]> pilppa.org Git - linux-2.6-omap-h63xx.git/blobdiff - drivers/net/bnx2x_main.c
Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6-omap-h63xx.git] / drivers / net / bnx2x_main.c
index 4be05847f86fe7d1fcd865a6108920a7ef108c54..d3e7775a9ccfaa2cc262c28ff4ba6c60b4a1ab6e 100644 (file)
@@ -1,6 +1,6 @@
 /* bnx2x_main.c: Broadcom Everest network driver.
  *
- * Copyright (c) 2007-2008 Broadcom Corporation
+ * Copyright (c) 2007-2009 Broadcom Corporation
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -38,9 +38,7 @@
 #include <linux/time.h>
 #include <linux/ethtool.h>
 #include <linux/mii.h>
-#ifdef NETIF_F_HW_VLAN_TX
-       #include <linux/if_vlan.h>
-#endif
+#include <linux/if_vlan.h>
 #include <net/ip.h>
 #include <net/tcp.h>
 #include <net/checksum.h>
@@ -59,8 +57,8 @@
 #include "bnx2x.h"
 #include "bnx2x_init.h"
 
-#define DRV_MODULE_VERSION     "1.45.23"
-#define DRV_MODULE_RELDATE     "2008/11/03"
+#define DRV_MODULE_VERSION     "1.45.26"
+#define DRV_MODULE_RELDATE     "2009/01/26"
 #define BNX2X_BC_VER           0x040200
 
 /* Time in jiffies before concluding the transmitter is hung */
@@ -71,7 +69,7 @@ static char version[] __devinitdata =
        DRV_MODULE_NAME " " DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
 
 MODULE_AUTHOR("Eliezer Tamir");
-MODULE_DESCRIPTION("Broadcom NetXtreme II BCM57710 Driver");
+MODULE_DESCRIPTION("Broadcom NetXtreme II BCM57710/57711/57711E Driver");
 MODULE_LICENSE("GPL");
 MODULE_VERSION(DRV_MODULE_VERSION);
 
@@ -95,6 +93,7 @@ MODULE_PARM_DESC(debug, "default debug msglevel");
 module_param(use_multi, int, 0);
 MODULE_PARM_DESC(use_multi, "use per-CPU queues");
 #endif
+static struct workqueue_struct *bnx2x_wq;
 
 enum bnx2x_board_type {
        BCM57710 = 0,
@@ -671,7 +670,8 @@ static void bnx2x_int_disable_sync(struct bnx2x *bp, int disable_hw)
                synchronize_irq(bp->pdev->irq);
 
        /* make sure sp_task is not running */
-       cancel_work_sync(&bp->sp_task);
+       cancel_delayed_work(&bp->sp_task);
+       flush_workqueue(bnx2x_wq);
 }
 
 /* fast path */
@@ -733,6 +733,24 @@ static u16 bnx2x_ack_int(struct bnx2x *bp)
  * fast path service functions
  */
 
+static inline int bnx2x_has_tx_work(struct bnx2x_fastpath *fp)
+{
+       u16 tx_cons_sb;
+
+       /* Tell compiler that status block fields can change */
+       barrier();
+       tx_cons_sb = le16_to_cpu(*fp->tx_cons_sb);
+       return (fp->tx_pkt_cons != tx_cons_sb);
+}
+
+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
  * return idx of last bd freed
  */
@@ -972,7 +990,7 @@ static inline void bnx2x_free_rx_sge(struct bnx2x *bp,
                return;
 
        pci_unmap_page(bp->pdev, pci_unmap_addr(sw_buf, mapping),
-                      BCM_PAGE_SIZE*PAGES_PER_SGE, PCI_DMA_FROMDEVICE);
+                      SGE_PAGE_SIZE*PAGES_PER_SGE, PCI_DMA_FROMDEVICE);
        __free_pages(page, PAGES_PER_SGE_SHIFT);
 
        sw_buf->page = NULL;
@@ -1000,7 +1018,7 @@ static inline int bnx2x_alloc_rx_sge(struct bnx2x *bp,
        if (unlikely(page == NULL))
                return -ENOMEM;
 
-       mapping = pci_map_page(bp->pdev, page, 0, BCM_PAGE_SIZE*PAGES_PER_SGE,
+       mapping = pci_map_page(bp->pdev, page, 0, SGE_PAGE_SIZE*PAGES_PER_SGE,
                               PCI_DMA_FROMDEVICE);
        if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
                __free_pages(page, PAGES_PER_SGE_SHIFT);
@@ -1096,9 +1114,9 @@ static void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
                                  struct eth_fast_path_rx_cqe *fp_cqe)
 {
        struct bnx2x *bp = fp->bp;
-       u16 sge_len = BCM_PAGE_ALIGN(le16_to_cpu(fp_cqe->pkt_len) -
+       u16 sge_len = SGE_PAGE_ALIGN(le16_to_cpu(fp_cqe->pkt_len) -
                                     le16_to_cpu(fp_cqe->len_on_bd)) >>
-                     BCM_PAGE_SHIFT;
+                     SGE_PAGE_SHIFT;
        u16 last_max, last_elem, first_elem;
        u16 delta = 0;
        u16 i;
@@ -1203,22 +1221,22 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
                               u16 cqe_idx)
 {
        struct sw_rx_page *rx_pg, old_rx_pg;
-       struct page *sge;
        u16 len_on_bd = le16_to_cpu(fp_cqe->len_on_bd);
        u32 i, frag_len, frag_size, pages;
        int err;
        int j;
 
        frag_size = le16_to_cpu(fp_cqe->pkt_len) - len_on_bd;
-       pages = BCM_PAGE_ALIGN(frag_size) >> BCM_PAGE_SHIFT;
+       pages = SGE_PAGE_ALIGN(frag_size) >> SGE_PAGE_SHIFT;
 
        /* This is needed in order to enable forwarding support */
        if (frag_size)
-               skb_shinfo(skb)->gso_size = min((u32)BCM_PAGE_SIZE,
+               skb_shinfo(skb)->gso_size = min((u32)SGE_PAGE_SIZE,
                                               max(frag_size, (u32)len_on_bd));
 
