Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6
[linux-2.6.git] / drivers / net / qlge / qlge_main.c
index 04bf212..707b391 100644 (file)
@@ -34,7 +34,6 @@
 #include <linux/etherdevice.h>
 #include <linux/ethtool.h>
 #include <linux/skbuff.h>
-#include <linux/rtnetlink.h>
 #include <linux/if_vlan.h>
 #include <linux/delay.h>
 #include <linux/mm.h>
@@ -58,8 +57,8 @@ static const u32 default_msg =
     NETIF_MSG_IFUP |
     NETIF_MSG_RX_ERR |
     NETIF_MSG_TX_ERR |
-    NETIF_MSG_TX_QUEUED |
-    NETIF_MSG_INTR | NETIF_MSG_TX_DONE | NETIF_MSG_RX_STATUS |
+/*  NETIF_MSG_TX_QUEUED | */
+/*  NETIF_MSG_INTR | NETIF_MSG_TX_DONE | NETIF_MSG_RX_STATUS | */
 /* NETIF_MSG_PKTDATA | */
     NETIF_MSG_HW | NETIF_MSG_WOL | 0;
 
@@ -70,12 +69,13 @@ MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 #define MSIX_IRQ 0
 #define MSI_IRQ 1
 #define LEG_IRQ 2
-static int irq_type = MSIX_IRQ;
-module_param(irq_type, int, MSIX_IRQ);
-MODULE_PARM_DESC(irq_type, "0 = MSI-X, 1 = MSI, 2 = Legacy.");
+static int qlge_irq_type = MSIX_IRQ;
+module_param(qlge_irq_type, int, MSIX_IRQ);
+MODULE_PARM_DESC(qlge_irq_type, "0 = MSI-X, 1 = MSI, 2 = Legacy.");
 
 static struct pci_device_id qlge_pci_tbl[] __devinitdata = {
-       {PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, QLGE_DEVICE_ID)},
+       {PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, QLGE_DEVICE_ID_8012)},
+       {PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, QLGE_DEVICE_ID_8000)},
        /* required last entry */
        {0,}
 };
@@ -213,6 +213,10 @@ int ql_write_cfg(struct ql_adapter *qdev, void *ptr, int size, u32 bit,
                return -ENOMEM;
        }
 
+       status = ql_sem_spinlock(qdev, SEM_ICB_MASK);
+       if (status)
+               return status;
+
        status = ql_wait_cfg(qdev, bit);
        if (status) {
                QPRINTK(qdev, IFUP, ERR,
@@ -220,12 +224,8 @@ int ql_write_cfg(struct ql_adapter *qdev, void *ptr, int size, u32 bit,
                goto exit;
        }
 
-       status = ql_sem_spinlock(qdev, SEM_ICB_MASK);
-       if (status)
-               goto exit;
        ql_write32(qdev, ICB_L, (u32) map);
        ql_write32(qdev, ICB_H, (u32) (map >> 32));
-       ql_sem_unlock(qdev, SEM_ICB_MASK);      /* does flush too */
 
        mask = CFG_Q_MASK | (bit << 16);
        value = bit | (q_id << CFG_Q_SHIFT);
@@ -236,6 +236,7 @@ int ql_write_cfg(struct ql_adapter *qdev, void *ptr, int size, u32 bit,
         */
        status = ql_wait_cfg(qdev, bit);
 exit:
+       ql_sem_unlock(qdev, SEM_ICB_MASK);      /* does flush too */
        pci_unmap_single(qdev->pdev, map, size, direction);
        return status;
 }
@@ -247,9 +248,6 @@ int ql_get_mac_addr_reg(struct ql_adapter *qdev, u32 type, u16 index,
        u32 offset = 0;
        int status;
 
-       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
-       if (status)
-               return status;
        switch (type) {
        case MAC_ADDR_TYPE_MULTI_MAC:
        case MAC_ADDR_TYPE_CAM_MAC:
@@ -308,7 +306,6 @@ int ql_get_mac_addr_reg(struct ql_adapter *qdev, u32 type, u16 index,
                status = -EPERM;
        }
 exit:
-       ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
        return status;
 }
 
@@ -321,11 +318,39 @@ static int ql_set_mac_addr_reg(struct ql_adapter *qdev, u8 *addr, u32 type,
        u32 offset = 0;
        int status = 0;
 
-       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
-       if (status)
-               return status;
        switch (type) {
        case MAC_ADDR_TYPE_MULTI_MAC:
+               {
+                       u32 upper = (addr[0] << 8) | addr[1];
+                       u32 lower = (addr[2] << 24) | (addr[3] << 16) |
+                                       (addr[4] << 8) | (addr[5]);
+
+                       status =
+                               ql_wait_reg_rdy(qdev,
+                               MAC_ADDR_IDX, MAC_ADDR_MW, 0);
+                       if (status)
+                               goto exit;
+                       ql_write32(qdev, MAC_ADDR_IDX, (offset++) |
+                               (index << MAC_ADDR_IDX_SHIFT) |
+                               type | MAC_ADDR_E);
+                       ql_write32(qdev, MAC_ADDR_DATA, lower);
+                       status =
+                               ql_wait_reg_rdy(qdev,
+                               MAC_ADDR_IDX, MAC_ADDR_MW, 0);
+                       if (status)
+                               goto exit;
+                       ql_write32(qdev, MAC_ADDR_IDX, (offset++) |
+                               (index << MAC_ADDR_IDX_SHIFT) |
+                               type | MAC_ADDR_E);
+
+                       ql_write32(qdev, MAC_ADDR_DATA, upper);
+                       status =
+                               ql_wait_reg_rdy(qdev,
+                               MAC_ADDR_IDX, MAC_ADDR_MW, 0);
+                       if (status)
+                               goto exit;
+                       break;
+               }
        case MAC_ADDR_TYPE_CAM_MAC:
                {
                        u32 cam_output;
@@ -334,7 +359,7 @@ static int ql_set_mac_addr_reg(struct ql_adapter *qdev, u8 *addr, u32 type,
                            (addr[2] << 24) | (addr[3] << 16) | (addr[4] << 8) |
                            (addr[5]);
 
-                       QPRINTK(qdev, IFUP, INFO,
+                       QPRINTK(qdev, IFUP, DEBUG,
                                "Adding %s address %pM"
                                " at index %d in the CAM.\n",
                                ((type ==
@@ -371,18 +396,14 @@ static int ql_set_mac_addr_reg(struct ql_adapter *qdev, u8 *addr, u32 type,
                           and possibly the function id.  Right now we hardcode
                           the route field to NIC core.
                         */
-                       if (type == MAC_ADDR_TYPE_CAM_MAC) {
-                               cam_output = (CAM_OUT_ROUTE_NIC |
-                                             (qdev->
-                                              func << CAM_OUT_FUNC_SHIFT) |
-                                             (qdev->
-                                              rss_ring_first_cq_id <<
-                                              CAM_OUT_CQ_ID_SHIFT));
-                               if (qdev->vlgrp)
-                                       cam_output |= CAM_OUT_RV;
-                               /* route to NIC core */
-                               ql_write32(qdev, MAC_ADDR_DATA, cam_output);
-                       }
+                       cam_output = (CAM_OUT_ROUTE_NIC |
+                                     (qdev->
+                                      func << CAM_OUT_FUNC_SHIFT) |
+                                       (0 << CAM_OUT_CQ_ID_SHIFT));
+                       if (qdev->vlgrp)
+                               cam_output |= CAM_OUT_RV;
+                       /* route to NIC core */
+                       ql_write32(qdev, MAC_ADDR_DATA, cam_output);
                        break;
                }
        case MAC_ADDR_TYPE_VLAN:
@@ -415,10 +436,60 @@ static int ql_set_mac_addr_reg(struct ql_adapter *qdev, u8 *addr, u32 type,
                status = -EPERM;
        }
 exit:
+       return status;
+}
+
+/* Set or clear MAC address in hardware. We sometimes
+ * have to clear it to prevent wrong frame routing
+ * especially in a bonding environment.
+ */
+static int ql_set_mac_addr(struct ql_adapter *qdev, int set)
+{
+       int status;
+       char zero_mac_addr[ETH_ALEN];
+       char *addr;
+
+       if (set) {
+               addr = &qdev->ndev->dev_addr[0];
+               QPRINTK(qdev, IFUP, DEBUG,
+                       "Set Mac addr %02x:%02x:%02x:%02x:%02x:%02x\n",
+                       addr[0], addr[1], addr[2], addr[3],
+                       addr[4], addr[5]);
+       } else {
+               memset(zero_mac_addr, 0, ETH_ALEN);
+               addr = &zero_mac_addr[0];
+               QPRINTK(qdev, IFUP, DEBUG,
+                               "Clearing MAC address on %s\n",
+                               qdev->ndev->name);
+       }
+       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
+       if (status)
+               return status;
+       status = ql_set_mac_addr_reg(qdev, (u8 *) addr,
+                       MAC_ADDR_TYPE_CAM_MAC, qdev->func * MAX_CQ);
        ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
+       if (status)
+               QPRINTK(qdev, IFUP, ERR, "Failed to init mac "
+                       "address.\n");
        return status;
 }
 
+void ql_link_on(struct ql_adapter *qdev)
+{
+       QPRINTK(qdev, LINK, ERR, "%s: Link is up.\n",
+                                qdev->ndev->name);
+       netif_carrier_on(qdev->ndev);
+       ql_set_mac_addr(qdev, 1);
+}
+
+void ql_link_off(struct ql_adapter *qdev)
+{
+       QPRINTK(qdev, LINK, ERR, "%s: Link is down.\n",
+                                qdev->ndev->name);
+       netif_carrier_off(qdev->ndev);
+       ql_set_mac_addr(qdev, 0);
+}
+
 /* Get a specific frame routing value from the CAM.
  * Used for debug and reg dump.
  */
@@ -426,10 +497,6 @@ int ql_get_routing_reg(struct ql_adapter *qdev, u32 index, u32 *value)
 {
        int status = 0;
 
-       status = ql_sem_spinlock(qdev, SEM_RT_IDX_MASK);
-       if (status)
-               goto exit;
-
        status = ql_wait_reg_rdy(qdev, RT_IDX, RT_IDX_MW, 0);
        if (status)
                goto exit;
@@ -441,7 +508,6 @@ int ql_get_routing_reg(struct ql_adapter *qdev, u32 index, u32 *value)
                goto exit;
        *value = ql_read32(qdev, RT_DATA);
 exit:
-       ql_sem_unlock(qdev, SEM_RT_IDX_MASK);
        return status;
 }
 
@@ -453,13 +519,9 @@ exit:
 static int ql_set_routing_reg(struct ql_adapter *qdev, u32 index, u32 mask,
                              int enable)
 {
-       int status;
+       int status = -EINVAL; /* Return error if no mask match. */
        u32 value = 0;
 
-       status = ql_sem_spinlock(qdev, SEM_RT_IDX_MASK);
-       if (status)
-               return status;
-
        QPRINTK(qdev, IFUP, DEBUG,
                "%s %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s mask %s the routing reg.\n",
                (enable ? "Adding" : "Removing"),
@@ -513,14 +575,14 @@ static int ql_set_routing_reg(struct ql_adapter *qdev, u32 index, u32 mask,
                }
        case RT_IDX_MCAST:      /* Pass up All Multicast frames. */
                {
-                       value = RT_IDX_DST_CAM_Q |      /* dest */
+                       value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_ALLMULTI_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_MCAST_MATCH:        /* Pass up matched Multicast frames. */
                {
-                       value = RT_IDX_DST_CAM_Q |      /* dest */
+                       value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_MCAST_MATCH_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
@@ -555,7 +617,6 @@ static int ql_set_routing_reg(struct ql_adapter *qdev, u32 index, u32 mask,
                ql_write32(qdev, RT_DATA, enable ? mask : 0);
        }
 exit:
-       ql_sem_unlock(qdev, SEM_RT_IDX_MASK);
        return status;
 }
 
@@ -604,7 +665,6 @@ u32 ql_enable_completion_interrupt(struct ql_adapter *qdev, u32 intr)
 static u32 ql_disable_completion_interrupt(struct ql_adapter *qdev, u32 intr)
 {
        u32 var = 0;
-       unsigned long hw_flags;
        struct intr_context *ctx;
 
        /* HW disables for us if we're MSIX multi interrupts and
@@ -614,14 +674,14 @@ static u32 ql_disable_completion_interrupt(struct ql_adapter *qdev, u32 intr)
                return 0;
 
        ctx = qdev->intr_context + intr;
-       spin_lock_irqsave(&qdev->hw_lock, hw_flags);
+       spin_lock(&qdev->hw_lock);
        if (!atomic_read(&ctx->irq_cnt)) {
                ql_write32(qdev, INTR_EN,
                ctx->intr_dis_mask);
                var = ql_read32(qdev, STS);
        }
        atomic_inc(&ctx->irq_cnt);
-       spin_unlock_irqrestore(&qdev->hw_lock, hw_flags);
+       spin_unlock(&qdev->hw_lock);
        return var;
 }
 
@@ -641,6 +701,28 @@ static void ql_enable_all_completion_interrupts(struct ql_adapter *qdev)
 
 }
 
+static int ql_validate_flash(struct ql_adapter *qdev, u32 size, const char *str)
+{
+       int status, i;
+       u16 csum = 0;
+       __le16 *flash = (__le16 *)&qdev->flash;
+
+       status = strncmp((char *)&qdev->flash, str, 4);
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR, "Invalid flash signature.\n");
+               return  status;
+       }
+
+       for (i = 0; i < size; i++)
+               csum += le16_to_cpu(*flash++);
+
+       if (csum)
+               QPRINTK(qdev, IFUP, ERR,
+                       "Invalid flash checksum, csum = 0x%.04x.\n", csum);
+
+       return csum;
+}
+
 static int ql_read_flash_word(struct ql_adapter *qdev, int offset, __le32 *data)
 {
        int status = 0;
@@ -665,23 +747,88 @@ exit:
        return status;
 }
 
-static int ql_get_flash_params(struct ql_adapter *qdev)
+static int ql_get_8000_flash_params(struct ql_adapter *qdev)
+{
+       u32 i, size;
+       int status;
+       __le32 *p = (__le32 *)&qdev->flash;
+       u32 offset;
+       u8 mac_addr[6];
+
+       /* Get flash offset for function and adjust
+        * for dword access.
+        */
+       if (!qdev->port)
+               offset = FUNC0_FLASH_OFFSET / sizeof(u32);
+       else
+               offset = FUNC1_FLASH_OFFSET / sizeof(u32);
+
+       if (ql_sem_spinlock(qdev, SEM_FLASH_MASK))
+               return -ETIMEDOUT;
+
+       size = sizeof(struct flash_params_8000) / sizeof(u32);
+       for (i = 0; i < size; i++, p++) {
+               status = ql_read_flash_word(qdev, i+offset, p);
+               if (status) {
+                       QPRINTK(qdev, IFUP, ERR, "Error reading flash.\n");
+                       goto exit;
+               }
+       }
+
+       status = ql_validate_flash(qdev,
+                       sizeof(struct flash_params_8000) / sizeof(u16),
+                       "8000");
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR, "Invalid flash.\n");
+               status = -EINVAL;
+               goto exit;
+       }
+
+       /* Extract either manufacturer or BOFM modified
+        * MAC address.
+        */
+       if (qdev->flash.flash_params_8000.data_type1 == 2)
+               memcpy(mac_addr,
+                       qdev->flash.flash_params_8000.mac_addr1,
+                       qdev->ndev->addr_len);
+       else
+               memcpy(mac_addr,
+                       qdev->flash.flash_params_8000.mac_addr,
+                       qdev->ndev->addr_len);
+
+       if (!is_valid_ether_addr(mac_addr)) {
+               QPRINTK(qdev, IFUP, ERR, "Invalid MAC address.\n");
+               status = -EINVAL;
+               goto exit;
+       }
+
+       memcpy(qdev->ndev->dev_addr,
+               mac_addr,
+               qdev->ndev->addr_len);
+
+exit:
+       ql_sem_unlock(qdev, SEM_FLASH_MASK);
+       return status;
+}
+
+static int ql_get_8012_flash_params(struct ql_adapter *qdev)
 {
        int i;
        int status;
        __le32 *p = (__le32 *)&qdev->flash;
        u32 offset = 0;
+       u32 size = sizeof(struct flash_params_8012) / sizeof(u32);
 
        /* Second function's parameters follow the first
         * function's.
         */
-       if (qdev->func)
-               offset = sizeof(qdev->flash) / sizeof(u32);
+       if (qdev->port)
+               offset = size;
 
        if (ql_sem_spinlock(qdev, SEM_FLASH_MASK))
                return -ETIMEDOUT;
 
-       for (i = 0; i < sizeof(qdev->flash) / sizeof(u32); i++, p++) {
+       for (i = 0; i < size; i++, p++) {
                status = ql_read_flash_word(qdev, i+offset, p);
                if (status) {
                        QPRINTK(qdev, IFUP, ERR, "Error reading flash.\n");
@@ -689,6 +836,25 @@ static int ql_get_flash_params(struct ql_adapter *qdev)
                }
 
