_ETH_MAC,
+			      FALSE);
+	if (rc < 0) {
+		PMD_DRV_LOG(NOTICE, "Failed to delete all ETH MACs (%d)", rc);
+	}
+
+	/* Clean up UC list  */
+	rc = bcm_del_all_macs(sc, &sc->sp_objs[0].mac_obj, ECORE_UC_LIST_MAC,
+			      TRUE);
+	if (rc < 0) {
+		PMD_DRV_LOG(NOTICE, "Failed to delete UC MACs list (%d)", rc);
+	}
+
+	/* Disable LLH */
+	REG_WR(sc, NIG_REG_LLH0_FUNC_EN + port * 8, 0);
+
+	/* Set "drop all" to stop Rx */
+
+	/*
+	 * We need to take the if_maddr_lock() here in order to prevent
+	 * a race between the completion code and this code.
+	 */
+
+	if (bcm_test_bit(ECORE_FILTER_RX_MODE_PENDING, &sc->sp_state)) {
+		bcm_set_bit(ECORE_FILTER_RX_MODE_SCHED, &sc->sp_state);
+	} else {
+		bcm_set_storm_rx_mode(sc);
+	}
+
+	/* Clean up multicast configuration */
+	rparam.mcast_obj = &sc->mcast_obj;
+	rc = ecore_config_mcast(sc, &rparam, ECORE_MCAST_CMD_DEL);
+	if (rc < 0) {
+		PMD_DRV_LOG(NOTICE,
+			    "Failed to send DEL MCAST command (%d)", rc);
+	}
+
+	/*
+	 * Send the UNLOAD_REQUEST to the MCP. This will return if
+	 * this function should perform FUNCTION, PORT, or COMMON HW
+	 * reset.
+	 */
+	reset_code = bcm_send_unload_req(sc, unload_mode);
+
+	/*
+	 * (assumption: No Attention from MCP at this stage)
+	 * PMF probably in the middle of TX disable/enable transaction
+	 */
+	rc = bcm_func_wait_started(sc);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "bcm_func_wait_started failed");
+	}
+
+	/*
+	 * Close multi and leading connections
+	 * Completions for ramrods are collected in a synchronous way
+	 */
+	for (i = 0; i < sc->num_queues; i++) {
+		if (bcm_stop_queue(sc, i)) {
+			goto unload_error;
+		}
+	}
+
+	/*
+	 * If SP settings didn't get completed so far - something
+	 * very wrong has happen.
+	 */
+	if (!bcm_wait_sp_comp(sc, ~0x0UL)) {
+		PMD_DRV_LOG(NOTICE, "Common slow path ramrods got stuck!");
+	}
+
+unload_error:
+
+	rc = bcm_func_stop(sc);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Function stop failed!");
+	}
+
+	/* disable HW interrupts */
+	bcm_int_disable_sync(sc, TRUE);
+
+	/* Reset the chip */
+	rc = bcm_reset_hw(sc, reset_code);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Hardware reset failed");
+	}
+
+	/* Report UNLOAD_DONE to MCP */
+	bcm_send_unload_done(sc, keep_link);
+}
+
+static void bcm_disable_close_the_gate(struct bcm_softc *sc)
+{
+	uint32_t val;
+
+	PMD_DRV_LOG(DEBUG, "Disabling 'close the gates'");
+
+	val = REG_RD(sc, MISC_REG_AEU_GENERAL_MASK);
+	val &= ~(MISC_AEU_GENERAL_MASK_REG_AEU_PXP_CLOSE_MASK |
+		 MISC_AEU_GENERAL_MASK_REG_AEU_NIG_CLOSE_MASK);
+	REG_WR(sc, MISC_REG_AEU_GENERAL_MASK, val);
+}
+
+/*
+ * Cleans the object that have internal lists without sending
+ * ramrods. Should be run when interrutps are disabled.
+ */
+static void bcm_squeeze_objects(struct bcm_softc *sc)
+{
+	unsigned long ramrod_flags = 0, vlan_mac_flags = 0;
+	struct ecore_mcast_ramrod_params rparam = { NULL };
+	struct ecore_vlan_mac_obj *mac_obj = &sc->sp_objs->mac_obj;
+	int rc;
+
+	/* Cleanup MACs' object first... */
+
+	/* Wait for completion of requested */
+	bcm_set_bit(RAMROD_COMP_WAIT, &ramrod_flags);
+	/* Perform a dry cleanup */
+	bcm_set_bit(RAMROD_DRV_CLR_ONLY, &ramrod_flags);
+
+	/* Clean ETH primary MAC */
+	bcm_set_bit(ECORE_ETH_MAC, &vlan_mac_flags);
+	rc = mac_obj->delete_all(sc, &sc->sp_objs->mac_obj, &vlan_mac_flags,
+				 &ramrod_flags);
+	if (rc != 0) {
+		PMD_DRV_LOG(NOTICE, "Failed to clean ETH MACs (%d)", rc);
+	}
+
+	/* Cleanup UC list */
+	vlan_mac_flags = 0;
+	bcm_set_bit(ECORE_UC_LIST_MAC, &vlan_mac_flags);
+	rc = mac_obj->delete_all(sc, mac_obj, &vlan_mac_flags, &ramrod_flags);
+	if (rc != 0) {
+		PMD_DRV_LOG(NOTICE, "Failed to clean UC list MACs (%d)", rc);
+	}
+
+	/* Now clean mcast object... */
+
+	rparam.mcast_obj = &sc->mcast_obj;
+	bcm_set_bit(RAMROD_DRV_CLR_ONLY, &rparam.ramrod_flags);
+
+	/* Add a DEL command... */
+	rc = ecore_config_mcast(sc, &rparam, ECORE_MCAST_CMD_DEL);
+	if (rc < 0) {
+		PMD_DRV_LOG(NOTICE,
+			    "Failed to send DEL MCAST command (%d)", rc);
+	}
+
+	/* now wait until all pending commands are cleared */
+
+	rc = ecore_config_mcast(sc, &rparam, ECORE_MCAST_CMD_CONT);
+	while (rc != 0) {
+		if (rc < 0) {
+			PMD_DRV_LOG(NOTICE,
+				    "Failed to clean MCAST object (%d)", rc);
+			return;
+		}
+
+		rc = ecore_config_mcast(sc, &rparam, ECORE_MCAST_CMD_CONT);
+	}
+}
+
+/* stop the controller */
+__attribute__ ((noinline))
+int
+bcm_nic_unload(struct bcm_softc *sc, uint32_t unload_mode, uint8_t keep_link)
+{
+	uint8_t global = FALSE;
+	uint32_t val;
+
+	PMD_DRV_LOG(DEBUG, "Starting NIC unload...");
+
+	/* stop the periodic callout */
+	bcm_periodic_stop(sc);
+
+	/* mark driver as unloaded in shmem2 */
+	if (IS_PF(sc) && SHMEM2_HAS(sc, drv_capabilities_flag)) {
+		val = SHMEM2_RD(sc, drv_capabilities_flag[SC_FW_MB_IDX(sc)]);
+		SHMEM2_WR(sc, drv_capabilities_flag[SC_FW_MB_IDX(sc)],
+			  val & ~DRV_FLAGS_CAPABILITIES_LOADED_L2);
+	}
+
+	if (IS_PF(sc) && sc->recovery_state != BCM_RECOVERY_DONE &&
+	    (sc->state == BCM_STATE_CLOSED || sc->state == BCM_STATE_ERROR)) {
+		/*
+		 * We can get here if the driver has been unloaded
+		 * during parity error recovery and is either waiting for a
+		 * leader to complete or for other functions to unload and
+		 * then ifconfig down has been issued. In this case we want to
+		 * unload and let other functions to complete a recovery
+		 * process.
+		 */
+		sc->recovery_state = BCM_RECOVERY_DONE;
+		sc->is_leader = 0;
+		bcm_release_leader_lock(sc);
+		mb();
+
+		PMD_DRV_LOG(NOTICE, "Can't unload in closed or error state");
+		return -1;
+	}
+
+	/*
+	 * Nothing to do during unload if previous bcm_nic_load()
+	 * did not completed succesfully - all resourses are released.
+	 */
+	if ((sc->state == BCM_STATE_CLOSED) || (sc->state == BCM_STATE_ERROR)) {
+		return 0;
+	}
+
+	sc->state = BCM_STATE_CLOSING_WAITING_HALT;
+	mb();
+
+	sc->rx_mode = BCM_RX_MODE_NONE;
+	bcm_set_rx_mode(sc);
+	mb();
+
+	if (IS_PF(sc)) {
+		/* set ALWAYS_ALIVE bit in shmem */
+		sc->fw_drv_pulse_wr_seq |= DRV_PULSE_ALWAYS_ALIVE;
+
+		bcm_drv_pulse(sc);
+
+		bcm_stats_handle(sc, STATS_EVENT_STOP);
+		bcm_save_statistics(sc);
+	}
+
+	/* wait till consumers catch up with producers in all queues */
+	bcm_drain_tx_queues(sc);
+
+	/* if VF indicate to PF this function is going down (PF will delete sp
+	 * elements and clear initializations
+	 */
+	if (IS_VF(sc)) {
+		bcm_vf_unload(sc);
+	} else if (unload_mode != UNLOAD_RECOVERY) {
+		/* if this is a normal/close unload need to clean up chip */
+		bcm_chip_cleanup(sc, unload_mode, keep_link);
+	} else {
+		/* Send the UNLOAD_REQUEST to the MCP */
+		bcm_send_unload_req(sc, unload_mode);
+
+		/*
+		 * Prevent transactions to host from the functions on the
+		 * engine that doesn't reset global blocks in case of global
+		 * attention once gloabl blocks are reset and gates are opened
+		 * (the engine which leader will perform the recovery
+		 * last).
+		 */
+		if (!CHIP_IS_E1x(sc)) {
+			bcm_pf_disable(sc);
+		}
+
+		/* disable HW interrupts */
+		bcm_int_disable_sync(sc, TRUE);
+
+		/* Report UNLOAD_DONE to MCP */
+		bcm_send_unload_done(sc, FALSE);
+	}
+
+	/*
+	 * At this stage no more interrupts will arrive so we may safely clean
+	 * the queue'able objects here in case they failed to get cleaned so far.
+	 */
+	if (IS_PF(sc)) {
+		bcm_squeeze_objects(sc);
+	}
+
+	/* There should be no more pending SP commands at this stage */
+	sc->sp_state = 0;
+
+	sc->port.pmf = 0;
+
+	if (IS_PF(sc)) {
+		bcm_free_mem(sc);
+	}
+
+	bcm_free_fw_stats_mem(sc);
+
+	sc->state = BCM_STATE_CLOSED;
+
+	/*
+	 * Check if there are pending parity attentions. If there are - set
+	 * RECOVERY_IN_PROGRESS.
+	 */
+	if (IS_PF(sc) && bcm_chk_parity_attn(sc, &global, FALSE)) {
+		bcm_set_reset_in_progress(sc);
+
+		/* Set RESET_IS_GLOBAL if needed */
+		if (global) {
+			bcm_set_reset_global(sc);
+		}
+	}
+
+	/*
+	 * The last driver must disable a "close the gate" if there is no
+	 * parity attention or "process kill" pending.
+	 */
+	if (IS_PF(sc) && !bcm_clear_pf_load(sc) &&
+	    bcm_reset_is_done(sc, SC_PATH(sc))) {
+		bcm_disable_close_the_gate(sc);
+	}
+
+	PMD_DRV_LOG(DEBUG, "Ended NIC unload");
+
+	return 0;
+}
+
+/*
+ * Encapsulte an mbuf cluster into the tx bd chain and makes the memory
+ * visible to the controller.
+ *
+ * If an mbuf is submitted to this routine and cannot be given to the
+ * controller (e.g. it has too many fragments) then the function may free
+ * the mbuf and return to the caller.
+ *
+ * Returns:
+ *   0 = Success, !0 = Failure
+ *   Note the side effect that an mbuf may be freed if it causes a problem.
+ */
+int bcm_tx_encap(struct bcm_tx_queue *txq, struct rte_mbuf **m_head, int m_pkts)
+{
+	struct rte_mbuf *m0;
+	struct eth_tx_start_bd *tx_start_bd;
+	uint16_t bd_prod, pkt_prod;
+	int m_tx;
+	struct bcm_softc *sc;
+	uint32_t nbds = 0;
+	struct bcm_fastpath *fp;
+
+	sc = txq->sc;
+	fp = &sc->fp[txq->queue_id];
+
+	bd_prod = txq->tx_bd_tail;
+	pkt_prod = txq->tx_pkt_tail;
+
+	for (m_tx = 0; m_tx < m_pkts; m_tx++) {
+
+		m0 = *m_head++;
+
+		if (unlikely(txq->nb_tx_avail < 3)) {
+			PMD_TX_LOG(ERR, "no enough bds %d/%d",
+				   bd_prod, txq->nb_tx_avail);
+			return -ENOMEM;
+		}
+
+		txq->sw_ring[TX_BD(pkt_prod, txq)] = m0;
+
+		tx_start_bd = &txq->tx_ring[TX_BD(bd_prod, txq)].start_bd;
+
+		tx_start_bd->addr =
+		    rte_cpu_to_le_64(RTE_MBUF_DATA_DMA_ADDR(m0));
+		tx_start_bd->nbytes = rte_cpu_to_le_16(m0->data_len);
+		tx_start_bd->bd_flags.as_bitfield = ETH_TX_BD_FLAGS_START_BD;
+		tx_start_bd->general_data =
+		    (1 << ETH_TX_START_BD_HDR_NBDS_SHIFT);
+
+		tx_start_bd->nbd = rte_cpu_to_le_16(2);
+
+		if (m0->ol_flags & PKT_TX_VLAN_PKT) {
+			tx_start_bd->vlan_or_ethertype =
+			    rte_cpu_to_le_16(m0->vlan_tci);
+			tx_start_bd->bd_flags.as_bitfield |=
+			    (X_ETH_OUTBAND_VLAN <<
+			     ETH_TX_BD_FLAGS_VLAN_MODE_SHIFT);
+		} else {
+			if (IS_PF(sc))
+				tx_start_bd->vlan_or_ethertype =
+				    rte_cpu_to_le_16(pkt_prod);
+			else {
+				struct ether_hdr *eh
+				    = rte_pktmbuf_mtod(m0, struct ether_hdr *);
+
+				tx_start_bd->vlan_or_ethertype = eh->ether_type;
+			}
+		}
+
+		bd_prod = NEXT_TX_BD(bd_prod);
+		if (IS_VF(sc)) {
+			struct eth_tx_parse_bd_e2 *tx_parse_bd;
+			uint8_t *data = rte_pktmbuf_mtod(m0, uint8_t *);
+
+			tx_parse_bd =
+			    &txq->tx_ring[TX_BD(bd_prod, txq)].parse_bd_e2;
+			tx_parse_bd->parsing_data =
+			    (1 << ETH_TX_PARSE_BD_E2_ETH_ADDR_TYPE_SHIFT);
+
+			rte_memcpy(&tx_parse_bd->data.mac_addr.dst_hi,
+				   &data[0], 2);
+			rte_memcpy(&tx_parse_bd->data.mac_addr.dst_mid,
+				   &data[2], 2);
+			rte_memcpy(&tx_parse_bd->data.mac_addr.dst_lo,
+				   &data[4], 2);
+			rte_memcpy(&tx_parse_bd->data.mac_addr.src_hi,
+				   &data[6], 2);
+			rte_memcpy(&tx_parse_bd->data.mac_addr.src_mid,
+				   &data[8], 2);
+			rte_memcpy(&tx_parse_bd->data.mac_addr.src_lo,
+				   &data[10], 2);
+
+			tx_parse_bd->data.mac_addr.dst_hi =
+			    rte_cpu_to_be_16(tx_parse_bd->data.mac_addr.dst_hi);
+			tx_parse_bd->data.mac_addr.dst_mid =
+			    rte_cpu_to_be_16(tx_parse_bd->data.
+					     mac_addr.dst_mid);
+			tx_parse_bd->data.mac_addr.dst_lo =
+			    rte_cpu_to_be_16(tx_parse_bd->data.mac_addr.dst_lo);
+			tx_parse_bd->data.mac_addr.src_hi =
+			    rte_cpu_to_be_16(tx_parse_bd->data.mac_addr.src_hi);
+			tx_parse_bd->data.mac_addr.src_mid =
+			    rte_cpu_to_be_16(tx_parse_bd->data.
+					     mac_addr.src_mid);
+			tx_parse_bd->data.mac_addr.src_lo =
+			    rte_cpu_to_be_16(tx_parse_bd->data.mac_addr.src_lo);
+
+			PMD_TX_LOG(DEBUG,
+				   "PBD dst %x %x %x src %x %x %x p_data %x",
+				   tx_parse_bd->data.mac_addr.dst_hi,
+				   tx_parse_bd->data.mac_addr.dst_mid,
+				   tx_parse_bd->data.mac_addr.dst_lo,
+				   tx_parse_bd->data.mac_addr.src_hi,
+				   tx_parse_bd->data.mac_addr.src_mid,
+				   tx_parse_bd->data.mac_addr.src_lo,
+				   tx_parse_bd->parsing_data);
+		}
+
+		PMD_TX_LOG(DEBUG,
+			   "start bd: nbytes %d flags %x vlan %x\n",
+			   tx_start_bd->nbytes,
+			   tx_start_bd->bd_flags.as_bitfield,
+			   tx_start_bd->vlan_or_ethertype);
+
+		bd_prod = NEXT_TX_BD(bd_prod);
+		pkt_prod++;
+
+		if (TX_IDX(bd_prod) < 2) {
+			nbds++;
+		}
+	}
+
+	txq->nb_tx_avail -= m_pkts << 1;
+	txq->tx_bd_tail = bd_prod;
+	txq->tx_pkt_tail = pkt_prod;
+
+	mb();
+	fp->tx_db.data.prod += (m_pkts << 1) + nbds;
+	DOORBELL(sc, txq->queue_id, fp->tx_db.raw);
+	mb();
+
+	return 0;
+}
+
+static uint16_t bcm_cid_ilt_lines(struct bcm_softc *sc)
+{
+	return L2_ILT_LINES(sc);
+}
+
+static void bcm_ilt_set_info(struct bcm_softc *sc)
+{
+	struct ilt_client_info *ilt_client;
+	struct ecore_ilt *ilt = sc->ilt;
+	uint16_t line = 0;
+
+	PMD_INIT_FUNC_TRACE();
+
+	ilt->start_line = FUNC_ILT_BASE(SC_FUNC(sc));
+
+	/* CDU */
+	ilt_client = &ilt->clients[ILT_CLIENT_CDU];
+	ilt_client->client_num = ILT_CLIENT_CDU;
+	ilt_client->page_size = CDU_ILT_PAGE_SZ;
+	ilt_client->flags = ILT_CLIENT_SKIP_MEM;
+	ilt_client->start = line;
+	line += bcm_cid_ilt_lines(sc);
+
+	if (CNIC_SUPPORT(sc)) {
+		line += CNIC_ILT_LINES;
+	}
+
+	ilt_client->end = (line - 1);
+
+	/* QM */
+	if (QM_INIT(sc->qm_cid_count)) {
+		ilt_client = &ilt->clients[ILT_CLIENT_QM];
+		ilt_client->client_num = ILT_CLIENT_QM;
+		ilt_client->page_size = QM_ILT_PAGE_SZ;
+		ilt_client->flags = 0;
+		ilt_client->start = line;
+
+		/* 4 bytes for each cid */
+		line += DIV_ROUND_UP(sc->qm_cid_count * QM_QUEUES_PER_FUNC * 4,
+				     QM_ILT_PAGE_SZ);
+
+		ilt_client->end = (line - 1);
+	}
+
+	if (CNIC_SUPPORT(sc)) {
+		/* SRC */
+		ilt_client = &ilt->clients[ILT_CLIENT_SRC];
+		ilt_client->client_num = ILT_CLIENT_SRC;
+		ilt_client->page_size = SRC_ILT_PAGE_SZ;
+		ilt_client->flags = 0;
+		ilt_client->start = line;
+		line += SRC_ILT_LINES;
+		ilt_client->end = (line - 1);
+
+		/* TM */
+		ilt_client = &ilt->clients[ILT_CLIENT_TM];
+		ilt_client->client_num = ILT_CLIENT_TM;
+		ilt_client->page_size = TM_ILT_PAGE_SZ;
+		ilt_client->flags = 0;
+		ilt_client->start = line;
+		line += TM_ILT_LINES;
+		ilt_client->end = (line - 1);
+	}
+
+	assert((line <= ILT_MAX_LINES));
+}
+
+static void bcm_set_fp_rx_buf_size(struct bcm_softc *sc)
+{
+	int i;
+
+	for (i = 0; i < sc->num_queues; i++) {
+		/* get the Rx buffer size for RX frames */
+		sc->fp[i].rx_buf_size =
+		    (IP_HEADER_ALIGNMENT_PADDING + ETH_OVERHEAD + sc->mtu);
+
+		/* get the mbuf allocation size for RX frames */
+		if (sc->fp[i].rx_buf_size <= MCLBYTES) {
+			sc->fp[i].mbuf_alloc_size = MCLBYTES;
+		} else if (sc->fp[i].rx_buf_size <= BCM_PAGE_SIZE) {
+			sc->fp[i].mbuf_alloc_size = PAGE_SIZE;
+		} else {
+			sc->fp[i].mbuf_alloc_size = MJUM9BYTES;
+		}
+	}
+}
+
+int bcm_alloc_ilt_mem(struct bcm_softc *sc)
+{
+
+	sc->ilt = rte_malloc("", sizeof(struct ecore_ilt), RTE_CACHE_LINE_SIZE);
+
+	return sc->ilt == NULL;
+}
+
+static int bcm_alloc_ilt_lines_mem(struct bcm_softc *sc)
+{
+	sc->ilt->lines = rte_calloc("",
+				    sizeof(struct ilt_line), ILT_MAX_LINES,
+				    RTE_CACHE_LINE_SIZE);
+	return sc->ilt->lines == NULL;
+}
+
+void bcm_free_ilt_mem(struct bcm_softc *sc)
+{
+	rte_free(sc->ilt);
+	sc->ilt = NULL;
+}
+
+static void bcm_free_ilt_lines_mem(struct bcm_softc *sc)
+{
+	if (sc->ilt->lines != NULL) {
+		rte_free(sc->ilt->lines);
+		sc->ilt->lines = NULL;
+	}
+}
+
+static void bcm_free_mem(struct bcm_softc *sc)
+{
+	uint32_t i;
+
+	for (i = 0; i < L2_ILT_LINES(sc); i++) {
+		sc->context[i].vcxt = NULL;
+		sc->context[i].size = 0;
+	}
+
+	ecore_ilt_mem_op(sc, ILT_MEMOP_FREE);
+
+	bcm_free_ilt_lines_mem(sc);
+}
+
+static int bcm_alloc_mem(struct bcm_softc *sc)
+{
+	int context_size;
+	int allocated;
+	int i;
+	char cdu_name[RTE_MEMZONE_NAMESIZE];
+
+	/*
+	 * Allocate memory for CDU context:
+	 * This memory is allocated separately and not in the generic ILT
+	 * functions because CDU differs in few aspects:
+	 * 1. There can be multiple entities allocating memory for context -
+	 * regular L2, CNIC, and SRIOV drivers. Each separately controls
+	 * its own ILT lines.
+	 * 2. Since CDU page-size is not a single 4KB page (which is the case
+	 * for the other ILT clients), to be efficient we want to support
+	 * allocation of sub-page-size in the last entry.
+	 * 3. Context pointers are used by the driver to pass to FW / update
+	 * the context (for the other ILT clients the pointers are used just to
+	 * free the memory during unload).
+	 */
+	context_size = (sizeof(union cdu_context) * BCM_L2_CID_COUNT(sc));
+	for (i = 0, allocated = 0; allocated < context_size; i++) {
+		sc->context[i].size = min(CDU_ILT_PAGE_SZ,
+					  (context_size - allocated));
+
+		snprintf(cdu_name, sizeof(cdu_name), "cdu_%d", i);
+		if (bcm_dma_alloc(sc, sc->context[i].size,
+				  &sc->context[i].vcxt_dma,
+				  cdu_name, BCM_PAGE_SIZE) != 0) {
+			bcm_free_mem(sc);
+			return -1;
+		}
+
+		sc->context[i].vcxt =
+		    (union cdu_context *)sc->context[i].vcxt_dma.vaddr;
+
+		allocated += sc->context[i].size;
+	}
+
+	bcm_alloc_ilt_lines_mem(sc);
+
+	if (ecore_ilt_mem_op(sc, ILT_MEMOP_ALLOC)) {
+		PMD_DRV_LOG(NOTICE, "ecore_ilt_mem_op ILT_MEMOP_ALLOC failed");
+		bcm_free_mem(sc);
+		return -1;
+	}
+
+	return 0;
+}
+
+static void bcm_free_fw_stats_mem(struct bcm_softc *sc)
+{
+	sc->fw_stats_num = 0;
+
+	sc->fw_stats_req_size = 0;
+	sc->fw_stats_req = NULL;
+	sc->fw_stats_req_mapping = 0;
+
+	sc->fw_stats_data_size = 0;
+	sc->fw_stats_data = NULL;
+	sc->fw_stats_data_mapping = 0;
+}
+
+static int bcm_alloc_fw_stats_mem(struct bcm_softc *sc)
+{
+	uint8_t num_queue_stats;
+	int num_groups, vf_headroom = 0;
+
+	/* number of queues for statistics is number of eth queues */
+	num_queue_stats = BCM_NUM_ETH_QUEUES(sc);
+
+	/*
+	 * Total number of FW statistics requests =
+	 *   1 for port stats + 1 for PF stats + num of queues
+	 */
+	sc->fw_stats_num = (2 + num_queue_stats);
+
+	/*
+	 * Request is built from stats_query_header and an array of
+	 * stats_query_cmd_group each of which contains STATS_QUERY_CMD_COUNT
+	 * rules. The real number or requests is configured in the
+	 * stats_query_header.
+	 */
+	num_groups = (sc->fw_stats_num + vf_headroom) / STATS_QUERY_CMD_COUNT;
+	if ((sc->fw_stats_num + vf_headroom) % STATS_QUERY_CMD_COUNT)
+		num_groups++;
+
+	sc->fw_stats_req_size =
+	    (sizeof(struct stats_query_header) +
+	     (num_groups * sizeof(struct stats_query_cmd_group)));
+
+	/*
+	 * Data for statistics requests + stats_counter.
+	 * stats_counter holds per-STORM counters that are incremented when
+	 * STORM has finished with the current request. Memory for FCoE
+	 * offloaded statistics are counted anyway, even if they will not be sent.
+	 * VF stats are not accounted for here as the data of VF stats is stored
+	 * in memory allocated by the VF, not here.
+	 */
+	sc->fw_stats_data_size =
+	    (sizeof(struct stats_counter) +
+	     sizeof(struct per_port_stats) + sizeof(struct per_pf_stats) +
+	     /* sizeof(struct fcoe_statistics_params) + */
+	     (sizeof(struct per_queue_stats) * num_queue_stats));
+
+	if (bcm_dma_alloc(sc, (sc->fw_stats_req_size + sc->fw_stats_data_size),
+			  &sc->fw_stats_dma, "fw_stats",
+			  RTE_CACHE_LINE_SIZE) != 0) {
+		bcm_free_fw_stats_mem(sc);
+		return -1;
+	}
+
+	/* set up the shortcuts */
+
+	sc->fw_stats_req = (struct bcm_fw_stats_req *)sc->fw_stats_dma.vaddr;
+	sc->fw_stats_req_mapping = sc->fw_stats_dma.paddr;
+
+	sc->fw_stats_data =
+	    (struct bcm_fw_stats_data *)((uint8_t *) sc->fw_stats_dma.vaddr +
+					 sc->fw_stats_req_size);
+	sc->fw_stats_data_mapping = (sc->fw_stats_dma.paddr +
+				     sc->fw_stats_req_size);
+
+	return 0;
+}
+
+/*
+ * Bits map:
+ * 0-7  - Engine0 load counter.
+ * 8-15 - Engine1 load counter.
+ * 16   - Engine0 RESET_IN_PROGRESS bit.
+ * 17   - Engine1 RESET_IN_PROGRESS bit.
+ * 18   - Engine0 ONE_IS_LOADED. Set when there is at least one active
+ *        function on the engine
+ * 19   - Engine1 ONE_IS_LOADED.
+ * 20   - Chip reset flow bit. When set none-leader must wait for both engines
+ *        leader to complete (check for both RESET_IN_PROGRESS bits and not
+ *        for just the one belonging to its engine).
+ */
+#define BCM_RECOVERY_GLOB_REG     MISC_REG_GENERIC_POR_1
+#define BCM_PATH0_LOAD_CNT_MASK   0x000000ff
+#define BCM_PATH0_LOAD_CNT_SHIFT  0
+#define BCM_PATH1_LOAD_CNT_MASK   0x0000ff00
+#define BCM_PATH1_LOAD_CNT_SHIFT  8
+#define BCM_PATH0_RST_IN_PROG_BIT 0x00010000
+#define BCM_PATH1_RST_IN_PROG_BIT 0x00020000
+#define BCM_GLOBAL_RESET_BIT      0x00040000
+
+/* set the GLOBAL_RESET bit, should be run under rtnl lock */
+static void bcm_set_reset_global(struct bcm_softc *sc)
+{
+	uint32_t val;
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val | BCM_GLOBAL_RESET_BIT);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+}
+
+/* clear the GLOBAL_RESET bit, should be run under rtnl lock */
+static void bcm_clear_reset_global(struct bcm_softc *sc)
+{
+	uint32_t val;
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val & (~BCM_GLOBAL_RESET_BIT));
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+}
+
+/* checks the GLOBAL_RESET bit, should be run under rtnl lock */
+static uint8_t bcm_reset_is_global(struct bcm_softc *sc)
+{
+	return (REG_RD(sc, BCM_RECOVERY_GLOB_REG) & BCM_GLOBAL_RESET_BIT);
+}
+
+/* clear RESET_IN_PROGRESS bit for the engine, should be run under rtnl lock */
+static void bcm_set_reset_done(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint32_t bit = SC_PATH(sc) ? BCM_PATH1_RST_IN_PROG_BIT :
+	    BCM_PATH0_RST_IN_PROG_BIT;
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+	/* Clear the bit */
+	val &= ~bit;
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val);
+
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+}
+
+/* set RESET_IN_PROGRESS for the engine, should be run under rtnl lock */
+static void bcm_set_reset_in_progress(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint32_t bit = SC_PATH(sc) ? BCM_PATH1_RST_IN_PROG_BIT :
+	    BCM_PATH0_RST_IN_PROG_BIT;
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+	/* Set the bit */
+	val |= bit;
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val);
+
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+}
+
+/* check RESET_IN_PROGRESS bit for an engine, should be run under rtnl lock */
+static uint8_t bcm_reset_is_done(struct bcm_softc *sc, int engine)
+{
+	uint32_t val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+	uint32_t bit = engine ? BCM_PATH1_RST_IN_PROG_BIT :
+	    BCM_PATH0_RST_IN_PROG_BIT;
+
+	/* return false if bit is set */
+	return (val & bit) ? FALSE : TRUE;
+}
+
+/* get the load status for an engine, should be run under rtnl lock */
+static uint8_t bcm_get_load_status(struct bcm_softc *sc, int engine)
+{
+	uint32_t mask = engine ? BCM_PATH1_LOAD_CNT_MASK :
+	    BCM_PATH0_LOAD_CNT_MASK;
+	uint32_t shift = engine ? BCM_PATH1_LOAD_CNT_SHIFT :
+	    BCM_PATH0_LOAD_CNT_SHIFT;
+	uint32_t val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+
+	val = ((val & mask) >> shift);
+
+	return (val != 0);
+}
+
+/* set pf load mark */
+static void bcm_set_pf_load(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint32_t val1;
+	uint32_t mask = SC_PATH(sc) ? BCM_PATH1_LOAD_CNT_MASK :
+	    BCM_PATH0_LOAD_CNT_MASK;
+	uint32_t shift = SC_PATH(sc) ? BCM_PATH1_LOAD_CNT_SHIFT :
+	    BCM_PATH0_LOAD_CNT_SHIFT;
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+
+	PMD_INIT_FUNC_TRACE();
+
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+
+	/* get the current counter value */
+	val1 = ((val & mask) >> shift);
+
+	/* set bit of this PF */
+	val1 |= (1 << SC_ABS_FUNC(sc));
+
+	/* clear the old value */
+	val &= ~mask;
+
+	/* set the new one */
+	val |= ((val1 << shift) & mask);
+
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val);
+
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+}
+
+/* clear pf load mark */
+static uint8_t bcm_clear_pf_load(struct bcm_softc *sc)
+{
+	uint32_t val1, val;
+	uint32_t mask = SC_PATH(sc) ? BCM_PATH1_LOAD_CNT_MASK :
+	    BCM_PATH0_LOAD_CNT_MASK;
+	uint32_t shift = SC_PATH(sc) ? BCM_PATH1_LOAD_CNT_SHIFT :
+	    BCM_PATH0_LOAD_CNT_SHIFT;
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+	val = REG_RD(sc, BCM_RECOVERY_GLOB_REG);
+
+	/* get the current counter value */
+	val1 = (val & mask) >> shift;
+
+	/* clear bit of that PF */
+	val1 &= ~(1 << SC_ABS_FUNC(sc));
+
+	/* clear the old value */
+	val &= ~mask;
+
+	/* set the new one */
+	val |= ((val1 << shift) & mask);
+
+	REG_WR(sc, BCM_RECOVERY_GLOB_REG, val);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RECOVERY_REG);
+	return val1 != 0;
+}
+
+/* send load requrest to mcp and analyze response */
+static int bcm_nic_load_request(struct bcm_softc *sc, uint32_t * load_code)
+{
+	PMD_INIT_FUNC_TRACE();
+
+	/* init fw_seq */
+	sc->fw_seq =
+	    (SHMEM_RD(sc, func_mb[SC_FW_MB_IDX(sc)].drv_mb_header) &
+	     DRV_MSG_SEQ_NUMBER_MASK);
+
+	PMD_DRV_LOG(DEBUG, "initial fw_seq 0x%04x", sc->fw_seq);
+
+#ifdef BCM_PULSE
+	/* get the current FW pulse sequence */
+	sc->fw_drv_pulse_wr_seq =
+	    (SHMEM_RD(sc, func_mb[SC_FW_MB_IDX(sc)].drv_pulse_mb) &
+	     DRV_PULSE_SEQ_MASK);
+#else
+	/* set ALWAYS_ALIVE bit in shmem */
+	sc->fw_drv_pulse_wr_seq |= DRV_PULSE_ALWAYS_ALIVE;
+	bcm_drv_pulse(sc);
+#endif
+
+	/* load request */
+	(*load_code) = bcm_fw_command(sc, DRV_MSG_CODE_LOAD_REQ,
+				      DRV_MSG_CODE_LOAD_REQ_WITH_LFA);
+
+	/* if the MCP fails to respond we must abort */
+	if (!(*load_code)) {
+		PMD_DRV_LOG(NOTICE, "MCP response failure!");
+		return -1;
+	}
+
+	/* if MCP refused then must abort */
+	if ((*load_code) == FW_MSG_CODE_DRV_LOAD_REFUSED) {
+		PMD_DRV_LOG(NOTICE, "MCP refused load request");
+		return -1;
+	}
+
+	return 0;
+}
+
+/*
+ * Check whether another PF has already loaded FW to chip. In virtualized
+ * environments a pf from anoth VM may have already initialized the device
+ * including loading FW.
+ */
+static int bcm_nic_load_analyze_req(struct bcm_softc *sc, uint32_t load_code)
+{
+	uint32_t my_fw, loaded_fw;
+
+	/* is another pf loaded on this engine? */
+	if ((load_code != FW_MSG_CODE_DRV_LOAD_COMMON_CHIP) &&
+	    (load_code != FW_MSG_CODE_DRV_LOAD_COMMON)) {
+		/* build my FW version dword */
+		my_fw = (BCM_5710_FW_MAJOR_VERSION +
+			 (BCM_5710_FW_MINOR_VERSION << 8) +
+			 (BCM_5710_FW_REVISION_VERSION << 16) +
+			 (BCM_5710_FW_ENGINEERING_VERSION << 24));
+
+		/* read loaded FW from chip */
+		loaded_fw = REG_RD(sc, XSEM_REG_PRAM);
+		PMD_DRV_LOG(DEBUG, "loaded FW 0x%08x / my FW 0x%08x",
+			    loaded_fw, my_fw);
+
+		/* abort nic load if version mismatch */
+		if (my_fw != loaded_fw) {
+			PMD_DRV_LOG(NOTICE,
+				    "FW 0x%08x already loaded (mine is 0x%08x)",
+				    loaded_fw, my_fw);
+			return -1;
+		}
+	}
+
+	return 0;
+}
+
+/* mark PMF if applicable */
+static void bcm_nic_load_pmf(struct bcm_softc *sc, uint32_t load_code)
+{
+	uint32_t ncsi_oem_data_addr;
+
+	PMD_INIT_FUNC_TRACE();
+
+	if ((load_code == FW_MSG_CODE_DRV_LOAD_COMMON) ||
+	    (load_code == FW_MSG_CODE_DRV_LOAD_COMMON_CHIP) ||
+	    (load_code == FW_MSG_CODE_DRV_LOAD_PORT)) {
+		/*
+		 * Barrier here for ordering between the writing to sc->port.pmf here
+		 * and reading it from the periodic task.
+		 */
+		sc->port.pmf = 1;
+		mb();
+	} else {
+		sc->port.pmf = 0;
+	}
+
+	PMD_DRV_LOG(DEBUG, "pmf %d", sc->port.pmf);
+
+	if (load_code == FW_MSG_CODE_DRV_LOAD_COMMON_CHIP) {
+		if (SHMEM2_HAS(sc, ncsi_oem_data_addr)) {
+			ncsi_oem_data_addr = SHMEM2_RD(sc, ncsi_oem_data_addr);
+			if (ncsi_oem_data_addr) {
+				REG_WR(sc,
+				       (ncsi_oem_data_addr +
+					offsetof(struct glob_ncsi_oem_data,
+						 driver_version)), 0);
+			}
+		}
+	}
+}
+
+static void bcm_read_mf_cfg(struct bcm_softc *sc)
+{
+	int n = (CHIP_IS_MODE_4_PORT(sc) ? 2 : 1);
+	int abs_func;
+	int vn;
+
+	if (BCM_NOMCP(sc)) {
+		return;		/* what should be the default bvalue in this case */
+	}
+
+	/*
+	 * The formula for computing the absolute function number is...
+	 * For 2 port configuration (4 functions per port):
+	 *   abs_func = 2 * vn + SC_PORT + SC_PATH
+	 * For 4 port configuration (2 functions per port):
+	 *   abs_func = 4 * vn + 2 * SC_PORT + SC_PATH
+	 */
+	for (vn = VN_0; vn < SC_MAX_VN_NUM(sc); vn++) {
+		abs_func = (n * (2 * vn + SC_PORT(sc)) + SC_PATH(sc));
+		if (abs_func >= E1H_FUNC_MAX) {
+			break;
+		}
+		sc->devinfo.mf_info.mf_config[vn] =
+		    MFCFG_RD(sc, func_mf_config[abs_func].config);
+	}
+
+	if (sc->devinfo.mf_info.mf_config[SC_VN(sc)] &
+	    FUNC_MF_CFG_FUNC_DISABLED) {
+		PMD_DRV_LOG(DEBUG, "mf_cfg function disabled");
+		sc->flags |= BCM_MF_FUNC_DIS;
+	} else {
+		PMD_DRV_LOG(DEBUG, "mf_cfg function enabled");
+		sc->flags &= ~BCM_MF_FUNC_DIS;
+	}
+}
+
+/* acquire split MCP access lock register */
+static int bcm_acquire_alr(struct bcm_softc *sc)
+{
+	uint32_t j, val;
+
+	for (j = 0; j < 1000; j++) {
+		val = (1UL << 31);
+		REG_WR(sc, GRCBASE_MCP + 0x9c, val);
+		val = REG_RD(sc, GRCBASE_MCP + 0x9c);
+		if (val & (1L << 31))
+			break;
+
+		DELAY(5000);
+	}
+
+	if (!(val & (1L << 31))) {
+		PMD_DRV_LOG(NOTICE, "Cannot acquire MCP access lock register");
+		return -1;
+	}
+
+	return 0;
+}
+
+/* release split MCP access lock register */
+static void bcm_release_alr(struct bcm_softc *sc)
+{
+	REG_WR(sc, GRCBASE_MCP + 0x9c, 0);
+}
+
+static void bcm_fan_failure(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	uint32_t ext_phy_config;
+
+	/* mark the failure */
+	ext_phy_config =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].external_phy_config);
+
+	ext_phy_config &= ~PORT_HW_CFG_XGXS_EXT_PHY_TYPE_MASK;
+	ext_phy_config |= PORT_HW_CFG_XGXS_EXT_PHY_TYPE_FAILURE;
+	SHMEM_WR(sc, dev_info.port_hw_config[port].external_phy_config,
+		 ext_phy_config);
+
+	/* log the failure */
+	PMD_DRV_LOG(INFO,
+		    "Fan Failure has caused the driver to shutdown "
+		    "the card to prevent permanent damage. "
+		    "Please contact OEM Support for assistance");
+
+	rte_panic("Schedule task to handle fan failure");
+}
+
+/* this function is called upon a link interrupt */
+static void bcm_link_attn(struct bcm_softc *sc)
+{
+	uint32_t pause_enabled = 0;
+	struct host_port_stats *pstats;
+	int cmng_fns;
+
+	/* Make sure that we are synced with the current statistics */
+	bcm_stats_handle(sc, STATS_EVENT_STOP);
+
+	elink_link_update(&sc->link_params, &sc->link_vars);
+
+	if (sc->link_vars.link_up) {
+
+		/* dropless flow control */
+		if (sc->dropless_fc) {
+			pause_enabled = 0;
+
+			if (sc->link_vars.flow_ctrl & ELINK_FLOW_CTRL_TX) {
+				pause_enabled = 1;
+			}
+
+			REG_WR(sc,
+			       (BAR_USTRORM_INTMEM +
+				USTORM_ETH_PAUSE_ENABLED_OFFSET(SC_PORT(sc))),
+			       pause_enabled);
+		}
+
+		if (sc->link_vars.mac_type != ELINK_MAC_TYPE_EMAC) {
+			pstats = BCM_SP(sc, port_stats);
+			/* reset old mac stats */
+			memset(&(pstats->mac_stx[0]), 0,
+			       sizeof(struct mac_stx));
+		}
+
+		if (sc->state == BCM_STATE_OPEN) {
+			bcm_stats_handle(sc, STATS_EVENT_LINK_UP);
+		}
+	}
+
+	if (sc->link_vars.link_up && sc->link_vars.line_speed) {
+		cmng_fns = bcm_get_cmng_fns_mode(sc);
+
+		if (cmng_fns != CMNG_FNS_NONE) {
+			bcm_cmng_fns_init(sc, FALSE, cmng_fns);
+			storm_memset_cmng(sc, &sc->cmng, SC_PORT(sc));
+		}
+	}
+
+	bcm_link_report(sc);
+
+	if (IS_MF(sc)) {
+		bcm_link_sync_notify(sc);
+	}
+}
+
+static void bcm_attn_int_asserted(struct bcm_softc *sc, uint32_t asserted)
+{
+	int port = SC_PORT(sc);
+	uint32_t aeu_addr = port ? MISC_REG_AEU_MASK_ATTN_FUNC_1 :
+	    MISC_REG_AEU_MASK_ATTN_FUNC_0;
+	uint32_t nig_int_mask_addr = port ? NIG_REG_MASK_INTERRUPT_PORT1 :
+	    NIG_REG_MASK_INTERRUPT_PORT0;
+	uint32_t aeu_mask;
+	uint32_t nig_mask = 0;
+	uint32_t reg_addr;
+	uint32_t igu_acked;
+	uint32_t cnt;
+
+	if (sc->attn_state & asserted) {
+		PMD_DRV_LOG(ERR, "IGU ERROR attn=0x%08x", asserted);
+	}
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_PORT0_ATT_MASK + port);
+
+	aeu_mask = REG_RD(sc, aeu_addr);
+
+	aeu_mask &= ~(asserted & 0x3ff);
+
+	REG_WR(sc, aeu_addr, aeu_mask);
+
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_PORT0_ATT_MASK + port);
+
+	sc->attn_state |= asserted;
+
+	if (asserted & ATTN_HARD_WIRED_MASK) {
+		if (asserted & ATTN_NIG_FOR_FUNC) {
+
+			/* save nig interrupt mask */
+			nig_mask = REG_RD(sc, nig_int_mask_addr);
+
+			/* If nig_mask is not set, no need to call the update function */
+			if (nig_mask) {
+				REG_WR(sc, nig_int_mask_addr, 0);
+
+				bcm_link_attn(sc);
+			}
+
+			/* handle unicore attn? */
+		}
+
+		if (asserted & ATTN_SW_TIMER_4_FUNC) {
+			PMD_DRV_LOG(DEBUG, "ATTN_SW_TIMER_4_FUNC!");
+		}
+
+		if (asserted & GPIO_2_FUNC) {
+			PMD_DRV_LOG(DEBUG, "GPIO_2_FUNC!");
+		}
+
+		if (asserted & GPIO_3_FUNC) {
+			PMD_DRV_LOG(DEBUG, "GPIO_3_FUNC!");
+		}
+
+		if (asserted & GPIO_4_FUNC) {
+			PMD_DRV_LOG(DEBUG, "GPIO_4_FUNC!");
+		}
+
+		if (port == 0) {
+			if (asserted & ATTN_GENERAL_ATTN_1) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_1!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_1, 0x0);
+			}
+			if (asserted & ATTN_GENERAL_ATTN_2) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_2!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_2, 0x0);
+			}
+			if (asserted & ATTN_GENERAL_ATTN_3) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_3!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_3, 0x0);
+			}
+		} else {
+			if (asserted & ATTN_GENERAL_ATTN_4) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_4!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_4, 0x0);
+			}
+			if (asserted & ATTN_GENERAL_ATTN_5) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_5!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_5, 0x0);
+			}
+			if (asserted & ATTN_GENERAL_ATTN_6) {
+				PMD_DRV_LOG(DEBUG, "ATTN_GENERAL_ATTN_6!");
+				REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_6, 0x0);
+			}
+		}
+	}
+	/* hardwired */
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		reg_addr =
+		    (HC_REG_COMMAND_REG + port * 32 +
+		     COMMAND_REG_ATTN_BITS_SET);
+	} else {
+		reg_addr = (BAR_IGU_INTMEM + IGU_CMD_ATTN_BIT_SET_UPPER * 8);
+	}
+
+	PMD_DRV_LOG(DEBUG, "about to mask 0x%08x at %s addr 0x%08x",
+		    asserted,
+		    (sc->devinfo.int_block == INT_BLOCK_HC) ? "HC" : "IGU",
+		    reg_addr);
+	REG_WR(sc, reg_addr, asserted);
+
+	/* now set back the mask */
+	if (asserted & ATTN_NIG_FOR_FUNC) {
+		/*
+		 * Verify that IGU ack through BAR was written before restoring
+		 * NIG mask. This loop should exit after 2-3 iterations max.
+		 */
+		if (sc->devinfo.int_block != INT_BLOCK_HC) {
+			cnt = 0;
+
+			do {
+				igu_acked =
+				    REG_RD(sc, IGU_REG_ATTENTION_ACK_BITS);
+			} while (((igu_acked & ATTN_NIG_FOR_FUNC) == 0)
+				 && (++cnt < MAX_IGU_ATTN_ACK_TO));
+
+			if (!igu_acked) {
+				PMD_DRV_LOG(ERR,
+					    "Failed to verify IGU ack on time");
+			}
+
+			mb();
+		}
+
+		REG_WR(sc, nig_int_mask_addr, nig_mask);
+
+	}
+}
+
+static void
+bcm_print_next_block(__rte_unused struct bcm_softc *sc, __rte_unused int idx,
+		     __rte_unused const char *blk)
+{
+	PMD_DRV_LOG(INFO, "%s%s", idx ? ", " : "", blk);
+}
+
+static int
+bcm_check_blocks_with_parity0(struct bcm_softc *sc, uint32_t sig, int par_num,
+			      uint8_t print)
+{
+	uint32_t cur_bit = 0;
+	int i = 0;
+
+	for (i = 0; sig; i++) {
+		cur_bit = ((uint32_t) 0x1 << i);
+		if (sig & cur_bit) {
+			switch (cur_bit) {
+			case AEU_INPUTS_ATTN_BITS_BRB_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "BRB");
+				break;
+			case AEU_INPUTS_ATTN_BITS_PARSER_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "PARSER");
+				break;
+			case AEU_INPUTS_ATTN_BITS_TSDM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "TSDM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_SEARCHER_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "SEARCHER");
+				break;
+			case AEU_INPUTS_ATTN_BITS_TCM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "TCM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_TSEMI_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "TSEMI");
+				break;
+			case AEU_INPUTS_ATTN_BITS_PBCLIENT_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "XPB");
+				break;
+			}
+
+			/* Clear the bit */
+			sig &= ~cur_bit;
+		}
+	}
+
+	return par_num;
+}
+
+static int
+bcm_check_blocks_with_parity1(struct bcm_softc *sc, uint32_t sig, int par_num,
+			      uint8_t * global, uint8_t print)
+{
+	int i = 0;
+	uint32_t cur_bit = 0;
+	for (i = 0; sig; i++) {
+		cur_bit = ((uint32_t) 0x1 << i);
+		if (sig & cur_bit) {
+			switch (cur_bit) {
+			case AEU_INPUTS_ATTN_BITS_PBF_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "PBF");
+				break;
+			case AEU_INPUTS_ATTN_BITS_QM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "QM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_TIMERS_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "TM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_XSDM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "XSDM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_XCM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "XCM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_XSEMI_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "XSEMI");
+				break;
+			case AEU_INPUTS_ATTN_BITS_DOORBELLQ_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "DOORBELLQ");
+				break;
+			case AEU_INPUTS_ATTN_BITS_NIG_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "NIG");
+				break;
+			case AEU_INPUTS_ATTN_BITS_VAUX_PCI_CORE_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "VAUX PCI CORE");
+				*global = TRUE;
+				break;
+			case AEU_INPUTS_ATTN_BITS_DEBUG_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "DEBUG");
+				break;
+			case AEU_INPUTS_ATTN_BITS_USDM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "USDM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_UCM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "UCM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_USEMI_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "USEMI");
+				break;
+			case AEU_INPUTS_ATTN_BITS_UPB_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "UPB");
+				break;
+			case AEU_INPUTS_ATTN_BITS_CSDM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "CSDM");
+				break;
+			case AEU_INPUTS_ATTN_BITS_CCM_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "CCM");
+				break;
+			}
+
+			/* Clear the bit */
+			sig &= ~cur_bit;
+		}
+	}
+
+	return par_num;
+}
+
+static int
+bcm_check_blocks_with_parity2(struct bcm_softc *sc, uint32_t sig, int par_num,
+			      uint8_t print)
+{
+	uint32_t cur_bit = 0;
+	int i = 0;
+
+	for (i = 0; sig; i++) {
+		cur_bit = ((uint32_t) 0x1 << i);
+		if (sig & cur_bit) {
+			switch (cur_bit) {
+			case AEU_INPUTS_ATTN_BITS_CSEMI_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "CSEMI");
+				break;
+			case AEU_INPUTS_ATTN_BITS_PXP_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "PXP");
+				break;
+			case AEU_IN_ATTN_BITS_PXPPCICLOCKCLIENT_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "PXPPCICLOCKCLIENT");
+				break;
+			case AEU_INPUTS_ATTN_BITS_CFC_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "CFC");
+				break;
+			case AEU_INPUTS_ATTN_BITS_CDU_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "CDU");
+				break;
+			case AEU_INPUTS_ATTN_BITS_DMAE_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "DMAE");
+				break;
+			case AEU_INPUTS_ATTN_BITS_IGU_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "IGU");
+				break;
+			case AEU_INPUTS_ATTN_BITS_MISC_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "MISC");
+				break;
+			}
+
+			/* Clear the bit */
+			sig &= ~cur_bit;
+		}
+	}
+
+	return par_num;
+}
+
+static int
+bcm_check_blocks_with_parity3(struct bcm_softc *sc, uint32_t sig, int par_num,
+			      uint8_t * global, uint8_t print)
+{
+	uint32_t cur_bit = 0;
+	int i = 0;
+
+	for (i = 0; sig; i++) {
+		cur_bit = ((uint32_t) 0x1 << i);
+		if (sig & cur_bit) {
+			switch (cur_bit) {
+			case AEU_INPUTS_ATTN_BITS_MCP_LATCHED_ROM_PARITY:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "MCP ROM");
+				*global = TRUE;
+				break;
+			case AEU_INPUTS_ATTN_BITS_MCP_LATCHED_UMP_RX_PARITY:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "MCP UMP RX");
+				*global = TRUE;
+				break;
+			case AEU_INPUTS_ATTN_BITS_MCP_LATCHED_UMP_TX_PARITY:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "MCP UMP TX");
+				*global = TRUE;
+				break;
+			case AEU_INPUTS_ATTN_BITS_MCP_LATCHED_SCPAD_PARITY:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "MCP SCPAD");
+				*global = TRUE;
+				break;
+			}
+
+			/* Clear the bit */
+			sig &= ~cur_bit;
+		}
+	}
+
+	return par_num;
+}
+
+static int
+bcm_check_blocks_with_parity4(struct bcm_softc *sc, uint32_t sig, int par_num,
+			      uint8_t print)
+{
+	uint32_t cur_bit = 0;
+	int i = 0;
+
+	for (i = 0; sig; i++) {
+		cur_bit = ((uint32_t) 0x1 << i);
+		if (sig & cur_bit) {
+			switch (cur_bit) {
+			case AEU_INPUTS_ATTN_BITS_PGLUE_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "PGLUE_B");
+				break;
+			case AEU_INPUTS_ATTN_BITS_ATC_PARITY_ERROR:
+				if (print)
+					bcm_print_next_block(sc, par_num++,
+							     "ATC");
+				break;
+			}
+
+			/* Clear the bit */
+			sig &= ~cur_bit;
+		}
+	}
+
+	return par_num;
+}
+
+static uint8_t
+bcm_parity_attn(struct bcm_softc *sc, uint8_t * global, uint8_t print,
+		uint32_t * sig)
+{
+	int par_num = 0;
+
+	if ((sig[0] & HW_PRTY_ASSERT_SET_0) ||
+	    (sig[1] & HW_PRTY_ASSERT_SET_1) ||
+	    (sig[2] & HW_PRTY_ASSERT_SET_2) ||
+	    (sig[3] & HW_PRTY_ASSERT_SET_3) ||
+	    (sig[4] & HW_PRTY_ASSERT_SET_4)) {
+		PMD_DRV_LOG(ERR,
+			    "Parity error: HW block parity attention:"
+			    "[0]:0x%08x [1]:0x%08x [2]:0x%08x [3]:0x%08x [4]:0x%08x",
+			    (uint32_t) (sig[0] & HW_PRTY_ASSERT_SET_0),
+			    (uint32_t) (sig[1] & HW_PRTY_ASSERT_SET_1),
+			    (uint32_t) (sig[2] & HW_PRTY_ASSERT_SET_2),
+			    (uint32_t) (sig[3] & HW_PRTY_ASSERT_SET_3),
+			    (uint32_t) (sig[4] & HW_PRTY_ASSERT_SET_4));
+
+		if (print)
+			PMD_DRV_LOG(INFO, "Parity errors detected in blocks: ");
+
+		par_num =
+		    bcm_check_blocks_with_parity0(sc, sig[0] &
+						  HW_PRTY_ASSERT_SET_0,
+						  par_num, print);
+		par_num =
+		    bcm_check_blocks_with_parity1(sc, sig[1] &
+						  HW_PRTY_ASSERT_SET_1,
+						  par_num, global, print);
+		par_num =
+		    bcm_check_blocks_with_parity2(sc, sig[2] &
+						  HW_PRTY_ASSERT_SET_2,
+						  par_num, print);
+		par_num =
+		    bcm_check_blocks_with_parity3(sc, sig[3] &
+						  HW_PRTY_ASSERT_SET_3,
+						  par_num, global, print);
+		par_num =
+		    bcm_check_blocks_with_parity4(sc, sig[4] &
+						  HW_PRTY_ASSERT_SET_4,
+						  par_num, print);
+
+		if (print)
+			PMD_DRV_LOG(INFO, "");
+
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+static uint8_t
+bcm_chk_parity_attn(struct bcm_softc *sc, uint8_t * global, uint8_t print)
+{
+	struct attn_route attn = { {0} };
+	int port = SC_PORT(sc);
+
+	attn.sig[0] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_1_FUNC_0 + port * 4);
+	attn.sig[1] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_2_FUNC_0 + port * 4);
+	attn.sig[2] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_3_FUNC_0 + port * 4);
+	attn.sig[3] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_4_FUNC_0 + port * 4);
+
+	if (!CHIP_IS_E1x(sc))
+		attn.sig[4] =
+		    REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_5_FUNC_0 + port * 4);
+
+	return bcm_parity_attn(sc, global, print, attn.sig);
+}
+
+static void bcm_attn_int_deasserted4(struct bcm_softc *sc, uint32_t attn)
+{
+	uint32_t val;
+
+	if (attn & AEU_INPUTS_ATTN_BITS_PGLUE_HW_INTERRUPT) {
+		val = REG_RD(sc, PGLUE_B_REG_PGLUE_B_INT_STS_CLR);
+		PMD_DRV_LOG(INFO, "ERROR: PGLUE hw attention 0x%08x", val);
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_ADDRESS_ERROR)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_ADDRESS_ERROR");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_INCORRECT_RCV_BEHAVIOR)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_INCORRECT_RCV_BEHAVIOR");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_WAS_ERROR_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_WAS_ERROR_ATTN");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_VF_LENGTH_VIOLATION_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_VF_LENGTH_VIOLATION_ATTN");
+		if (val &
+		    PGLUE_B_PGLUE_B_INT_STS_REG_VF_GRC_SPACE_VIOLATION_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_VF_GRC_SPACE_VIOLATION_ATTN");
+		if (val &
+		    PGLUE_B_PGLUE_B_INT_STS_REG_VF_MSIX_BAR_VIOLATION_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_VF_MSIX_BAR_VIOLATION_ATTN");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_TCPL_ERROR_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_TCPL_ERROR_ATTN");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_TCPL_IN_TWO_RCBS_ATTN)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_TCPL_IN_TWO_RCBS_ATTN");
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_CSSNOOP_FIFO_OVERFLOW)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: PGLUE_B_PGLUE_B_INT_STS_REG_CSSNOOP_FIFO_OVERFLOW");
+	}
+
+	if (attn & AEU_INPUTS_ATTN_BITS_ATC_HW_INTERRUPT) {
+		val = REG_RD(sc, ATC_REG_ATC_INT_STS_CLR);
+		PMD_DRV_LOG(INFO, "ERROR: ATC hw attention 0x%08x", val);
+		if (val & ATC_ATC_INT_STS_REG_ADDRESS_ERROR)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ADDRESS_ERROR");
+		if (val & ATC_ATC_INT_STS_REG_ATC_TCPL_TO_NOT_PEND)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ATC_TCPL_TO_NOT_PEND");
+		if (val & ATC_ATC_INT_STS_REG_ATC_GPA_MULTIPLE_HITS)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ATC_GPA_MULTIPLE_HITS");
+		if (val & ATC_ATC_INT_STS_REG_ATC_RCPL_TO_EMPTY_CNT)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ATC_RCPL_TO_EMPTY_CNT");
+		if (val & ATC_ATC_INT_STS_REG_ATC_TCPL_ERROR)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ATC_TCPL_ERROR");
+		if (val & ATC_ATC_INT_STS_REG_ATC_IREQ_LESS_THAN_STU)
+			PMD_DRV_LOG(INFO,
+				    "ERROR: ATC_ATC_INT_STS_REG_ATC_IREQ_LESS_THAN_STU");
+	}
+
+	if (attn & (AEU_INPUTS_ATTN_BITS_PGLUE_PARITY_ERROR |
+		    AEU_INPUTS_ATTN_BITS_ATC_PARITY_ERROR)) {
+		PMD_DRV_LOG(INFO,
+			    "ERROR: FATAL parity attention set4 0x%08x",
+			    (uint32_t) (attn &
+					(AEU_INPUTS_ATTN_BITS_PGLUE_PARITY_ERROR
+					 |
+					 AEU_INPUTS_ATTN_BITS_ATC_PARITY_ERROR)));
+	}
+}
+
+static void bcm_e1h_disable(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+
+	REG_WR(sc, NIG_REG_LLH0_FUNC_EN + port * 8, 0);
+}
+
+static void bcm_e1h_enable(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+
+	REG_WR(sc, NIG_REG_LLH0_FUNC_EN + port * 8, 1);
+}
+
+/*
+ * called due to MCP event (on pmf):
+ *   reread new bandwidth configuration
+ *   configure FW
+ *   notify others function about the change
+ */
+static void bcm_config_mf_bw(struct bcm_softc *sc)
+{
+	if (sc->link_vars.link_up) {
+		bcm_cmng_fns_init(sc, TRUE, CMNG_FNS_MINMAX);
+		bcm_link_sync_notify(sc);
+	}
+
+	storm_memset_cmng(sc, &sc->cmng, SC_PORT(sc));
+}
+
+static void bcm_set_mf_bw(struct bcm_softc *sc)
+{
+	bcm_config_mf_bw(sc);
+	bcm_fw_command(sc, DRV_MSG_CODE_SET_MF_BW_ACK, 0);
+}
+
+static void bcm_handle_eee_event(struct bcm_softc *sc)
+{
+	bcm_fw_command(sc, DRV_MSG_CODE_EEE_RESULTS_ACK, 0);
+}
+
+#define DRV_INFO_ETH_STAT_NUM_MACS_REQUIRED 3
+
+static void bcm_drv_info_ether_stat(struct bcm_softc *sc)
+{
+	struct eth_stats_info *ether_stat = &sc->sp->drv_info_to_mcp.ether_stat;
+
+	strncpy(ether_stat->version, BCM_DRIVER_VERSION,
+		ETH_STAT_INFO_VERSION_LEN);
+
+	sc->sp_objs[0].mac_obj.get_n_elements(sc, &sc->sp_objs[0].mac_obj,
+					      DRV_INFO_ETH_STAT_NUM_MACS_REQUIRED,
+					      ether_stat->mac_local + MAC_PAD,
+					      MAC_PAD, ETH_ALEN);
+
+	ether_stat->mtu_size = sc->mtu;
+
+	ether_stat->feature_flags |= FEATURE_ETH_CHKSUM_OFFLOAD_MASK;
+	ether_stat->promiscuous_mode = 0;	// (flags & PROMISC) ? 1 : 0;
+
+	ether_stat->txq_size = sc->tx_ring_size;
+	ether_stat->rxq_size = sc->rx_ring_size;
+}
+
+static void bcm_handle_drv_info_req(struct bcm_softc *sc)
+{
+	enum drv_info_opcode op_code;
+	uint32_t drv_info_ctl = SHMEM2_RD(sc, drv_info_control);
+
+	/* if drv_info version supported by MFW doesn't match - send NACK */
+	if ((drv_info_ctl & DRV_INFO_CONTROL_VER_MASK) != DRV_INFO_CUR_VER) {
+		bcm_fw_command(sc, DRV_MSG_CODE_DRV_INFO_NACK, 0);
+		return;
+	}
+
+	op_code = ((drv_info_ctl & DRV_INFO_CONTROL_OP_CODE_MASK) >>
+		   DRV_INFO_CONTROL_OP_CODE_SHIFT);
+
+	memset(&sc->sp->drv_info_to_mcp, 0, sizeof(union drv_info_to_mcp));
+
+	switch (op_code) {
+	case ETH_STATS_OPCODE:
+		bcm_drv_info_ether_stat(sc);
+		break;
+	case FCOE_STATS_OPCODE:
+	case ISCSI_STATS_OPCODE:
+	default:
+		/* if op code isn't supported - send NACK */
+		bcm_fw_command(sc, DRV_MSG_CODE_DRV_INFO_NACK, 0);
+		return;
+	}
+
+	/*
+	 * If we got drv_info attn from MFW then these fields are defined in
+	 * shmem2 for sure
+	 */
+	SHMEM2_WR(sc, drv_info_host_addr_lo,
+		  U64_LO(BCM_SP_MAPPING(sc, drv_info_to_mcp)));
+	SHMEM2_WR(sc, drv_info_host_addr_hi,
+		  U64_HI(BCM_SP_MAPPING(sc, drv_info_to_mcp)));
+
+	bcm_fw_command(sc, DRV_MSG_CODE_DRV_INFO_ACK, 0);
+}
+
+static void bcm_dcc_event(struct bcm_softc *sc, uint32_t dcc_event)
+{
+	if (dcc_event & DRV_STATUS_DCC_DISABLE_ENABLE_PF) {
+/*
+ * This is the only place besides the function initialization
+ * where the sc->flags can change so it is done without any
+ * locks
+ */
+		if (sc->devinfo.
+		    mf_info.mf_config[SC_VN(sc)] & FUNC_MF_CFG_FUNC_DISABLED) {
+			PMD_DRV_LOG(DEBUG, "mf_cfg function disabled");
+			sc->flags |= BCM_MF_FUNC_DIS;
+			bcm_e1h_disable(sc);
+		} else {
+			PMD_DRV_LOG(DEBUG, "mf_cfg function enabled");
+			sc->flags &= ~BCM_MF_FUNC_DIS;
+			bcm_e1h_enable(sc);
+		}
+		dcc_event &= ~DRV_STATUS_DCC_DISABLE_ENABLE_PF;
+	}
+
+	if (dcc_event & DRV_STATUS_DCC_BANDWIDTH_ALLOCATION) {
+		bcm_config_mf_bw(sc);
+		dcc_event &= ~DRV_STATUS_DCC_BANDWIDTH_ALLOCATION;
+	}
+
+	/* Report results to MCP */
+	if (dcc_event)
+		bcm_fw_command(sc, DRV_MSG_CODE_DCC_FAILURE, 0);
+	else
+		bcm_fw_command(sc, DRV_MSG_CODE_DCC_OK, 0);
+}
+
+static void bcm_pmf_update(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	uint32_t val;
+
+	sc->port.pmf = 1;
+
+	/*
+	 * We need the mb() to ensure the ordering between the writing to
+	 * sc->port.pmf here and reading it from the bcm_periodic_task().
+	 */
+	mb();
+
+	/* enable nig attention */
+	val = (0xff0f | (1 << (SC_VN(sc) + 4)));
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		REG_WR(sc, HC_REG_TRAILING_EDGE_0 + port * 8, val);
+		REG_WR(sc, HC_REG_LEADING_EDGE_0 + port * 8, val);
+	} else if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, IGU_REG_TRAILING_EDGE_LATCH, val);
+		REG_WR(sc, IGU_REG_LEADING_EDGE_LATCH, val);
+	}
+
+	bcm_stats_handle(sc, STATS_EVENT_PMF);
+}
+
+static int bcm_mc_assert(struct bcm_softc *sc)
+{
+	char last_idx;
+	int i, rc = 0;
+	__rte_unused uint32_t row0, row1, row2, row3;
+
+	/* XSTORM */
+	last_idx =
+	    REG_RD8(sc, BAR_XSTRORM_INTMEM + XSTORM_ASSERT_LIST_INDEX_OFFSET);
+	if (last_idx)
+		PMD_DRV_LOG(ERR, "XSTORM_ASSERT_LIST_INDEX 0x%x", last_idx);
+
+	/* print the asserts */
+	for (i = 0; i < STORM_ASSERT_ARRAY_SIZE; i++) {
+
+		row0 =
+		    REG_RD(sc,
+			   BAR_XSTRORM_INTMEM + XSTORM_ASSERT_LIST_OFFSET(i));
+		row1 =
+		    REG_RD(sc,
+			   BAR_XSTRORM_INTMEM + XSTORM_ASSERT_LIST_OFFSET(i) +
+			   4);
+		row2 =
+		    REG_RD(sc,
+			   BAR_XSTRORM_INTMEM + XSTORM_ASSERT_LIST_OFFSET(i) +
+			   8);
+		row3 =
+		    REG_RD(sc,
+			   BAR_XSTRORM_INTMEM + XSTORM_ASSERT_LIST_OFFSET(i) +
+			   12);
+
+		if (row0 != COMMON_ASM_INVALID_ASSERT_OPCODE) {
+			PMD_DRV_LOG(ERR,
+				    "XSTORM_ASSERT_INDEX 0x%x = 0x%08x 0x%08x 0x%08x 0x%08x",
+				    i, row3, row2, row1, row0);
+			rc++;
+		} else {
+			break;
+		}
+	}
+
+	/* TSTORM */
+	last_idx =
+	    REG_RD8(sc, BAR_TSTRORM_INTMEM + TSTORM_ASSERT_LIST_INDEX_OFFSET);
+	if (last_idx) {
+		PMD_DRV_LOG(ERR, "TSTORM_ASSERT_LIST_INDEX 0x%x", last_idx);
+	}
+
+	/* print the asserts */
+	for (i = 0; i < STORM_ASSERT_ARRAY_SIZE; i++) {
+
+		row0 =
+		    REG_RD(sc,
+			   BAR_TSTRORM_INTMEM + TSTORM_ASSERT_LIST_OFFSET(i));
+		row1 =
+		    REG_RD(sc,
+			   BAR_TSTRORM_INTMEM + TSTORM_ASSERT_LIST_OFFSET(i) +
+			   4);
+		row2 =
+		    REG_RD(sc,
+			   BAR_TSTRORM_INTMEM + TSTORM_ASSERT_LIST_OFFSET(i) +
+			   8);
+		row3 =
+		    REG_RD(sc,
+			   BAR_TSTRORM_INTMEM + TSTORM_ASSERT_LIST_OFFSET(i) +
+			   12);
+
+		if (row0 != COMMON_ASM_INVALID_ASSERT_OPCODE) {
+			PMD_DRV_LOG(ERR,
+				    "TSTORM_ASSERT_INDEX 0x%x = 0x%08x 0x%08x 0x%08x 0x%08x",
+				    i, row3, row2, row1, row0);
+			rc++;
+		} else {
+			break;
+		}
+	}
+
+	/* CSTORM */
+	last_idx =
+	    REG_RD8(sc, BAR_CSTRORM_INTMEM + CSTORM_ASSERT_LIST_INDEX_OFFSET);
+	if (last_idx) {
+		PMD_DRV_LOG(ERR, "CSTORM_ASSERT_LIST_INDEX 0x%x", last_idx);
+	}
+
+	/* print the asserts */
+	for (i = 0; i < STORM_ASSERT_ARRAY_SIZE; i++) {
+
+		row0 =
+		    REG_RD(sc,
+			   BAR_CSTRORM_INTMEM + CSTORM_ASSERT_LIST_OFFSET(i));
+		row1 =
+		    REG_RD(sc,
+			   BAR_CSTRORM_INTMEM + CSTORM_ASSERT_LIST_OFFSET(i) +
+			   4);
+		row2 =
+		    REG_RD(sc,
+			   BAR_CSTRORM_INTMEM + CSTORM_ASSERT_LIST_OFFSET(i) +
+			   8);
+		row3 =
+		    REG_RD(sc,
+			   BAR_CSTRORM_INTMEM + CSTORM_ASSERT_LIST_OFFSET(i) +
+			   12);
+
+		if (row0 != COMMON_ASM_INVALID_ASSERT_OPCODE) {
+			PMD_DRV_LOG(ERR,
+				    "CSTORM_ASSERT_INDEX 0x%x = 0x%08x 0x%08x 0x%08x 0x%08x",
+				    i, row3, row2, row1, row0);
+			rc++;
+		} else {
+			break;
+		}
+	}
+
+	/* USTORM */
+	last_idx =
+	    REG_RD8(sc, BAR_USTRORM_INTMEM + USTORM_ASSERT_LIST_INDEX_OFFSET);
+	if (last_idx) {
+		PMD_DRV_LOG(ERR, "USTORM_ASSERT_LIST_INDEX 0x%x", last_idx);
+	}
+
+	/* print the asserts */
+	for (i = 0; i < STORM_ASSERT_ARRAY_SIZE; i++) {
+
+		row0 =
+		    REG_RD(sc,
+			   BAR_USTRORM_INTMEM + USTORM_ASSERT_LIST_OFFSET(i));
+		row1 =
+		    REG_RD(sc,
+			   BAR_USTRORM_INTMEM + USTORM_ASSERT_LIST_OFFSET(i) +
+			   4);
+		row2 =
+		    REG_RD(sc,
+			   BAR_USTRORM_INTMEM + USTORM_ASSERT_LIST_OFFSET(i) +
+			   8);
+		row3 =
+		    REG_RD(sc,
+			   BAR_USTRORM_INTMEM + USTORM_ASSERT_LIST_OFFSET(i) +
+			   12);
+
+		if (row0 != COMMON_ASM_INVALID_ASSERT_OPCODE) {
+			PMD_DRV_LOG(ERR,
+				    "USTORM_ASSERT_INDEX 0x%x = 0x%08x 0x%08x 0x%08x 0x%08x",
+				    i, row3, row2, row1, row0);
+			rc++;
+		} else {
+			break;
+		}
+	}
+
+	return rc;
+}
+
+static void bcm_attn_int_deasserted3(struct bcm_softc *sc, uint32_t attn)
+{
+	int func = SC_FUNC(sc);
+	uint32_t val;
+
+	if (attn & EVEREST_GEN_ATTN_IN_USE_MASK) {
+
+		if (attn & BCM_PMF_LINK_ASSERT(sc)) {
+
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_12 + func * 4, 0);
+			bcm_read_mf_cfg(sc);
+			sc->devinfo.mf_info.mf_config[SC_VN(sc)] =
+			    MFCFG_RD(sc,
+				     func_mf_config[SC_ABS_FUNC(sc)].config);
+			val =
+			    SHMEM_RD(sc, func_mb[SC_FW_MB_IDX(sc)].drv_status);
+
+			if (val & DRV_STATUS_DCC_EVENT_MASK)
+				bcm_dcc_event(sc,
+					      (val &
+					       DRV_STATUS_DCC_EVENT_MASK));
+
+			if (val & DRV_STATUS_SET_MF_BW)
+				bcm_set_mf_bw(sc);
+
+			if (val & DRV_STATUS_DRV_INFO_REQ)
+				bcm_handle_drv_info_req(sc);
+
+			if ((sc->port.pmf == 0) && (val & DRV_STATUS_PMF))
+				bcm_pmf_update(sc);
+
+			if (val & DRV_STATUS_EEE_NEGOTIATION_RESULTS)
+				bcm_handle_eee_event(sc);
+
+			if (sc->link_vars.periodic_flags &
+			    ELINK_PERIODIC_FLAGS_LINK_EVENT) {
+				/* sync with link */
+				sc->link_vars.periodic_flags &=
+				    ~ELINK_PERIODIC_FLAGS_LINK_EVENT;
+				if (IS_MF(sc)) {
+					bcm_link_sync_notify(sc);
+				}
+				bcm_link_report(sc);
+			}
+
+			/*
+			 * Always call it here: bcm_link_report() will
+			 * prevent the link indication duplication.
+			 */
+			bcm_link_status_update(sc);
+
+		} else if (attn & BCM_MC_ASSERT_BITS) {
+
+			PMD_DRV_LOG(ERR, "MC assert!");
+			bcm_mc_assert(sc);
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_10, 0);
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_9, 0);
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_8, 0);
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_7, 0);
+			rte_panic("MC assert!");
+
+		} else if (attn & BCM_MCP_ASSERT) {
+
+			PMD_DRV_LOG(ERR, "MCP assert!");
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_11, 0);
+
+		} else {
+			PMD_DRV_LOG(ERR,
+				    "Unknown HW assert! (attn 0x%08x)", attn);
+		}
+	}
+
+	if (attn & EVEREST_LATCHED_ATTN_IN_USE_MASK) {
+		PMD_DRV_LOG(ERR, "LATCHED attention 0x%08x (masked)", attn);
+		if (attn & BCM_GRC_TIMEOUT) {
+			val = REG_RD(sc, MISC_REG_GRC_TIMEOUT_ATTN);
+			PMD_DRV_LOG(ERR, "GRC time-out 0x%08x", val);
+		}
+		if (attn & BCM_GRC_RSV) {
+			val = REG_RD(sc, MISC_REG_GRC_RSV_ATTN);
+			PMD_DRV_LOG(ERR, "GRC reserved 0x%08x", val);
+		}
+		REG_WR(sc, MISC_REG_AEU_CLR_LATCH_SIGNAL, 0x7ff);
+	}
+}
+
+static void bcm_attn_int_deasserted2(struct bcm_softc *sc, uint32_t attn)
+{
+	int port = SC_PORT(sc);
+	int reg_offset;
+	uint32_t val0, mask0, val1, mask1;
+	uint32_t val;
+
+	if (attn & AEU_INPUTS_ATTN_BITS_CFC_HW_INTERRUPT) {
+		val = REG_RD(sc, CFC_REG_CFC_INT_STS_CLR);
+		PMD_DRV_LOG(ERR, "CFC hw attention 0x%08x", val);
+/* CFC error attention */
+		if (val & 0x2) {
+			PMD_DRV_LOG(ERR, "FATAL error from CFC");
+		}
+	}
+
+	if (attn & AEU_INPUTS_ATTN_BITS_PXP_HW_INTERRUPT) {
+		val = REG_RD(sc, PXP_REG_PXP_INT_STS_CLR_0);
+		PMD_DRV_LOG(ERR, "PXP hw attention-0 0x%08x", val);
+/* RQ_USDMDP_FIFO_OVERFLOW */
+		if (val & 0x18000) {
+			PMD_DRV_LOG(ERR, "FATAL error from PXP");
+		}
+
+		if (!CHIP_IS_E1x(sc)) {
+			val = REG_RD(sc, PXP_REG_PXP_INT_STS_CLR_1);
+			PMD_DRV_LOG(ERR, "PXP hw attention-1 0x%08x", val);
+		}
+	}
+#define PXP2_EOP_ERROR_BIT  PXP2_PXP2_INT_STS_CLR_0_REG_WR_PGLUE_EOP_ERROR
+#define AEU_PXP2_HW_INT_BIT AEU_INPUTS_ATTN_BITS_PXPPCICLOCKCLIENT_HW_INTERRUPT
+
+	if (attn & AEU_PXP2_HW_INT_BIT) {
+/*  CQ47854 workaround do not panic on
+ *  PXP2_PXP2_INT_STS_0_REG_WR_PGLUE_EOP_ERROR
+ */
+		if (!CHIP_IS_E1x(sc)) {
+			mask0 = REG_RD(sc, PXP2_REG_PXP2_INT_MASK_0);
+			val1 = REG_RD(sc, PXP2_REG_PXP2_INT_STS_1);
+			mask1 = REG_RD(sc, PXP2_REG_PXP2_INT_MASK_1);
+			val0 = REG_RD(sc, PXP2_REG_PXP2_INT_STS_0);
+			/*
+			 * If the olny PXP2_EOP_ERROR_BIT is set in
+			 * STS0 and STS1 - clear it
+			 *
+			 * probably we lose additional attentions between
+			 * STS0 and STS_CLR0, in this case user will not
+			 * be notified about them
+			 */
+			if (val0 & mask0 & PXP2_EOP_ERROR_BIT &&
+			    !(val1 & mask1))
+				val0 = REG_RD(sc, PXP2_REG_PXP2_INT_STS_CLR_0);
+
+			/* print the register, since no one can restore it */
+			PMD_DRV_LOG(ERR,
+				    "PXP2_REG_PXP2_INT_STS_CLR_0 0x%08x", val0);
+
+			/*
+			 * if PXP2_PXP2_INT_STS_0_REG_WR_PGLUE_EOP_ERROR
+			 * then notify
+			 */
+			if (val0 & PXP2_EOP_ERROR_BIT) {
+				PMD_DRV_LOG(ERR, "PXP2_WR_PGLUE_EOP_ERROR");
+
+				/*
+				 * if only PXP2_PXP2_INT_STS_0_REG_WR_PGLUE_EOP_ERROR is
+				 * set then clear attention from PXP2 block without panic
+				 */
+				if (((val0 & mask0) == PXP2_EOP_ERROR_BIT) &&
+				    ((val1 & mask1) == 0))
+					attn &= ~AEU_PXP2_HW_INT_BIT;
+			}
+		}
+	}
+
+	if (attn & HW_INTERRUT_ASSERT_SET_2) {
+		reg_offset = (port ? MISC_REG_AEU_ENABLE1_FUNC_1_OUT_2 :
+			      MISC_REG_AEU_ENABLE1_FUNC_0_OUT_2);
+
+		val = REG_RD(sc, reg_offset);
+		val &= ~(attn & HW_INTERRUT_ASSERT_SET_2);
+		REG_WR(sc, reg_offset, val);
+
+		PMD_DRV_LOG(ERR,
+			    "FATAL HW block attention set2 0x%x",
+			    (uint32_t) (attn & HW_INTERRUT_ASSERT_SET_2));
+		rte_panic("HW block attention set2");
+	}
+}
+
+static void bcm_attn_int_deasserted1(struct bcm_softc *sc, uint32_t attn)
+{
+	int port = SC_PORT(sc);
+	int reg_offset;
+	uint32_t val;
+
+	if (attn & AEU_INPUTS_ATTN_BITS_DOORBELLQ_HW_INTERRUPT) {
+		val = REG_RD(sc, DORQ_REG_DORQ_INT_STS_CLR);
+		PMD_DRV_LOG(ERR, "DB hw attention 0x%08x", val);
+/* DORQ discard attention */
+		if (val & 0x2) {
+			PMD_DRV_LOG(ERR, "FATAL error from DORQ");
+		}
+	}
+
+	if (attn & HW_INTERRUT_ASSERT_SET_1) {
+		reg_offset = (port ? MISC_REG_AEU_ENABLE1_FUNC_1_OUT_1 :
+			      MISC_REG_AEU_ENABLE1_FUNC_0_OUT_1);
+
+		val = REG_RD(sc, reg_offset);
+		val &= ~(attn & HW_INTERRUT_ASSERT_SET_1);
+		REG_WR(sc, reg_offset, val);
+
+		PMD_DRV_LOG(ERR,
+			    "FATAL HW block attention set1 0x%08x",
+			    (uint32_t) (attn & HW_INTERRUT_ASSERT_SET_1));
+		rte_panic("HW block attention set1");
+	}
+}
+
+static void bcm_attn_int_deasserted0(struct bcm_softc *sc, uint32_t attn)
+{
+	int port = SC_PORT(sc);
+	int reg_offset;
+	uint32_t val;
+
+	reg_offset = (port) ? MISC_REG_AEU_ENABLE1_FUNC_1_OUT_0 :
+	    MISC_REG_AEU_ENABLE1_FUNC_0_OUT_0;
+
+	if (attn & AEU_INPUTS_ATTN_BITS_SPIO5) {
+		val = REG_RD(sc, reg_offset);
+		val &= ~AEU_INPUTS_ATTN_BITS_SPIO5;
+		REG_WR(sc, reg_offset, val);
+
+		PMD_DRV_LOG(WARN, "SPIO5 hw attention");
+
+/* Fan failure attention */
+		elink_hw_reset_phy(&sc->link_params);
+		bcm_fan_failure(sc);
+	}
+
+	if ((attn & sc->link_vars.aeu_int_mask) && sc->port.pmf) {
+		elink_handle_module_detect_int(&sc->link_params);
+	}
+
+	if (attn & HW_INTERRUT_ASSERT_SET_0) {
+		val = REG_RD(sc, reg_offset);
+		val &= ~(attn & HW_INTERRUT_ASSERT_SET_0);
+		REG_WR(sc, reg_offset, val);
+
+		rte_panic("FATAL HW block attention set0 0x%lx",
+			  (attn & HW_INTERRUT_ASSERT_SET_0));
+	}
+}
+
+static void bcm_attn_int_deasserted(struct bcm_softc *sc, uint32_t deasserted)
+{
+	struct attn_route attn;
+	struct attn_route *group_mask;
+	int port = SC_PORT(sc);
+	int index;
+	uint32_t reg_addr;
+	uint32_t val;
+	uint32_t aeu_mask;
+	uint8_t global = FALSE;
+
+	/*
+	 * Need to take HW lock because MCP or other port might also
+	 * try to handle this event.
