xref: /dflybsd-src/sys/dev/netif/msk/if_msk.c (revision c78f83cbcf27a59e0983cc4ebee3c3434f73d84d)

/******************************************************************************
 *
 * Name   : sky2.c
 * Project: Gigabit Ethernet Driver for FreeBSD 5.x/6.x
 * Version: $Revision: 1.23 $
 * Date   : $Date: 2005/12/22 09:04:11 $
 * Purpose: Main driver source file
 *
 *****************************************************************************/

/******************************************************************************
 *
 *	LICENSE:
 *	Copyright (C) Marvell International Ltd. and/or its affiliates
 *
 *	The computer program files contained in this folder ("Files")
 *	are provided to you under the BSD-type license terms provided
 *	below, and any use of such Files and any derivative works
 *	thereof created by you shall be governed by the following terms
 *	and conditions:
 *
 *	- Redistributions of source code must retain the above copyright
 *	  notice, this list of conditions and the following disclaimer.
 *	- 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.
 *	- Neither the name of Marvell nor the names of its contributors
 *	  may be used to endorse or promote products derived from this
 *	  software without specific prior written permission.
 *
 *	THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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
 *	COPYRIGHT OWNER 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.
 *	/LICENSE
 *
 *****************************************************************************/

/*-
 * Copyright (c) 1997, 1998, 1999, 2000
 *	Bill Paul <wpaul@ctr.columbia.edu>.  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, 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.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by Bill Paul.
 * 4. Neither the name of the author nor the names of any co-contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY Bill Paul 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 Bill Paul OR THE VOICES IN HIS HEAD
 * 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) 2003 Nathan L. Binkert <binkertn@umich.edu>
 *
 * Permission to use, copy, modify, and distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */

/* $FreeBSD: src/sys/dev/msk/if_msk.c,v 1.26 2007/12/05 09:41:58 remko Exp $ */
/* $DragonFly: src/sys/dev/netif/msk/if_msk.c,v 1.10 2008/11/23 04:28:27 sephe Exp $ */

/*
 * Device driver for the Marvell Yukon II Ethernet controller.
 * Due to lack of documentation, this driver is based on the code from
 * sk(4) and Marvell's myk(4) driver for FreeBSD 5.x.
 */

#include <sys/param.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/bus.h>
#include <sys/in_cksum.h>
#include <sys/interrupt.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/rman.h>
#include <sys/serialize.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>

#include <net/ethernet.h>
#include <net/if.h>
#include <net/bpf.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/ifq_var.h>
#include <net/vlan/if_vlan_var.h>

#include <netinet/ip.h>
#include <netinet/ip_var.h>

#include <dev/netif/mii_layer/miivar.h>

#include <bus/pci/pcireg.h>
#include <bus/pci/pcivar.h>

#include "if_mskreg.h"

/* "device miibus" required.  See GENERIC if you get errors here. */
#include "miibus_if.h"

#define MSK_CSUM_FEATURES	(CSUM_TCP | CSUM_UDP)

/*
 * Devices supported by this driver.
 */
static const struct msk_product {
	uint16_t	msk_vendorid;
	uint16_t	msk_deviceid;
	const char	*msk_name;
} msk_products[] = {
	{ VENDORID_SK, DEVICEID_SK_YUKON2,
	    "SK-9Sxx Gigabit Ethernet" },
	{ VENDORID_SK, DEVICEID_SK_YUKON2_EXPR,
	    "SK-9Exx Gigabit Ethernet"},
	{ VENDORID_MARVELL, DEVICEID_MRVL_8021CU,
	    "Marvell Yukon 88E8021CU Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8021X,
	    "Marvell Yukon 88E8021 SX/LX Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8022CU,
	    "Marvell Yukon 88E8022CU Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8022X,
	    "Marvell Yukon 88E8022 SX/LX Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8061CU,
	    "Marvell Yukon 88E8061CU Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8061X,
	    "Marvell Yukon 88E8061 SX/LX Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8062CU,
	    "Marvell Yukon 88E8062CU Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8062X,
	    "Marvell Yukon 88E8062 SX/LX Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8035,
	    "Marvell Yukon 88E8035 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8036,
	    "Marvell Yukon 88E8036 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8038,
	    "Marvell Yukon 88E8038 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_8039,
	    "Marvell Yukon 88E8039 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_4361,
	    "Marvell Yukon 88E8050 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_4360,
	    "Marvell Yukon 88E8052 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_4362,
	    "Marvell Yukon 88E8053 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_4363,
	    "Marvell Yukon 88E8055 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_4364,
	    "Marvell Yukon 88E8056 Gigabit Ethernet" },
	{ VENDORID_MARVELL, DEVICEID_MRVL_436A,
	    "Marvell Yukon 88E8058 Gigabit Ethernet" },
	{ VENDORID_DLINK, DEVICEID_DLINK_DGE550SX,
	    "D-Link 550SX Gigabit Ethernet" },
	{ VENDORID_DLINK, DEVICEID_DLINK_DGE560T,
	    "D-Link 560T Gigabit Ethernet" },
	{ 0, 0, NULL }
};

static const char *model_name[] = {
	"Yukon XL",
        "Yukon EC Ultra",
        "Yukon Unknown",
        "Yukon EC",
        "Yukon FE"
};

static int	mskc_probe(device_t);
static int	mskc_attach(device_t);
static int	mskc_detach(device_t);
static int	mskc_shutdown(device_t);
static int	mskc_suspend(device_t);
static int	mskc_resume(device_t);
static void	mskc_intr(void *);

static void	mskc_reset(struct msk_softc *);
static void	mskc_set_imtimer(struct msk_softc *);
static void	mskc_intr_hwerr(struct msk_softc *);
static int	mskc_handle_events(struct msk_softc *);
static void	mskc_phy_power(struct msk_softc *, int);
static int	mskc_setup_rambuffer(struct msk_softc *);
static int	mskc_status_dma_alloc(struct msk_softc *);
static void	mskc_status_dma_free(struct msk_softc *);
static int	mskc_sysctl_proc_limit(SYSCTL_HANDLER_ARGS);
static int	mskc_sysctl_intr_rate(SYSCTL_HANDLER_ARGS);

static int	msk_probe(device_t);
static int	msk_attach(device_t);
static int	msk_detach(device_t);
static int	msk_miibus_readreg(device_t, int, int);
static int	msk_miibus_writereg(device_t, int, int, int);
static void	msk_miibus_statchg(device_t);

static void	msk_init(void *);
static int	msk_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
static void	msk_start(struct ifnet *);
static void	msk_watchdog(struct ifnet *);
static int	msk_mediachange(struct ifnet *);
static void	msk_mediastatus(struct ifnet *, struct ifmediareq *);

static void	msk_tick(void *);
static void	msk_intr_phy(struct msk_if_softc *);
static void	msk_intr_gmac(struct msk_if_softc *);
static __inline void
		msk_rxput(struct msk_if_softc *);
static void	msk_handle_hwerr(struct msk_if_softc *, uint32_t);
static void	msk_rxeof(struct msk_if_softc *, uint32_t, int,
			  struct mbuf_chain *);
static void	msk_txeof(struct msk_if_softc *, int);
static void	msk_set_prefetch(struct msk_softc *, int, bus_addr_t, uint32_t);
static void	msk_set_rambuffer(struct msk_if_softc *);
static void	msk_stop(struct msk_if_softc *);

static void	msk_dmamap_mbuf_cb(void *, bus_dma_segment_t *, int,
				   bus_size_t, int);
static int	msk_txrx_dma_alloc(struct msk_if_softc *);
static void	msk_txrx_dma_free(struct msk_if_softc *);
static int	msk_init_rx_ring(struct msk_if_softc *);
static void	msk_init_tx_ring(struct msk_if_softc *);
static __inline void
		msk_discard_rxbuf(struct msk_if_softc *, int);
static int	msk_newbuf(struct msk_if_softc *, int);
static struct mbuf *
		msk_defrag(struct mbuf *, int, int);
static int	msk_encap(struct msk_if_softc *, struct mbuf **);

#ifdef MSK_JUMBO
static int msk_init_jumbo_rx_ring(struct msk_if_softc *);
static __inline void msk_discard_jumbo_rxbuf(struct msk_if_softc *, int);
static int msk_jumbo_newbuf(struct msk_if_softc *, int);
static void msk_jumbo_rxeof(struct msk_if_softc *, uint32_t, int);
static void *msk_jalloc(struct msk_if_softc *);
static void msk_jfree(void *, void *);
#endif

static int	msk_phy_readreg(struct msk_if_softc *, int, int);
static int	msk_phy_writereg(struct msk_if_softc *, int, int, int);

static void	msk_setmulti(struct msk_if_softc *);
static void	msk_setvlan(struct msk_if_softc *, struct ifnet *);
static void	msk_setpromisc(struct msk_if_softc *);

static int	msk_dmamem_create(device_t, bus_size_t, bus_dma_tag_t *,
				  void **, bus_addr_t *, bus_dmamap_t *);
static void	msk_dmamem_destroy(bus_dma_tag_t, void *, bus_dmamap_t);

static device_method_t mskc_methods[] = {
	/* Device interface */
	DEVMETHOD(device_probe,		mskc_probe),
	DEVMETHOD(device_attach,	mskc_attach),
	DEVMETHOD(device_detach,	mskc_detach),
	DEVMETHOD(device_suspend,	mskc_suspend),
	DEVMETHOD(device_resume,	mskc_resume),
	DEVMETHOD(device_shutdown,	mskc_shutdown),

	/* bus interface */
	DEVMETHOD(bus_print_child,	bus_generic_print_child),
	DEVMETHOD(bus_driver_added,	bus_generic_driver_added),

	{ NULL, NULL }
};

static DEFINE_CLASS_0(mskc, mskc_driver, mskc_methods, sizeof(struct msk_softc));
static devclass_t mskc_devclass;

static device_method_t msk_methods[] = {
	/* Device interface */
	DEVMETHOD(device_probe,		msk_probe),
	DEVMETHOD(device_attach,	msk_attach),
	DEVMETHOD(device_detach,	msk_detach),
	DEVMETHOD(device_shutdown,	bus_generic_shutdown),

	/* bus interface */
	DEVMETHOD(bus_print_child,	bus_generic_print_child),
	DEVMETHOD(bus_driver_added,	bus_generic_driver_added),

	/* MII interface */
	DEVMETHOD(miibus_readreg,	msk_miibus_readreg),
	DEVMETHOD(miibus_writereg,	msk_miibus_writereg),
	DEVMETHOD(miibus_statchg,	msk_miibus_statchg),

	{ NULL, NULL }
};

static DEFINE_CLASS_0(msk, msk_driver, msk_methods, sizeof(struct msk_if_softc));
static devclass_t msk_devclass;

DECLARE_DUMMY_MODULE(if_msk);
DRIVER_MODULE(if_msk, pci, mskc_driver, mskc_devclass, 0, 0);
DRIVER_MODULE(if_msk, mskc, msk_driver, msk_devclass, 0, 0);
DRIVER_MODULE(miibus, msk, miibus_driver, miibus_devclass, 0, 0);

static int	mskc_intr_rate = 0;
static int	mskc_process_limit = MSK_PROC_DEFAULT;

TUNABLE_INT("hw.mskc.intr_rate", &mskc_intr_rate);
TUNABLE_INT("hw.mskc.process_limit", &mskc_process_limit);

static int
msk_miibus_readreg(device_t dev, int phy, int reg)
{
	struct msk_if_softc *sc_if;

	if (phy != PHY_ADDR_MARV)
		return (0);

	sc_if = device_get_softc(dev);

	return (msk_phy_readreg(sc_if, phy, reg));
}

static int
msk_phy_readreg(struct msk_if_softc *sc_if, int phy, int reg)
{
	struct msk_softc *sc;
	int i, val;

	sc = sc_if->msk_softc;

        GMAC_WRITE_2(sc, sc_if->msk_port, GM_SMI_CTRL,
	    GM_SMI_CT_PHY_AD(phy) | GM_SMI_CT_REG_AD(reg) | GM_SMI_CT_OP_RD);

	for (i = 0; i < MSK_TIMEOUT; i++) {
		DELAY(1);
		val = GMAC_READ_2(sc, sc_if->msk_port, GM_SMI_CTRL);
		if ((val & GM_SMI_CT_RD_VAL) != 0) {
			val = GMAC_READ_2(sc, sc_if->msk_port, GM_SMI_DATA);
			break;
		}
	}

	if (i == MSK_TIMEOUT) {
		if_printf(sc_if->msk_ifp, "phy failed to come ready\n");
		val = 0;
	}

	return (val);
}

static int
msk_miibus_writereg(device_t dev, int phy, int reg, int val)
{
	struct msk_if_softc *sc_if;

	if (phy != PHY_ADDR_MARV)
		return (0);

	sc_if = device_get_softc(dev);

	return (msk_phy_writereg(sc_if, phy, reg, val));
}

static int
msk_phy_writereg(struct msk_if_softc *sc_if, int phy, int reg, int val)
{
	struct msk_softc *sc;
	int i;

	sc = sc_if->msk_softc;

	GMAC_WRITE_2(sc, sc_if->msk_port, GM_SMI_DATA, val);
        GMAC_WRITE_2(sc, sc_if->msk_port, GM_SMI_CTRL,
	    GM_SMI_CT_PHY_AD(phy) | GM_SMI_CT_REG_AD(reg));
	for (i = 0; i < MSK_TIMEOUT; i++) {
		DELAY(1);
		if ((GMAC_READ_2(sc, sc_if->msk_port, GM_SMI_CTRL) &
		    GM_SMI_CT_BUSY) == 0)
			break;
	}
	if (i == MSK_TIMEOUT)
		if_printf(sc_if->msk_ifp, "phy write timeout\n");

	return (0);
}

static void
msk_miibus_statchg(device_t dev)
{
	struct msk_if_softc *sc_if;
	struct msk_softc *sc;
	struct mii_data *mii;
	struct ifnet *ifp;
	uint32_t gmac;

	sc_if = device_get_softc(dev);
	sc = sc_if->msk_softc;

	mii = device_get_softc(sc_if->msk_miibus);
	ifp = sc_if->msk_ifp;

	if (mii->mii_media_status & IFM_ACTIVE) {
		if (IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE)
			sc_if->msk_link = 1;
	} else
		sc_if->msk_link = 0;

	if (sc_if->msk_link != 0) {
		/* Enable Tx FIFO Underrun. */
		CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, GMAC_IRQ_MSK),
		    GM_IS_TX_FF_UR | GM_IS_RX_FF_OR);
		/*
		 * Because mii(4) notify msk(4) that it detected link status
		 * change, there is no need to enable automatic
		 * speed/flow-control/duplex updates.
		 */
		gmac = GM_GPCR_AU_ALL_DIS;
		switch (IFM_SUBTYPE(mii->mii_media_active)) {
		case IFM_1000_SX:
		case IFM_1000_T:
			gmac |= GM_GPCR_SPEED_1000;
			break;
		case IFM_100_TX:
			gmac |= GM_GPCR_SPEED_100;
			break;
		case IFM_10_T:
			break;
		}

		if (((mii->mii_media_active & IFM_GMASK) & IFM_FDX) != 0)
			gmac |= GM_GPCR_DUP_FULL;
		/* Disable Rx flow control. */
		if (((mii->mii_media_active & IFM_GMASK) & IFM_FLAG0) == 0)
			gmac |= GM_GPCR_FC_RX_DIS;
		/* Disable Tx flow control. */
		if (((mii->mii_media_active & IFM_GMASK) & IFM_FLAG1) == 0)
			gmac |= GM_GPCR_FC_TX_DIS;
		gmac |= GM_GPCR_RX_ENA | GM_GPCR_TX_ENA;
		GMAC_WRITE_2(sc, sc_if->msk_port, GM_GP_CTRL, gmac);
		/* Read again to ensure writing. */
		GMAC_READ_2(sc, sc_if->msk_port, GM_GP_CTRL);

		gmac = GMC_PAUSE_ON;
		if (((mii->mii_media_active & IFM_GMASK) &
		    (IFM_FLAG0 | IFM_FLAG1)) == 0)
			gmac = GMC_PAUSE_OFF;
		/* Diable pause for 10/100 Mbps in half-duplex mode. */
		if ((((mii->mii_media_active & IFM_GMASK) & IFM_FDX) == 0) &&
		    (IFM_SUBTYPE(mii->mii_media_active) == IFM_100_TX ||
		    IFM_SUBTYPE(mii->mii_media_active) == IFM_10_T))
			gmac = GMC_PAUSE_OFF;
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, GMAC_CTRL), gmac);

