bnx2x: Update vlan_features
[safe/jmp/linux-2.6] / drivers / net / bnx2x_main.c
index 9c40b02..b404a9b 100644 (file)
 #include <linux/zlib.h>
 #include <linux/io.h>
 
-#include "bnx2x_reg.h"
-#include "bnx2x_fw_defs.h"
-#include "bnx2x_hsi.h"
-#include "bnx2x_link.h"
+
 #include "bnx2x.h"
 #include "bnx2x_init.h"
+#include "bnx2x_init_ops.h"
+#include "bnx2x_dump.h"
 
-#define DRV_MODULE_VERSION     "1.45.26"
-#define DRV_MODULE_RELDATE     "2009/01/26"
+#define DRV_MODULE_VERSION     "1.48.105-1"
+#define DRV_MODULE_RELDATE     "2009/04/22"
 #define BNX2X_BC_VER           0x040200
 
+#include <linux/firmware.h>
+#include "bnx2x_fw_file_hdr.h"
+/* FW files */
+#define FW_FILE_PREFIX_E1              "bnx2x-e1-"
+#define FW_FILE_PREFIX_E1H             "bnx2x-e1h-"
+
 /* Time in jiffies before concluding the transmitter is hung */
 #define TX_TIMEOUT             (5*HZ)
 
@@ -75,23 +80,29 @@ MODULE_VERSION(DRV_MODULE_VERSION);
 
 static int multi_mode = 1;
 module_param(multi_mode, int, 0);
+MODULE_PARM_DESC(multi_mode, " Use per-CPU queues");
 
 static int disable_tpa;
-static int poll;
-static int debug;
-static int load_count[3]; /* 0-common, 1-port0, 2-port1 */
-
 module_param(disable_tpa, int, 0);
+MODULE_PARM_DESC(disable_tpa, " Disable the TPA (LRO) feature");
 
 static int int_mode;
 module_param(int_mode, int, 0);
 MODULE_PARM_DESC(int_mode, " Force interrupt mode (1 INT#x; 2 MSI)");
 
+static int poll;
 module_param(poll, int, 0);
+MODULE_PARM_DESC(poll, " Use polling (for debug)");
+
+static int mrrs = -1;
+module_param(mrrs, int, 0);
+MODULE_PARM_DESC(mrrs, " Force Max Read Req Size (0..3) (for debug)");
+
+static int debug;
 module_param(debug, int, 0);
-MODULE_PARM_DESC(disable_tpa, "disable the TPA (LRO) feature");
-MODULE_PARM_DESC(poll, "use polling (for debug)");
-MODULE_PARM_DESC(debug, "default debug msglevel");
+MODULE_PARM_DESC(debug, " Default debug msglevel");
+
+static int load_count[3]; /* 0-common, 1-port0, 2-port1 */
 
 static struct workqueue_struct *bnx2x_wq;
 
@@ -213,7 +224,7 @@ void bnx2x_write_dmae(struct bnx2x *bp, dma_addr_t dma_addr, u32 dst_addr,
        dmae->comp_addr_hi = U64_HI(bnx2x_sp_mapping(bp, wb_comp));
        dmae->comp_val = DMAE_COMP_VAL;
 
-       DP(BNX2X_MSG_OFF, "dmae: opcode 0x%08x\n"
+       DP(BNX2X_MSG_OFF, "DMAE: opcode 0x%08x\n"
           DP_LEVEL "src_addr  [%x:%08x]  len [%d *4]  "
                    "dst_addr [%x:%08x (%08x)]\n"
           DP_LEVEL "comp_addr [%x:%08x]  comp_val 0x%08x\n",
@@ -234,7 +245,7 @@ void bnx2x_write_dmae(struct bnx2x *bp, dma_addr_t dma_addr, u32 dst_addr,
                DP(BNX2X_MSG_OFF, "wb_comp 0x%08x\n", *wb_comp);
 
                if (!cnt) {
-                       BNX2X_ERR("dmae timeout!\n");
+                       BNX2X_ERR("DMAE timeout!\n");
                        break;
                }
                cnt--;
@@ -289,7 +300,7 @@ void bnx2x_read_dmae(struct bnx2x *bp, u32 src_addr, u32 len32)
        dmae->comp_addr_hi = U64_HI(bnx2x_sp_mapping(bp, wb_comp));
        dmae->comp_val = DMAE_COMP_VAL;
 
-       DP(BNX2X_MSG_OFF, "dmae: opcode 0x%08x\n"
+       DP(BNX2X_MSG_OFF, "DMAE: opcode 0x%08x\n"
           DP_LEVEL "src_addr  [%x:%08x]  len [%d *4]  "
                    "dst_addr [%x:%08x (%08x)]\n"
           DP_LEVEL "comp_addr [%x:%08x]  comp_val 0x%08x\n",
@@ -306,7 +317,7 @@ void bnx2x_read_dmae(struct bnx2x *bp, u32 src_addr, u32 len32)
        while (*wb_comp != DMAE_COMP_VAL) {
 
                if (!cnt) {
-                       BNX2X_ERR("dmae timeout!\n");
+                       BNX2X_ERR("DMAE timeout!\n");
                        break;
                }
                cnt--;
@@ -468,7 +479,7 @@ static int bnx2x_mc_assert(struct bnx2x *bp)
 static void bnx2x_fw_dump(struct bnx2x *bp)
 {
        u32 mark, offset;
-       u32 data[9];
+       __be32 data[9];
        int word;
 
        mark = REG_RD(bp, MCP_REG_MCPR_SCRATCH + 0xf104);
@@ -502,82 +513,103 @@ static void bnx2x_panic_dump(struct bnx2x *bp)
 
        BNX2X_ERR("begin crash dump -----------------\n");
 
-       for_each_queue(bp, i) {
+       /* Indices */
+       /* Common */
+       BNX2X_ERR("def_c_idx(%u)  def_u_idx(%u)  def_x_idx(%u)"
+                 "  def_t_idx(%u)  def_att_idx(%u)  attn_state(%u)"
+                 "  spq_prod_idx(%u)\n",
+                 bp->def_c_idx, bp->def_u_idx, bp->def_x_idx, bp->def_t_idx,
+                 bp->def_att_idx, bp->attn_state, bp->spq_prod_idx);
+
+       /* Rx */
+       for_each_rx_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
-               struct eth_tx_db_data *hw_prods = fp->hw_tx_prods;
 
-               BNX2X_ERR("queue[%d]: tx_pkt_prod(%x)  tx_pkt_cons(%x)"
-                         "  tx_bd_prod(%x)  tx_bd_cons(%x)  *tx_cons_sb(%x)\n",
-                         i, fp->tx_pkt_prod, fp->tx_pkt_cons, fp->tx_bd_prod,
-                         fp->tx_bd_cons, le16_to_cpu(*fp->tx_cons_sb));
-               BNX2X_ERR("          rx_bd_prod(%x)  rx_bd_cons(%x)"
+               BNX2X_ERR("fp%d: rx_bd_prod(%x)  rx_bd_cons(%x)"
                          "  *rx_bd_cons_sb(%x)  rx_comp_prod(%x)"
                          "  rx_comp_cons(%x)  *rx_cons_sb(%x)\n",
-                         fp->rx_bd_prod, fp->rx_bd_cons,
+                         i, fp->rx_bd_prod, fp->rx_bd_cons,
                          le16_to_cpu(*fp->rx_bd_cons_sb), fp->rx_comp_prod,
                          fp->rx_comp_cons, le16_to_cpu(*fp->rx_cons_sb));
-               BNX2X_ERR("          rx_sge_prod(%x)  last_max_sge(%x)"
-                         "  fp_c_idx(%x)  *sb_c_idx(%x)  fp_u_idx(%x)"
-                         "  *sb_u_idx(%x)  bd data(%x,%x)\n",
-                         fp->rx_sge_prod, fp->last_max_sge, fp->fp_c_idx,
-                         fp->status_blk->c_status_block.status_block_index,
-                         fp->fp_u_idx,
-                         fp->status_blk->u_status_block.status_block_index,
-                         hw_prods->packets_prod, hw_prods->bds_prod);
-
-               start = TX_BD(le16_to_cpu(*fp->tx_cons_sb) - 10);
-               end = TX_BD(le16_to_cpu(*fp->tx_cons_sb) + 245);
-               for (j = start; j < end; j++) {
-                       struct sw_tx_bd *sw_bd = &fp->tx_buf_ring[j];
+               BNX2X_ERR("      rx_sge_prod(%x)  last_max_sge(%x)"
+                         "  fp_u_idx(%x) *sb_u_idx(%x)\n",
+                         fp->rx_sge_prod, fp->last_max_sge,
+                         le16_to_cpu(fp->fp_u_idx),
+                         fp->status_blk->u_status_block.status_block_index);
+       }
 
-                       BNX2X_ERR("packet[%x]=[%p,%x]\n", j,
-                                 sw_bd->skb, sw_bd->first_bd);
-               }
+       /* Tx */
+       for_each_tx_queue(bp, i) {
+               struct bnx2x_fastpath *fp = &bp->fp[i];
+               struct eth_tx_db_data *hw_prods = fp->hw_tx_prods;
 
-               start = TX_BD(fp->tx_bd_cons - 10);
-               end = TX_BD(fp->tx_bd_cons + 254);
-               for (j = start; j < end; j++) {
-                       u32 *tx_bd = (u32 *)&fp->tx_desc_ring[j];
+               BNX2X_ERR("fp%d: tx_pkt_prod(%x)  tx_pkt_cons(%x)"
+                         "  tx_bd_prod(%x)  tx_bd_cons(%x)  *tx_cons_sb(%x)\n",
+                         i, fp->tx_pkt_prod, fp->tx_pkt_cons, fp->tx_bd_prod,
+                         fp->tx_bd_cons, le16_to_cpu(*fp->tx_cons_sb));
+               BNX2X_ERR("      fp_c_idx(%x)  *sb_c_idx(%x)"
+                         "  bd data(%x,%x)\n", le16_to_cpu(fp->fp_c_idx),
+                         fp->status_blk->c_status_block.status_block_index,
+                         hw_prods->packets_prod, hw_prods->bds_prod);
+       }
 
-                       BNX2X_ERR("tx_bd[%x]=[%x:%x:%x:%x]\n",
-                                 j, tx_bd[0], tx_bd[1], tx_bd[2], tx_bd[3]);
-               }
+       /* Rings */
+       /* Rx */
+       for_each_rx_queue(bp, i) {
+               struct bnx2x_fastpath *fp = &bp->fp[i];
 
                start = RX_BD(le16_to_cpu(*fp->rx_cons_sb) - 10);
                end = RX_BD(le16_to_cpu(*fp->rx_cons_sb) + 503);
-               for (j = start; j < end; j++) {
+               for (j = start; j != end; j = RX_BD(j + 1)) {
                        u32 *rx_bd = (u32 *)&fp->rx_desc_ring[j];
                        struct sw_rx_bd *sw_bd = &fp->rx_buf_ring[j];
 
-                       BNX2X_ERR("rx_bd[%x]=[%x:%x]  sw_bd=[%p]\n",
-                                 j, rx_bd[1], rx_bd[0], sw_bd->skb);
+                       BNX2X_ERR("fp%d: rx_bd[%x]=[%x:%x]  sw_bd=[%p]\n",
+                                 i, j, rx_bd[1], rx_bd[0], sw_bd->skb);
                }
 
                start = RX_SGE(fp->rx_sge_prod);
                end = RX_SGE(fp->last_max_sge);
-               for (j = start; j < end; j++) {
+               for (j = start; j != end; j = RX_SGE(j + 1)) {
                        u32 *rx_sge = (u32 *)&fp->rx_sge_ring[j];
                        struct sw_rx_page *sw_page = &fp->rx_page_ring[j];
 
-                       BNX2X_ERR("rx_sge[%x]=[%x:%x]  sw_page=[%p]\n",
-                                 j, rx_sge[1], rx_sge[0], sw_page->page);
+                       BNX2X_ERR("fp%d: rx_sge[%x]=[%x:%x]  sw_page=[%p]\n",
+                                 i, j, rx_sge[1], rx_sge[0], sw_page->page);
                }
 
                start = RCQ_BD(fp->rx_comp_cons - 10);
                end = RCQ_BD(fp->rx_comp_cons + 503);
-               for (j = start; j < end; j++) {
+               for (j = start; j != end; j = RCQ_BD(j + 1)) {
                        u32 *cqe = (u32 *)&fp->rx_comp_ring[j];
 
-                       BNX2X_ERR("cqe[%x]=[%x:%x:%x:%x]\n",
-                                 j, cqe[0], cqe[1], cqe[2], cqe[3]);
+                       BNX2X_ERR("fp%d: cqe[%x]=[%x:%x:%x:%x]\n",
+                                 i, j, cqe[0], cqe[1], cqe[2], cqe[3]);
                }
        }
 
-       BNX2X_ERR("def_c_idx(%u)  def_u_idx(%u)  def_x_idx(%u)"
-                 "  def_t_idx(%u)  def_att_idx(%u)  attn_state(%u)"
-                 "  spq_prod_idx(%u)\n",
-                 bp->def_c_idx, bp->def_u_idx, bp->def_x_idx, bp->def_t_idx,
-                 bp->def_att_idx, bp->attn_state, bp->spq_prod_idx);
+       /* Tx */
+       for_each_tx_queue(bp, i) {
+               struct bnx2x_fastpath *fp = &bp->fp[i];
+
+               start = TX_BD(le16_to_cpu(*fp->tx_cons_sb) - 10);
+               end = TX_BD(le16_to_cpu(*fp->tx_cons_sb) + 245);
+               for (j = start; j != end; j = TX_BD(j + 1)) {
+                       struct sw_tx_bd *sw_bd = &fp->tx_buf_ring[j];
+
+                       BNX2X_ERR("fp%d: packet[%x]=[%p,%x]\n",
+                                 i, j, sw_bd->skb, sw_bd->first_bd);
+               }
+
+               start = TX_BD(fp->tx_bd_cons - 10);
+               end = TX_BD(fp->tx_bd_cons + 254);
+               for (j = start; j != end; j = TX_BD(j + 1)) {
+                       u32 *tx_bd = (u32 *)&fp->tx_desc_ring[j];
+
+                       BNX2X_ERR("fp%d: tx_bd[%x]=[%x:%x:%x:%x]\n",
+                                 i, j, tx_bd[0], tx_bd[1], tx_bd[2], tx_bd[3]);
+               }
+       }
 
        bnx2x_fw_dump(bp);
        bnx2x_mc_assert(bp);
@@ -620,20 +652,28 @@ static void bnx2x_int_enable(struct bnx2x *bp)
           val, port, addr, (msix ? "MSI-X" : (msi ? "MSI" : "INTx")));
 
        REG_WR(bp, addr, val);
+       /*
+        * Ensure that HC_CONFIG is written before leading/trailing edge config
+        */
+       mmiowb();
+       barrier();
 
        if (CHIP_IS_E1H(bp)) {
                /* init leading/trailing edge */
                if (IS_E1HMF(bp)) {
                        val = (0xee0f | (1 << (BP_E1HVN(bp) + 4)));
                        if (bp->port.pmf)
-                               /* enable nig attention */
-                               val |= 0x0100;
+                               /* enable nig and gpio3 attention */
+                               val |= 0x1100;
                } else
                        val = 0xffff;
 
                REG_WR(bp, HC_REG_TRAILING_EDGE_0 + port*8, val);
                REG_WR(bp, HC_REG_LEADING_EDGE_0 + port*8, val);
        }
+
+       /* Make sure that interrupts are indeed enabled from here on */
+       mmiowb();
 }
 
 static void bnx2x_int_disable(struct bnx2x *bp)
@@ -656,6 +696,7 @@ static void bnx2x_int_disable(struct bnx2x *bp)
        REG_WR(bp, addr, val);
        if (REG_RD(bp, addr) != val)
                BNX2X_ERR("BUG! proper val not read from IGU!\n");
+
 }
 
 static void bnx2x_int_disable_sync(struct bnx2x *bp, int disable_hw)
@@ -665,6 +706,8 @@ static void bnx2x_int_disable_sync(struct bnx2x *bp, int disable_hw)
 
