/*      $NetBSD: i82557.c,v 1.162 2024/06/29 12:11:11 riastradh Exp $   */

/*-
* Copyright (c) 1997, 1998, 1999, 2001, 2002 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
* NASA Ames Research Center.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/

/*
* Copyright (c) 1995, David Greenman
* Copyright (c) 2001 Jonathan Lemon <[email protected]>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice unmodified, this list of conditions, and the following
*    disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*      Id: if_fxp.c,v 1.113 2001/05/17 23:50:24 jlemon
*/

/*
* Device driver for the Intel i82557 fast Ethernet controller,
* and its successors, the i82558 and i82559.
*/

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: i82557.c,v 1.162 2024/06/29 12:11:11 riastradh Exp $");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <sys/errno.h>
#include <sys/device.h>
#include <sys/syslog.h>
#include <sys/proc.h>

#include <machine/endian.h>

#include <sys/rndsource.h>

#include <net/if.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_ether.h>

#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/tcp.h>
#include <netinet/udp.h>

#include <net/bpf.h>

#include <sys/bus.h>
#include <sys/intr.h>

#include <dev/mii/miivar.h>

#include <dev/ic/i82557reg.h>
#include <dev/ic/i82557var.h>

#include <dev/microcode/i8255x/rcvbundl.h>

/*
* NOTE!  On the Alpha, we have an alignment constraint.  The
* card DMAs the packet immediately following the RFA.  However,
* the first thing in the packet is a 14-byte Ethernet header.
* This means that the packet is misaligned.  To compensate,
* we actually offset the RFA 2 bytes into the cluster.  This
* aligns the packet after the Ethernet header at a 32-bit
* boundary.  HOWEVER!  This means that the RFA is misaligned!
*/
#define RFA_ALIGNMENT_FUDGE     2

/*
* The configuration byte map has several undefined fields which
* must be one or must be zero.  Set up a template for these bits
* only (assuming an i82557 chip), leaving the actual configuration
* for fxp_init().
*
* See the definition of struct fxp_cb_config for the bit definitions.
*/
const uint8_t fxp_cb_config_template[] = {
       0x0, 0x0,               /* cb_status */
       0x0, 0x0,               /* cb_command */
       0x0, 0x0, 0x0, 0x0,     /* link_addr */
       0x0,    /*  0 */
       0x0,    /*  1 */
       0x0,    /*  2 */
       0x0,    /*  3 */
       0x0,    /*  4 */
       0x0,    /*  5 */
       0x32,   /*  6 */
       0x0,    /*  7 */
       0x0,    /*  8 */
       0x0,    /*  9 */
       0x6,    /* 10 */
       0x0,    /* 11 */
       0x0,    /* 12 */
       0x0,    /* 13 */
       0xf2,   /* 14 */
       0x48,   /* 15 */
       0x0,    /* 16 */
       0x40,   /* 17 */
       0xf0,   /* 18 */
       0x0,    /* 19 */
       0x3f,   /* 20 */
       0x5,    /* 21 */
       0x0,    /* 22 */
       0x0,    /* 23 */
       0x0,    /* 24 */
       0x0,    /* 25 */
       0x0,    /* 26 */
       0x0,    /* 27 */
       0x0,    /* 28 */
       0x0,    /* 29 */
       0x0,    /* 30 */
       0x0,    /* 31 */
};

void    fxp_mii_initmedia(struct fxp_softc *);
void    fxp_mii_mediastatus(struct ifnet *, struct ifmediareq *);

void    fxp_80c24_initmedia(struct fxp_softc *);
int     fxp_80c24_mediachange(struct ifnet *);
void    fxp_80c24_mediastatus(struct ifnet *, struct ifmediareq *);

void    fxp_start(struct ifnet *);
int     fxp_ioctl(struct ifnet *, u_long, void *);
void    fxp_watchdog(struct ifnet *);
int     fxp_init(struct ifnet *);
void    fxp_stop(struct ifnet *, int);

void    fxp_txintr(struct fxp_softc *);
int     fxp_rxintr(struct fxp_softc *);

void    fxp_rx_hwcksum(struct fxp_softc *, struct mbuf *,
           const struct fxp_rfa *, u_int);

void    fxp_rxdrain(struct fxp_softc *);
int     fxp_add_rfabuf(struct fxp_softc *, bus_dmamap_t, int);
int     fxp_mdi_read(device_t, int, int, uint16_t *);
void    fxp_statchg(struct ifnet *);
int     fxp_mdi_write(device_t, int, int, uint16_t);
void    fxp_autosize_eeprom(struct fxp_softc*);
void    fxp_read_eeprom(struct fxp_softc *, uint16_t *, int, int);
void    fxp_write_eeprom(struct fxp_softc *, uint16_t *, int, int);
void    fxp_eeprom_update_cksum(struct fxp_softc *);
void    fxp_get_info(struct fxp_softc *, uint8_t *);
void    fxp_tick(void *);
void    fxp_mc_setup(struct fxp_softc *);
void    fxp_load_ucode(struct fxp_softc *);

int     fxp_copy_small = 0;

/*
* Variables for interrupt mitigating microcode.
*/
int     fxp_int_delay = 1000;           /* usec */
int     fxp_bundle_max = 6;             /* packets */

struct fxp_phytype {
       int     fp_phy;         /* type of PHY, -1 for MII at the end. */
       void    (*fp_init)(struct fxp_softc *);
} fxp_phytype_table[] = {
       { FXP_PHY_80C24,                fxp_80c24_initmedia },
       { -1,                           fxp_mii_initmedia },
};

/*
* Set initial transmit threshold at 64 (512 bytes). This is
* increased by 64 (512 bytes) at a time, to maximum of 192
* (1536 bytes), if an underrun occurs.
*/
static int tx_threshold = 64;

/*
* Wait for the previous command to be accepted (but not necessarily
* completed).
*/
static inline void
fxp_scb_wait(struct fxp_softc *sc)
{
       int i = 10000;

       while (CSR_READ_1(sc, FXP_CSR_SCB_COMMAND) && --i)
               delay(2);
       if (i == 0)
               log(LOG_WARNING,
                   "%s: WARNING: SCB timed out!\n", device_xname(sc->sc_dev));
}

/*
* Submit a command to the i82557.
*/
static inline void
fxp_scb_cmd(struct fxp_softc *sc, uint8_t cmd)
{

       CSR_WRITE_1(sc, FXP_CSR_SCB_COMMAND, cmd);
}

/*
* Finish attaching an i82557 interface.  Called by bus-specific front-end.
*/
void
fxp_attach(struct fxp_softc *sc)
{
       uint8_t enaddr[ETHER_ADDR_LEN];
       struct ifnet *ifp;
       bus_dma_segment_t seg;
       int rseg, i, error;
       struct fxp_phytype *fp;

       callout_init(&sc->sc_callout, 0);
       callout_setfunc(&sc->sc_callout, fxp_tick, sc);

       /*
        * Enable use of extended RFDs and IPCBs for 82550 and later chips.
        * Note: to use IPCB we need extended TXCB support too, and
        *       these feature flags should be set in each bus attachment.
        */
       if (sc->sc_flags & FXPF_EXT_RFA) {
               sc->sc_txcmd = htole16(FXP_CB_COMMAND_IPCBXMIT);
               sc->sc_rfa_size = RFA_EXT_SIZE;
       } else {
               sc->sc_txcmd = htole16(FXP_CB_COMMAND_XMIT);
               sc->sc_rfa_size = RFA_SIZE;
       }

       /*
        * Allocate the control data structures, and create and load the
        * DMA map for it.
        */
       if ((error = bus_dmamem_alloc(sc->sc_dmat,
           sizeof(struct fxp_control_data), PAGE_SIZE, 0, &seg, 1, &rseg,
           0)) != 0) {
               aprint_error_dev(sc->sc_dev,
                   "unable to allocate control data, error = %d\n",
                   error);
               goto fail_0;
       }

       if ((error = bus_dmamem_map(sc->sc_dmat, &seg, rseg,
           sizeof(struct fxp_control_data), (void **)&sc->sc_control_data,
           BUS_DMA_COHERENT)) != 0) {
               aprint_error_dev(sc->sc_dev,
                   "unable to map control data, error = %d\n", error);
               goto fail_1;
       }
       sc->sc_cdseg = seg;
       sc->sc_cdnseg = rseg;

       memset(sc->sc_control_data, 0, sizeof(struct fxp_control_data));

       if ((error = bus_dmamap_create(sc->sc_dmat,
           sizeof(struct fxp_control_data), 1,
           sizeof(struct fxp_control_data), 0, 0, &sc->sc_dmamap)) != 0) {
               aprint_error_dev(sc->sc_dev,
                   "unable to create control data DMA map, error = %d\n",
                   error);
               goto fail_2;
       }

       if ((error = bus_dmamap_load(sc->sc_dmat, sc->sc_dmamap,
           sc->sc_control_data, sizeof(struct fxp_control_data), NULL,
           0)) != 0) {
               aprint_error_dev(sc->sc_dev,
                   "can't load control data DMA map, error = %d\n",
                   error);
               goto fail_3;
       }

       /*
        * Create the transmit buffer DMA maps.
        */
       for (i = 0; i < FXP_NTXCB; i++) {
               if ((error = bus_dmamap_create(sc->sc_dmat, MCLBYTES,
                   (sc->sc_flags & FXPF_EXT_RFA) ?
                   FXP_IPCB_NTXSEG : FXP_NTXSEG,
                   MCLBYTES, 0, 0, &FXP_DSTX(sc, i)->txs_dmamap)) != 0) {
                       aprint_error_dev(sc->sc_dev,
                           "unable to create tx DMA map %d, error = %d\n",
                           i, error);
                       goto fail_4;
               }
       }

       /*
        * Create the receive buffer DMA maps.
        */
       for (i = 0; i < FXP_NRFABUFS; i++) {
               if ((error = bus_dmamap_create(sc->sc_dmat, MCLBYTES, 1,
                   MCLBYTES, 0, 0, &sc->sc_rxmaps[i])) != 0) {
                       aprint_error_dev(sc->sc_dev,
                           "unable to create rx DMA map %d, error = %d\n",
                           i, error);
                       goto fail_5;
               }
       }

       /* Initialize MAC address and media structures. */
       fxp_get_info(sc, enaddr);

       aprint_normal_dev(sc->sc_dev, "Ethernet address %s\n",
           ether_sprintf(enaddr));

       ifp = &sc->sc_ethercom.ec_if;