 #ifdef BNX2X_STOP_ON_ERROR
-       if (pages > 8*PAGES_PER_SGE) {
+       if (pages >
+           min((u32)8, (u32)MAX_SKB_FRAGS) * SGE_PAGE_SIZE * PAGES_PER_SGE) {
                BNX2X_ERR("SGL length is too long: %d. CQE index is %d\n",
                          pages, cqe_idx);
                BNX2X_ERR("fp_cqe->pkt_len = %d  fp_cqe->len_on_bd = %d\n",
@@ -1234,9 +1252,8 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 
                /* FW gives the indices of the SGE as if the ring is an array
                   (meaning that "next" element will consume 2 indices) */
-               frag_len = min(frag_size, (u32)(BCM_PAGE_SIZE*PAGES_PER_SGE));
+               frag_len = min(frag_size, (u32)(SGE_PAGE_SIZE*PAGES_PER_SGE));
                rx_pg = &fp->rx_page_ring[sge_idx];
-               sge = rx_pg->page;
                old_rx_pg = *rx_pg;
 
                /* If we fail to allocate a substitute page, we simply stop
@@ -1249,7 +1266,7 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 
                /* Unmap the page as we r going to pass it to the stack */
                pci_unmap_page(bp->pdev, pci_unmap_addr(&old_rx_pg, mapping),
-                             BCM_PAGE_SIZE*PAGES_PER_SGE, PCI_DMA_FROMDEVICE);
+                             SGE_PAGE_SIZE*PAGES_PER_SGE, PCI_DMA_FROMDEVICE);
 
                /* Add one frag and update the appropriate fields in the skb */
                skb_fill_page_desc(skb, j, old_rx_pg.page, 0, frag_len);
@@ -1282,6 +1299,13 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
        if (likely(new_skb)) {
                /* fix ip xsum and give it to the stack */
                /* (no need to map the new skb) */
+#ifdef BCM_VLAN
+               int is_vlan_cqe =
+                       (le16_to_cpu(cqe->fast_path_cqe.pars_flags.flags) &
+                        PARSING_FLAGS_VLAN);
+               int is_not_hwaccel_vlan_cqe =
+                       (is_vlan_cqe && (!(bp->flags & HW_VLAN_RX_FLAG)));
+#endif
 
                prefetch(skb);
                prefetch(((char *)(skb)) + 128);
@@ -1306,6 +1330,12 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
                        struct iphdr *iph;
 
                        iph = (struct iphdr *)skb->data;
+#ifdef BCM_VLAN
+                       /* If there is no Rx VLAN offloading -
+                          take VLAN tag into an account */
+                       if (unlikely(is_not_hwaccel_vlan_cqe))
+                               iph = (struct iphdr *)((u8 *)iph + VLAN_HLEN);
+#endif
                        iph->check = 0;
                        iph->check = ip_fast_csum((u8 *)iph, iph->ihl);
                }
@@ -1313,9 +1343,8 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
                if (!bnx2x_fill_frag_skb(bp, fp, skb,
                                         &cqe->fast_path_cqe, cqe_idx)) {
 #ifdef BCM_VLAN
-                       if ((bp->vlgrp != NULL) &&
-                           (le16_to_cpu(cqe->fast_path_cqe.pars_flags.flags) &
-                            PARSING_FLAGS_VLAN))
+                       if ((bp->vlgrp != NULL) && is_vlan_cqe &&
+                           (!is_not_hwaccel_vlan_cqe))
                                vlan_hwaccel_receive_skb(skb, bp->vlgrp,
                                                le16_to_cpu(cqe->fast_path_cqe.
                                                            vlan_tag));
@@ -1355,11 +1384,23 @@ static inline void bnx2x_update_rx_prod(struct bnx2x *bp,
        rx_prods.cqe_prod = rx_comp_prod;
        rx_prods.sge_prod = rx_sge_prod;
 
+       /*
+        * Make sure that the BD and SGE data is updated before updating the
+        * producers since FW might read the BD/SGE right after the producer
+        * is updated.
+        * This is only applicable for weak-ordered memory model archs such
+        * as IA-64. The following barrier is also mandatory since FW will
+        * assumes BDs must have buffers.
+        */
+       wmb();
+
        for (i = 0; i < sizeof(struct tstorm_eth_rx_producers)/4; i++)
                REG_WR(bp, BAR_TSTRORM_INTMEM +
                       TSTORM_RX_PRODS_OFFSET(BP_PORT(bp), FP_CL_ID(fp)) + i*4,
                       ((u32 *)&rx_prods)[i]);
 
+       mmiowb(); /* keep prod updates ordered */
+
        DP(NETIF_MSG_RX_STATUS,
           "Wrote: bd_prod %u  cqe_prod %u  sge_prod %u\n",
           bd_prod, rx_comp_prod, rx_sge_prod);
@@ -1415,7 +1456,7 @@ static int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
                DP(NETIF_MSG_RX_STATUS, "CQE type %x  err %x  status %x"
                   "  queue %x  vlan %x  len %u\n", CQE_TYPE(cqe_fp_flags),
                   cqe_fp_flags, cqe->fast_path_cqe.status_flags,
-                  cqe->fast_path_cqe.rss_hash_result,
+                  le32_to_cpu(cqe->fast_path_cqe.rss_hash_result),
                   le16_to_cpu(cqe->fast_path_cqe.vlan_tag),
                   le16_to_cpu(cqe->fast_path_cqe.pkt_len));
 
@@ -1547,7 +1588,7 @@ reuse_rx:
                }
 
 #ifdef BCM_VLAN
-               if ((bp->vlgrp != NULL) &&
+               if ((bp->vlgrp != NULL) && (bp->flags & HW_VLAN_RX_FLAG) &&
                    (le16_to_cpu(cqe->fast_path_cqe.pars_flags.flags) &
                     PARSING_FLAGS_VLAN))
                        vlan_hwaccel_receive_skb(skb, bp->vlgrp,
@@ -1580,7 +1621,6 @@ next_cqe:
        /* Update producers */
        bnx2x_update_rx_prod(bp, fp, bd_prod_fw, sw_comp_prod,
                             fp->rx_sge_prod);
-       mmiowb(); /* keep prod updates ordered */
 
        fp->rx_pkt += rx_pkt;
        fp->rx_calls++;
@@ -1660,7 +1700,7 @@ static irqreturn_t bnx2x_interrupt(int irq, void *dev_instance)
 
 
        if (unlikely(status & 0x1)) {
-               schedule_work(&bp->sp_task);
+               queue_delayed_work(bnx2x_wq, &bp->sp_task, 0);
 
                status &= ~0x1;
                if (!status)
@@ -1887,7 +1927,8 @@ static int bnx2x_set_spio(struct bnx2x *bp, int spio_num, u32 mode)
 
 static void bnx2x_calc_fc_adv(struct bnx2x *bp)
 {
-       switch (bp->link_vars.ieee_fc) {
+       switch (bp->link_vars.ieee_fc &
+               MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_MASK) {
        case MDIO_COMBO_IEEE0_AUTO_NEG_ADV_PAUSE_NONE:
                bp->port.advertising &= ~(ADVERTISED_Asym_Pause |
                                          ADVERTISED_Pause);
@@ -1957,10 +1998,11 @@ static u8 bnx2x_initial_phy_init(struct bnx2x *bp)
                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)
                        bnx2x_link_report(bp);
 