		/* Enable PHY interrupt for FIFO underrun/overflow. */
		msk_phy_writereg(sc_if, PHY_ADDR_MARV,
		    PHY_MARV_INT_MASK, PHY_M_IS_FIFO_ERROR);
	} else {
		/*
		 * Link state changed to down.
		 * Disable PHY interrupts.
		 */
		msk_phy_writereg(sc_if, PHY_ADDR_MARV, PHY_MARV_INT_MASK, 0);
		/* Disable Rx/Tx MAC. */
		gmac = GMAC_READ_2(sc, sc_if->msk_port, GM_GP_CTRL);
		gmac &= ~(GM_GPCR_RX_ENA | GM_GPCR_TX_ENA);
		GMAC_WRITE_2(sc, sc_if->msk_port, GM_GP_CTRL, gmac);
		/* Read again to ensure writing. */
		GMAC_READ_2(sc, sc_if->msk_port, GM_GP_CTRL);
	}
}

static void
msk_setmulti(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc;
	struct ifnet *ifp;
	struct ifmultiaddr *ifma;
	uint32_t mchash[2];
	uint32_t crc;
	uint16_t mode;

	sc = sc_if->msk_softc;
	ifp = sc_if->msk_ifp;

	bzero(mchash, sizeof(mchash));
	mode = GMAC_READ_2(sc, sc_if->msk_port, GM_RX_CTRL);
	mode |= GM_RXCR_UCF_ENA;
	if ((ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI)) != 0) {
		if ((ifp->if_flags & IFF_PROMISC) != 0)
			mode &= ~(GM_RXCR_UCF_ENA | GM_RXCR_MCF_ENA);
		else if ((ifp->if_flags & IFF_ALLMULTI) != 0) {
			mchash[0] = 0xffff;
			mchash[1] = 0xffff;
		}
	} else {
		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
			if (ifma->ifma_addr->sa_family != AF_LINK)
				continue;
			crc = ether_crc32_be(LLADDR((struct sockaddr_dl *)
			    ifma->ifma_addr), ETHER_ADDR_LEN);
			/* Just want the 6 least significant bits. */
			crc &= 0x3f;
			/* Set the corresponding bit in the hash table. */
			mchash[crc >> 5] |= 1 << (crc & 0x1f);
		}
		mode |= GM_RXCR_MCF_ENA;
	}

	GMAC_WRITE_2(sc, sc_if->msk_port, GM_MC_ADDR_H1,
	    mchash[0] & 0xffff);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_MC_ADDR_H2,
	    (mchash[0] >> 16) & 0xffff);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_MC_ADDR_H3,
	    mchash[1] & 0xffff);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_MC_ADDR_H4,
	    (mchash[1] >> 16) & 0xffff);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_RX_CTRL, mode);
}

static void
msk_setvlan(struct msk_if_softc *sc_if, struct ifnet *ifp)
{
	struct msk_softc *sc;

	sc = sc_if->msk_softc;
	if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T),
		    RX_VLAN_STRIP_ON);
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
		    TX_VLAN_TAG_ON);
	} else {
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T),
		    RX_VLAN_STRIP_OFF);
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
		    TX_VLAN_TAG_OFF);
	}
}

static void
msk_setpromisc(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc;
	struct ifnet *ifp;
	uint16_t mode;

	sc = sc_if->msk_softc;
	ifp = sc_if->msk_ifp;

	mode = GMAC_READ_2(sc, sc_if->msk_port, GM_RX_CTRL);
	if (ifp->if_flags & IFF_PROMISC)
		mode &= ~(GM_RXCR_UCF_ENA | GM_RXCR_MCF_ENA);
	else
		mode |= (GM_RXCR_UCF_ENA | GM_RXCR_MCF_ENA);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_RX_CTRL, mode);
}

static int
msk_init_rx_ring(struct msk_if_softc *sc_if)
{
	struct msk_ring_data *rd;
	struct msk_rxdesc *rxd;
	int i, prod;

	sc_if->msk_cdata.msk_rx_cons = 0;
	sc_if->msk_cdata.msk_rx_prod = 0;
	sc_if->msk_cdata.msk_rx_putwm = MSK_PUT_WM;

	rd = &sc_if->msk_rdata;
	bzero(rd->msk_rx_ring, sizeof(struct msk_rx_desc) * MSK_RX_RING_CNT);
	prod = sc_if->msk_cdata.msk_rx_prod;
	for (i = 0; i < MSK_RX_RING_CNT; i++) {
		rxd = &sc_if->msk_cdata.msk_rxdesc[prod];
		rxd->rx_m = NULL;
		rxd->rx_le = &rd->msk_rx_ring[prod];
		if (msk_newbuf(sc_if, prod) != 0)
			return (ENOBUFS);
		MSK_INC(prod, MSK_RX_RING_CNT);
	}

	/* Update prefetch unit. */
	sc_if->msk_cdata.msk_rx_prod = MSK_RX_RING_CNT - 1;
	CSR_WRITE_2(sc_if->msk_softc,
	    Y2_PREF_Q_ADDR(sc_if->msk_rxq, PREF_UNIT_PUT_IDX_REG),
	    sc_if->msk_cdata.msk_rx_prod);

	return (0);
}

#ifdef MSK_JUMBO
static int
msk_init_jumbo_rx_ring(struct msk_if_softc *sc_if)
{
	struct msk_ring_data *rd;
	struct msk_rxdesc *rxd;
	int i, prod;

	MSK_IF_LOCK_ASSERT(sc_if);

	sc_if->msk_cdata.msk_rx_cons = 0;
	sc_if->msk_cdata.msk_rx_prod = 0;
	sc_if->msk_cdata.msk_rx_putwm = MSK_PUT_WM;

	rd = &sc_if->msk_rdata;
	bzero(rd->msk_jumbo_rx_ring,
	    sizeof(struct msk_rx_desc) * MSK_JUMBO_RX_RING_CNT);
	prod = sc_if->msk_cdata.msk_rx_prod;
	for (i = 0; i < MSK_JUMBO_RX_RING_CNT; i++) {
		rxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[prod];
		rxd->rx_m = NULL;
		rxd->rx_le = &rd->msk_jumbo_rx_ring[prod];
		if (msk_jumbo_newbuf(sc_if, prod) != 0)
			return (ENOBUFS);
		MSK_INC(prod, MSK_JUMBO_RX_RING_CNT);
	}

	bus_dmamap_sync(sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
	    sc_if->msk_cdata.msk_jumbo_rx_ring_map,
	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);

	sc_if->msk_cdata.msk_rx_prod = MSK_JUMBO_RX_RING_CNT - 1;
	CSR_WRITE_2(sc_if->msk_softc,
	    Y2_PREF_Q_ADDR(sc_if->msk_rxq, PREF_UNIT_PUT_IDX_REG),
	    sc_if->msk_cdata.msk_rx_prod);

	return (0);
}
#endif

static void
msk_init_tx_ring(struct msk_if_softc *sc_if)
{
	struct msk_ring_data *rd;
	struct msk_txdesc *txd;
	int i;

	sc_if->msk_cdata.msk_tx_prod = 0;
	sc_if->msk_cdata.msk_tx_cons = 0;
	sc_if->msk_cdata.msk_tx_cnt = 0;

	rd = &sc_if->msk_rdata;
	bzero(rd->msk_tx_ring, sizeof(struct msk_tx_desc) * MSK_TX_RING_CNT);
	for (i = 0; i < MSK_TX_RING_CNT; i++) {
		txd = &sc_if->msk_cdata.msk_txdesc[i];
		txd->tx_m = NULL;
		txd->tx_le = &rd->msk_tx_ring[i];
	}
}

static __inline void
msk_discard_rxbuf(struct msk_if_softc *sc_if, int idx)
{
	struct msk_rx_desc *rx_le;
	struct msk_rxdesc *rxd;
	struct mbuf *m;

	rxd = &sc_if->msk_cdata.msk_rxdesc[idx];
	m = rxd->rx_m;
	rx_le = rxd->rx_le;
	rx_le->msk_control = htole32(m->m_len | OP_PACKET | HW_OWNER);
}

#ifdef MSK_JUMBO
static __inline void
msk_discard_jumbo_rxbuf(struct msk_if_softc *sc_if, int	idx)
{
	struct msk_rx_desc *rx_le;
	struct msk_rxdesc *rxd;
	struct mbuf *m;

	rxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[idx];
	m = rxd->rx_m;
	rx_le = rxd->rx_le;
	rx_le->msk_control = htole32(m->m_len | OP_PACKET | HW_OWNER);
}
#endif

static int
msk_newbuf(struct msk_if_softc *sc_if, int idx)
{
	struct msk_rx_desc *rx_le;
	struct msk_rxdesc *rxd;
	struct mbuf *m;
	struct msk_dmamap_arg ctx;
	bus_dma_segment_t seg;
	bus_dmamap_t map;

	m = m_getcl(MB_DONTWAIT, MT_DATA, M_PKTHDR);
	if (m == NULL)
		return (ENOBUFS);

	m->m_len = m->m_pkthdr.len = MCLBYTES;
	m_adj(m, ETHER_ALIGN);

	bzero(&ctx, sizeof(ctx));
	ctx.nseg = 1;
	ctx.segs = &seg;
	if (bus_dmamap_load_mbuf(sc_if->msk_cdata.msk_rx_tag,
	    sc_if->msk_cdata.msk_rx_sparemap, m, msk_dmamap_mbuf_cb, &ctx,
	    BUS_DMA_NOWAIT) != 0) {
		m_freem(m);
		return (ENOBUFS);
	}
	KASSERT(ctx.nseg == 1,
		("%s: %d segments returned!", __func__, ctx.nseg));

	rxd = &sc_if->msk_cdata.msk_rxdesc[idx];
	if (rxd->rx_m != NULL) {
		bus_dmamap_sync(sc_if->msk_cdata.msk_rx_tag, rxd->rx_dmamap,
		    BUS_DMASYNC_POSTREAD);
		bus_dmamap_unload(sc_if->msk_cdata.msk_rx_tag, rxd->rx_dmamap);
	}
	map = rxd->rx_dmamap;
	rxd->rx_dmamap = sc_if->msk_cdata.msk_rx_sparemap;
	sc_if->msk_cdata.msk_rx_sparemap = map;
	bus_dmamap_sync(sc_if->msk_cdata.msk_rx_tag, rxd->rx_dmamap,
	    BUS_DMASYNC_PREREAD);
	rxd->rx_m = m;
	rx_le = rxd->rx_le;
	rx_le->msk_addr = htole32(MSK_ADDR_LO(seg.ds_addr));
	rx_le->msk_control =
	    htole32(seg.ds_len | OP_PACKET | HW_OWNER);

	return (0);
}

#ifdef MSK_JUMBO
static int
msk_jumbo_newbuf(struct msk_if_softc *sc_if, int idx)
{
	struct msk_rx_desc *rx_le;
	struct msk_rxdesc *rxd;
	struct mbuf *m;
	bus_dma_segment_t segs[1];
	bus_dmamap_t map;
	int nsegs;
	void *buf;

	MGETHDR(m, M_DONTWAIT, MT_DATA);
	if (m == NULL)
		return (ENOBUFS);
	buf = msk_jalloc(sc_if);
	if (buf == NULL) {
		m_freem(m);
		return (ENOBUFS);
	}
	/* Attach the buffer to the mbuf. */
	MEXTADD(m, buf, MSK_JLEN, msk_jfree, (struct msk_if_softc *)sc_if, 0,
	    EXT_NET_DRV);
	if ((m->m_flags & M_EXT) == 0) {
		m_freem(m);
		return (ENOBUFS);
	}
	m->m_pkthdr.len = m->m_len = MSK_JLEN;
	m_adj(m, ETHER_ALIGN);

	if (bus_dmamap_load_mbuf_sg(sc_if->msk_cdata.msk_jumbo_rx_tag,
	    sc_if->msk_cdata.msk_jumbo_rx_sparemap, m, segs, &nsegs,
	    BUS_DMA_NOWAIT) != 0) {
		m_freem(m);
		return (ENOBUFS);
	}
	KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));

	rxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[idx];
	if (rxd->rx_m != NULL) {
		bus_dmamap_sync(sc_if->msk_cdata.msk_jumbo_rx_tag,
		    rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
		bus_dmamap_unload(sc_if->msk_cdata.msk_jumbo_rx_tag,
		    rxd->rx_dmamap);
	}
	map = rxd->rx_dmamap;
	rxd->rx_dmamap = sc_if->msk_cdata.msk_jumbo_rx_sparemap;
	sc_if->msk_cdata.msk_jumbo_rx_sparemap = map;
	bus_dmamap_sync(sc_if->msk_cdata.msk_jumbo_rx_tag, rxd->rx_dmamap,
	    BUS_DMASYNC_PREREAD);
	rxd->rx_m = m;
	rx_le = rxd->rx_le;
	rx_le->msk_addr = htole32(MSK_ADDR_LO(segs[0].ds_addr));
	rx_le->msk_control =
	    htole32(segs[0].ds_len | OP_PACKET | HW_OWNER);

	return (0);
}
#endif

/*
 * Set media options.
 */
static int
msk_mediachange(struct ifnet *ifp)
{
	struct msk_if_softc *sc_if = ifp->if_softc;
	struct mii_data	*mii;

	mii = device_get_softc(sc_if->msk_miibus);
	mii_mediachg(mii);

	return (0);
}

/*
 * Report current media status.
 */
static void
msk_mediastatus(struct ifnet *ifp, struct ifmediareq *ifmr)
{
	struct msk_if_softc *sc_if = ifp->if_softc;
	struct mii_data	*mii;

	mii = device_get_softc(sc_if->msk_miibus);
	mii_pollstat(mii);

	ifmr->ifm_active = mii->mii_media_active;
	ifmr->ifm_status = mii->mii_media_status;
}

static int
msk_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr)
{
	struct msk_if_softc *sc_if;
	struct ifreq *ifr;
	struct mii_data	*mii;
	int error, mask;

	sc_if = ifp->if_softc;
	ifr = (struct ifreq *)data;
	error = 0;

	switch(command) {
	case SIOCSIFMTU:
#ifdef MSK_JUMBO
		if (ifr->ifr_mtu > MSK_JUMBO_MTU || ifr->ifr_mtu < ETHERMIN) {
			error = EINVAL;
			break;
		}
		if (sc_if->msk_softc->msk_hw_id == CHIP_ID_YUKON_FE &&
		    ifr->ifr_mtu > MSK_MAX_FRAMELEN) {
			error = EINVAL;
			break;
		}
		ifp->if_mtu = ifr->ifr_mtu;
		if ((ifp->if_flags & IFF_RUNNING) != 0)
			msk_init(sc_if);
#else
		error = EOPNOTSUPP;
#endif
		break;

	case SIOCSIFFLAGS:
		if (ifp->if_flags & IFF_UP) {
			if (ifp->if_flags & IFF_RUNNING) {
				if (((ifp->if_flags ^ sc_if->msk_if_flags)
				    & IFF_PROMISC) != 0) {
					msk_setpromisc(sc_if);
					msk_setmulti(sc_if);
				}
			} else {
				if (sc_if->msk_detach == 0)
					msk_init(sc_if);
			}
		} else {
			if (ifp->if_flags & IFF_RUNNING)
				msk_stop(sc_if);
		}
		sc_if->msk_if_flags = ifp->if_flags;
		break;

	case SIOCADDMULTI:
	case SIOCDELMULTI:
		if (ifp->if_flags & IFF_RUNNING)
			msk_setmulti(sc_if);
		break;

	case SIOCGIFMEDIA:
	case SIOCSIFMEDIA:
		mii = device_get_softc(sc_if->msk_miibus);
		error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
		break;

	case SIOCSIFCAP:
		mask = ifr->ifr_reqcap ^ ifp->if_capenable;
		if ((mask & IFCAP_TXCSUM) != 0) {
			ifp->if_capenable ^= IFCAP_TXCSUM;
			if ((IFCAP_TXCSUM & ifp->if_capenable) != 0 &&
			    (IFCAP_TXCSUM & ifp->if_capabilities) != 0)
				ifp->if_hwassist |= MSK_CSUM_FEATURES;
			else
				ifp->if_hwassist &= ~MSK_CSUM_FEATURES;
		}
#ifdef notyet
		if ((mask & IFCAP_VLAN_HWTAGGING) != 0) {
			ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
			msk_setvlan(sc_if, ifp);
		}
#endif

		if (sc_if->msk_framesize > MSK_MAX_FRAMELEN &&
		    sc_if->msk_softc->msk_hw_id == CHIP_ID_YUKON_EC_U) {
			/*
			 * In Yukon EC Ultra, TSO & checksum offload is not
			 * supported for jumbo frame.
			 */
			ifp->if_hwassist &= ~MSK_CSUM_FEATURES;
			ifp->if_capenable &= ~IFCAP_TXCSUM;
		}
		break;

	default:
		error = ether_ioctl(ifp, command, data);
		break;
	}

	return (error);
}

static int
mskc_probe(device_t dev)
{
	const struct msk_product *mp;
	uint16_t vendor, devid;

	vendor = pci_get_vendor(dev);
	devid = pci_get_device(dev);
	for (mp = msk_products; mp->msk_name != NULL; ++mp) {
		if (vendor == mp->msk_vendorid && devid == mp->msk_deviceid) {
			device_set_desc(dev, mp->msk_name);
			return (0);
		}
	}
	return (ENXIO);
}

static int
mskc_setup_rambuffer(struct msk_softc *sc)
{
	int next;
	int i;
	uint8_t val;