       /*
        * Get info about our media interface, and initialize it.  Note
        * the table terminates itself with a phy of -1, indicating
        * that we're using MII.
        */
       for (fp = fxp_phytype_table; fp->fp_phy != -1; fp++)
               if (fp->fp_phy == sc->phy_primary_device)
                       break;
       (*fp->fp_init)(sc);

       strlcpy(ifp->if_xname, device_xname(sc->sc_dev), IFNAMSIZ);
       ifp->if_softc = sc;
       ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
       ifp->if_ioctl = fxp_ioctl;
       ifp->if_start = fxp_start;
       ifp->if_watchdog = fxp_watchdog;
       ifp->if_init = fxp_init;
       ifp->if_stop = fxp_stop;
       IFQ_SET_READY(&ifp->if_snd);

       if (sc->sc_flags & FXPF_EXT_RFA) {
               /*
                * Enable hardware cksum support by EXT_RFA and IPCB.
                *
                * IFCAP_CSUM_IPv4_Tx seems to have a problem,
                * at least, on i82550 rev.12.
                * specifically, it doesn't set ipv4 checksum properly
                * when sending UDP (and probably TCP) packets with
                * 20 byte ipv4 header + 1 or 2 byte data,
                * though ICMP packets seem working.
                * FreeBSD driver has related comments.
                * We've added a workaround to handle the bug by padding
                * such packets manually.
                */
               ifp->if_capabilities =
                   IFCAP_CSUM_IPv4_Tx  | IFCAP_CSUM_IPv4_Rx  |
                   IFCAP_CSUM_TCPv4_Tx | IFCAP_CSUM_TCPv4_Rx |
                   IFCAP_CSUM_UDPv4_Tx | IFCAP_CSUM_UDPv4_Rx;
               sc->sc_ethercom.ec_capabilities |= ETHERCAP_VLAN_HWTAGGING;
               sc->sc_ethercom.ec_capenable |= ETHERCAP_VLAN_HWTAGGING;
       } else if (sc->sc_flags & FXPF_82559_RXCSUM) {
               ifp->if_capabilities =
                   IFCAP_CSUM_TCPv4_Rx |
                   IFCAP_CSUM_UDPv4_Rx;
       }

       /*
        * We can support 802.1Q VLAN-sized frames.
        */
       sc->sc_ethercom.ec_capabilities |= ETHERCAP_VLAN_MTU;

       /*
        * Attach the interface.
        */
       if_attach(ifp);
       if_deferred_start_init(ifp, NULL);
       ether_ifattach(ifp, enaddr);
       rnd_attach_source(&sc->rnd_source, device_xname(sc->sc_dev),
           RND_TYPE_NET, RND_FLAG_DEFAULT);

#ifdef FXP_EVENT_COUNTERS
       evcnt_attach_dynamic(&sc->sc_ev_txstall, EVCNT_TYPE_MISC,
           NULL, device_xname(sc->sc_dev), "txstall");
       evcnt_attach_dynamic(&sc->sc_ev_txintr, EVCNT_TYPE_INTR,
           NULL, device_xname(sc->sc_dev), "txintr");
       evcnt_attach_dynamic(&sc->sc_ev_rxintr, EVCNT_TYPE_INTR,
           NULL, device_xname(sc->sc_dev), "rxintr");
       if (sc->sc_flags & FXPF_FC) {
               evcnt_attach_dynamic(&sc->sc_ev_txpause, EVCNT_TYPE_MISC,
                   NULL, device_xname(sc->sc_dev), "txpause");
               evcnt_attach_dynamic(&sc->sc_ev_rxpause, EVCNT_TYPE_MISC,
                   NULL, device_xname(sc->sc_dev), "rxpause");
       }
#endif /* FXP_EVENT_COUNTERS */

       /* The attach is successful. */
       sc->sc_flags |= FXPF_ATTACHED;

       return;

       /*
        * Free any resources we've allocated during the failed attach
        * attempt.  Do this in reverse order and fall though.
        */
fail_5:
       for (i = 0; i < FXP_NRFABUFS; i++) {
               if (sc->sc_rxmaps[i] != NULL)
                       bus_dmamap_destroy(sc->sc_dmat, sc->sc_rxmaps[i]);
       }
fail_4:
       for (i = 0; i < FXP_NTXCB; i++) {
               if (FXP_DSTX(sc, i)->txs_dmamap != NULL)
                       bus_dmamap_destroy(sc->sc_dmat,
                           FXP_DSTX(sc, i)->txs_dmamap);
       }
       bus_dmamap_unload(sc->sc_dmat, sc->sc_dmamap);
fail_3:
       bus_dmamap_destroy(sc->sc_dmat, sc->sc_dmamap);
fail_2:
       bus_dmamem_unmap(sc->sc_dmat, (void *)sc->sc_control_data,
           sizeof(struct fxp_control_data));
fail_1:
       bus_dmamem_free(sc->sc_dmat, &seg, rseg);
fail_0:
       return;
}

void
fxp_mii_initmedia(struct fxp_softc *sc)
{
       struct mii_data * const mii = &sc->sc_mii;
       int flags;

       sc->sc_flags |= FXPF_MII;

       mii->mii_ifp = &sc->sc_ethercom.ec_if;
       mii->mii_readreg = fxp_mdi_read;
       mii->mii_writereg = fxp_mdi_write;
       mii->mii_statchg = fxp_statchg;

       sc->sc_ethercom.ec_mii = mii;
       ifmedia_init(&mii->mii_media, IFM_IMASK, ether_mediachange,
           fxp_mii_mediastatus);

       flags = MIIF_NOISOLATE;
       if (sc->sc_flags & FXPF_FC)
               flags |= MIIF_FORCEANEG | MIIF_DOPAUSE;
       /*
        * The i82557 wedges if all of its PHYs are isolated!
        */
       mii_attach(sc->sc_dev, mii, 0xffffffff, MII_PHY_ANY,
           MII_OFFSET_ANY, flags);
       if (LIST_EMPTY(&mii->mii_phys)) {
               ifmedia_add(&mii->mii_media, IFM_ETHER | IFM_NONE, 0, NULL);
               ifmedia_set(&mii->mii_media, IFM_ETHER | IFM_NONE);
       } else
               ifmedia_set(&mii->mii_media, IFM_ETHER | IFM_AUTO);
}

void
fxp_80c24_initmedia(struct fxp_softc *sc)
{
       struct mii_data * const mii = &sc->sc_mii;

       /*
        * The Seeq 80c24 AutoDUPLEX(tm) Ethernet Interface Adapter
        * doesn't have a programming interface of any sort.  The
        * media is sensed automatically based on how the link partner
        * is configured.  This is, in essence, manual configuration.
        */
       aprint_normal_dev(sc->sc_dev,
           "Seeq 80c24 AutoDUPLEX media interface present\n");
       ifmedia_init(&mii->mii_media, 0, fxp_80c24_mediachange,
           fxp_80c24_mediastatus);
       ifmedia_add(&mii->mii_media, IFM_ETHER | IFM_MANUAL, 0, NULL);
       ifmedia_set(&mii->mii_media, IFM_ETHER | IFM_MANUAL);
}

/*
* Initialize the interface media.
*/
void
fxp_get_info(struct fxp_softc *sc, uint8_t *enaddr)
{
       uint16_t data, myea[ETHER_ADDR_LEN / 2];

       /*
        * Reset to a stable state.
        */
       CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SELECTIVE_RESET);
       DELAY(100);

       sc->sc_eeprom_size = 0;
       fxp_autosize_eeprom(sc);
       if (sc->sc_eeprom_size == 0) {
               aprint_error_dev(sc->sc_dev, "failed to detect EEPROM size\n");
               sc->sc_eeprom_size = 6; /* XXX panic here? */
       }
#ifdef DEBUG
       aprint_debug_dev(sc->sc_dev, "detected %d word EEPROM\n",
           1 << sc->sc_eeprom_size);
#endif

       /*
        * Get info about the primary PHY
        */
       fxp_read_eeprom(sc, &data, 6, 1);
       sc->phy_primary_device =
           (data & FXP_PHY_DEVICE_MASK) >> FXP_PHY_DEVICE_SHIFT;

       /*
        * Read MAC address.
        */
       fxp_read_eeprom(sc, myea, 0, 3);
       enaddr[0] = myea[0] & 0xff;
       enaddr[1] = myea[0] >> 8;
       enaddr[2] = myea[1] & 0xff;
       enaddr[3] = myea[1] >> 8;
       enaddr[4] = myea[2] & 0xff;
       enaddr[5] = myea[2] >> 8;

       /*
        * Systems based on the ICH2/ICH2-M chip from Intel, as well
        * as some i82559 designs, have a defect where the chip can
        * cause a PCI protocol violation if it receives a CU_RESUME
        * command when it is entering the IDLE state.
        *
        * The work-around is to disable Dynamic Standby Mode, so that
        * the chip never deasserts #CLKRUN, and always remains in the
        * active state.
        *
        * Unfortunately, the only way to disable Dynamic Standby is
        * to frob an EEPROM setting and reboot (the EEPROM setting
        * is only consulted when the PCI bus comes out of reset).
        *
        * See Intel 82801BA/82801BAM Specification Update, Errata #30.
        */
       if (sc->sc_flags & FXPF_HAS_RESUME_BUG) {
               fxp_read_eeprom(sc, &data, 10, 1);
               if (data & 0x02) {              /* STB enable */
                       aprint_error_dev(sc->sc_dev, "WARNING: "
                           "Disabling dynamic standby mode in EEPROM "
                           "to work around a\n");
                       aprint_normal_dev(sc->sc_dev,
                           "WARNING: hardware bug.  You must reset "
                           "the system before using this\n");
                       aprint_normal_dev(sc->sc_dev, "WARNING: interface.\n");
                       data &= ~0x02;
                       fxp_write_eeprom(sc, &data, 10, 1);
                       aprint_normal_dev(sc->sc_dev, "new EEPROM ID: 0x%04x\n",
                           data);
                       fxp_eeprom_update_cksum(sc);
               }
       }

       /* Receiver lock-up workaround detection. (FXPF_RECV_WORKAROUND) */
       /* Due to false positives we make it conditional on setting link1 */
       fxp_read_eeprom(sc, &data, 3, 1);
       if ((data & 0x03) != 0x03) {
               aprint_verbose_dev(sc->sc_dev,
                   "May need receiver lock-up workaround\n");
       }
}

static void
fxp_eeprom_shiftin(struct fxp_softc *sc, int data, int len)
{
       uint16_t reg;
       int x;

       for (x = 1 << (len - 1); x != 0; x >>= 1) {
               DELAY(40);
               if (data & x)
                       reg = FXP_EEPROM_EECS | FXP_EEPROM_EEDI;
               else
                       reg = FXP_EEPROM_EECS;
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
               DELAY(40);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL,
                   reg | FXP_EEPROM_EESK);
               DELAY(40);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
       }
       DELAY(40);
}