        }
+
+       status = ql_validate_flash(qdev,
+                       sizeof(struct flash_params_8012) / sizeof(u16),
+                       "8012");
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR, "Invalid flash.\n");
+               status = -EINVAL;
+               goto exit;
+       }
+
+       if (!is_valid_ether_addr(qdev->flash.flash_params_8012.mac_addr)) {
+               status = -EINVAL;
+               goto exit;
+       }
+
+       memcpy(qdev->ndev->dev_addr,
+               qdev->flash.flash_params_8012.mac_addr,
+               qdev->ndev->addr_len);
+
 exit:
        ql_sem_unlock(qdev, SEM_FLASH_MASK);
        return status;
@@ -759,13 +925,32 @@ exit:
        return status;
 }
 
+static int ql_8000_port_initialize(struct ql_adapter *qdev)
+{
+       int status;
+       /*
+        * Get MPI firmware version for driver banner
+        * and ethool info.
+        */
+       status = ql_mb_about_fw(qdev);
+       if (status)
+               goto exit;
+       status = ql_mb_get_fw_state(qdev);
+       if (status)
+               goto exit;
+       /* Wake up a worker to get/set the TX/RX frame sizes. */
+       queue_delayed_work(qdev->workqueue, &qdev->mpi_port_cfg_work, 0);
+exit:
+       return status;
+}
+
 /* Take the MAC Core out of reset.
  * Enable statistics counting.
  * Take the transmitter/receiver out of reset.
  * This functionality may be done in the MPI firmware at a
  * later date.
  */
-static int ql_port_initialize(struct ql_adapter *qdev)
+static int ql_8012_port_initialize(struct ql_adapter *qdev)
 {
        int status = 0;
        u32 data;
@@ -840,6 +1025,11 @@ end:
        return status;
 }
 
+static inline unsigned int ql_lbq_block_size(struct ql_adapter *qdev)
+{
+       return PAGE_SIZE << qdev->lbq_buf_order;
+}
+
 /* Get the next large buffer. */
 static struct bq_desc *ql_get_curr_lbuf(struct rx_ring *rx_ring)
 {
@@ -851,6 +1041,28 @@ static struct bq_desc *ql_get_curr_lbuf(struct rx_ring *rx_ring)
        return lbq_desc;
 }
 
+static struct bq_desc *ql_get_curr_lchunk(struct ql_adapter *qdev,
+               struct rx_ring *rx_ring)
+{
+       struct bq_desc *lbq_desc = ql_get_curr_lbuf(rx_ring);
+
+       pci_dma_sync_single_for_cpu(qdev->pdev,
+                                       pci_unmap_addr(lbq_desc, mapaddr),
+                                   rx_ring->lbq_buf_size,
+                                       PCI_DMA_FROMDEVICE);
+
+       /* If it's the last chunk of our master page then
+        * we unmap it.
+        */
+       if ((lbq_desc->p.pg_chunk.offset + rx_ring->lbq_buf_size)
+                                       == ql_lbq_block_size(qdev))
+               pci_unmap_page(qdev->pdev,
+                               lbq_desc->p.pg_chunk.map,
+                               ql_lbq_block_size(qdev),
+                               PCI_DMA_FROMDEVICE);
+       return lbq_desc;
+}
+
 /* Get the next small buffer. */
 static struct bq_desc *ql_get_curr_sbuf(struct rx_ring *rx_ring)
 {
@@ -878,43 +1090,84 @@ static void ql_write_cq_idx(struct rx_ring *rx_ring)
        ql_write_db_reg(rx_ring->cnsmr_idx, rx_ring->cnsmr_idx_db_reg);
 }
 
+static int ql_get_next_chunk(struct ql_adapter *qdev, struct rx_ring *rx_ring,
+                                               struct bq_desc *lbq_desc)
+{
+       if (!rx_ring->pg_chunk.page) {
+               u64 map;
+               rx_ring->pg_chunk.page = alloc_pages(__GFP_COLD | __GFP_COMP |
+                                               GFP_ATOMIC,
+                                               qdev->lbq_buf_order);
+               if (unlikely(!rx_ring->pg_chunk.page)) {
+                       QPRINTK(qdev, DRV, ERR,
+                               "page allocation failed.\n");
+                       return -ENOMEM;
+               }
+               rx_ring->pg_chunk.offset = 0;
+               map = pci_map_page(qdev->pdev, rx_ring->pg_chunk.page,
+                                       0, ql_lbq_block_size(qdev),
+                                       PCI_DMA_FROMDEVICE);
+               if (pci_dma_mapping_error(qdev->pdev, map)) {
+                       __free_pages(rx_ring->pg_chunk.page,
+                                       qdev->lbq_buf_order);
+                       QPRINTK(qdev, DRV, ERR,
+                               "PCI mapping failed.\n");
+                       return -ENOMEM;
+               }
+               rx_ring->pg_chunk.map = map;
+               rx_ring->pg_chunk.va = page_address(rx_ring->pg_chunk.page);
+       }
+
+       /* Copy the current master pg_chunk info
+        * to the current descriptor.
+        */
+       lbq_desc->p.pg_chunk = rx_ring->pg_chunk;
+
+       /* Adjust the master page chunk for next
+        * buffer get.
+        */
+       rx_ring->pg_chunk.offset += rx_ring->lbq_buf_size;
+       if (rx_ring->pg_chunk.offset == ql_lbq_block_size(qdev)) {
+               rx_ring->pg_chunk.page = NULL;
+               lbq_desc->p.pg_chunk.last_flag = 1;
+       } else {
+               rx_ring->pg_chunk.va += rx_ring->lbq_buf_size;
+               get_page(rx_ring->pg_chunk.page);
+               lbq_desc->p.pg_chunk.last_flag = 0;
+       }
+       return 0;
+}
 /* Process (refill) a large buffer queue. */
 static void ql_update_lbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
 {
-       int clean_idx = rx_ring->lbq_clean_idx;
+       u32 clean_idx = rx_ring->lbq_clean_idx;
+       u32 start_idx = clean_idx;
        struct bq_desc *lbq_desc;
        u64 map;
        int i;
 
-       while (rx_ring->lbq_free_cnt > 16) {
+       while (rx_ring->lbq_free_cnt > 32) {
                for (i = 0; i < 16; i++) {
                        QPRINTK(qdev, RX_STATUS, DEBUG,
                                "lbq: try cleaning clean_idx = %d.\n",
                                clean_idx);
                        lbq_desc = &rx_ring->lbq[clean_idx];
-                       if (lbq_desc->p.lbq_page == NULL) {
-                               QPRINTK(qdev, RX_STATUS, DEBUG,
-                                       "lbq: getting new page for index %d.\n",
-                                       lbq_desc->index);
-                               lbq_desc->p.lbq_page = alloc_page(GFP_ATOMIC);
-                               if (lbq_desc->p.lbq_page == NULL) {
-                                       QPRINTK(qdev, RX_STATUS, ERR,
-                                               "Couldn't get a page.\n");
-                                       return;
-                               }
-                               map = pci_map_page(qdev->pdev,
-                                                  lbq_desc->p.lbq_page,
-                                                  0, PAGE_SIZE,
-                                                  PCI_DMA_FROMDEVICE);
-                               if (pci_dma_mapping_error(qdev->pdev, map)) {
-                                       QPRINTK(qdev, RX_STATUS, ERR,
-                                               "PCI mapping failed.\n");
+                       if (ql_get_next_chunk(qdev, rx_ring, lbq_desc)) {
+                               QPRINTK(qdev, IFUP, ERR,
+                                       "Could not get a page chunk.\n");
                                        return;
                                }
+
+                       map = lbq_desc->p.pg_chunk.map +
+                               lbq_desc->p.pg_chunk.offset;
                                pci_unmap_addr_set(lbq_desc, mapaddr, map);
-                               pci_unmap_len_set(lbq_desc, maplen, PAGE_SIZE);
+                       pci_unmap_len_set(lbq_desc, maplen,
+                                       rx_ring->lbq_buf_size);
                                *lbq_desc->addr = cpu_to_le64(map);
-                       }
+
+                       pci_dma_sync_single_for_device(qdev->pdev, map,
+                                               rx_ring->lbq_buf_size,
+                                               PCI_DMA_FROMDEVICE);
                        clean_idx++;
                        if (clean_idx == rx_ring->lbq_len)
                                clean_idx = 0;
@@ -924,19 +1177,23 @@ static void ql_update_lbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                rx_ring->lbq_prod_idx += 16;
                if (rx_ring->lbq_prod_idx == rx_ring->lbq_len)
                        rx_ring->lbq_prod_idx = 0;
+               rx_ring->lbq_free_cnt -= 16;
+       }
+
+       if (start_idx != clean_idx) {
                QPRINTK(qdev, RX_STATUS, DEBUG,
                        "lbq: updating prod idx = %d.\n",
                        rx_ring->lbq_prod_idx);
                ql_write_db_reg(rx_ring->lbq_prod_idx,
                                rx_ring->lbq_prod_idx_db_reg);
-               rx_ring->lbq_free_cnt -= 16;
        }
 }
 
 /* Process (refill) a small buffer queue. */
 static void ql_update_sbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
 {
-       int clean_idx = rx_ring->sbq_clean_idx;
+       u32 clean_idx = rx_ring->sbq_clean_idx;
+       u32 start_idx = clean_idx;
        struct bq_desc *sbq_desc;
        u64 map;
        int i;
@@ -953,7 +1210,7 @@ static void ql_update_sbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                                        sbq_desc->index);
                                sbq_desc->p.skb =
                                    netdev_alloc_skb(qdev->ndev,
-                                                    rx_ring->sbq_buf_size);
+                                                    SMALL_BUFFER_SIZE);
                                if (sbq_desc->p.skb == NULL) {
                                        QPRINTK(qdev, PROBE, ERR,
                                                "Couldn't get an skb.\n");
@@ -963,16 +1220,18 @@ static void ql_update_sbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                                skb_reserve(sbq_desc->p.skb, QLGE_SB_PAD);
                                map = pci_map_single(qdev->pdev,
                                                     sbq_desc->p.skb->data,
-                                                    rx_ring->sbq_buf_size /
-                                                    2, PCI_DMA_FROMDEVICE);
+                                                    rx_ring->sbq_buf_size,
+                                                    PCI_DMA_FROMDEVICE);
                                if (pci_dma_mapping_error(qdev->pdev, map)) {
                                        QPRINTK(qdev, IFUP, ERR, "PCI mapping failed.\n");
                                        rx_ring->sbq_clean_idx = clean_idx;
+                                       dev_kfree_skb_any(sbq_desc->p.skb);
+                                       sbq_desc->p.skb = NULL;
                                        return;
                                }
                                pci_unmap_addr_set(sbq_desc, mapaddr, map);
                                pci_unmap_len_set(sbq_desc, maplen,
-                                                 rx_ring->sbq_buf_size / 2);
+                                                 rx_ring->sbq_buf_size);
                                *sbq_desc->addr = cpu_to_le64(map);
                        }
 
@@ -984,13 +1243,15 @@ static void ql_update_sbq(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                rx_ring->sbq_prod_idx += 16;
                if (rx_ring->sbq_prod_idx == rx_ring->sbq_len)
                        rx_ring->sbq_prod_idx = 0;
+               rx_ring->sbq_free_cnt -= 16;
+       }
+
+       if (start_idx != clean_idx) {
                QPRINTK(qdev, RX_STATUS, DEBUG,
                        "sbq: updating prod idx = %d.\n",
                        rx_ring->sbq_prod_idx);
                ql_write_db_reg(rx_ring->sbq_prod_idx,
                                rx_ring->sbq_prod_idx_db_reg);
-
-               rx_ring->sbq_free_cnt -= 16;
        }
 }
 
@@ -1282,27 +1543,24 @@ static struct sk_buff *ql_build_rx_skb(struct ql_adapter *qdev,
                         * chain it to the header buffer's skb and let
                         * it rip.
                         */
-                       lbq_desc = ql_get_curr_lbuf(rx_ring);
-                       pci_unmap_page(qdev->pdev,
-                                      pci_unmap_addr(lbq_desc,
-                                                     mapaddr),
-                                      pci_unmap_len(lbq_desc, maplen),
-                                      PCI_DMA_FROMDEVICE);
+                       lbq_desc = ql_get_curr_lchunk(qdev, rx_ring);
                        QPRINTK(qdev, RX_STATUS, DEBUG,
-                               "Chaining page to skb.\n");
-                       skb_fill_page_desc(skb, 0, lbq_desc->p.lbq_page,
-                                          0, length);
+                               "Chaining page at offset = %d,"
+                               "for %d bytes  to skb.\n",
+                               lbq_desc->p.pg_chunk.offset, length);
+                       skb_fill_page_desc(skb, 0, lbq_desc->p.pg_chunk.page,
+                                               lbq_desc->p.pg_chunk.offset,
+                                               length);
                        skb->len += length;
                        skb->data_len += length;
                        skb->truesize += length;
-                       lbq_desc->p.lbq_page = NULL;
                } else {
                        /*
                         * The headers and data are in a single large buffer. We
                         * copy it to a new skb and let it go. This can happen with
                         * jumbo mtu on a non-TCP/UDP frame.
                         */
-                       lbq_desc = ql_get_curr_lbuf(rx_ring);
+                       lbq_desc = ql_get_curr_lchunk(qdev, rx_ring);
                        skb = netdev_alloc_skb(qdev->ndev, length);
                        if (skb == NULL) {
                                QPRINTK(qdev, PROBE, DEBUG,
@@ -1317,13 +1575,14 @@ static struct sk_buff *ql_build_rx_skb(struct ql_adapter *qdev,
                        skb_reserve(skb, NET_IP_ALIGN);
                        QPRINTK(qdev, RX_STATUS, DEBUG,
                                "%d bytes of headers and data in large. Chain page to new skb and pull tail.\n", length);
-                       skb_fill_page_desc(skb, 0, lbq_desc->p.lbq_page,
-                                          0, length);
+                       skb_fill_page_desc(skb, 0,
+                                               lbq_desc->p.pg_chunk.page,
+                                               lbq_desc->p.pg_chunk.offset,
+                                               length);
                        skb->len += length;
                        skb->data_len += length;
                        skb->truesize += length;
                        length -= length;
-                       lbq_desc->p.lbq_page = NULL;
                        __pskb_pull_tail(skb,
                                (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) ?
                                VLAN_ETH_HLEN : ETH_HLEN);
@@ -1340,8 +1599,7 @@ static struct sk_buff *ql_build_rx_skb(struct ql_adapter *qdev,
                 *         frames.  If the MTU goes up we could
                 *          eventually be in trouble.
                 */
-               int size, offset, i = 0;
-               __le64 *bq, bq_array[8];
+               int size, i = 0;
                sbq_desc = ql_get_curr_sbuf(rx_ring);
                pci_unmap_single(qdev->pdev,
                                 pci_unmap_addr(sbq_desc, mapaddr),
@@ -1360,37 +1618,25 @@ static struct sk_buff *ql_build_rx_skb(struct ql_adapter *qdev,
                        QPRINTK(qdev, RX_STATUS, DEBUG,
                                "%d bytes of headers & data in chain of large.\n", length);
                        skb = sbq_desc->p.skb;
-                       bq = &bq_array[0];
-                       memcpy(bq, skb->data, sizeof(bq_array));
                        sbq_desc->p.skb = NULL;
                        skb_reserve(skb, NET_IP_ALIGN);
-               } else {
-                       QPRINTK(qdev, RX_STATUS, DEBUG,
-                               "Headers in small, %d bytes of data in chain of large.\n", length);
-                       bq = (__le64 *)sbq_desc->p.skb->data;
                }
                while (length > 0) {
-                       lbq_desc = ql_get_curr_lbuf(rx_ring);
-                       pci_unmap_page(qdev->pdev,
-                                      pci_unmap_addr(lbq_desc,
-                                                     mapaddr),
-                                      pci_unmap_len(lbq_desc,
-                                                    maplen),
-                                      PCI_DMA_FROMDEVICE);
-                       size = (length < PAGE_SIZE) ? length : PAGE_SIZE;
-                       offset = 0;
+                       lbq_desc = ql_get_curr_lchunk(qdev, rx_ring);
+                       size = (length < rx_ring->lbq_buf_size) ? length :
+                               rx_ring->lbq_buf_size;
 
                        QPRINTK(qdev, RX_STATUS, DEBUG,
                                "Adding page %d to skb for %d bytes.\n",
                                i, size);
-                       skb_fill_page_desc(skb, i, lbq_desc->p.lbq_page,
-                                          offset, size);
+                       skb_fill_page_desc(skb, i,
+                                               lbq_desc->p.pg_chunk.page,
+                                               lbq_desc->p.pg_chunk.offset,
+                                               size);
                        skb->len += size;
                        skb->data_len += size;
                        skb->truesize += size;
                        length -= size;
-                       lbq_desc->p.lbq_page = NULL;
-                       bq++;
                        i++;
                }
                __pskb_pull_tail(skb, (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) ?
@@ -1406,6 +1652,8 @@ static void ql_process_mac_rx_intr(struct ql_adapter *qdev,
 {
        struct net_device *ndev = qdev->ndev;
        struct sk_buff *skb = NULL;
+       u16 vlan_id = (le16_to_cpu(ib_mac_rsp->vlan_id) &
+                       IB_MAC_IOCB_RSP_VLAN_MASK)
 