-               bnx2x_calc_fc_adv(bp);
 
                return rc;
        }
@@ -2220,9 +2262,7 @@ static void bnx2x_link_attn(struct bnx2x *bp)
        /* Make sure that we are synced with the current statistics */
        bnx2x_stats_handle(bp, STATS_EVENT_STOP);
 
-       bnx2x_acquire_phy_lock(bp);
        bnx2x_link_update(&bp->link_params, &bp->link_vars);
-       bnx2x_release_phy_lock(bp);
 
        if (bp->link_vars.link_up) {
 
@@ -2471,6 +2511,8 @@ static void bnx2x_attn_int_asserted(struct bnx2x *bp, u32 asserted)
        if (asserted & ATTN_HARD_WIRED_MASK) {
                if (asserted & ATTN_NIG_FOR_FUNC) {
 
+                       bnx2x_acquire_phy_lock(bp);
+
                        /* save nig interrupt mask */
                        bp->nig_mask = REG_RD(bp, nig_int_mask_addr);
                        REG_WR(bp, nig_int_mask_addr, 0);
@@ -2526,8 +2568,10 @@ static void bnx2x_attn_int_asserted(struct bnx2x *bp, u32 asserted)
        REG_WR(bp, hc_addr, asserted);
 
        /* now set back the mask */
-       if (asserted & ATTN_NIG_FOR_FUNC)
+       if (asserted & ATTN_NIG_FOR_FUNC) {
                REG_WR(bp, nig_int_mask_addr, bp->nig_mask);
+               bnx2x_release_phy_lock(bp);
+       }
 }
 
 static inline void bnx2x_attn_int_deasserted0(struct bnx2x *bp, u32 attn)
@@ -2795,8 +2839,10 @@ static void bnx2x_attn_int_deasserted(struct bnx2x *bp, u32 deasserted)
 static void bnx2x_attn_int(struct bnx2x *bp)
 {
        /* read local copy of bits */
-       u32 attn_bits = bp->def_status_blk->atten_status_block.attn_bits;
-       u32 attn_ack = bp->def_status_blk->atten_status_block.attn_bits_ack;
+       u32 attn_bits = le32_to_cpu(bp->def_status_blk->atten_status_block.
+                                                               attn_bits);
+       u32 attn_ack = le32_to_cpu(bp->def_status_blk->atten_status_block.
+                                                               attn_bits_ack);
        u32 attn_state = bp->attn_state;
 
        /* look for changed bits */
@@ -2820,7 +2866,7 @@ static void bnx2x_attn_int(struct bnx2x *bp)
 
 static void bnx2x_sp_task(struct work_struct *work)
 {
-       struct bnx2x *bp = container_of(work, struct bnx2x, sp_task);
+       struct bnx2x *bp = container_of(work, struct bnx2x, sp_task.work);
        u16 status;
 
 
@@ -2844,7 +2890,7 @@ static void bnx2x_sp_task(struct work_struct *work)
        if (status & 0x2)
                bp->stats_pending = 0;
 
-       bnx2x_ack_sb(bp, DEF_SB_ID, ATTENTION_ID, bp->def_att_idx,
+       bnx2x_ack_sb(bp, DEF_SB_ID, ATTENTION_ID, le16_to_cpu(bp->def_att_idx),
                     IGU_INT_NOP, 1);
        bnx2x_ack_sb(bp, DEF_SB_ID, USTORM_ID, le16_to_cpu(bp->def_u_idx),
                     IGU_INT_NOP, 1);
@@ -2875,7 +2921,7 @@ static irqreturn_t bnx2x_msix_sp_int(int irq, void *dev_instance)
                return IRQ_HANDLED;
 #endif
 
-       schedule_work(&bp->sp_task);
+       queue_delayed_work(bnx2x_wq, &bp->sp_task, 0);
 
        return IRQ_HANDLED;
 }
@@ -2892,7 +2938,7 @@ static irqreturn_t bnx2x_msix_sp_int(int irq, void *dev_instance)
 #define ADD_64(s_hi, a_hi, s_lo, a_lo) \
        do { \
                s_lo += a_lo; \
-               s_hi += a_hi + (s_lo < a_lo) ? 1 : 0; \
+               s_hi += a_hi + ((s_lo < a_lo) ? 1 : 0); \
        } while (0)
 
 /* difference = minuend - subtrahend */
@@ -4496,7 +4542,7 @@ static void bnx2x_init_context(struct bnx2x *bp)
 
 static void bnx2x_init_ind_table(struct bnx2x *bp)
 {
-       int port = BP_PORT(bp);
+       int func = BP_FUNC(bp);
        int i;
 
        if (!is_multi(bp))
@@ -4505,10 +4551,8 @@ static void bnx2x_init_ind_table(struct bnx2x *bp)
        DP(NETIF_MSG_IFUP, "Initializing indirection table\n");
        for (i = 0; i < TSTORM_INDIRECTION_TABLE_SIZE; i++)
                REG_WR8(bp, BAR_TSTRORM_INTMEM +
-                       TSTORM_INDIRECTION_TABLE_OFFSET(port) + i,
-                       i % bp->num_queues);
-
-       REG_WR(bp, PRS_REG_A_PRSU_20, 0xf);
+                       TSTORM_INDIRECTION_TABLE_OFFSET(func) + i,
+                       BP_CL_ID(bp) + (i % bp->num_queues));
 }
 
 static void bnx2x_set_client_config(struct bnx2x *bp)
@@ -4517,12 +4561,12 @@ static void bnx2x_set_client_config(struct bnx2x *bp)
        int port = BP_PORT(bp);
        int i;
 
