eepro100.c

Go to the documentation of this file.

/*
 * eepro100.c -- This is a driver for Intel Fast Ethernet Controllers
 * (ifec).
 *
 * Originally written for Etherboot by:
 *
 *   Copyright (C) AW Computer Systems.
 *   written by R.E.Wolff -- R.E.Wolff@BitWizard.nl
 *
 *   AW Computer Systems is contributing to the free software community
 *   by paying for this driver and then putting the result under GPL.
 *
 *   If you need a Linux device driver, please contact BitWizard for a
 *   quote.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation; either version 2, or (at
 * your option) any later version.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
 * 02110-1301, USA.
 *
 *
 *              date       version  by      what
 *  Written:    May 29 1997  V0.10  REW     Initial revision.
 * changes:     May 31 1997  V0.90  REW     Works!
 *              Jun 1  1997  V0.91  REW     Cleanup
 *              Jun 2  1997  V0.92  REW     Add some code documentation
 *              Jul 25 1997  V1.00  REW     Tested by AW to work in a PROM
 *                                          Cleanup for publication
 *              Dez 11 2004  V1.10  Kiszka  Add RX ring buffer support
 *              Jun    2008  v2.0   mdeck   Updated to iPXE. Changed much.
 *
 * Cleanups and fixes by Thomas Miletich<thomas.miletich@gmail.com>
 *
 * This is the etherboot intel etherexpress Pro/100B driver.
 *
 * It was written from scratch, with Donald Beckers eepro100.c kernel
 * driver as a guideline. Mostly the 82557 related definitions and the
 * lower level routines have been cut-and-pasted into this source.
 *
 * The driver was finished before Intel got the NDA out of the closet.
 *
 * Datasheet is now published and available from 
 * ftp://download.intel.com/design/network/manuals/8255X_OpenSDM.pdf
 *    - Michael Brown
 * */

FILE_LICENCE ( GPL2_OR_LATER );

/*
 * General Theory of Operation
 *
 * Initialization
 *
 * ifec_pci_probe() is called by iPXE during initialization. Typical NIC
 * initialization is performed.  EEPROM data is read.
 *
 * Network Boot
 *
 * ifec_net_open() is called by iPXE before attempting to network boot from the
 * card.  Here, the Command Unit & Receive Unit are initialized.  The tx & rx
 * rings are setup.  The MAC address is programmed and the card is configured.
 *
 * Transmit
 *
 * ifec_net_transmit() enqueues a packet in the tx ring - active::tcbs[]  The tx
 * ring is composed of TCBs linked to each other into a ring.  A tx request
 * fills out the next available TCB with a pointer to the packet data.
 * The last enqueued tx is always at active::tcb_head.  Thus, a tx request fills
 * out the TCB following tcb_head.
 * active::tcb_tail points to the TCB we're awaiting completion of.
 * ifec_tx_process() checks tcb_tail, and once complete,
 * blindly increments tcb_tail to the next ring TCB.
 *
 * Receive
 *
 * priv::rfds[] is an array of Receive Frame Descriptors. The RFDs are linked
 * together to form a ring.
 * ifec_net_poll() calls ifec_rx_process(), which checks the next RFD for
 * data.  If we received a packet, we allocate a new io_buffer and copy the
 * packet data into it. If alloc_iob() fails, we don't touch the RFD and try
 * again on the next poll.
 */

/*
 * Debugging levels:
 *      - DBG() is for any errors, i.e. failed alloc_iob(), malloc_phys(),
 *        TX overflow, corrupted packets, ...
 *      - DBG2() is for successful events, like packet received,
 *        packet transmitted, and other general notifications.
 *      - DBGP() prints the name of each called function on entry
 */

#include <stdint.h>
#include <byteswap.h>
#include <errno.h>
#include <stdio.h>
#include <unistd.h>
#include <ipxe/ethernet.h>
#include <ipxe/if_ether.h>
#include <ipxe/iobuf.h>
#include <ipxe/malloc.h>
#include <ipxe/pci.h>
#include <ipxe/spi_bit.h>
#include <ipxe/timer.h>
#include <ipxe/nvs.h>
#include <ipxe/threewire.h>
#include <ipxe/netdevice.h>
#include "eepro100.h"

/****************************** Global data **********************************/

/*
 * This is the default configuration command data. The values were copied from
 * the Linux kernel initialization for the eepro100.
 */
static struct ifec_cfg ifec_cfg = {
        .status  = 0,
        .command = CmdConfigure | CmdSuspend,
        .link    = 0,        /* Filled in later */
        .byte = { 22,        /* How many bytes in this array */
                  ( TX_FIFO << 4 ) | RX_FIFO,  /* Rx & Tx FIFO limits */
                  0, 0,                        /* Adaptive Interframe Spacing */
                  RX_DMA_COUNT,                /* Rx DMA max byte count */
                  TX_DMA_COUNT + 0x80,         /* Tx DMA max byte count */
                  0x32,      /* Many bits. */
                  0x03,      /* Discard short receive & Underrun retries */
                  1,         /* 1=Use MII  0=Use AUI */
                  0,
                  0x2E,      /* NSAI, Preamble length, & Loopback*/
                  0,         /* Linear priority */
                  0x60,      /* L PRI MODE & Interframe spacing */
                  0, 0xf2,
                  0x48,      /* Promiscuous, Broadcast disable, CRS & CDT */
                  0, 0x40,
                  0xf2,      /* Stripping, Padding, Receive CRC Transfer */
                  0x80,      /* 0x40=Force full-duplex, 0x80=Allowfull-duplex*/
                  0x3f,      /* Multiple IA */
                  0x0D }     /* Multicast all */
};