        QL_DUMP_IB_MAC_RSP(ib_mac_rsp);
 
@@ -1413,6 +1661,32 @@ static void ql_process_mac_rx_intr(struct ql_adapter *qdev,
        if (unlikely(!skb)) {
                QPRINTK(qdev, RX_STATUS, DEBUG,
                        "No skb available, drop packet.\n");
+               rx_ring->rx_dropped++;
+               return;
+       }
+
+       /* Frame error, so drop the packet. */
+       if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_ERR_MASK) {
+               QPRINTK(qdev, DRV, ERR, "Receive error, flags2 = 0x%x\n",
+                                       ib_mac_rsp->flags2);
+               dev_kfree_skb_any(skb);
+               rx_ring->rx_errors++;
+               return;
+       }
+
+       /* The max framesize filter on this chip is set higher than
+        * MTU since FCoE uses 2k frames.
+        */
+       if (skb->len > ndev->mtu + ETH_HLEN) {
+               dev_kfree_skb_any(skb);
+               rx_ring->rx_dropped++;
+               return;
+       }
+
+       /* loopback self test for ethtool */
+       if (test_bit(QL_SELFTEST, &qdev->flags)) {
+               ql_check_lb_frame(qdev, skb);
+               dev_kfree_skb_any(skb);
                return;
        }
 
@@ -1426,36 +1700,56 @@ static void ql_process_mac_rx_intr(struct ql_adapter *qdev,
                        IB_MAC_IOCB_RSP_M_REG ? "Registered" : "",
                        (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                        IB_MAC_IOCB_RSP_M_PROM ? "Promiscuous" : "");
+               rx_ring->rx_multicast++;
        }
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_P) {
                QPRINTK(qdev, RX_STATUS, DEBUG, "Promiscuous Packet.\n");
        }
-       if (ib_mac_rsp->flags1 & (IB_MAC_IOCB_RSP_IE | IB_MAC_IOCB_RSP_TE)) {
-               QPRINTK(qdev, RX_STATUS, ERR,
-                       "Bad checksum for this %s packet.\n",
-                       ((ib_mac_rsp->
-                         flags2 & IB_MAC_IOCB_RSP_T) ? "TCP" : "UDP"));
-               skb->ip_summed = CHECKSUM_NONE;
-       } else if (qdev->rx_csum &&
-                  ((ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T) ||
-                   ((ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_U) &&
-                    !(ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_NU)))) {
-               QPRINTK(qdev, RX_STATUS, DEBUG, "RX checksum done!\n");
-               skb->ip_summed = CHECKSUM_UNNECESSARY;
-       }
-       qdev->stats.rx_packets++;
-       qdev->stats.rx_bytes += skb->len;
+
        skb->protocol = eth_type_trans(skb, ndev);
-       skb_record_rx_queue(skb, rx_ring - &qdev->rx_ring[0]);
-       if (qdev->vlgrp && (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V)) {
-               QPRINTK(qdev, RX_STATUS, DEBUG,
-                       "Passing a VLAN packet upstream.\n");
-               vlan_hwaccel_rx(skb, qdev->vlgrp,
-                               le16_to_cpu(ib_mac_rsp->vlan_id));
+       skb->ip_summed = CHECKSUM_NONE;
+
+       /* If rx checksum is on, and there are no
+        * csum or frame errors.
+        */
+       if (qdev->rx_csum &&
+               !(ib_mac_rsp->flags1 & IB_MAC_CSUM_ERR_MASK)) {
+               /* TCP frame. */
+               if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T) {
+                       QPRINTK(qdev, RX_STATUS, DEBUG,
+                                       "TCP checksum done!\n");
+                       skb->ip_summed = CHECKSUM_UNNECESSARY;
+               } else if ((ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_U) &&
+                               (ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_V4)) {
+               /* Unfragmented ipv4 UDP frame. */
+                       struct iphdr *iph = (struct iphdr *) skb->data;
+                       if (!(iph->frag_off &
+                               cpu_to_be16(IP_MF|IP_OFFSET))) {
+                               skb->ip_summed = CHECKSUM_UNNECESSARY;
+                               QPRINTK(qdev, RX_STATUS, DEBUG,
+                                               "TCP checksum done!\n");
+                       }
+               }
+       }
+
+       rx_ring->rx_packets++;
+       rx_ring->rx_bytes += skb->len;
+       skb_record_rx_queue(skb, rx_ring->cq_id);
+       if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
+               if (qdev->vlgrp &&
+                       (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) &&
+                       (vlan_id != 0))
+                       vlan_gro_receive(&rx_ring->napi, qdev->vlgrp,
+                               vlan_id, skb);
+               else
+                       napi_gro_receive(&rx_ring->napi, skb);
        } else {
-               QPRINTK(qdev, RX_STATUS, DEBUG,
-                       "Passing a normal packet upstream.\n");
-               netif_rx(skb);
+               if (qdev->vlgrp &&
+                       (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) &&
+                       (vlan_id != 0))
+                       vlan_hwaccel_receive_skb(skb, qdev->vlgrp, vlan_id);
+               else
+                       netif_receive_skb(skb);
        }
 }
 
@@ -1470,8 +1764,8 @@ static void ql_process_mac_tx_intr(struct ql_adapter *qdev,
        tx_ring = &qdev->tx_ring[mac_rsp->txq_idx];
        tx_ring_desc = &tx_ring->q[mac_rsp->tid];
        ql_unmap_send(qdev, tx_ring_desc, tx_ring_desc->map_cnt);
-       qdev->stats.tx_bytes += tx_ring_desc->map_cnt;
-       qdev->stats.tx_packets++;
+       tx_ring->tx_bytes += (tx_ring_desc->skb)->len;
+       tx_ring->tx_packets++;
        dev_kfree_skb(tx_ring_desc->skb);
        tx_ring_desc->skb = NULL;
 
@@ -1502,16 +1796,19 @@ static void ql_process_mac_tx_intr(struct ql_adapter *qdev,
 /* Fire up a handler to reset the MPI processor. */
 void ql_queue_fw_error(struct ql_adapter *qdev)
 {
-       netif_stop_queue(qdev->ndev);
-       netif_carrier_off(qdev->ndev);
+       ql_link_off(qdev);
        queue_delayed_work(qdev->workqueue, &qdev->mpi_reset_work, 0);
 }
 
 void ql_queue_asic_error(struct ql_adapter *qdev)
 {
-       netif_stop_queue(qdev->ndev);
-       netif_carrier_off(qdev->ndev);
+       ql_link_off(qdev);
        ql_disable_interrupts(qdev);
+       /* Clear adapter up bit to signal the recovery
+        * process that it shouldn't kill the reset worker
+        * thread
+        */
+       clear_bit(QL_ADAPTER_UP, &qdev->flags);
        queue_delayed_work(qdev->workqueue, &qdev->asic_reset_work, 0);
 }
 
@@ -1559,6 +1856,7 @@ static int ql_clean_outbound_rx_ring(struct rx_ring *rx_ring)
        struct ob_mac_iocb_rsp *net_rsp = NULL;
        int count = 0;
 
+       struct tx_ring *tx_ring;
        /* While there are entries in the completion queue. */
        while (prod != rx_ring->cnsmr_idx) {
 
@@ -1584,15 +1882,16 @@ static int ql_clean_outbound_rx_ring(struct rx_ring *rx_ring)
                prod = ql_read_sh_reg(rx_ring->prod_idx_sh_reg);
        }
        ql_write_cq_idx(rx_ring);
-       if (netif_queue_stopped(qdev->ndev) && net_rsp != NULL) {
-               struct tx_ring *tx_ring = &qdev->tx_ring[net_rsp->txq_idx];
+       tx_ring = &qdev->tx_ring[net_rsp->txq_idx];
+       if (__netif_subqueue_stopped(qdev->ndev, tx_ring->wq_id) &&
+                                       net_rsp != NULL) {
                if (atomic_read(&tx_ring->queue_stopped) &&
                    (atomic_read(&tx_ring->tx_count) > (tx_ring->wq_len / 4)))
                        /*
                         * The queue got stopped because the tx_ring was full.
                         * Wake it up, because it's now at least 25% empty.
                         */
-                       netif_wake_queue(qdev->ndev);
+                       netif_wake_subqueue(qdev->ndev, tx_ring->wq_id);
        }
 
        return count;
@@ -1647,19 +1946,49 @@ static int ql_napi_poll_msix(struct napi_struct *napi, int budget)
 {
        struct rx_ring *rx_ring = container_of(napi, struct rx_ring, napi);
        struct ql_adapter *qdev = rx_ring->qdev;
-       int work_done = ql_clean_inbound_rx_ring(rx_ring, budget);
+       struct rx_ring *trx_ring;
+       int i, work_done = 0;
+       struct intr_context *ctx = &qdev->intr_context[rx_ring->cq_id];
 
        QPRINTK(qdev, RX_STATUS, DEBUG, "Enter, NAPI POLL cq_id = %d.\n",
                rx_ring->cq_id);
 
+       /* Service the TX rings first.  They start
+        * right after the RSS rings. */
+       for (i = qdev->rss_ring_count; i < qdev->rx_ring_count; i++) {
+               trx_ring = &qdev->rx_ring[i];
+               /* If this TX completion ring belongs to this vector and
+                * it's not empty then service it.
+                */
+               if ((ctx->irq_mask & (1 << trx_ring->cq_id)) &&
+                       (ql_read_sh_reg(trx_ring->prod_idx_sh_reg) !=
+                                       trx_ring->cnsmr_idx)) {
+                       QPRINTK(qdev, INTR, DEBUG,
+                               "%s: Servicing TX completion ring %d.\n",
+                               __func__, trx_ring->cq_id);
+                       ql_clean_outbound_rx_ring(trx_ring);
+               }
+       }
+
+       /*
+        * Now service the RSS ring if it's active.
+        */
+       if (ql_read_sh_reg(rx_ring->prod_idx_sh_reg) !=
+                                       rx_ring->cnsmr_idx) {
+               QPRINTK(qdev, INTR, DEBUG,
+                       "%s: Servicing RX completion ring %d.\n",
+                       __func__, rx_ring->cq_id);
+               work_done = ql_clean_inbound_rx_ring(rx_ring, budget);
+       }
+
        if (work_done < budget) {
-               __napi_complete(napi);
+               napi_complete(napi);
                ql_enable_completion_interrupt(qdev, rx_ring->irq);
        }
        return work_done;
 }
 
-static void ql_vlan_rx_register(struct net_device *ndev, struct vlan_group *grp)
+static void qlge_vlan_rx_register(struct net_device *ndev, struct vlan_group *grp)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
 
@@ -1675,63 +2004,38 @@ static void ql_vlan_rx_register(struct net_device *ndev, struct vlan_group *grp)
        }
 }
 
-static void ql_vlan_rx_add_vid(struct net_device *ndev, u16 vid)
+static void qlge_vlan_rx_add_vid(struct net_device *ndev, u16 vid)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
        u32 enable_bit = MAC_ADDR_E;
+       int status;
 
-       spin_lock(&qdev->hw_lock);
+       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
+       if (status)
+               return;
        if (ql_set_mac_addr_reg
            (qdev, (u8 *) &enable_bit, MAC_ADDR_TYPE_VLAN, vid)) {
                QPRINTK(qdev, IFUP, ERR, "Failed to init vlan address.\n");
        }
-       spin_unlock(&qdev->hw_lock);
+       ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
 }
 
-static void ql_vlan_rx_kill_vid(struct net_device *ndev, u16 vid)
+static void qlge_vlan_rx_kill_vid(struct net_device *ndev, u16 vid)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
        u32 enable_bit = 0;
+       int status;
+
+       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
+       if (status)
+               return;
 
-       spin_lock(&qdev->hw_lock);
        if (ql_set_mac_addr_reg
            (qdev, (u8 *) &enable_bit, MAC_ADDR_TYPE_VLAN, vid)) {
                QPRINTK(qdev, IFUP, ERR, "Failed to clear vlan address.\n");
        }
-       spin_unlock(&qdev->hw_lock);
-
-}
-
-/* Worker thread to process a given rx_ring that is dedicated
- * to outbound completions.
- */
-static void ql_tx_clean(struct work_struct *work)
-{
-       struct rx_ring *rx_ring =
-           container_of(work, struct rx_ring, rx_work.work);
-       ql_clean_outbound_rx_ring(rx_ring);
-       ql_enable_completion_interrupt(rx_ring->qdev, rx_ring->irq);
-
-}
-
-/* Worker thread to process a given rx_ring that is dedicated
- * to inbound completions.
- */
-static void ql_rx_clean(struct work_struct *work)
-{
-       struct rx_ring *rx_ring =
-           container_of(work, struct rx_ring, rx_work.work);
-       ql_clean_inbound_rx_ring(rx_ring, 64);
-       ql_enable_completion_interrupt(rx_ring->qdev, rx_ring->irq);
-}
+       ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
 
-/* MSI-X Multiple Vector Interrupt Handler for outbound completions. */
-static irqreturn_t qlge_msix_tx_isr(int irq, void *dev_id)
-{
-       struct rx_ring *rx_ring = dev_id;
-       queue_delayed_work_on(rx_ring->cpu, rx_ring->qdev->q_workqueue,
-                             &rx_ring->rx_work, 0);
-       return IRQ_HANDLED;
 }
 
 /* MSI-X Multiple Vector Interrupt Handler for inbound completions. */
@@ -1753,7 +2057,6 @@ static irqreturn_t qlge_isr(int irq, void *dev_id)
        struct ql_adapter *qdev = rx_ring->qdev;
        struct intr_context *intr_context = &qdev->intr_context[0];
        u32 var;
-       int i;
        int work_done = 0;
 
        spin_lock(&qdev->hw_lock);
@@ -1781,54 +2084,33 @@ static irqreturn_t qlge_isr(int irq, void *dev_id)
        /*
         * Check MPI processor activity.
         */
-       if (var & STS_PI) {
+       if ((var & STS_PI) &&
+               (ql_read32(qdev, INTR_MASK) & INTR_MASK_PI)) {
                /*
                 * We've got an async event or mailbox completion.
                 * Handle it and clear the source of the interrupt.
                 */
                QPRINTK(qdev, INTR, ERR, "Got MPI processor interrupt.\n");
                ql_disable_completion_interrupt(qdev, intr_context->intr);
-               queue_delayed_work_on(smp_processor_id(), qdev->workqueue,
-                                     &qdev->mpi_work, 0);
+               ql_write32(qdev, INTR_MASK, (INTR_MASK_PI << 16));
+               queue_delayed_work_on(smp_processor_id(),
+                               qdev->workqueue, &qdev->mpi_work, 0);
                work_done++;
        }
 
        /*
-        * Check the default queue and wake handler if active.
+        * Get the bit-mask that shows the active queues for this
+        * pass.  Compare it to the queues that this irq services
+        * and call napi if there's a match.
         */
-       rx_ring = &qdev->rx_ring[0];
-       if (ql_read_sh_reg(rx_ring->prod_idx_sh_reg) != rx_ring->cnsmr_idx) {
-               QPRINTK(qdev, INTR, INFO, "Waking handler for rx_ring[0].\n");
+       var = ql_read32(qdev, ISR1);
+       if (var & intr_context->irq_mask) {
+               QPRINTK(qdev, INTR, INFO,
+                       "Waking handler for rx_ring[0].\n");
                ql_disable_completion_interrupt(qdev, intr_context->intr);
-               queue_delayed_work_on(smp_processor_id(), qdev->q_workqueue,
-                                     &rx_ring->rx_work, 0);
+               napi_schedule(&rx_ring->napi);
                work_done++;
        }
-
-       if (!test_bit(QL_MSIX_ENABLED, &qdev->flags)) {
-               /*
-                * Start the DPC for each active queue.
-                */
-               for (i = 1; i < qdev->rx_ring_count; i++) {
-                       rx_ring = &qdev->rx_ring[i];
-                       if (ql_read_sh_reg(rx_ring->prod_idx_sh_reg) !=
-                           rx_ring->cnsmr_idx) {
-                               QPRINTK(qdev, INTR, INFO,
-                                       "Waking handler for rx_ring[%d].\n", i);
-                               ql_disable_completion_interrupt(qdev,
-                                                               intr_context->
-                                                               intr);
-                               if (i < qdev->rss_ring_first_cq_id)
-                                       queue_delayed_work_on(rx_ring->cpu,
-                                                             qdev->q_workqueue,
-                                                             &rx_ring->rx_work,
-                                                             0);
-                               else
-                                       napi_schedule(&rx_ring->napi);
-                               work_done++;
-                       }
-               }
-       }
        ql_enable_completion_interrupt(qdev, intr_context->intr);
        return work_done ? IRQ_HANDLED : IRQ_NONE;
 }
@@ -1906,32 +2188,32 @@ static void ql_hw_csum_setup(struct sk_buff *skb,
                                    iph->daddr, len, iph->protocol, 0);
 }
 
