Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[safe/jmp/linux-2.6] / drivers / net / ixgb / ixgb_main.c
index fb4c1ad..c56ea69 100644 (file)
@@ -1,7 +1,7 @@
 /*******************************************************************************
 
   Intel PRO/10GbE Linux driver
-  Copyright(c) 1999 - 2006 Intel Corporation.
+  Copyright(c) 1999 - 2008 Intel Corporation.
 
   This program is free software; you can redistribute it and/or modify it
   under the terms and conditions of the GNU General Public License,
 char ixgb_driver_name[] = "ixgb";
 static char ixgb_driver_string[] = "Intel(R) PRO/10GbE Network Driver";
 
-#ifndef CONFIG_IXGB_NAPI
-#define DRIVERNAPI
-#else
 #define DRIVERNAPI "-NAPI"
-#endif
-#define DRV_VERSION            "1.0.112-k2"DRIVERNAPI
-char ixgb_driver_version[] = DRV_VERSION;
-static char ixgb_copyright[] = "Copyright (c) 1999-2006 Intel Corporation.";
+#define DRV_VERSION "1.0.135-k2" DRIVERNAPI
+const char ixgb_driver_version[] = DRV_VERSION;
+static const char ixgb_copyright[] = "Copyright (c) 1999-2008 Intel Corporation.";
+
+#define IXGB_CB_LENGTH 256
+static unsigned int copybreak __read_mostly = IXGB_CB_LENGTH;
+module_param(copybreak, uint, 0644);
+MODULE_PARM_DESC(copybreak,
+       "Maximum size of packet that is copied to a new buffer on receive");
 
 /* ixgb_pci_tbl - PCI Device ID Table
  *
@@ -48,14 +50,14 @@ static char ixgb_copyright[] = "Copyright (c) 1999-2006 Intel Corporation.";
  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  *   Class, Class Mask, private data (not used) }
  */
-static struct pci_device_id ixgb_pci_tbl[] = {
+static DEFINE_PCI_DEVICE_TABLE(ixgb_pci_tbl) = {
        {INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX,
         PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
        {INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_CX4,
         PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
        {INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_SR,
         PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
-       {INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_LR,  
+       {INTEL_VENDOR_ID, IXGB_DEVICE_ID_82597EX_LR,
         PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
 
        /* required last entry */
@@ -65,16 +67,6 @@ static struct pci_device_id ixgb_pci_tbl[] = {
 MODULE_DEVICE_TABLE(pci, ixgb_pci_tbl);
 
 /* Local Function Prototypes */
-
-int ixgb_up(struct ixgb_adapter *adapter);
-void ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog);
-void ixgb_reset(struct ixgb_adapter *adapter);
-int ixgb_setup_tx_resources(struct ixgb_adapter *adapter);
-int ixgb_setup_rx_resources(struct ixgb_adapter *adapter);
-void ixgb_free_tx_resources(struct ixgb_adapter *adapter);
-void ixgb_free_rx_resources(struct ixgb_adapter *adapter);
-void ixgb_update_stats(struct ixgb_adapter *adapter);
-
 static int ixgb_init_module(void);
 static void ixgb_exit_module(void);
 static int ixgb_probe(struct pci_dev *pdev, const struct pci_device_id *ent);
@@ -89,28 +81,25 @@ static void ixgb_clean_tx_ring(struct ixgb_adapter *adapter);
 static void ixgb_clean_rx_ring(struct ixgb_adapter *adapter);
 static void ixgb_set_multi(struct net_device *netdev);
 static void ixgb_watchdog(unsigned long data);
-static int ixgb_xmit_frame(struct sk_buff *skb, struct net_device *netdev);
+static netdev_tx_t ixgb_xmit_frame(struct sk_buff *skb,
+                                  struct net_device *netdev);
 static struct net_device_stats *ixgb_get_stats(struct net_device *netdev);
 static int ixgb_change_mtu(struct net_device *netdev, int new_mtu);
 static int ixgb_set_mac(struct net_device *netdev, void *p);
-static irqreturn_t ixgb_intr(int irq, void *data, struct pt_regs *regs);
-static boolean_t ixgb_clean_tx_irq(struct ixgb_adapter *adapter);
-
-#ifdef CONFIG_IXGB_NAPI
-static int ixgb_clean(struct net_device *netdev, int *budget);
-static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter,
-                                  int *work_done, int work_to_do);
-#else
-static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter);
-#endif
-static void ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter);
-void ixgb_set_ethtool_ops(struct net_device *netdev);
+static irqreturn_t ixgb_intr(int irq, void *data);
+static bool ixgb_clean_tx_irq(struct ixgb_adapter *adapter);
+
+static int ixgb_clean(struct napi_struct *, int);
+static bool ixgb_clean_rx_irq(struct ixgb_adapter *, int *, int);
+static void ixgb_alloc_rx_buffers(struct ixgb_adapter *, int);
+
 static void ixgb_tx_timeout(struct net_device *dev);
-static void ixgb_tx_timeout_task(struct net_device *dev);
+static void ixgb_tx_timeout_task(struct work_struct *work);
+
 static void ixgb_vlan_rx_register(struct net_device *netdev,
-                                 struct vlan_group *grp);
-static void ixgb_vlan_rx_add_vid(struct net_device *netdev, uint16_t vid);
-static void ixgb_vlan_rx_kill_vid(struct net_device *netdev, uint16_t vid);
+                                  struct vlan_group *grp);
+static void ixgb_vlan_rx_add_vid(struct net_device *netdev, u16 vid);
+static void ixgb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid);
 static void ixgb_restore_vlan(struct ixgb_adapter *adapter);
 
 #ifdef CONFIG_NET_POLL_CONTROLLER
@@ -119,13 +108,10 @@ static void ixgb_netpoll(struct net_device *dev);
 #endif
 
 static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
-                            enum pci_channel_state state);
+                             enum pci_channel_state state);
 static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev);
 static void ixgb_io_resume (struct pci_dev *pdev);
 
-/* Exported from other modules */
-extern void ixgb_check_options(struct ixgb_adapter *adapter);
-
 static struct pci_error_handlers ixgb_err_handler = {
        .error_detected = ixgb_io_error_detected,
        .slot_reset = ixgb_io_slot_reset,
@@ -150,14 +136,6 @@ static int debug = DEFAULT_DEBUG_LEVEL_SHIFT;
 module_param(debug, int, 0);
 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 
-/* some defines for controlling descriptor fetches in h/w */
-#define RXDCTL_WTHRESH_DEFAULT 15  /* chip writes back at this many or RXT0 */
-#define RXDCTL_PTHRESH_DEFAULT 0   /* chip considers prefech below
-                                    * this */
-#define RXDCTL_HTHRESH_DEFAULT 0   /* chip will only prefetch if tail
-                                    * is pushed this many descriptors
-                                    * from head */
-
 /**
  * ixgb_init_module - Driver Registration Routine
  *
@@ -201,7 +179,6 @@ module_exit(ixgb_exit_module);
 static void
 ixgb_irq_disable(struct ixgb_adapter *adapter)
 {
-       atomic_inc(&adapter->irq_sem);
        IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
        IXGB_WRITE_FLUSH(&adapter->hw);
        synchronize_irq(adapter->pdev->irq);
@@ -215,19 +192,19 @@ ixgb_irq_disable(struct ixgb_adapter *adapter)
 static void
 ixgb_irq_enable(struct ixgb_adapter *adapter)
 {
-       if(atomic_dec_and_test(&adapter->irq_sem)) {
-               IXGB_WRITE_REG(&adapter->hw, IMS,
-                              IXGB_INT_RXT0 | IXGB_INT_RXDMT0 | IXGB_INT_TXDW |
-                              IXGB_INT_LSC);
-               IXGB_WRITE_FLUSH(&adapter->hw);
-       }
+       u32 val = IXGB_INT_RXT0 | IXGB_INT_RXDMT0 |
+                 IXGB_INT_TXDW | IXGB_INT_LSC;
+       if (adapter->hw.subsystem_vendor_id == SUN_SUBVENDOR_ID)
+               val |= IXGB_INT_GPI0;
+       IXGB_WRITE_REG(&adapter->hw, IMS, val);
+       IXGB_WRITE_FLUSH(&adapter->hw);
 }
 
 int
 ixgb_up(struct ixgb_adapter *adapter)
 {
        struct net_device *netdev = adapter->netdev;
-       int err;
+       int err, irq_flags = IRQF_SHARED;
        int max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
        struct ixgb_hw *hw = &adapter->hw;
 
@@ -241,37 +218,32 @@ ixgb_up(struct ixgb_adapter *adapter)
        ixgb_configure_tx(adapter);
        ixgb_setup_rctl(adapter);
        ixgb_configure_rx(adapter);
-       ixgb_alloc_rx_buffers(adapter);
+       ixgb_alloc_rx_buffers(adapter, IXGB_DESC_UNUSED(&adapter->rx_ring));
 
        /* disable interrupts and get the hardware into a known state */
        IXGB_WRITE_REG(&adapter->hw, IMC, 0xffffffff);
 
-#ifdef CONFIG_PCI_MSI
-       {
-       boolean_t pcix = (IXGB_READ_REG(&adapter->hw, STATUS) & 
-                                                 IXGB_STATUS_PCIX_MODE) ? TRUE : FALSE;
-       adapter->have_msi = TRUE;
-
-       if (!pcix)
-          adapter->have_msi = FALSE;
-       else if((err = pci_enable_msi(adapter->pdev))) {
-               DPRINTK(PROBE, ERR,
-                "Unable to allocate MSI interrupt Error: %d\n", err);
-               adapter->have_msi = FALSE;
+       /* only enable MSI if bus is in PCI-X mode */
+       if (IXGB_READ_REG(&adapter->hw, STATUS) & IXGB_STATUS_PCIX_MODE) {
+               err = pci_enable_msi(adapter->pdev);
+               if (!err) {
+                       adapter->have_msi = 1;
+                       irq_flags = 0;
+               }
                /* proceed to try to request regular interrupt */
        }
-       }
 
-#endif
-       if((err = request_irq(adapter->pdev->irq, &ixgb_intr,
-                                 IRQF_SHARED | IRQF_SAMPLE_RANDOM,
-                                 netdev->name, netdev))) {
+       err = request_irq(adapter->pdev->irq, ixgb_intr, irq_flags,
+                         netdev->name, netdev);
+       if (err) {
+               if (adapter->have_msi)
+                       pci_disable_msi(adapter->pdev);
                DPRINTK(PROBE, ERR,
                 "Unable to allocate interrupt Error: %d\n", err);
                return err;
        }
 
-       if((hw->max_frame_size != max_frame) ||
+       if ((hw->max_frame_size != max_frame) ||
                (hw->max_frame_size !=
                (IXGB_READ_REG(hw, MFS) >> IXGB_MFS_SHIFT))) {
 
@@ -279,44 +251,48 @@ ixgb_up(struct ixgb_adapter *adapter)
 
                IXGB_WRITE_REG(hw, MFS, hw->max_frame_size << IXGB_MFS_SHIFT);
 
-               if(hw->max_frame_size >
+               if (hw->max_frame_size >
                   IXGB_MAX_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH) {
-                       uint32_t ctrl0 = IXGB_READ_REG(hw, CTRL0);
+                       u32 ctrl0 = IXGB_READ_REG(hw, CTRL0);
 
-                       if(!(ctrl0 & IXGB_CTRL0_JFE)) {
+                       if (!(ctrl0 & IXGB_CTRL0_JFE)) {
                                ctrl0 |= IXGB_CTRL0_JFE;
                                IXGB_WRITE_REG(hw, CTRL0, ctrl0);
                        }
                }
        }
 