static struct net_device_operations ifec_operations = {
        .open     = ifec_net_open,
        .close    = ifec_net_close,
        .transmit = ifec_net_transmit,
        .poll     = ifec_net_poll,
        .irq      = ifec_net_irq
};

/******************* iPXE PCI Device Driver API functions ********************/

/*
 * Initialize the PCI device.
 *
 * @v pci               The device's associated pci_device structure.
 * @v id                The PCI device + vendor id.
 * @ret rc              Returns zero if successfully initialized.
 *
 * This function is called very early on, while iPXE is initializing.
 * This is a iPXE PCI Device Driver API function.
 */
static int ifec_pci_probe ( struct pci_device *pci )
{
        struct net_device *netdev;
        struct ifec_private *priv;
        int rc;

        DBGP ( "ifec_pci_probe: " );

        if ( pci->ioaddr == 0 )
                return -EINVAL;

        netdev = alloc_etherdev ( sizeof(*priv) );
        if ( !netdev )
                return -ENOMEM;

        netdev_init ( netdev, &ifec_operations );
        priv = netdev->priv;

        pci_set_drvdata ( pci, netdev );
        netdev->dev = &pci->dev;

        /* enable bus master, etc */
        adjust_pci_device( pci );

        DBGP ( "pci " );

        memset ( priv, 0, sizeof(*priv) );
        priv->ioaddr = pci->ioaddr;

        ifec_reset ( netdev );
        DBGP ( "reset " );

        ifec_init_eeprom ( netdev );

        /* read MAC address */
        nvs_read ( &priv->eeprom.nvs, EEPROM_ADDR_MAC_0, netdev->hw_addr,
                   ETH_ALEN );
        /* read mdio_register */
        nvs_read ( &priv->eeprom.nvs, EEPROM_ADDR_MDIO_REGISTER,
                   &priv->mdio_register, 2 );

        if ( ( rc = register_netdev ( netdev ) ) != 0 )
                goto error;

        netdev_link_up ( netdev );

        DBGP ( "ints\n" );

        return 0;

error:
        ifec_reset     ( netdev );
        netdev_nullify ( netdev );
        netdev_put     ( netdev );

        return rc;
}

/*
 * Remove a device from the PCI device list.
 *
 * @v pci               PCI device to remove.
 *
 * This is a PCI Device Driver API function.
 */
static void ifec_pci_remove ( struct pci_device *pci )
{
        struct net_device *netdev = pci_get_drvdata ( pci );

        DBGP ( "ifec_pci_remove\n" );

        unregister_netdev ( netdev );
        ifec_reset        ( netdev );
        netdev_nullify    ( netdev );
        netdev_put        ( netdev );
}

/****************** iPXE Network Device Driver API functions *****************/

/*
 * Close a network device.
 *
 * @v netdev            Device to close.
 *
 * This is a iPXE Network Device Driver API function.
 */
static void ifec_net_close ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;
        unsigned short intr_status;

        DBGP ( "ifec_net_close\n" );

        /* disable interrupts */
        ifec_net_irq ( netdev, 0 );

        /* Ack & clear ints */
        intr_status = inw ( ioaddr + SCBStatus );
        outw ( intr_status, ioaddr + SCBStatus );
        inw ( ioaddr + SCBStatus );

        ifec_reset ( netdev );

        /* Free any resources */
        ifec_free ( netdev );
}

/* Interrupts to be masked */
#define INTERRUPT_MASK  ( SCBMaskEarlyRx | SCBMaskFlowCtl )

/*
 * Enable or disable IRQ masking.
 *
 * @v netdev            Device to control.
 * @v enable            Zero to mask off IRQ, non-zero to enable IRQ.
 *
 * This is a iPXE Network Driver API function.
 */
static void ifec_net_irq ( struct net_device *netdev, int enable )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;

        DBGP ( "ifec_net_irq\n" );

        outw ( enable ? INTERRUPT_MASK : SCBMaskAll, ioaddr + SCBCmd );
}

/*
 * Opens a network device.
 *
 * @v netdev            Device to be opened.
 * @ret rc              Non-zero if failed to open.
 *
 * This enables tx and rx on the device.
 * This is a iPXE Network Device Driver API function.
 */
static int ifec_net_open ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        struct ifec_ias *ias = NULL;
        struct ifec_cfg *cfg = NULL;
        int i, options;
        int rc = -ENOMEM;

        DBGP ( "ifec_net_open: " );

        /* Ensure interrupts are disabled. */
        ifec_net_irq ( netdev, 0 );