-static int qlge_send(struct sk_buff *skb, struct net_device *ndev)
+static netdev_tx_t qlge_send(struct sk_buff *skb, struct net_device *ndev)
 {
        struct tx_ring_desc *tx_ring_desc;
        struct ob_mac_iocb_req *mac_iocb_ptr;
        struct ql_adapter *qdev = netdev_priv(ndev);
        int tso;
        struct tx_ring *tx_ring;
-       u32 tx_ring_idx = (u32) QL_TXQ_IDX(qdev, skb);
+       u32 tx_ring_idx = (u32) skb->queue_mapping;
 
        tx_ring = &qdev->tx_ring[tx_ring_idx];
 
+       if (skb_padto(skb, ETH_ZLEN))
+               return NETDEV_TX_OK;
+
        if (unlikely(atomic_read(&tx_ring->tx_count) < 2)) {
                QPRINTK(qdev, TX_QUEUED, INFO,
                        "%s: shutting down tx queue %d du to lack of resources.\n",
                        __func__, tx_ring_idx);
-               netif_stop_queue(ndev);
+               netif_stop_subqueue(ndev, tx_ring->wq_id);
                atomic_inc(&tx_ring->queue_stopped);
+               tx_ring->tx_errors++;
                return NETDEV_TX_BUSY;
        }
        tx_ring_desc = &tx_ring->q[tx_ring->prod_idx];
        mac_iocb_ptr = tx_ring_desc->queue_entry;
-       memset((void *)mac_iocb_ptr, 0, sizeof(mac_iocb_ptr));
-       if (ql_map_send(qdev, mac_iocb_ptr, skb, tx_ring_desc) != NETDEV_TX_OK) {
-               QPRINTK(qdev, TX_QUEUED, ERR, "Could not map the segments.\n");
-               return NETDEV_TX_BUSY;
-       }
+       memset((void *)mac_iocb_ptr, 0, sizeof(*mac_iocb_ptr));
 
        mac_iocb_ptr->opcode = OPCODE_OB_MAC_IOCB;
        mac_iocb_ptr->tid = tx_ring_desc->index;
@@ -1957,6 +2239,13 @@ static int qlge_send(struct sk_buff *skb, struct net_device *ndev)
                ql_hw_csum_setup(skb,
                                 (struct ob_mac_tso_iocb_req *)mac_iocb_ptr);
        }
+       if (ql_map_send(qdev, mac_iocb_ptr, skb, tx_ring_desc) !=
+                       NETDEV_TX_OK) {
+               QPRINTK(qdev, TX_QUEUED, ERR,
+                               "Could not map the segments.\n");
+               tx_ring->tx_errors++;
+               return NETDEV_TX_BUSY;
+       }
        QL_DUMP_OB_MAC_IOCB(mac_iocb_ptr);
        tx_ring->prod_idx++;
        if (tx_ring->prod_idx == tx_ring->wq_len)
@@ -1964,7 +2253,6 @@ static int qlge_send(struct sk_buff *skb, struct net_device *ndev)
        wmb();
 
        ql_write_db_reg(tx_ring->prod_idx, tx_ring->prod_idx_db_reg);
-       ndev->trans_start = jiffies;
        QPRINTK(qdev, TX_QUEUED, DEBUG, "tx queued, slot %d, len %d\n",
                tx_ring->prod_idx, skb->len);
 
@@ -1972,6 +2260,7 @@ static int qlge_send(struct sk_buff *skb, struct net_device *ndev)
        return NETDEV_TX_OK;
 }
 
+
 static void ql_free_shadow_space(struct ql_adapter *qdev)
 {
        if (qdev->rx_ring_shadow_reg_area) {
@@ -2000,6 +2289,7 @@ static int ql_alloc_shadow_space(struct ql_adapter *qdev)
                        "Allocation of RX shadow space failed.\n");
                return -ENOMEM;
        }
+       memset(qdev->rx_ring_shadow_reg_area, 0, PAGE_SIZE);
        qdev->tx_ring_shadow_reg_area =
            pci_alloc_consistent(qdev->pdev, PAGE_SIZE,
                                 &qdev->tx_ring_shadow_reg_dma);
@@ -2008,6 +2298,7 @@ static int ql_alloc_shadow_space(struct ql_adapter *qdev)
                        "Allocation of TX shadow space failed.\n");
                goto err_wqp_sh_area;
        }
+       memset(qdev->tx_ring_shadow_reg_area, 0, PAGE_SIZE);
        return 0;
 
 err_wqp_sh_area:
@@ -2056,8 +2347,8 @@ static int ql_alloc_tx_resources(struct ql_adapter *qdev,
            pci_alloc_consistent(qdev->pdev, tx_ring->wq_size,
                                 &tx_ring->wq_base_dma);
 
-       if ((tx_ring->wq_base == NULL)
-           || tx_ring->wq_base_dma & (tx_ring->wq_size - 1)) {
+       if ((tx_ring->wq_base == NULL) ||
+           tx_ring->wq_base_dma & WQ_ADDR_ALIGN) {
                QPRINTK(qdev, IFUP, ERR, "tx_ring alloc failed.\n");
                return -ENOMEM;
        }
@@ -2075,62 +2366,30 @@ err:
 
 static void ql_free_lbq_buffers(struct ql_adapter *qdev, struct rx_ring *rx_ring)
 {
-       int i;
        struct bq_desc *lbq_desc;
 
-       for (i = 0; i < rx_ring->lbq_len; i++) {
-               lbq_desc = &rx_ring->lbq[i];
-               if (lbq_desc->p.lbq_page) {
+       uint32_t  curr_idx, clean_idx;
+
+       curr_idx = rx_ring->lbq_curr_idx;
+       clean_idx = rx_ring->lbq_clean_idx;
+       while (curr_idx != clean_idx) {
+               lbq_desc = &rx_ring->lbq[curr_idx];
+
+               if (lbq_desc->p.pg_chunk.last_flag) {
                        pci_unmap_page(qdev->pdev,
-                                      pci_unmap_addr(lbq_desc, mapaddr),
-                                      pci_unmap_len(lbq_desc, maplen),
+                               lbq_desc->p.pg_chunk.map,
+                               ql_lbq_block_size(qdev),
                                       PCI_DMA_FROMDEVICE);
-
-                       put_page(lbq_desc->p.lbq_page);
-                       lbq_desc->p.lbq_page = NULL;
+                       lbq_desc->p.pg_chunk.last_flag = 0;
                }
-       }
-}
 
-/*
- * Allocate and map a page for each element of the lbq.
- */
-static int ql_alloc_lbq_buffers(struct ql_adapter *qdev,
-                               struct rx_ring *rx_ring)
-{
-       int i;
-       struct bq_desc *lbq_desc;
-       u64 map;
-       __le64 *bq = rx_ring->lbq_base;
+               put_page(lbq_desc->p.pg_chunk.page);
+               lbq_desc->p.pg_chunk.page = NULL;
+
+               if (++curr_idx == rx_ring->lbq_len)
+                       curr_idx = 0;
 
-       for (i = 0; i < rx_ring->lbq_len; i++) {
-               lbq_desc = &rx_ring->lbq[i];
-               memset(lbq_desc, 0, sizeof(lbq_desc));
-               lbq_desc->addr = bq;
-               lbq_desc->index = i;
-               lbq_desc->p.lbq_page = alloc_page(GFP_ATOMIC);
-               if (unlikely(!lbq_desc->p.lbq_page)) {
-                       QPRINTK(qdev, IFUP, ERR, "failed alloc_page().\n");
-                       goto mem_error;
-               } else {
-                       map = pci_map_page(qdev->pdev,
-                                          lbq_desc->p.lbq_page,
-                                          0, PAGE_SIZE, PCI_DMA_FROMDEVICE);
-                       if (pci_dma_mapping_error(qdev->pdev, map)) {
-                               QPRINTK(qdev, IFUP, ERR,
-                                       "PCI mapping failed.\n");
-                               goto mem_error;
-                       }
-                       pci_unmap_addr_set(lbq_desc, mapaddr, map);
-                       pci_unmap_len_set(lbq_desc, maplen, PAGE_SIZE);
-                       *lbq_desc->addr = cpu_to_le64(map);
-               }
-               bq++;
        }
-       return 0;
-mem_error:
-       ql_free_lbq_buffers(qdev, rx_ring);
-       return -ENOMEM;
 }
 
 static void ql_free_sbq_buffers(struct ql_adapter *qdev, struct rx_ring *rx_ring)
@@ -2155,63 +2414,72 @@ static void ql_free_sbq_buffers(struct ql_adapter *qdev, struct rx_ring *rx_ring
        }
 }
 
-/* Allocate and map an skb for each element of the sbq. */
-static int ql_alloc_sbq_buffers(struct ql_adapter *qdev,
+/* Free all large and small rx buffers associated
+ * with the completion queues for this device.
+ */
+static void ql_free_rx_buffers(struct ql_adapter *qdev)
+{
+       int i;
+       struct rx_ring *rx_ring;
+
+       for (i = 0; i < qdev->rx_ring_count; i++) {
+               rx_ring = &qdev->rx_ring[i];
+               if (rx_ring->lbq)
+                       ql_free_lbq_buffers(qdev, rx_ring);
+               if (rx_ring->sbq)
+                       ql_free_sbq_buffers(qdev, rx_ring);
+       }
+}
+
+static void ql_alloc_rx_buffers(struct ql_adapter *qdev)
+{
+       struct rx_ring *rx_ring;
+       int i;
+
+       for (i = 0; i < qdev->rx_ring_count; i++) {
+               rx_ring = &qdev->rx_ring[i];
+               if (rx_ring->type != TX_Q)
+                       ql_update_buffer_queues(qdev, rx_ring);
+       }
+}
+
+static void ql_init_lbq_ring(struct ql_adapter *qdev,
+                               struct rx_ring *rx_ring)
+{
+       int i;
+       struct bq_desc *lbq_desc;
+       __le64 *bq = rx_ring->lbq_base;
+
+       memset(rx_ring->lbq, 0, rx_ring->lbq_len * sizeof(struct bq_desc));
+       for (i = 0; i < rx_ring->lbq_len; i++) {
+               lbq_desc = &rx_ring->lbq[i];
+               memset(lbq_desc, 0, sizeof(*lbq_desc));
+               lbq_desc->index = i;
+               lbq_desc->addr = bq;
+               bq++;
+       }
+}
+
+static void ql_init_sbq_ring(struct ql_adapter *qdev,
                                struct rx_ring *rx_ring)
 {
        int i;
        struct bq_desc *sbq_desc;
-       struct sk_buff *skb;
-       u64 map;
        __le64 *bq = rx_ring->sbq_base;
 
+       memset(rx_ring->sbq, 0, rx_ring->sbq_len * sizeof(struct bq_desc));
        for (i = 0; i < rx_ring->sbq_len; i++) {
                sbq_desc = &rx_ring->sbq[i];
-               memset(sbq_desc, 0, sizeof(sbq_desc));
+               memset(sbq_desc, 0, sizeof(*sbq_desc));
                sbq_desc->index = i;
                sbq_desc->addr = bq;
-               skb = netdev_alloc_skb(qdev->ndev, rx_ring->sbq_buf_size);
-               if (unlikely(!skb)) {
-                       /* Better luck next round */
-                       QPRINTK(qdev, IFUP, ERR,
-                               "small buff alloc failed for %d bytes at index %d.\n",
-                               rx_ring->sbq_buf_size, i);
-                       goto mem_err;
-               }
-               skb_reserve(skb, QLGE_SB_PAD);
-               sbq_desc->p.skb = skb;
-               /*
-                * Map only half the buffer. Because the
-                * other half may get some data copied to it
-                * when the completion arrives.
-                */
-               map = pci_map_single(qdev->pdev,
-                                    skb->data,
-                                    rx_ring->sbq_buf_size / 2,
-                                    PCI_DMA_FROMDEVICE);
-               if (pci_dma_mapping_error(qdev->pdev, map)) {
-                       QPRINTK(qdev, IFUP, ERR, "PCI mapping failed.\n");
-                       goto mem_err;
-               }
-               pci_unmap_addr_set(sbq_desc, mapaddr, map);
-               pci_unmap_len_set(sbq_desc, maplen, rx_ring->sbq_buf_size / 2);
-               *sbq_desc->addr = cpu_to_le64(map);
                bq++;
        }
-       return 0;
-mem_err:
-       ql_free_sbq_buffers(qdev, rx_ring);
-       return -ENOMEM;
 }
 
 static void ql_free_rx_resources(struct ql_adapter *qdev,
                                 struct rx_ring *rx_ring)
 {
-       if (rx_ring->sbq_len)
-               ql_free_sbq_buffers(qdev, rx_ring);
-       if (rx_ring->lbq_len)
-               ql_free_lbq_buffers(qdev, rx_ring);
-
        /* Free the small buffer queue. */
        if (rx_ring->sbq_base) {
                pci_free_consistent(qdev->pdev,
@@ -2289,11 +2557,7 @@ static int ql_alloc_rx_resources(struct ql_adapter *qdev,
                        goto err_mem;
                }
 
-               if (ql_alloc_sbq_buffers(qdev, rx_ring)) {
-                       QPRINTK(qdev, IFUP, ERR,
-                               "Small buffer allocation failed.\n");
-                       goto err_mem;
-               }
+               ql_init_sbq_ring(qdev, rx_ring);
        }
 
        if (rx_ring->lbq_len) {
@@ -2321,14 +2585,7 @@ static int ql_alloc_rx_resources(struct ql_adapter *qdev,
                        goto err_mem;
                }
 
-               /*
-                * Allocate the buffers.
-                */
-               if (ql_alloc_lbq_buffers(qdev, rx_ring)) {
-                       QPRINTK(qdev, IFUP, ERR,
-                               "Large buffer allocation failed.\n");
-                       goto err_mem;
-               }
+               ql_init_lbq_ring(qdev, rx_ring);
        }
 
        return 0;
@@ -2415,23 +2672,27 @@ static int ql_start_rx_ring(struct ql_adapter *qdev, struct rx_ring *rx_ring)
 {
        struct cqicb *cqicb = &rx_ring->cqicb;
        void *shadow_reg = qdev->rx_ring_shadow_reg_area +
-           (rx_ring->cq_id * sizeof(u64) * 4);
+               (rx_ring->cq_id * RX_RING_SHADOW_SPACE);
        u64 shadow_reg_dma = qdev->rx_ring_shadow_reg_dma +
-           (rx_ring->cq_id * sizeof(u64) * 4);
+               (rx_ring->cq_id * RX_RING_SHADOW_SPACE);
        void __iomem *doorbell_area =
            qdev->doorbell_area + (DB_PAGE_SIZE * (128 + rx_ring->cq_id));
        int err = 0;
        u16 bq_len;
+       u64 tmp;
+       __le64 *base_indirect_ptr;
+       int page_entries;
 
        /* Set up the shadow registers for this ring. */
        rx_ring->prod_idx_sh_reg = shadow_reg;
        rx_ring->prod_idx_sh_reg_dma = shadow_reg_dma;
+       *rx_ring->prod_idx_sh_reg = 0;
        shadow_reg += sizeof(u64);
        shadow_reg_dma += sizeof(u64);
        rx_ring->lbq_base_indirect = shadow_reg;
        rx_ring->lbq_base_indirect_dma = shadow_reg_dma;
-       shadow_reg += sizeof(u64);
-       shadow_reg_dma += sizeof(u64);
+       shadow_reg += (sizeof(u64) * MAX_DB_PAGES_PER_BQ(rx_ring->lbq_len));
+       shadow_reg_dma += (sizeof(u64) * MAX_DB_PAGES_PER_BQ(rx_ring->lbq_len));
        rx_ring->sbq_base_indirect = shadow_reg;
        rx_ring->sbq_base_indirect_dma = shadow_reg_dma;
 