	/* Get adapter SRAM size. */
	val = CSR_READ_1(sc, B2_E_0);
	sc->msk_ramsize = (val == 0) ? 128 : val * 4;
	if (bootverbose) {
		device_printf(sc->msk_dev,
		    "RAM buffer size : %dKB\n", sc->msk_ramsize);
	}
	/*
	 * Give receiver 2/3 of memory and round down to the multiple
	 * of 1024. Tx/Rx RAM buffer size of Yukon II shoud be multiple
	 * of 1024.
	 */
	sc->msk_rxqsize = rounddown((sc->msk_ramsize * 1024 * 2) / 3, 1024);
	sc->msk_txqsize = (sc->msk_ramsize * 1024) - sc->msk_rxqsize;
	for (i = 0, next = 0; i < sc->msk_num_port; i++) {
		sc->msk_rxqstart[i] = next;
		sc->msk_rxqend[i] = next + sc->msk_rxqsize - 1;
		next = sc->msk_rxqend[i] + 1;
		sc->msk_txqstart[i] = next;
		sc->msk_txqend[i] = next + sc->msk_txqsize - 1;
		next = sc->msk_txqend[i] + 1;
		if (bootverbose) {
			device_printf(sc->msk_dev,
			    "Port %d : Rx Queue %dKB(0x%08x:0x%08x)\n", i,
			    sc->msk_rxqsize / 1024, sc->msk_rxqstart[i],
			    sc->msk_rxqend[i]);
			device_printf(sc->msk_dev,
			    "Port %d : Tx Queue %dKB(0x%08x:0x%08x)\n", i,
			    sc->msk_txqsize / 1024, sc->msk_txqstart[i],
			    sc->msk_txqend[i]);
		}
	}

	return (0);
}

static void
mskc_phy_power(struct msk_softc *sc, int mode)
{
	uint32_t val;
	int i;

	switch (mode) {
	case MSK_PHY_POWERUP:
		/* Switch power to VCC (WA for VAUX problem). */
		CSR_WRITE_1(sc, B0_POWER_CTRL,
		    PC_VAUX_ENA | PC_VCC_ENA | PC_VAUX_OFF | PC_VCC_ON);
		/* Disable Core Clock Division, set Clock Select to 0. */
		CSR_WRITE_4(sc, B2_Y2_CLK_CTRL, Y2_CLK_DIV_DIS);

		val = 0;
		if (sc->msk_hw_id == CHIP_ID_YUKON_XL &&
		    sc->msk_hw_rev > CHIP_REV_YU_XL_A1) {
			/* Enable bits are inverted. */
			val = Y2_PCI_CLK_LNK1_DIS | Y2_COR_CLK_LNK1_DIS |
			      Y2_CLK_GAT_LNK1_DIS | Y2_PCI_CLK_LNK2_DIS |
			      Y2_COR_CLK_LNK2_DIS | Y2_CLK_GAT_LNK2_DIS;
		}
		/*
		 * Enable PCI & Core Clock, enable clock gating for both Links.
		 */
		CSR_WRITE_1(sc, B2_Y2_CLK_GATE, val);

		val = pci_read_config(sc->msk_dev, PCI_OUR_REG_1, 4);
		val &= ~(PCI_Y2_PHY1_POWD | PCI_Y2_PHY2_POWD);
		if (sc->msk_hw_id == CHIP_ID_YUKON_XL &&
		    sc->msk_hw_rev > CHIP_REV_YU_XL_A1) {
			/* Deassert Low Power for 1st PHY. */
			val |= PCI_Y2_PHY1_COMA;
			if (sc->msk_num_port > 1)
				val |= PCI_Y2_PHY2_COMA;
		} else if (sc->msk_hw_id == CHIP_ID_YUKON_EC_U) {
			uint32_t our;

			CSR_WRITE_2(sc, B0_CTST, Y2_HW_WOL_ON);

			/* Enable all clocks. */
			pci_write_config(sc->msk_dev, PCI_OUR_REG_3, 0, 4);
			our = pci_read_config(sc->msk_dev, PCI_OUR_REG_4, 4);
			our &= (PCI_FORCE_ASPM_REQUEST|PCI_ASPM_GPHY_LINK_DOWN|
			    PCI_ASPM_INT_FIFO_EMPTY|PCI_ASPM_CLKRUN_REQUEST);
			/* Set all bits to 0 except bits 15..12. */
			pci_write_config(sc->msk_dev, PCI_OUR_REG_4, our, 4);
			/* Set to default value. */
			pci_write_config(sc->msk_dev, PCI_OUR_REG_5, 0, 4);
		}
		/* Release PHY from PowerDown/COMA mode. */
		pci_write_config(sc->msk_dev, PCI_OUR_REG_1, val, 4);
		for (i = 0; i < sc->msk_num_port; i++) {
			CSR_WRITE_2(sc, MR_ADDR(i, GMAC_LINK_CTRL),
			    GMLC_RST_SET);
			CSR_WRITE_2(sc, MR_ADDR(i, GMAC_LINK_CTRL),
			    GMLC_RST_CLR);
		}
		break;
	case MSK_PHY_POWERDOWN:
		val = pci_read_config(sc->msk_dev, PCI_OUR_REG_1, 4);
		val |= PCI_Y2_PHY1_POWD | PCI_Y2_PHY2_POWD;
		if (sc->msk_hw_id == CHIP_ID_YUKON_XL &&
		    sc->msk_hw_rev > CHIP_REV_YU_XL_A1) {
			val &= ~PCI_Y2_PHY1_COMA;
			if (sc->msk_num_port > 1)
				val &= ~PCI_Y2_PHY2_COMA;
		}
		pci_write_config(sc->msk_dev, PCI_OUR_REG_1, val, 4);

		val = Y2_PCI_CLK_LNK1_DIS | Y2_COR_CLK_LNK1_DIS |
		      Y2_CLK_GAT_LNK1_DIS | Y2_PCI_CLK_LNK2_DIS |
		      Y2_COR_CLK_LNK2_DIS | Y2_CLK_GAT_LNK2_DIS;
		if (sc->msk_hw_id == CHIP_ID_YUKON_XL &&
		    sc->msk_hw_rev > CHIP_REV_YU_XL_A1) {
			/* Enable bits are inverted. */
			val = 0;
		}
		/*
		 * Disable PCI & Core Clock, disable clock gating for
		 * both Links.
		 */
		CSR_WRITE_1(sc, B2_Y2_CLK_GATE, val);
		CSR_WRITE_1(sc, B0_POWER_CTRL,
		    PC_VAUX_ENA | PC_VCC_ENA | PC_VAUX_ON | PC_VCC_OFF);
		break;
	default:
		break;
	}
}

static void
mskc_reset(struct msk_softc *sc)
{
	bus_addr_t addr;
	uint16_t status;
	uint32_t val;
	int i;

	CSR_WRITE_2(sc, B0_CTST, CS_RST_CLR);

	/* Disable ASF. */
	if (sc->msk_hw_id < CHIP_ID_YUKON_XL) {
		CSR_WRITE_4(sc, B28_Y2_ASF_STAT_CMD, Y2_ASF_RESET);
		CSR_WRITE_2(sc, B0_CTST, Y2_ASF_DISABLE);
	}
	/*
	 * Since we disabled ASF, S/W reset is required for Power Management.
	 */
	CSR_WRITE_2(sc, B0_CTST, CS_RST_SET);
	CSR_WRITE_2(sc, B0_CTST, CS_RST_CLR);

	/* Clear all error bits in the PCI status register. */
	status = pci_read_config(sc->msk_dev, PCIR_STATUS, 2);
	CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_ON);

	pci_write_config(sc->msk_dev, PCIR_STATUS, status |
	    PCIM_STATUS_PERR | PCIM_STATUS_SERR | PCIM_STATUS_RMABORT |
	    PCIM_STATUS_RTABORT | PCIM_STATUS_PERRREPORT, 2);
	CSR_WRITE_2(sc, B0_CTST, CS_MRST_CLR);

	switch (sc->msk_bustype) {
	case MSK_PEX_BUS:
		/* Clear all PEX errors. */
		CSR_PCI_WRITE_4(sc, PEX_UNC_ERR_STAT, 0xffffffff);
		val = CSR_PCI_READ_4(sc, PEX_UNC_ERR_STAT);
		if ((val & PEX_RX_OV) != 0) {
			sc->msk_intrmask &= ~Y2_IS_HW_ERR;
			sc->msk_intrhwemask &= ~Y2_IS_PCI_EXP;
		}
		break;
	case MSK_PCI_BUS:
	case MSK_PCIX_BUS:
		/* Set Cache Line Size to 2(8bytes) if configured to 0. */
		val = pci_read_config(sc->msk_dev, PCIR_CACHELNSZ, 1);
		if (val == 0)
			pci_write_config(sc->msk_dev, PCIR_CACHELNSZ, 2, 1);
		if (sc->msk_bustype == MSK_PCIX_BUS) {
			/* Set Cache Line Size opt. */
			val = pci_read_config(sc->msk_dev, PCI_OUR_REG_1, 4);
			val |= PCI_CLS_OPT;
			pci_write_config(sc->msk_dev, PCI_OUR_REG_1, val, 4);
		}
		break;
	}
	/* Set PHY power state. */
	mskc_phy_power(sc, MSK_PHY_POWERUP);

	/* Reset GPHY/GMAC Control */
	for (i = 0; i < sc->msk_num_port; i++) {
		/* GPHY Control reset. */
		CSR_WRITE_4(sc, MR_ADDR(i, GPHY_CTRL), GPC_RST_SET);
		CSR_WRITE_4(sc, MR_ADDR(i, GPHY_CTRL), GPC_RST_CLR);
		/* GMAC Control reset. */
		CSR_WRITE_4(sc, MR_ADDR(i, GMAC_CTRL), GMC_RST_SET);
		CSR_WRITE_4(sc, MR_ADDR(i, GMAC_CTRL), GMC_RST_CLR);
		CSR_WRITE_4(sc, MR_ADDR(i, GMAC_CTRL), GMC_F_LOOPB_OFF);
	}
	CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_OFF);

	/* LED On. */
	CSR_WRITE_2(sc, B0_CTST, Y2_LED_STAT_ON);

	/* Clear TWSI IRQ. */
	CSR_WRITE_4(sc, B2_I2C_IRQ, I2C_CLR_IRQ);

	/* Turn off hardware timer. */
	CSR_WRITE_1(sc, B2_TI_CTRL, TIM_STOP);
	CSR_WRITE_1(sc, B2_TI_CTRL, TIM_CLR_IRQ);

	/* Turn off descriptor polling. */
	CSR_WRITE_1(sc, B28_DPT_CTRL, DPT_STOP);

	/* Turn off time stamps. */
	CSR_WRITE_1(sc, GMAC_TI_ST_CTRL, GMT_ST_STOP);
	CSR_WRITE_1(sc, GMAC_TI_ST_CTRL, GMT_ST_CLR_IRQ);

	/* Configure timeout values. */
	for (i = 0; i < sc->msk_num_port; i++) {
		CSR_WRITE_2(sc, SELECT_RAM_BUFFER(i, B3_RI_CTRL), RI_RST_SET);
		CSR_WRITE_2(sc, SELECT_RAM_BUFFER(i, B3_RI_CTRL), RI_RST_CLR);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_R1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_XA1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_XS1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_R1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_XA1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_XS1),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_R2),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_XA2),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_WTO_XS2),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_R2),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_XA2),
		    MSK_RI_TO_53);
		CSR_WRITE_1(sc, SELECT_RAM_BUFFER(i, B3_RI_RTO_XS2),
		    MSK_RI_TO_53);
	}

	/* Disable all interrupts. */
	CSR_WRITE_4(sc, B0_HWE_IMSK, 0);
	CSR_READ_4(sc, B0_HWE_IMSK);
	CSR_WRITE_4(sc, B0_IMSK, 0);
	CSR_READ_4(sc, B0_IMSK);

        /*
         * On dual port PCI-X card, there is an problem where status
         * can be received out of order due to split transactions.
         */
	if (sc->msk_bustype == MSK_PCIX_BUS && sc->msk_num_port > 1) {
		uint16_t pcix_cmd;
		uint8_t pcix;

		pcix = pci_get_pcixcap_ptr(sc->msk_dev);

		pcix_cmd = pci_read_config(sc->msk_dev, pcix + 2, 2);
		/* Clear Max Outstanding Split Transactions. */
		pcix_cmd &= ~0x70;
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_ON);
		pci_write_config(sc->msk_dev, pcix + 2, pcix_cmd, 2);
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_OFF);
        }
	if (sc->msk_bustype == MSK_PEX_BUS) {
		uint16_t v, width;

		v = pci_read_config(sc->msk_dev, PEX_DEV_CTRL, 2);
		/* Change Max. Read Request Size to 4096 bytes. */
		v &= ~PEX_DC_MAX_RRS_MSK;
		v |= PEX_DC_MAX_RD_RQ_SIZE(5);
		pci_write_config(sc->msk_dev, PEX_DEV_CTRL, v, 2);
		width = pci_read_config(sc->msk_dev, PEX_LNK_STAT, 2);
		width = (width & PEX_LS_LINK_WI_MSK) >> 4;
		v = pci_read_config(sc->msk_dev, PEX_LNK_CAP, 2);
		v = (v & PEX_LS_LINK_WI_MSK) >> 4;
		if (v != width) {
			device_printf(sc->msk_dev,
			    "negotiated width of link(x%d) != "
			    "max. width of link(x%d)\n", width, v);
		}
	}

	/* Clear status list. */
	bzero(sc->msk_stat_ring,
	    sizeof(struct msk_stat_desc) * MSK_STAT_RING_CNT);
	sc->msk_stat_cons = 0;
	CSR_WRITE_4(sc, STAT_CTRL, SC_STAT_RST_SET);
	CSR_WRITE_4(sc, STAT_CTRL, SC_STAT_RST_CLR);
	/* Set the status list base address. */
	addr = sc->msk_stat_ring_paddr;
	CSR_WRITE_4(sc, STAT_LIST_ADDR_LO, MSK_ADDR_LO(addr));
	CSR_WRITE_4(sc, STAT_LIST_ADDR_HI, MSK_ADDR_HI(addr));
	/* Set the status list last index. */
	CSR_WRITE_2(sc, STAT_LAST_IDX, MSK_STAT_RING_CNT - 1);
	if (sc->msk_hw_id == CHIP_ID_YUKON_EC &&
	    sc->msk_hw_rev == CHIP_REV_YU_EC_A1) {
		/* WA for dev. #4.3 */
		CSR_WRITE_2(sc, STAT_TX_IDX_TH, ST_TXTH_IDX_MASK);
		/* WA for dev. #4.18 */
		CSR_WRITE_1(sc, STAT_FIFO_WM, 0x21);
		CSR_WRITE_1(sc, STAT_FIFO_ISR_WM, 0x07);
	} else {
		CSR_WRITE_2(sc, STAT_TX_IDX_TH, 0x0a);
		CSR_WRITE_1(sc, STAT_FIFO_WM, 0x10);
		if (sc->msk_hw_id == CHIP_ID_YUKON_XL &&
		    sc->msk_hw_rev == CHIP_REV_YU_XL_A0)
			CSR_WRITE_1(sc, STAT_FIFO_ISR_WM, 0x04);
		else
			CSR_WRITE_1(sc, STAT_FIFO_ISR_WM, 0x10);
		CSR_WRITE_4(sc, STAT_ISR_TIMER_INI, 0x0190);
	}
	/*
	 * Use default value for STAT_ISR_TIMER_INI, STAT_LEV_TIMER_INI.
	 */
	CSR_WRITE_4(sc, STAT_TX_TIMER_INI, MSK_USECS(sc, 1000));

	/* Enable status unit. */
	CSR_WRITE_4(sc, STAT_CTRL, SC_STAT_OP_ON);

	CSR_WRITE_1(sc, STAT_TX_TIMER_CTRL, TIM_START);
	CSR_WRITE_1(sc, STAT_LEV_TIMER_CTRL, TIM_START);
	CSR_WRITE_1(sc, STAT_ISR_TIMER_CTRL, TIM_START);
}

static int
msk_probe(device_t dev)
{
	struct msk_softc *sc = device_get_softc(device_get_parent(dev));
	char desc[100];