-       tstorm_client.mtu = bp->dev->mtu + ETH_OVREHEAD;
+       tstorm_client.mtu = bp->dev->mtu;
        tstorm_client.statistics_counter_id = BP_CL_ID(bp);
        tstorm_client.config_flags =
                                TSTORM_ETH_CLIENT_CONFIG_STATSITICS_ENABLE;
 #ifdef BCM_VLAN
-       if (bp->rx_mode && bp->vlgrp) {
+       if (bp->rx_mode && bp->vlgrp && (bp->flags & HW_VLAN_RX_FLAG)) {
                tstorm_client.config_flags |=
                                TSTORM_ETH_CLIENT_CONFIG_VLAN_REMOVAL_ENABLE;
                DP(NETIF_MSG_IFUP, "vlan removal enabled\n");
@@ -4531,7 +4575,7 @@ static void bnx2x_set_client_config(struct bnx2x *bp)
 
        if (bp->flags & TPA_ENABLE_FLAG) {
                tstorm_client.max_sges_for_packet =
-                       BCM_PAGE_ALIGN(tstorm_client.mtu) >> BCM_PAGE_SHIFT;
+                       SGE_PAGE_ALIGN(tstorm_client.mtu) >> SGE_PAGE_SHIFT;
                tstorm_client.max_sges_for_packet =
                        ((tstorm_client.max_sges_for_packet +
                          PAGES_PER_SGE - 1) & (~(PAGES_PER_SGE - 1))) >>
@@ -4714,10 +4758,11 @@ static void bnx2x_init_internal_func(struct bnx2x *bp)
                         bp->e1hov);
        }
 
-       /* Init CQ ring mapping and aggregation size */
-       max_agg_size = min((u32)(bp->rx_buf_size +
-                                8*BCM_PAGE_SIZE*PAGES_PER_SGE),
-                          (u32)0xffff);
+       /* Init CQ ring mapping and aggregation size, the FW limit is 8 frags */
+       max_agg_size =
+               min((u32)(min((u32)8, (u32)MAX_SKB_FRAGS) *
+                         SGE_PAGE_SIZE * PAGES_PER_SGE),
+                   (u32)0xffff);
        for_each_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
@@ -4785,6 +4830,15 @@ static void bnx2x_nic_init(struct bnx2x *bp, u32 load_code)
        bnx2x_init_context(bp);
        bnx2x_init_internal(bp, load_code);
        bnx2x_init_ind_table(bp);
+       bnx2x_stats_init(bp);
+
+       /* At this point, we are ready for interrupts */
+       atomic_set(&bp->intr_sem, 0);
+
+       /* flush all before enabling interrupts */
+       mb();
+       mmiowb();
+
        bnx2x_int_enable(bp);
 }
 
@@ -5101,12 +5155,21 @@ static void enable_blocks_attention(struct bnx2x *bp)
 }
 
 
+static void bnx2x_reset_common(struct bnx2x *bp)
+{
+       /* reset_common */
+       REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR,
+              0xd3ffff7f);
+       REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_CLEAR, 0x1403);
+}
+
 static int bnx2x_init_common(struct bnx2x *bp)
 {
        u32 val, i;
 
        DP(BNX2X_MSG_MCP, "starting common init  func %d\n", BP_FUNC(bp));
 
+       bnx2x_reset_common(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);
 
@@ -5134,7 +5197,6 @@ static int bnx2x_init_common(struct bnx2x *bp)
        REG_WR(bp, PXP2_REG_RQ_SRC_ENDIAN_M, 1);
        REG_WR(bp, PXP2_REG_RQ_CDU_ENDIAN_M, 1);
        REG_WR(bp, PXP2_REG_RQ_DBG_ENDIAN_M, 1);
-       REG_WR(bp, PXP2_REG_RQ_HC_ENDIAN_M, 1);
 
 /*     REG_WR(bp, PXP2_REG_RD_PBF_SWAP_MODE, 1); */
        REG_WR(bp, PXP2_REG_RD_QM_SWAP_MODE, 1);
@@ -5212,6 +5274,7 @@ static int bnx2x_init_common(struct bnx2x *bp)
        }
 
        bnx2x_init_block(bp, PRS_COMMON_START, PRS_COMMON_END);
+       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))
@@ -6087,8 +6150,8 @@ static void bnx2x_netif_start(struct bnx2x *bp)
 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)) {
-               bnx2x_napi_disable(bp);
                netif_tx_disable(bp->dev);
                bp->dev->trans_start = jiffies; /* prevent tx timeout */
        }
@@ -6108,7 +6171,7 @@ static void bnx2x_set_mac_addr_e1(struct bnx2x *bp, int set)
         * multicast 64-127:port0 128-191:port1
         */
        config->hdr.length_6b = 2;
-       config->hdr.offset = port ? 31 : 0;
+       config->hdr.offset = port ? 32 : 0;
        config->hdr.client_id = BP_CL_ID(bp);
        config->hdr.reserved1 = 0;
 
@@ -6272,7 +6335,7 @@ static void bnx2x_set_rx_mode(struct net_device *dev);
 static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 {
        u32 load_code;
-       int i, rc;
+       int i, rc = 0;
 #ifdef BNX2X_STOP_ON_ERROR
        if (unlikely(bp->panic))
                return -EPERM;
@@ -6280,48 +6343,6 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 
        bp->state = BNX2X_STATE_OPENING_WAIT4_LOAD;
 
-       /* Send LOAD_REQUEST command to MCP
-          Returns the type of LOAD command:
-          if it is the first port to be initialized
-          common blocks should be initialized, otherwise - not
-       */
-       if (!BP_NOMCP(bp)) {
-               load_code = bnx2x_fw_command(bp, DRV_MSG_CODE_LOAD_REQ);
-               if (!load_code) {
-                       BNX2X_ERR("MCP response failure, aborting\n");
-                       return -EBUSY;
-               }
-               if (load_code == FW_MSG_CODE_DRV_LOAD_REFUSED)
-                       return -EBUSY; /* other port in diagnostic mode */
-
-       } else {
-               int port = BP_PORT(bp);
-
-               DP(NETIF_MSG_IFUP, "NO MCP load counts before us %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",
-                  load_count[0], load_count[1], load_count[2]);
-               if (load_count[0] == 1)
-                       load_code = FW_MSG_CODE_DRV_LOAD_COMMON;
-               else if (load_count[1 + port] == 1)
-                       load_code = FW_MSG_CODE_DRV_LOAD_PORT;
-               else
-                       load_code = FW_MSG_CODE_DRV_LOAD_FUNCTION;
-       }
-
-       if ((load_code == FW_MSG_CODE_DRV_LOAD_COMMON) ||
-           (load_code == FW_MSG_CODE_DRV_LOAD_PORT))
-               bp->port.pmf = 1;
-       else
-               bp->port.pmf = 0;
-       DP(NETIF_MSG_LINK, "pmf %d\n", bp->port.pmf);
-
-       /* if we can't use MSI-X we only need one fp,
-        * so try to enable MSI-X with the requested number of fp's
-        * and fallback to inta with one fp
-        */
        if (use_inta) {
                bp->num_queues = 1;
 