-       mod_timer(&adapter->watchdog_timer, jiffies);
+       clear_bit(__IXGB_DOWN, &adapter->flags);
 
-#ifdef CONFIG_IXGB_NAPI
-       netif_poll_enable(netdev);
-#endif
+       napi_enable(&adapter->napi);
        ixgb_irq_enable(adapter);
 
+       netif_wake_queue(netdev);
+
+       mod_timer(&adapter->watchdog_timer, jiffies);
+
        return 0;
 }
 
 void
-ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog)
+ixgb_down(struct ixgb_adapter *adapter, bool kill_watchdog)
 {
        struct net_device *netdev = adapter->netdev;
 
+       /* prevent the interrupt handler from restarting watchdog */
+       set_bit(__IXGB_DOWN, &adapter->flags);
+
+       napi_disable(&adapter->napi);
+       /* waiting for NAPI to complete can re-enable interrupts */
        ixgb_irq_disable(adapter);
        free_irq(adapter->pdev->irq, netdev);
-#ifdef CONFIG_PCI_MSI
-       if(adapter->have_msi == TRUE)
+
+       if (adapter->have_msi)
                pci_disable_msi(adapter->pdev);
 
-#endif
-       if(kill_watchdog)
+       if (kill_watchdog)
                del_timer_sync(&adapter->watchdog_timer);
-#ifdef CONFIG_IXGB_NAPI
-       netif_poll_disable(netdev);
-#endif
+
        adapter->link_speed = 0;
        adapter->link_duplex = 0;
        netif_carrier_off(netdev);
@@ -330,12 +306,42 @@ ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog)
 void
 ixgb_reset(struct ixgb_adapter *adapter)
 {
+       struct ixgb_hw *hw = &adapter->hw;
 
-       ixgb_adapter_stop(&adapter->hw);
-       if(!ixgb_init_hw(&adapter->hw))
+       ixgb_adapter_stop(hw);
+       if (!ixgb_init_hw(hw))
                DPRINTK(PROBE, ERR, "ixgb_init_hw failed.\n");
+
+       /* restore frame size information */
+       IXGB_WRITE_REG(hw, MFS, hw->max_frame_size << IXGB_MFS_SHIFT);
+       if (hw->max_frame_size >
+           IXGB_MAX_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH) {
+               u32 ctrl0 = IXGB_READ_REG(hw, CTRL0);
+               if (!(ctrl0 & IXGB_CTRL0_JFE)) {
+                       ctrl0 |= IXGB_CTRL0_JFE;
+                       IXGB_WRITE_REG(hw, CTRL0, ctrl0);
+               }
+       }
 }
 
+static const struct net_device_ops ixgb_netdev_ops = {
+       .ndo_open               = ixgb_open,
+       .ndo_stop               = ixgb_close,
+       .ndo_start_xmit         = ixgb_xmit_frame,
+       .ndo_get_stats          = ixgb_get_stats,
+       .ndo_set_multicast_list = ixgb_set_multi,
+       .ndo_validate_addr      = eth_validate_addr,
+       .ndo_set_mac_address    = ixgb_set_mac,
+       .ndo_change_mtu         = ixgb_change_mtu,
+       .ndo_tx_timeout         = ixgb_tx_timeout,
+       .ndo_vlan_rx_register   = ixgb_vlan_rx_register,
+       .ndo_vlan_rx_add_vid    = ixgb_vlan_rx_add_vid,
+       .ndo_vlan_rx_kill_vid   = ixgb_vlan_rx_kill_vid,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+       .ndo_poll_controller    = ixgb_netpoll,
+#endif
+};
+
 /**
  * ixgb_probe - Device Initialization Routine
  * @pdev: PCI device information struct
@@ -349,27 +355,25 @@ ixgb_reset(struct ixgb_adapter *adapter)
  **/
 
 static int __devinit
-ixgb_probe(struct pci_dev *pdev,
-               const struct pci_device_id *ent)
+ixgb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
        struct net_device *netdev = NULL;
        struct ixgb_adapter *adapter;
        static int cards_found = 0;
-       unsigned long mmio_start;
-       int mmio_len;
        int pci_using_dac;
        int i;
        int err;
 
-       if((err = pci_enable_device(pdev)))
+       err = pci_enable_device(pdev);
+       if (err)
                return err;
 
-       if(!(err = pci_set_dma_mask(pdev, DMA_64BIT_MASK)) &&
-          !(err = pci_set_consistent_dma_mask(pdev, DMA_64BIT_MASK))) {
+       if (!(err = pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) &&
+           !(err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64)))) {
                pci_using_dac = 1;
        } else {
-               if((err = pci_set_dma_mask(pdev, DMA_32BIT_MASK)) ||
-                  (err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK))) {
+               if ((err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) ||
+                   (err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32)))) {
                        printk(KERN_ERR
                         "ixgb: No usable DMA configuration, aborting\n");
                        goto err_dma_mask;
@@ -377,18 +381,18 @@ ixgb_probe(struct pci_dev *pdev,
                pci_using_dac = 0;
        }
 
-       if((err = pci_request_regions(pdev, ixgb_driver_name)))
+       err = pci_request_regions(pdev, ixgb_driver_name);
+       if (err)
                goto err_request_regions;
 
        pci_set_master(pdev);
 
        netdev = alloc_etherdev(sizeof(struct ixgb_adapter));
-       if(!netdev) {
+       if (!netdev) {
                err = -ENOMEM;
                goto err_alloc_etherdev;
        }
 
-       SET_MODULE_OWNER(netdev);
        SET_NETDEV_DEV(netdev, &pdev->dev);
 
        pci_set_drvdata(pdev, netdev);
@@ -398,49 +402,27 @@ ixgb_probe(struct pci_dev *pdev,
        adapter->hw.back = adapter;
        adapter->msg_enable = netif_msg_init(debug, DEFAULT_DEBUG_LEVEL_SHIFT);
 
-       mmio_start = pci_resource_start(pdev, BAR_0);
-       mmio_len = pci_resource_len(pdev, BAR_0);
-
-       adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
-       if(!adapter->hw.hw_addr) {
+       adapter->hw.hw_addr = pci_ioremap_bar(pdev, BAR_0);
+       if (!adapter->hw.hw_addr) {
                err = -EIO;
                goto err_ioremap;
        }
 
-       for(i = BAR_1; i <= BAR_5; i++) {
-               if(pci_resource_len(pdev, i) == 0)
+       for (i = BAR_1; i <= BAR_5; i++) {
+               if (pci_resource_len(pdev, i) == 0)
                        continue;
-               if(pci_resource_flags(pdev, i) & IORESOURCE_IO) {
+               if (pci_resource_flags(pdev, i) & IORESOURCE_IO) {
                        adapter->hw.io_base = pci_resource_start(pdev, i);
                        break;
                }
        }
 
-       netdev->open = &ixgb_open;
-       netdev->stop = &ixgb_close;
-       netdev->hard_start_xmit = &ixgb_xmit_frame;
-       netdev->get_stats = &ixgb_get_stats;
-       netdev->set_multicast_list = &ixgb_set_multi;
-       netdev->set_mac_address = &ixgb_set_mac;
-       netdev->change_mtu = &ixgb_change_mtu;
+       netdev->netdev_ops = &ixgb_netdev_ops;
        ixgb_set_ethtool_ops(netdev);
-       netdev->tx_timeout = &ixgb_tx_timeout;
        netdev->watchdog_timeo = 5 * HZ;
-#ifdef CONFIG_IXGB_NAPI
-       netdev->poll = &ixgb_clean;
-       netdev->weight = 64;
-#endif
-       netdev->vlan_rx_register = ixgb_vlan_rx_register;
-       netdev->vlan_rx_add_vid = ixgb_vlan_rx_add_vid;
-       netdev->vlan_rx_kill_vid = ixgb_vlan_rx_kill_vid;
-#ifdef CONFIG_NET_POLL_CONTROLLER
-       netdev->poll_controller = ixgb_netpoll;
-#endif
+       netif_napi_add(netdev, &adapter->napi, ixgb_clean, 64);
 
        strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
-       netdev->mem_start = mmio_start;
-       netdev->mem_end = mmio_start + mmio_len;
-       netdev->base_addr = adapter->hw.io_base;
 
        adapter->bd_number = cards_found;
        adapter->link_speed = 0;
@@ -448,7 +430,8 @@ ixgb_probe(struct pci_dev *pdev,
 
        /* setup the private structure */
 
-       if((err = ixgb_sw_init(adapter)))
+       err = ixgb_sw_init(adapter);
+       if (err)
                goto err_sw_init;
 
        netdev->features = NETIF_F_SG |
@@ -456,19 +439,14 @@ ixgb_probe(struct pci_dev *pdev,
                           NETIF_F_HW_VLAN_TX |
                           NETIF_F_HW_VLAN_RX |
                           NETIF_F_HW_VLAN_FILTER;
-#ifdef NETIF_F_TSO
        netdev->features |= NETIF_F_TSO;
-#endif
-#ifdef NETIF_F_LLTX
-       netdev->features |= NETIF_F_LLTX;
-#endif
 
-       if(pci_using_dac)
+       if (pci_using_dac)
                netdev->features |= NETIF_F_HIGHDMA;
 
        /* make sure the EEPROM is good */
 
-       if(!ixgb_validate_eeprom_checksum(&adapter->hw)) {
+       if (!ixgb_validate_eeprom_checksum(&adapter->hw)) {
                DPRINTK(PROBE, ERR, "The EEPROM Checksum Is Not Valid\n");
                err = -EIO;
                goto err_eeprom;
@@ -477,7 +455,7 @@ ixgb_probe(struct pci_dev *pdev,
        ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
        memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
 
-       if(!is_valid_ether_addr(netdev->perm_addr)) {
+       if (!is_valid_ether_addr(netdev->perm_addr)) {
                DPRINTK(PROBE, ERR, "Invalid MAC Address\n");
                err = -EIO;
                goto err_eeprom;
@@ -489,17 +467,15 @@ ixgb_probe(struct pci_dev *pdev,
        adapter->watchdog_timer.function = &ixgb_watchdog;
        adapter->watchdog_timer.data = (unsigned long)adapter;
 
-       INIT_WORK(&adapter->tx_timeout_task,
-                 (void (*)(void *))ixgb_tx_timeout_task, netdev);
+       INIT_WORK(&adapter->tx_timeout_task, ixgb_tx_timeout_task);
 
        strcpy(netdev->name, "eth%d");
-       if((err = register_netdev(netdev)))
+       err = register_netdev(netdev);
+       if (err)
                goto err_register;
 
-       /* we're going to reset, so assume we have no link for now */
-
+       /* carrier off reporting is important to ethtool even BEFORE open */
        netif_carrier_off(netdev);
-       netif_stop_queue(netdev);
 
        DPRINTK(PROBE, INFO, "Intel(R) PRO/10GbE Network Connection\n");
        ixgb_check_options(adapter);
@@ -540,6 +516,8 @@ ixgb_remove(struct pci_dev *pdev)
        struct net_device *netdev = pci_get_drvdata(pdev);
        struct ixgb_adapter *adapter = netdev_priv(netdev);
 
+       flush_scheduled_work();
+
        unregister_netdev(netdev);
 
        iounmap(adapter->hw.hw_addr);
@@ -572,13 +550,13 @@ ixgb_sw_init(struct ixgb_adapter *adapter)
        hw->subsystem_id = pdev->subsystem_device;
 
        hw->max_frame_size = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
-       adapter->rx_buffer_len = hw->max_frame_size;
+       adapter->rx_buffer_len = hw->max_frame_size + 8; /* + 8 for errata */
 