	/*
	 * Not much to do here. We always know there will be
	 * at least one GMAC present, and if there are two,
	 * mskc_attach() will create a second device instance
	 * for us.
	 */
	ksnprintf(desc, sizeof(desc),
	    "Marvell Technology Group Ltd. %s Id 0x%02x Rev 0x%02x",
	    model_name[sc->msk_hw_id - CHIP_ID_YUKON_XL], sc->msk_hw_id,
	    sc->msk_hw_rev);
	device_set_desc_copy(dev, desc);

	return (0);
}

static int
msk_attach(device_t dev)
{
	struct msk_softc *sc = device_get_softc(device_get_parent(dev));
	struct msk_if_softc *sc_if = device_get_softc(dev);
	struct ifnet *ifp = &sc_if->arpcom.ac_if;
	int i, port, error;
	uint8_t eaddr[ETHER_ADDR_LEN];

	port = *(int *)device_get_ivars(dev);
	KKASSERT(port == MSK_PORT_A || port == MSK_PORT_B);

	kfree(device_get_ivars(dev), M_DEVBUF);
	device_set_ivars(dev, NULL);

	callout_init(&sc_if->msk_tick_ch);
	if_initname(ifp, device_get_name(dev), device_get_unit(dev));

	sc_if->msk_if_dev = dev;
	sc_if->msk_port = port;
	sc_if->msk_softc = sc;
	sc_if->msk_ifp = ifp;
	sc->msk_if[port] = sc_if;

	/* Setup Tx/Rx queue register offsets. */
	if (port == MSK_PORT_A) {
		sc_if->msk_txq = Q_XA1;
		sc_if->msk_txsq = Q_XS1;
		sc_if->msk_rxq = Q_R1;
	} else {
		sc_if->msk_txq = Q_XA2;
		sc_if->msk_txsq = Q_XS2;
		sc_if->msk_rxq = Q_R2;
	}

	error = msk_txrx_dma_alloc(sc_if);
	if (error)
		goto fail;

	ifp->if_softc = sc_if;
	ifp->if_mtu = ETHERMTU;
	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
	ifp->if_init = msk_init;
	ifp->if_ioctl = msk_ioctl;
	ifp->if_start = msk_start;
	ifp->if_watchdog = msk_watchdog;
	ifq_set_maxlen(&ifp->if_snd, MSK_TX_RING_CNT - 1);
	ifq_set_ready(&ifp->if_snd);

#ifdef notyet
	/*
	 * IFCAP_RXCSUM capability is intentionally disabled as the hardware
	 * has serious bug in Rx checksum offload for all Yukon II family
	 * hardware. It seems there is a workaround to make it work somtimes.
	 * However, the workaround also have to check OP code sequences to
	 * verify whether the OP code is correct. Sometimes it should compute
	 * IP/TCP/UDP checksum in driver in order to verify correctness of
	 * checksum computed by hardware. If you have to compute checksum
	 * with software to verify the hardware's checksum why have hardware
	 * compute the checksum? I think there is no reason to spend time to
	 * make Rx checksum offload work on Yukon II hardware.
	 */
	ifp->if_capabilities = IFCAP_TXCSUM | IFCAP_VLAN_MTU |
			       IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM;
	ifp->if_hwassist = MSK_CSUM_FEATURES;
	ifp->if_capenable = ifp->if_capabilities;
#endif

	/*
	 * Get station address for this interface. Note that
	 * dual port cards actually come with three station
	 * addresses: one for each port, plus an extra. The
	 * extra one is used by the SysKonnect driver software
	 * as a 'virtual' station address for when both ports
	 * are operating in failover mode. Currently we don't
	 * use this extra address.
	 */
	for (i = 0; i < ETHER_ADDR_LEN; i++)
		eaddr[i] = CSR_READ_1(sc, B2_MAC_1 + (port * 8) + i);

	sc_if->msk_framesize = ifp->if_mtu + ETHER_HDR_LEN + EVL_ENCAPLEN;

	/*
	 * Do miibus setup.
	 */
	error = mii_phy_probe(dev, &sc_if->msk_miibus,
			      msk_mediachange, msk_mediastatus);
	if (error) {
		device_printf(sc_if->msk_if_dev, "no PHY found!\n");
		goto fail;
	}

	/*
	 * Call MI attach routine.  Can't hold locks when calling into ether_*.
	 */
	ether_ifattach(ifp, eaddr, &sc->msk_serializer);
#if 0
	/*
	 * Tell the upper layer(s) we support long frames.
	 * Must appear after the call to ether_ifattach() because
	 * ether_ifattach() sets ifi_hdrlen to the default value.
	 */
        ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
#endif

	return 0;
fail:
	msk_detach(dev);
	sc->msk_if[port] = NULL;
	return (error);
}

/*
 * Attach the interface. Allocate softc structures, do ifmedia
 * setup and ethernet/BPF attach.
 */
static int
mskc_attach(device_t dev)
{
	struct msk_softc *sc;
	int error, *port, cpuid;

	sc = device_get_softc(dev);
	sc->msk_dev = dev;
	lwkt_serialize_init(&sc->msk_serializer);

	/*
	 * Initailize sysctl variables
	 */
	sc->msk_process_limit = mskc_process_limit;
	sc->msk_intr_rate = mskc_intr_rate;

#ifndef BURN_BRIDGES
	/*
	 * Handle power management nonsense.
	 */
	if (pci_get_powerstate(dev) != PCI_POWERSTATE_D0) {
		uint32_t irq, bar0, bar1;

		/* Save important PCI config data. */
		bar0 = pci_read_config(dev, PCIR_BAR(0), 4);
		bar1 = pci_read_config(dev, PCIR_BAR(1), 4);
		irq = pci_read_config(dev, PCIR_INTLINE, 4);

		/* Reset the power state. */
		device_printf(dev, "chip is in D%d power mode "
			      "-- setting to D0\n", pci_get_powerstate(dev));

		pci_set_powerstate(dev, PCI_POWERSTATE_D0);

		/* Restore PCI config data. */
		pci_write_config(dev, PCIR_BAR(0), bar0, 4);
		pci_write_config(dev, PCIR_BAR(1), bar1, 4);
		pci_write_config(dev, PCIR_INTLINE, irq, 4);
	}
#endif	/* BURN_BRIDGES */

	/*
	 * Map control/status registers.
	 */
	pci_enable_busmaster(dev);

	/*
	 * Allocate I/O resource
	 */
#ifdef MSK_USEIOSPACE
	sc->msk_res_type = SYS_RES_IOPORT;
	sc->msk_res_rid = PCIR_BAR(1);
#else
	sc->msk_res_type = SYS_RES_MEMORY;
	sc->msk_res_rid = PCIR_BAR(0);
#endif
	sc->msk_res = bus_alloc_resource_any(dev, sc->msk_res_type,
					     &sc->msk_res_rid, RF_ACTIVE);
	if (sc->msk_res == NULL) {
		if (sc->msk_res_type == SYS_RES_MEMORY) {
			sc->msk_res_type = SYS_RES_IOPORT;
			sc->msk_res_rid = PCIR_BAR(1);
		} else {
			sc->msk_res_type = SYS_RES_MEMORY;
			sc->msk_res_rid = PCIR_BAR(0);
		}
		sc->msk_res = bus_alloc_resource_any(dev, sc->msk_res_type,
						     &sc->msk_res_rid,
						     RF_ACTIVE);
		if (sc->msk_res == NULL) {
			device_printf(dev, "couldn't allocate %s resources\n",
			sc->msk_res_type == SYS_RES_MEMORY ? "memory" : "I/O");
			return (ENXIO);
		}
	}
	sc->msk_res_bt = rman_get_bustag(sc->msk_res);
	sc->msk_res_bh = rman_get_bushandle(sc->msk_res);

	/*
	 * Allocate IRQ
	 */
	sc->msk_irq_rid = 0;
	sc->msk_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ,
					     &sc->msk_irq_rid,
					     RF_SHAREABLE | RF_ACTIVE);
	if (sc->msk_irq == NULL) {
		device_printf(dev, "couldn't allocate IRQ resources\n");
		error = ENXIO;
		goto fail;
	}

	CSR_WRITE_2(sc, B0_CTST, CS_RST_CLR);
	sc->msk_hw_id = CSR_READ_1(sc, B2_CHIP_ID);
	sc->msk_hw_rev = (CSR_READ_1(sc, B2_MAC_CFG) >> 4) & 0x0f;
	/* Bail out if chip is not recognized. */
	if (sc->msk_hw_id < CHIP_ID_YUKON_XL ||
	    sc->msk_hw_id > CHIP_ID_YUKON_FE) {
		device_printf(dev, "unknown device: id=0x%02x, rev=0x%02x\n",
		    sc->msk_hw_id, sc->msk_hw_rev);
		error = ENXIO;
		goto fail;
	}

	/*
	 * Create sysctl tree
	 */
	sysctl_ctx_init(&sc->msk_sysctl_ctx);
	sc->msk_sysctl_tree = SYSCTL_ADD_NODE(&sc->msk_sysctl_ctx,
					      SYSCTL_STATIC_CHILDREN(_hw),
					      OID_AUTO,
					      device_get_nameunit(dev),
					      CTLFLAG_RD, 0, "");
	if (sc->msk_sysctl_tree == NULL) {
		device_printf(dev, "can't add sysctl node\n");
		error = ENXIO;
		goto fail;
	}

	SYSCTL_ADD_PROC(&sc->msk_sysctl_ctx,
			SYSCTL_CHILDREN(sc->msk_sysctl_tree),
			OID_AUTO, "process_limit", CTLTYPE_INT | CTLFLAG_RW,
			&sc->msk_process_limit, 0, mskc_sysctl_proc_limit,
			"I", "max number of Rx events to process");
	SYSCTL_ADD_PROC(&sc->msk_sysctl_ctx,
			SYSCTL_CHILDREN(sc->msk_sysctl_tree),
			OID_AUTO, "intr_rate", CTLTYPE_INT | CTLFLAG_RW,
			sc, 0, mskc_sysctl_intr_rate,
			"I", "max number of interrupt per second");

	/* Soft reset. */
	CSR_WRITE_2(sc, B0_CTST, CS_RST_SET);
	CSR_WRITE_2(sc, B0_CTST, CS_RST_CLR);
	sc->msk_pmd = CSR_READ_1(sc, B2_PMD_TYP);
	if (sc->msk_pmd == 'L' || sc->msk_pmd == 'S')
		sc->msk_coppertype = 0;
	else
		sc->msk_coppertype = 1;
	/* Check number of MACs. */
	sc->msk_num_port = 1;
	if ((CSR_READ_1(sc, B2_Y2_HW_RES) & CFG_DUAL_MAC_MSK) ==
	    CFG_DUAL_MAC_MSK) {
		if (!(CSR_READ_1(sc, B2_Y2_CLK_GATE) & Y2_STATUS_LNK2_INAC))
			sc->msk_num_port++;
	}

	/* Check bus type. */
	if (pci_is_pcie(sc->msk_dev) == 0)
		sc->msk_bustype = MSK_PEX_BUS;
	else if (pci_is_pcix(sc->msk_dev) == 0)
		sc->msk_bustype = MSK_PCIX_BUS;
	else
		sc->msk_bustype = MSK_PCI_BUS;

	switch (sc->msk_hw_id) {
	case CHIP_ID_YUKON_EC:
	case CHIP_ID_YUKON_EC_U:
		sc->msk_clock = 125;	/* 125 Mhz */
		break;
	case CHIP_ID_YUKON_FE:
		sc->msk_clock = 100;	/* 100 Mhz */
		break;
	case CHIP_ID_YUKON_XL:
		sc->msk_clock = 156;	/* 156 Mhz */
		break;
	default:
		sc->msk_clock = 156;	/* 156 Mhz */
		break;
	}

	error = mskc_status_dma_alloc(sc);
	if (error)
		goto fail;

	/* Set base interrupt mask. */
	sc->msk_intrmask = Y2_IS_HW_ERR | Y2_IS_STAT_BMU;
	sc->msk_intrhwemask = Y2_IS_TIST_OV | Y2_IS_MST_ERR |
	    Y2_IS_IRQ_STAT | Y2_IS_PCI_EXP | Y2_IS_PCI_NEXP;

	/* Reset the adapter. */
	mskc_reset(sc);

	error = mskc_setup_rambuffer(sc);
	if (error)
		goto fail;

	sc->msk_devs[MSK_PORT_A] = device_add_child(dev, "msk", -1);
	if (sc->msk_devs[MSK_PORT_A] == NULL) {
		device_printf(dev, "failed to add child for PORT_A\n");
		error = ENXIO;
		goto fail;
	}
	port = kmalloc(sizeof(*port), M_DEVBUF, M_WAITOK);
	*port = MSK_PORT_A;
	device_set_ivars(sc->msk_devs[MSK_PORT_A], port);

	if (sc->msk_num_port > 1) {
		sc->msk_devs[MSK_PORT_B] = device_add_child(dev, "msk", -1);
		if (sc->msk_devs[MSK_PORT_B] == NULL) {
			device_printf(dev, "failed to add child for PORT_B\n");
			error = ENXIO;
			goto fail;
		}
		port = kmalloc(sizeof(*port), M_DEVBUF, M_WAITOK);
		*port = MSK_PORT_B;
		device_set_ivars(sc->msk_devs[MSK_PORT_B], port);
	}

	bus_generic_attach(dev);

	error = bus_setup_intr(dev, sc->msk_irq, INTR_MPSAFE,
			       mskc_intr, sc, &sc->msk_intrhand,
			       &sc->msk_serializer);
	if (error) {
		device_printf(dev, "couldn't set up interrupt handler\n");
		goto fail;
	}

	cpuid = ithread_cpuid(rman_get_start(sc->msk_irq));
	KKASSERT(cpuid >= 0 && cpuid < ncpus);

	if (sc->msk_if[0] != NULL)
		sc->msk_if[0]->msk_ifp->if_cpuid = cpuid;
	if (sc->msk_if[1] != NULL)
		sc->msk_if[1]->msk_ifp->if_cpuid = cpuid;
	return 0;
fail:
	mskc_detach(dev);
	return (error);
}

/*
 * Shutdown hardware and free up resources. This can be called any
 * time after the mutex has been initialized. It is called in both
 * the error case in attach and the normal detach case so it needs
 * to be careful about only freeing resources that have actually been
 * allocated.
 */
static int
msk_detach(device_t dev)
{
	struct msk_if_softc *sc_if = device_get_softc(dev);

	if (device_is_attached(dev)) {
		struct msk_softc *sc = sc_if->msk_softc;
		struct ifnet *ifp = &sc_if->arpcom.ac_if;

		lwkt_serialize_enter(ifp->if_serializer);

		if (sc->msk_intrhand != NULL) {
			if (sc->msk_if[MSK_PORT_A] != NULL)
				msk_stop(sc->msk_if[MSK_PORT_A]);
			if (sc->msk_if[MSK_PORT_B] != NULL)
				msk_stop(sc->msk_if[MSK_PORT_B]);

			bus_teardown_intr(sc->msk_dev, sc->msk_irq,
					  sc->msk_intrhand);
			sc->msk_intrhand = NULL;
		}

		lwkt_serialize_exit(ifp->if_serializer);

		ether_ifdetach(ifp);
	}

	if (sc_if->msk_miibus != NULL)
		device_delete_child(dev, sc_if->msk_miibus);

	msk_txrx_dma_free(sc_if);
	return (0);
}

static int
mskc_detach(device_t dev)
{
	struct msk_softc *sc = device_get_softc(dev);
	int *port, i;

#ifdef INVARIANTS
	if (device_is_attached(dev)) {
		KASSERT(sc->msk_intrhand == NULL,
			("intr is not torn down yet\n"));
	}
#endif

	for (i = 0; i < sc->msk_num_port; ++i) {
		if (sc->msk_devs[i] != NULL) {
			port = device_get_ivars(sc->msk_devs[i]);
			if (port != NULL) {
				kfree(port, M_DEVBUF);
				device_set_ivars(sc->msk_devs[i], NULL);
			}
			device_delete_child(dev, sc->msk_devs[i]);
		}
	}

	/* Disable all interrupts. */
	CSR_WRITE_4(sc, B0_IMSK, 0);
	CSR_READ_4(sc, B0_IMSK);
	CSR_WRITE_4(sc, B0_HWE_IMSK, 0);
	CSR_READ_4(sc, B0_HWE_IMSK);

	/* LED Off. */
	CSR_WRITE_2(sc, B0_CTST, Y2_LED_STAT_OFF);

