// SPDX-License-Identifier: GPL-2.0-only
/*
 * QLogic qlge NIC HBA Driver
 * Copyright (c)  2003-2008 QLogic Corporation
 * Author:     Linux qlge network device driver by
 *                      Ron Mercer <ron.mercer@qlogic.com>
 */
#include <linux/kernel.h>
#include <linux/bitops.h>
#include <linux/types.h>
#include <linux/module.h>
#include <linux/list.h>
#include <linux/pci.h>
#include <linux/dma-mapping.h>
#include <linux/pagemap.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/dmapool.h>
#include <linux/mempool.h>
#include <linux/spinlock.h>
#include <linux/kthread.h>
#include <linux/interrupt.h>
#include <linux/errno.h>
#include <linux/ioport.h>
#include <linux/in.h>
#include <linux/ip.h>
#include <linux/ipv6.h>
#include <net/ipv6.h>
#include <linux/tcp.h>
#include <linux/udp.h>
#include <linux/if_arp.h>
#include <linux/if_ether.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/ethtool.h>
#include <linux/if_vlan.h>
#include <linux/skbuff.h>
#include <linux/delay.h>
#include <linux/mm.h>
#include <linux/vmalloc.h>
#include <linux/prefetch.h>
#include <net/ip6_checksum.h>

#include "qlge.h"
#include "qlge_devlink.h"

char qlge_driver_name[] = DRV_NAME;
const char qlge_driver_version[] = DRV_VERSION;

MODULE_AUTHOR("Ron Mercer <ron.mercer@qlogic.com>");
MODULE_DESCRIPTION(DRV_STRING " ");
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_VERSION);

static const u32 default_msg =
        NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK |
        NETIF_MSG_IFDOWN |
        NETIF_MSG_IFUP |
        NETIF_MSG_RX_ERR |
        NETIF_MSG_TX_ERR |
        NETIF_MSG_HW | NETIF_MSG_WOL | 0;

static int debug = -1;  /* defaults above */
module_param(debug, int, 0664);
MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");

#define MSIX_IRQ 0
#define MSI_IRQ 1
#define LEG_IRQ 2
static int qlge_irq_type = MSIX_IRQ;
module_param(qlge_irq_type, int, 0664);
MODULE_PARM_DESC(qlge_irq_type, "0 = MSI-X, 1 = MSI, 2 = Legacy.");

static int qlge_mpi_coredump;
module_param(qlge_mpi_coredump, int, 0);
MODULE_PARM_DESC(qlge_mpi_coredump,
                 "Option to enable MPI firmware dump. Default is OFF - Do Not allocate memory. ");

static int qlge_force_coredump;
module_param(qlge_force_coredump, int, 0);
MODULE_PARM_DESC(qlge_force_coredump,
                 "Option to allow force of firmware core dump. Default is OFF - Do not allow.");

static const struct pci_device_id qlge_pci_tbl[] = {
        {PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, QLGE_DEVICE_ID_8012)},
        {PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, QLGE_DEVICE_ID_8000)},
        /* required last entry */
        {0,}
};

MODULE_DEVICE_TABLE(pci, qlge_pci_tbl);

static int qlge_wol(struct qlge_adapter *);
static void qlge_set_multicast_list(struct net_device *);
static int qlge_adapter_down(struct qlge_adapter *);
static int qlge_adapter_up(struct qlge_adapter *);

/* This hardware semaphore causes exclusive access to
 * resources shared between the NIC driver, MPI firmware,
 * FCOE firmware and the FC driver.
 */
static int qlge_sem_trylock(struct qlge_adapter *qdev, u32 sem_mask)
{
        u32 sem_bits = 0;

        switch (sem_mask) {
        case SEM_XGMAC0_MASK:
                sem_bits = SEM_SET << SEM_XGMAC0_SHIFT;
                break;
        case SEM_XGMAC1_MASK:
                sem_bits = SEM_SET << SEM_XGMAC1_SHIFT;
                break;
        case SEM_ICB_MASK:
                sem_bits = SEM_SET << SEM_ICB_SHIFT;
                break;
        case SEM_MAC_ADDR_MASK:
                sem_bits = SEM_SET << SEM_MAC_ADDR_SHIFT;
                break;
        case SEM_FLASH_MASK:
                sem_bits = SEM_SET << SEM_FLASH_SHIFT;
                break;
        case SEM_PROBE_MASK:
                sem_bits = SEM_SET << SEM_PROBE_SHIFT;
                break;
        case SEM_RT_IDX_MASK:
                sem_bits = SEM_SET << SEM_RT_IDX_SHIFT;
                break;
        case SEM_PROC_REG_MASK:
                sem_bits = SEM_SET << SEM_PROC_REG_SHIFT;
                break;
        default:
                netif_alert(qdev, probe, qdev->ndev, "bad Semaphore mask!.\n");
                return -EINVAL;
        }

        qlge_write32(qdev, SEM, sem_bits | sem_mask);
        return !(qlge_read32(qdev, SEM) & sem_bits);
}

int qlge_sem_spinlock(struct qlge_adapter *qdev, u32 sem_mask)
{
        unsigned int wait_count = 30;

        do {
                if (!qlge_sem_trylock(qdev, sem_mask))
                        return 0;
                udelay(100);
        } while (--wait_count);
        return -ETIMEDOUT;
}

void qlge_sem_unlock(struct qlge_adapter *qdev, u32 sem_mask)
{
        qlge_write32(qdev, SEM, sem_mask);
        qlge_read32(qdev, SEM); /* flush */
}

/* This function waits for a specific bit to come ready
 * in a given register.  It is used mostly by the initialize
 * process, but is also used in kernel thread API such as
 * netdev->set_multi, netdev->set_mac_address, netdev->vlan_rx_add_vid.
 */
int qlge_wait_reg_rdy(struct qlge_adapter *qdev, u32 reg, u32 bit, u32 err_bit)
{
        u32 temp;
        int count;

        for (count = 0; count < UDELAY_COUNT; count++) {
                temp = qlge_read32(qdev, reg);

                /* check for errors */
                if (temp & err_bit) {
                        netif_alert(qdev, probe, qdev->ndev,
                                    "register 0x%.08x access error, value = 0x%.08x!.\n",
                                    reg, temp);
                        return -EIO;
                } else if (temp & bit) {
                        return 0;
                }
                udelay(UDELAY_DELAY);
        }
        netif_alert(qdev, probe, qdev->ndev,
                    "Timed out waiting for reg %x to come ready.\n", reg);
        return -ETIMEDOUT;
}

/* The CFG register is used to download TX and RX control blocks
 * to the chip. This function waits for an operation to complete.
 */
static int qlge_wait_cfg(struct qlge_adapter *qdev, u32 bit)
{
        int count;
        u32 temp;

        for (count = 0; count < UDELAY_COUNT; count++) {
                temp = qlge_read32(qdev, CFG);
                if (temp & CFG_LE)
                        return -EIO;
                if (!(temp & bit))
                        return 0;
                udelay(UDELAY_DELAY);
        }
        return -ETIMEDOUT;
}

/* Used to issue init control blocks to hw. Maps control block,
 * sets address, triggers download, waits for completion.
 */
int qlge_write_cfg(struct qlge_adapter *qdev, void *ptr, int size, u32 bit,
                   u16 q_id)
{
        u64 map;
        int status = 0;
        int direction;
        u32 mask;
        u32 value;

        if (bit & (CFG_LRQ | CFG_LR | CFG_LCQ))
                direction = DMA_TO_DEVICE;
        else
                direction = DMA_FROM_DEVICE;

        map = dma_map_single(&qdev->pdev->dev, ptr, size, direction);
        if (dma_mapping_error(&qdev->pdev->dev, map)) {
                netif_err(qdev, ifup, qdev->ndev, "Couldn't map DMA area.\n");
                return -ENOMEM;
        }

        status = qlge_sem_spinlock(qdev, SEM_ICB_MASK);
        if (status)
                goto lock_failed;

        status = qlge_wait_cfg(qdev, bit);
        if (status) {
                netif_err(qdev, ifup, qdev->ndev,
                          "Timed out waiting for CFG to come ready.\n");
                goto exit;
        }

        qlge_write32(qdev, ICB_L, (u32)map);
        qlge_write32(qdev, ICB_H, (u32)(map >> 32));

        mask = CFG_Q_MASK | (bit << 16);
        value = bit | (q_id << CFG_Q_SHIFT);
        qlge_write32(qdev, CFG, (mask | value));

        /*
         * Wait for the bit to clear after signaling hw.
         */
        status = qlge_wait_cfg(qdev, bit);
exit:
        qlge_sem_unlock(qdev, SEM_ICB_MASK);    /* does flush too */
lock_failed:
        dma_unmap_single(&qdev->pdev->dev, map, size, direction);
        return status;
}

/* Get a specific MAC address from the CAM.  Used for debug and reg dump. */
int qlge_get_mac_addr_reg(struct qlge_adapter *qdev, u32 type, u16 index,
                          u32 *value)
{
        u32 offset = 0;
        int status;

        switch (type) {
        case MAC_ADDR_TYPE_MULTI_MAC:
        case MAC_ADDR_TYPE_CAM_MAC: {
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   MAC_ADDR_ADR | MAC_ADDR_RS |
                                   type); /* type */
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MR, 0);
                if (status)
                        break;
                *value++ = qlge_read32(qdev, MAC_ADDR_DATA);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   MAC_ADDR_ADR | MAC_ADDR_RS |
                                   type); /* type */
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MR, 0);
                if (status)
                        break;
                *value++ = qlge_read32(qdev, MAC_ADDR_DATA);
                if (type == MAC_ADDR_TYPE_CAM_MAC) {
                        status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX,
                                                   MAC_ADDR_MW, 0);
                        if (status)
                                break;
                        qlge_write32(qdev, MAC_ADDR_IDX,
                                     (offset++) | /* offset */
                                           (index
                                            << MAC_ADDR_IDX_SHIFT) | /* index */
                                           MAC_ADDR_ADR |
                                           MAC_ADDR_RS | type); /* type */
                        status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX,
                                                   MAC_ADDR_MR, 0);
                        if (status)
                                break;
                        *value++ = qlge_read32(qdev, MAC_ADDR_DATA);
                }
                break;
        }
        case MAC_ADDR_TYPE_VLAN:
        case MAC_ADDR_TYPE_MULTI_FLTR:
        default:
                netif_crit(qdev, ifup, qdev->ndev,
                           "Address type %d not yet supported.\n", type);
                status = -EPERM;
        }
        return status;
}

/* Set up a MAC, multicast or VLAN address for the
 * inbound frame matching.
 */
static int qlge_set_mac_addr_reg(struct qlge_adapter *qdev, u8 *addr, u32 type,
                                 u16 index)
{
        u32 offset = 0;
        int status = 0;