        /* disable interrupt handling */
        atomic_inc(&bp->intr_sem);
+       smp_wmb(); /* Ensure that bp->intr_sem update is SMP-safe */
+
        if (disable_hw)
                /* prevent the HW from sending interrupts */
                bnx2x_int_disable(bp);
@@ -706,6 +749,10 @@ static inline void bnx2x_ack_sb(struct bnx2x *bp, u8 sb_id,
        DP(BNX2X_MSG_OFF, "write 0x%08x to HC addr 0x%x\n",
           (*(u32 *)&igu_ack), hc_addr);
        REG_WR(bp, hc_addr, (*(u32 *)&igu_ack));
+
+       /* Make sure that ACK is written */
+       mmiowb();
+       barrier();
 }
 
 static inline u16 bnx2x_update_fpsb_idx(struct bnx2x_fastpath *fp)
@@ -757,7 +804,6 @@ static inline int bnx2x_has_tx_work_unload(struct bnx2x_fastpath *fp)
        /* Tell compiler that consumer and producer can change */
        barrier();
        return (fp->tx_pkt_prod != fp->tx_pkt_cons);
-
 }
 
 /* free skb in the packet ring at pos idx
@@ -851,7 +897,7 @@ static inline u16 bnx2x_tx_avail(struct bnx2x_fastpath *fp)
        return (s16)(fp->bp->tx_ring_size) - used;
 }
 
-static void bnx2x_tx_int(struct bnx2x_fastpath *fp, int work)
+static void bnx2x_tx_int(struct bnx2x_fastpath *fp)
 {
        struct bnx2x *bp = fp->bp;
        struct netdev_queue *txq;
@@ -885,26 +931,24 @@ static void bnx2x_tx_int(struct bnx2x_fastpath *fp, int work)
                bd_cons = bnx2x_free_tx_pkt(bp, fp, pkt_cons);
                sw_cons++;
                done++;
-
-               if (done == work)
-                       break;
        }
 
        fp->tx_pkt_cons = sw_cons;
        fp->tx_bd_cons = bd_cons;
 
-       /* Need to make the tx_bd_cons update visible to start_xmit()
-        * before checking for netif_tx_queue_stopped().  Without the
-        * memory barrier, there is a small possibility that start_xmit()
-        * will miss it and cause the queue to be stopped forever.
-        */
-       smp_mb();
-
        /* TBD need a thresh? */
        if (unlikely(netif_tx_queue_stopped(txq))) {
 
                __netif_tx_lock(txq, smp_processor_id());
 
+               /* Need to make the tx_bd_cons update visible to start_xmit()
+                * before checking for netif_tx_queue_stopped().  Without the
+                * memory barrier, there is a small possibility that
+                * start_xmit() will miss it and cause the queue to be stopped
+                * forever.
+                */
+               smp_mb();
+
                if ((netif_tx_queue_stopped(txq)) &&
                    (bp->state == BNX2X_STATE_OPEN) &&
                    (bnx2x_tx_avail(fp) >= MAX_SKB_FRAGS + 3))
@@ -924,12 +968,12 @@ static void bnx2x_sp_event(struct bnx2x_fastpath *fp,
 
        DP(BNX2X_MSG_SP,
           "fp %d  cid %d  got ramrod #%d  state is %x  type is %d\n",
-          FP_IDX(fp), cid, command, bp->state,
+          fp->index, cid, command, bp->state,
           rr_cqe->ramrod_cqe.ramrod_type);
 
        bp->spq_left++;
 
-       if (FP_IDX(fp)) {
+       if (fp->index) {
                switch (command | fp->state) {
                case (RAMROD_CMD_ID_ETH_CLIENT_SETUP |
                                                BNX2X_FP_STATE_OPENING):
@@ -1406,7 +1450,7 @@ static inline void bnx2x_update_rx_prod(struct bnx2x *bp,
 
        for (i = 0; i < sizeof(struct ustorm_eth_rx_producers)/4; i++)
                REG_WR(bp, BAR_USTRORM_INTMEM +
-                      USTORM_RX_PRODS_OFFSET(BP_PORT(bp), FP_CL_ID(fp)) + i*4,
+                      USTORM_RX_PRODS_OFFSET(BP_PORT(bp), fp->cl_id) + i*4,
                       ((u32 *)&rx_prods)[i]);
 
        mmiowb(); /* keep prod updates ordered */
@@ -1447,7 +1491,7 @@ static int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 
        DP(NETIF_MSG_RX_STATUS,
           "queue[%d]:  hw_comp_cons %u  sw_comp_cons %u\n",
-          FP_IDX(fp), hw_comp_cons, sw_comp_cons);
+          fp->index, hw_comp_cons, sw_comp_cons);
 
        while (sw_comp_cons != hw_comp_cons) {
                struct sw_rx_bd *rx_buf = NULL;
@@ -1516,7 +1560,7 @@ static int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
                                                    len, cqe, comp_ring_cons);
 #ifdef BNX2X_STOP_ON_ERROR
                                        if (bp->panic)
-                                               return -EINVAL;
+                                               return 0;
 #endif
 
                                        bnx2x_update_sge_prod(fp,
@@ -1643,7 +1687,7 @@ static irqreturn_t bnx2x_msix_fp_int(int irq, void *fp_cookie)
 {
        struct bnx2x_fastpath *fp = fp_cookie;
        struct bnx2x *bp = fp->bp;
-       int index = FP_IDX(fp);
+       int index = fp->index;
 
        /* Return here if interrupt is disabled */
        if (unlikely(atomic_read(&bp->intr_sem) != 0)) {
@@ -1652,8 +1696,8 @@ static irqreturn_t bnx2x_msix_fp_int(int irq, void *fp_cookie)
        }
 
        DP(BNX2X_MSG_FP, "got an MSI-X interrupt on IDX:SB [%d:%d]\n",
-          index, FP_SB_ID(fp));
-       bnx2x_ack_sb(bp, FP_SB_ID(fp), USTORM_ID, 0, IGU_INT_DISABLE, 0);
+          index, fp->sb_id);
+       bnx2x_ack_sb(bp, fp->sb_id, USTORM_ID, 0, IGU_INT_DISABLE, 0);
 
 #ifdef BNX2X_STOP_ON_ERROR
        if (unlikely(bp->panic))
@@ -1681,7 +1725,7 @@ static irqreturn_t bnx2x_interrupt(int irq, void *dev_instance)
                DP(NETIF_MSG_INTR, "not our interrupt!\n");
                return IRQ_NONE;
        }
-       DP(NETIF_MSG_INTR, "got an interrupt  status %u\n", status);
+       DP(NETIF_MSG_INTR, "got an interrupt  status 0x%x\n", status);
 
        /* Return here if interrupt is disabled */
        if (unlikely(atomic_read(&bp->intr_sem) != 0)) {
@@ -1816,26 +1860,50 @@ static int bnx2x_release_hw_lock(struct bnx2x *bp, u32 resource)
 /* HW Lock for shared dual port PHYs */
 static void bnx2x_acquire_phy_lock(struct bnx2x *bp)
 {
-       u32 ext_phy_type = XGXS_EXT_PHY_TYPE(bp->link_params.ext_phy_config);
-
        mutex_lock(&bp->port.phy_mutex);
 
-       if ((ext_phy_type == PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072) ||
-           (ext_phy_type == PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073))
-               bnx2x_acquire_hw_lock(bp, HW_LOCK_RESOURCE_8072_MDIO);
+       if (bp->port.need_hw_lock)
+               bnx2x_acquire_hw_lock(bp, HW_LOCK_RESOURCE_MDIO);
 }
 
 static void bnx2x_release_phy_lock(struct bnx2x *bp)
 {
-       u32 ext_phy_type = XGXS_EXT_PHY_TYPE(bp->link_params.ext_phy_config);
-
-       if ((ext_phy_type == PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072) ||
-           (ext_phy_type == PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073))
-               bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_8072_MDIO);
+       if (bp->port.need_hw_lock)
+               bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_MDIO);
 
        mutex_unlock(&bp->port.phy_mutex);
 }
 
+int bnx2x_get_gpio(struct bnx2x *bp, int gpio_num, u8 port)
+{
+       /* The GPIO should be swapped if swap register is set and active */
+       int gpio_port = (REG_RD(bp, NIG_REG_PORT_SWAP) &&
+                        REG_RD(bp, NIG_REG_STRAP_OVERRIDE)) ^ port;
+       int gpio_shift = gpio_num +
+                       (gpio_port ? MISC_REGISTERS_GPIO_PORT_SHIFT : 0);
+       u32 gpio_mask = (1 << gpio_shift);
+       u32 gpio_reg;
+       int value;
+
+       if (gpio_num > MISC_REGISTERS_GPIO_3) {
+               BNX2X_ERR("Invalid GPIO %d\n", gpio_num);
+               return -EINVAL;
+       }
+
+       /* read GPIO value */
+       gpio_reg = REG_RD(bp, MISC_REG_GPIO);
+
+       /* get the requested pin value */
+       if ((gpio_reg & gpio_mask) == gpio_mask)
+               value = 1;
+       else
+               value = 0;
+
+       DP(NETIF_MSG_LINK, "pin %d  value 0x%x\n", gpio_num, value);
+
+       return value;
+}
+
 int bnx2x_set_gpio(struct bnx2x *bp, int gpio_num, u32 mode, u8 port)
 {
        /* The GPIO should be swapped if swap register is set and active */
@@ -1889,6 +1957,52 @@ int bnx2x_set_gpio(struct bnx2x *bp, int gpio_num, u32 mode, u8 port)
        return 0;
 }
 
+int bnx2x_set_gpio_int(struct bnx2x *bp, int gpio_num, u32 mode, u8 port)
+{
+       /* The GPIO should be swapped if swap register is set and active */
+       int gpio_port = (REG_RD(bp, NIG_REG_PORT_SWAP) &&
+                        REG_RD(bp, NIG_REG_STRAP_OVERRIDE)) ^ port;
+       int gpio_shift = gpio_num +
+                       (gpio_port ? MISC_REGISTERS_GPIO_PORT_SHIFT : 0);
+       u32 gpio_mask = (1 << gpio_shift);
+       u32 gpio_reg;
+
+       if (gpio_num > MISC_REGISTERS_GPIO_3) {
+               BNX2X_ERR("Invalid GPIO %d\n", gpio_num);
+               return -EINVAL;
+       }
+
+       bnx2x_acquire_hw_lock(bp, HW_LOCK_RESOURCE_GPIO);
+       /* read GPIO int */
+       gpio_reg = REG_RD(bp, MISC_REG_GPIO_INT);
+
+       switch (mode) {
+       case MISC_REGISTERS_GPIO_INT_OUTPUT_CLR:
+               DP(NETIF_MSG_LINK, "Clear GPIO INT %d (shift %d) -> "
+                                  "output low\n", gpio_num, gpio_shift);
+               /* 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:
+               DP(NETIF_MSG_LINK, "Set GPIO INT %d (shift %d) -> "
+                                  "output high\n", gpio_num, gpio_shift);
+               /* 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(bp, MISC_REG_GPIO_INT, gpio_reg);
+       bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_GPIO);
+
+       return 0;
+}
+
 static int bnx2x_set_spio(struct bnx2x *bp, int spio_num, u32 mode)
 {
        u32 spio_mask = (1 << spio_num);
@@ -1943,13 +2057,16 @@ static void bnx2x_calc_fc_adv(struct bnx2x *bp)
                bp->port.advertising &= ~(ADVERTISED_Asym_Pause |
                                          ADVERTISED_Pause);
                break;
+
        case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_BOTH:
                bp->port.advertising |= (ADVERTISED_Asym_Pause |
                                         ADVERTISED_Pause);
                break;
+
        case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_ASYMMETRIC:
                bp->port.advertising |= ADVERTISED_Asym_Pause;
                break;
+
        default:
                bp->port.advertising &= ~(ADVERTISED_Asym_Pause |
                                          ADVERTISED_Pause);
@@ -1974,7 +2091,8 @@ static void bnx2x_link_report(struct bnx2x *bp)
                if (bp->link_vars.flow_ctrl != BNX2X_FLOW_CTRL_NONE) {
                        if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_RX) {
                                printk(", receive ");
-                               if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_TX)
+                               if (bp->link_vars.flow_ctrl &
+                                   BNX2X_FLOW_CTRL_TX)
                                        printk("& transmit ");
                        } else {
                                printk(", transmit ");
@@ -1989,7 +2107,7 @@ static void bnx2x_link_report(struct bnx2x *bp)
        }
 }
 
-static u8 bnx2x_initial_phy_init(struct bnx2x *bp)
+static u8 bnx2x_initial_phy_init(struct bnx2x *bp, int load_mode)
 {
        if (!BP_NOMCP(bp)) {
                u8 rc;
@@ -2005,18 +2123,24 @@ static u8 bnx2x_initial_phy_init(struct bnx2x *bp)
                        bp->link_params.req_fc_auto_adv = BNX2X_FLOW_CTRL_BOTH;
 
                bnx2x_acquire_phy_lock(bp);
+
+               if (load_mode == LOAD_DIAG)
+                       bp->link_params.loopback_mode = LOOPBACK_XGXS_10;
+
                rc = bnx2x_phy_init(&bp->link_params, &bp->link_vars);
+
                bnx2x_release_phy_lock(bp);
 
                bnx2x_calc_fc_adv(bp);
 
-               if (bp->link_vars.link_up)
+               if (CHIP_REV_IS_SLOW(bp) && bp->link_vars.link_up) {
+                       bnx2x_stats_handle(bp, STATS_EVENT_LINK_UP);
                        bnx2x_link_report(bp);
-
+               }
 
                return rc;
        }
-       BNX2X_ERR("Bootcode is missing -not initializing link\n");
+       BNX2X_ERR("Bootcode is missing - can not initialize link\n");
        return -EINVAL;
 }
 
@@ -2029,17 +2153,17 @@ static void bnx2x_link_set(struct bnx2x *bp)
 
                bnx2x_calc_fc_adv(bp);
        } else
-               BNX2X_ERR("Bootcode is missing -not setting link\n");
+               BNX2X_ERR("Bootcode is missing - can not set link\n");
 }
 
 static void bnx2x__link_reset(struct bnx2x *bp)
 {
        if (!BP_NOMCP(bp)) {
                bnx2x_acquire_phy_lock(bp);
-               bnx2x_link_reset(&bp->link_params, &bp->link_vars);
+               bnx2x_link_reset(&bp->link_params, &bp->link_vars, 1);
                bnx2x_release_phy_lock(bp);
        } else
-               BNX2X_ERR("Bootcode is missing -not resetting link\n");
+               BNX2X_ERR("Bootcode is missing - can not reset link\n");
 }
 
 static u8 bnx2x_link_test(struct bnx2x *bp)
@@ -2319,9 +2443,14 @@ static int bnx2x_sp_post(struct bnx2x *bp, int command, int cid,
                bp->spq_prod_idx++;
        }
 
+       /* Make sure that BD data is updated before writing the producer */
+       wmb();
+
        REG_WR(bp, BAR_XSTRORM_INTMEM + XSTORM_SPQ_PROD_OFFSET(func),
               bp->spq_prod_idx);
 
+       mmiowb();
+
        spin_unlock_bh(&bp->spq_lock);
        return 0;
 }
@@ -2488,6 +2617,22 @@ static void bnx2x_attn_int_asserted(struct bnx2x *bp, u32 asserted)
        }
 }
 
+static inline void bnx2x_fan_failure(struct bnx2x *bp)
+{
+       int port = BP_PORT(bp);
+
+       /* mark the failure */
+       bp->link_params.ext_phy_config &= ~PORT_HW_CFG_XGXS_EXT_PHY_TYPE_MASK;
+       bp->link_params.ext_phy_config |= PORT_HW_CFG_XGXS_EXT_PHY_TYPE_FAILURE;
+       SHMEM_WR(bp, dev_info.port_hw_config[port].external_phy_config,
+                bp->link_params.ext_phy_config);
+
+       /* log the failure */
+       printk(KERN_ERR PFX "Fan Failure on Network Controller %s has caused"
+              " the driver to shutdown the card to prevent permanent"
+              " damage.  Please contact Dell Support for assistance\n",
+              bp->dev->name);
+}
 static inline void bnx2x_attn_int_deasserted0(struct bnx2x *bp, u32 attn)
 {
        int port = BP_PORT(bp);
@@ -2505,36 +2650,28 @@ static inline void bnx2x_attn_int_deasserted0(struct bnx2x *bp, u32 attn)
 
                BNX2X_ERR("SPIO5 hw attention\n");
 