        /* Initialize Command Unit and Receive Unit base addresses. */
        ifec_scb_cmd ( netdev, 0, RUAddrLoad );
        ifec_scb_cmd ( netdev, virt_to_bus ( &priv->stats ), CUStatsAddr );
        ifec_scb_cmd ( netdev, 0, CUCmdBase );

        /* Initialize both rings */
        if ( ( rc = ifec_rx_setup ( netdev ) ) != 0 )
                goto error;
        if ( ( rc = ifec_tx_setup ( netdev ) ) != 0 )
                goto error;

        /* Initialize MDIO */
        options = 0x00; /* 0x40 = 10mbps half duplex, 0x00 = Autosense */
        ifec_mdio_setup ( netdev, options );

        /* Prepare MAC address w/ Individual Address Setup (ias) command.*/
        ias = malloc_phys ( sizeof ( *ias ), CB_ALIGN );
        if ( !ias ) {
                rc = -ENOMEM;
                goto error;
        }
        ias->command      = CmdIASetup;
        ias->status       = 0;
        memcpy ( ias->ia, netdev->ll_addr, ETH_ALEN );

        /* Prepare operating parameters w/ a configure command. */
        cfg = malloc_phys ( sizeof ( *cfg ), CB_ALIGN );
        if ( !cfg ) {
                rc = -ENOMEM;
                goto error;
        }
        memcpy ( cfg, &ifec_cfg, sizeof ( *cfg ) );
        cfg->link     = virt_to_bus ( priv->tcbs );
        cfg->byte[19] = ( options & 0x10 ) ? 0xC0 : 0x80;
        ias->link     = virt_to_bus ( cfg );

        /* Issue the ias and configure commands. */
        ifec_scb_cmd ( netdev, virt_to_bus ( ias ), CUStart );
        ifec_scb_cmd_wait ( netdev );
        priv->configured = 1;

        /* Wait up to 10 ms for configuration to initiate */
        for ( i = 10; i && !cfg->status; i-- )
                mdelay ( 1 );
        if ( ! cfg->status ) {
                DBG ( "Failed to initiate!\n" );
                goto error;
        }
        free_phys ( ias, sizeof ( *ias ) );
        free_phys ( cfg, sizeof ( *cfg ) );
        DBG2 ( "cfg " );

        /* Enable rx by sending ring address to card */
        if ( priv->rfds[0] != NULL ) {
                ifec_scb_cmd ( netdev, virt_to_bus( priv->rfds[0] ), RUStart );
                ifec_scb_cmd_wait ( netdev );
        }
        DBG2 ( "rx_start\n" );

        return 0;

error:
        free_phys ( cfg, sizeof ( *cfg ) );
        free_phys ( ias, sizeof ( *ias ) );
        ifec_free ( netdev );
        ifec_reset ( netdev );
        return rc;
}

/*
 * This function allows a driver to process events during operation.
 *
 * @v netdev            Device being polled.
 *
 * This is called periodically by iPXE to let the driver check the status of
 * transmitted packets and to allow the driver to check for received packets.
 * This is a iPXE Network Device Driver API function.
 */
static void ifec_net_poll ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        unsigned short intr_status;

        DBGP ( "ifec_net_poll\n" );

        /* acknowledge interrupts ASAP */
        intr_status = inw ( priv->ioaddr + SCBStatus );
        outw ( intr_status, priv->ioaddr + SCBStatus );
        inw ( priv->ioaddr + SCBStatus );

        DBG2 ( "poll - status: 0x%04X\n", intr_status );

        /* anything to do here? */
        if ( ( intr_status & ( ~INTERRUPT_MASK ) ) == 0 )
                return;

        /* process received and transmitted packets */
        ifec_tx_process ( netdev );
        ifec_rx_process ( netdev );

        ifec_check_ru_status ( netdev, intr_status );

        return;
}

/*
 * This transmits a packet.
 *
 * @v netdev            Device to transmit from.
 * @v iobuf             Data to transmit.
 * @ret rc              Non-zero if failed to transmit.
 *
 * This is a iPXE Network Driver API function.
 */
static int ifec_net_transmit ( struct net_device *netdev,
                               struct io_buffer *iobuf )
{
        struct ifec_private *priv = netdev->priv;
        struct ifec_tcb *tcb = priv->tcb_head->next;
        unsigned long ioaddr = priv->ioaddr;

        DBGP ( "ifec_net_transmit\n" );

        /* Wait for TCB to become available. */
        if ( tcb->status || tcb->iob ) {
                DBG ( "TX overflow\n" );
                return -ENOBUFS;
        }

        DBG2 ( "transmitting packet (%zd bytes). status = %hX, cmd=%hX\n",
                iob_len ( iobuf ), tcb->status, inw ( ioaddr + SCBCmd ) );

        tcb->command   = CmdSuspend | CmdTx | CmdTxFlex;
        tcb->count     = 0x01208000;
        tcb->tbd_addr0 = virt_to_bus ( iobuf->data );
        tcb->tbd_size0 = 0x3FFF & iob_len ( iobuf );
        tcb->iob = iobuf;

        ifec_tx_wake ( netdev );