-       if((hw->device_id == IXGB_DEVICE_ID_82597EX)
+       if ((hw->device_id == IXGB_DEVICE_ID_82597EX)
           || (hw->device_id == IXGB_DEVICE_ID_82597EX_CX4)
           || (hw->device_id == IXGB_DEVICE_ID_82597EX_LR)
           || (hw->device_id == IXGB_DEVICE_ID_82597EX_SR))
-                       hw->mac_type = ixgb_82597;
+               hw->mac_type = ixgb_82597;
        else {
                /* should never have loaded on this device */
                DPRINTK(PROBE, ERR, "unsupported device id\n");
@@ -587,9 +565,7 @@ ixgb_sw_init(struct ixgb_adapter *adapter)
        /* enable flow control to be programmed */
        hw->fc.send_xon = 1;
 
-       atomic_set(&adapter->irq_sem, 1);
-       spin_lock_init(&adapter->tx_lock);
-
+       set_bit(__IXGB_DOWN, &adapter->flags);
        return 0;
 }
 
@@ -613,18 +589,24 @@ ixgb_open(struct net_device *netdev)
        int err;
 
        /* allocate transmit descriptors */
-
-       if((err = ixgb_setup_tx_resources(adapter)))
+       err = ixgb_setup_tx_resources(adapter);
+       if (err)
                goto err_setup_tx;
 
+       netif_carrier_off(netdev);
+
        /* allocate receive descriptors */
 
-       if((err = ixgb_setup_rx_resources(adapter)))
+       err = ixgb_setup_rx_resources(adapter);
+       if (err)
                goto err_setup_rx;
 
-       if((err = ixgb_up(adapter)))
+       err = ixgb_up(adapter);
+       if (err)
                goto err_up;
 
+       netif_start_queue(netdev);
+
        return 0;
 
 err_up:
@@ -654,7 +636,7 @@ ixgb_close(struct net_device *netdev)
 {
        struct ixgb_adapter *adapter = netdev_priv(netdev);
 
-       ixgb_down(adapter, TRUE);
+       ixgb_down(adapter, true);
 
        ixgb_free_tx_resources(adapter);
        ixgb_free_rx_resources(adapter);
@@ -678,7 +660,7 @@ ixgb_setup_tx_resources(struct ixgb_adapter *adapter)
 
        size = sizeof(struct ixgb_buffer) * txdr->count;
        txdr->buffer_info = vmalloc(size);
-       if(!txdr->buffer_info) {
+       if (!txdr->buffer_info) {
                DPRINTK(PROBE, ERR,
                 "Unable to allocate transmit descriptor ring memory\n");
                return -ENOMEM;
@@ -688,10 +670,10 @@ ixgb_setup_tx_resources(struct ixgb_adapter *adapter)
        /* round up to nearest 4K */
 
        txdr->size = txdr->count * sizeof(struct ixgb_tx_desc);
-       IXGB_ROUNDUP(txdr->size, 4096);
+       txdr->size = ALIGN(txdr->size, 4096);
 
        txdr->desc = pci_alloc_consistent(pdev, txdr->size, &txdr->dma);
-       if(!txdr->desc) {
+       if (!txdr->desc) {
                vfree(txdr->buffer_info);
                DPRINTK(PROBE, ERR,
                 "Unable to allocate transmit descriptor memory\n");
@@ -715,13 +697,13 @@ ixgb_setup_tx_resources(struct ixgb_adapter *adapter)
 static void
 ixgb_configure_tx(struct ixgb_adapter *adapter)
 {
-       uint64_t tdba = adapter->tx_ring.dma;
-       uint32_t tdlen = adapter->tx_ring.count * sizeof(struct ixgb_tx_desc);
-       uint32_t tctl;
+       u64 tdba = adapter->tx_ring.dma;
+       u32 tdlen = adapter->tx_ring.count * sizeof(struct ixgb_tx_desc);
+       u32 tctl;
        struct ixgb_hw *hw = &adapter->hw;
 
-       /* Setup the Base and Length of the Tx Descriptor Ring 
-        * tx_ring.dma can be either a 32 or 64 bit value 
+       /* Setup the Base and Length of the Tx Descriptor Ring
+        * tx_ring.dma can be either a 32 or 64 bit value
         */
 
        IXGB_WRITE_REG(hw, TDBAL, (tdba & 0x00000000ffffffffULL));
@@ -747,8 +729,8 @@ ixgb_configure_tx(struct ixgb_adapter *adapter)
 
        /* Setup Transmit Descriptor Settings for this adapter */
        adapter->tx_cmd_type =
-               IXGB_TX_DESC_TYPE 
-               (adapter->tx_int_delay_enable ? IXGB_TX_DESC_CMD_IDE : 0);
+               IXGB_TX_DESC_TYPE |
+               (adapter->tx_int_delay_enable ? IXGB_TX_DESC_CMD_IDE : 0);
 }
 
 /**
@@ -767,7 +749,7 @@ ixgb_setup_rx_resources(struct ixgb_adapter *adapter)
 
        size = sizeof(struct ixgb_buffer) * rxdr->count;
        rxdr->buffer_info = vmalloc(size);
-       if(!rxdr->buffer_info) {
+       if (!rxdr->buffer_info) {
                DPRINTK(PROBE, ERR,
                 "Unable to allocate receive descriptor ring\n");
                return -ENOMEM;
@@ -777,11 +759,11 @@ ixgb_setup_rx_resources(struct ixgb_adapter *adapter)
        /* Round up to nearest 4K */
 
        rxdr->size = rxdr->count * sizeof(struct ixgb_rx_desc);
-       IXGB_ROUNDUP(rxdr->size, 4096);
+       rxdr->size = ALIGN(rxdr->size, 4096);
 
        rxdr->desc = pci_alloc_consistent(pdev, rxdr->size, &rxdr->dma);
 
-       if(!rxdr->desc) {
+       if (!rxdr->desc) {
                vfree(rxdr->buffer_info);
                DPRINTK(PROBE, ERR,
                 "Unable to allocate receive descriptors\n");
@@ -803,15 +785,15 @@ ixgb_setup_rx_resources(struct ixgb_adapter *adapter)
 static void
 ixgb_setup_rctl(struct ixgb_adapter *adapter)
 {
-       uint32_t rctl;
+       u32 rctl;
 
        rctl = IXGB_READ_REG(&adapter->hw, RCTL);
 
        rctl &= ~(3 << IXGB_RCTL_MO_SHIFT);
 
        rctl |=
-               IXGB_RCTL_BAM | IXGB_RCTL_RDMTS_1_2 | 
-               IXGB_RCTL_RXEN | IXGB_RCTL_CFF | 
+               IXGB_RCTL_BAM | IXGB_RCTL_RDMTS_1_2 |
+               IXGB_RCTL_RXEN | IXGB_RCTL_CFF |
                (adapter->hw.mc_filter_type << IXGB_RCTL_MO_SHIFT);
 
        rctl |= IXGB_RCTL_SECRC;
@@ -838,12 +820,11 @@ ixgb_setup_rctl(struct ixgb_adapter *adapter)
 static void
 ixgb_configure_rx(struct ixgb_adapter *adapter)
 {
-       uint64_t rdba = adapter->rx_ring.dma;
-       uint32_t rdlen = adapter->rx_ring.count * sizeof(struct ixgb_rx_desc);
+       u64 rdba = adapter->rx_ring.dma;
+       u32 rdlen = adapter->rx_ring.count * sizeof(struct ixgb_rx_desc);
        struct ixgb_hw *hw = &adapter->hw;
-       uint32_t rctl;
-       uint32_t rxcsum;
-       uint32_t rxdctl;
+       u32 rctl;
+       u32 rxcsum;
 
        /* make sure receives are disabled while setting up the descriptors */
 
@@ -865,21 +846,15 @@ ixgb_configure_rx(struct ixgb_adapter *adapter)
        IXGB_WRITE_REG(hw, RDH, 0);
        IXGB_WRITE_REG(hw, RDT, 0);
 
-       /* set up pre-fetching of receive buffers so we get some before we
-        * run out (default hardware behavior is to run out before fetching
-        * more).  This sets up to fetch if HTHRESH rx descriptors are avail
-        * and the descriptors in hw cache are below PTHRESH.  This avoids
-        * the hardware behavior of fetching <=512 descriptors in a single
-        * burst that pre-empts all other activity, usually causing fifo
-        * overflows. */
-       /* use WTHRESH to burst write 16 descriptors or burst when RXT0 */
-       rxdctl = RXDCTL_WTHRESH_DEFAULT << IXGB_RXDCTL_WTHRESH_SHIFT |
-                RXDCTL_HTHRESH_DEFAULT << IXGB_RXDCTL_HTHRESH_SHIFT |
-                RXDCTL_PTHRESH_DEFAULT << IXGB_RXDCTL_PTHRESH_SHIFT;
-       IXGB_WRITE_REG(hw, RXDCTL, rxdctl);
+       /* due to the hardware errata with RXDCTL, we are unable to use any of
+        * the performance enhancing features of it without causing other
+        * subtle bugs, some of the bugs could include receive length
+        * corruption at high data rates (WTHRESH > 0) and/or receive
+        * descriptor ring irregularites (particularly in hardware cache) */
+       IXGB_WRITE_REG(hw, RXDCTL, 0);
 
        /* Enable Receive Checksum Offload for TCP and UDP */
-       if(adapter->rx_csum == TRUE) {
+       if (adapter->rx_csum) {
                rxcsum = IXGB_READ_REG(hw, RXCSUM);
                rxcsum |= IXGB_RXCSUM_TUOFL;
                IXGB_WRITE_REG(hw, RXCSUM, rxcsum);
@@ -915,19 +890,23 @@ ixgb_free_tx_resources(struct ixgb_adapter *adapter)
 
 static void
 ixgb_unmap_and_free_tx_resource(struct ixgb_adapter *adapter,
-                                       struct ixgb_buffer *buffer_info)
+                                struct ixgb_buffer *buffer_info)
 {
-       struct pci_dev *pdev = adapter->pdev;
-
-       if (buffer_info->dma)
-               pci_unmap_page(pdev, buffer_info->dma, buffer_info->length,
-                              PCI_DMA_TODEVICE);
+       if (buffer_info->dma) {
+               if (buffer_info->mapped_as_page)
+                       pci_unmap_page(adapter->pdev, buffer_info->dma,
+                                      buffer_info->length, PCI_DMA_TODEVICE);
+               else
+                       pci_unmap_single(adapter->pdev, buffer_info->dma,
+                                        buffer_info->length,
+                                        PCI_DMA_TODEVICE);
+               buffer_info->dma = 0;
+       }
 
-       if (buffer_info->skb)
+       if (buffer_info->skb) {
                dev_kfree_skb_any(buffer_info->skb);
-
-       buffer_info->skb = NULL;
-       buffer_info->dma = 0;
+               buffer_info->skb = NULL;
+       }
        buffer_info->time_stamp = 0;
        /* these fields must always be initialized in tx
         * buffer_info->length = 0;
@@ -949,7 +928,7 @@ ixgb_clean_tx_ring(struct ixgb_adapter *adapter)
 