+               /* Fan failure attention */
                switch (XGXS_EXT_PHY_TYPE(bp->link_params.ext_phy_config)) {
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_SFX7101:
-                       /* Fan failure attention */
-
-                       /* The PHY reset is controlled by GPIO 1 */
-                       bnx2x_set_gpio(bp, MISC_REGISTERS_GPIO_1,
-                                      MISC_REGISTERS_GPIO_OUTPUT_LOW, port);
                        /* Low power mode is controlled by GPIO 2 */
                        bnx2x_set_gpio(bp, MISC_REGISTERS_GPIO_2,
                                       MISC_REGISTERS_GPIO_OUTPUT_LOW, port);
-                       /* mark the failure */
-                       bp->link_params.ext_phy_config &=
-                                       ~PORT_HW_CFG_XGXS_EXT_PHY_TYPE_MASK;
-                       bp->link_params.ext_phy_config |=
-                                       PORT_HW_CFG_XGXS_EXT_PHY_TYPE_FAILURE;
-                       SHMEM_WR(bp,
-                                dev_info.port_hw_config[port].
-                                                       external_phy_config,
-                                bp->link_params.ext_phy_config);
-                       /* log the failure */
-                       printk(KERN_ERR PFX "Fan Failure on Network"
-                              " Controller %s has caused the driver to"
-                              " shutdown the card to prevent permanent"
-                              " damage.  Please contact Dell Support for"
-                              " assistance\n", bp->dev->name);
+                       /* The PHY reset is controlled by GPIO 1 */
+                       bnx2x_set_gpio(bp, MISC_REGISTERS_GPIO_1,
+                                      MISC_REGISTERS_GPIO_OUTPUT_LOW, port);
                        break;
 
                default:
                        break;
                }
+               bnx2x_fan_failure(bp);
+       }
+
+       if (attn & (AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_0 |
+                   AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_1)) {
+               bnx2x_acquire_phy_lock(bp);
+               bnx2x_handle_module_detect_int(&bp->link_params);
+               bnx2x_release_phy_lock(bp);
        }
 
        if (attn & HW_INTERRUT_ASSERT_SET_0) {
@@ -2553,7 +2690,7 @@ static inline void bnx2x_attn_int_deasserted1(struct bnx2x *bp, u32 attn)
 {
        u32 val;
 
-       if (attn & BNX2X_DOORQ_ASSERT) {
+       if (attn & AEU_INPUTS_ATTN_BITS_DOORBELLQ_HW_INTERRUPT) {
 
                val = REG_RD(bp, DORQ_REG_DORQ_INT_STS_CLR);
                BNX2X_ERR("DB hw attention 0x%x\n", val);
@@ -2912,8 +3049,8 @@ static irqreturn_t bnx2x_msix_sp_int(int irq, void *dev_instance)
 
 #define UPDATE_EXTEND_TSTAT(s, t) \
        do { \
-               diff = le32_to_cpu(tclient->s) - old_tclient->s; \
-               old_tclient->s = le32_to_cpu(tclient->s); \
+               diff = le32_to_cpu(tclient->s) - le32_to_cpu(old_tclient->s); \
+               old_tclient->s = tclient->s; \
                ADD_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff); \
        } while (0)
 
@@ -2926,8 +3063,8 @@ static irqreturn_t bnx2x_msix_sp_int(int irq, void *dev_instance)
 
 #define UPDATE_EXTEND_XSTAT(s, t) \
        do { \
-               diff = le32_to_cpu(xclient->s) - old_xclient->s; \
-               old_xclient->s = le32_to_cpu(xclient->s); \
+               diff = le32_to_cpu(xclient->s) - le32_to_cpu(old_xclient->s); \
+               old_xclient->s = xclient->s; \
                ADD_EXTEND_64(qstats->t##_hi, qstats->t##_lo, diff); \
        } while (0)
 
@@ -3436,7 +3573,10 @@ static void bnx2x_bmac_stats_update(struct bnx2x *bp)
        struct bmac_stats *new = bnx2x_sp(bp, mac_stats.bmac_stats);
        struct host_port_stats *pstats = bnx2x_sp(bp, port_stats);
        struct bnx2x_eth_stats *estats = &bp->eth_stats;
-       struct regpair diff;
+       struct {
+               u32 lo;
+               u32 hi;
+       } diff;
 
        UPDATE_STAT64(rx_stat_grerb, rx_stat_ifhcinbadoctets);
        UPDATE_STAT64(rx_stat_grfcs, rx_stat_dot3statsfcserrors);
@@ -3542,7 +3682,10 @@ static int bnx2x_hw_stats_update(struct bnx2x *bp)
        struct nig_stats *old = &(bp->port.old_nig_stats);
        struct host_port_stats *pstats = bnx2x_sp(bp, port_stats);
        struct bnx2x_eth_stats *estats = &bp->eth_stats;
-       struct regpair diff;
+       struct {
+               u32 lo;
+               u32 hi;
+       } diff;
        u32 nig_timer_max;
 
        if (bp->link_vars.mac_type == MAC_TYPE_BMAC)
@@ -3552,7 +3695,7 @@ static int bnx2x_hw_stats_update(struct bnx2x *bp)
                bnx2x_emac_stats_update(bp);
 
        else { /* unreached */
-               BNX2X_ERR("stats updated by dmae but no MAC active\n");
+               BNX2X_ERR("stats updated by DMAE but no MAC active\n");
                return -1;
        }
 
@@ -3906,12 +4049,12 @@ static void bnx2x_stats_update(struct bnx2x *bp)
                        "mac_discard %u  mac_filter_discard %u  "
                        "xxovrflow_discard %u  brb_truncate_discard %u  "
                        "ttl0_discard %u\n",
-                      old_tclient->checksum_discard,
+                      le32_to_cpu(old_tclient->checksum_discard),
                       bnx2x_hilo(&qstats->etherstatsoverrsizepkts_hi),
                       bnx2x_hilo(&qstats->no_buff_discard_hi),
                       estats->mac_discard, estats->mac_filter_discard,
                       estats->xxoverflow_discard, estats->brb_truncate_discard,
-                      old_tclient->ttl0_discard);
+                      le32_to_cpu(old_tclient->ttl0_discard));
 
                for_each_queue(bp, i) {
                        printk(KERN_DEBUG "[%d]: %lu\t%lu\t%lu\n", i,
@@ -4060,7 +4203,7 @@ static void bnx2x_timer(unsigned long data)
                struct bnx2x_fastpath *fp = &bp->fp[0];
                int rc;
 
-               bnx2x_tx_int(fp, 1000);
+               bnx2x_tx_int(fp);
                rc = bnx2x_rx_int(fp, 1000);
        }
 
@@ -4108,10 +4251,10 @@ static void bnx2x_zero_sb(struct bnx2x *bp, int sb_id)
 {
        int port = BP_PORT(bp);
 
-       bnx2x_init_fill(bp, BAR_USTRORM_INTMEM +
+       bnx2x_init_fill(bp, USTORM_INTMEM_ADDR +
                        USTORM_SB_HOST_STATUS_BLOCK_OFFSET(port, sb_id), 0,
                        sizeof(struct ustorm_status_block)/4);
-       bnx2x_init_fill(bp, BAR_CSTRORM_INTMEM +
+       bnx2x_init_fill(bp, CSTORM_INTMEM_ADDR +
                        CSTORM_SB_HOST_STATUS_BLOCK_OFFSET(port, sb_id), 0,
                        sizeof(struct cstorm_status_block)/4);
 }
@@ -4165,18 +4308,18 @@ static void bnx2x_zero_def_sb(struct bnx2x *bp)
 {
        int func = BP_FUNC(bp);
 
-       bnx2x_init_fill(bp, BAR_USTRORM_INTMEM +
+       bnx2x_init_fill(bp, TSTORM_INTMEM_ADDR +
+                       TSTORM_DEF_SB_HOST_STATUS_BLOCK_OFFSET(func), 0,
+                       sizeof(struct tstorm_def_status_block)/4);
+       bnx2x_init_fill(bp, USTORM_INTMEM_ADDR +
                        USTORM_DEF_SB_HOST_STATUS_BLOCK_OFFSET(func), 0,
                        sizeof(struct ustorm_def_status_block)/4);
-       bnx2x_init_fill(bp, BAR_CSTRORM_INTMEM +
+       bnx2x_init_fill(bp, CSTORM_INTMEM_ADDR +
                        CSTORM_DEF_SB_HOST_STATUS_BLOCK_OFFSET(func), 0,
                        sizeof(struct cstorm_def_status_block)/4);
-       bnx2x_init_fill(bp, BAR_XSTRORM_INTMEM +
+       bnx2x_init_fill(bp, XSTORM_INTMEM_ADDR +
                        XSTORM_DEF_SB_HOST_STATUS_BLOCK_OFFSET(func), 0,
                        sizeof(struct xstorm_def_status_block)/4);
-       bnx2x_init_fill(bp, BAR_TSTRORM_INTMEM +
-                       TSTORM_DEF_SB_HOST_STATUS_BLOCK_OFFSET(func), 0,
-                       sizeof(struct tstorm_def_status_block)/4);
 }
 
 static void bnx2x_init_def_sb(struct bnx2x *bp,
@@ -4311,7 +4454,7 @@ static void bnx2x_update_coalesce(struct bnx2x *bp)
                REG_WR16(bp, BAR_USTRORM_INTMEM +
                         USTORM_SB_HC_DISABLE_OFFSET(port, sb_id,
                                                     U_SB_ETH_RX_CQ_INDEX),
-                        bp->rx_ticks ? 0 : 1);
+                        (bp->rx_ticks/12) ? 0 : 1);
 
                /* HC_INDEX_C_ETH_TX_CQ_CONS */
                REG_WR8(bp, BAR_CSTRORM_INTMEM +
@@ -4321,7 +4464,7 @@ static void bnx2x_update_coalesce(struct bnx2x *bp)
                REG_WR16(bp, BAR_CSTRORM_INTMEM +
                         CSTORM_SB_HC_DISABLE_OFFSET(port, sb_id,
                                                     C_SB_ETH_TX_CQ_INDEX),
-                        bp->tx_ticks ? 0 : 1);
+                        (bp->tx_ticks/12) ? 0 : 1);
        }
 }
 
@@ -4342,8 +4485,7 @@ static inline void bnx2x_free_tpa_pool(struct bnx2x *bp,
                if (fp->tpa_state[i] == BNX2X_TPA_START)
                        pci_unmap_single(bp->pdev,
                                         pci_unmap_addr(rx_buf, mapping),
-                                        bp->rx_buf_size,
-                                        PCI_DMA_FROMDEVICE);
+                                        bp->rx_buf_size, PCI_DMA_FROMDEVICE);
 
                dev_kfree_skb(skb);
                rx_buf->skb = NULL;
@@ -4553,11 +4695,11 @@ static void bnx2x_init_context(struct bnx2x *bp)
                struct eth_context *context = bnx2x_sp(bp, context[i].eth);
                struct bnx2x_fastpath *fp = &bp->fp[i];
                u8 cl_id = fp->cl_id;
-               u8 sb_id = FP_SB_ID(fp);
+               u8 sb_id = fp->sb_id;
 
                context->ustorm_st_context.common.sb_index_numbers =
                                                BNX2X_RX_SB_INDEX_NUM;
-               context->ustorm_st_context.common.clientId = FP_CL_ID(fp);
+               context->ustorm_st_context.common.clientId = cl_id;
                context->ustorm_st_context.common.status_block_id = sb_id;
                context->ustorm_st_context.common.flags =
                        (USTORM_ETH_ST_CONTEXT_CONFIG_ENABLE_MC_ALIGNMENT |
@@ -4598,7 +4740,7 @@ static void bnx2x_init_context(struct bnx2x *bp)
                                                U64_HI(fp->tx_prods_mapping);
                context->xstorm_st_context.db_data_addr_lo =
                                                U64_LO(fp->tx_prods_mapping);
-               context->xstorm_st_context.statistics_data = (fp->cl_id |
+               context->xstorm_st_context.statistics_data = (cl_id |
                                XSTORM_ETH_ST_CONTEXT_STATISTICS_ENABLE);
                context->cstorm_st_context.sb_index_number =
                                                C_SB_ETH_TX_CQ_INDEX;
@@ -4624,7 +4766,7 @@ static void bnx2x_init_ind_table(struct bnx2x *bp)
        for (i = 0; i < TSTORM_INDIRECTION_TABLE_SIZE; i++)
                REG_WR8(bp, BAR_TSTRORM_INTMEM +
                        TSTORM_INDIRECTION_TABLE_OFFSET(func) + i,
-                       BP_CL_ID(bp) + (i % bp->num_rx_queues));
+                       bp->fp->cl_id + (i % bp->num_rx_queues));
 }
 
 static void bnx2x_set_client_config(struct bnx2x *bp)
@@ -4688,18 +4830,22 @@ static void bnx2x_set_storm_rx_mode(struct bnx2x *bp)
                tstorm_mac_filter.mcast_drop_all = mask;
                tstorm_mac_filter.bcast_drop_all = mask;
                break;
+
        case BNX2X_RX_MODE_NORMAL:
                tstorm_mac_filter.bcast_accept_all = mask;
                break;
+
        case BNX2X_RX_MODE_ALLMULTI:
                tstorm_mac_filter.mcast_accept_all = mask;
                tstorm_mac_filter.bcast_accept_all = mask;
                break;
+
        case BNX2X_RX_MODE_PROMISC:
                tstorm_mac_filter.ucast_accept_all = mask;
                tstorm_mac_filter.mcast_accept_all = mask;
                tstorm_mac_filter.bcast_accept_all = mask;
                break;
+
        default:
                BNX2X_ERR("BAD rx mode (%d)\n", mode);
                break;
@@ -4910,14 +5056,14 @@ static void bnx2x_init_internal_func(struct bnx2x *bp)
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                REG_WR(bp, BAR_USTRORM_INTMEM +
-                      USTORM_CQE_PAGE_BASE_OFFSET(port, FP_CL_ID(fp)),
+                      USTORM_CQE_PAGE_BASE_OFFSET(port, fp->cl_id),
                       U64_LO(fp->rx_comp_mapping));
                REG_WR(bp, BAR_USTRORM_INTMEM +
-                      USTORM_CQE_PAGE_BASE_OFFSET(port, FP_CL_ID(fp)) + 4,
+                      USTORM_CQE_PAGE_BASE_OFFSET(port, fp->cl_id) + 4,
                       U64_HI(fp->rx_comp_mapping));
 
                REG_WR16(bp, BAR_USTRORM_INTMEM +
-                        USTORM_MAX_AGG_SIZE_OFFSET(port, FP_CL_ID(fp)),
+                        USTORM_MAX_AGG_SIZE_OFFSET(port, fp->cl_id),
                         max_agg_size);
        }
 
@@ -5027,13 +5173,17 @@ static void bnx2x_nic_init(struct bnx2x *bp, u32 load_code)
                fp->cl_id = BP_L_ID(bp) + i;
                fp->sb_id = fp->cl_id;
                DP(NETIF_MSG_IFUP,
-                  "bnx2x_init_sb(%p,%p) index %d  cl_id %d  sb %d\n",
-                  bp, fp->status_blk, i, FP_CL_ID(fp), FP_SB_ID(fp));
+                  "queue[%d]:  bnx2x_init_sb(%p,%p)  cl_id %d  sb %d\n",
+                  i, bp, fp->status_blk, fp->cl_id, fp->sb_id);
                bnx2x_init_sb(bp, fp->status_blk, fp->status_blk_mapping,
-                             FP_SB_ID(fp));
+                             fp->sb_id);
                bnx2x_update_fpsb_idx(fp);
        }
 
+       /* ensure status block indices were read */
+       rmb();
+
+
        bnx2x_init_def_sb(bp, bp->def_status_blk, bp->def_status_blk_mapping,
                          DEF_SB_ID);
        bnx2x_update_dsb_idx(bp);
@@ -5054,6 +5204,11 @@ static void bnx2x_nic_init(struct bnx2x *bp, u32 load_code)
        mmiowb();
 
        bnx2x_int_enable(bp);
+
+       /* Check for SPIO5 */
+       bnx2x_attn_int_deasserted0(bp,
+               REG_RD(bp, MISC_REG_AEU_AFTER_INVERT_1_FUNC_0 + BP_PORT(bp)*4) &
+                                  AEU_INPUTS_ATTN_BITS_SPIO5);
 }
 
 /* end of nic init */
@@ -5109,13 +5264,15 @@ static void bnx2x_gunzip_end(struct bnx2x *bp)
        }
 }
 
-static int bnx2x_gunzip(struct bnx2x *bp, u8 *zbuf, int len)
+static int bnx2x_gunzip(struct bnx2x *bp, const u8 *zbuf, int len)
 {
        int n, rc;
 
        /* check gzip header */
-       if ((zbuf[0] != 0x1f) || (zbuf[1] != 0x8b) || (zbuf[2] != Z_DEFLATED))
+       if ((zbuf[0] != 0x1f) || (zbuf[1] != 0x8b) || (zbuf[2] != Z_DEFLATED)) {
+               BNX2X_ERR("Bad gzip header\n");
                return -EINVAL;
+       }
 
        n = 10;
 