        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 = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | (index << MAC_ADDR_IDX_SHIFT) | type |
                                   MAC_ADDR_E);
                qlge_write32(qdev, MAC_ADDR_DATA, lower);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | (index << MAC_ADDR_IDX_SHIFT) | type |
                                   MAC_ADDR_E);

                qlge_write32(qdev, MAC_ADDR_DATA, upper);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                break;
        }
        case MAC_ADDR_TYPE_CAM_MAC: {
                u32 cam_output;
                u32 upper = (addr[0] << 8) | addr[1];
                u32 lower = (addr[2] << 24) | (addr[3] << 16) | (addr[4] << 8) |
                            (addr[5]);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   type); /* type */
                qlge_write32(qdev, MAC_ADDR_DATA, lower);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset++) | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   type); /* type */
                qlge_write32(qdev, MAC_ADDR_DATA, upper);
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             (offset) | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   type); /* type */
                /* This field should also include the queue id
                 * and possibly the function id.  Right now we hardcode
                 * the route field to NIC core.
                 */
                cam_output = (CAM_OUT_ROUTE_NIC |
                              (qdev->func << CAM_OUT_FUNC_SHIFT) |
                              (0 << CAM_OUT_CQ_ID_SHIFT));
                if (qdev->ndev->features & NETIF_F_HW_VLAN_CTAG_RX)
                        cam_output |= CAM_OUT_RV;
                /* route to NIC core */
                qlge_write32(qdev, MAC_ADDR_DATA, cam_output);
                break;
        }
        case MAC_ADDR_TYPE_VLAN: {
                u32 enable_bit = *((u32 *)&addr[0]);
                /* For VLAN, the addr actually holds a bit that
                 * either enables or disables the vlan id we are
                 * addressing. It's either MAC_ADDR_E on or off.
                 * That's bit-27 we're talking about.
                 */
                status = qlge_wait_reg_rdy(qdev, MAC_ADDR_IDX, MAC_ADDR_MW, 0);
                if (status)
                        break;
                qlge_write32(qdev, MAC_ADDR_IDX,
                             offset | /* offset */
                                   (index << MAC_ADDR_IDX_SHIFT) | /* index */
                                   type | /* type */
                                   enable_bit); /* enable/disable */
                break;
        }
        case MAC_ADDR_TYPE_MULTI_FLTR:
        default:
                netif_crit(qdev, ifup, qdev->ndev,
                           "Address type %d not yet supported.\n", type);
                status = -EPERM;
        }
        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 qlge_set_mac_addr(struct qlge_adapter *qdev, int set)
{
        int status;
        char zero_mac_addr[ETH_ALEN];
        char *addr;

        if (set) {
                addr = &qdev->current_mac_addr[0];
                netif_printk(qdev, ifup, KERN_DEBUG, qdev->ndev,
                             "Set Mac addr %pM\n", addr);
        } else {
                eth_zero_addr(zero_mac_addr);
                addr = &zero_mac_addr[0];
                netif_printk(qdev, ifup, KERN_DEBUG, qdev->ndev,
                             "Clearing MAC address\n");
        }
        status = qlge_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
        if (status)
                return status;
        status = qlge_set_mac_addr_reg(qdev, (u8 *)addr,
                                       MAC_ADDR_TYPE_CAM_MAC,
                                       qdev->func * MAX_CQ);
        qlge_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
        if (status)
                netif_err(qdev, ifup, qdev->ndev,
                          "Failed to init mac address.\n");
        return status;
}

void qlge_link_on(struct qlge_adapter *qdev)
{
        netif_err(qdev, link, qdev->ndev, "Link is up.\n");
        netif_carrier_on(qdev->ndev);
        qlge_set_mac_addr(qdev, 1);
}

void qlge_link_off(struct qlge_adapter *qdev)
{
        netif_err(qdev, link, qdev->ndev, "Link is down.\n");
        netif_carrier_off(qdev->ndev);
        qlge_set_mac_addr(qdev, 0);
}

/* Get a specific frame routing value from the CAM.
 * Used for debug and reg dump.
 */
int qlge_get_routing_reg(struct qlge_adapter *qdev, u32 index, u32 *value)
{
        int status = 0;

        status = qlge_wait_reg_rdy(qdev, RT_IDX, RT_IDX_MW, 0);
        if (status)
                goto exit;

        qlge_write32(qdev, RT_IDX,
                     RT_IDX_TYPE_NICQ | RT_IDX_RS | (index << RT_IDX_IDX_SHIFT));
        status = qlge_wait_reg_rdy(qdev, RT_IDX, RT_IDX_MR, 0);
        if (status)
                goto exit;
        *value = qlge_read32(qdev, RT_DATA);
exit:
        return status;
}

/* The NIC function for this chip has 16 routing indexes.  Each one can be used
 * to route different frame types to various inbound queues.  We send broadcast/
 * multicast/error frames to the default queue for slow handling,
 * and CAM hit/RSS frames to the fast handling queues.
 */
static int qlge_set_routing_reg(struct qlge_adapter *qdev, u32 index, u32 mask,
                                int enable)
{
        int status = -EINVAL; /* Return error if no mask match. */
        u32 value = 0;

        switch (mask) {
        case RT_IDX_CAM_HIT:
                {
                        value = RT_IDX_DST_CAM_Q |      /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_CAM_HIT_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_VALID:      /* Promiscuous Mode frames. */
                {
                        value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_PROMISCUOUS_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_ERR:        /* Pass up MAC,IP,TCP/UDP error frames. */
                {
                        value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_ALL_ERR_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_IP_CSUM_ERR: /* Pass up IP CSUM error frames. */
                {
                        value = RT_IDX_DST_DFLT_Q | /* dest */
                                RT_IDX_TYPE_NICQ | /* type */
                                (RT_IDX_IP_CSUM_ERR_SLOT <<
                                RT_IDX_IDX_SHIFT); /* index */
                        break;
                }
        case RT_IDX_TU_CSUM_ERR: /* Pass up TCP/UDP CSUM error frames. */
                {
                        value = RT_IDX_DST_DFLT_Q | /* dest */
                                RT_IDX_TYPE_NICQ | /* type */
                                (RT_IDX_TCP_UDP_CSUM_ERR_SLOT <<
                                RT_IDX_IDX_SHIFT); /* index */
                        break;
                }
        case RT_IDX_BCAST:      /* Pass up Broadcast frames to default Q. */
                {
                        value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_BCAST_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_MCAST:      /* Pass up All Multicast frames. */
                {
                        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_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_MCAST_MATCH_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case RT_IDX_RSS_MATCH:  /* Pass up matched RSS frames. */
                {
                        value = RT_IDX_DST_RSS |        /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (RT_IDX_RSS_MATCH_SLOT << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        case 0:         /* Clear the E-bit on an entry. */
                {
                        value = RT_IDX_DST_DFLT_Q |     /* dest */
                            RT_IDX_TYPE_NICQ |  /* type */
                            (index << RT_IDX_IDX_SHIFT);/* index */
                        break;
                }
        default:
                netif_err(qdev, ifup, qdev->ndev,
                          "Mask type %d not yet supported.\n", mask);
                status = -EPERM;
                goto exit;
        }

        if (value) {
                status = qlge_wait_reg_rdy(qdev, RT_IDX, RT_IDX_MW, 0);
                if (status)
                        goto exit;
                value |= (enable ? RT_IDX_E : 0);
                qlge_write32(qdev, RT_IDX, value);
                qlge_write32(qdev, RT_DATA, enable ? mask : 0);
        }
exit:
        return status;
}

static void qlge_enable_interrupts(struct qlge_adapter *qdev)
{
        qlge_write32(qdev, INTR_EN, (INTR_EN_EI << 16) | INTR_EN_EI);
}

static void qlge_disable_interrupts(struct qlge_adapter *qdev)
{
        qlge_write32(qdev, INTR_EN, (INTR_EN_EI << 16));
}

static void qlge_enable_completion_interrupt(struct qlge_adapter *qdev, u32 intr)
{
        struct intr_context *ctx = &qdev->intr_context[intr];

        qlge_write32(qdev, INTR_EN, ctx->intr_en_mask);
}

static void qlge_disable_completion_interrupt(struct qlge_adapter *qdev, u32 intr)
{
        struct intr_context *ctx = &qdev->intr_context[intr];

        qlge_write32(qdev, INTR_EN, ctx->intr_dis_mask);
}

static void qlge_enable_all_completion_interrupts(struct qlge_adapter *qdev)
{
        int i;

        for (i = 0; i < qdev->intr_count; i++)
                qlge_enable_completion_interrupt(qdev, i);
}

static int qlge_validate_flash(struct qlge_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) {
                netif_err(qdev, ifup, qdev->ndev, "Invalid flash signature.\n");
                return  status;
        }

        for (i = 0; i < size; i++)
                csum += le16_to_cpu(*flash++);

        if (csum)
                netif_err(qdev, ifup, qdev->ndev,
                          "Invalid flash checksum, csum = 0x%.04x.\n", csum);

        return csum;
}

static int qlge_read_flash_word(struct qlge_adapter *qdev, int offset, __le32 *data)
{
        int status = 0;
        /* wait for reg to come ready */
        status = qlge_wait_reg_rdy(qdev,
                                   FLASH_ADDR, FLASH_ADDR_RDY, FLASH_ADDR_ERR);
        if (status)
                goto exit;
        /* set up for reg read */
        qlge_write32(qdev, FLASH_ADDR, FLASH_ADDR_R | offset);
        /* wait for reg to come ready */
        status = qlge_wait_reg_rdy(qdev,
                                   FLASH_ADDR, FLASH_ADDR_RDY, FLASH_ADDR_ERR);
        if (status)
                goto exit;
        /* This data is stored on flash as an array of
         * __le32.  Since qlge_read32() returns cpu endian
         * we need to swap it back.
         */
        *data = cpu_to_le32(qlge_read32(qdev, FLASH_DATA));
exit:
        return status;
}

static int qlge_get_8000_flash_params(struct qlge_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 (qlge_sem_spinlock(qdev, SEM_FLASH_MASK))
                return -ETIMEDOUT;

        size = sizeof(struct flash_params_8000) / sizeof(u32);
        for (i = 0; i < size; i++, p++) {
                status = qlge_read_flash_word(qdev, i + offset, p);
                if (status) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "Error reading flash.\n");
                        goto exit;
                }
        }

        status = qlge_validate_flash(qdev,
                                     sizeof(struct flash_params_8000) /
                                   sizeof(u16),
                                   "8000");
        if (status) {
                netif_err(qdev, ifup, qdev->ndev, "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)) {
                netif_err(qdev, ifup, qdev->ndev, "Invalid MAC address.\n");
                status = -EINVAL;
                goto exit;
        }

        memcpy(qdev->ndev->dev_addr,
               mac_addr,
               qdev->ndev->addr_len);

exit:
        qlge_sem_unlock(qdev, SEM_FLASH_MASK);
        return status;
}

static int qlge_get_8012_flash_params(struct qlge_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->port)
                offset = size;

        if (qlge_sem_spinlock(qdev, SEM_FLASH_MASK))
                return -ETIMEDOUT;

        for (i = 0; i < size; i++, p++) {
                status = qlge_read_flash_word(qdev, i + offset, p);
                if (status) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "Error reading flash.\n");
                        goto exit;
                }
        }

        status = qlge_validate_flash(qdev,
                                     sizeof(struct flash_params_8012) /
                                       sizeof(u16),
                                     "8012");
        if (status) {
                netif_err(qdev, ifup, qdev->ndev, "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:
        qlge_sem_unlock(qdev, SEM_FLASH_MASK);
        return status;
}

/* xgmac register are located behind the xgmac_addr and xgmac_data
 * register pair.  Each read/write requires us to wait for the ready
 * bit before reading/writing the data.
 */
static int qlge_write_xgmac_reg(struct qlge_adapter *qdev, u32 reg, u32 data)
{
        int status;
        /* wait for reg to come ready */
        status = qlge_wait_reg_rdy(qdev,
                                   XGMAC_ADDR, XGMAC_ADDR_RDY, XGMAC_ADDR_XME);
        if (status)
                return status;
        /* write the data to the data reg */
        qlge_write32(qdev, XGMAC_DATA, data);
        /* trigger the write */
        qlge_write32(qdev, XGMAC_ADDR, reg);
        return status;
}

/* xgmac register are located behind the xgmac_addr and xgmac_data
 * register pair.  Each read/write requires us to wait for the ready
 * bit before reading/writing the data.
 */
int qlge_read_xgmac_reg(struct qlge_adapter *qdev, u32 reg, u32 *data)
{
        int status = 0;
        /* wait for reg to come ready */
        status = qlge_wait_reg_rdy(qdev,
                                   XGMAC_ADDR, XGMAC_ADDR_RDY, XGMAC_ADDR_XME);
        if (status)
                goto exit;
        /* set up for reg read */
        qlge_write32(qdev, XGMAC_ADDR, reg | XGMAC_ADDR_R);
        /* wait for reg to come ready */
        status = qlge_wait_reg_rdy(qdev,
                                   XGMAC_ADDR, XGMAC_ADDR_RDY, XGMAC_ADDR_XME);
        if (status)
                goto exit;
        /* get the data */
        *data = qlge_read32(qdev, XGMAC_DATA);
exit:
        return status;
}

/* This is used for reading the 64-bit statistics regs. */
int qlge_read_xgmac_reg64(struct qlge_adapter *qdev, u32 reg, u64 *data)
{
        int status = 0;
        u32 hi = 0;
        u32 lo = 0;

        status = qlge_read_xgmac_reg(qdev, reg, &lo);
        if (status)
                goto exit;

        status = qlge_read_xgmac_reg(qdev, reg + 4, &hi);
        if (status)
                goto exit;