        /* Free all the Tx ring sk_buffs */
 
-       for(i = 0; i < tx_ring->count; i++) {
+       for (i = 0; i < tx_ring->count; i++) {
                buffer_info = &tx_ring->buffer_info[i];
                ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
        }
@@ -1007,17 +986,19 @@ ixgb_clean_rx_ring(struct ixgb_adapter *adapter)
 
        /* Free all the Rx ring sk_buffs */
 
-       for(i = 0; i < rx_ring->count; i++) {
+       for (i = 0; i < rx_ring->count; i++) {
                buffer_info = &rx_ring->buffer_info[i];
-               if(buffer_info->skb) {
-
+               if (buffer_info->dma) {
                        pci_unmap_single(pdev,
                                         buffer_info->dma,
                                         buffer_info->length,
                                         PCI_DMA_FROMDEVICE);
+                       buffer_info->dma = 0;
+                       buffer_info->length = 0;
+               }
 
+               if (buffer_info->skb) {
                        dev_kfree_skb(buffer_info->skb);
-
                        buffer_info->skb = NULL;
                }
        }
@@ -1050,7 +1031,7 @@ ixgb_set_mac(struct net_device *netdev, void *p)
        struct ixgb_adapter *adapter = netdev_priv(netdev);
        struct sockaddr *addr = p;
 
-       if(!is_valid_ether_addr(addr->sa_data))
+       if (!is_valid_ether_addr(addr->sa_data))
                return -EADDRNOTAVAIL;
 
        memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
@@ -1076,34 +1057,40 @@ ixgb_set_multi(struct net_device *netdev)
        struct ixgb_adapter *adapter = netdev_priv(netdev);
        struct ixgb_hw *hw = &adapter->hw;
        struct dev_mc_list *mc_ptr;
-       uint32_t rctl;
+       u32 rctl;
        int i;
 
        /* Check for Promiscuous and All Multicast modes */
 
        rctl = IXGB_READ_REG(hw, RCTL);
 
-       if(netdev->flags & IFF_PROMISC) {
+       if (netdev->flags & IFF_PROMISC) {
                rctl |= (IXGB_RCTL_UPE | IXGB_RCTL_MPE);
-       } else if(netdev->flags & IFF_ALLMULTI) {
-               rctl |= IXGB_RCTL_MPE;
-               rctl &= ~IXGB_RCTL_UPE;
+               rctl &= ~IXGB_RCTL_VFE;
        } else {
-               rctl &= ~(IXGB_RCTL_UPE | IXGB_RCTL_MPE);
+               if (netdev->flags & IFF_ALLMULTI) {
+                       rctl |= IXGB_RCTL_MPE;
+                       rctl &= ~IXGB_RCTL_UPE;
+               } else {
+                       rctl &= ~(IXGB_RCTL_UPE | IXGB_RCTL_MPE);
+               }
+               rctl |= IXGB_RCTL_VFE;
        }
 
-       if(netdev->mc_count > IXGB_MAX_NUM_MULTICAST_ADDRESSES) {
+       if (netdev->mc_count > IXGB_MAX_NUM_MULTICAST_ADDRESSES) {
                rctl |= IXGB_RCTL_MPE;
                IXGB_WRITE_REG(hw, RCTL, rctl);
        } else {
-               uint8_t mta[netdev->mc_count * IXGB_ETH_LENGTH_OF_ADDRESS];
+               u8 mta[IXGB_MAX_NUM_MULTICAST_ADDRESSES *
+                           IXGB_ETH_LENGTH_OF_ADDRESS];
 
                IXGB_WRITE_REG(hw, RCTL, rctl);
 
-               for(i = 0, mc_ptr = netdev->mc_list; mc_ptr;
-                       i++, mc_ptr = mc_ptr->next)
+               for (i = 0, mc_ptr = netdev->mc_list;
+                    mc_ptr;
+                    i++, mc_ptr = mc_ptr->next)
                        memcpy(&mta[i * IXGB_ETH_LENGTH_OF_ADDRESS],
-                                  mc_ptr->dmi_addr, IXGB_ETH_LENGTH_OF_ADDRESS);
+                              mc_ptr->dmi_addr, IXGB_ETH_LENGTH_OF_ADDRESS);
 
                ixgb_mc_addr_list_update(hw, mta, netdev->mc_count, 0);
        }
@@ -1128,40 +1115,47 @@ ixgb_watchdog(unsigned long data)
                netif_stop_queue(netdev);
        }
 
-       if(adapter->hw.link_up) {
-               if(!netif_carrier_ok(netdev)) {
-                       DPRINTK(LINK, INFO,
-                               "NIC Link is Up 10000 Mbps Full Duplex\n");
+       if (adapter->hw.link_up) {
+               if (!netif_carrier_ok(netdev)) {
+                       printk(KERN_INFO "ixgb: %s NIC Link is Up 10 Gbps "
+                              "Full Duplex, Flow Control: %s\n",
+                              netdev->name,
+                              (adapter->hw.fc.type == ixgb_fc_full) ?
+                               "RX/TX" :
+                               ((adapter->hw.fc.type == ixgb_fc_rx_pause) ?
+                                "RX" :
+                                ((adapter->hw.fc.type == ixgb_fc_tx_pause) ?
+                                 "TX" : "None")));
                        adapter->link_speed = 10000;
                        adapter->link_duplex = FULL_DUPLEX;
                        netif_carrier_on(netdev);
-                       netif_wake_queue(netdev);
                }
        } else {
-               if(netif_carrier_ok(netdev)) {
+               if (netif_carrier_ok(netdev)) {
                        adapter->link_speed = 0;
                        adapter->link_duplex = 0;
-                       DPRINTK(LINK, INFO, "NIC Link is Down\n");
+                       printk(KERN_INFO "ixgb: %s NIC Link is Down\n",
+                              netdev->name);
                        netif_carrier_off(netdev);
-                       netif_stop_queue(netdev);
-
                }
        }
 
        ixgb_update_stats(adapter);
 
-       if(!netif_carrier_ok(netdev)) {
-               if(IXGB_DESC_UNUSED(txdr) + 1 < txdr->count) {
+       if (!netif_carrier_ok(netdev)) {
+               if (IXGB_DESC_UNUSED(txdr) + 1 < txdr->count) {
                        /* We've lost link, so the controller stops DMA,
                         * but we've got queued Tx work that's never going
                         * to get done, so reset controller to flush Tx.
                         * (Do the reset outside of interrupt context). */
                        schedule_work(&adapter->tx_timeout_task);
+                       /* return immediately since reset is imminent */
+                       return;
                }
        }
 
        /* Force detection of hung controller every watchdog period */
-       adapter->detect_tx_hung = TRUE;
+       adapter->detect_tx_hung = true;
 
        /* generate an interrupt to force clean up of any stragglers */
        IXGB_WRITE_REG(&adapter->hw, ICS, IXGB_INT_TXDW);
@@ -1177,33 +1171,35 @@ ixgb_watchdog(unsigned long data)
 static int
 ixgb_tso(struct ixgb_adapter *adapter, struct sk_buff *skb)
 {
-#ifdef NETIF_F_TSO
        struct ixgb_context_desc *context_desc;
        unsigned int i;
-       uint8_t ipcss, ipcso, tucss, tucso, hdr_len;
-       uint16_t ipcse, tucse, mss;
+       u8 ipcss, ipcso, tucss, tucso, hdr_len;
+       u16 ipcse, tucse, mss;
        int err;
 
        if (likely(skb_is_gso(skb))) {
                struct ixgb_buffer *buffer_info;
+               struct iphdr *iph;
+
                if (skb_header_cloned(skb)) {
                        err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
                        if (err)
                                return err;
                }
 
-               hdr_len = ((skb->h.raw - skb->data) + (skb->h.th->doff << 2));
+               hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
                mss = skb_shinfo(skb)->gso_size;
-               skb->nh.iph->tot_len = 0;
-               skb->nh.iph->check = 0;
-               skb->h.th->check = ~csum_tcpudp_magic(skb->nh.iph->saddr,
-                                                     skb->nh.iph->daddr,
-                                                     0, IPPROTO_TCP, 0);
-               ipcss = skb->nh.raw - skb->data;
-               ipcso = (void *)&(skb->nh.iph->check) - (void *)skb->data;
-               ipcse = skb->h.raw - skb->data - 1;
-               tucss = skb->h.raw - skb->data;
-               tucso = (void *)&(skb->h.th->check) - (void *)skb->data;
+               iph = ip_hdr(skb);
+               iph->tot_len = 0;
+               iph->check = 0;
+               tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
+                                                        iph->daddr, 0,
+                                                        IPPROTO_TCP, 0);
+               ipcss = skb_network_offset(skb);
+               ipcso = (void *)&(iph->check) - (void *)skb->data;
+               ipcse = skb_transport_offset(skb) - 1;
+               tucss = skb_transport_offset(skb);
+               tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
                tucse = 0;
 
                i = adapter->tx_ring.next_to_use;
@@ -1221,7 +1217,7 @@ ixgb_tso(struct ixgb_adapter *adapter, struct sk_buff *skb)
                context_desc->hdr_len = hdr_len;
                context_desc->status = 0;
                context_desc->cmd_type_len = cpu_to_le32(
-                                                 IXGB_CONTEXT_DESC_TYPE 
+                                                 IXGB_CONTEXT_DESC_TYPE
                                                | IXGB_CONTEXT_DESC_CMD_TSE
                                                | IXGB_CONTEXT_DESC_CMD_IP
                                                | IXGB_CONTEXT_DESC_CMD_TCP
@@ -1229,27 +1225,26 @@ ixgb_tso(struct ixgb_adapter *adapter, struct sk_buff *skb)
                                                | (skb->len - (hdr_len)));
 
 
-               if(++i == adapter->tx_ring.count) i = 0;
+               if (++i == adapter->tx_ring.count) i = 0;
                adapter->tx_ring.next_to_use = i;
 
                return 1;
        }
-#endif
 
        return 0;
 }
 
-static boolean_t
+static bool
 ixgb_tx_csum(struct ixgb_adapter *adapter, struct sk_buff *skb)
 {
        struct ixgb_context_desc *context_desc;
        unsigned int i;
-       uint8_t css, cso;
+       u8 css, cso;
 
-       if(likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
+       if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
                struct ixgb_buffer *buffer_info;
-               css = skb->h.raw - skb->data;
-               cso = (skb->h.raw + skb->csum) - skb->data;
+               css = skb_transport_offset(skb);
+               cso = css + skb->csum_offset;
 
                i = adapter->tx_ring.next_to_use;
                context_desc = IXGB_CONTEXT_DESC(adapter->tx_ring, i);
@@ -1260,7 +1255,7 @@ ixgb_tx_csum(struct ixgb_adapter *adapter, struct sk_buff *skb)
                context_desc->tucso = cso;
                context_desc->tucse = 0;
                /* zero out any previously existing data in one instruction */
-               *(uint32_t *)&(context_desc->ipcss) = 0;
+               *(u32 *)&(context_desc->ipcss) = 0;
                context_desc->status = 0;
                context_desc->hdr_len = 0;
                context_desc->mss = 0;
@@ -1268,13 +1263,13 @@ ixgb_tx_csum(struct ixgb_adapter *adapter, struct sk_buff *skb)
                        cpu_to_le32(IXGB_CONTEXT_DESC_TYPE
                                    | IXGB_TX_DESC_CMD_IDE);
 
-               if(++i == adapter->tx_ring.count) i = 0;
+               if (++i == adapter->tx_ring.count) i = 0;
                adapter->tx_ring.next_to_use = i;
 