@@ -5124,7 +5281,7 @@ static int bnx2x_gunzip(struct bnx2x *bp, u8 *zbuf, int len)
        if (zbuf[3] & FNAME)
                while ((zbuf[n++] != 0) && (n < len));
 
-       bp->strm->next_in = zbuf + n;
+       bp->strm->next_in = (typeof(bp->strm->next_in))zbuf + n;
        bp->strm->avail_in = len - n;
        bp->strm->next_out = bp->gunzip_buf;
        bp->strm->avail_out = FW_BUF_SIZE;
@@ -5246,8 +5403,8 @@ static int bnx2x_int_mem_test(struct bnx2x *bp)
        msleep(50);
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET, 0x03);
        msleep(50);
-       bnx2x_init_block(bp, BRB1_COMMON_START, BRB1_COMMON_END);
-       bnx2x_init_block(bp, PRS_COMMON_START, PRS_COMMON_END);
+       bnx2x_init_block(bp, BRB1_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, PRS_BLOCK, COMMON_STAGE);
 
        DP(NETIF_MSG_HW, "part2\n");
 
@@ -5311,8 +5468,8 @@ static int bnx2x_int_mem_test(struct bnx2x *bp)
        msleep(50);
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET, 0x03);
        msleep(50);
-       bnx2x_init_block(bp, BRB1_COMMON_START, BRB1_COMMON_END);
-       bnx2x_init_block(bp, PRS_COMMON_START, PRS_COMMON_END);
+       bnx2x_init_block(bp, BRB1_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, PRS_BLOCK, COMMON_STAGE);
 #ifndef BCM_ISCSI
        /* set NIC mode */
        REG_WR(bp, PRS_REG_NIC_MODE, 1);
@@ -5377,6 +5534,58 @@ static void bnx2x_reset_common(struct bnx2x *bp)
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_CLEAR, 0x1403);
 }
 
+
+static void bnx2x_setup_fan_failure_detection(struct bnx2x *bp)
+{
+       u32 val;
+       u8 port;
+       u8 is_required = 0;
+
+       val = SHMEM_RD(bp, 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++) {
+                       u32 phy_type =
+                               SHMEM_RD(bp, dev_info.port_hw_config[port].
+                                        external_phy_config) &
+                               PORT_HW_CFG_XGXS_EXT_PHY_TYPE_MASK;
+                       is_required |=
+                               ((phy_type ==
+                                 PORT_HW_CFG_XGXS_EXT_PHY_TYPE_SFX7101) ||
+                                (phy_type ==
+                                 PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8481));
+               }
+
+       DP(NETIF_MSG_HW, "fan detection setting: %d\n", is_required);
+
+       if (is_required == 0)
+               return;
+
+       /* Fan failure is indicated by SPIO 5 */
+       bnx2x_set_spio(bp, MISC_REGISTERS_SPIO_5,
+                      MISC_REGISTERS_SPIO_INPUT_HI_Z);
+
+       /* set to active low mode */
+       val = REG_RD(bp, MISC_REG_SPIO_INT);
+       val |= ((1 << MISC_REGISTERS_SPIO_5) <<
+                               MISC_REGISTERS_SPIO_INT_OLD_SET_POS);
+       REG_WR(bp, MISC_REG_SPIO_INT, val);
+
+       /* enable interrupt to signal the IGU */
+       val = REG_RD(bp, MISC_REG_SPIO_EVENT_EN);
+       val |= (1 << MISC_REGISTERS_SPIO_5);
+       REG_WR(bp, MISC_REG_SPIO_EVENT_EN, val);
+}
+
 static int bnx2x_init_common(struct bnx2x *bp)
 {
        u32 val, i;
@@ -5387,7 +5596,7 @@ static int bnx2x_init_common(struct bnx2x *bp)
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET, 0xffffffff);
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_SET, 0xfffc);
 
-       bnx2x_init_block(bp, MISC_COMMON_START, MISC_COMMON_END);
+       bnx2x_init_block(bp, MISC_BLOCK, COMMON_STAGE);
        if (CHIP_IS_E1H(bp))
                REG_WR(bp, MISC_REG_E1HMF_MODE, IS_E1HMF(bp));
 
@@ -5395,14 +5604,14 @@ static int bnx2x_init_common(struct bnx2x *bp)
        msleep(30);
        REG_WR(bp, MISC_REG_LCPLL_CTRL_REG_2, 0x0);
 
-       bnx2x_init_block(bp, PXP_COMMON_START, PXP_COMMON_END);
+       bnx2x_init_block(bp, PXP_BLOCK, COMMON_STAGE);
        if (CHIP_IS_E1(bp)) {
                /* enable HW interrupt from PXP on USDM overflow
                   bit 16 on INT_MASK_0 */
                REG_WR(bp, PXP_REG_PXP_INT_MASK_0, 0);
        }
 
-       bnx2x_init_block(bp, PXP2_COMMON_START, PXP2_COMMON_END);
+       bnx2x_init_block(bp, PXP2_BLOCK, COMMON_STAGE);
        bnx2x_init_pxp(bp);
 
 #ifdef __BIG_ENDIAN
@@ -5448,87 +5657,60 @@ static int bnx2x_init_common(struct bnx2x *bp)
        REG_WR(bp, PXP2_REG_RQ_DISABLE_INPUTS, 0);
        REG_WR(bp, PXP2_REG_RD_DISABLE_INPUTS, 0);
 
-       bnx2x_init_block(bp, DMAE_COMMON_START, DMAE_COMMON_END);
+       bnx2x_init_block(bp, DMAE_BLOCK, COMMON_STAGE);
 
        /* clean the DMAE memory */
        bp->dmae_ready = 1;
        bnx2x_init_fill(bp, TSEM_REG_PRAM, 0, 8);
 
-       bnx2x_init_block(bp, TCM_COMMON_START, TCM_COMMON_END);
-       bnx2x_init_block(bp, UCM_COMMON_START, UCM_COMMON_END);
-       bnx2x_init_block(bp, CCM_COMMON_START, CCM_COMMON_END);
-       bnx2x_init_block(bp, XCM_COMMON_START, XCM_COMMON_END);
+       bnx2x_init_block(bp, TCM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, UCM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, CCM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, XCM_BLOCK, COMMON_STAGE);
 
        bnx2x_read_dmae(bp, XSEM_REG_PASSIVE_BUFFER, 3);
        bnx2x_read_dmae(bp, CSEM_REG_PASSIVE_BUFFER, 3);
        bnx2x_read_dmae(bp, TSEM_REG_PASSIVE_BUFFER, 3);
        bnx2x_read_dmae(bp, USEM_REG_PASSIVE_BUFFER, 3);
 
-       bnx2x_init_block(bp, QM_COMMON_START, QM_COMMON_END);
+       bnx2x_init_block(bp, QM_BLOCK, COMMON_STAGE);
        /* soft reset pulse */
        REG_WR(bp, QM_REG_SOFT_RESET, 1);
        REG_WR(bp, QM_REG_SOFT_RESET, 0);
 
 #ifdef BCM_ISCSI
-       bnx2x_init_block(bp, TIMERS_COMMON_START, TIMERS_COMMON_END);
+       bnx2x_init_block(bp, TIMERS_BLOCK, COMMON_STAGE);
 #endif
 
-       bnx2x_init_block(bp, DQ_COMMON_START, DQ_COMMON_END);
+       bnx2x_init_block(bp, DQ_BLOCK, COMMON_STAGE);
        REG_WR(bp, DORQ_REG_DPM_CID_OFST, BCM_PAGE_SHIFT);
        if (!CHIP_REV_IS_SLOW(bp)) {
                /* enable hw interrupt from doorbell Q */
                REG_WR(bp, DORQ_REG_DORQ_INT_MASK, 0);
        }
 
-       bnx2x_init_block(bp, BRB1_COMMON_START, BRB1_COMMON_END);
-       bnx2x_init_block(bp, PRS_COMMON_START, PRS_COMMON_END);
+       bnx2x_init_block(bp, BRB1_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, PRS_BLOCK, COMMON_STAGE);
        REG_WR(bp, PRS_REG_A_PRSU_20, 0xf);
        /* set NIC mode */
        REG_WR(bp, PRS_REG_NIC_MODE, 1);
        if (CHIP_IS_E1H(bp))
                REG_WR(bp, PRS_REG_E1HOV_MODE, IS_E1HMF(bp));
 
-       bnx2x_init_block(bp, TSDM_COMMON_START, TSDM_COMMON_END);
-       bnx2x_init_block(bp, CSDM_COMMON_START, CSDM_COMMON_END);
-       bnx2x_init_block(bp, USDM_COMMON_START, USDM_COMMON_END);
-       bnx2x_init_block(bp, XSDM_COMMON_START, XSDM_COMMON_END);
+       bnx2x_init_block(bp, TSDM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, CSDM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, USDM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, XSDM_BLOCK, COMMON_STAGE);
 
-       if (CHIP_IS_E1H(bp)) {
-               bnx2x_init_fill(bp, TSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp,
-                               TSTORM_INTMEM_ADDR + STORM_INTMEM_SIZE_E1H/2,
-                               0, STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp, CSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp,
-                               CSTORM_INTMEM_ADDR + STORM_INTMEM_SIZE_E1H/2,
-                               0, STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp, XSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp,
-                               XSTORM_INTMEM_ADDR + STORM_INTMEM_SIZE_E1H/2,
-                               0, STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp, USTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1H/2);
-               bnx2x_init_fill(bp,
-                               USTORM_INTMEM_ADDR + STORM_INTMEM_SIZE_E1H/2,
-                               0, STORM_INTMEM_SIZE_E1H/2);
-       } else { /* E1 */
-               bnx2x_init_fill(bp, TSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1);
-               bnx2x_init_fill(bp, CSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1);
-               bnx2x_init_fill(bp, XSTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1);
-               bnx2x_init_fill(bp, USTORM_INTMEM_ADDR, 0,
-                               STORM_INTMEM_SIZE_E1);
-       }
-
-       bnx2x_init_block(bp, TSEM_COMMON_START, TSEM_COMMON_END);
-       bnx2x_init_block(bp, USEM_COMMON_START, USEM_COMMON_END);
-       bnx2x_init_block(bp, CSEM_COMMON_START, CSEM_COMMON_END);
-       bnx2x_init_block(bp, XSEM_COMMON_START, XSEM_COMMON_END);
+       bnx2x_init_fill(bp, TSTORM_INTMEM_ADDR, 0, STORM_INTMEM_SIZE(bp));
+       bnx2x_init_fill(bp, USTORM_INTMEM_ADDR, 0, STORM_INTMEM_SIZE(bp));
+       bnx2x_init_fill(bp, CSTORM_INTMEM_ADDR, 0, STORM_INTMEM_SIZE(bp));
+       bnx2x_init_fill(bp, XSTORM_INTMEM_ADDR, 0, STORM_INTMEM_SIZE(bp));
+
+       bnx2x_init_block(bp, TSEM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, USEM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, CSEM_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, XSEM_BLOCK, COMMON_STAGE);
 
        /* sync semi rtc */
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR,
@@ -5536,16 +5718,16 @@ static int bnx2x_init_common(struct bnx2x *bp)
        REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_SET,
               0x80000000);
 
-       bnx2x_init_block(bp, UPB_COMMON_START, UPB_COMMON_END);
-       bnx2x_init_block(bp, XPB_COMMON_START, XPB_COMMON_END);
-       bnx2x_init_block(bp, PBF_COMMON_START, PBF_COMMON_END);
+       bnx2x_init_block(bp, UPB_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, XPB_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, PBF_BLOCK, COMMON_STAGE);
 
        REG_WR(bp, SRC_REG_SOFT_RST, 1);
        for (i = SRC_REG_KEYRSS0_0; i <= SRC_REG_KEYRSS1_9; i += 4) {
                REG_WR(bp, i, 0xc0cac01a);
                /* TODO: replace with something meaningful */
        }
-       bnx2x_init_block(bp, SRCH_COMMON_START, SRCH_COMMON_END);
+       bnx2x_init_block(bp, SRCH_BLOCK, COMMON_STAGE);
        REG_WR(bp, SRC_REG_SOFT_RST, 0);
 
        if (sizeof(union cdu_context) != 1024)
@@ -5553,7 +5735,7 @@ static int bnx2x_init_common(struct bnx2x *bp)
                printk(KERN_ALERT PFX "please adjust the size of"
                       " cdu_context(%ld)\n", (long)sizeof(union cdu_context));
 
-       bnx2x_init_block(bp, CDU_COMMON_START, CDU_COMMON_END);
+       bnx2x_init_block(bp, CDU_BLOCK, COMMON_STAGE);
        val = (4 << 24) + (0 << 12) + 1024;
        REG_WR(bp, CDU_REG_CDU_GLOBAL_PARAMS, val);
        if (CHIP_IS_E1(bp)) {
@@ -5562,7 +5744,7 @@ static int bnx2x_init_common(struct bnx2x *bp)
                REG_WR(bp, CDU_REG_CDU_DEBUG, 0);
        }
 
-       bnx2x_init_block(bp, CFC_COMMON_START, CFC_COMMON_END);
+       bnx2x_init_block(bp, CFC_BLOCK, COMMON_STAGE);
        REG_WR(bp, CFC_REG_INIT_REG, 0x7FF);
        /* enable context validation interrupt from CFC */
        REG_WR(bp, CFC_REG_CFC_INT_MASK, 0);
@@ -5570,20 +5752,25 @@ static int bnx2x_init_common(struct bnx2x *bp)
        /* set the thresholds to prevent CFC/CDU race */
        REG_WR(bp, CFC_REG_DEBUG0, 0x20020000);
 
-       bnx2x_init_block(bp, HC_COMMON_START, HC_COMMON_END);
-       bnx2x_init_block(bp, MISC_AEU_COMMON_START, MISC_AEU_COMMON_END);
+       bnx2x_init_block(bp, HC_BLOCK, COMMON_STAGE);
+       bnx2x_init_block(bp, MISC_AEU_BLOCK, COMMON_STAGE);
 
        /* PXPCS COMMON comes here */
+       bnx2x_init_block(bp, PXPCS_BLOCK, COMMON_STAGE);
        /* Reset PCIE errors for debug */
        REG_WR(bp, 0x2814, 0xffffffff);
        REG_WR(bp, 0x3820, 0xffffffff);
 
        /* EMAC0 COMMON comes here */
+       bnx2x_init_block(bp, EMAC0_BLOCK, COMMON_STAGE);
        /* EMAC1 COMMON comes here */
+       bnx2x_init_block(bp, EMAC1_BLOCK, COMMON_STAGE);
        /* DBU COMMON comes here */
+       bnx2x_init_block(bp, DBU_BLOCK, COMMON_STAGE);
        /* DBG COMMON comes here */
+       bnx2x_init_block(bp, DBG_BLOCK, COMMON_STAGE);
 
-       bnx2x_init_block(bp, NIG_COMMON_START, NIG_COMMON_END);
+       bnx2x_init_block(bp, NIG_BLOCK, COMMON_STAGE);
        if (CHIP_IS_E1H(bp)) {
                REG_WR(bp, NIG_REG_LLH_MF_MODE, IS_E1HMF(bp));
                REG_WR(bp, NIG_REG_LLH_E1HOV_MODE, IS_E1HMF(bp));
@@ -5622,27 +5809,18 @@ static int bnx2x_init_common(struct bnx2x *bp)
        }
 
        switch (XGXS_EXT_PHY_TYPE(bp->link_params.ext_phy_config)) {
-       case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_SFX7101:
-               /* Fan failure is indicated by SPIO 5 */
-               bnx2x_set_spio(bp, MISC_REGISTERS_SPIO_5,
-                              MISC_REGISTERS_SPIO_INPUT_HI_Z);
-
-               /* set to active low mode */
-               val = REG_RD(bp, MISC_REG_SPIO_INT);
-               val |= ((1 << MISC_REGISTERS_SPIO_5) <<
-                                       MISC_REGISTERS_SPIO_INT_OLD_SET_POS);
-               REG_WR(bp, MISC_REG_SPIO_INT, val);
-
-               /* enable interrupt to signal the IGU */
-               val = REG_RD(bp, MISC_REG_SPIO_EVENT_EN);
-               val |= (1 << MISC_REGISTERS_SPIO_5);
-               REG_WR(bp, MISC_REG_SPIO_EVENT_EN, val);
+       case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072:
+       case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073:
+       case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8726:
+               bp->port.need_hw_lock = 1;
                break;
 
        default:
                break;
        }
 
+       bnx2x_setup_fan_failure_detection(bp);
+
        /* clear PXP2 attentions */
        REG_RD(bp, PXP2_REG_PXP2_INT_STS_CLR_0);
 