        *data = (u64)lo | ((u64)hi << 32);

exit:
        return status;
}

static int qlge_8000_port_initialize(struct qlge_adapter *qdev)
{
        int status;
        /*
         * Get MPI firmware version for driver banner
         * and ethool info.
         */
        status = qlge_mb_about_fw(qdev);
        if (status)
                goto exit;
        status = qlge_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 qlge_8012_port_initialize(struct qlge_adapter *qdev)
{
        int status = 0;
        u32 data;

        if (qlge_sem_trylock(qdev, qdev->xg_sem_mask)) {
                /* Another function has the semaphore, so
                 * wait for the port init bit to come ready.
                 */
                netif_info(qdev, link, qdev->ndev,
                           "Another function has the semaphore, so wait for the port init bit to come ready.\n");
                status = qlge_wait_reg_rdy(qdev, STS, qdev->port_init, 0);
                if (status) {
                        netif_crit(qdev, link, qdev->ndev,
                                   "Port initialize timed out.\n");
                }
                return status;
        }

        netif_info(qdev, link, qdev->ndev, "Got xgmac semaphore!.\n");
        /* Set the core reset. */
        status = qlge_read_xgmac_reg(qdev, GLOBAL_CFG, &data);
        if (status)
                goto end;
        data |= GLOBAL_CFG_RESET;
        status = qlge_write_xgmac_reg(qdev, GLOBAL_CFG, data);
        if (status)
                goto end;

        /* Clear the core reset and turn on jumbo for receiver. */
        data &= ~GLOBAL_CFG_RESET;      /* Clear core reset. */
        data |= GLOBAL_CFG_JUMBO;       /* Turn on jumbo. */
        data |= GLOBAL_CFG_TX_STAT_EN;
        data |= GLOBAL_CFG_RX_STAT_EN;
        status = qlge_write_xgmac_reg(qdev, GLOBAL_CFG, data);
        if (status)
                goto end;

        /* Enable transmitter, and clear it's reset. */
        status = qlge_read_xgmac_reg(qdev, TX_CFG, &data);
        if (status)
                goto end;
        data &= ~TX_CFG_RESET;  /* Clear the TX MAC reset. */
        data |= TX_CFG_EN;      /* Enable the transmitter. */
        status = qlge_write_xgmac_reg(qdev, TX_CFG, data);
        if (status)
                goto end;

        /* Enable receiver and clear it's reset. */
        status = qlge_read_xgmac_reg(qdev, RX_CFG, &data);
        if (status)
                goto end;
        data &= ~RX_CFG_RESET;  /* Clear the RX MAC reset. */
        data |= RX_CFG_EN;      /* Enable the receiver. */
        status = qlge_write_xgmac_reg(qdev, RX_CFG, data);
        if (status)
                goto end;

        /* Turn on jumbo. */
        status =
            qlge_write_xgmac_reg(qdev, MAC_TX_PARAMS, MAC_TX_PARAMS_JUMBO | (0x2580 << 16));
        if (status)
                goto end;
        status =
            qlge_write_xgmac_reg(qdev, MAC_RX_PARAMS, 0x2580);
        if (status)
                goto end;

        /* Signal to the world that the port is enabled.        */
        qlge_write32(qdev, STS, ((qdev->port_init << 16) | qdev->port_init));
end:
        qlge_sem_unlock(qdev, qdev->xg_sem_mask);
        return status;
}

static inline unsigned int qlge_lbq_block_size(struct qlge_adapter *qdev)
{
        return PAGE_SIZE << qdev->lbq_buf_order;
}

static struct qlge_bq_desc *qlge_get_curr_buf(struct qlge_bq *bq)
{
        struct qlge_bq_desc *bq_desc;

        bq_desc = &bq->queue[bq->next_to_clean];
        bq->next_to_clean = QLGE_BQ_WRAP(bq->next_to_clean + 1);

        return bq_desc;
}

static struct qlge_bq_desc *qlge_get_curr_lchunk(struct qlge_adapter *qdev,
                                                 struct rx_ring *rx_ring)
{
        struct qlge_bq_desc *lbq_desc = qlge_get_curr_buf(&rx_ring->lbq);

        dma_sync_single_for_cpu(&qdev->pdev->dev, lbq_desc->dma_addr,
                                qdev->lbq_buf_size, DMA_FROM_DEVICE);

        if ((lbq_desc->p.pg_chunk.offset + qdev->lbq_buf_size) ==
            qlge_lbq_block_size(qdev)) {
                /* last chunk of the master page */
                dma_unmap_page(&qdev->pdev->dev, lbq_desc->dma_addr,
                               qlge_lbq_block_size(qdev), DMA_FROM_DEVICE);
        }

        return lbq_desc;
}

/* Update an rx ring index. */
static void qlge_update_cq(struct rx_ring *rx_ring)
{
        rx_ring->cnsmr_idx++;
        rx_ring->curr_entry++;
        if (unlikely(rx_ring->cnsmr_idx == rx_ring->cq_len)) {
                rx_ring->cnsmr_idx = 0;
                rx_ring->curr_entry = rx_ring->cq_base;
        }
}

static void qlge_write_cq_idx(struct rx_ring *rx_ring)
{
        qlge_write_db_reg(rx_ring->cnsmr_idx, rx_ring->cnsmr_idx_db_reg);
}

static const char * const bq_type_name[] = {

        [QLGE_SB] = "sbq",
        [QLGE_LB] = "lbq",
};

/* return 0 or negative error */
static int qlge_refill_sb(struct rx_ring *rx_ring,
                          struct qlge_bq_desc *sbq_desc, gfp_t gfp)
{
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct sk_buff *skb;

        if (sbq_desc->p.skb)
                return 0;

        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                     "ring %u sbq: getting new skb for index %d.\n",
                     rx_ring->cq_id, sbq_desc->index);

        skb = __netdev_alloc_skb(qdev->ndev, SMALL_BUFFER_SIZE, gfp);
        if (!skb)
                return -ENOMEM;
        skb_reserve(skb, QLGE_SB_PAD);

        sbq_desc->dma_addr = dma_map_single(&qdev->pdev->dev, skb->data,
                                            SMALL_BUF_MAP_SIZE,
                                            DMA_FROM_DEVICE);
        if (dma_mapping_error(&qdev->pdev->dev, sbq_desc->dma_addr)) {
                netif_err(qdev, ifup, qdev->ndev, "PCI mapping failed.\n");
                dev_kfree_skb_any(skb);
                return -EIO;
        }
        *sbq_desc->buf_ptr = cpu_to_le64(sbq_desc->dma_addr);

        sbq_desc->p.skb = skb;
        return 0;
}

/* return 0 or negative error */
static int qlge_refill_lb(struct rx_ring *rx_ring,
                          struct qlge_bq_desc *lbq_desc, gfp_t gfp)
{
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct qlge_page_chunk *master_chunk = &rx_ring->master_chunk;

        if (!master_chunk->page) {
                struct page *page;
                dma_addr_t dma_addr;

                page = alloc_pages(gfp | __GFP_COMP, qdev->lbq_buf_order);
                if (unlikely(!page))
                        return -ENOMEM;
                dma_addr = dma_map_page(&qdev->pdev->dev, page, 0,
                                        qlge_lbq_block_size(qdev),
                                        DMA_FROM_DEVICE);
                if (dma_mapping_error(&qdev->pdev->dev, dma_addr)) {
                        __free_pages(page, qdev->lbq_buf_order);
                        netif_err(qdev, drv, qdev->ndev,
                                  "PCI mapping failed.\n");
                        return -EIO;
                }
                master_chunk->page = page;
                master_chunk->va = page_address(page);
                master_chunk->offset = 0;
                rx_ring->chunk_dma_addr = dma_addr;
        }

        lbq_desc->p.pg_chunk = *master_chunk;
        lbq_desc->dma_addr = rx_ring->chunk_dma_addr;
        *lbq_desc->buf_ptr = cpu_to_le64(lbq_desc->dma_addr +
                                         lbq_desc->p.pg_chunk.offset);

        /* Adjust the master page chunk for next
         * buffer get.
         */
        master_chunk->offset += qdev->lbq_buf_size;
        if (master_chunk->offset == qlge_lbq_block_size(qdev)) {
                master_chunk->page = NULL;
        } else {
                master_chunk->va += qdev->lbq_buf_size;
                get_page(master_chunk->page);
        }

        return 0;
}

/* return 0 or negative error */
static int qlge_refill_bq(struct qlge_bq *bq, gfp_t gfp)
{
        struct rx_ring *rx_ring = QLGE_BQ_CONTAINER(bq);
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct qlge_bq_desc *bq_desc;
        int refill_count;
        int retval;
        int i;

        refill_count = QLGE_BQ_WRAP(QLGE_BQ_ALIGN(bq->next_to_clean - 1) -
                                    bq->next_to_use);
        if (!refill_count)
                return 0;

        i = bq->next_to_use;
        bq_desc = &bq->queue[i];
        i -= QLGE_BQ_LEN;
        do {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "ring %u %s: try cleaning idx %d\n",
                             rx_ring->cq_id, bq_type_name[bq->type], i);

                if (bq->type == QLGE_SB)
                        retval = qlge_refill_sb(rx_ring, bq_desc, gfp);
                else
                        retval = qlge_refill_lb(rx_ring, bq_desc, gfp);
                if (retval < 0) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "ring %u %s: Could not get a page chunk, idx %d\n",
                                  rx_ring->cq_id, bq_type_name[bq->type], i);
                        break;
                }

                bq_desc++;
                i++;
                if (unlikely(!i)) {
                        bq_desc = &bq->queue[0];
                        i -= QLGE_BQ_LEN;
                }
                refill_count--;
        } while (refill_count);
        i += QLGE_BQ_LEN;

        if (bq->next_to_use != i) {
                if (QLGE_BQ_ALIGN(bq->next_to_use) != QLGE_BQ_ALIGN(i)) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "ring %u %s: updating prod idx = %d.\n",
                                     rx_ring->cq_id, bq_type_name[bq->type],
                                     i);
                        qlge_write_db_reg(i, bq->prod_idx_db_reg);
                }
                bq->next_to_use = i;
        }

        return retval;
}

static void qlge_update_buffer_queues(struct rx_ring *rx_ring, gfp_t gfp,
                                      unsigned long delay)
{
        bool sbq_fail, lbq_fail;

        sbq_fail = !!qlge_refill_bq(&rx_ring->sbq, gfp);
        lbq_fail = !!qlge_refill_bq(&rx_ring->lbq, gfp);

        /* Minimum number of buffers needed to be able to receive at least one
         * frame of any format:
         * sbq: 1 for header + 1 for data
         * lbq: mtu 9000 / lb size
         * Below this, the queue might stall.
         */
        if ((sbq_fail && QLGE_BQ_HW_OWNED(&rx_ring->sbq) < 2) ||
            (lbq_fail && QLGE_BQ_HW_OWNED(&rx_ring->lbq) <
             DIV_ROUND_UP(9000, LARGE_BUFFER_MAX_SIZE)))
                /* Allocations can take a long time in certain cases (ex.
                 * reclaim). Therefore, use a workqueue for long-running
                 * work items.
                 */
                queue_delayed_work_on(smp_processor_id(), system_long_wq,
                                      &rx_ring->refill_work, delay);
}

static void qlge_slow_refill(struct work_struct *work)
{
        struct rx_ring *rx_ring = container_of(work, struct rx_ring,
                                               refill_work.work);
        struct napi_struct *napi = &rx_ring->napi;

        napi_disable(napi);
        qlge_update_buffer_queues(rx_ring, GFP_KERNEL, HZ / 2);
        napi_enable(napi);

        local_bh_disable();
        /* napi_disable() might have prevented incomplete napi work from being
         * rescheduled.
         */
        napi_schedule(napi);
        /* trigger softirq processing */
        local_bh_enable();
}

/* Unmaps tx buffers.  Can be called from send() if a pci mapping
 * fails at some stage, or from the interrupt when a tx completes.
 */
static void qlge_unmap_send(struct qlge_adapter *qdev,
                            struct tx_ring_desc *tx_ring_desc, int mapped)
{
        int i;

        for (i = 0; i < mapped; i++) {
                if (i == 0 || (i == 7 && mapped > 7)) {
                        /*
                         * Unmap the skb->data area, or the
                         * external sglist (AKA the Outbound
                         * Address List (OAL)).
                         * If its the zeroeth element, then it's
                         * the skb->data area.  If it's the 7th
                         * element and there is more than 6 frags,
                         * then its an OAL.
                         */
                        if (i == 7) {
                                netif_printk(qdev, tx_done, KERN_DEBUG,
                                             qdev->ndev,
                                             "unmapping OAL area.\n");
                        }
                        dma_unmap_single(&qdev->pdev->dev,
                                         dma_unmap_addr(&tx_ring_desc->map[i],
                                                        mapaddr),
                                         dma_unmap_len(&tx_ring_desc->map[i],
                                                       maplen),
                                         DMA_TO_DEVICE);
                } else {
                        netif_printk(qdev, tx_done, KERN_DEBUG, qdev->ndev,
                                     "unmapping frag %d.\n", i);
                        dma_unmap_page(&qdev->pdev->dev,
                                       dma_unmap_addr(&tx_ring_desc->map[i],
                                                      mapaddr),
                                       dma_unmap_len(&tx_ring_desc->map[i],
                                                     maplen), DMA_TO_DEVICE);
                }
        }
}