        /* Append to end of ring. */
        priv->tcb_head = tcb;

        return 0;
}

/*************************** Local support functions *************************/

/* Define what each GPIO Pin does */
static const uint16_t ifec_ee_bits[] = {
        [SPI_BIT_SCLK]  = EE_SHIFT_CLK,
        [SPI_BIT_MOSI]  = EE_DATA_WRITE,
        [SPI_BIT_MISO]  = EE_DATA_READ,
        [SPI_BIT_SS(0)] = EE_ENB,
};

/*
 * Read a single bit from the GPIO pins used for SPI.
 * should be called by SPI bitbash functions only
 *
 * @v basher            Bitbash device
 * @v bit_id            Line to be read
 */
static int ifec_spi_read_bit ( struct bit_basher *basher,
                               unsigned int bit_id )
{
        struct ifec_private *priv =
                container_of ( basher, struct ifec_private, spi.basher );
        unsigned long ee_addr = priv->ioaddr + CSREeprom;
        unsigned int ret = 0;
        uint16_t mask;

        DBGP ( "ifec_spi_read_bit\n" );

        mask = ifec_ee_bits[bit_id];
        ret = inw (ee_addr);

        return ( ret & mask ) ? 1 : 0;
}

/*
 * Write a single bit to the GPIO pins used for SPI.
 * should be called by SPI bitbash functions only
 *
 * @v basher            Bitbash device
 * @v bit_id            Line to write to
 * @v data              Value to write
 */
static void ifec_spi_write_bit ( struct bit_basher *basher,
                                 unsigned int bit_id,
                                 unsigned long data )
{
        struct ifec_private *priv =
                container_of ( basher, struct ifec_private, spi.basher );
        unsigned long ee_addr = priv->ioaddr + CSREeprom;
        short val;
        uint16_t mask = ifec_ee_bits[bit_id];

        DBGP ( "ifec_spi_write_bit\n" );

        val = inw ( ee_addr );
        val &= ~mask;
        val |= data & mask;

        outw ( val, ee_addr );
}

/* set function pointer to SPI read- and write-bit functions */
static struct bit_basher_operations ifec_basher_ops = {
        .read = ifec_spi_read_bit,
        .write = ifec_spi_write_bit,
};

/*
 * Initialize the eeprom stuff
 *
 * @v netdev            Network device
 */
static void ifec_init_eeprom ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;

        DBGP ( "ifec_init_eeprom\n" );

        priv->spi.basher.op = &ifec_basher_ops;
        priv->spi.bus.mode = SPI_MODE_THREEWIRE;
        init_spi_bit_basher ( &priv->spi );

        priv->eeprom.bus = &priv->spi.bus;

        /* init as 93c46(93c14 compatible) first, to set the command len,
         * block size and word len. Needs to be set for address len detection.
         */
        init_at93c46 ( &priv->eeprom, 16 );

        /* detect address length, */
        threewire_detect_address_len ( &priv->eeprom );

        /* address len == 8 means 93c66 instead of 93c46 */
        if ( priv->eeprom.address_len == 8 )
                init_at93c66 ( &priv->eeprom, 16 );
}

/*
 * Support function: ifec_mdio_read
 *
 * This probably reads a register in the "physical media interface chip".
 * -- REW
 */
static int ifec_mdio_read ( struct net_device *netdev, int phy_id,
                                                       int location )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;
        int val;
        int boguscnt = 64*4;     /* <64 usec. to complete, typ 27 ticks */

        DBGP ( "ifec_mdio_read\n" );

        outl ( 0x08000000 | ( location << 16 ) | ( phy_id << 21 ),
               ioaddr + CSRCtrlMDI );
        do {
                udelay ( 16 );

                val = inl ( ioaddr + CSRCtrlMDI );

                if ( --boguscnt < 0 ) {
                        DBG ( " ifec_mdio_read() time out with val = %X.\n",
                                 val );
                        break;
                }
        } while (! ( val & 0x10000000 ) );
        return val & 0xffff;
}

/*
 * Initializes MDIO.
 *
 * @v netdev            Network device
 * @v options           MDIO options
 */
static void ifec_mdio_setup ( struct net_device *netdev, int options )
{
        struct ifec_private *priv = netdev->priv;
        unsigned short mdio_register = priv->mdio_register;

        DBGP ( "ifec_mdio_setup\n" );

        if (   ( (mdio_register>>8) & 0x3f ) == DP83840
            || ( (mdio_register>>8) & 0x3f ) == DP83840A ) {
                int mdi_reg23 = ifec_mdio_read ( netdev, mdio_register
                                                  & 0x1f, 23 ) | 0x0422;
                if (CONGENB)
                        mdi_reg23 |= 0x0100;
                DBG2 ( "DP83840 specific setup, setting register 23 to "
                                                         "%hX.\n", mdi_reg23 );
                ifec_mdio_write ( netdev, mdio_register & 0x1f, 23, mdi_reg23 );
        }
        DBG2 ( "dp83840 " );
        if ( options != 0 ) {
                ifec_mdio_write ( netdev, mdio_register & 0x1f, 0,
                                           ( (options & 0x20) ? 0x2000 : 0 ) |
                                           ( (options & 0x10) ? 0x0100 : 0 ) );
                DBG2 ( "set mdio_register. " );
        }
}