@@ -5661,6 +5839,7 @@ static int bnx2x_init_common(struct bnx2x *bp)
 static int bnx2x_init_port(struct bnx2x *bp)
 {
        int port = BP_PORT(bp);
+       int init_stage = port ? PORT1_STAGE : PORT0_STAGE;
        u32 low, high;
        u32 val;
 
@@ -5669,7 +5848,9 @@ static int bnx2x_init_port(struct bnx2x *bp)
        REG_WR(bp, NIG_REG_MASK_INTERRUPT_PORT0 + port*4, 0);
 
        /* Port PXP comes here */
+       bnx2x_init_block(bp, PXP_BLOCK, init_stage);
        /* Port PXP2 comes here */
+       bnx2x_init_block(bp, PXP2_BLOCK, init_stage);
 #ifdef BCM_ISCSI
        /* Port0  1
         * Port1  385 */
@@ -5696,21 +5877,19 @@ static int bnx2x_init_port(struct bnx2x *bp)
        REG_WR(bp, PXP2_REG_PSWRQ_SRC0_L2P + func*4, PXP_ONE_ILT(i));
 #endif
        /* Port CMs come here */
-       bnx2x_init_block(bp, (port ? XCM_PORT1_START : XCM_PORT0_START),
-                            (port ? XCM_PORT1_END : XCM_PORT0_END));
+       bnx2x_init_block(bp, XCM_BLOCK, init_stage);
 
        /* Port QM comes here */
 #ifdef BCM_ISCSI
        REG_WR(bp, TM_REG_LIN0_SCAN_TIME + func*4, 1024/64*20);
        REG_WR(bp, TM_REG_LIN0_MAX_ACTIVE_CID + func*4, 31);
 
-       bnx2x_init_block(bp, func ? TIMERS_PORT1_START : TIMERS_PORT0_START,
-                            func ? TIMERS_PORT1_END : TIMERS_PORT0_END);
+       bnx2x_init_block(bp, TIMERS_BLOCK, init_stage);
 #endif
        /* Port DQ comes here */
+       bnx2x_init_block(bp, DQ_BLOCK, init_stage);
 
-       bnx2x_init_block(bp, (port ? BRB1_PORT1_START : BRB1_PORT0_START),
-                            (port ? BRB1_PORT1_END : BRB1_PORT0_END));
+       bnx2x_init_block(bp, BRB1_BLOCK, init_stage);
        if (CHIP_REV_IS_SLOW(bp) && !CHIP_IS_E1H(bp)) {
                /* no pause for emulation and FPGA */
                low = 0;
@@ -5735,23 +5914,27 @@ static int bnx2x_init_port(struct bnx2x *bp)
 
 
        /* Port PRS comes here */
+       bnx2x_init_block(bp, PRS_BLOCK, init_stage);
        /* Port TSDM comes here */
+       bnx2x_init_block(bp, TSDM_BLOCK, init_stage);
        /* Port CSDM comes here */
+       bnx2x_init_block(bp, CSDM_BLOCK, init_stage);
        /* Port USDM comes here */
+       bnx2x_init_block(bp, USDM_BLOCK, init_stage);
        /* Port XSDM comes here */
-       bnx2x_init_block(bp, port ? TSEM_PORT1_START : TSEM_PORT0_START,
-                            port ? TSEM_PORT1_END : TSEM_PORT0_END);
-       bnx2x_init_block(bp, port ? USEM_PORT1_START : USEM_PORT0_START,
-                            port ? USEM_PORT1_END : USEM_PORT0_END);
-       bnx2x_init_block(bp, port ? CSEM_PORT1_START : CSEM_PORT0_START,
-                            port ? CSEM_PORT1_END : CSEM_PORT0_END);
-       bnx2x_init_block(bp, port ? XSEM_PORT1_START : XSEM_PORT0_START,
-                            port ? XSEM_PORT1_END : XSEM_PORT0_END);
+       bnx2x_init_block(bp, XSDM_BLOCK, init_stage);
+
+       bnx2x_init_block(bp, TSEM_BLOCK, init_stage);
+       bnx2x_init_block(bp, USEM_BLOCK, init_stage);
+       bnx2x_init_block(bp, CSEM_BLOCK, init_stage);
+       bnx2x_init_block(bp, XSEM_BLOCK, init_stage);
+
        /* Port UPB comes here */
+       bnx2x_init_block(bp, UPB_BLOCK, init_stage);
        /* Port XPB comes here */
+       bnx2x_init_block(bp, XPB_BLOCK, init_stage);
 
-       bnx2x_init_block(bp, port ? PBF_PORT1_START : PBF_PORT0_START,
-                            port ? PBF_PORT1_END : PBF_PORT0_END);
+       bnx2x_init_block(bp, PBF_BLOCK, init_stage);
 
        /* configure PBF to work without PAUSE mtu 9000 */
        REG_WR(bp, PBF_REG_P0_PAUSE_ENABLE + port*4, 0);
@@ -5781,18 +5964,17 @@ static int bnx2x_init_port(struct bnx2x *bp)
        /* Port SRCH comes here */
 #endif
        /* Port CDU comes here */
+       bnx2x_init_block(bp, CDU_BLOCK, init_stage);
        /* Port CFC comes here */
+       bnx2x_init_block(bp, CFC_BLOCK, init_stage);
 
        if (CHIP_IS_E1(bp)) {
                REG_WR(bp, HC_REG_LEADING_EDGE_0 + port*8, 0);
                REG_WR(bp, HC_REG_TRAILING_EDGE_0 + port*8, 0);
        }
-       bnx2x_init_block(bp, port ? HC_PORT1_START : HC_PORT0_START,
-                            port ? HC_PORT1_END : HC_PORT0_END);
+       bnx2x_init_block(bp, HC_BLOCK, init_stage);
 
-       bnx2x_init_block(bp, port ? MISC_AEU_PORT1_START :
-                                   MISC_AEU_PORT0_START,
-                            port ? MISC_AEU_PORT1_END : MISC_AEU_PORT0_END);
+       bnx2x_init_block(bp, MISC_AEU_BLOCK, init_stage);
        /* 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
@@ -5801,12 +5983,17 @@ static int bnx2x_init_port(struct bnx2x *bp)
               (IS_E1HMF(bp) ? 0xF7 : 0x7));
 
        /* Port PXPCS comes here */
+       bnx2x_init_block(bp, PXPCS_BLOCK, init_stage);
        /* Port EMAC0 comes here */
+       bnx2x_init_block(bp, EMAC0_BLOCK, init_stage);
        /* Port EMAC1 comes here */
+       bnx2x_init_block(bp, EMAC1_BLOCK, init_stage);
        /* Port DBU comes here */
+       bnx2x_init_block(bp, DBU_BLOCK, init_stage);
        /* Port DBG comes here */
-       bnx2x_init_block(bp, port ? NIG_PORT1_START : NIG_PORT0_START,
-                            port ? NIG_PORT1_END : NIG_PORT0_END);
+       bnx2x_init_block(bp, DBG_BLOCK, init_stage);
+
+       bnx2x_init_block(bp, NIG_BLOCK, init_stage);
 
        REG_WR(bp, NIG_REG_XGXS_SERDES0_MODE_SEL + port*4, 1);
 
@@ -5826,9 +6013,42 @@ static int bnx2x_init_port(struct bnx2x *bp)
        }
 
        /* Port MCP comes here */
+       bnx2x_init_block(bp, MCP_BLOCK, init_stage);
        /* Port DMAE comes here */
+       bnx2x_init_block(bp, DMAE_BLOCK, init_stage);
 
        switch (XGXS_EXT_PHY_TYPE(bp->link_params.ext_phy_config)) {
+       case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8726:
+               {
+               u32 swap_val, swap_override, aeu_gpio_mask, offset;
+
+               bnx2x_set_gpio(bp, MISC_REGISTERS_GPIO_3,
+                              MISC_REGISTERS_GPIO_INPUT_HI_Z, port);
+
+               /* The GPIO should be swapped if the swap register is
+                  set and active */
+               swap_val = REG_RD(bp, NIG_REG_PORT_SWAP);
+               swap_override = REG_RD(bp, NIG_REG_STRAP_OVERRIDE);
+
+               /* Select function upon port-swap configuration */
+               if (port == 0) {
+                       offset = MISC_REG_AEU_ENABLE1_FUNC_0_OUT_0;
+                       aeu_gpio_mask = (swap_val && swap_override) ?
+                               AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_1 :
+                               AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_0;
+               } else {
+                       offset = MISC_REG_AEU_ENABLE1_FUNC_1_OUT_0;
+                       aeu_gpio_mask = (swap_val && swap_override) ?
+                               AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_0 :
+                               AEU_INPUTS_ATTN_BITS_GPIO3_FUNCTION_1;
+               }
+               val = REG_RD(bp, offset);
+               /* add GPIO3 to group */
+               val |= aeu_gpio_mask;
+               REG_WR(bp, offset, val);
+               }
+               break;
+
        case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_SFX7101:
                /* add SPIO 5 to group 0 */
                val = REG_RD(bp, MISC_REG_AEU_ENABLE1_FUNC_0_OUT_0);
@@ -5900,7 +6120,7 @@ static int bnx2x_init_func(struct bnx2x *bp)
        if (CHIP_IS_E1H(bp)) {
                for (i = 0; i < 9; i++)
                        bnx2x_init_block(bp,
-                                        cm_start[func][i], cm_end[func][i]);
+                                        cm_blocks[i], FUNC0_STAGE + func);
 
                REG_WR(bp, NIG_REG_LLH0_FUNC_EN + port*8, 1);
                REG_WR(bp, NIG_REG_LLH0_FUNC_VLAN_ID + port*8, bp->e1hov);
@@ -5913,7 +6133,7 @@ static int bnx2x_init_func(struct bnx2x *bp)
                REG_WR(bp, HC_REG_LEADING_EDGE_0 + port*8, 0);
                REG_WR(bp, HC_REG_TRAILING_EDGE_0 + port*8, 0);
        }
-       bnx2x_init_block(bp, hc_limits[func][0], hc_limits[func][1]);
+       bnx2x_init_block(bp, HC_BLOCK, FUNC0_STAGE + func);
 
        /* Reset PCIE errors for debug */
        REG_WR(bp, 0x2114, 0xffffffff);
@@ -6255,8 +6475,7 @@ static void bnx2x_free_rx_skbs(struct bnx2x *bp)
 
                        pci_unmap_single(bp->pdev,
                                         pci_unmap_addr(rx_buf, mapping),
-                                        bp->rx_buf_size,
-                                        PCI_DMA_FROMDEVICE);
+                                        bp->rx_buf_size, PCI_DMA_FROMDEVICE);
 
                        rx_buf->skb = NULL;
                        dev_kfree_skb(skb);
@@ -6425,7 +6644,12 @@ static void bnx2x_napi_disable(struct bnx2x *bp)
 
 static void bnx2x_netif_start(struct bnx2x *bp)
 {
-       if (atomic_dec_and_test(&bp->intr_sem)) {
+       int intr_sem;
+
+       intr_sem = atomic_dec_and_test(&bp->intr_sem);
+       smp_wmb(); /* Ensure that bp->intr_sem update is SMP-safe */
+
+       if (intr_sem) {
                if (netif_running(bp->dev)) {
                        bnx2x_napi_enable(bp);
                        bnx2x_int_enable(bp);
@@ -6439,10 +6663,8 @@ static void bnx2x_netif_stop(struct bnx2x *bp, int disable_hw)
 {
        bnx2x_int_disable_sync(bp, disable_hw);
        bnx2x_napi_disable(bp);
-       if (netif_running(bp->dev)) {
-               netif_tx_disable(bp->dev);
-               bp->dev->trans_start = jiffies; /* prevent tx timeout */
-       }
+       netif_tx_disable(bp->dev);
+       bp->dev->trans_start = jiffies; /* prevent tx timeout */
 }
 
 /*
@@ -6460,7 +6682,7 @@ static void bnx2x_set_mac_addr_e1(struct bnx2x *bp, int set)
         */
        config->hdr.length = 2;
        config->hdr.offset = port ? 32 : 0;
-       config->hdr.client_id = BP_CL_ID(bp);
+       config->hdr.client_id = bp->fp->cl_id;
        config->hdr.reserved1 = 0;
 
        /* primary MAC */
@@ -6485,9 +6707,9 @@ static void bnx2x_set_mac_addr_e1(struct bnx2x *bp, int set)
           config->config_table[0].cam_entry.lsb_mac_addr);
 
        /* broadcast */
-       config->config_table[1].cam_entry.msb_mac_addr = 0xffff;
-       config->config_table[1].cam_entry.middle_mac_addr = 0xffff;
-       config->config_table[1].cam_entry.lsb_mac_addr = 0xffff;
+       config->config_table[1].cam_entry.msb_mac_addr = cpu_to_le16(0xffff);
+       config->config_table[1].cam_entry.middle_mac_addr = cpu_to_le16(0xffff);
+       config->config_table[1].cam_entry.lsb_mac_addr = cpu_to_le16(0xffff);
        config->config_table[1].cam_entry.flags = cpu_to_le16(port);
        if (set)
                config->config_table[1].target_table_entry.flags =
@@ -6518,7 +6740,7 @@ static void bnx2x_set_mac_addr_e1h(struct bnx2x *bp, int set)
         */
        config->hdr.length = 1;
        config->hdr.offset = BP_FUNC(bp);
-       config->hdr.client_id = BP_CL_ID(bp);
+       config->hdr.client_id = bp->fp->cl_id;
        config->hdr.reserved1 = 0;
 
        /* primary MAC */
@@ -6552,7 +6774,7 @@ static int bnx2x_wait_ramrod(struct bnx2x *bp, int state, int idx,
                             int *state_p, int poll)
 {
        /* can take a while if any port is running */
-       int cnt = 500;
+       int cnt = 5000;
 
        DP(NETIF_MSG_IFUP, "%s for state to become %x on IDX [%d]\n",
           poll ? "polling" : "waiting", state, idx);
@@ -6570,8 +6792,12 @@ static int bnx2x_wait_ramrod(struct bnx2x *bp, int state, int idx,
                }
 
                mb(); /* state is changed by bnx2x_sp_event() */
-               if (*state_p == state)
+               if (*state_p == state) {
+#ifdef BNX2X_STOP_ON_ERROR
+                       DP(NETIF_MSG_IFUP, "exit  (cnt %d)\n", 5000 - cnt);
+#endif
                        return 0;
+               }
 
                msleep(1);
        }
@@ -6751,11 +6977,11 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
        } else {
                int port = BP_PORT(bp);
 
-               DP(NETIF_MSG_IFUP, "NO MCP load counts before us %d, %d, %d\n",
+               DP(NETIF_MSG_IFUP, "NO MCP - load counts      %d, %d, %d\n",
                   load_count[0], load_count[1], load_count[2]);
                load_count[0]++;
                load_count[1 + port]++;
-               DP(NETIF_MSG_IFUP, "NO MCP new load counts       %d, %d, %d\n",
+               DP(NETIF_MSG_IFUP, "NO MCP - new load counts  %d, %d, %d\n",
                   load_count[0], load_count[1], load_count[2]);
                if (load_count[0] == 1)
                        load_code = FW_MSG_CODE_DRV_LOAD_COMMON;
@@ -6802,7 +7028,7 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 
        if (CHIP_IS_E1H(bp))
                if (bp->mf_config & FUNC_MF_CFG_FUNC_DISABLED) {
-                       BNX2X_ERR("!!!  mf_cfg function disabled\n");
+                       DP(NETIF_MSG_IFUP, "mf_cfg function disabled\n");
                        bp->state = BNX2X_STATE_DISABLED;
                }
 
@@ -6819,7 +7045,7 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                bnx2x_set_mac_addr_e1h(bp, 1);
 
        if (bp->port.pmf)
-               bnx2x_initial_phy_init(bp);
+               bnx2x_initial_phy_init(bp, load_mode);
 
        /* Start fast path */
        switch (load_mode) {
@@ -6875,8 +7101,6 @@ load_error1:
                netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
 
-       /* TBD we really need to reset the chip
-          if we want to recover from this */
        return rc;
 }
 
@@ -6906,7 +7130,7 @@ static int bnx2x_stop_multi(struct bnx2x *bp, int index)
 
 static int bnx2x_stop_leading(struct bnx2x *bp)
 {
-       u16 dsb_sp_prod_idx;
+       __le16 dsb_sp_prod_idx;
        /* if the other port is handling traffic,
           this can take a lot of time */
        int cnt = 500;
@@ -6916,7 +7140,7 @@ static int bnx2x_stop_leading(struct bnx2x *bp)
 