@@ -2467,7 +2728,15 @@ static int ql_start_rx_ring(struct ql_adapter *qdev, struct rx_ring *rx_ring)
            FLAGS_LI;           /* Load irq delay values */
        if (rx_ring->lbq_len) {
                cqicb->flags |= FLAGS_LL;       /* Load lbq values */
-               *((u64 *) rx_ring->lbq_base_indirect) = rx_ring->lbq_base_dma;
+               tmp = (u64)rx_ring->lbq_base_dma;
+               base_indirect_ptr = (__le64 *) rx_ring->lbq_base_indirect;
+               page_entries = 0;
+               do {
+                       *base_indirect_ptr = cpu_to_le64(tmp);
+                       tmp += DB_PAGE_SIZE;
+                       base_indirect_ptr++;
+                       page_entries++;
+               } while (page_entries < MAX_DB_PAGES_PER_BQ(rx_ring->lbq_len));
                cqicb->lbq_addr =
                    cpu_to_le64(rx_ring->lbq_base_indirect_dma);
                bq_len = (rx_ring->lbq_buf_size == 65536) ? 0 :
@@ -2476,57 +2745,39 @@ static int ql_start_rx_ring(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                bq_len = (rx_ring->lbq_len == 65536) ? 0 :
                        (u16) rx_ring->lbq_len;
                cqicb->lbq_len = cpu_to_le16(bq_len);
-               rx_ring->lbq_prod_idx = rx_ring->lbq_len - 16;
+               rx_ring->lbq_prod_idx = 0;
                rx_ring->lbq_curr_idx = 0;
-               rx_ring->lbq_clean_idx = rx_ring->lbq_prod_idx;
-               rx_ring->lbq_free_cnt = 16;
+               rx_ring->lbq_clean_idx = 0;
+               rx_ring->lbq_free_cnt = rx_ring->lbq_len;
        }
        if (rx_ring->sbq_len) {
                cqicb->flags |= FLAGS_LS;       /* Load sbq values */
-               *((u64 *) rx_ring->sbq_base_indirect) = rx_ring->sbq_base_dma;
+               tmp = (u64)rx_ring->sbq_base_dma;
+               base_indirect_ptr = (__le64 *) rx_ring->sbq_base_indirect;
+               page_entries = 0;
+               do {
+                       *base_indirect_ptr = cpu_to_le64(tmp);
+                       tmp += DB_PAGE_SIZE;
+                       base_indirect_ptr++;
+                       page_entries++;
+               } while (page_entries < MAX_DB_PAGES_PER_BQ(rx_ring->sbq_len));
                cqicb->sbq_addr =
                    cpu_to_le64(rx_ring->sbq_base_indirect_dma);
                cqicb->sbq_buf_size =
-                   cpu_to_le16(((rx_ring->sbq_buf_size / 2) + 8) & 0xfffffff8);
+                   cpu_to_le16((u16)(rx_ring->sbq_buf_size));
                bq_len = (rx_ring->sbq_len == 65536) ? 0 :
                        (u16) rx_ring->sbq_len;
                cqicb->sbq_len = cpu_to_le16(bq_len);
-               rx_ring->sbq_prod_idx = rx_ring->sbq_len - 16;
+               rx_ring->sbq_prod_idx = 0;
                rx_ring->sbq_curr_idx = 0;
-               rx_ring->sbq_clean_idx = rx_ring->sbq_prod_idx;
-               rx_ring->sbq_free_cnt = 16;
+               rx_ring->sbq_clean_idx = 0;
+               rx_ring->sbq_free_cnt = rx_ring->sbq_len;
        }
        switch (rx_ring->type) {
        case TX_Q:
-               /* If there's only one interrupt, then we use
-                * worker threads to process the outbound
-                * completion handling rx_rings. We do this so
-                * they can be run on multiple CPUs. There is
-                * room to play with this more where we would only
-                * run in a worker if there are more than x number
-                * of outbound completions on the queue and more
-                * than one queue active.  Some threshold that
-                * would indicate a benefit in spite of the cost
-                * of a context switch.
-                * If there's more than one interrupt, then the
-                * outbound completions are processed in the ISR.
-                */
-               if (!test_bit(QL_MSIX_ENABLED, &qdev->flags))
-                       INIT_DELAYED_WORK(&rx_ring->rx_work, ql_tx_clean);
-               else {
-                       /* With all debug warnings on we see a WARN_ON message
-                        * when we free the skb in the interrupt context.
-                        */
-                       INIT_DELAYED_WORK(&rx_ring->rx_work, ql_tx_clean);
-               }
                cqicb->irq_delay = cpu_to_le16(qdev->tx_coalesce_usecs);
                cqicb->pkt_delay = cpu_to_le16(qdev->tx_max_coalesced_frames);
                break;
-       case DEFAULT_Q:
-               INIT_DELAYED_WORK(&rx_ring->rx_work, ql_rx_clean);
-               cqicb->irq_delay = 0;
-               cqicb->pkt_delay = 0;
-               break;
        case RX_Q:
                /* Inbound completion handling rx_rings run in
                 * separate NAPI contexts.
@@ -2540,24 +2791,13 @@ static int ql_start_rx_ring(struct ql_adapter *qdev, struct rx_ring *rx_ring)
                QPRINTK(qdev, IFUP, DEBUG, "Invalid rx_ring->type = %d.\n",
                        rx_ring->type);
        }
-       QPRINTK(qdev, IFUP, INFO, "Initializing rx work queue.\n");
+       QPRINTK(qdev, IFUP, DEBUG, "Initializing rx work queue.\n");
        err = ql_write_cfg(qdev, cqicb, sizeof(struct cqicb),
                           CFG_LCQ, rx_ring->cq_id);
        if (err) {
                QPRINTK(qdev, IFUP, ERR, "Failed to load CQICB.\n");
                return err;
        }
-       QPRINTK(qdev, IFUP, INFO, "Successfully loaded CQICB.\n");
-       /*
-        * Advance the producer index for the buffer queues.
-        */
-       wmb();
-       if (rx_ring->lbq_len)
-               ql_write_db_reg(rx_ring->lbq_prod_idx,
-                               rx_ring->lbq_prod_idx_db_reg);
-       if (rx_ring->sbq_len)
-               ql_write_db_reg(rx_ring->sbq_prod_idx,
-                               rx_ring->sbq_prod_idx_db_reg);
        return err;
 }
 
@@ -2598,13 +2838,13 @@ static int ql_start_tx_ring(struct ql_adapter *qdev, struct tx_ring *tx_ring)
 
        ql_init_tx_ring(qdev, tx_ring);
 
-       err = ql_write_cfg(qdev, wqicb, sizeof(wqicb), CFG_LRQ,
+       err = ql_write_cfg(qdev, wqicb, sizeof(*wqicb), CFG_LRQ,
                           (u16) tx_ring->wq_id);
        if (err) {
                QPRINTK(qdev, IFUP, ERR, "Failed to load tx_ring.\n");
                return err;
        }
-       QPRINTK(qdev, IFUP, INFO, "Successfully loaded WQICB.\n");
+       QPRINTK(qdev, IFUP, DEBUG, "Successfully loaded WQICB.\n");
        return err;
 }
 
@@ -2621,45 +2861,58 @@ static void ql_disable_msix(struct ql_adapter *qdev)
        }
 }
 
+/* We start by trying to get the number of vectors
+ * stored in qdev->intr_count. If we don't get that
+ * many then we reduce the count and try again.
+ */
 static void ql_enable_msix(struct ql_adapter *qdev)
 {
-       int i;
+       int i, err;
 
-       qdev->intr_count = 1;
        /* Get the MSIX vectors. */
-       if (irq_type == MSIX_IRQ) {
+       if (qlge_irq_type == MSIX_IRQ) {
                /* Try to alloc space for the msix struct,
                 * if it fails then go to MSI/legacy.
                 */
-               qdev->msi_x_entry = kcalloc(qdev->rx_ring_count,
+               qdev->msi_x_entry = kcalloc(qdev->intr_count,
                                            sizeof(struct msix_entry),
                                            GFP_KERNEL);
                if (!qdev->msi_x_entry) {
-                       irq_type = MSI_IRQ;
+                       qlge_irq_type = MSI_IRQ;
                        goto msi;
                }
 
-               for (i = 0; i < qdev->rx_ring_count; i++)
+               for (i = 0; i < qdev->intr_count; i++)
                        qdev->msi_x_entry[i].entry = i;
 
-               if (!pci_enable_msix
-                   (qdev->pdev, qdev->msi_x_entry, qdev->rx_ring_count)) {
+               /* Loop to get our vectors.  We start with
+                * what we want and settle for what we get.
+                */
+               do {
+                       err = pci_enable_msix(qdev->pdev,
+                               qdev->msi_x_entry, qdev->intr_count);
+                       if (err > 0)
+                               qdev->intr_count = err;
+               } while (err > 0);
+
+               if (err < 0) {
+                       kfree(qdev->msi_x_entry);
+                       qdev->msi_x_entry = NULL;
+                       QPRINTK(qdev, IFUP, WARNING,
+                               "MSI-X Enable failed, trying MSI.\n");
+                       qdev->intr_count = 1;
+                       qlge_irq_type = MSI_IRQ;
+               } else if (err == 0) {
                        set_bit(QL_MSIX_ENABLED, &qdev->flags);
-                       qdev->intr_count = qdev->rx_ring_count;
                        QPRINTK(qdev, IFUP, INFO,
                                "MSI-X Enabled, got %d vectors.\n",
                                qdev->intr_count);
                        return;
-               } else {
-                       kfree(qdev->msi_x_entry);
-                       qdev->msi_x_entry = NULL;
-                       QPRINTK(qdev, IFUP, WARNING,
-                               "MSI-X Enable failed, trying MSI.\n");
-                       irq_type = MSI_IRQ;
                }
        }
 msi:
-       if (irq_type == MSI_IRQ) {
+       qdev->intr_count = 1;
+       if (qlge_irq_type == MSI_IRQ) {
                if (!pci_enable_msi(qdev->pdev)) {
                        set_bit(QL_MSI_ENABLED, &qdev->flags);
                        QPRINTK(qdev, IFUP, INFO,
@@ -2667,10 +2920,75 @@ msi:
                        return;
                }
        }
-       irq_type = LEG_IRQ;
+       qlge_irq_type = LEG_IRQ;
        QPRINTK(qdev, IFUP, DEBUG, "Running with legacy interrupts.\n");
 }
 
+/* Each vector services 1 RSS ring and and 1 or more
+ * TX completion rings.  This function loops through
+ * the TX completion rings and assigns the vector that
+ * will service it.  An example would be if there are
+ * 2 vectors (so 2 RSS rings) and 8 TX completion rings.
+ * This would mean that vector 0 would service RSS ring 0
+ * and TX competion rings 0,1,2 and 3.  Vector 1 would
+ * service RSS ring 1 and TX completion rings 4,5,6 and 7.
+ */
+static void ql_set_tx_vect(struct ql_adapter *qdev)
+{
+       int i, j, vect;
+       u32 tx_rings_per_vector = qdev->tx_ring_count / qdev->intr_count;
+
+       if (likely(test_bit(QL_MSIX_ENABLED, &qdev->flags))) {
+               /* Assign irq vectors to TX rx_rings.*/
+               for (vect = 0, j = 0, i = qdev->rss_ring_count;
+                                        i < qdev->rx_ring_count; i++) {
+                       if (j == tx_rings_per_vector) {
+                               vect++;
+                               j = 0;
+                       }
+                       qdev->rx_ring[i].irq = vect;
+                       j++;
+               }
+       } else {
+               /* For single vector all rings have an irq
+                * of zero.
+                */
+               for (i = 0; i < qdev->rx_ring_count; i++)
+                       qdev->rx_ring[i].irq = 0;
+       }
+}
+
+/* Set the interrupt mask for this vector.  Each vector
+ * will service 1 RSS ring and 1 or more TX completion
+ * rings.  This function sets up a bit mask per vector
+ * that indicates which rings it services.
+ */
+static void ql_set_irq_mask(struct ql_adapter *qdev, struct intr_context *ctx)
+{
+       int j, vect = ctx->intr;
+       u32 tx_rings_per_vector = qdev->tx_ring_count / qdev->intr_count;
+
+       if (likely(test_bit(QL_MSIX_ENABLED, &qdev->flags))) {
+               /* Add the RSS ring serviced by this vector
+                * to the mask.
+                */
+               ctx->irq_mask = (1 << qdev->rx_ring[vect].cq_id);
+               /* Add the TX ring(s) serviced by this vector
+                * to the mask. */
+               for (j = 0; j < tx_rings_per_vector; j++) {
+                       ctx->irq_mask |=
+                       (1 << qdev->rx_ring[qdev->rss_ring_count +
+                       (vect * tx_rings_per_vector) + j].cq_id);
+               }
+       } else {
+               /* For single vector we just shift each queue's
+                * ID into the mask.
+                */
+               for (j = 0; j < qdev->rx_ring_count; j++)
+                       ctx->irq_mask |= (1 << qdev->rx_ring[j].cq_id);
+       }
+}
+
 /*
  * Here we build the intr_context structures based on
  * our rx_ring count and intr vector count.
@@ -2682,18 +3000,19 @@ static void ql_resolve_queues_to_irqs(struct ql_adapter *qdev)
        int i = 0;
        struct intr_context *intr_context = &qdev->intr_context[0];
 
-       ql_enable_msix(qdev);
-
        if (likely(test_bit(QL_MSIX_ENABLED, &qdev->flags))) {
                /* Each rx_ring has it's
                 * own intr_context since we have separate
                 * vectors for each queue.
-                * This only true when MSI-X is enabled.
                 */
                for (i = 0; i < qdev->intr_count; i++, intr_context++) {
                        qdev->rx_ring[i].irq = i;
                        intr_context->intr = i;
                        intr_context->qdev = qdev;
+                       /* Set up this vector's bit-mask that indicates
+                        * which queues it services.
+                        */
+                       ql_set_irq_mask(qdev, intr_context);
                        /*
                         * We set up each vectors enable/disable/read bits so
                         * there's no bit/mask calculations in the critical path.
@@ -2710,21 +3029,14 @@ static void ql_resolve_queues_to_irqs(struct ql_adapter *qdev)
                            INTR_EN_TYPE_MASK | INTR_EN_INTR_MASK |
                            INTR_EN_TYPE_READ | INTR_EN_IHD_MASK | INTR_EN_IHD |
                            i;
-
                        if (i == 0) {
-                               /*
-                                * Default queue handles bcast/mcast plus
-                                * async events.  Needs buffers.
+                               /* The first vector/queue handles
+                                * broadcast/multicast, fatal errors,
+                                * and firmware events.  This in addition
+                                * to normal inbound NAPI processing.
                                 */
                                intr_context->handler = qlge_isr;
-                               sprintf(intr_context->name, "%s-default-queue",
-                                       qdev->ndev->name);
-                       } else if (i < qdev->rss_ring_first_cq_id) {
-                               /*
-                                * Outbound queue is for outbound completions only.
-                                */
-                               intr_context->handler = qlge_msix_tx_isr;
-                               sprintf(intr_context->name, "%s-tx-%d",
+                               sprintf(intr_context->name, "%s-rx-%d",
                                        qdev->ndev->name, i);
                        } else {
                                /*
@@ -2758,9 +3070,17 @@ static void ql_resolve_queues_to_irqs(struct ql_adapter *qdev)
                 */
                intr_context->handler = qlge_isr;
                sprintf(intr_context->name, "%s-single_irq", qdev->ndev->name);
-               for (i = 0; i < qdev->rx_ring_count; i++)
-                       qdev->rx_ring[i].irq = 0;
+               /* Set up this vector's bit-mask that indicates
+                * which queues it services. In this case there is
+                * a single vector so it will service all RSS and
+                * TX completion rings.
+                */
+               ql_set_irq_mask(qdev, intr_context);
        }
+       /* Tell the TX completion rings which MSIx vector
+        * they will be using.
+        */
+       ql_set_tx_vect(qdev);
 }
 
 static void ql_free_irq(struct ql_adapter *qdev)
@@ -2773,11 +3093,11 @@ static void ql_free_irq(struct ql_adapter *qdev)
                        if (test_bit(QL_MSIX_ENABLED, &qdev->flags)) {
                                free_irq(qdev->msi_x_entry[i].vector,
                                         &qdev->rx_ring[i]);
-                               QPRINTK(qdev, IFDOWN, ERR,
+                               QPRINTK(qdev, IFDOWN, DEBUG,
                                        "freeing msix interrupt %d.\n", i);
                        } else {
                                free_irq(qdev->pdev->irq, &qdev->rx_ring[0]);
-                               QPRINTK(qdev, IFDOWN, ERR,
+                               QPRINTK(qdev, IFDOWN, DEBUG,
                                        "freeing msi interrupt %d.\n", i);
                        }
                }
@@ -2808,7 +3128,7 @@ static int ql_request_irq(struct ql_adapter *qdev)
                                        i);
                                goto err_irq;
                        } else {
-                               QPRINTK(qdev, IFUP, INFO,
+                               QPRINTK(qdev, IFUP, DEBUG,
                                        "Hooked intr %d, queue type %s%s%s, with name %s.\n",
                                        i,
                                        qdev->rx_ring[i].type ==
@@ -2858,69 +3178,90 @@ err_irq:
 
 static int ql_start_rss(struct ql_adapter *qdev)
 {
+       u8 init_hash_seed[] = {0x6d, 0x5a, 0x56, 0xda, 0x25, 0x5b, 0x0e, 0xc2,
+                               0x41, 0x67, 0x25, 0x3d, 0x43, 0xa3, 0x8f,
+                               0xb0, 0xd0, 0xca, 0x2b, 0xcb, 0xae, 0x7b,
+                               0x30, 0xb4, 0x77, 0xcb, 0x2d, 0xa3, 0x80,
+                               0x30, 0xf2, 0x0c, 0x6a, 0x42, 0xb7, 0x3b,
+                               0xbe, 0xac, 0x01, 0xfa};
        struct ricb *ricb = &qdev->ricb;
        int status = 0;
        int i;
        u8 *hash_id = (u8 *) ricb->hash_cq_id;
 