	/* Put hardware reset. */
	CSR_WRITE_2(sc, B0_CTST, CS_RST_SET);

	mskc_status_dma_free(sc);

	if (sc->msk_irq != NULL) {
		bus_release_resource(dev, SYS_RES_IRQ, sc->msk_irq_rid,
				     sc->msk_irq);
	}
	if (sc->msk_res != NULL) {
		bus_release_resource(dev, sc->msk_res_type, sc->msk_res_rid,
				     sc->msk_res);
	}

	if (sc->msk_sysctl_tree != NULL)
		sysctl_ctx_free(&sc->msk_sysctl_ctx);

	return (0);
}

static void
msk_dmamap_mbuf_cb(void *arg, bus_dma_segment_t *segs, int nseg,
		   bus_size_t mapsz __unused, int error)
{
	struct msk_dmamap_arg *ctx = arg;
	int i;

	if (error)
		return;

	if (ctx->nseg < nseg) {
		ctx->nseg = 0;
		return;
	}

	ctx->nseg = nseg;
	for (i = 0; i < ctx->nseg; ++i)
		ctx->segs[i] = segs[i];
}

/* Create status DMA region. */
static int
mskc_status_dma_alloc(struct msk_softc *sc)
{
	bus_dmamem_t dmem;
	int error;

	error = bus_dmamem_coherent(NULL/* XXX parent */, MSK_STAT_ALIGN, 0,
			BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR,
			MSK_STAT_RING_SZ, BUS_DMA_WAITOK | BUS_DMA_ZERO, &dmem);
	if (error) {
		device_printf(sc->msk_dev,
		    "failed to create status coherent DMA memory\n");
		return error;
	}
	sc->msk_stat_tag = dmem.dmem_tag;
	sc->msk_stat_map = dmem.dmem_map;
	sc->msk_stat_ring = dmem.dmem_addr;
	sc->msk_stat_ring_paddr = dmem.dmem_busaddr;

	return (0);
}

static void
mskc_status_dma_free(struct msk_softc *sc)
{
	/* Destroy status block. */
	if (sc->msk_stat_tag) {
		bus_dmamap_unload(sc->msk_stat_tag, sc->msk_stat_map);
		bus_dmamem_free(sc->msk_stat_tag, sc->msk_stat_ring,
				sc->msk_stat_map);
		bus_dma_tag_destroy(sc->msk_stat_tag);
		sc->msk_stat_tag = NULL;
	}
}

static int
msk_txrx_dma_alloc(struct msk_if_softc *sc_if)
{
	int error, i, j;
#ifdef MSK_JUMBO
	struct msk_rxdesc *jrxd;
	struct msk_jpool_entry *entry;
	uint8_t *ptr;
#endif

	/* Create parent DMA tag. */
	/*
	 * XXX
	 * It seems that Yukon II supports full 64bits DMA operations. But
	 * it needs two descriptors(list elements) for 64bits DMA operations.
	 * Since we don't know what DMA address mappings(32bits or 64bits)
	 * would be used in advance for each mbufs, we limits its DMA space
	 * to be in range of 32bits address space. Otherwise, we should check
	 * what DMA address is used and chain another descriptor for the
	 * 64bits DMA operation. This also means descriptor ring size is
	 * variable. Limiting DMA address to be in 32bit address space greatly
	 * simplyfies descriptor handling and possibly would increase
	 * performance a bit due to efficient handling of descriptors.
	 * Apart from harassing checksum offloading mechanisms, it seems
	 * it's really bad idea to use a seperate descriptor for 64bit
	 * DMA operation to save small descriptor memory. Anyway, I've
	 * never seen these exotic scheme on ethernet interface hardware.
	 */
	error = bus_dma_tag_create(
		    NULL,			/* parent */
		    1, 0,			/* alignment, boundary */
		    BUS_SPACE_MAXADDR_32BIT,	/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    BUS_SPACE_MAXSIZE_32BIT,	/* maxsize */
		    0,				/* nsegments */
		    BUS_SPACE_MAXSIZE_32BIT,	/* maxsegsize */
		    0,				/* flags */
		    &sc_if->msk_cdata.msk_parent_tag);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create parent DMA tag\n");
		return error;
	}

	/* Create DMA stuffs for Tx ring. */
	error = msk_dmamem_create(sc_if->msk_if_dev, MSK_TX_RING_SZ,
				  &sc_if->msk_cdata.msk_tx_ring_tag,
				  (void **)&sc_if->msk_rdata.msk_tx_ring,
				  &sc_if->msk_rdata.msk_tx_ring_paddr,
				  &sc_if->msk_cdata.msk_tx_ring_map);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create TX ring DMA stuffs\n");
		return error;
	}

	/* Create DMA stuffs for Rx ring. */
	error = msk_dmamem_create(sc_if->msk_if_dev, MSK_RX_RING_SZ,
				  &sc_if->msk_cdata.msk_rx_ring_tag,
				  (void **)&sc_if->msk_rdata.msk_rx_ring,
				  &sc_if->msk_rdata.msk_rx_ring_paddr,
				  &sc_if->msk_cdata.msk_rx_ring_map);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create RX ring DMA stuffs\n");
		return error;
	}

	/* Create tag for Tx buffers. */
	error = bus_dma_tag_create(sc_if->msk_cdata.msk_parent_tag,/* parent */
		    1, 0,			/* alignment, boundary */
		    BUS_SPACE_MAXADDR,		/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    MSK_JUMBO_FRAMELEN,		/* maxsize */
		    MSK_MAXTXSEGS,		/* nsegments */
		    MSK_MAXSGSIZE,		/* maxsegsize */
		    BUS_DMA_ALLOCNOW | BUS_DMA_WAITOK |
		    BUS_DMA_ONEBPAGE,		/* flags */
		    &sc_if->msk_cdata.msk_tx_tag);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create Tx DMA tag\n");
		return error;
	}

	/* Create DMA maps for Tx buffers. */
	for (i = 0; i < MSK_TX_RING_CNT; i++) {
		struct msk_txdesc *txd = &sc_if->msk_cdata.msk_txdesc[i];

		error = bus_dmamap_create(sc_if->msk_cdata.msk_tx_tag,
				BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE,
				&txd->tx_dmamap);
		if (error) {
			device_printf(sc_if->msk_if_dev,
				      "failed to create %dth Tx dmamap\n", i);

			for (j = 0; j < i; ++j) {
				txd = &sc_if->msk_cdata.msk_txdesc[j];
				bus_dmamap_destroy(sc_if->msk_cdata.msk_tx_tag,
						   txd->tx_dmamap);
			}
			bus_dma_tag_destroy(sc_if->msk_cdata.msk_tx_tag);
			sc_if->msk_cdata.msk_tx_tag = NULL;

			return error;
		}
	}

	/* Create tag for Rx buffers. */
	error = bus_dma_tag_create(sc_if->msk_cdata.msk_parent_tag,/* parent */
		    1, 0,			/* alignment, boundary */
		    BUS_SPACE_MAXADDR,		/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    MCLBYTES,			/* maxsize */
		    1,				/* nsegments */
		    MCLBYTES,			/* maxsegsize */
		    BUS_DMA_ALLOCNOW | BUS_DMA_WAITOK,/* flags */
		    &sc_if->msk_cdata.msk_rx_tag);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create Rx DMA tag\n");
		return error;
	}

	/* Create DMA maps for Rx buffers. */
	error = bus_dmamap_create(sc_if->msk_cdata.msk_rx_tag, BUS_DMA_WAITOK,
				  &sc_if->msk_cdata.msk_rx_sparemap);
	if (error) {
		device_printf(sc_if->msk_if_dev,
			      "failed to create spare Rx dmamap\n");
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_rx_tag);
		sc_if->msk_cdata.msk_rx_tag = NULL;
		return error;
	}
	for (i = 0; i < MSK_RX_RING_CNT; i++) {
		struct msk_rxdesc *rxd = &sc_if->msk_cdata.msk_rxdesc[i];

		error = bus_dmamap_create(sc_if->msk_cdata.msk_rx_tag,
					  BUS_DMA_WAITOK, &rxd->rx_dmamap);
		if (error) {
			device_printf(sc_if->msk_if_dev,
				      "failed to create %dth Rx dmamap\n", i);

			for (j = 0; j < i; ++j) {
				rxd = &sc_if->msk_cdata.msk_rxdesc[j];
				bus_dmamap_destroy(sc_if->msk_cdata.msk_rx_tag,
						   rxd->rx_dmamap);
			}
			bus_dma_tag_destroy(sc_if->msk_cdata.msk_rx_tag);
			sc_if->msk_cdata.msk_rx_tag = NULL;

			return error;
		}
	}

#ifdef MSK_JUMBO
	SLIST_INIT(&sc_if->msk_jfree_listhead);
	SLIST_INIT(&sc_if->msk_jinuse_listhead);

	/* Create tag for jumbo Rx ring. */
	error = bus_dma_tag_create(sc_if->msk_cdata.msk_parent_tag,/* parent */
		    MSK_RING_ALIGN, 0,		/* alignment, boundary */
		    BUS_SPACE_MAXADDR,		/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    MSK_JUMBO_RX_RING_SZ,	/* maxsize */
		    1,				/* nsegments */
		    MSK_JUMBO_RX_RING_SZ,	/* maxsegsize */
		    0,				/* flags */
		    NULL, NULL,			/* lockfunc, lockarg */
		    &sc_if->msk_cdata.msk_jumbo_rx_ring_tag);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to create jumbo Rx ring DMA tag\n");
		goto fail;
	}

	/* Allocate DMA'able memory and load the DMA map for jumbo Rx ring. */
	error = bus_dmamem_alloc(sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
	    (void **)&sc_if->msk_rdata.msk_jumbo_rx_ring,
	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
	    &sc_if->msk_cdata.msk_jumbo_rx_ring_map);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to allocate DMA'able memory for jumbo Rx ring\n");
		goto fail;
	}

	ctx.msk_busaddr = 0;
	error = bus_dmamap_load(sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
	    sc_if->msk_cdata.msk_jumbo_rx_ring_map,
	    sc_if->msk_rdata.msk_jumbo_rx_ring, MSK_JUMBO_RX_RING_SZ,
	    msk_dmamap_cb, &ctx, 0);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to load DMA'able memory for jumbo Rx ring\n");
		goto fail;
	}
	sc_if->msk_rdata.msk_jumbo_rx_ring_paddr = ctx.msk_busaddr;

	/* Create tag for jumbo buffer blocks. */
	error = bus_dma_tag_create(sc_if->msk_cdata.msk_parent_tag,/* parent */
		    PAGE_SIZE, 0,		/* alignment, boundary */
		    BUS_SPACE_MAXADDR,		/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    MSK_JMEM,			/* maxsize */
		    1,				/* nsegments */
		    MSK_JMEM,			/* maxsegsize */
		    0,				/* flags */
		    NULL, NULL,			/* lockfunc, lockarg */
		    &sc_if->msk_cdata.msk_jumbo_tag);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to create jumbo Rx buffer block DMA tag\n");
		goto fail;
	}

	/* Create tag for jumbo Rx buffers. */
	error = bus_dma_tag_create(sc_if->msk_cdata.msk_parent_tag,/* parent */
		    PAGE_SIZE, 0,		/* alignment, boundary */
		    BUS_SPACE_MAXADDR,		/* lowaddr */
		    BUS_SPACE_MAXADDR,		/* highaddr */
		    NULL, NULL,			/* filter, filterarg */
		    MCLBYTES * MSK_MAXRXSEGS,	/* maxsize */
		    MSK_MAXRXSEGS,		/* nsegments */
		    MSK_JLEN,			/* maxsegsize */
		    0,				/* flags */
		    NULL, NULL,			/* lockfunc, lockarg */
		    &sc_if->msk_cdata.msk_jumbo_rx_tag);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to create jumbo Rx DMA tag\n");
		goto fail;
	}

	/* Create DMA maps for jumbo Rx buffers. */
	if ((error = bus_dmamap_create(sc_if->msk_cdata.msk_jumbo_rx_tag, 0,
	    &sc_if->msk_cdata.msk_jumbo_rx_sparemap)) != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to create spare jumbo Rx dmamap\n");
		goto fail;
	}
	for (i = 0; i < MSK_JUMBO_RX_RING_CNT; i++) {
		jrxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[i];
		jrxd->rx_m = NULL;
		jrxd->rx_dmamap = NULL;
		error = bus_dmamap_create(sc_if->msk_cdata.msk_jumbo_rx_tag, 0,
		    &jrxd->rx_dmamap);
		if (error != 0) {
			device_printf(sc_if->msk_if_dev,
			    "failed to create jumbo Rx dmamap\n");
			goto fail;
		}
	}

	/* Allocate DMA'able memory and load the DMA map for jumbo buf. */
	error = bus_dmamem_alloc(sc_if->msk_cdata.msk_jumbo_tag,
	    (void **)&sc_if->msk_rdata.msk_jumbo_buf,
	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
	    &sc_if->msk_cdata.msk_jumbo_map);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to allocate DMA'able memory for jumbo buf\n");
		goto fail;
	}

	ctx.msk_busaddr = 0;
	error = bus_dmamap_load(sc_if->msk_cdata.msk_jumbo_tag,
	    sc_if->msk_cdata.msk_jumbo_map, sc_if->msk_rdata.msk_jumbo_buf,
	    MSK_JMEM, msk_dmamap_cb, &ctx, 0);
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "failed to load DMA'able memory for jumbobuf\n");
		goto fail;
	}
	sc_if->msk_rdata.msk_jumbo_buf_paddr = ctx.msk_busaddr;

	/*
	 * Now divide it up into 9K pieces and save the addresses
	 * in an array.
	 */
	ptr = sc_if->msk_rdata.msk_jumbo_buf;
	for (i = 0; i < MSK_JSLOTS; i++) {
		sc_if->msk_cdata.msk_jslots[i] = ptr;
		ptr += MSK_JLEN;
		entry = malloc(sizeof(struct msk_jpool_entry),
		    M_DEVBUF, M_WAITOK);
		if (entry == NULL) {
			device_printf(sc_if->msk_if_dev,
			    "no memory for jumbo buffers!\n");
			error = ENOMEM;
			goto fail;
		}
		entry->slot = i;
		SLIST_INSERT_HEAD(&sc_if->msk_jfree_listhead, entry,
		    jpool_entries);
	}
#endif
	return 0;
}

static void
msk_txrx_dma_free(struct msk_if_softc *sc_if)
{
	struct msk_txdesc *txd;
	struct msk_rxdesc *rxd;
#ifdef MSK_JUMBO
	struct msk_rxdesc *jrxd;
	struct msk_jpool_entry *entry;
#endif
	int i;

#ifdef MSK_JUMBO
	MSK_JLIST_LOCK(sc_if);
	while ((entry = SLIST_FIRST(&sc_if->msk_jinuse_listhead))) {
		device_printf(sc_if->msk_if_dev,
		    "asked to free buffer that is in use!\n");
		SLIST_REMOVE_HEAD(&sc_if->msk_jinuse_listhead, jpool_entries);
		SLIST_INSERT_HEAD(&sc_if->msk_jfree_listhead, entry,
		    jpool_entries);
	}

	while (!SLIST_EMPTY(&sc_if->msk_jfree_listhead)) {
		entry = SLIST_FIRST(&sc_if->msk_jfree_listhead);
		SLIST_REMOVE_HEAD(&sc_if->msk_jfree_listhead, jpool_entries);
		free(entry, M_DEVBUF);
	}
	MSK_JLIST_UNLOCK(sc_if);

	/* Destroy jumbo buffer block. */
	if (sc_if->msk_cdata.msk_jumbo_map)
		bus_dmamap_unload(sc_if->msk_cdata.msk_jumbo_tag,
		    sc_if->msk_cdata.msk_jumbo_map);

	if (sc_if->msk_rdata.msk_jumbo_buf) {
		bus_dmamem_free(sc_if->msk_cdata.msk_jumbo_tag,
		    sc_if->msk_rdata.msk_jumbo_buf,
		    sc_if->msk_cdata.msk_jumbo_map);
		sc_if->msk_rdata.msk_jumbo_buf = NULL;
		sc_if->msk_cdata.msk_jumbo_map = NULL;
	}

	/* Jumbo Rx ring. */
	if (sc_if->msk_cdata.msk_jumbo_rx_ring_tag) {
		if (sc_if->msk_cdata.msk_jumbo_rx_ring_map)
			bus_dmamap_unload(sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
			    sc_if->msk_cdata.msk_jumbo_rx_ring_map);
		if (sc_if->msk_cdata.msk_jumbo_rx_ring_map &&
		    sc_if->msk_rdata.msk_jumbo_rx_ring)
			bus_dmamem_free(sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
			    sc_if->msk_rdata.msk_jumbo_rx_ring,
			    sc_if->msk_cdata.msk_jumbo_rx_ring_map);
		sc_if->msk_rdata.msk_jumbo_rx_ring = NULL;
		sc_if->msk_cdata.msk_jumbo_rx_ring_map = NULL;
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_jumbo_rx_ring_tag);
		sc_if->msk_cdata.msk_jumbo_rx_ring_tag = NULL;
	}