@@ -6336,7 +6357,15 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                else
                        bp->num_queues = 1;
 
-               if (bnx2x_enable_msix(bp)) {
+               DP(NETIF_MSG_IFUP,
+                  "set number of queues to %d\n", bp->num_queues);
+
+               /* if we can't use MSI-X we only need one fp,
+                * so try to enable MSI-X with the requested number of fp's
+                * and fallback to MSI or legacy INTx with one fp
+                */
+               rc = bnx2x_enable_msix(bp);
+               if (rc) {
                        /* failed to enable MSI-X */
                        bp->num_queues = 1;
                        if (use_multi)
@@ -6344,8 +6373,6 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                                          " to enable MSI-X\n");
                }
        }
-       DP(NETIF_MSG_IFUP,
-          "set number of queues to %d\n", bp->num_queues);
 
        if (bnx2x_alloc_mem(bp))
                return -ENOMEM;
@@ -6354,30 +6381,85 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                bnx2x_fp(bp, i, disable_tpa) =
                                        ((bp->flags & TPA_ENABLE_FLAG) == 0);
 
+       for_each_queue(bp, i)
+               netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
+                              bnx2x_poll, 128);
+
+#ifdef BNX2X_STOP_ON_ERROR
+       for_each_queue(bp, i) {
+               struct bnx2x_fastpath *fp = &bp->fp[i];
+
+               fp->poll_no_work = 0;
+               fp->poll_calls = 0;
+               fp->poll_max_calls = 0;
+               fp->poll_complete = 0;
+               fp->poll_exit = 0;
+       }
+#endif
+       bnx2x_napi_enable(bp);
+
        if (bp->flags & USING_MSIX_FLAG) {
                rc = bnx2x_req_msix_irqs(bp);
                if (rc) {
                        pci_disable_msix(bp->pdev);
-                       goto load_error;
+                       goto load_error1;
                }
+               printk(KERN_INFO PFX "%s: using MSI-X\n", bp->dev->name);
        } else {
                bnx2x_ack_int(bp);
                rc = bnx2x_req_irq(bp);
                if (rc) {
-                       BNX2X_ERR("IRQ request failed, aborting\n");
-                       goto load_error;
+                       BNX2X_ERR("IRQ request failed  rc %d, aborting\n", rc);
+                       goto load_error1;
                }
        }
 
-       for_each_queue(bp, i)
-               netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
-                              bnx2x_poll, 128);
+       /* Send LOAD_REQUEST command to MCP
+          Returns the type of LOAD command:
+          if it is the first port to be initialized
+          common blocks should be initialized, otherwise - not
+       */
+       if (!BP_NOMCP(bp)) {
+               load_code = bnx2x_fw_command(bp, DRV_MSG_CODE_LOAD_REQ);
+               if (!load_code) {
+                       BNX2X_ERR("MCP response failure, aborting\n");
+                       rc = -EBUSY;
+                       goto load_error2;
+               }
+               if (load_code == FW_MSG_CODE_DRV_LOAD_REFUSED) {
+                       rc = -EBUSY; /* other port in diagnostic mode */
+                       goto load_error2;
+               }
+
+       } else {
+               int port = BP_PORT(bp);
+
+               DP(NETIF_MSG_IFUP, "NO MCP load counts before us %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",
+                  load_count[0], load_count[1], load_count[2]);
+               if (load_count[0] == 1)
+                       load_code = FW_MSG_CODE_DRV_LOAD_COMMON;
+               else if (load_count[1 + port] == 1)
+                       load_code = FW_MSG_CODE_DRV_LOAD_PORT;
+               else
+                       load_code = FW_MSG_CODE_DRV_LOAD_FUNCTION;
+       }
+
+       if ((load_code == FW_MSG_CODE_DRV_LOAD_COMMON) ||
+           (load_code == FW_MSG_CODE_DRV_LOAD_PORT))
+               bp->port.pmf = 1;
+       else
+               bp->port.pmf = 0;
+       DP(NETIF_MSG_LINK, "pmf %d\n", bp->port.pmf);
 
        /* Initialize HW */
        rc = bnx2x_init_hw(bp, load_code);
        if (rc) {
                BNX2X_ERR("HW init failed, aborting\n");
-               goto load_int_disable;
+               goto load_error2;
        }
 
        /* Setup NIC internals and enable interrupts */
@@ -6389,25 +6471,16 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                if (!load_code) {
                        BNX2X_ERR("MCP response failure, aborting\n");
                        rc = -EBUSY;
-                       goto load_rings_free;
+                       goto load_error3;
                }
        }
 
-       bnx2x_stats_init(bp);
-
        bp->state = BNX2X_STATE_OPENING_WAIT4_PORT;
 
-       /* Enable Rx interrupt handling before sending the ramrod
-          as it's completed on Rx FP queue */
-       bnx2x_napi_enable(bp);
-
-       /* Enable interrupt handling */
-       atomic_set(&bp->intr_sem, 0);
-
        rc = bnx2x_setup_leading(bp);
        if (rc) {
                BNX2X_ERR("Setup leading failed!\n");
-               goto load_netif_stop;
+               goto load_error3;
        }
 
        if (CHIP_IS_E1H(bp))
@@ -6420,7 +6493,7 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
                for_each_nondefault_queue(bp, i) {
                        rc = bnx2x_setup_multi(bp, i);
                        if (rc)
-                               goto load_netif_stop;
+                               goto load_error3;
                }
 
        if (CHIP_IS_E1(bp))
@@ -6436,18 +6509,18 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
        case LOAD_NORMAL:
                /* Tx queue should be only reenabled */
                netif_wake_queue(bp->dev);
+               /* Initialize the receive filter. */
                bnx2x_set_rx_mode(bp->dev);
                break;
 
        case LOAD_OPEN:
                netif_start_queue(bp->dev);
+               /* Initialize the receive filter. */
                bnx2x_set_rx_mode(bp->dev);
-               if (bp->flags & USING_MSIX_FLAG)
-                       printk(KERN_INFO PFX "%s: using MSI-X\n",
-                              bp->dev->name);
                break;
 
        case LOAD_DIAG:
+               /* Initialize the receive filter. */
                bnx2x_set_rx_mode(bp->dev);
                bp->state = BNX2X_STATE_DIAG;
                break;
@@ -6465,20 +6538,25 @@ static int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 
        return 0;
 