-       memset((void *)ricb, 0, sizeof(ricb));
+       memset((void *)ricb, 0, sizeof(*ricb));
 
-       ricb->base_cq = qdev->rss_ring_first_cq_id | RSS_L4K;
+       ricb->base_cq = RSS_L4K;
        ricb->flags =
-           (RSS_L6K | RSS_LI | RSS_LB | RSS_LM | RSS_RI4 | RSS_RI6 | RSS_RT4 |
-            RSS_RT6);
-       ricb->mask = cpu_to_le16(qdev->rss_ring_count - 1);
+               (RSS_L6K | RSS_LI | RSS_LB | RSS_LM | RSS_RT4 | RSS_RT6);
+       ricb->mask = cpu_to_le16((u16)(0x3ff));
 
        /*
         * Fill out the Indirection Table.
         */
-       for (i = 0; i < 32; i++)
-               hash_id[i] = i & 1;
+       for (i = 0; i < 1024; i++)
+               hash_id[i] = (i & (qdev->rss_ring_count - 1));
 
-       /*
-        * Random values for the IPv6 and IPv4 Hash Keys.
-        */
-       get_random_bytes((void *)&ricb->ipv6_hash_key[0], 40);
-       get_random_bytes((void *)&ricb->ipv4_hash_key[0], 16);
+       memcpy((void *)&ricb->ipv6_hash_key[0], init_hash_seed, 40);
+       memcpy((void *)&ricb->ipv4_hash_key[0], init_hash_seed, 16);
 
-       QPRINTK(qdev, IFUP, INFO, "Initializing RSS.\n");
+       QPRINTK(qdev, IFUP, DEBUG, "Initializing RSS.\n");
 
-       status = ql_write_cfg(qdev, ricb, sizeof(ricb), CFG_LR, 0);
+       status = ql_write_cfg(qdev, ricb, sizeof(*ricb), CFG_LR, 0);
        if (status) {
                QPRINTK(qdev, IFUP, ERR, "Failed to load RICB.\n");
                return status;
        }
-       QPRINTK(qdev, IFUP, INFO, "Successfully loaded RICB.\n");
+       QPRINTK(qdev, IFUP, DEBUG, "Successfully loaded RICB.\n");
        return status;
 }
 
-/* Initialize the frame-to-queue routing. */
-static int ql_route_initialize(struct ql_adapter *qdev)
+static int ql_clear_routing_entries(struct ql_adapter *qdev)
 {
-       int status = 0;
-       int i;
+       int i, status = 0;
 
+       status = ql_sem_spinlock(qdev, SEM_RT_IDX_MASK);
+       if (status)
+               return status;
        /* Clear all the entries in the routing table. */
        for (i = 0; i < 16; i++) {
                status = ql_set_routing_reg(qdev, i, 0, 0);
                if (status) {
                        QPRINTK(qdev, IFUP, ERR,
-                               "Failed to init routing register for CAM packets.\n");
-                       return status;
+                               "Failed to init routing register for CAM "
+                               "packets.\n");
+                       break;
                }
        }
+       ql_sem_unlock(qdev, SEM_RT_IDX_MASK);
+       return status;
+}
+
+/* Initialize the frame-to-queue routing. */
+static int ql_route_initialize(struct ql_adapter *qdev)
+{
+       int status = 0;
+
+       /* Clear all the entries in the routing table. */
+       status = ql_clear_routing_entries(qdev);
+       if (status)
+               return status;
+
+       status = ql_sem_spinlock(qdev, SEM_RT_IDX_MASK);
+       if (status)
+               return status;
 
        status = ql_set_routing_reg(qdev, RT_IDX_ALL_ERR_SLOT, RT_IDX_ERR, 1);
        if (status) {
                QPRINTK(qdev, IFUP, ERR,
                        "Failed to init routing register for error packets.\n");
-               return status;
+               goto exit;
        }
        status = ql_set_routing_reg(qdev, RT_IDX_BCAST_SLOT, RT_IDX_BCAST, 1);
        if (status) {
                QPRINTK(qdev, IFUP, ERR,
                        "Failed to init routing register for broadcast packets.\n");
-               return status;
+               goto exit;
        }
        /* If we have more than one inbound queue, then turn on RSS in the
         * routing block.
@@ -2931,17 +3272,40 @@ static int ql_route_initialize(struct ql_adapter *qdev)
                if (status) {
                        QPRINTK(qdev, IFUP, ERR,
                                "Failed to init routing register for MATCH RSS packets.\n");
-                       return status;
+                       goto exit;
                }
        }
 
        status = ql_set_routing_reg(qdev, RT_IDX_CAM_HIT_SLOT,
                                    RT_IDX_CAM_HIT, 1);
-       if (status) {
+       if (status)
                QPRINTK(qdev, IFUP, ERR,
                        "Failed to init routing register for CAM packets.\n");
+exit:
+       ql_sem_unlock(qdev, SEM_RT_IDX_MASK);
+       return status;
+}
+
+int ql_cam_route_initialize(struct ql_adapter *qdev)
+{
+       int status, set;
+
+       /* If check if the link is up and use to
+        * determine if we are setting or clearing
+        * the MAC address in the CAM.
+        */
+       set = ql_read32(qdev, STS);
+       set &= qdev->port_link_up;
+       status = ql_set_mac_addr(qdev, set);
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR, "Failed to init mac address.\n");
                return status;
        }
+
+       status = ql_route_initialize(qdev);
+       if (status)
+               QPRINTK(qdev, IFUP, ERR, "Failed to init routing table.\n");
+
        return status;
 }
 
@@ -2958,9 +3322,9 @@ static int ql_adapter_initialize(struct ql_adapter *qdev)
        mask = value << 16;
        ql_write32(qdev, SYS, mask | value);
 
-       /* Set the default queue. */
-       value = NIC_RCV_CFG_DFQ;
-       mask = NIC_RCV_CFG_DFQ_MASK;
+       /* Set the default queue, and VLAN behavior. */
+       value = NIC_RCV_CFG_DFQ | NIC_RCV_CFG_RV;
+       mask = NIC_RCV_CFG_DFQ_MASK | (NIC_RCV_CFG_RV << 16);
        ql_write32(qdev, NIC_RCV_CFG, (mask | value));
 
        /* Set the MPI interrupt to enabled. */
@@ -2976,7 +3340,30 @@ static int ql_adapter_initialize(struct ql_adapter *qdev)
        ql_write32(qdev, FSC, mask | value);
 
        ql_write32(qdev, SPLT_HDR, SPLT_HDR_EP |
-               min(SMALL_BUFFER_SIZE, MAX_SPLIT_SIZE));
+               min(SMALL_BUF_MAP_SIZE, MAX_SPLIT_SIZE));
+
+       /* Set RX packet routing to use port/pci function on which the
+        * packet arrived on in addition to usual frame routing.
+        * This is helpful on bonding where both interfaces can have
+        * the same MAC address.
+        */
+       ql_write32(qdev, RST_FO, RST_FO_RR_MASK | RST_FO_RR_RCV_FUNC_CQ);
+       /* Reroute all packets to our Interface.
+        * They may have been routed to MPI firmware
+        * due to WOL.
+        */
+       value = ql_read32(qdev, MGMT_RCV_CFG);
+       value &= ~MGMT_RCV_CFG_RM;
+       mask = 0xffff0000;
+
+       /* Sticky reg needs clearing due to WOL. */
+       ql_write32(qdev, MGMT_RCV_CFG, mask);
+       ql_write32(qdev, MGMT_RCV_CFG, mask | value);
+
+       /* Default WOL is enable on Mezz cards */
+       if (qdev->pdev->subsystem_device == 0x0068 ||
+                       qdev->pdev->subsystem_device == 0x0180)
+               qdev->wol = WAKE_MAGIC;
 
        /* Start up the rx queues. */
        for (i = 0; i < qdev->rx_ring_count; i++) {
@@ -3009,28 +3396,22 @@ static int ql_adapter_initialize(struct ql_adapter *qdev)
                }
        }
 
-       status = ql_port_initialize(qdev);
-       if (status) {
+       /* Initialize the port and set the max framesize. */
+       status = qdev->nic_ops->port_initialize(qdev);
+       if (status)
                QPRINTK(qdev, IFUP, ERR, "Failed to start port.\n");
-               return status;
-       }
-
-       status = ql_set_mac_addr_reg(qdev, (u8 *) qdev->ndev->perm_addr,
-                                    MAC_ADDR_TYPE_CAM_MAC, qdev->func);
-       if (status) {
-               QPRINTK(qdev, IFUP, ERR, "Failed to init mac address.\n");
-               return status;
-       }
 
-       status = ql_route_initialize(qdev);
+       /* Set up the MAC address and frame routing filter. */
+       status = ql_cam_route_initialize(qdev);
        if (status) {
-               QPRINTK(qdev, IFUP, ERR, "Failed to init routing table.\n");
+               QPRINTK(qdev, IFUP, ERR,
+                               "Failed to init CAM/Routing tables.\n");
                return status;
        }
 
        /* Start NAPI for the RSS queues. */
-       for (i = qdev->rss_ring_first_cq_id; i < qdev->rx_ring_count; i++) {
-               QPRINTK(qdev, IFUP, INFO, "Enabling NAPI for rx_ring[%d].\n",
+       for (i = 0; i < qdev->rss_ring_count; i++) {
+               QPRINTK(qdev, IFUP, DEBUG, "Enabling NAPI for rx_ring[%d].\n",
                        i);
                napi_enable(&qdev->rx_ring[i].napi);
        }
@@ -3042,38 +3423,42 @@ static int ql_adapter_initialize(struct ql_adapter *qdev)
 static int ql_adapter_reset(struct ql_adapter *qdev)
 {
        u32 value;
-       int max_wait_time;
        int status = 0;
-       int resetCnt = 0;
+       unsigned long end_jiffies;
+
+       /* Clear all the entries in the routing table. */
+       status = ql_clear_routing_entries(qdev);
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR, "Failed to clear routing bits.\n");
+               return status;
+       }
+
+       end_jiffies = jiffies +
+               max((unsigned long)1, usecs_to_jiffies(30));
+
+       /* Stop management traffic. */
+       ql_mb_set_mgmnt_traffic_ctl(qdev, MB_SET_MPI_TFK_STOP);
+
+       /* Wait for the NIC and MGMNT FIFOs to empty. */
+       ql_wait_fifo_empty(qdev);
 
-#define MAX_RESET_CNT   1
-issueReset:
-       resetCnt++;
-       QPRINTK(qdev, IFDOWN, DEBUG, "Issue soft reset to chip.\n");
        ql_write32(qdev, RST_FO, (RST_FO_FR << 16) | RST_FO_FR);
-       /* Wait for reset to complete. */
-       max_wait_time = 3;
-       QPRINTK(qdev, IFDOWN, DEBUG, "Wait %d seconds for reset to complete.\n",
-               max_wait_time);
+
        do {
                value = ql_read32(qdev, RST_FO);
                if ((value & RST_FO_FR) == 0)
                        break;
+               cpu_relax();
+       } while (time_before(jiffies, end_jiffies));
 
-               ssleep(1);
-       } while ((--max_wait_time));
        if (value & RST_FO_FR) {
                QPRINTK(qdev, IFDOWN, ERR,
-                       "Stuck in SoftReset:  FSC_SR:0x%08x\n", value);
-               if (resetCnt < MAX_RESET_CNT)
-                       goto issueReset;
-       }
-       if (max_wait_time == 0) {
+                       "ETIMEDOUT!!! errored out of resetting the chip!\n");
                status = -ETIMEDOUT;
-               QPRINTK(qdev, IFDOWN, ERR,
-                       "ETIMEOUT!!! errored out of resetting the chip!\n");
        }
 
+       /* Resume management traffic. */
+       ql_mb_set_mgmnt_traffic_ctl(qdev, MB_SET_MPI_TFK_RESUME);
        return status;
 }
 
@@ -3082,9 +3467,10 @@ static void ql_display_dev_info(struct net_device *ndev)
        struct ql_adapter *qdev = (struct ql_adapter *)netdev_priv(ndev);
 
        QPRINTK(qdev, PROBE, INFO,
-               "Function #%d, NIC Roll %d, NIC Rev = %d, "
+               "Function #%d, Port %d, NIC Roll %d, NIC Rev = %d, "
                "XG Roll = %d, XG Rev = %d.\n",
                qdev->func,
+               qdev->port,
                qdev->chip_rev_id & 0x0000000f,
                qdev->chip_rev_id >> 4 & 0x0000000f,
                qdev->chip_rev_id >> 8 & 0x0000000f,
@@ -3092,40 +3478,70 @@ static void ql_display_dev_info(struct net_device *ndev)
        QPRINTK(qdev, PROBE, INFO, "MAC address %pM\n", ndev->dev_addr);
 }
 
+int ql_wol(struct ql_adapter *qdev)
+{
+       int status = 0;
+       u32 wol = MB_WOL_DISABLE;
+
+       /* The CAM is still intact after a reset, but if we
+        * are doing WOL, then we may need to program the
+        * routing regs. We would also need to issue the mailbox
+        * commands to instruct the MPI what to do per the ethtool
+        * settings.
+        */
+
+       if (qdev->wol & (WAKE_ARP | WAKE_MAGICSECURE | WAKE_PHY | WAKE_UCAST |
+                       WAKE_MCAST | WAKE_BCAST)) {
+               QPRINTK(qdev, IFDOWN, ERR,
+                       "Unsupported WOL paramter. qdev->wol = 0x%x.\n",
+                       qdev->wol);
+               return -EINVAL;
+       }
+
+       if (qdev->wol & WAKE_MAGIC) {
+               status = ql_mb_wol_set_magic(qdev, 1);
+               if (status) {
+                       QPRINTK(qdev, IFDOWN, ERR,
+                               "Failed to set magic packet on %s.\n",
+                               qdev->ndev->name);
+                       return status;
+               } else
+                       QPRINTK(qdev, DRV, INFO,
+                               "Enabled magic packet successfully on %s.\n",
+                               qdev->ndev->name);
+
+               wol |= MB_WOL_MAGIC_PKT;
+       }
+
+       if (qdev->wol) {
+               wol |= MB_WOL_MODE_ON;
+               status = ql_mb_wol_mode(qdev, wol);
+               QPRINTK(qdev, DRV, ERR, "WOL %s (wol code 0x%x) on %s\n",
+                       (status == 0) ? "Sucessfully set" : "Failed", wol,
+                       qdev->ndev->name);
+       }
+
+       return status;
+}
+
 static int ql_adapter_down(struct ql_adapter *qdev)
 {
-       struct net_device *ndev = qdev->ndev;
        int i, status = 0;
-       struct rx_ring *rx_ring;
 
-       netif_stop_queue(ndev);
-       netif_carrier_off(ndev);
+       ql_link_off(qdev);
 
-       cancel_delayed_work_sync(&qdev->asic_reset_work);
+       /* Don't kill the reset worker thread if we
+        * are in the process of recovery.
+        */
+       if (test_bit(QL_ADAPTER_UP, &qdev->flags))
+               cancel_delayed_work_sync(&qdev->asic_reset_work);
        cancel_delayed_work_sync(&qdev->mpi_reset_work);
        cancel_delayed_work_sync(&qdev->mpi_work);
+       cancel_delayed_work_sync(&qdev->mpi_idc_work);
+       cancel_delayed_work_sync(&qdev->mpi_port_cfg_work);
 
-       /* The default queue at index 0 is always processed in
-        * a workqueue.
-        */
-       cancel_delayed_work_sync(&qdev->rx_ring[0].rx_work);
-
-       /* The rest of the rx_rings are processed in
-        * a workqueue only if it's a single interrupt
-        * environment (MSI/Legacy).
-        */
-       for (i = 1; i < qdev->rx_ring_count; i++) {
-               rx_ring = &qdev->rx_ring[i];
-               /* Only the RSS rings use NAPI on multi irq
-                * environment.  Outbound completion processing
-                * is done in interrupt context.
-                */
-               if (i >= qdev->rss_ring_first_cq_id) {
-                       napi_disable(&rx_ring->napi);
-               } else {
-                       cancel_delayed_work_sync(&rx_ring->rx_work);
-               }
-       }
+       for (i = 0; i < qdev->rss_ring_count; i++)
+               napi_disable(&qdev->rx_ring[i].napi);
 
        clear_bit(QL_ADAPTER_UP, &qdev->flags);
 
@@ -3133,12 +3549,17 @@ static int ql_adapter_down(struct ql_adapter *qdev)
 
        ql_tx_ring_clean(qdev);
 
-       spin_lock(&qdev->hw_lock);
+       /* Call netif_napi_del() from common point.
+        */
+       for (i = 0; i < qdev->rss_ring_count; i++)
+               netif_napi_del(&qdev->rx_ring[i].napi);
+
+       ql_free_rx_buffers(qdev);
+
        status = ql_adapter_reset(qdev);
        if (status)
                QPRINTK(qdev, IFDOWN, ERR, "reset(func #%d) FAILED!\n",
                        qdev->func);
-       spin_unlock(&qdev->hw_lock);
        return status;
 }
 