/*
* Figure out EEPROM size.
*
* 559's can have either 64-word or 256-word EEPROMs, the 558
* datasheet only talks about 64-word EEPROMs, and the 557 datasheet
* talks about the existence of 16 to 256 word EEPROMs.
*
* The only known sizes are 64 and 256, where the 256 version is used
* by CardBus cards to store CIS information.
*
* The address is shifted in msb-to-lsb, and after the last
* address-bit the EEPROM is supposed to output a `dummy zero' bit,
* after which follows the actual data. We try to detect this zero, by
* probing the data-out bit in the EEPROM control register just after
* having shifted in a bit. If the bit is zero, we assume we've
* shifted enough address bits. The data-out should be tri-state,
* before this, which should translate to a logical one.
*
* Other ways to do this would be to try to read a register with known
* contents with a varying number of address bits, but no such
* register seem to be available. The high bits of register 10 are 01
* on the 558 and 559, but apparently not on the 557.
*
* The Linux driver computes a checksum on the EEPROM data, but the
* value of this checksum is not very well documented.
*/

void
fxp_autosize_eeprom(struct fxp_softc *sc)
{
       int x;

       CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
       DELAY(40);

       /* Shift in read opcode. */
       fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_READ, 3);

       /*
        * Shift in address, wait for the dummy zero following a correct
        * address shift.
        */
       for (x = 1; x <= 8; x++) {
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               DELAY(40);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL,
                   FXP_EEPROM_EECS | FXP_EEPROM_EESK);
               DELAY(40);
               if ((CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) &
                   FXP_EEPROM_EEDO) == 0)
                       break;
               DELAY(40);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               DELAY(40);
       }
       CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
       DELAY(40);
       if (x != 6 && x != 8) {
#ifdef DEBUG
               printf("%s: strange EEPROM size (%d)\n",
                   device_xname(sc->sc_dev), 1 << x);
#endif
       } else
               sc->sc_eeprom_size = x;
}

/*
* Read from the serial EEPROM. Basically, you manually shift in
* the read opcode (one bit at a time) and then shift in the address,
* and then you shift out the data (all of this one bit at a time).
* The word size is 16 bits, so you have to provide the address for
* every 16 bits of data.
*/
void
fxp_read_eeprom(struct fxp_softc *sc, uint16_t *data, int offset, int words)
{
       uint16_t reg;
       int i, x;

       for (i = 0; i < words; i++) {
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);

               /* Shift in read opcode. */
               fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_READ, 3);

               /* Shift in address. */
               fxp_eeprom_shiftin(sc, i + offset, sc->sc_eeprom_size);

               reg = FXP_EEPROM_EECS;
               data[i] = 0;

               /* Shift out data. */
               for (x = 16; x > 0; x--) {
                       CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL,
                           reg | FXP_EEPROM_EESK);
                       DELAY(40);
                       if (CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) &
                           FXP_EEPROM_EEDO)
                               data[i] |= (1 << (x - 1));
                       CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
                       DELAY(40);
               }
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
               DELAY(40);
       }
}

/*
* Write data to the serial EEPROM.
*/
void
fxp_write_eeprom(struct fxp_softc *sc, uint16_t *data, int offset, int words)
{
       int i, j;

       for (i = 0; i < words; i++) {
               /* Erase/write enable. */
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_ERASE, 3);
               fxp_eeprom_shiftin(sc, 0x3 << (sc->sc_eeprom_size - 2),
                   sc->sc_eeprom_size);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
               DELAY(4);

               /* Shift in write opcode, address, data. */
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_WRITE, 3);
               fxp_eeprom_shiftin(sc, i + offset, sc->sc_eeprom_size);
               fxp_eeprom_shiftin(sc, data[i], 16);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
               DELAY(4);

               /* Wait for the EEPROM to finish up. */
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               DELAY(4);
               for (j = 0; j < 1000; j++) {
                       if (CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) &
                           FXP_EEPROM_EEDO)
                               break;
                       DELAY(50);
               }
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
               DELAY(4);

               /* Erase/write disable. */
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
               fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_ERASE, 3);
               fxp_eeprom_shiftin(sc, 0, sc->sc_eeprom_size);
               CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
               DELAY(4);
       }
}

/*
* Update the checksum of the EEPROM.
*/
void
fxp_eeprom_update_cksum(struct fxp_softc *sc)
{
       int i;
       uint16_t data, cksum;

       cksum = 0;
       for (i = 0; i < (1 << sc->sc_eeprom_size) - 1; i++) {
               fxp_read_eeprom(sc, &data, i, 1);
               cksum += data;
       }
       i = (1 << sc->sc_eeprom_size) - 1;
       cksum = 0xbaba - cksum;
       fxp_read_eeprom(sc, &data, i, 1);
       fxp_write_eeprom(sc, &cksum, i, 1);
       log(LOG_INFO, "%s: EEPROM checksum @ 0x%x: 0x%04x -> 0x%04x\n",
           device_xname(sc->sc_dev), i, data, cksum);
}

/*
* Start packet transmission on the interface.
*/
void
fxp_start(struct ifnet *ifp)
{
       struct fxp_softc *sc = ifp->if_softc;
       struct mbuf *m0, *m;
       struct fxp_txdesc *txd;
       struct fxp_txsoft *txs;
       bus_dmamap_t dmamap;
       int error, lasttx, nexttx, opending, seg, nsegs, len;

       /*
        * If we want a re-init, bail out now.
        */
       if (sc->sc_flags & FXPF_WANTINIT) {
               ifp->if_flags |= IFF_OACTIVE;
               return;
       }

       if ((ifp->if_flags & (IFF_RUNNING | IFF_OACTIVE)) != IFF_RUNNING)
               return;

       /*
        * Remember the previous txpending and the current lasttx.
        */
       opending = sc->sc_txpending;
       lasttx = sc->sc_txlast;

       /*
        * Loop through the send queue, setting up transmit descriptors
        * until we drain the queue, or use up all available transmit
        * descriptors.
        */
       for (;;) {
               struct fxp_tbd *tbdp;
               int csum_flags;

               /*
                * Grab a packet off the queue.
                */
               IFQ_POLL(&ifp->if_snd, m0);
               if (m0 == NULL)
                       break;
               m = NULL;

               if (sc->sc_txpending == FXP_NTXCB - 1) {
                       FXP_EVCNT_INCR(&sc->sc_ev_txstall);
                       break;
               }

               /*
                * Get the next available transmit descriptor.
                */
               nexttx = FXP_NEXTTX(sc->sc_txlast);
               txd = FXP_CDTX(sc, nexttx);
               txs = FXP_DSTX(sc, nexttx);
               dmamap = txs->txs_dmamap;

               /*
                * Load the DMA map.  If this fails, the packet either
                * didn't fit in the allotted number of frags, or we were
                * short on resources.  In this case, we'll copy and try
                * again.
                */
               if (bus_dmamap_load_mbuf(sc->sc_dmat, dmamap, m0,
                   BUS_DMA_WRITE | BUS_DMA_NOWAIT) != 0) {
                       MGETHDR(m, M_DONTWAIT, MT_DATA);
                       if (m == NULL) {
                               log(LOG_ERR, "%s: unable to allocate Tx mbuf\n",
                                   device_xname(sc->sc_dev));
                               break;
                       }
                       MCLAIM(m, &sc->sc_ethercom.ec_tx_mowner);
                       if (m0->m_pkthdr.len > MHLEN) {
                               MCLGET(m, M_DONTWAIT);
                               if ((m->m_flags & M_EXT) == 0) {
                                       log(LOG_ERR, "%s: unable to allocate "
                                           "Tx cluster\n",
                                           device_xname(sc->sc_dev));
                                       m_freem(m);
                                       break;
                               }
                       }
                       m_copydata(m0, 0, m0->m_pkthdr.len, mtod(m, void *));
                       m->m_pkthdr.len = m->m_len = m0->m_pkthdr.len;
                       error = bus_dmamap_load_mbuf(sc->sc_dmat, dmamap,
                           m, BUS_DMA_WRITE | BUS_DMA_NOWAIT);
                       if (error) {
                               log(LOG_ERR, "%s: unable to load Tx buffer, "
                                   "error = %d\n",
                                   device_xname(sc->sc_dev), error);
                               break;
                       }
               }

               IFQ_DEQUEUE(&ifp->if_snd, m0);
               csum_flags = m0->m_pkthdr.csum_flags;
               if (m != NULL) {
                       m_freem(m0);
                       m0 = m;
               }

               /* Initialize the fraglist. */
               tbdp = txd->txd_tbd;
               len = m0->m_pkthdr.len;
               nsegs = dmamap->dm_nsegs;
               if (sc->sc_flags & FXPF_EXT_RFA)
                       tbdp++;
               for (seg = 0; seg < nsegs; seg++) {
                       tbdp[seg].tb_addr =
                           htole32(dmamap->dm_segs[seg].ds_addr);
                       tbdp[seg].tb_size =
                           htole32(dmamap->dm_segs[seg].ds_len);
               }
               if (__predict_false(len <= FXP_IP4CSUMTX_PADLEN &&
                   (csum_flags & M_CSUM_IPv4) != 0)) {
                       /*
                        * Pad short packets to avoid ip4csum-tx bug.
                        *
                        * XXX Should we still consider if such short
                        *     (36 bytes or less) packets might already
                        *     occupy FXP_IPCB_NTXSEG (15) fragments here?
                        */
                       KASSERT(nsegs < FXP_IPCB_NTXSEG);
                       nsegs++;
                       tbdp[seg].tb_addr = htole32(FXP_CDTXPADADDR(sc));
                       tbdp[seg].tb_size =
                           htole32(FXP_IP4CSUMTX_PADLEN + 1 - len);
               }

               /* Sync the DMA map. */
               bus_dmamap_sync(sc->sc_dmat, dmamap, 0, dmamap->dm_mapsize,
                   BUS_DMASYNC_PREWRITE);

               /*
                * Store a pointer to the packet so we can free it later.
                */
               txs->txs_mbuf = m0;

               /*
                * Initialize the transmit descriptor.
                */
               /* BIG_ENDIAN: no need to swap to store 0 */
               txd->txd_txcb.cb_status = 0;
               txd->txd_txcb.cb_command =
                   sc->sc_txcmd | htole16(FXP_CB_COMMAND_SF);
               txd->txd_txcb.tx_threshold = tx_threshold;
               txd->txd_txcb.tbd_number = nsegs;

               KASSERT((csum_flags & (M_CSUM_TCPv6 | M_CSUM_UDPv6)) == 0);
               if (sc->sc_flags & FXPF_EXT_RFA) {
                       struct fxp_ipcb *ipcb;
                       /*
                        * Deal with TCP/IP checksum offload. Note that
                        * in order for TCP checksum offload to work,
                        * the pseudo header checksum must have already
                        * been computed and stored in the checksum field
                        * in the TCP header. The stack should have
                        * already done this for us.
                        */
                       ipcb = &txd->txd_u.txdu_ipcb;
                       memset(ipcb, 0, sizeof(*ipcb));
                       /*
                        * always do hardware parsing.
                        */
                       ipcb->ipcb_ip_activation_high =
                           FXP_IPCB_HARDWAREPARSING_ENABLE;
                       /*
                        * ip checksum offloading.
                        */
                       if (csum_flags & M_CSUM_IPv4) {
                               ipcb->ipcb_ip_schedule |=
                                   FXP_IPCB_IP_CHECKSUM_ENABLE;
                       }
                       /*
                        * TCP/UDP checksum offloading.
                        */
                       if (csum_flags & (M_CSUM_TCPv4 | M_CSUM_UDPv4)) {
                               ipcb->ipcb_ip_schedule |=
                                   FXP_IPCB_TCPUDP_CHECKSUM_ENABLE;
                       }