+	 */
+	bcm_acquire_alr(sc);
+
+	if (bcm_chk_parity_attn(sc, &global, TRUE)) {
+		sc->recovery_state = BCM_RECOVERY_INIT;
+
+/* disable HW interrupts */
+		bcm_int_disable(sc);
+		bcm_release_alr(sc);
+		return;
+	}
+
+	attn.sig[0] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_1_FUNC_0 + port * 4);
+	attn.sig[1] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_2_FUNC_0 + port * 4);
+	attn.sig[2] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_3_FUNC_0 + port * 4);
+	attn.sig[3] = REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_4_FUNC_0 + port * 4);
+	if (!CHIP_IS_E1x(sc)) {
+		attn.sig[4] =
+		    REG_RD(sc, MISC_REG_AEU_AFTER_INVERT_5_FUNC_0 + port * 4);
+	} else {
+		attn.sig[4] = 0;
+	}
+
+	for (index = 0; index < MAX_DYNAMIC_ATTN_GRPS; index++) {
+		if (deasserted & (1 << index)) {
+			group_mask = &sc->attn_group[index];
+
+			bcm_attn_int_deasserted4(sc,
+						 attn.
+						 sig[4] & group_mask->sig[4]);
+			bcm_attn_int_deasserted3(sc,
+						 attn.
+						 sig[3] & group_mask->sig[3]);
+			bcm_attn_int_deasserted1(sc,
+						 attn.
+						 sig[1] & group_mask->sig[1]);
+			bcm_attn_int_deasserted2(sc,
+						 attn.
+						 sig[2] & group_mask->sig[2]);
+			bcm_attn_int_deasserted0(sc,
+						 attn.
+						 sig[0] & group_mask->sig[0]);
+		}
+	}
+
+	bcm_release_alr(sc);
+
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		reg_addr = (HC_REG_COMMAND_REG + port * 32 +
+			    COMMAND_REG_ATTN_BITS_CLR);
+	} else {
+		reg_addr = (BAR_IGU_INTMEM + IGU_CMD_ATTN_BIT_CLR_UPPER * 8);
+	}
+
+	val = ~deasserted;
+	PMD_DRV_LOG(DEBUG,
+		    "about to mask 0x%08x at %s addr 0x%08x", val,
+		    (sc->devinfo.int_block == INT_BLOCK_HC) ? "HC" : "IGU",
+		    reg_addr);
+	REG_WR(sc, reg_addr, val);
+
+	if (~sc->attn_state & deasserted) {
+		PMD_DRV_LOG(ERR, "IGU error");
+	}
+
+	reg_addr = port ? MISC_REG_AEU_MASK_ATTN_FUNC_1 :
+	    MISC_REG_AEU_MASK_ATTN_FUNC_0;
+
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_PORT0_ATT_MASK + port);
+
+	aeu_mask = REG_RD(sc, reg_addr);
+
+	aeu_mask |= (deasserted & 0x3ff);
+
+	REG_WR(sc, reg_addr, aeu_mask);
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_PORT0_ATT_MASK + port);
+
+	sc->attn_state &= ~deasserted;
+}
+
+static void bcm_attn_int(struct bcm_softc *sc)
+{
+	/* read local copy of bits */
+	uint32_t attn_bits = le32toh(sc->def_sb->atten_status_block.attn_bits);
+	uint32_t attn_ack =
+	    le32toh(sc->def_sb->atten_status_block.attn_bits_ack);
+	uint32_t attn_state = sc->attn_state;
+
+	/* look for changed bits */
+	uint32_t asserted = attn_bits & ~attn_ack & ~attn_state;
+	uint32_t deasserted = ~attn_bits & attn_ack & attn_state;
+
+	PMD_DRV_LOG(DEBUG,
+		    "attn_bits 0x%08x attn_ack 0x%08x asserted 0x%08x deasserted 0x%08x",
+		    attn_bits, attn_ack, asserted, deasserted);
+
+	if (~(attn_bits ^ attn_ack) & (attn_bits ^ attn_state)) {
+		PMD_DRV_LOG(ERR, "BAD attention state");
+	}
+
+	/* handle bits that were raised */
+	if (asserted) {
+		bcm_attn_int_asserted(sc, asserted);
+	}
+
+	if (deasserted) {
+		bcm_attn_int_deasserted(sc, deasserted);
+	}
+}
+
+static uint16_t bcm_update_dsb_idx(struct bcm_softc *sc)
+{
+	struct host_sp_status_block *def_sb = sc->def_sb;
+	uint16_t rc = 0;
+
+	mb();			/* status block is written to by the chip */
+
+	if (sc->def_att_idx != def_sb->atten_status_block.attn_bits_index) {
+		sc->def_att_idx = def_sb->atten_status_block.attn_bits_index;
+		rc |= BCM_DEF_SB_ATT_IDX;
+	}
+
+	if (sc->def_idx != def_sb->sp_sb.running_index) {
+		sc->def_idx = def_sb->sp_sb.running_index;
+		rc |= BCM_DEF_SB_IDX;
+	}
+
+	mb();
+
+	return rc;
+}
+
+static struct ecore_queue_sp_obj *bcm_cid_to_q_obj(struct bcm_softc *sc,
+							  uint32_t cid)
+{
+	return &sc->sp_objs[CID_TO_FP(cid, sc)].q_obj;
+}
+
+static void bcm_handle_mcast_eqe(struct bcm_softc *sc)
+{
+	struct ecore_mcast_ramrod_params rparam;
+	int rc;
+
+	memset(&rparam, 0, sizeof(rparam));
+
+	rparam.mcast_obj = &sc->mcast_obj;
+
+	/* clear pending state for the last command */
+	sc->mcast_obj.raw.clear_pending(&sc->mcast_obj.raw);
+
+	/* if there are pending mcast commands - send them */
+	if (sc->mcast_obj.check_pending(&sc->mcast_obj)) {
+		rc = ecore_config_mcast(sc, &rparam, ECORE_MCAST_CMD_CONT);
+		if (rc < 0) {
+			PMD_DRV_LOG(INFO,
+				    "Failed to send pending mcast commands (%d)",
+				    rc);
+		}
+	}
+}
+
+static void
+bcm_handle_classification_eqe(struct bcm_softc *sc, union event_ring_elem *elem)
+{
+	unsigned long ramrod_flags = 0;
+	int rc = 0;
+	uint32_t cid = elem->message.data.eth_event.echo & BCM_SWCID_MASK;
+	struct ecore_vlan_mac_obj *vlan_mac_obj;
+
+	/* always push next commands out, don't wait here */
+	bcm_set_bit(RAMROD_CONT, &ramrod_flags);
+
+	switch (le32toh(elem->message.data.eth_event.echo) >> BCM_SWCID_SHIFT) {
+	case ECORE_FILTER_MAC_PENDING:
+		PMD_DRV_LOG(DEBUG, "Got SETUP_MAC completions");
+		vlan_mac_obj = &sc->sp_objs[cid].mac_obj;
+		break;
+
+	case ECORE_FILTER_MCAST_PENDING:
+		PMD_DRV_LOG(DEBUG, "Got SETUP_MCAST completions");
+		bcm_handle_mcast_eqe(sc);
+		return;
+
+	default:
+		PMD_DRV_LOG(NOTICE, "Unsupported classification command: %d",
+			    elem->message.data.eth_event.echo);
+		return;
+	}
+
+	rc = vlan_mac_obj->complete(sc, vlan_mac_obj, elem, &ramrod_flags);
+
+	if (rc < 0) {
+		PMD_DRV_LOG(NOTICE, "Failed to schedule new commands (%d)", rc);
+	} else if (rc > 0) {
+		PMD_DRV_LOG(DEBUG, "Scheduled next pending commands...");
+	}
+}
+
+static void bcm_handle_rx_mode_eqe(struct bcm_softc *sc)
+{
+	bcm_clear_bit(ECORE_FILTER_RX_MODE_PENDING, &sc->sp_state);
+
+	/* send rx_mode command again if was requested */
+	if (bcm_test_and_clear_bit(ECORE_FILTER_RX_MODE_SCHED, &sc->sp_state)) {
+		bcm_set_storm_rx_mode(sc);
+	}
+}
+
+static void bcm_update_eq_prod(struct bcm_softc *sc, uint16_t prod)
+{
+	storm_memset_eq_prod(sc, prod, SC_FUNC(sc));
+	wmb();			/* keep prod updates ordered */
+}
+
+static void bcm_eq_int(struct bcm_softc *sc)
+{
+	uint16_t hw_cons, sw_cons, sw_prod;
+	union event_ring_elem *elem;
+	uint8_t echo;
+	uint32_t cid;
+	uint8_t opcode;
+	int spqe_cnt = 0;
+	struct ecore_queue_sp_obj *q_obj;
+	struct ecore_func_sp_obj *f_obj = &sc->func_obj;
+	struct ecore_raw_obj *rss_raw = &sc->rss_conf_obj.raw;
+
+	hw_cons = le16toh(*sc->eq_cons_sb);
+
+	/*
+	 * The hw_cons range is 1-255, 257 - the sw_cons range is 0-254, 256.
+	 * when we get to the next-page we need to adjust so the loop
+	 * condition below will be met. The next element is the size of a
+	 * regular element and hence incrementing by 1
+	 */
+	if ((hw_cons & EQ_DESC_MAX_PAGE) == EQ_DESC_MAX_PAGE) {
+		hw_cons++;
+	}
+
+	/*
+	 * This function may never run in parallel with itself for a
+	 * specific sc and no need for a read memory barrier here.
+	 */
+	sw_cons = sc->eq_cons;
+	sw_prod = sc->eq_prod;
+
+	for (;
+	     sw_cons != hw_cons;
+	     sw_prod = NEXT_EQ_IDX(sw_prod), sw_cons = NEXT_EQ_IDX(sw_cons)) {
+
+		elem = &sc->eq[EQ_DESC(sw_cons)];
+
+/* elem CID originates from FW, actually LE */
+		cid = SW_CID(elem->message.data.cfc_del_event.cid);
+		opcode = elem->message.opcode;
+
+/* handle eq element */
+		switch (opcode) {
+		case EVENT_RING_OPCODE_STAT_QUERY:
+			PMD_DRV_LOG(DEBUG, "got statistics completion event %d",
+				    sc->stats_comp++);
+			/* nothing to do with stats comp */
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_CFC_DEL:
+			/* handle according to cid range */
+			/* we may want to verify here that the sc state is HALTING */
+			PMD_DRV_LOG(DEBUG, "got delete ramrod for MULTI[%d]",
+				    cid);
+			q_obj = bcm_cid_to_q_obj(sc, cid);
+			if (q_obj->complete_cmd(sc, q_obj, ECORE_Q_CMD_CFC_DEL)) {
+				break;
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_STOP_TRAFFIC:
+			PMD_DRV_LOG(DEBUG, "got STOP TRAFFIC");
+			if (f_obj->complete_cmd(sc, f_obj, ECORE_F_CMD_TX_STOP)) {
+				break;
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_START_TRAFFIC:
+			PMD_DRV_LOG(DEBUG, "got START TRAFFIC");
+			if (f_obj->complete_cmd
+			    (sc, f_obj, ECORE_F_CMD_TX_START)) {
+				break;
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_FUNCTION_UPDATE:
+			echo = elem->message.data.function_update_event.echo;
+			if (echo == SWITCH_UPDATE) {
+				PMD_DRV_LOG(DEBUG,
+					    "got FUNC_SWITCH_UPDATE ramrod");
+				if (f_obj->complete_cmd(sc, f_obj,
+							ECORE_F_CMD_SWITCH_UPDATE))
+				{
+					break;
+				}
+			} else {
+				PMD_DRV_LOG(DEBUG,
+					    "AFEX: ramrod completed FUNCTION_UPDATE");
+				f_obj->complete_cmd(sc, f_obj,
+						    ECORE_F_CMD_AFEX_UPDATE);
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_FORWARD_SETUP:
+			q_obj = &bcm_fwd_sp_obj(sc, q_obj);
+			if (q_obj->complete_cmd(sc, q_obj,
+						ECORE_Q_CMD_SETUP_TX_ONLY)) {
+				break;
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_FUNCTION_START:
+			PMD_DRV_LOG(DEBUG, "got FUNC_START ramrod");
+			if (f_obj->complete_cmd(sc, f_obj, ECORE_F_CMD_START)) {
+				break;
+			}
+			goto next_spqe;
+
+		case EVENT_RING_OPCODE_FUNCTION_STOP:
+			PMD_DRV_LOG(DEBUG, "got FUNC_STOP ramrod");
+			if (f_obj->complete_cmd(sc, f_obj, ECORE_F_CMD_STOP)) {
+				break;
+			}
+			goto next_spqe;
+		}
+
+		switch (opcode | sc->state) {
+		case (EVENT_RING_OPCODE_RSS_UPDATE_RULES | BCM_STATE_OPEN):
+		case (EVENT_RING_OPCODE_RSS_UPDATE_RULES | BCM_STATE_OPENING_WAITING_PORT):
+			cid =
+			    elem->message.data.eth_event.echo & BCM_SWCID_MASK;
+			PMD_DRV_LOG(DEBUG, "got RSS_UPDATE ramrod. CID %d",
+				    cid);
+			rss_raw->clear_pending(rss_raw);
+			break;
+
+		case (EVENT_RING_OPCODE_SET_MAC | BCM_STATE_OPEN):
+		case (EVENT_RING_OPCODE_SET_MAC | BCM_STATE_DIAG):
+		case (EVENT_RING_OPCODE_SET_MAC | BCM_STATE_CLOSING_WAITING_HALT):
+		case (EVENT_RING_OPCODE_CLASSIFICATION_RULES | BCM_STATE_OPEN):
+		case (EVENT_RING_OPCODE_CLASSIFICATION_RULES | BCM_STATE_DIAG):
+		case (EVENT_RING_OPCODE_CLASSIFICATION_RULES | BCM_STATE_CLOSING_WAITING_HALT):
+			PMD_DRV_LOG(DEBUG,
+				    "got (un)set mac ramrod");
+			bcm_handle_classification_eqe(sc, elem);
+			break;
+
+		case (EVENT_RING_OPCODE_MULTICAST_RULES | BCM_STATE_OPEN):
+		case (EVENT_RING_OPCODE_MULTICAST_RULES | BCM_STATE_DIAG):
+		case (EVENT_RING_OPCODE_MULTICAST_RULES | BCM_STATE_CLOSING_WAITING_HALT):
+			PMD_DRV_LOG(DEBUG,
+				    "got mcast ramrod");
+			bcm_handle_mcast_eqe(sc);
+			break;
+
+		case (EVENT_RING_OPCODE_FILTERS_RULES | BCM_STATE_OPEN):
+		case (EVENT_RING_OPCODE_FILTERS_RULES | BCM_STATE_DIAG):
+		case (EVENT_RING_OPCODE_FILTERS_RULES | BCM_STATE_CLOSING_WAITING_HALT):
+			PMD_DRV_LOG(DEBUG,
+				    "got rx_mode ramrod");
+			bcm_handle_rx_mode_eqe(sc);
+			break;
+
+		default:
+			/* unknown event log error and continue */
+			PMD_DRV_LOG(INFO, "Unknown EQ event %d, sc->state 0x%x",
+				    elem->message.opcode, sc->state);
+		}
+
+next_spqe:
+		spqe_cnt++;
+	}			/* for */
+
+	mb();
+	atomic_add_acq_long(&sc->eq_spq_left, spqe_cnt);
+
+	sc->eq_cons = sw_cons;
+	sc->eq_prod = sw_prod;
+
+	/* make sure that above mem writes were issued towards the memory */
+	wmb();
+
+	/* update producer */
+	bcm_update_eq_prod(sc, sc->eq_prod);
+}
+
+static int bcm_handle_sp_tq(struct bcm_softc *sc)
+{
+	uint16_t status;
+	int rc = 0;
+
+	/* what work needs to be performed? */
+	status = bcm_update_dsb_idx(sc);
+
+	/* HW attentions */
+	if (status & BCM_DEF_SB_ATT_IDX) {
+		PMD_DRV_LOG(DEBUG, "---> ATTN INTR <---");
+		bcm_attn_int(sc);
+		status &= ~BCM_DEF_SB_ATT_IDX;
+		rc = 1;
+	}
+
+	/* SP events: STAT_QUERY and others */
+	if (status & BCM_DEF_SB_IDX) {
+/* handle EQ completions */
+		PMD_DRV_LOG(DEBUG, "---> EQ INTR <---");
+		bcm_eq_int(sc);
+		bcm_ack_sb(sc, sc->igu_dsb_id, USTORM_ID,
+			   le16toh(sc->def_idx), IGU_INT_NOP, 1);
+		status &= ~BCM_DEF_SB_IDX;
+	}
+
+	/* if status is non zero then something went wrong */
+	if (unlikely(status)) {
+		PMD_DRV_LOG(INFO,
+			    "Got an unknown SP interrupt! (0x%04x)", status);
+	}
+
+	/* ack status block only if something was actually handled */
+	bcm_ack_sb(sc, sc->igu_dsb_id, ATTENTION_ID,
+		   le16toh(sc->def_att_idx), IGU_INT_ENABLE, 1);
+
+	return rc;
+}
+
+static void bcm_handle_fp_tq(struct bcm_fastpath *fp, int scan_fp)
+{
+	struct bcm_softc *sc = fp->sc;
+	uint8_t more_rx = FALSE;
+
+	/* update the fastpath index */
+	bcm_update_fp_sb_idx(fp);
+
+	if (scan_fp) {
+		if (bcm_has_rx_work(fp)) {
+			more_rx = bcm_rxeof(sc, fp);
+		}
+
+		if (more_rx) {
+			/* still more work to do */
+			bcm_handle_fp_tq(fp, scan_fp);
+			return;
+		}
+	}
+
+	bcm_ack_sb(sc, fp->igu_sb_id, USTORM_ID,
+		   le16toh(fp->fp_hc_idx), IGU_INT_ENABLE, 1);
+}
+
+/*
+ * Legacy interrupt entry point.
+ *
+ * Verifies that the controller generated the interrupt and
+ * then calls a separate routine to handle the various
+ * interrupt causes: link, RX, and TX.
+ */
+int bcm_intr_legacy(struct bcm_softc *sc, int scan_fp)
+{
+	struct bcm_fastpath *fp;
+	uint32_t status, mask;
+	int i, rc = 0;
+
+	/*
+	 * 0 for ustorm, 1 for cstorm
+	 * the bits returned from ack_int() are 0-15
+	 * bit 0 = attention status block
+	 * bit 1 = fast path status block
+	 * a mask of 0x2 or more = tx/rx event
+	 * a mask of 1 = slow path event
+	 */
+
+	status = bcm_ack_int(sc);
+
+	/* the interrupt is not for us */
+	if (unlikely(status == 0)) {
+		return 0;
+	}
+
+	PMD_DRV_LOG(DEBUG, "Interrupt status 0x%04x", status);
+	//bcm_dump_status_block(sc);
+
+	FOR_EACH_ETH_QUEUE(sc, i) {
+		fp = &sc->fp[i];
+		mask = (0x2 << (fp->index + CNIC_SUPPORT(sc)));
+		if (status & mask) {
+			bcm_handle_fp_tq(fp, scan_fp);
+			status &= ~mask;
+		}
+	}
+
+	if (unlikely(status & 0x1)) {
+		rc = bcm_handle_sp_tq(sc);
+		status &= ~0x1;
+	}
+
+	if (unlikely(status)) {
+		PMD_DRV_LOG(WARN,
+			    "Unexpected fastpath status (0x%08x)!", status);
+	}
+
+	return rc;
+}
+
+static int bcm_init_hw_common_chip(struct bcm_softc *sc);
+static int bcm_init_hw_common(struct bcm_softc *sc);
+static int bcm_init_hw_port(struct bcm_softc *sc);
+static int bcm_init_hw_func(struct bcm_softc *sc);
+static void bcm_reset_common(struct bcm_softc *sc);
+static void bcm_reset_port(struct bcm_softc *sc);
+static void bcm_reset_func(struct bcm_softc *sc);
+static int bcm_init_firmware(struct bcm_softc *sc);
+static void bcm_release_firmware(struct bcm_softc *sc);
+
+static struct
+ecore_func_sp_drv_ops bcm_func_sp_drv = {
+	.init_hw_cmn_chip = bcm_init_hw_common_chip,
+	.init_hw_cmn = bcm_init_hw_common,
+	.init_hw_port = bcm_init_hw_port,
+	.init_hw_func = bcm_init_hw_func,
+
+	.reset_hw_cmn = bcm_reset_common,
+	.reset_hw_port = bcm_reset_port,
+	.reset_hw_func = bcm_reset_func,
+
+	.init_fw = bcm_init_firmware,
+	.release_fw = bcm_release_firmware,
+};
+
+static void bcm_init_func_obj(struct bcm_softc *sc)
+{
+	sc->dmae_ready = 0;
+
+	PMD_INIT_FUNC_TRACE();
+
+	ecore_init_func_obj(sc,
+			    &sc->func_obj,
+			    BCM_SP(sc, func_rdata), (phys_addr_t) ((void *)
+								   BCM_SP_MAPPING(sc, func_rdata)), BCM_SP(sc, func_afex_rdata), (phys_addr_t) ((void *)
+																		BCM_SP_MAPPING(sc, func_afex_rdata)), &bcm_func_sp_drv);
+}
+
+static int bcm_init_hw(struct bcm_softc *sc, uint32_t load_code)
+{
+	struct ecore_func_state_params func_params = { NULL };
+	int rc;
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* prepare the parameters for function state transitions */
+	bcm_set_bit(RAMROD_COMP_WAIT, &func_params.ramrod_flags);
+
+	func_params.f_obj = &sc->func_obj;
+	func_params.cmd = ECORE_F_CMD_HW_INIT;
+
+	func_params.params.hw_init.load_phase = load_code;
+
+	/*
+	 * Via a plethora of function pointers, we will eventually reach
+	 * bcm_init_hw_common(), bcm_init_hw_port(), or bcm_init_hw_func().
+	 */
+	rc = ecore_func_state_change(sc, &func_params);
+
+	return rc;
+}
+
+static void
+bcm_fill(struct bcm_softc *sc, uint32_t addr, int fill, uint32_t len)
+{
+	uint32_t i;
+
+	if (!(len % 4) && !(addr % 4)) {
+		for (i = 0; i < len; i += 4) {
+			REG_WR(sc, (addr + i), fill);
+		}
+	} else {
+		for (i = 0; i < len; i++) {
+			REG_WR8(sc, (addr + i), fill);
+		}
+	}
+}
+
+/* writes FP SP data to FW - data_size in dwords */
+static void
+bcm_wr_fp_sb_data(struct bcm_softc *sc, int fw_sb_id, uint32_t * sb_data_p,
+		  uint32_t data_size)
+{
+	uint32_t index;
+
+	for (index = 0; index < data_size; index++) {
+		REG_WR(sc,
+		       (BAR_CSTRORM_INTMEM +
+			CSTORM_STATUS_BLOCK_DATA_OFFSET(fw_sb_id) +
+			(sizeof(uint32_t) * index)), *(sb_data_p + index));
+	}
+}
+
+static void bcm_zero_fp_sb(struct bcm_softc *sc, int fw_sb_id)
+{
+	struct hc_status_block_data_e2 sb_data_e2;
+	struct hc_status_block_data_e1x sb_data_e1x;
+	uint32_t *sb_data_p;
+	uint32_t data_size = 0;
+
+	if (!CHIP_IS_E1x(sc)) {
+		memset(&sb_data_e2, 0, sizeof(struct hc_status_block_data_e2));
+		sb_data_e2.common.state = SB_DISABLED;
+		sb_data_e2.common.p_func.vf_valid = FALSE;
+		sb_data_p = (uint32_t *) & sb_data_e2;
+		data_size = (sizeof(struct hc_status_block_data_e2) /
+			     sizeof(uint32_t));
+	} else {
+		memset(&sb_data_e1x, 0,
+		       sizeof(struct hc_status_block_data_e1x));
+		sb_data_e1x.common.state = SB_DISABLED;
+		sb_data_e1x.common.p_func.vf_valid = FALSE;
+		sb_data_p = (uint32_t *) & sb_data_e1x;
+		data_size = (sizeof(struct hc_status_block_data_e1x) /
+			     sizeof(uint32_t));
+	}
+
+	bcm_wr_fp_sb_data(sc, fw_sb_id, sb_data_p, data_size);
+
+	bcm_fill(sc,
+		 (BAR_CSTRORM_INTMEM + CSTORM_STATUS_BLOCK_OFFSET(fw_sb_id)), 0,
+		 CSTORM_STATUS_BLOCK_SIZE);
+	bcm_fill(sc, (BAR_CSTRORM_INTMEM + CSTORM_SYNC_BLOCK_OFFSET(fw_sb_id)),
+		 0, CSTORM_SYNC_BLOCK_SIZE);
+}
+
+static void
+bcm_wr_sp_sb_data(struct bcm_softc *sc,
+		  struct hc_sp_status_block_data *sp_sb_data)
+{
+	uint32_t i;
+
+	for (i = 0;
+	     i < (sizeof(struct hc_sp_status_block_data) / sizeof(uint32_t));
+	     i++) {
+		REG_WR(sc,
+		       (BAR_CSTRORM_INTMEM +
+			CSTORM_SP_STATUS_BLOCK_DATA_OFFSET(SC_FUNC(sc)) +
+			(i * sizeof(uint32_t))),
+		       *((uint32_t *) sp_sb_data + i));
+	}
+}
+
+static void bcm_zero_sp_sb(struct bcm_softc *sc)
+{
+	struct hc_sp_status_block_data sp_sb_data;
+
+	memset(&sp_sb_data, 0, sizeof(struct hc_sp_status_block_data));
+
+	sp_sb_data.state = SB_DISABLED;
+	sp_sb_data.p_func.vf_valid = FALSE;
+
+	bcm_wr_sp_sb_data(sc, &sp_sb_data);
+
+	bcm_fill(sc,
+		 (BAR_CSTRORM_INTMEM +
+		  CSTORM_SP_STATUS_BLOCK_OFFSET(SC_FUNC(sc))),
+		 0, CSTORM_SP_STATUS_BLOCK_SIZE);
+	bcm_fill(sc,
+		 (BAR_CSTRORM_INTMEM +
+		  CSTORM_SP_SYNC_BLOCK_OFFSET(SC_FUNC(sc))),
+		 0, CSTORM_SP_SYNC_BLOCK_SIZE);
+}
+
+static void
+bcm_setup_ndsb_state_machine(struct hc_status_block_sm *hc_sm, int igu_sb_id,
+			     int igu_seg_id)
+{
+	hc_sm->igu_sb_id = igu_sb_id;
+	hc_sm->igu_seg_id = igu_seg_id;
+	hc_sm->timer_value = 0xFF;
+	hc_sm->time_to_expire = 0xFFFFFFFF;
+}
+
+static void bcm_map_sb_state_machines(struct hc_index_data *index_data)
+{
+	/* zero out state machine indices */
+
+	/* rx indices */
+	index_data[HC_INDEX_ETH_RX_CQ_CONS].flags &= ~HC_INDEX_DATA_SM_ID;
+
+	/* tx indices */
+	index_data[HC_INDEX_OOO_TX_CQ_CONS].flags &= ~HC_INDEX_DATA_SM_ID;
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS0].flags &= ~HC_INDEX_DATA_SM_ID;
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS1].flags &= ~HC_INDEX_DATA_SM_ID;
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS2].flags &= ~HC_INDEX_DATA_SM_ID;
+
+	/* map indices */
+
+	/* rx indices */
+	index_data[HC_INDEX_ETH_RX_CQ_CONS].flags |=
+	    (SM_RX_ID << HC_INDEX_DATA_SM_ID_SHIFT);
+
+	/* tx indices */
+	index_data[HC_INDEX_OOO_TX_CQ_CONS].flags |=
+	    (SM_TX_ID << HC_INDEX_DATA_SM_ID_SHIFT);
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS0].flags |=
+	    (SM_TX_ID << HC_INDEX_DATA_SM_ID_SHIFT);
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS1].flags |=
+	    (SM_TX_ID << HC_INDEX_DATA_SM_ID_SHIFT);
+	index_data[HC_INDEX_ETH_TX_CQ_CONS_COS2].flags |=
+	    (SM_TX_ID << HC_INDEX_DATA_SM_ID_SHIFT);
+}
+
+static void
+bcm_init_sb(struct bcm_softc *sc, phys_addr_t busaddr, int vfid,
+	    uint8_t vf_valid, int fw_sb_id, int igu_sb_id)
+{
+	struct hc_status_block_data_e2 sb_data_e2;
+	struct hc_status_block_data_e1x sb_data_e1x;
+	struct hc_status_block_sm *hc_sm_p;
+	uint32_t *sb_data_p;
+	int igu_seg_id;
+	int data_size;
+
+	if (CHIP_INT_MODE_IS_BC(sc)) {
+		igu_seg_id = HC_SEG_ACCESS_NORM;
+	} else {
+		igu_seg_id = IGU_SEG_ACCESS_NORM;
+	}
+
+	bcm_zero_fp_sb(sc, fw_sb_id);
+
+	if (!CHIP_IS_E1x(sc)) {
+		memset(&sb_data_e2, 0, sizeof(struct hc_status_block_data_e2));
+		sb_data_e2.common.state = SB_ENABLED;
+		sb_data_e2.common.p_func.pf_id = SC_FUNC(sc);
+		sb_data_e2.common.p_func.vf_id = vfid;
+		sb_data_e2.common.p_func.vf_valid = vf_valid;
+		sb_data_e2.common.p_func.vnic_id = SC_VN(sc);
+		sb_data_e2.common.same_igu_sb_1b = TRUE;
+		sb_data_e2.common.host_sb_addr.hi = U64_HI(busaddr);
+		sb_data_e2.common.host_sb_addr.lo = U64_LO(busaddr);
+		hc_sm_p = sb_data_e2.common.state_machine;
+		sb_data_p = (uint32_t *) & sb_data_e2;
+		data_size = (sizeof(struct hc_status_block_data_e2) /
+			     sizeof(uint32_t));
+		bcm_map_sb_state_machines(sb_data_e2.index_data);
+	} else {
+		memset(&sb_data_e1x, 0,
+		       sizeof(struct hc_status_block_data_e1x));
+		sb_data_e1x.common.state = SB_ENABLED;
+		sb_data_e1x.common.p_func.pf_id = SC_FUNC(sc);
+		sb_data_e1x.common.p_func.vf_id = 0xff;
+		sb_data_e1x.common.p_func.vf_valid = FALSE;
+		sb_data_e1x.common.p_func.vnic_id = SC_VN(sc);
+		sb_data_e1x.common.same_igu_sb_1b = TRUE;
+		sb_data_e1x.common.host_sb_addr.hi = U64_HI(busaddr);
+		sb_data_e1x.common.host_sb_addr.lo = U64_LO(busaddr);
+		hc_sm_p = sb_data_e1x.common.state_machine;
+		sb_data_p = (uint32_t *) & sb_data_e1x;
+		data_size = (sizeof(struct hc_status_block_data_e1x) /
+			     sizeof(uint32_t));
+		bcm_map_sb_state_machines(sb_data_e1x.index_data);
+	}
+
+	bcm_setup_ndsb_state_machine(&hc_sm_p[SM_RX_ID], igu_sb_id, igu_seg_id);
+	bcm_setup_ndsb_state_machine(&hc_sm_p[SM_TX_ID], igu_sb_id, igu_seg_id);
+
+	/* write indices to HW - PCI guarantees endianity of regpairs */
+	bcm_wr_fp_sb_data(sc, fw_sb_id, sb_data_p, data_size);
+}
+
+static uint8_t bcm_fp_qzone_id(struct bcm_fastpath *fp)
+{
+	if (CHIP_IS_E1x(fp->sc)) {
+		return (fp->cl_id + SC_PORT(fp->sc) * ETH_MAX_RX_CLIENTS_E1H);
+	} else {
+		return (fp->cl_id);
+	}
+}
+
+static uint32_t
+bcm_rx_ustorm_prods_offset(struct bcm_softc *sc, struct bcm_fastpath *fp)
+{
+	uint32_t offset = BAR_USTRORM_INTMEM;
+
+	if (IS_VF(sc)) {
+		return (PXP_VF_ADDR_USDM_QUEUES_START +
+			(sc->acquire_resp.resc.hw_qid[fp->index] *
+			 sizeof(struct ustorm_queue_zone_data)));
+	} else if (!CHIP_IS_E1x(sc)) {
+		offset += USTORM_RX_PRODS_E2_OFFSET(fp->cl_qzone_id);
+	} else {
+		offset += USTORM_RX_PRODS_E1X_OFFSET(SC_PORT(sc), fp->cl_id);
+	}
+
+	return offset;
+}
+
+static void bcm_init_eth_fp(struct bcm_softc *sc, int idx)
+{
+	struct bcm_fastpath *fp = &sc->fp[idx];
+	uint32_t cids[ECORE_MULTI_TX_COS] = { 0 };
+	unsigned long q_type = 0;
+	int cos;
+
+	fp->sc = sc;
+	fp->index = idx;
+
+	fp->igu_sb_id = (sc->igu_base_sb + idx + CNIC_SUPPORT(sc));
+	fp->fw_sb_id = (sc->base_fw_ndsb + idx + CNIC_SUPPORT(sc));
+
+	if (CHIP_IS_E1x(sc))
+		fp->cl_id = SC_L_ID(sc) + idx;
+	else
+/* want client ID same as IGU SB ID for non-E1 */
+		fp->cl_id = fp->igu_sb_id;
+	fp->cl_qzone_id = bcm_fp_qzone_id(fp);
+
+	/* setup sb indices */
+	if (!CHIP_IS_E1x(sc)) {
+		fp->sb_index_values = fp->status_block.e2_sb->sb.index_values;
+		fp->sb_running_index = fp->status_block.e2_sb->sb.running_index;
+	} else {
+		fp->sb_index_values = fp->status_block.e1x_sb->sb.index_values;
+		fp->sb_running_index =
+		    fp->status_block.e1x_sb->sb.running_index;
+	}
+
+	/* init shortcut */
+	fp->ustorm_rx_prods_offset = bcm_rx_ustorm_prods_offset(sc, fp);
+
+	fp->rx_cq_cons_sb = &fp->sb_index_values[HC_INDEX_ETH_RX_CQ_CONS];
+
+	for (cos = 0; cos < sc->max_cos; cos++) {
+		cids[cos] = idx;
+	}
+	fp->tx_cons_sb = &fp->sb_index_values[HC_INDEX_ETH_TX_CQ_CONS_COS0];
+
+	/* nothing more for a VF to do */
+	if (IS_VF(sc)) {
+		return;
+	}
+
+	bcm_init_sb(sc, fp->sb_dma.paddr, BCM_VF_ID_INVALID, FALSE,
+		    fp->fw_sb_id, fp->igu_sb_id);
+
+	bcm_update_fp_sb_idx(fp);
+
+	/* Configure Queue State object */
+	bcm_set_bit(ECORE_Q_TYPE_HAS_RX, &q_type);
+	bcm_set_bit(ECORE_Q_TYPE_HAS_TX, &q_type);
+
+	ecore_init_queue_obj(sc,
+			     &sc->sp_objs[idx].q_obj,
+			     fp->cl_id,
+			     cids,
+			     sc->max_cos,
+			     SC_FUNC(sc),
+			     BCM_SP(sc, q_rdata), (phys_addr_t) ((void *)
+								 BCM_SP_MAPPING
+								 (sc, q_rdata)),
+			     q_type);
+
+	/* configure classification DBs */
+	ecore_init_mac_obj(sc,
+			   &sc->sp_objs[idx].mac_obj,
+			   fp->cl_id,
+			   idx,
+			   SC_FUNC(sc),
+			   BCM_SP(sc, mac_rdata), (phys_addr_t) ((void *)
+								 BCM_SP_MAPPING
+								 (sc,
+								  mac_rdata)),
+			   ECORE_FILTER_MAC_PENDING, &sc->sp_state,
+			   ECORE_OBJ_TYPE_RX_TX, &sc->macs_pool);
+}
+
+static void
+bcm_update_rx_prod(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		   uint16_t rx_bd_prod, uint16_t rx_cq_prod)
+{
+	union ustorm_eth_rx_producers rx_prods;
+	uint32_t i;
+
+	/* update producers */
+	rx_prods.prod.bd_prod = rx_bd_prod;
+	rx_prods.prod.cqe_prod = rx_cq_prod;
+	rx_prods.prod.reserved = 0;
+
+	/*
+	 * Make sure that the BD and SGE data is updated before updating the
+	 * producers since FW might read the BD/SGE right after the producer
+	 * is updated.
+	 * This is only applicable for weak-ordered memory model archs such
+	 * as IA-64. The following barrier is also mandatory since FW will
+	 * assumes BDs must have buffers.
+	 */
+	wmb();
+
+	for (i = 0; i < (sizeof(rx_prods) / 4); i++) {
+		REG_WR(sc,
+		       (fp->ustorm_rx_prods_offset + (i * 4)),
+		       rx_prods.raw_data[i]);
+	}
+
+	wmb();			/* keep prod updates ordered */
+}
+
+static void bcm_init_rx_rings(struct bcm_softc *sc)
+{
+	struct bcm_fastpath *fp;
+	int i;
+	struct bcm_rx_queue *rxq;
+
+	for (i = 0; i < sc->num_queues; i++) {
+		fp = &sc->fp[i];
+		rxq = sc->rx_queues[fp->index];
+		if (!rxq) {
+			PMD_RX_LOG(ERR, "RX queue is NULL");
+			return;
+		}
+
+/*
+ * Activate the BD ring...
+ * Warning, this will generate an interrupt (to the TSTORM)
+ * so this can only be done after the chip is initialized
+ */
+		bcm_update_rx_prod(sc, fp, rxq->rx_bd_tail, rxq->rx_cq_tail);
+
+		if (i != 0) {
+			continue;
+		}
+	}
+}
+
+static void bcm_init_tx_ring_one(struct bcm_fastpath *fp)
+{
+	struct bcm_tx_queue *txq = fp->sc->tx_queues[fp->index];
+
+	fp->tx_db.data.header.header = 1 << DOORBELL_HDR_DB_TYPE_SHIFT;
+	fp->tx_db.data.zero_fill1 = 0;
+	fp->tx_db.data.prod = 0;
+
+	if (!txq) {
+		PMD_TX_LOG(ERR, "ERROR: TX queue is NULL");
+		return;
+	}
+
+	txq->tx_pkt_tail = 0;
+	txq->tx_pkt_head = 0;
+	txq->tx_bd_tail = 0;
+	txq->tx_bd_head = 0;
+}
+
+static void bcm_init_tx_rings(struct bcm_softc *sc)
+{
+	int i;
+
+	for (i = 0; i < sc->num_queues; i++) {
+		bcm_init_tx_ring_one(&sc->fp[i]);
+	}
+}
+
+static void bcm_init_def_sb(struct bcm_softc *sc)
+{
+	struct host_sp_status_block *def_sb = sc->def_sb;
+	phys_addr_t mapping = sc->def_sb_dma.paddr;
+	int igu_sp_sb_index;
+	int igu_seg_id;
+	int port = SC_PORT(sc);
+	int func = SC_FUNC(sc);
+	int reg_offset, reg_offset_en5;
+	uint64_t section;
+	int index, sindex;
+	struct hc_sp_status_block_data sp_sb_data;
+
+	memset(&sp_sb_data, 0, sizeof(struct hc_sp_status_block_data));
+
+	if (CHIP_INT_MODE_IS_BC(sc)) {
+		igu_sp_sb_index = DEF_SB_IGU_ID;
+		igu_seg_id = HC_SEG_ACCESS_DEF;
+	} else {
+		igu_sp_sb_index = sc->igu_dsb_id;
+		igu_seg_id = IGU_SEG_ACCESS_DEF;
+	}
+
+	/* attentions */
+	section = ((uint64_t) mapping +
+		   offsetof(struct host_sp_status_block, atten_status_block));
+	def_sb->atten_status_block.status_block_id = igu_sp_sb_index;
+	sc->attn_state = 0;
+
+	reg_offset = (port) ? MISC_REG_AEU_ENABLE1_FUNC_1_OUT_0 :
+	    MISC_REG_AEU_ENABLE1_FUNC_0_OUT_0;
+
+	reg_offset_en5 = (port) ? MISC_REG_AEU_ENABLE5_FUNC_1_OUT_0 :
+	    MISC_REG_AEU_ENABLE5_FUNC_0_OUT_0;
+
+	for (index = 0; index < MAX_DYNAMIC_ATTN_GRPS; index++) {
+/* take care of sig[0]..sig[4] */
+		for (sindex = 0; sindex < 4; sindex++) {
+			sc->attn_group[index].sig[sindex] =
+			    REG_RD(sc,
+				   (reg_offset + (sindex * 0x4) +
+				    (0x10 * index)));
+		}
+
+		if (!CHIP_IS_E1x(sc)) {
+			/*
+			 * enable5 is separate from the rest of the registers,
+			 * and the address skip is 4 and not 16 between the
+			 * different groups
+			 */
+			sc->attn_group[index].sig[4] =
+			    REG_RD(sc, (reg_offset_en5 + (0x4 * index)));
+		} else {
+			sc->attn_group[index].sig[4] = 0;
+		}
+	}
+
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		reg_offset =
+		    port ? HC_REG_ATTN_MSG1_ADDR_L : HC_REG_ATTN_MSG0_ADDR_L;
+		REG_WR(sc, reg_offset, U64_LO(section));
+		REG_WR(sc, (reg_offset + 4), U64_HI(section));
+	} else if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, IGU_REG_ATTN_MSG_ADDR_L, U64_LO(section));
+		REG_WR(sc, IGU_REG_ATTN_MSG_ADDR_H, U64_HI(section));
+	}
+
+	section = ((uint64_t) mapping +
+		   offsetof(struct host_sp_status_block, sp_sb));
+
+	bcm_zero_sp_sb(sc);
+
+	/* PCI guarantees endianity of regpair */
+	sp_sb_data.state = SB_ENABLED;
+	sp_sb_data.host_sb_addr.lo = U64_LO(section);
+	sp_sb_data.host_sb_addr.hi = U64_HI(section);
+	sp_sb_data.igu_sb_id = igu_sp_sb_index;
+	sp_sb_data.igu_seg_id = igu_seg_id;
+	sp_sb_data.p_func.pf_id = func;
+	sp_sb_data.p_func.vnic_id = SC_VN(sc);
+	sp_sb_data.p_func.vf_id = 0xff;
+
+	bcm_wr_sp_sb_data(sc, &sp_sb_data);
+
+	bcm_ack_sb(sc, sc->igu_dsb_id, USTORM_ID, 0, IGU_INT_ENABLE, 0);
+}
+
+static void bcm_init_sp_ring(struct bcm_softc *sc)
+{
+	atomic_store_rel_long(&sc->cq_spq_left, MAX_SPQ_PENDING);
+	sc->spq_prod_idx = 0;
+	sc->dsb_sp_prod =
+	    &sc->def_sb->sp_sb.index_values[HC_SP_INDEX_ETH_DEF_CONS];
+	sc->spq_prod_bd = sc->spq;
+	sc->spq_last_bd = (sc->spq_prod_bd + MAX_SP_DESC_CNT);
+}
+
+static void bcm_init_eq_ring(struct bcm_softc *sc)
+{
+	union event_ring_elem *elem;
+	int i;
+
+	for (i = 1; i <= NUM_EQ_PAGES; i++) {
+		elem = &sc->eq[EQ_DESC_CNT_PAGE * i - 1];
+
+		elem->next_page.addr.hi = htole32(U64_HI(sc->eq_dma.paddr +
+							 BCM_PAGE_SIZE *
+							 (i % NUM_EQ_PAGES)));
+		elem->next_page.addr.lo = htole32(U64_LO(sc->eq_dma.paddr +
+							 BCM_PAGE_SIZE *
+							 (i % NUM_EQ_PAGES)));
+	}
+
+	sc->eq_cons = 0;
+	sc->eq_prod = NUM_EQ_DESC;
+	sc->eq_cons_sb = &sc->def_sb->sp_sb.index_values[HC_SP_INDEX_EQ_CONS];
+
+	atomic_store_rel_long(&sc->eq_spq_left,
+			      (min((MAX_SP_DESC_CNT - MAX_SPQ_PENDING),
+				   NUM_EQ_DESC) - 1));
+}
+
+static void bcm_init_internal_common(struct bcm_softc *sc)
+{
+	int i;
+
+	if (IS_MF_SI(sc)) {
+/*
+ * In switch independent mode, the TSTORM needs to accept
+ * packets that failed classification, since approximate match
+ * mac addresses aren't written to NIG LLH.
+ */
+		REG_WR8(sc,
+			(BAR_TSTRORM_INTMEM +
+			 TSTORM_ACCEPT_CLASSIFY_FAILED_OFFSET), 2);
+	} else
+		REG_WR8(sc,
+			(BAR_TSTRORM_INTMEM +
+			 TSTORM_ACCEPT_CLASSIFY_FAILED_OFFSET), 0);
+
+	/*
+	 * Zero this manually as its initialization is currently missing
+	 * in the initTool.
+	 */
+	for (i = 0; i < (USTORM_AGG_DATA_SIZE >> 2); i++) {
+		REG_WR(sc,
+		       (BAR_USTRORM_INTMEM + USTORM_AGG_DATA_OFFSET + (i * 4)),
+		       0);
+	}
+
+	if (!CHIP_IS_E1x(sc)) {
+		REG_WR8(sc, (BAR_CSTRORM_INTMEM + CSTORM_IGU_MODE_OFFSET),
+			CHIP_INT_MODE_IS_BC(sc) ? HC_IGU_BC_MODE :
+			HC_IGU_NBC_MODE);
+	}
+}
+
+static void bcm_init_internal(struct bcm_softc *sc, uint32_t load_code)
+{
+	switch (load_code) {
+	case FW_MSG_CODE_DRV_LOAD_COMMON:
+	case FW_MSG_CODE_DRV_LOAD_COMMON_CHIP:
+		bcm_init_internal_common(sc);
+		/* no break */
+
+	case FW_MSG_CODE_DRV_LOAD_PORT:
+		/* nothing to do */
+		/* no break */
+
+	case FW_MSG_CODE_DRV_LOAD_FUNCTION:
+		/* internal memory per function is initialized inside bcm_pf_init */
+		break;
+
+	default:
+		PMD_DRV_LOG(NOTICE, "Unknown load_code (0x%x) from MCP",
+			    load_code);
+		break;
+	}
+}
+
+static void
+storm_memset_func_cfg(struct bcm_softc *sc,
+		      struct tstorm_eth_function_common_config *tcfg,
+		      uint16_t abs_fid)
+{
+	uint32_t addr;
+	size_t size;
+
+	addr = (BAR_TSTRORM_INTMEM +
+		TSTORM_FUNCTION_COMMON_CONFIG_OFFSET(abs_fid));
+	size = sizeof(struct tstorm_eth_function_common_config);
+	ecore_storm_memset_struct(sc, addr, size, (uint32_t *) tcfg);
+}
+
+static void bcm_func_init(struct bcm_softc *sc, struct bcm_func_init_params *p)
+{
+	struct tstorm_eth_function_common_config tcfg = { 0 };
+
+	if (CHIP_IS_E1x(sc)) {
+		storm_memset_func_cfg(sc, &tcfg, p->func_id);
+	}
+
+	/* Enable the function in the FW */
+	storm_memset_vf_to_pf(sc, p->func_id, p->pf_id);
+	storm_memset_func_en(sc, p->func_id, 1);
+
+	/* spq */
+	if (p->func_flgs & FUNC_FLG_SPQ) {
+		storm_memset_spq_addr(sc, p->spq_map, p->func_id);
+		REG_WR(sc,
+		       (XSEM_REG_FAST_MEMORY +
+			XSTORM_SPQ_PROD_OFFSET(p->func_id)), p->spq_prod);
+	}
+}
+
+/*
+ * Calculates the sum of vn_min_rates.
+ * It's needed for further normalizing of the min_rates.
+ * Returns:
+ *   sum of vn_min_rates.
+ *     or
+ *   0 - if all the min_rates are 0.
+ * In the later case fainess algorithm should be deactivated.
+ * If all min rates are not zero then those that are zeroes will be set to 1.
+ */
+static void bcm_calc_vn_min(struct bcm_softc *sc, struct cmng_init_input *input)
+{
+	uint32_t vn_cfg;
+	uint32_t vn_min_rate;
+	int all_zero = 1;
+	int vn;
+
+	for (vn = VN_0; vn < SC_MAX_VN_NUM(sc); vn++) {
+		vn_cfg = sc->devinfo.mf_info.mf_config[vn];
+		vn_min_rate = (((vn_cfg & FUNC_MF_CFG_MIN_BW_MASK) >>
+				FUNC_MF_CFG_MIN_BW_SHIFT) * 100);
+
+		if (vn_cfg & FUNC_MF_CFG_FUNC_HIDE) {
+			/* skip hidden VNs */
+			vn_min_rate = 0;
+		} else if (!vn_min_rate) {
+			/* If min rate is zero - set it to 100 */
+			vn_min_rate = DEF_MIN_RATE;
+		} else {
+			all_zero = 0;
+		}
+
+		input->vnic_min_rate[vn] = vn_min_rate;
+	}
+
+	/* if ETS or all min rates are zeros - disable fairness */
+	if (all_zero) {
+		input->flags.cmng_enables &= ~CMNG_FLAGS_PER_PORT_FAIRNESS_VN;
+	} else {
+		input->flags.cmng_enables |= CMNG_FLAGS_PER_PORT_FAIRNESS_VN;
+	}
+}
+
+static uint16_t
+bcm_extract_max_cfg(__rte_unused struct bcm_softc *sc, uint32_t mf_cfg)
+{
+	uint16_t max_cfg = ((mf_cfg & FUNC_MF_CFG_MAX_BW_MASK) >>
+			    FUNC_MF_CFG_MAX_BW_SHIFT);
+
+	if (!max_cfg) {
+		PMD_DRV_LOG(DEBUG,
+			    "Max BW configured to 0 - using 100 instead");
+		max_cfg = 100;
+	}
+
+	return max_cfg;
+}
+
+static void
+bcm_calc_vn_max(struct bcm_softc *sc, int vn, struct cmng_init_input *input)
+{
+	uint16_t vn_max_rate;
+	uint32_t vn_cfg = sc->devinfo.mf_info.mf_config[vn];
+	uint32_t max_cfg;
+
+	if (vn_cfg & FUNC_MF_CFG_FUNC_HIDE) {
+		vn_max_rate = 0;
+	} else {
+		max_cfg = bcm_extract_max_cfg(sc, vn_cfg);
+
+		if (IS_MF_SI(sc)) {
+			/* max_cfg in percents of linkspeed */
+			vn_max_rate =
+			    ((sc->link_vars.line_speed * max_cfg) / 100);
+		} else {	/* SD modes */
+			/* max_cfg is absolute in 100Mb units */
+			vn_max_rate = (max_cfg * 100);
+		}
+	}
+
+	input->vnic_max_rate[vn] = vn_max_rate;
+}
+
+static void
+bcm_cmng_fns_init(struct bcm_softc *sc, uint8_t read_cfg, uint8_t cmng_type)
+{
+	struct cmng_init_input input;
+	int vn;
+
+	memset(&input, 0, sizeof(struct cmng_init_input));
+
+	input.port_rate = sc->link_vars.line_speed;
+
+	if (cmng_type == CMNG_FNS_MINMAX) {
+/* read mf conf from shmem */
+		if (read_cfg) {
+			bcm_read_mf_cfg(sc);
+		}
+
+/* get VN min rate and enable fairness if not 0 */
+		bcm_calc_vn_min(sc, &input);
+
+/* get VN max rate */
+		if (sc->port.pmf) {
+			for (vn = VN_0; vn < SC_MAX_VN_NUM(sc); vn++) {
+				bcm_calc_vn_max(sc, vn, &input);
+			}
+		}
+
+/* always enable rate shaping and fairness */
+		input.flags.cmng_enables |= CMNG_FLAGS_PER_PORT_RATE_SHAPING_VN;
+
+		ecore_init_cmng(&input, &sc->cmng);
+		return;
+	}
+}
+
+static int bcm_get_cmng_fns_mode(struct bcm_softc *sc)
+{
+	if (CHIP_REV_IS_SLOW(sc)) {
+		return CMNG_FNS_NONE;
+	}
+
+	if (IS_MF(sc)) {
+		return CMNG_FNS_MINMAX;
+	}
+
+	return CMNG_FNS_NONE;
+}
+
+static void
+storm_memset_cmng(struct bcm_softc *sc, struct cmng_init *cmng, uint8_t port)
+{
+	int vn;
+	int func;
+	uint32_t addr;
+	size_t size;
+
+	addr = (BAR_XSTRORM_INTMEM + XSTORM_CMNG_PER_PORT_VARS_OFFSET(port));
+	size = sizeof(struct cmng_struct_per_port);
+	ecore_storm_memset_struct(sc, addr, size, (uint32_t *) & cmng->port);
+
+	for (vn = VN_0; vn < SC_MAX_VN_NUM(sc); vn++) {
+		func = func_by_vn(sc, vn);
+
+		addr = (BAR_XSTRORM_INTMEM +
+			XSTORM_RATE_SHAPING_PER_VN_VARS_OFFSET(func));
+		size = sizeof(struct rate_shaping_vars_per_vn);
+		ecore_storm_memset_struct(sc, addr, size,
+					  (uint32_t *) & cmng->
+					  vnic.vnic_max_rate[vn]);
+
+		addr = (BAR_XSTRORM_INTMEM +
+			XSTORM_FAIRNESS_PER_VN_VARS_OFFSET(func));
+		size = sizeof(struct fairness_vars_per_vn);
+		ecore_storm_memset_struct(sc, addr, size,
+					  (uint32_t *) & cmng->
+					  vnic.vnic_min_rate[vn]);
+	}
+}
+
+static void bcm_pf_init(struct bcm_softc *sc)
+{
+	struct bcm_func_init_params func_init;
+	struct event_ring_data eq_data;
+	uint16_t flags;
+
+	memset(&eq_data, 0, sizeof(struct event_ring_data));
+	memset(&func_init, 0, sizeof(struct bcm_func_init_params));
+
+	if (!CHIP_IS_E1x(sc)) {
+/* reset IGU PF statistics: MSIX + ATTN */
+/* PF */
+		REG_WR(sc,
+		       (IGU_REG_STATISTIC_NUM_MESSAGE_SENT +
+			(BCM_IGU_STAS_MSG_VF_CNT * 4) +
+			((CHIP_IS_MODE_4_PORT(sc) ? SC_FUNC(sc) : SC_VN(sc)) *
+			 4)), 0);
+/* ATTN */
+		REG_WR(sc,
+		       (IGU_REG_STATISTIC_NUM_MESSAGE_SENT +
+			(BCM_IGU_STAS_MSG_VF_CNT * 4) +
+			(BCM_IGU_STAS_MSG_PF_CNT * 4) +
+			((CHIP_IS_MODE_4_PORT(sc) ? SC_FUNC(sc) : SC_VN(sc)) *
+			 4)), 0);
+	}
+
+	/* function setup flags */
+	flags = (FUNC_FLG_STATS | FUNC_FLG_LEADING | FUNC_FLG_SPQ);
+
+	func_init.func_flgs = flags;
+	func_init.pf_id = SC_FUNC(sc);
+	func_init.func_id = SC_FUNC(sc);
+	func_init.spq_map = sc->spq_dma.paddr;
+	func_init.spq_prod = sc->spq_prod_idx;
+
+	bcm_func_init(sc, &func_init);
+
+	memset(&sc->cmng, 0, sizeof(struct cmng_struct_per_port));
+
+	/*
+	 * Congestion management values depend on the link rate.
+	 * There is no active link so initial link rate is set to 10Gbps.
+	 * When the link comes up the congestion management values are
+	 * re-calculated according to the actual link rate.
+	 */
+	sc->link_vars.line_speed = SPEED_10000;
+	bcm_cmng_fns_init(sc, TRUE, bcm_get_cmng_fns_mode(sc));
+
+	/* Only the PMF sets the HW */
+	if (sc->port.pmf) {
+		storm_memset_cmng(sc, &sc->cmng, SC_PORT(sc));
+	}
+
+	/* init Event Queue - PCI bus guarantees correct endainity */
+	eq_data.base_addr.hi = U64_HI(sc->eq_dma.paddr);
+	eq_data.base_addr.lo = U64_LO(sc->eq_dma.paddr);
+	eq_data.producer = sc->eq_prod;
+	eq_data.index_id = HC_SP_INDEX_EQ_CONS;
+	eq_data.sb_id = DEF_SB_ID;
+	storm_memset_eq_data(sc, &eq_data, SC_FUNC(sc));
+}
+
+static void bcm_hc_int_enable(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	uint32_t addr = (port) ? HC_REG_CONFIG_1 : HC_REG_CONFIG_0;
+	uint32_t val = REG_RD(sc, addr);
+	uint8_t msix = (sc->interrupt_mode == INTR_MODE_MSIX)
+	    || (sc->interrupt_mode == INTR_MODE_SINGLE_MSIX);
+	uint8_t single_msix = (sc->interrupt_mode == INTR_MODE_SINGLE_MSIX);
+	uint8_t msi = (sc->interrupt_mode == INTR_MODE_MSI);
+
+	if (msix) {
+		val &= ~(HC_CONFIG_0_REG_SINGLE_ISR_EN_0 |
+			 HC_CONFIG_0_REG_INT_LINE_EN_0);
+		val |= (HC_CONFIG_0_REG_MSI_MSIX_INT_EN_0 |
+			HC_CONFIG_0_REG_ATTN_BIT_EN_0);
+		if (single_msix) {
+			val |= HC_CONFIG_0_REG_SINGLE_ISR_EN_0;
+		}
+	} else if (msi) {
+		val &= ~HC_CONFIG_0_REG_INT_LINE_EN_0;
+		val |= (HC_CONFIG_0_REG_SINGLE_ISR_EN_0 |
+			HC_CONFIG_0_REG_MSI_MSIX_INT_EN_0 |
+			HC_CONFIG_0_REG_ATTN_BIT_EN_0);
+	} else {
+		val |= (HC_CONFIG_0_REG_SINGLE_ISR_EN_0 |
+			HC_CONFIG_0_REG_MSI_MSIX_INT_EN_0 |
+			HC_CONFIG_0_REG_INT_LINE_EN_0 |
+			HC_CONFIG_0_REG_ATTN_BIT_EN_0);
+
+		REG_WR(sc, addr, val);
+
+		val &= ~HC_CONFIG_0_REG_MSI_MSIX_INT_EN_0;
+	}
+
+	REG_WR(sc, addr, val);
+
+	/* ensure that HC_CONFIG is written before leading/trailing edge config */
+	mb();
+
+	/* init leading/trailing edge */
+	if (IS_MF(sc)) {
+		val = (0xee0f | (1 << (SC_VN(sc) + 4)));
+		if (sc->port.pmf) {
+			/* enable nig and gpio3 attention */
+			val |= 0x1100;
+		}
+	} else {
+		val = 0xffff;
+	}
+
+	REG_WR(sc, (HC_REG_TRAILING_EDGE_0 + port * 8), val);
+	REG_WR(sc, (HC_REG_LEADING_EDGE_0 + port * 8), val);
+
+	/* make sure that interrupts are indeed enabled from here on */
+	mb();
+}
+
+static void bcm_igu_int_enable(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint8_t msix = (sc->interrupt_mode == INTR_MODE_MSIX)
+	    || (sc->interrupt_mode == INTR_MODE_SINGLE_MSIX);
+	uint8_t single_msix = (sc->interrupt_mode == INTR_MODE_SINGLE_MSIX);
+	uint8_t msi = (sc->interrupt_mode == INTR_MODE_MSI);
+
+	val = REG_RD(sc, IGU_REG_PF_CONFIGURATION);
+
+	if (msix) {
+		val &= ~(IGU_PF_CONF_INT_LINE_EN | IGU_PF_CONF_SINGLE_ISR_EN);
+		val |= (IGU_PF_CONF_MSI_MSIX_EN | IGU_PF_CONF_ATTN_BIT_EN);
+		if (single_msix) {
+			val |= IGU_PF_CONF_SINGLE_ISR_EN;
+		}
+	} else if (msi) {
+		val &= ~IGU_PF_CONF_INT_LINE_EN;
+		val |= (IGU_PF_CONF_MSI_MSIX_EN |
+			IGU_PF_CONF_ATTN_BIT_EN | IGU_PF_CONF_SINGLE_ISR_EN);
+	} else {
+		val &= ~IGU_PF_CONF_MSI_MSIX_EN;
+		val |= (IGU_PF_CONF_INT_LINE_EN |
+			IGU_PF_CONF_ATTN_BIT_EN | IGU_PF_CONF_SINGLE_ISR_EN);
+	}
+
+	/* clean previous status - need to configure igu prior to ack */
+	if ((!msix) || single_msix) {
+		REG_WR(sc, IGU_REG_PF_CONFIGURATION, val);
+		bcm_ack_int(sc);
+	}
+
+	val |= IGU_PF_CONF_FUNC_EN;
+
+	PMD_DRV_LOG(DEBUG, "write 0x%x to IGU mode %s",
+		    val, ((msix) ? "MSI-X" : ((msi) ? "MSI" : "INTx")));
+
+	REG_WR(sc, IGU_REG_PF_CONFIGURATION, val);
+
+	mb();
+
+	/* init leading/trailing edge */
+	if (IS_MF(sc)) {
+		val = (0xee0f | (1 << (SC_VN(sc) + 4)));
+		if (sc->port.pmf) {
+			/* enable nig and gpio3 attention */
+			val |= 0x1100;
+		}
+	} else {
+		val = 0xffff;
+	}
+
+	REG_WR(sc, IGU_REG_TRAILING_EDGE_LATCH, val);
+	REG_WR(sc, IGU_REG_LEADING_EDGE_LATCH, val);
+
+	/* make sure that interrupts are indeed enabled from here on */
+	mb();
+}
+
+static void bcm_int_enable(struct bcm_softc *sc)
+{
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		bcm_hc_int_enable(sc);
+	} else {
+		bcm_igu_int_enable(sc);
+	}
+}
+
+static void bcm_hc_int_disable(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	uint32_t addr = (port) ? HC_REG_CONFIG_1 : HC_REG_CONFIG_0;
+	uint32_t val = REG_RD(sc, addr);
+
+	val &= ~(HC_CONFIG_0_REG_SINGLE_ISR_EN_0 |
+		 HC_CONFIG_0_REG_MSI_MSIX_INT_EN_0 |
+		 HC_CONFIG_0_REG_INT_LINE_EN_0 | HC_CONFIG_0_REG_ATTN_BIT_EN_0);
+	/* flush all outstanding writes */
+	mb();
+
+	REG_WR(sc, addr, val);
+	if (REG_RD(sc, addr) != val) {
+		PMD_DRV_LOG(ERR, "proper val not read from HC IGU!");
+	}
+}
+
+static void bcm_igu_int_disable(struct bcm_softc *sc)
+{
+	uint32_t val = REG_RD(sc, IGU_REG_PF_CONFIGURATION);
+
+	val &= ~(IGU_PF_CONF_MSI_MSIX_EN |
+		 IGU_PF_CONF_INT_LINE_EN | IGU_PF_CONF_ATTN_BIT_EN);
+
+	PMD_DRV_LOG(DEBUG, "write %x to IGU", val);
+
+	/* flush all outstanding writes */
+	mb();
+
+	REG_WR(sc, IGU_REG_PF_CONFIGURATION, val);
+	if (REG_RD(sc, IGU_REG_PF_CONFIGURATION) != val) {
+		PMD_DRV_LOG(ERR, "proper val not read from IGU!");
+	}
+}
+
+static void bcm_int_disable(struct bcm_softc *sc)
+{
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		bcm_hc_int_disable(sc);
+	} else {
+		bcm_igu_int_disable(sc);
+	}
+}
+
+static void bcm_nic_init(struct bcm_softc *sc, int load_code)
+{
+	int i;
+
+	PMD_INIT_FUNC_TRACE();
+
+	for (i = 0; i < sc->num_queues; i++) {
+		bcm_init_eth_fp(sc, i);
+	}
+
+	rmb();			/* ensure status block indices were read */
+
+	bcm_init_rx_rings(sc);
+	bcm_init_tx_rings(sc);
+
+	if (IS_VF(sc)) {
+		bcm_memset_stats(sc);
+		return;
+	}
+
+	/* initialize MOD_ABS interrupts */
+	elink_init_mod_abs_int(sc, &sc->link_vars,
+			       sc->devinfo.chip_id,
+			       sc->devinfo.shmem_base,
+			       sc->devinfo.shmem2_base, SC_PORT(sc));
+
+	bcm_init_def_sb(sc);
+	bcm_update_dsb_idx(sc);
+	bcm_init_sp_ring(sc);
+	bcm_init_eq_ring(sc);
+	bcm_init_internal(sc, load_code);
+	bcm_pf_init(sc);
+	bcm_stats_init(sc);
+
+	/* flush all before enabling interrupts */
+	mb();
+
+	bcm_int_enable(sc);
+
+	/* check for SPIO5 */
+	bcm_attn_int_deasserted0(sc,
+				 REG_RD(sc,
+					(MISC_REG_AEU_AFTER_INVERT_1_FUNC_0 +
+					 SC_PORT(sc) * 4)) &
+				 AEU_INPUTS_ATTN_BITS_SPIO5);
+}
+
+static void bcm_init_objs(struct bcm_softc *sc)
+{
+	/* mcast rules must be added to tx if tx switching is enabled */
+	ecore_obj_type o_type;
+	if (sc->flags & BCM_TX_SWITCHING)
+		o_type = ECORE_OBJ_TYPE_RX_TX;
+	else
+		o_type = ECORE_OBJ_TYPE_RX;
+
+	/* RX_MODE controlling object */
+	ecore_init_rx_mode_obj(sc, &sc->rx_mode_obj);
+
+	/* multicast configuration controlling object */
+	ecore_init_mcast_obj(sc,
+			     &sc->mcast_obj,
+			     sc->fp[0].cl_id,
+			     sc->fp[0].index,
+			     SC_FUNC(sc),
+			     SC_FUNC(sc),
+			     BCM_SP(sc, mcast_rdata), (phys_addr_t) ((void *)
+								     BCM_SP_MAPPING(sc, mcast_rdata)), ECORE_FILTER_MCAST_PENDING, &sc->sp_state, o_type);
+
+	/* Setup CAM credit pools */
+	ecore_init_mac_credit_pool(sc,
+				   &sc->macs_pool,
+				   SC_FUNC(sc),
+				   CHIP_IS_E1x(sc) ? VNICS_PER_PORT(sc) :
+				   VNICS_PER_PATH(sc));
+
+	ecore_init_vlan_credit_pool(sc,
+				    &sc->vlans_pool,
+				    SC_ABS_FUNC(sc) >> 1,
+				    CHIP_IS_E1x(sc) ? VNICS_PER_PORT(sc) :
+				    VNICS_PER_PATH(sc));
+
+	/* RSS configuration object */
+	ecore_init_rss_config_obj(&sc->rss_conf_obj,
+				  sc->fp[0].cl_id,
+				  sc->fp[0].index,
+				  SC_FUNC(sc),
+				  SC_FUNC(sc),
+				  BCM_SP(sc, rss_rdata), (phys_addr_t) ((void *)
+									BCM_SP_MAPPING(sc, rss_rdata)), ECORE_FILTER_RSS_CONF_PENDING, &sc->sp_state, ECORE_OBJ_TYPE_RX);
+}
+
+/*
+ * Initialize the function. This must be called before sending CLIENT_SETUP
+ * for the first client.
+ */
+static int bcm_func_start(struct bcm_softc *sc)
+{
+	struct ecore_func_state_params func_params = { NULL };
+	struct ecore_func_start_params *start_params =
+	    &func_params.params.start;
+
+	/* Prepare parameters for function state transitions */
+	bcm_set_bit(RAMROD_COMP_WAIT, &func_params.ramrod_flags);
+
+	func_params.f_obj = &sc->func_obj;
+	func_params.cmd = ECORE_F_CMD_START;
+
+	/* Function parameters */
+	start_params->mf_mode = sc->devinfo.mf_info.mf_mode;
+	start_params->sd_vlan_tag = OVLAN(sc);
+
+	if (CHIP_IS_E2(sc) || CHIP_IS_E3(sc)) {
+		start_params->network_cos_mode = STATIC_COS;
+	} else {		/* CHIP_IS_E1X */
+		start_params->network_cos_mode = FW_WRR;
+	}
+
+	start_params->gre_tunnel_mode = 0;
+	start_params->gre_tunnel_rss = 0;
+
+	return ecore_func_state_change(sc, &func_params);
+}
+
+static int bcm_set_power_state(struct bcm_softc *sc, uint8_t state)
+{
+	uint16_t pmcsr;
+
+	/* If there is no power capability, silently succeed */
+	if (!(sc->devinfo.pcie_cap_flags & BCM_PM_CAPABLE_FLAG)) {
+		PMD_DRV_LOG(WARN, "No power capability");
+		return 0;
+	}
+
+	pci_read(sc, (sc->devinfo.pcie_pm_cap_reg + PCIR_POWER_STATUS), &pmcsr,
+		 2);
+
+	switch (state) {
+	case PCI_PM_D0:
+		pci_write_word(sc,
+			       (sc->devinfo.pcie_pm_cap_reg +
+				PCIR_POWER_STATUS),
+			       ((pmcsr & ~PCIM_PSTAT_DMASK) | PCIM_PSTAT_PME));
+
+		if (pmcsr & PCIM_PSTAT_DMASK) {
+			/* delay required during transition out of D3hot */
+			DELAY(20000);
+		}
+
+		break;
+
+	case PCI_PM_D3hot:
+		/* don't shut down the power for emulation and FPGA */
+		if (CHIP_REV_IS_SLOW(sc)) {
+			return 0;
+		}
+
+		pmcsr &= ~PCIM_PSTAT_DMASK;
+		pmcsr |= PCIM_PSTAT_D3;
+
+		if (sc->wol) {
+			pmcsr |= PCIM_PSTAT_PMEENABLE;
+		}
+
+		pci_write_long(sc,
+			       (sc->devinfo.pcie_pm_cap_reg +
+				PCIR_POWER_STATUS), pmcsr);
+
+		/*
+		 * No more memory access after this point until device is brought back
+		 * to D0 state.
+		 */
+		break;
+
+	default:
+		PMD_DRV_LOG(NOTICE, "Can't support PCI power state = %d",
+			    state);
+		return -1;
+	}
+
+	return 0;
+}
+
+/* return true if succeeded to acquire the lock */
+static uint8_t bcm_trylock_hw_lock(struct bcm_softc *sc, uint32_t resource)
+{
+	uint32_t lock_status;
+	uint32_t resource_bit = (1 << resource);
+	int func = SC_FUNC(sc);
+	uint32_t hw_lock_control_reg;
+
+	/* Validating that the resource is within range */
+	if (resource > HW_LOCK_MAX_RESOURCE_VALUE) {
+		PMD_DRV_LOG(INFO,
+			    "resource(0x%x) > HW_LOCK_MAX_RESOURCE_VALUE(0x%x)",
+			    resource, HW_LOCK_MAX_RESOURCE_VALUE);
+		return FALSE;
+	}
+
+	if (func <= 5) {
+		hw_lock_control_reg = (MISC_REG_DRIVER_CONTROL_1 + func * 8);
+	} else {
+		hw_lock_control_reg =
+		    (MISC_REG_DRIVER_CONTROL_7 + (func - 6) * 8);
+	}
+
+	/* try to acquire the lock */
+	REG_WR(sc, hw_lock_control_reg + 4, resource_bit);
+	lock_status = REG_RD(sc, hw_lock_control_reg);
+	if (lock_status & resource_bit) {
+		return TRUE;
+	}
+
+	PMD_DRV_LOG(NOTICE, "Failed to get a resource lock 0x%x", resource);
+
+	return FALSE;
+}
+
+/*
+ * Get the recovery leader resource id according to the engine this function
+ * belongs to. Currently only only 2 engines is supported.
+ */
+static int bcm_get_leader_lock_resource(struct bcm_softc *sc)
+{
+	if (SC_PATH(sc)) {
+		return HW_LOCK_RESOURCE_RECOVERY_LEADER_1;
+	} else {
+		return HW_LOCK_RESOURCE_RECOVERY_LEADER_0;
+	}
+}
+
+/* try to acquire a leader lock for current engine */
+static uint8_t bcm_trylock_leader_lock(struct bcm_softc *sc)
+{
+	return bcm_trylock_hw_lock(sc, bcm_get_leader_lock_resource(sc));
+}
+
+static int bcm_release_leader_lock(struct bcm_softc *sc)
+{
+	return bcm_release_hw_lock(sc, bcm_get_leader_lock_resource(sc));
+}
+
+/* close gates #2, #3 and #4 */
+static void bcm_set_234_gates(struct bcm_softc *sc, uint8_t close)
+{
+	uint32_t val;
+
+	/* gates #2 and #4a are closed/opened */
+	/* #4 */
+	REG_WR(sc, PXP_REG_HST_DISCARD_DOORBELLS, ! !close);
+	/* #2 */
+	REG_WR(sc, PXP_REG_HST_DISCARD_INTERNAL_WRITES, ! !close);
+
+	/* #3 */
+	if (CHIP_IS_E1x(sc)) {
+/* prevent interrupts from HC on both ports */
+		val = REG_RD(sc, HC_REG_CONFIG_1);
+		if (close)
+			REG_WR(sc, HC_REG_CONFIG_1, (val & ~(uint32_t)
+						     HC_CONFIG_1_REG_BLOCK_DISABLE_1));
+		else
+			REG_WR(sc, HC_REG_CONFIG_1,
+			       (val | HC_CONFIG_1_REG_BLOCK_DISABLE_1));
+
+		val = REG_RD(sc, HC_REG_CONFIG_0);
+		if (close)
+			REG_WR(sc, HC_REG_CONFIG_0, (val & ~(uint32_t)
+						     HC_CONFIG_0_REG_BLOCK_DISABLE_0));
+		else
+			REG_WR(sc, HC_REG_CONFIG_0,
+			       (val | HC_CONFIG_0_REG_BLOCK_DISABLE_0));
+
+	} else {
+/* Prevent incomming interrupts in IGU */
+		val = REG_RD(sc, IGU_REG_BLOCK_CONFIGURATION);
+
+		if (close)
+			REG_WR(sc, IGU_REG_BLOCK_CONFIGURATION,
+			       (val & ~(uint32_t)
+				IGU_BLOCK_CONFIGURATION_REG_BLOCK_ENABLE));
+		else
+			REG_WR(sc, IGU_REG_BLOCK_CONFIGURATION,
+			       (val |
+				IGU_BLOCK_CONFIGURATION_REG_BLOCK_ENABLE));
+	}
+
+	wmb();
+}
+
+/* poll for pending writes bit, it should get cleared in no more than 1s */
+static int bcm_er_poll_igu_vq(struct bcm_softc *sc)
+{
+	uint32_t cnt = 1000;
+	uint32_t pend_bits = 0;
+
+	do {
+		pend_bits = REG_RD(sc, IGU_REG_PENDING_BITS_STATUS);
+
+		if (pend_bits == 0) {
+			break;
+		}
+
+		DELAY(1000);
+	} while (cnt-- > 0);
+
+	if (cnt <= 0) {
+		PMD_DRV_LOG(NOTICE, "Still pending IGU requests bits=0x%08x!",
+			    pend_bits);
+		return -1;
+	}
+
+	return 0;
+}
+
+#define SHARED_MF_CLP_MAGIC  0x80000000	/* 'magic' bit */
+
+static void bcm_clp_reset_prep(struct bcm_softc *sc, uint32_t * magic_val)
+{
+	/* Do some magic... */
+	uint32_t val = MFCFG_RD(sc, shared_mf_config.clp_mb);
+	*magic_val = val & SHARED_MF_CLP_MAGIC;
+	MFCFG_WR(sc, shared_mf_config.clp_mb, val | SHARED_MF_CLP_MAGIC);
+}
+
+/* restore the value of the 'magic' bit */
+static void bcm_clp_reset_done(struct bcm_softc *sc, uint32_t magic_val)
+{
+	/* Restore the 'magic' bit value... */
+	uint32_t val = MFCFG_RD(sc, shared_mf_config.clp_mb);
+	MFCFG_WR(sc, shared_mf_config.clp_mb,
+		 (val & (~SHARED_MF_CLP_MAGIC)) | magic_val);
+}
+
+/* prepare for MCP reset, takes care of CLP configurations */
+static void bcm_reset_mcp_prep(struct bcm_softc *sc, uint32_t * magic_val)
+{
+	uint32_t shmem;
+	uint32_t validity_offset;
+
+	/* set `magic' bit in order to save MF config */
+	bcm_clp_reset_prep(sc, magic_val);
+
+	/* get shmem offset */
+	shmem = REG_RD(sc, MISC_REG_SHARED_MEM_ADDR);
+	validity_offset =
+	    offsetof(struct shmem_region, validity_map[SC_PORT(sc)]);
+
+	/* Clear validity map flags */
+	if (shmem > 0) {
+		REG_WR(sc, shmem + validity_offset, 0);
+	}
+}
+
+#define MCP_TIMEOUT      5000	/* 5 seconds (in ms) */
+#define MCP_ONE_TIMEOUT  100	/* 100 ms */
+
+static void bcm_mcp_wait_one(struct bcm_softc *sc)
+{
+	/* special handling for emulation and FPGA (10 times longer) */
+	if (CHIP_REV_IS_SLOW(sc)) {
+		DELAY((MCP_ONE_TIMEOUT * 10) * 1000);
+	} else {
+		DELAY((MCP_ONE_TIMEOUT) * 1000);
+	}
+}
+
+/* initialize shmem_base and waits for validity signature to appear */
+static int bcm_init_shmem(struct bcm_softc *sc)
+{
+	int cnt = 0;
+	uint32_t val = 0;
+
+	do {
+		sc->devinfo.shmem_base =
+		    sc->link_params.shmem_base =
+		    REG_RD(sc, MISC_REG_SHARED_MEM_ADDR);
+
+		if (sc->devinfo.shmem_base) {
+			val = SHMEM_RD(sc, validity_map[SC_PORT(sc)]);
+			if (val & SHR_MEM_VALIDITY_MB)
+				return 0;
+		}
+
+		bcm_mcp_wait_one(sc);
+
+	} while (cnt++ < (MCP_TIMEOUT / MCP_ONE_TIMEOUT));
+
+	PMD_DRV_LOG(NOTICE, "BAD MCP validity signature");
+
+	return -1;
+}
+
+static int bcm_reset_mcp_comp(struct bcm_softc *sc, uint32_t magic_val)
+{
+	int rc = bcm_init_shmem(sc);
+
+	/* Restore the `magic' bit value */
+	bcm_clp_reset_done(sc, magic_val);
+
+	return rc;
+}
+
+static void bcm_pxp_prep(struct bcm_softc *sc)
+{
+	REG_WR(sc, PXP2_REG_RD_START_INIT, 0);
+	REG_WR(sc, PXP2_REG_RQ_RBC_DONE, 0);
+	wmb();
+}
+
+/*
+ * Reset the whole chip except for:
+ *      - PCIE core
+ *      - PCI Glue, PSWHST, PXP/PXP2 RF (all controlled by one reset bit)
+ *      - IGU
+ *      - MISC (including AEU)
+ *      - GRC
+ *      - RBCN, RBCP
+ */
+static void bcm_process_kill_chip_reset(struct bcm_softc *sc, uint8_t global)
+{
+	uint32_t not_reset_mask1, reset_mask1, not_reset_mask2, reset_mask2;
+	uint32_t global_bits2, stay_reset2;
+
+	/*
+	 * Bits that have to be set in reset_mask2 if we want to reset 'global'
+	 * (per chip) blocks.