	/* Jumbo Rx buffers. */
	if (sc_if->msk_cdata.msk_jumbo_rx_tag) {
		for (i = 0; i < MSK_JUMBO_RX_RING_CNT; i++) {
			jrxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[i];
			if (jrxd->rx_dmamap) {
				bus_dmamap_destroy(
				    sc_if->msk_cdata.msk_jumbo_rx_tag,
				    jrxd->rx_dmamap);
				jrxd->rx_dmamap = NULL;
			}
		}
		if (sc_if->msk_cdata.msk_jumbo_rx_sparemap) {
			bus_dmamap_destroy(sc_if->msk_cdata.msk_jumbo_rx_tag,
			    sc_if->msk_cdata.msk_jumbo_rx_sparemap);
			sc_if->msk_cdata.msk_jumbo_rx_sparemap = 0;
		}
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_jumbo_rx_tag);
		sc_if->msk_cdata.msk_jumbo_rx_tag = NULL;
	}
#endif

	/* Tx ring. */
	msk_dmamem_destroy(sc_if->msk_cdata.msk_tx_ring_tag,
			   sc_if->msk_rdata.msk_tx_ring,
			   sc_if->msk_cdata.msk_tx_ring_map);

	/* Rx ring. */
	msk_dmamem_destroy(sc_if->msk_cdata.msk_rx_ring_tag,
			   sc_if->msk_rdata.msk_rx_ring,
			   sc_if->msk_cdata.msk_rx_ring_map);

	/* Tx buffers. */
	if (sc_if->msk_cdata.msk_tx_tag) {
		for (i = 0; i < MSK_TX_RING_CNT; i++) {
			txd = &sc_if->msk_cdata.msk_txdesc[i];
			bus_dmamap_destroy(sc_if->msk_cdata.msk_tx_tag,
					   txd->tx_dmamap);
		}
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_tx_tag);
		sc_if->msk_cdata.msk_tx_tag = NULL;
	}

	/* Rx buffers. */
	if (sc_if->msk_cdata.msk_rx_tag) {
		for (i = 0; i < MSK_RX_RING_CNT; i++) {
			rxd = &sc_if->msk_cdata.msk_rxdesc[i];
			bus_dmamap_destroy(sc_if->msk_cdata.msk_rx_tag,
					   rxd->rx_dmamap);
		}
		bus_dmamap_destroy(sc_if->msk_cdata.msk_rx_tag,
				   sc_if->msk_cdata.msk_rx_sparemap);
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_rx_tag);
		sc_if->msk_cdata.msk_rx_tag = NULL;
	}

	if (sc_if->msk_cdata.msk_parent_tag) {
		bus_dma_tag_destroy(sc_if->msk_cdata.msk_parent_tag);
		sc_if->msk_cdata.msk_parent_tag = NULL;
	}
}

#ifdef MSK_JUMBO
/*
 * Allocate a jumbo buffer.
 */
static void *
msk_jalloc(struct msk_if_softc *sc_if)
{
	struct msk_jpool_entry *entry;

	MSK_JLIST_LOCK(sc_if);

	entry = SLIST_FIRST(&sc_if->msk_jfree_listhead);

	if (entry == NULL) {
		MSK_JLIST_UNLOCK(sc_if);
		return (NULL);
	}

	SLIST_REMOVE_HEAD(&sc_if->msk_jfree_listhead, jpool_entries);
	SLIST_INSERT_HEAD(&sc_if->msk_jinuse_listhead, entry, jpool_entries);

	MSK_JLIST_UNLOCK(sc_if);

	return (sc_if->msk_cdata.msk_jslots[entry->slot]);
}

/*
 * Release a jumbo buffer.
 */
static void
msk_jfree(void *buf, void *args)
{
	struct msk_if_softc *sc_if;
	struct msk_jpool_entry *entry;
	int i;

	/* Extract the softc struct pointer. */
	sc_if = (struct msk_if_softc *)args;
	KASSERT(sc_if != NULL, ("%s: can't find softc pointer!", __func__));

	MSK_JLIST_LOCK(sc_if);
	/* Calculate the slot this buffer belongs to. */
	i = ((vm_offset_t)buf
	     - (vm_offset_t)sc_if->msk_rdata.msk_jumbo_buf) / MSK_JLEN;
	KASSERT(i >= 0 && i < MSK_JSLOTS,
	    ("%s: asked to free buffer that we don't manage!", __func__));

	entry = SLIST_FIRST(&sc_if->msk_jinuse_listhead);
	KASSERT(entry != NULL, ("%s: buffer not in use!", __func__));
	entry->slot = i;
	SLIST_REMOVE_HEAD(&sc_if->msk_jinuse_listhead, jpool_entries);
	SLIST_INSERT_HEAD(&sc_if->msk_jfree_listhead, entry, jpool_entries);
	if (SLIST_EMPTY(&sc_if->msk_jinuse_listhead))
		wakeup(sc_if);

	MSK_JLIST_UNLOCK(sc_if);
}
#endif

/*
 * It's copy of ath_defrag(ath(4)).
 *
 * Defragment an mbuf chain, returning at most maxfrags separate
 * mbufs+clusters.  If this is not possible NULL is returned and
 * the original mbuf chain is left in it's present (potentially
 * modified) state.  We use two techniques: collapsing consecutive
 * mbufs and replacing consecutive mbufs by a cluster.
 */
static struct mbuf *
msk_defrag(struct mbuf *m0, int how, int maxfrags)
{
	struct mbuf *m, *n, *n2, **prev;
	u_int curfrags;

	/*
	 * Calculate the current number of frags.
	 */
	curfrags = 0;
	for (m = m0; m != NULL; m = m->m_next)
		curfrags++;
	/*
	 * First, try to collapse mbufs.  Note that we always collapse
	 * towards the front so we don't need to deal with moving the
	 * pkthdr.  This may be suboptimal if the first mbuf has much
	 * less data than the following.
	 */
	m = m0;
again:
	for (;;) {
		n = m->m_next;
		if (n == NULL)
			break;
		if (n->m_len < M_TRAILINGSPACE(m)) {
			bcopy(mtod(n, void *), mtod(m, char *) + m->m_len,
				n->m_len);
			m->m_len += n->m_len;
			m->m_next = n->m_next;
			m_free(n);
			if (--curfrags <= maxfrags)
				return (m0);
		} else
			m = n;
	}
	KASSERT(maxfrags > 1,
		("maxfrags %u, but normal collapse failed", maxfrags));
	/*
	 * Collapse consecutive mbufs to a cluster.
	 */
	prev = &m0->m_next;		/* NB: not the first mbuf */
	while ((n = *prev) != NULL) {
		if ((n2 = n->m_next) != NULL &&
		    n->m_len + n2->m_len < MCLBYTES) {
			m = m_getcl(how, MT_DATA, 0);
			if (m == NULL)
				goto bad;
			bcopy(mtod(n, void *), mtod(m, void *), n->m_len);
			bcopy(mtod(n2, void *), mtod(m, char *) + n->m_len,
				n2->m_len);
			m->m_len = n->m_len + n2->m_len;
			m->m_next = n2->m_next;
			*prev = m;
			m_free(n);
			m_free(n2);
			if (--curfrags <= maxfrags)	/* +1 cl -2 mbufs */
				return m0;
			/*
			 * Still not there, try the normal collapse
			 * again before we allocate another cluster.
			 */
			goto again;
		}
		prev = &n->m_next;
	}
	/*
	 * No place where we can collapse to a cluster; punt.
	 * This can occur if, for example, you request 2 frags
	 * but the packet requires that both be clusters (we
	 * never reallocate the first mbuf to avoid moving the
	 * packet header).
	 */
bad:
	return (NULL);
}

static int
msk_encap(struct msk_if_softc *sc_if, struct mbuf **m_head)
{
	struct msk_txdesc *txd, *txd_last;
	struct msk_tx_desc *tx_le;
	struct mbuf *m;
	bus_dmamap_t map;
	struct msk_dmamap_arg ctx;
	bus_dma_segment_t txsegs[MSK_MAXTXSEGS];
	uint32_t control, prod, si;
	uint16_t offset, tcp_offset;
	int error, i;

	tcp_offset = offset = 0;
	m = *m_head;
	if (m->m_pkthdr.csum_flags & MSK_CSUM_FEATURES) {
		/*
		 * Since mbuf has no protocol specific structure information
		 * in it we have to inspect protocol information here to
		 * setup TSO and checksum offload. I don't know why Marvell
		 * made a such decision in chip design because other GigE
		 * hardwares normally takes care of all these chores in
		 * hardware. However, TSO performance of Yukon II is very
		 * good such that it's worth to implement it.
		 */
		struct ether_header *eh;
		struct ip *ip;

		/* TODO check for M_WRITABLE(m) */

		offset = sizeof(struct ether_header);
		m = m_pullup(m, offset);
		if (m == NULL) {
			*m_head = NULL;
			return (ENOBUFS);
		}
		eh = mtod(m, struct ether_header *);
		/* Check if hardware VLAN insertion is off. */
		if (eh->ether_type == htons(ETHERTYPE_VLAN)) {
			offset = sizeof(struct ether_vlan_header);
			m = m_pullup(m, offset);
			if (m == NULL) {
				*m_head = NULL;
				return (ENOBUFS);
			}
		}
		m = m_pullup(m, offset + sizeof(struct ip));
		if (m == NULL) {
			*m_head = NULL;
			return (ENOBUFS);
		}
		ip = (struct ip *)(mtod(m, char *) + offset);
		offset += (ip->ip_hl << 2);
		tcp_offset = offset;
		/*
		 * It seems that Yukon II has Tx checksum offload bug for
		 * small TCP packets that's less than 60 bytes in size
		 * (e.g. TCP window probe packet, pure ACK packet).
		 * Common work around like padding with zeros to make the
		 * frame minimum ethernet frame size didn't work at all.
		 * Instead of disabling checksum offload completely we
		 * resort to S/W checksum routine when we encounter short
		 * TCP frames.
		 * Short UDP packets appear to be handled correctly by
		 * Yukon II.
		 */
		if (m->m_pkthdr.len < MSK_MIN_FRAMELEN &&
		    (m->m_pkthdr.csum_flags & CSUM_TCP) != 0) {
			uint16_t csum;

			csum = in_cksum_skip(m, ntohs(ip->ip_len) + offset -
			    (ip->ip_hl << 2), offset);
			*(uint16_t *)(m->m_data + offset +
			    m->m_pkthdr.csum_data) = csum;
			m->m_pkthdr.csum_flags &= ~CSUM_TCP;
		}
		*m_head = m;
	}

	prod = sc_if->msk_cdata.msk_tx_prod;
	txd = &sc_if->msk_cdata.msk_txdesc[prod];
	txd_last = txd;
	map = txd->tx_dmamap;
	bzero(&ctx, sizeof(ctx));
	ctx.nseg = MSK_MAXTXSEGS;
	ctx.segs = txsegs;
	error = bus_dmamap_load_mbuf(sc_if->msk_cdata.msk_tx_tag, map,
	    *m_head, msk_dmamap_mbuf_cb, &ctx, BUS_DMA_NOWAIT);
	if (error == 0 && ctx.nseg == 0) {
		bus_dmamap_unload(sc_if->msk_cdata.msk_tx_tag, map);
		error = EFBIG;
	}
	if (error == EFBIG) {
		m = msk_defrag(*m_head, MB_DONTWAIT, MSK_MAXTXSEGS);
		if (m == NULL) {
			m_freem(*m_head);
			*m_head = NULL;
			return (ENOBUFS);
		}
		*m_head = m;

		bzero(&ctx, sizeof(ctx));
		ctx.nseg = MSK_MAXTXSEGS;
		ctx.segs = txsegs;
		error = bus_dmamap_load_mbuf(sc_if->msk_cdata.msk_tx_tag,
		    map, *m_head, msk_dmamap_mbuf_cb, &ctx, BUS_DMA_NOWAIT);
		if (error == 0 && ctx.nseg == 0) {
			bus_dmamap_unload(sc_if->msk_cdata.msk_tx_tag, map);
			error = EFBIG;
		}
		if (error != 0) {
			m_freem(*m_head);
			*m_head = NULL;
			return (error);
		}
	} else if (error != 0) {
		return (error);
	}

	/* Check number of available descriptors. */
	if (sc_if->msk_cdata.msk_tx_cnt + ctx.nseg >=
	    (MSK_TX_RING_CNT - MSK_RESERVED_TX_DESC_CNT)) {
		bus_dmamap_unload(sc_if->msk_cdata.msk_tx_tag, map);
		return (ENOBUFS);
	}

	control = 0;
	tx_le = NULL;

#ifdef notyet
	/* Check if we have a VLAN tag to insert. */
	if ((m->m_flags & M_VLANTAG) != 0) {
		tx_le = &sc_if->msk_rdata.msk_tx_ring[prod];
		tx_le->msk_addr = htole32(0);
		tx_le->msk_control = htole32(OP_VLAN | HW_OWNER |
		    htons(m->m_pkthdr.ether_vtag));
		sc_if->msk_cdata.msk_tx_cnt++;
		MSK_INC(prod, MSK_TX_RING_CNT);
		control |= INS_VLAN;
	}
#endif
	/* Check if we have to handle checksum offload. */
	if (m->m_pkthdr.csum_flags & MSK_CSUM_FEATURES) {
		tx_le = &sc_if->msk_rdata.msk_tx_ring[prod];
		tx_le->msk_addr = htole32(((tcp_offset + m->m_pkthdr.csum_data)
		    & 0xffff) | ((uint32_t)tcp_offset << 16));
		tx_le->msk_control = htole32(1 << 16 | (OP_TCPLISW | HW_OWNER));
		control = CALSUM | WR_SUM | INIT_SUM | LOCK_SUM;
		if ((m->m_pkthdr.csum_flags & CSUM_UDP) != 0)
			control |= UDPTCP;
		sc_if->msk_cdata.msk_tx_cnt++;
		MSK_INC(prod, MSK_TX_RING_CNT);
	}

	si = prod;
	tx_le = &sc_if->msk_rdata.msk_tx_ring[prod];
	tx_le->msk_addr = htole32(MSK_ADDR_LO(txsegs[0].ds_addr));
	tx_le->msk_control = htole32(txsegs[0].ds_len | control |
	    OP_PACKET);
	sc_if->msk_cdata.msk_tx_cnt++;
	MSK_INC(prod, MSK_TX_RING_CNT);

	for (i = 1; i < ctx.nseg; i++) {
		tx_le = &sc_if->msk_rdata.msk_tx_ring[prod];
		tx_le->msk_addr = htole32(MSK_ADDR_LO(txsegs[i].ds_addr));
		tx_le->msk_control = htole32(txsegs[i].ds_len | control |
		    OP_BUFFER | HW_OWNER);
		sc_if->msk_cdata.msk_tx_cnt++;
		MSK_INC(prod, MSK_TX_RING_CNT);
	}
	/* Update producer index. */
	sc_if->msk_cdata.msk_tx_prod = prod;

	/* Set EOP on the last desciptor. */
	prod = (prod + MSK_TX_RING_CNT - 1) % MSK_TX_RING_CNT;
	tx_le = &sc_if->msk_rdata.msk_tx_ring[prod];
	tx_le->msk_control |= htole32(EOP);

	/* Turn the first descriptor ownership to hardware. */
	tx_le = &sc_if->msk_rdata.msk_tx_ring[si];
	tx_le->msk_control |= htole32(HW_OWNER);

	txd = &sc_if->msk_cdata.msk_txdesc[prod];
	map = txd_last->tx_dmamap;
	txd_last->tx_dmamap = txd->tx_dmamap;
	txd->tx_dmamap = map;
	txd->tx_m = m;

	/* Sync descriptors. */
	bus_dmamap_sync(sc_if->msk_cdata.msk_tx_tag, map, BUS_DMASYNC_PREWRITE);

	return (0);
}

static void
msk_start(struct ifnet *ifp)
{
        struct msk_if_softc *sc_if;
        struct mbuf *m_head;
	int enq;

	sc_if = ifp->if_softc;

	ASSERT_SERIALIZED(ifp->if_serializer);

	if (!sc_if->msk_link) {
		ifq_purge(&ifp->if_snd);
		return;
	}

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

	for (enq = 0; !ifq_is_empty(&ifp->if_snd) &&
	    sc_if->msk_cdata.msk_tx_cnt <
	    (MSK_TX_RING_CNT - MSK_RESERVED_TX_DESC_CNT); ) {
		m_head = ifq_dequeue(&ifp->if_snd, NULL);
		if (m_head == NULL)
			break;

		/*
		 * Pack the data into the transmit ring. If we
		 * don't have room, set the OACTIVE flag and wait
		 * for the NIC to drain the ring.
		 */
		if (msk_encap(sc_if, &m_head) != 0) {
			if (m_head == NULL)
				break;
			m_freem(m_head);
			ifp->if_flags |= IFF_OACTIVE;
			break;
		}

		enq++;
		/*
		 * If there's a BPF listener, bounce a copy of this frame
		 * to him.
		 */
		BPF_MTAP(ifp, m_head);
	}

	if (enq > 0) {
		/* Transmit */
		CSR_WRITE_2(sc_if->msk_softc,
		    Y2_PREF_Q_ADDR(sc_if->msk_txq, PREF_UNIT_PUT_IDX_REG),
		    sc_if->msk_cdata.msk_tx_prod);