/*
 * Support function: ifec_mdio_write
 *
 * This probably writes to the "physical media interface chip".
 * -- REW
 */
static int ifec_mdio_write ( struct net_device *netdev,
                             int phy_id, int location, int value )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;
        int val;
        int boguscnt = 64*4;     /* <64 usec. to complete, typ 27 ticks */

        DBGP ( "ifec_mdio_write\n" );

        outl ( 0x04000000 | ( location << 16 ) | ( phy_id << 21 ) | value,
               ioaddr + CSRCtrlMDI );
        do {
                udelay ( 16 );

                val = inl ( ioaddr + CSRCtrlMDI );
                if ( --boguscnt < 0 ) {
                        DBG ( " ifec_mdio_write() time out with val = %X.\n",
                              val );
                        break;
                }
        } while (! ( val & 0x10000000 ) );
        return val & 0xffff;
}

/*
 * Resets the hardware.
 *
 * @v netdev            Network device
 */
static void ifec_reset ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;

        DBGP ( "ifec_reset\n" );

        /* do partial reset first */
        outl ( PortPartialReset, ioaddr + CSRPort );
        inw ( ioaddr + SCBStatus );
        udelay ( 20 );

        /* full reset */
        outl ( PortReset, ioaddr + CSRPort );
        inw ( ioaddr + SCBStatus );
        udelay ( 20 );

        /* disable interrupts again */
        ifec_net_irq ( netdev, 0 );
}

/*
 * free()s the tx/rx rings.
 *
 * @v netdev            Network device
 */
static void ifec_free ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        int i;

        DBGP ( "ifec_free\n" );

        /* free all allocated receive io_buffers */
        for ( i = 0; i < RFD_COUNT; i++ ) {
                free_iob ( priv->rx_iobs[i] );
                priv->rx_iobs[i] = NULL;
                priv->rfds[i] = NULL;
        }

        /* free TX ring buffer */
        free_phys ( priv->tcbs, TX_RING_BYTES );

        priv->tcbs = NULL;
}

/*
 * Initializes an RFD.
 *
 * @v rfd               RFD struct to initialize
 * @v command           Command word
 * @v link              Link value
 */
static void ifec_rfd_init ( struct ifec_rfd *rfd, s16 command, u32 link )
{
        DBGP ( "ifec_rfd_init\n" );

        rfd->status      = 0;
        rfd->command     = command;
        rfd->rx_buf_addr = 0xFFFFFFFF;
        rfd->count       = 0;
        rfd->size        = RFD_PACKET_LEN;
        rfd->link        = link;
}

/*
 * Send address of new RFD to card
 *
 * @v netdev            Network device
 */
static void ifec_reprime_ru ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        int cur_rx = priv->cur_rx;
        
        DBGP ( "ifec_reprime_ru\n" );
        
        if ( priv->rfds[cur_rx] != NULL ) {
                ifec_scb_cmd ( netdev, virt_to_bus ( priv->rfds[cur_rx] ),
                               RUStart );
                ifec_scb_cmd_wait ( netdev );
        }
}

/*
 * Check if reprime of RU needed
 *
 * @v netdev            Network device
 */
static void ifec_check_ru_status ( struct net_device *netdev,
                                   unsigned short intr_status )
{
        struct ifec_private *priv = netdev->priv;

        DBGP ( "ifec_check_ru_status\n" );

        /*
        * The chip may have suspended reception for various reasons.
        * Check for that, and re-prime it should this be the case.
        */
        switch ( ( intr_status >> 2 ) & 0xf ) {
                case 0:  /* Idle */
                case 4:  /* Ready */
                        break;
                case 1:  /* Suspended */
                case 2:  /* No resources (RFDs) */
                case 9:  /* Suspended with no more RBDs */
                case 10: /* No resources due to no RBDs */
                case 12: /* Ready with no RBDs */
                        DBG ( "ifec_net_poll: RU reprimed.\n" );
                        ifec_reprime_ru ( netdev );
                        break;
                default:
                        /* reserved values */
                        DBG ( "ifec_net_poll: RU state anomaly: %i\n",
                              ( inw ( priv->ioaddr + SCBStatus ) >> 2 ) & 0xf );
                        break;
        }
}

#define RFD_STATUS ( RFD_OK | RFDRxCol | RFDRxErr | RFDShort | \
                     RFDDMAOverrun | RFDNoBufs | RFDCRCError )
/*
 * Looks for received packets in the rx ring, reports success or error to
 * the core accordingly. Starts reallocation of rx ring.
 *
 * @v netdev            Network device
 */
static void ifec_rx_process ( struct net_device *netdev )
{
        struct ifec_private *priv   = netdev->priv;
        int cur_rx = priv->cur_rx;
        struct io_buffer *iob = priv->rx_iobs[cur_rx];
        struct ifec_rfd *rfd = priv->rfds[cur_rx];
        unsigned int rx_len;
        s16 status;