                       /*
                        * request VLAN tag insertion if needed.
                        */
                       if (vlan_has_tag(m0)) {
                               ipcb->ipcb_vlan_id = htobe16(vlan_get_tag(m0));
                               ipcb->ipcb_ip_activation_high |=
                                   FXP_IPCB_INSERTVLAN_ENABLE;
                       }
               } else {
                       KASSERT((csum_flags &
                           (M_CSUM_IPv4 | M_CSUM_TCPv4 | M_CSUM_UDPv4)) == 0);
               }

               FXP_CDTXSYNC(sc, nexttx,
                   BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

               /* Advance the tx pointer. */
               sc->sc_txpending++;
               sc->sc_txlast = nexttx;

               /*
                * Pass packet to bpf if there is a listener.
                */
               bpf_mtap(ifp, m0, BPF_D_OUT);
       }

       if (sc->sc_txpending == FXP_NTXCB - 1) {
               /* No more slots; notify upper layer. */
               ifp->if_flags |= IFF_OACTIVE;
       }

       if (sc->sc_txpending != opending) {
               /*
                * We enqueued packets.  If the transmitter was idle,
                * reset the txdirty pointer.
                */
               if (opending == 0)
                       sc->sc_txdirty = FXP_NEXTTX(lasttx);

               /*
                * Cause the chip to interrupt and suspend command
                * processing once the last packet we've enqueued
                * has been transmitted.
                *
                * To avoid a race between updating status bits
                * by the fxp chip and clearing command bits
                * by this function on machines which don't have
                * atomic methods to clear/set bits in memory
                * smaller than 32bits (both cb_status and cb_command
                * members are uint16_t and in the same 32bit word),
                * we have to prepare a dummy TX descriptor which has
                * NOP command and just causes a TX completion interrupt.
                */
               sc->sc_txpending++;
               sc->sc_txlast = FXP_NEXTTX(sc->sc_txlast);
               txd = FXP_CDTX(sc, sc->sc_txlast);
               /* BIG_ENDIAN: no need to swap to store 0 */
               txd->txd_txcb.cb_status = 0;
               txd->txd_txcb.cb_command = htole16(FXP_CB_COMMAND_NOP |
                   FXP_CB_COMMAND_I | FXP_CB_COMMAND_S);
               FXP_CDTXSYNC(sc, sc->sc_txlast,
                   BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

               /*
                * The entire packet chain is set up.  Clear the suspend bit
                * on the command prior to the first packet we set up.
                */
               FXP_CDTXSYNC(sc, lasttx,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
               FXP_CDTX(sc, lasttx)->txd_txcb.cb_command &=
                   htole16(~FXP_CB_COMMAND_S);
               FXP_CDTXSYNC(sc, lasttx,
                   BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

               /*
                * Issue a Resume command in case the chip was suspended.
                */
               fxp_scb_wait(sc);
               fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_RESUME);

               /* Set a watchdog timer in case the chip flakes out. */
               ifp->if_timer = 5;
       }
}

/*
* Process interface interrupts.
*/
int
fxp_intr(void *arg)
{
       struct fxp_softc *sc = arg;
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       bus_dmamap_t rxmap;
       int claimed = 0, rnr;
       uint8_t statack, rndstat = 0;

       if (!device_is_active(sc->sc_dev) || sc->sc_enabled == 0)
               return (0);
       /*
        * If the interface isn't running, don't try to
        * service the interrupt.. just ack it and bail.
        */
       if ((ifp->if_flags & IFF_RUNNING) == 0) {
               statack = CSR_READ_1(sc, FXP_CSR_SCB_STATACK);
               if (statack) {
                       claimed = 1;
                       CSR_WRITE_1(sc, FXP_CSR_SCB_STATACK, statack);
               }
               return (claimed);
       }

       while ((statack = CSR_READ_1(sc, FXP_CSR_SCB_STATACK)) != 0) {
               rndstat = statack;
               claimed = 1;

               /*
                * First ACK all the interrupts in this pass.
                */
               CSR_WRITE_1(sc, FXP_CSR_SCB_STATACK, statack);

               /*
                * Process receiver interrupts. If a no-resource (RNR)
                * condition exists, get whatever packets we can and
                * re-start the receiver.
                */
               rnr = (statack & (FXP_SCB_STATACK_RNR | FXP_SCB_STATACK_SWI)) ?
                   1 : 0;
               if (statack & (FXP_SCB_STATACK_FR | FXP_SCB_STATACK_RNR |
                   FXP_SCB_STATACK_SWI)) {
                       FXP_EVCNT_INCR(&sc->sc_ev_rxintr);
                       rnr |= fxp_rxintr(sc);
               }

               /*
                * Free any finished transmit mbuf chains.
                */
               if (statack & (FXP_SCB_STATACK_CXTNO | FXP_SCB_STATACK_CNA)) {
                       FXP_EVCNT_INCR(&sc->sc_ev_txintr);
                       fxp_txintr(sc);

                       /*
                        * Try to get more packets going.
                        */
                       if_schedule_deferred_start(ifp);

                       if (sc->sc_txpending == 0) {
                               /*
                                * Tell them that they can re-init now.
                                */
                               if (sc->sc_flags & FXPF_WANTINIT)
                                       wakeup(sc);
                       }
               }

               if (rnr) {
                       fxp_scb_wait(sc);
                       fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_ABORT);
                       rxmap = M_GETCTX(sc->sc_rxq.ifq_head, bus_dmamap_t);
                       fxp_scb_wait(sc);
                       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL,
                           rxmap->dm_segs[0].ds_addr +
                           RFA_ALIGNMENT_FUDGE);
                       fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_START);
               }
       }

       if (claimed)
               rnd_add_uint32(&sc->rnd_source, rndstat);
       return (claimed);
}

/*
* Handle transmit completion interrupts.
*/
void
fxp_txintr(struct fxp_softc *sc)
{
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       struct fxp_txdesc *txd;
       struct fxp_txsoft *txs;
       int i;
       uint16_t txstat;

       ifp->if_flags &= ~IFF_OACTIVE;
       for (i = sc->sc_txdirty; sc->sc_txpending != 0;
           i = FXP_NEXTTX(i), sc->sc_txpending--) {
               txd = FXP_CDTX(sc, i);
               txs = FXP_DSTX(sc, i);

               FXP_CDTXSYNC(sc, i,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);

               /* skip dummy NOP TX descriptor */
               if ((le16toh(txd->txd_txcb.cb_command) & FXP_CB_COMMAND_CMD)
                   == FXP_CB_COMMAND_NOP)
                       continue;

               txstat = le16toh(txd->txd_txcb.cb_status);

               if ((txstat & FXP_CB_STATUS_C) == 0)
                       break;

               bus_dmamap_sync(sc->sc_dmat, txs->txs_dmamap,
                   0, txs->txs_dmamap->dm_mapsize,
                   BUS_DMASYNC_POSTWRITE);
               bus_dmamap_unload(sc->sc_dmat, txs->txs_dmamap);
               m_freem(txs->txs_mbuf);
               txs->txs_mbuf = NULL;
       }

       /* Update the dirty transmit buffer pointer. */
       sc->sc_txdirty = i;

       /*
        * Cancel the watchdog timer if there are no pending
        * transmissions.
        */
       if (sc->sc_txpending == 0)
               ifp->if_timer = 0;
}

/*
* fxp_rx_hwcksum: check status of H/W offloading for received packets.
*/

void
fxp_rx_hwcksum(struct fxp_softc *sc, struct mbuf *m, const struct fxp_rfa *rfa,
   u_int len)
{
       uint32_t csum_data;
       int csum_flags;

       /*
        * check H/W Checksumming.
        */

       csum_flags = 0;
       csum_data = 0;

       if ((sc->sc_flags & FXPF_EXT_RFA) != 0) {
               uint8_t csum_stat;

               csum_stat = rfa->cksum_stat;
               if ((rfa->rfa_status & htole16(FXP_RFA_STATUS_PARSE)) == 0)
                       goto out;

               if (csum_stat & FXP_RFDX_CS_IP_CSUM_BIT_VALID) {
                       csum_flags = M_CSUM_IPv4;
                       if ((csum_stat & FXP_RFDX_CS_IP_CSUM_VALID) == 0)
                               csum_flags |= M_CSUM_IPv4_BAD;
               }

               if (csum_stat & FXP_RFDX_CS_TCPUDP_CSUM_BIT_VALID) {
                       csum_flags |= (M_CSUM_TCPv4 | M_CSUM_UDPv4); /* XXX */
                       if ((csum_stat & FXP_RFDX_CS_TCPUDP_CSUM_VALID) == 0)
                               csum_flags |= M_CSUM_TCP_UDP_BAD;
               }

       } else if ((sc->sc_flags & FXPF_82559_RXCSUM) != 0) {
               struct ifnet *ifp = &sc->sc_ethercom.ec_if;
               struct ether_header *eh;
               struct ip *ip;
               struct udphdr *uh;
               u_int hlen, pktlen;

               if (len < ETHER_HDR_LEN + sizeof(struct ip))
                       goto out;
               pktlen = len - ETHER_HDR_LEN;
               eh = mtod(m, struct ether_header *);
               if (ntohs(eh->ether_type) != ETHERTYPE_IP)
                       goto out;
               ip = (struct ip *)((uint8_t *)eh + ETHER_HDR_LEN);
               if (ip->ip_v != IPVERSION)
                       goto out;

               hlen = ip->ip_hl << 2;
               if (hlen < sizeof(struct ip))
                       goto out;

               /*
                * Bail if too short, has random trailing garbage, truncated,
                * fragment, or has ethernet pad.
                */
               if (ntohs(ip->ip_len) < hlen ||
                   ntohs(ip->ip_len) != pktlen ||
                   (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) != 0)
                       goto out;

               switch (ip->ip_p) {
               case IPPROTO_TCP:
                       if ((ifp->if_csum_flags_rx & M_CSUM_TCPv4) == 0 ||
                           pktlen < (hlen + sizeof(struct tcphdr)))
                               goto out;
                       csum_flags =
                           M_CSUM_TCPv4 | M_CSUM_DATA | M_CSUM_NO_PSEUDOHDR;
                       break;
               case IPPROTO_UDP:
                       if ((ifp->if_csum_flags_rx & M_CSUM_UDPv4) == 0 ||
                           pktlen < (hlen + sizeof(struct udphdr)))
                               goto out;
                       uh = (struct udphdr *)((uint8_t *)ip + hlen);
                       if (uh->uh_sum == 0)
                               goto out;       /* no checksum */
                       csum_flags =
                           M_CSUM_UDPv4 | M_CSUM_DATA | M_CSUM_NO_PSEUDOHDR;
                       break;
               default:
                       goto out;
               }