@@ -3146,21 +3567,22 @@ static int ql_adapter_up(struct ql_adapter *qdev)
 {
        int err = 0;
 
-       spin_lock(&qdev->hw_lock);
        err = ql_adapter_initialize(qdev);
        if (err) {
                QPRINTK(qdev, IFUP, INFO, "Unable to initialize adapter.\n");
-               spin_unlock(&qdev->hw_lock);
                goto err_init;
        }
-       spin_unlock(&qdev->hw_lock);
        set_bit(QL_ADAPTER_UP, &qdev->flags);
+       ql_alloc_rx_buffers(qdev);
+       /* If the port is initialized and the
+        * link is up the turn on the carrier.
+        */
+       if ((ql_read32(qdev, STS) & qdev->port_init) &&
+                       (ql_read32(qdev, STS) & qdev->port_link_up))
+               ql_link_on(qdev);
        ql_enable_interrupts(qdev);
        ql_enable_all_completion_interrupts(qdev);
-       if ((ql_read32(qdev, STS) & qdev->port_init)) {
-               netif_carrier_on(qdev->ndev);
-               netif_start_queue(qdev->ndev);
-       }
+       netif_tx_start_all_queues(qdev->ndev);
 
        return 0;
 err_init:
@@ -3168,28 +3590,6 @@ err_init:
        return err;
 }
 
-static int ql_cycle_adapter(struct ql_adapter *qdev)
-{
-       int status;
-
-       status = ql_adapter_down(qdev);
-       if (status)
-               goto error;
-
-       status = ql_adapter_up(qdev);
-       if (status)
-               goto error;
-
-       return status;
-error:
-       QPRINTK(qdev, IFUP, ALERT,
-               "Driver up/down cycle failed, closing device\n");
-       rtnl_lock();
-       dev_close(qdev->ndev);
-       rtnl_unlock();
-       return status;
-}
-
 static void ql_release_adapter_resources(struct ql_adapter *qdev)
 {
        ql_free_mem_resources(qdev);
@@ -3205,11 +3605,6 @@ static int ql_get_adapter_resources(struct ql_adapter *qdev)
                return -ENOMEM;
        }
        status = ql_request_irq(qdev);
-       if (status)
-               goto err_irq;
-       return status;
-err_irq:
-       ql_free_mem_resources(qdev);
        return status;
 }
 
@@ -3233,47 +3628,29 @@ static int ql_configure_rings(struct ql_adapter *qdev)
        int i;
        struct rx_ring *rx_ring;
        struct tx_ring *tx_ring;
-       int cpu_cnt = num_online_cpus();
-
-       /*
-        * For each processor present we allocate one
-        * rx_ring for outbound completions, and one
-        * rx_ring for inbound completions.  Plus there is
-        * always the one default queue.  For the CPU
-        * counts we end up with the following rx_rings:
-        * rx_ring count =
-        *  one default queue +
-        *  (CPU count * outbound completion rx_ring) +
-        *  (CPU count * inbound (RSS) completion rx_ring)
-        * To keep it simple we limit the total number of
-        * queues to < 32, so we truncate CPU to 8.
-        * This limitation can be removed when requested.
-        */
-
-       if (cpu_cnt > MAX_CPUS)
-               cpu_cnt = MAX_CPUS;
-
-       /*
-        * rx_ring[0] is always the default queue.
+       int cpu_cnt = min(MAX_CPUS, (int)num_online_cpus());
+       unsigned int lbq_buf_len = (qdev->ndev->mtu > 1500) ?
+               LARGE_BUFFER_MAX_SIZE : LARGE_BUFFER_MIN_SIZE;
+
+       qdev->lbq_buf_order = get_order(lbq_buf_len);
+
+       /* In a perfect world we have one RSS ring for each CPU
+        * and each has it's own vector.  To do that we ask for
+        * cpu_cnt vectors.  ql_enable_msix() will adjust the
+        * vector count to what we actually get.  We then
+        * allocate an RSS ring for each.
+        * Essentially, we are doing min(cpu_count, msix_vector_count).
         */
-       /* Allocate outbound completion ring for each CPU. */
+       qdev->intr_count = cpu_cnt;
+       ql_enable_msix(qdev);
+       /* Adjust the RSS ring count to the actual vector count. */
+       qdev->rss_ring_count = qdev->intr_count;
        qdev->tx_ring_count = cpu_cnt;
-       /* Allocate inbound completion (RSS) ring for each CPU. */
-       qdev->rss_ring_count = cpu_cnt;
-       /* cq_id for the first inbound ring handler. */
-       qdev->rss_ring_first_cq_id = cpu_cnt + 1;
-       /*
-        * qdev->rx_ring_count:
-        * Total number of rx_rings.  This includes the one
-        * default queue, a number of outbound completion
-        * handler rx_rings, and the number of inbound
-        * completion handler rx_rings.
-        */
-       qdev->rx_ring_count = qdev->tx_ring_count + qdev->rss_ring_count + 1;
+       qdev->rx_ring_count = qdev->tx_ring_count + qdev->rss_ring_count;
 
        for (i = 0; i < qdev->tx_ring_count; i++) {
                tx_ring = &qdev->tx_ring[i];
-               memset((void *)tx_ring, 0, sizeof(tx_ring));
+               memset((void *)tx_ring, 0, sizeof(*tx_ring));
                tx_ring->qdev = qdev;
                tx_ring->wq_id = i;
                tx_ring->wq_len = qdev->tx_ring_size;
@@ -3282,21 +3659,20 @@ static int ql_configure_rings(struct ql_adapter *qdev)
 
                /*
                 * The completion queue ID for the tx rings start
-                * immediately after the default Q ID, which is zero.
+                * immediately after the rss rings.
                 */
-               tx_ring->cq_id = i + 1;
+               tx_ring->cq_id = qdev->rss_ring_count + i;
        }
 
        for (i = 0; i < qdev->rx_ring_count; i++) {
                rx_ring = &qdev->rx_ring[i];
-               memset((void *)rx_ring, 0, sizeof(rx_ring));
+               memset((void *)rx_ring, 0, sizeof(*rx_ring));
                rx_ring->qdev = qdev;
                rx_ring->cq_id = i;
                rx_ring->cpu = i % cpu_cnt;     /* CPU to run handler on. */
-               if (i == 0) {   /* Default queue at index 0. */
+               if (i < qdev->rss_ring_count) {
                        /*
-                        * Default queue handles bcast/mcast plus
-                        * async events.  Needs buffers.
+                        * Inbound (RSS) queues.
                         */
                        rx_ring->cq_len = qdev->rx_ring_size;
                        rx_ring->cq_size =
@@ -3304,13 +3680,16 @@ static int ql_configure_rings(struct ql_adapter *qdev)
                        rx_ring->lbq_len = NUM_LARGE_BUFFERS;
                        rx_ring->lbq_size =
                            rx_ring->lbq_len * sizeof(__le64);
-                       rx_ring->lbq_buf_size = LARGE_BUFFER_SIZE;
+                       rx_ring->lbq_buf_size = (u16)lbq_buf_len;
+                       QPRINTK(qdev, IFUP, DEBUG,
+                               "lbq_buf_size %d, order = %d\n",
+                               rx_ring->lbq_buf_size, qdev->lbq_buf_order);
                        rx_ring->sbq_len = NUM_SMALL_BUFFERS;
                        rx_ring->sbq_size =
                            rx_ring->sbq_len * sizeof(__le64);
-                       rx_ring->sbq_buf_size = SMALL_BUFFER_SIZE * 2;
-                       rx_ring->type = DEFAULT_Q;
-               } else if (i < qdev->rss_ring_first_cq_id) {
+                       rx_ring->sbq_buf_size = SMALL_BUF_MAP_SIZE;
+                       rx_ring->type = RX_Q;
+               } else {
                        /*
                         * Outbound queue handles outbound completions only.
                         */
@@ -3325,22 +3704,6 @@ static int ql_configure_rings(struct ql_adapter *qdev)
                        rx_ring->sbq_size = 0;
                        rx_ring->sbq_buf_size = 0;
                        rx_ring->type = TX_Q;
-               } else {        /* Inbound completions (RSS) queues */
-                       /*
-                        * Inbound queues handle unicast frames only.
-                        */
-                       rx_ring->cq_len = qdev->rx_ring_size;
-                       rx_ring->cq_size =
-                           rx_ring->cq_len * sizeof(struct ql_net_rsp_iocb);
-                       rx_ring->lbq_len = NUM_LARGE_BUFFERS;
-                       rx_ring->lbq_size =
-                           rx_ring->lbq_len * sizeof(__le64);
-                       rx_ring->lbq_buf_size = LARGE_BUFFER_SIZE;
-                       rx_ring->sbq_len = NUM_SMALL_BUFFERS;
-                       rx_ring->sbq_size =
-                           rx_ring->sbq_len * sizeof(__le64);
-                       rx_ring->sbq_buf_size = SMALL_BUFFER_SIZE * 2;
-                       rx_ring->type = RX_Q;
                }
        }
        return 0;
@@ -3351,6 +3714,10 @@ static int qlge_open(struct net_device *ndev)
        int err = 0;
        struct ql_adapter *qdev = netdev_priv(ndev);
 
+       err = ql_adapter_reset(qdev);
+       if (err)
+               return err;
+
        err = ql_configure_rings(qdev);
        if (err)
                return err;
@@ -3370,9 +3737,60 @@ error_up:
        return err;
 }
 
+static int ql_change_rx_buffers(struct ql_adapter *qdev)
+{
+       struct rx_ring *rx_ring;
+       int i, status;
+       u32 lbq_buf_len;
+
+       /* Wait for an oustanding reset to complete. */
+       if (!test_bit(QL_ADAPTER_UP, &qdev->flags)) {
+               int i = 3;
+               while (i-- && !test_bit(QL_ADAPTER_UP, &qdev->flags)) {
+                       QPRINTK(qdev, IFUP, ERR,
+                                "Waiting for adapter UP...\n");
+                       ssleep(1);
+               }
+
+               if (!i) {
+                       QPRINTK(qdev, IFUP, ERR,
+                        "Timed out waiting for adapter UP\n");
+                       return -ETIMEDOUT;
+               }
+       }
+
+       status = ql_adapter_down(qdev);
+       if (status)
+               goto error;
+
+       /* Get the new rx buffer size. */
+       lbq_buf_len = (qdev->ndev->mtu > 1500) ?
+               LARGE_BUFFER_MAX_SIZE : LARGE_BUFFER_MIN_SIZE;
+       qdev->lbq_buf_order = get_order(lbq_buf_len);
+
+       for (i = 0; i < qdev->rss_ring_count; i++) {
+               rx_ring = &qdev->rx_ring[i];
+               /* Set the new size. */
+               rx_ring->lbq_buf_size = lbq_buf_len;
+       }
+
+       status = ql_adapter_up(qdev);
+       if (status)
+               goto error;
+
+       return status;
+error:
+       QPRINTK(qdev, IFUP, ALERT,
+               "Driver up/down cycle failed, closing device.\n");
+       set_bit(QL_ADAPTER_UP, &qdev->flags);
+       dev_close(qdev->ndev);
+       return status;
+}
+
 static int qlge_change_mtu(struct net_device *ndev, int new_mtu)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
+       int status;
 
        if (ndev->mtu == 1500 && new_mtu == 9000) {
                QPRINTK(qdev, IFUP, ERR, "Changing to jumbo MTU.\n");
@@ -3383,24 +3801,71 @@ static int qlge_change_mtu(struct net_device *ndev, int new_mtu)
                return 0;
        } else
                return -EINVAL;
+
+       queue_delayed_work(qdev->workqueue,
+                       &qdev->mpi_port_cfg_work, 3*HZ);
+
+       if (!netif_running(qdev->ndev)) {
+               ndev->mtu = new_mtu;
+               return 0;
+       }
+
        ndev->mtu = new_mtu;
-       return 0;
+       status = ql_change_rx_buffers(qdev);
+       if (status) {
+               QPRINTK(qdev, IFUP, ERR,
+                       "Changing MTU failed.\n");
+       }
+
+       return status;
 }
 
 static struct net_device_stats *qlge_get_stats(struct net_device
                                               *ndev)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
-       return &qdev->stats;
+       struct rx_ring *rx_ring = &qdev->rx_ring[0];
+       struct tx_ring *tx_ring = &qdev->tx_ring[0];
+       unsigned long pkts, mcast, dropped, errors, bytes;
+       int i;
+
+       /* Get RX stats. */
+       pkts = mcast = dropped = errors = bytes = 0;
+       for (i = 0; i < qdev->rss_ring_count; i++, rx_ring++) {
+                       pkts += rx_ring->rx_packets;
+                       bytes += rx_ring->rx_bytes;
+                       dropped += rx_ring->rx_dropped;
+                       errors += rx_ring->rx_errors;
+                       mcast += rx_ring->rx_multicast;
+       }
+       ndev->stats.rx_packets = pkts;
+       ndev->stats.rx_bytes = bytes;
+       ndev->stats.rx_dropped = dropped;
+       ndev->stats.rx_errors = errors;
+       ndev->stats.multicast = mcast;
+
+       /* Get TX stats. */
+       pkts = errors = bytes = 0;
+       for (i = 0; i < qdev->tx_ring_count; i++, tx_ring++) {
+                       pkts += tx_ring->tx_packets;
+                       bytes += tx_ring->tx_bytes;
+                       errors += tx_ring->tx_errors;
+       }
+       ndev->stats.tx_packets = pkts;
+       ndev->stats.tx_bytes = bytes;
+       ndev->stats.tx_errors = errors;
+       return &ndev->stats;
 }
 
 static void qlge_set_multicast_list(struct net_device *ndev)
 {
        struct ql_adapter *qdev = (struct ql_adapter *)netdev_priv(ndev);
        struct dev_mc_list *mc_ptr;
-       int i;
+       int i, status;
 
-       spin_lock(&qdev->hw_lock);
+       status = ql_sem_spinlock(qdev, SEM_RT_IDX_MASK);
+       if (status)
+               return;
        /*
         * Set or clear promiscuous mode if a
         * transition is taking place.
@@ -3455,14 +3920,19 @@ static void qlge_set_multicast_list(struct net_device *ndev)
        }
 
        if (ndev->mc_count) {
+               status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
+               if (status)
+                       goto exit;
                for (i = 0, mc_ptr = ndev->mc_list; mc_ptr;
                     i++, mc_ptr = mc_ptr->next)
                        if (ql_set_mac_addr_reg(qdev, (u8 *) mc_ptr->dmi_addr,
                                                MAC_ADDR_TYPE_MULTI_MAC, i)) {
                                QPRINTK(qdev, HW, ERR,
                                        "Failed to loadmulticast address.\n");
+                               ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
                                goto exit;
                        }
+               ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
                if (ql_set_routing_reg
                    (qdev, RT_IDX_MCAST_MATCH_SLOT, RT_IDX_MCAST_MATCH, 1)) {
                        QPRINTK(qdev, HW, ERR,
@@ -3472,51 +3942,123 @@ static void qlge_set_multicast_list(struct net_device *ndev)
                }
        }
 exit:
-       spin_unlock(&qdev->hw_lock);
+       ql_sem_unlock(qdev, SEM_RT_IDX_MASK);
 }
 
 static int qlge_set_mac_address(struct net_device *ndev, void *p)
 {
        struct ql_adapter *qdev = (struct ql_adapter *)netdev_priv(ndev);
        struct sockaddr *addr = p;
-       int ret = 0;
-
-       if (netif_running(ndev))
-               return -EBUSY;
+       int status;
 
        if (!is_valid_ether_addr(addr->sa_data))
                return -EADDRNOTAVAIL;
        memcpy(ndev->dev_addr, addr->sa_data, ndev->addr_len);
 