        DBGP ( "ifec_rx_process\n" );

        /* Process any received packets */
        while ( iob && rfd && ( status = rfd->status ) ) {
                rx_len = rfd->count & RFDMaskCount;

                DBG2 ( "Got a packet: Len = %d, cur_rx = %d.\n", rx_len,
                       cur_rx );
                DBGIO_HD ( (void*)rfd->packet, 0x30 );

                if ( ( status & ( RFD_STATUS & ~RFDShort ) ) != RFD_OK ) {
                        DBG ( "Corrupted packet received. "
                              "Status = %#08hx\n", status );
                        netdev_rx_err ( netdev, iob, -EINVAL );
                } else {
                        /* Hand off the packet to the network subsystem */
                        iob_put ( iob, rx_len );
                        DBG2 ( "Received packet: %p, len: %d\n", iob, rx_len );
                        netdev_rx ( netdev, iob );
                }

                /* make sure we don't reuse this RFD */
                priv->rx_iobs[cur_rx] = NULL;
                priv->rfds[cur_rx] = NULL;

                /* Next RFD */
                priv->cur_rx = ( cur_rx + 1 ) % RFD_COUNT;
                cur_rx = priv->cur_rx;
                iob = priv->rx_iobs[cur_rx];
                rfd = priv->rfds[cur_rx];
        }

        ifec_refill_rx_ring ( netdev );
}

/*
 * Allocates io_buffer, set pointers in ifec_private structure accordingly,
 * reserves space for RFD header in io_buffer.
 *
 * @v netdev            Network device
 * @v cur               Descriptor number to work on
 * @v cmd               Value to set cmd field in RFD to
 * @v link              Pointer to ned RFD
 * @ret rc              0 on success, negative on failure
 */
static int ifec_get_rx_desc ( struct net_device *netdev, int cur, int cmd,
                              int link )
{
        struct ifec_private *priv = netdev->priv;
        struct ifec_rfd *rfd  = priv->rfds[cur];

        DBGP ( "ifec_get_rx_desc\n" );

        priv->rx_iobs[cur] = alloc_iob ( sizeof ( *rfd ) );
        if ( ! priv->rx_iobs[cur] ) {
                DBG ( "alloc_iob failed. desc. nr: %d\n", cur );
                priv->rfds[cur] = NULL;
                return -ENOMEM;
        }

        /* Initialize new tail. */
        priv->rfds[cur] = priv->rx_iobs[cur]->data;
        ifec_rfd_init ( priv->rfds[cur], cmd, link );
        iob_reserve ( priv->rx_iobs[cur], RFD_HEADER_LEN );

        return 0;
}

/*
 * Allocate new descriptor entries and initialize them if needed
 *
 * @v netdev            Network device
 */
static void ifec_refill_rx_ring ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        int i, cur_rx;
        unsigned short intr_status;

        DBGP ( "ifec_refill_rx_ring\n" );

        for ( i = 0; i < RFD_COUNT; i++ ) {
                cur_rx = ( priv->cur_rx + i ) % RFD_COUNT;
                /* only refill if empty */
                if ( priv->rfds[cur_rx] != NULL ||
                     priv->rx_iobs[cur_rx] != NULL )
                        continue;

                DBG2 ( "refilling RFD %d\n", cur_rx );

                if ( ifec_get_rx_desc ( netdev, cur_rx,
                     CmdSuspend | CmdEndOfList, 0 ) == 0 ) {
                        if ( i > 0 ) {
                                int prev_rx = ( ( ( cur_rx + RFD_COUNT ) - 1 )
                                                % RFD_COUNT );
                                struct ifec_rfd *rfd = priv->rfds[prev_rx];

                                rfd->command = 0;
                                rfd->link = virt_to_bus ( priv->rfds[cur_rx] );
                        }
                }
        }

        intr_status = inw ( priv->ioaddr + SCBStatus );
        ifec_check_ru_status ( netdev, intr_status );
}

/*
 * Initial allocation & initialization of the rx ring.
 *
 * @v netdev            Device of rx ring.
 * @ret rc              Non-zero if error occurred
 */
static int ifec_rx_setup ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        int i;

        DBGP ( "ifec_rx_setup\n" );

        priv->cur_rx = 0;

        /* init values for ifec_refill_rx_ring() */
        for ( i = 0; i < RFD_COUNT; i++ ) {
                priv->rfds[i] = NULL;
                priv->rx_iobs[i] = NULL;
        }
        ifec_refill_rx_ring ( netdev );

        return 0;
}

/*
 * Initiates a SCB command.
 *
 * @v netdev            Network device
 * @v ptr               General pointer value for command.
 * @v cmd               Command to issue.
 * @ret rc              Non-zero if command not issued.
 */
static int ifec_scb_cmd ( struct net_device *netdev, u32 ptr, u8 cmd )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;
        int rc;