-load_netif_stop:
-       bnx2x_napi_disable(bp);
-load_rings_free:
+load_error3:
+       bnx2x_int_disable_sync(bp, 1);
+       if (!BP_NOMCP(bp)) {
+               bnx2x_fw_command(bp, DRV_MSG_CODE_UNLOAD_REQ_WOL_MCP);
+               bnx2x_fw_command(bp, DRV_MSG_CODE_UNLOAD_DONE);
+       }
+       bp->port.pmf = 0;
        /* Free SKBs, SGEs, TPA pool and driver internals */
        bnx2x_free_skbs(bp);
        for_each_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
-load_int_disable:
-       bnx2x_int_disable_sync(bp, 1);
+load_error2:
        /* Release IRQs */
        bnx2x_free_irq(bp);
-load_error:
+load_error1:
+       bnx2x_napi_disable(bp);
+       for_each_queue(bp, i)
+               netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
-       bp->port.pmf = 0;
 
        /* TBD we really need to reset the chip
           if we want to recover from this */
@@ -6551,6 +6629,7 @@ static int bnx2x_stop_leading(struct bnx2x *bp)
                }
                cnt--;
                msleep(1);
+               rmb(); /* Refresh the dsb_sp_prod */
        }
        bp->state = BNX2X_STATE_CLOSING_WAIT4_UNLOAD;
        bp->fp[0].state = BNX2X_FP_STATE_CLOSED;
@@ -6602,14 +6681,6 @@ static void bnx2x_reset_port(struct bnx2x *bp)
        /* TODO: Close Doorbell port? */
 }
 
-static void bnx2x_reset_common(struct bnx2x *bp)
-{
-       /* reset_common */
-       REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_1_CLEAR,
-              0xd3ffff7f);
-       REG_WR(bp, GRCBASE_MISC + MISC_REGISTERS_RESET_REG_2_CLEAR, 0x1403);
-}
-
 static void bnx2x_reset_chip(struct bnx2x *bp, u32 reset_code)
 {
        DP(BNX2X_MSG_MCP, "function %d  reset_code %x\n",
@@ -6650,20 +6721,22 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
        bnx2x_set_storm_rx_mode(bp);
 
        bnx2x_netif_stop(bp, 1);
-       if (!netif_running(bp->dev))
-               bnx2x_napi_disable(bp);
+
        del_timer_sync(&bp->timer);
        SHMEM_WR(bp, func_mb[BP_FUNC(bp)].drv_pulse_mb,
                 (DRV_PULSE_ALWAYS_ALIVE | bp->fw_drv_pulse_wr_seq));
        bnx2x_stats_handle(bp, STATS_EVENT_STOP);
 
+       /* Release IRQs */
+       bnx2x_free_irq(bp);
+
        /* Wait until tx fast path tasks complete */
        for_each_queue(bp, i) {
                struct bnx2x_fastpath *fp = &bp->fp[i];
 
                cnt = 1000;
                smp_rmb();
-               while (BNX2X_HAS_TX_WORK(fp)) {
+               while (bnx2x_has_tx_work_unload(fp)) {
 
                        bnx2x_tx_int(fp, 1000);
                        if (!cnt) {
@@ -6684,9 +6757,6 @@ static int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
        /* Give HW time to discard old tx messages */
        msleep(1);
 
-       /* Release IRQs */
-       bnx2x_free_irq(bp);
-
        if (CHIP_IS_E1(bp)) {
                struct mac_configuration_cmd *config =
                                                bnx2x_sp(bp, mcast_config);
@@ -6795,6 +6865,8 @@ unload_error:
        bnx2x_free_skbs(bp);
        for_each_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
+       for_each_queue(bp, i)
+               netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
 
        bp->state = BNX2X_STATE_CLOSED;
@@ -6847,10 +6919,6 @@ static void __devinit bnx2x_undi_unload(struct bnx2x *bp)
                 */
                bnx2x_acquire_hw_lock(bp, HW_LOCK_RESOURCE_UNDI);
                val = REG_RD(bp, DORQ_REG_NORM_CID_OFST);
-               if (val == 0x7)
-                       REG_WR(bp, DORQ_REG_NORM_CID_OFST, 0);
-               bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_UNDI);
-
                if (val == 0x7) {
                        u32 reset_code = DRV_MSG_CODE_UNLOAD_REQ_WOL_DIS;
                        /* save our func */
@@ -6858,6 +6926,9 @@ static void __devinit bnx2x_undi_unload(struct bnx2x *bp)
                        u32 swap_en;
                        u32 swap_val;
 
+                       /* clear the UNDI indication */
+                       REG_WR(bp, DORQ_REG_NORM_CID_OFST, 0);
+
                        BNX2X_DEV_INFO("UNDI is active! reset device\n");
 
                        /* try unload UNDI on port 0 */
@@ -6883,6 +6954,9 @@ static void __devinit bnx2x_undi_unload(struct bnx2x *bp)
                                bnx2x_fw_command(bp, reset_code);
                        }
 
+                       /* now it's safe to release the lock */
+                       bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_UNDI);
+
                        REG_WR(bp, (BP_PORT(bp) ? HC_REG_CONFIG_1 :
                                    HC_REG_CONFIG_0), 0x1000);
 
@@ -6927,7 +7001,9 @@ static void __devinit bnx2x_undi_unload(struct bnx2x *bp)
                        bp->fw_seq =
                               (SHMEM_RD(bp, func_mb[bp->func].drv_mb_header) &
                                DRV_MSG_SEQ_NUMBER_MASK);
-               }
+
+               } else
+                       bnx2x_release_hw_lock(bp, HW_LOCK_RESOURCE_UNDI);
        }
 }
 