-               return TRUE;
+               return true;
        }
 
-       return FALSE;
+       return false;
 }
 
 #define IXGB_MAX_TXD_PWR       14
@@ -1285,67 +1280,101 @@ ixgb_tx_map(struct ixgb_adapter *adapter, struct sk_buff *skb,
            unsigned int first)
 {
        struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
+       struct pci_dev *pdev = adapter->pdev;
        struct ixgb_buffer *buffer_info;
-       int len = skb->len;
+       int len = skb_headlen(skb);
        unsigned int offset = 0, size, count = 0, i;
-
+       unsigned int mss = skb_shinfo(skb)->gso_size;
        unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
        unsigned int f;
 
-       len -= skb->data_len;
-
        i = tx_ring->next_to_use;
 
-       while(len) {
+       while (len) {
                buffer_info = &tx_ring->buffer_info[i];
                size = min(len, IXGB_MAX_DATA_PER_TXD);
+               /* Workaround for premature desc write-backs
+                * in TSO mode.  Append 4-byte sentinel desc */
+               if (unlikely(mss && !nr_frags && size == len && size > 8))
+                       size -= 4;
+
                buffer_info->length = size;
                WARN_ON(buffer_info->dma != 0);
-               buffer_info->dma =
-                       pci_map_single(adapter->pdev,
-                               skb->data + offset,
-                               size,
-                               PCI_DMA_TODEVICE);
                buffer_info->time_stamp = jiffies;
+               buffer_info->mapped_as_page = false;
+               buffer_info->dma = pci_map_single(pdev, skb->data + offset,
+                                                 size, PCI_DMA_TODEVICE);
+               if (pci_dma_mapping_error(pdev, buffer_info->dma))
+                       goto dma_error;
                buffer_info->next_to_watch = 0;
 
                len -= size;
                offset += size;
                count++;
-               if(++i == tx_ring->count) i = 0;
+               if (len) {
+                       i++;
+                       if (i == tx_ring->count)
+                               i = 0;
+               }
        }
 
-       for(f = 0; f < nr_frags; f++) {
+       for (f = 0; f < nr_frags; f++) {
                struct skb_frag_struct *frag;
 
                frag = &skb_shinfo(skb)->frags[f];
                len = frag->size;
-               offset = 0;
+               offset = frag->page_offset;
+
+               while (len) {
+                       i++;
+                       if (i == tx_ring->count)
+                               i = 0;
 
-               while(len) {
                        buffer_info = &tx_ring->buffer_info[i];
                        size = min(len, IXGB_MAX_DATA_PER_TXD);
+
+                       /* Workaround for premature desc write-backs
+                        * in TSO mode.  Append 4-byte sentinel desc */
+                       if (unlikely(mss && (f == (nr_frags - 1))
+                                    && size == len && size > 8))
+                               size -= 4;
+
                        buffer_info->length = size;
-                       buffer_info->dma =
-                               pci_map_page(adapter->pdev,
-                                       frag->page,
-                                       frag->page_offset + offset,
-                                       size,
-                                       PCI_DMA_TODEVICE);
                        buffer_info->time_stamp = jiffies;
+                       buffer_info->mapped_as_page = true;
+                       buffer_info->dma =
+                               pci_map_page(pdev, frag->page,
+                                            offset, size,
+                                            PCI_DMA_TODEVICE);
+                       if (pci_dma_mapping_error(pdev, buffer_info->dma))
+                               goto dma_error;
                        buffer_info->next_to_watch = 0;
 
                        len -= size;
                        offset += size;
                        count++;
-                       if(++i == tx_ring->count) i = 0;
                }
        }
-       i = (i == 0) ? tx_ring->count - 1 : i - 1;
        tx_ring->buffer_info[i].skb = skb;
        tx_ring->buffer_info[first].next_to_watch = i;
 
        return count;
+
+dma_error:
+       dev_err(&pdev->dev, "TX DMA map failed\n");
+       buffer_info->dma = 0;
+       if (count)
+               count--;
+
+       while (count--) {
+               if (i==0)
+                       i += tx_ring->count;
+               i--;
+               buffer_info = &tx_ring->buffer_info[i];
+               ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
+       }
+
+       return 0;
 }
 
 static void
@@ -1354,26 +1383,25 @@ ixgb_tx_queue(struct ixgb_adapter *adapter, int count, int vlan_id,int tx_flags)
        struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
        struct ixgb_tx_desc *tx_desc = NULL;
        struct ixgb_buffer *buffer_info;
-       uint32_t cmd_type_len = adapter->tx_cmd_type;
-       uint8_t status = 0;
-       uint8_t popts = 0;
+       u32 cmd_type_len = adapter->tx_cmd_type;
+       u8 status = 0;
+       u8 popts = 0;
        unsigned int i;
 
-       if(tx_flags & IXGB_TX_FLAGS_TSO) {
+       if (tx_flags & IXGB_TX_FLAGS_TSO) {
                cmd_type_len |= IXGB_TX_DESC_CMD_TSE;
                popts |= (IXGB_TX_DESC_POPTS_IXSM | IXGB_TX_DESC_POPTS_TXSM);
        }
 
-       if(tx_flags & IXGB_TX_FLAGS_CSUM)
+       if (tx_flags & IXGB_TX_FLAGS_CSUM)
                popts |= IXGB_TX_DESC_POPTS_TXSM;
 
-       if(tx_flags & IXGB_TX_FLAGS_VLAN) {
+       if (tx_flags & IXGB_TX_FLAGS_VLAN)
                cmd_type_len |= IXGB_TX_DESC_CMD_VLE;
-       }
 
        i = tx_ring->next_to_use;
 
-       while(count--) {
+       while (count--) {
                buffer_info = &tx_ring->buffer_info[i];
                tx_desc = IXGB_TX_DESC(*tx_ring, i);
                tx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
@@ -1383,11 +1411,11 @@ ixgb_tx_queue(struct ixgb_adapter *adapter, int count, int vlan_id,int tx_flags)
                tx_desc->popts = popts;
                tx_desc->vlan = cpu_to_le16(vlan_id);
 
-               if(++i == tx_ring->count) i = 0;
+               if (++i == tx_ring->count) i = 0;
        }
 
-       tx_desc->cmd_type_len |= cpu_to_le32(IXGB_TX_DESC_CMD_EOP 
-                               | IXGB_TX_DESC_CMD_RS );
+       tx_desc->cmd_type_len |=
+               cpu_to_le32(IXGB_TX_DESC_CMD_EOP | IXGB_TX_DESC_CMD_RS);
 
        /* Force memory writes to complete before letting h/w
         * know there are new descriptors to fetch.  (Only
@@ -1399,82 +1427,99 @@ ixgb_tx_queue(struct ixgb_adapter *adapter, int count, int vlan_id,int tx_flags)
        IXGB_WRITE_REG(&adapter->hw, TDT, i);
 }
 
+static int __ixgb_maybe_stop_tx(struct net_device *netdev, int size)
+{
+       struct ixgb_adapter *adapter = netdev_priv(netdev);
+       struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
+
+       netif_stop_queue(netdev);
+       /* Herbert's original patch had:
+        *  smp_mb__after_netif_stop_queue();
+        * but since that doesn't exist yet, just open code it. */
+       smp_mb();
+
+       /* We need to check again in a case another CPU has just
+        * made room available. */
+       if (likely(IXGB_DESC_UNUSED(tx_ring) < size))
+               return -EBUSY;
+
+       /* A reprieve! */
+       netif_start_queue(netdev);
+       ++adapter->restart_queue;
+       return 0;
+}
+
+static int ixgb_maybe_stop_tx(struct net_device *netdev,
+                              struct ixgb_desc_ring *tx_ring, int size)
+{
+       if (likely(IXGB_DESC_UNUSED(tx_ring) >= size))
+               return 0;
+       return __ixgb_maybe_stop_tx(netdev, size);
+}
+
+
 /* Tx Descriptors needed, worst case */
 #define TXD_USE_COUNT(S) (((S) >> IXGB_MAX_TXD_PWR) + \
                         (((S) & (IXGB_MAX_DATA_PER_TXD - 1)) ? 1 : 0))
-#define DESC_NEEDED TXD_USE_COUNT(IXGB_MAX_DATA_PER_TXD) + \
-       MAX_SKB_FRAGS * TXD_USE_COUNT(PAGE_SIZE) + 1
+#define DESC_NEEDED TXD_USE_COUNT(IXGB_MAX_DATA_PER_TXD) /* skb->date */ + \
+       MAX_SKB_FRAGS * TXD_USE_COUNT(PAGE_SIZE) + 1 /* for context */ \
+       + 1 /* one more needed for sentinel TSO workaround */
 
-static int
+static netdev_tx_t
 ixgb_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
 {
        struct ixgb_adapter *adapter = netdev_priv(netdev);
        unsigned int first;
        unsigned int tx_flags = 0;
-       unsigned long flags;
        int vlan_id = 0;
+       int count = 0;
        int tso;
 
-       if(skb->len <= 0) {
-               dev_kfree_skb_any(skb);
-               return 0;
+       if (test_bit(__IXGB_DOWN, &adapter->flags)) {
+               dev_kfree_skb(skb);
+               return NETDEV_TX_OK;
        }
 
-#ifdef NETIF_F_LLTX
-       local_irq_save(flags);
-       if (!spin_trylock(&adapter->tx_lock)) {
-               /* Collision - tell upper layer to requeue */
-               local_irq_restore(flags);
-               return NETDEV_TX_LOCKED;
+       if (skb->len <= 0) {
+               dev_kfree_skb(skb);
+               return NETDEV_TX_OK;
        }
-#else
-       spin_lock_irqsave(&adapter->tx_lock, flags);
-#endif
 
-       if(unlikely(IXGB_DESC_UNUSED(&adapter->tx_ring) < DESC_NEEDED)) {
-               netif_stop_queue(netdev);
-               spin_unlock_irqrestore(&adapter->tx_lock, flags);
+       if (unlikely(ixgb_maybe_stop_tx(netdev, &adapter->tx_ring,
+                     DESC_NEEDED)))
                return NETDEV_TX_BUSY;
-       }
 
-#ifndef NETIF_F_LLTX
-       spin_unlock_irqrestore(&adapter->tx_lock, flags);
-#endif
-
-       if(adapter->vlgrp && vlan_tx_tag_present(skb)) {
+       if (adapter->vlgrp && vlan_tx_tag_present(skb)) {
                tx_flags |= IXGB_TX_FLAGS_VLAN;
                vlan_id = vlan_tx_tag_get(skb);
        }
 
        first = adapter->tx_ring.next_to_use;
-       
+
        tso = ixgb_tso(adapter, skb);
        if (tso < 0) {
-               dev_kfree_skb_any(skb);
-#ifdef NETIF_F_LLTX
-               spin_unlock_irqrestore(&adapter->tx_lock, flags);
-#endif
+               dev_kfree_skb(skb);
                return NETDEV_TX_OK;
        }
 
        if (likely(tso))
                tx_flags |= IXGB_TX_FLAGS_TSO;
-       else if(ixgb_tx_csum(adapter, skb))
+       else if (ixgb_tx_csum(adapter, skb))
                tx_flags |= IXGB_TX_FLAGS_CSUM;
 
-       ixgb_tx_queue(adapter, ixgb_tx_map(adapter, skb, first), vlan_id,
-                       tx_flags);
+       count = ixgb_tx_map(adapter, skb, first);
 
-       netdev->trans_start = jiffies;
+       if (count) {
+               ixgb_tx_queue(adapter, count, vlan_id, tx_flags);
+               /* Make sure there is space in the ring for the next send. */
+               ixgb_maybe_stop_tx(netdev, &adapter->tx_ring, DESC_NEEDED);
 