+	 */
+	global_bits2 =
+	    MISC_REGISTERS_RESET_REG_2_RST_MCP_N_RESET_CMN_CPU |
+	    MISC_REGISTERS_RESET_REG_2_RST_MCP_N_RESET_CMN_CORE;
+
+	/*
+	 * Don't reset the following blocks.
+	 * Important: per port blocks (such as EMAC, BMAC, UMAC) can't be
+	 *            reset, as in 4 port device they might still be owned
+	 *            by the MCP (there is only one leader per path).
+	 */
+	not_reset_mask1 =
+	    MISC_REGISTERS_RESET_REG_1_RST_HC |
+	    MISC_REGISTERS_RESET_REG_1_RST_PXPV |
+	    MISC_REGISTERS_RESET_REG_1_RST_PXP;
+
+	not_reset_mask2 =
+	    MISC_REGISTERS_RESET_REG_2_RST_PCI_MDIO |
+	    MISC_REGISTERS_RESET_REG_2_RST_EMAC0_HARD_CORE |
+	    MISC_REGISTERS_RESET_REG_2_RST_EMAC1_HARD_CORE |
+	    MISC_REGISTERS_RESET_REG_2_RST_MISC_CORE |
+	    MISC_REGISTERS_RESET_REG_2_RST_RBCN |
+	    MISC_REGISTERS_RESET_REG_2_RST_GRC |
+	    MISC_REGISTERS_RESET_REG_2_RST_MCP_N_RESET_REG_HARD_CORE |
+	    MISC_REGISTERS_RESET_REG_2_RST_MCP_N_HARD_CORE_RST_B |
+	    MISC_REGISTERS_RESET_REG_2_RST_ATC |
+	    MISC_REGISTERS_RESET_REG_2_PGLC |
+	    MISC_REGISTERS_RESET_REG_2_RST_BMAC0 |
+	    MISC_REGISTERS_RESET_REG_2_RST_BMAC1 |
+	    MISC_REGISTERS_RESET_REG_2_RST_EMAC0 |
+	    MISC_REGISTERS_RESET_REG_2_RST_EMAC1 |
+	    MISC_REGISTERS_RESET_REG_2_UMAC0 | MISC_REGISTERS_RESET_REG_2_UMAC1;
+
+	/*
+	 * Keep the following blocks in reset:
+	 *  - all xxMACs are handled by the elink code.
+	 */
+	stay_reset2 =
+	    MISC_REGISTERS_RESET_REG_2_XMAC |
+	    MISC_REGISTERS_RESET_REG_2_XMAC_SOFT;
+
+	/* Full reset masks according to the chip */
+	reset_mask1 = 0xffffffff;
+
+	if (CHIP_IS_E1H(sc))
+		reset_mask2 = 0x1ffff;
+	else if (CHIP_IS_E2(sc))
+		reset_mask2 = 0xfffff;
+	else			/* CHIP_IS_E3 */
+		reset_mask2 = 0x3ffffff;
+
+	/* Don't reset global blocks unless we need to */
+	if (!global)
+		reset_mask2 &= ~global_bits2;
+
+	/*
+	 * In case of attention in the QM, we need to reset PXP
+	 * (MISC_REGISTERS_RESET_REG_2_RST_PXP_RQ_RD_WR) before QM
+	 * because otherwise QM reset would release 'close the gates' shortly
+	 * before resetting the PXP, then the PSWRQ would send a write
+	 * request to PGLUE. Then when PXP is reset, PGLUE would try to
+	 * read the payload data from PSWWR, but PSWWR would not
+	 * respond. The write queue in PGLUE would stuck, dmae commands
+	 * would not return. Therefore it's important to reset the second
+	 * reset register (containing the
+	 * MISC_REGISTERS_RESET_REG_2_RST_PXP_RQ_RD_WR bit) before the
+	 * first one (containing the MISC_REGISTERS_RESET_REG_1_RST_QM
+	 * bit).
+	 */
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_CLEAR,
+	       reset_mask2 & (~not_reset_mask2));
+
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR,
+	       reset_mask1 & (~not_reset_mask1));
+
+	mb();
+	wmb();
+
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_SET,
+	       reset_mask2 & (~stay_reset2));
+
+	mb();
+	wmb();
+
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET, reset_mask1);
+	wmb();
+}
+
+static int bcm_process_kill(struct bcm_softc *sc, uint8_t global)
+{
+	int cnt = 1000;
+	uint32_t val = 0;
+	uint32_t sr_cnt, blk_cnt, port_is_idle_0, port_is_idle_1, pgl_exp_rom2;
+	uint32_t tags_63_32 = 0;
+
+	/* Empty the Tetris buffer, wait for 1s */
+	do {
+		sr_cnt = REG_RD(sc, PXP2_REG_RD_SR_CNT);
+		blk_cnt = REG_RD(sc, PXP2_REG_RD_BLK_CNT);
+		port_is_idle_0 = REG_RD(sc, PXP2_REG_RD_PORT_IS_IDLE_0);
+		port_is_idle_1 = REG_RD(sc, PXP2_REG_RD_PORT_IS_IDLE_1);
+		pgl_exp_rom2 = REG_RD(sc, PXP2_REG_PGL_EXP_ROM2);
+		if (CHIP_IS_E3(sc)) {
+			tags_63_32 = REG_RD(sc, PGLUE_B_REG_TAGS_63_32);
+		}
+
+		if ((sr_cnt == 0x7e) && (blk_cnt == 0xa0) &&
+		    ((port_is_idle_0 & 0x1) == 0x1) &&
+		    ((port_is_idle_1 & 0x1) == 0x1) &&
+		    (pgl_exp_rom2 == 0xffffffff) &&
+		    (!CHIP_IS_E3(sc) || (tags_63_32 == 0xffffffff)))
+			break;
+		DELAY(1000);
+	} while (cnt-- > 0);
+
+	if (cnt <= 0) {
+		PMD_DRV_LOG(NOTICE,
+			    "ERROR: Tetris buffer didn't get empty or there "
+			    "are still outstanding read requests after 1s! "
+			    "sr_cnt=0x%08x, blk_cnt=0x%08x, port_is_idle_0=0x%08x, "
+			    "port_is_idle_1=0x%08x, pgl_exp_rom2=0x%08x",
+			    sr_cnt, blk_cnt, port_is_idle_0, port_is_idle_1,
+			    pgl_exp_rom2);
+		return -1;
+	}
+
+	mb();
+
+	/* Close gates #2, #3 and #4 */
+	bcm_set_234_gates(sc, TRUE);
+
+	/* Poll for IGU VQs for 57712 and newer chips */
+	if (!CHIP_IS_E1x(sc) && bcm_er_poll_igu_vq(sc)) {
+		return -1;
+	}
+
+	/* clear "unprepared" bit */
+	REG_WR(sc, MISC_REG_UNPREPARED, 0);
+	mb();
+
+	/* Make sure all is written to the chip before the reset */
+	wmb();
+
+	/*
+	 * Wait for 1ms to empty GLUE and PCI-E core queues,
+	 * PSWHST, GRC and PSWRD Tetris buffer.
+	 */
+	DELAY(1000);
+
+	/* Prepare to chip reset: */
+	/* MCP */
+	if (global) {
+		bcm_reset_mcp_prep(sc, &val);
+	}
+
+	/* PXP */
+	bcm_pxp_prep(sc);
+	mb();
+
+	/* reset the chip */
+	bcm_process_kill_chip_reset(sc, global);
+	mb();
+
+	/* Recover after reset: */
+	/* MCP */
+	if (global &&bcm_reset_mcp_comp(sc, val)) {
+		return -1;
+	}
+
+	/* Open the gates #2, #3 and #4 */
+	bcm_set_234_gates(sc, FALSE);
+
+	return 0;
+}
+
+static int bcm_leader_reset(struct bcm_softc *sc)
+{
+	int rc = 0;
+	uint8_t global = bcm_reset_is_global(sc);
+	uint32_t load_code;
+
+	/*
+	 * If not going to reset MCP, load "fake" driver to reset HW while
+	 * driver is owner of the HW.
+	 */
+	if (!global &&!BCM_NOMCP(sc)) {
+		load_code = bcm_fw_command(sc, DRV_MSG_CODE_LOAD_REQ,
+					   DRV_MSG_CODE_LOAD_REQ_WITH_LFA);
+		if (!load_code) {
+			PMD_DRV_LOG(NOTICE, "MCP response failure, aborting");
+			rc = -1;
+			goto exit_leader_reset;
+		}
+
+		if ((load_code != FW_MSG_CODE_DRV_LOAD_COMMON_CHIP) &&
+		    (load_code != FW_MSG_CODE_DRV_LOAD_COMMON)) {
+			PMD_DRV_LOG(NOTICE,
+				    "MCP unexpected response, aborting");
+			rc = -1;
+			goto exit_leader_reset2;
+		}
+
+		load_code = bcm_fw_command(sc, DRV_MSG_CODE_LOAD_DONE, 0);
+		if (!load_code) {
+			PMD_DRV_LOG(NOTICE, "MCP response failure, aborting");
+			rc = -1;
+			goto exit_leader_reset2;
+		}
+	}
+
+	/* try to recover after the failure */
+	if (bcm_process_kill(sc, global)) {
+		PMD_DRV_LOG(NOTICE, "Something bad occurred on engine %d!",
+			    SC_PATH(sc));
+		rc = -1;
+		goto exit_leader_reset2;
+	}
+
+	/*
+	 * Clear the RESET_IN_PROGRESS and RESET_GLOBAL bits and update the driver
+	 * state.
+	 */
+	bcm_set_reset_done(sc);
+	if (global) {
+		bcm_clear_reset_global(sc);
+	}
+
+exit_leader_reset2:
+
+	/* unload "fake driver" if it was loaded */
+	if (!global &&!BCM_NOMCP(sc)) {
+		bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_REQ_WOL_MCP, 0);
+		bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_DONE, 0);
+	}
+
+exit_leader_reset:
+
+	sc->is_leader = 0;
+	bcm_release_leader_lock(sc);
+
+	mb();
+	return rc;
+}
+
+/*
+ * prepare INIT transition, parameters configured:
+ *   - HC configuration
+ *   - Queue's CDU context
+ */
+static void
+bcm_pf_q_prep_init(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		   struct ecore_queue_init_params *init_params)
+{
+	uint8_t cos;
+	int cxt_index, cxt_offset;
+
+	bcm_set_bit(ECORE_Q_FLG_HC, &init_params->rx.flags);
+	bcm_set_bit(ECORE_Q_FLG_HC, &init_params->tx.flags);
+
+	bcm_set_bit(ECORE_Q_FLG_HC_EN, &init_params->rx.flags);
+	bcm_set_bit(ECORE_Q_FLG_HC_EN, &init_params->tx.flags);
+
+	/* HC rate */
+	init_params->rx.hc_rate =
+	    sc->hc_rx_ticks ? (1000000 / sc->hc_rx_ticks) : 0;
+	init_params->tx.hc_rate =
+	    sc->hc_tx_ticks ? (1000000 / sc->hc_tx_ticks) : 0;
+
+	/* FW SB ID */
+	init_params->rx.fw_sb_id = init_params->tx.fw_sb_id = fp->fw_sb_id;
+
+	/* CQ index among the SB indices */
+	init_params->rx.sb_cq_index = HC_INDEX_ETH_RX_CQ_CONS;
+	init_params->tx.sb_cq_index = HC_INDEX_ETH_FIRST_TX_CQ_CONS;
+
+	/* set maximum number of COSs supported by this queue */
+	init_params->max_cos = sc->max_cos;
+
+	/* set the context pointers queue object */
+	for (cos = FIRST_TX_COS_INDEX; cos < init_params->max_cos; cos++) {
+		cxt_index = fp->index / ILT_PAGE_CIDS;
+		cxt_offset = fp->index - (cxt_index * ILT_PAGE_CIDS);
+		init_params->cxts[cos] =
+		    &sc->context[cxt_index].vcxt[cxt_offset].eth;
+	}
+}
+
+/* set flags that are common for the Tx-only and not normal connections */
+static unsigned long
+bcm_get_common_flags(struct bcm_softc *sc, uint8_t zero_stats)
+{
+	unsigned long flags = 0;
+
+	/* PF driver will always initialize the Queue to an ACTIVE state */
+	bcm_set_bit(ECORE_Q_FLG_ACTIVE, &flags);
+
+	/*
+	 * tx only connections collect statistics (on the same index as the
+	 * parent connection). The statistics are zeroed when the parent
+	 * connection is initialized.
+	 */
+
+	bcm_set_bit(ECORE_Q_FLG_STATS, &flags);
+	if (zero_stats) {
+		bcm_set_bit(ECORE_Q_FLG_ZERO_STATS, &flags);
+	}
+
+	/*
+	 * tx only connections can support tx-switching, though their
+	 * CoS-ness doesn't survive the loopback
+	 */
+	if (sc->flags & BCM_TX_SWITCHING) {
+		bcm_set_bit(ECORE_Q_FLG_TX_SWITCH, &flags);
+	}
+
+	bcm_set_bit(ECORE_Q_FLG_PCSUM_ON_PKT, &flags);
+
+	return flags;
+}
+
+static unsigned long bcm_get_q_flags(struct bcm_softc *sc, uint8_t leading)
+{
+	unsigned long flags = 0;
+
+	if (IS_MF_SD(sc)) {
+		bcm_set_bit(ECORE_Q_FLG_OV, &flags);
+	}
+
+	if (leading) {
+		bcm_set_bit(ECORE_Q_FLG_LEADING_RSS, &flags);
+		bcm_set_bit(ECORE_Q_FLG_MCAST, &flags);
+	}
+
+	bcm_set_bit(ECORE_Q_FLG_VLAN, &flags);
+
+	/* merge with common flags */
+	return flags | bcm_get_common_flags(sc, TRUE);
+}
+
+static void
+bcm_pf_q_prep_general(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		      struct ecore_general_setup_params *gen_init, uint8_t cos)
+{
+	gen_init->stat_id = bcm_stats_id(fp);
+	gen_init->spcl_id = fp->cl_id;
+	gen_init->mtu = sc->mtu;
+	gen_init->cos = cos;
+}
+
+static void
+bcm_pf_rx_q_prep(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		 struct rxq_pause_params *pause,
+		 struct ecore_rxq_setup_params *rxq_init)
+{
+	struct bcm_rx_queue *rxq;
+
+	rxq = sc->rx_queues[fp->index];
+	if (!rxq) {
+		PMD_RX_LOG(ERR, "RX queue is NULL");
+		return;
+	}
+	/* pause */
+	pause->bd_th_lo = BD_TH_LO(sc);
+	pause->bd_th_hi = BD_TH_HI(sc);
+
+	pause->rcq_th_lo = RCQ_TH_LO(sc);
+	pause->rcq_th_hi = RCQ_TH_HI(sc);
+
+	/* validate rings have enough entries to cross high thresholds */
+	if (sc->dropless_fc &&
+	    pause->bd_th_hi + FW_PREFETCH_CNT > sc->rx_ring_size) {
+		PMD_DRV_LOG(WARN, "rx bd ring threshold limit");
+	}
+
+	if (sc->dropless_fc &&
+	    pause->rcq_th_hi + FW_PREFETCH_CNT > USABLE_RCQ_ENTRIES(rxq)) {
+		PMD_DRV_LOG(WARN, "rcq ring threshold limit");
+	}
+
+	pause->pri_map = 1;
+
+	/* rxq setup */
+	rxq_init->dscr_map = (phys_addr_t) ((void *)rxq->rx_ring_phys_addr);
+	rxq_init->rcq_map = (phys_addr_t) ((void *)rxq->cq_ring_phys_addr);
+	rxq_init->rcq_np_map = (phys_addr_t) ((void *)(rxq->cq_ring_phys_addr +
+						       BCM_PAGE_SIZE));
+
+	/*
+	 * This should be a maximum number of data bytes that may be
+	 * placed on the BD (not including paddings).
+	 */
+	rxq_init->buf_sz = (fp->rx_buf_size - IP_HEADER_ALIGNMENT_PADDING);
+
+	rxq_init->cl_qzone_id = fp->cl_qzone_id;
+	rxq_init->rss_engine_id = SC_FUNC(sc);
+	rxq_init->mcast_engine_id = SC_FUNC(sc);
+
+	rxq_init->cache_line_log = BCM_RX_ALIGN_SHIFT;
+	rxq_init->fw_sb_id = fp->fw_sb_id;
+
+	rxq_init->sb_cq_index = HC_INDEX_ETH_RX_CQ_CONS;
+
+	/*
+	 * configure silent vlan removal
+	 * if multi function mode is afex, then mask default vlan
+	 */
+	if (IS_MF_AFEX(sc)) {
+		rxq_init->silent_removal_value =
+		    sc->devinfo.mf_info.afex_def_vlan_tag;
+		rxq_init->silent_removal_mask = EVL_VLID_MASK;
+	}
+}
+
+static void
+bcm_pf_tx_q_prep(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		 struct ecore_txq_setup_params *txq_init, uint8_t cos)
+{
+	struct bcm_tx_queue *txq = fp->sc->tx_queues[fp->index];
+
+	if (!txq) {
+		PMD_TX_LOG(ERR, "ERROR: TX queue is NULL");
+		return;
+	}
+	txq_init->dscr_map = (phys_addr_t) ((void *)txq->tx_ring_phys_addr);
+	txq_init->sb_cq_index = HC_INDEX_ETH_FIRST_TX_CQ_CONS + cos;
+	txq_init->traffic_type = LLFC_TRAFFIC_TYPE_NW;
+	txq_init->fw_sb_id = fp->fw_sb_id;
+
+	/*
+	 * set the TSS leading client id for TX classfication to the
+	 * leading RSS client id
+	 */
+	txq_init->tss_leading_cl_id = BCM_FP(sc, 0, cl_id);
+}
+
+/*
+ * This function performs 2 steps in a queue state machine:
+ *   1) RESET->INIT
+ *   2) INIT->SETUP
+ */
+static int
+bcm_setup_queue(struct bcm_softc *sc, struct bcm_fastpath *fp, uint8_t leading)
+{
+	struct ecore_queue_state_params q_params = { NULL };
+	struct ecore_queue_setup_params *setup_params = &q_params.params.setup;
+	int rc;
+
+	PMD_DRV_LOG(DEBUG, "setting up queue %d", fp->index);
+
+	bcm_ack_sb(sc, fp->igu_sb_id, USTORM_ID, 0, IGU_INT_ENABLE, 0);
+
+	q_params.q_obj = &BCM_SP_OBJ(sc, fp).q_obj;
+
+	/* we want to wait for completion in this context */
+	bcm_set_bit(RAMROD_COMP_WAIT, &q_params.ramrod_flags);
+
+	/* prepare the INIT parameters */
+	bcm_pf_q_prep_init(sc, fp, &q_params.params.init);
+
+	/* Set the command */
+	q_params.cmd = ECORE_Q_CMD_INIT;
+
+	/* Change the state to INIT */
+	rc = ecore_queue_state_change(sc, &q_params);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Queue(%d) INIT failed", fp->index);
+		return rc;
+	}
+
+	PMD_DRV_LOG(DEBUG, "init complete");
+
+	/* now move the Queue to the SETUP state */
+	memset(setup_params, 0, sizeof(*setup_params));
+
+	/* set Queue flags */
+	setup_params->flags = bcm_get_q_flags(sc, leading);
+
+	/* set general SETUP parameters */
+	bcm_pf_q_prep_general(sc, fp, &setup_params->gen_params,
+			      FIRST_TX_COS_INDEX);
+
+	bcm_pf_rx_q_prep(sc, fp,
+			 &setup_params->pause_params,
+			 &setup_params->rxq_params);
+
+	bcm_pf_tx_q_prep(sc, fp, &setup_params->txq_params, FIRST_TX_COS_INDEX);
+
+	/* Set the command */
+	q_params.cmd = ECORE_Q_CMD_SETUP;
+
+	/* change the state to SETUP */
+	rc = ecore_queue_state_change(sc, &q_params);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Queue(%d) SETUP failed", fp->index);
+		return rc;
+	}
+
+	return rc;
+}
+
+static int bcm_setup_leading(struct bcm_softc *sc)
+{
+	if (IS_PF(sc))
+		return bcm_setup_queue(sc, &sc->fp[0], TRUE);
+	else			/* VF */
+		return bcm_vf_setup_queue(sc, &sc->fp[0], TRUE);
+}
+
+static int
+bcm_config_rss_pf(struct bcm_softc *sc, struct ecore_rss_config_obj *rss_obj,
+		  uint8_t config_hash)
+{
+	struct ecore_config_rss_params params = { NULL };
+	uint32_t i;
+
+	/*
+	 * Although RSS is meaningless when there is a single HW queue we
+	 * still need it enabled in order to have HW Rx hash generated.
+	 */
+
+	params.rss_obj = rss_obj;
+
+	bcm_set_bit(RAMROD_COMP_WAIT, &params.ramrod_flags);
+
+	bcm_set_bit(ECORE_RSS_MODE_REGULAR, &params.rss_flags);
+
+	/* RSS configuration */
+	bcm_set_bit(ECORE_RSS_IPV4, &params.rss_flags);
+	bcm_set_bit(ECORE_RSS_IPV4_TCP, &params.rss_flags);
+	bcm_set_bit(ECORE_RSS_IPV6, &params.rss_flags);
+	bcm_set_bit(ECORE_RSS_IPV6_TCP, &params.rss_flags);
+	if (rss_obj->udp_rss_v4) {
+		bcm_set_bit(ECORE_RSS_IPV4_UDP, &params.rss_flags);
+	}
+	if (rss_obj->udp_rss_v6) {
+		bcm_set_bit(ECORE_RSS_IPV6_UDP, &params.rss_flags);
+	}
+
+	/* Hash bits */
+	params.rss_result_mask = MULTI_MASK;
+
+	(void)rte_memcpy(params.ind_table, rss_obj->ind_table,
+			 sizeof(params.ind_table));
+
+	if (config_hash) {
+/* RSS keys */
+		for (i = 0; i < sizeof(params.rss_key) / 4; i++) {
+			params.rss_key[i] = (uint32_t) rte_rand();
+		}
+
+		bcm_set_bit(ECORE_RSS_SET_SRCH, &params.rss_flags);
+	}
+
+	if (IS_PF(sc))
+		return ecore_config_rss(sc, &params);
+	else
+		return bcm_vf_config_rss(sc, &params);
+}
+
+static int bcm_config_rss_eth(struct bcm_softc *sc, uint8_t config_hash)
+{
+	return bcm_config_rss_pf(sc, &sc->rss_conf_obj, config_hash);
+}
+
+static int bcm_init_rss_pf(struct bcm_softc *sc)
+{
+	uint8_t num_eth_queues = BCM_NUM_ETH_QUEUES(sc);
+	uint32_t i;
+
+	/*
+	 * Prepare the initial contents of the indirection table if
+	 * RSS is enabled
+	 */
+	for (i = 0; i < sizeof(sc->rss_conf_obj.ind_table); i++) {
+		sc->rss_conf_obj.ind_table[i] =
+		    (sc->fp->cl_id + (i % num_eth_queues));
+	}
+
+	if (sc->udp_rss) {
+		sc->rss_conf_obj.udp_rss_v4 = sc->rss_conf_obj.udp_rss_v6 = 1;
+	}
+
+	/*
+	 * For 57711 SEARCHER configuration (rss_keys) is
+	 * per-port, so if explicit configuration is needed, do it only
+	 * for a PMF.
+	 *
+	 * For 57712 and newer it's a per-function configuration.
+	 */
+	return bcm_config_rss_eth(sc, sc->port.pmf || !CHIP_IS_E1x(sc));
+}
+
+static int
+bcm_set_mac_one(struct bcm_softc *sc, uint8_t * mac,
+		struct ecore_vlan_mac_obj *obj, uint8_t set, int mac_type,
+		unsigned long *ramrod_flags)
+{
+	struct ecore_vlan_mac_ramrod_params ramrod_param;
+	int rc;
+
+	memset(&ramrod_param, 0, sizeof(ramrod_param));
+
+	/* fill in general parameters */
+	ramrod_param.vlan_mac_obj = obj;
+	ramrod_param.ramrod_flags = *ramrod_flags;
+
+	/* fill a user request section if needed */
+	if (!bcm_test_bit(RAMROD_CONT, ramrod_flags)) {
+		(void)rte_memcpy(ramrod_param.user_req.u.mac.mac, mac,
+				 ETH_ALEN);
+
+		bcm_set_bit(mac_type, &ramrod_param.user_req.vlan_mac_flags);
+
+/* Set the command: ADD or DEL */
+		ramrod_param.user_req.cmd = (set) ? ECORE_VLAN_MAC_ADD :
+		    ECORE_VLAN_MAC_DEL;
+	}
+
+	rc = ecore_config_vlan_mac(sc, &ramrod_param);
+
+	if (rc == ECORE_EXISTS) {
+		PMD_DRV_LOG(INFO, "Failed to schedule ADD operations (EEXIST)");
+/* do not treat adding same MAC as error */
+		rc = 0;
+	} else if (rc < 0) {
+		PMD_DRV_LOG(ERR,
+			    "%s MAC failed (%d)", (set ? "Set" : "Delete"), rc);
+	}
+
+	return rc;
+}
+
+static int bcm_set_eth_mac(struct bcm_softc *sc, uint8_t set)
+{
+	unsigned long ramrod_flags = 0;
+
+	PMD_DRV_LOG(DEBUG, "Adding Ethernet MAC");
+
+	bcm_set_bit(RAMROD_COMP_WAIT, &ramrod_flags);
+
+	/* Eth MAC is set on RSS leading client (fp[0]) */
+	return bcm_set_mac_one(sc, sc->link_params.mac_addr,
+			       &sc->sp_objs->mac_obj,
+			       set, ECORE_ETH_MAC, &ramrod_flags);
+}
+
+static int bcm_get_cur_phy_idx(struct bcm_softc *sc)
+{
+	uint32_t sel_phy_idx = 0;
+
+	if (sc->link_params.num_phys <= 1) {
+		return ELINK_INT_PHY;
+	}
+
+	if (sc->link_vars.link_up) {
+		sel_phy_idx = ELINK_EXT_PHY1;
+/* In case link is SERDES, check if the ELINK_EXT_PHY2 is the one */
+		if ((sc->link_vars.link_status & LINK_STATUS_SERDES_LINK) &&
+		    (sc->link_params.phy[ELINK_EXT_PHY2].supported &
+		     ELINK_SUPPORTED_FIBRE))
+			sel_phy_idx = ELINK_EXT_PHY2;
+	} else {
+		switch (elink_phy_selection(&sc->link_params)) {
+		case PORT_HW_CFG_PHY_SELECTION_HARDWARE_DEFAULT:
+		case PORT_HW_CFG_PHY_SELECTION_FIRST_PHY:
+		case PORT_HW_CFG_PHY_SELECTION_FIRST_PHY_PRIORITY:
+			sel_phy_idx = ELINK_EXT_PHY1;
+			break;
+		case PORT_HW_CFG_PHY_SELECTION_SECOND_PHY:
+		case PORT_HW_CFG_PHY_SELECTION_SECOND_PHY_PRIORITY:
+			sel_phy_idx = ELINK_EXT_PHY2;
+			break;
+		}
+	}
+
+	return sel_phy_idx;
+}
+
+static int bcm_get_link_cfg_idx(struct bcm_softc *sc)
+{
+	uint32_t sel_phy_idx = bcm_get_cur_phy_idx(sc);
+
+	/*
+	 * The selected activated PHY is always after swapping (in case PHY
+	 * swapping is enabled). So when swapping is enabled, we need to reverse
+	 * the configuration
+	 */
+
+	if (sc->link_params.multi_phy_config & PORT_HW_CFG_PHY_SWAPPED_ENABLED) {
+		if (sel_phy_idx == ELINK_EXT_PHY1)
+			sel_phy_idx = ELINK_EXT_PHY2;
+		else if (sel_phy_idx == ELINK_EXT_PHY2)
+			sel_phy_idx = ELINK_EXT_PHY1;
+	}
+
+	return ELINK_LINK_CONFIG_IDX(sel_phy_idx);
+}
+
+static void bcm_set_requested_fc(struct bcm_softc *sc)
+{
+	/*
+	 * Initialize link parameters structure variables
+	 * It is recommended to turn off RX FC for jumbo frames
+	 * for better performance
+	 */
+	if (CHIP_IS_E1x(sc) && (sc->mtu > 5000)) {
+		sc->link_params.req_fc_auto_adv = ELINK_FLOW_CTRL_TX;
+	} else {
+		sc->link_params.req_fc_auto_adv = ELINK_FLOW_CTRL_BOTH;
+	}
+}
+
+static void bcm_calc_fc_adv(struct bcm_softc *sc)
+{
+	uint8_t cfg_idx = bcm_get_link_cfg_idx(sc);
+	switch (sc->link_vars.ieee_fc &
+		MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_MASK) {
+	case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_NONE:
+	default:
+		sc->port.advertising[cfg_idx] &= ~(ADVERTISED_Asym_Pause |
+						   ADVERTISED_Pause);
+		break;
+
+	case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_BOTH:
+		sc->port.advertising[cfg_idx] |= (ADVERTISED_Asym_Pause |
+						  ADVERTISED_Pause);
+		break;
+
+	case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_ASYMMETRIC:
+		sc->port.advertising[cfg_idx] |= ADVERTISED_Asym_Pause;
+		break;
+	}
+}
+
+static uint16_t bcm_get_mf_speed(struct bcm_softc *sc)
+{
+	uint16_t line_speed = sc->link_vars.line_speed;
+	if (IS_MF(sc)) {
+		uint16_t maxCfg = bcm_extract_max_cfg(sc,
+						      sc->devinfo.
+						      mf_info.mf_config[SC_VN
+									(sc)]);
+
+/* calculate the current MAX line speed limit for the MF devices */
+		if (IS_MF_SI(sc)) {
+			line_speed = (line_speed * maxCfg) / 100;
+		} else {	/* SD mode */
+			uint16_t vn_max_rate = maxCfg * 100;
+
+			if (vn_max_rate < line_speed) {
+				line_speed = vn_max_rate;
+			}
+		}
+	}
+
+	return line_speed;
+}
+
+static void
+bcm_fill_report_data(struct bcm_softc *sc, struct bcm_link_report_data *data)
+{
+	uint16_t line_speed = bcm_get_mf_speed(sc);
+
+	memset(data, 0, sizeof(*data));
+
+	/* fill the report data with the effective line speed */
+	data->line_speed = line_speed;
+
+	/* Link is down */
+	if (!sc->link_vars.link_up || (sc->flags & BCM_MF_FUNC_DIS)) {
+		bcm_set_bit(BCM_LINK_REPORT_LINK_DOWN,
+			    &data->link_report_flags);
+	}
+
+	/* Full DUPLEX */
+	if (sc->link_vars.duplex == DUPLEX_FULL) {
+		bcm_set_bit(BCM_LINK_REPORT_FULL_DUPLEX,
+			    &data->link_report_flags);
+	}
+
+	/* Rx Flow Control is ON */
+	if (sc->link_vars.flow_ctrl & ELINK_FLOW_CTRL_RX) {
+		bcm_set_bit(BCM_LINK_REPORT_RX_FC_ON, &data->link_report_flags);
+	}
+
+	/* Tx Flow Control is ON */
+	if (sc->link_vars.flow_ctrl & ELINK_FLOW_CTRL_TX) {
+		bcm_set_bit(BCM_LINK_REPORT_TX_FC_ON, &data->link_report_flags);
+	}
+}
+
+/* report link status to OS, should be called under phy_lock */
+static void bcm_link_report(struct bcm_softc *sc)
+{
+	struct bcm_link_report_data cur_data;
+
+	/* reread mf_cfg */
+	if (IS_PF(sc)) {
+		bcm_read_mf_cfg(sc);
+	}
+
+	/* Read the current link report info */
+	bcm_fill_report_data(sc, &cur_data);
+
+	/* Don't report link down or exactly the same link status twice */
+	if (!memcmp(&cur_data, &sc->last_reported_link, sizeof(cur_data)) ||
+	    (bcm_test_bit(BCM_LINK_REPORT_LINK_DOWN,
+			  &sc->last_reported_link.link_report_flags) &&
+	     bcm_test_bit(BCM_LINK_REPORT_LINK_DOWN,
+			  &cur_data.link_report_flags))) {
+		return;
+	}
+
+	sc->link_cnt++;
+
+	/* report new link params and remember the state for the next time */
+	(void)rte_memcpy(&sc->last_reported_link, &cur_data, sizeof(cur_data));
+
+	if (bcm_test_bit(BCM_LINK_REPORT_LINK_DOWN,
+			 &cur_data.link_report_flags)) {
+		PMD_DRV_LOG(INFO, "NIC Link is Down");
+	} else {
+		__rte_unused const char *duplex;
+		__rte_unused const char *flow;
+
+		if (bcm_test_and_clear_bit(BCM_LINK_REPORT_FULL_DUPLEX,
+					   &cur_data.link_report_flags)) {
+			duplex = "full";
+		} else {
+			duplex = "half";
+		}
+
+/*
+ * Handle the FC at the end so that only these flags would be
+ * possibly set. This way we may easily check if there is no FC
+ * enabled.
+ */
+		if (cur_data.link_report_flags) {
+			if (bcm_test_bit(BCM_LINK_REPORT_RX_FC_ON,
+					 &cur_data.link_report_flags) &&
+			    bcm_test_bit(BCM_LINK_REPORT_TX_FC_ON,
+					 &cur_data.link_report_flags)) {
+				flow = "ON - receive & transmit";
+			} else if (bcm_test_bit(BCM_LINK_REPORT_RX_FC_ON,
+						&cur_data.link_report_flags) &&
+				   !bcm_test_bit(BCM_LINK_REPORT_TX_FC_ON,
+						 &cur_data.link_report_flags)) {
+				flow = "ON - receive";
+			} else if (!bcm_test_bit(BCM_LINK_REPORT_RX_FC_ON,
+						 &cur_data.link_report_flags) &&
+				   bcm_test_bit(BCM_LINK_REPORT_TX_FC_ON,
+						&cur_data.link_report_flags)) {
+				flow = "ON - transmit";
+			} else {
+				flow = "none";	/* possible? */
+			}
+		} else {
+			flow = "none";
+		}
+
+		PMD_DRV_LOG(INFO,
+			    "NIC Link is Up, %d Mbps %s duplex, Flow control: %s",
+			    cur_data.line_speed, duplex, flow);
+	}
+}
+
+void bcm_link_status_update(struct bcm_softc *sc)
+{
+	if (sc->state != BCM_STATE_OPEN) {
+		return;
+	}
+
+	if (IS_PF(sc) && !CHIP_REV_IS_SLOW(sc)) {
+		elink_link_status_update(&sc->link_params, &sc->link_vars);
+	} else {
+		sc->port.supported[0] |= (ELINK_SUPPORTED_10baseT_Half |
+					  ELINK_SUPPORTED_10baseT_Full |
+					  ELINK_SUPPORTED_100baseT_Half |
+					  ELINK_SUPPORTED_100baseT_Full |
+					  ELINK_SUPPORTED_1000baseT_Full |
+					  ELINK_SUPPORTED_2500baseX_Full |
+					  ELINK_SUPPORTED_10000baseT_Full |
+					  ELINK_SUPPORTED_TP |
+					  ELINK_SUPPORTED_FIBRE |
+					  ELINK_SUPPORTED_Autoneg |
+					  ELINK_SUPPORTED_Pause |
+					  ELINK_SUPPORTED_Asym_Pause);
+		sc->port.advertising[0] = sc->port.supported[0];
+
+		sc->link_params.sc = sc;
+		sc->link_params.port = SC_PORT(sc);
+		sc->link_params.req_duplex[0] = DUPLEX_FULL;
+		sc->link_params.req_flow_ctrl[0] = ELINK_FLOW_CTRL_NONE;
+		sc->link_params.req_line_speed[0] = SPEED_10000;
+		sc->link_params.speed_cap_mask[0] = 0x7f0000;
+		sc->link_params.switch_cfg = ELINK_SWITCH_CFG_10G;
+
+		if (CHIP_REV_IS_FPGA(sc)) {
+			sc->link_vars.mac_type = ELINK_MAC_TYPE_EMAC;
+			sc->link_vars.line_speed = ELINK_SPEED_1000;
+			sc->link_vars.link_status = (LINK_STATUS_LINK_UP |
+						     LINK_STATUS_SPEED_AND_DUPLEX_1000TFD);
+		} else {
+			sc->link_vars.mac_type = ELINK_MAC_TYPE_BMAC;
+			sc->link_vars.line_speed = ELINK_SPEED_10000;
+			sc->link_vars.link_status = (LINK_STATUS_LINK_UP |
+						     LINK_STATUS_SPEED_AND_DUPLEX_10GTFD);
+		}
+
+		sc->link_vars.link_up = 1;
+
+		sc->link_vars.duplex = DUPLEX_FULL;
+		sc->link_vars.flow_ctrl = ELINK_FLOW_CTRL_NONE;
+
+		if (IS_PF(sc)) {
+			REG_WR(sc,
+			       NIG_REG_EGRESS_DRAIN0_MODE +
+			       sc->link_params.port * 4, 0);
+			bcm_stats_handle(sc, STATS_EVENT_LINK_UP);
+			bcm_link_report(sc);
+		}
+	}
+
+	if (IS_PF(sc)) {
+		if (sc->link_vars.link_up) {
+			bcm_stats_handle(sc, STATS_EVENT_LINK_UP);
+		} else {
+			bcm_stats_handle(sc, STATS_EVENT_STOP);
+		}
+		bcm_link_report(sc);
+	} else {
+		bcm_link_report(sc);
+		bcm_stats_handle(sc, STATS_EVENT_LINK_UP);
+	}
+}
+
+static void bcm_periodic_start(struct bcm_softc *sc)
+{
+	atomic_store_rel_long(&sc->periodic_flags, PERIODIC_GO);
+}
+
+static void bcm_periodic_stop(struct bcm_softc *sc)
+{
+	atomic_store_rel_long(&sc->periodic_flags, PERIODIC_STOP);
+}
+
+static int bcm_initial_phy_init(struct bcm_softc *sc, int load_mode)
+{
+	int rc, cfg_idx = bcm_get_link_cfg_idx(sc);
+	uint16_t req_line_speed = sc->link_params.req_line_speed[cfg_idx];
+	struct elink_params *lp = &sc->link_params;
+
+	bcm_set_requested_fc(sc);
+
+	if (CHIP_REV_IS_SLOW(sc)) {
+		uint32_t bond = CHIP_BOND_ID(sc);
+		uint32_t feat = 0;
+
+		if (CHIP_IS_E2(sc) && CHIP_IS_MODE_4_PORT(sc)) {
+			feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_BMAC;
+		} else if (bond & 0x4) {
+			if (CHIP_IS_E3(sc)) {
+				feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_XMAC;
+			} else {
+				feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_BMAC;
+			}
+		} else if (bond & 0x8) {
+			if (CHIP_IS_E3(sc)) {
+				feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_UMAC;
+			} else {
+				feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_EMAC;
+			}
+		}
+
+/* disable EMAC for E3 and above */
+		if (bond & 0x2) {
+			feat |= ELINK_FEATURE_CONFIG_EMUL_DISABLE_EMAC;
+		}
+
+		sc->link_params.feature_config_flags |= feat;
+	}
+
+	if (load_mode == LOAD_DIAG) {
+		lp->loopback_mode = ELINK_LOOPBACK_XGXS;
+/* Prefer doing PHY loopback at 10G speed, if possible */
+		if (lp->req_line_speed[cfg_idx] < ELINK_SPEED_10000) {
+			if (lp->speed_cap_mask[cfg_idx] &
+			    PORT_HW_CFG_SPEED_CAPABILITY_D0_10G) {
+				lp->req_line_speed[cfg_idx] = ELINK_SPEED_10000;
+			} else {
+				lp->req_line_speed[cfg_idx] = ELINK_SPEED_1000;
+			}
+		}
+	}
+
+	if (load_mode == LOAD_LOOPBACK_EXT) {
+		lp->loopback_mode = ELINK_LOOPBACK_EXT;
+	}
+
+	rc = elink_phy_init(&sc->link_params, &sc->link_vars);
+
+	bcm_calc_fc_adv(sc);
+
+	if (sc->link_vars.link_up) {
+		bcm_stats_handle(sc, STATS_EVENT_LINK_UP);
+		bcm_link_report(sc);
+	}
+
+	if (!CHIP_REV_IS_SLOW(sc)) {
+		bcm_periodic_start(sc);
+	}
+
+	sc->link_params.req_line_speed[cfg_idx] = req_line_speed;
+	return rc;
+}
+
+/* update flags in shmem */
+static void
+bcm_update_drv_flags(struct bcm_softc *sc, uint32_t flags, uint32_t set)
+{
+	uint32_t drv_flags;
+
+	if (SHMEM2_HAS(sc, drv_flags)) {
+		bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_DRV_FLAGS);
+		drv_flags = SHMEM2_RD(sc, drv_flags);
+
+		if (set) {
+			drv_flags |= flags;
+		} else {
+			drv_flags &= ~flags;
+		}
+
+		SHMEM2_WR(sc, drv_flags, drv_flags);
+
+		bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_DRV_FLAGS);
+	}
+}
+
+/* periodic timer callout routine, only runs when the interface is up */
+void bcm_periodic_callout(struct bcm_softc *sc)
+{
+	if ((sc->state != BCM_STATE_OPEN) ||
+	    (atomic_load_acq_long(&sc->periodic_flags) == PERIODIC_STOP)) {
+		PMD_DRV_LOG(WARN, "periodic callout exit (state=0x%x)",
+			    sc->state);
+		return;
+	}
+	if (!CHIP_REV_IS_SLOW(sc)) {
+/*
+ * This barrier is needed to ensure the ordering between the writing
+ * to the sc->port.pmf in the bcm_nic_load() or bcm_pmf_update() and
+ * the reading here.
+ */
+		mb();
+		if (sc->port.pmf) {
+			elink_period_func(&sc->link_params, &sc->link_vars);
+		}
+	}
+#ifdef BCM_PULSE
+	if (IS_PF(sc) && !BCM_NOMCP(sc)) {
+		int mb_idx = SC_FW_MB_IDX(sc);
+		uint32_t drv_pulse;
+		uint32_t mcp_pulse;
+
+		++sc->fw_drv_pulse_wr_seq;
+		sc->fw_drv_pulse_wr_seq &= DRV_PULSE_SEQ_MASK;
+
+		drv_pulse = sc->fw_drv_pulse_wr_seq;
+		bcm_drv_pulse(sc);
+
+		mcp_pulse = (SHMEM_RD(sc, func_mb[mb_idx].mcp_pulse_mb) &
+			     MCP_PULSE_SEQ_MASK);
+
+/*
+ * The delta between driver pulse and mcp response should
+ * be 1 (before mcp response) or 0 (after mcp response).
+ */
+		if ((drv_pulse != mcp_pulse) &&
+		    (drv_pulse != ((mcp_pulse + 1) & MCP_PULSE_SEQ_MASK))) {
+			/* someone lost a heartbeat... */
+			PMD_DRV_LOG(ERR,
+				    "drv_pulse (0x%x) != mcp_pulse (0x%x)",
+				    drv_pulse, mcp_pulse);
+		}
+	}
+#endif
+}
+
+/* start the controller */
+static __attribute__ ((noinline))
+int bcm_nic_load(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint32_t load_code = 0;
+	int i, rc = 0;
+
+	PMD_INIT_FUNC_TRACE();
+
+	sc->state = BCM_STATE_OPENING_WAITING_LOAD;
+
+	if (IS_PF(sc)) {
+/* must be called before memory allocation and HW init */
+		bcm_ilt_set_info(sc);
+	}
+
+	bcm_set_fp_rx_buf_size(sc);
+
+	if (IS_PF(sc)) {
+		if (bcm_alloc_mem(sc) != 0) {
+			sc->state = BCM_STATE_CLOSED;
+			rc = -ENOMEM;
+			goto bcm_nic_load_error0;
+		}
+	}
+
+	if (bcm_alloc_fw_stats_mem(sc) != 0) {
+		sc->state = BCM_STATE_CLOSED;
+		rc = -ENOMEM;
+		goto bcm_nic_load_error0;
+	}
+
+	if (IS_VF(sc)) {
+		rc = bcm_vf_init(sc);
+		if (rc) {
+			sc->state = BCM_STATE_ERROR;
+			goto bcm_nic_load_error0;
+		}
+	}
+
+	if (IS_PF(sc)) {
+/* set pf load just before approaching the MCP */
+		bcm_set_pf_load(sc);
+
+/* if MCP exists send load request and analyze response */
+		if (!BCM_NOMCP(sc)) {
+			/* attempt to load pf */
+			if (bcm_nic_load_request(sc, &load_code) != 0) {
+				sc->state = BCM_STATE_CLOSED;
+				rc = -ENXIO;
+				goto bcm_nic_load_error1;
+			}
+
+			/* what did the MCP say? */
+			if (bcm_nic_load_analyze_req(sc, load_code) != 0) {
+				bcm_fw_command(sc, DRV_MSG_CODE_LOAD_DONE, 0);
+				sc->state = BCM_STATE_CLOSED;
+				rc = -ENXIO;
+				goto bcm_nic_load_error2;
+			}
+		} else {
+			PMD_DRV_LOG(INFO, "Device has no MCP!");
+			load_code = bcm_nic_load_no_mcp(sc);
+		}
+
+/* mark PMF if applicable */
+		bcm_nic_load_pmf(sc, load_code);
+
+/* Init Function state controlling object */
+		bcm_init_func_obj(sc);
+
+/* Initialize HW */
+		if (bcm_init_hw(sc, load_code) != 0) {
+			PMD_DRV_LOG(NOTICE, "HW init failed");
+			bcm_fw_command(sc, DRV_MSG_CODE_LOAD_DONE, 0);
+			sc->state = BCM_STATE_CLOSED;
+			rc = -ENXIO;
+			goto bcm_nic_load_error2;
+		}
+	}
+
+	bcm_nic_init(sc, load_code);
+
+	/* Init per-function objects */
+	if (IS_PF(sc)) {
+		bcm_init_objs(sc);
+
+/* set AFEX default VLAN tag to an invalid value */
+		sc->devinfo.mf_info.afex_def_vlan_tag = -1;
+
+		sc->state = BCM_STATE_OPENING_WAITING_PORT;
+		rc = bcm_func_start(sc);
+		if (rc) {
+			PMD_DRV_LOG(NOTICE, "Function start failed!");
+			bcm_fw_command(sc, DRV_MSG_CODE_LOAD_DONE, 0);
+			sc->state = BCM_STATE_ERROR;
+			goto bcm_nic_load_error3;
+		}
+
+/* send LOAD_DONE command to MCP */
+		if (!BCM_NOMCP(sc)) {
+			load_code =
+			    bcm_fw_command(sc, DRV_MSG_CODE_LOAD_DONE, 0);
+			if (!load_code) {
+				PMD_DRV_LOG(NOTICE,
+					    "MCP response failure, aborting");
+				sc->state = BCM_STATE_ERROR;
+				rc = -ENXIO;
+				goto bcm_nic_load_error3;
+			}
+		}
+	}
+
+	rc = bcm_setup_leading(sc);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Setup leading failed!");
+		sc->state = BCM_STATE_ERROR;
+		goto bcm_nic_load_error3;
+	}
+
+	FOR_EACH_NONDEFAULT_ETH_QUEUE(sc, i) {
+		if (IS_PF(sc))
+			rc = bcm_setup_queue(sc, &sc->fp[i], FALSE);
+		else		/* IS_VF(sc) */
+			rc = bcm_vf_setup_queue(sc, &sc->fp[i], FALSE);
+
+		if (rc) {
+			PMD_DRV_LOG(NOTICE, "Queue(%d) setup failed", i);
+			sc->state = BCM_STATE_ERROR;
+			goto bcm_nic_load_error3;
+		}
+	}
+
+	rc = bcm_init_rss_pf(sc);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "PF RSS init failed");
+		sc->state = BCM_STATE_ERROR;
+		goto bcm_nic_load_error3;
+	}
+
+	/* now when Clients are configured we are ready to work */
+	sc->state = BCM_STATE_OPEN;
+
+	/* Configure a ucast MAC */
+	if (IS_PF(sc)) {
+		rc = bcm_set_eth_mac(sc, TRUE);
+	} else {		/* IS_VF(sc) */
+		rc = bcm_vf_set_mac(sc, TRUE);
+	}
+
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Setting Ethernet MAC failed");
+		sc->state = BCM_STATE_ERROR;
+		goto bcm_nic_load_error3;
+	}
+
+	if (sc->port.pmf) {
+		rc = bcm_initial_phy_init(sc, LOAD_OPEN);
+		if (rc) {
+			sc->state = BCM_STATE_ERROR;
+			goto bcm_nic_load_error3;
+		}
+	}
+
+	sc->link_params.feature_config_flags &=
+	    ~ELINK_FEATURE_CONFIG_BOOT_FROM_SAN;
+
+	/* start the Tx */
+	switch (LOAD_OPEN) {
+	case LOAD_NORMAL:
+	case LOAD_OPEN:
+		break;
+
+	case LOAD_DIAG:
+	case LOAD_LOOPBACK_EXT:
+		sc->state = BCM_STATE_DIAG;
+		break;
+
+	default:
+		break;
+	}
+
+	if (sc->port.pmf) {
+		bcm_update_drv_flags(sc, 1 << DRV_FLAGS_PORT_MASK, 0);
+	} else {
+		bcm_link_status_update(sc);
+	}
+
+	if (IS_PF(sc) && SHMEM2_HAS(sc, drv_capabilities_flag)) {
+/* mark driver is loaded in shmem2 */
+		val = SHMEM2_RD(sc, drv_capabilities_flag[SC_FW_MB_IDX(sc)]);
+		SHMEM2_WR(sc, drv_capabilities_flag[SC_FW_MB_IDX(sc)],
+			  (val |
+			   DRV_FLAGS_CAPABILITIES_LOADED_SUPPORTED |
+			   DRV_FLAGS_CAPABILITIES_LOADED_L2));
+	}
+
+	/* start fast path */
+	/* Initialize Rx filter */
+	bcm_set_rx_mode(sc);
+
+	/* wait for all pending SP commands to complete */
+	if (IS_PF(sc) && !bcm_wait_sp_comp(sc, ~0x0UL)) {
+		PMD_DRV_LOG(NOTICE, "Timeout waiting for all SPs to complete!");
+		bcm_periodic_stop(sc);
+		bcm_nic_unload(sc, UNLOAD_CLOSE, FALSE);
+		return -ENXIO;
+	}
+
+	PMD_DRV_LOG(DEBUG, "NIC successfully loaded");
+
+	return 0;
+
+bcm_nic_load_error3:
+
+	if (IS_PF(sc)) {
+		bcm_int_disable_sync(sc, 1);
+
+/* clean out queued objects */
+		bcm_squeeze_objects(sc);
+	}
+
+bcm_nic_load_error2:
+
+	if (IS_PF(sc) && !BCM_NOMCP(sc)) {
+		bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_REQ_WOL_MCP, 0);
+		bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_DONE, 0);
+	}
+
+	sc->port.pmf = 0;
+
+bcm_nic_load_error1:
+
+	/* clear pf_load status, as it was already set */
+	if (IS_PF(sc)) {
+		bcm_clear_pf_load(sc);
+	}
+
+bcm_nic_load_error0:
+
+	bcm_free_fw_stats_mem(sc);
+	bcm_free_mem(sc);
+
+	return rc;
+}
+
+/*
+* Handles controller initialization.
+*/
+int bcm_init(struct bcm_softc *sc)
+{
+	int other_engine = SC_PATH(sc) ? 0 : 1;
+	uint8_t other_load_status, load_status;
+	uint8_t global = FALSE;
+	int rc;
+
+	/* Check if the driver is still running and bail out if it is. */
+	if (sc->link_vars.link_up) {
+		PMD_DRV_LOG(DEBUG, "Init called while driver is running!");
+		rc = 0;
+		goto bcm_init_done;
+	}
+
+	bcm_set_power_state(sc, PCI_PM_D0);
+
+	/*
+	 * If parity occurred during the unload, then attentions and/or
+	 * RECOVERY_IN_PROGRESS may still be set. If so we want the first function
+	 * loaded on the current engine to complete the recovery. Parity recovery
+	 * is only relevant for PF driver.
+	 */
+	if (IS_PF(sc)) {
+		other_load_status = bcm_get_load_status(sc, other_engine);
+		load_status = bcm_get_load_status(sc, SC_PATH(sc));
+
+		if (!bcm_reset_is_done(sc, SC_PATH(sc)) ||
+		    bcm_chk_parity_attn(sc, &global, TRUE)) {
+			do {
+				/*
+				 * If there are attentions and they are in global blocks, set
+				 * the GLOBAL_RESET bit regardless whether it will be this
+				 * function that will complete the recovery or not.
+				 */
+				if (global) {
+					bcm_set_reset_global(sc);
+				}
+
+				/*
+				 * Only the first function on the current engine should try
+				 * to recover in open. In case of attentions in global blocks
+				 * only the first in the chip should try to recover.
+				 */
+				if ((!load_status
+				     && (!global ||!other_load_status))
+				    && bcm_trylock_leader_lock(sc)
+				    && !bcm_leader_reset(sc)) {
+					PMD_DRV_LOG(INFO,
+						    "Recovered during init");
+					break;
+				}
+
+				/* recovery has failed... */
+				bcm_set_power_state(sc, PCI_PM_D3hot);
+
+				sc->recovery_state = BCM_RECOVERY_FAILED;
+
+				PMD_DRV_LOG(NOTICE,
+					    "Recovery flow hasn't properly "
+					    "completed yet, try again later. "
+					    "If you still see this message after a "
+					    "few retries then power cycle is required.");
+
+				rc = -ENXIO;
+				goto bcm_init_done;
+			} while (0);
+		}
+	}
+
+	sc->recovery_state = BCM_RECOVERY_DONE;
+
+	rc = bcm_nic_load(sc);
+
+bcm_init_done:
+
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "Initialization failed, "
+			    "stack notified driver is NOT running!");
+	}
+
+	return rc;
+}
+
+static void bcm_get_function_num(struct bcm_softc *sc)
+{
+	uint32_t val = 0;
+
+	/*
+	 * Read the ME register to get the function number. The ME register
+	 * holds the relative-function number and absolute-function number. The
+	 * absolute-function number appears only in E2 and above. Before that
+	 * these bits always contained zero, therefore we cannot blindly use them.
+	 */
+
+	val = REG_RD(sc, BAR_ME_REGISTER);
+
+	sc->pfunc_rel =
+	    (uint8_t) ((val & ME_REG_PF_NUM) >> ME_REG_PF_NUM_SHIFT);
+	sc->path_id =
+	    (uint8_t) ((val & ME_REG_ABS_PF_NUM) >> ME_REG_ABS_PF_NUM_SHIFT) &
+	    1;
+
+	if (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) {
+		sc->pfunc_abs = ((sc->pfunc_rel << 1) | sc->path_id);
+	} else {
+		sc->pfunc_abs = (sc->pfunc_rel | sc->path_id);
+	}
+
+	PMD_DRV_LOG(DEBUG,
+		    "Relative function %d, Absolute function %d, Path %d",
+		    sc->pfunc_rel, sc->pfunc_abs, sc->path_id);
+}
+
+static uint32_t bcm_get_shmem_mf_cfg_base(struct bcm_softc *sc)
+{
+	uint32_t shmem2_size;
+	uint32_t offset;
+	uint32_t mf_cfg_offset_value;
+
+	/* Non 57712 */
+	offset = (SHMEM_ADDR(sc, func_mb) +
+		  (MAX_FUNC_NUM * sizeof(struct drv_func_mb)));
+
+	/* 57712 plus */
+	if (sc->devinfo.shmem2_base != 0) {
+		shmem2_size = SHMEM2_RD(sc, size);
+		if (shmem2_size > offsetof(struct shmem2_region, mf_cfg_addr)) {
+			mf_cfg_offset_value = SHMEM2_RD(sc, mf_cfg_addr);
+			if (SHMEM_MF_CFG_ADDR_NONE != mf_cfg_offset_value) {
+				offset = mf_cfg_offset_value;
+			}
+		}
+	}
+
+	return offset;
+}
+
+static uint32_t bcm_pcie_capability_read(struct bcm_softc *sc, int reg)
+{
+	uint32_t ret;
+	struct bcm_pci_cap *caps;
+
+	/* ensure PCIe capability is enabled */
+	caps = pci_find_cap(sc, PCIY_EXPRESS, BCM_PCI_CAP);
+	if (NULL != caps) {
+		PMD_DRV_LOG(DEBUG, "Found PCIe capability: "
+			    "id=0x%04X type=0x%04X addr=0x%08X",
+			    caps->id, caps->type, caps->addr);
+		pci_read(sc, (caps->addr + reg), &ret, 2);
+		return ret;
+	}
+
+	PMD_DRV_LOG(WARN, "PCIe capability NOT FOUND!!!");
+
+	return 0;
+}
+
+static uint8_t bcm_is_pcie_pending(struct bcm_softc *sc)
+{
+	return (bcm_pcie_capability_read(sc, PCIR_EXPRESS_DEVICE_STA) &
+		PCIM_EXP_STA_TRANSACTION_PND);
+}
+
+/*
+* Walk the PCI capabiites list for the device to find what features are
+* supported. These capabilites may be enabled/disabled by firmware so it's
+* best to walk the list rather than make assumptions.
+*/
+static void bcm_probe_pci_caps(struct bcm_softc *sc)
+{
+	PMD_INIT_FUNC_TRACE();
+
+	struct bcm_pci_cap *caps;
+	uint16_t link_status;
+#ifdef RTE_LIBRTE_BCM_DEBUG
+	int reg = 0;
+#endif
+
+	/* check if PCI Power Management is enabled */
+	caps = pci_find_cap(sc, PCIY_PMG, BCM_PCI_CAP);
+	if (NULL != caps) {
+		PMD_DRV_LOG(DEBUG, "Found PM capability: "
+			    "id=0x%04X type=0x%04X addr=0x%08X",
+			    caps->id, caps->type, caps->addr);
+
+		sc->devinfo.pcie_cap_flags |= BCM_PM_CAPABLE_FLAG;
+		sc->devinfo.pcie_pm_cap_reg = caps->addr;
+	}
+
+	link_status = bcm_pcie_capability_read(sc, PCIR_EXPRESS_LINK_STA);
+
+	sc->devinfo.pcie_link_speed = (link_status & PCIM_LINK_STA_SPEED);
+	sc->devinfo.pcie_link_width =
+	    ((link_status & PCIM_LINK_STA_WIDTH) >> 4);
+
+	PMD_DRV_LOG(DEBUG, "PCIe link speed=%d width=%d",
+		    sc->devinfo.pcie_link_speed, sc->devinfo.pcie_link_width);
+
+	sc->devinfo.pcie_cap_flags |= BCM_PCIE_CAPABLE_FLAG;
+
+	/* check if MSI capability is enabled */
+	caps = pci_find_cap(sc, PCIY_MSI, BCM_PCI_CAP);
+	if (NULL != caps) {
+		PMD_DRV_LOG(DEBUG, "Found MSI capability at 0x%04x", reg);
+
+		sc->devinfo.pcie_cap_flags |= BCM_MSI_CAPABLE_FLAG;
+		sc->devinfo.pcie_msi_cap_reg = caps->addr;
+	}
+
+	/* check if MSI-X capability is enabled */
+	caps = pci_find_cap(sc, PCIY_MSIX, BCM_PCI_CAP);
+	if (NULL != caps) {
+		PMD_DRV_LOG(DEBUG, "Found MSI-X capability at 0x%04x", reg);
+
+		sc->devinfo.pcie_cap_flags |= BCM_MSIX_CAPABLE_FLAG;
+		sc->devinfo.pcie_msix_cap_reg = caps->addr;
+	}
+}
+
+static int bcm_get_shmem_mf_cfg_info_sd(struct bcm_softc *sc)
+{
+	struct bcm_mf_info *mf_info = &sc->devinfo.mf_info;
+	uint32_t val;
+
+	/* get the outer vlan if we're in switch-dependent mode */
+
+	val = MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].e1hov_tag);
+	mf_info->ext_id = (uint16_t) val;
+
+	mf_info->multi_vnics_mode = 1;
+
+	if (!VALID_OVLAN(mf_info->ext_id)) {
+		PMD_DRV_LOG(NOTICE, "Invalid VLAN (%d)", mf_info->ext_id);
+		return 1;
+	}
+
+	/* get the capabilities */
+	if ((mf_info->mf_config[SC_VN(sc)] & FUNC_MF_CFG_PROTOCOL_MASK) ==
+	    FUNC_MF_CFG_PROTOCOL_ISCSI) {
+		mf_info->mf_protos_supported |= MF_PROTO_SUPPORT_ISCSI;
+	} else if ((mf_info->mf_config[SC_VN(sc)] & FUNC_MF_CFG_PROTOCOL_MASK)
+		   == FUNC_MF_CFG_PROTOCOL_FCOE) {
+		mf_info->mf_protos_supported |= MF_PROTO_SUPPORT_FCOE;
+	} else {
+		mf_info->mf_protos_supported |= MF_PROTO_SUPPORT_ETHERNET;
+	}
+
+	mf_info->vnics_per_port =
+	    (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) ? 2 : 4;
+
+	return 0;
+}
+
+static uint32_t bcm_get_shmem_ext_proto_support_flags(struct bcm_softc *sc)
+{
+	uint32_t retval = 0;
+	uint32_t val;
+
+	val = MFCFG_RD(sc, func_ext_config[SC_ABS_FUNC(sc)].func_cfg);
+
+	if (val & MACP_FUNC_CFG_FLAGS_ENABLED) {
+		if (val & MACP_FUNC_CFG_FLAGS_ETHERNET) {
+			retval |= MF_PROTO_SUPPORT_ETHERNET;
+		}
+		if (val & MACP_FUNC_CFG_FLAGS_ISCSI_OFFLOAD) {
+			retval |= MF_PROTO_SUPPORT_ISCSI;
+		}
+		if (val & MACP_FUNC_CFG_FLAGS_FCOE_OFFLOAD) {
+			retval |= MF_PROTO_SUPPORT_FCOE;
+		}
+	}
+
+	return retval;
+}
+
+static int bcm_get_shmem_mf_cfg_info_si(struct bcm_softc *sc)
+{
+	struct bcm_mf_info *mf_info = &sc->devinfo.mf_info;
+	uint32_t val;
+
+	/*
+	 * There is no outer vlan if we're in switch-independent mode.
+	 * If the mac is valid then assume multi-function.
+	 */
+
+	val = MFCFG_RD(sc, func_ext_config[SC_ABS_FUNC(sc)].func_cfg);
+
+	mf_info->multi_vnics_mode = ((val & MACP_FUNC_CFG_FLAGS_MASK) != 0);
+
+	mf_info->mf_protos_supported =
+	    bcm_get_shmem_ext_proto_support_flags(sc);
+
+	mf_info->vnics_per_port =
+	    (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) ? 2 : 4;
+
+	return 0;
+}
+
+static int bcm_get_shmem_mf_cfg_info_niv(struct bcm_softc *sc)
+{
+	struct bcm_mf_info *mf_info = &sc->devinfo.mf_info;
+	uint32_t e1hov_tag;
+	uint32_t func_config;
+	uint32_t niv_config;
+
+	mf_info->multi_vnics_mode = 1;
+
+	e1hov_tag = MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].e1hov_tag);
+	func_config = MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].config);
+	niv_config = MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].afex_config);
+
+	mf_info->ext_id =
+	    (uint16_t) ((e1hov_tag & FUNC_MF_CFG_E1HOV_TAG_MASK) >>
+			FUNC_MF_CFG_E1HOV_TAG_SHIFT);
+
+	mf_info->default_vlan =
+	    (uint16_t) ((e1hov_tag & FUNC_MF_CFG_AFEX_VLAN_MASK) >>
+			FUNC_MF_CFG_AFEX_VLAN_SHIFT);
+
+	mf_info->niv_allowed_priorities =
+	    (uint8_t) ((niv_config & FUNC_MF_CFG_AFEX_COS_FILTER_MASK) >>
+		       FUNC_MF_CFG_AFEX_COS_FILTER_SHIFT);
+
+	mf_info->niv_default_cos =
+	    (uint8_t) ((func_config & FUNC_MF_CFG_TRANSMIT_PRIORITY_MASK) >>
+		       FUNC_MF_CFG_TRANSMIT_PRIORITY_SHIFT);
+
+	mf_info->afex_vlan_mode =
+	    ((niv_config & FUNC_MF_CFG_AFEX_VLAN_MODE_MASK) >>
+	     FUNC_MF_CFG_AFEX_VLAN_MODE_SHIFT);
+
+	mf_info->niv_mba_enabled =
+	    ((niv_config & FUNC_MF_CFG_AFEX_MBA_ENABLED_MASK) >>
+	     FUNC_MF_CFG_AFEX_MBA_ENABLED_SHIFT);
+
+	mf_info->mf_protos_supported =
+	    bcm_get_shmem_ext_proto_support_flags(sc);
+
+	mf_info->vnics_per_port =
+	    (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) ? 2 : 4;
+
+	return 0;
+}
+
+static int bcm_check_valid_mf_cfg(struct bcm_softc *sc)
+{
+	struct bcm_mf_info *mf_info = &sc->devinfo.mf_info;
+	uint32_t mf_cfg1;
+	uint32_t mf_cfg2;
+	uint32_t ovlan1;
+	uint32_t ovlan2;
+	uint8_t i, j;
+
+	/* various MF mode sanity checks... */
+
+	if (mf_info->mf_config[SC_VN(sc)] & FUNC_MF_CFG_FUNC_HIDE) {
+		PMD_DRV_LOG(NOTICE,
+			    "Enumerated function %d is marked as hidden",
+			    SC_PORT(sc));
+		return 1;
+	}
+
+	if ((mf_info->vnics_per_port > 1) && !mf_info->multi_vnics_mode) {
+		PMD_DRV_LOG(NOTICE, "vnics_per_port=%d multi_vnics_mode=%d",
+			    mf_info->vnics_per_port, mf_info->multi_vnics_mode);
+		return 1;
+	}
+
+	if (mf_info->mf_mode == MULTI_FUNCTION_SD) {
+/* vnic id > 0 must have valid ovlan in switch-dependent mode */
+		if ((SC_VN(sc) > 0) && !VALID_OVLAN(OVLAN(sc))) {
+			PMD_DRV_LOG(NOTICE, "mf_mode=SD vnic_id=%d ovlan=%d",
+				    SC_VN(sc), OVLAN(sc));
+			return 1;
+		}
+
+		if (!VALID_OVLAN(OVLAN(sc)) && mf_info->multi_vnics_mode) {
+			PMD_DRV_LOG(NOTICE,
+				    "mf_mode=SD multi_vnics_mode=%d ovlan=%d",
+				    mf_info->multi_vnics_mode, OVLAN(sc));
+			return 1;
+		}
+
+/*
+ * Verify all functions are either MF or SF mode. If MF, make sure
+ * sure that all non-hidden functions have a valid ovlan. If SF,
+ * make sure that all non-hidden functions have an invalid ovlan.
+ */
+		FOREACH_ABS_FUNC_IN_PORT(sc, i) {
+			mf_cfg1 = MFCFG_RD(sc, func_mf_config[i].config);
+			ovlan1 = MFCFG_RD(sc, func_mf_config[i].e1hov_tag);
+			if (!(mf_cfg1 & FUNC_MF_CFG_FUNC_HIDE) &&
+			    (((mf_info->multi_vnics_mode)
+			      && !VALID_OVLAN(ovlan1))
+			     || ((!mf_info->multi_vnics_mode)
+				 && VALID_OVLAN(ovlan1)))) {
+				PMD_DRV_LOG(NOTICE,
+					    "mf_mode=SD function %d MF config "
+					    "mismatch, multi_vnics_mode=%d ovlan=%d",
+					    i, mf_info->multi_vnics_mode,
+					    ovlan1);
+				return 1;
+			}
+		}
+
+/* Verify all funcs on the same port each have a different ovlan. */
+		FOREACH_ABS_FUNC_IN_PORT(sc, i) {
+			mf_cfg1 = MFCFG_RD(sc, func_mf_config[i].config);
+			ovlan1 = MFCFG_RD(sc, func_mf_config[i].e1hov_tag);
+			/* iterate from the next function on the port to the max func */
+			for (j = i + 2; j < MAX_FUNC_NUM; j += 2) {
+				mf_cfg2 =
+				    MFCFG_RD(sc, func_mf_config[j].config);
+				ovlan2 =
+				    MFCFG_RD(sc, func_mf_config[j].e1hov_tag);
+				if (!(mf_cfg1 & FUNC_MF_CFG_FUNC_HIDE)
+				    && VALID_OVLAN(ovlan1)
+				    && !(mf_cfg2 & FUNC_MF_CFG_FUNC_HIDE)
+				    && VALID_OVLAN(ovlan2)
+				    && (ovlan1 == ovlan2)) {
+					PMD_DRV_LOG(NOTICE,
+						    "mf_mode=SD functions %d and %d "
+						    "have the same ovlan (%d)",
+						    i, j, ovlan1);
+					return 1;
+				}
+			}
+		}
+	}
+	/* MULTI_FUNCTION_SD */
+	return 0;
+}
+
+static int bcm_get_mf_cfg_info(struct bcm_softc *sc)
+{
+	struct bcm_mf_info *mf_info = &sc->devinfo.mf_info;
+	uint32_t val, mac_upper;
+	uint8_t i, vnic;
+
+	/* initialize mf_info defaults */
+	mf_info->vnics_per_port = 1;
+	mf_info->multi_vnics_mode = FALSE;
+	mf_info->path_has_ovlan = FALSE;
+	mf_info->mf_mode = SINGLE_FUNCTION;
+
+	if (!CHIP_IS_MF_CAP(sc)) {
+		return 0;
+	}
+
+	if (sc->devinfo.mf_cfg_base == SHMEM_MF_CFG_ADDR_NONE) {
+		PMD_DRV_LOG(NOTICE, "Invalid mf_cfg_base!");
+		return 1;
+	}
+
+	/* get the MF mode (switch dependent / independent / single-function) */
+
+	val = SHMEM_RD(sc, dev_info.shared_feature_config.config);
+
+	switch (val & SHARED_FEAT_CFG_FORCE_SF_MODE_MASK) {
+	case SHARED_FEAT_CFG_FORCE_SF_MODE_SWITCH_INDEPT:
+
+		mac_upper =
+		    MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].mac_upper);
+
+		/* check for legal upper mac bytes */
+		if (mac_upper != FUNC_MF_CFG_UPPERMAC_DEFAULT) {
+			mf_info->mf_mode = MULTI_FUNCTION_SI;
+		} else {
+			PMD_DRV_LOG(NOTICE,
+				    "Invalid config for Switch Independent mode");
+		}
+
+		break;
+
+	case SHARED_FEAT_CFG_FORCE_SF_MODE_MF_ALLOWED:
+	case SHARED_FEAT_CFG_FORCE_SF_MODE_SPIO4:
+
+		/* get outer vlan configuration */
+		val = MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].e1hov_tag);
+
+		if ((val & FUNC_MF_CFG_E1HOV_TAG_MASK) !=
+		    FUNC_MF_CFG_E1HOV_TAG_DEFAULT) {
+			mf_info->mf_mode = MULTI_FUNCTION_SD;
+		} else {
+			PMD_DRV_LOG(NOTICE,
+				    "Invalid config for Switch Dependent mode");
+		}
+
+		break;
+
+	case SHARED_FEAT_CFG_FORCE_SF_MODE_FORCED_SF:
+
+		/* not in MF mode, vnics_per_port=1 and multi_vnics_mode=FALSE */
+		return 0;
+
+	case SHARED_FEAT_CFG_FORCE_SF_MODE_AFEX_MODE:
+
+		/*
+		 * Mark MF mode as NIV if MCP version includes NPAR-SD support
+		 * and the MAC address is valid.
+		 */
+		mac_upper =
+		    MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].mac_upper);
+
+		if ((SHMEM2_HAS(sc, afex_driver_support)) &&
+		    (mac_upper != FUNC_MF_CFG_UPPERMAC_DEFAULT)) {
+			mf_info->mf_mode = MULTI_FUNCTION_AFEX;
+		} else {
+			PMD_DRV_LOG(NOTICE, "Invalid config for AFEX mode");
+		}
+
+		break;
+
+	default:
+
+		PMD_DRV_LOG(NOTICE, "Unknown MF mode (0x%08x)",
+			    (val & SHARED_FEAT_CFG_FORCE_SF_MODE_MASK));
+
+		return 1;
+	}
+
+	/* set path mf_mode (which could be different than function mf_mode) */
+	if (mf_info->mf_mode == MULTI_FUNCTION_SD) {
+		mf_info->path_has_ovlan = TRUE;
+	} else if (mf_info->mf_mode == SINGLE_FUNCTION) {
+/*
+ * Decide on path multi vnics mode. If we're not in MF mode and in
+ * 4-port mode, this is good enough to check vnic-0 of the other port
+ * on the same path
+ */
+		if (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) {
+			uint8_t other_port = !(PORT_ID(sc) & 1);
+			uint8_t abs_func_other_port =
+			    (SC_PATH(sc) + (2 * other_port));
+
+			val =
+			    MFCFG_RD(sc,
+				     func_mf_config
+				     [abs_func_other_port].e1hov_tag);
+
+			mf_info->path_has_ovlan = VALID_OVLAN((uint16_t) val);
+		}
+	}
+
+	if (mf_info->mf_mode == SINGLE_FUNCTION) {
+/* invalid MF config */
+		if (SC_VN(sc) >= 1) {
+			PMD_DRV_LOG(NOTICE, "VNIC ID >= 1 in SF mode");
+			return 1;
+		}
+
+		return 0;
+	}
+
+	/* get the MF configuration */
+	mf_info->mf_config[SC_VN(sc)] =
+	    MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].config);
+
+	switch (mf_info->mf_mode) {
+	case MULTI_FUNCTION_SD:
+
+		bcm_get_shmem_mf_cfg_info_sd(sc);
+		break;
+
+	case MULTI_FUNCTION_SI:
+
+		bcm_get_shmem_mf_cfg_info_si(sc);
+		break;
+
+	case MULTI_FUNCTION_AFEX:
+
+		bcm_get_shmem_mf_cfg_info_niv(sc);
+		break;
+
+	default:
+
+		PMD_DRV_LOG(NOTICE, "Get MF config failed (mf_mode=0x%08x)",
+			    mf_info->mf_mode);
+		return 1;
+	}
+
+	/* get the congestion management parameters */
+
+	vnic = 0;
+	FOREACH_ABS_FUNC_IN_PORT(sc, i) {
+/* get min/max bw */
+		val = MFCFG_RD(sc, func_mf_config[i].config);
+		mf_info->min_bw[vnic] =
+		    ((val & FUNC_MF_CFG_MIN_BW_MASK) >>
+		     FUNC_MF_CFG_MIN_BW_SHIFT);
+		mf_info->max_bw[vnic] =
+		    ((val & FUNC_MF_CFG_MAX_BW_MASK) >>
+		     FUNC_MF_CFG_MAX_BW_SHIFT);
+		vnic++;
+	}
+
+	return bcm_check_valid_mf_cfg(sc);
+}
+
+static int bcm_get_shmem_info(struct bcm_softc *sc)
+{
+	int port;
+	uint32_t mac_hi, mac_lo, val;
+
+	PMD_INIT_FUNC_TRACE();
+
+	port = SC_PORT(sc);
+	mac_hi = mac_lo = 0;
+
+	sc->link_params.sc = sc;
+	sc->link_params.port = port;
+
+	/* get the hardware config info */
+	sc->devinfo.hw_config = SHMEM_RD(sc, dev_info.shared_hw_config.config);
+	sc->devinfo.hw_config2 =
+	    SHMEM_RD(sc, dev_info.shared_hw_config.config2);
+
+	sc->link_params.hw_led_mode =
+	    ((sc->devinfo.hw_config & SHARED_HW_CFG_LED_MODE_MASK) >>
+	     SHARED_HW_CFG_LED_MODE_SHIFT);
+
+	/* get the port feature config */
+	sc->port.config =
+	    SHMEM_RD(sc, dev_info.port_feature_config[port].config);
+
+	/* get the link params */
+	sc->link_params.speed_cap_mask[ELINK_INT_PHY] =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].speed_capability_mask)
+	    & PORT_HW_CFG_SPEED_CAPABILITY_D0_MASK;
+	sc->link_params.speed_cap_mask[ELINK_EXT_PHY1] =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].speed_capability_mask2)
+	    & PORT_HW_CFG_SPEED_CAPABILITY_D0_MASK;
+
+	/* get the lane config */
+	sc->link_params.lane_config =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].lane_config);
+
+	/* get the link config */
+	val = SHMEM_RD(sc, dev_info.port_feature_config[port].link_config);
+	sc->port.link_config[ELINK_INT_PHY] = val;
+	sc->link_params.switch_cfg = (val & PORT_FEATURE_CONNECTED_SWITCH_MASK);
+	sc->port.link_config[ELINK_EXT_PHY1] =
+	    SHMEM_RD(sc, dev_info.port_feature_config[port].link_config2);
+
+	/* get the override preemphasis flag and enable it or turn it off */
+	val = SHMEM_RD(sc, dev_info.shared_feature_config.config);
+	if (val & SHARED_FEAT_CFG_OVERRIDE_PREEMPHASIS_CFG_ENABLED) {
+		sc->link_params.feature_config_flags |=
+		    ELINK_FEATURE_CONFIG_OVERRIDE_PREEMPHASIS_ENABLED;
+	} else {
+		sc->link_params.feature_config_flags &=
+		    ~ELINK_FEATURE_CONFIG_OVERRIDE_PREEMPHASIS_ENABLED;
+	}
+
+	/* get the initial value of the link params */
+	sc->link_params.multi_phy_config =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].multi_phy_config);
+
+	/* get external phy info */
+	sc->port.ext_phy_config =
+	    SHMEM_RD(sc, dev_info.port_hw_config[port].external_phy_config);
+
+	/* get the multifunction configuration */
+	bcm_get_mf_cfg_info(sc);
+
+	/* get the mac address */
+	if (IS_MF(sc)) {
+		mac_hi =
+		    MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].mac_upper);
+		mac_lo =
+		    MFCFG_RD(sc, func_mf_config[SC_ABS_FUNC(sc)].mac_lower);
+	} else {
+		mac_hi = SHMEM_RD(sc, dev_info.port_hw_config[port].mac_upper);
+		mac_lo = SHMEM_RD(sc, dev_info.port_hw_config[port].mac_lower);
+	}
+
+	if ((mac_lo == 0) && (mac_hi == 0)) {
+		*sc->mac_addr_str = 0;
+		PMD_DRV_LOG(NOTICE, "No Ethernet address programmed!");
+	} else {
+		sc->link_params.mac_addr[0] = (uint8_t) (mac_hi >> 8);
+		sc->link_params.mac_addr[1] = (uint8_t) (mac_hi);
+		sc->link_params.mac_addr[2] = (uint8_t) (mac_lo >> 24);
+		sc->link_params.mac_addr[3] = (uint8_t) (mac_lo >> 16);
+		sc->link_params.mac_addr[4] = (uint8_t) (mac_lo >> 8);
+		sc->link_params.mac_addr[5] = (uint8_t) (mac_lo);
+		snprintf(sc->mac_addr_str, sizeof(sc->mac_addr_str),
+			 "%02x:%02x:%02x:%02x:%02x:%02x",
+			 sc->link_params.mac_addr[0],
+			 sc->link_params.mac_addr[1],
+			 sc->link_params.mac_addr[2],
+			 sc->link_params.mac_addr[3],
+			 sc->link_params.mac_addr[4],
+			 sc->link_params.mac_addr[5]);
+		PMD_DRV_LOG(DEBUG, "Ethernet address: %s", sc->mac_addr_str);
+	}
+
+	return 0;
+}
+
+static void bcm_media_detect(struct bcm_softc *sc)
+{
+	uint32_t phy_idx = bcm_get_cur_phy_idx(sc);
+	switch (sc->link_params.phy[phy_idx].media_type) {
+	case ELINK_ETH_PHY_SFPP_10G_FIBER:
+	case ELINK_ETH_PHY_SFP_1G_FIBER:
+	case ELINK_ETH_PHY_XFP_FIBER:
+	case ELINK_ETH_PHY_KR:
+	case ELINK_ETH_PHY_CX4:
+		PMD_DRV_LOG(INFO, "Found 10GBase-CX4 media.");
+		sc->media = IFM_10G_CX4;
+		break;
+	case ELINK_ETH_PHY_DA_TWINAX:
+		PMD_DRV_LOG(INFO, "Found 10Gb Twinax media.");
+		sc->media = IFM_10G_TWINAX;
+		break;
+	case ELINK_ETH_PHY_BASE_T:
+		PMD_DRV_LOG(INFO, "Found 10GBase-T media.");
+		sc->media = IFM_10G_T;
+		break;
+	case ELINK_ETH_PHY_NOT_PRESENT:
+		PMD_DRV_LOG(INFO, "Media not present.");
+		sc->media = 0;
+		break;
+	case ELINK_ETH_PHY_UNSPECIFIED:
+	default:
+		PMD_DRV_LOG(INFO, "Unknown media!");
+		sc->media = 0;
+		break;
+	}
+}
+
+#define GET_FIELD(value, fname)                     \
+(((value) & (fname##_MASK)) >> (fname##_SHIFT))
+#define IGU_FID(val) GET_FIELD((val), IGU_REG_MAPPING_MEMORY_FID)
+#define IGU_VEC(val) GET_FIELD((val), IGU_REG_MAPPING_MEMORY_VECTOR)
+
+static int bcm_get_igu_cam_info(struct bcm_softc *sc)
+{
+	int pfid = SC_FUNC(sc);
+	int igu_sb_id;
+	uint32_t val;
+	uint8_t fid, igu_sb_cnt = 0;
+
+	sc->igu_base_sb = 0xff;
+
+	if (CHIP_INT_MODE_IS_BC(sc)) {
+		int vn = SC_VN(sc);
+		igu_sb_cnt = sc->igu_sb_cnt;
+		sc->igu_base_sb = ((CHIP_IS_MODE_4_PORT(sc) ? pfid : vn) *
+				   FP_SB_MAX_E1x);
+		sc->igu_dsb_id = (E1HVN_MAX * FP_SB_MAX_E1x +
+				  (CHIP_IS_MODE_4_PORT(sc) ? pfid : vn));
+		return 0;
+	}
+
+	/* IGU in normal mode - read CAM */
+	for (igu_sb_id = 0;
+	     igu_sb_id < IGU_REG_MAPPING_MEMORY_SIZE; igu_sb_id++) {
+		val = REG_RD(sc, IGU_REG_MAPPING_MEMORY + igu_sb_id * 4);
+		if (!(val & IGU_REG_MAPPING_MEMORY_VALID)) {
+			continue;
+		}
+		fid = IGU_FID(val);
+		if ((fid & IGU_FID_ENCODE_IS_PF)) {
+			if ((fid & IGU_FID_PF_NUM_MASK) != pfid) {
+				continue;
+			}
+			if (IGU_VEC(val) == 0) {
+				/* default status block */
+				sc->igu_dsb_id = igu_sb_id;
+			} else {
+				if (sc->igu_base_sb == 0xff) {
+					sc->igu_base_sb = igu_sb_id;
+				}
+				igu_sb_cnt++;
+			}
+		}
+	}
+
+	/*
+	 * Due to new PF resource allocation by MFW T7.4 and above, it's optional
+	 * that number of CAM entries will not be equal to the value advertised in
+	 * PCI. Driver should use the minimal value of both as the actual status
+	 * block count
+	 */
+	sc->igu_sb_cnt = min(sc->igu_sb_cnt, igu_sb_cnt);
+
+	if (igu_sb_cnt == 0) {
+		PMD_DRV_LOG(ERR, "CAM configuration error");
+		return -1;
+	}
+
+	return 0;
+}
+
+/*
+* Gather various information from the device config space, the device itself,
+* shmem, and the user input.