@@ -6944,7 +7020,7 @@ static void __devinit bnx2x_get_common_hwinfo(struct bnx2x *bp)
        id |= ((val & 0xf) << 12);
        val = REG_RD(bp, MISC_REG_CHIP_METAL);
        id |= ((val & 0xff) << 4);
-       REG_RD(bp, MISC_REG_BOND_ID);
+       val = REG_RD(bp, MISC_REG_BOND_ID);
        id |= (val & 0xf);
        bp->common.chip_id = id;
        bp->link_params.chip_id = bp->common.chip_id;
@@ -7501,7 +7577,7 @@ static int __devinit bnx2x_init_bp(struct bnx2x *bp)
 
        mutex_init(&bp->port.phy_mutex);
 
-       INIT_WORK(&bp->sp_task, bnx2x_sp_task);
+       INIT_DELAYED_WORK(&bp->sp_task, bnx2x_sp_task);
        INIT_WORK(&bp->reset_task, bnx2x_reset_task);
 
        rc = bnx2x_get_hwinfo(bp);
@@ -8076,6 +8152,9 @@ static int bnx2x_get_eeprom(struct net_device *dev,
        struct bnx2x *bp = netdev_priv(dev);
        int rc;
 
+       if (!netif_running(dev))
+               return -EAGAIN;
+
        DP(BNX2X_MSG_NVM, "ethtool_eeprom: cmd %d\n"
           DP_LEVEL "  magic 0x%x  offset 0x%x (%d)  len 0x%x (%d)\n",
           eeprom->cmd, eeprom->magic, eeprom->offset, eeprom->offset,
@@ -8678,18 +8757,17 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode, u8 link_up)
 
        if (loopback_mode == BNX2X_MAC_LOOPBACK) {
                bp->link_params.loopback_mode = LOOPBACK_BMAC;
-               bnx2x_acquire_phy_lock(bp);
                bnx2x_phy_init(&bp->link_params, &bp->link_vars);
-               bnx2x_release_phy_lock(bp);
 
        } else if (loopback_mode == BNX2X_PHY_LOOPBACK) {
+               u16 cnt = 1000;
                bp->link_params.loopback_mode = LOOPBACK_XGXS_10;
-               bnx2x_acquire_phy_lock(bp);
                bnx2x_phy_init(&bp->link_params, &bp->link_vars);
-               bnx2x_release_phy_lock(bp);
                /* wait until link state is restored */
-               bnx2x_wait_for_link(bp, link_up);
-
+               if (link_up)
+                       while (cnt-- && bnx2x_test_link(&bp->link_params,
+                                                       &bp->link_vars))
+                               msleep(10);
        } else
                return -EINVAL;
 
@@ -8727,6 +8805,8 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode, u8 link_up)
        tx_bd->general_data = ((UNICAST_ADDRESS <<
                                ETH_TX_BD_ETH_ADDR_TYPE_SHIFT) | 1);
 
+       wmb();
+
        fp->hw_tx_prods->bds_prod =
                cpu_to_le16(le16_to_cpu(fp->hw_tx_prods->bds_prod) + 1);
        mb(); /* FW restriction: must not reorder writing nbd and packets */
@@ -8778,7 +8858,6 @@ test_loopback_rx_exit:
        /* Update producers */
        bnx2x_update_rx_prod(bp, fp, fp->rx_bd_prod, fp->rx_comp_prod,
                             fp->rx_sge_prod);
-       mmiowb(); /* keep prod updates ordered */
 
 test_loopback_exit:
        bp->link_params.loopback_mode = LOOPBACK_NONE;
@@ -8794,6 +8873,7 @@ static int bnx2x_test_loopback(struct bnx2x *bp, u8 link_up)
                return BNX2X_LOOPBACK_FAILED;
 
        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");
@@ -8805,6 +8885,7 @@ static int bnx2x_test_loopback(struct bnx2x *bp, u8 link_up)
                rc |= BNX2X_PHY_LOOPBACK_FAILED;
        }
 
+       bnx2x_release_phy_lock(bp);
        bnx2x_netif_start(bp);
 
        return rc;
@@ -8878,7 +8959,10 @@ static int bnx2x_test_intr(struct bnx2x *bp)
                return -ENODEV;
 
        config->hdr.length_6b = 0;
-       config->hdr.offset = 0;
+       if (CHIP_IS_E1(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.reserved1 = 0;
 
@@ -9243,6 +9327,18 @@ static int bnx2x_set_power_state(struct bnx2x *bp, pci_power_t state)
        return 0;
 }
 
+static inline int bnx2x_has_rx_work(struct bnx2x_fastpath *fp)
+{
+       u16 rx_cons_sb;
+
+       /* Tell compiler that status block fields can change */
+       barrier();
+       rx_cons_sb = le16_to_cpu(*fp->rx_cons_sb);
+       if ((rx_cons_sb & MAX_RCQ_DESC_CNT) == MAX_RCQ_DESC_CNT)
+               rx_cons_sb++;
+       return (fp->rx_comp_cons != rx_cons_sb);
+}
+
 /*
  * net_device service functions
  */
@@ -9253,7 +9349,6 @@ static int bnx2x_poll(struct napi_struct *napi, int budget)
                                                 napi);
        struct bnx2x *bp = fp->bp;
        int work_done = 0;
-       u16 rx_cons_sb;
 
 #ifdef BNX2X_STOP_ON_ERROR
        if (unlikely(bp->panic))
@@ -9266,19 +9361,12 @@ static int bnx2x_poll(struct napi_struct *napi, int budget)
 
        bnx2x_update_fpsb_idx(fp);
 
-       if (BNX2X_HAS_TX_WORK(fp))
+       if (bnx2x_has_tx_work(fp))
                bnx2x_tx_int(fp, budget);
 
-       rx_cons_sb = le16_to_cpu(*fp->rx_cons_sb);
-       if ((rx_cons_sb & MAX_RCQ_DESC_CNT) == MAX_RCQ_DESC_CNT)
-               rx_cons_sb++;
-       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 */
-       rx_cons_sb = le16_to_cpu(*fp->rx_cons_sb);
-       if ((rx_cons_sb & MAX_RCQ_DESC_CNT) == MAX_RCQ_DESC_CNT)
-               rx_cons_sb++;
 
        /* must not complete if we consumed full budget */
        if ((work_done < budget) && !BNX2X_HAS_WORK(fp)) {
@@ -9389,6 +9477,7 @@ static inline u32 bnx2x_xmit_type(struct bnx2x *bp, struct sk_buff *skb)
        return rc;
 }
 