               /* Extract computed checksum. */
               csum_data = be16dec(mtod(m, uint8_t *) + len);

               /*
                * The computed checksum includes IP headers,
                * so we have to deduct them.
                */
#if 0
               /*
                * But in TCP/UDP layer we can assume the IP header is valid,
                * i.e. a sum of the whole IP header should be 0xffff,
                * so we don't have to bother to deduct it.
                */
               if (hlen > 0) {
                       uint32_t hsum;
                       const uint16_t *iphdr;
                       hsum = 0;
                       iphdr = (uint16_t *)ip;

                       while (hlen > 1) {
                               hsum += ntohs(*iphdr++);
                               hlen -= sizeof(uint16_t);
                       }
                       while (hsum >> 16)
                               hsum = (hsum >> 16) + (hsum & 0xffff);

                       csum_data += (uint16_t)~hsum;

                       while (csum_data >> 16)
                               csum_data =
                                   (csum_data >> 16) + (csum_data & 0xffff);
               }
#endif
       }
out:
       m->m_pkthdr.csum_flags = csum_flags;
       m->m_pkthdr.csum_data = csum_data;
}

/*
* Handle receive interrupts.
*/
int
fxp_rxintr(struct fxp_softc *sc)
{
       struct ethercom *ec = &sc->sc_ethercom;
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       struct mbuf *m, *m0;
       bus_dmamap_t rxmap;
       struct fxp_rfa *rfa;
       int rnr;
       uint16_t len, rxstat;

       rnr = 0;

       for (;;) {
               m = sc->sc_rxq.ifq_head;
               rfa = FXP_MTORFA(m);
               rxmap = M_GETCTX(m, bus_dmamap_t);

               FXP_RFASYNC(sc, m,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);

               rxstat = le16toh(rfa->rfa_status);

               if ((rxstat & FXP_RFA_STATUS_RNR) != 0)
                       rnr = 1;

               if ((rxstat & FXP_RFA_STATUS_C) == 0) {
                       /*
                        * We have processed all of the
                        * receive buffers.
                        */
                       FXP_RFASYNC(sc, m, BUS_DMASYNC_PREREAD);
                       return rnr;
               }

               IF_DEQUEUE(&sc->sc_rxq, m);

               FXP_RXBUFSYNC(sc, m, BUS_DMASYNC_POSTREAD);

               len = le16toh(rfa->actual_size) &
                   (m->m_ext.ext_size - 1);
               if ((sc->sc_flags & FXPF_82559_RXCSUM) != 0) {
                       /* Adjust for appended checksum bytes. */
                       len -= sizeof(uint16_t);
               }

               if (len < sizeof(struct ether_header)) {
                       /*
                        * Runt packet; drop it now.
                        */
                       FXP_INIT_RFABUF(sc, m);
                       continue;
               }

               /*
                * If support for 802.1Q VLAN sized frames is
                * enabled, we need to do some additional error
                * checking (as we are saving bad frames, in
                * order to receive the larger ones).
                */
               if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) != 0 &&
                   (rxstat & (FXP_RFA_STATUS_OVERRUN |
                              FXP_RFA_STATUS_RNR |
                              FXP_RFA_STATUS_ALIGN |
                              FXP_RFA_STATUS_CRC)) != 0) {
                       FXP_INIT_RFABUF(sc, m);
                       continue;
               }

               /*
                * check VLAN tag stripping.
                */
               if ((sc->sc_flags & FXPF_EXT_RFA) != 0 &&
                   (rfa->rfa_status & htole16(FXP_RFA_STATUS_VLAN)) != 0)
                       vlan_set_tag(m, be16toh(rfa->vlan_id));

               /* Do checksum checking. */
               if ((ifp->if_csum_flags_rx &
                   (M_CSUM_TCPv4 | M_CSUM_UDPv4)) != 0)
                       fxp_rx_hwcksum(sc, m, rfa, len);

               /*
                * If the packet is small enough to fit in a
                * single header mbuf, allocate one and copy
                * the data into it.  This greatly reduces
                * memory consumption when we receive lots
                * of small packets.
                *
                * Otherwise, we add a new buffer to the receive
                * chain.  If this fails, we drop the packet and
                * recycle the old buffer.
                */
               if (fxp_copy_small != 0 && len <= MHLEN) {
                       MGETHDR(m0, M_DONTWAIT, MT_DATA);
                       if (m0 == NULL)
                               goto dropit;
                       MCLAIM(m0, &sc->sc_ethercom.ec_rx_mowner);
                       memcpy(mtod(m0, void *),
                           mtod(m, void *), len);
                       m0->m_pkthdr.csum_flags = m->m_pkthdr.csum_flags;
                       m0->m_pkthdr.csum_data = m->m_pkthdr.csum_data;
                       FXP_INIT_RFABUF(sc, m);
                       m = m0;
               } else {
                       if (fxp_add_rfabuf(sc, rxmap, 1) != 0) {
dropit:
                               if_statinc(ifp, if_ierrors);
                               FXP_INIT_RFABUF(sc, m);
                               continue;
                       }
               }

               m_set_rcvif(m, ifp);
               m->m_pkthdr.len = m->m_len = len;

               /* Pass it on. */
               if_percpuq_enqueue(ifp->if_percpuq, m);
       }
}

/*
* Update packet in/out/collision statistics. The i82557 doesn't
* allow you to access these counters without doing a fairly
* expensive DMA to get _all_ of the statistics it maintains, so
* we do this operation here only once per second. The statistics
* counters in the kernel are updated from the previous dump-stats
* DMA and then a new dump-stats DMA is started. The on-chip
* counters are zeroed when the DMA completes. If we can't start
* the DMA immediately, we don't wait - we just prepare to read
* them again next time.
*/
void
fxp_tick(void *arg)
{
       struct fxp_softc *sc = arg;
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       struct fxp_stats *sp = &sc->sc_control_data->fcd_stats;
       int s;

       if (!device_is_active(sc->sc_dev))
               return;

       s = splnet();

       net_stat_ref_t nsr = IF_STAT_GETREF(ifp);

       FXP_CDSTATSSYNC(sc, BUS_DMASYNC_POSTREAD);

       if_statadd_ref(ifp, nsr, if_opackets, le32toh(sp->tx_good));
       if_statadd_ref(ifp, nsr, if_collisions,
           le32toh(sp->tx_total_collisions));
       if (sp->rx_good) {
               sc->sc_rxidle = 0;
       } else if (sc->sc_flags & FXPF_RECV_WORKAROUND) {
               sc->sc_rxidle++;
       }
       if_statadd_ref(ifp, nsr, if_ierrors,
           le32toh(sp->rx_crc_errors) +
           le32toh(sp->rx_alignment_errors) +
           le32toh(sp->rx_rnr_errors) +
           le32toh(sp->rx_overrun_errors));
       /*
        * If any transmit underruns occurred, bump up the transmit
        * threshold by another 512 bytes (64 * 8).
        */
       if (sp->tx_underruns) {
               if_statadd_ref(ifp, nsr, if_oerrors,
                   le32toh(sp->tx_underruns));
               if (tx_threshold < 192)
                       tx_threshold += 64;
       }
#ifdef FXP_EVENT_COUNTERS
       if (sc->sc_flags & FXPF_FC) {
               sc->sc_ev_txpause.ev_count += sp->tx_pauseframes;
               sc->sc_ev_rxpause.ev_count += sp->rx_pauseframes;
       }
#endif

       IF_STAT_PUTREF(ifp);

       /*
        * If we haven't received any packets in FXP_MAX_RX_IDLE seconds,
        * then assume the receiver has locked up and attempt to clear
        * the condition by reprogramming the multicast filter (actually,
        * resetting the interface). This is a work-around for a bug in
        * the 82557 where the receiver locks up if it gets certain types
        * of garbage in the synchronization bits prior to the packet header.
        * This bug is supposed to only occur in 10Mbps mode, but has been
        * seen to occur in 100Mbps mode as well (perhaps due to a 10/100
        * speed transition).
        */
       if (sc->sc_rxidle > FXP_MAX_RX_IDLE) {
               (void) fxp_init(ifp);
               splx(s);
               return;
       }
       /*
        * If there is no pending command, start another stats
        * dump. Otherwise punt for now.
        */
       if (CSR_READ_1(sc, FXP_CSR_SCB_COMMAND) == 0) {
               /*
                * Start another stats dump.
                */
               FXP_CDSTATSSYNC(sc, BUS_DMASYNC_PREREAD);
               fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_DUMPRESET);
       } else {
               /*
                * A previous command is still waiting to be accepted.
                * Just zero our copy of the stats and wait for the
                * next timer event to update them.
                */
               /* BIG_ENDIAN: no swap required to store 0 */
               sp->tx_good = 0;
               sp->tx_underruns = 0;
               sp->tx_total_collisions = 0;

               sp->rx_good = 0;
               sp->rx_crc_errors = 0;
               sp->rx_alignment_errors = 0;
               sp->rx_rnr_errors = 0;
               sp->rx_overrun_errors = 0;
               if (sc->sc_flags & FXPF_FC) {
                       sp->tx_pauseframes = 0;
                       sp->rx_pauseframes = 0;
               }
       }

       if (sc->sc_flags & FXPF_MII) {
               /* Tick the MII clock. */
               mii_tick(&sc->sc_mii);
       }

       splx(s);

       /*
        * Schedule another timeout one second from now.
        */
       callout_schedule(&sc->sc_callout, hz);
}

/*
* Drain the receive queue.
*/
void
fxp_rxdrain(struct fxp_softc *sc)
{
       bus_dmamap_t rxmap;
       struct mbuf *m;

       for (;;) {
               IF_DEQUEUE(&sc->sc_rxq, m);
               if (m == NULL)
                       break;
               rxmap = M_GETCTX(m, bus_dmamap_t);
               bus_dmamap_unload(sc->sc_dmat, rxmap);
               FXP_RXMAP_PUT(sc, rxmap);
               m_freem(m);
       }
}

/*
* Stop the interface. Cancels the statistics updater and resets
* the interface.
*/
void
fxp_stop(struct ifnet *ifp, int disable)
{
       struct fxp_softc *sc = ifp->if_softc;
       struct fxp_txsoft *txs;
       int i;

       /*
        * Turn down interface (done early to avoid bad interactions
        * between panics, shutdown hooks, and the watchdog timer)
        */
       ifp->if_timer = 0;
       ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);

       /*
        * Cancel stats updater.
        */
       callout_stop(&sc->sc_callout);
       if (sc->sc_flags & FXPF_MII) {
               /* Down the MII. */
               mii_down(&sc->sc_mii);
       }

       /*
        * Issue software reset.  This unloads any microcode that
        * might already be loaded.
        */
       sc->sc_flags &= ~FXPF_UCODE_LOADED;
       CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SOFTWARE_RESET);
       DELAY(50);