+*/
+static int bcm_get_device_info(struct bcm_softc *sc)
+{
+	uint32_t val;
+	int rc;
+
+	/* get the chip revision (chip metal comes from pci config space) */
+	sc->devinfo.chip_id = sc->link_params.chip_id =
+	    (((REG_RD(sc, MISC_REG_CHIP_NUM) & 0xffff) << 16) |
+	     ((REG_RD(sc, MISC_REG_CHIP_REV) & 0xf) << 12) |
+	     (((REG_RD(sc, PCICFG_OFFSET + PCI_ID_VAL3) >> 24) & 0xf) << 4) |
+	     ((REG_RD(sc, MISC_REG_BOND_ID) & 0xf) << 0));
+
+	/* force 57811 according to MISC register */
+	if (REG_RD(sc, MISC_REG_CHIP_TYPE) & MISC_REG_CHIP_TYPE_57811_MASK) {
+		if (CHIP_IS_57810(sc)) {
+			sc->devinfo.chip_id = ((CHIP_NUM_57811 << 16) |
+					       (sc->
+						devinfo.chip_id & 0x0000ffff));
+		} else if (CHIP_IS_57810_MF(sc)) {
+			sc->devinfo.chip_id = ((CHIP_NUM_57811_MF << 16) |
+					       (sc->
+						devinfo.chip_id & 0x0000ffff));
+		}
+		sc->devinfo.chip_id |= 0x1;
+	}
+
+	PMD_DRV_LOG(DEBUG,
+		    "chip_id=0x%08x (num=0x%04x rev=0x%01x metal=0x%02x bond=0x%01x)",
+		    sc->devinfo.chip_id,
+		    ((sc->devinfo.chip_id >> 16) & 0xffff),
+		    ((sc->devinfo.chip_id >> 12) & 0xf),
+		    ((sc->devinfo.chip_id >> 4) & 0xff),
+		    ((sc->devinfo.chip_id >> 0) & 0xf));
+
+	val = (REG_RD(sc, 0x2874) & 0x55);
+	if ((sc->devinfo.chip_id & 0x1) || (CHIP_IS_E1H(sc) && (val == 0x55))) {
+		sc->flags |= BCM_ONE_PORT_FLAG;
+		PMD_DRV_LOG(DEBUG, "single port device");
+	}
+
+	/* set the doorbell size */
+	sc->doorbell_size = (1 << BCM_DB_SHIFT);
+
+	/* determine whether the device is in 2 port or 4 port mode */
+	sc->devinfo.chip_port_mode = CHIP_PORT_MODE_NONE;	/* E1h */
+	if (CHIP_IS_E2E3(sc)) {
+/*
+ * Read port4mode_en_ovwr[0]:
+ *   If 1, four port mode is in port4mode_en_ovwr[1].
+ *   If 0, four port mode is in port4mode_en[0].
+ */
+		val = REG_RD(sc, MISC_REG_PORT4MODE_EN_OVWR);
+		if (val & 1) {
+			val = ((val >> 1) & 1);
+		} else {
+			val = REG_RD(sc, MISC_REG_PORT4MODE_EN);
+		}
+
+		sc->devinfo.chip_port_mode =
+		    (val) ? CHIP_4_PORT_MODE : CHIP_2_PORT_MODE;
+
+		PMD_DRV_LOG(DEBUG, "Port mode = %s", (val) ? "4" : "2");
+	}
+
+	/* get the function and path info for the device */
+	bcm_get_function_num(sc);
+
+	/* get the shared memory base address */
+	sc->devinfo.shmem_base =
+	    sc->link_params.shmem_base = REG_RD(sc, MISC_REG_SHARED_MEM_ADDR);
+	sc->devinfo.shmem2_base =
+	    REG_RD(sc, (SC_PATH(sc) ? MISC_REG_GENERIC_CR_1 :
+			MISC_REG_GENERIC_CR_0));
+
+	if (!sc->devinfo.shmem_base) {
+/* this should ONLY prevent upcoming shmem reads */
+		PMD_DRV_LOG(INFO, "MCP not active");
+		sc->flags |= BCM_NO_MCP_FLAG;
+		return 0;
+	}
+
+	/* make sure the shared memory contents are valid */
+	val = SHMEM_RD(sc, validity_map[SC_PORT(sc)]);
+	if ((val & (SHR_MEM_VALIDITY_DEV_INFO | SHR_MEM_VALIDITY_MB)) !=
+	    (SHR_MEM_VALIDITY_DEV_INFO | SHR_MEM_VALIDITY_MB)) {
+		PMD_DRV_LOG(NOTICE, "Invalid SHMEM validity signature: 0x%08x",
+			    val);
+		return 0;
+	}
+
+	/* get the bootcode version */
+	sc->devinfo.bc_ver = SHMEM_RD(sc, dev_info.bc_rev);
+	snprintf(sc->devinfo.bc_ver_str,
+		 sizeof(sc->devinfo.bc_ver_str),
+		 "%d.%d.%d",
+		 ((sc->devinfo.bc_ver >> 24) & 0xff),
+		 ((sc->devinfo.bc_ver >> 16) & 0xff),
+		 ((sc->devinfo.bc_ver >> 8) & 0xff));
+	PMD_DRV_LOG(INFO, "Bootcode version: %s", sc->devinfo.bc_ver_str);
+
+	/* get the bootcode shmem address */
+	sc->devinfo.mf_cfg_base = bcm_get_shmem_mf_cfg_base(sc);
+
+	/* clean indirect addresses as they're not used */
+	pci_write_long(sc, PCICFG_GRC_ADDRESS, 0);
+	if (IS_PF(sc)) {
+		REG_WR(sc, PXP2_REG_PGL_ADDR_88_F0, 0);
+		REG_WR(sc, PXP2_REG_PGL_ADDR_8C_F0, 0);
+		REG_WR(sc, PXP2_REG_PGL_ADDR_90_F0, 0);
+		REG_WR(sc, PXP2_REG_PGL_ADDR_94_F0, 0);
+		if (CHIP_IS_E1x(sc)) {
+			REG_WR(sc, PXP2_REG_PGL_ADDR_88_F1, 0);
+			REG_WR(sc, PXP2_REG_PGL_ADDR_8C_F1, 0);
+			REG_WR(sc, PXP2_REG_PGL_ADDR_90_F1, 0);
+			REG_WR(sc, PXP2_REG_PGL_ADDR_94_F1, 0);
+		}
+
+/*
+ * Enable internal target-read (in case we are probed after PF
+ * FLR). Must be done prior to any BAR read access. Only for
+ * 57712 and up
+ */
+		if (!CHIP_IS_E1x(sc)) {
+			REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_TARGET_READ,
+			       1);
+		}
+	}
+
+	/* get the nvram size */
+	val = REG_RD(sc, MCP_REG_MCPR_NVM_CFG4);
+	sc->devinfo.flash_size =
+	    (NVRAM_1MB_SIZE << (val & MCPR_NVM_CFG4_FLASH_SIZE));
+
+	bcm_set_power_state(sc, PCI_PM_D0);
+	/* get various configuration parameters from shmem */
+	bcm_get_shmem_info(sc);
+
+	/* initialize IGU parameters */
+	if (CHIP_IS_E1x(sc)) {
+		sc->devinfo.int_block = INT_BLOCK_HC;
+		sc->igu_dsb_id = DEF_SB_IGU_ID;
+		sc->igu_base_sb = 0;
+	} else {
+		sc->devinfo.int_block = INT_BLOCK_IGU;
+
+/* do not allow device reset during IGU info preocessing */
+		bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RESET);
+
+		val = REG_RD(sc, IGU_REG_BLOCK_CONFIGURATION);
+
+		if (val & IGU_BLOCK_CONFIGURATION_REG_BACKWARD_COMP_EN) {
+			int tout = 5000;
+
+			val &= ~(IGU_BLOCK_CONFIGURATION_REG_BACKWARD_COMP_EN);
+			REG_WR(sc, IGU_REG_BLOCK_CONFIGURATION, val);
+			REG_WR(sc, IGU_REG_RESET_MEMORIES, 0x7f);
+
+			while (tout && REG_RD(sc, IGU_REG_RESET_MEMORIES)) {
+				tout--;
+				DELAY(1000);
+			}
+
+			if (REG_RD(sc, IGU_REG_RESET_MEMORIES)) {
+				PMD_DRV_LOG(NOTICE,
+					    "FORCING IGU Normal Mode failed!!!");
+				bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RESET);
+				return -1;
+			}
+		}
+
+		if (val & IGU_BLOCK_CONFIGURATION_REG_BACKWARD_COMP_EN) {
+			PMD_DRV_LOG(DEBUG, "IGU Backward Compatible Mode");
+			sc->devinfo.int_block |= INT_BLOCK_MODE_BW_COMP;
+		} else {
+			PMD_DRV_LOG(DEBUG, "IGU Normal Mode");
+		}
+
+		rc = bcm_get_igu_cam_info(sc);
+
+		bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RESET);
+
+		if (rc) {
+			return rc;
+		}
+	}
+
+	/*
+	 * Get base FW non-default (fast path) status block ID. This value is
+	 * used to initialize the fw_sb_id saved on the fp/queue structure to
+	 * determine the id used by the FW.
+	 */
+	if (CHIP_IS_E1x(sc)) {
+		sc->base_fw_ndsb =
+		    ((SC_PORT(sc) * FP_SB_MAX_E1x) + SC_L_ID(sc));
+	} else {
+/*
+ * 57712+ - We currently use one FW SB per IGU SB (Rx and Tx of
+ * the same queue are indicated on the same IGU SB). So we prefer
+ * FW and IGU SBs to be the same value.
+ */
+		sc->base_fw_ndsb = sc->igu_base_sb;
+	}
+
+	elink_phy_probe(&sc->link_params);
+
+	return 0;
+}
+
+static void
+bcm_link_settings_supported(struct bcm_softc *sc, uint32_t switch_cfg)
+{
+	uint32_t cfg_size = 0;
+	uint32_t idx;
+	uint8_t port = SC_PORT(sc);
+
+	/* aggregation of supported attributes of all external phys */
+	sc->port.supported[0] = 0;
+	sc->port.supported[1] = 0;
+
+	switch (sc->link_params.num_phys) {
+	case 1:
+		sc->port.supported[0] =
+		    sc->link_params.phy[ELINK_INT_PHY].supported;
+		cfg_size = 1;
+		break;
+	case 2:
+		sc->port.supported[0] =
+		    sc->link_params.phy[ELINK_EXT_PHY1].supported;
+		cfg_size = 1;
+		break;
+	case 3:
+		if (sc->link_params.multi_phy_config &
+		    PORT_HW_CFG_PHY_SWAPPED_ENABLED) {
+			sc->port.supported[1] =
+			    sc->link_params.phy[ELINK_EXT_PHY1].supported;
+			sc->port.supported[0] =
+			    sc->link_params.phy[ELINK_EXT_PHY2].supported;
+		} else {
+			sc->port.supported[0] =
+			    sc->link_params.phy[ELINK_EXT_PHY1].supported;
+			sc->port.supported[1] =
+			    sc->link_params.phy[ELINK_EXT_PHY2].supported;
+		}
+		cfg_size = 2;
+		break;
+	}
+
+	if (!(sc->port.supported[0] || sc->port.supported[1])) {
+		PMD_DRV_LOG(ERR,
+			    "Invalid phy config in NVRAM (PHY1=0x%08x PHY2=0x%08x)",
+			    SHMEM_RD(sc,
+				     dev_info.port_hw_config
+				     [port].external_phy_config),
+			    SHMEM_RD(sc,
+				     dev_info.port_hw_config
+				     [port].external_phy_config2));
+		return;
+	}
+
+	if (CHIP_IS_E3(sc))
+		sc->port.phy_addr = REG_RD(sc, MISC_REG_WC0_CTRL_PHY_ADDR);
+	else {
+		switch (switch_cfg) {
+		case ELINK_SWITCH_CFG_1G:
+			sc->port.phy_addr =
+			    REG_RD(sc,
+				   NIG_REG_SERDES0_CTRL_PHY_ADDR + port * 0x10);
+			break;
+		case ELINK_SWITCH_CFG_10G:
+			sc->port.phy_addr =
+			    REG_RD(sc,
+				   NIG_REG_XGXS0_CTRL_PHY_ADDR + port * 0x18);
+			break;
+		default:
+			PMD_DRV_LOG(ERR,
+				    "Invalid switch config in"
+				    "link_config=0x%08x",
+				    sc->port.link_config[0]);
+			return;
+		}
+	}
+
+	PMD_DRV_LOG(INFO, "PHY addr 0x%08x", sc->port.phy_addr);
+
+	/* mask what we support according to speed_cap_mask per configuration */
+	for (idx = 0; idx < cfg_size; idx++) {
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_10M_HALF)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_10baseT_Half;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_10M_FULL)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_10baseT_Full;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_100M_HALF)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_100baseT_Half;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_100M_FULL)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_100baseT_Full;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_1G)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_1000baseT_Full;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_2_5G)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_2500baseX_Full;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_10G)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_10000baseT_Full;
+		}
+
+		if (!(sc->link_params.speed_cap_mask[idx] &
+		      PORT_HW_CFG_SPEED_CAPABILITY_D0_20G)) {
+			sc->port.supported[idx] &=
+			    ~ELINK_SUPPORTED_20000baseKR2_Full;
+		}
+	}
+
+	PMD_DRV_LOG(INFO, "PHY supported 0=0x%08x 1=0x%08x",
+		    sc->port.supported[0], sc->port.supported[1]);
+}
+
+static void bcm_link_settings_requested(struct bcm_softc *sc)
+{
+	uint32_t link_config;
+	uint32_t idx;
+	uint32_t cfg_size = 0;
+
+	sc->port.advertising[0] = 0;
+	sc->port.advertising[1] = 0;
+
+	switch (sc->link_params.num_phys) {
+	case 1:
+	case 2:
+		cfg_size = 1;
+		break;
+	case 3:
+		cfg_size = 2;
+		break;
+	}
+
+	for (idx = 0; idx < cfg_size; idx++) {
+		sc->link_params.req_duplex[idx] = DUPLEX_FULL;
+		link_config = sc->port.link_config[idx];
+
+		switch (link_config & PORT_FEATURE_LINK_SPEED_MASK) {
+		case PORT_FEATURE_LINK_SPEED_AUTO:
+			if (sc->port.supported[idx] & ELINK_SUPPORTED_Autoneg) {
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_AUTO_NEG;
+				sc->port.advertising[idx] |=
+				    sc->port.supported[idx];
+				if (sc->link_params.phy[ELINK_EXT_PHY1].type ==
+				    PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM84833)
+					sc->port.advertising[idx] |=
+					    (ELINK_SUPPORTED_100baseT_Half |
+					     ELINK_SUPPORTED_100baseT_Full);
+			} else {
+				/* force 10G, no AN */
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_10000;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_10000baseT_Full |
+				     ADVERTISED_FIBRE);
+				continue;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_10M_FULL:
+			if (sc->
+			    port.supported[idx] & ELINK_SUPPORTED_10baseT_Full)
+			{
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_10;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_10baseT_Full | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_10M_HALF:
+			if (sc->
+			    port.supported[idx] & ELINK_SUPPORTED_10baseT_Half)
+			{
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_10;
+				sc->link_params.req_duplex[idx] = DUPLEX_HALF;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_10baseT_Half | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_100M_FULL:
+			if (sc->
+			    port.supported[idx] & ELINK_SUPPORTED_100baseT_Full)
+			{
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_100;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_100baseT_Full | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_100M_HALF:
+			if (sc->
+			    port.supported[idx] & ELINK_SUPPORTED_100baseT_Half)
+			{
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_100;
+				sc->link_params.req_duplex[idx] = DUPLEX_HALF;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_100baseT_Half | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_1G:
+			if (sc->port.supported[idx] &
+			    ELINK_SUPPORTED_1000baseT_Full) {
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_1000;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_1000baseT_Full | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_2_5G:
+			if (sc->port.supported[idx] &
+			    ELINK_SUPPORTED_2500baseX_Full) {
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_2500;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_2500baseX_Full | ADVERTISED_TP);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_10G_CX4:
+			if (sc->port.supported[idx] &
+			    ELINK_SUPPORTED_10000baseT_Full) {
+				sc->link_params.req_line_speed[idx] =
+				    ELINK_SPEED_10000;
+				sc->port.advertising[idx] |=
+				    (ADVERTISED_10000baseT_Full |
+				     ADVERTISED_FIBRE);
+			} else {
+				PMD_DRV_LOG(ERR,
+					    "Invalid NVRAM config link_config=0x%08x "
+					    "speed_cap_mask=0x%08x",
+					    link_config,
+					    sc->
+					    link_params.speed_cap_mask[idx]);
+				return;
+			}
+			break;
+
+		case PORT_FEATURE_LINK_SPEED_20G:
+			sc->link_params.req_line_speed[idx] = ELINK_SPEED_20000;
+			break;
+
+		default:
+			PMD_DRV_LOG(ERR,
+				    "Invalid NVRAM config link_config=0x%08x "
+				    "speed_cap_mask=0x%08x", link_config,
+				    sc->link_params.speed_cap_mask[idx]);
+			sc->link_params.req_line_speed[idx] =
+			    ELINK_SPEED_AUTO_NEG;
+			sc->port.advertising[idx] = sc->port.supported[idx];
+			break;
+		}
+
+		sc->link_params.req_flow_ctrl[idx] =
+		    (link_config & PORT_FEATURE_FLOW_CONTROL_MASK);
+
+		if (sc->link_params.req_flow_ctrl[idx] == ELINK_FLOW_CTRL_AUTO) {
+			if (!
+			    (sc->
+			     port.supported[idx] & ELINK_SUPPORTED_Autoneg)) {
+				sc->link_params.req_flow_ctrl[idx] =
+				    ELINK_FLOW_CTRL_NONE;
+			} else {
+				bcm_set_requested_fc(sc);
+			}
+		}
+	}
+}
+
+static void bcm_get_phy_info(struct bcm_softc *sc)
+{
+	uint8_t port = SC_PORT(sc);
+	uint32_t eee_mode;
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* shmem data already read in bcm_get_shmem_info() */
+
+	bcm_link_settings_supported(sc, sc->link_params.switch_cfg);
+	bcm_link_settings_requested(sc);
+
+	/* configure link feature according to nvram value */
+	eee_mode =
+	    (((SHMEM_RD(sc, dev_info.port_feature_config[port].eee_power_mode))
+	      & PORT_FEAT_CFG_EEE_POWER_MODE_MASK) >>
+	     PORT_FEAT_CFG_EEE_POWER_MODE_SHIFT);
+	if (eee_mode != PORT_FEAT_CFG_EEE_POWER_MODE_DISABLED) {
+		sc->link_params.eee_mode = (ELINK_EEE_MODE_ADV_LPI |
+					    ELINK_EEE_MODE_ENABLE_LPI |
+					    ELINK_EEE_MODE_OUTPUT_TIME);
+	} else {
+		sc->link_params.eee_mode = 0;
+	}
+
+	/* get the media type */
+	bcm_media_detect(sc);
+}
+
+static void bcm_set_modes_bitmap(struct bcm_softc *sc)
+{
+	uint32_t flags = MODE_ASIC | MODE_PORT2;
+
+	if (CHIP_IS_E2(sc)) {
+		flags |= MODE_E2;
+	} else if (CHIP_IS_E3(sc)) {
+		flags |= MODE_E3;
+		if (CHIP_REV(sc) == CHIP_REV_Ax) {
+			flags |= MODE_E3_A0;
+		} else {	/*if (CHIP_REV(sc) == CHIP_REV_Bx) */
+
+			flags |= MODE_E3_B0 | MODE_COS3;
+		}
+	}
+
+	if (IS_MF(sc)) {
+		flags |= MODE_MF;
+		switch (sc->devinfo.mf_info.mf_mode) {
+		case MULTI_FUNCTION_SD:
+			flags |= MODE_MF_SD;
+			break;
+		case MULTI_FUNCTION_SI:
+			flags |= MODE_MF_SI;
+			break;
+		case MULTI_FUNCTION_AFEX:
+			flags |= MODE_MF_AFEX;
+			break;
+		}
+	} else {
+		flags |= MODE_SF;
+	}
+
+#if defined(__LITTLE_ENDIAN)
+	flags |= MODE_LITTLE_ENDIAN;
+#else /* __BIG_ENDIAN */
+	flags |= MODE_BIG_ENDIAN;
+#endif
+
+	INIT_MODE_FLAGS(sc) = flags;
+}
+
+int bcm_alloc_hsi_mem(struct bcm_softc *sc)
+{
+	struct bcm_fastpath *fp;
+	char buf[32];
+	uint32_t i;
+
+	if (IS_PF(sc)) {
+/************************/
+/* DEFAULT STATUS BLOCK */
+/************************/
+
+		if (bcm_dma_alloc(sc, sizeof(struct host_sp_status_block),
+				  &sc->def_sb_dma, "def_sb",
+				  RTE_CACHE_LINE_SIZE) != 0) {
+			return -1;
+		}
+
+		sc->def_sb =
+		    (struct host_sp_status_block *)sc->def_sb_dma.vaddr;
+/***************/
+/* EVENT QUEUE */
+/***************/
+
+		if (bcm_dma_alloc(sc, BCM_PAGE_SIZE,
+				  &sc->eq_dma, "ev_queue",
+				  RTE_CACHE_LINE_SIZE) != 0) {
+			sc->def_sb = NULL;
+			return -1;
+		}
+
+		sc->eq = (union event_ring_elem *)sc->eq_dma.vaddr;
+
+/*************/
+/* SLOW PATH */
+/*************/
+
+		if (bcm_dma_alloc(sc, sizeof(struct bcm_slowpath),
+				  &sc->sp_dma, "sp",
+				  RTE_CACHE_LINE_SIZE) != 0) {
+			sc->eq = NULL;
+			sc->def_sb = NULL;
+			return -1;
+		}
+
+		sc->sp = (struct bcm_slowpath *)sc->sp_dma.vaddr;
+
+/*******************/
+/* SLOW PATH QUEUE */
+/*******************/
+
+		if (bcm_dma_alloc(sc, BCM_PAGE_SIZE,
+				  &sc->spq_dma, "sp_queue",
+				  RTE_CACHE_LINE_SIZE) != 0) {
+			sc->sp = NULL;
+			sc->eq = NULL;
+			sc->def_sb = NULL;
+			return -1;
+		}
+
+		sc->spq = (struct eth_spe *)sc->spq_dma.vaddr;
+
+/***************************/
+/* FW DECOMPRESSION BUFFER */
+/***************************/
+
+		if (bcm_dma_alloc(sc, FW_BUF_SIZE, &sc->gz_buf_dma,
+				  "fw_dec_buf", RTE_CACHE_LINE_SIZE) != 0) {
+			sc->spq = NULL;
+			sc->sp = NULL;
+			sc->eq = NULL;
+			sc->def_sb = NULL;
+			return -1;
+		}
+
+		sc->gz_buf = (void *)sc->gz_buf_dma.vaddr;
+	}
+
+	/*************/
+	/* FASTPATHS */
+	/*************/
+
+	/* allocate DMA memory for each fastpath structure */
+	for (i = 0; i < sc->num_queues; i++) {
+		fp = &sc->fp[i];
+		fp->sc = sc;
+		fp->index = i;
+
+/*******************/
+/* FP STATUS BLOCK */
+/*******************/
+
+		snprintf(buf, sizeof(buf), "fp_%d_sb", i);
+		if (bcm_dma_alloc(sc, sizeof(union bcm_host_hc_status_block),
+				  &fp->sb_dma, buf, RTE_CACHE_LINE_SIZE) != 0) {
+			PMD_DRV_LOG(NOTICE, "Failed to alloc %s", buf);
+			return -1;
+		} else {
+			if (CHIP_IS_E2E3(sc)) {
+				fp->status_block.e2_sb =
+				    (struct host_hc_status_block_e2 *)
+				    fp->sb_dma.vaddr;
+			} else {
+				fp->status_block.e1x_sb =
+				    (struct host_hc_status_block_e1x *)
+				    fp->sb_dma.vaddr;
+			}
+		}
+	}
+
+	return 0;
+}
+
+void bcm_free_hsi_mem(struct bcm_softc *sc)
+{
+	struct bcm_fastpath *fp;
+	int i;
+
+	for (i = 0; i < sc->num_queues; i++) {
+		fp = &sc->fp[i];
+
+/*******************/
+/* FP STATUS BLOCK */
+/*******************/
+
+		memset(&fp->status_block, 0, sizeof(fp->status_block));
+	}
+
+	/***************************/
+	/* FW DECOMPRESSION BUFFER */
+	/***************************/
+
+	sc->gz_buf = NULL;
+
+	/*******************/
+	/* SLOW PATH QUEUE */
+	/*******************/
+
+	sc->spq = NULL;
+
+	/*************/
+	/* SLOW PATH */
+	/*************/
+
+	sc->sp = NULL;
+
+	/***************/
+	/* EVENT QUEUE */
+	/***************/
+
+	sc->eq = NULL;
+
+	/************************/
+	/* DEFAULT STATUS BLOCK */
+	/************************/
+
+	sc->def_sb = NULL;
+
+}
+
+/*
+* Previous driver DMAE transaction may have occurred when pre-boot stage
+* ended and boot began. This would invalidate the addresses of the
+* transaction, resulting in was-error bit set in the PCI causing all
+* hw-to-host PCIe transactions to timeout. If this happened we want to clear
+* the interrupt which detected this from the pglueb and the was-done bit
+*/
+static void bcm_prev_interrupted_dmae(struct bcm_softc *sc)
+{
+	uint32_t val;
+
+	if (!CHIP_IS_E1x(sc)) {
+		val = REG_RD(sc, PGLUE_B_REG_PGLUE_B_INT_STS);
+		if (val & PGLUE_B_PGLUE_B_INT_STS_REG_WAS_ERROR_ATTN) {
+			REG_WR(sc, PGLUE_B_REG_WAS_ERROR_PF_7_0_CLR,
+			       1 << SC_FUNC(sc));
+		}
+	}
+}
+
+static int bcm_prev_mcp_done(struct bcm_softc *sc)
+{
+	uint32_t rc = bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_DONE,
+				     DRV_MSG_CODE_UNLOAD_SKIP_LINK_RESET);
+	if (!rc) {
+		PMD_DRV_LOG(NOTICE, "MCP response failure, aborting");
+		return -1;
+	}
+
+	return 0;
+}
+
+static struct bcm_prev_list_node *bcm_prev_path_get_entry(struct bcm_softc *sc)
+{
+	struct bcm_prev_list_node *tmp;
+
+	LIST_FOREACH(tmp, &bcm_prev_list, node) {
+		if ((sc->pcie_bus == tmp->bus) &&
+		    (sc->pcie_device == tmp->slot) &&
+		    (SC_PATH(sc) == tmp->path)) {
+			return tmp;
+		}
+	}
+
+	return NULL;
+}
+
+static uint8_t bcm_prev_is_path_marked(struct bcm_softc *sc)
+{
+	struct bcm_prev_list_node *tmp;
+	int rc = FALSE;
+
+	rte_spinlock_lock(&bcm_prev_mtx);
+
+	tmp = bcm_prev_path_get_entry(sc);
+	if (tmp) {
+		if (tmp->aer) {
+			PMD_DRV_LOG(DEBUG,
+				    "Path %d/%d/%d was marked by AER",
+				    sc->pcie_bus, sc->pcie_device, SC_PATH(sc));
+		} else {
+			rc = TRUE;
+			PMD_DRV_LOG(DEBUG,
+				    "Path %d/%d/%d was already cleaned from previous drivers",
+				    sc->pcie_bus, sc->pcie_device, SC_PATH(sc));
+		}
+	}
+
+	rte_spinlock_unlock(&bcm_prev_mtx);
+
+	return rc;
+}
+
+static int bcm_prev_mark_path(struct bcm_softc *sc, uint8_t after_undi)
+{
+	struct bcm_prev_list_node *tmp;
+
+	rte_spinlock_lock(&bcm_prev_mtx);
+
+	/* Check whether the entry for this path already exists */
+	tmp = bcm_prev_path_get_entry(sc);
+	if (tmp) {
+		if (!tmp->aer) {
+			PMD_DRV_LOG(DEBUG,
+				    "Re-marking AER in path %d/%d/%d",
+				    sc->pcie_bus, sc->pcie_device, SC_PATH(sc));
+		} else {
+			PMD_DRV_LOG(DEBUG,
+				    "Removing AER indication from path %d/%d/%d",
+				    sc->pcie_bus, sc->pcie_device, SC_PATH(sc));
+			tmp->aer = 0;
+		}
+
+		rte_spinlock_unlock(&bcm_prev_mtx);
+		return 0;
+	}
+
+	rte_spinlock_unlock(&bcm_prev_mtx);
+
+	/* Create an entry for this path and add it */
+	tmp = rte_malloc("", sizeof(struct bcm_prev_list_node),
+			 RTE_CACHE_LINE_SIZE);
+	if (!tmp) {
+		PMD_DRV_LOG(NOTICE, "Failed to allocate 'bcm_prev_list_node'");
+		return -1;
+	}
+
+	tmp->bus = sc->pcie_bus;
+	tmp->slot = sc->pcie_device;
+	tmp->path = SC_PATH(sc);
+	tmp->aer = 0;
+	tmp->undi = after_undi ? (1 << SC_PORT(sc)) : 0;
+
+	rte_spinlock_lock(&bcm_prev_mtx);
+
+	LIST_INSERT_HEAD(&bcm_prev_list, tmp, node);
+
+	rte_spinlock_unlock(&bcm_prev_mtx);
+
+	return 0;
+}
+
+static int bcm_do_flr(struct bcm_softc *sc)
+{
+	int i;
+
+	/* only E2 and onwards support FLR */
+	if (CHIP_IS_E1x(sc)) {
+		PMD_DRV_LOG(WARN, "FLR not supported in E1H");
+		return -1;
+	}
+
+	/* only bootcode REQ_BC_VER_4_INITIATE_FLR and onwards support flr */
+	if (sc->devinfo.bc_ver < REQ_BC_VER_4_INITIATE_FLR) {
+		PMD_DRV_LOG(WARN,
+			    "FLR not supported by BC_VER: 0x%08x",
+			    sc->devinfo.bc_ver);
+		return -1;
+	}
+
+	/* Wait for Transaction Pending bit clean */
+	for (i = 0; i < 4; i++) {
+		if (i) {
+			DELAY(((1 << (i - 1)) * 100) * 1000);
+		}
+
+		if (!bcm_is_pcie_pending(sc)) {
+			goto clear;
+		}
+	}
+
+	PMD_DRV_LOG(NOTICE, "PCIE transaction is not cleared, "
+		    "proceeding with reset anyway");
+
+clear:
+	bcm_fw_command(sc, DRV_MSG_CODE_INITIATE_FLR, 0);
+
+	return 0;
+}
+
+struct bcm_mac_vals {
+	uint32_t xmac_addr;
+	uint32_t xmac_val;
+	uint32_t emac_addr;
+	uint32_t emac_val;
+	uint32_t umac_addr;
+	uint32_t umac_val;
+	uint32_t bmac_addr;
+	uint32_t bmac_val[2];
+};
+
+static void
+bcm_prev_unload_close_mac(struct bcm_softc *sc, struct bcm_mac_vals *vals)
+{
+	uint32_t val, base_addr, offset, mask, reset_reg;
+	uint8_t mac_stopped = FALSE;
+	uint8_t port = SC_PORT(sc);
+	uint32_t wb_data[2];
+
+	/* reset addresses as they also mark which values were changed */
+	vals->bmac_addr = 0;
+	vals->umac_addr = 0;
+	vals->xmac_addr = 0;
+	vals->emac_addr = 0;
+
+	reset_reg = REG_RD(sc, MISC_REG_RESET_REG_2);
+
+	if (!CHIP_IS_E3(sc)) {
+		val = REG_RD(sc, NIG_REG_BMAC0_REGS_OUT_EN + port * 4);
+		mask = MISC_REGISTERS_RESET_REG_2_RST_BMAC0 << port;
+		if ((mask & reset_reg) && val) {
+			base_addr = SC_PORT(sc) ? NIG_REG_INGRESS_BMAC1_MEM
+			    : NIG_REG_INGRESS_BMAC0_MEM;
+			offset = CHIP_IS_E2(sc) ? BIGMAC2_REGISTER_BMAC_CONTROL
+			    : BIGMAC_REGISTER_BMAC_CONTROL;
+
+			/*
+			 * use rd/wr since we cannot use dmae. This is safe
+			 * since MCP won't access the bus due to the request
+			 * to unload, and no function on the path can be
+			 * loaded at this time.
+			 */
+			wb_data[0] = REG_RD(sc, base_addr + offset);
+			wb_data[1] = REG_RD(sc, base_addr + offset + 0x4);
+			vals->bmac_addr = base_addr + offset;
+			vals->bmac_val[0] = wb_data[0];
+			vals->bmac_val[1] = wb_data[1];
+			wb_data[0] &= ~ELINK_BMAC_CONTROL_RX_ENABLE;
+			REG_WR(sc, vals->bmac_addr, wb_data[0]);
+			REG_WR(sc, vals->bmac_addr + 0x4, wb_data[1]);
+		}
+
+		vals->emac_addr = NIG_REG_NIG_EMAC0_EN + SC_PORT(sc) * 4;
+		vals->emac_val = REG_RD(sc, vals->emac_addr);
+		REG_WR(sc, vals->emac_addr, 0);
+		mac_stopped = TRUE;
+	} else {
+		if (reset_reg & MISC_REGISTERS_RESET_REG_2_XMAC) {
+			base_addr = SC_PORT(sc) ? GRCBASE_XMAC1 : GRCBASE_XMAC0;
+			val = REG_RD(sc, base_addr + XMAC_REG_PFC_CTRL_HI);
+			REG_WR(sc, base_addr + XMAC_REG_PFC_CTRL_HI,
+			       val & ~(1 << 1));
+			REG_WR(sc, base_addr + XMAC_REG_PFC_CTRL_HI,
+			       val | (1 << 1));
+			vals->xmac_addr = base_addr + XMAC_REG_CTRL;
+			vals->xmac_val = REG_RD(sc, vals->xmac_addr);
+			REG_WR(sc, vals->xmac_addr, 0);
+			mac_stopped = TRUE;
+		}
+
+		mask = MISC_REGISTERS_RESET_REG_2_UMAC0 << port;
+		if (mask & reset_reg) {
+			base_addr = SC_PORT(sc) ? GRCBASE_UMAC1 : GRCBASE_UMAC0;
+			vals->umac_addr = base_addr + UMAC_REG_COMMAND_CONFIG;
+			vals->umac_val = REG_RD(sc, vals->umac_addr);
+			REG_WR(sc, vals->umac_addr, 0);
+			mac_stopped = TRUE;
+		}
+	}
+
+	if (mac_stopped) {
+		DELAY(20000);
+	}
+}
+
+#define BCM_PREV_UNDI_PROD_ADDR(p)  (BAR_TSTRORM_INTMEM + 0x1508 + ((p) << 4))
+#define BCM_PREV_UNDI_RCQ(val)      ((val) & 0xffff)
+#define BCM_PREV_UNDI_BD(val)       ((val) >> 16 & 0xffff)
+#define BCM_PREV_UNDI_PROD(rcq, bd) ((bd) << 16 | (rcq))
+
+static void
+bcm_prev_unload_undi_inc(struct bcm_softc *sc, uint8_t port, uint8_t inc)
+{
+	uint16_t rcq, bd;
+	uint32_t tmp_reg = REG_RD(sc, BCM_PREV_UNDI_PROD_ADDR(port));
+
+	rcq = BCM_PREV_UNDI_RCQ(tmp_reg) + inc;
+	bd = BCM_PREV_UNDI_BD(tmp_reg) + inc;
+
+	tmp_reg = BCM_PREV_UNDI_PROD(rcq, bd);
+	REG_WR(sc, BCM_PREV_UNDI_PROD_ADDR(port), tmp_reg);
+}
+
+static int bcm_prev_unload_common(struct bcm_softc *sc)
+{
+	uint32_t reset_reg, tmp_reg = 0, rc;
+	uint8_t prev_undi = FALSE;
+	struct bcm_mac_vals mac_vals;
+	uint32_t timer_count = 1000;
+	uint32_t prev_brb;
+
+	/*
+	 * It is possible a previous function received 'common' answer,
+	 * but hasn't loaded yet, therefore creating a scenario of
+	 * multiple functions receiving 'common' on the same path.
+	 */
+	memset(&mac_vals, 0, sizeof(mac_vals));
+
+	if (bcm_prev_is_path_marked(sc)) {
+		return bcm_prev_mcp_done(sc);
+	}
+
+	reset_reg = REG_RD(sc, MISC_REG_RESET_REG_1);
+
+	/* Reset should be performed after BRB is emptied */
+	if (reset_reg & MISC_REGISTERS_RESET_REG_1_RST_BRB1) {
+		/* Close the MAC Rx to prevent BRB from filling up */
+		bcm_prev_unload_close_mac(sc, &mac_vals);
+
+		/* close LLH filters towards the BRB */
+		elink_set_rx_filter(&sc->link_params, 0);
+
+		/*
+		 * Check if the UNDI driver was previously loaded.
+		 * UNDI driver initializes CID offset for normal bell to 0x7
+		 */
+		if (reset_reg & MISC_REGISTERS_RESET_REG_1_RST_DORQ) {
+			tmp_reg = REG_RD(sc, DORQ_REG_NORM_CID_OFST);
+			if (tmp_reg == 0x7) {
+				PMD_DRV_LOG(DEBUG, "UNDI previously loaded");
+				prev_undi = TRUE;
+				/* clear the UNDI indication */
+				REG_WR(sc, DORQ_REG_NORM_CID_OFST, 0);
+				/* clear possible idle check errors */
+				REG_RD(sc, NIG_REG_NIG_INT_STS_CLR_0);
+			}
+		}
+
+		/* wait until BRB is empty */
+		tmp_reg = REG_RD(sc, BRB1_REG_NUM_OF_FULL_BLOCKS);
+		while (timer_count) {
+			prev_brb = tmp_reg;
+
+			tmp_reg = REG_RD(sc, BRB1_REG_NUM_OF_FULL_BLOCKS);
+			if (!tmp_reg) {
+				break;
+			}
+
+			PMD_DRV_LOG(DEBUG, "BRB still has 0x%08x", tmp_reg);
+
+			/* reset timer as long as BRB actually gets emptied */
+			if (prev_brb > tmp_reg) {
+				timer_count = 1000;
+			} else {
+				timer_count--;
+			}
+
+			/* If UNDI resides in memory, manually increment it */
+			if (prev_undi) {
+				bcm_prev_unload_undi_inc(sc, SC_PORT(sc), 1);
+			}
+
+			DELAY(10);
+		}
+
+		if (!timer_count) {
+			PMD_DRV_LOG(NOTICE, "Failed to empty BRB");
+		}
+	}
+
+	/* No packets are in the pipeline, path is ready for reset */
+	bcm_reset_common(sc);
+
+	if (mac_vals.xmac_addr) {
+		REG_WR(sc, mac_vals.xmac_addr, mac_vals.xmac_val);
+	}
+	if (mac_vals.umac_addr) {
+		REG_WR(sc, mac_vals.umac_addr, mac_vals.umac_val);
+	}
+	if (mac_vals.emac_addr) {
+		REG_WR(sc, mac_vals.emac_addr, mac_vals.emac_val);
+	}
+	if (mac_vals.bmac_addr) {
+		REG_WR(sc, mac_vals.bmac_addr, mac_vals.bmac_val[0]);
+		REG_WR(sc, mac_vals.bmac_addr + 4, mac_vals.bmac_val[1]);
+	}
+
+	rc = bcm_prev_mark_path(sc, prev_undi);
+	if (rc) {
+		bcm_prev_mcp_done(sc);
+		return rc;
+	}
+
+	return bcm_prev_mcp_done(sc);
+}
+
+static int bcm_prev_unload_uncommon(struct bcm_softc *sc)
+{
+	int rc;
+
+	/* Test if previous unload process was already finished for this path */
+	if (bcm_prev_is_path_marked(sc)) {
+		return bcm_prev_mcp_done(sc);
+	}
+
+	/*
+	 * If function has FLR capabilities, and existing FW version matches
+	 * the one required, then FLR will be sufficient to clean any residue
+	 * left by previous driver
+	 */
+	rc = bcm_nic_load_analyze_req(sc, FW_MSG_CODE_DRV_LOAD_FUNCTION);
+	if (!rc) {
+		/* fw version is good */
+		rc = bcm_do_flr(sc);
+	}
+
+	if (!rc) {
+		/* FLR was performed */
+		return 0;
+	}
+
+	PMD_DRV_LOG(INFO, "Could not FLR");
+
+	/* Close the MCP request, return failure */
+	rc = bcm_prev_mcp_done(sc);
+	if (!rc) {
+		rc = BCM_PREV_WAIT_NEEDED;
+	}
+
+	return rc;
+}
+
+static int bcm_prev_unload(struct bcm_softc *sc)
+{
+	int time_counter = 10;
+	uint32_t fw, hw_lock_reg, hw_lock_val;
+	uint32_t rc = 0;
+
+	/*
+	 * Clear HW from errors which may have resulted from an interrupted
+	 * DMAE transaction.
+	 */
+	bcm_prev_interrupted_dmae(sc);
+
+	/* Release previously held locks */
+	if (SC_FUNC(sc) <= 5)
+		hw_lock_reg = (MISC_REG_DRIVER_CONTROL_1 + SC_FUNC(sc) * 8);
+	else
+		hw_lock_reg =
+		    (MISC_REG_DRIVER_CONTROL_7 + (SC_FUNC(sc) - 6) * 8);
+
+	hw_lock_val = (REG_RD(sc, hw_lock_reg));
+	if (hw_lock_val) {
+		if (hw_lock_val & HW_LOCK_RESOURCE_NVRAM) {
+			REG_WR(sc, MCP_REG_MCPR_NVM_SW_ARB,
+			       (MCPR_NVM_SW_ARB_ARB_REQ_CLR1 << SC_PORT(sc)));
+		}
+		REG_WR(sc, hw_lock_reg, 0xffffffff);
+	}
+
+	if (MCPR_ACCESS_LOCK_LOCK & REG_RD(sc, MCP_REG_MCPR_ACCESS_LOCK)) {
+		REG_WR(sc, MCP_REG_MCPR_ACCESS_LOCK, 0);
+	}
+
+	do {
+		/* Lock MCP using an unload request */
+		fw = bcm_fw_command(sc, DRV_MSG_CODE_UNLOAD_REQ_WOL_DIS, 0);
+		if (!fw) {
+			PMD_DRV_LOG(NOTICE, "MCP response failure, aborting");
+			rc = -1;
+			break;
+		}
+
+		if (fw == FW_MSG_CODE_DRV_UNLOAD_COMMON) {
+			rc = bcm_prev_unload_common(sc);
+			break;
+		}
+
+		/* non-common reply from MCP might require looping */
+		rc = bcm_prev_unload_uncommon(sc);
+		if (rc != BCM_PREV_WAIT_NEEDED) {
+			break;
+		}
+
+		DELAY(20000);
+	} while (--time_counter);
+
+	if (!time_counter || rc) {
+		PMD_DRV_LOG(NOTICE, "Failed to unload previous driver!");
+		rc = -1;
+	}
+
+	return rc;
+}
+
+static void
+bcm_dcbx_set_state(struct bcm_softc *sc, uint8_t dcb_on, uint32_t dcbx_enabled)
+{
+	if (!CHIP_IS_E1x(sc)) {
+		sc->dcb_state = dcb_on;
+		sc->dcbx_enabled = dcbx_enabled;
+	} else {
+		sc->dcb_state = FALSE;
+		sc->dcbx_enabled = BCM_DCBX_ENABLED_INVALID;
+	}
+	PMD_DRV_LOG(DEBUG,
+		    "DCB state [%s:%s]",
+		    dcb_on ? "ON" : "OFF",
+		    (dcbx_enabled == BCM_DCBX_ENABLED_OFF) ? "user-mode" :
+		    (dcbx_enabled ==
+		     BCM_DCBX_ENABLED_ON_NEG_OFF) ? "on-chip static"
+		    : (dcbx_enabled ==
+		       BCM_DCBX_ENABLED_ON_NEG_ON) ?
+		    "on-chip with negotiation" : "invalid");
+}
+
+static int bcm_set_qm_cid_count(struct bcm_softc *sc)
+{
+	int cid_count = BCM_L2_MAX_CID(sc);
+
+	if (CNIC_SUPPORT(sc)) {
+		cid_count += CNIC_CID_MAX;
+	}
+
+	return roundup(cid_count, QM_CID_ROUND);
+}
+
+static void bcm_init_multi_cos(struct bcm_softc *sc)
+{
+	int pri, cos;
+
+	uint32_t pri_map = 0;
+
+	for (pri = 0; pri < BCM_MAX_PRIORITY; pri++) {
+		cos = ((pri_map & (0xf << (pri * 4))) >> (pri * 4));
+		if (cos < sc->max_cos) {
+			sc->prio_to_cos[pri] = cos;
+		} else {
+			PMD_DRV_LOG(WARN,
+				    "Invalid COS %d for priority %d "
+				    "(max COS is %d), setting to 0", cos, pri,
+				    (sc->max_cos - 1));
+			sc->prio_to_cos[pri] = 0;
+		}
+	}
+}
+
+static int bcm_pci_get_caps(struct bcm_softc *sc)
+{
+	struct {
+		uint8_t id;
+		uint8_t next;
+	} pci_cap;
+	uint16_t status;
+	struct bcm_pci_cap *cap;
+
+	cap = sc->pci_caps = rte_zmalloc("caps", sizeof(struct bcm_pci_cap),
+					 RTE_CACHE_LINE_SIZE);
+	if (!cap) {
+		PMD_DRV_LOG(NOTICE, "Failed to allocate memory");
+		return -ENOMEM;
+	}
+
+	pci_read(sc, PCI_STATUS, &status, 2);
+	if (!(status & PCI_STATUS_CAP_LIST)) {
+		PMD_DRV_LOG(NOTICE, "PCIe capability reading failed");
+		return -1;
+	}
+
+	pci_read(sc, PCI_CAPABILITY_LIST, &pci_cap.next, 1);
+	while (pci_cap.next) {
+		cap->addr = pci_cap.next & ~3;
+		pci_read(sc, pci_cap.next & ~3, &pci_cap, 2);
+		if (pci_cap.id == 0xff)
+			break;
+		cap->id = pci_cap.id;
+		cap->type = BCM_PCI_CAP;
+		cap->next = rte_zmalloc("pci_cap",
+					sizeof(struct bcm_pci_cap),
+					RTE_CACHE_LINE_SIZE);
+		if (!cap->next) {
+			PMD_DRV_LOG(NOTICE, "Failed to allocate memory");
+			return -ENOMEM;
+		}
+		cap = cap->next;
+	}
+
+	return 0;
+}
+
+static void bcm_init_rte(struct bcm_softc *sc)
+{
+	sc->max_tx_queues = 128;
+	sc->max_rx_queues = 128;
+}
+
+#define FW_HEADER_LEN 104
+#define FW_NAME_57711 "/lib/firmware/bnx2x/bnx2x-e1h-7.2.51.0.fw"
+#define FW_NAME_57810 "/lib/firmware/bnx2x/bnx2x-e2-7.2.51.0.fw"
+
+void bcm_load_firmware(struct bcm_softc *sc)
+{
+	const char *fwname;
+	int f;
+	struct stat st;
+
+	fwname =
+	    sc->devinfo.device_id ==
+	    BROADCOM_DEV_ID_57711 ? FW_NAME_57711 : FW_NAME_57810;
+	f = open(fwname, O_RDONLY);
+	if (f < 0) {
+		PMD_DRV_LOG(NOTICE, "Can't open firmware file");
+		return;
+	}
+
+	if (fstat(f, &st) < 0) {
+		PMD_DRV_LOG(NOTICE, "Can't stat firmware file");
+		close(f);
+		return;
+	}
+
+	sc->firmware = rte_zmalloc("bcm_fw", st.st_size, RTE_CACHE_LINE_SIZE);
+	if (!sc->firmware) {
+		PMD_DRV_LOG(NOTICE, "Can't allocate memory for firmware");
+		close(f);
+		return;
+	}
+
+	if (read(f, sc->firmware, st.st_size) != st.st_size) {
+		PMD_DRV_LOG(NOTICE, "Can't read firmware data");
+		close(f);
+		return;
+	}
+	close(f);
+
+	sc->fw_len = st.st_size;
+	if (sc->fw_len < FW_HEADER_LEN) {
+		PMD_DRV_LOG(NOTICE, "Invalid fw size: %lu", sc->fw_len);
+		return;
+	}
+	PMD_DRV_LOG(DEBUG, "fw_len = %lu", sc->fw_len);
+}
+
+static void
+bcm_data_to_init_ops(uint8_t * data, struct raw_op *dst, uint32_t len)
+{
+	uint32_t *src = (uint32_t *) data;
+	uint32_t i, j, tmp;
+
+	for (i = 0, j = 0; i < len / 8; ++i, j += 2) {
+		tmp = rte_be_to_cpu_32(src[j]);
+		dst[i].op = (tmp >> 24) & 0xFF;
+		dst[i].offset = tmp & 0xFFFFFF;
+		dst[i].raw_data = rte_be_to_cpu_32(src[j + 1]);
+	}
+}
+
+static void
+bcm_data_to_init_offsets(uint8_t * data, uint16_t * dst, uint32_t len)
+{
+	uint16_t *src = (uint16_t *) data;
+	uint32_t i;
+
+	for (i = 0; i < len / 2; ++i)
+		dst[i] = rte_be_to_cpu_16(src[i]);
+}
+
+static void bcm_data_to_init_data(uint8_t * data, uint32_t * dst, uint32_t len)
+{
+	uint32_t *src = (uint32_t *) data;
+	uint32_t i;
+
+	for (i = 0; i < len / 4; ++i)
+		dst[i] = rte_be_to_cpu_32(src[i]);
+}
+
+static void bcm_data_to_iro_array(uint8_t * data, struct iro *dst, uint32_t len)
+{
+	uint32_t *src = (uint32_t *) data;
+	uint32_t i, j, tmp;
+
+	for (i = 0, j = 0; i < len / sizeof(struct iro); ++i, ++j) {
+		dst[i].base = rte_be_to_cpu_32(src[j++]);
+		tmp = rte_be_to_cpu_32(src[j]);
+		dst[i].m1 = (tmp >> 16) & 0xFFFF;
+		dst[i].m2 = tmp & 0xFFFF;
+		++j;
+		tmp = rte_be_to_cpu_32(src[j]);
+		dst[i].m3 = (tmp >> 16) & 0xFFFF;
+		dst[i].size = tmp & 0xFFFF;
+	}
+}
+
+/*
+* Device attach function.
+*
+* Allocates device resources, performs secondary chip identification, and
+* initializes driver instance variables. This function is called from driver
+* load after a successful probe.
+*
+* Returns:
+*   0 = Success, >0 = Failure
+*/
+int bcm_attach(struct bcm_softc *sc)
+{
+	int rc;
+
+	PMD_DRV_LOG(DEBUG, "Starting attach...");
+
+	rc = bcm_pci_get_caps(sc);
+	if (rc) {
+		PMD_DRV_LOG(NOTICE, "PCIe caps reading was failed");
+		return rc;
+	}
+
+	sc->state = BCM_STATE_CLOSED;
+
+	/* Init RTE stuff */
+	bcm_init_rte(sc);
+
+	pci_write_long(sc, PCICFG_GRC_ADDRESS, PCICFG_VENDOR_ID_OFFSET);
+
+	sc->igu_base_addr = IS_VF(sc) ? PXP_VF_ADDR_IGU_START : BAR_IGU_INTMEM;
+
+	/* get PCI capabilites */
+	bcm_probe_pci_caps(sc);
+
+	if (sc->devinfo.pcie_msix_cap_reg != 0) {
+		uint32_t val;
+		pci_read(sc,
+			 (sc->devinfo.pcie_msix_cap_reg + PCIR_MSIX_CTRL), &val,
+			 2);
+		sc->igu_sb_cnt = (val & PCIM_MSIXCTRL_TABLE_SIZE);
+	} else {
+		sc->igu_sb_cnt = 1;
+	}
+
+	if (IS_PF(sc)) {
+/* get device info and set params */
+		if (bcm_get_device_info(sc) != 0) {
+			PMD_DRV_LOG(NOTICE, "getting device info");
+			return -ENXIO;
+		}
+
+/* get phy settings from shmem and 'and' against admin settings */
+		bcm_get_phy_info(sc);
+	} else {
+/* Left mac of VF unfilled, PF should set it for VF */
+		memset(sc->link_params.mac_addr, 0, ETHER_ADDR_LEN);
+	}
+
+	sc->wol = 0;
+
+	/* set the default MTU (changed via ifconfig) */
+	sc->mtu = ETHER_MTU;
+
+	bcm_set_modes_bitmap(sc);
+
+	/* need to reset chip if UNDI was active */
+	if (IS_PF(sc) && !BCM_NOMCP(sc)) {
+/* init fw_seq */
+		sc->fw_seq =
+		    (SHMEM_RD(sc, func_mb[SC_FW_MB_IDX(sc)].drv_mb_header) &
+		     DRV_MSG_SEQ_NUMBER_MASK);
+		bcm_prev_unload(sc);
+	}
+
+	bcm_dcbx_set_state(sc, FALSE, BCM_DCBX_ENABLED_OFF);
+
+	/* calculate qm_cid_count */
+	sc->qm_cid_count = bcm_set_qm_cid_count(sc);
+
+	sc->max_cos = 1;
+	bcm_init_multi_cos(sc);
+
+	return 0;
+}
+
+static void
+bcm_igu_ack_sb(struct bcm_softc *sc, uint8_t igu_sb_id, uint8_t segment,
+	       uint16_t index, uint8_t op, uint8_t update)
+{
+	uint32_t igu_addr = sc->igu_base_addr;
+	igu_addr += (IGU_CMD_INT_ACK_BASE + igu_sb_id) * 8;
+	bcm_igu_ack_sb_gen(sc, segment, index, op, update, igu_addr);
+}
+
+static void
+bcm_ack_sb(struct bcm_softc *sc, uint8_t igu_sb_id, uint8_t storm,
+	   uint16_t index, uint8_t op, uint8_t update)
+{
+	if (unlikely(sc->devinfo.int_block == INT_BLOCK_HC))
+		bcm_hc_ack_sb(sc, igu_sb_id, storm, index, op, update);
+	else {
+		uint8_t segment;
+		if (CHIP_INT_MODE_IS_BC(sc)) {
+			segment = storm;
+		} else if (igu_sb_id != sc->igu_dsb_id) {
+			segment = IGU_SEG_ACCESS_DEF;
+		} else if (storm == ATTENTION_ID) {
+			segment = IGU_SEG_ACCESS_ATTN;
+		} else {
+			segment = IGU_SEG_ACCESS_DEF;
+		}
+		bcm_igu_ack_sb(sc, igu_sb_id, segment, index, op, update);
+	}
+}
+
+static void
+bcm_igu_clear_sb_gen(struct bcm_softc *sc, uint8_t func, uint8_t idu_sb_id,
+		     uint8_t is_pf)
+{
+	uint32_t data, ctl, cnt = 100;
+	uint32_t igu_addr_data = IGU_REG_COMMAND_REG_32LSB_DATA;
+	uint32_t igu_addr_ctl = IGU_REG_COMMAND_REG_CTRL;
+	uint32_t igu_addr_ack = IGU_REG_CSTORM_TYPE_0_SB_CLEANUP +
+	    (idu_sb_id / 32) * 4;
+	uint32_t sb_bit = 1 << (idu_sb_id % 32);
+	uint32_t func_encode = func |
+	    (is_pf ? 1 : 0) << IGU_FID_ENCODE_IS_PF_SHIFT;
+	uint32_t addr_encode = IGU_CMD_E2_PROD_UPD_BASE + idu_sb_id;
+
+	/* Not supported in BC mode */
+	if (CHIP_INT_MODE_IS_BC(sc)) {
+		return;
+	}
+
+	data = ((IGU_USE_REGISTER_cstorm_type_0_sb_cleanup <<
+		 IGU_REGULAR_CLEANUP_TYPE_SHIFT) |
+		IGU_REGULAR_CLEANUP_SET | IGU_REGULAR_BCLEANUP);
+
+	ctl = ((addr_encode << IGU_CTRL_REG_ADDRESS_SHIFT) |
+	       (func_encode << IGU_CTRL_REG_FID_SHIFT) |
+	       (IGU_CTRL_CMD_TYPE_WR << IGU_CTRL_REG_TYPE_SHIFT));
+
+	REG_WR(sc, igu_addr_data, data);
+
+	mb();
+
+	PMD_DRV_LOG(DEBUG, "write 0x%08x to IGU(via GRC) addr 0x%x",
+		    ctl, igu_addr_ctl);
+	REG_WR(sc, igu_addr_ctl, ctl);
+
+	mb();
+
+	/* wait for clean up to finish */
+	while (!(REG_RD(sc, igu_addr_ack) & sb_bit) && --cnt) {
+		DELAY(20000);
+	}
+
+	if (!(REG_RD(sc, igu_addr_ack) & sb_bit)) {
+		PMD_DRV_LOG(DEBUG,
+			    "Unable to finish IGU cleanup: "
+			    "idu_sb_id %d offset %d bit %d (cnt %d)",
+			    idu_sb_id, idu_sb_id / 32, idu_sb_id % 32, cnt);
+	}
+}
+
+static void bcm_igu_clear_sb(struct bcm_softc *sc, uint8_t idu_sb_id)
+{
+	bcm_igu_clear_sb_gen(sc, SC_FUNC(sc), idu_sb_id, TRUE /*PF*/);
+}
+
+/*******************/
+/* ECORE CALLBACKS */
+/*******************/
+
+static void bcm_reset_common(struct bcm_softc *sc)
+{
+	uint32_t val = 0x1400;
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* reset_common */
+	REG_WR(sc, (GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR),
+	       0xd3ffff7f);
+
+	if (CHIP_IS_E3(sc)) {
+		val |= MISC_REGISTERS_RESET_REG_2_MSTAT0;
+		val |= MISC_REGISTERS_RESET_REG_2_MSTAT1;
+	}
+
+	REG_WR(sc, (GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_CLEAR), val);
+}
+
+static void bcm_common_init_phy(struct bcm_softc *sc)
+{
+	uint32_t shmem_base[2];
+	uint32_t shmem2_base[2];
+
+	/* Avoid common init in case MFW supports LFA */
+	if (SHMEM2_RD(sc, size) >
+	    (uint32_t) offsetof(struct shmem2_region,
+				lfa_host_addr[SC_PORT(sc)])) {
+		return;
+	}
+
+	shmem_base[0] = sc->devinfo.shmem_base;
+	shmem2_base[0] = sc->devinfo.shmem2_base;
+
+	if (!CHIP_IS_E1x(sc)) {
+		shmem_base[1] = SHMEM2_RD(sc, other_shmem_base_addr);
+		shmem2_base[1] = SHMEM2_RD(sc, other_shmem2_base_addr);
+	}
+
+	elink_common_init_phy(sc, shmem_base, shmem2_base,
+			      sc->devinfo.chip_id, 0);
+}
+
+static void bcm_pf_disable(struct bcm_softc *sc)
+{
+	uint32_t val = REG_RD(sc, IGU_REG_PF_CONFIGURATION);
+
+	val &= ~IGU_PF_CONF_FUNC_EN;
+
+	REG_WR(sc, IGU_REG_PF_CONFIGURATION, val);
+	REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER, 0);
+	REG_WR(sc, CFC_REG_WEAK_ENABLE_PF, 0);
+}
+
+static void bcm_init_pxp(struct bcm_softc *sc)
+{
+	uint16_t devctl;
+	int r_order, w_order;
+
+	devctl = bcm_pcie_capability_read(sc, PCIR_EXPRESS_DEVICE_CTL);
+
+	w_order = ((devctl & PCIM_EXP_CTL_MAX_PAYLOAD) >> 5);
+	r_order = ((devctl & PCIM_EXP_CTL_MAX_READ_REQUEST) >> 12);
+
+	ecore_init_pxp_arb(sc, r_order, w_order);
+}
+
+static uint32_t bcm_get_pretend_reg(struct bcm_softc *sc)
+{
+	uint32_t base = PXP2_REG_PGL_PRETEND_FUNC_F0;
+	uint32_t stride = (PXP2_REG_PGL_PRETEND_FUNC_F1 - base);
+	return (base + (SC_ABS_FUNC(sc)) * stride);
+}
+
+/*
+ * Called only on E1H or E2.
+ * When pretending to be PF, the pretend value is the function number 0..7.
+ * When pretending to be VF, the pretend val is the PF-num:VF-valid:ABS-VFID
+ * combination.
+ */
+static int bcm_pretend_func(struct bcm_softc *sc, uint16_t pretend_func_val)
+{
+	uint32_t pretend_reg;
+
+	if (CHIP_IS_E1H(sc) && (pretend_func_val > E1H_FUNC_MAX))
+		return -1;
+
+	/* get my own pretend register */
+	pretend_reg = bcm_get_pretend_reg(sc);
+	REG_WR(sc, pretend_reg, pretend_func_val);
+	REG_RD(sc, pretend_reg);
+	return 0;
+}
+
+static void bcm_setup_fan_failure_detection(struct bcm_softc *sc)
+{
+	int is_required;
+	uint32_t val;
+	int port;
+
+	is_required = 0;
+	val = (SHMEM_RD(sc, dev_info.shared_hw_config.config2) &
+	       SHARED_HW_CFG_FAN_FAILURE_MASK);
+
+	if (val == SHARED_HW_CFG_FAN_FAILURE_ENABLED) {
+		is_required = 1;
+	}
+	/*
+	 * The fan failure mechanism is usually related to the PHY type since
+	 * the power consumption of the board is affected by the PHY. Currently,
+	 * fan is required for most designs with SFX7101, BCM8727 and BCM8481.
+	 */
+	else if (val == SHARED_HW_CFG_FAN_FAILURE_PHY_TYPE) {
+		for (port = PORT_0; port < PORT_MAX; port++) {
+			is_required |= elink_fan_failure_det_req(sc,
+								 sc->
+								 devinfo.shmem_base,
+								 sc->
+								 devinfo.shmem2_base,
+								 port);
+		}
+	}
+
+	if (is_required == 0) {
+		return;
+	}
+
+	/* Fan failure is indicated by SPIO 5 */
+	bcm_set_spio(sc, MISC_SPIO_SPIO5, MISC_SPIO_INPUT_HI_Z);
+
+	/* set to active low mode */
+	val = REG_RD(sc, MISC_REG_SPIO_INT);
+	val |= (MISC_SPIO_SPIO5 << MISC_SPIO_INT_OLD_SET_POS);
+	REG_WR(sc, MISC_REG_SPIO_INT, val);
+
+	/* enable interrupt to signal the IGU */
+	val = REG_RD(sc, MISC_REG_SPIO_EVENT_EN);
+	val |= MISC_SPIO_SPIO5;
+	REG_WR(sc, MISC_REG_SPIO_EVENT_EN, val);
+}
+
+static void bcm_enable_blocks_attention(struct bcm_softc *sc)
+{
+	uint32_t val;
+
+	REG_WR(sc, PXP_REG_PXP_INT_MASK_0, 0);
+	if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, PXP_REG_PXP_INT_MASK_1, 0x40);
+	} else {
+		REG_WR(sc, PXP_REG_PXP_INT_MASK_1, 0);
+	}
+	REG_WR(sc, DORQ_REG_DORQ_INT_MASK, 0);
+	REG_WR(sc, CFC_REG_CFC_INT_MASK, 0);
+	/*
+	 * mask read length error interrupts in brb for parser
+	 * (parsing unit and 'checksum and crc' unit)
+	 * these errors are legal (PU reads fixed length and CAC can cause
+	 * read length error on truncated packets)
+	 */
+	REG_WR(sc, BRB1_REG_BRB1_INT_MASK, 0xFC00);
+	REG_WR(sc, QM_REG_QM_INT_MASK, 0);
+	REG_WR(sc, TM_REG_TM_INT_MASK, 0);
+	REG_WR(sc, XSDM_REG_XSDM_INT_MASK_0, 0);
+	REG_WR(sc, XSDM_REG_XSDM_INT_MASK_1, 0);
+	REG_WR(sc, XCM_REG_XCM_INT_MASK, 0);
+	/*      REG_WR(sc, XSEM_REG_XSEM_INT_MASK_0, 0); */
+	/*      REG_WR(sc, XSEM_REG_XSEM_INT_MASK_1, 0); */
+	REG_WR(sc, USDM_REG_USDM_INT_MASK_0, 0);
+	REG_WR(sc, USDM_REG_USDM_INT_MASK_1, 0);
+	REG_WR(sc, UCM_REG_UCM_INT_MASK, 0);
+	/*      REG_WR(sc, USEM_REG_USEM_INT_MASK_0, 0); */
+	/*      REG_WR(sc, USEM_REG_USEM_INT_MASK_1, 0); */
+	REG_WR(sc, GRCBASE_UPB + PB_REG_PB_INT_MASK, 0);
+	REG_WR(sc, CSDM_REG_CSDM_INT_MASK_0, 0);
+	REG_WR(sc, CSDM_REG_CSDM_INT_MASK_1, 0);
+	REG_WR(sc, CCM_REG_CCM_INT_MASK, 0);
+	/*      REG_WR(sc, CSEM_REG_CSEM_INT_MASK_0, 0); */
+	/*      REG_WR(sc, CSEM_REG_CSEM_INT_MASK_1, 0); */
+
+	val = (PXP2_PXP2_INT_MASK_0_REG_PGL_CPL_AFT |
+	       PXP2_PXP2_INT_MASK_0_REG_PGL_CPL_OF |
+	       PXP2_PXP2_INT_MASK_0_REG_PGL_PCIE_ATTN);
+	if (!CHIP_IS_E1x(sc)) {
+		val |= (PXP2_PXP2_INT_MASK_0_REG_PGL_READ_BLOCKED |
+			PXP2_PXP2_INT_MASK_0_REG_PGL_WRITE_BLOCKED);
+	}
+	REG_WR(sc, PXP2_REG_PXP2_INT_MASK_0, val);
+
+	REG_WR(sc, TSDM_REG_TSDM_INT_MASK_0, 0);
+	REG_WR(sc, TSDM_REG_TSDM_INT_MASK_1, 0);
+	REG_WR(sc, TCM_REG_TCM_INT_MASK, 0);
+	/*      REG_WR(sc, TSEM_REG_TSEM_INT_MASK_0, 0); */
+
+	if (!CHIP_IS_E1x(sc)) {
+/* enable VFC attentions: bits 11 and 12, bits 31:13 reserved */
+		REG_WR(sc, TSEM_REG_TSEM_INT_MASK_1, 0x07ff);
+	}
+
+	REG_WR(sc, CDU_REG_CDU_INT_MASK, 0);
+	REG_WR(sc, DMAE_REG_DMAE_INT_MASK, 0);
+	/*      REG_WR(sc, MISC_REG_MISC_INT_MASK, 0); */
+	REG_WR(sc, PBF_REG_PBF_INT_MASK, 0x18);	/* bit 3,4 masked */
+}
+
+/**
+ * bcm_init_hw_common - initialize the HW at the COMMON phase.
+ *
+ * @sc:     driver handle
+ */
+static int bcm_init_hw_common(struct bcm_softc *sc)
+{
+	uint8_t abs_func_id;
+	uint32_t val;
+
+	PMD_DRV_LOG(DEBUG, "starting common init for func %d", SC_ABS_FUNC(sc));
+
+	/*
+	 * take the RESET lock to protect undi_unload flow from accessing
+	 * registers while we are resetting the chip
+	 */
+	bcm_acquire_hw_lock(sc, HW_LOCK_RESOURCE_RESET);
+
+	bcm_reset_common(sc);
+
+	REG_WR(sc, (GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET), 0xffffffff);
+
+	val = 0xfffc;
+	if (CHIP_IS_E3(sc)) {
+		val |= MISC_REGISTERS_RESET_REG_2_MSTAT0;
+		val |= MISC_REGISTERS_RESET_REG_2_MSTAT1;
+	}
+
+	REG_WR(sc, (GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_SET), val);
+
+	bcm_release_hw_lock(sc, HW_LOCK_RESOURCE_RESET);
+
+	ecore_init_block(sc, BLOCK_MISC, PHASE_COMMON);
+
+	if (!CHIP_IS_E1x(sc)) {
+/*
+ * 4-port mode or 2-port mode we need to turn off master-enable for
+ * everyone. After that we turn it back on for self. So, we disregard
+ * multi-function, and always disable all functions on the given path,
+ * this means 0,2,4,6 for path 0 and 1,3,5,7 for path 1
+ */
+		for (abs_func_id = SC_PATH(sc);
+		     abs_func_id < (E2_FUNC_MAX * 2); abs_func_id += 2) {
+			if (abs_func_id == SC_ABS_FUNC(sc)) {
+				REG_WR(sc,
+				       PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER,
+				       1);
+				continue;
+			}
+
+			bcm_pretend_func(sc, abs_func_id);
+
+			/* clear pf enable */
+			bcm_pf_disable(sc);
+
+			bcm_pretend_func(sc, SC_ABS_FUNC(sc));
+		}
+	}
+
+	ecore_init_block(sc, BLOCK_PXP, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_PXP2, PHASE_COMMON);
+	bcm_init_pxp(sc);
+
+#ifdef __BIG_ENDIAN
+	REG_WR(sc, PXP2_REG_RQ_QM_ENDIAN_M, 1);
+	REG_WR(sc, PXP2_REG_RQ_TM_ENDIAN_M, 1);
+	REG_WR(sc, PXP2_REG_RQ_SRC_ENDIAN_M, 1);
+	REG_WR(sc, PXP2_REG_RQ_CDU_ENDIAN_M, 1);
+	REG_WR(sc, PXP2_REG_RQ_DBG_ENDIAN_M, 1);
+	/* make sure this value is 0 */
+	REG_WR(sc, PXP2_REG_RQ_HC_ENDIAN_M, 0);
+
+	//REG_WR(sc, PXP2_REG_RD_PBF_SWAP_MODE, 1);
+	REG_WR(sc, PXP2_REG_RD_QM_SWAP_MODE, 1);
+	REG_WR(sc, PXP2_REG_RD_TM_SWAP_MODE, 1);
+	REG_WR(sc, PXP2_REG_RD_SRC_SWAP_MODE, 1);
+	REG_WR(sc, PXP2_REG_RD_CDURD_SWAP_MODE, 1);
+#endif
+
+	ecore_ilt_init_page_size(sc, INITOP_SET);
+
+	if (CHIP_REV_IS_FPGA(sc) && CHIP_IS_E1H(sc)) {
+		REG_WR(sc, PXP2_REG_PGL_TAGS_LIMIT, 0x1);
+	}
+
+	/* let the HW do it's magic... */
+	DELAY(100000);
+
+	/* finish PXP init */
+
+	val = REG_RD(sc, PXP2_REG_RQ_CFG_DONE);
+	if (val != 1) {
+		PMD_DRV_LOG(NOTICE, "PXP2 CFG failed");
+		return -1;
+	}
+	val = REG_RD(sc, PXP2_REG_RD_INIT_DONE);
+	if (val != 1) {
+		PMD_DRV_LOG(NOTICE, "PXP2 RD_INIT failed");
+		return -1;
+	}
+
+	/*
+	 * Timer bug workaround for E2 only. We need to set the entire ILT to have
+	 * entries with value "0" and valid bit on. This needs to be done by the
+	 * first PF that is loaded in a path (i.e. common phase)
+	 */
+	if (!CHIP_IS_E1x(sc)) {
+/*
+ * In E2 there is a bug in the timers block that can cause function 6 / 7
+ * (i.e. vnic3) to start even if it is marked as "scan-off".
+ * This occurs when a different function (func2,3) is being marked
+ * as "scan-off". Real-life scenario for example: if a driver is being
+ * load-unloaded while func6,7 are down. This will cause the timer to access
+ * the ilt, translate to a logical address and send a request to read/write.
+ * Since the ilt for the function that is down is not valid, this will cause
+ * a translation error which is unrecoverable.
+ * The Workaround is intended to make sure that when this happens nothing
+ * fatal will occur. The workaround:
+ *  1.  First PF driver which loads on a path will:
+ *      a.  After taking the chip out of reset, by using pretend,
+ *          it will write "0" to the following registers of
+ *          the other vnics.
+ *          REG_WR(pdev, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER, 0);
+ *          REG_WR(pdev, CFC_REG_WEAK_ENABLE_PF,0);
+ *          REG_WR(pdev, CFC_REG_STRONG_ENABLE_PF,0);
+ *          And for itself it will write '1' to
+ *          PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER to enable
+ *          dmae-operations (writing to pram for example.)
+ *          note: can be done for only function 6,7 but cleaner this
+ *            way.
+ *      b.  Write zero+valid to the entire ILT.
+ *      c.  Init the first_timers_ilt_entry, last_timers_ilt_entry of
+ *          VNIC3 (of that port). The range allocated will be the
+ *          entire ILT. This is needed to prevent  ILT range error.
+ *  2.  Any PF driver load flow:
+ *      a.  ILT update with the physical addresses of the allocated
+ *          logical pages.
+ *      b.  Wait 20msec. - note that this timeout is needed to make
+ *          sure there are no requests in one of the PXP internal
+ *          queues with "old" ILT addresses.
+ *      c.  PF enable in the PGLC.
+ *      d.  Clear the was_error of the PF in the PGLC. (could have
+ *          occurred while driver was down)
+ *      e.  PF enable in the CFC (WEAK + STRONG)
+ *      f.  Timers scan enable
+ *  3.  PF driver unload flow:
+ *      a.  Clear the Timers scan_en.
+ *      b.  Polling for scan_on=0 for that PF.
+ *      c.  Clear the PF enable bit in the PXP.
+ *      d.  Clear the PF enable in the CFC (WEAK + STRONG)
+ *      e.  Write zero+valid to all ILT entries (The valid bit must
+ *          stay set)
+ *      f.  If this is VNIC 3 of a port then also init
+ *          first_timers_ilt_entry to zero and last_timers_ilt_entry
+ *          to the last enrty in the ILT.
+ *
+ *      Notes:
+ *      Currently the PF error in the PGLC is non recoverable.
+ *      In the future the there will be a recovery routine for this error.
+ *      Currently attention is masked.
+ *      Having an MCP lock on the load/unload process does not guarantee that
+ *      there is no Timer disable during Func6/7 enable. This is because the
+ *      Timers scan is currently being cleared by the MCP on FLR.
+ *      Step 2.d can be done only for PF6/7 and the driver can also check if
+ *      there is error before clearing it. But the flow above is simpler and
+ *      more general.
+ *      All ILT entries are written by zero+valid and not just PF6/7
+ *      ILT entries since in the future the ILT entries allocation for
+ *      PF-s might be dynamic.