		/* Set a timeout in case the chip goes out to lunch. */
		ifp->if_timer = MSK_TX_TIMEOUT;
	}
}

static void
msk_watchdog(struct ifnet *ifp)
{
	struct msk_if_softc *sc_if = ifp->if_softc;
	uint32_t ridx;
	int idx;

	ASSERT_SERIALIZED(ifp->if_serializer);

	if (sc_if->msk_link == 0) {
		if (bootverbose)
			if_printf(sc_if->msk_ifp, "watchdog timeout "
			   "(missed link)\n");
		ifp->if_oerrors++;
		msk_init(sc_if);
		return;
	}

	/*
	 * Reclaim first as there is a possibility of losing Tx completion
	 * interrupts.
	 */
	ridx = sc_if->msk_port == MSK_PORT_A ? STAT_TXA1_RIDX : STAT_TXA2_RIDX;
	idx = CSR_READ_2(sc_if->msk_softc, ridx);
	if (sc_if->msk_cdata.msk_tx_cons != idx) {
		msk_txeof(sc_if, idx);
		if (sc_if->msk_cdata.msk_tx_cnt == 0) {
			if_printf(ifp, "watchdog timeout (missed Tx interrupts) "
			    "-- recovering\n");
			if (!ifq_is_empty(&ifp->if_snd))
				if_devstart(ifp);
			return;
		}
	}

	if_printf(ifp, "watchdog timeout\n");
	ifp->if_oerrors++;
	msk_init(sc_if);
	if (!ifq_is_empty(&ifp->if_snd))
		if_devstart(ifp);
}

static int
mskc_shutdown(device_t dev)
{
	struct msk_softc *sc = device_get_softc(dev);
	int i;

	lwkt_serialize_enter(&sc->msk_serializer);

	for (i = 0; i < sc->msk_num_port; i++) {
		if (sc->msk_if[i] != NULL)
			msk_stop(sc->msk_if[i]);
	}

	/* Disable all interrupts. */
	CSR_WRITE_4(sc, B0_IMSK, 0);
	CSR_READ_4(sc, B0_IMSK);
	CSR_WRITE_4(sc, B0_HWE_IMSK, 0);
	CSR_READ_4(sc, B0_HWE_IMSK);

	/* Put hardware reset. */
	CSR_WRITE_2(sc, B0_CTST, CS_RST_SET);

	lwkt_serialize_exit(&sc->msk_serializer);
	return (0);
}

static int
mskc_suspend(device_t dev)
{
	struct msk_softc *sc = device_get_softc(dev);
	int i;

	lwkt_serialize_enter(&sc->msk_serializer);

	for (i = 0; i < sc->msk_num_port; i++) {
		if (sc->msk_if[i] != NULL && sc->msk_if[i]->msk_ifp != NULL &&
		    ((sc->msk_if[i]->msk_ifp->if_flags & IFF_RUNNING) != 0))
			msk_stop(sc->msk_if[i]);
	}

	/* Disable all interrupts. */
	CSR_WRITE_4(sc, B0_IMSK, 0);
	CSR_READ_4(sc, B0_IMSK);
	CSR_WRITE_4(sc, B0_HWE_IMSK, 0);
	CSR_READ_4(sc, B0_HWE_IMSK);

	mskc_phy_power(sc, MSK_PHY_POWERDOWN);

	/* Put hardware reset. */
	CSR_WRITE_2(sc, B0_CTST, CS_RST_SET);
	sc->msk_suspended = 1;

	lwkt_serialize_exit(&sc->msk_serializer);

	return (0);
}

static int
mskc_resume(device_t dev)
{
	struct msk_softc *sc = device_get_softc(dev);
	int i;

	lwkt_serialize_enter(&sc->msk_serializer);

	mskc_reset(sc);
	for (i = 0; i < sc->msk_num_port; i++) {
		if (sc->msk_if[i] != NULL && sc->msk_if[i]->msk_ifp != NULL &&
		    ((sc->msk_if[i]->msk_ifp->if_flags & IFF_UP) != 0))
			msk_init(sc->msk_if[i]);
	}
	sc->msk_suspended = 0;

	lwkt_serialize_exit(&sc->msk_serializer);

	return (0);
}

static void
msk_rxeof(struct msk_if_softc *sc_if, uint32_t status, int len,
	  struct mbuf_chain *chain)
{
	struct mbuf *m;
	struct ifnet *ifp;
	struct msk_rxdesc *rxd;
	int cons, rxlen;

	ifp = sc_if->msk_ifp;

	cons = sc_if->msk_cdata.msk_rx_cons;
	do {
		rxlen = status >> 16;
		if ((status & GMR_FS_VLAN) != 0 &&
		    (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
			rxlen -= EVL_ENCAPLEN;
		if (len > sc_if->msk_framesize ||
		    ((status & GMR_FS_ANY_ERR) != 0) ||
		    ((status & GMR_FS_RX_OK) == 0) || (rxlen != len)) {
			/* Don't count flow-control packet as errors. */
			if ((status & GMR_FS_GOOD_FC) == 0)
				ifp->if_ierrors++;
			msk_discard_rxbuf(sc_if, cons);
			break;
		}
		rxd = &sc_if->msk_cdata.msk_rxdesc[cons];
		m = rxd->rx_m;
		if (msk_newbuf(sc_if, cons) != 0) {
			ifp->if_iqdrops++;
			/* Reuse old buffer. */
			msk_discard_rxbuf(sc_if, cons);
			break;
		}
		m->m_pkthdr.rcvif = ifp;
		m->m_pkthdr.len = m->m_len = len;
		ifp->if_ipackets++;
#ifdef notyet
		/* Check for VLAN tagged packets. */
		if ((status & GMR_FS_VLAN) != 0 &&
		    (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
			m->m_pkthdr.ether_vtag = sc_if->msk_vtag;
			m->m_flags |= M_VLANTAG;
		}
#endif

		ether_input_chain(ifp, m, chain);
	} while (0);

	MSK_INC(sc_if->msk_cdata.msk_rx_cons, MSK_RX_RING_CNT);
	MSK_INC(sc_if->msk_cdata.msk_rx_prod, MSK_RX_RING_CNT);
}

#ifdef MSK_JUMBO
static void
msk_jumbo_rxeof(struct msk_if_softc *sc_if, uint32_t status, int len)
{
	struct mbuf *m;
	struct ifnet *ifp;
	struct msk_rxdesc *jrxd;
	int cons, rxlen;

	ifp = sc_if->msk_ifp;

	MSK_IF_LOCK_ASSERT(sc_if);

	cons = sc_if->msk_cdata.msk_rx_cons;
	do {
		rxlen = status >> 16;
		if ((status & GMR_FS_VLAN) != 0 &&
		    (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
			rxlen -= ETHER_VLAN_ENCAP_LEN;
		if (len > sc_if->msk_framesize ||
		    ((status & GMR_FS_ANY_ERR) != 0) ||
		    ((status & GMR_FS_RX_OK) == 0) || (rxlen != len)) {
			/* Don't count flow-control packet as errors. */
			if ((status & GMR_FS_GOOD_FC) == 0)
				ifp->if_ierrors++;
			msk_discard_jumbo_rxbuf(sc_if, cons);
			break;
		}
		jrxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[cons];
		m = jrxd->rx_m;
		if (msk_jumbo_newbuf(sc_if, cons) != 0) {
			ifp->if_iqdrops++;
			/* Reuse old buffer. */
			msk_discard_jumbo_rxbuf(sc_if, cons);
			break;
		}
		m->m_pkthdr.rcvif = ifp;
		m->m_pkthdr.len = m->m_len = len;
		ifp->if_ipackets++;
		/* Check for VLAN tagged packets. */
		if ((status & GMR_FS_VLAN) != 0 &&
		    (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
			m->m_pkthdr.ether_vtag = sc_if->msk_vtag;
			m->m_flags |= M_VLANTAG;
		}
		MSK_IF_UNLOCK(sc_if);
		(*ifp->if_input)(ifp, m);
		MSK_IF_LOCK(sc_if);
	} while (0);

	MSK_INC(sc_if->msk_cdata.msk_rx_cons, MSK_JUMBO_RX_RING_CNT);
	MSK_INC(sc_if->msk_cdata.msk_rx_prod, MSK_JUMBO_RX_RING_CNT);
}
#endif

static void
msk_txeof(struct msk_if_softc *sc_if, int idx)
{
	struct msk_txdesc *txd;
	struct msk_tx_desc *cur_tx;
	struct ifnet *ifp;
	uint32_t control;
	int cons, prog;

	ifp = sc_if->msk_ifp;

	/*
	 * Go through our tx ring and free mbufs for those
	 * frames that have been sent.
	 */
	cons = sc_if->msk_cdata.msk_tx_cons;
	prog = 0;
	for (; cons != idx; MSK_INC(cons, MSK_TX_RING_CNT)) {
		if (sc_if->msk_cdata.msk_tx_cnt <= 0)
			break;
		prog++;
		cur_tx = &sc_if->msk_rdata.msk_tx_ring[cons];
		control = le32toh(cur_tx->msk_control);
		sc_if->msk_cdata.msk_tx_cnt--;
		ifp->if_flags &= ~IFF_OACTIVE;
		if ((control & EOP) == 0)
			continue;
		txd = &sc_if->msk_cdata.msk_txdesc[cons];
		bus_dmamap_sync(sc_if->msk_cdata.msk_tx_tag, txd->tx_dmamap,
		    BUS_DMASYNC_POSTWRITE);
		bus_dmamap_unload(sc_if->msk_cdata.msk_tx_tag, txd->tx_dmamap);

		ifp->if_opackets++;
		KASSERT(txd->tx_m != NULL, ("%s: freeing NULL mbuf!",
		    __func__));
		m_freem(txd->tx_m);
		txd->tx_m = NULL;
	}

	if (prog > 0) {
		sc_if->msk_cdata.msk_tx_cons = cons;
		if (sc_if->msk_cdata.msk_tx_cnt == 0)
			ifp->if_timer = 0;
		/* No need to sync LEs as we didn't update LEs. */
	}
}

static void
msk_tick(void *xsc_if)
{
	struct msk_if_softc *sc_if = xsc_if;
	struct ifnet *ifp = &sc_if->arpcom.ac_if;
	struct mii_data *mii;

	lwkt_serialize_enter(ifp->if_serializer);

	mii = device_get_softc(sc_if->msk_miibus);

	mii_tick(mii);
	callout_reset(&sc_if->msk_tick_ch, hz, msk_tick, sc_if);

	lwkt_serialize_exit(ifp->if_serializer);
}

static void
msk_intr_phy(struct msk_if_softc *sc_if)
{
	uint16_t status;

	msk_phy_readreg(sc_if, PHY_ADDR_MARV, PHY_MARV_INT_STAT);
	status = msk_phy_readreg(sc_if, PHY_ADDR_MARV, PHY_MARV_INT_STAT);
	/* Handle FIFO Underrun/Overflow? */
	if (status & PHY_M_IS_FIFO_ERROR) {
		device_printf(sc_if->msk_if_dev,
		    "PHY FIFO underrun/overflow.\n");
	}
}

static void
msk_intr_gmac(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc;
	uint8_t status;

	sc = sc_if->msk_softc;
	status = CSR_READ_1(sc, MR_ADDR(sc_if->msk_port, GMAC_IRQ_SRC));

	/* GMAC Rx FIFO overrun. */
	if ((status & GM_IS_RX_FF_OR) != 0) {
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T),
		    GMF_CLI_RX_FO);
		device_printf(sc_if->msk_if_dev, "Rx FIFO overrun!\n");
	}
	/* GMAC Tx FIFO underrun. */
	if ((status & GM_IS_TX_FF_UR) != 0) {
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
		    GMF_CLI_TX_FU);
		device_printf(sc_if->msk_if_dev, "Tx FIFO underrun!\n");
		/*
		 * XXX
		 * In case of Tx underrun, we may need to flush/reset
		 * Tx MAC but that would also require resynchronization
		 * with status LEs. Reintializing status LEs would
		 * affect other port in dual MAC configuration so it
		 * should be avoided as possible as we can.
		 * Due to lack of documentation it's all vague guess but
		 * it needs more investigation.
		 */
	}
}

static void
msk_handle_hwerr(struct msk_if_softc *sc_if, uint32_t status)
{
	struct msk_softc *sc;

	sc = sc_if->msk_softc;
	if ((status & Y2_IS_PAR_RD1) != 0) {
		device_printf(sc_if->msk_if_dev,
		    "RAM buffer read parity error\n");
		/* Clear IRQ. */
		CSR_WRITE_2(sc, SELECT_RAM_BUFFER(sc_if->msk_port, B3_RI_CTRL),
		    RI_CLR_RD_PERR);
	}
	if ((status & Y2_IS_PAR_WR1) != 0) {
		device_printf(sc_if->msk_if_dev,
		    "RAM buffer write parity error\n");
		/* Clear IRQ. */
		CSR_WRITE_2(sc, SELECT_RAM_BUFFER(sc_if->msk_port, B3_RI_CTRL),
		    RI_CLR_WR_PERR);
	}
	if ((status & Y2_IS_PAR_MAC1) != 0) {
		device_printf(sc_if->msk_if_dev, "Tx MAC parity error\n");
		/* Clear IRQ. */
		CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
		    GMF_CLI_TX_PE);
	}
	if ((status & Y2_IS_PAR_RX1) != 0) {
		device_printf(sc_if->msk_if_dev, "Rx parity error\n");
		/* Clear IRQ. */
		CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR), BMU_CLR_IRQ_PAR);
	}
	if ((status & (Y2_IS_TCP_TXS1 | Y2_IS_TCP_TXA1)) != 0) {
		device_printf(sc_if->msk_if_dev, "TCP segmentation error\n");
		/* Clear IRQ. */
		CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR), BMU_CLR_IRQ_TCP);
	}
}

static void
mskc_intr_hwerr(struct msk_softc *sc)
{
	uint32_t status;
	uint32_t tlphead[4];

	status = CSR_READ_4(sc, B0_HWE_ISRC);
	/* Time Stamp timer overflow. */
	if ((status & Y2_IS_TIST_OV) != 0)
		CSR_WRITE_1(sc, GMAC_TI_ST_CTRL, GMT_ST_CLR_IRQ);
	if ((status & Y2_IS_PCI_NEXP) != 0) {
		/*
		 * PCI Express Error occured which is not described in PEX
		 * spec.
		 * This error is also mapped either to Master Abort(
		 * Y2_IS_MST_ERR) or Target Abort (Y2_IS_IRQ_STAT) bit and
		 * can only be cleared there.
                 */
		device_printf(sc->msk_dev,
		    "PCI Express protocol violation error\n");
	}

	if ((status & (Y2_IS_MST_ERR | Y2_IS_IRQ_STAT)) != 0) {
		uint16_t v16;

		if ((status & Y2_IS_MST_ERR) != 0)
			device_printf(sc->msk_dev,
			    "unexpected IRQ Status error\n");
		else
			device_printf(sc->msk_dev,
			    "unexpected IRQ Master error\n");
		/* Reset all bits in the PCI status register. */
		v16 = pci_read_config(sc->msk_dev, PCIR_STATUS, 2);
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_ON);
		pci_write_config(sc->msk_dev, PCIR_STATUS, v16 |
		    PCIM_STATUS_PERR | PCIM_STATUS_SERR | PCIM_STATUS_RMABORT |
		    PCIM_STATUS_RTABORT | PCIM_STATUS_PERRREPORT, 2);
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_OFF);
	}

	/* Check for PCI Express Uncorrectable Error. */
	if ((status & Y2_IS_PCI_EXP) != 0) {
		uint32_t v32;

		/*
		 * On PCI Express bus bridges are called root complexes (RC).
		 * PCI Express errors are recognized by the root complex too,
		 * which requests the system to handle the problem. After
		 * error occurence it may be that no access to the adapter
		 * may be performed any longer.
		 */

		v32 = CSR_PCI_READ_4(sc, PEX_UNC_ERR_STAT);
		if ((v32 & PEX_UNSUP_REQ) != 0) {
			/* Ignore unsupported request error. */
			if (bootverbose) {
				device_printf(sc->msk_dev,
				    "Uncorrectable PCI Express error\n");
			}
		}
		if ((v32 & (PEX_FATAL_ERRORS | PEX_POIS_TLP)) != 0) {
			int i;