       /*
        * Release any xmit buffers.
        */
       for (i = 0; i < FXP_NTXCB; i++) {
               txs = FXP_DSTX(sc, i);
               if (txs->txs_mbuf != NULL) {
                       bus_dmamap_unload(sc->sc_dmat, txs->txs_dmamap);
                       m_freem(txs->txs_mbuf);
                       txs->txs_mbuf = NULL;
               }
       }
       sc->sc_txpending = 0;

       if (disable) {
               fxp_rxdrain(sc);
               fxp_disable(sc);
       }

}

/*
* Watchdog/transmission transmit timeout handler. Called when a
* transmission is started on the interface, but no interrupt is
* received before the timeout. This usually indicates that the
* card has wedged for some reason.
*/
void
fxp_watchdog(struct ifnet *ifp)
{
       struct fxp_softc *sc = ifp->if_softc;

       log(LOG_ERR, "%s: device timeout\n", device_xname(sc->sc_dev));
       if_statinc(ifp, if_oerrors);

       (void) fxp_init(ifp);
}

/*
* Initialize the interface.  Must be called at splnet().
*/
int
fxp_init(struct ifnet *ifp)
{
       struct fxp_softc *sc = ifp->if_softc;
       struct fxp_cb_config *cbp;
       struct fxp_cb_ias *cb_ias;
       struct fxp_txdesc *txd;
       bus_dmamap_t rxmap;
       int i, prm, save_bf, lrxen, vlan_drop, allm, error = 0;
       uint16_t status;

       if ((error = fxp_enable(sc)) != 0)
               goto out;

       /*
        * Cancel any pending I/O
        */
       fxp_stop(ifp, 0);

       /*
        * XXX just setting sc_flags to 0 here clears any FXPF_MII
        * flag, and this prevents the MII from detaching resulting in
        * a panic. The flags field should perhaps be split in runtime
        * flags and more static information. For now, just clear the
        * only other flag set.
        */

       sc->sc_flags &= ~FXPF_WANTINIT;

       /*
        * Initialize base of CBL and RFA memory. Loading with zero
        * sets it up for regular linear addressing.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, 0);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_BASE);

       fxp_scb_wait(sc);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_BASE);

       /*
        * Initialize the multicast filter.  Do this now, since we might
        * have to setup the config block differently.
        */
       fxp_mc_setup(sc);

       prm = (ifp->if_flags & IFF_PROMISC) ? 1 : 0;
       allm = (ifp->if_flags & IFF_ALLMULTI) ? 1 : 0;

       /*
        * In order to support receiving 802.1Q VLAN frames, we have to
        * enable "save bad frames", since they are 4 bytes larger than
        * the normal Ethernet maximum frame length.  On i82558 and later,
        * we have a better mechanism for this.
        */
       save_bf = 0;
       lrxen = 0;
       vlan_drop = 0;
       if (sc->sc_ethercom.ec_capenable & ETHERCAP_VLAN_MTU) {
               if (sc->sc_rev < FXP_REV_82558_A4)
                       save_bf = 1;
               else
                       lrxen = 1;
               if (sc->sc_rev >= FXP_REV_82550)
                       vlan_drop = 1;
       }

       /*
        * Initialize base of dump-stats buffer.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL,
           sc->sc_cddma + FXP_CDSTATSOFF);
       FXP_CDSTATSSYNC(sc, BUS_DMASYNC_PREREAD);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_DUMP_ADR);

       cbp = &sc->sc_control_data->fcd_configcb;
       memset(cbp, 0, sizeof(struct fxp_cb_config));

       /*
        * Load microcode for this controller.
        */
       fxp_load_ucode(sc);

       if ((sc->sc_ethercom.ec_if.if_flags & IFF_LINK1))
               sc->sc_flags |= FXPF_RECV_WORKAROUND;
       else
               sc->sc_flags &= ~FXPF_RECV_WORKAROUND;

       /*
        * This copy is kind of disgusting, but there are a bunch of must be
        * zero and must be one bits in this structure and this is the easiest
        * way to initialize them all to proper values.
        */
       memcpy(cbp, fxp_cb_config_template, sizeof(fxp_cb_config_template));

       /* BIG_ENDIAN: no need to swap to store 0 */
       cbp->cb_status =        0;
       cbp->cb_command =       htole16(FXP_CB_COMMAND_CONFIG |
                                   FXP_CB_COMMAND_EL);
       /* BIG_ENDIAN: no need to swap to store 0xffffffff */
       cbp->link_addr =        0xffffffff; /* (no) next command */
                                       /* bytes in config block */
       cbp->byte_count =       (sc->sc_flags & FXPF_EXT_RFA) ?
                               FXP_EXT_CONFIG_LEN : FXP_CONFIG_LEN;
       cbp->rx_fifo_limit =    8;      /* rx fifo threshold (32 bytes) */
       cbp->tx_fifo_limit =    0;      /* tx fifo threshold (0 bytes) */
       cbp->adaptive_ifs =     0;      /* (no) adaptive interframe spacing */
       cbp->mwi_enable =       (sc->sc_flags & FXPF_MWI) ? 1 : 0;
       cbp->type_enable =      0;      /* actually reserved */
       cbp->read_align_en =    (sc->sc_flags & FXPF_READ_ALIGN) ? 1 : 0;
       cbp->end_wr_on_cl =     (sc->sc_flags & FXPF_WRITE_ALIGN) ? 1 : 0;
       cbp->rx_dma_bytecount = 0;      /* (no) rx DMA max */
       cbp->tx_dma_bytecount = 0;      /* (no) tx DMA max */
       cbp->dma_mbce =         0;      /* (disable) dma max counters */
       cbp->late_scb =         0;      /* (don't) defer SCB update */
       cbp->tno_int_or_tco_en =0;      /* (disable) tx not okay interrupt */
       cbp->ci_int =           1;      /* interrupt on CU idle */
       cbp->ext_txcb_dis =     (sc->sc_flags & FXPF_EXT_TXCB) ? 0 : 1;
       cbp->ext_stats_dis =    1;      /* disable extended counters */
       cbp->keep_overrun_rx =  0;      /* don't pass overrun frames to host */
       cbp->save_bf =          save_bf;/* save bad frames */
       cbp->disc_short_rx =    !prm;   /* discard short packets */
       cbp->underrun_retry =   1;      /* retry mode (1) on DMA underrun */
       cbp->ext_rfa =          (sc->sc_flags & FXPF_EXT_RFA) ? 1 : 0;
       cbp->two_frames =       0;      /* do not limit FIFO to 2 frames */
       cbp->dyn_tbd =          0;      /* (no) dynamic TBD mode */
                                       /* interface mode */
       cbp->mediatype =        (sc->sc_flags & FXPF_MII) ? 1 : 0;
       cbp->csma_dis =         0;      /* (don't) disable link */
       cbp->tcp_udp_cksum =    (sc->sc_flags & FXPF_82559_RXCSUM) ? 1 : 0;
                                       /* (don't) enable RX checksum */
       cbp->vlan_tco =         0;      /* (don't) enable vlan wakeup */
       cbp->link_wake_en =     0;      /* (don't) assert PME# on link change */
       cbp->arp_wake_en =      0;      /* (don't) assert PME# on arp */
       cbp->mc_wake_en =       0;      /* (don't) assert PME# on mcmatch */
       cbp->nsai =             1;      /* (don't) disable source addr insert */
       cbp->preamble_length =  2;      /* (7 byte) preamble */
       cbp->loopback =         0;      /* (don't) loopback */
       cbp->linear_priority =  0;      /* (normal CSMA/CD operation) */
       cbp->linear_pri_mode =  0;      /* (wait after xmit only) */
       cbp->interfrm_spacing = 6;      /* (96 bits of) interframe spacing */
       cbp->promiscuous =      prm;    /* promiscuous mode */
       cbp->bcast_disable =    0;      /* (don't) disable broadcasts */
       cbp->wait_after_win =   0;      /* (don't) enable modified backoff alg*/
       cbp->ignore_ul =        0;      /* consider U/L bit in IA matching */
       cbp->crc16_en =         0;      /* (don't) enable crc-16 algorithm */
       cbp->crscdt =           (sc->sc_flags & FXPF_MII) ? 0 : 1;
       cbp->stripping =        !prm;   /* truncate rx packet to byte count */
       cbp->padding =          1;      /* (do) pad short tx packets */
       cbp->rcv_crc_xfer =     0;      /* (don't) xfer CRC to host */
       cbp->long_rx_en =       lrxen;  /* long packet receive enable */
       cbp->ia_wake_en =       0;      /* (don't) wake up on address match */
       cbp->magic_pkt_dis =    0;      /* (don't) disable magic packet */
                                       /* must set wake_en in PMCSR also */
       cbp->force_fdx =        0;      /* (don't) force full duplex */
       cbp->fdx_pin_en =       1;      /* (enable) FDX# pin */
       cbp->multi_ia =         0;      /* (don't) accept multiple IAs */
       cbp->mc_all =           allm;   /* accept all multicasts */
       cbp->ext_rx_mode =      (sc->sc_flags & FXPF_EXT_RFA) ? 1 : 0;
       cbp->vlan_drop_en =     vlan_drop;

       if (!(sc->sc_flags & FXPF_FC)) {
               /*
                * The i82557 has no hardware flow control, the values
                * here are the defaults for the chip.
                */
               cbp->fc_delay_lsb =     0;
               cbp->fc_delay_msb =     0x40;
               cbp->pri_fc_thresh =    3;
               cbp->tx_fc_dis =        0;
               cbp->rx_fc_restop =     0;
               cbp->rx_fc_restart =    0;
               cbp->fc_filter =        0;
               cbp->pri_fc_loc =       1;
       } else {
               cbp->fc_delay_lsb =     0x1f;
               cbp->fc_delay_msb =     0x01;
               cbp->pri_fc_thresh =    3;
               cbp->tx_fc_dis =        0;      /* enable transmit FC */
               cbp->rx_fc_restop =     1;      /* enable FC restop frames */
               cbp->rx_fc_restart =    1;      /* enable FC restart frames */
               cbp->fc_filter =        !prm;   /* drop FC frames to host */
               cbp->pri_fc_loc =       1;      /* FC pri location (byte31) */
               cbp->ext_stats_dis =    0;      /* enable extended stats */
       }

       FXP_CDCONFIGSYNC(sc, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

       /*
        * Start the config command/DMA.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->sc_cddma + FXP_CDCONFIGOFF);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
       /* ...and wait for it to complete. */
       for (i = 1000; i > 0; i--) {
               FXP_CDCONFIGSYNC(sc,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
               status = le16toh(cbp->cb_status);
               FXP_CDCONFIGSYNC(sc, BUS_DMASYNC_PREREAD);
               if ((status & FXP_CB_STATUS_C) != 0)
                       break;
               DELAY(1);
       }
       if (i == 0) {
               log(LOG_WARNING, "%s: line %d: dmasync timeout\n",
                   device_xname(sc->sc_dev), __LINE__);
               return (ETIMEDOUT);
       }