+ */
+		struct ilt_client_info ilt_cli;
+		struct ecore_ilt ilt;
+
+		memset(&ilt_cli, 0, sizeof(struct ilt_client_info));
+		memset(&ilt, 0, sizeof(struct ecore_ilt));
+
+/* initialize dummy TM client */
+		ilt_cli.start = 0;
+		ilt_cli.end = ILT_NUM_PAGE_ENTRIES - 1;
+		ilt_cli.client_num = ILT_CLIENT_TM;
+
+/*
+ * Step 1: set zeroes to all ilt page entries with valid bit on
+ * Step 2: set the timers first/last ilt entry to point
+ * to the entire range to prevent ILT range error for 3rd/4th
+ * vnic (this code assumes existence of the vnic)
+ *
+ * both steps performed by call to ecore_ilt_client_init_op()
+ * with dummy TM client
+ *
+ * we must use pretend since PXP2_REG_RQ_##blk##_FIRST_ILT
+ * and his brother are split registers
+ */
+
+		bcm_pretend_func(sc, (SC_PATH(sc) + 6));
+		ecore_ilt_client_init_op_ilt(sc, &ilt, &ilt_cli, INITOP_CLEAR);
+		bcm_pretend_func(sc, SC_ABS_FUNC(sc));
+
+		REG_WR(sc, PXP2_REG_RQ_DRAM_ALIGN, BCM_PXP_DRAM_ALIGN);
+		REG_WR(sc, PXP2_REG_RQ_DRAM_ALIGN_RD, BCM_PXP_DRAM_ALIGN);
+		REG_WR(sc, PXP2_REG_RQ_DRAM_ALIGN_SEL, 1);
+	}
+
+	REG_WR(sc, PXP2_REG_RQ_DISABLE_INPUTS, 0);
+	REG_WR(sc, PXP2_REG_RD_DISABLE_INPUTS, 0);
+
+	if (!CHIP_IS_E1x(sc)) {
+		int factor = 0;
+
+		ecore_init_block(sc, BLOCK_PGLUE_B, PHASE_COMMON);
+		ecore_init_block(sc, BLOCK_ATC, PHASE_COMMON);
+
+/* let the HW do it's magic... */
+		do {
+			DELAY(200000);
+			val = REG_RD(sc, ATC_REG_ATC_INIT_DONE);
+		} while (factor-- && (val != 1));
+
+		if (val != 1) {
+			PMD_DRV_LOG(NOTICE, "ATC_INIT failed");
+			return -1;
+		}
+	}
+
+	ecore_init_block(sc, BLOCK_DMAE, PHASE_COMMON);
+
+	/* clean the DMAE memory */
+	sc->dmae_ready = 1;
+	ecore_init_fill(sc, TSEM_REG_PRAM, 0, 8);
+
+	ecore_init_block(sc, BLOCK_TCM, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_UCM, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_CCM, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_XCM, PHASE_COMMON);
+
+	bcm_read_dmae(sc, XSEM_REG_PASSIVE_BUFFER, 3);
+	bcm_read_dmae(sc, CSEM_REG_PASSIVE_BUFFER, 3);
+	bcm_read_dmae(sc, TSEM_REG_PASSIVE_BUFFER, 3);
+	bcm_read_dmae(sc, USEM_REG_PASSIVE_BUFFER, 3);
+
+	ecore_init_block(sc, BLOCK_QM, PHASE_COMMON);
+
+	/* QM queues pointers table */
+	ecore_qm_init_ptr_table(sc, sc->qm_cid_count, INITOP_SET);
+
+	/* soft reset pulse */
+	REG_WR(sc, QM_REG_SOFT_RESET, 1);
+	REG_WR(sc, QM_REG_SOFT_RESET, 0);
+
+	if (CNIC_SUPPORT(sc))
+		ecore_init_block(sc, BLOCK_TM, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_DORQ, PHASE_COMMON);
+	REG_WR(sc, DORQ_REG_DPM_CID_OFST, BCM_DB_SHIFT);
+
+	if (!CHIP_REV_IS_SLOW(sc)) {
+/* enable hw interrupt from doorbell Q */
+		REG_WR(sc, DORQ_REG_DORQ_INT_MASK, 0);
+	}
+
+	ecore_init_block(sc, BLOCK_BRB1, PHASE_COMMON);
+
+	ecore_init_block(sc, BLOCK_PRS, PHASE_COMMON);
+	REG_WR(sc, PRS_REG_A_PRSU_20, 0xf);
+	REG_WR(sc, PRS_REG_E1HOV_MODE, sc->devinfo.mf_info.path_has_ovlan);
+
+	if (!CHIP_IS_E1x(sc) && !CHIP_IS_E3B0(sc)) {
+		if (IS_MF_AFEX(sc)) {
+			/*
+			 * configure that AFEX and VLAN headers must be
+			 * received in AFEX mode
+			 */
+			REG_WR(sc, PRS_REG_HDRS_AFTER_BASIC, 0xE);
+			REG_WR(sc, PRS_REG_MUST_HAVE_HDRS, 0xA);
+			REG_WR(sc, PRS_REG_HDRS_AFTER_TAG_0, 0x6);
+			REG_WR(sc, PRS_REG_TAG_ETHERTYPE_0, 0x8926);
+			REG_WR(sc, PRS_REG_TAG_LEN_0, 0x4);
+		} else {
+			/*
+			 * Bit-map indicating which L2 hdrs may appear
+			 * after the basic Ethernet header
+			 */
+			REG_WR(sc, PRS_REG_HDRS_AFTER_BASIC,
+			       sc->devinfo.mf_info.path_has_ovlan ? 7 : 6);
+		}
+	}
+
+	ecore_init_block(sc, BLOCK_TSDM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_CSDM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_USDM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_XSDM, PHASE_COMMON);
+
+	if (!CHIP_IS_E1x(sc)) {
+/* reset VFC memories */
+		REG_WR(sc, TSEM_REG_FAST_MEMORY + VFC_REG_MEMORIES_RST,
+		       VFC_MEMORIES_RST_REG_CAM_RST |
+		       VFC_MEMORIES_RST_REG_RAM_RST);
+		REG_WR(sc, XSEM_REG_FAST_MEMORY + VFC_REG_MEMORIES_RST,
+		       VFC_MEMORIES_RST_REG_CAM_RST |
+		       VFC_MEMORIES_RST_REG_RAM_RST);
+
+		DELAY(20000);
+	}
+
+	ecore_init_block(sc, BLOCK_TSEM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_USEM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_CSEM, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_XSEM, PHASE_COMMON);
+
+	/* sync semi rtc */
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR, 0x80000000);
+	REG_WR(sc, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET, 0x80000000);
+
+	ecore_init_block(sc, BLOCK_UPB, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_XPB, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_PBF, PHASE_COMMON);
+
+	if (!CHIP_IS_E1x(sc)) {
+		if (IS_MF_AFEX(sc)) {
+			/*
+			 * configure that AFEX and VLAN headers must be
+			 * sent in AFEX mode
+			 */
+			REG_WR(sc, PBF_REG_HDRS_AFTER_BASIC, 0xE);
+			REG_WR(sc, PBF_REG_MUST_HAVE_HDRS, 0xA);
+			REG_WR(sc, PBF_REG_HDRS_AFTER_TAG_0, 0x6);
+			REG_WR(sc, PBF_REG_TAG_ETHERTYPE_0, 0x8926);
+			REG_WR(sc, PBF_REG_TAG_LEN_0, 0x4);
+		} else {
+			REG_WR(sc, PBF_REG_HDRS_AFTER_BASIC,
+			       sc->devinfo.mf_info.path_has_ovlan ? 7 : 6);
+		}
+	}
+
+	REG_WR(sc, SRC_REG_SOFT_RST, 1);
+
+	ecore_init_block(sc, BLOCK_SRC, PHASE_COMMON);
+
+	if (CNIC_SUPPORT(sc)) {
+		REG_WR(sc, SRC_REG_KEYSEARCH_0, 0x63285672);
+		REG_WR(sc, SRC_REG_KEYSEARCH_1, 0x24b8f2cc);
+		REG_WR(sc, SRC_REG_KEYSEARCH_2, 0x223aef9b);
+		REG_WR(sc, SRC_REG_KEYSEARCH_3, 0x26001e3a);
+		REG_WR(sc, SRC_REG_KEYSEARCH_4, 0x7ae91116);
+		REG_WR(sc, SRC_REG_KEYSEARCH_5, 0x5ce5230b);
+		REG_WR(sc, SRC_REG_KEYSEARCH_6, 0x298d8adf);
+		REG_WR(sc, SRC_REG_KEYSEARCH_7, 0x6eb0ff09);
+		REG_WR(sc, SRC_REG_KEYSEARCH_8, 0x1830f82f);
+		REG_WR(sc, SRC_REG_KEYSEARCH_9, 0x01e46be7);
+	}
+	REG_WR(sc, SRC_REG_SOFT_RST, 0);
+
+	if (sizeof(union cdu_context) != 1024) {
+/* we currently assume that a context is 1024 bytes */
+		PMD_DRV_LOG(NOTICE,
+			    "please adjust the size of cdu_context(%ld)",
+			    (long)sizeof(union cdu_context));
+	}
+
+	ecore_init_block(sc, BLOCK_CDU, PHASE_COMMON);
+	val = (4 << 24) + (0 << 12) + 1024;
+	REG_WR(sc, CDU_REG_CDU_GLOBAL_PARAMS, val);
+
+	ecore_init_block(sc, BLOCK_CFC, PHASE_COMMON);
+
+	REG_WR(sc, CFC_REG_INIT_REG, 0x7FF);
+	/* enable context validation interrupt from CFC */
+	REG_WR(sc, CFC_REG_CFC_INT_MASK, 0);
+
+	/* set the thresholds to prevent CFC/CDU race */
+	REG_WR(sc, CFC_REG_DEBUG0, 0x20020000);
+	ecore_init_block(sc, BLOCK_HC, PHASE_COMMON);
+
+	if (!CHIP_IS_E1x(sc) && BCM_NOMCP(sc)) {
+		REG_WR(sc, IGU_REG_RESET_MEMORIES, 0x36);
+	}
+
+	ecore_init_block(sc, BLOCK_IGU, PHASE_COMMON);
+	ecore_init_block(sc, BLOCK_MISC_AEU, PHASE_COMMON);
+
+	/* Reset PCIE errors for debug */
+	REG_WR(sc, 0x2814, 0xffffffff);
+	REG_WR(sc, 0x3820, 0xffffffff);
+
+	if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, PCICFG_OFFSET + PXPCS_TL_CONTROL_5,
+		       (PXPCS_TL_CONTROL_5_ERR_UNSPPORT1 |
+			PXPCS_TL_CONTROL_5_ERR_UNSPPORT));
+		REG_WR(sc, PCICFG_OFFSET + PXPCS_TL_FUNC345_STAT,
+		       (PXPCS_TL_FUNC345_STAT_ERR_UNSPPORT4 |
+			PXPCS_TL_FUNC345_STAT_ERR_UNSPPORT3 |
+			PXPCS_TL_FUNC345_STAT_ERR_UNSPPORT2));
+		REG_WR(sc, PCICFG_OFFSET + PXPCS_TL_FUNC678_STAT,
+		       (PXPCS_TL_FUNC678_STAT_ERR_UNSPPORT7 |
+			PXPCS_TL_FUNC678_STAT_ERR_UNSPPORT6 |
+			PXPCS_TL_FUNC678_STAT_ERR_UNSPPORT5));
+	}
+
+	ecore_init_block(sc, BLOCK_NIG, PHASE_COMMON);
+
+	/* in E3 this done in per-port section */
+	if (!CHIP_IS_E3(sc))
+		REG_WR(sc, NIG_REG_LLH_MF_MODE, IS_MF(sc));
+
+	if (CHIP_IS_E1H(sc)) {
+/* not applicable for E2 (and above ...) */
+		REG_WR(sc, NIG_REG_LLH_E1HOV_MODE, IS_MF_SD(sc));
+	}
+
+	if (CHIP_REV_IS_SLOW(sc)) {
+		DELAY(200000);
+	}
+
+	/* finish CFC init */
+	val = reg_poll(sc, CFC_REG_LL_INIT_DONE, 1, 100, 10);
+	if (val != 1) {
+		PMD_DRV_LOG(NOTICE, "CFC LL_INIT failed");
+		return -1;
+	}
+	val = reg_poll(sc, CFC_REG_AC_INIT_DONE, 1, 100, 10);
+	if (val != 1) {
+		PMD_DRV_LOG(NOTICE, "CFC AC_INIT failed");
+		return -1;
+	}
+	val = reg_poll(sc, CFC_REG_CAM_INIT_DONE, 1, 100, 10);
+	if (val != 1) {
+		PMD_DRV_LOG(NOTICE, "CFC CAM_INIT failed");
+		return -1;
+	}
+	REG_WR(sc, CFC_REG_DEBUG0, 0);
+
+	bcm_setup_fan_failure_detection(sc);
+
+	/* clear PXP2 attentions */
+	REG_RD(sc, PXP2_REG_PXP2_INT_STS_CLR_0);
+
+	bcm_enable_blocks_attention(sc);
+
+	if (!CHIP_REV_IS_SLOW(sc)) {
+		ecore_enable_blocks_parity(sc);
+	}
+
+	if (!BCM_NOMCP(sc)) {
+		if (CHIP_IS_E1x(sc)) {
+			bcm_common_init_phy(sc);
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * bcm_init_hw_common_chip - init HW at the COMMON_CHIP phase.
+ *
+ * @sc:     driver handle
+ */
+static int bcm_init_hw_common_chip(struct bcm_softc *sc)
+{
+	int rc = bcm_init_hw_common(sc);
+
+	if (rc) {
+		return rc;
+	}
+
+	/* In E2 2-PORT mode, same ext phy is used for the two paths */
+	if (!BCM_NOMCP(sc)) {
+		bcm_common_init_phy(sc);
+	}
+
+	return 0;
+}
+
+static int bcm_init_hw_port(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	int init_phase = port ? PHASE_PORT1 : PHASE_PORT0;
+	uint32_t low, high;
+	uint32_t val;
+
+	PMD_DRV_LOG(DEBUG, "starting port init for port %d", port);
+
+	REG_WR(sc, NIG_REG_MASK_INTERRUPT_PORT0 + port * 4, 0);
+
+	ecore_init_block(sc, BLOCK_MISC, init_phase);
+	ecore_init_block(sc, BLOCK_PXP, init_phase);
+	ecore_init_block(sc, BLOCK_PXP2, init_phase);
+
+	/*
+	 * Timers bug workaround: disables the pf_master bit in pglue at
+	 * common phase, we need to enable it here before any dmae access are
+	 * attempted. Therefore we manually added the enable-master to the
+	 * port phase (it also happens in the function phase)
+	 */
+	if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER, 1);
+	}
+
+	ecore_init_block(sc, BLOCK_ATC, init_phase);
+	ecore_init_block(sc, BLOCK_DMAE, init_phase);
+	ecore_init_block(sc, BLOCK_PGLUE_B, init_phase);
+	ecore_init_block(sc, BLOCK_QM, init_phase);
+
+	ecore_init_block(sc, BLOCK_TCM, init_phase);
+	ecore_init_block(sc, BLOCK_UCM, init_phase);
+	ecore_init_block(sc, BLOCK_CCM, init_phase);
+	ecore_init_block(sc, BLOCK_XCM, init_phase);
+
+	/* QM cid (connection) count */
+	ecore_qm_init_cid_count(sc, sc->qm_cid_count, INITOP_SET);
+
+	if (CNIC_SUPPORT(sc)) {
+		ecore_init_block(sc, BLOCK_TM, init_phase);
+		REG_WR(sc, TM_REG_LIN0_SCAN_TIME + port * 4, 20);
+		REG_WR(sc, TM_REG_LIN0_MAX_ACTIVE_CID + port * 4, 31);
+	}
+
+	ecore_init_block(sc, BLOCK_DORQ, init_phase);
+
+	ecore_init_block(sc, BLOCK_BRB1, init_phase);
+
+	if (CHIP_IS_E1H(sc)) {
+		if (IS_MF(sc)) {
+			low = (BCM_ONE_PORT(sc) ? 160 : 246);
+		} else if (sc->mtu > 4096) {
+			if (BCM_ONE_PORT(sc)) {
+				low = 160;
+			} else {
+				val = sc->mtu;
+				/* (24*1024 + val*4)/256 */
+				low = (96 + (val / 64) + ((val % 64) ? 1 : 0));
+			}
+		} else {
+			low = (BCM_ONE_PORT(sc) ? 80 : 160);
+		}
+		high = (low + 56);	/* 14*1024/256 */
+		REG_WR(sc, BRB1_REG_PAUSE_LOW_THRESHOLD_0 + port * 4, low);
+		REG_WR(sc, BRB1_REG_PAUSE_HIGH_THRESHOLD_0 + port * 4, high);
+	}
+
+	if (CHIP_IS_MODE_4_PORT(sc)) {
+		REG_WR(sc, SC_PORT(sc) ?
+		       BRB1_REG_MAC_GUARANTIED_1 :
+		       BRB1_REG_MAC_GUARANTIED_0, 40);
+	}
+
+	ecore_init_block(sc, BLOCK_PRS, init_phase);
+	if (CHIP_IS_E3B0(sc)) {
+		if (IS_MF_AFEX(sc)) {
+			/* configure headers for AFEX mode */
+			if (SC_PORT(sc)) {
+				REG_WR(sc, PRS_REG_HDRS_AFTER_BASIC_PORT_1,
+				       0xE);
+				REG_WR(sc, PRS_REG_HDRS_AFTER_TAG_0_PORT_1,
+				       0x6);
+				REG_WR(sc, PRS_REG_MUST_HAVE_HDRS_PORT_1, 0xA);
+			} else {
+				REG_WR(sc, PRS_REG_HDRS_AFTER_BASIC_PORT_0,
+				       0xE);
+				REG_WR(sc, PRS_REG_HDRS_AFTER_TAG_0_PORT_0,
+				       0x6);
+				REG_WR(sc, PRS_REG_MUST_HAVE_HDRS_PORT_0, 0xA);
+			}
+		} else {
+			/* Ovlan exists only if we are in multi-function +
+			 * switch-dependent mode, in switch-independent there
+			 * is no ovlan headers
+			 */
+			REG_WR(sc, SC_PORT(sc) ?
+			       PRS_REG_HDRS_AFTER_BASIC_PORT_1 :
+			       PRS_REG_HDRS_AFTER_BASIC_PORT_0,
+			       (sc->devinfo.mf_info.path_has_ovlan ? 7 : 6));
+		}
+	}
+
+	ecore_init_block(sc, BLOCK_TSDM, init_phase);
+	ecore_init_block(sc, BLOCK_CSDM, init_phase);
+	ecore_init_block(sc, BLOCK_USDM, init_phase);
+	ecore_init_block(sc, BLOCK_XSDM, init_phase);
+
+	ecore_init_block(sc, BLOCK_TSEM, init_phase);
+	ecore_init_block(sc, BLOCK_USEM, init_phase);
+	ecore_init_block(sc, BLOCK_CSEM, init_phase);
+	ecore_init_block(sc, BLOCK_XSEM, init_phase);
+
+	ecore_init_block(sc, BLOCK_UPB, init_phase);
+	ecore_init_block(sc, BLOCK_XPB, init_phase);
+
+	ecore_init_block(sc, BLOCK_PBF, init_phase);
+
+	if (CHIP_IS_E1x(sc)) {
+/* configure PBF to work without PAUSE mtu 9000 */
+		REG_WR(sc, PBF_REG_P0_PAUSE_ENABLE + port * 4, 0);
+
+/* update threshold */
+		REG_WR(sc, PBF_REG_P0_ARB_THRSH + port * 4, (9040 / 16));
+/* update init credit */
+		REG_WR(sc, PBF_REG_P0_INIT_CRD + port * 4,
+		       (9040 / 16) + 553 - 22);
+
+/* probe changes */
+		REG_WR(sc, PBF_REG_INIT_P0 + port * 4, 1);
+		DELAY(50);
+		REG_WR(sc, PBF_REG_INIT_P0 + port * 4, 0);
+	}
+
+	if (CNIC_SUPPORT(sc)) {
+		ecore_init_block(sc, BLOCK_SRC, init_phase);
+	}
+
+	ecore_init_block(sc, BLOCK_CDU, init_phase);
+	ecore_init_block(sc, BLOCK_CFC, init_phase);
+	ecore_init_block(sc, BLOCK_HC, init_phase);
+	ecore_init_block(sc, BLOCK_IGU, init_phase);
+	ecore_init_block(sc, BLOCK_MISC_AEU, init_phase);
+	/* init aeu_mask_attn_func_0/1:
+	 *  - SF mode: bits 3-7 are masked. only bits 0-2 are in use
+	 *  - MF mode: bit 3 is masked. bits 0-2 are in use as in SF
+	 *             bits 4-7 are used for "per vn group attention" */
+	val = IS_MF(sc) ? 0xF7 : 0x7;
+	val |= 0x10;
+	REG_WR(sc, MISC_REG_AEU_MASK_ATTN_FUNC_0 + port * 4, val);
+
+	ecore_init_block(sc, BLOCK_NIG, init_phase);
+
+	if (!CHIP_IS_E1x(sc)) {
+/* Bit-map indicating which L2 hdrs may appear after the
+ * basic Ethernet header
+ */
+		if (IS_MF_AFEX(sc)) {
+			REG_WR(sc, SC_PORT(sc) ?
+			       NIG_REG_P1_HDRS_AFTER_BASIC :
+			       NIG_REG_P0_HDRS_AFTER_BASIC, 0xE);
+		} else {
+			REG_WR(sc, SC_PORT(sc) ?
+			       NIG_REG_P1_HDRS_AFTER_BASIC :
+			       NIG_REG_P0_HDRS_AFTER_BASIC,
+			       IS_MF_SD(sc) ? 7 : 6);
+		}
+
+		if (CHIP_IS_E3(sc)) {
+			REG_WR(sc, SC_PORT(sc) ?
+			       NIG_REG_LLH1_MF_MODE :
+			       NIG_REG_LLH_MF_MODE, IS_MF(sc));
+		}
+	}
+	if (!CHIP_IS_E3(sc)) {
+		REG_WR(sc, NIG_REG_XGXS_SERDES0_MODE_SEL + port * 4, 1);
+	}
+
+	/* 0x2 disable mf_ov, 0x1 enable */
+	REG_WR(sc, NIG_REG_LLH0_BRB1_DRV_MASK_MF + port * 4,
+	       (IS_MF_SD(sc) ? 0x1 : 0x2));
+
+	if (!CHIP_IS_E1x(sc)) {
+		val = 0;
+		switch (sc->devinfo.mf_info.mf_mode) {
+		case MULTI_FUNCTION_SD:
+			val = 1;
+			break;
+		case MULTI_FUNCTION_SI:
+		case MULTI_FUNCTION_AFEX:
+			val = 2;
+			break;
+		}
+
+		REG_WR(sc, (SC_PORT(sc) ? NIG_REG_LLH1_CLS_TYPE :
+			    NIG_REG_LLH0_CLS_TYPE), val);
+	}
+	REG_WR(sc, NIG_REG_LLFC_ENABLE_0 + port * 4, 0);
+	REG_WR(sc, NIG_REG_LLFC_OUT_EN_0 + port * 4, 0);
+	REG_WR(sc, NIG_REG_PAUSE_ENABLE_0 + port * 4, 1);
+
+	/* If SPIO5 is set to generate interrupts, enable it for this port */
+	val = REG_RD(sc, MISC_REG_SPIO_EVENT_EN);
+	if (val & MISC_SPIO_SPIO5) {
+		uint32_t reg_addr = (port ? MISC_REG_AEU_ENABLE1_FUNC_1_OUT_0 :
+				     MISC_REG_AEU_ENABLE1_FUNC_0_OUT_0);
+		val = REG_RD(sc, reg_addr);
+		val |= AEU_INPUTS_ATTN_BITS_SPIO5;
+		REG_WR(sc, reg_addr, val);
+	}
+
+	return 0;
+}
+
+static uint32_t
+bcm_flr_clnup_reg_poll(struct bcm_softc *sc, uint32_t reg,
+		       uint32_t expected, uint32_t poll_count)
+{
+	uint32_t cur_cnt = poll_count;
+	uint32_t val;
+
+	while ((val = REG_RD(sc, reg)) != expected && cur_cnt--) {
+		DELAY(FLR_WAIT_INTERVAL);
+	}
+
+	return val;
+}
+
+static int
+bcm_flr_clnup_poll_hw_counter(struct bcm_softc *sc, uint32_t reg,
+			      __rte_unused const char *msg, uint32_t poll_cnt)
+{
+	uint32_t val = bcm_flr_clnup_reg_poll(sc, reg, 0, poll_cnt);
+
+	if (val != 0) {
+		PMD_DRV_LOG(NOTICE, "%s usage count=%d", msg, val);
+		return -1;
+	}
+
+	return 0;
+}
+
+/* Common routines with VF FLR cleanup */
+static uint32_t bcm_flr_clnup_poll_count(struct bcm_softc *sc)
+{
+	/* adjust polling timeout */
+	if (CHIP_REV_IS_EMUL(sc)) {
+		return (FLR_POLL_CNT * 2000);
+	}
+
+	if (CHIP_REV_IS_FPGA(sc)) {
+		return (FLR_POLL_CNT * 120);
+	}
+
+	return FLR_POLL_CNT;
+}
+
+static int bcm_poll_hw_usage_counters(struct bcm_softc *sc, uint32_t poll_cnt)
+{
+	/* wait for CFC PF usage-counter to zero (includes all the VFs) */
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  CFC_REG_NUM_LCIDS_INSIDE_PF,
+					  "CFC PF usage counter timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	/* Wait for DQ PF usage-counter to zero (until DQ cleanup) */
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  DORQ_REG_PF_USAGE_CNT,
+					  "DQ PF usage counter timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	/* Wait for QM PF usage-counter to zero (until DQ cleanup) */
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  QM_REG_PF_USG_CNT_0 + 4 * SC_FUNC(sc),
+					  "QM PF usage counter timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	/* Wait for Timer PF usage-counters to zero (until DQ cleanup) */
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  TM_REG_LIN0_VNIC_UC + 4 * SC_PORT(sc),
+					  "Timers VNIC usage counter timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  TM_REG_LIN0_NUM_SCANS +
+					  4 * SC_PORT(sc),
+					  "Timers NUM_SCANS usage counter timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	/* Wait DMAE PF usage counter to zero */
+	if (bcm_flr_clnup_poll_hw_counter(sc,
+					  dmae_reg_go_c[INIT_DMAE_C(sc)],
+					  "DMAE dommand register timed out",
+					  poll_cnt)) {
+		return -1;
+	}
+
+	return 0;
+}
+
+#define OP_GEN_PARAM(param)                                            \
+	(((param) << SDM_OP_GEN_COMP_PARAM_SHIFT) & SDM_OP_GEN_COMP_PARAM)
+#define OP_GEN_TYPE(type)                                           \
+	(((type) << SDM_OP_GEN_COMP_TYPE_SHIFT) & SDM_OP_GEN_COMP_TYPE)
+#define OP_GEN_AGG_VECT(index)                                             \
+	(((index) << SDM_OP_GEN_AGG_VECT_IDX_SHIFT) & SDM_OP_GEN_AGG_VECT_IDX)
+
+static int
+bcm_send_final_clnup(struct bcm_softc *sc, uint8_t clnup_func,
+		     uint32_t poll_cnt)
+{
+	uint32_t op_gen_command = 0;
+	uint32_t comp_addr = (BAR_CSTRORM_INTMEM +
+			      CSTORM_FINAL_CLEANUP_COMPLETE_OFFSET(clnup_func));
+	int ret = 0;
+
+	if (REG_RD(sc, comp_addr)) {
+		PMD_DRV_LOG(NOTICE,
+			    "Cleanup complete was not 0 before sending");
+		return -1;
+	}
+
+	op_gen_command |= OP_GEN_PARAM(XSTORM_AGG_INT_FINAL_CLEANUP_INDEX);
+	op_gen_command |= OP_GEN_TYPE(XSTORM_AGG_INT_FINAL_CLEANUP_COMP_TYPE);
+	op_gen_command |= OP_GEN_AGG_VECT(clnup_func);
+	op_gen_command |= 1 << SDM_OP_GEN_AGG_VECT_IDX_VALID_SHIFT;
+
+	REG_WR(sc, XSDM_REG_OPERATION_GEN, op_gen_command);
+
+	if (bcm_flr_clnup_reg_poll(sc, comp_addr, 1, poll_cnt) != 1) {
+		PMD_DRV_LOG(NOTICE, "FW final cleanup did not succeed");
+		PMD_DRV_LOG(DEBUG, "At timeout completion address contained %x",
+			    (REG_RD(sc, comp_addr)));
+		rte_panic("FLR cleanup failed");
+		return -1;
+	}
+
+	/* Zero completion for nxt FLR */
+	REG_WR(sc, comp_addr, 0);
+
+	return ret;
+}
+
+static void
+bcm_pbf_pN_buf_flushed(struct bcm_softc *sc, struct pbf_pN_buf_regs *regs,
+		       uint32_t poll_count)
+{
+	uint32_t init_crd, crd, crd_start, crd_freed, crd_freed_start;
+	uint32_t cur_cnt = poll_count;
+
+	crd_freed = crd_freed_start = REG_RD(sc, regs->crd_freed);
+	crd = crd_start = REG_RD(sc, regs->crd);
+	init_crd = REG_RD(sc, regs->init_crd);
+
+	while ((crd != init_crd) &&
+	       ((uint32_t) ((int32_t) crd_freed - (int32_t) crd_freed_start) <
+		(init_crd - crd_start))) {
+		if (cur_cnt--) {
+			DELAY(FLR_WAIT_INTERVAL);
+			crd = REG_RD(sc, regs->crd);
+			crd_freed = REG_RD(sc, regs->crd_freed);
+		} else {
+			break;
+		}
+	}
+}
+
+static void
+bcm_pbf_pN_cmd_flushed(struct bcm_softc *sc, struct pbf_pN_cmd_regs *regs,
+		       uint32_t poll_count)
+{
+	uint32_t occup, to_free, freed, freed_start;
+	uint32_t cur_cnt = poll_count;
+
+	occup = to_free = REG_RD(sc, regs->lines_occup);
+	freed = freed_start = REG_RD(sc, regs->lines_freed);
+
+	while (occup &&
+	       ((uint32_t) ((int32_t) freed - (int32_t) freed_start) <
+		to_free)) {
+		if (cur_cnt--) {
+			DELAY(FLR_WAIT_INTERVAL);
+			occup = REG_RD(sc, regs->lines_occup);
+			freed = REG_RD(sc, regs->lines_freed);
+		} else {
+			break;
+		}
+	}
+}
+
+static void bcm_tx_hw_flushed(struct bcm_softc *sc, uint32_t poll_count)
+{
+	struct pbf_pN_cmd_regs cmd_regs[] = {
+		{0, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_OCCUPANCY_Q0 : PBF_REG_P0_TQ_OCCUPANCY,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_LINES_FREED_CNT_Q0 : PBF_REG_P0_TQ_LINES_FREED_CNT},
+		{1, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_OCCUPANCY_Q1 : PBF_REG_P1_TQ_OCCUPANCY,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_LINES_FREED_CNT_Q1 : PBF_REG_P1_TQ_LINES_FREED_CNT},
+		{4, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_OCCUPANCY_LB_Q : PBF_REG_P4_TQ_OCCUPANCY,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_TQ_LINES_FREED_CNT_LB_Q :
+		 PBF_REG_P4_TQ_LINES_FREED_CNT}
+	};
+
+	struct pbf_pN_buf_regs buf_regs[] = {
+		{0, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INIT_CRD_Q0 : PBF_REG_P0_INIT_CRD,
+		 (CHIP_IS_E3B0(sc)) ? PBF_REG_CREDIT_Q0 : PBF_REG_P0_CREDIT,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INTERNAL_CRD_FREED_CNT_Q0 :
+		 PBF_REG_P0_INTERNAL_CRD_FREED_CNT},
+		{1, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INIT_CRD_Q1 : PBF_REG_P1_INIT_CRD,
+		 (CHIP_IS_E3B0(sc)) ? PBF_REG_CREDIT_Q1 : PBF_REG_P1_CREDIT,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INTERNAL_CRD_FREED_CNT_Q1 :
+		 PBF_REG_P1_INTERNAL_CRD_FREED_CNT},
+		{4, (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INIT_CRD_LB_Q : PBF_REG_P4_INIT_CRD,
+		 (CHIP_IS_E3B0(sc)) ? PBF_REG_CREDIT_LB_Q : PBF_REG_P4_CREDIT,
+		 (CHIP_IS_E3B0(sc)) ?
+		 PBF_REG_INTERNAL_CRD_FREED_CNT_LB_Q :
+		 PBF_REG_P4_INTERNAL_CRD_FREED_CNT},
+	};
+
+	uint32_t i;
+
+	/* Verify the command queues are flushed P0, P1, P4 */
+	for (i = 0; i < ARRAY_SIZE(cmd_regs); i++) {
+		bcm_pbf_pN_cmd_flushed(sc, &cmd_regs[i], poll_count);
+	}
+
+	/* Verify the transmission buffers are flushed P0, P1, P4 */
+	for (i = 0; i < ARRAY_SIZE(buf_regs); i++) {
+		bcm_pbf_pN_buf_flushed(sc, &buf_regs[i], poll_count);
+	}
+}
+
+static void bcm_hw_enable_status(struct bcm_softc *sc)
+{
+	__rte_unused uint32_t val;
+
+	val = REG_RD(sc, CFC_REG_WEAK_ENABLE_PF);
+	PMD_DRV_LOG(DEBUG, "CFC_REG_WEAK_ENABLE_PF is 0x%x", val);
+
+	val = REG_RD(sc, PBF_REG_DISABLE_PF);
+	PMD_DRV_LOG(DEBUG, "PBF_REG_DISABLE_PF is 0x%x", val);
+
+	val = REG_RD(sc, IGU_REG_PCI_PF_MSI_EN);
+	PMD_DRV_LOG(DEBUG, "IGU_REG_PCI_PF_MSI_EN is 0x%x", val);
+
+	val = REG_RD(sc, IGU_REG_PCI_PF_MSIX_EN);
+	PMD_DRV_LOG(DEBUG, "IGU_REG_PCI_PF_MSIX_EN is 0x%x", val);
+
+	val = REG_RD(sc, IGU_REG_PCI_PF_MSIX_FUNC_MASK);
+	PMD_DRV_LOG(DEBUG, "IGU_REG_PCI_PF_MSIX_FUNC_MASK is 0x%x", val);
+
+	val = REG_RD(sc, PGLUE_B_REG_SHADOW_BME_PF_7_0_CLR);
+	PMD_DRV_LOG(DEBUG, "PGLUE_B_REG_SHADOW_BME_PF_7_0_CLR is 0x%x", val);
+
+	val = REG_RD(sc, PGLUE_B_REG_FLR_REQUEST_PF_7_0_CLR);
+	PMD_DRV_LOG(DEBUG, "PGLUE_B_REG_FLR_REQUEST_PF_7_0_CLR is 0x%x", val);
+
+	val = REG_RD(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER);
+	PMD_DRV_LOG(DEBUG, "PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER is 0x%x",
+		    val);
+}
+
+/**
+ *	bcm_pf_flr_clnup
+ *	a. re-enable target read on the PF
+ *	b. poll cfc per function usgae counter
+ *	c. poll the qm perfunction usage counter
+ *	d. poll the tm per function usage counter
+ *	e. poll the tm per function scan-done indication
+ *	f. clear the dmae channel associated wit hthe PF
+ *	g. zero the igu 'trailing edge' and 'leading edge' regs (attentions)
+ *	h. call the common flr cleanup code with -1 (pf indication)
+ */
+static int bcm_pf_flr_clnup(struct bcm_softc *sc)
+{
+	uint32_t poll_cnt = bcm_flr_clnup_poll_count(sc);
+
+	/* Re-enable PF target read access */
+	REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_TARGET_READ, 1);
+
+	/* Poll HW usage counters */
+	if (bcm_poll_hw_usage_counters(sc, poll_cnt)) {
+		return -1;
+	}
+
+	/* Zero the igu 'trailing edge' and 'leading edge' */
+
+	/* Send the FW cleanup command */
+	if (bcm_send_final_clnup(sc, (uint8_t) SC_FUNC(sc), poll_cnt)) {
+		return -1;
+	}
+
+	/* ATC cleanup */
+
+	/* Verify TX hw is flushed */
+	bcm_tx_hw_flushed(sc, poll_cnt);
+
+	/* Wait 100ms (not adjusted according to platform) */
+	DELAY(100000);
+
+	/* Verify no pending pci transactions */
+	if (bcm_is_pcie_pending(sc)) {
+		PMD_DRV_LOG(NOTICE, "PCIE Transactions still pending");
+	}
+
+	/* Debug */
+	bcm_hw_enable_status(sc);
+
+	/*
+	 * Master enable - Due to WB DMAE writes performed before this
+	 * register is re-initialized as part of the regular function init
+	 */
+	REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER, 1);
+
+	return 0;
+}
+
+static int bcm_init_hw_func(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	int func = SC_FUNC(sc);
+	int init_phase = PHASE_PF0 + func;
+	struct ecore_ilt *ilt = sc->ilt;
+	uint16_t cdu_ilt_start;
+	uint32_t addr, val;
+	uint32_t main_mem_base, main_mem_size, main_mem_prty_clr;
+	int main_mem_width, rc;
+	uint32_t i;
+
+	PMD_DRV_LOG(DEBUG, "starting func init for func %d", func);
+
+	/* FLR cleanup */
+	if (!CHIP_IS_E1x(sc)) {
+		rc = bcm_pf_flr_clnup(sc);
+		if (rc) {
+			PMD_DRV_LOG(NOTICE, "FLR cleanup failed!");
+			return rc;
+		}
+	}
+
+	/* set MSI reconfigure capability */
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		addr = (port ? HC_REG_CONFIG_1 : HC_REG_CONFIG_0);
+		val = REG_RD(sc, addr);
+		val |= HC_CONFIG_0_REG_MSI_ATTN_EN_0;
+		REG_WR(sc, addr, val);
+	}
+
+	ecore_init_block(sc, BLOCK_PXP, init_phase);
+	ecore_init_block(sc, BLOCK_PXP2, init_phase);
+
+	ilt = sc->ilt;
+	cdu_ilt_start = ilt->clients[ILT_CLIENT_CDU].start;
+
+	for (i = 0; i < L2_ILT_LINES(sc); i++) {
+		ilt->lines[cdu_ilt_start + i].page = sc->context[i].vcxt;
+		ilt->lines[cdu_ilt_start + i].page_mapping =
+		    (phys_addr_t) ((void *)sc->context[i].vcxt_dma.paddr);
+		ilt->lines[cdu_ilt_start + i].size = sc->context[i].size;
+	}
+	ecore_ilt_init_op(sc, INITOP_SET);
+
+	REG_WR(sc, PRS_REG_NIC_MODE, 1);
+
+	if (!CHIP_IS_E1x(sc)) {
+		uint32_t pf_conf = IGU_PF_CONF_FUNC_EN;
+
+/* Turn on a single ISR mode in IGU if driver is going to use
+ * INT#x or MSI
+ */
+		if ((sc->interrupt_mode != INTR_MODE_MSIX)
+		    || (sc->interrupt_mode != INTR_MODE_SINGLE_MSIX)) {
+			pf_conf |= IGU_PF_CONF_SINGLE_ISR_EN;
+		}
+
+/*
+ * Timers workaround bug: function init part.
+ * Need to wait 20msec after initializing ILT,
+ * needed to make sure there are no requests in
+ * one of the PXP internal queues with "old" ILT addresses
+ */
+		DELAY(20000);
+
+/*
+ * Master enable - Due to WB DMAE writes performed before this
+ * register is re-initialized as part of the regular function
+ * init
+ */
+		REG_WR(sc, PGLUE_B_REG_INTERNAL_PFID_ENABLE_MASTER, 1);
+/* Enable the function in IGU */
+		REG_WR(sc, IGU_REG_PF_CONFIGURATION, pf_conf);
+	}
+
+	sc->dmae_ready = 1;
+
+	ecore_init_block(sc, BLOCK_PGLUE_B, init_phase);
+
+	if (!CHIP_IS_E1x(sc))
+		REG_WR(sc, PGLUE_B_REG_WAS_ERROR_PF_7_0_CLR, func);
+
+	ecore_init_block(sc, BLOCK_ATC, init_phase);
+	ecore_init_block(sc, BLOCK_DMAE, init_phase);
+	ecore_init_block(sc, BLOCK_NIG, init_phase);
+	ecore_init_block(sc, BLOCK_SRC, init_phase);
+	ecore_init_block(sc, BLOCK_MISC, init_phase);
+	ecore_init_block(sc, BLOCK_TCM, init_phase);
+	ecore_init_block(sc, BLOCK_UCM, init_phase);
+	ecore_init_block(sc, BLOCK_CCM, init_phase);
+	ecore_init_block(sc, BLOCK_XCM, init_phase);
+	ecore_init_block(sc, BLOCK_TSEM, init_phase);
+	ecore_init_block(sc, BLOCK_USEM, init_phase);
+	ecore_init_block(sc, BLOCK_CSEM, init_phase);
+	ecore_init_block(sc, BLOCK_XSEM, init_phase);
+
+	if (!CHIP_IS_E1x(sc))
+		REG_WR(sc, QM_REG_PF_EN, 1);
+
+	if (!CHIP_IS_E1x(sc)) {
+		REG_WR(sc, TSEM_REG_VFPF_ERR_NUM, BCM_MAX_NUM_OF_VFS + func);
+		REG_WR(sc, USEM_REG_VFPF_ERR_NUM, BCM_MAX_NUM_OF_VFS + func);
+		REG_WR(sc, CSEM_REG_VFPF_ERR_NUM, BCM_MAX_NUM_OF_VFS + func);
+		REG_WR(sc, XSEM_REG_VFPF_ERR_NUM, BCM_MAX_NUM_OF_VFS + func);
+	}
+	ecore_init_block(sc, BLOCK_QM, init_phase);
+
+	ecore_init_block(sc, BLOCK_TM, init_phase);
+	ecore_init_block(sc, BLOCK_DORQ, init_phase);
+
+	ecore_init_block(sc, BLOCK_BRB1, init_phase);
+	ecore_init_block(sc, BLOCK_PRS, init_phase);
+	ecore_init_block(sc, BLOCK_TSDM, init_phase);
+	ecore_init_block(sc, BLOCK_CSDM, init_phase);
+	ecore_init_block(sc, BLOCK_USDM, init_phase);
+	ecore_init_block(sc, BLOCK_XSDM, init_phase);
+	ecore_init_block(sc, BLOCK_UPB, init_phase);
+	ecore_init_block(sc, BLOCK_XPB, init_phase);
+	ecore_init_block(sc, BLOCK_PBF, init_phase);
+	if (!CHIP_IS_E1x(sc))
+		REG_WR(sc, PBF_REG_DISABLE_PF, 0);
+
+	ecore_init_block(sc, BLOCK_CDU, init_phase);
+
+	ecore_init_block(sc, BLOCK_CFC, init_phase);
+
+	if (!CHIP_IS_E1x(sc))
+		REG_WR(sc, CFC_REG_WEAK_ENABLE_PF, 1);
+
+	if (IS_MF(sc)) {
+		REG_WR(sc, NIG_REG_LLH0_FUNC_EN + port * 8, 1);
+		REG_WR(sc, NIG_REG_LLH0_FUNC_VLAN_ID + port * 8, OVLAN(sc));
+	}
+
+	ecore_init_block(sc, BLOCK_MISC_AEU, init_phase);
+
+	/* HC init per function */
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		if (CHIP_IS_E1H(sc)) {
+			REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_12 + func * 4, 0);
+
+			REG_WR(sc, HC_REG_LEADING_EDGE_0 + port * 8, 0);
+			REG_WR(sc, HC_REG_TRAILING_EDGE_0 + port * 8, 0);
+		}
+		ecore_init_block(sc, BLOCK_HC, init_phase);
+
+	} else {
+		uint32_t num_segs, sb_idx, prod_offset;
+
+		REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_12 + func * 4, 0);
+
+		if (!CHIP_IS_E1x(sc)) {
+			REG_WR(sc, IGU_REG_LEADING_EDGE_LATCH, 0);
+			REG_WR(sc, IGU_REG_TRAILING_EDGE_LATCH, 0);
+		}
+
+		ecore_init_block(sc, BLOCK_IGU, init_phase);
+
+		if (!CHIP_IS_E1x(sc)) {
+			int dsb_idx = 0;
+	/**
+	 * Producer memory:
+	 * E2 mode: address 0-135 match to the mapping memory;
+	 * 136 - PF0 default prod; 137 - PF1 default prod;
+	 * 138 - PF2 default prod; 139 - PF3 default prod;
+	 * 140 - PF0 attn prod;    141 - PF1 attn prod;
+	 * 142 - PF2 attn prod;    143 - PF3 attn prod;
+	 * 144-147 reserved.
+	 *
+	 * E1.5 mode - In backward compatible mode;
+	 * for non default SB; each even line in the memory
+	 * holds the U producer and each odd line hold
+	 * the C producer. The first 128 producers are for
+	 * NDSB (PF0 - 0-31; PF1 - 32-63 and so on). The last 20
+	 * producers are for the DSB for each PF.
+	 * Each PF has five segments: (the order inside each
+	 * segment is PF0; PF1; PF2; PF3) - 128-131 U prods;
+	 * 132-135 C prods; 136-139 X prods; 140-143 T prods;
+	 * 144-147 attn prods;
+	 */
+			/* non-default-status-blocks */
+			num_segs = CHIP_INT_MODE_IS_BC(sc) ?
+			    IGU_BC_NDSB_NUM_SEGS : IGU_NORM_NDSB_NUM_SEGS;
+			for (sb_idx = 0; sb_idx < sc->igu_sb_cnt; sb_idx++) {
+				prod_offset = (sc->igu_base_sb + sb_idx) *
+				    num_segs;
+
+				for (i = 0; i < num_segs; i++) {
+					addr = IGU_REG_PROD_CONS_MEMORY +
+					    (prod_offset + i) * 4;
+					REG_WR(sc, addr, 0);
+				}
+				/* send consumer update with value 0 */
+				bcm_ack_sb(sc, sc->igu_base_sb + sb_idx,
+					   USTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_igu_clear_sb(sc, sc->igu_base_sb + sb_idx);
+			}
+
+			/* default-status-blocks */
+			num_segs = CHIP_INT_MODE_IS_BC(sc) ?
+			    IGU_BC_DSB_NUM_SEGS : IGU_NORM_DSB_NUM_SEGS;
+
+			if (CHIP_IS_MODE_4_PORT(sc))
+				dsb_idx = SC_FUNC(sc);
+			else
+				dsb_idx = SC_VN(sc);
+
+			prod_offset = (CHIP_INT_MODE_IS_BC(sc) ?
+				       IGU_BC_BASE_DSB_PROD + dsb_idx :
+				       IGU_NORM_BASE_DSB_PROD + dsb_idx);
+
+			/*
+			 * igu prods come in chunks of E1HVN_MAX (4) -
+			 * does not matters what is the current chip mode
+			 */
+			for (i = 0; i < (num_segs * E1HVN_MAX); i += E1HVN_MAX) {
+				addr = IGU_REG_PROD_CONS_MEMORY +
+				    (prod_offset + i) * 4;
+				REG_WR(sc, addr, 0);
+			}
+			/* send consumer update with 0 */
+			if (CHIP_INT_MODE_IS_BC(sc)) {
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   USTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   CSTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   XSTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   TSTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   ATTENTION_ID, 0, IGU_INT_NOP, 1);
+			} else {
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   USTORM_ID, 0, IGU_INT_NOP, 1);
+				bcm_ack_sb(sc, sc->igu_dsb_id,
+					   ATTENTION_ID, 0, IGU_INT_NOP, 1);
+			}
+			bcm_igu_clear_sb(sc, sc->igu_dsb_id);
+
+			/* !!! these should become driver const once
+			   rf-tool supports split-68 const */
+			REG_WR(sc, IGU_REG_SB_INT_BEFORE_MASK_LSB, 0);
+			REG_WR(sc, IGU_REG_SB_INT_BEFORE_MASK_MSB, 0);
+			REG_WR(sc, IGU_REG_SB_MASK_LSB, 0);
+			REG_WR(sc, IGU_REG_SB_MASK_MSB, 0);
+			REG_WR(sc, IGU_REG_PBA_STATUS_LSB, 0);
+			REG_WR(sc, IGU_REG_PBA_STATUS_MSB, 0);
+		}
+	}
+
+	/* Reset PCIE errors for debug */
+	REG_WR(sc, 0x2114, 0xffffffff);
+	REG_WR(sc, 0x2120, 0xffffffff);
+
+	if (CHIP_IS_E1x(sc)) {
+		main_mem_size = HC_REG_MAIN_MEMORY_SIZE / 2;	/*dwords */
+		main_mem_base = HC_REG_MAIN_MEMORY +
+		    SC_PORT(sc) * (main_mem_size * 4);
+		main_mem_prty_clr = HC_REG_HC_PRTY_STS_CLR;
+		main_mem_width = 8;
+
+		val = REG_RD(sc, main_mem_prty_clr);
+		if (val) {
+			PMD_DRV_LOG(DEBUG,
+				    "Parity errors in HC block during function init (0x%x)!",
+				    val);
+		}
+
+/* Clear "false" parity errors in MSI-X table */
+		for (i = main_mem_base;
+		     i < main_mem_base + main_mem_size * 4;
+		     i += main_mem_width) {
+			bcm_read_dmae(sc, i, main_mem_width / 4);
+			bcm_write_dmae(sc, BCM_SP_MAPPING(sc, wb_data),
+				       i, main_mem_width / 4);
+		}
+/* Clear HC parity attention */
+		REG_RD(sc, main_mem_prty_clr);
+	}
+
+	/* Enable STORMs SP logging */
+	REG_WR8(sc, BAR_USTRORM_INTMEM +
+		USTORM_RECORD_SLOW_PATH_OFFSET(SC_FUNC(sc)), 1);
+	REG_WR8(sc, BAR_TSTRORM_INTMEM +
+		TSTORM_RECORD_SLOW_PATH_OFFSET(SC_FUNC(sc)), 1);
+	REG_WR8(sc, BAR_CSTRORM_INTMEM +
+		CSTORM_RECORD_SLOW_PATH_OFFSET(SC_FUNC(sc)), 1);
+	REG_WR8(sc, BAR_XSTRORM_INTMEM +
+		XSTORM_RECORD_SLOW_PATH_OFFSET(SC_FUNC(sc)), 1);
+
+	elink_phy_probe(&sc->link_params);
+
+	return 0;
+}
+
+static void bcm_link_reset(struct bcm_softc *sc)
+{
+	if (!BCM_NOMCP(sc)) {
+		elink_lfa_reset(&sc->link_params, &sc->link_vars);
+	} else {
+		if (!CHIP_REV_IS_SLOW(sc)) {
+			PMD_DRV_LOG(WARN,
+				    "Bootcode is missing - cannot reset link");
+		}
+	}
+}
+
+static void bcm_reset_port(struct bcm_softc *sc)
+{
+	int port = SC_PORT(sc);
+	uint32_t val;
+
+	/* reset physical Link */
+	bcm_link_reset(sc);
+
+	REG_WR(sc, NIG_REG_MASK_INTERRUPT_PORT0 + port * 4, 0);
+
+	/* Do not rcv packets to BRB */
+	REG_WR(sc, NIG_REG_LLH0_BRB1_DRV_MASK + port * 4, 0x0);
+	/* Do not direct rcv packets that are not for MCP to the BRB */
+	REG_WR(sc, (port ? NIG_REG_LLH1_BRB1_NOT_MCP :
+		    NIG_REG_LLH0_BRB1_NOT_MCP), 0x0);
+
+	/* Configure AEU */
+	REG_WR(sc, MISC_REG_AEU_MASK_ATTN_FUNC_0 + port * 4, 0);
+
+	DELAY(100000);
+
+	/* Check for BRB port occupancy */
+	val = REG_RD(sc, BRB1_REG_PORT_NUM_OCC_BLOCKS_0 + port * 4);
+	if (val) {
+		PMD_DRV_LOG(DEBUG,
+			    "BRB1 is not empty, %d blocks are occupied", val);
+	}
+}
+
+static void bcm_ilt_wr(struct bcm_softc *sc, uint32_t index, phys_addr_t addr)
+{
+	int reg;
+	uint32_t wb_write[2];
+
+	reg = PXP2_REG_RQ_ONCHIP_AT_B0 + index * 8;
+
+	wb_write[0] = ONCHIP_ADDR1(addr);
+	wb_write[1] = ONCHIP_ADDR2(addr);
+	REG_WR_DMAE(sc, reg, wb_write, 2);
+}
+
+static void bcm_clear_func_ilt(struct bcm_softc *sc, uint32_t func)
+{
+	uint32_t i, base = FUNC_ILT_BASE(func);
+	for (i = base; i < base + ILT_PER_FUNC; i++) {
+		bcm_ilt_wr(sc, i, 0);
+	}
+}
+
+static void bcm_reset_func(struct bcm_softc *sc)
+{
+	struct bcm_fastpath *fp;
+	int port = SC_PORT(sc);
+	int func = SC_FUNC(sc);
+	int i;
+
+	/* Disable the function in the FW */
+	REG_WR8(sc, BAR_XSTRORM_INTMEM + XSTORM_FUNC_EN_OFFSET(func), 0);
+	REG_WR8(sc, BAR_CSTRORM_INTMEM + CSTORM_FUNC_EN_OFFSET(func), 0);
+	REG_WR8(sc, BAR_TSTRORM_INTMEM + TSTORM_FUNC_EN_OFFSET(func), 0);
+	REG_WR8(sc, BAR_USTRORM_INTMEM + USTORM_FUNC_EN_OFFSET(func), 0);
+
+	/* FP SBs */
+	FOR_EACH_ETH_QUEUE(sc, i) {
+		fp = &sc->fp[i];
+		REG_WR8(sc, BAR_CSTRORM_INTMEM +
+			CSTORM_STATUS_BLOCK_DATA_STATE_OFFSET(fp->fw_sb_id),
+			SB_DISABLED);
+	}
+
+	/* SP SB */
+	REG_WR8(sc, BAR_CSTRORM_INTMEM +
+		CSTORM_SP_STATUS_BLOCK_DATA_STATE_OFFSET(func), SB_DISABLED);
+
+	for (i = 0; i < XSTORM_SPQ_DATA_SIZE / 4; i++) {
+		REG_WR(sc, BAR_XSTRORM_INTMEM + XSTORM_SPQ_DATA_OFFSET(func),
+		       0);
+	}
+
+	/* Configure IGU */
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		REG_WR(sc, HC_REG_LEADING_EDGE_0 + port * 8, 0);
+		REG_WR(sc, HC_REG_TRAILING_EDGE_0 + port * 8, 0);
+	} else {
+		REG_WR(sc, IGU_REG_LEADING_EDGE_LATCH, 0);
+		REG_WR(sc, IGU_REG_TRAILING_EDGE_LATCH, 0);
+	}
+
+	if (CNIC_LOADED(sc)) {
+/* Disable Timer scan */
+		REG_WR(sc, TM_REG_EN_LINEAR0_TIMER + port * 4, 0);
+/*
+ * Wait for at least 10ms and up to 2 second for the timers
+ * scan to complete
+ */
+		for (i = 0; i < 200; i++) {
+			DELAY(10000);
+			if (!REG_RD(sc, TM_REG_LIN0_SCAN_ON + port * 4))
+				break;
+		}
+	}
+
+	/* Clear ILT */
+	bcm_clear_func_ilt(sc, func);
+
+	/*
+	 * Timers workaround bug for E2: if this is vnic-3,
+	 * we need to set the entire ilt range for this timers.
+	 */
+	if (!CHIP_IS_E1x(sc) && SC_VN(sc) == 3) {
+		struct ilt_client_info ilt_cli;
+/* use dummy TM client */
+		memset(&ilt_cli, 0, sizeof(struct ilt_client_info));
+		ilt_cli.start = 0;
+		ilt_cli.end = ILT_NUM_PAGE_ENTRIES - 1;
+		ilt_cli.client_num = ILT_CLIENT_TM;
+
+		ecore_ilt_boundry_init_op(sc, &ilt_cli, 0);
+	}
+
+	/* this assumes that reset_port() called before reset_func() */
+	if (!CHIP_IS_E1x(sc)) {
+		bcm_pf_disable(sc);
+	}
+
+	sc->dmae_ready = 0;
+}
+
+static void bcm_release_firmware(struct bcm_softc *sc)
+{
+	rte_free(sc->init_ops);
+	rte_free(sc->init_ops_offsets);
+	rte_free(sc->init_data);
+	rte_free(sc->iro_array);
+}
+
+static int bcm_init_firmware(struct bcm_softc *sc)
+{
+	uint32_t len, i;
+	uint8_t *p = sc->firmware;
+	uint32_t off[24];
+
+	for (i = 0; i < 24; ++i)
+		off[i] = rte_be_to_cpu_32(*((uint32_t *) sc->firmware + i));
+
+	len = off[0];
+	sc->init_ops = rte_zmalloc("", len, RTE_CACHE_LINE_SIZE);
+	if (!sc->init_ops)
+		goto alloc_failed;
+	bcm_data_to_init_ops(p + off[1], sc->init_ops, len);
+
+	len = off[2];
+	sc->init_ops_offsets = rte_zmalloc("", len, RTE_CACHE_LINE_SIZE);
+	if (!sc->init_ops_offsets)
+		goto alloc_failed;
+	bcm_data_to_init_offsets(p + off[3], sc->init_ops_offsets, len);
+
+	len = off[4];
+	sc->init_data = rte_zmalloc("", len, RTE_CACHE_LINE_SIZE);
+	if (!sc->init_data)
+		goto alloc_failed;
+	bcm_data_to_init_data(p + off[5], sc->init_data, len);
+
+	sc->tsem_int_table_data = p + off[7];
+	sc->tsem_pram_data = p + off[9];
+	sc->usem_int_table_data = p + off[11];
+	sc->usem_pram_data = p + off[13];
+	sc->csem_int_table_data = p + off[15];
+	sc->csem_pram_data = p + off[17];
+	sc->xsem_int_table_data = p + off[19];
+	sc->xsem_pram_data = p + off[21];
+
+	len = off[22];
+	sc->iro_array = rte_zmalloc("", len, RTE_CACHE_LINE_SIZE);
+	if (!sc->iro_array)
+		goto alloc_failed;
+	bcm_data_to_iro_array(p + off[23], sc->iro_array, len);
+
+	return 0;
+
+alloc_failed:
+	bcm_release_firmware(sc);
+	return -1;
+}
+
+static int cut_gzip_prefix(const uint8_t * zbuf, int len)
+{
+#define MIN_PREFIX_SIZE (10)
+
+	int n = MIN_PREFIX_SIZE;
+	uint16_t xlen;
+
+	if (!(zbuf[0] == 0x1f && zbuf[1] == 0x8b && zbuf[2] == Z_DEFLATED) ||
+	    len <= MIN_PREFIX_SIZE) {
+		return -1;
+	}
+
+	/* optional extra fields are present */
+	if (zbuf[3] & 0x4) {
+		xlen = zbuf[13];
+		xlen <<= 8;
+		xlen += zbuf[12];
+
+		n += xlen;
+	}
+	/* file name is present */
+	if (zbuf[3] & 0x8) {
+		while ((zbuf[n++] != 0) && (n < len)) ;
+	}
+
+	return n;
+}
+
+static int ecore_gunzip(struct bcm_softc *sc, const uint8_t * zbuf, int len)
+{
+	int ret;
+	int data_begin = cut_gzip_prefix(zbuf, len);
+
+	PMD_DRV_LOG(DEBUG, "ecore_gunzip %d", len);
+
+	if (data_begin <= 0) {
+		PMD_DRV_LOG(NOTICE, "bad gzip prefix");
+		return -1;
+	}
+
+	memset(&zlib_stream, 0, sizeof(zlib_stream));
+	zlib_stream.next_in = zbuf + data_begin;
+	zlib_stream.avail_in = len - data_begin;
+	zlib_stream.next_out = sc->gz_buf;
+	zlib_stream.avail_out = FW_BUF_SIZE;
+
+	ret = inflateInit2(&zlib_stream, -MAX_WBITS);
+	if (ret != Z_OK) {
+		PMD_DRV_LOG(NOTICE, "zlib inflateInit2 error");
+		return ret;
+	}
+
+	ret = inflate(&zlib_stream, Z_FINISH);
+	if ((ret != Z_STREAM_END) && (ret != Z_OK)) {
+		PMD_DRV_LOG(NOTICE, "zlib inflate error: %d %s", ret,
+			    zlib_stream.msg);
+	}
+
+	sc->gz_outlen = zlib_stream.total_out;
+	if (sc->gz_outlen & 0x3) {
+		PMD_DRV_LOG(NOTICE, "firmware is not aligned. gz_outlen == %d",
+			    sc->gz_outlen);
+	}
+	sc->gz_outlen >>= 2;
+
+	inflateEnd(&zlib_stream);
+
+	if (ret == Z_STREAM_END)
+		return 0;
+
+	return ret;
+}
+
+static void
+ecore_write_dmae_phys_len(struct bcm_softc *sc, phys_addr_t phys_addr,
+			  uint32_t addr, uint32_t len)
+{
+	bcm_write_dmae_phys_len(sc, phys_addr, addr, len);
+}
+
+void
+ecore_storm_memset_struct(struct bcm_softc *sc, uint32_t addr, size_t size,
+			  uint32_t * data)
+{
+	uint8_t i;
+	for (i = 0; i < size / 4; i++) {
+		REG_WR(sc, addr + (i * 4), data[i]);
+	}
+}
+
+static const char *get_ext_phy_type(uint32_t ext_phy_type)
+{
+	uint32_t phy_type_idx = ext_phy_type >> 8;
+	static const char *types[] =
+	    { "DIRECT", "BCM-8071", "BCM-8072", "BCM-8073",
+		"BCM-8705", "BCM-8706", "BCM-8726", "BCM-8481", "SFX-7101",
+		"BCM-8727",
+		"BCM-8727-NOC", "BCM-84823", "NOT_CONN", "FAILURE"
+	};
+
+	if (phy_type_idx < 12)
+		return types[phy_type_idx];
+	else if (PORT_HW_CFG_XGXS_EXT_PHY_TYPE_NOT_CONN == ext_phy_type)
+		return types[12];
+	else
+		return types[13];
+}
+
+static const char *get_state(uint32_t state)
+{
+	uint32_t state_idx = state >> 12;
+	static const char *states[] = { "CLOSED", "OPENING_WAIT4_LOAD",
+		"OPENING_WAIT4_PORT", "OPEN", "CLOSING_WAIT4_HALT",
+		"CLOSING_WAIT4_DELETE", "CLOSING_WAIT4_UNLOAD",
+		"UNKNOWN", "UNKNOWN", "UNKNOWN", "UNKNOWN", "UNKNOWN",
+		"UNKNOWN", "DISABLED", "DIAG", "ERROR", "UNDEFINED"
+	};
+
+	if (state_idx <= 0xF)
+		return states[state_idx];
+	else
+		return states[0x10];
+}
+
+static const char *get_recovery_state(uint32_t state)
+{
+	static const char *states[] = { "NONE", "DONE", "INIT",
+		"WAIT", "FAILED", "NIC_LOADING"
+	};
+	return states[state];
+}
+
+static const char *get_rx_mode(uint32_t mode)
+{
+	static const char *modes[] = { "NONE", "NORMAL", "ALLMULTI",
+		"PROMISC", "MAX_MULTICAST", "ERROR"
+	};
+
+	if (mode < 0x4)
+		return modes[mode];
+	else if (BCM_MAX_MULTICAST == mode)
+		return modes[4];
+	else
+		return modes[5];
+}
+
+#define BCM_INFO_STR_MAX 256
+static const char *get_bcm_flags(uint32_t flags)
+{
+	int i;
+	static const char *flag[] = { "ONE_PORT ", "NO_ISCSI ",
+		"NO_FCOE ", "NO_WOL ", "USING_DAC ", "USING_MSIX ",
+		"USING_MSI ", "DISABLE_MSI ", "UNKNOWN ", "NO_MCP ",
+		"SAFC_TX_FLAG ", "MF_FUNC_DIS ", "TX_SWITCHING "
+	};
+	static char flag_str[BCM_INFO_STR_MAX];
+	memset(flag_str, 0, BCM_INFO_STR_MAX);
+
+	for (i = 0; i < 5; i++)
+		if (flags & (1 << i)) {
+			strcat(flag_str, flag[i]);
+			flags ^= (1 << i);
+		}
+	if (flags) {
+		static char unknown[BCM_INFO_STR_MAX];
+		snprintf(unknown, 32, "Unknown flag mask %x", flags);
+		strcat(flag_str, unknown);
+	}
+	return flag_str;
+}
+
+/*
+ * Prints useful adapter info.
+ */
+void bcm_print_adapter_info(struct bcm_softc *sc)
+{
+	int i = 0;
+	__rte_unused uint32_t ext_phy_type;
+
+	PMD_INIT_FUNC_TRACE();
+	if (sc->link_vars.phy_flags & PHY_XGXS_FLAG)
+		ext_phy_type = ELINK_XGXS_EXT_PHY_TYPE(REG_RD(sc,
+							      sc->
+							      devinfo.shmem_base
+							      + offsetof(struct
+									 shmem_region,
+									 dev_info.port_hw_config
+									 [0].external_phy_config)));
+	else
+		ext_phy_type = ELINK_SERDES_EXT_PHY_TYPE(REG_RD(sc,
+								sc->
+								devinfo.shmem_base
+								+
+								offsetof(struct
+									 shmem_region,
+									 dev_info.port_hw_config
+									 [0].external_phy_config)));
+
+	PMD_INIT_LOG(DEBUG, "\n\n===================================\n");
+	/* Hardware chip info. */
+	PMD_INIT_LOG(DEBUG, "%10s : %#08x\n", "ASIC", sc->devinfo.chip_id);
+	PMD_INIT_LOG(DEBUG, "%10s : %c%d\n", "Rev", (CHIP_REV(sc) >> 12) + 'A',
+		     (CHIP_METAL(sc) >> 4));
+
+	/* Bus info. */
+	PMD_INIT_LOG(DEBUG, "%10s : %d, ", "Bus PCIe", sc->devinfo.pcie_link_width);
+	switch (sc->devinfo.pcie_link_speed) {
+	case 1:
+		PMD_INIT_LOG(DEBUG, "2.5 Gbps\n");
+		break;
+	case 2:
+		PMD_INIT_LOG(DEBUG, "5 Gbps\n");
+		break;
+	case 4:
+		PMD_INIT_LOG(DEBUG, "8 Gbps\n");
+		break;
+	default:
+		PMD_INIT_LOG(DEBUG, "Unknown link speed\n");
+	}
+
+	/* Device features. */
+	PMD_INIT_LOG(DEBUG, "%10s : ", "Flags");
+
+	/* Miscellaneous flags. */
+	if (sc->devinfo.pcie_cap_flags & BCM_MSI_CAPABLE_FLAG) {
+		PMD_INIT_LOG(DEBUG, "MSI");
+		i++;
+	}
+
+	if (sc->devinfo.pcie_cap_flags & BCM_MSIX_CAPABLE_FLAG) {
+		if (i > 0)
+			PMD_INIT_LOG(DEBUG, "|");
+		PMD_INIT_LOG(DEBUG, "MSI-X");
+		i++;
+	}
+
+	PMD_INIT_LOG(DEBUG, "\n");
+
+	if (IS_PF(sc)) {
+		PMD_INIT_LOG(DEBUG, "\n%10s : ", "Queues");
+		switch (sc->sp->rss_rdata.rss_mode) {
+		case ETH_RSS_MODE_DISABLED:
+			PMD_INIT_LOG(DEBUG, "None\n");
+			break;
+		case ETH_RSS_MODE_REGULAR:
+			PMD_INIT_LOG(DEBUG, "RSS : %d\n", sc->num_queues);
+			break;
+		default:
+			PMD_INIT_LOG(DEBUG, "Unknown\n");
+			break;
+		}
+	}
+
+	/* Firmware versions and device features. */
+	PMD_INIT_LOG(DEBUG, "%10s : %d.%d.%d\n%10s : %s\n",
+		     "Firmware",
+		     BCM_5710_FW_MAJOR_VERSION,
+		     BCM_5710_FW_MINOR_VERSION,
+		     BCM_5710_FW_REVISION_VERSION,
+		     "Bootcode", sc->devinfo.bc_ver_str);
+
+	PMD_INIT_LOG(DEBUG, "===================================\n");
+	PMD_INIT_LOG(DEBUG, "%10s : %u\n", "Bcm Func", sc->pcie_func);
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "Bcm Flags", get_bcm_flags(sc->flags));
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "DMAE Is",
+		     (sc->dmae_ready ? "Ready" : "Not Ready"));
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "OVLAN", (OVLAN(sc) ? "YES" : "NO"));
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "MF", (IS_MF(sc) ? "YES" : "NO"));
+	PMD_INIT_LOG(DEBUG, "%10s : %u\n", "MTU", sc->mtu);
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "PHY Type", get_ext_phy_type(ext_phy_type));
+	PMD_INIT_LOG(DEBUG, "%10s : ", "MAC Addr");
+	for (i = 0; i < 6; i++)
+		PMD_INIT_LOG(DEBUG, "%x%s", sc->link_params.mac_addr[i],
+			     i < 5 ? ":" : "\n");
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "RX Mode", get_rx_mode(sc->rx_mode));
+	PMD_INIT_LOG(DEBUG, "%10s : %s\n", "State", get_state(sc->state));
+	if (sc->recovery_state)
+		PMD_INIT_LOG(DEBUG, "%10s : %s\n", "Recovery",
+			     get_recovery_state(sc->recovery_state));
+	PMD_INIT_LOG(DEBUG, "%10s : CQ = %lx,  EQ = %lx\n", "SPQ Left",
+		     sc->cq_spq_left, sc->eq_spq_left);
+	PMD_INIT_LOG(DEBUG, "%10s : %x\n", "Switch", sc->link_params.switch_cfg);
+	PMD_INIT_LOG(DEBUG, "===================================\n\n");
+}
diff --git a/lib/librte_pmd_bcm/bcm.h b/lib/librte_pmd_bcm/bcm.h
new file mode 100644
index 0000000..4e07dc9
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm.h
@@ -0,0 +1,1998 @@ 
+/*-
+ * Copyright (c) 2007-2013 Broadcom Corporation. All rights reserved.
+ *
+ * Eric Davis        <edavis@broadcom.com>
+ * David Christensen <davidch@broadcom.com>
+ * Gary Zambrano     <zambrano@broadcom.com>
+ *
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ * All rights reserved.
+ *
+ * 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 Broadcom Corporation nor the name of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written consent.
+ *
+ * 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 OWNER 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.
+ */
+
+#ifndef __BCM_H__
+#define __BCM_H__
+
+#include "bcm_ethdev.h"
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#ifndef LITTLE_ENDIAN
+#define LITTLE_ENDIAN
+#endif
+#ifndef __LITTLE_ENDIAN
+#define __LITTLE_ENDIAN
+#endif
+#undef BIG_ENDIAN
+#undef __BIG_ENDIAN
+#else /* _BIG_ENDIAN */
+#ifndef BIG_ENDIAN
+#define BIG_ENDIAN
+#endif
+#ifndef __BIG_ENDIAN
+#define __BIG_ENDIAN
+#endif
+#undef LITTLE_ENDIAN
+#undef __LITTLE_ENDIAN
+#endif
+
+#include "ecore_mfw_req.h"
+#include "ecore_fw_defs.h"
+#include "ecore_hsi.h"
+#include "ecore_reg.h"
+#include "bcm_stats.h"
+#include "bcm_vfpf.h"
+
+#include "elink.h"
+
+#include <linux/pci_regs.h>
+
+#define PCIY_PMG                       PCI_CAP_ID_PM
+#define PCIY_MSI                       PCI_CAP_ID_MSI
+#define PCIY_EXPRESS                   PCI_CAP_ID_EXP
+#define PCIY_MSIX                      PCI_CAP_ID_MSIX
+#define PCIR_EXPRESS_DEVICE_STA        PCI_EXP_TYPE_RC_EC
+#define PCIM_EXP_STA_TRANSACTION_PND   PCI_EXP_DEVSTA_TRPND
+#define PCIR_EXPRESS_LINK_STA          PCI_EXP_LNKSTA
+#define PCIM_LINK_STA_WIDTH            PCI_EXP_LNKSTA_NLW
+#define PCIM_LINK_STA_SPEED            PCI_EXP_LNKSTA_CLS
+#define PCIR_EXPRESS_DEVICE_CTL        PCI_EXP_DEVCTL
+#define PCIM_EXP_CTL_MAX_PAYLOAD       PCI_EXP_DEVCTL_PAYLOAD
+#define PCIM_EXP_CTL_MAX_READ_REQUEST  PCI_EXP_DEVCTL_READRQ
+#define PCIR_POWER_STATUS              PCI_PM_CTRL
+#define PCIM_PSTAT_DMASK               PCI_PM_CTRL_STATE_MASK
+#define PCIM_PSTAT_PME                 PCI_PM_CTRL_PME_STATUS
+#define PCIM_PSTAT_D3                  0x3
+#define PCIM_PSTAT_PMEENABLE           PCI_PM_CTRL_PME_ENABLE
+#define PCIR_MSIX_CTRL                 PCI_MSIX_FLAGS
+#define PCIM_MSIXCTRL_TABLE_SIZE       PCI_MSIX_FLAGS_QSIZE
+
+#define IFM_10G_CX4                    20 /* 10GBase CX4 copper */
+#define IFM_10G_TWINAX                 22 /* 10GBase Twinax copper */
+#define IFM_10G_T                      26 /* 10GBase-T - RJ45 */
+
+#define PCIR_EXPRESS_DEVICE_STA        PCI_EXP_TYPE_RC_EC
+#define PCIM_EXP_STA_TRANSACTION_PND   PCI_EXP_DEVSTA_TRPND
+#define PCIR_EXPRESS_LINK_STA          PCI_EXP_LNKSTA
+#define PCIM_LINK_STA_WIDTH            PCI_EXP_LNKSTA_NLW
+#define PCIM_LINK_STA_SPEED            PCI_EXP_LNKSTA_CLS
+#define PCIR_EXPRESS_DEVICE_CTL        PCI_EXP_DEVCTL
+#define PCIM_EXP_CTL_MAX_PAYLOAD       PCI_EXP_DEVCTL_PAYLOAD
+#define PCIM_EXP_CTL_MAX_READ_REQUEST  PCI_EXP_DEVCTL_READRQ
+
+#ifndef ARRAY_SIZE
+#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]))
+#endif
+#ifndef ARRSIZE
+#define ARRSIZE(arr) (sizeof(arr) / sizeof((arr)[0]))
+#endif
+#ifndef DIV_ROUND_UP
+#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#endif
+#ifndef roundup
+#define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
+#endif
+#ifndef ilog2
+static inline
+int bcm_ilog2(int x)
+{
+	int log = 0;
+	x >>= 1;
+
+	while(x) {
+		log++;
+		x >>= 1;
+	}
+	return log;
+}
+#define ilog2(x) bcm_ilog2(x)
+#endif
+
+#include "ecore_sp.h"
+
+struct bcm_device_type
+{
+    uint16_t bcm_vid;
+    uint16_t bcm_did;
+    uint16_t bcm_svid;
+    uint16_t bcm_sdid;
+    char     *bcm_name;
+};
+
+#define RTE_MBUF_DATA_DMA_ADDR(mb) \
+	(uint64_t) ((mb)->buf_physaddr + (mb)->data_off)
+
+#define BCM_PAGE_SHIFT       12
+#define BCM_PAGE_SIZE        (1 << BCM_PAGE_SHIFT)
+#define BCM_PAGE_MASK        (~(BCM_PAGE_SIZE - 1))
+#define BCM_PAGE_ALIGN(addr) ((addr + BCM_PAGE_SIZE - 1) & BCM_PAGE_MASK)
+
+#if BCM_PAGE_SIZE != 4096
+#error Page sizes other than 4KB are unsupported!
+#endif
+
+#define U64_LO(addr) ((uint32_t)(((uint64_t)(addr)) & 0xFFFFFFFF))
+#define U64_HI(addr) ((uint32_t)(((uint64_t)(addr)) >> 32))
+#define HILO_U64(hi, lo) ((((uint64_t)(hi)) << 32) + (lo))
+
+/* dropless fc FW/HW related params */
+#define BRB_SIZE(sc)         (CHIP_IS_E3(sc) ? 1024 : 512)
+#define MAX_AGG_QS(sc)       ETH_MAX_AGGREGATION_QUEUES_E1H_E2
+#define FW_DROP_LEVEL(sc)    (3 + MAX_SPQ_PENDING + MAX_AGG_QS(sc))
+#define FW_PREFETCH_CNT      16U
+#define DROPLESS_FC_HEADROOM 100
+
+#ifndef MCLSHIFT
+#define MCLSHIFT                              11
+#endif
+#define MCLBYTES                              (1 << MCLSHIFT)
+
+#if BCM_PAGE_SIZE < 2048
+#define MJUMPAGESIZE    MCLBYTES
+#elif BCM_PAGE_SIZE <= 8192
+#define MJUMPAGESIZE    BCM_PAGE_SIZE
+#else
+#define MJUMPAGESIZE    (8 * 1024)
+#endif
+#define MJUM9BYTES      (9 * 1024)
+#define MJUM16BYTES     (16 * 1024)
+
+/*
+ * Transmit Buffer Descriptor (tx_bd) definitions*
+ */
+/* NUM_TX_PAGES must be a power of 2. */
+#define TOTAL_TX_BD_PER_PAGE     (BCM_PAGE_SIZE / sizeof(union eth_tx_bd_types)) /*  256 */
+#define USABLE_TX_BD_PER_PAGE    (TOTAL_TX_BD_PER_PAGE - 1)                      /*  255 */
+
+#define TOTAL_TX_BD(q)           (TOTAL_TX_BD_PER_PAGE * q->nb_tx_pages)         /*  512 */
+#define USABLE_TX_BD(q)          (USABLE_TX_BD_PER_PAGE * q->nb_tx_pages)        /*  510 */
+#define MAX_TX_BD(q)             (TOTAL_TX_BD(q) - 1)                            /*  511 */
+
+#define NEXT_TX_BD(x)                                                   \
+	((((x) & USABLE_TX_BD_PER_PAGE) ==                              \
+	  (USABLE_TX_BD_PER_PAGE - 1)) ? (x) + 2 : (x) + 1)
+
+#define TX_BD(x, q)             ((x) & MAX_TX_BD(q))
+#define TX_PAGE(x)              (((x) & ~USABLE_TX_BD_PER_PAGE) >> 8)
+#define TX_IDX(x)               ((x) & USABLE_TX_BD_PER_PAGE)
+
+/*
+ * Trigger pending transmits when the number of available BDs is greater
+ * than 1/8 of the total number of usable BDs.
+ */
+#define BCM_TX_CLEANUP_THRESHOLD(q) (USABLE_TX_BD(q) / 8)
+#define BCM_TX_TIMEOUT 5
+
+/*
+ * Receive Buffer Descriptor (rx_bd) definitions*
+ */
+//#define NUM_RX_PAGES            1
+#define TOTAL_RX_BD_PER_PAGE    (BCM_PAGE_SIZE / sizeof(struct eth_rx_bd))      /*  512 */
+#define USABLE_RX_BD_PER_PAGE   (TOTAL_RX_BD_PER_PAGE - 2)                      /*  510 */
+#define RX_BD_PER_PAGE_MASK     (TOTAL_RX_BD_PER_PAGE - 1)                      /*  511 */
+#define TOTAL_RX_BD(q)          (TOTAL_RX_BD_PER_PAGE * q->nb_rx_pages)         /*  512 */
+#define USABLE_RX_BD(q)         (USABLE_RX_BD_PER_PAGE * q->nb_rx_pages)        /*  510 */
+#define MAX_RX_BD(q)            (TOTAL_RX_BD(q) - 1)                            /*  511 */
+#define RX_BD_NEXT_PAGE_DESC_CNT 2
+
+#define NEXT_RX_BD(x)                                                   \
+        ((((x) & RX_BD_PER_PAGE_MASK) ==                                \
+        (USABLE_RX_BD_PER_PAGE - 1)) ? (x) + 3 : (x) + 1)
+
+/* x & 0x3ff */
+#define RX_BD(x, q)             ((x) & MAX_RX_BD(q))
+#define RX_PAGE(x)              (((x) & ~RX_BD_PER_PAGE_MASK) >> 9)
+#define RX_IDX(x)               ((x) & RX_BD_PER_PAGE_MASK)
+
+/*
+ * Receive Completion Queue definitions*
+ */
+//#define NUM_RCQ_PAGES           (NUM_RX_PAGES * 4)
+#define TOTAL_RCQ_ENTRIES_PER_PAGE (BCM_PAGE_SIZE / sizeof(union eth_rx_cqe))   /*  128 */
+#define USABLE_RCQ_ENTRIES_PER_PAGE (TOTAL_RCQ_ENTRIES_PER_PAGE - 1)            /*  127 */
+#define TOTAL_RCQ_ENTRIES(q)    (TOTAL_RCQ_ENTRIES_PER_PAGE * q->nb_cq_pages)   /*  512 */
+#define USABLE_RCQ_ENTRIES(q)   (USABLE_RCQ_ENTRIES_PER_PAGE * q->nb_cq_pages)  /*  508 */
+#define MAX_RCQ_ENTRIES(q)      (TOTAL_RCQ_ENTRIES(q) - 1)                      /*  511 */
+#define RCQ_NEXT_PAGE_DESC_CNT 1
+
+#define NEXT_RCQ_IDX(x)                                                 \
+        ((((x) & USABLE_RCQ_ENTRIES_PER_PAGE) ==                        \
+        (USABLE_RCQ_ENTRIES_PER_PAGE - 1)) ? (x) + 2 : (x) + 1)
+
+#define CQE_BD_REL                                                      \
+	(sizeof(union eth_rx_cqe) / sizeof(struct eth_rx_bd))
+
+#define RCQ_BD_PAGES(q)                                                 \
+	(q->nb_rx_pages * CQE_BD_REL)
+
+#define RCQ_ENTRY(x, q)         ((x) & MAX_RCQ_ENTRIES(q))
+#define RCQ_PAGE(x)             (((x) & ~USABLE_RCQ_ENTRIES_PER_PAGE) >> 7)
+#define RCQ_IDX(x)              ((x) & USABLE_RCQ_ENTRIES_PER_PAGE)
+
+/*
+ * dropless fc calculations for BDs
+ * Number of BDs should be as number of buffers in BRB:
+ * Low threshold takes into account RX_BD_NEXT_PAGE_DESC_CNT
+ * "next" elements on each page
+ */
+#define NUM_BD_REQ(sc) \
+    BRB_SIZE(sc)
+#define NUM_BD_PG_REQ(sc)                                                  \
+    ((NUM_BD_REQ(sc) + USABLE_RX_BD_PER_PAGE - 1) / USABLE_RX_BD_PER_PAGE)
+#define BD_TH_LO(sc)                                \
+    (NUM_BD_REQ(sc) +                               \
+     NUM_BD_PG_REQ(sc) * RX_BD_NEXT_PAGE_DESC_CNT + \
+     FW_DROP_LEVEL(sc))
+#define BD_TH_HI(sc)                      \
+    (BD_TH_LO(sc) + DROPLESS_FC_HEADROOM)
+#define MIN_RX_AVAIL(sc)                           \
+    ((sc)->dropless_fc ? BD_TH_HI(sc) + 128 : 128)
+
+/*
+ * dropless fc calculations for RCQs
+ * Number of RCQs should be as number of buffers in BRB:
+ * Low threshold takes into account RCQ_NEXT_PAGE_DESC_CNT
+ * "next" elements on each page
+ */
+#define NUM_RCQ_REQ(sc) \
+    BRB_SIZE(sc)
+#define NUM_RCQ_PG_REQ(sc)                                              \
+    ((NUM_RCQ_REQ(sc) + USABLE_RCQ_ENTRIES_PER_PAGE - 1) / USABLE_RCQ_ENTRIES_PER_PAGE)
+#define RCQ_TH_LO(sc)                              \
+    (NUM_RCQ_REQ(sc) +                             \
+     NUM_RCQ_PG_REQ(sc) * RCQ_NEXT_PAGE_DESC_CNT + \
+     FW_DROP_LEVEL(sc))
+#define RCQ_TH_HI(sc)                      \
+    (RCQ_TH_LO(sc) + DROPLESS_FC_HEADROOM)
+
+/* Load / Unload modes */
+#define LOAD_NORMAL       0
+#define LOAD_OPEN         1
+#define LOAD_DIAG         2
+#define LOAD_LOOPBACK_EXT 3
+#define UNLOAD_NORMAL     0
+#define UNLOAD_CLOSE      1
+#define UNLOAD_RECOVERY   2
+
+/* Some constants... */
+//#define MAX_PATH_NUM       2
+//#define E2_MAX_NUM_OF_VFS  64
+//#define E1H_FUNC_MAX       8
+//#define E2_FUNC_MAX        4   /* per path */
+#define MAX_VNIC_NUM       4
+#define MAX_FUNC_NUM       8   /* common to all chips */
+//#define MAX_NDSB           HC_SB_MAX_SB_E2 /* max non-default status block */
+#define MAX_RSS_CHAINS     16 /* a constant for HW limit */
+#define MAX_MSI_VECTOR     8  /* a constant for HW limit */
+
+#define ILT_NUM_PAGE_ENTRIES 3072
+/*
+ * 57711 we use whole table since we have 8 functions.