        /* Send HALT ramrod */
        bp->fp[0].state = BNX2X_FP_STATE_HALTING;
-       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_HALT, 0, 0, BP_CL_ID(bp), 0);
+       bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_HALT, 0, 0, bp->fp->cl_id, 0);
 
        /* Wait for completion */
        rc = bnx2x_wait_ramrod(bp, BNX2X_FP_STATE_HALTED, 0,
@@ -6940,9 +7164,8 @@ static int bnx2x_stop_leading(struct bnx2x *bp)
                           *bp->dsb_sp_prod, dsb_sp_prod_idx);
 #ifdef BNX2X_STOP_ON_ERROR
                        bnx2x_panic();
-#else
-                       rc = -EBUSY;
 #endif
+                       rc = -EBUSY;
                        break;
                }
                cnt--;
@@ -7051,10 +7274,9 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                cnt = 1000;
-               smp_rmb();
                while (bnx2x_has_tx_work_unload(fp)) {
 
-                       bnx2x_tx_int(fp, 1000);
+                       bnx2x_tx_int(fp);
                        if (!cnt) {
                                BNX2X_ERR("timeout waiting for queue[%d]\n",
                                          i);
@@ -7067,7 +7289,6 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
                        }
                        cnt--;
                        msleep(1);
-                       smp_rmb();
                }
        }
        /* Give HW time to discard old tx messages */
@@ -7087,7 +7308,7 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
                        config->hdr.offset = BNX2X_MAX_EMUL_MULTI*(1 + port);
                else
                        config->hdr.offset = BNX2X_MAX_MULTICAST*(1 + port);
-               config->hdr.client_id = BP_CL_ID(bp);
+               config->hdr.client_id = bp->fp->cl_id;
                config->hdr.reserved1 = 0;
 
                bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_SET_MAC, 0,
@@ -7151,11 +7372,11 @@ unload_error:
        if (!BP_NOMCP(bp))
                reset_code = bnx2x_fw_command(bp, reset_code);
        else {
-               DP(NETIF_MSG_IFDOWN, "NO MCP load counts      %d, %d, %d\n",
+               DP(NETIF_MSG_IFDOWN, "NO MCP load counts      %d, %d, %d\n",
                   load_count[0], load_count[1], load_count[2]);
                load_count[0]--;
                load_count[1 + port]--;
-               DP(NETIF_MSG_IFDOWN, "NO MCP new load counts  %d, %d, %d\n",
+               DP(NETIF_MSG_IFDOWN, "NO MCP new load counts  %d, %d, %d\n",
                   load_count[0], load_count[1], load_count[2]);
                if (load_count[0] == 0)
                        reset_code = FW_MSG_CODE_DRV_UNLOAD_COMMON;
@@ -7175,6 +7396,7 @@ unload_error:
        /* Report UNLOAD_DONE to MCP */
        if (!BP_NOMCP(bp))
                bnx2x_fw_command(bp, DRV_MSG_CODE_UNLOAD_DONE);
+
        bp->port.pmf = 0;
 
        /* Free SKBs, SGEs, TPA pool and driver internals */
@@ -7436,6 +7658,15 @@ static void __devinit bnx2x_get_common_hwinfo(struct bnx2x *bp)
                                        SHARED_HW_CFG_LED_MODE_MASK) >>
                                       SHARED_HW_CFG_LED_MODE_SHIFT);
 
+       bp->link_params.feature_config_flags = 0;
+       val = SHMEM_RD(bp, dev_info.shared_feature_config.config);
+       if (val & SHARED_FEAT_CFG_OVERRIDE_PREEMPHASIS_CFG_ENABLED)
+               bp->link_params.feature_config_flags |=
+                               FEATURE_CONFIG_OVERRIDE_PREEMPHASIS_ENABLED;
+       else
+               bp->link_params.feature_config_flags &=
+                               ~FEATURE_CONFIG_OVERRIDE_PREEMPHASIS_ENABLED;
+
        val = SHMEM_RD(bp, dev_info.bc_rev) >> 8;
        bp->common.bc_ver = val;
        BNX2X_DEV_INFO("bc_ver %X\n", val);
@@ -7454,7 +7685,7 @@ static void __devinit bnx2x_get_common_hwinfo(struct bnx2x *bp)
                bp->flags |= NO_WOL_FLAG;
        }
        BNX2X_DEV_INFO("%sWoL capable\n",
-                      (bp->flags & NO_WOL_FLAG) ? "Not " : "");
+                      (bp->flags & NO_WOL_FLAG) ? "not " : "");
 
        val = SHMEM_RD(bp, dev_info.shared_hw_config.part_num);
        val2 = SHMEM_RD(bp, dev_info.shared_hw_config.part_num[4]);
@@ -7547,48 +7778,60 @@ static void __devinit bnx2x_link_settings_supported(struct bnx2x *bp,
                                               SUPPORTED_Asym_Pause);
                        break;
 
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8705:
-                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8705)\n",
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8072)\n",
                                       ext_phy_type);
 
                        bp->port.supported |= (SUPPORTED_10000baseT_Full |
+                                              SUPPORTED_1000baseT_Full |
                                               SUPPORTED_FIBRE |
+                                              SUPPORTED_Autoneg |
                                               SUPPORTED_Pause |
                                               SUPPORTED_Asym_Pause);
                        break;
 
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8706:
-                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8706)\n",
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8073)\n",
                                       ext_phy_type);
 
                        bp->port.supported |= (SUPPORTED_10000baseT_Full |
+                                              SUPPORTED_2500baseX_Full |
                                               SUPPORTED_1000baseT_Full |
                                               SUPPORTED_FIBRE |
+                                              SUPPORTED_Autoneg |
                                               SUPPORTED_Pause |
                                               SUPPORTED_Asym_Pause);
                        break;
 
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072:
-                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8072)\n",
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8705:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8705)\n",
                                       ext_phy_type);
 
                        bp->port.supported |= (SUPPORTED_10000baseT_Full |
-                                              SUPPORTED_1000baseT_Full |
                                               SUPPORTED_FIBRE |
-                                              SUPPORTED_Autoneg |
                                               SUPPORTED_Pause |
                                               SUPPORTED_Asym_Pause);
                        break;
 
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073:
-                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8073)\n",
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8706:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8706)\n",
                                       ext_phy_type);
 
                        bp->port.supported |= (SUPPORTED_10000baseT_Full |
-                                              SUPPORTED_2500baseX_Full |
                                               SUPPORTED_1000baseT_Full |
                                               SUPPORTED_FIBRE |
+                                              SUPPORTED_Pause |
+                                              SUPPORTED_Asym_Pause);
+                       break;
+
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8726:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (8726)\n",
+                                      ext_phy_type);
+
+                       bp->port.supported |= (SUPPORTED_10000baseT_Full |
+                                              SUPPORTED_1000baseT_Full |
                                               SUPPORTED_Autoneg |
+                                              SUPPORTED_FIBRE |
                                               SUPPORTED_Pause |
                                               SUPPORTED_Asym_Pause);
                        break;
@@ -7604,6 +7847,22 @@ static void __devinit bnx2x_link_settings_supported(struct bnx2x *bp,
                                               SUPPORTED_Asym_Pause);
                        break;
 
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8481:
+                       BNX2X_DEV_INFO("ext_phy_type 0x%x (BCM8481)\n",
+                                      ext_phy_type);
+
+                       bp->port.supported |= (SUPPORTED_10baseT_Half |
+                                              SUPPORTED_10baseT_Full |
+                                              SUPPORTED_100baseT_Half |
+                                              SUPPORTED_100baseT_Full |
+                                              SUPPORTED_1000baseT_Full |
+                                              SUPPORTED_10000baseT_Full |
+                                              SUPPORTED_TP |
+                                              SUPPORTED_Autoneg |
+                                              SUPPORTED_Pause |
+                                              SUPPORTED_Asym_Pause);
+                       break;
+
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_FAILURE:
                        BNX2X_ERR("XGXS PHY Failure detected 0x%x\n",
                                  bp->link_params.ext_phy_config);
@@ -7829,12 +8088,12 @@ static void __devinit bnx2x_get_port_hwinfo(struct bnx2x *bp)
 {
        int port = BP_PORT(bp);
        u32 val, val2;
+       u32 config;
+       u16 i;
 
        bp->link_params.bp = bp;
        bp->link_params.port = port;
 
-       bp->link_params.serdes_config =
-               SHMEM_RD(bp, dev_info.port_hw_config[port].serdes_config);
        bp->link_params.lane_config =
                SHMEM_RD(bp, dev_info.port_hw_config[port].lane_config);
        bp->link_params.ext_phy_config =
@@ -7847,10 +8106,35 @@ static void __devinit bnx2x_get_port_hwinfo(struct bnx2x *bp)
        bp->port.link_config =
                SHMEM_RD(bp, dev_info.port_feature_config[port].link_config);
 
-       BNX2X_DEV_INFO("serdes_config 0x%08x  lane_config 0x%08x\n"
-            KERN_INFO "  ext_phy_config 0x%08x  speed_cap_mask 0x%08x"
-                      "  link_config 0x%08x\n",
-                      bp->link_params.serdes_config,
+       /* Get the 4 lanes xgxs config rx and tx */
+       for (i = 0; i < 2; i++) {
+               val = SHMEM_RD(bp,
+                          dev_info.port_hw_config[port].xgxs_config_rx[i<<1]);
+               bp->link_params.xgxs_config_rx[i << 1] = ((val>>16) & 0xffff);
+               bp->link_params.xgxs_config_rx[(i << 1) + 1] = (val & 0xffff);
+
+               val = SHMEM_RD(bp,
+                          dev_info.port_hw_config[port].xgxs_config_tx[i<<1]);
+               bp->link_params.xgxs_config_tx[i << 1] = ((val>>16) & 0xffff);
+               bp->link_params.xgxs_config_tx[(i << 1) + 1] = (val & 0xffff);
+       }
+
+       config = SHMEM_RD(bp, dev_info.port_feature_config[port].config);
+       if (config & PORT_FEAT_CFG_OPT_MDL_ENFRCMNT_ENABLED)
+               bp->link_params.feature_config_flags |=
+                               FEATURE_CONFIG_MODULE_ENFORCMENT_ENABLED;
+       else
+               bp->link_params.feature_config_flags &=
+                               ~FEATURE_CONFIG_MODULE_ENFORCMENT_ENABLED;
+
+       /* If the device is capable of WoL, set the default state according
+        * to the HW
+        */
+       bp->wol = (!(bp->flags & NO_WOL_FLAG) &&
+                  (config & PORT_FEATURE_WOL_ENABLED));
+
+       BNX2X_DEV_INFO("lane_config 0x%08x  ext_phy_config 0x%08x"
+                      "  speed_cap_mask 0x%08x  link_config 0x%08x\n",
                       bp->link_params.lane_config,
                       bp->link_params.ext_phy_config,
                       bp->link_params.speed_cap_mask, bp->port.link_config);
@@ -7897,7 +8181,7 @@ static int __devinit bnx2x_get_hwinfo(struct bnx2x *bp)
                                       "(0x%04x)\n",
                                       func, bp->e1hov, bp->e1hov);
                } else {
-                       BNX2X_DEV_INFO("Single function mode\n");
+                       BNX2X_DEV_INFO("single function mode\n");
                        if (BP_E1HVN(bp)) {
                                BNX2X_ERR("!!!  No valid E1HOV for func %d,"
                                          "  aborting\n", func);
@@ -7952,6 +8236,7 @@ static int __devinit bnx2x_init_bp(struct bnx2x *bp)
 
        /* Disable interrupt handling until HW is initialized */
        atomic_set(&bp->intr_sem, 1);
+       smp_wmb(); /* Ensure that bp->intr_sem update is SMP-safe */
 
        mutex_init(&bp->port.phy_mutex);
 
@@ -7990,6 +8275,7 @@ static int __devinit bnx2x_init_bp(struct bnx2x *bp)
                bp->dev->features |= NETIF_F_LRO;
        }
 
+       bp->mrrs = mrrs;
 
        bp->tx_ring_size = MAX_TX_AVAIL;
        bp->rx_ring_size = MAX_RX_AVAIL;
@@ -8045,14 +8331,16 @@ static int bnx2x_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
 
                switch (ext_phy_type) {
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_DIRECT:
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8705:
-               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8706:
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8072:
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8073:
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8705:
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8706:
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8726:
                        cmd->port = PORT_FIBRE;
                        break;
 
                case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_SFX7101:
+               case PORT_HW_CFG_XGXS_EXT_PHY_TYPE_BCM8481:
                        cmd->port = PORT_TP;
                        break;
 
@@ -8274,6 +8562,84 @@ static void bnx2x_get_drvinfo(struct net_device *dev,
        info->regdump_len = 0;
 }
 
+#define IS_E1_ONLINE(info)     (((info) & RI_E1_ONLINE) == RI_E1_ONLINE)
+#define IS_E1H_ONLINE(info)    (((info) & RI_E1H_ONLINE) == RI_E1H_ONLINE)
+
+static int bnx2x_get_regs_len(struct net_device *dev)
+{
+       static u32 regdump_len;
+       struct bnx2x *bp = netdev_priv(dev);
+       int i;
+
+       if (regdump_len)
+               return regdump_len;
+
+       if (CHIP_IS_E1(bp)) {
+               for (i = 0; i < REGS_COUNT; i++)
+                       if (IS_E1_ONLINE(reg_addrs[i].info))
+                               regdump_len += reg_addrs[i].size;
+
+               for (i = 0; i < WREGS_COUNT_E1; i++)
+                       if (IS_E1_ONLINE(wreg_addrs_e1[i].info))
+                               regdump_len += wreg_addrs_e1[i].size *
+                                       (1 + wreg_addrs_e1[i].read_regs_count);
+
+       } else { /* E1H */
+               for (i = 0; i < REGS_COUNT; i++)
+                       if (IS_E1H_ONLINE(reg_addrs[i].info))
+                               regdump_len += reg_addrs[i].size;
+
+               for (i = 0; i < WREGS_COUNT_E1H; i++)
+                       if (IS_E1H_ONLINE(wreg_addrs_e1h[i].info))
+                               regdump_len += wreg_addrs_e1h[i].size *
+                                       (1 + wreg_addrs_e1h[i].read_regs_count);
+       }
+       regdump_len *= 4;
+       regdump_len += sizeof(struct dump_hdr);
+
+       return regdump_len;
+}
+
+static void bnx2x_get_regs(struct net_device *dev,
+                          struct ethtool_regs *regs, void *_p)
+{
+       u32 *p = _p, i, j;
+       struct bnx2x *bp = netdev_priv(dev);
+       struct dump_hdr dump_hdr = {0};
+
+       regs->version = 0;
+       memset(p, 0, regs->len);
+
+       if (!netif_running(bp->dev))
+               return;
+
+       dump_hdr.hdr_size = (sizeof(struct dump_hdr) / 4) - 1;
+       dump_hdr.dump_sign = dump_sign_all;
+       dump_hdr.xstorm_waitp = REG_RD(bp, XSTORM_WAITP_ADDR);
+       dump_hdr.tstorm_waitp = REG_RD(bp, TSTORM_WAITP_ADDR);
+       dump_hdr.ustorm_waitp = REG_RD(bp, USTORM_WAITP_ADDR);
+       dump_hdr.cstorm_waitp = REG_RD(bp, CSTORM_WAITP_ADDR);
+       dump_hdr.info = CHIP_IS_E1(bp) ? RI_E1_ONLINE : RI_E1H_ONLINE;
+
+       memcpy(p, &dump_hdr, sizeof(struct dump_hdr));
+       p += dump_hdr.hdr_size + 1;
+
+       if (CHIP_IS_E1(bp)) {
+               for (i = 0; i < REGS_COUNT; i++)
+                       if (IS_E1_ONLINE(reg_addrs[i].info))
+                               for (j = 0; j < reg_addrs[i].size; j++)
+                                       *p++ = REG_RD(bp,
+                                                     reg_addrs[i].addr + j*4);
+
+       } else { /* E1H */
+               for (i = 0; i < REGS_COUNT; i++)
+                       if (IS_E1H_ONLINE(reg_addrs[i].info))
+                               for (j = 0; j < reg_addrs[i].size; j++)
+                                       *p++ = REG_RD(bp,
+                                                     reg_addrs[i].addr + j*4);
+       }
+}
+
 static void bnx2x_get_wol(struct net_device *dev, struct ethtool_wolinfo *wol)
 {
        struct bnx2x *bp = netdev_priv(dev);
@@ -8339,6 +8705,14 @@ static int bnx2x_nway_reset(struct net_device *dev)
        return 0;
 }
 