-#ifdef NETIF_F_LLTX
-       /* Make sure there is space in the ring for the next send. */
-       if(unlikely(IXGB_DESC_UNUSED(&adapter->tx_ring) < DESC_NEEDED))
-               netif_stop_queue(netdev);
-
-       spin_unlock_irqrestore(&adapter->tx_lock, flags);
+       } else {
+               dev_kfree_skb_any(skb);
+               adapter->tx_ring.buffer_info[first].time_stamp = 0;
+               adapter->tx_ring.next_to_use = first;
+       }
 
-#endif
        return NETDEV_TX_OK;
 }
 
@@ -1493,12 +1538,13 @@ ixgb_tx_timeout(struct net_device *netdev)
 }
 
 static void
-ixgb_tx_timeout_task(struct net_device *netdev)
+ixgb_tx_timeout_task(struct work_struct *work)
 {
-       struct ixgb_adapter *adapter = netdev_priv(netdev);
+       struct ixgb_adapter *adapter =
+               container_of(work, struct ixgb_adapter, tx_timeout_task);
 
        adapter->tx_timeout_count++;
-       ixgb_down(adapter, TRUE);
+       ixgb_down(adapter, true);
        ixgb_up(adapter);
 }
 
@@ -1513,9 +1559,7 @@ ixgb_tx_timeout_task(struct net_device *netdev)
 static struct net_device_stats *
 ixgb_get_stats(struct net_device *netdev)
 {
-       struct ixgb_adapter *adapter = netdev_priv(netdev);
-
-       return &adapter->net_stats;
+       return &netdev->stats;
 }
 
 /**
@@ -1533,21 +1577,25 @@ ixgb_change_mtu(struct net_device *netdev, int new_mtu)
        int max_frame = new_mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
        int old_max_frame = netdev->mtu + ENET_HEADER_SIZE + ENET_FCS_LENGTH;
 
-
-       if((max_frame < IXGB_MIN_ENET_FRAME_SIZE_WITHOUT_FCS + ENET_FCS_LENGTH)
-          || (max_frame > IXGB_MAX_JUMBO_FRAME_SIZE + ENET_FCS_LENGTH)) {
+       /* MTU < 68 is an error for IPv4 traffic, just don't allow it */
+       if ((new_mtu < 68) ||
+           (max_frame > IXGB_MAX_JUMBO_FRAME_SIZE + ENET_FCS_LENGTH)) {
                DPRINTK(PROBE, ERR, "Invalid MTU setting %d\n", new_mtu);
                return -EINVAL;
        }
 
-       adapter->rx_buffer_len = max_frame;
+       if (old_max_frame == max_frame)
+               return 0;
+
+       if (netif_running(netdev))
+               ixgb_down(adapter, true);
+
+       adapter->rx_buffer_len = max_frame + 8; /* + 8 for errata */
 
        netdev->mtu = new_mtu;
 
-       if ((old_max_frame != max_frame) && netif_running(netdev)) {
-               ixgb_down(adapter, TRUE);
+       if (netif_running(netdev))
                ixgb_up(adapter);
-       }
 
        return 0;
 }
@@ -1564,24 +1612,24 @@ ixgb_update_stats(struct ixgb_adapter *adapter)
        struct pci_dev *pdev = adapter->pdev;
 
        /* Prevent stats update while adapter is being reset */
-       if (pdev->error_state && pdev->error_state != pci_channel_io_normal)
+       if (pci_channel_offline(pdev))
                return;
 
-       if((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
+       if ((netdev->flags & IFF_PROMISC) || (netdev->flags & IFF_ALLMULTI) ||
           (netdev->mc_count > IXGB_MAX_NUM_MULTICAST_ADDRESSES)) {
                u64 multi = IXGB_READ_REG(&adapter->hw, MPRCL);
                u32 bcast_l = IXGB_READ_REG(&adapter->hw, BPRCL);
                u32 bcast_h = IXGB_READ_REG(&adapter->hw, BPRCH);
-               u64 bcast = ((u64)bcast_h << 32) | bcast_l; 
+               u64 bcast = ((u64)bcast_h << 32) | bcast_l;
 
                multi |= ((u64)IXGB_READ_REG(&adapter->hw, MPRCH) << 32);
                /* fix up multicast stats by removing broadcasts */
-               if(multi >= bcast)
+               if (multi >= bcast)
                        multi -= bcast;
-               
+
                adapter->stats.mprcl += (multi & 0xFFFFFFFF);
                adapter->stats.mprch += (multi >> 32);
-               adapter->stats.bprcl += bcast_l; 
+               adapter->stats.bprcl += bcast_l;
                adapter->stats.bprch += bcast_h;
        } else {
                adapter->stats.mprcl += IXGB_READ_REG(&adapter->hw, MPRCL);
@@ -1648,16 +1696,16 @@ ixgb_update_stats(struct ixgb_adapter *adapter)
 
        /* Fill out the OS statistics structure */
 
-       adapter->net_stats.rx_packets = adapter->stats.gprcl;
-       adapter->net_stats.tx_packets = adapter->stats.gptcl;
-       adapter->net_stats.rx_bytes = adapter->stats.gorcl;
-       adapter->net_stats.tx_bytes = adapter->stats.gotcl;
-       adapter->net_stats.multicast = adapter->stats.mprcl;
-       adapter->net_stats.collisions = 0;
+       netdev->stats.rx_packets = adapter->stats.gprcl;
+       netdev->stats.tx_packets = adapter->stats.gptcl;
+       netdev->stats.rx_bytes = adapter->stats.gorcl;
+       netdev->stats.tx_bytes = adapter->stats.gotcl;
+       netdev->stats.multicast = adapter->stats.mprcl;
+       netdev->stats.collisions = 0;
 
        /* ignore RLEC as it reports errors for padded (<64bytes) frames
         * with a length in the type/len field */
-       adapter->net_stats.rx_errors =
+       netdev->stats.rx_errors =
            /* adapter->stats.rnbc + */ adapter->stats.crcerrs +
            adapter->stats.ruc +
            adapter->stats.roc /*+ adapter->stats.rlec */  +
@@ -1665,21 +1713,21 @@ ixgb_update_stats(struct ixgb_adapter *adapter)
            adapter->stats.ecbc + adapter->stats.mpc;
 
        /* see above
-        * adapter->net_stats.rx_length_errors = adapter->stats.rlec;
+        * netdev->stats.rx_length_errors = adapter->stats.rlec;
         */
 
-       adapter->net_stats.rx_crc_errors = adapter->stats.crcerrs;
-       adapter->net_stats.rx_fifo_errors = adapter->stats.mpc;
-       adapter->net_stats.rx_missed_errors = adapter->stats.mpc;
-       adapter->net_stats.rx_over_errors = adapter->stats.mpc;
-
-       adapter->net_stats.tx_errors = 0;
-       adapter->net_stats.rx_frame_errors = 0;
-       adapter->net_stats.tx_aborted_errors = 0;
-       adapter->net_stats.tx_carrier_errors = 0;
-       adapter->net_stats.tx_fifo_errors = 0;
-       adapter->net_stats.tx_heartbeat_errors = 0;
-       adapter->net_stats.tx_window_errors = 0;
+       netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
+       netdev->stats.rx_fifo_errors = adapter->stats.mpc;
+       netdev->stats.rx_missed_errors = adapter->stats.mpc;
+       netdev->stats.rx_over_errors = adapter->stats.mpc;
+
+       netdev->stats.tx_errors = 0;
+       netdev->stats.rx_frame_errors = 0;
+       netdev->stats.tx_aborted_errors = 0;
+       netdev->stats.tx_carrier_errors = 0;
+       netdev->stats.tx_fifo_errors = 0;
+       netdev->stats.tx_heartbeat_errors = 0;
+       netdev->stats.tx_window_errors = 0;
 }
 
 #define IXGB_MAX_INTR 10
@@ -1687,88 +1735,65 @@ ixgb_update_stats(struct ixgb_adapter *adapter)
  * ixgb_intr - Interrupt Handler
  * @irq: interrupt number
  * @data: pointer to a network interface device structure
- * @pt_regs: CPU registers structure
  **/
 
 static irqreturn_t
-ixgb_intr(int irq, void *data, struct pt_regs *regs)
+ixgb_intr(int irq, void *data)
 {
        struct net_device *netdev = data;
        struct ixgb_adapter *adapter = netdev_priv(netdev);
        struct ixgb_hw *hw = &adapter->hw;
-       uint32_t icr = IXGB_READ_REG(hw, ICR);
-#ifndef CONFIG_IXGB_NAPI
-       unsigned int i;
-#endif
+       u32 icr = IXGB_READ_REG(hw, ICR);
 
-       if(unlikely(!icr))
+       if (unlikely(!icr))
                return IRQ_NONE;  /* Not our interrupt */
 
-       if(unlikely(icr & (IXGB_INT_RXSEQ | IXGB_INT_LSC))) {
-               mod_timer(&adapter->watchdog_timer, jiffies);
-       }
+       if (unlikely(icr & (IXGB_INT_RXSEQ | IXGB_INT_LSC)))
+               if (!test_bit(__IXGB_DOWN, &adapter->flags))
+                       mod_timer(&adapter->watchdog_timer, jiffies);
 
-#ifdef CONFIG_IXGB_NAPI
-       if(netif_rx_schedule_prep(netdev)) {
+       if (napi_schedule_prep(&adapter->napi)) {
 
-               /* Disable interrupts and register for poll. The flush 
+               /* Disable interrupts and register for poll. The flush
                  of the posted write is intentionally left out.
                */
 
-               atomic_inc(&adapter->irq_sem);
                IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
-               __netif_rx_schedule(netdev);
+               __napi_schedule(&adapter->napi);
        }
-#else
-       /* yes, that is actually a & and it is meant to make sure that
-        * every pass through this for loop checks both receive and
-        * transmit queues for completed descriptors, intended to
-        * avoid starvation issues and assist tx/rx fairness. */
-       for(i = 0; i < IXGB_MAX_INTR; i++)
-               if(!ixgb_clean_rx_irq(adapter) &
-                  !ixgb_clean_tx_irq(adapter))
-                       break;
-#endif 
        return IRQ_HANDLED;
 }
 
-#ifdef CONFIG_IXGB_NAPI
 /**
  * ixgb_clean - NAPI Rx polling callback
  * @adapter: board private structure
  **/
 
 static int
-ixgb_clean(struct net_device *netdev, int *budget)
+ixgb_clean(struct napi_struct *napi, int budget)
 {
-       struct ixgb_adapter *adapter = netdev_priv(netdev);
-       int work_to_do = min(*budget, netdev->quota);
-       int tx_cleaned;
+       struct ixgb_adapter *adapter = container_of(napi, struct ixgb_adapter, napi);
        int work_done = 0;
 
-       tx_cleaned = ixgb_clean_tx_irq(adapter);
-       ixgb_clean_rx_irq(adapter, &work_done, work_to_do);
+       ixgb_clean_tx_irq(adapter);
+       ixgb_clean_rx_irq(adapter, &work_done, budget);
 
-       *budget -= work_done;
-       netdev->quota -= work_done;
-
-       /* if no Tx and not enough Rx work done, exit the polling mode */
-       if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) {
-               netif_rx_complete(netdev);
-               ixgb_irq_enable(adapter);
-               return 0;
+       /* If budget not fully consumed, exit the polling mode */
+       if (work_done < budget) {
+               napi_complete(napi);
+               if (!test_bit(__IXGB_DOWN, &adapter->flags))
+                       ixgb_irq_enable(adapter);
        }
 