       /*
        * Initialize the station address.
        */
       cb_ias = &sc->sc_control_data->fcd_iascb;
       /* BIG_ENDIAN: no need to swap to store 0 */
       cb_ias->cb_status = 0;
       cb_ias->cb_command = htole16(FXP_CB_COMMAND_IAS | FXP_CB_COMMAND_EL);
       /* BIG_ENDIAN: no need to swap to store 0xffffffff */
       cb_ias->link_addr = 0xffffffff;
       memcpy(cb_ias->macaddr, CLLADDR(ifp->if_sadl), ETHER_ADDR_LEN);

       FXP_CDIASSYNC(sc, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

       /*
        * Start the IAS (Individual Address Setup) command/DMA.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->sc_cddma + FXP_CDIASOFF);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
       /* ...and wait for it to complete. */
       for (i = 1000; i > 0; i--) {
               FXP_CDIASSYNC(sc,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
               status = le16toh(cb_ias->cb_status);
               FXP_CDIASSYNC(sc, BUS_DMASYNC_PREREAD);
               if ((status & FXP_CB_STATUS_C) != 0)
                       break;
               DELAY(1);
       }
       if (i == 0) {
               log(LOG_WARNING, "%s: line %d: dmasync timeout\n",
                   device_xname(sc->sc_dev), __LINE__);
               return (ETIMEDOUT);
       }

       /*
        * Initialize the transmit descriptor ring.  txlast is initialized
        * to the end of the list so that it will wrap around to the first
        * descriptor when the first packet is transmitted.
        */
       for (i = 0; i < FXP_NTXCB; i++) {
               txd = FXP_CDTX(sc, i);
               memset(txd, 0, sizeof(*txd));
               txd->txd_txcb.cb_command =
                   htole16(FXP_CB_COMMAND_NOP | FXP_CB_COMMAND_S);
               txd->txd_txcb.link_addr =
                   htole32(FXP_CDTXADDR(sc, FXP_NEXTTX(i)));
               if (sc->sc_flags & FXPF_EXT_TXCB)
                       txd->txd_txcb.tbd_array_addr =
                           htole32(FXP_CDTBDADDR(sc, i) +
                                   (2 * sizeof(struct fxp_tbd)));
               else
                       txd->txd_txcb.tbd_array_addr =
                           htole32(FXP_CDTBDADDR(sc, i));
               FXP_CDTXSYNC(sc, i,
                   BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
       }
       sc->sc_txpending = 0;
       sc->sc_txdirty = 0;
       sc->sc_txlast = FXP_NTXCB - 1;

       /*
        * Initialize the receive buffer list.
        */
       sc->sc_rxq.ifq_maxlen = FXP_NRFABUFS;
       while (sc->sc_rxq.ifq_len < FXP_NRFABUFS) {
               rxmap = FXP_RXMAP_GET(sc);
               if ((error = fxp_add_rfabuf(sc, rxmap, 0)) != 0) {
                       log(LOG_ERR, "%s: unable to allocate or map rx "
                           "buffer %d, error = %d\n",
                           device_xname(sc->sc_dev),
                           sc->sc_rxq.ifq_len, error);
                       /*
                        * XXX Should attempt to run with fewer receive
                        * XXX buffers instead of just failing.
                        */
                       FXP_RXMAP_PUT(sc, rxmap);
                       fxp_rxdrain(sc);
                       goto out;
               }
       }
       sc->sc_rxidle = 0;

       /*
        * Give the transmit ring to the chip.  We do this by pointing
        * the chip at the last descriptor (which is a NOP|SUSPEND), and
        * issuing a start command.  It will execute the NOP and then
        * suspend, pointing at the first descriptor.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, FXP_CDTXADDR(sc, sc->sc_txlast));
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);

       /*
        * Initialize receiver buffer area - RFA.
        */
#if 0   /* initialization will be done by FXP_SCB_INTRCNTL_REQUEST_SWI later */
       rxmap = M_GETCTX(sc->sc_rxq.ifq_head, bus_dmamap_t);
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL,
           rxmap->dm_segs[0].ds_addr + RFA_ALIGNMENT_FUDGE);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_START);
#endif

       if (sc->sc_flags & FXPF_MII) {
               /*
                * Set current media.
                */
               if ((error = mii_ifmedia_change(&sc->sc_mii)) != 0)
                       goto out;
       }

       /*
        * ...all done!
        */
       ifp->if_flags |= IFF_RUNNING;
       ifp->if_flags &= ~IFF_OACTIVE;

       /*
        * Request a software generated interrupt that will be used to
        * (re)start the RU processing.  If we direct the chip to start
        * receiving from the start of queue now, instead of letting the
        * interrupt handler first process all received packets, we run
        * the risk of having it overwrite mbuf clusters while they are
        * being processed or after they have been returned to the pool.
        */
       CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTRCNTL_REQUEST_SWI);

       /*
        * Start the one second timer.
        */
       callout_schedule(&sc->sc_callout, hz);

       /*
        * Attempt to start output on the interface.
        */
       fxp_start(ifp);

out:
       if (error) {
               ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
               ifp->if_timer = 0;
               log(LOG_ERR, "%s: interface not running\n",
                   device_xname(sc->sc_dev));
       }
       return (error);
}

/*
* Notify the world which media we're using.
*/
void
fxp_mii_mediastatus(struct ifnet *ifp, struct ifmediareq *ifmr)
{
       struct fxp_softc *sc = ifp->if_softc;

       if (sc->sc_enabled == 0) {
               ifmr->ifm_active = IFM_ETHER | IFM_NONE;
               ifmr->ifm_status = 0;
               return;
       }

       ether_mediastatus(ifp, ifmr);
}

int
fxp_80c24_mediachange(struct ifnet *ifp)
{

       /* Nothing to do here. */
       return (0);
}

void
fxp_80c24_mediastatus(struct ifnet *ifp, struct ifmediareq *ifmr)
{
       struct fxp_softc *sc = ifp->if_softc;

       /*
        * Media is currently-selected media.  We cannot determine
        * the link status.
        */
       ifmr->ifm_status = 0;
       ifmr->ifm_active = sc->sc_mii.mii_media.ifm_cur->ifm_media;
}

/*
* Add a buffer to the end of the RFA buffer list.
* Return 0 if successful, error code on failure.
*
* The RFA struct is stuck at the beginning of mbuf cluster and the
* data pointer is fixed up to point just past it.
*/
int
fxp_add_rfabuf(struct fxp_softc *sc, bus_dmamap_t rxmap, int unload)
{
       struct mbuf *m;
       int error;

       MGETHDR(m, M_DONTWAIT, MT_DATA);
       if (m == NULL)
               return (ENOBUFS);

       MCLAIM(m, &sc->sc_ethercom.ec_rx_mowner);
       MCLGET(m, M_DONTWAIT);
       if ((m->m_flags & M_EXT) == 0) {
               m_freem(m);
               return (ENOBUFS);
       }

       if (unload)
               bus_dmamap_unload(sc->sc_dmat, rxmap);

       M_SETCTX(m, rxmap);

       m->m_len = m->m_pkthdr.len = m->m_ext.ext_size;
       error = bus_dmamap_load_mbuf(sc->sc_dmat, rxmap, m,
           BUS_DMA_READ | BUS_DMA_NOWAIT);
       if (error) {
               /* XXX XXX XXX */
               aprint_error_dev(sc->sc_dev,
                   "can't load rx DMA map %d, error = %d\n",
                   sc->sc_rxq.ifq_len, error);
               panic("fxp_add_rfabuf");
       }

       FXP_INIT_RFABUF(sc, m);

       return (0);
}

int
fxp_mdi_read(device_t self, int phy, int reg, uint16_t *value)
{
       struct fxp_softc *sc = device_private(self);
       int count = 10000;
       uint32_t data;

       CSR_WRITE_4(sc, FXP_CSR_MDICONTROL,
           (FXP_MDI_READ << 26) | (reg << 16) | (phy << 21));

       while (((data = CSR_READ_4(sc, FXP_CSR_MDICONTROL)) &
           0x10000000) == 0 && count--)
               DELAY(10);

       if (count <= 0) {
               log(LOG_WARNING,
                   "%s: fxp_mdi_read: timed out\n", device_xname(self));
               return ETIMEDOUT;
       }

       *value = data & 0xffff;
       return 0;
}

void
fxp_statchg(struct ifnet *ifp)
{

       /* Nothing to do. */
}

int
fxp_mdi_write(device_t self, int phy, int reg, uint16_t value)
{
       struct fxp_softc *sc = device_private(self);
       int count = 10000;

       CSR_WRITE_4(sc, FXP_CSR_MDICONTROL,
           (FXP_MDI_WRITE << 26) | (reg << 16) | (phy << 21) | value);

       while ((CSR_READ_4(sc, FXP_CSR_MDICONTROL) & 0x10000000) == 0 &&
           count--)
               DELAY(10);

       if (count <= 0) {
               log(LOG_WARNING,
                   "%s: fxp_mdi_write: timed out\n", device_xname(self));
               return ETIMEDOUT;
       }

       return 0;
}

int
fxp_ioctl(struct ifnet *ifp, u_long cmd, void *data)
{
       struct fxp_softc *sc = ifp->if_softc;
       int s, error;

       s = splnet();

       switch (cmd) {
       default:
               if ((error = ether_ioctl(ifp, cmd, data)) != ENETRESET)
                       break;

               error = 0;

               if (cmd != SIOCADDMULTI && cmd != SIOCDELMULTI)
                       ;
               else if (ifp->if_flags & IFF_RUNNING) {
                       /*
                        * Multicast list has changed; set the
                        * hardware filter accordingly.
                        */
                       while (sc->sc_txpending) {
                               sc->sc_flags |= FXPF_WANTINIT;
                               tsleep(sc, PSOCK, "fxp_init", 0);
                       }
                       error = fxp_init(ifp);
               }
               break;
       }

       /* Try to get more packets going. */
       if (sc->sc_enabled)
               fxp_start(ifp);

       splx(s);
       return (error);
}

/*
* Program the multicast filter.
*
* This function must be called at splnet().
*/
void
fxp_mc_setup(struct fxp_softc *sc)
{
       struct fxp_cb_mcs *mcsp = &sc->sc_control_data->fcd_mcscb;
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       struct ethercom *ec = &sc->sc_ethercom;
       struct ether_multi *enm;
       struct ether_multistep step;
       int count, nmcasts;
       uint16_t status;

#ifdef DIAGNOSTIC
       if (sc->sc_txpending)
               panic("fxp_mc_setup: pending transmissions");
#endif


       if (ifp->if_flags & IFF_PROMISC) {
               ifp->if_flags |= IFF_ALLMULTI;
               return;
       } else {
               ifp->if_flags &= ~IFF_ALLMULTI;
       }