/* Map the buffers for this transmit.  This will return
 * NETDEV_TX_BUSY or NETDEV_TX_OK based on success.
 */
static int qlge_map_send(struct qlge_adapter *qdev,
                         struct qlge_ob_mac_iocb_req *mac_iocb_ptr,
                         struct sk_buff *skb, struct tx_ring_desc *tx_ring_desc)
{
        int len = skb_headlen(skb);
        dma_addr_t map;
        int frag_idx, err, map_idx = 0;
        struct tx_buf_desc *tbd = mac_iocb_ptr->tbd;
        int frag_cnt = skb_shinfo(skb)->nr_frags;

        if (frag_cnt) {
                netif_printk(qdev, tx_queued, KERN_DEBUG, qdev->ndev,
                             "frag_cnt = %d.\n", frag_cnt);
        }
        /*
         * Map the skb buffer first.
         */
        map = dma_map_single(&qdev->pdev->dev, skb->data, len, DMA_TO_DEVICE);

        err = dma_mapping_error(&qdev->pdev->dev, map);
        if (err) {
                netif_err(qdev, tx_queued, qdev->ndev,
                          "PCI mapping failed with error: %d\n", err);

                return NETDEV_TX_BUSY;
        }

        tbd->len = cpu_to_le32(len);
        tbd->addr = cpu_to_le64(map);
        dma_unmap_addr_set(&tx_ring_desc->map[map_idx], mapaddr, map);
        dma_unmap_len_set(&tx_ring_desc->map[map_idx], maplen, len);
        map_idx++;

        /*
         * This loop fills the remainder of the 8 address descriptors
         * in the IOCB.  If there are more than 7 fragments, then the
         * eighth address desc will point to an external list (OAL).
         * When this happens, the remainder of the frags will be stored
         * in this list.
         */
        for (frag_idx = 0; frag_idx < frag_cnt; frag_idx++, map_idx++) {
                skb_frag_t *frag = &skb_shinfo(skb)->frags[frag_idx];

                tbd++;
                if (frag_idx == 6 && frag_cnt > 7) {
                        /* Let's tack on an sglist.
                         * Our control block will now
                         * look like this:
                         * iocb->seg[0] = skb->data
                         * iocb->seg[1] = frag[0]
                         * iocb->seg[2] = frag[1]
                         * iocb->seg[3] = frag[2]
                         * iocb->seg[4] = frag[3]
                         * iocb->seg[5] = frag[4]
                         * iocb->seg[6] = frag[5]
                         * iocb->seg[7] = ptr to OAL (external sglist)
                         * oal->seg[0] = frag[6]
                         * oal->seg[1] = frag[7]
                         * oal->seg[2] = frag[8]
                         * oal->seg[3] = frag[9]
                         * oal->seg[4] = frag[10]
                         *      etc...
                         */
                        /* Tack on the OAL in the eighth segment of IOCB. */
                        map = dma_map_single(&qdev->pdev->dev, &tx_ring_desc->oal,
                                             sizeof(struct qlge_oal),
                                             DMA_TO_DEVICE);
                        err = dma_mapping_error(&qdev->pdev->dev, map);
                        if (err) {
                                netif_err(qdev, tx_queued, qdev->ndev,
                                          "PCI mapping outbound address list with error: %d\n",
                                          err);
                                goto map_error;
                        }

                        tbd->addr = cpu_to_le64(map);
                        /*
                         * The length is the number of fragments
                         * that remain to be mapped times the length
                         * of our sglist (OAL).
                         */
                        tbd->len =
                            cpu_to_le32((sizeof(struct tx_buf_desc) *
                                         (frag_cnt - frag_idx)) | TX_DESC_C);
                        dma_unmap_addr_set(&tx_ring_desc->map[map_idx], mapaddr,
                                           map);
                        dma_unmap_len_set(&tx_ring_desc->map[map_idx], maplen,
                                          sizeof(struct qlge_oal));
                        tbd = (struct tx_buf_desc *)&tx_ring_desc->oal;
                        map_idx++;
                }

                map = skb_frag_dma_map(&qdev->pdev->dev, frag, 0, skb_frag_size(frag),
                                       DMA_TO_DEVICE);

                err = dma_mapping_error(&qdev->pdev->dev, map);
                if (err) {
                        netif_err(qdev, tx_queued, qdev->ndev,
                                  "PCI mapping frags failed with error: %d.\n",
                                  err);
                        goto map_error;
                }

                tbd->addr = cpu_to_le64(map);
                tbd->len = cpu_to_le32(skb_frag_size(frag));
                dma_unmap_addr_set(&tx_ring_desc->map[map_idx], mapaddr, map);
                dma_unmap_len_set(&tx_ring_desc->map[map_idx], maplen,
                                  skb_frag_size(frag));
        }
        /* Save the number of segments we've mapped. */
        tx_ring_desc->map_cnt = map_idx;
        /* Terminate the last segment. */
        tbd->len = cpu_to_le32(le32_to_cpu(tbd->len) | TX_DESC_E);
        return NETDEV_TX_OK;

map_error:
        /*
         * If the first frag mapping failed, then i will be zero.
         * This causes the unmap of the skb->data area.  Otherwise
         * we pass in the number of frags that mapped successfully
         * so they can be umapped.
         */
        qlge_unmap_send(qdev, tx_ring_desc, map_idx);
        return NETDEV_TX_BUSY;
}

/* Categorizing receive firmware frame errors */
static void qlge_categorize_rx_err(struct qlge_adapter *qdev, u8 rx_err,
                                   struct rx_ring *rx_ring)
{
        struct nic_stats *stats = &qdev->nic_stats;

        stats->rx_err_count++;
        rx_ring->rx_errors++;

        switch (rx_err & IB_MAC_IOCB_RSP_ERR_MASK) {
        case IB_MAC_IOCB_RSP_ERR_CODE_ERR:
                stats->rx_code_err++;
                break;
        case IB_MAC_IOCB_RSP_ERR_OVERSIZE:
                stats->rx_oversize_err++;
                break;
        case IB_MAC_IOCB_RSP_ERR_UNDERSIZE:
                stats->rx_undersize_err++;
                break;
        case IB_MAC_IOCB_RSP_ERR_PREAMBLE:
                stats->rx_preamble_err++;
                break;
        case IB_MAC_IOCB_RSP_ERR_FRAME_LEN:
                stats->rx_frame_len_err++;
                break;
        case IB_MAC_IOCB_RSP_ERR_CRC:
                stats->rx_crc_err++;
                break;
        default:
                break;
        }
}

/**
 * qlge_update_mac_hdr_len - helper routine to update the mac header length
 * based on vlan tags if present
 */
static void qlge_update_mac_hdr_len(struct qlge_adapter *qdev,
                                    struct qlge_ib_mac_iocb_rsp *ib_mac_rsp,
                                    void *page, size_t *len)
{
        u16 *tags;

        if (qdev->ndev->features & NETIF_F_HW_VLAN_CTAG_RX)
                return;
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) {
                tags = (u16 *)page;
                /* Look for stacked vlan tags in ethertype field */
                if (tags[6] == ETH_P_8021Q &&
                    tags[8] == ETH_P_8021Q)
                        *len += 2 * VLAN_HLEN;
                else
                        *len += VLAN_HLEN;
        }
}

/* Process an inbound completion from an rx ring. */
static void qlge_process_mac_rx_gro_page(struct qlge_adapter *qdev,
                                         struct rx_ring *rx_ring,
                                         struct qlge_ib_mac_iocb_rsp *ib_mac_rsp,
                                         u32 length, u16 vlan_id)
{
        struct sk_buff *skb;
        struct qlge_bq_desc *lbq_desc = qlge_get_curr_lchunk(qdev, rx_ring);
        struct napi_struct *napi = &rx_ring->napi;

        /* Frame error, so drop the packet. */
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_ERR_MASK) {
                qlge_categorize_rx_err(qdev, ib_mac_rsp->flags2, rx_ring);
                put_page(lbq_desc->p.pg_chunk.page);
                return;
        }
        napi->dev = qdev->ndev;

        skb = napi_get_frags(napi);
        if (!skb) {
                netif_err(qdev, drv, qdev->ndev,
                          "Couldn't get an skb, exiting.\n");
                rx_ring->rx_dropped++;
                put_page(lbq_desc->p.pg_chunk.page);
                return;
        }
        prefetch(lbq_desc->p.pg_chunk.va);
        __skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
                             lbq_desc->p.pg_chunk.page,
                             lbq_desc->p.pg_chunk.offset,
                             length);

        skb->len += length;
        skb->data_len += length;
        skb->truesize += length;
        skb_shinfo(skb)->nr_frags++;

        rx_ring->rx_packets++;
        rx_ring->rx_bytes += length;
        skb->ip_summed = CHECKSUM_UNNECESSARY;
        skb_record_rx_queue(skb, rx_ring->cq_id);
        if (vlan_id != 0xffff)
                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
        napi_gro_frags(napi);
}

/* Process an inbound completion from an rx ring. */
static void qlge_process_mac_rx_page(struct qlge_adapter *qdev,
                                     struct rx_ring *rx_ring,
                                     struct qlge_ib_mac_iocb_rsp *ib_mac_rsp,
                                     u32 length, u16 vlan_id)
{
        struct net_device *ndev = qdev->ndev;
        struct sk_buff *skb = NULL;
        void *addr;
        struct qlge_bq_desc *lbq_desc = qlge_get_curr_lchunk(qdev, rx_ring);
        struct napi_struct *napi = &rx_ring->napi;
        size_t hlen = ETH_HLEN;

        skb = netdev_alloc_skb(ndev, length);
        if (!skb) {
                rx_ring->rx_dropped++;
                put_page(lbq_desc->p.pg_chunk.page);
                return;
        }

        addr = lbq_desc->p.pg_chunk.va;
        prefetch(addr);

        /* Frame error, so drop the packet. */
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_ERR_MASK) {
                qlge_categorize_rx_err(qdev, ib_mac_rsp->flags2, rx_ring);
                goto err_out;
        }

        /* Update the MAC header length*/
        qlge_update_mac_hdr_len(qdev, ib_mac_rsp, addr, &hlen);

        /* The max framesize filter on this chip is set higher than
         * MTU since FCoE uses 2k frames.
         */
        if (skb->len > ndev->mtu + hlen) {
                netif_err(qdev, drv, qdev->ndev,
                          "Segment too small, dropping.\n");
                rx_ring->rx_dropped++;
                goto err_out;
        }
        skb_put_data(skb, addr, hlen);
        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                     "%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.pg_chunk.page,
                           lbq_desc->p.pg_chunk.offset + hlen, length - hlen);
        skb->len += length - hlen;
        skb->data_len += length - hlen;
        skb->truesize += length - hlen;

        rx_ring->rx_packets++;
        rx_ring->rx_bytes += skb->len;
        skb->protocol = eth_type_trans(skb, ndev);
        skb_checksum_none_assert(skb);

        if ((ndev->features & NETIF_F_RXCSUM) &&
            !(ib_mac_rsp->flags1 & IB_MAC_CSUM_ERR_MASK)) {
                /* TCP frame. */
                if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "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 *)((u8 *)addr + hlen);
                        if (!(iph->frag_off &
                              htons(IP_MF | IP_OFFSET))) {
                                skb->ip_summed = CHECKSUM_UNNECESSARY;
                                netif_printk(qdev, rx_status, KERN_DEBUG,
                                             qdev->ndev,
                                             "UDP checksum done!\n");
                        }
                }
        }

        skb_record_rx_queue(skb, rx_ring->cq_id);
        if (vlan_id != 0xffff)
                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
        if (skb->ip_summed == CHECKSUM_UNNECESSARY)
                napi_gro_receive(napi, skb);
        else
                netif_receive_skb(skb);
        return;
err_out:
        dev_kfree_skb_any(skb);
        put_page(lbq_desc->p.pg_chunk.page);
}