-       return 1;
+       return work_done;
 }
-#endif
 
 /**
  * ixgb_clean_tx_irq - Reclaim resources after transmit completes
  * @adapter: board private structure
  **/
 
-static boolean_t
+static bool
 ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
 {
        struct ixgb_desc_ring *tx_ring = &adapter->tx_ring;
@@ -1776,29 +1801,29 @@ ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
        struct ixgb_tx_desc *tx_desc, *eop_desc;
        struct ixgb_buffer *buffer_info;
        unsigned int i, eop;
-       boolean_t cleaned = FALSE;
+       bool cleaned = false;
 
        i = tx_ring->next_to_clean;
        eop = tx_ring->buffer_info[i].next_to_watch;
        eop_desc = IXGB_TX_DESC(*tx_ring, eop);
 
-       while(eop_desc->status & IXGB_TX_DESC_STATUS_DD) {
+       while (eop_desc->status & IXGB_TX_DESC_STATUS_DD) {
 
-               for(cleaned = FALSE; !cleaned; ) {
+               for (cleaned = false; !cleaned; ) {
                        tx_desc = IXGB_TX_DESC(*tx_ring, i);
                        buffer_info = &tx_ring->buffer_info[i];
 
-                       if (tx_desc->popts
-                           & (IXGB_TX_DESC_POPTS_TXSM |
-                              IXGB_TX_DESC_POPTS_IXSM))
+                       if (tx_desc->popts &
+                          (IXGB_TX_DESC_POPTS_TXSM |
+                           IXGB_TX_DESC_POPTS_IXSM))
                                adapter->hw_csum_tx_good++;
 
                        ixgb_unmap_and_free_tx_resource(adapter, buffer_info);
 
-                       *(uint32_t *)&(tx_desc->status) = 0;
+                       *(u32 *)&(tx_desc->status) = 0;
 
                        cleaned = (i == eop);
-                       if(++i == tx_ring->count) i = 0;
+                       if (++i == tx_ring->count) i = 0;
                }
 
                eop = tx_ring->buffer_info[i].next_to_watch;
@@ -1807,19 +1832,24 @@ ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
 
        tx_ring->next_to_clean = i;
 
-       if (unlikely(netif_queue_stopped(netdev))) {
-               spin_lock(&adapter->tx_lock);
-               if (netif_queue_stopped(netdev) && netif_carrier_ok(netdev) &&
-                   (IXGB_DESC_UNUSED(tx_ring) >= DESC_NEEDED))
+       if (unlikely(cleaned && netif_carrier_ok(netdev) &&
+                    IXGB_DESC_UNUSED(tx_ring) >= DESC_NEEDED)) {
+               /* Make sure that anybody stopping the queue after this
+                * sees the new next_to_clean. */
+               smp_mb();
+
+               if (netif_queue_stopped(netdev) &&
+                   !(test_bit(__IXGB_DOWN, &adapter->flags))) {
                        netif_wake_queue(netdev);
-               spin_unlock(&adapter->tx_lock);
+                       ++adapter->restart_queue;
+               }
        }
 
-       if(adapter->detect_tx_hung) {
+       if (adapter->detect_tx_hung) {
                /* detect a transmit hang in hardware, this serializes the
                 * check with the clearing of time_stamp and movement of i */
-               adapter->detect_tx_hung = FALSE;
-               if (tx_ring->buffer_info[eop].dma &&
+               adapter->detect_tx_hung = false;
+               if (tx_ring->buffer_info[eop].time_stamp &&
                   time_after(jiffies, tx_ring->buffer_info[eop].time_stamp + HZ)
                   && !(IXGB_READ_REG(&adapter->hw, STATUS) &
                        IXGB_STATUS_TXOFF)) {
@@ -1858,13 +1888,13 @@ ixgb_clean_tx_irq(struct ixgb_adapter *adapter)
 
 static void
 ixgb_rx_checksum(struct ixgb_adapter *adapter,
-                struct ixgb_rx_desc *rx_desc,
-                struct sk_buff *skb)
+                 struct ixgb_rx_desc *rx_desc,
+                 struct sk_buff *skb)
 {
        /* Ignore Checksum bit is set OR
         * TCP Checksum has not been calculated
         */
-       if((rx_desc->status & IXGB_RX_DESC_STATUS_IXSM) ||
+       if ((rx_desc->status & IXGB_RX_DESC_STATUS_IXSM) ||
           (!(rx_desc->status & IXGB_RX_DESC_STATUS_TCPCS))) {
                skb->ip_summed = CHECKSUM_NONE;
                return;
@@ -1872,7 +1902,7 @@ ixgb_rx_checksum(struct ixgb_adapter *adapter,
 
        /* At this point we know the hardware did the TCP checksum */
        /* now look at the TCP checksum error bit */
-       if(rx_desc->errors & IXGB_RX_DESC_ERRORS_TCPE) {
+       if (rx_desc->errors & IXGB_RX_DESC_ERRORS_TCPE) {
                /* let the stack verify checksum errors */
                skb->ip_summed = CHECKSUM_NONE;
                adapter->hw_csum_rx_error++;
@@ -1888,64 +1918,60 @@ ixgb_rx_checksum(struct ixgb_adapter *adapter,
  * @adapter: board private structure
  **/
 
-static boolean_t
-#ifdef CONFIG_IXGB_NAPI
+static bool
 ixgb_clean_rx_irq(struct ixgb_adapter *adapter, int *work_done, int work_to_do)
-#else
-ixgb_clean_rx_irq(struct ixgb_adapter *adapter)
-#endif
 {
        struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
        struct net_device *netdev = adapter->netdev;
        struct pci_dev *pdev = adapter->pdev;
        struct ixgb_rx_desc *rx_desc, *next_rxd;
        struct ixgb_buffer *buffer_info, *next_buffer, *next2_buffer;
-       uint32_t length;
+       u32 length;
        unsigned int i, j;
-       boolean_t cleaned = FALSE;
+       int cleaned_count = 0;
+       bool cleaned = false;
 
        i = rx_ring->next_to_clean;
        rx_desc = IXGB_RX_DESC(*rx_ring, i);
        buffer_info = &rx_ring->buffer_info[i];
 
-       while(rx_desc->status & IXGB_RX_DESC_STATUS_DD) {
-               struct sk_buff *skb, *next_skb;
+       while (rx_desc->status & IXGB_RX_DESC_STATUS_DD) {
+               struct sk_buff *skb;
                u8 status;
 
-#ifdef CONFIG_IXGB_NAPI
-               if(*work_done >= work_to_do)
+               if (*work_done >= work_to_do)
                        break;
 
                (*work_done)++;
-#endif
                status = rx_desc->status;
                skb = buffer_info->skb;
                buffer_info->skb = NULL;
 
-               prefetch(skb->data);
+               prefetch(skb->data - NET_IP_ALIGN);
 
-               if(++i == rx_ring->count) i = 0;
+               if (++i == rx_ring->count) i = 0;
                next_rxd = IXGB_RX_DESC(*rx_ring, i);
                prefetch(next_rxd);
 
-               if((j = i + 1) == rx_ring->count) j = 0;
+               if ((j = i + 1) == rx_ring->count) j = 0;
                next2_buffer = &rx_ring->buffer_info[j];
                prefetch(next2_buffer);
 
                next_buffer = &rx_ring->buffer_info[i];
-               next_skb = next_buffer->skb;
-               prefetch(next_skb);
 
-               cleaned = TRUE;
+               cleaned = true;
+               cleaned_count++;
 
                pci_unmap_single(pdev,
                                 buffer_info->dma,
                                 buffer_info->length,
                                 PCI_DMA_FROMDEVICE);
+               buffer_info->dma = 0;
 
                length = le16_to_cpu(rx_desc->length);
+               rx_desc->length = 0;
 
-               if(unlikely(!(status & IXGB_RX_DESC_STATUS_EOP))) {
+               if (unlikely(!(status & IXGB_RX_DESC_STATUS_EOP))) {
 
                        /* All receives must fit into a single buffer */
 
@@ -1956,11 +1982,9 @@ ixgb_clean_rx_irq(struct ixgb_adapter *adapter)
                        goto rxdesc_done;
                }
 
-               if (unlikely(rx_desc->errors
-                            & (IXGB_RX_DESC_ERRORS_CE | IXGB_RX_DESC_ERRORS_SE
-                               | IXGB_RX_DESC_ERRORS_P |
-                               IXGB_RX_DESC_ERRORS_RXE))) {
-
+               if (unlikely(rx_desc->errors &
+                   (IXGB_RX_DESC_ERRORS_CE | IXGB_RX_DESC_ERRORS_SE |
+                    IXGB_RX_DESC_ERRORS_P | IXGB_RX_DESC_ERRORS_RXE))) {
                        dev_kfree_skb_irq(skb);
                        goto rxdesc_done;
                }
@@ -1968,15 +1992,16 @@ ixgb_clean_rx_irq(struct ixgb_adapter *adapter)
                /* code added for copybreak, this should improve
                 * performance for small packets with large amounts
                 * of reassembly being done in the stack */
-#define IXGB_CB_LENGTH 256
-               if (length < IXGB_CB_LENGTH) {
+               if (length < copybreak) {
                        struct sk_buff *new_skb =
-                           netdev_alloc_skb(netdev, length + NET_IP_ALIGN);
+                           netdev_alloc_skb_ip_align(netdev, length);
                        if (new_skb) {
-                               skb_reserve(new_skb, NET_IP_ALIGN);
-                               memcpy(new_skb->data - NET_IP_ALIGN,
-                                      skb->data - NET_IP_ALIGN,
-                                      length + NET_IP_ALIGN);
+                               skb_copy_to_linear_data_offset(new_skb,
+                                                              -NET_IP_ALIGN,
+                                                              (skb->data -
+                                                               NET_IP_ALIGN),
+                                                              (length +
+                                                               NET_IP_ALIGN));
                                /* save the skb in buffer_info as good */
                                buffer_info->skb = skb;
                                skb = new_skb;
@@ -1991,29 +2016,23 @@ ixgb_clean_rx_irq(struct ixgb_adapter *adapter)
                ixgb_rx_checksum(adapter, rx_desc, skb);
 
                skb->protocol = eth_type_trans(skb, netdev);
-#ifdef CONFIG_IXGB_NAPI
-               if(adapter->vlgrp && (status & IXGB_RX_DESC_STATUS_VP)) {
+               if (adapter->vlgrp && (status & IXGB_RX_DESC_STATUS_VP)) {
                        vlan_hwaccel_receive_skb(skb, adapter->vlgrp,
-                               le16_to_cpu(rx_desc->special) &
-                                       IXGB_RX_DESC_SPECIAL_VLAN_MASK);
+                                               le16_to_cpu(rx_desc->special));
                } else {
                        netif_receive_skb(skb);
                }
-#else /* CONFIG_IXGB_NAPI */
-               if(adapter->vlgrp && (status & IXGB_RX_DESC_STATUS_VP)) {
-                       vlan_hwaccel_rx(skb, adapter->vlgrp,
-                               le16_to_cpu(rx_desc->special) &
-                                       IXGB_RX_DESC_SPECIAL_VLAN_MASK);
-               } else {
-                       netif_rx(skb);
-               }
-#endif /* CONFIG_IXGB_NAPI */
-               netdev->last_rx = jiffies;
 
 rxdesc_done:
                /* clean up descriptor, might be written over by hw */
                rx_desc->status = 0;
 