        DBGP ( "ifec_scb_cmd\n" );

        rc = ifec_scb_cmd_wait ( netdev );      /* Wait until ready */
        if ( !rc ) {
                outl ( ptr, ioaddr + SCBPointer );
                outb ( cmd, ioaddr + SCBCmd );          /* Issue command */
        }
        return rc;
}

/*
 * Wait for command unit to accept a command.
 *
 * @v cmd_ioaddr        I/O address of command register.
 * @ret rc              Non-zero if command timed out.
 */
static int ifec_scb_cmd_wait ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long cmd_ioaddr = priv->ioaddr + SCBCmd;
        int rc, wait = CU_CMD_TIMEOUT;

        DBGP ( "ifec_scb_cmd_wait\n" );

        for ( ; wait && ( rc = inb ( cmd_ioaddr ) ); wait-- )
                udelay ( 1 );

        if ( !wait )
                DBG ( "ifec_scb_cmd_wait timeout!\n" );
        return rc;
}

/*
 * Check status of transmitted packets & perform tx completions.
 *
 * @v netdev            Network device.
 */
static void ifec_tx_process ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        struct ifec_tcb *tcb = priv->tcb_tail;
        s16 status;

        DBGP ( "ifec_tx_process\n" );

        /* Check status of transmitted packets */
        while ( ( status = tcb->status ) && tcb->iob ) {
                if ( status & TCB_U ) {
                        /* report error to iPXE */
                        DBG ( "ifec_tx_process : tx error!\n " );
                        netdev_tx_complete_err ( netdev, tcb->iob, -EINVAL );
                } else {
                        /* report successful transmit */
                        netdev_tx_complete ( netdev, tcb->iob );
                }
                DBG2 ( "tx completion\n" );

                tcb->iob = NULL;
                tcb->status = 0;

                priv->tcb_tail = tcb->next;     /* Next TCB */
                tcb = tcb->next;
        }
}

/*
 * Allocates & initialize tx resources.
 *
 * @v netdev            Network device.
 * @ret rc              Non-zero if error occurred.
 */
static int ifec_tx_setup ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        struct ifec_tcb *tcb;
        int i;

        DBGP ( "ifec_tx_setup\n" );

        /* allocate tx ring */
        priv->tcbs = malloc_phys ( TX_RING_BYTES, CB_ALIGN );
        if ( !priv->tcbs ) {
                DBG ( "TX-ring allocation failed\n" );
                return -ENOMEM;
        }

        tcb = priv->tcb_tail = priv->tcbs;
        priv->tx_curr = priv->tx_tail = 0;
        priv->tx_cnt = 0;

        for ( i = 0; i < TCB_COUNT; i++, tcb++ ) {
                tcb->status    = 0;
                tcb->count     = 0x01208000;
                tcb->iob       = NULL;
                tcb->tbda_addr = virt_to_bus ( &tcb->tbd_addr0 );
                tcb->link      = virt_to_bus ( tcb + 1 );
                tcb->next      = tcb + 1;
        }
        /* We point tcb_head at the last TCB, so the first ifec_net_transmit()
         * will use the first (head->next) TCB to transmit. */
        priv->tcb_head = --tcb;
        tcb->link = virt_to_bus ( priv->tcbs );
        tcb->next = priv->tcbs;
        
        return 0;
}

/*
 * Wake up the Command Unit and issue a Resume/Start.
 *
 * @v netdev            Network device containing Command Unit
 *
 * The time between clearing the S bit and issuing Resume must be as short as
 * possible to prevent a race condition. As noted in linux eepro100.c :
 *   Note: Watch out for the potential race condition here: imagine
 *      erasing the previous suspend
 *              the chip processes the previous command
 *              the chip processes the final command, and suspends
 *      doing the CU_RESUME
 *              the chip processes the next-yet-valid post-final-command.
 *   So blindly sending a CU_RESUME is only safe if we do it immediately after
 *   erasing the previous CmdSuspend, without the possibility of an intervening
 *   delay.
 */
void ifec_tx_wake ( struct net_device *netdev )
{
        struct ifec_private *priv = netdev->priv;
        unsigned long ioaddr = priv->ioaddr;
        struct ifec_tcb *tcb = priv->tcb_head->next;

        DBGP ( "ifec_tx_wake\n" );

        /* For the special case of the first transmit, we issue a START. The
         * card won't RESUME after the configure command. */
        if ( priv->configured ) {
                priv->configured = 0;
                ifec_scb_cmd ( netdev, virt_to_bus ( tcb ), CUStart );
                ifec_scb_cmd_wait ( netdev );
                return;
        }

        /* Resume if suspended. */
        switch ( ( inw ( ioaddr + SCBStatus ) >> 6 ) & 0x3 ) {
        case 0:  /* Idle - We should not reach this state. */
                DBG2 ( "ifec_tx_wake: tx idle!\n" );
                ifec_scb_cmd ( netdev, virt_to_bus ( tcb ), CUStart );
                ifec_scb_cmd_wait ( netdev );
                return;
        case 1:  /* Suspended */
                DBG2 ( "s" );
                break;
        default: /* Active */
                DBG2 ( "a" );
        }
        ifec_scb_cmd_wait ( netdev );
        outl ( 0, ioaddr + SCBPointer );
        priv->tcb_head->command &= ~CmdSuspend;
        /* Immediately issue Resume command */
        outb ( CUResume, ioaddr + SCBCmd );
        ifec_scb_cmd_wait ( netdev );
}