       /*
        * Initialize multicast setup descriptor.
        */
       nmcasts = 0;
       ETHER_LOCK(ec);
       ETHER_FIRST_MULTI(step, ec, enm);
       while (enm != NULL) {
               /*
                * Check for too many multicast addresses or if we're
                * listening to a range.  Either way, we simply have
                * to accept all multicasts.
                */
               if (nmcasts >= MAXMCADDR ||
                   memcmp(enm->enm_addrlo, enm->enm_addrhi,
                   ETHER_ADDR_LEN) != 0) {
                       /*
                        * Callers of this function must do the
                        * right thing with this.  If we're called
                        * from outside fxp_init(), the caller must
                        * detect if the state if IFF_ALLMULTI changes.
                        * If it does, the caller must then call
                        * fxp_init(), since allmulti is handled by
                        * the config block.
                        */
                       ifp->if_flags |= IFF_ALLMULTI;
                       ETHER_UNLOCK(ec);
                       return;
               }
               memcpy(&mcsp->mc_addr[nmcasts][0], enm->enm_addrlo,
                   ETHER_ADDR_LEN);
               nmcasts++;
               ETHER_NEXT_MULTI(step, enm);
       }
       ETHER_UNLOCK(ec);

       /* BIG_ENDIAN: no need to swap to store 0 */
       mcsp->cb_status = 0;
       mcsp->cb_command = htole16(FXP_CB_COMMAND_MCAS | FXP_CB_COMMAND_EL);
       mcsp->link_addr = htole32(FXP_CDTXADDR(sc, FXP_NEXTTX(sc->sc_txlast)));
       mcsp->mc_cnt = htole16(nmcasts * ETHER_ADDR_LEN);

       FXP_CDMCSSYNC(sc, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

       /*
        * Wait until the command unit is not active.  This should never
        * happen since nothing is queued, but make sure anyway.
        */
       count = 100;
       while ((CSR_READ_1(sc, FXP_CSR_SCB_RUSCUS) >> 6) ==
           FXP_SCB_CUS_ACTIVE && --count)
               DELAY(1);
       if (count == 0) {
               log(LOG_WARNING, "%s: line %d: command queue timeout\n",
                   device_xname(sc->sc_dev), __LINE__);
               return;
       }

       /*
        * Start the multicast setup command/DMA.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->sc_cddma + FXP_CDMCSOFF);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);

       /* ...and wait for it to complete. */
       for (count = 1000; count > 0; count--) {
               FXP_CDMCSSYNC(sc,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
               status = le16toh(mcsp->cb_status);
               FXP_CDMCSSYNC(sc, BUS_DMASYNC_PREREAD);
               if ((status & FXP_CB_STATUS_C) != 0)
                       break;
               DELAY(1);
       }
       if (count == 0) {
               log(LOG_WARNING, "%s: line %d: dmasync timeout\n",
                   device_xname(sc->sc_dev), __LINE__);
               return;
       }
}

static const uint32_t fxp_ucode_d101a[] = D101_A_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d101b0[] = D101_B0_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d101ma[] = D101M_B_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d101s[] = D101S_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d102[] = D102_B_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d102c[] = D102_C_RCVBUNDLE_UCODE;
static const uint32_t fxp_ucode_d102e[] = D102_E_RCVBUNDLE_UCODE;

#define UCODE(x)        x, sizeof(x)/sizeof(uint32_t)

static const struct ucode {
       int32_t         revision;
       const uint32_t  *ucode;
       size_t          length;
       uint16_t        int_delay_offset;
       uint16_t        bundle_max_offset;
} ucode_table[] = {
       { FXP_REV_82558_A4, UCODE(fxp_ucode_d101a),
         D101_CPUSAVER_DWORD, 0 },

       { FXP_REV_82558_B0, UCODE(fxp_ucode_d101b0),
         D101_CPUSAVER_DWORD, 0 },

       { FXP_REV_82559_A0, UCODE(fxp_ucode_d101ma),
         D101M_CPUSAVER_DWORD, D101M_CPUSAVER_BUNDLE_MAX_DWORD },

       { FXP_REV_82559S_A, UCODE(fxp_ucode_d101s),
         D101S_CPUSAVER_DWORD, D101S_CPUSAVER_BUNDLE_MAX_DWORD },

       { FXP_REV_82550, UCODE(fxp_ucode_d102),
         D102_B_CPUSAVER_DWORD, D102_B_CPUSAVER_BUNDLE_MAX_DWORD },

       { FXP_REV_82550_C, UCODE(fxp_ucode_d102c),
         D102_C_CPUSAVER_DWORD, D102_C_CPUSAVER_BUNDLE_MAX_DWORD },

       { FXP_REV_82551_F, UCODE(fxp_ucode_d102e),
           D102_E_CPUSAVER_DWORD, D102_E_CPUSAVER_BUNDLE_MAX_DWORD },

       { FXP_REV_82551_10, UCODE(fxp_ucode_d102e),
           D102_E_CPUSAVER_DWORD, D102_E_CPUSAVER_BUNDLE_MAX_DWORD },

       { 0, NULL, 0, 0, 0 }
};

void
fxp_load_ucode(struct fxp_softc *sc)
{
       const struct ucode *uc;
       struct fxp_cb_ucode *cbp = &sc->sc_control_data->fcd_ucode;
       int count, i;
       uint16_t status;

       if (sc->sc_flags & FXPF_UCODE_LOADED)
               return;

       /*
        * Only load the uCode if the user has requested that
        * we do so.
        */
       if ((sc->sc_ethercom.ec_if.if_flags & IFF_LINK0) == 0) {
               sc->sc_int_delay = 0;
               sc->sc_bundle_max = 0;
               return;
       }

       for (uc = ucode_table; uc->ucode != NULL; uc++) {
               if (sc->sc_rev == uc->revision)
                       break;
       }
       if (uc->ucode == NULL)
               return;

       /* BIG ENDIAN: no need to swap to store 0 */
       cbp->cb_status = 0;
       cbp->cb_command = htole16(FXP_CB_COMMAND_UCODE | FXP_CB_COMMAND_EL);
       cbp->link_addr = 0xffffffff;            /* (no) next command */
       for (i = 0; i < uc->length; i++)
               cbp->ucode[i] = htole32(uc->ucode[i]);

       if (uc->int_delay_offset)
               *(volatile uint16_t *) &cbp->ucode[uc->int_delay_offset] =
                   htole16(fxp_int_delay + (fxp_int_delay / 2));

       if (uc->bundle_max_offset)
               *(volatile uint16_t *) &cbp->ucode[uc->bundle_max_offset] =
                   htole16(fxp_bundle_max);

       FXP_CDUCODESYNC(sc, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

       /*
        * Download the uCode to the chip.
        */
       fxp_scb_wait(sc);
       CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->sc_cddma + FXP_CDUCODEOFF);
       fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);

       /* ...and wait for it to complete. */
       for (count = 10000; count > 0; count--) {
               FXP_CDUCODESYNC(sc,
                   BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
               status = le16toh(cbp->cb_status);
               FXP_CDUCODESYNC(sc, BUS_DMASYNC_PREREAD);
               if ((status & FXP_CB_STATUS_C) != 0)
                       break;
               DELAY(2);
       }
       if (count == 0) {
               sc->sc_int_delay = 0;
               sc->sc_bundle_max = 0;
               log(LOG_WARNING, "%s: timeout loading microcode\n",
                   device_xname(sc->sc_dev));
               return;
       }

       if (sc->sc_int_delay != fxp_int_delay ||
           sc->sc_bundle_max != fxp_bundle_max) {
               sc->sc_int_delay = fxp_int_delay;
               sc->sc_bundle_max = fxp_bundle_max;
               log(LOG_INFO, "%s: Microcode loaded: int delay: %d usec, "
                   "max bundle: %d\n", device_xname(sc->sc_dev),
                   sc->sc_int_delay,
                   uc->bundle_max_offset == 0 ? 0 : sc->sc_bundle_max);
       }

       sc->sc_flags |= FXPF_UCODE_LOADED;
}

int
fxp_enable(struct fxp_softc *sc)
{

       if (sc->sc_enabled == 0 && sc->sc_enable != NULL) {
               if ((*sc->sc_enable)(sc) != 0) {
                       log(LOG_ERR, "%s: device enable failed\n",
                           device_xname(sc->sc_dev));
                       return (EIO);
               }
       }

       sc->sc_enabled = 1;
       return (0);
}

void
fxp_disable(struct fxp_softc *sc)
{

       if (sc->sc_enabled != 0 && sc->sc_disable != NULL) {
               (*sc->sc_disable)(sc);
               sc->sc_enabled = 0;
       }
}

/*
* fxp_activate:
*
*      Handle device activation/deactivation requests.
*/
int
fxp_activate(device_t self, enum devact act)
{
       struct fxp_softc *sc = device_private(self);

       switch (act) {
       case DVACT_DEACTIVATE:
               if_deactivate(&sc->sc_ethercom.ec_if);
               return 0;
       default:
               return EOPNOTSUPP;
       }
}

/*
* fxp_detach:
*
*      Detach an i82557 interface.
*/
int
fxp_detach(struct fxp_softc *sc, int flags)
{
       struct ifnet *ifp = &sc->sc_ethercom.ec_if;
       int i, s;

       /* Succeed now if there's no work to do. */
       if ((sc->sc_flags & FXPF_ATTACHED) == 0)
               return (0);

       s = splnet();
       /* Stop the interface. Callouts are stopped in it. */
       fxp_stop(ifp, 1);
       splx(s);

       /* Destroy our callout. */
       callout_destroy(&sc->sc_callout);

       if (sc->sc_flags & FXPF_MII) {
               /* Detach all PHYs */
               mii_detach(&sc->sc_mii, MII_PHY_ANY, MII_OFFSET_ANY);
       }

       rnd_detach_source(&sc->rnd_source);
       ether_ifdetach(ifp);
       if_detach(ifp);

       /* Delete all remaining media. */
       ifmedia_fini(&sc->sc_mii.mii_media);

       for (i = 0; i < FXP_NRFABUFS; i++) {
               bus_dmamap_unload(sc->sc_dmat, sc->sc_rxmaps[i]);
               bus_dmamap_destroy(sc->sc_dmat, sc->sc_rxmaps[i]);
       }

       for (i = 0; i < FXP_NTXCB; i++) {
               bus_dmamap_unload(sc->sc_dmat, FXP_DSTX(sc, i)->txs_dmamap);
               bus_dmamap_destroy(sc->sc_dmat, FXP_DSTX(sc, i)->txs_dmamap);
       }

       bus_dmamap_unload(sc->sc_dmat, sc->sc_dmamap);
       bus_dmamap_destroy(sc->sc_dmat, sc->sc_dmamap);
       bus_dmamem_unmap(sc->sc_dmat, (void *)sc->sc_control_data,
           sizeof(struct fxp_control_data));
       bus_dmamem_free(sc->sc_dmat, &sc->sc_cdseg, sc->sc_cdnseg);

       return (0);
}