+static u32
+bnx2x_get_link(struct net_device *dev)
+{
+       struct bnx2x *bp = netdev_priv(dev);
+
+       return bp->link_vars.link_up;
+}
+
 static int bnx2x_get_eeprom_len(struct net_device *dev)
 {
        struct bnx2x *bp = netdev_priv(dev);
@@ -8432,7 +8806,7 @@ static void bnx2x_disable_nvram_access(struct bnx2x *bp)
                        MCPR_NVM_ACCESS_ENABLE_WR_EN)));
 }
 
-static int bnx2x_nvram_read_dword(struct bnx2x *bp, u32 offset, u32 *ret_val,
+static int bnx2x_nvram_read_dword(struct bnx2x *bp, u32 offset, __be32 *ret_val,
                                  u32 cmd_flags)
 {
        int count, i, rc;
@@ -8468,8 +8842,7 @@ static int bnx2x_nvram_read_dword(struct bnx2x *bp, u32 offset, u32 *ret_val,
                        /* we read nvram data in cpu order
                         * but ethtool sees it as an array of bytes
                         * converting to big-endian will do the work */
-                       val = cpu_to_be32(val);
-                       *ret_val = val;
+                       *ret_val = cpu_to_be32(val);
                        rc = 0;
                        break;
                }
@@ -8483,7 +8856,7 @@ static int bnx2x_nvram_read(struct bnx2x *bp, u32 offset, u8 *ret_buf,
 {
        int rc;
        u32 cmd_flags;
-       u32 val;
+       __be32 val;
 
        if ((offset & 0x03) || (buf_size & 0x03) || (buf_size == 0)) {
                DP(BNX2X_MSG_NVM,
@@ -8602,7 +8975,7 @@ static int bnx2x_nvram_write1(struct bnx2x *bp, u32 offset, u8 *data_buf,
        int rc;
        u32 cmd_flags;
        u32 align_offset;
-       u32 val;
+       __be32 val;
 
        if (offset + buf_size > bp->common.flash_size) {
                DP(BNX2X_MSG_NVM, "Invalid parameter: offset (0x%x) +"
@@ -8731,7 +9104,7 @@ static int bnx2x_set_eeprom(struct net_device *dev,
                        if ((bp->state == BNX2X_STATE_OPEN) ||
                            (bp->state == BNX2X_STATE_DISABLED)) {
                                rc |= bnx2x_link_reset(&bp->link_params,
-                                                      &bp->link_vars);
+                                                      &bp->link_vars, 1);
                                rc |= bnx2x_phy_init(&bp->link_params,
                                                     &bp->link_vars);
                        }
@@ -8764,12 +9137,12 @@ static int bnx2x_set_coalesce(struct net_device *dev,
        struct bnx2x *bp = netdev_priv(dev);
 
        bp->rx_ticks = (u16) coal->rx_coalesce_usecs;
-       if (bp->rx_ticks > 3000)
-               bp->rx_ticks = 3000;
+       if (bp->rx_ticks > BNX2X_MAX_COALESCE_TOUT)
+               bp->rx_ticks = BNX2X_MAX_COALESCE_TOUT;
 
        bp->tx_ticks = (u16) coal->tx_coalesce_usecs;
-       if (bp->tx_ticks > 0x3000)
-               bp->tx_ticks = 0x3000;
+       if (bp->tx_ticks > BNX2X_MAX_COALESCE_TOUT)
+               bp->tx_ticks = BNX2X_MAX_COALESCE_TOUT;
 
        if (netif_running(dev))
                bnx2x_update_coalesce(bp);
@@ -8821,7 +9194,8 @@ static void bnx2x_get_pauseparam(struct net_device *dev,
 {
        struct bnx2x *bp = netdev_priv(dev);
 
-       epause->autoneg = (bp->link_params.req_flow_ctrl == BNX2X_FLOW_CTRL_AUTO) &&
+       epause->autoneg = (bp->link_params.req_flow_ctrl ==
+                          BNX2X_FLOW_CTRL_AUTO) &&
                          (bp->link_params.req_line_speed == SPEED_AUTO_NEG);
 
        epause->rx_pause = ((bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_RX) ==
@@ -8936,9 +9310,17 @@ static int bnx2x_set_tso(struct net_device *dev, u32 data)
        if (data) {
                dev->features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
                dev->features |= NETIF_F_TSO6;
+#ifdef BCM_VLAN
+               dev->vlan_features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
+               dev->vlan_features |= NETIF_F_TSO6;
+#endif
        } else {
                dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO_ECN);
                dev->features &= ~NETIF_F_TSO6;
+#ifdef BCM_VLAN
+               dev->vlan_features &= ~(NETIF_F_TSO | NETIF_F_TSO_ECN);
+               dev->vlan_features &= ~NETIF_F_TSO6;
+#endif
        }
 
        return 0;
@@ -9141,23 +9523,23 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode, u8 link_up)
        u16 len;
        int rc = -ENODEV;
 
-       if (loopback_mode == BNX2X_MAC_LOOPBACK) {
+       /* check the loopback mode */
+       switch (loopback_mode) {
+       case BNX2X_PHY_LOOPBACK:
+               if (bp->link_params.loopback_mode != LOOPBACK_XGXS_10)
+                       return -EINVAL;
+               break;
+       case BNX2X_MAC_LOOPBACK:
                bp->link_params.loopback_mode = LOOPBACK_BMAC;
                bnx2x_phy_init(&bp->link_params, &bp->link_vars);
-
-       } else if (loopback_mode == BNX2X_PHY_LOOPBACK) {
-               u16 cnt = 1000;
-               bp->link_params.loopback_mode = LOOPBACK_XGXS_10;
-               bnx2x_phy_init(&bp->link_params, &bp->link_vars);
-               /* wait until link state is restored */
-               if (link_up)
-                       while (cnt-- && bnx2x_test_link(&bp->link_params,
-                                                       &bp->link_vars))
-                               msleep(10);
-       } else
+               break;
+       default:
                return -EINVAL;
+       }
 
-       pkt_size = 1514;
+       /* prepare the loopback packet */
+       pkt_size = (((bp->dev->mtu < ETH_MAX_PACKET_SIZE) ?
+                    bp->dev->mtu : ETH_MAX_PACKET_SIZE) + ETH_HLEN);
        skb = netdev_alloc_skb(bp->dev, bp->rx_buf_size);
        if (!skb) {
                rc = -ENOMEM;
@@ -9169,6 +9551,7 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode, u8 link_up)
        for (i = ETH_HLEN; i < pkt_size; i++)
                packet[i] = (unsigned char) (i & 0xff);
 
+       /* send the loopback packet */
        num_pkts = 0;
        tx_start_idx = le16_to_cpu(*fp->tx_cons_sb);
        rx_start_idx = le16_to_cpu(*fp->rx_cons_sb);
@@ -9193,12 +9576,10 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode, u8 link_up)
 
        wmb();
 
-       fp->hw_tx_prods->bds_prod =
-               cpu_to_le16(le16_to_cpu(fp->hw_tx_prods->bds_prod) + 1);
+       le16_add_cpu(&fp->hw_tx_prods->bds_prod, 1);
        mb(); /* FW restriction: must not reorder writing nbd and packets */
-       fp->hw_tx_prods->packets_prod =
-               cpu_to_le32(le32_to_cpu(fp->hw_tx_prods->packets_prod) + 1);
-       DOORBELL(bp, FP_IDX(fp), 0);
+       le32_add_cpu(&fp->hw_tx_prods->packets_prod, 1);
+       DOORBELL(bp, fp->index, 0);
 
        mmiowb();
 
@@ -9253,7 +9634,7 @@ test_loopback_exit:
 
 static int bnx2x_test_loopback(struct bnx2x *bp, u8 link_up)
 {
-       int rc = 0;
+       int rc = 0, res;
 
        if (!netif_running(bp->dev))
                return BNX2X_LOOPBACK_FAILED;
@@ -9261,14 +9642,16 @@ static int bnx2x_test_loopback(struct bnx2x *bp, u8 link_up)
        bnx2x_netif_stop(bp, 1);
        bnx2x_acquire_phy_lock(bp);
 
-       if (bnx2x_run_loopback(bp, BNX2X_MAC_LOOPBACK, link_up)) {
-               DP(NETIF_MSG_PROBE, "MAC loopback failed\n");
-               rc |= BNX2X_MAC_LOOPBACK_FAILED;
+       res = bnx2x_run_loopback(bp, BNX2X_PHY_LOOPBACK, link_up);
+       if (res) {
+               DP(NETIF_MSG_PROBE, "  PHY loopback failed  (res %d)\n", res);
+               rc |= BNX2X_PHY_LOOPBACK_FAILED;
        }
 
-       if (bnx2x_run_loopback(bp, BNX2X_PHY_LOOPBACK, link_up)) {
-               DP(NETIF_MSG_PROBE, "PHY loopback failed\n");
-               rc |= BNX2X_PHY_LOOPBACK_FAILED;
+       res = bnx2x_run_loopback(bp, BNX2X_MAC_LOOPBACK, link_up);
+       if (res) {
+               DP(NETIF_MSG_PROBE, "  MAC loopback failed  (res %d)\n", res);
+               rc |= BNX2X_MAC_LOOPBACK_FAILED;
        }
 
        bnx2x_release_phy_lock(bp);
@@ -9295,14 +9678,14 @@ static int bnx2x_test_nvram(struct bnx2x *bp)
                { 0x778,  0x70 },
                {     0,     0 }
        };
-       u32 buf[0x350 / 4];
+       __be32 buf[0x350 / 4];
        u8 *data = (u8 *)buf;
        int i, rc;
        u32 magic, csum;
 
        rc = bnx2x_nvram_read(bp, 0, data, 4);
        if (rc) {
-               DP(NETIF_MSG_PROBE, "magic value read (rc -%d)\n", -rc);
+               DP(NETIF_MSG_PROBE, "magic value read (rc %d)\n", rc);
                goto test_nvram_exit;
        }
 
@@ -9319,7 +9702,7 @@ static int bnx2x_test_nvram(struct bnx2x *bp)
                                      nvram_tbl[i].size);
                if (rc) {
                        DP(NETIF_MSG_PROBE,
-                          "nvram_tbl[%d] read data (rc -%d)\n", i, -rc);
+                          "nvram_tbl[%d] read data (rc %d)\n", i, rc);
                        goto test_nvram_exit;
                }
 
@@ -9349,7 +9732,7 @@ static int bnx2x_test_intr(struct bnx2x *bp)
                config->hdr.offset = (BP_PORT(bp) ? 32 : 0);
        else
                config->hdr.offset = BP_FUNC(bp);
-       config->hdr.client_id = BP_CL_ID(bp);
+       config->hdr.client_id = bp->fp->cl_id;
        config->hdr.reserved1 = 0;
 
        rc = bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_SET_MAC, 0,
@@ -9384,8 +9767,15 @@ static void bnx2x_self_test(struct net_device *dev,
                etest->flags &= ~ETH_TEST_FL_OFFLINE;
 
        if (etest->flags & ETH_TEST_FL_OFFLINE) {
+               int port = BP_PORT(bp);
+               u32 val;
                u8 link_up;
 
+               /* save current value of input enable for TX port IF */
+               val = REG_RD(bp, NIG_REG_EGRESS_UMP0_IN_EN + port*4);
+               /* disable input for TX port IF */
+               REG_WR(bp, NIG_REG_EGRESS_UMP0_IN_EN + port*4, 0);
+
                link_up = bp->link_vars.link_up;
                bnx2x_nic_unload(bp, UNLOAD_NORMAL);
                bnx2x_nic_load(bp, LOAD_DIAG);
@@ -9405,6 +9795,10 @@ static void bnx2x_self_test(struct net_device *dev,
                        etest->flags |= ETH_TEST_FL_FAILED;
 
                bnx2x_nic_unload(bp, UNLOAD_NORMAL);
+
+               /* restore input for TX port IF */
+               REG_WR(bp, NIG_REG_EGRESS_UMP0_IN_EN + port*4, val);
+
                bnx2x_nic_load(bp, LOAD_NORMAL);
                /* wait until link state is restored */
                bnx2x_wait_for_link(bp, link_up);
@@ -9728,12 +10122,14 @@ static struct ethtool_ops bnx2x_ethtool_ops = {
        .get_settings           = bnx2x_get_settings,
        .set_settings           = bnx2x_set_settings,
        .get_drvinfo            = bnx2x_get_drvinfo,
+       .get_regs_len           = bnx2x_get_regs_len,
+       .get_regs               = bnx2x_get_regs,
        .get_wol                = bnx2x_get_wol,
        .set_wol                = bnx2x_set_wol,
        .get_msglevel           = bnx2x_get_msglevel,
        .set_msglevel           = bnx2x_set_msglevel,
        .nway_reset             = bnx2x_nway_reset,
-       .get_link               = ethtool_op_get_link,
+       .get_link               = bnx2x_get_link,
        .get_eeprom_len         = bnx2x_get_eeprom_len,
        .get_eeprom             = bnx2x_get_eeprom,
        .set_eeprom             = bnx2x_set_eeprom,
@@ -9840,25 +10236,41 @@ static int bnx2x_poll(struct napi_struct *napi, int budget)
        bnx2x_update_fpsb_idx(fp);
 
        if (bnx2x_has_tx_work(fp))
-               bnx2x_tx_int(fp, budget);
+               bnx2x_tx_int(fp);
 
-       if (bnx2x_has_rx_work(fp))
+       if (bnx2x_has_rx_work(fp)) {
                work_done = bnx2x_rx_int(fp, budget);
-       rmb(); /* BNX2X_HAS_WORK() reads the status block */
 
-       /* must not complete if we consumed full budget */
-       if ((work_done < budget) && !BNX2X_HAS_WORK(fp)) {
+               /* must not complete if we consumed full budget */
+               if (work_done >= budget)
+                       goto poll_again;
+       }
 
+       /* BNX2X_HAS_WORK() reads the status block, thus we need to
+        * ensure that status block indices have been actually read
+        * (bnx2x_update_fpsb_idx) prior to this check (BNX2X_HAS_WORK)
+        * so that we won't write the "newer" value of the status block to IGU
+        * (if there was a DMA right after BNX2X_HAS_WORK and
+        * if there is no rmb, the memory reading (bnx2x_update_fpsb_idx)
+        * may be postponed to right before bnx2x_ack_sb). In this case
+        * there will never be another interrupt until there is another update
+        * of the status block, while there is still unhandled work.
+        */
+       rmb();
+
+       if (!BNX2X_HAS_WORK(fp)) {
 #ifdef BNX2X_STOP_ON_ERROR
 poll_panic:
 #endif
                napi_complete(napi);
 
-               bnx2x_ack_sb(bp, FP_SB_ID(fp), USTORM_ID,
+               bnx2x_ack_sb(bp, fp->sb_id, USTORM_ID,
                             le16_to_cpu(fp->fp_u_idx), IGU_INT_NOP, 1);
-               bnx2x_ack_sb(bp, FP_SB_ID(fp), CSTORM_ID,
+               bnx2x_ack_sb(bp, fp->sb_id, CSTORM_ID,
                             le16_to_cpu(fp->fp_c_idx), IGU_INT_ENABLE, 1);
        }
+
+poll_again:
        return work_done;
 }
 
@@ -9934,7 +10346,7 @@ static inline u32 bnx2x_xmit_type(struct bnx2x *bp, struct sk_buff *skb)
                rc = XMIT_PLAIN;
 
        else {
-               if (skb->protocol == ntohs(ETH_P_IPV6)) {
+               if (skb->protocol == htons(ETH_P_IPV6)) {
                        rc = XMIT_CSUM_V6;
                        if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
                                rc |= XMIT_CSUM_TCP;
@@ -9956,7 +10368,9 @@ static inline u32 bnx2x_xmit_type(struct bnx2x *bp, struct sk_buff *skb)
 }
 
 #if (MAX_SKB_FRAGS >= MAX_FETCH_BD - 3)
-/* check if packet requires linearization (packet is too fragmented) */
+/* check if packet requires linearization (packet is too fragmented)
+   no need to check fragmentation if page size > 8K (there will be no
+   violation to FW restrictions) */
 static int bnx2x_pkt_req_lin(struct bnx2x *bp, struct sk_buff *skb,
                             u32 xmit_type)
 {
@@ -10015,7 +10429,6 @@ static int bnx2x_pkt_req_lin(struct bnx2x *bp, struct sk_buff *skb,
                                wnd_sum -=
                                        skb_shinfo(skb)->frags[wnd_idx].size;
                        }
-
                } else {
                        /* in non-LSO too fragmented packet should always
                           be linearized */
@@ -10078,8 +10491,9 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
           ip_hdr(skb)->protocol, skb_shinfo(skb)->gso_type, xmit_type);
 