/* Process an inbound completion from an rx ring. */
static void qlge_process_mac_rx_skb(struct qlge_adapter *qdev,
                                    struct rx_ring *rx_ring,
                                    struct qlge_ib_mac_iocb_rsp *ib_mac_rsp,
                                    u32 length, u16 vlan_id)
{
        struct qlge_bq_desc *sbq_desc = qlge_get_curr_buf(&rx_ring->sbq);
        struct net_device *ndev = qdev->ndev;
        struct sk_buff *skb, *new_skb;

        skb = sbq_desc->p.skb;
        /* Allocate new_skb and copy */
        new_skb = netdev_alloc_skb(qdev->ndev, length + NET_IP_ALIGN);
        if (!new_skb) {
                rx_ring->rx_dropped++;
                return;
        }
        skb_reserve(new_skb, NET_IP_ALIGN);

        dma_sync_single_for_cpu(&qdev->pdev->dev, sbq_desc->dma_addr,
                                SMALL_BUF_MAP_SIZE, DMA_FROM_DEVICE);

        skb_put_data(new_skb, skb->data, length);

        skb = new_skb;

        /* Frame error, so drop the packet. */
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_ERR_MASK) {
                qlge_categorize_rx_err(qdev, ib_mac_rsp->flags2, rx_ring);
                dev_kfree_skb_any(skb);
                return;
        }

        /* loopback self test for ethtool */
        if (test_bit(QL_SELFTEST, &qdev->flags)) {
                qlge_check_lb_frame(qdev, skb);
                dev_kfree_skb_any(skb);
                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;
        }

        prefetch(skb->data);
        if (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "%s Multicast.\n",
                             (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                             IB_MAC_IOCB_RSP_M_HASH ? "Hash" :
                             (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                             IB_MAC_IOCB_RSP_M_REG ? "Registered" :
                             (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                             IB_MAC_IOCB_RSP_M_PROM ? "Promiscuous" : "");
        }
        if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_P)
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "Promiscuous Packet.\n");

        rx_ring->rx_packets++;
        rx_ring->rx_bytes += skb->len;
        skb->protocol = eth_type_trans(skb, ndev);
        skb_checksum_none_assert(skb);

        /* If rx checksum is on, and there are no
         * csum or frame errors.
         */
        if ((ndev->features & NETIF_F_RXCSUM) &&
            !(ib_mac_rsp->flags1 & IB_MAC_CSUM_ERR_MASK)) {
                /* TCP frame. */
                if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "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 &
                              htons(IP_MF | IP_OFFSET))) {
                                skb->ip_summed = CHECKSUM_UNNECESSARY;
                                netif_printk(qdev, rx_status, KERN_DEBUG,
                                             qdev->ndev,
                                             "UDP checksum done!\n");
                        }
                }
        }

        skb_record_rx_queue(skb, rx_ring->cq_id);
        if (vlan_id != 0xffff)
                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
        if (skb->ip_summed == CHECKSUM_UNNECESSARY)
                napi_gro_receive(&rx_ring->napi, skb);
        else
                netif_receive_skb(skb);
}

static void qlge_realign_skb(struct sk_buff *skb, int len)
{
        void *temp_addr = skb->data;

        /* Undo the skb_reserve(skb,32) we did before
         * giving to hardware, and realign data on
         * a 2-byte boundary.
         */
        skb->data -= QLGE_SB_PAD - NET_IP_ALIGN;
        skb->tail -= QLGE_SB_PAD - NET_IP_ALIGN;
        memmove(skb->data, temp_addr, len);
}

/*
 * This function builds an skb for the given inbound
 * completion.  It will be rewritten for readability in the near
 * future, but for not it works well.
 */
static struct sk_buff *qlge_build_rx_skb(struct qlge_adapter *qdev,
                                         struct rx_ring *rx_ring,
                                         struct qlge_ib_mac_iocb_rsp *ib_mac_rsp)
{
        u32 length = le32_to_cpu(ib_mac_rsp->data_len);
        u32 hdr_len = le32_to_cpu(ib_mac_rsp->hdr_len);
        struct qlge_bq_desc *lbq_desc, *sbq_desc;
        struct sk_buff *skb = NULL;
        size_t hlen = ETH_HLEN;

        /*
         * Handle the header buffer if present.
         */
        if (ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HV &&
            ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HS) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "Header of %d bytes in small buffer.\n", hdr_len);
                /*
                 * Headers fit nicely into a small buffer.
                 */
                sbq_desc = qlge_get_curr_buf(&rx_ring->sbq);
                dma_unmap_single(&qdev->pdev->dev, sbq_desc->dma_addr,
                                 SMALL_BUF_MAP_SIZE, DMA_FROM_DEVICE);
                skb = sbq_desc->p.skb;
                qlge_realign_skb(skb, hdr_len);
                skb_put(skb, hdr_len);
                sbq_desc->p.skb = NULL;
        }

        /*
         * Handle the data buffer(s).
         */
        if (unlikely(!length)) {        /* Is there data too? */
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "No Data buffer in this packet.\n");
                return skb;
        }

        if (ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_DS) {
                if (ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HS) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "Headers in small, data of %d bytes in small, combine them.\n",
                                     length);
                        /*
                         * Data is less than small buffer size so it's
                         * stuffed in a small buffer.
                         * For this case we append the data
                         * from the "data" small buffer to the "header" small
                         * buffer.
                         */
                        sbq_desc = qlge_get_curr_buf(&rx_ring->sbq);
                        dma_sync_single_for_cpu(&qdev->pdev->dev,
                                                sbq_desc->dma_addr,
                                                SMALL_BUF_MAP_SIZE,
                                                DMA_FROM_DEVICE);
                        skb_put_data(skb, sbq_desc->p.skb->data, length);
                } else {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "%d bytes in a single small buffer.\n",
                                     length);
                        sbq_desc = qlge_get_curr_buf(&rx_ring->sbq);
                        skb = sbq_desc->p.skb;
                        qlge_realign_skb(skb, length);
                        skb_put(skb, length);
                        dma_unmap_single(&qdev->pdev->dev, sbq_desc->dma_addr,
                                         SMALL_BUF_MAP_SIZE,
                                         DMA_FROM_DEVICE);
                        sbq_desc->p.skb = NULL;
                }
        } else if (ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_DL) {
                if (ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HS) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "Header in small, %d bytes in large. Chain large to small!\n",
                                     length);
                        /*
                         * The data is in a single large buffer.  We
                         * chain it to the header buffer's skb and let
                         * it rip.
                         */
                        lbq_desc = qlge_get_curr_lchunk(qdev, rx_ring);
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "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;
                } 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 = qlge_get_curr_lchunk(qdev, rx_ring);
                        skb = netdev_alloc_skb(qdev->ndev, length);
                        if (!skb) {
                                netif_printk(qdev, probe, KERN_DEBUG, qdev->ndev,
                                             "No skb available, drop the packet.\n");
                                return NULL;
                        }
                        dma_unmap_page(&qdev->pdev->dev, lbq_desc->dma_addr,
                                       qdev->lbq_buf_size,
                                       DMA_FROM_DEVICE);
                        skb_reserve(skb, NET_IP_ALIGN);
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "%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.pg_chunk.page,
                                           lbq_desc->p.pg_chunk.offset,
                                           length);
                        skb->len += length;
                        skb->data_len += length;
                        skb->truesize += length;
                        qlge_update_mac_hdr_len(qdev, ib_mac_rsp,
                                                lbq_desc->p.pg_chunk.va,
                                                &hlen);
                        __pskb_pull_tail(skb, hlen);
                }
        } else {
                /*
                 * The data is in a chain of large buffers
                 * pointed to by a small buffer.  We loop
                 * thru and chain them to the our small header
                 * buffer's skb.
                 * frags:  There are 18 max frags and our small
                 *         buffer will hold 32 of them. The thing is,
                 *         we'll use 3 max for our 9000 byte jumbo
                 *         frames.  If the MTU goes up we could
                 *          eventually be in trouble.
                 */
                int size, i = 0;

                sbq_desc = qlge_get_curr_buf(&rx_ring->sbq);
                dma_unmap_single(&qdev->pdev->dev, sbq_desc->dma_addr,
                                 SMALL_BUF_MAP_SIZE, DMA_FROM_DEVICE);
                if (!(ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HS)) {
                        /*
                         * This is an non TCP/UDP IP frame, so
                         * the headers aren't split into a small
                         * buffer.  We have to use the small buffer
                         * that contains our sg list as our skb to
                         * send upstairs. Copy the sg list here to
                         * a local buffer and use it to find the
                         * pages to chain.
                         */
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "%d bytes of headers & data in chain of large.\n",
                                     length);
                        skb = sbq_desc->p.skb;
                        sbq_desc->p.skb = NULL;
                        skb_reserve(skb, NET_IP_ALIGN);
                }
                do {
                        lbq_desc = qlge_get_curr_lchunk(qdev, rx_ring);
                        size = min(length, qdev->lbq_buf_size);

                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "Adding page %d to skb for %d bytes.\n",
                                     i, 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;
                        i++;
                } while (length > 0);
                qlge_update_mac_hdr_len(qdev, ib_mac_rsp, lbq_desc->p.pg_chunk.va,
                                        &hlen);
                __pskb_pull_tail(skb, hlen);
        }
        return skb;
}

/* Process an inbound completion from an rx ring. */
static void qlge_process_mac_split_rx_intr(struct qlge_adapter *qdev,
                                           struct rx_ring *rx_ring,
                                           struct qlge_ib_mac_iocb_rsp *ib_mac_rsp,
                                           u16 vlan_id)
{
        struct net_device *ndev = qdev->ndev;
        struct sk_buff *skb = NULL;

        skb = qlge_build_rx_skb(qdev, rx_ring, ib_mac_rsp);
        if (unlikely(!skb)) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "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) {
                qlge_categorize_rx_err(qdev, ib_mac_rsp->flags2, rx_ring);
                dev_kfree_skb_any(skb);
                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)) {
                qlge_check_lb_frame(qdev, skb);
                dev_kfree_skb_any(skb);
                return;
        }

        prefetch(skb->data);
        if (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev, "%s Multicast.\n",
                             (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                             IB_MAC_IOCB_RSP_M_HASH ? "Hash" :
                             (ib_mac_rsp->flags1 & IB_MAC_IOCB_RSP_M_MASK) ==
                             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) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "Promiscuous Packet.\n");
        }

        skb->protocol = eth_type_trans(skb, ndev);
        skb_checksum_none_assert(skb);

        /* If rx checksum is on, and there are no
         * csum or frame errors.
         */
        if ((ndev->features & NETIF_F_RXCSUM) &&
            !(ib_mac_rsp->flags1 & IB_MAC_CSUM_ERR_MASK)) {
                /* TCP frame. */
                if (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T) {
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "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 &
                              htons(IP_MF | IP_OFFSET))) {
                                skb->ip_summed = CHECKSUM_UNNECESSARY;
                                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                             "TCP checksum done!\n");
                        }
                }
        }

        rx_ring->rx_packets++;
        rx_ring->rx_bytes += skb->len;
        skb_record_rx_queue(skb, rx_ring->cq_id);
        if (vlan_id != 0xffff)
                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
        if (skb->ip_summed == CHECKSUM_UNNECESSARY)
                napi_gro_receive(&rx_ring->napi, skb);
        else
                netif_receive_skb(skb);
}