+#if (MAX_SKB_FRAGS >= MAX_FETCH_BD - 3)
 /* check if packet requires linearization (packet is too fragmented) */
 static int bnx2x_pkt_req_lin(struct bnx2x *bp, struct sk_buff *skb,
                             u32 xmit_type)
@@ -9466,6 +9555,7 @@ exit_lbl:
 
        return to_copy;
 }
+#endif
 
 /* called with netif_tx_lock
  * bnx2x_tx_int() runs without netif_tx_lock unless it needs to call
@@ -9506,6 +9596,7 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
           skb->ip_summed, skb->protocol, ipv6_hdr(skb)->nexthdr,
           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) */
        if (bnx2x_pkt_req_lin(bp, skb, xmit_type)) {
@@ -9518,6 +9609,7 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
                        return NETDEV_TX_OK;
                }
        }
+#endif
 
        /*
        Please read carefully. First we use one BD which we mark as start,
@@ -9549,11 +9641,14 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
           "sending pkt %u @%p  next_idx %u  bd %u @%p\n",
           pkt_prod, tx_buf, fp->tx_pkt_prod, bd_prod, tx_bd);
 
-       if ((bp->vlgrp != NULL) && vlan_tx_tag_present(skb)) {
+#ifdef BCM_VLAN
+       if ((bp->vlgrp != NULL) && vlan_tx_tag_present(skb) &&
+           (bp->flags & HW_VLAN_TX_FLAG)) {
                tx_bd->vlan = cpu_to_le16(vlan_tx_tag_get(skb));
                tx_bd->bd_flags.as_bitfield |= ETH_TX_BD_FLAGS_VLAN_TAG;
                vlan_off += 4;
        } else
+#endif
                tx_bd->vlan = cpu_to_le16(pkt_prod);
 
        if (xmit_type) {
@@ -9705,6 +9800,15 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
        DP(NETIF_MSG_TX_QUEUED, "doorbell: nbd %d  bd %u\n", nbd, bd_prod);
 
+       /*
+        * Make sure that the BD data is updated before updating the producer
+        * since FW might read the BD right after the producer is updated.
+        * This is only applicable for weak-ordered memory model archs such
+        * as IA-64. The following barrier is also mandatory since FW will
+        * assumes packets must have BDs.
+        */
+       wmb();
+
        fp->hw_tx_prods->bds_prod =
                cpu_to_le16(le16_to_cpu(fp->hw_tx_prods->bds_prod) + nbd);
        mb(); /* FW restriction: must not reorder writing nbd and packets */
@@ -9718,6 +9822,9 @@ static int bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
        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
+                  if we put Tx into XOFF state. */
+               smp_mb();
                netif_stop_queue(dev);
                bp->eth_stats.driver_xoff++;
                if (bnx2x_tx_avail(fp) >= MAX_SKB_FRAGS + 3)
@@ -9733,6 +9840,8 @@ static int bnx2x_open(struct net_device *dev)
 {
        struct bnx2x *bp = netdev_priv(dev);
 
+       netif_carrier_off(dev);
+
        bnx2x_set_power_state(bp, PCI_D0);
 
        return bnx2x_nic_load(bp, LOAD_OPEN);
@@ -9816,7 +9925,7 @@ static void bnx2x_set_rx_mode(struct net_device *dev)
                                for (; i < old; i++) {
                                        if (CAM_IS_INVALID(config->
                                                           config_table[i])) {
-                                               i--; /* already invalidated */
+                                               /* already invalidated */
                                                break;
                                        }
                                        /* invalidate */
@@ -9987,6 +10096,16 @@ static void bnx2x_vlan_rx_register(struct net_device *dev,
        struct bnx2x *bp = netdev_priv(dev);
 
        bp->vlgrp = vlgrp;
+
+       /* Set flags according to the required capabilities */
+       bp->flags &= ~(HW_VLAN_RX_FLAG | HW_VLAN_TX_FLAG);
+
+       if (dev->features & NETIF_F_HW_VLAN_TX)
+               bp->flags |= HW_VLAN_TX_FLAG;
+
+       if (dev->features & NETIF_F_HW_VLAN_RX)
+               bp->flags |= HW_VLAN_RX_FLAG;
+
        if (netif_running(dev))
                bnx2x_set_client_config(bp);
 }
@@ -10143,6 +10262,7 @@ static int __devinit bnx2x_init_dev(struct pci_dev *pdev,
                dev->features |= NETIF_F_HIGHDMA;
 #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);
 #endif
        dev->features |= (NETIF_F_TSO | NETIF_F_TSO_ECN);
        dev->features |= NETIF_F_TSO6;
@@ -10215,22 +10335,18 @@ static int __devinit bnx2x_init_one(struct pci_dev *pdev,
                return rc;
        }
 
-       rc = register_netdev(dev);
-       if (rc) {
-               dev_err(&pdev->dev, "Cannot register net device\n");
-               goto init_one_exit;
-       }
-
        pci_set_drvdata(pdev, dev);
 
        rc = bnx2x_init_bp(bp);
+       if (rc)
+               goto init_one_exit;
+
+       rc = register_netdev(dev);
        if (rc) {
-               unregister_netdev(dev);
+               dev_err(&pdev->dev, "Cannot register net device\n");
                goto init_one_exit;
        }
 
-       netif_carrier_off(dev);
-
        bp->common.name = board_info[ent->driver_data].name;
        printk(KERN_INFO "%s: %s (%c%d) PCI-E x%d %s found at mem %lx,"
               " IRQ %d, ", dev->name, bp->common.name,
@@ -10378,6 +10494,8 @@ static int bnx2x_eeh_nic_unload(struct bnx2x *bp)
        bnx2x_free_skbs(bp);
        for_each_queue(bp, i)
                bnx2x_free_rx_sge_range(bp, bp->fp + i, NUM_RX_SGE);
+       for_each_queue(bp, i)
+               netif_napi_del(&bnx2x_fp(bp, i, napi));
        bnx2x_free_mem(bp);
 
        bp->state = BNX2X_STATE_CLOSED;
@@ -10519,12 +10637,20 @@ static struct pci_driver bnx2x_pci_driver = {
 
 static int __init bnx2x_init(void)
 {
+       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);
 }
 
 static void __exit bnx2x_cleanup(void)
 {
        pci_unregister_driver(&bnx2x_pci_driver);
+
+       destroy_workqueue(bnx2x_wq);
 }
 
 module_init(bnx2x_init);