+ * 57712 we have only 4 functions, but use same size per func, so only half
+ * of the table is used.
+ */
+#define ILT_PER_FUNC        (ILT_NUM_PAGE_ENTRIES / 8)
+#define FUNC_ILT_BASE(func) (func * ILT_PER_FUNC)
+/*
+ * the phys address is shifted right 12 bits and has an added
+ * 1=valid bit added to the 53rd bit
+ * then since this is a wide register(TM)
+ * we split it into two 32 bit writes
+ */
+#define ONCHIP_ADDR1(x) ((uint32_t)(((uint64_t)x >> 12) & 0xFFFFFFFF))
+#define ONCHIP_ADDR2(x) ((uint32_t)((1 << 20) | ((uint64_t)x >> 44)))
+
+/* L2 header size + 2*VLANs (8 bytes) + LLC SNAP (8 bytes) */
+#define ETH_HLEN                  14
+#define ETH_OVERHEAD              (ETH_HLEN + 8 + 8)
+#define ETH_MIN_PACKET_SIZE       60
+#define ETH_MAX_PACKET_SIZE       ETHERMTU /* 1500 */
+#define ETH_MAX_JUMBO_PACKET_SIZE 9600
+/* TCP with Timestamp Option (32) + IPv6 (40) */
+
+/* max supported alignment is 256 (8 shift) */
+#define BCM_RX_ALIGN_SHIFT 8
+/* FW uses 2 cache lines alignment for start packet and size  */
+#define BCM_FW_RX_ALIGN_START (1 << BCM_RX_ALIGN_SHIFT)
+#define BCM_FW_RX_ALIGN_END   (1 << BCM_RX_ALIGN_SHIFT)
+
+#define BCM_PXP_DRAM_ALIGN (BCM_RX_ALIGN_SHIFT - 5)
+
+struct bcm_bar {
+	void* base_addr;
+};
+
+/* Used to manage DMA allocations. */
+struct bcm_dma {
+	struct bcm_softc        *sc;
+	phys_addr_t             paddr;
+	void                    *vaddr;
+	int                     nseg;
+	char                    msg[RTE_MEMZONE_NAMESIZE - 6];
+};
+
+/* attn group wiring */
+#define MAX_DYNAMIC_ATTN_GRPS 8
+
+struct attn_route {
+    uint32_t sig[5];
+};
+
+struct iro {
+    uint32_t base;
+    uint16_t m1;
+    uint16_t m2;
+    uint16_t m3;
+    uint16_t size;
+};
+
+union bcm_host_hc_status_block {
+    /* pointer to fp status block e2 */
+    struct host_hc_status_block_e2  *e2_sb;
+    /* pointer to fp status block e1x */
+    struct host_hc_status_block_e1x *e1x_sb;
+};
+
+union bcm_db_prod {
+    struct doorbell_set_prod data;
+    uint32_t                 raw;
+};
+
+struct bcm_sw_tx_bd {
+    struct mbuf  *m;
+    uint16_t     first_bd;
+    uint8_t      flags;
+/* set on the first BD descriptor when there is a split BD */
+#define BCM_TSO_SPLIT_BD (1 << 0)
+};
+
+/*
+ * This is the HSI fastpath data structure. There can be up to MAX_RSS_CHAIN
+ * instances of the fastpath structure when using multiple queues.
+ */
+struct bcm_fastpath {
+	/* pointer back to parent structure */
+	struct bcm_softc *sc;
+
+	/* status block */
+	struct bcm_dma                 sb_dma;
+	union bcm_host_hc_status_block status_block;
+
+	phys_addr_t tx_desc_mapping;
+
+	phys_addr_t rx_desc_mapping;
+	phys_addr_t rx_comp_mapping;
+
+	uint16_t *sb_index_values;
+	uint16_t *sb_running_index;
+	uint32_t ustorm_rx_prods_offset;
+
+	uint8_t igu_sb_id; /* status block number in HW */
+	uint8_t fw_sb_id;  /* status block number in FW */
+
+	uint32_t rx_buf_size;
+	int mbuf_alloc_size;
+
+	int state;
+#define BCM_FP_STATE_CLOSED  0x01
+#define BCM_FP_STATE_IRQ     0x02
+#define BCM_FP_STATE_OPENING 0x04
+#define BCM_FP_STATE_OPEN    0x08
+#define BCM_FP_STATE_HALTING 0x10
+#define BCM_FP_STATE_HALTED  0x20
+
+	/* reference back to this fastpath queue number */
+	uint8_t index; /* this is also the 'cid' */
+#define FP_IDX(fp) (fp->index)
+
+	/* ethernet client ID (each fastpath set of RX/TX/CQE is a client) */
+	uint8_t cl_id;
+#define FP_CL_ID(fp) (fp->cl_id)
+	uint8_t cl_qzone_id;
+
+	uint16_t fp_hc_idx;
+
+	union bcm_db_prod tx_db;
+
+	struct tstorm_per_queue_stats old_tclient;
+	struct ustorm_per_queue_stats old_uclient;
+	struct xstorm_per_queue_stats old_xclient;
+	struct bcm_eth_q_stats        eth_q_stats;
+	struct bcm_eth_q_stats_old    eth_q_stats_old;
+
+	/* Pointer to the receive consumer in the status block */
+	uint16_t *rx_cq_cons_sb;
+
+	/* Pointer to the transmit consumer in the status block */
+	uint16_t *tx_cons_sb;
+
+	/* transmit timeout until chip reset */
+	int watchdog_timer;
+
+}; /* struct bcm_fastpath */
+
+#define BCM_MAX_NUM_OF_VFS 64
+#define BCM_VF_ID_INVALID  0xFF
+
+/* maximum number of fast-path interrupt contexts */
+#define FP_SB_MAX_E1x 16
+#define FP_SB_MAX_E2  HC_SB_MAX_SB_E2
+
+union cdu_context {
+    struct eth_context eth;
+    char pad[1024];
+};
+
+/* CDU host DB constants */
+#define CDU_ILT_PAGE_SZ_HW 2
+#define CDU_ILT_PAGE_SZ    (8192 << CDU_ILT_PAGE_SZ_HW) /* 32K */
+#define ILT_PAGE_CIDS      (CDU_ILT_PAGE_SZ / sizeof(union cdu_context))
+
+#define CNIC_ISCSI_CID_MAX 256
+#define CNIC_FCOE_CID_MAX  2048
+#define CNIC_CID_MAX       (CNIC_ISCSI_CID_MAX + CNIC_FCOE_CID_MAX)
+#define CNIC_ILT_LINES     DIV_ROUND_UP(CNIC_CID_MAX, ILT_PAGE_CIDS)
+
+#define QM_ILT_PAGE_SZ_HW  0
+#define QM_ILT_PAGE_SZ     (4096 << QM_ILT_PAGE_SZ_HW) /* 4K */
+#define QM_CID_ROUND       1024
+
+/* TM (timers) host DB constants */
+#define TM_ILT_PAGE_SZ_HW  0
+#define TM_ILT_PAGE_SZ     (4096 << TM_ILT_PAGE_SZ_HW) /* 4K */
+/*#define TM_CONN_NUM        (CNIC_STARTING_CID+CNIC_ISCSI_CXT_MAX) */
+#define TM_CONN_NUM        1024
+#define TM_ILT_SZ          (8 * TM_CONN_NUM)
+#define TM_ILT_LINES       DIV_ROUND_UP(TM_ILT_SZ, TM_ILT_PAGE_SZ)
+
+/* SRC (Searcher) host DB constants */
+#define SRC_ILT_PAGE_SZ_HW 0
+#define SRC_ILT_PAGE_SZ    (4096 << SRC_ILT_PAGE_SZ_HW) /* 4K */
+#define SRC_HASH_BITS      10
+#define SRC_CONN_NUM       (1 << SRC_HASH_BITS) /* 1024 */
+#define SRC_ILT_SZ         (sizeof(struct src_ent) * SRC_CONN_NUM)
+#define SRC_T2_SZ          SRC_ILT_SZ
+#define SRC_ILT_LINES      DIV_ROUND_UP(SRC_ILT_SZ, SRC_ILT_PAGE_SZ)
+
+struct hw_context {
+    struct bcm_dma    vcxt_dma;
+    union cdu_context *vcxt;
+    //phys_addr_t        cxt_mapping;
+    size_t            size;
+};
+
+#define SM_RX_ID 0
+#define SM_TX_ID 1
+
+/* defines for multiple tx priority indices */
+#define FIRST_TX_ONLY_COS_INDEX 1
+#define FIRST_TX_COS_INDEX      0
+
+#define CID_TO_FP(cid, sc) ((cid) % BCM_NUM_NON_CNIC_QUEUES(sc))
+
+#define HC_INDEX_ETH_RX_CQ_CONS       1
+#define HC_INDEX_OOO_TX_CQ_CONS       4
+#define HC_INDEX_ETH_TX_CQ_CONS_COS0  5
+#define HC_INDEX_ETH_TX_CQ_CONS_COS1  6
+#define HC_INDEX_ETH_TX_CQ_CONS_COS2  7
+#define HC_INDEX_ETH_FIRST_TX_CQ_CONS HC_INDEX_ETH_TX_CQ_CONS_COS0
+
+/* congestion management fairness mode */
+#define CMNG_FNS_NONE   0
+#define CMNG_FNS_MINMAX 1
+
+/* CMNG constants, as derived from system spec calculations */
+/* default MIN rate in case VNIC min rate is configured to zero - 100Mbps */
+#define DEF_MIN_RATE 100
+/* resolution of the rate shaping timer - 400 usec */
+#define RS_PERIODIC_TIMEOUT_USEC 400
+/* number of bytes in single QM arbitration cycle -
+ * coefficient for calculating the fairness timer */
+#define QM_ARB_BYTES 160000
+/* resolution of Min algorithm 1:100 */
+#define MIN_RES 100
+/* how many bytes above threshold for the minimal credit of Min algorithm*/
+#define MIN_ABOVE_THRESH 32768
+/* fairness algorithm integration time coefficient -
+ * for calculating the actual Tfair */
+#define T_FAIR_COEF ((MIN_ABOVE_THRESH + QM_ARB_BYTES) * 8 * MIN_RES)
+/* memory of fairness algorithm - 2 cycles */
+#define FAIR_MEM 2
+
+#define HC_SEG_ACCESS_DEF   0 /* Driver decision 0-3 */
+#define HC_SEG_ACCESS_ATTN  4
+#define HC_SEG_ACCESS_NORM  0 /* Driver decision 0-1 */
+
+/*
+ * The total number of L2 queues, MSIX vectors and HW contexts (CIDs) is
+ * control by the number of fast-path status blocks supported by the
+ * device (HW/FW). Each fast-path status block (FP-SB) aka non-default
+ * status block represents an independent interrupts context that can
+ * serve a regular L2 networking queue. However special L2 queues such
+ * as the FCoE queue do not require a FP-SB and other components like
+ * the CNIC may consume FP-SB reducing the number of possible L2 queues
+ *
+ * If the maximum number of FP-SB available is X then:
+ * a. If CNIC is supported it consumes 1 FP-SB thus the max number of
+ *    regular L2 queues is Y=X-1
+ * b. in MF mode the actual number of L2 queues is Y= (X-1/MF_factor)
+ * c. If the FCoE L2 queue is supported the actual number of L2 queues
+ *    is Y+1
+ * d. The number of irqs (MSIX vectors) is either Y+1 (one extra for
+ *    slow-path interrupts) or Y+2 if CNIC is supported (one additional
+ *    FP interrupt context for the CNIC).
+ * e. The number of HW context (CID count) is always X or X+1 if FCoE
+ *    L2 queue is supported. the cid for the FCoE L2 queue is always X.
+ *
+ * So this is quite simple for now as no ULPs are supported yet. :-)
+ */
+#define BCM_NUM_QUEUES(sc)          ((sc)->num_queues)
+#define BCM_NUM_ETH_QUEUES(sc)      BCM_NUM_QUEUES(sc)
+#define BCM_NUM_NON_CNIC_QUEUES(sc) BCM_NUM_QUEUES(sc)
+#define BCM_NUM_RX_QUEUES(sc)       BCM_NUM_QUEUES(sc)
+
+#define FOR_EACH_QUEUE(sc, var)                          \
+    for ((var) = 0; (var) < BCM_NUM_QUEUES(sc); (var)++)
+
+#define FOR_EACH_NONDEFAULT_QUEUE(sc, var)               \
+    for ((var) = 1; (var) < BCM_NUM_QUEUES(sc); (var)++)
+
+#define FOR_EACH_ETH_QUEUE(sc, var)                          \
+    for ((var) = 0; (var) < BCM_NUM_ETH_QUEUES(sc); (var)++)
+
+#define FOR_EACH_NONDEFAULT_ETH_QUEUE(sc, var)               \
+    for ((var) = 1; (var) < BCM_NUM_ETH_QUEUES(sc); (var)++)
+
+#define FOR_EACH_COS_IN_TX_QUEUE(sc, var)           \
+    for ((var) = 0; (var) < (sc)->max_cos; (var)++)
+
+#define FOR_EACH_CNIC_QUEUE(sc, var)     \
+    for ((var) = BCM_NUM_ETH_QUEUES(sc); \
+         (var) < BCM_NUM_QUEUES(sc);     \
+         (var)++)
+
+enum {
+    OOO_IDX_OFFSET,
+    FCOE_IDX_OFFSET,
+    FWD_IDX_OFFSET,
+};
+
+#define FCOE_IDX(sc)              (BCM_NUM_NON_CNIC_QUEUES(sc) + FCOE_IDX_OFFSET)
+#define bcm_fcoe_fp(sc)           (&sc->fp[FCOE_IDX(sc)])
+#define bcm_fcoe(sc, var)         (bcm_fcoe_fp(sc)->var)
+#define bcm_fcoe_inner_sp_obj(sc) (&sc->sp_objs[FCOE_IDX(sc)])
+#define bcm_fcoe_sp_obj(sc, var)  (bcm_fcoe_inner_sp_obj(sc)->var)
+#define bcm_fcoe_tx(sc, var)      (bcm_fcoe_fp(sc)->txdata_ptr[FIRST_TX_COS_INDEX]->var)
+
+#define OOO_IDX(sc)               (BCM_NUM_NON_CNIC_QUEUES(sc) + OOO_IDX_OFFSET)
+#define bcm_ooo_fp(sc)            (&sc->fp[OOO_IDX(sc)])
+#define bcm_ooo(sc, var)          (bcm_ooo_fp(sc)->var)
+#define bcm_ooo_inner_sp_obj(sc)  (&sc->sp_objs[OOO_IDX(sc)])
+#define bcm_ooo_sp_obj(sc, var)   (bcm_ooo_inner_sp_obj(sc)->var)
+
+#define FWD_IDX(sc)               (BCM_NUM_NON_CNIC_QUEUES(sc) + FWD_IDX_OFFSET)
+#define bcm_fwd_fp(sc)            (&sc->fp[FWD_IDX(sc)])
+#define bcm_fwd(sc, var)          (bcm_fwd_fp(sc)->var)
+#define bcm_fwd_inner_sp_obj(sc)  (&sc->sp_objs[FWD_IDX(sc)])
+#define bcm_fwd_sp_obj(sc, var)   (bcm_fwd_inner_sp_obj(sc)->var)
+#define bcm_fwd_txdata(fp)        (fp->txdata_ptr[FIRST_TX_COS_INDEX])
+
+#define IS_ETH_FP(fp)    ((fp)->index < BCM_NUM_ETH_QUEUES((fp)->sc))
+#define IS_FCOE_FP(fp)   ((fp)->index == FCOE_IDX((fp)->sc))
+#define IS_FCOE_IDX(idx) ((idx) == FCOE_IDX(sc))
+#define IS_FWD_FP(fp)    ((fp)->index == FWD_IDX((fp)->sc))
+#define IS_FWD_IDX(idx)  ((idx) == FWD_IDX(sc))
+#define IS_OOO_FP(fp)    ((fp)->index == OOO_IDX((fp)->sc))
+#define IS_OOO_IDX(idx)  ((idx) == OOO_IDX(sc))
+
+enum {
+    BCM_PORT_QUERY_IDX,
+    BCM_PF_QUERY_IDX,
+    BCM_FCOE_QUERY_IDX,
+    BCM_FIRST_QUEUE_QUERY_IDX,
+};
+
+struct bcm_fw_stats_req {
+    struct stats_query_header hdr;
+    struct stats_query_entry  query[FP_SB_MAX_E1x +
+                                    BCM_FIRST_QUEUE_QUERY_IDX];
+};
+
+struct bcm_fw_stats_data {
+    struct stats_counter          storm_counters;
+    struct per_port_stats         port;
+    struct per_pf_stats           pf;
+    struct per_queue_stats        queue_stats[1];
+};
+
+/* IGU MSIX STATISTICS on 57712: 64 for VFs; 4 for PFs; 4 for Attentions */
+#define BCM_IGU_STAS_MSG_VF_CNT 64
+#define BCM_IGU_STAS_MSG_PF_CNT 4
+
+#define MAX_DMAE_C 8
+
+/*
+ * This is the slowpath data structure. It is mapped into non-paged memory
+ * so that the hardware can access it's contents directly and must be page
+ * aligned.
+ */
+struct bcm_slowpath {
+
+    /* used by the DMAE command executer */
+    struct dmae_command dmae[MAX_DMAE_C];
+
+    /* statistics completion */
+    uint32_t stats_comp;
+
+    /* firmware defined statistics blocks */
+    union mac_stats        mac_stats;
+    struct nig_stats       nig_stats;
+    struct host_port_stats port_stats;
+    struct host_func_stats func_stats;
+
+    /* DMAE completion value and data source/sink */
+    uint32_t wb_comp;
+    uint32_t wb_data[4];
+
+    union {
+        struct mac_configuration_cmd          e1x;
+        struct eth_classify_rules_ramrod_data e2;
+    } mac_rdata;
+
+    union {
+        struct tstorm_eth_mac_filter_config e1x;
+        struct eth_filter_rules_ramrod_data e2;
+    } rx_mode_rdata;
+
+    struct eth_rss_update_ramrod_data rss_rdata;
+
+    union {
+        struct mac_configuration_cmd           e1;
+        struct eth_multicast_rules_ramrod_data e2;
+    } mcast_rdata;
+
+    union {
+        struct function_start_data        func_start;
+        struct flow_control_configuration pfc_config; /* for DCBX ramrod */
+    } func_rdata;
+
+    /* Queue State related ramrods */
+    union {
+        struct client_init_ramrod_data   init_data;
+        struct client_update_ramrod_data update_data;
+    } q_rdata;
+
+    /*
+     * AFEX ramrod can not be a part of func_rdata union because these
+     * events might arrive in parallel to other events from func_rdata.
+     * If they were defined in the same union the data can get corrupted.
+     */
+    struct afex_vif_list_ramrod_data func_afex_rdata;
+
+    union drv_info_to_mcp drv_info_to_mcp;
+}; /* struct bcm_slowpath */
+
+/*
+ * Port specifc data structure.
+ */
+struct bcm_port {
+    /*
+     * Port Management Function (for 57711E only).
+     * When this field is set the driver instance is
+     * responsible for managing port specifc
+     * configurations such as handling link attentions.
+     */
+    uint32_t pmf;
+
+    /* Ethernet maximum transmission unit. */
+    uint16_t ether_mtu;
+
+    uint32_t link_config[ELINK_LINK_CONFIG_SIZE];
+
+    uint32_t ext_phy_config;
+
+    /* Port feature config.*/
+    uint32_t config;
+
+    /* Defines the features supported by the PHY. */
+    uint32_t supported[ELINK_LINK_CONFIG_SIZE];
+
+    /* Defines the features advertised by the PHY. */
+    uint32_t advertising[ELINK_LINK_CONFIG_SIZE];
+#define ADVERTISED_10baseT_Half    (1 << 1)
+#define ADVERTISED_10baseT_Full    (1 << 2)
+#define ADVERTISED_100baseT_Half   (1 << 3)
+#define ADVERTISED_100baseT_Full   (1 << 4)
+#define ADVERTISED_1000baseT_Half  (1 << 5)
+#define ADVERTISED_1000baseT_Full  (1 << 6)
+#define ADVERTISED_TP              (1 << 7)
+#define ADVERTISED_FIBRE           (1 << 8)
+#define ADVERTISED_Autoneg         (1 << 9)
+#define ADVERTISED_Asym_Pause      (1 << 10)
+#define ADVERTISED_Pause           (1 << 11)
+#define ADVERTISED_2500baseX_Full  (1 << 15)
+#define ADVERTISED_10000baseT_Full (1 << 16)
+
+    uint32_t    phy_addr;
+
+    /*
+     * MCP scratchpad address for port specific statistics.
+     * The device is responsible for writing statistcss
+     * back to the MCP for use with management firmware such
+     * as UMP/NC-SI.
+     */
+    uint32_t port_stx;
+
+    struct nig_stats old_nig_stats;
+}; /* struct bcm_port */
+
+struct bcm_mf_info {
+    uint32_t mf_config[E1HVN_MAX];
+
+    uint32_t vnics_per_port;   /* 1, 2 or 4 */
+    uint32_t multi_vnics_mode; /* can be set even if vnics_per_port = 1 */
+    uint32_t path_has_ovlan;   /* MF mode in the path (can be different than the MF mode of the function */
+
+#define IS_MULTI_VNIC(sc)  ((sc)->devinfo.mf_info.multi_vnics_mode)
+#define VNICS_PER_PORT(sc) ((sc)->devinfo.mf_info.vnics_per_port)
+#define VNICS_PER_PATH(sc)                                  \
+    ((sc)->devinfo.mf_info.vnics_per_port *                 \
+     ((CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) ? 2 : 1 ))
+
+    uint8_t min_bw[MAX_VNIC_NUM];
+    uint8_t max_bw[MAX_VNIC_NUM];
+
+    uint16_t ext_id; /* vnic outer vlan or VIF ID */
+#define VALID_OVLAN(ovlan) ((ovlan) <= 4096)
+#define INVALID_VIF_ID 0xFFFF
+#define OVLAN(sc) ((sc)->devinfo.mf_info.ext_id)
+#define VIF_ID(sc) ((sc)->devinfo.mf_info.ext_id)
+
+    uint16_t default_vlan;
+#define NIV_DEFAULT_VLAN(sc) ((sc)->devinfo.mf_info.default_vlan)
+
+    uint8_t niv_allowed_priorities;
+#define NIV_ALLOWED_PRIORITIES(sc) ((sc)->devinfo.mf_info.niv_allowed_priorities)
+
+    uint8_t niv_default_cos;
+#define NIV_DEFAULT_COS(sc) ((sc)->devinfo.mf_info.niv_default_cos)
+
+    uint8_t niv_mba_enabled;
+
+    enum mf_cfg_afex_vlan_mode afex_vlan_mode;
+#define AFEX_VLAN_MODE(sc) ((sc)->devinfo.mf_info.afex_vlan_mode)
+    int                        afex_def_vlan_tag;
+    uint32_t                   pending_max;
+
+    uint16_t flags;
+#define MF_INFO_VALID_MAC       0x0001
+
+    uint16_t mf_ov;
+    uint8_t mf_mode; /* Switch-Dependent or Switch-Independent */
+#define IS_MF(sc)                        \
+    (IS_MULTI_VNIC(sc) &&                \
+     ((sc)->devinfo.mf_info.mf_mode != 0))
+#define IS_MF_SD(sc)                                     \
+    (IS_MULTI_VNIC(sc) &&                                \
+     ((sc)->devinfo.mf_info.mf_mode == MULTI_FUNCTION_SD))
+#define IS_MF_SI(sc)                                     \
+    (IS_MULTI_VNIC(sc) &&                                \
+     ((sc)->devinfo.mf_info.mf_mode == MULTI_FUNCTION_SI))
+#define IS_MF_AFEX(sc)                              \
+    (IS_MULTI_VNIC(sc) &&                           \
+     ((sc)->devinfo.mf_info.mf_mode == MULTI_FUNCTION_AFEX))
+#define IS_MF_SD_MODE(sc)   IS_MF_SD(sc)
+#define IS_MF_SI_MODE(sc)   IS_MF_SI(sc)
+#define IS_MF_AFEX_MODE(sc) IS_MF_AFEX(sc)
+
+    uint32_t mf_protos_supported;
+    #define MF_PROTO_SUPPORT_ETHERNET 0x1
+    #define MF_PROTO_SUPPORT_ISCSI    0x2
+    #define MF_PROTO_SUPPORT_FCOE     0x4
+}; /* struct bcm_mf_info */
+
+/* Device information data structure. */
+struct bcm_devinfo {
+    /* PCIe info */
+    uint16_t vendor_id;
+    uint16_t device_id;
+    uint16_t subvendor_id;
+    uint16_t subdevice_id;
+
+    /*
+     * chip_id = 0b'CCCCCCCCCCCCCCCCRRRRMMMMMMMMBBBB'
+     *   C = Chip Number   (bits 16-31)
+     *   R = Chip Revision (bits 12-15)
+     *   M = Chip Metal    (bits 4-11)
+     *   B = Chip Bond ID  (bits 0-3)
+     */
+    uint32_t chip_id;
+#define CHIP_ID(sc)           ((sc)->devinfo.chip_id & 0xffff0000)
+#define CHIP_NUM(sc)          ((sc)->devinfo.chip_id >> 16)
+/* device ids */
+#define CHIP_NUM_57711        0x164f
+#define CHIP_NUM_57711E       0x1650
+#define CHIP_NUM_57712        0x1662
+#define CHIP_NUM_57712_MF     0x1663
+#define CHIP_NUM_57712_VF     0x166f
+#define CHIP_NUM_57800        0x168a
+#define CHIP_NUM_57800_MF     0x16a5
+#define CHIP_NUM_57800_VF     0x16a9
+#define CHIP_NUM_57810        0x168e
+#define CHIP_NUM_57810_MF     0x16ae
+#define CHIP_NUM_57810_VF     0x16af
+#define CHIP_NUM_57811        0x163d
+#define CHIP_NUM_57811_MF     0x163e
+#define CHIP_NUM_57811_VF     0x163f
+#define CHIP_NUM_57840_OBS    0x168d
+#define CHIP_NUM_57840_OBS_MF 0x16ab
+#define CHIP_NUM_57840_4_10   0x16a1
+#define CHIP_NUM_57840_2_20   0x16a2
+#define CHIP_NUM_57840_MF     0x16a4
+#define CHIP_NUM_57840_VF     0x16ad
+
+#define CHIP_REV_SHIFT      12
+#define CHIP_REV_MASK       (0xF << CHIP_REV_SHIFT)
+#define CHIP_REV(sc)        ((sc)->devinfo.chip_id & CHIP_REV_MASK)
+
+#define CHIP_REV_Ax         (0x0 << CHIP_REV_SHIFT)
+#define CHIP_REV_Bx         (0x1 << CHIP_REV_SHIFT)
+#define CHIP_REV_Cx         (0x2 << CHIP_REV_SHIFT)
+
+#define CHIP_REV_IS_SLOW(sc)    \
+    (CHIP_REV(sc) > 0x00005000)
+#define CHIP_REV_IS_FPGA(sc)                              \
+    (CHIP_REV_IS_SLOW(sc) && (CHIP_REV(sc) & 0x00001000))
+#define CHIP_REV_IS_EMUL(sc)                               \
+    (CHIP_REV_IS_SLOW(sc) && !(CHIP_REV(sc) & 0x00001000))
+#define CHIP_REV_IS_ASIC(sc) \
+    (!CHIP_REV_IS_SLOW(sc))
+
+#define CHIP_METAL(sc)      ((sc->devinfo.chip_id) & 0x00000ff0)
+#define CHIP_BOND_ID(sc)    ((sc->devinfo.chip_id) & 0x0000000f)
+
+#define CHIP_IS_57711(sc)   (CHIP_NUM(sc) == CHIP_NUM_57711)
+#define CHIP_IS_57711E(sc)  (CHIP_NUM(sc) == CHIP_NUM_57711E)
+#define CHIP_IS_E1H(sc)     ((CHIP_IS_57711(sc)) || \
+                             (CHIP_IS_57711E(sc)))
+#define CHIP_IS_E1x(sc)     CHIP_IS_E1H(sc)
+
+#define CHIP_IS_57712(sc)    (CHIP_NUM(sc) == CHIP_NUM_57712)
+#define CHIP_IS_57712_MF(sc) (CHIP_NUM(sc) == CHIP_NUM_57712_MF)
+#define CHIP_IS_57712_VF(sc) (CHIP_NUM(sc) == CHIP_NUM_57712_VF)
+#define CHIP_IS_E2(sc)       (CHIP_IS_57712(sc) ||  \
+                              CHIP_IS_57712_MF(sc))
+
+#define CHIP_IS_57800(sc)    (CHIP_NUM(sc) == CHIP_NUM_57800)
+#define CHIP_IS_57800_MF(sc) (CHIP_NUM(sc) == CHIP_NUM_57800_MF)
+#define CHIP_IS_57800_VF(sc) (CHIP_NUM(sc) == CHIP_NUM_57800_VF)
+#define CHIP_IS_57810(sc)    (CHIP_NUM(sc) == CHIP_NUM_57810)
+#define CHIP_IS_57810_MF(sc) (CHIP_NUM(sc) == CHIP_NUM_57810_MF)
+#define CHIP_IS_57810_VF(sc) (CHIP_NUM(sc) == CHIP_NUM_57810_VF)
+#define CHIP_IS_57811(sc)    (CHIP_NUM(sc) == CHIP_NUM_57811)
+#define CHIP_IS_57811_MF(sc) (CHIP_NUM(sc) == CHIP_NUM_57811_MF)
+#define CHIP_IS_57811_VF(sc) (CHIP_NUM(sc) == CHIP_NUM_57811_VF)
+#define CHIP_IS_57840(sc)    ((CHIP_NUM(sc) == CHIP_NUM_57840_OBS)  || \
+                              (CHIP_NUM(sc) == CHIP_NUM_57840_4_10) || \
+                              (CHIP_NUM(sc) == CHIP_NUM_57840_2_20))
+#define CHIP_IS_57840_MF(sc) ((CHIP_NUM(sc) == CHIP_NUM_57840_OBS_MF) || \
+                              (CHIP_NUM(sc) == CHIP_NUM_57840_MF))
+#define CHIP_IS_57840_VF(sc) (CHIP_NUM(sc) == CHIP_NUM_57840_VF)
+
+#define CHIP_IS_E3(sc)      (CHIP_IS_57800(sc)    || \
+                             CHIP_IS_57800_MF(sc) || \
+                             CHIP_IS_57800_VF(sc) || \
+                             CHIP_IS_57810(sc)    || \
+                             CHIP_IS_57810_MF(sc) || \
+                             CHIP_IS_57810_VF(sc) || \
+                             CHIP_IS_57811(sc)    || \
+                             CHIP_IS_57811_MF(sc) || \
+                             CHIP_IS_57811_VF(sc) || \
+                             CHIP_IS_57840(sc)    || \
+                             CHIP_IS_57840_MF(sc) || \
+                             CHIP_IS_57840_VF(sc))
+#define CHIP_IS_E3A0(sc)    (CHIP_IS_E3(sc) &&              \
+                             (CHIP_REV(sc) == CHIP_REV_Ax))
+#define CHIP_IS_E3B0(sc)    (CHIP_IS_E3(sc) &&              \
+                             (CHIP_REV(sc) == CHIP_REV_Bx))
+
+#define USES_WARPCORE(sc)   (CHIP_IS_E3(sc))
+#define CHIP_IS_E2E3(sc)    (CHIP_IS_E2(sc) || \
+                             CHIP_IS_E3(sc))
+
+#define CHIP_IS_MF_CAP(sc)  (CHIP_IS_57711E(sc)  ||  \
+                             CHIP_IS_57712_MF(sc) || \
+                             CHIP_IS_E3(sc))
+
+#define IS_VF(sc)           ((sc)->flags & BCM_IS_VF_FLAG)
+#define IS_PF(sc)           (!IS_VF(sc))
+
+/*
+ * This define is used in two main places:
+ * 1. In the early stages of nic_load, to know if to configure Parser/Searcher
+ * to nic-only mode or to offload mode. Offload mode is configured if either
+ * the chip is E1x (where NIC_MODE register is not applicable), or if cnic
+ * already registered for this port (which means that the user wants storage
+ * services).
+ * 2. During cnic-related load, to know if offload mode is already configured
+ * in the HW or needs to be configrued. Since the transition from nic-mode to
+ * offload-mode in HW causes traffic coruption, nic-mode is configured only
+ * in ports on which storage services where never requested.
+ */
+#define CONFIGURE_NIC_MODE(sc) (!CHIP_IS_E1x(sc) && !CNIC_ENABLED(sc))
+
+    uint8_t  chip_port_mode;
+#define CHIP_4_PORT_MODE        0x0
+#define CHIP_2_PORT_MODE        0x1
+#define CHIP_PORT_MODE_NONE     0x2
+#define CHIP_PORT_MODE(sc)      ((sc)->devinfo.chip_port_mode)
+#define CHIP_IS_MODE_4_PORT(sc) (CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE)
+
+    uint8_t int_block;
+#define INT_BLOCK_HC            0
+#define INT_BLOCK_IGU           1
+#define INT_BLOCK_MODE_NORMAL   0
+#define INT_BLOCK_MODE_BW_COMP  2
+#define CHIP_INT_MODE_IS_NBC(sc)                          \
+    (!CHIP_IS_E1x(sc) &&                                  \
+     !((sc)->devinfo.int_block & INT_BLOCK_MODE_BW_COMP))
+#define CHIP_INT_MODE_IS_BC(sc) (!CHIP_INT_MODE_IS_NBC(sc))
+
+    uint32_t shmem_base;
+    uint32_t shmem2_base;
+    uint32_t bc_ver;
+    char bc_ver_str[32];
+    uint32_t mf_cfg_base; /* bootcode shmem address in BAR memory */
+    struct bcm_mf_info mf_info;
+
+    uint32_t flash_size;
+#define NVRAM_1MB_SIZE      0x20000
+#define NVRAM_TIMEOUT_COUNT 30000
+#define NVRAM_PAGE_SIZE     256
+
+    /* PCIe capability information */
+    uint32_t pcie_cap_flags;
+#define BCM_PM_CAPABLE_FLAG     0x00000001
+#define BCM_PCIE_CAPABLE_FLAG   0x00000002
+#define BCM_MSI_CAPABLE_FLAG    0x00000004
+#define BCM_MSIX_CAPABLE_FLAG   0x00000008
+    uint16_t pcie_pm_cap_reg;
+    uint16_t pcie_link_width;
+    uint16_t pcie_link_speed;
+    uint16_t pcie_msi_cap_reg;
+    uint16_t pcie_msix_cap_reg;
+
+    /* device configuration read from bootcode shared memory */
+    uint32_t hw_config;
+    uint32_t hw_config2;
+}; /* struct bcm_devinfo */
+
+struct bcm_sp_objs {
+    struct ecore_vlan_mac_obj mac_obj; /* MACs object */
+    struct ecore_queue_sp_obj q_obj; /* Queue State object */
+}; /* struct bcm_sp_objs */
+
+/*
+ * Data that will be used to create a link report message. We will keep the
+ * data used for the last link report in order to prevent reporting the same
+ * link parameters twice.
+ */
+struct bcm_link_report_data {
+    uint16_t      line_speed;        /* Effective line speed */
+    unsigned long link_report_flags; /* BCM_LINK_REPORT_XXX flags */
+};
+enum {
+    BCM_LINK_REPORT_FULL_DUPLEX,
+    BCM_LINK_REPORT_LINK_DOWN,
+    BCM_LINK_REPORT_RX_FC_ON,
+    BCM_LINK_REPORT_TX_FC_ON
+};
+
+#define BCM_RX_CHAIN_PAGE_SZ    BCM_PAGE_SIZE
+
+struct bcm_pci_cap {
+	struct bcm_pci_cap *next;
+	uint16_t id;
+	uint16_t type;
+	uint16_t addr;
+};
+
+struct bcm_vfdb;
+
+/* Top level device private data structure. */
+struct bcm_softc {
+
+	void            **rx_queues;
+	void            **tx_queues;
+	uint32_t        max_tx_queues;
+	uint32_t        max_rx_queues;
+	const struct rte_pci_device *pci_dev;
+	uint32_t        pci_val;
+	struct bcm_pci_cap *pci_caps;
+#define BCM_INTRS_POLL_PERIOD   1
+
+	void            *firmware;
+	uint64_t        fw_len;
+
+	/* MAC address operations */
+	struct bcm_mac_ops mac_ops;
+
+	/* structures for VF mbox/response/bulletin */
+	struct bcm_vf_mbx_msg	*vf2pf_mbox;
+	struct bcm_dma		vf2pf_mbox_mapping;
+	struct vf_acquire_resp_tlv acquire_resp;
+	struct bcm_vf_bulletin	*pf2vf_bulletin;
+	struct bcm_dma		pf2vf_bulletin_mapping;
+	struct bcm_vf_bulletin	old_bulletin;
+
+	int             media;
+
+	int             state; /* device state */
+#define BCM_STATE_CLOSED                 0x0000
+#define BCM_STATE_OPENING_WAITING_LOAD   0x1000
+#define BCM_STATE_OPENING_WAITING_PORT   0x2000
+#define BCM_STATE_OPEN                   0x3000
+#define BCM_STATE_CLOSING_WAITING_HALT   0x4000
+#define BCM_STATE_CLOSING_WAITING_DELETE 0x5000
+#define BCM_STATE_CLOSING_WAITING_UNLOAD 0x6000
+#define BCM_STATE_DISABLED               0xD000
+#define BCM_STATE_DIAG                   0xE000
+#define BCM_STATE_ERROR                  0xF000
+
+	int flags;
+#define BCM_ONE_PORT_FLAG     0x1
+#define BCM_NO_FCOE_FLAG      0x2
+#define BCM_NO_WOL_FLAG       0x4
+#define BCM_NO_MCP_FLAG       0x8
+#define BCM_NO_ISCSI_OOO_FLAG 0x10
+#define BCM_NO_ISCSI_FLAG     0x20
+#define BCM_MF_FUNC_DIS       0x40
+#define BCM_TX_SWITCHING      0x80
+#define BCM_IS_VF_FLAG        0x100
+
+#define BCM_ONE_PORT(sc)      (sc->flags & BCM_ONE_PORT_FLAG)
+#define BCM_NOFCOE(sc)        (sc->flags & BCM_NO_FCOE_FLAG)
+#define BCM_NOMCP(sc)         (sc->flags & BCM_NO_MCP_FLAG)
+
+#define MAX_BARS 5
+	struct bcm_bar bar[MAX_BARS]; /* map BARs 0, 2, 4 */
+
+	uint16_t doorbell_size;
+
+	/* periodic timer callout */
+#define PERIODIC_STOP 0
+#define PERIODIC_GO   1
+	volatile unsigned long periodic_flags;
+
+	struct bcm_fastpath fp[MAX_RSS_CHAINS];
+	struct bcm_sp_objs  sp_objs[MAX_RSS_CHAINS];
+
+	uint8_t  unit; /* driver instance number */
+
+	int pcie_bus;    /* PCIe bus number */
+	int pcie_device; /* PCIe device/slot number */
+	int pcie_func;   /* PCIe function number */
+
+	uint8_t pfunc_rel; /* function relative */
+	uint8_t pfunc_abs; /* function absolute */
+	uint8_t path_id;   /* function absolute */
+#define SC_PATH(sc)     (sc->path_id)
+#define SC_PORT(sc)     (sc->pfunc_rel & 1)
+#define SC_FUNC(sc)     (sc->pfunc_rel)
+#define SC_ABS_FUNC(sc) (sc->pfunc_abs)
+#define SC_VN(sc)       (sc->pfunc_rel >> 1)
+#define SC_L_ID(sc)     (SC_VN(sc) << 2)
+#define PORT_ID(sc)     SC_PORT(sc)
+#define PATH_ID(sc)     SC_PATH(sc)
+#define VNIC_ID(sc)     SC_VN(sc)
+#define FUNC_ID(sc)     SC_FUNC(sc)
+#define ABS_FUNC_ID(sc) SC_ABS_FUNC(sc)
+#define SC_FW_MB_IDX_VN(sc, vn)                                \
+	(SC_PORT(sc) + (vn) *                                      \
+	 ((CHIP_IS_E1x(sc) || (CHIP_IS_MODE_4_PORT(sc))) ? 2 : 1))
+#define SC_FW_MB_IDX(sc) SC_FW_MB_IDX_VN(sc, SC_VN(sc))
+
+	int if_capen; /* enabled interface capabilities */
+
+	struct bcm_devinfo devinfo;
+	char fw_ver_str[32];
+	char mf_mode_str[32];
+	char pci_link_str[32];
+
+	struct iro *iro_array;
+
+	int dmae_ready;
+#define DMAE_READY(sc) (sc->dmae_ready)
+
+	struct ecore_credit_pool_obj vlans_pool;
+	struct ecore_credit_pool_obj macs_pool;
+	struct ecore_rx_mode_obj     rx_mode_obj;
+	struct ecore_mcast_obj       mcast_obj;
+	struct ecore_rss_config_obj  rss_conf_obj;
+	struct ecore_func_sp_obj     func_obj;
+
+	uint16_t fw_seq;
+	uint16_t fw_drv_pulse_wr_seq;
+	uint32_t func_stx;
+
+	struct elink_params         link_params;
+	struct elink_vars           link_vars;
+	uint32_t                    link_cnt;
+	struct bcm_link_report_data last_reported_link;
+	char mac_addr_str[32];
+
+	uint32_t tx_ring_size;
+	uint32_t rx_ring_size;
+	int wol;
+
+	int is_leader;
+	int recovery_state;
+#define BCM_RECOVERY_DONE        1
+#define BCM_RECOVERY_INIT        2
+#define BCM_RECOVERY_WAIT        3
+#define BCM_RECOVERY_FAILED      4
+#define BCM_RECOVERY_NIC_LOADING 5
+
+	uint32_t rx_mode;
+#define BCM_RX_MODE_NONE     0
+#define BCM_RX_MODE_NORMAL   1
+#define BCM_RX_MODE_ALLMULTI 2
+#define BCM_RX_MODE_PROMISC  3
+#define BCM_MAX_MULTICAST    64
+
+	struct bcm_port port;
+
+	struct cmng_init cmng;
+
+	/* user configs */
+	uint8_t  num_queues;
+	int      hc_rx_ticks;
+	int      hc_tx_ticks;
+	uint32_t rx_budget;
+	int      interrupt_mode;
+#define INTR_MODE_INTX 0
+#define INTR_MODE_MSI  1
+#define INTR_MODE_MSIX 2
+#define INTR_MODE_SINGLE_MSIX 3
+	int      udp_rss;
+
+	uint8_t         igu_dsb_id;
+	uint8_t         igu_base_sb;
+	uint8_t         igu_sb_cnt;
+	uint32_t        igu_base_addr;
+	uint8_t         base_fw_ndsb;
+#define DEF_SB_IGU_ID 16
+#define DEF_SB_ID     HC_SP_SB_ID
+
+	/* default status block */
+	struct bcm_dma              def_sb_dma;
+	struct host_sp_status_block *def_sb;
+	uint16_t                    def_idx;
+	uint16_t                    def_att_idx;
+	uint32_t                    attn_state;
+	struct attn_route           attn_group[MAX_DYNAMIC_ATTN_GRPS];
+
+	/* general SP events - stats query, cfc delete, etc */
+#define HC_SP_INDEX_ETH_DEF_CONS         3
+	/* EQ completions */
+#define HC_SP_INDEX_EQ_CONS              7
+	/* FCoE L2 connection completions */
+#define HC_SP_INDEX_ETH_FCOE_TX_CQ_CONS  6
+#define HC_SP_INDEX_ETH_FCOE_RX_CQ_CONS  4
+	/* iSCSI L2 */
+#define HC_SP_INDEX_ETH_ISCSI_CQ_CONS    5
+#define HC_SP_INDEX_ETH_ISCSI_RX_CQ_CONS 1
+
+	/* event queue */
+	struct bcm_dma        eq_dma;
+	union event_ring_elem *eq;
+	uint16_t              eq_prod;
+	uint16_t              eq_cons;
+	uint16_t              *eq_cons_sb;
+#define NUM_EQ_PAGES     1 /* must be a power of 2 */
+#define EQ_DESC_CNT_PAGE (BCM_PAGE_SIZE / sizeof(union event_ring_elem))
+#define EQ_DESC_MAX_PAGE (EQ_DESC_CNT_PAGE - 1)
+#define NUM_EQ_DESC      (EQ_DESC_CNT_PAGE * NUM_EQ_PAGES)
+#define EQ_DESC_MASK     (NUM_EQ_DESC - 1)
+#define MAX_EQ_AVAIL     (EQ_DESC_MAX_PAGE * NUM_EQ_PAGES - 2)
+	/* depends on EQ_DESC_CNT_PAGE being a power of 2 */
+#define NEXT_EQ_IDX(x)                                      \
+	((((x) & EQ_DESC_MAX_PAGE) == (EQ_DESC_MAX_PAGE - 1)) ? \
+	 ((x) + 2) : ((x) + 1))
+	/* depends on the above and on NUM_EQ_PAGES being a power of 2 */
+#define EQ_DESC(x) ((x) & EQ_DESC_MASK)
+
+	/* slow path */
+	struct bcm_dma      sp_dma;
+	struct bcm_slowpath *sp;
+	unsigned long       sp_state;
+
+	/* slow path queue */
+	struct bcm_dma spq_dma;
+	struct eth_spe *spq;
+#define SP_DESC_CNT     (BCM_PAGE_SIZE / sizeof(struct eth_spe))
+#define MAX_SP_DESC_CNT (SP_DESC_CNT - 1)
+#define MAX_SPQ_PENDING 8
+
+	uint16_t       spq_prod_idx;
+	struct eth_spe *spq_prod_bd;
+	struct eth_spe *spq_last_bd;
+	uint16_t       *dsb_sp_prod;
+
+	volatile unsigned long eq_spq_left; /* COMMON_xxx ramrod credit */
+	volatile unsigned long cq_spq_left; /* ETH_xxx ramrod credit */
+
+	/* fw decompression buffer */
+	struct bcm_dma gz_buf_dma;
+	void           *gz_buf;
+	uint32_t       gz_outlen;
+#define GUNZIP_BUF(sc)    (sc->gz_buf)
+#define GUNZIP_OUTLEN(sc) (sc->gz_outlen)
+#define GUNZIP_PHYS(sc)   (phys_addr_t)((void *)(sc->gz_buf_dma.paddr))
+#define FW_BUF_SIZE       0x40000
+
+	struct raw_op *init_ops;
+	uint16_t *init_ops_offsets; /* init block offsets inside init_ops */
+	uint32_t *init_data;        /* data blob, 32 bit granularity */
+	uint32_t       init_mode_flags;
+#define INIT_MODE_FLAGS(sc) (sc->init_mode_flags)
+	/* PRAM blobs - raw data */
+	const uint8_t *tsem_int_table_data;
+	const uint8_t *tsem_pram_data;
+	const uint8_t *usem_int_table_data;
+	const uint8_t *usem_pram_data;
+	const uint8_t *xsem_int_table_data;
+	const uint8_t *xsem_pram_data;
+	const uint8_t *csem_int_table_data;
+	const uint8_t *csem_pram_data;
+#define INIT_OPS(sc)                 (sc->init_ops)
+#define INIT_OPS_OFFSETS(sc)         (sc->init_ops_offsets)
+#define INIT_DATA(sc)                (sc->init_data)
+#define INIT_TSEM_INT_TABLE_DATA(sc) (sc->tsem_int_table_data)
+#define INIT_TSEM_PRAM_DATA(sc)      (sc->tsem_pram_data)
+#define INIT_USEM_INT_TABLE_DATA(sc) (sc->usem_int_table_data)
+#define INIT_USEM_PRAM_DATA(sc)      (sc->usem_pram_data)
+#define INIT_XSEM_INT_TABLE_DATA(sc) (sc->xsem_int_table_data)
+#define INIT_XSEM_PRAM_DATA(sc)      (sc->xsem_pram_data)
+#define INIT_CSEM_INT_TABLE_DATA(sc) (sc->csem_int_table_data)
+#define INIT_CSEM_PRAM_DATA(sc)      (sc->csem_pram_data)
+
+#define PHY_FW_VER_LEN			20
+	char			fw_ver[32];
+
+	/* ILT
+	 * For max 196 cids (64*3 + non-eth), 32KB ILT page size and 1KB
+	 * context size we need 8 ILT entries.
+	 */
+#define ILT_MAX_L2_LINES 8
+	struct hw_context context[ILT_MAX_L2_LINES];
+	struct ecore_ilt *ilt;
+#define ILT_MAX_LINES 256
+
+	/* max supported number of RSS queues: IGU SBs minus one for CNIC */
+#define BCM_MAX_RSS_COUNT(sc) ((sc)->igu_sb_cnt - CNIC_SUPPORT(sc))
+	/* max CID count: Max RSS * Max_Tx_Multi_Cos + FCoE + iSCSI */
+#define BCM_L2_MAX_CID(sc)                                              \
+	(BCM_MAX_RSS_COUNT(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
+#define BCM_L2_CID_COUNT(sc)                                             \
+	(BCM_NUM_ETH_QUEUES(sc) * ECORE_MULTI_TX_COS + 2 * CNIC_SUPPORT(sc))
+#define L2_ILT_LINES(sc)                                \
+	(DIV_ROUND_UP(BCM_L2_CID_COUNT(sc), ILT_PAGE_CIDS))
+
+	int qm_cid_count;
+
+	uint8_t dropless_fc;
+
+	/* total number of FW statistics requests */
+	uint8_t fw_stats_num;
+	/*
+	 * This is a memory buffer that will contain both statistics ramrod
+	 * request and data.
+	 */
+	struct bcm_dma fw_stats_dma;
+	/*
+	 * FW statistics request shortcut (points at the beginning of fw_stats
+	 * buffer).
+	 */
+	int                     fw_stats_req_size;
+	struct bcm_fw_stats_req *fw_stats_req;
+	phys_addr_t              fw_stats_req_mapping;
+	/*
+	 * FW statistics data shortcut (points at the beginning of fw_stats
+	 * buffer + fw_stats_req_size).
+	 */
+	int                      fw_stats_data_size;
+	struct bcm_fw_stats_data *fw_stats_data;
+	phys_addr_t               fw_stats_data_mapping;
+
+	/* tracking a pending STAT_QUERY ramrod */
+	uint16_t stats_pending;
+	/* number of completed statistics ramrods */
+	uint16_t stats_comp;
+	uint16_t stats_counter;
+	uint8_t  stats_init;
+	int      stats_state;
+
+	struct bcm_eth_stats         eth_stats;
+	struct host_func_stats       func_stats;
+	struct bcm_eth_stats_old     eth_stats_old;
+	struct bcm_net_stats_old     net_stats_old;
+	struct bcm_fw_port_stats_old fw_stats_old;
+
+	struct dmae_command stats_dmae; /* used by dmae command loader */
+	int                 executer_idx;
+
+	int mtu;
+
+	/* DCB support on/off */
+	int dcb_state;
+#define BCM_DCB_STATE_OFF 0
+#define BCM_DCB_STATE_ON  1
+	/* DCBX engine mode */
+	int dcbx_enabled;
+#define BCM_DCBX_ENABLED_OFF        0
+#define BCM_DCBX_ENABLED_ON_NEG_OFF 1
+#define BCM_DCBX_ENABLED_ON_NEG_ON  2
+#define BCM_DCBX_ENABLED_INVALID    -1
+
+	uint8_t cnic_support;
+	uint8_t cnic_enabled;
+	uint8_t cnic_loaded;
+#define CNIC_SUPPORT(sc) 0 /* ((sc)->cnic_support) */
+#define CNIC_ENABLED(sc) 0 /* ((sc)->cnic_enabled) */
+#define CNIC_LOADED(sc)  0 /* ((sc)->cnic_loaded) */
+
+	/* multiple tx classes of service */
+	uint8_t max_cos;
+#define BCM_MAX_PRIORITY 8
+	/* priority to cos mapping */
+	uint8_t prio_to_cos[BCM_MAX_PRIORITY];
+
+	int panic;
+}; /* struct bcm_softc */
+
+/* IOCTL sub-commands for edebug and firmware upgrade */
+#define BCM_IOC_RD_NVRAM        1
+#define BCM_IOC_WR_NVRAM        2
+#define BCM_IOC_STATS_SHOW_NUM  3
+#define BCM_IOC_STATS_SHOW_STR  4
+#define BCM_IOC_STATS_SHOW_CNT  5
+
+struct bcm_nvram_data {
+    uint32_t op; /* ioctl sub-command */
+    uint32_t offset;
+    uint32_t len;
+    uint32_t value[1]; /* variable */
+};
+
+union bcm_stats_show_data {
+    uint32_t op; /* ioctl sub-command */
+
+    struct {
+        uint32_t num; /* return number of stats */
+        uint32_t len; /* length of each string item */
+    } desc;
+
+    /* variable length... */
+    char str[1]; /* holds names of desc.num stats, each desc.len in length */
+
+    /* variable length... */
+    uint64_t stats[1]; /* holds all stats */
+};
+
+/* function init flags */
+#define FUNC_FLG_RSS     0x0001
+#define FUNC_FLG_STATS   0x0002
+/* FUNC_FLG_UNMATCHED       0x0004 */
+#define FUNC_FLG_SPQ     0x0010
+#define FUNC_FLG_LEADING 0x0020 /* PF only */
+
+struct bcm_func_init_params {
+    phys_addr_t fw_stat_map; /* (dma) valid if FUNC_FLG_STATS */
+    phys_addr_t spq_map;     /* (dma) valid if FUNC_FLG_SPQ */
+    uint16_t   func_flgs;
+    uint16_t   func_id;     /* abs function id */
+    uint16_t   pf_id;
+    uint16_t   spq_prod;    /* valid if FUNC_FLG_SPQ */
+};
+
+/* memory resources reside at BARs 0, 2, 4 */
+/* Run `pciconf -lb` to see mappings */
+#define BAR0 0
+#define BAR1 2
+#define BAR2 4
+
+#ifdef RTE_LIBRTE_BCM_DEBUG
+uint8_t bcm_reg_read8(struct bcm_softc *sc, size_t offset);
+uint16_t bcm_reg_read16(struct bcm_softc *sc, size_t offset);
+uint32_t bcm_reg_read32(struct bcm_softc *sc, size_t offset);
+
+void bcm_reg_write8(struct bcm_softc *sc, size_t offset, uint8_t val);
+void bcm_reg_write16(struct bcm_softc *sc, size_t offset, uint16_t val);
+void bcm_reg_write32(struct bcm_softc *sc, size_t offset, uint32_t val);
+#else
+#define bcm_reg_write8(sc, offset, val)\
+	*((volatile uint8_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)) = val
+
+#define bcm_reg_write16(sc, offset, val)\
+	*((volatile uint16_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)) = val
+
+#define bcm_reg_write32(sc, offset, val)\
+	*((volatile uint32_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)) = val
+
+#define bcm_reg_read8(sc, offset)\
+	(*((volatile uint8_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)))
+
+#define bcm_reg_read16(sc, offset)\
+	(*((volatile uint16_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)))
+
+#define bcm_reg_read32(sc, offset)\
+	(*((volatile uint32_t*)((uint64_t)sc->bar[BAR0].base_addr + offset)))
+#endif
+
+#define REG_ADDR(sc, offset) (((uint64_t)sc->bar[BAR0].base_addr) + (offset))
+
+#define REG_RD8(sc, offset)  bcm_reg_read8(sc, (offset))
+#define REG_RD16(sc, offset) bcm_reg_read16(sc, (offset))
+#define REG_RD32(sc, offset) bcm_reg_read32(sc, (offset))
+
+#define REG_WR8(sc, offset, val)  bcm_reg_write8(sc, (offset), val)
+#define REG_WR16(sc, offset, val) bcm_reg_write16(sc, (offset), val)
+#define REG_WR32(sc, offset, val) bcm_reg_write32(sc, (offset), val)
+
+#define REG_RD(sc, offset)      REG_RD32(sc, offset)
+#define REG_WR(sc, offset, val) REG_WR32(sc, offset, val)
+
+#define BCM_SP(sc, var) (&(sc)->sp->var)
+#define BCM_SP_MAPPING(sc, var) \
+    (sc->sp_dma.paddr + offsetof(struct bcm_slowpath, var))
+
+#define BCM_FP(sc, nr, var) ((sc)->fp[(nr)].var)
+#define BCM_SP_OBJ(sc, fp) ((sc)->sp_objs[(fp)->index])
+
+#define bcm_fp(sc, nr, var)   ((sc)->fp[nr].var)
+
+#define REG_RD_DMAE(sc, offset, valp, len32)               \
+    do {                                                   \
+        (void)bcm_read_dmae(sc, offset, len32);                  \
+        (void)rte_memcpy(valp, BCM_SP(sc, wb_data[0]), (len32) * 4); \
+    } while (0)
+
+#define REG_WR_DMAE(sc, offset, valp, len32)                            \
+    do {                                                                \
+        (void)rte_memcpy(BCM_SP(sc, wb_data[0]), valp, (len32) * 4);              \
+        (void)bcm_write_dmae(sc, BCM_SP_MAPPING(sc, wb_data), offset, len32); \
+    } while (0)
+
+#define REG_WR_DMAE_LEN(sc, offset, valp, len32) \
+    REG_WR_DMAE(sc, offset, valp, len32)
+
+#define REG_RD_DMAE_LEN(sc, offset, valp, len32) \
+    REG_RD_DMAE(sc, offset, valp, len32)
+
+#define VIRT_WR_DMAE_LEN(sc, data, addr, len32, le32_swap)         \
+    do {                                                           \
+        /* if (le32_swap) {                                     */ \
+        /*    PMD_PWARN_LOG(sc, "VIRT_WR_DMAE_LEN with le32_swap=1"); */ \
+        /* }                                                    */ \
+        rte_memcpy(GUNZIP_BUF(sc), data, len32 * 4);                   \
+        ecore_write_big_buf_wb(sc, addr, len32);                   \
+    } while (0)
+
+#define BCM_DB_MIN_SHIFT 3   /* 8 bytes */
+#define BCM_DB_SHIFT     7   /* 128 bytes */
+#if (BCM_DB_SHIFT < BCM_DB_MIN_SHIFT)
+#error "Minimum DB doorbell stride is 8"
+#endif
+#define DPM_TRIGGER_TYPE 0x40
+
+/* Doorbell macro */
+#define BCM_DB_WRITE(db_bar, val) \
+	*((volatile uint32_t *)(db_bar)) = (val)
+
+#define BCM_DB_READ(db_bar) \
+	*((volatile uint32_t *)(db_bar))
+
+#define DOORBELL_ADDR(sc, offset) \
+	(volatile uint32_t *)(((char *)(sc)->bar[BAR1].base_addr + (offset)))
+
+#define DOORBELL(sc, cid, val) \
+	if (IS_PF(sc)) \
+	BCM_DB_WRITE((DOORBELL_ADDR(sc, sc->doorbell_size * (cid) + DPM_TRIGGER_TYPE)), (val)); \
+	else \
+	BCM_DB_WRITE((DOORBELL_ADDR(sc, sc->doorbell_size * (cid))), (val)) \
+
+#define SHMEM_ADDR(sc, field)                                       \
+    (sc->devinfo.shmem_base + offsetof(struct shmem_region, field))
+#define SHMEM_RD(sc, field)      REG_RD(sc, SHMEM_ADDR(sc, field))
+#define SHMEM_RD16(sc, field)    REG_RD16(sc, SHMEM_ADDR(sc, field))
+#define SHMEM_WR(sc, field, val) REG_WR(sc, SHMEM_ADDR(sc, field), val)
+
+#define SHMEM2_ADDR(sc, field)                                        \
+    (sc->devinfo.shmem2_base + offsetof(struct shmem2_region, field))
+#define SHMEM2_HAS(sc, field)                                            \
+    (sc->devinfo.shmem2_base && (REG_RD(sc, SHMEM2_ADDR(sc, size)) >     \
+                                 offsetof(struct shmem2_region, field)))
+#define SHMEM2_RD(sc, field)      REG_RD(sc, SHMEM2_ADDR(sc, field))
+#define SHMEM2_WR(sc, field, val) REG_WR(sc, SHMEM2_ADDR(sc, field), val)
+
+#define MFCFG_ADDR(sc, field)                                  \
+    (sc->devinfo.mf_cfg_base + offsetof(struct mf_cfg, field))
+#define MFCFG_RD(sc, field)      REG_RD(sc, MFCFG_ADDR(sc, field))
+#define MFCFG_RD16(sc, field)    REG_RD16(sc, MFCFG_ADDR(sc, field))
+#define MFCFG_WR(sc, field, val) REG_WR(sc, MFCFG_ADDR(sc, field), val)
+
+/* DMAE command defines */
+
+#define DMAE_TIMEOUT      -1
+#define DMAE_PCI_ERROR    -2 /* E2 and onward */
+#define DMAE_NOT_RDY      -3
+#define DMAE_PCI_ERR_FLAG 0x80000000
+
+#define DMAE_SRC_PCI      0
+#define DMAE_SRC_GRC      1
+
+#define DMAE_DST_NONE     0
+#define DMAE_DST_PCI      1
+#define DMAE_DST_GRC      2
+
+#define DMAE_COMP_PCI     0
+#define DMAE_COMP_GRC     1
+
+#define DMAE_COMP_REGULAR 0
+#define DMAE_COM_SET_ERR  1
+
+#define DMAE_CMD_SRC_PCI (DMAE_SRC_PCI << DMAE_COMMAND_SRC_SHIFT)
+#define DMAE_CMD_SRC_GRC (DMAE_SRC_GRC << DMAE_COMMAND_SRC_SHIFT)
+#define DMAE_CMD_DST_PCI (DMAE_DST_PCI << DMAE_COMMAND_DST_SHIFT)
+#define DMAE_CMD_DST_GRC (DMAE_DST_GRC << DMAE_COMMAND_DST_SHIFT)
+
+#define DMAE_CMD_C_DST_PCI (DMAE_COMP_PCI << DMAE_COMMAND_C_DST_SHIFT)
+#define DMAE_CMD_C_DST_GRC (DMAE_COMP_GRC << DMAE_COMMAND_C_DST_SHIFT)
+
+#define DMAE_CMD_ENDIANITY_NO_SWAP   (0 << DMAE_COMMAND_ENDIANITY_SHIFT)
+#define DMAE_CMD_ENDIANITY_B_SWAP    (1 << DMAE_COMMAND_ENDIANITY_SHIFT)
+#define DMAE_CMD_ENDIANITY_DW_SWAP   (2 << DMAE_COMMAND_ENDIANITY_SHIFT)
+#define DMAE_CMD_ENDIANITY_B_DW_SWAP (3 << DMAE_COMMAND_ENDIANITY_SHIFT)
+
+#define DMAE_CMD_PORT_0 0
+#define DMAE_CMD_PORT_1 DMAE_COMMAND_PORT
+
+#define DMAE_SRC_PF 0
+#define DMAE_SRC_VF 1
+
+#define DMAE_DST_PF 0
+#define DMAE_DST_VF 1
+
+#define DMAE_C_SRC 0
+#define DMAE_C_DST 1
+
+#define DMAE_LEN32_RD_MAX     0x80
+#define DMAE_LEN32_WR_MAX(sc) 0x2000
+
+#define DMAE_COMP_VAL 0x60d0d0ae /* E2 and beyond, upper bit indicates error */
+
+#define MAX_DMAE_C_PER_PORT 8
+#define INIT_DMAE_C(sc)     ((SC_PORT(sc) * MAX_DMAE_C_PER_PORT) + SC_VN(sc))
+#define PMF_DMAE_C(sc)      ((SC_PORT(sc) * MAX_DMAE_C_PER_PORT) + E1HVN_MAX)
+
+static const uint32_t dmae_reg_go_c[] = {
+    DMAE_REG_GO_C0,  DMAE_REG_GO_C1,  DMAE_REG_GO_C2,  DMAE_REG_GO_C3,
+    DMAE_REG_GO_C4,  DMAE_REG_GO_C5,  DMAE_REG_GO_C6,  DMAE_REG_GO_C7,
+    DMAE_REG_GO_C8,  DMAE_REG_GO_C9,  DMAE_REG_GO_C10, DMAE_REG_GO_C11,
+    DMAE_REG_GO_C12, DMAE_REG_GO_C13, DMAE_REG_GO_C14, DMAE_REG_GO_C15
+};
+
+#define ATTN_NIG_FOR_FUNC     (1L << 8)
+#define ATTN_SW_TIMER_4_FUNC  (1L << 9)
+#define GPIO_2_FUNC           (1L << 10)
+#define GPIO_3_FUNC           (1L << 11)
+#define GPIO_4_FUNC           (1L << 12)
+#define ATTN_GENERAL_ATTN_1   (1L << 13)
+#define ATTN_GENERAL_ATTN_2   (1L << 14)
+#define ATTN_GENERAL_ATTN_3   (1L << 15)
+#define ATTN_GENERAL_ATTN_4   (1L << 13)
+#define ATTN_GENERAL_ATTN_5   (1L << 14)
+#define ATTN_GENERAL_ATTN_6   (1L << 15)
+#define ATTN_HARD_WIRED_MASK  0xff00
+#define ATTENTION_ID          4
+
+#define AEU_IN_ATTN_BITS_PXPPCICLOCKCLIENT_PARITY_ERROR \
+    AEU_INPUTS_ATTN_BITS_PXPPCICLOCKCLIENT_PARITY_ERROR
+
+#define MAX_IGU_ATTN_ACK_TO 100
+
+#define STORM_ASSERT_ARRAY_SIZE 50
+
+#define BCM_PMF_LINK_ASSERT(sc) \
+    GENERAL_ATTEN_OFFSET(LINK_SYNC_ATTENTION_BIT_FUNC_0 + SC_FUNC(sc))
+
+#define BCM_MC_ASSERT_BITS \
+    (GENERAL_ATTEN_OFFSET(TSTORM_FATAL_ASSERT_ATTENTION_BIT) | \
+     GENERAL_ATTEN_OFFSET(USTORM_FATAL_ASSERT_ATTENTION_BIT) | \
+     GENERAL_ATTEN_OFFSET(CSTORM_FATAL_ASSERT_ATTENTION_BIT) | \
+     GENERAL_ATTEN_OFFSET(XSTORM_FATAL_ASSERT_ATTENTION_BIT))
+
+#define BCM_MCP_ASSERT \
+    GENERAL_ATTEN_OFFSET(MCP_FATAL_ASSERT_ATTENTION_BIT)
+
+#define BCM_GRC_TIMEOUT GENERAL_ATTEN_OFFSET(LATCHED_ATTN_TIMEOUT_GRC)
+#define BCM_GRC_RSV     (GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RBCR) | \
+                         GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RBCT) | \
+                         GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RBCN) | \
+                         GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RBCU) | \
+                         GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RBCP) | \
+                         GENERAL_ATTEN_OFFSET(LATCHED_ATTN_RSVD_GRC))
+
+#define MULTI_MASK 0x7f
+
+#define PFS_PER_PORT(sc)                               \
+    ((CHIP_PORT_MODE(sc) == CHIP_4_PORT_MODE) ? 2 : 4)
+#define SC_MAX_VN_NUM(sc) PFS_PER_PORT(sc)
+
+#define FIRST_ABS_FUNC_IN_PORT(sc)                    \
+    ((CHIP_PORT_MODE(sc) == CHIP_PORT_MODE_NONE) ?    \
+     PORT_ID(sc) : (PATH_ID(sc) + (2 * PORT_ID(sc))))
+
+#define FOREACH_ABS_FUNC_IN_PORT(sc, i)            \
+    for ((i) = FIRST_ABS_FUNC_IN_PORT(sc);         \
+         (i) < MAX_FUNC_NUM;                       \
+         (i) += (MAX_FUNC_NUM / PFS_PER_PORT(sc)))
+
+#define BCM_SWCID_SHIFT 17
+#define BCM_SWCID_MASK  ((0x1 << BCM_SWCID_SHIFT) - 1)
+
+#define SW_CID(x)  (le32toh(x) & BCM_SWCID_MASK)
+#define CQE_CMD(x) (le32toh(x) >> COMMON_RAMROD_ETH_RX_CQE_CMD_ID_SHIFT)
+
+#define CQE_TYPE(cqe_fp_flags)   ((cqe_fp_flags) & ETH_FAST_PATH_RX_CQE_TYPE)
+#define CQE_TYPE_START(cqe_type) ((cqe_type) == RX_ETH_CQE_TYPE_ETH_START_AGG)
+#define CQE_TYPE_STOP(cqe_type)  ((cqe_type) == RX_ETH_CQE_TYPE_ETH_STOP_AGG)
+#define CQE_TYPE_SLOW(cqe_type)  ((cqe_type) == RX_ETH_CQE_TYPE_ETH_RAMROD)
+#define CQE_TYPE_FAST(cqe_type)  ((cqe_type) == RX_ETH_CQE_TYPE_ETH_FASTPATH)
+
+/* must be used on a CID before placing it on a HW ring */
+#define HW_CID(sc, x) \
+    ((SC_PORT(sc) << 23) | (SC_VN(sc) << BCM_SWCID_SHIFT) | (x))
+
+#define SPEED_10    10
+#define SPEED_100   100
+#define SPEED_1000  1000
+#define SPEED_2500  2500
+#define SPEED_10000 10000
+
+#define PCI_PM_D0    1
+#define PCI_PM_D3hot 2
+
+int  bcm_test_bit(int nr, volatile unsigned long * addr);
+void bcm_set_bit(unsigned int nr, volatile unsigned long * addr);
+void bcm_clear_bit(int nr, volatile unsigned long * addr);
+int  bcm_test_and_clear_bit(int nr, volatile unsigned long * addr);
+int  bcm_cmpxchg(volatile int *addr, int old, int new);
+
+int bcm_dma_alloc(struct bcm_softc *sc, size_t size,
+		struct bcm_dma *dma, const char *msg, uint32_t align);
+
+uint32_t bcm_dmae_opcode_add_comp(uint32_t opcode, uint8_t comp_type);
+uint32_t bcm_dmae_opcode_clr_src_reset(uint32_t opcode);
+uint32_t bcm_dmae_opcode(struct bcm_softc *sc, uint8_t src_type,
+                         uint8_t dst_type, uint8_t with_comp,
+                         uint8_t comp_type);
+void bcm_post_dmae(struct bcm_softc *sc, struct dmae_command *dmae, int idx);
+void bcm_read_dmae(struct bcm_softc *sc, uint32_t src_addr, uint32_t len32);
+void bcm_write_dmae(struct bcm_softc *sc, phys_addr_t dma_addr,
+                    uint32_t dst_addr, uint32_t len32);
+void bcm_set_ctx_validation(struct bcm_softc *sc, struct eth_context *cxt,
+                            uint32_t cid);
+void bcm_update_coalesce_sb_index(struct bcm_softc *sc, uint8_t fw_sb_id,
+                                  uint8_t sb_index, uint8_t disable,
+                                  uint16_t usec);
+
+int bcm_sp_post(struct bcm_softc *sc, int command, int cid,
+                uint32_t data_hi, uint32_t data_lo, int cmd_type);
+
+void ecore_init_e1h_firmware(struct bcm_softc *sc);
+void ecore_init_e2_firmware(struct bcm_softc *sc);
+
+void ecore_storm_memset_struct(struct bcm_softc *sc, uint32_t addr,
+                               size_t size, uint32_t *data);
+
+#define CATC_TRIGGER(sc, data) REG_WR((sc), 0x2000, (data));
+#define CATC_TRIGGER_START(sc) CATC_TRIGGER((sc), 0xcafecafe)
+
+#define BCM_MAC_FMT		"%pM"
+#define BCM_MAC_PRN_LIST(mac)	(mac)
+
+/***********/
+/* INLINES */
+/***********/
+
+static inline uint32_t
+reg_poll(struct bcm_softc *sc, uint32_t reg, uint32_t expected, int ms, int wait)
+{
+    uint32_t val;
+    do {
+        val = REG_RD(sc, reg);
+        if (val == expected) {
+            break;
+        }
+        ms -= wait;
+        DELAY(wait * 1000);
+    } while (ms > 0);
+
+    return val;
+}
+
+static inline void
+bcm_update_fp_sb_idx(struct bcm_fastpath *fp)
+{
+	mb(); /* status block is written to by the chip */
+	fp->fp_hc_idx = fp->sb_running_index[SM_RX_ID];
+}
+
+static inline void
+bcm_igu_ack_sb_gen(struct bcm_softc *sc, uint8_t segment,
+	uint16_t index, uint8_t op, uint8_t update, uint32_t igu_addr)
+{
+	struct igu_regular cmd_data = {0};
+
+	cmd_data.sb_id_and_flags =
+		((index << IGU_REGULAR_SB_INDEX_SHIFT) |
+		 (segment << IGU_REGULAR_SEGMENT_ACCESS_SHIFT) |
+		 (update << IGU_REGULAR_BUPDATE_SHIFT) |
+		 (op << IGU_REGULAR_ENABLE_INT_SHIFT));
+
+	REG_WR(sc, igu_addr, cmd_data.sb_id_and_flags);
+
+	/* Make sure that ACK is written */
+	mb();
+}
+
+static inline void
+bcm_hc_ack_sb(struct bcm_softc *sc, uint8_t sb_id, uint8_t storm,
+		uint16_t index, uint8_t op, uint8_t update)
+{
+	uint32_t hc_addr = (HC_REG_COMMAND_REG + SC_PORT(sc) * 32 +
+			COMMAND_REG_INT_ACK);
+	union igu_ack_register igu_ack;
+
+	igu_ack.sb.status_block_index = index;
+	igu_ack.sb.sb_id_and_flags =
+		((sb_id << IGU_ACK_REGISTER_STATUS_BLOCK_ID_SHIFT) |
+		 (storm << IGU_ACK_REGISTER_STORM_ID_SHIFT) |
+		 (update << IGU_ACK_REGISTER_UPDATE_INDEX_SHIFT) |
+		 (op << IGU_ACK_REGISTER_INTERRUPT_MODE_SHIFT));
+
+	REG_WR(sc, hc_addr, igu_ack.raw_data);
+
+	/* Make sure that ACK is written */
+	mb();
+}
+
+static inline uint32_t
+bcm_hc_ack_int(struct bcm_softc *sc)
+{
+	uint32_t hc_addr = (HC_REG_COMMAND_REG + SC_PORT(sc) * 32 +
+			COMMAND_REG_SIMD_MASK);
+	uint32_t result = REG_RD(sc, hc_addr);
+
+	mb();
+	return result;
+}
+
+static inline uint32_t
+bcm_igu_ack_int(struct bcm_softc *sc)
+{
+	uint32_t igu_addr = (BAR_IGU_INTMEM + IGU_REG_SISR_MDPC_WMASK_LSB_UPPER * 8);
+	uint32_t result = REG_RD(sc, igu_addr);
+
+	/* PMD_PDEBUG_LOG(sc, DBG_INTR, "read 0x%08x from IGU addr 0x%x",
+			result, igu_addr); */
+
+	mb();
+	return result;
+}
+
+static inline uint32_t
+bcm_ack_int(struct bcm_softc *sc)
+{
+	mb();
+	if (sc->devinfo.int_block == INT_BLOCK_HC) {
+		return bcm_hc_ack_int(sc);
+	} else {
+		return bcm_igu_ack_int(sc);
+	}
+}
+
+static inline int
+func_by_vn(struct bcm_softc *sc, int vn)
+{
+    return (2 * vn + SC_PORT(sc));
+}
+
+/*
+ * send notification to other functions.
+ */
+static inline void
+bcm_link_sync_notify(struct bcm_softc *sc)
+{
+	int func, vn;
+
+	/* Set the attention towards other drivers on the same port */
+	for (vn = VN_0; vn < SC_MAX_VN_NUM(sc); vn++) {
+		if (vn == SC_VN(sc))
+			continue;
+
+		func = func_by_vn(sc, vn);
+		REG_WR(sc, MISC_REG_AEU_GENERAL_ATTN_0 +
+				(LINK_SYNC_ATTENTION_BIT_FUNC_0 + func) * 4, 1);
+	}
+}
+
+/*
+ * Statistics ID are global per chip/path, while Client IDs for E1x
+ * are per port.