/* Process an inbound completion from an rx ring. */
static unsigned long qlge_process_mac_rx_intr(struct qlge_adapter *qdev,
                                              struct rx_ring *rx_ring,
                                              struct qlge_ib_mac_iocb_rsp *ib_mac_rsp)
{
        u32 length = le32_to_cpu(ib_mac_rsp->data_len);
        u16 vlan_id = ((ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_V) &&
                       (qdev->ndev->features & NETIF_F_HW_VLAN_CTAG_RX)) ?
                ((le16_to_cpu(ib_mac_rsp->vlan_id) &
                  IB_MAC_IOCB_RSP_VLAN_MASK)) : 0xffff;

        if (ib_mac_rsp->flags4 & IB_MAC_IOCB_RSP_HV) {
                /* The data and headers are split into
                 * separate buffers.
                 */
                qlge_process_mac_split_rx_intr(qdev, rx_ring, ib_mac_rsp,
                                               vlan_id);
        } else if (ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_DS) {
                /* The data fit in a single small buffer.
                 * Allocate a new skb, copy the data and
                 * return the buffer to the free pool.
                 */
                qlge_process_mac_rx_skb(qdev, rx_ring, ib_mac_rsp, length,
                                        vlan_id);
        } else if ((ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_DL) &&
                   !(ib_mac_rsp->flags1 & IB_MAC_CSUM_ERR_MASK) &&
                   (ib_mac_rsp->flags2 & IB_MAC_IOCB_RSP_T)) {
                /* TCP packet in a page chunk that's been checksummed.
                 * Tack it on to our GRO skb and let it go.
                 */
                qlge_process_mac_rx_gro_page(qdev, rx_ring, ib_mac_rsp, length,
                                             vlan_id);
        } else if (ib_mac_rsp->flags3 & IB_MAC_IOCB_RSP_DL) {
                /* Non-TCP packet in a page chunk. Allocate an
                 * skb, tack it on frags, and send it up.
                 */
                qlge_process_mac_rx_page(qdev, rx_ring, ib_mac_rsp, length,
                                         vlan_id);
        } else {
                /* Non-TCP/UDP large frames that span multiple buffers
                 * can be processed corrrectly by the split frame logic.
                 */
                qlge_process_mac_split_rx_intr(qdev, rx_ring, ib_mac_rsp,
                                               vlan_id);
        }

        return (unsigned long)length;
}

/* Process an outbound completion from an rx ring. */
static void qlge_process_mac_tx_intr(struct qlge_adapter *qdev,
                                     struct qlge_ob_mac_iocb_rsp *mac_rsp)
{
        struct tx_ring *tx_ring;
        struct tx_ring_desc *tx_ring_desc;

        tx_ring = &qdev->tx_ring[mac_rsp->txq_idx];
        tx_ring_desc = &tx_ring->q[mac_rsp->tid];

        qlge_unmap_send(qdev, tx_ring_desc, tx_ring_desc->map_cnt);
        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;

        if (unlikely(mac_rsp->flags1 & (OB_MAC_IOCB_RSP_E |
                                        OB_MAC_IOCB_RSP_S |
                                        OB_MAC_IOCB_RSP_L |
                                        OB_MAC_IOCB_RSP_P | OB_MAC_IOCB_RSP_B))) {
                if (mac_rsp->flags1 & OB_MAC_IOCB_RSP_E) {
                        netif_warn(qdev, tx_done, qdev->ndev,
                                   "Total descriptor length did not match transfer length.\n");
                }
                if (mac_rsp->flags1 & OB_MAC_IOCB_RSP_S) {
                        netif_warn(qdev, tx_done, qdev->ndev,
                                   "Frame too short to be valid, not sent.\n");
                }
                if (mac_rsp->flags1 & OB_MAC_IOCB_RSP_L) {
                        netif_warn(qdev, tx_done, qdev->ndev,
                                   "Frame too long, but sent anyway.\n");
                }
                if (mac_rsp->flags1 & OB_MAC_IOCB_RSP_B) {
                        netif_warn(qdev, tx_done, qdev->ndev,
                                   "PCI backplane error. Frame not sent.\n");
                }
        }
        atomic_inc(&tx_ring->tx_count);
}

/* Fire up a handler to reset the MPI processor. */
void qlge_queue_fw_error(struct qlge_adapter *qdev)
{
        qlge_link_off(qdev);
        queue_delayed_work(qdev->workqueue, &qdev->mpi_reset_work, 0);
}

void qlge_queue_asic_error(struct qlge_adapter *qdev)
{
        qlge_link_off(qdev);
        qlge_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);
        /* Set asic recovery bit to indicate reset process that we are
         * in fatal error recovery process rather than normal close
         */
        set_bit(QL_ASIC_RECOVERY, &qdev->flags);
        queue_delayed_work(qdev->workqueue, &qdev->asic_reset_work, 0);
}

static void qlge_process_chip_ae_intr(struct qlge_adapter *qdev,
                                      struct qlge_ib_ae_iocb_rsp *ib_ae_rsp)
{
        switch (ib_ae_rsp->event) {
        case MGMT_ERR_EVENT:
                netif_err(qdev, rx_err, qdev->ndev,
                          "Management Processor Fatal Error.\n");
                qlge_queue_fw_error(qdev);
                return;

        case CAM_LOOKUP_ERR_EVENT:
                netdev_err(qdev->ndev, "Multiple CAM hits lookup occurred.\n");
                netdev_err(qdev->ndev, "This event shouldn't occur.\n");
                qlge_queue_asic_error(qdev);
                return;

        case SOFT_ECC_ERROR_EVENT:
                netdev_err(qdev->ndev, "Soft ECC error detected.\n");
                qlge_queue_asic_error(qdev);
                break;

        case PCI_ERR_ANON_BUF_RD:
                netdev_err(qdev->ndev,
                           "PCI error occurred when reading anonymous buffers from rx_ring %d.\n",
                           ib_ae_rsp->q_id);
                qlge_queue_asic_error(qdev);
                break;

        default:
                netif_err(qdev, drv, qdev->ndev, "Unexpected event %d.\n",
                          ib_ae_rsp->event);
                qlge_queue_asic_error(qdev);
                break;
        }
}

static int qlge_clean_outbound_rx_ring(struct rx_ring *rx_ring)
{
        struct qlge_adapter *qdev = rx_ring->qdev;
        u32 prod = qlge_read_sh_reg(rx_ring->prod_idx_sh_reg);
        struct qlge_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) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "cq_id = %d, prod = %d, cnsmr = %d\n",
                             rx_ring->cq_id, prod, rx_ring->cnsmr_idx);

                net_rsp = (struct qlge_ob_mac_iocb_rsp *)rx_ring->curr_entry;
                rmb();
                switch (net_rsp->opcode) {
                case OPCODE_OB_MAC_TSO_IOCB:
                case OPCODE_OB_MAC_IOCB:
                        qlge_process_mac_tx_intr(qdev, net_rsp);
                        break;
                default:
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "Hit default case, not handled! dropping the packet, opcode = %x.\n",
                                     net_rsp->opcode);
                }
                count++;
                qlge_update_cq(rx_ring);
                prod = qlge_read_sh_reg(rx_ring->prod_idx_sh_reg);
        }
        if (!net_rsp)
                return 0;
        qlge_write_cq_idx(rx_ring);
        tx_ring = &qdev->tx_ring[net_rsp->txq_idx];
        if (__netif_subqueue_stopped(qdev->ndev, tx_ring->wq_id)) {
                if ((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_subqueue(qdev->ndev, tx_ring->wq_id);
        }

        return count;
}

static int qlge_clean_inbound_rx_ring(struct rx_ring *rx_ring, int budget)
{
        struct qlge_adapter *qdev = rx_ring->qdev;
        u32 prod = qlge_read_sh_reg(rx_ring->prod_idx_sh_reg);
        struct qlge_net_rsp_iocb *net_rsp;
        int count = 0;

        /* While there are entries in the completion queue. */
        while (prod != rx_ring->cnsmr_idx) {
                netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                             "cq_id = %d, prod = %d, cnsmr = %d\n",
                             rx_ring->cq_id, prod, rx_ring->cnsmr_idx);

                net_rsp = rx_ring->curr_entry;
                rmb();
                switch (net_rsp->opcode) {
                case OPCODE_IB_MAC_IOCB:
                        qlge_process_mac_rx_intr(qdev, rx_ring,
                                                 (struct qlge_ib_mac_iocb_rsp *)
                                                 net_rsp);
                        break;

                case OPCODE_IB_AE_IOCB:
                        qlge_process_chip_ae_intr(qdev, (struct qlge_ib_ae_iocb_rsp *)
                                                  net_rsp);
                        break;
                default:
                        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                                     "Hit default case, not handled! dropping the packet, opcode = %x.\n",
                                     net_rsp->opcode);
                        break;
                }
                count++;
                qlge_update_cq(rx_ring);
                prod = qlge_read_sh_reg(rx_ring->prod_idx_sh_reg);
                if (count == budget)
                        break;
        }
        qlge_update_buffer_queues(rx_ring, GFP_ATOMIC, 0);
        qlge_write_cq_idx(rx_ring);
        return count;
}

static int qlge_napi_poll_msix(struct napi_struct *napi, int budget)
{
        struct rx_ring *rx_ring = container_of(napi, struct rx_ring, napi);
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct rx_ring *trx_ring;
        int i, work_done = 0;
        struct intr_context *ctx = &qdev->intr_context[rx_ring->cq_id];

        netif_printk(qdev, rx_status, KERN_DEBUG, qdev->ndev,
                     "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)) &&
                    (qlge_read_sh_reg(trx_ring->prod_idx_sh_reg) !=
                     trx_ring->cnsmr_idx)) {
                        netif_printk(qdev, intr, KERN_DEBUG, qdev->ndev,
                                     "%s: Servicing TX completion ring %d.\n",
                                     __func__, trx_ring->cq_id);
                        qlge_clean_outbound_rx_ring(trx_ring);
                }
        }

        /*
         * Now service the RSS ring if it's active.
         */
        if (qlge_read_sh_reg(rx_ring->prod_idx_sh_reg) !=
            rx_ring->cnsmr_idx) {
                netif_printk(qdev, intr, KERN_DEBUG, qdev->ndev,
                             "%s: Servicing RX completion ring %d.\n",
                             __func__, rx_ring->cq_id);
                work_done = qlge_clean_inbound_rx_ring(rx_ring, budget);
        }

        if (work_done < budget) {
                napi_complete_done(napi, work_done);
                qlge_enable_completion_interrupt(qdev, rx_ring->irq);
        }
        return work_done;
}

static void qlge_vlan_mode(struct net_device *ndev, netdev_features_t features)
{
        struct qlge_adapter *qdev = netdev_to_qdev(ndev);

        if (features & NETIF_F_HW_VLAN_CTAG_RX) {
                qlge_write32(qdev, NIC_RCV_CFG, NIC_RCV_CFG_VLAN_MASK |
                             NIC_RCV_CFG_VLAN_MATCH_AND_NON);
        } else {
                qlge_write32(qdev, NIC_RCV_CFG, NIC_RCV_CFG_VLAN_MASK);
        }
}

/**
 * qlge_update_hw_vlan_features - helper routine to reinitialize the adapter
 * based on the features to enable/disable hardware vlan accel
 */
static int qlge_update_hw_vlan_features(struct net_device *ndev,
                                        netdev_features_t features)
{
        struct qlge_adapter *qdev = netdev_to_qdev(ndev);
        bool need_restart = netif_running(ndev);
        int status = 0;

        if (need_restart) {
                status = qlge_adapter_down(qdev);
                if (status) {
                        netif_err(qdev, link, qdev->ndev,
                                  "Failed to bring down the adapter\n");
                        return status;
                }
        }

        /* update the features with resent change */
        ndev->features = features;

        if (need_restart) {
                status = qlge_adapter_up(qdev);
                if (status) {
                        netif_err(qdev, link, qdev->ndev,
                                  "Failed to bring up the adapter\n");
                        return status;
                }
        }

        return status;
}

static int qlge_set_features(struct net_device *ndev,
                             netdev_features_t features)
{
        netdev_features_t changed = ndev->features ^ features;
        int err;

        if (changed & NETIF_F_HW_VLAN_CTAG_RX) {
                /* Update the behavior of vlan accel in the adapter */
                err = qlge_update_hw_vlan_features(ndev, features);
                if (err)
                        return err;

                qlge_vlan_mode(ndev, features);
        }

        return 0;
}

static int __qlge_vlan_rx_add_vid(struct qlge_adapter *qdev, u16 vid)
{
        u32 enable_bit = MAC_ADDR_E;
        int err;

        err = qlge_set_mac_addr_reg(qdev, (u8 *)&enable_bit,
                                    MAC_ADDR_TYPE_VLAN, vid);
        if (err)
                netif_err(qdev, ifup, qdev->ndev,
                          "Failed to init vlan address.\n");
        return err;
}

static int qlge_vlan_rx_add_vid(struct net_device *ndev, __be16 proto, u16 vid)
{
        struct qlge_adapter *qdev = netdev_to_qdev(ndev);
        int status;
        int err;

        status = qlge_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
        if (status)
                return status;

        err = __qlge_vlan_rx_add_vid(qdev, vid);
        set_bit(vid, qdev->active_vlans);

        qlge_sem_unlock(qdev, SEM_MAC_ADDR_MASK);

        return err;
}

static int __qlge_vlan_rx_kill_vid(struct qlge_adapter *qdev, u16 vid)
{
        u32 enable_bit = 0;
        int err;

        err = qlge_set_mac_addr_reg(qdev, (u8 *)&enable_bit,
                                    MAC_ADDR_TYPE_VLAN, vid);
        if (err)
                netif_err(qdev, ifup, qdev->ndev,
                          "Failed to clear vlan address.\n");
        return err;
}

static int qlge_vlan_rx_kill_vid(struct net_device *ndev, __be16 proto, u16 vid)
{
        struct qlge_adapter *qdev = netdev_to_qdev(ndev);
        int status;
        int err;

        status = qlge_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
        if (status)
                return status;

        err = __qlge_vlan_rx_kill_vid(qdev, vid);
        clear_bit(vid, qdev->active_vlans);

        qlge_sem_unlock(qdev, SEM_MAC_ADDR_MASK);

        return err;
}

static void qlge_restore_vlan(struct qlge_adapter *qdev)
{
        int status;
        u16 vid;

        status = qlge_sem_spinlock(qdev, SEM_MAC_ADDR_MASK);
        if (status)
                return;

        for_each_set_bit(vid, qdev->active_vlans, VLAN_N_VID)
                __qlge_vlan_rx_add_vid(qdev, vid);

        qlge_sem_unlock(qdev, SEM_MAC_ADDR_MASK);
}

/* MSI-X Multiple Vector Interrupt Handler for inbound completions. */
static irqreturn_t qlge_msix_rx_isr(int irq, void *dev_id)
{
        struct rx_ring *rx_ring = dev_id;

        napi_schedule(&rx_ring->napi);
        return IRQ_HANDLED;
}

/* This handles a fatal error, MPI activity, and the default
 * rx_ring in an MSI-X multiple vector environment.
 * In MSI/Legacy environment it also process the rest of
 * the rx_rings.
 */
static irqreturn_t qlge_isr(int irq, void *dev_id)
{
        struct rx_ring *rx_ring = dev_id;
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct intr_context *intr_context = &qdev->intr_context[0];
        u32 var;
        int work_done = 0;