/*********************************************************************/

static struct pci_device_id ifec_nics[] = {
PCI_ROM(0x8086, 0x1029, "id1029",        "Intel EtherExpressPro100 ID1029", 0),
PCI_ROM(0x8086, 0x1030, "id1030",        "Intel EtherExpressPro100 ID1030", 0),
PCI_ROM(0x8086, 0x1031, "82801cam",      "Intel 82801CAM (ICH3) Chipset Ethernet Controller", 0),
PCI_ROM(0x8086, 0x1032, "eepro100-1032", "Intel PRO/100 VE Network Connection", 0),
PCI_ROM(0x8086, 0x1033, "eepro100-1033", "Intel PRO/100 VM Network Connection", 0),
PCI_ROM(0x8086, 0x1034, "eepro100-1034", "Intel PRO/100 VM Network Connection", 0),
PCI_ROM(0x8086, 0x1035, "eepro100-1035", "Intel 82801CAM (ICH3) Chipset Ethernet Controller", 0),
PCI_ROM(0x8086, 0x1036, "eepro100-1036", "Intel 82801CAM (ICH3) Chipset Ethernet Controller", 0),
PCI_ROM(0x8086, 0x1037, "eepro100-1037", "Intel 82801CAM (ICH3) Chipset Ethernet Controller", 0),
PCI_ROM(0x8086, 0x1038, "id1038",        "Intel PRO/100 VM Network Connection", 0),
PCI_ROM(0x8086, 0x1039, "82562et",       "Intel PRO100 VE 82562ET", 0),
PCI_ROM(0x8086, 0x103a, "id103a",        "Intel Corporation 82559 InBusiness 10/100", 0),
PCI_ROM(0x8086, 0x103b, "82562etb",      "Intel PRO100 VE 82562ETB", 0),
PCI_ROM(0x8086, 0x103c, "eepro100-103c", "Intel PRO/100 VM Network Connection", 0),
PCI_ROM(0x8086, 0x103d, "eepro100-103d", "Intel PRO/100 VE Network Connection", 0),
PCI_ROM(0x8086, 0x103e, "eepro100-103e", "Intel PRO/100 VM Network Connection", 0),
PCI_ROM(0x8086, 0x1050, "82562ez",       "Intel 82562EZ Network Connection", 0),
PCI_ROM(0x8086, 0x1051, "prove",         "Intel PRO/100 VE Network Connection", 0),
PCI_ROM(0x8086, 0x1059, "82551qm",       "Intel PRO/100 M Mobile Connection", 0),
PCI_ROM(0x8086, 0x1065, "82562-3",       "Intel 82562 based Fast Ethernet Connection", 0),
PCI_ROM(0x8086, 0x1092, "82562-3",       "Intel Pro/100 VE Network", 0),
PCI_ROM(0x8086, 0x10fe, "82552",         "Intel 82552 10/100 Network Connection", 0),
PCI_ROM(0x8086, 0x1209, "82559er",       "Intel EtherExpressPro100 82559ER", 0),
PCI_ROM(0x8086, 0x1227, "82865",         "Intel 82865 EtherExpress PRO/100A", 0),
PCI_ROM(0x8086, 0x1228, "82556",         "Intel 82556 EtherExpress PRO/100 Smart", 0),
PCI_ROM(0x8086, 0x1229, "eepro100",      "Intel EtherExpressPro100", 0),
PCI_ROM(0x8086, 0x2449, "82562em",       "Intel EtherExpressPro100 82562EM", 0),
PCI_ROM(0x8086, 0x2459, "82562-1",       "Intel 82562 based Fast Ethernet Connection", 0),
PCI_ROM(0x8086, 0x245d, "82562-2",       "Intel 82562 based Fast Ethernet Connection", 0),
PCI_ROM(0x8086, 0x27dc, "eepro100-27dc", "Intel 82801G (ICH7) Chipset Ethernet Controller", 0),
PCI_ROM(0x8086, 0x5200, "eepro100-5200", "Intel EtherExpress PRO/100 Intelligent Server", 0),
PCI_ROM(0x8086, 0x5201, "eepro100-5201", "Intel EtherExpress PRO/100 Intelligent Server", 0),
};

/* Cards with device ids 0x1030 to 0x103F, 0x2449, 0x2459 or 0x245D might need
 * a workaround for hardware bug on 10 mbit half duplex (see linux driver eepro100.c)
 * 2003/03/17 gbaum */

struct pci_driver ifec_driver __pci_driver = {
        .ids      = ifec_nics,
        .id_count = ( sizeof (ifec_nics) / sizeof (ifec_nics[0]) ),
        .probe    = ifec_pci_probe,
        .remove   = ifec_pci_remove
};

/*
 * Local variables:
 *  c-basic-offset: 8
 *  c-indent-level: 8
 *  tab-width: 8
 * End:
 */