+ */
+static inline uint8_t
+bcm_stats_id(struct bcm_fastpath *fp)
+{
+    struct bcm_softc *sc = fp->sc;
+
+    if (!CHIP_IS_E1x(sc)) {
+        return fp->cl_id;
+    }
+
+    return (fp->cl_id + SC_PORT(sc) * FP_SB_MAX_E1x);
+}
+
+int bcm_init(struct bcm_softc *sc);
+void bcm_load_firmware(struct bcm_softc *sc);
+int bcm_attach(struct bcm_softc *sc);
+int bcm_nic_unload(struct bcm_softc *sc, uint32_t unload_mode, uint8_t keep_link);
+int bcm_alloc_hsi_mem(struct bcm_softc *sc);
+int bcm_alloc_ilt_mem(struct bcm_softc *sc);
+void bcm_free_ilt_mem(struct bcm_softc *sc);
+void bcm_dump_tx_chain(struct bcm_fastpath * fp, int bd_prod, int count);
+int bcm_tx_encap(struct bcm_tx_queue *txq, struct rte_mbuf **m_head, int m_pkts);
+uint8_t bcm_txeof(struct bcm_softc *sc, struct bcm_fastpath *fp);
+void bcm_print_adapter_info(struct bcm_softc *sc);
+int bcm_intr_legacy(struct bcm_softc *sc, int scan_fp);
+void bcm_link_status_update(struct bcm_softc *sc);
+int bcm_complete_sp(struct bcm_softc *sc);
+int bcm_set_storm_rx_mode(struct bcm_softc *sc);
+void bcm_periodic_callout(struct bcm_softc *sc);
+
+int bcm_vf_get_resources(struct bcm_softc *sc, uint8_t tx_count, uint8_t rx_count);
+void bcm_vf_close(struct bcm_softc *sc);
+int bcm_vf_init(struct bcm_softc *sc);
+void bcm_vf_unload(struct bcm_softc *sc);
+int bcm_vf_setup_queue(struct bcm_softc *sc, struct bcm_fastpath *fp,
+	int leading);
+void bcm_free_hsi_mem(struct bcm_softc *sc);
+int bcm_vf_set_rx_mode(struct bcm_softc *sc);
+int bcm_fill_accept_flags(struct bcm_softc *sc, uint32_t rx_mode,
+	unsigned long *rx_accept_flags, unsigned long *tx_accept_flags);
+int bcm_check_bull(struct bcm_softc *sc);
+
+//#define BCM_PULSE
+
+#define BCM_PCI_CAP  1
+#define BCM_PCI_ECAP 2
+
+static inline struct bcm_pci_cap*
+pci_find_cap(struct bcm_softc *sc, uint8_t id, uint8_t type)
+{
+	struct bcm_pci_cap *cap = sc->pci_caps;
+
+	while (cap) {
+		if (cap->id == id && cap->type == type)
+			return cap;
+		cap = cap->next;
+	}
+
+	return NULL;
+}
+
+static inline int is_valid_ether_addr(uint8_t *addr)
+{
+	if (!(addr[0] | addr[1] | addr[2] | addr[3] | addr[4] | addr[5]))
+		return 0;
+	else
+		return 1;
+}
+
+static inline void
+bcm_set_rx_mode(struct bcm_softc *sc)
+{
+	if (sc->state == BCM_STATE_OPEN) {
+		if (IS_PF(sc)) {
+			bcm_set_storm_rx_mode(sc);
+		} else {
+			sc->rx_mode = BCM_RX_MODE_PROMISC;
+			bcm_vf_set_rx_mode(sc);
+		}
+	} else {
+		PMD_DRV_LOG(NOTICE, "Card is not ready to change mode");
+	}
+}
+
+static inline int pci_read(struct bcm_softc *sc, size_t addr,
+			   void *val, uint8_t size)
+{
+	if (rte_eal_pci_read_config(sc->pci_dev, val, size, addr) <= 0) {
+		PMD_DRV_LOG(ERR, "Can't read from PCI config space");
+		return ENXIO;
+	}
+
+	return 0;
+}
+
+static inline int pci_write_word(struct bcm_softc *sc, size_t addr, off_t val)
+{
+	uint16_t val16 = val;
+
+	if (rte_eal_pci_write_config(sc->pci_dev, &val16,
+				     sizeof(val16), addr) <= 0) {
+		PMD_DRV_LOG(ERR, "Can't write to PCI config space");
+		return ENXIO;
+	}
+
+	return 0;
+}
+
+static inline int pci_write_long(struct bcm_softc *sc, size_t addr, off_t val)
+{
+	uint32_t val32 = val;
+	if (rte_eal_pci_write_config(sc->pci_dev, &val32,
+				     sizeof(val32), addr) <= 0) {
+		PMD_DRV_LOG(ERR, "Can't write to PCI config space");
+		return ENXIO;
+	}
+
+	return 0;
+}
+
+#endif /* __BCM_H__ */
diff --git a/lib/librte_pmd_bcm/bcm_ethdev.c b/lib/librte_pmd_bcm/bcm_ethdev.c
new file mode 100644
index 0000000..c56d93b
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_ethdev.c
@@ -0,0 +1,544 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#include "bcm.h"
+#include "bcm_rxtx.h"
+
+#include <rte_dev.h>
+
+/*
+ * The set of PCI devices this driver supports
+ */
+static struct rte_pci_id pci_id_bcm_map[] = {
+#define RTE_PCI_DEV_ID_DECL_BCM(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
+#include "rte_pci_dev_ids.h"
+	{ .vendor_id = 0, }
+};
+
+static struct rte_pci_id pci_id_bcmvf_map[] = {
+#define RTE_PCI_DEV_ID_DECL_BCMVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
+#include "rte_pci_dev_ids.h"
+	{ .vendor_id = 0, }
+};
+
+static void
+bcm_link_update(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+	bcm_link_status_update(sc);
+	mb();
+	dev->data->dev_link.link_speed = sc->link_vars.line_speed;
+	switch (sc->link_vars.duplex) {
+		case DUPLEX_FULL:
+			dev->data->dev_link.link_duplex = ETH_LINK_FULL_DUPLEX;
+			break;
+		case DUPLEX_HALF:
+			dev->data->dev_link.link_duplex = ETH_LINK_HALF_DUPLEX;
+			break;
+		default:
+			dev->data->dev_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
+	}
+	dev->data->dev_link.link_status = sc->link_vars.link_up;
+}
+
+static void
+bcm_interrupt_action(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+	uint32_t link_status;
+
+	PMD_DRV_LOG(INFO, "Interrupt handled");
+
+	if (bcm_intr_legacy(sc, 0))
+		DELAY_MS(250);
+	if (sc->periodic_flags & PERIODIC_GO)
+		bcm_periodic_callout(sc);
+	link_status = REG_RD(sc, sc->link_params.shmem_base +
+			offsetof(struct shmem_region,
+				port_mb[sc->link_params.port].link_status));
+	if ((link_status & LINK_STATUS_LINK_UP) != dev->data->dev_link.link_status)
+		bcm_link_update(dev);
+}
+
+static __rte_unused void
+bcm_interrupt_handler(__rte_unused struct rte_intr_handle *handle, void *param)
+{
+	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+
+	bcm_interrupt_action(dev);
+	rte_intr_enable(&(dev->pci_dev->intr_handle));
+}
+
+/*
+ * Devops - helper functions can be called from user application
+ */
+
+static int
+bcm_dev_configure(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+	int mp_ncpus = sysconf(_SC_NPROCESSORS_CONF);
+	int ret;
+
+	PMD_INIT_FUNC_TRACE();
+
+	if (dev->data->dev_conf.rxmode.jumbo_frame)
+		sc->mtu = dev->data->dev_conf.rxmode.max_rx_pkt_len;
+
+	if (dev->data->nb_tx_queues > dev->data->nb_rx_queues) {
+		PMD_DRV_LOG(ERR, "The number of TX queues is greater than number of RX queues");
+		return -EINVAL;
+	}
+
+	sc->num_queues = MAX(dev->data->nb_rx_queues, dev->data->nb_tx_queues);
+	if (sc->num_queues > mp_ncpus) {
+		PMD_DRV_LOG(ERR, "The number of queues is more than number of CPUs");
+		return -EINVAL;
+	}
+
+	PMD_DRV_LOG(DEBUG, "num_queues=%d, mtu=%d",
+		       sc->num_queues, sc->mtu);
+
+	/* allocate ilt */
+	if (bcm_alloc_ilt_mem(sc) != 0) {
+		PMD_DRV_LOG(ERR, "bcm_alloc_ilt_mem was failed");
+		return -ENXIO;
+	}
+
+	/* allocate the host hardware/software hsi structures */
+	if (bcm_alloc_hsi_mem(sc) != 0) {
+		PMD_DRV_LOG(ERR, "bcm_alloc_hsi_mem was failed");
+		bcm_free_ilt_mem(sc);
+		return -ENXIO;
+	}
+
+	if (IS_VF(sc)) {
+		if (bcm_dma_alloc(sc, sizeof(struct bcm_vf_mbx_msg),
+				  &sc->vf2pf_mbox_mapping, "vf2pf_mbox",
+				  RTE_CACHE_LINE_SIZE) != 0)
+			return -ENOMEM;
+
+		sc->vf2pf_mbox = (struct bcm_vf_mbx_msg *)sc->vf2pf_mbox_mapping.vaddr;
+		if (bcm_dma_alloc(sc, sizeof(struct bcm_vf_bulletin),
+				  &sc->pf2vf_bulletin_mapping, "vf2pf_bull",
+				  RTE_CACHE_LINE_SIZE) != 0)
+			return -ENOMEM;
+
+		sc->pf2vf_bulletin = (struct bcm_vf_bulletin *)sc->pf2vf_bulletin_mapping.vaddr;
+
+		ret = bcm_vf_get_resources(sc, sc->num_queues, sc->num_queues);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static int
+bcm_dev_start(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+	int ret = 0;
+
+	PMD_INIT_FUNC_TRACE();
+
+	ret = bcm_init(sc);
+	if (ret) {
+		PMD_DRV_LOG(DEBUG, "bcm_init failed (%d)", ret);
+		return -1;
+	}
+
+	if (IS_PF(sc)) {
+		rte_intr_callback_register(&(dev->pci_dev->intr_handle),
+				bcm_interrupt_handler, (void *)dev);
+
+		if(rte_intr_enable(&(dev->pci_dev->intr_handle)))
+			PMD_DRV_LOG(ERR, "rte_intr_enable failed");
+	}
+
+	ret = bcm_dev_rx_init(dev);
+	if (ret != 0) {
+		PMD_DRV_LOG(DEBUG, "bcm_dev_rx_init returned error code");
+		return -3;
+	}
+
+	/* Print important adapter info for the user. */
+	bcm_print_adapter_info(sc);
+
+	DELAY_MS(2500);
+
+	return ret;
+}
+
+static void
+bcm_dev_stop(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+	int ret = 0;
+
+	PMD_INIT_FUNC_TRACE();
+
+	if (IS_PF(sc)) {
+		rte_intr_disable(&(dev->pci_dev->intr_handle));
+		rte_intr_callback_unregister(&(dev->pci_dev->intr_handle),
+				bcm_interrupt_handler, (void *)dev);
+	}
+
+	ret = bcm_nic_unload(sc, UNLOAD_NORMAL, FALSE);
+	if (ret) {
+		PMD_DRV_LOG(DEBUG, "bcm_nic_unload failed (%d)", ret);
+		return;
+	}
+
+	return;
+}
+
+static void
+bcm_dev_close(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+
+	if (IS_VF(sc))
+		bcm_vf_close(sc);
+
+	bcm_dev_clear_queues(dev);
+	memset(&(dev->data->dev_link), 0 , sizeof(struct rte_eth_link));
+
+	/* free the host hardware/software hsi structures */
+	bcm_free_hsi_mem(sc);
+
+	/* free ilt */
+	bcm_free_ilt_mem(sc);
+}
+
+static void
+bcm_promisc_enable(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+	sc->rx_mode = BCM_RX_MODE_PROMISC;
+	bcm_set_rx_mode(sc);
+}
+
+static void
+bcm_promisc_disable(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+	sc->rx_mode = BCM_RX_MODE_NORMAL;
+	bcm_set_rx_mode(sc);
+}
+
+static void
+bcm_dev_allmulticast_enable(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+	sc->rx_mode = BCM_RX_MODE_ALLMULTI;
+	bcm_set_rx_mode(sc);
+}
+
+static void
+bcm_dev_allmulticast_disable(struct rte_eth_dev *dev)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+	sc->rx_mode = BCM_RX_MODE_NORMAL;
+	bcm_set_rx_mode(sc);
+}
+
+static int
+bcm_dev_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
+{
+	PMD_INIT_FUNC_TRACE();
+
+	int old_link_status = dev->data->dev_link.link_status;
+
+	bcm_link_update(dev);
+
+	return old_link_status == dev->data->dev_link.link_status ? -1 : 0;
+}
+
+static int
+bcmvf_dev_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
+{
+	int old_link_status = dev->data->dev_link.link_status;
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	bcm_link_update(dev);
+
+	bcm_check_bull(sc);
+	if (sc->old_bulletin.valid_bitmap & (1 << CHANNEL_DOWN)) {
+		PMD_DRV_LOG(ERR, "PF indicated channel is down."
+				"VF device is no longer operational");
+		dev->data->dev_link.link_status = 0;
+	}
+
+	return old_link_status == dev->data->dev_link.link_status ? -1 : 0;
+}
+
+static void
+bcm_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	PMD_INIT_FUNC_TRACE();
+
+	bcm_stats_handle(sc, STATS_EVENT_UPDATE);
+
+	memset(stats, 0, sizeof (struct rte_eth_stats));
+
+	stats->ipackets =
+		HILO_U64(sc->eth_stats.total_unicast_packets_received_hi,
+				sc->eth_stats.total_unicast_packets_received_lo) +
+		HILO_U64(sc->eth_stats.total_multicast_packets_received_hi,
+				sc->eth_stats.total_multicast_packets_received_lo) +
+		HILO_U64(sc->eth_stats.total_broadcast_packets_received_hi,
+				sc->eth_stats.total_broadcast_packets_received_lo);
+
+	stats->opackets =
+		HILO_U64(sc->eth_stats.total_unicast_packets_transmitted_hi,
+				sc->eth_stats.total_unicast_packets_transmitted_lo) +
+		HILO_U64(sc->eth_stats.total_multicast_packets_transmitted_hi,
+				sc->eth_stats.total_multicast_packets_transmitted_lo) +
+		HILO_U64(sc->eth_stats.total_broadcast_packets_transmitted_hi,
+				sc->eth_stats.total_broadcast_packets_transmitted_lo);
+
+	stats->ibytes =
+		HILO_U64(sc->eth_stats.total_bytes_received_hi,
+				sc->eth_stats.total_bytes_received_lo);
+
+	stats->obytes =
+		HILO_U64(sc->eth_stats.total_bytes_transmitted_hi,
+				sc->eth_stats.total_bytes_transmitted_lo);
+
+	stats->ierrors =
+		HILO_U64(sc->eth_stats.error_bytes_received_hi,
+				sc->eth_stats.error_bytes_received_lo);
+
+	stats->oerrors = 0;
+
+	stats->rx_nombuf =
+		HILO_U64(sc->eth_stats.no_buff_discard_hi,
+				sc->eth_stats.no_buff_discard_lo);
+}
+
+static void
+bcm_dev_infos_get(struct rte_eth_dev *dev, __rte_unused struct rte_eth_dev_info *dev_info)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+	dev_info->max_rx_queues  = sc->max_rx_queues;
+	dev_info->max_tx_queues  = sc->max_tx_queues;
+	dev_info->min_rx_bufsize = BCM_MIN_RX_BUF_SIZE;
+	dev_info->max_rx_pktlen  = BCM_MAX_RX_PKT_LEN;
+	dev_info->max_mac_addrs  = BCM_MAX_MAC_ADDRS;
+}
+
+static void
+bcm_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
+		uint32_t index, uint32_t pool)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	if (sc->mac_ops.mac_addr_add)
+		sc->mac_ops.mac_addr_add(dev, mac_addr, index, pool);
+}
+
+static void
+bcm_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
+{
+	struct bcm_softc *sc = dev->data->dev_private;
+
+	if (sc->mac_ops.mac_addr_remove)
+		sc->mac_ops.mac_addr_remove(dev, index);
+}
+
+static struct eth_dev_ops bcm_eth_dev_ops = {
+	.dev_configure                = bcm_dev_configure,
+	.dev_start                    = bcm_dev_start,
+	.dev_stop                     = bcm_dev_stop,
+	.dev_close                    = bcm_dev_close,
+	.promiscuous_enable           = bcm_promisc_enable,
+	.promiscuous_disable          = bcm_promisc_disable,
+	.allmulticast_enable          = bcm_dev_allmulticast_enable,
+	.allmulticast_disable         = bcm_dev_allmulticast_disable,
+	.link_update                  = bcm_dev_link_update,
+	.stats_get                    = bcm_dev_stats_get,
+	.dev_infos_get                = bcm_dev_infos_get,
+	.rx_queue_setup               = bcm_dev_rx_queue_setup,
+	.rx_queue_release             = bcm_dev_rx_queue_release,
+	.tx_queue_setup               = bcm_dev_tx_queue_setup,
+	.tx_queue_release             = bcm_dev_tx_queue_release,
+	.mac_addr_add                 = bcm_mac_addr_add,
+	.mac_addr_remove              = bcm_mac_addr_remove,
+};
+
+/*
+ * dev_ops for virtual function
+ */
+static struct eth_dev_ops bcmvf_eth_dev_ops = {
+	.dev_configure                = bcm_dev_configure,
+	.dev_start                    = bcm_dev_start,
+	.dev_stop                     = bcm_dev_stop,
+	.dev_close                    = bcm_dev_close,
+	.promiscuous_enable           = bcm_promisc_enable,
+	.promiscuous_disable          = bcm_promisc_disable,
+	.allmulticast_enable          = bcm_dev_allmulticast_enable,
+	.allmulticast_disable         = bcm_dev_allmulticast_disable,
+	.link_update                  = bcmvf_dev_link_update,
+	.stats_get                    = bcm_dev_stats_get,
+	.dev_infos_get                = bcm_dev_infos_get,
+	.rx_queue_setup               = bcm_dev_rx_queue_setup,
+	.rx_queue_release             = bcm_dev_rx_queue_release,
+	.tx_queue_setup               = bcm_dev_tx_queue_setup,
+	.tx_queue_release             = bcm_dev_tx_queue_release,
+	.mac_addr_add                 = bcm_mac_addr_add,
+	.mac_addr_remove              = bcm_mac_addr_remove,
+};
+
+
+static int
+bcm_common_dev_init(struct rte_eth_dev *eth_dev, int is_vf)
+{
+	int ret = 0;
+	struct rte_pci_device *pci_dev;
+	struct bcm_softc *sc;
+
+	PMD_INIT_FUNC_TRACE();
+
+	eth_dev->dev_ops = is_vf ? &bcmvf_eth_dev_ops : &bcm_eth_dev_ops;
+	pci_dev = eth_dev->pci_dev;
+	sc = eth_dev->data->dev_private;
+	sc->pcie_bus    = pci_dev->addr.bus;
+	sc->pcie_device = pci_dev->addr.devid;
+
+	if (is_vf)
+		sc->flags = BCM_IS_VF_FLAG;
+
+	sc->devinfo.vendor_id    = pci_dev->id.vendor_id;
+        sc->devinfo.device_id    = pci_dev->id.device_id;
+        sc->devinfo.subvendor_id = pci_dev->id.subsystem_vendor_id;
+        sc->devinfo.subdevice_id = pci_dev->id.subsystem_device_id;
+
+	sc->pcie_func = pci_dev->addr.function;
+	sc->bar[BAR0].base_addr = (void *)pci_dev->mem_resource[0].addr;
+	if (is_vf)
+		sc->bar[BAR1].base_addr = (void *)
+			((uint64_t)pci_dev->mem_resource[0].addr + PXP_VF_ADDR_DB_START);
+	else
+		sc->bar[BAR1].base_addr = pci_dev->mem_resource[2].addr;
+
+	assert(sc->bar[BAR0].base_addr);
+	assert(sc->bar[BAR1].base_addr);
+
+	bcm_load_firmware(sc);
+	assert(sc->firmware);
+
+	if (eth_dev->data->dev_conf.rx_adv_conf.rss_conf.rss_hf & ETH_RSS_IPV4_UDP)
+		sc->udp_rss = 1;
+
+	sc->rx_budget = BCM_RX_BUDGET;
+	sc->hc_rx_ticks = BCM_RX_TICKS;
+	sc->hc_tx_ticks = BCM_TX_TICKS;
+
+	sc->interrupt_mode = INTR_MODE_SINGLE_MSIX;
+	sc->rx_mode = BCM_RX_MODE_NORMAL;
+
+	sc->pci_dev = pci_dev;
+	ret = bcm_attach(sc);
+	if (ret) {
+		PMD_DRV_LOG(ERR, "bcm_attach failed (%d)", ret);
+	}
+
+	eth_dev->data->mac_addrs = (struct ether_addr *)sc->link_params.mac_addr;
+
+	PMD_DRV_LOG(INFO, "pcie_bus=%d, pcie_device=%d",
+			sc->pcie_bus, sc->pcie_device);
+	PMD_DRV_LOG(INFO, "bar0.addr=%p, bar1.addr=%p",
+			sc->bar[BAR0].base_addr, sc->bar[BAR1].base_addr);
+	PMD_DRV_LOG(INFO, "port=%d, path=%d, vnic=%d, func=%d",
+			PORT_ID(sc), PATH_ID(sc), VNIC_ID(sc), FUNC_ID(sc));
+	PMD_DRV_LOG(INFO, "portID=%d vendorID=0x%x deviceID=0x%x",
+			eth_dev->data->port_id, pci_dev->id.vendor_id, pci_dev->id.device_id);
+
+	return ret;
+}
+
+static int
+eth_bcm_dev_init(__rte_unused struct eth_driver *eth_drv,
+                     struct rte_eth_dev *eth_dev)
+{
+	PMD_INIT_FUNC_TRACE();
+	return bcm_common_dev_init(eth_dev, 0);
+}
+
+static int
+eth_bcmvf_dev_init(__rte_unused struct eth_driver *eth_drv,
+                     struct rte_eth_dev *eth_dev)
+{
+	PMD_INIT_FUNC_TRACE();
+	return bcm_common_dev_init(eth_dev, 1);
+}
+
+static struct eth_driver rte_bcm_pmd = {
+	.pci_drv = {
+		.name = "rte_bcm_pmd",
+		.id_table = pci_id_bcm_map,
+		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	},
+	.eth_dev_init = eth_bcm_dev_init,
+	.dev_private_size = sizeof(struct bcm_softc),
+};
+
+/*
+ * virtual function driver struct
+ */
+static struct eth_driver rte_bcmvf_pmd = {
+	.pci_drv = {
+		.name = "rte_bcmvf_pmd",
+		.id_table = pci_id_bcmvf_map,
+		.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+	},
+	.eth_dev_init = eth_bcmvf_dev_init,
+	.dev_private_size = sizeof(struct bcm_softc),
+};
+
+static int rte_bcm_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
+{
+	PMD_INIT_FUNC_TRACE();
+	rte_eth_driver_register(&rte_bcm_pmd);
+
+	return 0;
+}
+
+static int rte_bcmvf_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
+{
+	PMD_INIT_FUNC_TRACE();
+	rte_eth_driver_register(&rte_bcmvf_pmd);
+
+	return 0;
+}
+
+static struct rte_driver rte_bcm_driver = {
+	.type = PMD_PDEV,
+	.init = rte_bcm_pmd_init,
+};
+
+static struct rte_driver rte_bcmvf_driver = {
+	.type = PMD_PDEV,
+	.init = rte_bcmvf_pmd_init,
+};
+
+PMD_REGISTER_DRIVER(rte_bcm_driver);
+PMD_REGISTER_DRIVER(rte_bcmvf_driver);
diff --git a/lib/librte_pmd_bcm/bcm_ethdev.h b/lib/librte_pmd_bcm/bcm_ethdev.h
new file mode 100644
index 0000000..da3abe3
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_ethdev.h
@@ -0,0 +1,79 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#ifndef PMD_BCM_ETHDEV_H
+#define PMD_BCM_ETHDEV_H
+
+#include <sys/queue.h>
+#include <sys/param.h>
+#include <sys/user.h>
+#include <sys/stat.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <stdint.h>
+#include <string.h>
+#include <unistd.h>
+#include <stdarg.h>
+#include <inttypes.h>
+#include <assert.h>
+
+#include <rte_byteorder.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_log.h>
+#include <rte_debug.h>
+#include <rte_pci.h>
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_spinlock.h>
+#include <rte_memzone.h>
+#include <rte_eal.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+
+#include "bcm_rxtx.h"
+#include "bcm_logs.h"
+
+#define DELAY(x) rte_delay_us(x)
+#define DELAY_MS(x) rte_delay_ms(x)
+#define usec_delay(x) DELAY(x)
+#define msec_delay(x) DELAY(1000*(x))
+
+#define FALSE               0
+#define TRUE                1
+
+#define false               0
+#define true                1
+#define min(a,b)        RTE_MIN(a,b)
+
+#define mb()    rte_mb()
+#define wmb()   rte_wmb()
+#define rmb()   rte_rmb()
+
+
+#define MAX_QUEUES sysconf(_SC_NPROCESSORS_CONF)
+
+#define BCM_MIN_RX_BUF_SIZE 1024
+#define BCM_MAX_RX_PKT_LEN  15872
+#define BCM_MAX_MAC_ADDRS   1
+
+/* Hardware RX tick timer (usecs) */
+#define BCM_RX_TICKS 25
+/* Hardware TX tick timer (usecs) */
+#define BCM_TX_TICKS 50
+/* Maximum number of Rx packets to process at a time */
+#define BCM_RX_BUDGET 0xffffffff
+
+#endif
+
+/* MAC address operations */
+struct bcm_mac_ops {
+	void (*mac_addr_add)(struct rte_eth_dev *dev, struct ether_addr *addr,
+			uint16_t index, uint32_t pool);                           /* not implemented yet */
+	void (*mac_addr_remove)(struct rte_eth_dev *dev, uint16_t index); /* not implemented yet */
+};
diff --git a/lib/librte_pmd_bcm/bcm_logs.h b/lib/librte_pmd_bcm/bcm_logs.h
new file mode 100644
index 0000000..d562415
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_logs.h
@@ -0,0 +1,51 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#ifndef _PMD_LOGS_H_
+#define _PMD_LOGS_H_
+
+#define PMD_INIT_LOG(level, fmt, args...) \
+	rte_log(RTE_LOG_ ## level, RTE_LOGTYPE_PMD, \
+		"PMD: %s(): " fmt "\n", __func__, ##args)
+
+#ifdef RTE_LIBRTE_BCM_DEBUG_INIT
+#define PMD_INIT_FUNC_TRACE() PMD_INIT_LOG(DEBUG, " >>")
+#else
+#define PMD_INIT_FUNC_TRACE() do { } while(0)
+#endif
+
+#ifdef RTE_LIBRTE_BCM_DEBUG_RX
+#define PMD_RX_LOG(level, fmt, args...) \
+	RTE_LOG(level, PMD, "%s(): " fmt "\n", __func__, ## args)
+#else
+#define PMD_RX_LOG(level, fmt, args...) do { } while(0)
+#endif
+
+#ifdef RTE_LIBRTE_BCM_DEBUG_TX
+#define PMD_TX_LOG(level, fmt, args...) \
+	RTE_LOG(level, PMD, "%s(): " fmt "\n", __func__, ## args)
+#else
+#define PMD_TX_LOG(level, fmt, args...) do { } while(0)
+#endif
+
+#ifdef RTE_LIBRTE_BCM_DEBUG_TX_FREE
+#define PMD_TX_FREE_LOG(level, fmt, args...) \
+	RTE_LOG(level, PMD, "%s(): " fmt "\n", __func__, ## args)
+#else
+#define PMD_TX_FREE_LOG(level, fmt, args...) do { } while(0)
+#endif
+
+#ifdef RTE_LIBRTE_BCM_DEBUG_DRIVER
+#define PMD_DRV_LOG_RAW(level, fmt, args...) \
+	RTE_LOG(level, PMD, "%s(): " fmt, __func__, ## args)
+#else
+#define PMD_DRV_LOG_RAW(level, fmt, args...) do { } while (0)
+#endif
+
+#define PMD_DRV_LOG(level, fmt, args...) \
+	PMD_DRV_LOG_RAW(level, fmt "\n", ## args)
+
+#endif /* _PMD_LOGS_H_ */
diff --git a/lib/librte_pmd_bcm/bcm_rxtx.c b/lib/librte_pmd_bcm/bcm_rxtx.c
new file mode 100644
index 0000000..f263e95
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_rxtx.c
@@ -0,0 +1,487 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#include "bcm.h"
+#include "bcm_rxtx.h"
+
+static inline struct rte_mbuf *
+bcm_rxmbuf_alloc(struct rte_mempool *mp)
+{
+	struct rte_mbuf *m;
+
+	m = __rte_mbuf_raw_alloc(mp);
+	__rte_mbuf_sanity_check(m, 0);
+
+	return m;
+}
+
+static const struct rte_memzone *
+ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
+                      uint16_t queue_id, uint32_t ring_size, int socket_id)
+{
+	char z_name[RTE_MEMZONE_NAMESIZE];
+	const struct rte_memzone *mz;
+
+	snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
+			dev->driver->pci_drv.name, ring_name, dev->data->port_id, queue_id);
+
+	mz = rte_memzone_lookup(z_name);
+	if (mz)
+		return mz;
+
+	return rte_memzone_reserve_aligned(z_name, ring_size, socket_id, 0, BCM_PAGE_SIZE);
+}
+
+static void
+bcm_rx_queue_release(struct bcm_rx_queue *rx_queue)
+{
+	uint16_t i;
+	struct rte_mbuf **sw_ring;
+
+	if (NULL != rx_queue) {
+
+		sw_ring = rx_queue->sw_ring;
+		if (NULL != sw_ring) {
+			for (i = 0; i < rx_queue->nb_rx_desc; i++) {
+				if (NULL != sw_ring[i])
+					rte_pktmbuf_free(sw_ring[i]);
+			}
+			rte_free(sw_ring);
+		}
+		rte_free(rx_queue);
+	}
+}
+
+void
+bcm_dev_rx_queue_release(void *rxq)
+{
+        bcm_rx_queue_release(rxq);
+}
+
+int
+bcm_dev_rx_queue_setup(struct rte_eth_dev *dev,
+                       uint16_t queue_idx,
+                       uint16_t nb_desc,
+                       unsigned int socket_id,
+                       const struct rte_eth_rxconf *rx_conf,
+                       struct rte_mempool *mp)
+{
+	uint16_t j, idx;
+	const struct rte_memzone *dma;
+	struct bcm_rx_queue *rxq;
+	uint32_t dma_size;
+	struct rte_mbuf *mbuf;
+	struct bcm_softc *sc = dev->data->dev_private;
+	struct bcm_fastpath *fp = &sc->fp[queue_idx];
+	struct eth_rx_cqe_next_page *nextpg;
+	phys_addr_t *rx_bd;
+	phys_addr_t busaddr;
+
+	/* First allocate the rx queue data structure */
+	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct bcm_rx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (NULL == rxq) {
+		PMD_INIT_LOG(ERR, "rte_zmalloc for rxq failed!");
+		return (-ENOMEM);
+	}
+	rxq->sc = sc;
+	rxq->mb_pool = mp;
+	rxq->queue_id = queue_idx;
+	rxq->port_id = dev->data->port_id;
+	rxq->crc_len = (uint8_t)((dev->data->dev_conf.rxmode.hw_strip_crc) ? 0 : ETHER_CRC_LEN);
+
+	rxq->nb_rx_pages = 1;
+	while (USABLE_RX_BD(rxq) < nb_desc)
+		rxq->nb_rx_pages <<= 1;
+
+	rxq->nb_rx_desc  = TOTAL_RX_BD(rxq);
+	sc->rx_ring_size = USABLE_RX_BD(rxq);
+	rxq->nb_cq_pages = RCQ_BD_PAGES(rxq);
+
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh ?
+		rx_conf->rx_free_thresh : DEFAULT_RX_FREE_THRESH;
+
+	PMD_INIT_LOG(DEBUG, "fp[%02d] req_bd=%u, thresh=%u, usable_bd=%lu, "
+		       "total_bd=%lu, rx_pages=%u, cq_pages=%u",
+		       queue_idx, nb_desc, rxq->rx_free_thresh, USABLE_RX_BD(rxq),
+		       TOTAL_RX_BD(rxq), rxq->nb_rx_pages, rxq->nb_cq_pages);
+
+	/* Allocate RX ring hardware descriptors */
+	dma_size = rxq->nb_rx_desc * sizeof(struct eth_rx_bd);
+	dma = ring_dma_zone_reserve(dev, "hw_ring", queue_idx, dma_size, socket_id);
+	if (NULL == dma) {
+		PMD_RX_LOG(ERR, "ring_dma_zone_reserve for rx_ring failed!");
+		bcm_rx_queue_release(rxq);
+		return (-ENOMEM);
+	}
+	fp->rx_desc_mapping = rxq->rx_ring_phys_addr = (uint64_t)dma->phys_addr;
+	rxq->rx_ring = (uint64_t*)dma->addr;
+	memset((void *)rxq->rx_ring, 0, dma_size);
+
+	/* Link the RX chain pages. */
+	for (j = 1; j <= rxq->nb_rx_pages; j++) {
+		rx_bd = &rxq->rx_ring[TOTAL_RX_BD_PER_PAGE * j - 2];
+		busaddr = rxq->rx_ring_phys_addr + BCM_PAGE_SIZE * (j % rxq->nb_rx_pages);
+		*rx_bd = busaddr;
+	}
+
+	/* Allocate software ring */
+	dma_size = rxq->nb_rx_desc * sizeof(struct bcm_rx_entry);
+	rxq->sw_ring = rte_zmalloc_socket("sw_ring", dma_size,
+					  RTE_CACHE_LINE_SIZE,
+					  socket_id);
+	if (NULL == rxq->sw_ring) {
+		PMD_RX_LOG(ERR, "rte_zmalloc for sw_ring failed!");
+		bcm_rx_queue_release(rxq);
+		return (-ENOMEM);
+	}
+
+	/* Initialize software ring entries */
+	rxq->rx_mbuf_alloc = 0;
+	for (idx = 0; idx < rxq->nb_rx_desc; idx = NEXT_RX_BD(idx)) {
+		mbuf = bcm_rxmbuf_alloc(mp);
+		if (NULL == mbuf) {
+			PMD_RX_LOG(ERR, "RX mbuf alloc failed queue_id=%u, idx=%d",
+				   (unsigned)rxq->queue_id, idx);
+			bcm_rx_queue_release(rxq);
+			return (-ENOMEM);
+		}
+		rxq->sw_ring[idx] = mbuf;
+		rxq->rx_ring[idx] = mbuf->buf_physaddr;
+		rxq->rx_mbuf_alloc++;
+	}
+	rxq->pkt_first_seg = NULL;
+	rxq->pkt_last_seg = NULL;
+	rxq->rx_bd_head = 0;
+	rxq->rx_bd_tail = idx;
+
+	/* Allocate CQ chain. */
+	dma_size = BCM_RX_CHAIN_PAGE_SZ * rxq->nb_cq_pages;
+	dma = ring_dma_zone_reserve(dev, "bcm_rcq", queue_idx, dma_size, socket_id);
+	if (NULL == dma) {
+		PMD_RX_LOG(ERR, "RCQ  alloc failed");
+		return (-ENOMEM);
+	}
+	fp->rx_comp_mapping = rxq->cq_ring_phys_addr = (uint64_t)dma->phys_addr;
+	rxq->cq_ring = (union eth_rx_cqe*)dma->addr;
+
+	/* Link the CQ chain pages. */
+	for (j = 1; j <= rxq->nb_cq_pages; j++) {
+		nextpg = &rxq->cq_ring[TOTAL_RCQ_ENTRIES_PER_PAGE * j - 1].next_page_cqe;
+		busaddr = rxq->cq_ring_phys_addr + BCM_PAGE_SIZE * (j % rxq->nb_cq_pages);
+		nextpg->addr_hi = rte_cpu_to_le_32(U64_HI(busaddr));
+		nextpg->addr_lo = rte_cpu_to_le_32(U64_LO(busaddr));
+	}
+	rxq->rx_cq_head = 0;
+	rxq->rx_cq_tail = TOTAL_RCQ_ENTRIES(rxq);
+
+	dev->data->rx_queues[queue_idx] = rxq;
+	if (!sc->rx_queues) sc->rx_queues = dev->data->rx_queues;
+
+	return 0;
+}
+
+static void
+bcm_tx_queue_release(struct bcm_tx_queue *tx_queue)
+{
+	uint16_t i;
+	struct rte_mbuf **sw_ring;
+
+	if (NULL != tx_queue) {
+
+		sw_ring = tx_queue->sw_ring;
+		if (NULL != sw_ring) {
+			for (i = 0; i < tx_queue->nb_tx_desc; i++) {
+				if (NULL != sw_ring[i])
+					rte_pktmbuf_free(sw_ring[i]);
+			}
+			rte_free(sw_ring);
+		}
+		rte_free(tx_queue);
+	}
+}
+
+void
+bcm_dev_tx_queue_release(void *txq)
+{
+	bcm_tx_queue_release(txq);
+}
+
+static uint16_t
+bcm_xmit_pkts(void *p_txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+	struct bcm_tx_queue *txq;
+	struct bcm_softc *sc;
+	struct bcm_fastpath *fp;
+	uint32_t burst, nb_tx;
+	struct rte_mbuf **m = tx_pkts;
+	int ret;
+
+	txq = p_txq;
+	sc = txq->sc;
+	fp = &sc->fp[txq->queue_id];
+
+	nb_tx = nb_pkts;
+
+	do {
+		burst = RTE_MIN(nb_pkts, RTE_PMD_BCM_TX_MAX_BURST);
+
+		ret = bcm_tx_encap(txq, m, burst);
+		if (unlikely(ret)) {
+			PMD_TX_LOG(ERR, "tx_encap failed!");
+		}
+
+		bcm_update_fp_sb_idx(fp);
+
+		if ((txq->nb_tx_desc - txq->nb_tx_avail) > txq->tx_free_thresh) {
+			bcm_txeof(sc, fp);
+		}
+
+		if (unlikely(ret == ENOMEM)) {
+			break;
+		}
+
+		m += burst;
+		nb_pkts -= burst;
+
+	} while (nb_pkts);
+
+	return nb_tx - nb_pkts;
+}
+
+int
+bcm_dev_tx_queue_setup(struct rte_eth_dev *dev,
+		       uint16_t queue_idx,
+		       uint16_t nb_desc,
+		       unsigned int socket_id,
+		       const struct rte_eth_txconf *tx_conf)
+{
+	uint16_t i;
+	unsigned int tsize;
+	const struct rte_memzone *tz;
+	struct bcm_tx_queue *txq;
+	struct eth_tx_next_bd *tx_n_bd;
+	uint64_t busaddr;
+	struct bcm_softc *sc = dev->data->dev_private;
+	struct bcm_fastpath *fp = &sc->fp[queue_idx];
+
+	/* First allocate the tx queue data structure */
+	txq = rte_zmalloc("ethdev TX queue", sizeof(struct bcm_tx_queue),
+			  RTE_CACHE_LINE_SIZE);
+	if (txq == NULL)
+		return (-ENOMEM);
+	txq->sc = sc;
+
+	txq->nb_tx_pages = 1;
+	while (USABLE_TX_BD(txq) < nb_desc)
+		txq->nb_tx_pages <<= 1;
+
+	txq->nb_tx_desc  = TOTAL_TX_BD(txq);
+	sc->tx_ring_size = TOTAL_TX_BD(txq);
+
+	txq->tx_free_thresh = tx_conf->tx_free_thresh ?
+		tx_conf->tx_free_thresh : DEFAULT_TX_FREE_THRESH;
+
+	PMD_INIT_LOG(DEBUG, "fp[%02d] req_bd=%u, thresh=%u, usable_bd=%lu, "
+		     "total_bd=%lu, tx_pages=%u",
+		     queue_idx, nb_desc, txq->tx_free_thresh, USABLE_TX_BD(txq),
+		     TOTAL_TX_BD(txq), txq->nb_tx_pages);
+
+	/* Allocate TX ring hardware descriptors */
+	tsize = txq->nb_tx_desc * sizeof(union eth_tx_bd_types);
+	tz = ring_dma_zone_reserve(dev, "tx_hw_ring", queue_idx, tsize, socket_id);
+	if (tz == NULL) {
+		bcm_tx_queue_release(txq);
+		return (-ENOMEM);
+	}
+	fp->tx_desc_mapping = txq->tx_ring_phys_addr = (uint64_t)tz->phys_addr;
+	txq->tx_ring = (union eth_tx_bd_types *) tz->addr;
+	memset(txq->tx_ring, 0, tsize);
+
+	/* Allocate software ring */
+	tsize = txq->nb_tx_desc * sizeof(struct rte_mbuf *);
+	txq->sw_ring = rte_zmalloc("tx_sw_ring", tsize,
+				   RTE_CACHE_LINE_SIZE);
+	if (txq->sw_ring == NULL) {
+		bcm_tx_queue_release(txq);
+		return (-ENOMEM);
+	}
+
+	/* PMD_DRV_LOG(DEBUG, "sw_ring=%p hw_ring=%p dma_addr=0x%"PRIx64,
+	   txq->sw_ring, txq->tx_ring, txq->tx_ring_phys_addr); */
+
+	/* Link TX pages */
+	for (i = 1; i <= txq->nb_tx_pages; i++) {
+		tx_n_bd = &txq->tx_ring[TOTAL_TX_BD_PER_PAGE * i - 1].next_bd;
+		busaddr = txq->tx_ring_phys_addr + BCM_PAGE_SIZE * (i % txq->nb_tx_pages);
+		tx_n_bd->addr_hi = rte_cpu_to_le_32(U64_HI(busaddr));
+		tx_n_bd->addr_lo = rte_cpu_to_le_32(U64_LO(busaddr));
+		/* PMD_DRV_LOG(DEBUG, "link tx page %lu", (TOTAL_TX_BD_PER_PAGE * i - 1)); */
+	}
+
+	txq->queue_id = queue_idx;
+	txq->port_id = dev->data->port_id;
+	txq->tx_pkt_tail = 0;
+	txq->tx_pkt_head = 0;
+	txq->tx_bd_tail = 0;
+	txq->tx_bd_head = 0;
+	txq->nb_tx_avail = txq->nb_tx_desc;
+	dev->tx_pkt_burst = bcm_xmit_pkts;
+	dev->data->tx_queues[queue_idx] = txq;
+	if (!sc->tx_queues) sc->tx_queues = dev->data->tx_queues;
+
+	return 0;
+}
+
+static inline void
+bcm_upd_rx_prod_fast(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		uint16_t rx_bd_prod, uint16_t rx_cq_prod)
+{
+	union ustorm_eth_rx_producers rx_prods;
+
+	rx_prods.prod.bd_prod  = rx_bd_prod;
+	rx_prods.prod.cqe_prod = rx_cq_prod;
+
+	REG_WR(sc, fp->ustorm_rx_prods_offset, rx_prods.raw_data[0]);
+}
+
+static uint16_t
+bcm_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+	struct bcm_rx_queue *rxq = p_rxq;
+	struct bcm_softc *sc = rxq->sc;
+	struct bcm_fastpath *fp = &sc->fp[rxq->queue_id];
+	uint32_t nb_rx = 0;
+	uint16_t hw_cq_cons, sw_cq_cons, sw_cq_prod;
+	uint16_t bd_cons, bd_prod;
+	struct rte_mbuf *new_mb;
+	uint16_t rx_pref;
+	struct eth_fast_path_rx_cqe *cqe_fp;
+	uint16_t len, pad;
+	struct rte_mbuf *rx_mb = NULL;
+
+	hw_cq_cons = le16toh(*fp->rx_cq_cons_sb);
+	if ((hw_cq_cons & USABLE_RCQ_ENTRIES_PER_PAGE) ==
+			USABLE_RCQ_ENTRIES_PER_PAGE) {
+		++hw_cq_cons;
+	}
+
+	bd_cons = rxq->rx_bd_head;
+	bd_prod = rxq->rx_bd_tail;
+	sw_cq_cons = rxq->rx_cq_head;
+	sw_cq_prod = rxq->rx_cq_tail;
+
+	while (nb_rx < nb_pkts && sw_cq_cons != hw_cq_cons) {
+
+		bd_prod &= MAX_RX_BD(rxq);
+		bd_cons &= MAX_RX_BD(rxq);
+
+		cqe_fp = &rxq->cq_ring[sw_cq_cons & MAX_RX_BD(rxq)].fast_path_cqe;
+
+		if (unlikely(CQE_TYPE_SLOW(cqe_fp->type_error_flags & ETH_FAST_PATH_RX_CQE_TYPE))) {
+			PMD_RX_LOG(ERR, "slowpath event during traffic processing");
+			break;
+		}
+
+		if (unlikely(cqe_fp->type_error_flags & ETH_FAST_PATH_RX_CQE_PHY_DECODE_ERR_FLG)) {
+			PMD_RX_LOG(ERR, "flags 0x%x rx packet %u",
+					cqe_fp->type_error_flags, sw_cq_cons);
+			goto next_rx;
+		}
+
+		len = cqe_fp->pkt_len_or_gro_seg_len;
+		pad = cqe_fp->placement_offset;
+
+		new_mb = bcm_rxmbuf_alloc(rxq->mb_pool);
+		if (unlikely(!new_mb)) {
+			PMD_RX_LOG(ERR, "mbuf alloc fail fp[%02d]", fp->index);
+			goto next_rx;
+		}
+
+		rx_mb = rxq->sw_ring[bd_cons];
+		rxq->sw_ring[bd_cons] = new_mb;
+		rxq->rx_ring[bd_prod] = new_mb->buf_physaddr;
+
+		rx_pref = NEXT_RX_BD(bd_cons) & MAX_RX_BD(rxq);
+		rte_prefetch0(rxq->sw_ring[rx_pref]);
+		if ((rx_pref & 0x3) == 0) {
+			rte_prefetch0(&rxq->rx_ring[rx_pref]);
+			rte_prefetch0(&rxq->sw_ring[rx_pref]);
+		}
+
+		rx_mb->data_off = pad;
+		rx_mb->nb_segs = 1;
+		rx_mb->next = NULL;
+		rx_mb->pkt_len = rx_mb->data_len = len;
+		rx_mb->port = rxq->port_id;
+		rx_mb->buf_len = len + pad;
+		rte_prefetch1(rte_pktmbuf_mtod(rx_mb, void *));
+
+		/*
+		 * If we received a packet with a vlan tag,
+		 * attach that information to the packet.
+		 */
+		if (cqe_fp->pars_flags.flags & PARSING_FLAGS_VLAN) {
+			rx_mb->vlan_tci = cqe_fp->vlan_tag;
+			rx_mb->ol_flags |= PKT_RX_VLAN_PKT;
+		}
+
+		rx_pkts[nb_rx] = rx_mb;
+		nb_rx++;
+
+		/* limit spinning on the queue */
+		if (unlikely(nb_rx == sc->rx_budget)) {
+			PMD_RX_LOG(ERR, "Limit spinning on the queue");
+			break;
+		}
+
+next_rx:
+		bd_cons    = NEXT_RX_BD(bd_cons);
+		bd_prod    = NEXT_RX_BD(bd_prod);
+		sw_cq_prod = NEXT_RCQ_IDX(sw_cq_prod);
+		sw_cq_cons = NEXT_RCQ_IDX(sw_cq_cons);
+	}
+	rxq->rx_bd_head = bd_cons;
+	rxq->rx_bd_tail = bd_prod;
+	rxq->rx_cq_head = sw_cq_cons;
+	rxq->rx_cq_tail = sw_cq_prod;
+
+	bcm_upd_rx_prod_fast(sc, fp, bd_prod, sw_cq_prod);
+
+	return nb_rx;
+}
+
+int
+bcm_dev_rx_init(struct rte_eth_dev *dev)
+{
+	dev->rx_pkt_burst = bcm_recv_pkts;
+
+	return 0;
+}
+
+void
+bcm_dev_clear_queues(struct rte_eth_dev *dev)
+{
+	uint8_t i;
+
+	PMD_INIT_FUNC_TRACE();
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		struct bcm_tx_queue *txq = dev->data->tx_queues[i];
+		if (txq != NULL) {
+			bcm_tx_queue_release(txq);
+			dev->data->tx_queues[i] = NULL;
+		}
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		struct bcm_rx_queue *rxq = dev->data->rx_queues[i];
+		if (rxq != NULL) {
+			bcm_rx_queue_release(rxq);
+			dev->data->rx_queues[i] = NULL;
+		}
+	}
+}
diff --git a/lib/librte_pmd_bcm/bcm_rxtx.h b/lib/librte_pmd_bcm/bcm_rxtx.h
new file mode 100644
index 0000000..978b801
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_rxtx.h
@@ -0,0 +1,85 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#ifndef _BCM_RXTX_H_
+#define _BCM_RXTX_H_
+
+
+#define DEFAULT_RX_FREE_THRESH   0
+#define DEFAULT_TX_FREE_THRESH   512
+#define RTE_PMD_BCM_TX_MAX_BURST 1
+
+/**
+ * Structure associated with each descriptor of the RX ring of a RX queue.
+ */
+struct bcm_rx_entry {
+        struct rte_mbuf     *mbuf;                /**< mbuf associated with RX descriptor. */
+};
+
+/**
+ * Structure associated with each RX queue.
+ */
+struct bcm_rx_queue {
+        struct rte_mempool         *mb_pool;             /**< mbuf pool to populate RX ring. */
+        union eth_rx_cqe           *cq_ring;             /**< RCQ ring virtual address. */
+        uint64_t                   cq_ring_phys_addr;    /**< RCQ ring DMA address. */
+        uint64_t                   *rx_ring;             /**< RX ring virtual address. */
+        uint64_t                   rx_ring_phys_addr;    /**< RX ring DMA address. */
+        struct rte_mbuf            **sw_ring;            /**< address of RX software ring. */
+        struct rte_mbuf            *pkt_first_seg;       /**< First segment of current packet. */
+        struct rte_mbuf            *pkt_last_seg;        /**< Last segment of current packet. */
+        uint16_t                   nb_cq_pages;          /**< number of RCQ pages. */
+        uint16_t                   nb_rx_desc;           /**< number of RX descriptors. */
+        uint16_t                   nb_rx_pages;          /**< number of RX pages. */
+        uint16_t                   rx_bd_head;           /**< Index of current rx bd. */
+        uint16_t                   rx_bd_tail;           /**< Index of last rx bd. */
+        uint16_t                   rx_cq_head;           /**< Index of current rcq bd. */
+        uint16_t                   rx_cq_tail;           /**< Index of last rcq bd. */
+        uint16_t                   nb_rx_hold;           /**< number of held free RX desc. */
+        uint16_t                   rx_free_thresh;       /**< max free RX desc to hold. */
+        uint16_t                   queue_id;             /**< RX queue index. */
+        uint8_t                    port_id;              /**< Device port identifier. */
+        uint8_t                    crc_len;              /**< 0 if CRC stripped, 4 otherwise. */
+        struct bcm_softc           *sc;                  /**< Ptr to dev_private data. */
+        uint64_t                   rx_mbuf_alloc;        /**< Number of allocated mbufs. */
+};
+
+/**
+ * Structure associated with each TX queue.
+ */
+struct bcm_tx_queue {
+	/** TX ring virtual address. */
+	union eth_tx_bd_types      *tx_ring;             /**< TX ring virtual address. */
+	uint64_t                   tx_ring_phys_addr;    /**< TX ring DMA address. */
+	struct rte_mbuf            **sw_ring;            /**< virtual address of SW ring. */
+	uint16_t                   tx_pkt_tail;          /**< Index of current tx pkt. */
+	uint16_t                   tx_pkt_head;          /**< Index of last pkt counted by txeof. */
+	uint16_t                   tx_bd_tail;           /**< Index of current tx bd. */
+	uint16_t                   tx_bd_head;           /**< Index of last bd counted by txeof. */
+	uint16_t                   nb_tx_desc;           /**< number of TX descriptors. */
+	uint16_t                   tx_free_thresh;       /**< minimum TX before freeing. */
+	uint16_t                   nb_tx_avail;          /**< Number of TX descriptors available. */
+	uint16_t                   nb_tx_pages;          /**< number of TX pages */
+	uint16_t                   queue_id;             /**< TX queue index. */
+	uint8_t                    port_id;              /**< Device port identifier. */
+	struct bcm_softc           *sc;                  /**< Ptr to dev_private data */
+};
+
+int bcm_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
+                              uint16_t nb_rx_desc, unsigned int socket_id,
+                              const struct rte_eth_rxconf *rx_conf,
+                              struct rte_mempool *mb_pool);
+
+int bcm_dev_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id,
+                              uint16_t nb_tx_desc, unsigned int socket_id,
+                              const struct rte_eth_txconf *tx_conf);
+
+void bcm_dev_rx_queue_release(void *rxq);
+void bcm_dev_tx_queue_release(void *txq);
+int bcm_dev_rx_init(struct rte_eth_dev *dev);
+void bcm_dev_clear_queues(struct rte_eth_dev *dev);
+
+#endif /* _BCM_RXTX_H_ */
diff --git a/lib/librte_pmd_bcm/bcm_stats.c b/lib/librte_pmd_bcm/bcm_stats.c
new file mode 100644
index 0000000..73d9c1c
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_stats.c
@@ -0,0 +1,1619 @@ 
+/*-
+ * Copyright (c) 2007-2013 Broadcom Corporation. All rights reserved.
+ *
+ * Eric Davis        <edavis@broadcom.com>
+ * David Christensen <davidch@broadcom.com>
+ * Gary Zambrano     <zambrano@broadcom.com>
+ *
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ * All rights reserved.
+ *
+ * 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 Broadcom Corporation nor the name of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written consent.
+ *
+ * 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 OWNER 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.
+ */
+
+#include "bcm.h"
+#include "bcm_stats.h"
+
+#ifdef __i386__
+#define BITS_PER_LONG 32
+#else
+#define BITS_PER_LONG 64
+#endif
+
+static inline long
+bcm_hilo(uint32_t *hiref)
+{
+    uint32_t lo = *(hiref + 1);
+#if (BITS_PER_LONG == 64)
+    uint32_t hi = *hiref;
+    return HILO_U64(hi, lo);
+#else
+    return lo;
+#endif
+}
+
+static inline uint16_t
+bcm_get_port_stats_dma_len(struct bcm_softc *sc)
+{
+	uint16_t res = 0;
+	uint32_t size;
+
+	/* 'newest' convention - shmem2 contains the size of the port stats */
+	if (SHMEM2_HAS(sc, sizeof_port_stats)) {
+		size = SHMEM2_RD(sc, sizeof_port_stats);
+		if (size) {
+			res = size;
+		}
+
+		/* prevent newer BC from causing buffer overflow */
+		if (res > sizeof(struct host_port_stats)) {
+			res = sizeof(struct host_port_stats);
+		}
+	}
+
+	/*
+	 * Older convention - all BCs support the port stats fields up until
+	 * the 'not_used' field
+	 */
+	if (!res) {
+		res = (offsetof(struct host_port_stats, not_used) + 4);
+
+		/* if PFC stats are supported by the MFW, DMA them as well */
+		if (sc->devinfo.bc_ver >= REQ_BC_VER_4_PFC_STATS_SUPPORTED) {
+			res += (offsetof(struct host_port_stats, pfc_frames_rx_lo) -
+				offsetof(struct host_port_stats, pfc_frames_tx_hi) + 4);
+		}
+	}
+
+	res >>= 2;
+
+	return res;
+}
+
+/*
+ * Init service functions
+ */
+
+/*
+ * Post the next statistics ramrod. Protect it with the lock in
+ * order to ensure the strict order between statistics ramrods
+ * (each ramrod has a sequence number passed in a
+ * sc->fw_stats_req->hdr.drv_stats_counter and ramrods must be
+ * sent in order).
+ */
+static void
+bcm_storm_stats_post(struct bcm_softc *sc)
+{
+	int rc;
+
+	if (!sc->stats_pending) {
+		if (sc->stats_pending) {
+			return;
+		}
+
+		sc->fw_stats_req->hdr.drv_stats_counter =
+			htole16(sc->stats_counter++);
+
+		PMD_DRV_LOG(DEBUG,
+				"sending statistics ramrod %d",
+				le16toh(sc->fw_stats_req->hdr.drv_stats_counter));
+
+		/* adjust the ramrod to include VF queues statistics */
+
+		/* send FW stats ramrod */
+		rc = bcm_sp_post(sc, RAMROD_CMD_ID_COMMON_STAT_QUERY, 0,
+				U64_HI(sc->fw_stats_req_mapping),
+				U64_LO(sc->fw_stats_req_mapping),
+				NONE_CONNECTION_TYPE);
+		if (rc == 0) {
+			sc->stats_pending = 1;
+		}
+	}
+}
+
+static void
+bcm_hw_stats_post(struct bcm_softc *sc)
+{
+	struct dmae_command *dmae = &sc->stats_dmae;
+	uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+	int loader_idx;
+	uint32_t opcode;
+
+	*stats_comp = DMAE_COMP_VAL;
+	if (CHIP_REV_IS_SLOW(sc)) {
+		return;
+	}
+
+	/* Update MCP's statistics if possible */
+	if (sc->func_stx) {
+		rte_memcpy(BCM_SP(sc, func_stats), &sc->func_stats,
+				sizeof(sc->func_stats));
+	}
+
+	/* loader */
+	if (sc->executer_idx) {
+		loader_idx = PMF_DMAE_C(sc);
+		opcode =  bcm_dmae_opcode(sc, DMAE_SRC_PCI, DMAE_DST_GRC,
+				TRUE, DMAE_COMP_GRC);
+		opcode = bcm_dmae_opcode_clr_src_reset(opcode);
+
+		memset(dmae, 0, sizeof(struct dmae_command));
+		dmae->opcode = opcode;
+		dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, dmae[0]));
+		dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, dmae[0]));
+		dmae->dst_addr_lo = ((DMAE_REG_CMD_MEM +
+					sizeof(struct dmae_command) *
+					(loader_idx + 1)) >> 2);
+		dmae->dst_addr_hi = 0;
+		dmae->len = sizeof(struct dmae_command) >> 2;
+		dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx + 1] >> 2);
+		dmae->comp_addr_hi = 0;
+		dmae->comp_val = 1;
+
+		*stats_comp = 0;
+		bcm_post_dmae(sc, dmae, loader_idx);
+	} else if (sc->func_stx) {
+		*stats_comp = 0;
+		bcm_post_dmae(sc, dmae, INIT_DMAE_C(sc));
+	}
+}
+
+static int
+bcm_stats_comp(struct bcm_softc *sc)
+{
+	uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+	int cnt = 10;
+
+	while (*stats_comp != DMAE_COMP_VAL) {
+		if (!cnt) {
+			PMD_DRV_LOG(ERR, "Timeout waiting for stats finished");
+			break;
+		}
+
+		cnt--;
+		DELAY(1000);
+	}
+
+	return 1;
+}
+
+/*
+ * Statistics service functions
+ */
+
+static void
+bcm_stats_pmf_update(struct bcm_softc *sc)
+{
+	struct dmae_command *dmae;
+	uint32_t opcode;
+	int loader_idx = PMF_DMAE_C(sc);
+	uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+	if (sc->devinfo.bc_ver <= 0x06001400) {
+		/*
+		 * Bootcode v6.0.21 fixed a GRC timeout that occurs when accessing
+		 * BRB registers while the BRB block is in reset. The DMA transfer
+		 * below triggers this issue resulting in the DMAE to stop
+		 * functioning. Skip this initial stats transfer for old bootcode
+		 * versions <= 6.0.20.
+		 */
+		return;
+	}
+	/* sanity */
+	if (!sc->port.pmf || !sc->port.port_stx) {
+		PMD_DRV_LOG(ERR, "BUG!");
+		return;
+	}
+
+	sc->executer_idx = 0;
+
+	opcode = bcm_dmae_opcode(sc, DMAE_SRC_GRC, DMAE_DST_PCI, FALSE, 0);
+
+	dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+	dmae->opcode = bcm_dmae_opcode_add_comp(opcode, DMAE_COMP_GRC);
+	dmae->src_addr_lo = (sc->port.port_stx >> 2);
+	dmae->src_addr_hi = 0;
+	dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, port_stats));
+	dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, port_stats));
+	dmae->len = DMAE_LEN32_RD_MAX;
+	dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+	dmae->comp_addr_hi = 0;
+	dmae->comp_val = 1;
+
+	dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+	dmae->opcode = bcm_dmae_opcode_add_comp(opcode, DMAE_COMP_PCI);
+	dmae->src_addr_lo = ((sc->port.port_stx >> 2) + DMAE_LEN32_RD_MAX);
+	dmae->src_addr_hi = 0;
+	dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, port_stats) +
+			DMAE_LEN32_RD_MAX * 4);
+	dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, port_stats) +
+			DMAE_LEN32_RD_MAX * 4);
+	dmae->len = (bcm_get_port_stats_dma_len(sc) - DMAE_LEN32_RD_MAX);
+
+	dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+	dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+	dmae->comp_val = DMAE_COMP_VAL;
+
+	*stats_comp = 0;
+	bcm_hw_stats_post(sc);
+	bcm_stats_comp(sc);
+}
+
+static void
+bcm_port_stats_init(struct bcm_softc *sc)
+{
+    struct dmae_command *dmae;
+    int port = SC_PORT(sc);
+    uint32_t opcode;
+    int loader_idx = PMF_DMAE_C(sc);
+    uint32_t mac_addr;
+    uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+    /* sanity */
+    if (!sc->link_vars.link_up || !sc->port.pmf) {
+        PMD_DRV_LOG(ERR, "BUG!");
+        return;
+    }
+
+    sc->executer_idx = 0;
+
+    /* MCP */
+    opcode = bcm_dmae_opcode(sc, DMAE_SRC_PCI, DMAE_DST_GRC,
+                             TRUE, DMAE_COMP_GRC);
+
+    if (sc->port.port_stx) {
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, port_stats));
+        dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, port_stats));
+        dmae->dst_addr_lo = sc->port.port_stx >> 2;
+        dmae->dst_addr_hi = 0;
+        dmae->len = bcm_get_port_stats_dma_len(sc);
+        dmae->comp_addr_lo = dmae_reg_go_c[loader_idx] >> 2;
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+    }
+
+    if (sc->func_stx) {
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, func_stats));
+        dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, func_stats));
+        dmae->dst_addr_lo = (sc->func_stx >> 2);
+        dmae->dst_addr_hi = 0;
+        dmae->len = (sizeof(struct host_func_stats) >> 2);
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+    }
+
+    /* MAC */
+    opcode = bcm_dmae_opcode(sc, DMAE_SRC_GRC, DMAE_DST_PCI,
+                             TRUE, DMAE_COMP_GRC);
+
+    /* EMAC is special */
+    if (sc->link_vars.mac_type == ELINK_MAC_TYPE_EMAC) {
+        mac_addr = (port ? GRCBASE_EMAC1 : GRCBASE_EMAC0);
+
+        /* EMAC_REG_EMAC_RX_STAT_AC (EMAC_REG_EMAC_RX_STAT_AC_COUNT)*/
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = (mac_addr + EMAC_REG_EMAC_RX_STAT_AC) >> 2;
+        dmae->src_addr_hi = 0;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, mac_stats));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, mac_stats));
+        dmae->len = EMAC_REG_EMAC_RX_STAT_AC_COUNT;
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+
+        /* EMAC_REG_EMAC_RX_STAT_AC_28 */
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = ((mac_addr + EMAC_REG_EMAC_RX_STAT_AC_28) >> 2);
+        dmae->src_addr_hi = 0;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, mac_stats) +
+                                   offsetof(struct emac_stats,
+                                            rx_stat_falsecarriererrors));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, mac_stats) +
+                                   offsetof(struct emac_stats,
+                                            rx_stat_falsecarriererrors));
+        dmae->len = 1;
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+
+        /* EMAC_REG_EMAC_TX_STAT_AC (EMAC_REG_EMAC_TX_STAT_AC_COUNT)*/
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = ((mac_addr + EMAC_REG_EMAC_TX_STAT_AC) >> 2);
+        dmae->src_addr_hi = 0;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, mac_stats) +
+                                   offsetof(struct emac_stats,
+                                            tx_stat_ifhcoutoctets));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, mac_stats) +
+                                   offsetof(struct emac_stats,
+                                            tx_stat_ifhcoutoctets));
+        dmae->len = EMAC_REG_EMAC_TX_STAT_AC_COUNT;
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+    } else {
+        uint32_t tx_src_addr_lo, rx_src_addr_lo;
+        uint16_t rx_len, tx_len;
+
+        /* configure the params according to MAC type */
+        switch (sc->link_vars.mac_type) {
+        case ELINK_MAC_TYPE_BMAC:
+            mac_addr = (port) ? NIG_REG_INGRESS_BMAC1_MEM :
+                                NIG_REG_INGRESS_BMAC0_MEM;
+
+            /* BIGMAC_REGISTER_TX_STAT_GTPKT ..
+               BIGMAC_REGISTER_TX_STAT_GTBYT */
+            if (CHIP_IS_E1x(sc)) {
+                tx_src_addr_lo =
+                    ((mac_addr + BIGMAC_REGISTER_TX_STAT_GTPKT) >> 2);
+                tx_len = ((8 + BIGMAC_REGISTER_TX_STAT_GTBYT -
+                           BIGMAC_REGISTER_TX_STAT_GTPKT) >> 2);
+                rx_src_addr_lo =
+                    ((mac_addr + BIGMAC_REGISTER_RX_STAT_GR64) >> 2);
+                rx_len = ((8 + BIGMAC_REGISTER_RX_STAT_GRIPJ -
+                           BIGMAC_REGISTER_RX_STAT_GR64) >> 2);
+            } else {
+                tx_src_addr_lo =
+                    ((mac_addr + BIGMAC2_REGISTER_TX_STAT_GTPOK) >> 2);
+                tx_len = ((8 + BIGMAC2_REGISTER_TX_STAT_GTBYT -
+                           BIGMAC2_REGISTER_TX_STAT_GTPOK) >> 2);
+                rx_src_addr_lo =
+                    ((mac_addr + BIGMAC2_REGISTER_RX_STAT_GR64) >> 2);
+                rx_len = ((8 + BIGMAC2_REGISTER_RX_STAT_GRIPJ -
+                           BIGMAC2_REGISTER_RX_STAT_GR64) >> 2);
+            }
+
+            break;
+
+        case ELINK_MAC_TYPE_UMAC: /* handled by MSTAT */
+        case ELINK_MAC_TYPE_XMAC: /* handled by MSTAT */
+        default:
+            mac_addr = (port) ? GRCBASE_MSTAT1 : GRCBASE_MSTAT0;
+            tx_src_addr_lo = ((mac_addr + MSTAT_REG_TX_STAT_GTXPOK_LO) >> 2);
+            rx_src_addr_lo = ((mac_addr + MSTAT_REG_RX_STAT_GR64_LO) >> 2);
+            tx_len =
+                (sizeof(sc->sp->mac_stats.mstat_stats.stats_tx) >> 2);
+            rx_len =
+                (sizeof(sc->sp->mac_stats.mstat_stats.stats_rx) >> 2);
+            break;
+        }
+
+        /* TX stats */
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo = tx_src_addr_lo;
+        dmae->src_addr_hi = 0;
+        dmae->len = tx_len;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, mac_stats));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, mac_stats));
+        dmae->comp_addr_lo = dmae_reg_go_c[loader_idx] >> 2;
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+
+        /* RX stats */
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_hi = 0;
+        dmae->src_addr_lo = rx_src_addr_lo;
+        dmae->dst_addr_lo =
+            U64_LO(BCM_SP_MAPPING(sc, mac_stats) + (tx_len << 2));
+        dmae->dst_addr_hi =
+            U64_HI(BCM_SP_MAPPING(sc, mac_stats) + (tx_len << 2));
+        dmae->len = rx_len;
+        dmae->comp_addr_lo = dmae_reg_go_c[loader_idx] >> 2;
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+    }
+
+    /* NIG */
+    if (!CHIP_IS_E3(sc)) {
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo =
+            (port ? NIG_REG_STAT1_EGRESS_MAC_PKT0 :
+                    NIG_REG_STAT0_EGRESS_MAC_PKT0) >> 2;
+        dmae->src_addr_hi = 0;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, nig_stats) +
+                                   offsetof(struct nig_stats,
+                                            egress_mac_pkt0_lo));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, nig_stats) +
+                                   offsetof(struct nig_stats,
+                                            egress_mac_pkt0_lo));
+        dmae->len = ((2 * sizeof(uint32_t)) >> 2);
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = opcode;
+        dmae->src_addr_lo =
+            (port ? NIG_REG_STAT1_EGRESS_MAC_PKT1 :
+                    NIG_REG_STAT0_EGRESS_MAC_PKT1) >> 2;
+        dmae->src_addr_hi = 0;
+        dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, nig_stats) +
+                                   offsetof(struct nig_stats,
+                                            egress_mac_pkt1_lo));
+        dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, nig_stats) +
+                                   offsetof(struct nig_stats,
+                                            egress_mac_pkt1_lo));
+        dmae->len = ((2 * sizeof(uint32_t)) >> 2);
+        dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+        dmae->comp_addr_hi = 0;
+        dmae->comp_val = 1;
+    }
+
+    dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+    dmae->opcode = bcm_dmae_opcode(sc, DMAE_SRC_GRC, DMAE_DST_PCI,
+                                   TRUE, DMAE_COMP_PCI);
+    dmae->src_addr_lo =
+        (port ? NIG_REG_STAT1_BRB_DISCARD :
+                NIG_REG_STAT0_BRB_DISCARD) >> 2;
+    dmae->src_addr_hi = 0;
+    dmae->dst_addr_lo = U64_LO(BCM_SP_MAPPING(sc, nig_stats));
+    dmae->dst_addr_hi = U64_HI(BCM_SP_MAPPING(sc, nig_stats));
+    dmae->len = (sizeof(struct nig_stats) - 4*sizeof(uint32_t)) >> 2;
+
+    dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_val = DMAE_COMP_VAL;
+
+    *stats_comp = 0;
+}
+
+static void
+bcm_func_stats_init(struct bcm_softc *sc)
+{
+    struct dmae_command *dmae = &sc->stats_dmae;
+    uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+    /* sanity */
+    if (!sc->func_stx) {
+        PMD_DRV_LOG(ERR, "BUG!");
+        return;
+    }
+
+    sc->executer_idx = 0;
+    memset(dmae, 0, sizeof(struct dmae_command));
+
+    dmae->opcode = bcm_dmae_opcode(sc, DMAE_SRC_PCI, DMAE_DST_GRC,
+                                   TRUE, DMAE_COMP_PCI);
+    dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, func_stats));
+    dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, func_stats));
+    dmae->dst_addr_lo = (sc->func_stx >> 2);
+    dmae->dst_addr_hi = 0;
+    dmae->len = (sizeof(struct host_func_stats) >> 2);
+    dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_val = DMAE_COMP_VAL;
+
+    *stats_comp = 0;
+}
+
+static void
+bcm_stats_start(struct bcm_softc *sc)
+{
+    /*
+     * VFs travel through here as part of the statistics FSM, but no action
+     * is required
+     */
+    if (IS_VF(sc)) {
+        return;
+    }
+
+    if (sc->port.pmf) {
+        bcm_port_stats_init(sc);
+    }
+
+    else if (sc->func_stx) {
+        bcm_func_stats_init(sc);
+    }
+
+    bcm_hw_stats_post(sc);
+    bcm_storm_stats_post(sc);
+}
+
+static void
+bcm_stats_pmf_start(struct bcm_softc *sc)
+{
+    bcm_stats_comp(sc);
+    bcm_stats_pmf_update(sc);
+    bcm_stats_start(sc);
+}
+
+static void
+bcm_stats_restart(struct bcm_softc *sc)
+{
+    /*
+     * VFs travel through here as part of the statistics FSM, but no action
+     * is required
+     */
+    if (IS_VF(sc)) {
+        return;
+    }
+
+    bcm_stats_comp(sc);
+    bcm_stats_start(sc);
+}
+
+static void
+bcm_bmac_stats_update(struct bcm_softc *sc)
+{
+    struct host_port_stats *pstats = BCM_SP(sc, port_stats);
+    struct bcm_eth_stats *estats = &sc->eth_stats;
+    struct {
+        uint32_t lo;
+        uint32_t hi;
+    } diff;
+
+    if (CHIP_IS_E1x(sc)) {
+        struct bmac1_stats *new = BCM_SP(sc, mac_stats.bmac1_stats);
+
+        /* the macros below will use "bmac1_stats" type */
+        UPDATE_STAT64(rx_stat_grerb, rx_stat_ifhcinbadoctets);
+        UPDATE_STAT64(rx_stat_grfcs, rx_stat_dot3statsfcserrors);
+        UPDATE_STAT64(rx_stat_grund, rx_stat_etherstatsundersizepkts);
+        UPDATE_STAT64(rx_stat_grovr, rx_stat_dot3statsframestoolong);
+        UPDATE_STAT64(rx_stat_grfrg, rx_stat_etherstatsfragments);
+        UPDATE_STAT64(rx_stat_grjbr, rx_stat_etherstatsjabbers);
+        UPDATE_STAT64(rx_stat_grxcf, rx_stat_maccontrolframesreceived);
+        UPDATE_STAT64(rx_stat_grxpf, rx_stat_xoffstateentered);
+        UPDATE_STAT64(rx_stat_grxpf, rx_stat_mac_xpf);
+
+        UPDATE_STAT64(tx_stat_gtxpf, tx_stat_outxoffsent);
+        UPDATE_STAT64(tx_stat_gtxpf, tx_stat_flowcontroldone);
+        UPDATE_STAT64(tx_stat_gt64, tx_stat_etherstatspkts64octets);
+        UPDATE_STAT64(tx_stat_gt127,
+                      tx_stat_etherstatspkts65octetsto127octets);
+        UPDATE_STAT64(tx_stat_gt255,
+                      tx_stat_etherstatspkts128octetsto255octets);
+        UPDATE_STAT64(tx_stat_gt511,
+                      tx_stat_etherstatspkts256octetsto511octets);
+        UPDATE_STAT64(tx_stat_gt1023,
+                      tx_stat_etherstatspkts512octetsto1023octets);
+        UPDATE_STAT64(tx_stat_gt1518,
+                      tx_stat_etherstatspkts1024octetsto1522octets);
+        UPDATE_STAT64(tx_stat_gt2047, tx_stat_mac_2047);
+        UPDATE_STAT64(tx_stat_gt4095, tx_stat_mac_4095);
+        UPDATE_STAT64(tx_stat_gt9216, tx_stat_mac_9216);
+        UPDATE_STAT64(tx_stat_gt16383, tx_stat_mac_16383);
+        UPDATE_STAT64(tx_stat_gterr,
+                      tx_stat_dot3statsinternalmactransmiterrors);
+        UPDATE_STAT64(tx_stat_gtufl, tx_stat_mac_ufl);
+    } else {
+        struct bmac2_stats *new = BCM_SP(sc, mac_stats.bmac2_stats);
+        struct bcm_fw_port_stats_old *fwstats = &sc->fw_stats_old;
+
+        /* the macros below will use "bmac2_stats" type */
+        UPDATE_STAT64(rx_stat_grerb, rx_stat_ifhcinbadoctets);
+        UPDATE_STAT64(rx_stat_grfcs, rx_stat_dot3statsfcserrors);
+        UPDATE_STAT64(rx_stat_grund, rx_stat_etherstatsundersizepkts);
+        UPDATE_STAT64(rx_stat_grovr, rx_stat_dot3statsframestoolong);
+        UPDATE_STAT64(rx_stat_grfrg, rx_stat_etherstatsfragments);
+        UPDATE_STAT64(rx_stat_grjbr, rx_stat_etherstatsjabbers);
+        UPDATE_STAT64(rx_stat_grxcf, rx_stat_maccontrolframesreceived);
+        UPDATE_STAT64(rx_stat_grxpf, rx_stat_xoffstateentered);
+        UPDATE_STAT64(rx_stat_grxpf, rx_stat_mac_xpf);
+        UPDATE_STAT64(tx_stat_gtxpf, tx_stat_outxoffsent);
+        UPDATE_STAT64(tx_stat_gtxpf, tx_stat_flowcontroldone);
+        UPDATE_STAT64(tx_stat_gt64, tx_stat_etherstatspkts64octets);
+        UPDATE_STAT64(tx_stat_gt127,
+                      tx_stat_etherstatspkts65octetsto127octets);
+        UPDATE_STAT64(tx_stat_gt255,
+                      tx_stat_etherstatspkts128octetsto255octets);
+        UPDATE_STAT64(tx_stat_gt511,
+                      tx_stat_etherstatspkts256octetsto511octets);
+        UPDATE_STAT64(tx_stat_gt1023,
+                      tx_stat_etherstatspkts512octetsto1023octets);
+        UPDATE_STAT64(tx_stat_gt1518,
+                      tx_stat_etherstatspkts1024octetsto1522octets);
+        UPDATE_STAT64(tx_stat_gt2047, tx_stat_mac_2047);
+        UPDATE_STAT64(tx_stat_gt4095, tx_stat_mac_4095);
+        UPDATE_STAT64(tx_stat_gt9216, tx_stat_mac_9216);
+        UPDATE_STAT64(tx_stat_gt16383, tx_stat_mac_16383);
+        UPDATE_STAT64(tx_stat_gterr,
+                      tx_stat_dot3statsinternalmactransmiterrors);
+        UPDATE_STAT64(tx_stat_gtufl, tx_stat_mac_ufl);
+
+        /* collect PFC stats */
+        pstats->pfc_frames_tx_hi = new->tx_stat_gtpp_hi;
+        pstats->pfc_frames_tx_lo = new->tx_stat_gtpp_lo;
+        ADD_64(pstats->pfc_frames_tx_hi, fwstats->pfc_frames_tx_hi,
+               pstats->pfc_frames_tx_lo, fwstats->pfc_frames_tx_lo);
+
+        pstats->pfc_frames_rx_hi = new->rx_stat_grpp_hi;
+        pstats->pfc_frames_rx_lo = new->rx_stat_grpp_lo;
+        ADD_64(pstats->pfc_frames_rx_hi, fwstats->pfc_frames_rx_hi,
+               pstats->pfc_frames_rx_lo, fwstats->pfc_frames_rx_lo);
+    }
+
+    estats->pause_frames_received_hi = pstats->mac_stx[1].rx_stat_mac_xpf_hi;
+    estats->pause_frames_received_lo = pstats->mac_stx[1].rx_stat_mac_xpf_lo;
+
+    estats->pause_frames_sent_hi = pstats->mac_stx[1].tx_stat_outxoffsent_hi;
+    estats->pause_frames_sent_lo = pstats->mac_stx[1].tx_stat_outxoffsent_lo;
+
+    estats->pfc_frames_received_hi = pstats->pfc_frames_rx_hi;
+    estats->pfc_frames_received_lo = pstats->pfc_frames_rx_lo;
+    estats->pfc_frames_sent_hi = pstats->pfc_frames_tx_hi;
+    estats->pfc_frames_sent_lo = pstats->pfc_frames_tx_lo;
+}
+
+static void
+bcm_mstat_stats_update(struct bcm_softc *sc)
+{
+    struct host_port_stats *pstats = BCM_SP(sc, port_stats);
+    struct bcm_eth_stats *estats = &sc->eth_stats;
+    struct mstat_stats *new = BCM_SP(sc, mac_stats.mstat_stats);
+
+    ADD_STAT64(stats_rx.rx_grerb, rx_stat_ifhcinbadoctets);
+    ADD_STAT64(stats_rx.rx_grfcs, rx_stat_dot3statsfcserrors);
+    ADD_STAT64(stats_rx.rx_grund, rx_stat_etherstatsundersizepkts);
+    ADD_STAT64(stats_rx.rx_grovr, rx_stat_dot3statsframestoolong);
+    ADD_STAT64(stats_rx.rx_grfrg, rx_stat_etherstatsfragments);
+    ADD_STAT64(stats_rx.rx_grxcf, rx_stat_maccontrolframesreceived);
+    ADD_STAT64(stats_rx.rx_grxpf, rx_stat_xoffstateentered);
+    ADD_STAT64(stats_rx.rx_grxpf, rx_stat_mac_xpf);
+    ADD_STAT64(stats_tx.tx_gtxpf, tx_stat_outxoffsent);
+    ADD_STAT64(stats_tx.tx_gtxpf, tx_stat_flowcontroldone);
+
+    /* collect pfc stats */
+    ADD_64(pstats->pfc_frames_tx_hi, new->stats_tx.tx_gtxpp_hi,
+           pstats->pfc_frames_tx_lo, new->stats_tx.tx_gtxpp_lo);
+    ADD_64(pstats->pfc_frames_rx_hi, new->stats_rx.rx_grxpp_hi,
+           pstats->pfc_frames_rx_lo, new->stats_rx.rx_grxpp_lo);
+
+    ADD_STAT64(stats_tx.tx_gt64, tx_stat_etherstatspkts64octets);
+    ADD_STAT64(stats_tx.tx_gt127, tx_stat_etherstatspkts65octetsto127octets);
+    ADD_STAT64(stats_tx.tx_gt255, tx_stat_etherstatspkts128octetsto255octets);
+    ADD_STAT64(stats_tx.tx_gt511, tx_stat_etherstatspkts256octetsto511octets);
+    ADD_STAT64(stats_tx.tx_gt1023,
+               tx_stat_etherstatspkts512octetsto1023octets);
+    ADD_STAT64(stats_tx.tx_gt1518,
+               tx_stat_etherstatspkts1024octetsto1522octets);
+    ADD_STAT64(stats_tx.tx_gt2047, tx_stat_mac_2047);
+
+    ADD_STAT64(stats_tx.tx_gt4095, tx_stat_mac_4095);
+    ADD_STAT64(stats_tx.tx_gt9216, tx_stat_mac_9216);
+    ADD_STAT64(stats_tx.tx_gt16383, tx_stat_mac_16383);
+
+    ADD_STAT64(stats_tx.tx_gterr, tx_stat_dot3statsinternalmactransmiterrors);
+    ADD_STAT64(stats_tx.tx_gtufl, tx_stat_mac_ufl);
+
+    estats->etherstatspkts1024octetsto1522octets_hi =
+        pstats->mac_stx[1].tx_stat_etherstatspkts1024octetsto1522octets_hi;
+    estats->etherstatspkts1024octetsto1522octets_lo =
+        pstats->mac_stx[1].tx_stat_etherstatspkts1024octetsto1522octets_lo;
+
+    estats->etherstatspktsover1522octets_hi =
+        pstats->mac_stx[1].tx_stat_mac_2047_hi;
+    estats->etherstatspktsover1522octets_lo =
+        pstats->mac_stx[1].tx_stat_mac_2047_lo;
+
+    ADD_64(estats->etherstatspktsover1522octets_hi,
+           pstats->mac_stx[1].tx_stat_mac_4095_hi,
+           estats->etherstatspktsover1522octets_lo,
+           pstats->mac_stx[1].tx_stat_mac_4095_lo);
+
+    ADD_64(estats->etherstatspktsover1522octets_hi,
+           pstats->mac_stx[1].tx_stat_mac_9216_hi,
+           estats->etherstatspktsover1522octets_lo,
+           pstats->mac_stx[1].tx_stat_mac_9216_lo);
+
+    ADD_64(estats->etherstatspktsover1522octets_hi,
+           pstats->mac_stx[1].tx_stat_mac_16383_hi,
+           estats->etherstatspktsover1522octets_lo,
+           pstats->mac_stx[1].tx_stat_mac_16383_lo);
+
+    estats->pause_frames_received_hi = pstats->mac_stx[1].rx_stat_mac_xpf_hi;
+    estats->pause_frames_received_lo = pstats->mac_stx[1].rx_stat_mac_xpf_lo;
+
+    estats->pause_frames_sent_hi = pstats->mac_stx[1].tx_stat_outxoffsent_hi;
+    estats->pause_frames_sent_lo = pstats->mac_stx[1].tx_stat_outxoffsent_lo;
+
+    estats->pfc_frames_received_hi = pstats->pfc_frames_rx_hi;
+    estats->pfc_frames_received_lo = pstats->pfc_frames_rx_lo;
+    estats->pfc_frames_sent_hi = pstats->pfc_frames_tx_hi;
+    estats->pfc_frames_sent_lo = pstats->pfc_frames_tx_lo;
+}
+
+static void
+bcm_emac_stats_update(struct bcm_softc *sc)
+{
+    struct emac_stats *new = BCM_SP(sc, mac_stats.emac_stats);
+    struct host_port_stats *pstats = BCM_SP(sc, port_stats);
+    struct bcm_eth_stats *estats = &sc->eth_stats;
+
+    UPDATE_EXTEND_STAT(rx_stat_ifhcinbadoctets);
+    UPDATE_EXTEND_STAT(tx_stat_ifhcoutbadoctets);
+    UPDATE_EXTEND_STAT(rx_stat_dot3statsfcserrors);
+    UPDATE_EXTEND_STAT(rx_stat_dot3statsalignmenterrors);
+    UPDATE_EXTEND_STAT(rx_stat_dot3statscarriersenseerrors);
+    UPDATE_EXTEND_STAT(rx_stat_falsecarriererrors);
+    UPDATE_EXTEND_STAT(rx_stat_etherstatsundersizepkts);
+    UPDATE_EXTEND_STAT(rx_stat_dot3statsframestoolong);
+    UPDATE_EXTEND_STAT(rx_stat_etherstatsfragments);
+    UPDATE_EXTEND_STAT(rx_stat_etherstatsjabbers);
+    UPDATE_EXTEND_STAT(rx_stat_maccontrolframesreceived);
+    UPDATE_EXTEND_STAT(rx_stat_xoffstateentered);
+    UPDATE_EXTEND_STAT(rx_stat_xonpauseframesreceived);
+    UPDATE_EXTEND_STAT(rx_stat_xoffpauseframesreceived);
+    UPDATE_EXTEND_STAT(tx_stat_outxonsent);
+    UPDATE_EXTEND_STAT(tx_stat_outxoffsent);
+    UPDATE_EXTEND_STAT(tx_stat_flowcontroldone);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatscollisions);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statssinglecollisionframes);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statsmultiplecollisionframes);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statsdeferredtransmissions);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statsexcessivecollisions);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statslatecollisions);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts64octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts65octetsto127octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts128octetsto255octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts256octetsto511octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts512octetsto1023octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspkts1024octetsto1522octets);
+    UPDATE_EXTEND_STAT(tx_stat_etherstatspktsover1522octets);
+    UPDATE_EXTEND_STAT(tx_stat_dot3statsinternalmactransmiterrors);
+
+    estats->pause_frames_received_hi =
+        pstats->mac_stx[1].rx_stat_xonpauseframesreceived_hi;
+    estats->pause_frames_received_lo =
+        pstats->mac_stx[1].rx_stat_xonpauseframesreceived_lo;
+    ADD_64(estats->pause_frames_received_hi,
+           pstats->mac_stx[1].rx_stat_xoffpauseframesreceived_hi,
+           estats->pause_frames_received_lo,
+           pstats->mac_stx[1].rx_stat_xoffpauseframesreceived_lo);
+
+    estats->pause_frames_sent_hi =
+        pstats->mac_stx[1].tx_stat_outxonsent_hi;
+    estats->pause_frames_sent_lo =
+        pstats->mac_stx[1].tx_stat_outxonsent_lo;
+    ADD_64(estats->pause_frames_sent_hi,
+           pstats->mac_stx[1].tx_stat_outxoffsent_hi,
+           estats->pause_frames_sent_lo,
+           pstats->mac_stx[1].tx_stat_outxoffsent_lo);
+}
+
+static int
+bcm_hw_stats_update(struct bcm_softc *sc)
+{
+    struct nig_stats *new = BCM_SP(sc, nig_stats);
+    struct nig_stats *old = &(sc->port.old_nig_stats);
+    struct host_port_stats *pstats = BCM_SP(sc, port_stats);
+    struct bcm_eth_stats *estats = &sc->eth_stats;
+    uint32_t lpi_reg, nig_timer_max;
+    struct {
+        uint32_t lo;
+        uint32_t hi;
+    } diff;
+
+    switch (sc->link_vars.mac_type) {
+    case ELINK_MAC_TYPE_BMAC:
+        bcm_bmac_stats_update(sc);
+        break;
+
+    case ELINK_MAC_TYPE_EMAC:
+        bcm_emac_stats_update(sc);
+        break;
+
+    case ELINK_MAC_TYPE_UMAC:
+    case ELINK_MAC_TYPE_XMAC:
+        bcm_mstat_stats_update(sc);
+        break;
+
+    case ELINK_MAC_TYPE_NONE: /* unreached */
+        PMD_DRV_LOG(DEBUG,
+              "stats updated by DMAE but no MAC active");
+        return -1;
+
+    default: /* unreached */
+        PMD_DRV_LOG(ERR, "stats update failed, unknown MAC type");
+    }
+
+    ADD_EXTEND_64(pstats->brb_drop_hi, pstats->brb_drop_lo,
+                  new->brb_discard - old->brb_discard);
+    ADD_EXTEND_64(estats->brb_truncate_hi, estats->brb_truncate_lo,
+                  new->brb_truncate - old->brb_truncate);
+
+    if (!CHIP_IS_E3(sc)) {
+        UPDATE_STAT64_NIG(egress_mac_pkt0,
+                          etherstatspkts1024octetsto1522octets);
+        UPDATE_STAT64_NIG(egress_mac_pkt1,
+                          etherstatspktsover1522octets);
+    }
+
+    rte_memcpy(old, new, sizeof(struct nig_stats));
+
+    rte_memcpy(&(estats->rx_stat_ifhcinbadoctets_hi), &(pstats->mac_stx[1]),
+           sizeof(struct mac_stx));
+    estats->brb_drop_hi = pstats->brb_drop_hi;
+    estats->brb_drop_lo = pstats->brb_drop_lo;
+
+    pstats->host_port_stats_counter++;
+
+    if (CHIP_IS_E3(sc)) {
+        lpi_reg = (SC_PORT(sc)) ?