        /* Experience shows that when using INTx interrupts, interrupts must
         * be masked manually.
         * When using MSI mode, INTR_EN_EN must be explicitly disabled
         * (even though it is auto-masked), otherwise a later command to
         * enable it is not effective.
         */
        if (!test_bit(QL_MSIX_ENABLED, &qdev->flags))
                qlge_disable_completion_interrupt(qdev, 0);

        var = qlge_read32(qdev, STS);

        /*
         * Check for fatal error.
         */
        if (var & STS_FE) {
                qlge_disable_completion_interrupt(qdev, 0);
                qlge_queue_asic_error(qdev);
                netdev_err(qdev->ndev, "Got fatal error, STS = %x.\n", var);
                var = qlge_read32(qdev, ERR_STS);
                netdev_err(qdev->ndev, "Resetting chip. Error Status Register = 0x%x\n", var);
                return IRQ_HANDLED;
        }

        /*
         * Check MPI processor activity.
         */
        if ((var & STS_PI) &&
            (qlge_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.
                 */
                netif_err(qdev, intr, qdev->ndev,
                          "Got MPI processor interrupt.\n");
                qlge_write32(qdev, INTR_MASK, (INTR_MASK_PI << 16));
                queue_delayed_work_on(smp_processor_id(),
                                      qdev->workqueue, &qdev->mpi_work, 0);
                work_done++;
        }

        /*
         * 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.
         */
        var = qlge_read32(qdev, ISR1);
        if (var & intr_context->irq_mask) {
                netif_info(qdev, intr, qdev->ndev,
                           "Waking handler for rx_ring[0].\n");
                napi_schedule(&rx_ring->napi);
                work_done++;
        } else {
                /* Experience shows that the device sometimes signals an
                 * interrupt but no work is scheduled from this function.
                 * Nevertheless, the interrupt is auto-masked. Therefore, we
                 * systematically re-enable the interrupt if we didn't
                 * schedule napi.
                 */
                qlge_enable_completion_interrupt(qdev, 0);
        }

        return work_done ? IRQ_HANDLED : IRQ_NONE;
}

static int qlge_tso(struct sk_buff *skb, struct qlge_ob_mac_tso_iocb_req *mac_iocb_ptr)
{
        if (skb_is_gso(skb)) {
                int err;
                __be16 l3_proto = vlan_get_protocol(skb);

                err = skb_cow_head(skb, 0);
                if (err < 0)
                        return err;

                mac_iocb_ptr->opcode = OPCODE_OB_MAC_TSO_IOCB;
                mac_iocb_ptr->flags3 |= OB_MAC_TSO_IOCB_IC;
                mac_iocb_ptr->frame_len = cpu_to_le32((u32)skb->len);
                mac_iocb_ptr->total_hdrs_len =
                        cpu_to_le16(skb_transport_offset(skb) + tcp_hdrlen(skb));
                mac_iocb_ptr->net_trans_offset =
                        cpu_to_le16(skb_network_offset(skb) |
                                    skb_transport_offset(skb)
                                    << OB_MAC_TRANSPORT_HDR_SHIFT);
                mac_iocb_ptr->mss = cpu_to_le16(skb_shinfo(skb)->gso_size);
                mac_iocb_ptr->flags2 |= OB_MAC_TSO_IOCB_LSO;
                if (likely(l3_proto == htons(ETH_P_IP))) {
                        struct iphdr *iph = ip_hdr(skb);

                        iph->check = 0;
                        mac_iocb_ptr->flags1 |= OB_MAC_TSO_IOCB_IP4;
                        tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr,
                                                                 iph->daddr, 0,
                                                                 IPPROTO_TCP,
                                                                 0);
                } else if (l3_proto == htons(ETH_P_IPV6)) {
                        mac_iocb_ptr->flags1 |= OB_MAC_TSO_IOCB_IP6;
                        tcp_hdr(skb)->check =
                                ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
                                                 &ipv6_hdr(skb)->daddr,
                                                 0, IPPROTO_TCP, 0);
                }
                return 1;
        }
        return 0;
}

static void qlge_hw_csum_setup(struct sk_buff *skb,
                               struct qlge_ob_mac_tso_iocb_req *mac_iocb_ptr)
{
        int len;
        struct iphdr *iph = ip_hdr(skb);
        __sum16 *check;

        mac_iocb_ptr->opcode = OPCODE_OB_MAC_TSO_IOCB;
        mac_iocb_ptr->frame_len = cpu_to_le32((u32)skb->len);
        mac_iocb_ptr->net_trans_offset =
                cpu_to_le16(skb_network_offset(skb) |
                            skb_transport_offset(skb) << OB_MAC_TRANSPORT_HDR_SHIFT);

        mac_iocb_ptr->flags1 |= OB_MAC_TSO_IOCB_IP4;
        len = (ntohs(iph->tot_len) - (iph->ihl << 2));
        if (likely(iph->protocol == IPPROTO_TCP)) {
                check = &(tcp_hdr(skb)->check);
                mac_iocb_ptr->flags2 |= OB_MAC_TSO_IOCB_TC;
                mac_iocb_ptr->total_hdrs_len =
                        cpu_to_le16(skb_transport_offset(skb) +
                                    (tcp_hdr(skb)->doff << 2));
        } else {
                check = &(udp_hdr(skb)->check);
                mac_iocb_ptr->flags2 |= OB_MAC_TSO_IOCB_UC;
                mac_iocb_ptr->total_hdrs_len =
                        cpu_to_le16(skb_transport_offset(skb) +
                                    sizeof(struct udphdr));
        }
        *check = ~csum_tcpudp_magic(iph->saddr,
                                    iph->daddr, len, iph->protocol, 0);
}

static netdev_tx_t qlge_send(struct sk_buff *skb, struct net_device *ndev)
{
        struct qlge_adapter *qdev = netdev_to_qdev(ndev);
        struct qlge_ob_mac_iocb_req *mac_iocb_ptr;
        struct tx_ring_desc *tx_ring_desc;
        int tso;
        struct tx_ring *tx_ring;
        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)) {
                netif_info(qdev, tx_queued, qdev->ndev,
                           "%s: BUG! shutting down tx queue %d due to lack of resources.\n",
                           __func__, tx_ring_idx);
                netif_stop_subqueue(ndev, tx_ring->wq_id);
                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));

        mac_iocb_ptr->opcode = OPCODE_OB_MAC_IOCB;
        mac_iocb_ptr->tid = tx_ring_desc->index;
        /* We use the upper 32-bits to store the tx queue for this IO.
         * When we get the completion we can use it to establish the context.
         */
        mac_iocb_ptr->txq_idx = tx_ring_idx;
        tx_ring_desc->skb = skb;

        mac_iocb_ptr->frame_len = cpu_to_le16((u16)skb->len);

        if (skb_vlan_tag_present(skb)) {
                netif_printk(qdev, tx_queued, KERN_DEBUG, qdev->ndev,
                             "Adding a vlan tag %d.\n", skb_vlan_tag_get(skb));
                mac_iocb_ptr->flags3 |= OB_MAC_IOCB_V;
                mac_iocb_ptr->vlan_tci = cpu_to_le16(skb_vlan_tag_get(skb));
        }
        tso = qlge_tso(skb, (struct qlge_ob_mac_tso_iocb_req *)mac_iocb_ptr);
        if (tso < 0) {
                dev_kfree_skb_any(skb);
                return NETDEV_TX_OK;
        } else if (unlikely(!tso) && (skb->ip_summed == CHECKSUM_PARTIAL)) {
                qlge_hw_csum_setup(skb,
                                   (struct qlge_ob_mac_tso_iocb_req *)mac_iocb_ptr);
        }
        if (qlge_map_send(qdev, mac_iocb_ptr, skb, tx_ring_desc) !=
            NETDEV_TX_OK) {
                netif_err(qdev, tx_queued, qdev->ndev,
                          "Could not map the segments.\n");
                tx_ring->tx_errors++;
                return NETDEV_TX_BUSY;
        }

        tx_ring->prod_idx++;
        if (tx_ring->prod_idx == tx_ring->wq_len)
                tx_ring->prod_idx = 0;
        wmb();

        qlge_write_db_reg_relaxed(tx_ring->prod_idx, tx_ring->prod_idx_db_reg);
        netif_printk(qdev, tx_queued, KERN_DEBUG, qdev->ndev,
                     "tx queued, slot %d, len %d\n",
                     tx_ring->prod_idx, skb->len);

        atomic_dec(&tx_ring->tx_count);

        if (unlikely(atomic_read(&tx_ring->tx_count) < 2)) {
                netif_stop_subqueue(ndev, tx_ring->wq_id);
                if ((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_subqueue(qdev->ndev, tx_ring->wq_id);
        }
        return NETDEV_TX_OK;
}

static void qlge_free_shadow_space(struct qlge_adapter *qdev)
{
        if (qdev->rx_ring_shadow_reg_area) {
                dma_free_coherent(&qdev->pdev->dev,
                                  PAGE_SIZE,
                                  qdev->rx_ring_shadow_reg_area,
                                  qdev->rx_ring_shadow_reg_dma);
                qdev->rx_ring_shadow_reg_area = NULL;
        }
        if (qdev->tx_ring_shadow_reg_area) {
                dma_free_coherent(&qdev->pdev->dev,
                                  PAGE_SIZE,
                                  qdev->tx_ring_shadow_reg_area,
                                  qdev->tx_ring_shadow_reg_dma);
                qdev->tx_ring_shadow_reg_area = NULL;
        }
}

static int qlge_alloc_shadow_space(struct qlge_adapter *qdev)
{
        qdev->rx_ring_shadow_reg_area =
                dma_alloc_coherent(&qdev->pdev->dev, PAGE_SIZE,
                                   &qdev->rx_ring_shadow_reg_dma, GFP_ATOMIC);
        if (!qdev->rx_ring_shadow_reg_area) {
                netif_err(qdev, ifup, qdev->ndev,
                          "Allocation of RX shadow space failed.\n");
                return -ENOMEM;
        }

        qdev->tx_ring_shadow_reg_area =
                dma_alloc_coherent(&qdev->pdev->dev, PAGE_SIZE,
                                   &qdev->tx_ring_shadow_reg_dma, GFP_ATOMIC);
        if (!qdev->tx_ring_shadow_reg_area) {
                netif_err(qdev, ifup, qdev->ndev,
                          "Allocation of TX shadow space failed.\n");
                goto err_wqp_sh_area;
        }
        return 0;

err_wqp_sh_area:
        dma_free_coherent(&qdev->pdev->dev,
                          PAGE_SIZE,
                          qdev->rx_ring_shadow_reg_area,
                          qdev->rx_ring_shadow_reg_dma);
        return -ENOMEM;
}