+               /* return some buffers to hardware, one at a time is too slow */
+               if (unlikely(cleaned_count >= IXGB_RX_BUFFER_WRITE)) {
+                       ixgb_alloc_rx_buffers(adapter, cleaned_count);
+                       cleaned_count = 0;
+               }
+
                /* use prefetched values */
                rx_desc = next_rxd;
                buffer_info = next_buffer;
@@ -2021,7 +2040,9 @@ rxdesc_done:
 
        rx_ring->next_to_clean = i;
 
-       ixgb_alloc_rx_buffers(adapter);
+       cleaned_count = IXGB_DESC_UNUSED(rx_ring);
+       if (cleaned_count)
+               ixgb_alloc_rx_buffers(adapter, cleaned_count);
 
        return cleaned;
 }
@@ -2032,7 +2053,7 @@ rxdesc_done:
  **/
 
 static void
-ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter)
+ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter, int cleaned_count)
 {
        struct ixgb_desc_ring *rx_ring = &adapter->rx_ring;
        struct net_device *netdev = adapter->netdev;
@@ -2041,17 +2062,15 @@ ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter)
        struct ixgb_buffer *buffer_info;
        struct sk_buff *skb;
        unsigned int i;
-       int num_group_tail_writes;
        long cleancount;
 
        i = rx_ring->next_to_use;
        buffer_info = &rx_ring->buffer_info[i];
        cleancount = IXGB_DESC_UNUSED(rx_ring);
 
-       num_group_tail_writes = IXGB_RX_BUFFER_WRITE;
 
        /* leave three descriptors unused */
-       while(--cleancount > 2) {
+       while (--cleancount > 2 && cleaned_count--) {
                /* recycle! its good for you */
                skb = buffer_info->skb;
                if (skb) {
@@ -2059,20 +2078,13 @@ ixgb_alloc_rx_buffers(struct ixgb_adapter *adapter)
                        goto map_skb;
                }
 
-               skb = netdev_alloc_skb(netdev, adapter->rx_buffer_len
-                                      + NET_IP_ALIGN);
+               skb = netdev_alloc_skb_ip_align(netdev, adapter->rx_buffer_len);
                if (unlikely(!skb)) {
                        /* Better luck next round */
                        adapter->alloc_rx_buff_failed++;
                        break;
                }
 
-               /* Make buffer alignment 2 beyond a 16 byte boundary
-                * this will result in a 16 byte aligned IP header after
-                * the 14 byte MAC header is removed
-                */
-               skb_reserve(skb, NET_IP_ALIGN);
-
                buffer_info->skb = skb;
                buffer_info->length = adapter->rx_buffer_len;
 map_skb:
@@ -2084,12 +2096,12 @@ map_skb:
                rx_desc = IXGB_RX_DESC(*rx_ring, i);
                rx_desc->buff_addr = cpu_to_le64(buffer_info->dma);
                /* guarantee DD bit not set now before h/w gets descriptor
-                * this is the rest of the workaround for h/w double 
+                * this is the rest of the workaround for h/w double
                 * writeback. */
                rx_desc->status = 0;
 
 
-               if(++i == rx_ring->count) i = 0;
+               if (++i == rx_ring->count) i = 0;
                buffer_info = &rx_ring->buffer_info[i];
        }
 
@@ -2109,7 +2121,7 @@ map_skb:
 
 /**
  * ixgb_vlan_rx_register - enables or disables vlan tagging/stripping.
- * 
+ *
  * @param netdev network interface device structure
  * @param grp indicates to enable or disable tagging/stripping
  **/
@@ -2117,12 +2129,12 @@ static void
 ixgb_vlan_rx_register(struct net_device *netdev, struct vlan_group *grp)
 {
        struct ixgb_adapter *adapter = netdev_priv(netdev);
-       uint32_t ctrl, rctl;
+       u32 ctrl, rctl;
 
        ixgb_irq_disable(adapter);
        adapter->vlgrp = grp;
 
-       if(grp) {
+       if (grp) {
                /* enable VLAN tag insert/strip */
                ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
                ctrl |= IXGB_CTRL0_VME;
@@ -2131,7 +2143,6 @@ ixgb_vlan_rx_register(struct net_device *netdev, struct vlan_group *grp)
                /* enable VLAN receive filtering */
 
                rctl = IXGB_READ_REG(&adapter->hw, RCTL);
-               rctl |= IXGB_RCTL_VFE;
                rctl &= ~IXGB_RCTL_CFIEN;
                IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
        } else {
@@ -2140,22 +2151,18 @@ ixgb_vlan_rx_register(struct net_device *netdev, struct vlan_group *grp)
                ctrl = IXGB_READ_REG(&adapter->hw, CTRL0);
                ctrl &= ~IXGB_CTRL0_VME;
                IXGB_WRITE_REG(&adapter->hw, CTRL0, ctrl);
-
-               /* disable VLAN filtering */
-
-               rctl = IXGB_READ_REG(&adapter->hw, RCTL);
-               rctl &= ~IXGB_RCTL_VFE;
-               IXGB_WRITE_REG(&adapter->hw, RCTL, rctl);
        }
 
-       ixgb_irq_enable(adapter);
+       /* don't enable interrupts unless we are UP */
+       if (adapter->netdev->flags & IFF_UP)
+               ixgb_irq_enable(adapter);
 }
 
 static void
-ixgb_vlan_rx_add_vid(struct net_device *netdev, uint16_t vid)
+ixgb_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
 {
        struct ixgb_adapter *adapter = netdev_priv(netdev);
-       uint32_t vfta, index;
+       u32 vfta, index;
 
        /* add VID to filter table */
 
@@ -2166,19 +2173,20 @@ ixgb_vlan_rx_add_vid(struct net_device *netdev, uint16_t vid)
 }
 
 static void
-ixgb_vlan_rx_kill_vid(struct net_device *netdev, uint16_t vid)
+ixgb_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
 {
        struct ixgb_adapter *adapter = netdev_priv(netdev);
-       uint32_t vfta, index;
+       u32 vfta, index;
 
        ixgb_irq_disable(adapter);
 
-       if(adapter->vlgrp)
-               adapter->vlgrp->vlan_devices[vid] = NULL;
+       vlan_group_set_device(adapter->vlgrp, vid, NULL);
 
-       ixgb_irq_enable(adapter);
+       /* don't enable interrupts unless we are UP */
+       if (adapter->netdev->flags & IFF_UP)
+               ixgb_irq_enable(adapter);
 
-       /* remove VID from filter table*/
+       /* remove VID from filter table */
 
        index = (vid >> 5) & 0x7F;
        vfta = IXGB_READ_REG_ARRAY(&adapter->hw, VFTA, index);
@@ -2191,10 +2199,10 @@ ixgb_restore_vlan(struct ixgb_adapter *adapter)
 {
        ixgb_vlan_rx_register(adapter->netdev, adapter->vlgrp);
 
-       if(adapter->vlgrp) {
-               uint16_t vid;
-               for(vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
-                       if(!adapter->vlgrp->vlan_devices[vid])
+       if (adapter->vlgrp) {
+               u16 vid;
+               for (vid = 0; vid < VLAN_GROUP_ARRAY_LEN; vid++) {
+                       if (!vlan_group_get_device(adapter->vlgrp, vid))
                                continue;
                        ixgb_vlan_rx_add_vid(adapter->netdev, vid);
                }
@@ -2213,7 +2221,7 @@ static void ixgb_netpoll(struct net_device *dev)
        struct ixgb_adapter *adapter = netdev_priv(dev);
 
        disable_irq(adapter->pdev->irq);
-       ixgb_intr(adapter->pdev->irq, dev, NULL);
+       ixgb_intr(adapter->pdev->irq, dev);
        enable_irq(adapter->pdev->irq);
 }
 #endif
@@ -2226,14 +2234,19 @@ static void ixgb_netpoll(struct net_device *dev)
  * This callback is called by the PCI subsystem whenever
  * a PCI bus error is detected.
  */
-static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
-                                    enum pci_channel_state state)
+static pci_ers_result_t ixgb_io_error_detected(struct pci_dev *pdev,
+                                               enum pci_channel_state state)
 {
        struct net_device *netdev = pci_get_drvdata(pdev);
-       struct ixgb_adapter *adapter = netdev->priv;
+       struct ixgb_adapter *adapter = netdev_priv(netdev);
 
-       if(netif_running(netdev))
-               ixgb_down(adapter, TRUE);
+       netif_device_detach(netdev);
+
+       if (state == pci_channel_io_perm_failure)
+               return PCI_ERS_RESULT_DISCONNECT;
+
+       if (netif_running(netdev))
+               ixgb_down(adapter, true);
 
        pci_disable_device(pdev);
 
@@ -2245,17 +2258,17 @@ static pci_ers_result_t ixgb_io_error_detected (struct pci_dev *pdev,
  * ixgb_io_slot_reset - called after the pci bus has been reset.
  * @pdev    pointer to pci device with error
  *
- * This callback is called after the PCI buss has been reset.
+ * This callback is called after the PCI bus has been reset.
  * Basically, this tries to restart the card from scratch.
  * This is a shortened version of the device probe/discovery code,
  * it resembles the first-half of the ixgb_probe() routine.
  */
-static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev)
+static pci_ers_result_t ixgb_io_slot_reset(struct pci_dev *pdev)
 {
        struct net_device *netdev = pci_get_drvdata(pdev);
-       struct ixgb_adapter *adapter = netdev->priv;
+       struct ixgb_adapter *adapter = netdev_priv(netdev);
 
-       if(pci_enable_device(pdev)) {
+       if (pci_enable_device(pdev)) {
                DPRINTK(PROBE, ERR, "Cannot re-enable PCI device after reset.\n");
                return PCI_ERS_RESULT_DISCONNECT;
        }
@@ -2271,14 +2284,14 @@ static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev)
        ixgb_reset(adapter);
 
        /* Make sure the EEPROM is good */
-       if(!ixgb_validate_eeprom_checksum(&adapter->hw)) {
+       if (!ixgb_validate_eeprom_checksum(&adapter->hw)) {
                DPRINTK(PROBE, ERR, "After reset, the EEPROM checksum is not valid.\n");
                return PCI_ERS_RESULT_DISCONNECT;
        }
        ixgb_get_ee_mac_addr(&adapter->hw, netdev->dev_addr);
        memcpy(netdev->perm_addr, netdev->dev_addr, netdev->addr_len);
 
-       if(!is_valid_ether_addr(netdev->perm_addr)) {
+       if (!is_valid_ether_addr(netdev->perm_addr)) {
                DPRINTK(PROBE, ERR, "After reset, invalid MAC address.\n");
                return PCI_ERS_RESULT_DISCONNECT;
        }
@@ -2294,15 +2307,15 @@ static pci_ers_result_t ixgb_io_slot_reset (struct pci_dev *pdev)
  * normal operation. Implementation resembles the second-half
  * of the ixgb_probe() routine.
  */
-static void ixgb_io_resume (struct pci_dev *pdev)
+static void ixgb_io_resume(struct pci_dev *pdev)
 {
        struct net_device *netdev = pci_get_drvdata(pdev);
-       struct ixgb_adapter *adapter = netdev->priv;
+       struct ixgb_adapter *adapter = netdev_priv(netdev);
 
        pci_set_master(pdev);
 
-       if(netif_running(netdev)) {
-               if(ixgb_up(adapter)) {
+       if (netif_running(netdev)) {
+               if (ixgb_up(adapter)) {
                        printk ("ixgb: can't bring device back up after reset\n");
                        return;
                }