+                      MISC_REG_CPMU_LP_SM_ENT_CNT_P1 :
+                      MISC_REG_CPMU_LP_SM_ENT_CNT_P0;
+        estats->eee_tx_lpi += REG_RD(sc, lpi_reg);
+    }
+
+    if (!BCM_NOMCP(sc)) {
+        nig_timer_max = SHMEM_RD(sc, port_mb[SC_PORT(sc)].stat_nig_timer);
+        if (nig_timer_max != estats->nig_timer_max) {
+            estats->nig_timer_max = nig_timer_max;
+            PMD_DRV_LOG(ERR, "invalid NIG timer max (%u)",
+                  estats->nig_timer_max);
+        }
+    }
+
+    return 0;
+}
+
+static int
+bcm_storm_stats_validate_counters(struct bcm_softc *sc)
+{
+    struct stats_counter *counters = &sc->fw_stats_data->storm_counters;
+    uint16_t cur_stats_counter;
+
+    /*
+     * Make sure we use the value of the counter
+     * used for sending the last stats ramrod.
+     */
+    cur_stats_counter = (sc->stats_counter - 1);
+
+    /* are storm stats valid? */
+    if (le16toh(counters->xstats_counter) != cur_stats_counter) {
+        PMD_DRV_LOG(DEBUG,
+              "stats not updated by xstorm, "
+              "counter 0x%x != stats_counter 0x%x",
+              le16toh(counters->xstats_counter), sc->stats_counter);
+        return -EAGAIN;
+    }
+
+    if (le16toh(counters->ustats_counter) != cur_stats_counter) {
+        PMD_DRV_LOG(DEBUG,
+              "stats not updated by ustorm, "
+              "counter 0x%x != stats_counter 0x%x",
+              le16toh(counters->ustats_counter), sc->stats_counter);
+        return -EAGAIN;
+    }
+
+    if (le16toh(counters->cstats_counter) != cur_stats_counter) {
+        PMD_DRV_LOG(DEBUG,
+              "stats not updated by cstorm, "
+              "counter 0x%x != stats_counter 0x%x",
+              le16toh(counters->cstats_counter), sc->stats_counter);
+        return -EAGAIN;
+    }
+
+    if (le16toh(counters->tstats_counter) != cur_stats_counter) {
+        PMD_DRV_LOG(DEBUG,
+              "stats not updated by tstorm, "
+              "counter 0x%x != stats_counter 0x%x",
+              le16toh(counters->tstats_counter), sc->stats_counter);
+        return -EAGAIN;
+    }
+
+    return 0;
+}
+
+static int
+bcm_storm_stats_update(struct bcm_softc *sc)
+{
+	struct tstorm_per_port_stats *tport =
+		&sc->fw_stats_data->port.tstorm_port_statistics;
+	struct tstorm_per_pf_stats *tfunc =
+		&sc->fw_stats_data->pf.tstorm_pf_statistics;
+	struct host_func_stats *fstats = &sc->func_stats;
+	struct bcm_eth_stats *estats = &sc->eth_stats;
+	struct bcm_eth_stats_old *estats_old = &sc->eth_stats_old;
+	int i;
+
+	/* vfs stat counter is managed by pf */
+	if (IS_PF(sc) && bcm_storm_stats_validate_counters(sc)) {
+		return -EAGAIN;
+	}
+
+	estats->error_bytes_received_hi = 0;
+	estats->error_bytes_received_lo = 0;
+
+	for (i = 0; i < sc->num_queues; i++) {
+		struct bcm_fastpath *fp = &sc->fp[i];
+		struct tstorm_per_queue_stats *tclient =
+			&sc->fw_stats_data->queue_stats[i].tstorm_queue_statistics;
+		struct tstorm_per_queue_stats *old_tclient = &fp->old_tclient;
+		struct ustorm_per_queue_stats *uclient =
+			&sc->fw_stats_data->queue_stats[i].ustorm_queue_statistics;
+		struct ustorm_per_queue_stats *old_uclient = &fp->old_uclient;
+		struct xstorm_per_queue_stats *xclient =
+			&sc->fw_stats_data->queue_stats[i].xstorm_queue_statistics;
+		struct xstorm_per_queue_stats *old_xclient = &fp->old_xclient;
+		struct bcm_eth_q_stats *qstats = &fp->eth_q_stats;
+		struct bcm_eth_q_stats_old *qstats_old = &fp->eth_q_stats_old;
+
+		uint32_t diff;
+
+		/* PMD_DRV_LOG(DEBUG,
+				"queue[%d]: ucast_sent 0x%x bcast_sent 0x%x mcast_sent 0x%x",
+				i, xclient->ucast_pkts_sent, xclient->bcast_pkts_sent,
+				xclient->mcast_pkts_sent);
+
+		PMD_DRV_LOG(DEBUG, "---------------"); */
+
+		UPDATE_QSTAT(tclient->rcv_bcast_bytes,
+				total_broadcast_bytes_received);
+		UPDATE_QSTAT(tclient->rcv_mcast_bytes,
+				total_multicast_bytes_received);
+		UPDATE_QSTAT(tclient->rcv_ucast_bytes,
+				total_unicast_bytes_received);
+
+		/*
+		 * sum to total_bytes_received all
+		 * unicast/multicast/broadcast
+		 */
+		qstats->total_bytes_received_hi =
+			qstats->total_broadcast_bytes_received_hi;
+		qstats->total_bytes_received_lo =
+			qstats->total_broadcast_bytes_received_lo;
+
+		ADD_64(qstats->total_bytes_received_hi,
+				qstats->total_multicast_bytes_received_hi,
+				qstats->total_bytes_received_lo,
+				qstats->total_multicast_bytes_received_lo);
+
+		ADD_64(qstats->total_bytes_received_hi,
+				qstats->total_unicast_bytes_received_hi,
+				qstats->total_bytes_received_lo,
+				qstats->total_unicast_bytes_received_lo);
+
+		qstats->valid_bytes_received_hi = qstats->total_bytes_received_hi;
+		qstats->valid_bytes_received_lo = qstats->total_bytes_received_lo;
+
+		UPDATE_EXTEND_TSTAT(rcv_ucast_pkts, total_unicast_packets_received);
+		UPDATE_EXTEND_TSTAT(rcv_mcast_pkts, total_multicast_packets_received);
+		UPDATE_EXTEND_TSTAT(rcv_bcast_pkts, total_broadcast_packets_received);
+		UPDATE_EXTEND_E_TSTAT(pkts_too_big_discard,
+				etherstatsoverrsizepkts, 32);
+		UPDATE_EXTEND_E_TSTAT(no_buff_discard, no_buff_discard, 16);
+
+		SUB_EXTEND_USTAT(ucast_no_buff_pkts, total_unicast_packets_received);
+		SUB_EXTEND_USTAT(mcast_no_buff_pkts,
+				total_multicast_packets_received);
+		SUB_EXTEND_USTAT(bcast_no_buff_pkts,
+				total_broadcast_packets_received);
+		UPDATE_EXTEND_E_USTAT(ucast_no_buff_pkts, no_buff_discard);
+		UPDATE_EXTEND_E_USTAT(mcast_no_buff_pkts, no_buff_discard);
+		UPDATE_EXTEND_E_USTAT(bcast_no_buff_pkts, no_buff_discard);
+
+		UPDATE_QSTAT(xclient->bcast_bytes_sent,
+				total_broadcast_bytes_transmitted);
+		UPDATE_QSTAT(xclient->mcast_bytes_sent,
+				total_multicast_bytes_transmitted);
+		UPDATE_QSTAT(xclient->ucast_bytes_sent,
+				total_unicast_bytes_transmitted);
+
+		/*
+		 * sum to total_bytes_transmitted all
+		 * unicast/multicast/broadcast
+		 */
+		qstats->total_bytes_transmitted_hi =
+			qstats->total_unicast_bytes_transmitted_hi;
+		qstats->total_bytes_transmitted_lo =
+			qstats->total_unicast_bytes_transmitted_lo;
+
+		ADD_64(qstats->total_bytes_transmitted_hi,
+				qstats->total_broadcast_bytes_transmitted_hi,
+				qstats->total_bytes_transmitted_lo,
+				qstats->total_broadcast_bytes_transmitted_lo);
+
+		ADD_64(qstats->total_bytes_transmitted_hi,
+				qstats->total_multicast_bytes_transmitted_hi,
+				qstats->total_bytes_transmitted_lo,
+				qstats->total_multicast_bytes_transmitted_lo);
+
+		UPDATE_EXTEND_XSTAT(ucast_pkts_sent,
+				total_unicast_packets_transmitted);
+		UPDATE_EXTEND_XSTAT(mcast_pkts_sent,
+				total_multicast_packets_transmitted);
+		UPDATE_EXTEND_XSTAT(bcast_pkts_sent,
+				total_broadcast_packets_transmitted);
+
+		UPDATE_EXTEND_TSTAT(checksum_discard,
+				total_packets_received_checksum_discarded);
+		UPDATE_EXTEND_TSTAT(ttl0_discard,
+				total_packets_received_ttl0_discarded);
+
+		UPDATE_EXTEND_XSTAT(error_drop_pkts,
+				total_transmitted_dropped_packets_error);
+
+		UPDATE_FSTAT_QSTAT(total_bytes_received);
+		UPDATE_FSTAT_QSTAT(total_bytes_transmitted);
+		UPDATE_FSTAT_QSTAT(total_unicast_packets_received);
+		UPDATE_FSTAT_QSTAT(total_multicast_packets_received);
+		UPDATE_FSTAT_QSTAT(total_broadcast_packets_received);
+		UPDATE_FSTAT_QSTAT(total_unicast_packets_transmitted);
+		UPDATE_FSTAT_QSTAT(total_multicast_packets_transmitted);
+		UPDATE_FSTAT_QSTAT(total_broadcast_packets_transmitted);
+		UPDATE_FSTAT_QSTAT(valid_bytes_received);
+	}
+
+	ADD_64(estats->total_bytes_received_hi,
+			estats->rx_stat_ifhcinbadoctets_hi,
+			estats->total_bytes_received_lo,
+			estats->rx_stat_ifhcinbadoctets_lo);
+
+	ADD_64_LE(estats->total_bytes_received_hi,
+			tfunc->rcv_error_bytes.hi,
+			estats->total_bytes_received_lo,
+			tfunc->rcv_error_bytes.lo);
+
+	ADD_64_LE(estats->error_bytes_received_hi,
+			tfunc->rcv_error_bytes.hi,
+			estats->error_bytes_received_lo,
+			tfunc->rcv_error_bytes.lo);
+
+	UPDATE_ESTAT(etherstatsoverrsizepkts, rx_stat_dot3statsframestoolong);
+
+	ADD_64(estats->error_bytes_received_hi,
+			estats->rx_stat_ifhcinbadoctets_hi,
+			estats->error_bytes_received_lo,
+			estats->rx_stat_ifhcinbadoctets_lo);
+
+	if (sc->port.pmf) {
+		struct bcm_fw_port_stats_old *fwstats = &sc->fw_stats_old;
+		UPDATE_FW_STAT(mac_filter_discard);
+		UPDATE_FW_STAT(mf_tag_discard);
+		UPDATE_FW_STAT(brb_truncate_discard);
+		UPDATE_FW_STAT(mac_discard);
+	}
+
+	fstats->host_func_stats_start = ++fstats->host_func_stats_end;
+
+	sc->stats_pending = 0;
+
+	return 0;
+}
+
+static void
+bcm_drv_stats_update(struct bcm_softc *sc)
+{
+    struct bcm_eth_stats *estats = &sc->eth_stats;
+    int i;
+
+    for (i = 0; i < sc->num_queues; i++) {
+        struct bcm_eth_q_stats *qstats = &sc->fp[i].eth_q_stats;
+        struct bcm_eth_q_stats_old *qstats_old = &sc->fp[i].eth_q_stats_old;
+
+        UPDATE_ESTAT_QSTAT(rx_calls);
+        UPDATE_ESTAT_QSTAT(rx_pkts);
+        UPDATE_ESTAT_QSTAT(rx_soft_errors);
+        UPDATE_ESTAT_QSTAT(rx_hw_csum_errors);
+        UPDATE_ESTAT_QSTAT(rx_ofld_frames_csum_ip);
+        UPDATE_ESTAT_QSTAT(rx_ofld_frames_csum_tcp_udp);
+        UPDATE_ESTAT_QSTAT(rx_budget_reached);
+        UPDATE_ESTAT_QSTAT(tx_pkts);
+        UPDATE_ESTAT_QSTAT(tx_soft_errors);
+        UPDATE_ESTAT_QSTAT(tx_ofld_frames_csum_ip);
+        UPDATE_ESTAT_QSTAT(tx_ofld_frames_csum_tcp);
+        UPDATE_ESTAT_QSTAT(tx_ofld_frames_csum_udp);
+        UPDATE_ESTAT_QSTAT(tx_encap_failures);
+        UPDATE_ESTAT_QSTAT(tx_hw_queue_full);
+        UPDATE_ESTAT_QSTAT(tx_hw_max_queue_depth);
+        UPDATE_ESTAT_QSTAT(tx_dma_mapping_failure);
+        UPDATE_ESTAT_QSTAT(tx_max_drbr_queue_depth);
+        UPDATE_ESTAT_QSTAT(tx_window_violation_std);
+        UPDATE_ESTAT_QSTAT(tx_chain_lost_mbuf);
+        UPDATE_ESTAT_QSTAT(tx_frames_deferred);
+        UPDATE_ESTAT_QSTAT(tx_queue_xoff);
+
+        /* mbuf driver statistics */
+        UPDATE_ESTAT_QSTAT(mbuf_defrag_attempts);
+        UPDATE_ESTAT_QSTAT(mbuf_defrag_failures);
+        UPDATE_ESTAT_QSTAT(mbuf_rx_bd_alloc_failed);
+        UPDATE_ESTAT_QSTAT(mbuf_rx_bd_mapping_failed);
+
+        /* track the number of allocated mbufs */
+        UPDATE_ESTAT_QSTAT(mbuf_alloc_tx);
+        UPDATE_ESTAT_QSTAT(mbuf_alloc_rx);
+    }
+}
+
+static uint8_t
+bcm_edebug_stats_stopped(struct bcm_softc *sc)
+{
+    uint32_t val;
+
+    if (SHMEM2_HAS(sc, edebug_driver_if[1])) {
+        val = SHMEM2_RD(sc, edebug_driver_if[1]);
+
+        if (val == EDEBUG_DRIVER_IF_OP_CODE_DISABLE_STAT) {
+            return TRUE;
+        }
+    }
+
+    return FALSE;
+}
+
+static void
+bcm_stats_update(struct bcm_softc *sc)
+{
+	uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+	if (bcm_edebug_stats_stopped(sc)) {
+		return;
+	}
+
+	if (IS_PF(sc)) {
+
+		bcm_storm_stats_update(sc);
+		bcm_hw_stats_post(sc);
+		bcm_storm_stats_post(sc);
+		DELAY_MS(5);
+
+		if (*stats_comp != DMAE_COMP_VAL) {
+			return;
+		}
+
+		if (sc->port.pmf) {
+			bcm_hw_stats_update(sc);
+		}
+
+		if (bcm_storm_stats_update(sc)) {
+			if (sc->stats_pending++ == 3) {
+				rte_panic("storm stats not updated for 3 times");
+			}
+			return;
+		}
+	} else {
+		/*
+		 * VF doesn't collect HW statistics, and doesn't get completions,
+		 * performs only update.
+		 */
+		bcm_storm_stats_update(sc);
+	}
+
+	bcm_drv_stats_update(sc);
+}
+
+static void
+bcm_port_stats_stop(struct bcm_softc *sc)
+{
+    struct dmae_command *dmae;
+    uint32_t opcode;
+    int loader_idx = PMF_DMAE_C(sc);
+    uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+    sc->executer_idx = 0;
+
+    opcode = bcm_dmae_opcode(sc, DMAE_SRC_PCI, DMAE_DST_GRC, FALSE, 0);
+
+    if (sc->port.port_stx) {
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+
+        if (sc->func_stx) {
+            dmae->opcode = bcm_dmae_opcode_add_comp(opcode, DMAE_COMP_GRC);
+        } else {
+            dmae->opcode = bcm_dmae_opcode_add_comp(opcode, DMAE_COMP_PCI);
+        }
+
+        dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, port_stats));
+        dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, port_stats));
+        dmae->dst_addr_lo = sc->port.port_stx >> 2;
+        dmae->dst_addr_hi = 0;
+        dmae->len = bcm_get_port_stats_dma_len(sc);
+        if (sc->func_stx) {
+            dmae->comp_addr_lo = (dmae_reg_go_c[loader_idx] >> 2);
+            dmae->comp_addr_hi = 0;
+            dmae->comp_val = 1;
+        } else {
+            dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+            dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+            dmae->comp_val = DMAE_COMP_VAL;
+
+            *stats_comp = 0;
+        }
+    }
+
+    if (sc->func_stx) {
+        dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+        dmae->opcode = bcm_dmae_opcode_add_comp(opcode, DMAE_COMP_PCI);
+        dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, func_stats));
+        dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, func_stats));
+        dmae->dst_addr_lo = (sc->func_stx >> 2);
+        dmae->dst_addr_hi = 0;
+        dmae->len = (sizeof(struct host_func_stats) >> 2);
+        dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+        dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+        dmae->comp_val = DMAE_COMP_VAL;
+
+        *stats_comp = 0;
+    }
+}
+
+static void
+bcm_stats_stop(struct bcm_softc *sc)
+{
+    uint8_t update = FALSE;
+
+    bcm_stats_comp(sc);
+
+    if (sc->port.pmf) {
+        update = bcm_hw_stats_update(sc) == 0;
+    }
+
+    update |= bcm_storm_stats_update(sc) == 0;
+
+    if (update) {
+
+        if (sc->port.pmf) {
+            bcm_port_stats_stop(sc);
+        }
+
+        bcm_hw_stats_post(sc);
+        bcm_stats_comp(sc);
+    }
+}
+
+static void
+bcm_stats_do_nothing(__rte_unused struct bcm_softc *sc)
+{
+    return;
+}
+
+static const struct {
+    void (*action)(struct bcm_softc *sc);
+    enum bcm_stats_state next_state;
+} bcm_stats_stm[STATS_STATE_MAX][STATS_EVENT_MAX] = {
+    {
+    /* DISABLED PMF */ { bcm_stats_pmf_update, STATS_STATE_DISABLED },
+    /*      LINK_UP */ { bcm_stats_start,      STATS_STATE_ENABLED },
+    /*      UPDATE  */ { bcm_stats_do_nothing, STATS_STATE_DISABLED },
+    /*      STOP    */ { bcm_stats_do_nothing, STATS_STATE_DISABLED }
+    },
+    {
+    /* ENABLED  PMF */ { bcm_stats_pmf_start,  STATS_STATE_ENABLED },
+    /*      LINK_UP */ { bcm_stats_restart,    STATS_STATE_ENABLED },
+    /*      UPDATE  */ { bcm_stats_update,     STATS_STATE_ENABLED },
+    /*      STOP    */ { bcm_stats_stop,       STATS_STATE_DISABLED }
+    }
+};
+
+void bcm_stats_handle(struct bcm_softc *sc, enum bcm_stats_event event)
+{
+	enum bcm_stats_state state;
+
+	if (unlikely(sc->panic)) {
+		return;
+	}
+
+	state = sc->stats_state;
+	sc->stats_state = bcm_stats_stm[state][event].next_state;
+
+	bcm_stats_stm[state][event].action(sc);
+
+	if (event != STATS_EVENT_UPDATE) {
+		PMD_DRV_LOG(DEBUG,
+				"state %d -> event %d -> state %d",
+				state, event, sc->stats_state);
+	}
+}
+
+static void
+bcm_port_stats_base_init(struct bcm_softc *sc)
+{
+    struct dmae_command *dmae;
+    uint32_t *stats_comp = BCM_SP(sc, stats_comp);
+
+    /* sanity */
+    if (!sc->port.pmf || !sc->port.port_stx) {
+        PMD_DRV_LOG(ERR, "BUG!");
+        return;
+    }
+
+    sc->executer_idx = 0;
+
+    dmae = BCM_SP(sc, dmae[sc->executer_idx++]);
+    dmae->opcode = bcm_dmae_opcode(sc, DMAE_SRC_PCI, DMAE_DST_GRC,
+                                   TRUE, DMAE_COMP_PCI);
+    dmae->src_addr_lo = U64_LO(BCM_SP_MAPPING(sc, port_stats));
+    dmae->src_addr_hi = U64_HI(BCM_SP_MAPPING(sc, port_stats));
+    dmae->dst_addr_lo = (sc->port.port_stx >> 2);
+    dmae->dst_addr_hi = 0;
+    dmae->len = bcm_get_port_stats_dma_len(sc);
+    dmae->comp_addr_lo = U64_LO(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_addr_hi = U64_HI(BCM_SP_MAPPING(sc, stats_comp));
+    dmae->comp_val = DMAE_COMP_VAL;
+
+    *stats_comp = 0;
+    bcm_hw_stats_post(sc);
+    bcm_stats_comp(sc);
+}
+
+/*
+ * This function will prepare the statistics ramrod data the way
+ * we will only have to increment the statistics counter and
+ * send the ramrod each time we have to.
+ */
+static void
+bcm_prep_fw_stats_req(struct bcm_softc *sc)
+{
+    int i;
+    int first_queue_query_index;
+    struct stats_query_header *stats_hdr = &sc->fw_stats_req->hdr;
+    phys_addr_t cur_data_offset;
+    struct stats_query_entry *cur_query_entry;
+
+    stats_hdr->cmd_num = sc->fw_stats_num;
+    stats_hdr->drv_stats_counter = 0;
+
+    /*
+     * The storm_counters struct contains the counters of completed
+     * statistics requests per storm which are incremented by FW
+     * each time it completes hadning a statistics ramrod. We will
+     * check these counters in the timer handler and discard a
+     * (statistics) ramrod completion.
+     */
+    cur_data_offset = (sc->fw_stats_data_mapping +
+                       offsetof(struct bcm_fw_stats_data, storm_counters));
+
+    stats_hdr->stats_counters_addrs.hi = htole32(U64_HI(cur_data_offset));
+    stats_hdr->stats_counters_addrs.lo = htole32(U64_LO(cur_data_offset));
+
+    /*
+     * Prepare the first stats ramrod (will be completed with
+     * the counters equal to zero) - init counters to somethig different.
+     */
+    memset(&sc->fw_stats_data->storm_counters, 0xff,
+           sizeof(struct stats_counter));
+
+    /**** Port FW statistics data ****/
+    cur_data_offset = (sc->fw_stats_data_mapping +
+                       offsetof(struct bcm_fw_stats_data, port));
+
+    cur_query_entry = &sc->fw_stats_req->query[BCM_PORT_QUERY_IDX];
+
+    cur_query_entry->kind = STATS_TYPE_PORT;
+    /* For port query index is a DONT CARE */
+    cur_query_entry->index = SC_PORT(sc);
+    /* For port query funcID is a DONT CARE */
+    cur_query_entry->funcID = htole16(SC_FUNC(sc));
+    cur_query_entry->address.hi = htole32(U64_HI(cur_data_offset));
+    cur_query_entry->address.lo = htole32(U64_LO(cur_data_offset));
+
+    /**** PF FW statistics data ****/
+    cur_data_offset = (sc->fw_stats_data_mapping +
+                       offsetof(struct bcm_fw_stats_data, pf));
+
+    cur_query_entry = &sc->fw_stats_req->query[BCM_PF_QUERY_IDX];
+
+    cur_query_entry->kind = STATS_TYPE_PF;
+    /* For PF query index is a DONT CARE */
+    cur_query_entry->index = SC_PORT(sc);
+    cur_query_entry->funcID = htole16(SC_FUNC(sc));
+    cur_query_entry->address.hi = htole32(U64_HI(cur_data_offset));
+    cur_query_entry->address.lo = htole32(U64_LO(cur_data_offset));
+
+    /**** Clients' queries ****/
+    cur_data_offset = (sc->fw_stats_data_mapping +
+                       offsetof(struct bcm_fw_stats_data, queue_stats));
+
+    /*
+     * First queue query index depends whether FCoE offloaded request will
+     * be included in the ramrod
+     */
+        first_queue_query_index = (BCM_FIRST_QUEUE_QUERY_IDX - 1);
+
+    for (i = 0; i < sc->num_queues; i++) {
+        cur_query_entry =
+            &sc->fw_stats_req->query[first_queue_query_index + i];
+
+        cur_query_entry->kind = STATS_TYPE_QUEUE;
+        cur_query_entry->index = bcm_stats_id(&sc->fp[i]);
+        cur_query_entry->funcID = htole16(SC_FUNC(sc));
+        cur_query_entry->address.hi = htole32(U64_HI(cur_data_offset));
+        cur_query_entry->address.lo = htole32(U64_LO(cur_data_offset));
+
+        cur_data_offset += sizeof(struct per_queue_stats);
+    }
+}
+
+void bcm_memset_stats(struct bcm_softc *sc)
+{
+	int i;
+
+	/* function stats */
+	for (i = 0; i < sc->num_queues; i++) {
+		struct bcm_fastpath *fp = &sc->fp[i];
+
+		memset(&fp->old_tclient, 0,
+				sizeof(fp->old_tclient));
+		memset(&fp->old_uclient, 0,
+				sizeof(fp->old_uclient));
+		memset(&fp->old_xclient, 0,
+				sizeof(fp->old_xclient));
+		if (sc->stats_init) {
+			memset(&fp->eth_q_stats, 0,
+					sizeof(fp->eth_q_stats));
+			memset(&fp->eth_q_stats_old, 0,
+					sizeof(fp->eth_q_stats_old));
+		}
+	}
+
+	if (sc->stats_init) {
+		memset(&sc->net_stats_old, 0, sizeof(sc->net_stats_old));
+		memset(&sc->fw_stats_old, 0, sizeof(sc->fw_stats_old));
+		memset(&sc->eth_stats_old, 0, sizeof(sc->eth_stats_old));
+		memset(&sc->eth_stats, 0, sizeof(sc->eth_stats));
+		memset(&sc->func_stats, 0, sizeof(sc->func_stats));
+	}
+
+	sc->stats_state = STATS_STATE_DISABLED;
+
+	if (sc->port.pmf && sc->port.port_stx)
+		bcm_port_stats_base_init(sc);
+
+	/* mark the end of statistics initializiation */
+	sc->stats_init = false;
+}
+
+void
+bcm_stats_init(struct bcm_softc *sc)
+{
+	int /*abs*/port = SC_PORT(sc);
+	int mb_idx = SC_FW_MB_IDX(sc);
+	int i;
+
+	sc->stats_pending = 0;
+	sc->executer_idx = 0;
+	sc->stats_counter = 0;
+
+	sc->stats_init = TRUE;
+
+	/* port and func stats for management */
+	if (!BCM_NOMCP(sc)) {
+		sc->port.port_stx = SHMEM_RD(sc, port_mb[port].port_stx);
+		sc->func_stx = SHMEM_RD(sc, func_mb[mb_idx].fw_mb_param);
+	} else {
+		sc->port.port_stx = 0;
+		sc->func_stx = 0;
+	}
+
+	PMD_DRV_LOG(DEBUG, "port_stx 0x%x func_stx 0x%x",
+			sc->port.port_stx, sc->func_stx);
+
+	/* pmf should retrieve port statistics from SP on a non-init*/
+	if (!sc->stats_init && sc->port.pmf && sc->port.port_stx) {
+		bcm_stats_handle(sc, STATS_EVENT_PMF);
+	}
+
+	port = SC_PORT(sc);
+	/* port stats */
+	memset(&(sc->port.old_nig_stats), 0, sizeof(struct nig_stats));
+	sc->port.old_nig_stats.brb_discard =
+		REG_RD(sc, NIG_REG_STAT0_BRB_DISCARD + port*0x38);
+	sc->port.old_nig_stats.brb_truncate =
+		REG_RD(sc, NIG_REG_STAT0_BRB_TRUNCATE + port*0x38);
+	if (!CHIP_IS_E3(sc)) {
+		REG_RD_DMAE(sc, NIG_REG_STAT0_EGRESS_MAC_PKT0 + port*0x50,
+				&(sc->port.old_nig_stats.egress_mac_pkt0_lo), 2);
+		REG_RD_DMAE(sc, NIG_REG_STAT0_EGRESS_MAC_PKT1 + port*0x50,
+				&(sc->port.old_nig_stats.egress_mac_pkt1_lo), 2);
+	}
+
+	/* function stats */
+	for (i = 0; i < sc->num_queues; i++) {
+		memset(&sc->fp[i].old_tclient, 0, sizeof(sc->fp[i].old_tclient));
+		memset(&sc->fp[i].old_uclient, 0, sizeof(sc->fp[i].old_uclient));
+		memset(&sc->fp[i].old_xclient, 0, sizeof(sc->fp[i].old_xclient));
+		if (sc->stats_init) {
+			memset(&sc->fp[i].eth_q_stats, 0,
+					sizeof(sc->fp[i].eth_q_stats));
+			memset(&sc->fp[i].eth_q_stats_old, 0,
+					sizeof(sc->fp[i].eth_q_stats_old));
+		}
+	}
+
+	/* prepare statistics ramrod data */
+	bcm_prep_fw_stats_req(sc);
+
+	if (sc->stats_init) {
+		memset(&sc->net_stats_old, 0, sizeof(sc->net_stats_old));
+		memset(&sc->fw_stats_old, 0, sizeof(sc->fw_stats_old));
+		memset(&sc->eth_stats_old, 0, sizeof(sc->eth_stats_old));
+		memset(&sc->eth_stats, 0, sizeof(sc->eth_stats));
+		memset(&sc->func_stats, 0, sizeof(sc->func_stats));
+
+		/* Clean SP from previous statistics */
+		if (sc->func_stx) {
+			memset(BCM_SP(sc, func_stats), 0, sizeof(struct host_func_stats));
+			bcm_func_stats_init(sc);
+			bcm_hw_stats_post(sc);
+			bcm_stats_comp(sc);
+		}
+	}
+
+	sc->stats_state = STATS_STATE_DISABLED;
+
+	if (sc->port.pmf && sc->port.port_stx) {
+		bcm_port_stats_base_init(sc);
+	}
+
+	/* mark the end of statistics initializiation */
+	sc->stats_init = FALSE;
+}
+
+void
+bcm_save_statistics(struct bcm_softc *sc)
+{
+	int i;
+
+	/* save queue statistics */
+	for (i = 0; i < sc->num_queues; i++) {
+		struct bcm_fastpath *fp = &sc->fp[i];
+		struct bcm_eth_q_stats *qstats = &fp->eth_q_stats;
+		struct bcm_eth_q_stats_old *qstats_old = &fp->eth_q_stats_old;
+
+		UPDATE_QSTAT_OLD(total_unicast_bytes_received_hi);
+		UPDATE_QSTAT_OLD(total_unicast_bytes_received_lo);
+		UPDATE_QSTAT_OLD(total_broadcast_bytes_received_hi);
+		UPDATE_QSTAT_OLD(total_broadcast_bytes_received_lo);
+		UPDATE_QSTAT_OLD(total_multicast_bytes_received_hi);
+		UPDATE_QSTAT_OLD(total_multicast_bytes_received_lo);
+		UPDATE_QSTAT_OLD(total_unicast_bytes_transmitted_hi);
+		UPDATE_QSTAT_OLD(total_unicast_bytes_transmitted_lo);
+		UPDATE_QSTAT_OLD(total_broadcast_bytes_transmitted_hi);
+		UPDATE_QSTAT_OLD(total_broadcast_bytes_transmitted_lo);
+		UPDATE_QSTAT_OLD(total_multicast_bytes_transmitted_hi);
+		UPDATE_QSTAT_OLD(total_multicast_bytes_transmitted_lo);
+	}
+
+	/* store port firmware statistics */
+	if (sc->port.pmf) {
+		struct bcm_eth_stats *estats = &sc->eth_stats;
+		struct bcm_fw_port_stats_old *fwstats = &sc->fw_stats_old;
+		struct host_port_stats *pstats = BCM_SP(sc, port_stats);
+
+		fwstats->pfc_frames_rx_hi = pstats->pfc_frames_rx_hi;
+		fwstats->pfc_frames_rx_lo = pstats->pfc_frames_rx_lo;
+		fwstats->pfc_frames_tx_hi = pstats->pfc_frames_tx_hi;
+		fwstats->pfc_frames_tx_lo = pstats->pfc_frames_tx_lo;
+
+		if (IS_MF(sc)) {
+			UPDATE_FW_STAT_OLD(mac_filter_discard);
+			UPDATE_FW_STAT_OLD(mf_tag_discard);
+			UPDATE_FW_STAT_OLD(brb_truncate_discard);
+			UPDATE_FW_STAT_OLD(mac_discard);
+		}
+	}
+}
diff --git a/lib/librte_pmd_bcm/bcm_stats.h b/lib/librte_pmd_bcm/bcm_stats.h
new file mode 100644
index 0000000..da8240f
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_stats.h
@@ -0,0 +1,633 @@ 
+/*-
+ * Copyright (c) 2007-2013 Broadcom Corporation. All rights reserved.
+ *
+ * Eric Davis        <edavis@broadcom.com>
+ * David Christensen <davidch@broadcom.com>
+ * Gary Zambrano     <zambrano@broadcom.com>
+ *
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ * All rights reserved.
+ *
+ * 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 Broadcom Corporation nor the name of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written consent.
+ *
+ * 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 OWNER 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.
+ */
+
+#ifndef BCM_STATS_H
+#define BCM_STATS_H
+
+#include <sys/types.h>
+
+struct nig_stats {
+    uint32_t brb_discard;
+    uint32_t brb_packet;
+    uint32_t brb_truncate;
+    uint32_t flow_ctrl_discard;
+    uint32_t flow_ctrl_octets;
+    uint32_t flow_ctrl_packet;
+    uint32_t mng_discard;
+    uint32_t mng_octet_inp;
+    uint32_t mng_octet_out;
+    uint32_t mng_packet_inp;
+    uint32_t mng_packet_out;
+    uint32_t pbf_octets;
+    uint32_t pbf_packet;
+    uint32_t safc_inp;
+    uint32_t egress_mac_pkt0_lo;
+    uint32_t egress_mac_pkt0_hi;
+    uint32_t egress_mac_pkt1_lo;
+    uint32_t egress_mac_pkt1_hi;
+};
+
+
+enum bcm_stats_event {
+    STATS_EVENT_PMF = 0,
+    STATS_EVENT_LINK_UP,
+    STATS_EVENT_UPDATE,
+    STATS_EVENT_STOP,
+    STATS_EVENT_MAX
+};
+
+enum bcm_stats_state {
+    STATS_STATE_DISABLED = 0,
+    STATS_STATE_ENABLED,
+    STATS_STATE_MAX
+};
+
+struct bcm_eth_stats {
+    uint32_t total_bytes_received_hi;
+    uint32_t total_bytes_received_lo;
+    uint32_t total_bytes_transmitted_hi;
+    uint32_t total_bytes_transmitted_lo;
+    uint32_t total_unicast_packets_received_hi;
+    uint32_t total_unicast_packets_received_lo;
+    uint32_t total_multicast_packets_received_hi;
+    uint32_t total_multicast_packets_received_lo;
+    uint32_t total_broadcast_packets_received_hi;
+    uint32_t total_broadcast_packets_received_lo;
+    uint32_t total_unicast_packets_transmitted_hi;
+    uint32_t total_unicast_packets_transmitted_lo;
+    uint32_t total_multicast_packets_transmitted_hi;
+    uint32_t total_multicast_packets_transmitted_lo;
+    uint32_t total_broadcast_packets_transmitted_hi;
+    uint32_t total_broadcast_packets_transmitted_lo;
+    uint32_t valid_bytes_received_hi;
+    uint32_t valid_bytes_received_lo;
+
+    uint32_t error_bytes_received_hi;
+    uint32_t error_bytes_received_lo;
+    uint32_t etherstatsoverrsizepkts_hi;
+    uint32_t etherstatsoverrsizepkts_lo;
+    uint32_t no_buff_discard_hi;
+    uint32_t no_buff_discard_lo;
+
+    uint32_t rx_stat_ifhcinbadoctets_hi;
+    uint32_t rx_stat_ifhcinbadoctets_lo;
+    uint32_t tx_stat_ifhcoutbadoctets_hi;
+    uint32_t tx_stat_ifhcoutbadoctets_lo;
+    uint32_t rx_stat_dot3statsfcserrors_hi;
+    uint32_t rx_stat_dot3statsfcserrors_lo;
+    uint32_t rx_stat_dot3statsalignmenterrors_hi;
+    uint32_t rx_stat_dot3statsalignmenterrors_lo;
+    uint32_t rx_stat_dot3statscarriersenseerrors_hi;
+    uint32_t rx_stat_dot3statscarriersenseerrors_lo;
+    uint32_t rx_stat_falsecarriererrors_hi;
+    uint32_t rx_stat_falsecarriererrors_lo;
+    uint32_t rx_stat_etherstatsundersizepkts_hi;
+    uint32_t rx_stat_etherstatsundersizepkts_lo;
+    uint32_t rx_stat_dot3statsframestoolong_hi;
+    uint32_t rx_stat_dot3statsframestoolong_lo;
+    uint32_t rx_stat_etherstatsfragments_hi;
+    uint32_t rx_stat_etherstatsfragments_lo;
+    uint32_t rx_stat_etherstatsjabbers_hi;
+    uint32_t rx_stat_etherstatsjabbers_lo;
+    uint32_t rx_stat_maccontrolframesreceived_hi;
+    uint32_t rx_stat_maccontrolframesreceived_lo;
+    uint32_t rx_stat_bmac_xpf_hi;
+    uint32_t rx_stat_bmac_xpf_lo;
+    uint32_t rx_stat_bmac_xcf_hi;
+    uint32_t rx_stat_bmac_xcf_lo;
+    uint32_t rx_stat_xoffstateentered_hi;
+    uint32_t rx_stat_xoffstateentered_lo;
+    uint32_t rx_stat_xonpauseframesreceived_hi;
+    uint32_t rx_stat_xonpauseframesreceived_lo;
+    uint32_t rx_stat_xoffpauseframesreceived_hi;
+    uint32_t rx_stat_xoffpauseframesreceived_lo;
+    uint32_t tx_stat_outxonsent_hi;
+    uint32_t tx_stat_outxonsent_lo;
+    uint32_t tx_stat_outxoffsent_hi;
+    uint32_t tx_stat_outxoffsent_lo;
+    uint32_t tx_stat_flowcontroldone_hi;
+    uint32_t tx_stat_flowcontroldone_lo;
+    uint32_t tx_stat_etherstatscollisions_hi;
+    uint32_t tx_stat_etherstatscollisions_lo;
+    uint32_t tx_stat_dot3statssinglecollisionframes_hi;
+    uint32_t tx_stat_dot3statssinglecollisionframes_lo;
+    uint32_t tx_stat_dot3statsmultiplecollisionframes_hi;
+    uint32_t tx_stat_dot3statsmultiplecollisionframes_lo;
+    uint32_t tx_stat_dot3statsdeferredtransmissions_hi;
+    uint32_t tx_stat_dot3statsdeferredtransmissions_lo;
+    uint32_t tx_stat_dot3statsexcessivecollisions_hi;
+    uint32_t tx_stat_dot3statsexcessivecollisions_lo;
+    uint32_t tx_stat_dot3statslatecollisions_hi;
+    uint32_t tx_stat_dot3statslatecollisions_lo;
+    uint32_t tx_stat_etherstatspkts64octets_hi;
+    uint32_t tx_stat_etherstatspkts64octets_lo;
+    uint32_t tx_stat_etherstatspkts65octetsto127octets_hi;
+    uint32_t tx_stat_etherstatspkts65octetsto127octets_lo;
+    uint32_t tx_stat_etherstatspkts128octetsto255octets_hi;
+    uint32_t tx_stat_etherstatspkts128octetsto255octets_lo;
+    uint32_t tx_stat_etherstatspkts256octetsto511octets_hi;
+    uint32_t tx_stat_etherstatspkts256octetsto511octets_lo;
+    uint32_t tx_stat_etherstatspkts512octetsto1023octets_hi;
+    uint32_t tx_stat_etherstatspkts512octetsto1023octets_lo;
+    uint32_t tx_stat_etherstatspkts1024octetsto1522octets_hi;
+    uint32_t tx_stat_etherstatspkts1024octetsto1522octets_lo;
+    uint32_t tx_stat_etherstatspktsover1522octets_hi;
+    uint32_t tx_stat_etherstatspktsover1522octets_lo;
+    uint32_t tx_stat_bmac_2047_hi;
+    uint32_t tx_stat_bmac_2047_lo;
+    uint32_t tx_stat_bmac_4095_hi;
+    uint32_t tx_stat_bmac_4095_lo;
+    uint32_t tx_stat_bmac_9216_hi;
+    uint32_t tx_stat_bmac_9216_lo;
+    uint32_t tx_stat_bmac_16383_hi;
+    uint32_t tx_stat_bmac_16383_lo;
+    uint32_t tx_stat_dot3statsinternalmactransmiterrors_hi;
+    uint32_t tx_stat_dot3statsinternalmactransmiterrors_lo;
+    uint32_t tx_stat_bmac_ufl_hi;
+    uint32_t tx_stat_bmac_ufl_lo;
+
+    uint32_t pause_frames_received_hi;
+    uint32_t pause_frames_received_lo;
+    uint32_t pause_frames_sent_hi;
+    uint32_t pause_frames_sent_lo;
+
+    uint32_t etherstatspkts1024octetsto1522octets_hi;
+    uint32_t etherstatspkts1024octetsto1522octets_lo;
+    uint32_t etherstatspktsover1522octets_hi;
+    uint32_t etherstatspktsover1522octets_lo;
+
+    uint32_t brb_drop_hi;
+    uint32_t brb_drop_lo;
+    uint32_t brb_truncate_hi;
+    uint32_t brb_truncate_lo;
+
+    uint32_t mac_filter_discard;
+    uint32_t mf_tag_discard;
+    uint32_t brb_truncate_discard;
+    uint32_t mac_discard;
+
+    uint32_t nig_timer_max;
+
+    /* PFC */
+    uint32_t pfc_frames_received_hi;
+    uint32_t pfc_frames_received_lo;
+    uint32_t pfc_frames_sent_hi;
+    uint32_t pfc_frames_sent_lo;
+
+    /* Recovery */
+    uint32_t recoverable_error;
+    uint32_t unrecoverable_error;
+
+    /* src: Clear-on-Read register; Will not survive PMF Migration */
+    uint32_t eee_tx_lpi;
+
+    /* receive path driver statistics */
+    uint32_t rx_calls;
+    uint32_t rx_pkts;
+    uint32_t rx_soft_errors;
+    uint32_t rx_hw_csum_errors;
+    uint32_t rx_ofld_frames_csum_ip;
+    uint32_t rx_ofld_frames_csum_tcp_udp;
+    uint32_t rx_budget_reached;
+
+    /* tx path driver statistics */
+    uint32_t tx_pkts;
+    uint32_t tx_soft_errors;
+    uint32_t tx_ofld_frames_csum_ip;
+    uint32_t tx_ofld_frames_csum_tcp;
+    uint32_t tx_ofld_frames_csum_udp;
+    uint32_t tx_encap_failures;
+    uint32_t tx_hw_queue_full;
+    uint32_t tx_hw_max_queue_depth;
+    uint32_t tx_dma_mapping_failure;
+    uint32_t tx_max_drbr_queue_depth;
+    uint32_t tx_window_violation_std;
+    uint32_t tx_chain_lost_mbuf;
+    uint32_t tx_frames_deferred;
+    uint32_t tx_queue_xoff;
+
+    /* mbuf driver statistics */
+    uint32_t mbuf_defrag_attempts;
+    uint32_t mbuf_defrag_failures;
+    uint32_t mbuf_rx_bd_alloc_failed;
+    uint32_t mbuf_rx_bd_mapping_failed;
+
+    /* track the number of allocated mbufs */
+    uint32_t mbuf_alloc_tx;
+    uint32_t mbuf_alloc_rx;
+};
+
+
+struct bcm_eth_q_stats {
+    uint32_t total_unicast_bytes_received_hi;
+    uint32_t total_unicast_bytes_received_lo;
+    uint32_t total_broadcast_bytes_received_hi;
+    uint32_t total_broadcast_bytes_received_lo;
+    uint32_t total_multicast_bytes_received_hi;
+    uint32_t total_multicast_bytes_received_lo;
+    uint32_t total_bytes_received_hi;
+    uint32_t total_bytes_received_lo;
+    uint32_t total_unicast_bytes_transmitted_hi;
+    uint32_t total_unicast_bytes_transmitted_lo;
+    uint32_t total_broadcast_bytes_transmitted_hi;
+    uint32_t total_broadcast_bytes_transmitted_lo;
+    uint32_t total_multicast_bytes_transmitted_hi;
+    uint32_t total_multicast_bytes_transmitted_lo;
+    uint32_t total_bytes_transmitted_hi;
+    uint32_t total_bytes_transmitted_lo;
+    uint32_t total_unicast_packets_received_hi;
+    uint32_t total_unicast_packets_received_lo;
+    uint32_t total_multicast_packets_received_hi;
+    uint32_t total_multicast_packets_received_lo;
+    uint32_t total_broadcast_packets_received_hi;
+    uint32_t total_broadcast_packets_received_lo;
+    uint32_t total_unicast_packets_transmitted_hi;
+    uint32_t total_unicast_packets_transmitted_lo;
+    uint32_t total_multicast_packets_transmitted_hi;
+    uint32_t total_multicast_packets_transmitted_lo;
+    uint32_t total_broadcast_packets_transmitted_hi;
+    uint32_t total_broadcast_packets_transmitted_lo;
+    uint32_t valid_bytes_received_hi;
+    uint32_t valid_bytes_received_lo;
+
+    uint32_t etherstatsoverrsizepkts_hi;
+    uint32_t etherstatsoverrsizepkts_lo;
+    uint32_t no_buff_discard_hi;
+    uint32_t no_buff_discard_lo;
+
+    uint32_t total_packets_received_checksum_discarded_hi;
+    uint32_t total_packets_received_checksum_discarded_lo;
+    uint32_t total_packets_received_ttl0_discarded_hi;
+    uint32_t total_packets_received_ttl0_discarded_lo;
+    uint32_t total_transmitted_dropped_packets_error_hi;
+    uint32_t total_transmitted_dropped_packets_error_lo;
+
+    /* receive path driver statistics */
+    uint32_t rx_calls;
+    uint32_t rx_pkts;
+    uint32_t rx_soft_errors;
+    uint32_t rx_hw_csum_errors;
+    uint32_t rx_ofld_frames_csum_ip;
+    uint32_t rx_ofld_frames_csum_tcp_udp;
+    uint32_t rx_budget_reached;
+
+    /* tx path driver statistics */
+    uint32_t tx_pkts;
+    uint32_t tx_soft_errors;
+    uint32_t tx_ofld_frames_csum_ip;
+    uint32_t tx_ofld_frames_csum_tcp;
+    uint32_t tx_ofld_frames_csum_udp;
+    uint32_t tx_encap_failures;
+    uint32_t tx_hw_queue_full;
+    uint32_t tx_hw_max_queue_depth;
+    uint32_t tx_dma_mapping_failure;
+    uint32_t tx_max_drbr_queue_depth;
+    uint32_t tx_window_violation_std;
+    uint32_t tx_chain_lost_mbuf;
+    uint32_t tx_frames_deferred;
+    uint32_t tx_queue_xoff;
+
+    /* mbuf driver statistics */
+    uint32_t mbuf_defrag_attempts;
+    uint32_t mbuf_defrag_failures;
+    uint32_t mbuf_rx_bd_alloc_failed;
+    uint32_t mbuf_rx_bd_mapping_failed;
+
+    /* track the number of allocated mbufs */
+    uint32_t mbuf_alloc_tx;
+    uint32_t mbuf_alloc_rx;
+};
+
+struct bcm_eth_stats_old {
+    uint32_t rx_stat_dot3statsframestoolong_hi;
+    uint32_t rx_stat_dot3statsframestoolong_lo;
+};
+
+struct bcm_eth_q_stats_old {
+    /* Fields to perserve over fw reset*/
+    uint32_t total_unicast_bytes_received_hi;
+    uint32_t total_unicast_bytes_received_lo;
+    uint32_t total_broadcast_bytes_received_hi;
+    uint32_t total_broadcast_bytes_received_lo;
+    uint32_t total_multicast_bytes_received_hi;
+    uint32_t total_multicast_bytes_received_lo;
+    uint32_t total_unicast_bytes_transmitted_hi;
+    uint32_t total_unicast_bytes_transmitted_lo;
+    uint32_t total_broadcast_bytes_transmitted_hi;
+    uint32_t total_broadcast_bytes_transmitted_lo;
+    uint32_t total_multicast_bytes_transmitted_hi;
+    uint32_t total_multicast_bytes_transmitted_lo;
+
+    /* Fields to perserve last of */
+    uint32_t total_bytes_received_hi;
+    uint32_t total_bytes_received_lo;
+    uint32_t total_bytes_transmitted_hi;
+    uint32_t total_bytes_transmitted_lo;
+    uint32_t total_unicast_packets_received_hi;
+    uint32_t total_unicast_packets_received_lo;
+    uint32_t total_multicast_packets_received_hi;
+    uint32_t total_multicast_packets_received_lo;
+    uint32_t total_broadcast_packets_received_hi;
+    uint32_t total_broadcast_packets_received_lo;
+    uint32_t total_unicast_packets_transmitted_hi;
+    uint32_t total_unicast_packets_transmitted_lo;
+    uint32_t total_multicast_packets_transmitted_hi;
+    uint32_t total_multicast_packets_transmitted_lo;
+    uint32_t total_broadcast_packets_transmitted_hi;
+    uint32_t total_broadcast_packets_transmitted_lo;
+    uint32_t valid_bytes_received_hi;
+    uint32_t valid_bytes_received_lo;
+
+    /* receive path driver statistics */
+    uint32_t rx_calls_old;
+    uint32_t rx_pkts_old;
+    uint32_t rx_soft_errors_old;
+    uint32_t rx_hw_csum_errors_old;
+    uint32_t rx_ofld_frames_csum_ip_old;
+    uint32_t rx_ofld_frames_csum_tcp_udp_old;
+    uint32_t rx_budget_reached_old;
+
+    /* tx path driver statistics */
+    uint32_t tx_pkts_old;
+    uint32_t tx_soft_errors_old;
+    uint32_t tx_ofld_frames_csum_ip_old;
+    uint32_t tx_ofld_frames_csum_tcp_old;
+    uint32_t tx_ofld_frames_csum_udp_old;
+    uint32_t tx_encap_failures_old;
+    uint32_t tx_hw_queue_full_old;
+    uint32_t tx_hw_max_queue_depth_old;
+    uint32_t tx_dma_mapping_failure_old;
+    uint32_t tx_max_drbr_queue_depth_old;
+    uint32_t tx_window_violation_std_old;
+    uint32_t tx_chain_lost_mbuf_old;
+    uint32_t tx_frames_deferred_old;
+    uint32_t tx_queue_xoff_old;
+
+    /* mbuf driver statistics */
+    uint32_t mbuf_defrag_attempts_old;
+    uint32_t mbuf_defrag_failures_old;
+    uint32_t mbuf_rx_bd_alloc_failed_old;
+    uint32_t mbuf_rx_bd_mapping_failed_old;
+
+    /* track the number of allocated mbufs */
+    int mbuf_alloc_tx_old;
+    int mbuf_alloc_rx_old;
+};
+
+struct bcm_net_stats_old {
+    uint32_t rx_dropped;
+};
+
+struct bcm_fw_port_stats_old {
+    uint32_t pfc_frames_tx_hi;
+    uint32_t pfc_frames_tx_lo;
+    uint32_t pfc_frames_rx_hi;
+    uint32_t pfc_frames_rx_lo;
+
+    uint32_t mac_filter_discard;
+    uint32_t mf_tag_discard;
+    uint32_t brb_truncate_discard;
+    uint32_t mac_discard;
+};
+
+/* sum[hi:lo] += add[hi:lo] */
+#define ADD_64(s_hi, a_hi, s_lo, a_lo)          \
+    do {                                        \
+        s_lo += a_lo;                           \
+        s_hi += a_hi + ((s_lo < a_lo) ? 1 : 0); \
+    } while (0)
+
+#define LE32_0 ((uint32_t) 0)
+#define LE16_0 ((uint16_t) 0)
+
+/* The _force is for cases where high value is 0 */
+#define ADD_64_LE(s_hi, a_hi_le, s_lo, a_lo_le) \
+        ADD_64(s_hi, le32toh(a_hi_le),          \
+               s_lo, le32toh(a_lo_le))
+
+#define ADD_64_LE16(s_hi, a_hi_le, s_lo, a_lo_le) \
+        ADD_64(s_hi, le16toh(a_hi_le),            \
+               s_lo, le16toh(a_lo_le))
+
+/* difference = minuend - subtrahend */
+#define DIFF_64(d_hi, m_hi, s_hi, d_lo, m_lo, s_lo)  \
+    do {                                             \
+        if (m_lo < s_lo) {                           \
+            /* underflow */                          \
+            d_hi = m_hi - s_hi;                      \
+            if (d_hi > 0) {                          \
+                /* we can 'loan' 1 */                \
+                d_hi--;                              \
+                d_lo = m_lo + (UINT_MAX - s_lo) + 1; \
+            } else {                                 \
+                /* m_hi <= s_hi */                   \
+                d_hi = 0;                            \
+                d_lo = 0;                            \
+            }                                        \
+        } else {                                     \
+            /* m_lo >= s_lo */                       \
+            if (m_hi < s_hi) {                       \
+                d_hi = 0;                            \
+                d_lo = 0;                            \
+            } else {                                 \
+                /* m_hi >= s_hi */                   \
+                d_hi = m_hi - s_hi;                  \
+                d_lo = m_lo - s_lo;                  \
+            }                                        \
+        }                                            \
+    } while (0)
+
+#define UPDATE_STAT64(s, t)                                      \
+    do {                                                         \
+        DIFF_64(diff.hi, new->s##_hi, pstats->mac_stx[0].t##_hi, \
+            diff.lo, new->s##_lo, pstats->mac_stx[0].t##_lo);    \
+        pstats->mac_stx[0].t##_hi = new->s##_hi;                 \
+        pstats->mac_stx[0].t##_lo = new->s##_lo;                 \
+        ADD_64(pstats->mac_stx[1].t##_hi, diff.hi,               \
+               pstats->mac_stx[1].t##_lo, diff.lo);              \
+    } while (0)
+
+#define UPDATE_STAT64_NIG(s, t)                    \
+    do {                                           \
+        DIFF_64(diff.hi, new->s##_hi, old->s##_hi, \
+            diff.lo, new->s##_lo, old->s##_lo);    \
+        ADD_64(estats->t##_hi, diff.hi,            \
+               estats->t##_lo, diff.lo);           \
+    } while (0)
+
+/* sum[hi:lo] += add */
+#define ADD_EXTEND_64(s_hi, s_lo, a) \
+    do {                             \
+        s_lo += a;                   \
+        s_hi += (s_lo < a) ? 1 : 0;  \
+    } while (0)
+
+#define ADD_STAT64(diff, t)                                \
+    do {                                                   \
+        ADD_64(pstats->mac_stx[1].t##_hi, new->diff##_hi,  \
+               pstats->mac_stx[1].t##_lo, new->diff##_lo); \
+    } while (0)
+
+#define UPDATE_EXTEND_STAT(s)                    \
+    do {                                         \
+        ADD_EXTEND_64(pstats->mac_stx[1].s##_hi, \
+                  pstats->mac_stx[1].s##_lo,     \
+                  new->s);                       \
+    } while (0)
+
+#define UPDATE_EXTEND_TSTAT_X(s, t, size)                    \
+    do {                                                     \
+        diff = le##size##toh(tclient->s) -                   \
+               le##size##toh(old_tclient->s);                \
+        old_tclient->s = tclient->s;                         \
+        ADD_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff); \
+    } while (0)
+
+#define UPDATE_EXTEND_TSTAT(s, t) UPDATE_EXTEND_TSTAT_X(s, t, 32)
+
+#define UPDATE_EXTEND_E_TSTAT(s, t, size)                    \
+    do {                                                     \
+        UPDATE_EXTEND_TSTAT_X(s, t, size);                   \
+        ADD_EXTEND_64(estats->t##_hi, estats->t##_lo, diff); \
+    } while (0)
+
+#define UPDATE_EXTEND_USTAT(s, t)                             \
+    do {                                                      \
+        diff = le32toh(uclient->s) - le32toh(old_uclient->s); \
+        old_uclient->s = uclient->s;                          \
+        ADD_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff);  \
+    } while (0)
+
+#define UPDATE_EXTEND_E_USTAT(s, t)                          \
+    do {                                                     \
+        UPDATE_EXTEND_USTAT(s, t);                           \
+        ADD_EXTEND_64(estats->t##_hi, estats->t##_lo, diff); \
+    } while (0)
+
+#define UPDATE_EXTEND_XSTAT(s, t)                             \
+    do {                                                      \
+        diff = le32toh(xclient->s) - le32toh(old_xclient->s); \
+        old_xclient->s = xclient->s;                          \
+        ADD_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff);  \
+    } while (0)
+
+#define UPDATE_QSTAT(s, t)                                   \
+    do {                                                     \
+        qstats->t##_hi = qstats_old->t##_hi + le32toh(s.hi); \
+        qstats->t##_lo = qstats_old->t##_lo + le32toh(s.lo); \
+    } while (0)
+
+#define UPDATE_QSTAT_OLD(f)        \
+    do {                           \
+        qstats_old->f = qstats->f; \
+    } while (0)
+
+#define UPDATE_ESTAT_QSTAT_64(s)                        \
+    do {                                                \
+        ADD_64(estats->s##_hi, qstats->s##_hi,          \
+               estats->s##_lo, qstats->s##_lo);         \
+        SUB_64(estats->s##_hi, qstats_old->s##_hi_old,  \
+               estats->s##_lo, qstats_old->s##_lo_old); \
+        qstats_old->s##_hi_old = qstats->s##_hi;        \
+        qstats_old->s##_lo_old = qstats->s##_lo;        \
+    } while (0)
+
+#define UPDATE_ESTAT_QSTAT(s)             \
+    do {                                  \
+        estats->s += qstats->s;           \
+        estats->s -= qstats_old->s##_old; \
+        qstats_old->s##_old = qstats->s;  \
+    } while (0)
+
+#define UPDATE_FSTAT_QSTAT(s)                       \
+    do {                                            \
+        ADD_64(fstats->s##_hi, qstats->s##_hi,      \
+               fstats->s##_lo, qstats->s##_lo);     \
+        SUB_64(fstats->s##_hi, qstats_old->s##_hi,  \
+               fstats->s##_lo, qstats_old->s##_lo); \
+        estats->s##_hi = fstats->s##_hi;            \
+        estats->s##_lo = fstats->s##_lo;            \
+        qstats_old->s##_hi = qstats->s##_hi;        \
+        qstats_old->s##_lo = qstats->s##_lo;        \
+    } while (0)
+
+#define UPDATE_FW_STAT(s)                           \
+    do {                                            \
+        estats->s = le32toh(tport->s) + fwstats->s; \
+    } while (0)
+
+#define UPDATE_FW_STAT_OLD(f)   \
+    do {                        \
+        fwstats->f = estats->f; \
+    } while (0)
+
+#define UPDATE_ESTAT(s, t)                          \
+    do {                                            \
+        SUB_64(estats->s##_hi, estats_old->t##_hi,  \
+               estats->s##_lo, estats_old->t##_lo); \
+        ADD_64(estats->s##_hi, estats->t##_hi,      \
+               estats->s##_lo, estats->t##_lo);     \
+        estats_old->t##_hi = estats->t##_hi;        \
+        estats_old->t##_lo = estats->t##_lo;        \
+    } while (0)
+
+/* minuend -= subtrahend */
+#define SUB_64(m_hi, s_hi, m_lo, s_lo)               \
+    do {                                             \
+        DIFF_64(m_hi, m_hi, s_hi, m_lo, m_lo, s_lo); \
+    } while (0)
+
+/* minuend[hi:lo] -= subtrahend */
+#define SUB_EXTEND_64(m_hi, m_lo, s)    \
+    do {                                \
+        uint32_t s_hi = 0;              \
+        SUB_64(m_hi, s_hi, m_lo, s);    \
+    } while (0)
+
+#define SUB_EXTEND_USTAT(s, t)                                \
+    do {                                                      \
+        diff = le32toh(uclient->s) - le32toh(old_uclient->s); \
+        SUB_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff);  \
+    } while (0)
+
+struct bcm_softc;
+void bcm_stats_init(struct bcm_softc *sc);
+void bcm_stats_handle(struct bcm_softc *sc, enum bcm_stats_event event);
+void bcm_save_statistics(struct bcm_softc *sc);
+void bcm_memset_stats(struct bcm_softc *sc);
+
+#endif /* BCM_STATS_H */
+
diff --git a/lib/librte_pmd_bcm/bcm_vfpf.c b/lib/librte_pmd_bcm/bcm_vfpf.c
new file mode 100644
index 0000000..eef10de
--- /dev/null
+++ b/lib/librte_pmd_bcm/bcm_vfpf.c
@@ -0,0 +1,597 @@ 
+/*
+ * Copyright (c) 2013-2015 Brocade Communications Systems, Inc.
+ *
+ * All rights reserved.
+ */
+
+#include "bcm.h"
+
+/* calculate the crc in the bulletin board */
+static inline uint32_t
+bcm_vf_crc(struct bcm_vf_bulletin *bull)
+{
+	uint32_t crc_sz = sizeof(bull->crc), length = bull->length - crc_sz;
+
+	return ECORE_CRC32_LE(0, (uint8_t *)bull + crc_sz, length);
+}
+
+/* Checks are there mac/channel updates for VF
+ * returns TRUE if something was updated
+*/
+int
+bcm_check_bull(struct bcm_softc *sc)
+{
+	struct bcm_vf_bulletin *bull;
+	uint8_t tries = 0;
+	uint16_t old_version = sc->old_bulletin.version;
+	uint64_t valid_bitmap;
+
+	bull = sc->pf2vf_bulletin;
+	if (old_version == bull->version) {
+		return FALSE;
+	} else {
+		/* Check the crc until we get the correct data */
+		while (tries < BCM_VF_BULLETIN_TRIES) {
+			bull = sc->pf2vf_bulletin;
+			if (bull->crc == bcm_vf_crc(bull))
+				break;
+
+			PMD_DRV_LOG(ERR, "bad crc on bulletin board. contained %x computed %x",
+					bull->crc, bcm_vf_crc(bull));
+			++tries;
+		}
+		if (tries == BCM_VF_BULLETIN_TRIES) {
+			PMD_DRV_LOG(ERR, "pf to vf bulletin board crc was wrong %d consecutive times. Aborting",
+					tries);
+			return FALSE;
+		}
+	}
+
+	valid_bitmap = bull->valid_bitmap;
+
+	/* check the mac address and VLAN and allocate memory if valid */
+	if (valid_bitmap & (1 << MAC_ADDR_VALID) && memcmp(bull->mac, sc->old_bulletin.mac, ETH_ALEN))
+		rte_memcpy(&sc->link_params.mac_addr, bull->mac, ETH_ALEN);
+	if (valid_bitmap & (1 << VLAN_VALID))
+		rte_memcpy(&bull->vlan, &sc->old_bulletin.vlan, VLAN_HLEN);
+
+	sc->old_bulletin = *bull;
+
+	return TRUE;
+}
+
+/* add tlv to a buffer */
+#define BCM_TLV_APPEND(_tlvs, _offset, _type, _length) \
+	((struct vf_first_tlv *)((uint64_t)_tlvs + _offset))->type   = _type; \
+	((struct vf_first_tlv *)((uint64_t)_tlvs + _offset))->length = _length
+
+/* Initiliaze header of the first tlv and clear mailbox*/
+static void
+bcm_init_first_tlv(struct bcm_softc *sc, struct vf_first_tlv *tlv,
+	uint16_t type, uint16_t len)
+{
+	struct bcm_vf_mbx_msg *mbox = sc->vf2pf_mbox;
+	PMD_DRV_LOG(DEBUG, "Preparing %d tlv for sending", type);
+
+	memset(mbox, 0, sizeof(struct bcm_vf_mbx_msg));
+
+	BCM_TLV_APPEND(tlv, 0, type, len);
+
+	/* Initialize header of the first tlv */
+	tlv->reply_offset = sizeof(mbox->query);
+}
+
+#define BCM_VF_CMD_ADDR_LO PXP_VF_ADDR_CSDM_GLOBAL_START
+#define BCM_VF_CMD_ADDR_HI BCM_VF_CMD_ADDR_LO + 4
+#define BCM_VF_CMD_TRIGGER BCM_VF_CMD_ADDR_HI + 4
+#define BCM_VF_CHANNEL_DELAY 100
+#define BCM_VF_CHANNEL_TRIES 100
+
+static int
+bcm_do_req4pf(struct bcm_softc *sc, phys_addr_t phys_addr)
+{
+	uint8_t *status = &sc->vf2pf_mbox->resp.common_reply.status;
+	uint8_t i;
+
+	if (!*status) {
+		bcm_check_bull(sc);
+		if (sc->old_bulletin.valid_bitmap & (1 << CHANNEL_DOWN)) {
+			PMD_DRV_LOG(ERR, "channel is down. Aborting message sending");
+			*status = BCM_VF_STATUS_SUCCESS;
+			return 0;
+		}
+
+		REG_WR(sc, BCM_VF_CMD_ADDR_LO, U64_LO(phys_addr));
+		REG_WR(sc, BCM_VF_CMD_ADDR_HI, U64_HI(phys_addr));
+
+		/* memory barrier to ensure that FW can read phys_addr */
+		wmb();
+
+		REG_WR8(sc, BCM_VF_CMD_TRIGGER, 1);
+
+		/* Do several attempts until PF completes
+		 * "." is used to show progress
+		 */
+		for (i = 0; i < BCM_VF_CHANNEL_TRIES; i++) {
+			DELAY_MS(BCM_VF_CHANNEL_DELAY);
+			if (*status)
+				break;
+		}
+
+		if (i == BCM_VF_CHANNEL_TRIES) {
+			PMD_DRV_LOG(ERR, "Response from PF timed out");
+			return -EAGAIN;
+		}
+
+		if (BCM_VF_STATUS_SUCCESS != *status) {
+			PMD_DRV_LOG(ERR, "Bad reply from PF : %u",
+					*status);
+			return -EINVAL;
+		}
+	} else {
+		PMD_DRV_LOG(ERR, "status should be zero before message"
+				"to pf was sent");
+		return -EINVAL;
+	}
+
+	PMD_DRV_LOG(DEBUG, "Response from PF was received");
+	return 0;
+}
+
+static inline uint16_t bcm_check_me_flags(uint32_t val)
+{
+	if (((val) & ME_REG_VF_VALID) && (!((val) & ME_REG_VF_ERR)))
+		return ME_REG_VF_VALID;
+	else
+		return 0;
+}
+
+#define BCM_ME_ANSWER_DELAY 100
+#define BCM_ME_ANSWER_TRIES 10
+
+static inline int bcm_read_vf_id(struct bcm_softc *sc)
+{
+	uint32_t val;
+	uint8_t i = 0;
+
+	while (i <= BCM_ME_ANSWER_TRIES) {
+		val = BCM_DB_READ(DOORBELL_ADDR(sc, 0));
+		if (bcm_check_me_flags(val))
+			return VF_ID(val);
+
+		DELAY_MS(BCM_ME_ANSWER_DELAY);
+		i++;
+	}
+
+	return -EINVAL;
+}
+
+#define BCM_VF_OBTAIN_MAX_TRIES 3
+#define BCM_VF_OBTAIN_MAC_FILTERS 1
+#define BCM_VF_OBTAIN_MC_FILTERS 10
+
+struct bcm_obtain_status {
+	int success;
+	int err_code;
+};
+
+static
+struct bcm_obtain_status bcm_loop_obtain_resources(struct bcm_softc *sc)
+{
+	int tries = 0;
+	struct vf_acquire_resp_tlv *resp = &sc->vf2pf_mbox->resp.acquire_resp,
+								 *sc_resp = &sc->acquire_resp;
+	struct vf_resource_query    *res_query;
+	struct vf_resc            *resc;
+	struct bcm_obtain_status     status;
+	int res_obtained = false;
+
+	do {
+		PMD_DRV_LOG(DEBUG, "trying to get resources");
+
+		if ( bcm_do_req4pf(sc, sc->vf2pf_mbox_mapping.paddr) ) {
+			/* timeout */
+			status.success = 0;
+			status.err_code = 0;
+			return status;
+		}
+
+		memcpy(sc_resp, resp, sizeof(sc->acquire_resp));
+
+		tries++;
+
+		/* check PF to request acceptance */
+		if (sc_resp->status == BCM_VF_STATUS_SUCCESS) {
+			PMD_DRV_LOG(DEBUG, "resources obtained successfully");
+			res_obtained = true;
+		} else if (sc_resp->status == BCM_VF_STATUS_NO_RESOURCES &&
+			tries < BCM_VF_OBTAIN_MAX_TRIES) {
+			PMD_DRV_LOG(DEBUG,
+			   "PF cannot allocate requested amount of resources");
+
+			res_query = &sc->vf2pf_mbox->query[0].acquire.res_query;
+			resc     = &sc_resp->resc;
+
+			/* PF refused our request. Try to decrease request params */
+			res_query->num_txqs         = min(res_query->num_txqs, resc->num_txqs);
+			res_query->num_rxqs         = min(res_query->num_rxqs, resc->num_rxqs);
+			res_query->num_sbs          = min(res_query->num_sbs, resc->num_sbs);
+			res_query->num_mac_filters  = min(res_query->num_mac_filters, resc->num_mac_filters);
+			res_query->num_vlan_filters = min(res_query->num_vlan_filters, resc->num_vlan_filters);
+			res_query->num_mc_filters   = min(res_query->num_mc_filters, resc->num_mc_filters);
+
+			memset(&sc->vf2pf_mbox->resp, 0, sizeof(union resp_tlvs));
+		} else {
+			PMD_DRV_LOG(ERR, "Resources cannot be obtained. Status of handling: %d. Aborting",
+					sc_resp->status);
+			status.success = 0;
+			status.err_code = -EAGAIN;
+			return status;
+		}
+	} while (!res_obtained);
+
+	status.success = 1;
+	return status;
+}
+
+int bcm_vf_get_resources(struct bcm_softc *sc, uint8_t tx_count, uint8_t rx_count)
+{
+	struct vf_acquire_tlv *acq = &sc->vf2pf_mbox->query[0].acquire;
+	int vf_id;
+	struct bcm_obtain_status obtain_status;
+
+	bcm_vf_close(sc);
+	bcm_init_first_tlv(sc, &acq->first_tlv, BCM_VF_TLV_ACQUIRE, sizeof(*acq));
+
+	vf_id = bcm_read_vf_id(sc);
+	if (vf_id < 0)
+		return -EAGAIN;
+
+	acq->vf_id = vf_id;
+
+	acq->res_query.num_rxqs = rx_count;
+	acq->res_query.num_txqs = tx_count;
+	acq->res_query.num_sbs = sc->igu_sb_cnt;
+	acq->res_query.num_mac_filters = BCM_VF_OBTAIN_MAC_FILTERS;
+	acq->res_query.num_mc_filters = BCM_VF_OBTAIN_MC_FILTERS;
+
+	acq->bulletin_addr = sc->pf2vf_bulletin_mapping.paddr;
+
+	BCM_TLV_APPEND(acq, acq->first_tlv.length, BCM_VF_TLV_LIST_END,
+			sizeof(struct channel_list_end_tlv));
+
+	/* requesting the resources in loop */
+	obtain_status = bcm_loop_obtain_resources(sc);
+	if (!obtain_status.success)
+		return obtain_status.err_code;
+
+	struct vf_acquire_resp_tlv sc_resp = sc->acquire_resp;
+
+	sc->devinfo.chip_id        |= (sc_resp.chip_num & 0xFFFF);
+	sc->devinfo.int_block       = INT_BLOCK_IGU;
+	sc->devinfo.chip_port_mode  = CHIP_2_PORT_MODE;
+	sc->devinfo.mf_info.mf_ov   = 0;
+	sc->devinfo.mf_info.mf_mode = 0;
+	sc->devinfo.flash_size      = 0;
+
+	sc->igu_sb_cnt  = sc_resp.resc.num_sbs;
+	sc->igu_base_sb = sc_resp.resc.hw_sbs[0] & 0xFF;
+	sc->igu_dsb_id  = -1;
+
+	sc->link_params.chip_id = sc->devinfo.chip_id;
+	sc->doorbell_size = sc_resp.db_size;
+	sc->flags |= BCM_NO_WOL_FLAG | BCM_NO_ISCSI_OOO_FLAG | BCM_NO_ISCSI_FLAG | BCM_NO_FCOE_FLAG;
+
+	PMD_DRV_LOG(DEBUG, "status block count = %d, base status block = %x",
+		sc->igu_sb_cnt, sc->igu_base_sb);
+	strncpy(sc->fw_ver, sc_resp.fw_ver, sizeof(sc->fw_ver));
+
+	if (is_valid_ether_addr(sc_resp.resc.current_mac_addr))
+		(void)rte_memcpy(sc->link_params.mac_addr,
+		       sc_resp.resc.current_mac_addr,
+		       ETH_ALEN);
+
+	return 0;
+}
+
+/* Ask PF to release VF's resources */
+void
+bcm_vf_close(struct bcm_softc *sc)
+{
+	struct vf_release_tlv *query;
+	int vf_id = bcm_read_vf_id(sc);
+	int ret;
+
+	if (vf_id >= 0) {
+		query = &sc->vf2pf_mbox->query[0].release;
+		bcm_init_first_tlv(sc, &query->first_tlv, BCM_VF_TLV_RELEASE,
+				sizeof(*query));
+
+		query->vf_id = vf_id;
+		BCM_TLV_APPEND(query, query->first_tlv.length, BCM_VF_TLV_LIST_END,
+				sizeof(struct channel_list_end_tlv));
+
+		ret = bcm_do_req4pf(sc, sc->vf2pf_mbox_mapping.paddr);
+
+		if (ret) {
+			PMD_DRV_LOG(ERR, "Failed to release VF");
+		}
+	}
+}
+
+/* Let PF know the VF status blocks phys_addrs */
+int
+bcm_vf_init(struct bcm_softc *sc)
+{
+	struct vf_init_tlv *query;
+	int i, ret;
+
+	query = &sc->vf2pf_mbox->query[0].init;
+	bcm_init_first_tlv(sc, &query->first_tlv, BCM_VF_TLV_INIT,
+			sizeof(*query));
+
+	FOR_EACH_QUEUE(sc, i) {
+		query->sb_addr[i] = (unsigned long)(sc->fp[i].sb_dma.paddr);
+	}
+
+	query->stats_step = sizeof(struct per_queue_stats);
+	query->stats_addr = sc->fw_stats_data_mapping +
+		offsetof(struct bcm_fw_stats_data, queue_stats);
+
+	BCM_TLV_APPEND(query, query->first_tlv.length, BCM_VF_TLV_LIST_END,
+			sizeof(struct channel_list_end_tlv));
+
+	ret = bcm_do_req4pf(sc, sc->vf2pf_mbox_mapping.paddr);
+
+	if (ret) {
+		PMD_DRV_LOG(ERR, "Failed to init VF");
+		return ret;
+	}
+
+	PMD_DRV_LOG(DEBUG, "VF was initialized");
+	return 0;
+}
+
+void
+bcm_vf_unload(struct bcm_softc *sc)
+{
+	struct vf_close_tlv *query;
+	struct vf_q_op_tlv *query_op;
+	int i, vf_id, ret;
+
+	vf_id = bcm_read_vf_id(sc);
+	if (vf_id > 0) {
+		FOR_EACH_QUEUE(sc, i) {
+			query_op = &sc->vf2pf_mbox->query[0].q_op;
+			bcm_init_first_tlv(sc, &query_op->first_tlv,
+					BCM_VF_TLV_TEARDOWN_Q,
+					sizeof(*query_op));
+
+			query_op->vf_qid = i;
+
+			BCM_TLV_APPEND(query_op, query_op->first_tlv.length,
+					BCM_VF_TLV_LIST_END,
+					sizeof(struct channel_list_end_tlv));
+
+			ret = bcm_do_req4pf(sc, sc->vf2pf_mbox_mapping.paddr);
+			if (ret)
+				PMD_DRV_LOG(ERR,
+						"Bad reply for vf_q %d teardown", i);
+		}
+
+		bcm_vf_set_mac(sc, false);
+
+		query = &sc->vf2pf_mbox->query[0].close;
+		bcm_init_first_tlv(sc, &query->first_tlv, BCM_VF_TLV_CLOSE,
+				sizeof(*query));
+
+		query->vf_id = vf_id;
+
+		BCM_TLV_APPEND(query, query->first_tlv.length,
+				BCM_VF_TLV_LIST_END,
+				sizeof(struct channel_list_end_tlv));
+
+		ret = bcm_do_req4pf(sc, sc->vf2pf_mbox_mapping.paddr);
+
+		if (ret)
+			PMD_DRV_LOG(ERR,
+				"Bad reply from PF for close message");
+	}
+}
+
+static inline uint16_t
+bcm_vf_q_flags(uint8_t leading)
+{
+	uint16_t flags = leading ? BCM_VF_Q_FLAG_LEADING_RSS : 0;
+
+	flags |= BCM_VF_Q_FLAG_CACHE_ALIGN;
+	flags |= BCM_VF_Q_FLAG_STATS;
+	flags |= BCM_VF_Q_FLAG_VLAN;
+
+	return flags;
+}
+
+static void
+bcm_vf_rx_q_prep(struct bcm_softc *sc, struct bcm_fastpath *fp,
+		struct vf_rxq_params *rxq_init, uint16_t flags)
+{
+	struct bcm_rx_queue *rxq;
+
+	rxq = sc->rx_queues[fp->index];
+	if (!rxq) {
+		PMD_DRV_LOG(ERR, "RX queue %d is NULL", fp->index);
+		return;
+	}
+
+	rxq_init->rcq_addr = rxq->cq_ring_phys_addr;
+	rxq_init->rcq_np_addr = rxq->cq_ring_phys_addr + BCM_PAGE_SIZE;
+	rxq_init->rxq_addr = rxq->rx_ring_phys_addr;
+	rxq_init->vf_sb_i