static void qlge_init_tx_ring(struct qlge_adapter *qdev, struct tx_ring *tx_ring)
{
        struct tx_ring_desc *tx_ring_desc;
        int i;
        struct qlge_ob_mac_iocb_req *mac_iocb_ptr;

        mac_iocb_ptr = tx_ring->wq_base;
        tx_ring_desc = tx_ring->q;
        for (i = 0; i < tx_ring->wq_len; i++) {
                tx_ring_desc->index = i;
                tx_ring_desc->skb = NULL;
                tx_ring_desc->queue_entry = mac_iocb_ptr;
                mac_iocb_ptr++;
                tx_ring_desc++;
        }
        atomic_set(&tx_ring->tx_count, tx_ring->wq_len);
}

static void qlge_free_tx_resources(struct qlge_adapter *qdev,
                                   struct tx_ring *tx_ring)
{
        if (tx_ring->wq_base) {
                dma_free_coherent(&qdev->pdev->dev, tx_ring->wq_size,
                                  tx_ring->wq_base, tx_ring->wq_base_dma);
                tx_ring->wq_base = NULL;
        }
        kfree(tx_ring->q);
        tx_ring->q = NULL;
}

static int qlge_alloc_tx_resources(struct qlge_adapter *qdev,
                                   struct tx_ring *tx_ring)
{
        tx_ring->wq_base =
                dma_alloc_coherent(&qdev->pdev->dev, tx_ring->wq_size,
                                   &tx_ring->wq_base_dma, GFP_ATOMIC);

        if (!tx_ring->wq_base ||
            tx_ring->wq_base_dma & WQ_ADDR_ALIGN)
                goto pci_alloc_err;

        tx_ring->q =
                kmalloc_array(tx_ring->wq_len, sizeof(struct tx_ring_desc),
                              GFP_KERNEL);
        if (!tx_ring->q)
                goto err;

        return 0;
err:
        dma_free_coherent(&qdev->pdev->dev, tx_ring->wq_size,
                          tx_ring->wq_base, tx_ring->wq_base_dma);
        tx_ring->wq_base = NULL;
pci_alloc_err:
        netif_err(qdev, ifup, qdev->ndev, "tx_ring alloc failed.\n");
        return -ENOMEM;
}

static void qlge_free_lbq_buffers(struct qlge_adapter *qdev, struct rx_ring *rx_ring)
{
        struct qlge_bq *lbq = &rx_ring->lbq;
        unsigned int last_offset;

        last_offset = qlge_lbq_block_size(qdev) - qdev->lbq_buf_size;
        while (lbq->next_to_clean != lbq->next_to_use) {
                struct qlge_bq_desc *lbq_desc =
                        &lbq->queue[lbq->next_to_clean];

                if (lbq_desc->p.pg_chunk.offset == last_offset)
                        dma_unmap_page(&qdev->pdev->dev, lbq_desc->dma_addr,
                                       qlge_lbq_block_size(qdev),
                                       DMA_FROM_DEVICE);
                put_page(lbq_desc->p.pg_chunk.page);

                lbq->next_to_clean = QLGE_BQ_WRAP(lbq->next_to_clean + 1);
        }

        if (rx_ring->master_chunk.page) {
                dma_unmap_page(&qdev->pdev->dev, rx_ring->chunk_dma_addr,
                               qlge_lbq_block_size(qdev), DMA_FROM_DEVICE);
                put_page(rx_ring->master_chunk.page);
                rx_ring->master_chunk.page = NULL;
        }
}

static void qlge_free_sbq_buffers(struct qlge_adapter *qdev, struct rx_ring *rx_ring)
{
        int i;

        for (i = 0; i < QLGE_BQ_LEN; i++) {
                struct qlge_bq_desc *sbq_desc = &rx_ring->sbq.queue[i];

                if (!sbq_desc) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "sbq_desc %d is NULL.\n", i);
                        return;
                }
                if (sbq_desc->p.skb) {
                        dma_unmap_single(&qdev->pdev->dev, sbq_desc->dma_addr,
                                         SMALL_BUF_MAP_SIZE,
                                         DMA_FROM_DEVICE);
                        dev_kfree_skb(sbq_desc->p.skb);
                        sbq_desc->p.skb = NULL;
                }
        }
}

/* Free all large and small rx buffers associated
 * with the completion queues for this device.
 */
static void qlge_free_rx_buffers(struct qlge_adapter *qdev)
{
        int i;

        for (i = 0; i < qdev->rx_ring_count; i++) {
                struct rx_ring *rx_ring = &qdev->rx_ring[i];

                if (rx_ring->lbq.queue)
                        qlge_free_lbq_buffers(qdev, rx_ring);
                if (rx_ring->sbq.queue)
                        qlge_free_sbq_buffers(qdev, rx_ring);
        }
}

static void qlge_alloc_rx_buffers(struct qlge_adapter *qdev)
{
        int i;

        for (i = 0; i < qdev->rss_ring_count; i++)
                qlge_update_buffer_queues(&qdev->rx_ring[i], GFP_KERNEL,
                                          HZ / 2);
}

static int qlge_init_bq(struct qlge_bq *bq)
{
        struct rx_ring *rx_ring = QLGE_BQ_CONTAINER(bq);
        struct qlge_adapter *qdev = rx_ring->qdev;
        struct qlge_bq_desc *bq_desc;
        __le64 *buf_ptr;
        int i;

        bq->base = dma_alloc_coherent(&qdev->pdev->dev, QLGE_BQ_SIZE,
                                      &bq->base_dma, GFP_ATOMIC);
        if (!bq->base) {
                netif_err(qdev, ifup, qdev->ndev,
                          "ring %u %s allocation failed.\n", rx_ring->cq_id,
                          bq_type_name[bq->type]);
                return -ENOMEM;
        }

        bq->queue = kmalloc_array(QLGE_BQ_LEN, sizeof(struct qlge_bq_desc),
                                  GFP_KERNEL);
        if (!bq->queue)
                return -ENOMEM;

        buf_ptr = bq->base;
        bq_desc = &bq->queue[0];
        for (i = 0; i < QLGE_BQ_LEN; i++, buf_ptr++, bq_desc++) {
                bq_desc->p.skb = NULL;
                bq_desc->index = i;
                bq_desc->buf_ptr = buf_ptr;
        }

        return 0;
}

static void qlge_free_rx_resources(struct qlge_adapter *qdev,
                                   struct rx_ring *rx_ring)
{
        /* Free the small buffer queue. */
        if (rx_ring->sbq.base) {
                dma_free_coherent(&qdev->pdev->dev, QLGE_BQ_SIZE,
                                  rx_ring->sbq.base, rx_ring->sbq.base_dma);
                rx_ring->sbq.base = NULL;
        }

        /* Free the small buffer queue control blocks. */
        kfree(rx_ring->sbq.queue);
        rx_ring->sbq.queue = NULL;

        /* Free the large buffer queue. */
        if (rx_ring->lbq.base) {
                dma_free_coherent(&qdev->pdev->dev, QLGE_BQ_SIZE,
                                  rx_ring->lbq.base, rx_ring->lbq.base_dma);
                rx_ring->lbq.base = NULL;
        }

        /* Free the large buffer queue control blocks. */
        kfree(rx_ring->lbq.queue);
        rx_ring->lbq.queue = NULL;

        /* Free the rx queue. */
        if (rx_ring->cq_base) {
                dma_free_coherent(&qdev->pdev->dev,
                                  rx_ring->cq_size,
                                  rx_ring->cq_base, rx_ring->cq_base_dma);
                rx_ring->cq_base = NULL;
        }
}

/* Allocate queues and buffers for this completions queue based
 * on the values in the parameter structure.
 */
static int qlge_alloc_rx_resources(struct qlge_adapter *qdev,
                                   struct rx_ring *rx_ring)
{
        /*
         * Allocate the completion queue for this rx_ring.
         */
        rx_ring->cq_base =
                dma_alloc_coherent(&qdev->pdev->dev, rx_ring->cq_size,
                                   &rx_ring->cq_base_dma, GFP_ATOMIC);

        if (!rx_ring->cq_base) {
                netif_err(qdev, ifup, qdev->ndev, "rx_ring alloc failed.\n");
                return -ENOMEM;
        }

        if (rx_ring->cq_id < qdev->rss_ring_count &&
            (qlge_init_bq(&rx_ring->sbq) || qlge_init_bq(&rx_ring->lbq))) {
                qlge_free_rx_resources(qdev, rx_ring);
                return -ENOMEM;
        }

        return 0;
}

static void qlge_tx_ring_clean(struct qlge_adapter *qdev)
{
        struct tx_ring *tx_ring;
        struct tx_ring_desc *tx_ring_desc;
        int i, j;

        /*
         * Loop through all queues and free
         * any resources.
         */
        for (j = 0; j < qdev->tx_ring_count; j++) {
                tx_ring = &qdev->tx_ring[j];
                for (i = 0; i < tx_ring->wq_len; i++) {
                        tx_ring_desc = &tx_ring->q[i];
                        if (tx_ring_desc && tx_ring_desc->skb) {
                                netif_err(qdev, ifdown, qdev->ndev,
                                          "Freeing lost SKB %p, from queue %d, index %d.\n",
                                          tx_ring_desc->skb, j,
                                          tx_ring_desc->index);
                                qlge_unmap_send(qdev, tx_ring_desc,
                                                tx_ring_desc->map_cnt);
                                dev_kfree_skb(tx_ring_desc->skb);
                                tx_ring_desc->skb = NULL;
                        }
                }
        }
}

static void qlge_free_mem_resources(struct qlge_adapter *qdev)
{
        int i;

        for (i = 0; i < qdev->tx_ring_count; i++)
                qlge_free_tx_resources(qdev, &qdev->tx_ring[i]);
        for (i = 0; i < qdev->rx_ring_count; i++)
                qlge_free_rx_resources(qdev, &qdev->rx_ring[i]);
        qlge_free_shadow_space(qdev);
}

static int qlge_alloc_mem_resources(struct qlge_adapter *qdev)
{
        int i;

        /* Allocate space for our shadow registers and such. */
        if (qlge_alloc_shadow_space(qdev))
                return -ENOMEM;

        for (i = 0; i < qdev->rx_ring_count; i++) {
                if (qlge_alloc_rx_resources(qdev, &qdev->rx_ring[i]) != 0) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "RX resource allocation failed.\n");
                        goto err_mem;
                }
        }
        /* Allocate tx queue resources */
        for (i = 0; i < qdev->tx_ring_count; i++) {
                if (qlge_alloc_tx_resources(qdev, &qdev->tx_ring[i]) != 0) {
                        netif_err(qdev, ifup, qdev->ndev,
                                  "TX resource allocation failed.\n");
                        goto err_mem;
                }
        }
        return 0;

err_mem:
        qlge_free_mem_resources(qdev);
        return -ENOMEM;
}

/* Set up the rx ring control block and pass it to the chip.
 * The control block is defined as
 * "Completion Queue Initialization Control Block", or cqicb.
 */
static int qlge_start_rx_ring(struct qlge_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 * RX_RING_SHADOW_SPACE);
        u64 shadow_reg_dma = qdev->rx_ring_shadow_reg_dma +
                (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;
        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) * MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN));
        shadow_reg_dma += (sizeof(u64) * MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN));
        rx_ring->sbq.base_indirect = shadow_reg;
        rx_ring->sbq.base_indirect_dma = shadow_reg_dma;

        /* PCI doorbell mem area + 0x00 for consumer index register */
        rx_ring->cnsmr_idx_db_reg = (u32 __iomem *)doorbell_area;
        rx_ring->cnsmr_idx = 0;
        rx_ring->curr_entry = rx_ring->cq_base;

        /* PCI doorbell mem area + 0x04 for valid register */
        rx_ring->valid_db_reg = doorbell_area + 0x04;

        /* PCI doorbell mem area + 0x18 for large buffer consumer */
        rx_ring->lbq.prod_idx_db_reg = (u32 __iomem *)(doorbell_area + 0x18);

        /* PCI doorbell mem area + 0x1c */
        rx_ring->sbq.prod_idx_db_reg = (u32 __iomem *)(doorbell_area + 0x1c);

        memset((void *)cqicb, 0, sizeof(struct cqicb));
        cqicb->msix_vect = rx_ring->irq;

        cqicb->len = cpu_to_le16(QLGE_FIT16(rx_ring->cq_len) | LEN_V |