 #if (MAX_SKB_FRAGS >= MAX_FETCH_BD - 3)
-       /* First, check if we need to linearize the skb
-          (due to FW restrictions) */
+       /* First, check if we need to linearize the skb (due to FW
+          restrictions). No need to check fragmentation if page size > 8K
+          (there will be no violation to FW restrictions) */
        if (bnx2x_pkt_req_lin(bp, skb, xmit_type)) {
                /* Statistics of linearization */
                bp->lin_cnt++;
@@ -10144,9 +10558,9 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
                hlen = (skb_network_header(skb) - skb->data + vlan_off) / 2;
 
                /* for now NS flag is not used in Linux */
-               pbd->global_data = (hlen |
-                                   ((skb->protocol == ntohs(ETH_P_8021Q)) <<
-                                    ETH_TX_PARSE_BD_LLC_SNAP_EN_SHIFT));
+               pbd->global_data =
+                       (hlen | ((skb->protocol == cpu_to_be16(ETH_P_8021Q)) <<
+                                ETH_TX_PARSE_BD_LLC_SNAP_EN_SHIFT));
 
                pbd->ip_hlen = (skb_transport_header(skb) -
                                skb_network_header(skb)) / 2;
@@ -10290,17 +10704,14 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
         */
        wmb();
 
-       fp->hw_tx_prods->bds_prod =
-               cpu_to_le16(le16_to_cpu(fp->hw_tx_prods->bds_prod) + nbd);
+       le16_add_cpu(&fp->hw_tx_prods->bds_prod, nbd);
        mb(); /* FW restriction: must not reorder writing nbd and packets */
-       fp->hw_tx_prods->packets_prod =
-               cpu_to_le32(le32_to_cpu(fp->hw_tx_prods->packets_prod) + 1);
-       DOORBELL(bp, FP_IDX(fp), 0);
+       le32_add_cpu(&fp->hw_tx_prods->packets_prod, 1);
+       DOORBELL(bp, fp->index, 0);
 
        mmiowb();
 
        fp->tx_bd_prod += nbd;
-       dev->trans_start = jiffies;
 
        if (unlikely(bnx2x_tx_avail(fp) < MAX_SKB_FRAGS + 3)) {
                /* We want bnx2x_tx_int to "see" the updated tx_bd_prod
@@ -10342,7 +10753,7 @@ static int bnx2x_close(struct net_device *dev)
        return 0;
 }
 
-/* called with netif_tx_lock from set_multicast */
+/* called with netif_tx_lock from dev_mcast.c */
 static void bnx2x_set_rx_mode(struct net_device *dev)
 {
        struct bnx2x *bp = netdev_priv(dev);
@@ -10608,7 +11019,7 @@ static const struct net_device_ops bnx2x_netdev_ops = {
        .ndo_open               = bnx2x_open,
        .ndo_stop               = bnx2x_close,
        .ndo_start_xmit         = bnx2x_start_xmit,
-       .ndo_set_multicast_list = bnx2x_set_rx_mode,
+       .ndo_set_multicast_list = bnx2x_set_rx_mode,
        .ndo_set_mac_address    = bnx2x_change_mac_addr,
        .ndo_validate_addr      = eth_validate_addr,
        .ndo_do_ioctl           = bnx2x_ioctl,
@@ -10622,7 +11033,6 @@ static const struct net_device_ops bnx2x_netdev_ops = {
 #endif
 };
 
-
 static int __devinit bnx2x_init_dev(struct pci_dev *pdev,
                                    struct net_device *dev)
 {
@@ -10685,16 +11095,16 @@ static int __devinit bnx2x_init_dev(struct pci_dev *pdev,
                goto err_out_release;
        }
 
-       if (pci_set_dma_mask(pdev, DMA_64BIT_MASK) == 0) {
+       if (pci_set_dma_mask(pdev, DMA_BIT_MASK(64)) == 0) {
                bp->flags |= USING_DAC_FLAG;
-               if (pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK) != 0) {
+               if (pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64)) != 0) {
                        printk(KERN_ERR PFX "pci_set_consistent_dma_mask"
                               " failed, aborting\n");
                        rc = -EIO;
                        goto err_out_release;
                }
 
-       } else if (pci_set_dma_mask(pdev, DMA_32BIT_MASK) != 0) {
+       } else if (pci_set_dma_mask(pdev, DMA_BIT_MASK(32)) != 0) {
                printk(KERN_ERR PFX "System does not support DMA,"
                       " aborting\n");
                rc = -EIO;
@@ -10741,12 +11151,19 @@ static int __devinit bnx2x_init_dev(struct pci_dev *pdev,
        dev->features |= NETIF_F_HW_CSUM;
        if (bp->flags & USING_DAC_FLAG)
                dev->features |= NETIF_F_HIGHDMA;
+       dev->features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
+       dev->features |= NETIF_F_TSO6;
 #ifdef BCM_VLAN
        dev->features |= (NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX);
        bp->flags |= (HW_VLAN_RX_FLAG | HW_VLAN_TX_FLAG);
+
+       dev->vlan_features |= NETIF_F_SG;
+       dev->vlan_features |= NETIF_F_HW_CSUM;
+       if (bp->flags & USING_DAC_FLAG)
+               dev->vlan_features |= NETIF_F_HIGHDMA;
+       dev->vlan_features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
+       dev->vlan_features |= NETIF_F_TSO6;
 #endif
-       dev->features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
-       dev->features |= NETIF_F_TSO6;
 
        return 0;
 
@@ -10788,6 +11205,190 @@ static int __devinit bnx2x_get_pcie_speed(struct bnx2x *bp)
        val = (val & PCICFG_LINK_SPEED) >> PCICFG_LINK_SPEED_SHIFT;
        return val;
 }
+static int __devinit bnx2x_check_firmware(struct bnx2x *bp)
+{
+       struct bnx2x_fw_file_hdr *fw_hdr;
+       struct bnx2x_fw_file_section *sections;
+       u16 *ops_offsets;
+       u32 offset, len, num_ops;
+       int i;
+       const struct firmware *firmware = bp->firmware;
+       const u8 * fw_ver;
+
+       if (firmware->size < sizeof(struct bnx2x_fw_file_hdr))
+               return -EINVAL;
+
+       fw_hdr = (struct bnx2x_fw_file_hdr *)firmware->data;
+       sections = (struct bnx2x_fw_file_section *)fw_hdr;
+
+       /* Make sure none of the offsets and sizes make us read beyond
+        * the end of the firmware data */
+       for (i = 0; i < sizeof(*fw_hdr) / sizeof(*sections); i++) {
+               offset = be32_to_cpu(sections[i].offset);
+               len = be32_to_cpu(sections[i].len);
+               if (offset + len > firmware->size) {
+                       printk(KERN_ERR PFX "Section %d length is out of bounds\n", i);
+                       return -EINVAL;
+               }
+       }
+
+       /* Likewise for the init_ops offsets */
+       offset = be32_to_cpu(fw_hdr->init_ops_offsets.offset);
+       ops_offsets = (u16 *)(firmware->data + offset);
+       num_ops = be32_to_cpu(fw_hdr->init_ops.len) / sizeof(struct raw_op);
+
+       for (i = 0; i < be32_to_cpu(fw_hdr->init_ops_offsets.len) / 2; i++) {
+               if (be16_to_cpu(ops_offsets[i]) > num_ops) {
+                       printk(KERN_ERR PFX "Section offset %d is out of bounds\n", i);
+                       return -EINVAL;
+               }
+       }
+
+       /* Check FW version */
+       offset = be32_to_cpu(fw_hdr->fw_version.offset);
+       fw_ver = firmware->data + offset;
+       if ((fw_ver[0] != BCM_5710_FW_MAJOR_VERSION) ||
+           (fw_ver[1] != BCM_5710_FW_MINOR_VERSION) ||
+           (fw_ver[2] != BCM_5710_FW_REVISION_VERSION) ||
+           (fw_ver[3] != BCM_5710_FW_ENGINEERING_VERSION)) {
+               printk(KERN_ERR PFX "Bad FW version:%d.%d.%d.%d."
+                                   " Should be %d.%d.%d.%d\n",
+                      fw_ver[0], fw_ver[1], fw_ver[2],
+                      fw_ver[3], BCM_5710_FW_MAJOR_VERSION,
+                      BCM_5710_FW_MINOR_VERSION,
+                      BCM_5710_FW_REVISION_VERSION,
+                      BCM_5710_FW_ENGINEERING_VERSION);
+                return -EINVAL;
+       }
+
+       return 0;
+}
+
+static void inline be32_to_cpu_n(const u8 *_source, u8 *_target, u32 n)
+{
+       u32 i;
+       const __be32 *source = (const __be32*)_source;
+       u32 *target = (u32*)_target;
+
+       for (i = 0; i < n/4; i++)
+               target[i] = be32_to_cpu(source[i]);
+}
+
+/*
+   Ops array is stored in the following format:
+   {op(8bit), offset(24bit, big endian), data(32bit, big endian)}
+ */
+static void inline bnx2x_prep_ops(const u8 *_source, u8 *_target, u32 n)
+{
+       u32 i, j, tmp;
+       const __be32 *source = (const __be32*)_source;
+       struct raw_op *target = (struct raw_op*)_target;
+
+       for (i = 0, j = 0; i < n/8; i++, j+=2) {
+               tmp = be32_to_cpu(source[j]);
+               target[i].op = (tmp >> 24) & 0xff;
+               target[i].offset =  tmp & 0xffffff;
+               target[i].raw_data = be32_to_cpu(source[j+1]);
+       }
+}
+static void inline be16_to_cpu_n(const u8 *_source, u8 *_target, u32 n)
+{
+       u32 i;
+       u16 *target = (u16*)_target;
+       const __be16 *source = (const __be16*)_source;
+
+       for (i = 0; i < n/2; i++)
+               target[i] = be16_to_cpu(source[i]);
+}
+
+#define BNX2X_ALLOC_AND_SET(arr, lbl, func) \
+       do {   \
+               u32 len = be32_to_cpu(fw_hdr->arr.len);   \
+               bp->arr = kmalloc(len, GFP_KERNEL);  \
+               if (!bp->arr) { \
+                       printk(KERN_ERR PFX "Failed to allocate %d bytes for "#arr"\n", len); \
+                       goto lbl; \
+               } \
+               func(bp->firmware->data + \
+                       be32_to_cpu(fw_hdr->arr.offset), \
+                       (u8*)bp->arr, len); \
+       } while (0)
+
+
+static int __devinit bnx2x_init_firmware(struct bnx2x *bp, struct device *dev)
+{
+       char fw_file_name[40] = {0};
+        int rc, offset;
+       struct bnx2x_fw_file_hdr *fw_hdr;
+
+       /* Create a FW file name */
+       if (CHIP_IS_E1(bp))
+                offset = sprintf(fw_file_name, FW_FILE_PREFIX_E1);
+       else
+               offset = sprintf(fw_file_name, FW_FILE_PREFIX_E1H);
+
+       sprintf(fw_file_name + offset, "%d.%d.%d.%d.fw",
+               BCM_5710_FW_MAJOR_VERSION,
+                BCM_5710_FW_MINOR_VERSION,
+                BCM_5710_FW_REVISION_VERSION,
+                BCM_5710_FW_ENGINEERING_VERSION);
+
+       printk(KERN_INFO PFX "Loading %s\n", fw_file_name);
+
+       rc = request_firmware(&bp->firmware, fw_file_name, dev);
+       if (rc) {
+               printk(KERN_ERR PFX "Can't load firmware file %s\n", fw_file_name);
+               goto request_firmware_exit;
+       }
+
+       rc = bnx2x_check_firmware(bp);
+       if (rc) {
+               printk(KERN_ERR PFX "Corrupt firmware file %s\n", fw_file_name);
+               goto request_firmware_exit;
+       }
+
+       fw_hdr = (struct bnx2x_fw_file_hdr *)bp->firmware->data;
+
+       /* Initialize the pointers to the init arrays */
+       /* Blob */
+       BNX2X_ALLOC_AND_SET(init_data, request_firmware_exit, be32_to_cpu_n);
+
+       /* Opcodes */
+       BNX2X_ALLOC_AND_SET(init_ops, init_ops_alloc_err, bnx2x_prep_ops);
+
+       /* Offsets */
+       BNX2X_ALLOC_AND_SET(init_ops_offsets, init_offsets_alloc_err, be16_to_cpu_n);
+
+       /* STORMs firmware */
+       bp->tsem_int_table_data = bp->firmware->data +
+               be32_to_cpu(fw_hdr->tsem_int_table_data.offset);
+       bp->tsem_pram_data      = bp->firmware->data +
+               be32_to_cpu(fw_hdr->tsem_pram_data.offset);
+       bp->usem_int_table_data = bp->firmware->data +
+               be32_to_cpu(fw_hdr->usem_int_table_data.offset);
+       bp->usem_pram_data      = bp->firmware->data +
+               be32_to_cpu(fw_hdr->usem_pram_data.offset);
+       bp->xsem_int_table_data = bp->firmware->data +
+               be32_to_cpu(fw_hdr->xsem_int_table_data.offset);
+       bp->xsem_pram_data      = bp->firmware->data +
+               be32_to_cpu(fw_hdr->xsem_pram_data.offset);
+       bp->csem_int_table_data = bp->firmware->data +
+               be32_to_cpu(fw_hdr->csem_int_table_data.offset);
+       bp->csem_pram_data      = bp->firmware->data +
+               be32_to_cpu(fw_hdr->csem_pram_data.offset);
+
+       return 0;
+init_offsets_alloc_err:
+       kfree(bp->init_ops);
+init_ops_alloc_err:
+       kfree(bp->init_data);
+request_firmware_exit:
+       release_firmware(bp->firmware);
+
+       return rc;
+}
+
+
 
 static int __devinit bnx2x_init_one(struct pci_dev *pdev,
                                    const struct pci_device_id *ent)
@@ -10822,6 +11423,13 @@ static int __devinit bnx2x_init_one(struct pci_dev *pdev,
        if (rc)
                goto init_one_exit;
 
+       /* Set init arrays */
+       rc = bnx2x_init_firmware(bp, &pdev->dev);
+       if (rc) {
+               printk(KERN_ERR PFX "Error loading firmware\n");
+               goto init_one_exit;
+       }
+
        rc = register_netdev(dev);
        if (rc) {
                dev_err(&pdev->dev, "Cannot register net device\n");
@@ -10835,6 +11443,7 @@ static int __devinit bnx2x_init_one(struct pci_dev *pdev,
               (bnx2x_get_pcie_speed(bp) == 2) ? "5GHz (Gen2)" : "2.5GHz",
               dev->base_addr, bp->pdev->irq);
        printk(KERN_CONT "node addr %pM\n", dev->dev_addr);
+
        return 0;
 
 init_one_exit:
@@ -10868,6 +11477,11 @@ static void __devexit bnx2x_remove_one(struct pci_dev *pdev)
 
        unregister_netdev(dev);
 
+       kfree(bp->init_ops_offsets);
+       kfree(bp->init_ops);
+       kfree(bp->init_data);
+       release_firmware(bp->firmware);
+
        if (bp->regview)
                iounmap(bp->regview);
 
@@ -11101,8 +11715,8 @@ static void bnx2x_io_resume(struct pci_dev *pdev)
 
 static struct pci_error_handlers bnx2x_err_handler = {
        .error_detected = bnx2x_io_error_detected,
-       .slot_reset = bnx2x_io_slot_reset,
-       .resume = bnx2x_io_resume,
+       .slot_reset     = bnx2x_io_slot_reset,
+       .resume         = bnx2x_io_resume,
 };
 
 static struct pci_driver bnx2x_pci_driver = {
@@ -11117,13 +11731,20 @@ static struct pci_driver bnx2x_pci_driver = {
 
 static int __init bnx2x_init(void)
 {
+       int ret;
+
        bnx2x_wq = create_singlethread_workqueue("bnx2x");
        if (bnx2x_wq == NULL) {
                printk(KERN_ERR PFX "Cannot create workqueue\n");
                return -ENOMEM;
        }
 
-       return pci_register_driver(&bnx2x_pci_driver);
+       ret = pci_register_driver(&bnx2x_pci_driver);
+       if (ret) {
+               printk(KERN_ERR PFX "Cannot register driver\n");
+               destroy_workqueue(bnx2x_wq);
+       }
+       return ret;
 }
 
 static void __exit bnx2x_cleanup(void)
@@ -11136,3 +11757,4 @@ static void __exit bnx2x_cleanup(void)
 module_init(bnx2x_init);
 module_exit(bnx2x_cleanup);
 
+