-       spin_lock(&qdev->hw_lock);
-       if (ql_set_mac_addr_reg(qdev, (u8 *) ndev->dev_addr,
-                       MAC_ADDR_TYPE_CAM_MAC, qdev->func)) {/* Unicast */
+       status = ql_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
+       if (status)
+               return status;
+       status = ql_set_mac_addr_reg(qdev, (u8 *) ndev->dev_addr,
+                       MAC_ADDR_TYPE_CAM_MAC, qdev->func * MAX_CQ);
+       if (status)
                QPRINTK(qdev, HW, ERR, "Failed to load MAC address.\n");
-               ret = -1;
-       }
-       spin_unlock(&qdev->hw_lock);
-
-       return ret;
+       ql_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
+       return status;
 }
 
 static void qlge_tx_timeout(struct net_device *ndev)
 {
        struct ql_adapter *qdev = (struct ql_adapter *)netdev_priv(ndev);
-       queue_delayed_work(qdev->workqueue, &qdev->asic_reset_work, 0);
+       ql_queue_asic_error(qdev);
 }
 
 static void ql_asic_reset_work(struct work_struct *work)
 {
        struct ql_adapter *qdev =
            container_of(work, struct ql_adapter, asic_reset_work.work);
-       ql_cycle_adapter(qdev);
+       int status;
+       rtnl_lock();
+       status = ql_adapter_down(qdev);
+       if (status)
+               goto error;
+
+       status = ql_adapter_up(qdev);
+       if (status)
+               goto error;
+
+       /* Restore rx mode. */
+       clear_bit(QL_ALLMULTI, &qdev->flags);
+       clear_bit(QL_PROMISCUOUS, &qdev->flags);
+       qlge_set_multicast_list(qdev->ndev);
+
+       rtnl_unlock();
+       return;
+error:
+       QPRINTK(qdev, IFUP, ALERT,
+               "Driver up/down cycle failed, closing device\n");
+
+       set_bit(QL_ADAPTER_UP, &qdev->flags);
+       dev_close(qdev->ndev);
+       rtnl_unlock();
+}
+
+static struct nic_operations qla8012_nic_ops = {
+       .get_flash              = ql_get_8012_flash_params,
+       .port_initialize        = ql_8012_port_initialize,
+};
+
+static struct nic_operations qla8000_nic_ops = {
+       .get_flash              = ql_get_8000_flash_params,
+       .port_initialize        = ql_8000_port_initialize,
+};
+
+/* Find the pcie function number for the other NIC
+ * on this chip.  Since both NIC functions share a
+ * common firmware we have the lowest enabled function
+ * do any common work.  Examples would be resetting
+ * after a fatal firmware error, or doing a firmware
+ * coredump.
+ */
+static int ql_get_alt_pcie_func(struct ql_adapter *qdev)
+{
+       int status = 0;
+       u32 temp;
+       u32 nic_func1, nic_func2;
+
+       status = ql_read_mpi_reg(qdev, MPI_TEST_FUNC_PORT_CFG,
+                       &temp);
+       if (status)
+               return status;
+
+       nic_func1 = ((temp >> MPI_TEST_NIC1_FUNC_SHIFT) &
+                       MPI_TEST_NIC_FUNC_MASK);
+       nic_func2 = ((temp >> MPI_TEST_NIC2_FUNC_SHIFT) &
+                       MPI_TEST_NIC_FUNC_MASK);
+
+       if (qdev->func == nic_func1)
+               qdev->alt_func = nic_func2;
+       else if (qdev->func == nic_func2)
+               qdev->alt_func = nic_func1;
+       else
+               status = -EIO;
+
+       return status;
 }
 
-static void ql_get_board_info(struct ql_adapter *qdev)
+static int ql_get_board_info(struct ql_adapter *qdev)
 {
+       int status;
        qdev->func =
            (ql_read32(qdev, STS) & STS_FUNC_ID_MASK) >> STS_FUNC_ID_SHIFT;
-       if (qdev->func) {
+       if (qdev->func > 3)
+               return -EIO;
+
+       status = ql_get_alt_pcie_func(qdev);
+       if (status)
+               return status;
+
+       qdev->port = (qdev->func < qdev->alt_func) ? 0 : 1;
+       if (qdev->port) {
                qdev->xg_sem_mask = SEM_XGMAC1_MASK;
                qdev->port_link_up = STS_PL1;
                qdev->port_init = STS_PI1;
@@ -3530,6 +4072,12 @@ static void ql_get_board_info(struct ql_adapter *qdev)
                qdev->mailbox_out = PROC_ADDR_MPI_RISC | PROC_ADDR_FUNC0_MBO;
        }
        qdev->chip_rev_id = ql_read32(qdev, REV_ID);
+       qdev->device_id = qdev->pdev->device;
+       if (qdev->device_id == QLGE_DEVICE_ID_8012)
+               qdev->nic_ops = &qla8012_nic_ops;
+       else if (qdev->device_id == QLGE_DEVICE_ID_8000)
+               qdev->nic_ops = &qla8000_nic_ops;
+       return status;
 }
 
 static void ql_release_all(struct pci_dev *pdev)
@@ -3541,10 +4089,7 @@ static void ql_release_all(struct pci_dev *pdev)
                destroy_workqueue(qdev->workqueue);
                qdev->workqueue = NULL;
        }
-       if (qdev->q_workqueue) {
-               destroy_workqueue(qdev->q_workqueue);
-               qdev->q_workqueue = NULL;
-       }
+
        if (qdev->reg_base)
                iounmap(qdev->reg_base);
        if (qdev->doorbell_area)
@@ -3557,44 +4102,40 @@ static int __devinit ql_init_device(struct pci_dev *pdev,
                                    struct net_device *ndev, int cards_found)
 {
        struct ql_adapter *qdev = netdev_priv(ndev);
-       int pos, err = 0;
-       u16 val16;
+       int err = 0;
 
-       memset((void *)qdev, 0, sizeof(qdev));
+       memset((void *)qdev, 0, sizeof(*qdev));
        err = pci_enable_device(pdev);
        if (err) {
                dev_err(&pdev->dev, "PCI device enable failed.\n");
                return err;
        }
 
-       pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
-       if (pos <= 0) {
-               dev_err(&pdev->dev, PFX "Cannot find PCI Express capability, "
-                       "aborting.\n");
+       qdev->ndev = ndev;
+       qdev->pdev = pdev;
+       pci_set_drvdata(pdev, ndev);
+
+       /* Set PCIe read request size */
+       err = pcie_set_readrq(pdev, 4096);
+       if (err) {
+               dev_err(&pdev->dev, "Set readrq failed.\n");
                goto err_out;
-       } else {
-               pci_read_config_word(pdev, pos + PCI_EXP_DEVCTL, &val16);
-               val16 &= ~PCI_EXP_DEVCTL_NOSNOOP_EN;
-               val16 |= (PCI_EXP_DEVCTL_CERE |
-                         PCI_EXP_DEVCTL_NFERE |
-                         PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE);
-               pci_write_config_word(pdev, pos + PCI_EXP_DEVCTL, val16);
        }
 
        err = pci_request_regions(pdev, DRV_NAME);
        if (err) {
                dev_err(&pdev->dev, "PCI region request failed.\n");
-               goto err_out;
+               return err;
        }
 
        pci_set_master(pdev);
-       if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
+       if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
                set_bit(QL_DMA64, &qdev->flags);
-               err = pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK);
+               err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
        } else {
-               err = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
+               err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
                if (!err)
-                      err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
+                      err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
        }
 
        if (err) {
@@ -3602,7 +4143,9 @@ static int __devinit ql_init_device(struct pci_dev *pdev,
                goto err_out;
        }
 
-       pci_set_drvdata(pdev, ndev);
+       /* Set PCIe reset type for EEH to fundamental. */
+       pdev->needs_freset = 1;
+       pci_save_state(pdev);
        qdev->reg_base =
            ioremap_nocache(pci_resource_start(pdev, 1),
                            pci_resource_len(pdev, 1));
@@ -3622,24 +4165,23 @@ static int __devinit ql_init_device(struct pci_dev *pdev,
                goto err_out;
        }
 
-       ql_get_board_info(qdev);
-       qdev->ndev = ndev;
-       qdev->pdev = pdev;
+       err = ql_get_board_info(qdev);
+       if (err) {
+               dev_err(&pdev->dev, "Register access failed.\n");
+               err = -EIO;
+               goto err_out;
+       }
        qdev->msg_enable = netif_msg_init(debug, default_msg);
        spin_lock_init(&qdev->hw_lock);
        spin_lock_init(&qdev->stats_lock);
 
        /* make sure the EEPROM is good */
-       err = ql_get_flash_params(qdev);
+       err = qdev->nic_ops->get_flash(qdev);
        if (err) {
                dev_err(&pdev->dev, "Invalid FLASH.\n");
                goto err_out;
        }
 
-       if (!is_valid_ether_addr(qdev->flash.mac_addr))
-               goto err_out;
-
-       memcpy(ndev->dev_addr, qdev->flash.mac_addr, ndev->addr_len);
        memcpy(ndev->perm_addr, ndev->dev_addr, ndev->addr_len);
 
        /* Set up the default ring sizes. */
@@ -3656,12 +4198,13 @@ static int __devinit ql_init_device(struct pci_dev *pdev,
         * Set up the operating parameters.
         */
        qdev->rx_csum = 1;
-
-       qdev->q_workqueue = create_workqueue(ndev->name);
        qdev->workqueue = create_singlethread_workqueue(ndev->name);
        INIT_DELAYED_WORK(&qdev->asic_reset_work, ql_asic_reset_work);
        INIT_DELAYED_WORK(&qdev->mpi_reset_work, ql_mpi_reset_work);
        INIT_DELAYED_WORK(&qdev->mpi_work, ql_mpi_work);
+       INIT_DELAYED_WORK(&qdev->mpi_port_cfg_work, ql_mpi_port_cfg_work);
+       INIT_DELAYED_WORK(&qdev->mpi_idc_work, ql_mpi_idc_work);
+       init_completion(&qdev->ide_completion);
 
        if (!cards_found) {
                dev_info(&pdev->dev, "%s\n", DRV_STRING);
@@ -3675,7 +4218,6 @@ err_out:
        return err;
 }
 
-
 static const struct net_device_ops qlge_netdev_ops = {
        .ndo_open               = qlge_open,
        .ndo_stop               = qlge_close,
@@ -3686,9 +4228,9 @@ static const struct net_device_ops qlge_netdev_ops = {
        .ndo_set_mac_address    = qlge_set_mac_address,
        .ndo_validate_addr      = eth_validate_addr,
        .ndo_tx_timeout         = qlge_tx_timeout,
-       .ndo_vlan_rx_register   = ql_vlan_rx_register,
-       .ndo_vlan_rx_add_vid    = ql_vlan_rx_add_vid,
-       .ndo_vlan_rx_kill_vid   = ql_vlan_rx_kill_vid,
+       .ndo_vlan_rx_register   = qlge_vlan_rx_register,
+       .ndo_vlan_rx_add_vid    = qlge_vlan_rx_add_vid,
+       .ndo_vlan_rx_kill_vid   = qlge_vlan_rx_kill_vid,
 };
 
 static int __devinit qlge_probe(struct pci_dev *pdev,
@@ -3699,7 +4241,8 @@ static int __devinit qlge_probe(struct pci_dev *pdev,
        static int cards_found = 0;
        int err = 0;
 
-       ndev = alloc_etherdev(sizeof(struct ql_adapter));
+       ndev = alloc_etherdev_mq(sizeof(struct ql_adapter),
+                       min(MAX_CPUS, (int)num_online_cpus()));
        if (!ndev)
                return -ENOMEM;
 
@@ -3719,6 +4262,7 @@ static int __devinit qlge_probe(struct pci_dev *pdev,
                          | NETIF_F_TSO_ECN
                          | NETIF_F_HW_VLAN_TX
                          | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_FILTER);
+       ndev->features |= NETIF_F_GRO;
 
        if (test_bit(QL_DMA64, &qdev->flags))
                ndev->features |= NETIF_F_HIGHDMA;
@@ -3740,13 +4284,23 @@ static int __devinit qlge_probe(struct pci_dev *pdev,
                pci_disable_device(pdev);
                return err;
        }
-       netif_carrier_off(ndev);
-       netif_stop_queue(ndev);
+       ql_link_off(qdev);
        ql_display_dev_info(ndev);
+       atomic_set(&qdev->lb_count, 0);
        cards_found++;
        return 0;
 }
 
+netdev_tx_t ql_lb_send(struct sk_buff *skb, struct net_device *ndev)
+{
+       return qlge_send(skb, ndev);
+}
+
+int ql_clean_lb_rx_ring(struct rx_ring *rx_ring, int budget)
+{
+       return ql_clean_inbound_rx_ring(rx_ring, budget);
+}
+
 static void __devexit qlge_remove(struct pci_dev *pdev)
 {
        struct net_device *ndev = pci_get_drvdata(pdev);
@@ -3756,6 +4310,33 @@ static void __devexit qlge_remove(struct pci_dev *pdev)
        free_netdev(ndev);
 }
 
+/* Clean up resources without touching hardware. */
+static void ql_eeh_close(struct net_device *ndev)
+{
+       int i;
+       struct ql_adapter *qdev = netdev_priv(ndev);
+
+       if (netif_carrier_ok(ndev)) {
+               netif_carrier_off(ndev);
+               netif_stop_queue(ndev);
+       }
+
+       if (test_bit(QL_ADAPTER_UP, &qdev->flags))
+               cancel_delayed_work_sync(&qdev->asic_reset_work);
+       cancel_delayed_work_sync(&qdev->mpi_reset_work);
+       cancel_delayed_work_sync(&qdev->mpi_work);
+       cancel_delayed_work_sync(&qdev->mpi_idc_work);
+       cancel_delayed_work_sync(&qdev->mpi_port_cfg_work);
+
+       for (i = 0; i < qdev->rss_ring_count; i++)
+               netif_napi_del(&qdev->rx_ring[i].napi);
+
+       clear_bit(QL_ADAPTER_UP, &qdev->flags);
+       ql_tx_ring_clean(qdev);
+       ql_free_rx_buffers(qdev);
+       ql_release_adapter_resources(qdev);
+}
+
 /*
  * This callback is called by the PCI subsystem whenever
  * a PCI bus error is detected.
@@ -3764,12 +4345,21 @@ static pci_ers_result_t qlge_io_error_detected(struct pci_dev *pdev,
                                               enum pci_channel_state state)
 {
        struct net_device *ndev = pci_get_drvdata(pdev);
-       struct ql_adapter *qdev = netdev_priv(ndev);
-
-       if (netif_running(ndev))
-               ql_adapter_down(qdev);
 
-       pci_disable_device(pdev);
+       switch (state) {
+       case pci_channel_io_normal:
+               return PCI_ERS_RESULT_CAN_RECOVER;
+       case pci_channel_io_frozen:
+               netif_device_detach(ndev);
+               if (netif_running(ndev))
+                       ql_eeh_close(ndev);
+               pci_disable_device(pdev);
+               return PCI_ERS_RESULT_NEED_RESET;
+       case pci_channel_io_perm_failure:
+               dev_err(&pdev->dev,
+                       "%s: pci_channel_io_perm_failure.\n", __func__);
+               return PCI_ERS_RESULT_DISCONNECT;
+       }
 
        /* Request a slot reset. */
        return PCI_ERS_RESULT_NEED_RESET;
@@ -3786,26 +4376,15 @@ static pci_ers_result_t qlge_io_slot_reset(struct pci_dev *pdev)
        struct net_device *ndev = pci_get_drvdata(pdev);
        struct ql_adapter *qdev = netdev_priv(ndev);
 
+       pdev->error_state = pci_channel_io_normal;
+
+       pci_restore_state(pdev);
        if (pci_enable_device(pdev)) {
                QPRINTK(qdev, IFUP, ERR,
                        "Cannot re-enable PCI device after reset.\n");
                return PCI_ERS_RESULT_DISCONNECT;
        }
-
        pci_set_master(pdev);
-
-       netif_carrier_off(ndev);
-       netif_stop_queue(ndev);
-       ql_adapter_reset(qdev);
-
-       /* Make sure the EEPROM is good */
-       memcpy(ndev->perm_addr, ndev->dev_addr, ndev->addr_len);
-
-       if (!is_valid_ether_addr(ndev->perm_addr)) {
-               QPRINTK(qdev, IFUP, ERR, "After reset, invalid MAC address.\n");
-               return PCI_ERS_RESULT_DISCONNECT;
-       }
-
        return PCI_ERS_RESULT_RECOVERED;
 }
 
@@ -3813,17 +4392,21 @@ static void qlge_io_resume(struct pci_dev *pdev)
 {
        struct net_device *ndev = pci_get_drvdata(pdev);
        struct ql_adapter *qdev = netdev_priv(ndev);
+       int err = 0;
 
-       pci_set_master(pdev);
-
+       if (ql_adapter_reset(qdev))
+               QPRINTK(qdev, DRV, ERR, "reset FAILED!\n");
        if (netif_running(ndev)) {
-               if (ql_adapter_up(qdev)) {
+               err = qlge_open(ndev);
+               if (err) {
                        QPRINTK(qdev, IFUP, ERR,
                                "Device initialization failed after reset.\n");
                        return;
                }
+       } else {
+               QPRINTK(qdev, IFUP, ERR,
+                       "Device was not running prior to EEH.\n");
        }
-
        netif_device_attach(ndev);
 }
 
@@ -3837,7 +4420,7 @@ static int qlge_suspend(struct pci_dev *pdev, pm_message_t state)
 {
        struct net_device *ndev = pci_get_drvdata(pdev);
        struct ql_adapter *qdev = netdev_priv(ndev);
-       int err, i;
+       int err;
 
        netif_device_detach(ndev);
 
@@ -3847,9 +4430,7 @@ static int qlge_suspend(struct pci_dev *pdev, pm_message_t state)
                        return err;
        }
 
-       for (i = qdev->rss_ring_first_cq_id; i < qdev->rx_ring_count; i++)
-               netif_napi_del(&qdev->rx_ring[i].napi);
-
+       ql_wol(qdev);
        err = pci_save_state(pdev);
        if (err)
                return err;