			/* Get TLP header form Log Registers. */
			for (i = 0; i < 4; i++)
				tlphead[i] = CSR_PCI_READ_4(sc,
				    PEX_HEADER_LOG + i * 4);
			/* Check for vendor defined broadcast message. */
			if (!(tlphead[0] == 0x73004001 && tlphead[1] == 0x7f)) {
				sc->msk_intrhwemask &= ~Y2_IS_PCI_EXP;
				CSR_WRITE_4(sc, B0_HWE_IMSK,
				    sc->msk_intrhwemask);
				CSR_READ_4(sc, B0_HWE_IMSK);
			}
		}
		/* Clear the interrupt. */
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_ON);
		CSR_PCI_WRITE_4(sc, PEX_UNC_ERR_STAT, 0xffffffff);
		CSR_WRITE_1(sc, B2_TST_CTRL1, TST_CFG_WRITE_OFF);
	}

	if ((status & Y2_HWE_L1_MASK) != 0 && sc->msk_if[MSK_PORT_A] != NULL)
		msk_handle_hwerr(sc->msk_if[MSK_PORT_A], status);
	if ((status & Y2_HWE_L2_MASK) != 0 && sc->msk_if[MSK_PORT_B] != NULL)
		msk_handle_hwerr(sc->msk_if[MSK_PORT_B], status >> 8);
}

static __inline void
msk_rxput(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc;

	sc = sc_if->msk_softc;
#ifdef MSK_JUMBO
	if (sc_if->msk_framesize > (MCLBYTES - ETHER_HDR_LEN)) {
		bus_dmamap_sync(
		    sc_if->msk_cdata.msk_jumbo_rx_ring_tag,
		    sc_if->msk_cdata.msk_jumbo_rx_ring_map,
		    BUS_DMASYNC_PREWRITE);
	}
#endif
	CSR_WRITE_2(sc, Y2_PREF_Q_ADDR(sc_if->msk_rxq,
	    PREF_UNIT_PUT_IDX_REG), sc_if->msk_cdata.msk_rx_prod);
}

static int
mskc_handle_events(struct msk_softc *sc)
{
	struct msk_if_softc *sc_if;
	int rxput[2];
	struct msk_stat_desc *sd;
	uint32_t control, status;
	int cons, idx, len, port, rxprog;
	struct mbuf_chain chain[MAXCPU];

	idx = CSR_READ_2(sc, STAT_PUT_IDX);
	if (idx == sc->msk_stat_cons)
		return (0);

	ether_input_chain_init(chain);

	rxput[MSK_PORT_A] = rxput[MSK_PORT_B] = 0;

	rxprog = 0;
	for (cons = sc->msk_stat_cons; cons != idx;) {
		sd = &sc->msk_stat_ring[cons];
		control = le32toh(sd->msk_control);
		if ((control & HW_OWNER) == 0)
			break;
		/*
		 * Marvell's FreeBSD driver updates status LE after clearing
		 * HW_OWNER. However we don't have a way to sync single LE
		 * with bus_dma(9) API. bus_dma(9) provides a way to sync
		 * an entire DMA map. So don't sync LE until we have a better
		 * way to sync LEs.
		 */
		control &= ~HW_OWNER;
		sd->msk_control = htole32(control);
		status = le32toh(sd->msk_status);
		len = control & STLE_LEN_MASK;
		port = (control >> 16) & 0x01;
		sc_if = sc->msk_if[port];
		if (sc_if == NULL) {
			device_printf(sc->msk_dev, "invalid port opcode "
			    "0x%08x\n", control & STLE_OP_MASK);
			continue;
		}

		switch (control & STLE_OP_MASK) {
		case OP_RXVLAN:
			sc_if->msk_vtag = ntohs(len);
			break;
		case OP_RXCHKSVLAN:
			sc_if->msk_vtag = ntohs(len);
			break;
		case OP_RXSTAT:
#ifdef MSK_JUMBO
			if (sc_if->msk_framesize > (MCLBYTES - ETHER_HDR_LEN))
				msk_jumbo_rxeof(sc_if, status, len);
			else
#endif
				msk_rxeof(sc_if, status, len, chain);
			rxprog++;
			/*
			 * Because there is no way to sync single Rx LE
			 * put the DMA sync operation off until the end of
			 * event processing.
			 */
			rxput[port]++;
			/* Update prefetch unit if we've passed water mark. */
			if (rxput[port] >= sc_if->msk_cdata.msk_rx_putwm) {
				msk_rxput(sc_if);
				rxput[port] = 0;
			}
			break;
		case OP_TXINDEXLE:
			if (sc->msk_if[MSK_PORT_A] != NULL) {
				msk_txeof(sc->msk_if[MSK_PORT_A],
				    status & STLE_TXA1_MSKL);
			}
			if (sc->msk_if[MSK_PORT_B] != NULL) {
				msk_txeof(sc->msk_if[MSK_PORT_B],
				    ((status & STLE_TXA2_MSKL) >>
				    STLE_TXA2_SHIFTL) |
				    ((len & STLE_TXA2_MSKH) <<
				    STLE_TXA2_SHIFTH));
			}
			break;
		default:
			device_printf(sc->msk_dev, "unhandled opcode 0x%08x\n",
			    control & STLE_OP_MASK);
			break;
		}
		MSK_INC(cons, MSK_STAT_RING_CNT);
		if (rxprog > sc->msk_process_limit)
			break;
	}

	if (rxprog > 0)
		ether_input_dispatch(chain);

	sc->msk_stat_cons = cons;
	/* XXX We should sync status LEs here. See above notes. */

	if (rxput[MSK_PORT_A] > 0)
		msk_rxput(sc->msk_if[MSK_PORT_A]);
	if (rxput[MSK_PORT_B] > 0)
		msk_rxput(sc->msk_if[MSK_PORT_B]);

	return (sc->msk_stat_cons != CSR_READ_2(sc, STAT_PUT_IDX));
}

/* Legacy interrupt handler for shared interrupt. */
static void
mskc_intr(void *xsc)
{
	struct msk_softc *sc;
	struct msk_if_softc *sc_if0, *sc_if1;
	struct ifnet *ifp0, *ifp1;
	uint32_t status;

	sc = xsc;
	ASSERT_SERIALIZED(&sc->msk_serializer);

	/* Reading B0_Y2_SP_ISRC2 masks further interrupts. */
	status = CSR_READ_4(sc, B0_Y2_SP_ISRC2);
	if (status == 0 || status == 0xffffffff || sc->msk_suspended != 0 ||
	    (status & sc->msk_intrmask) == 0) {
		CSR_WRITE_4(sc, B0_Y2_SP_ICR, 2);
		return;
	}

	sc_if0 = sc->msk_if[MSK_PORT_A];
	sc_if1 = sc->msk_if[MSK_PORT_B];
	ifp0 = ifp1 = NULL;
	if (sc_if0 != NULL)
		ifp0 = sc_if0->msk_ifp;
	if (sc_if1 != NULL)
		ifp1 = sc_if1->msk_ifp;

	if ((status & Y2_IS_IRQ_PHY1) != 0 && sc_if0 != NULL)
		msk_intr_phy(sc_if0);
	if ((status & Y2_IS_IRQ_PHY2) != 0 && sc_if1 != NULL)
		msk_intr_phy(sc_if1);
	if ((status & Y2_IS_IRQ_MAC1) != 0 && sc_if0 != NULL)
		msk_intr_gmac(sc_if0);
	if ((status & Y2_IS_IRQ_MAC2) != 0 && sc_if1 != NULL)
		msk_intr_gmac(sc_if1);
	if ((status & (Y2_IS_CHK_RX1 | Y2_IS_CHK_RX2)) != 0) {
		device_printf(sc->msk_dev, "Rx descriptor error\n");
		sc->msk_intrmask &= ~(Y2_IS_CHK_RX1 | Y2_IS_CHK_RX2);
		CSR_WRITE_4(sc, B0_IMSK, sc->msk_intrmask);
		CSR_READ_4(sc, B0_IMSK);
	}
        if ((status & (Y2_IS_CHK_TXA1 | Y2_IS_CHK_TXA2)) != 0) {
		device_printf(sc->msk_dev, "Tx descriptor error\n");
		sc->msk_intrmask &= ~(Y2_IS_CHK_TXA1 | Y2_IS_CHK_TXA2);
		CSR_WRITE_4(sc, B0_IMSK, sc->msk_intrmask);
		CSR_READ_4(sc, B0_IMSK);
	}
	if ((status & Y2_IS_HW_ERR) != 0)
		mskc_intr_hwerr(sc);

	while (mskc_handle_events(sc) != 0)
		;
	if ((status & Y2_IS_STAT_BMU) != 0)
		CSR_WRITE_4(sc, STAT_CTRL, SC_STAT_CLR_IRQ);

	/* Reenable interrupts. */
	CSR_WRITE_4(sc, B0_Y2_SP_ICR, 2);

	if (ifp0 != NULL && (ifp0->if_flags & IFF_RUNNING) != 0 &&
	    !ifq_is_empty(&ifp0->if_snd))
		if_devstart(ifp0);
	if (ifp1 != NULL && (ifp1->if_flags & IFF_RUNNING) != 0 &&
	    !ifq_is_empty(&ifp1->if_snd))
		if_devstart(ifp1);
}

static void
msk_init(void *xsc)
{
	struct msk_if_softc *sc_if = xsc;
	struct msk_softc *sc = sc_if->msk_softc;
	struct ifnet *ifp = sc_if->msk_ifp;
	struct mii_data	 *mii;
	uint16_t eaddr[ETHER_ADDR_LEN / 2];
	uint16_t gmac;
	int error, i;

	ASSERT_SERIALIZED(ifp->if_serializer);

	mii = device_get_softc(sc_if->msk_miibus);

	error = 0;
	/* Cancel pending I/O and free all Rx/Tx buffers. */
	msk_stop(sc_if);

	sc_if->msk_framesize = ifp->if_mtu + ETHER_HDR_LEN + EVL_ENCAPLEN;
	if (sc_if->msk_framesize > MSK_MAX_FRAMELEN &&
	    sc_if->msk_softc->msk_hw_id == CHIP_ID_YUKON_EC_U) {
		/*
		 * In Yukon EC Ultra, TSO & checksum offload is not
		 * supported for jumbo frame.
		 */
		ifp->if_hwassist &= ~MSK_CSUM_FEATURES;
		ifp->if_capenable &= ~IFCAP_TXCSUM;
	}

	/*
	 * Initialize GMAC first.
	 * Without this initialization, Rx MAC did not work as expected
	 * and Rx MAC garbled status LEs and it resulted in out-of-order
	 * or duplicated frame delivery which in turn showed very poor
	 * Rx performance.(I had to write a packet analysis code that
	 * could be embeded in driver to diagnose this issue.)
	 * I've spent almost 2 months to fix this issue. If I have had
	 * datasheet for Yukon II I wouldn't have encountered this. :-(
	 */
	gmac = GM_GPCR_SPEED_100 | GM_GPCR_SPEED_1000 | GM_GPCR_DUP_FULL;
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_GP_CTRL, gmac);

	/* Dummy read the Interrupt Source Register. */
	CSR_READ_1(sc, MR_ADDR(sc_if->msk_port, GMAC_IRQ_SRC));

	/* Set MIB Clear Counter Mode. */
	gmac = GMAC_READ_2(sc, sc_if->msk_port, GM_PHY_ADDR);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_PHY_ADDR, gmac | GM_PAR_MIB_CLR);
	/* Read all MIB Counters with Clear Mode set. */
	for (i = 0; i < GM_MIB_CNT_SIZE; i++)
		GMAC_READ_2(sc, sc_if->msk_port, GM_MIB_CNT_BASE + 8 * i);
	/* Clear MIB Clear Counter Mode. */
	gmac &= ~GM_PAR_MIB_CLR;
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_PHY_ADDR, gmac);

	/* Disable FCS. */
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_RX_CTRL, GM_RXCR_CRC_DIS);

	/* Setup Transmit Control Register. */
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_TX_CTRL, TX_COL_THR(TX_COL_DEF));

	/* Setup Transmit Flow Control Register. */
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_TX_FLOW_CTRL, 0xffff);

	/* Setup Transmit Parameter Register. */
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_TX_PARAM,
	    TX_JAM_LEN_VAL(TX_JAM_LEN_DEF) | TX_JAM_IPG_VAL(TX_JAM_IPG_DEF) |
	    TX_IPG_JAM_DATA(TX_IPG_JAM_DEF) | TX_BACK_OFF_LIM(TX_BOF_LIM_DEF));

	gmac = DATA_BLIND_VAL(DATA_BLIND_DEF) |
	    GM_SMOD_VLAN_ENA | IPG_DATA_VAL(IPG_DATA_DEF);

	if (sc_if->msk_framesize > MSK_MAX_FRAMELEN)
		gmac |= GM_SMOD_JUMBO_ENA;
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_SERIAL_MODE, gmac);

	/* Set station address. */
        bcopy(IF_LLADDR(ifp), eaddr, ETHER_ADDR_LEN);
        for (i = 0; i < ETHER_ADDR_LEN /2; i++)
		GMAC_WRITE_2(sc, sc_if->msk_port, GM_SRC_ADDR_1L + i * 4,
		    eaddr[i]);
        for (i = 0; i < ETHER_ADDR_LEN /2; i++)
		GMAC_WRITE_2(sc, sc_if->msk_port, GM_SRC_ADDR_2L + i * 4,
		    eaddr[i]);

	/* Disable interrupts for counter overflows. */
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_TX_IRQ_MSK, 0);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_RX_IRQ_MSK, 0);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_TR_IRQ_MSK, 0);

	/* Configure Rx MAC FIFO. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T), GMF_RST_SET);
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T), GMF_RST_CLR);
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T),
	    GMF_OPER_ON | GMF_RX_F_FL_ON);

	/* Set promiscuous mode. */
	msk_setpromisc(sc_if);

	/* Set multicast filter. */
	msk_setmulti(sc_if);

	/* Flush Rx MAC FIFO on any flow control or error. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_FL_MSK),
	    GMR_FS_ANY_ERR);

	/* Set Rx FIFO flush threshold to 64 bytes. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_FL_THR),
	    RX_GMF_FL_THR_DEF);

	/* Configure Tx MAC FIFO. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T), GMF_RST_SET);
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T), GMF_RST_CLR);
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T), GMF_OPER_ON);

	/* Configure hardware VLAN tag insertion/stripping. */
	msk_setvlan(sc_if, ifp);

	if (sc->msk_hw_id == CHIP_ID_YUKON_EC_U) {
		/* Set Rx Pause threshould. */
		CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, RX_GMF_LP_THR),
		    MSK_ECU_LLPP);
		CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, RX_GMF_UP_THR),
		    MSK_ECU_ULPP);
		if (sc_if->msk_framesize > MSK_MAX_FRAMELEN) {
			/*
			 * Set Tx GMAC FIFO Almost Empty Threshold.
			 */
			CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_AE_THR),
			    MSK_ECU_JUMBO_WM << 16 | MSK_ECU_AE_THR);
			/* Disable Store & Forward mode for Tx. */
			CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
			    TX_JUMBO_ENA | TX_STFW_DIS);
		} else {
			/* Enable Store & Forward mode for Tx. */
			CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T),
			    TX_JUMBO_DIS | TX_STFW_ENA);
		}
	}

	/*
	 * Disable Force Sync bit and Alloc bit in Tx RAM interface
	 * arbiter as we don't use Sync Tx queue.
	 */
	CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, TXA_CTRL),
	    TXA_DIS_FSYNC | TXA_DIS_ALLOC | TXA_STOP_RC);
	/* Enable the RAM Interface Arbiter. */
	CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, TXA_CTRL), TXA_ENA_ARB);

	/* Setup RAM buffer. */
	msk_set_rambuffer(sc_if);

	/* Disable Tx sync Queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txsq, RB_CTRL), RB_RST_SET);

	/* Setup Tx Queue Bus Memory Interface. */
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR), BMU_CLR_RESET);
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR), BMU_OPER_INIT);
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR), BMU_FIFO_OP_ON);
	CSR_WRITE_2(sc, Q_ADDR(sc_if->msk_txq, Q_WM), MSK_BMU_TX_WM);
	if (sc->msk_hw_id == CHIP_ID_YUKON_EC_U &&
	    sc->msk_hw_rev == CHIP_REV_YU_EC_U_A0) {
		/* Fix for Yukon-EC Ultra: set BMU FIFO level */
		CSR_WRITE_2(sc, Q_ADDR(sc_if->msk_txq, Q_AL), MSK_ECU_TXFF_LEV);
	}

	/* Setup Rx Queue Bus Memory Interface. */
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR), BMU_CLR_RESET);
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR), BMU_OPER_INIT);
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR), BMU_FIFO_OP_ON);
	CSR_WRITE_2(sc, Q_ADDR(sc_if->msk_rxq, Q_WM), MSK_BMU_RX_WM);
        if (sc->msk_hw_id == CHIP_ID_YUKON_EC_U &&
	    sc->msk_hw_rev >= CHIP_REV_YU_EC_U_A1) {
		/* MAC Rx RAM Read is controlled by hardware. */
                CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_F), F_M_RX_RAM_DIS);
	}

	msk_set_prefetch(sc, sc_if->msk_txq,
	    sc_if->msk_rdata.msk_tx_ring_paddr, MSK_TX_RING_CNT - 1);
	msk_init_tx_ring(sc_if);

	/* Disable Rx checksum offload and RSS hash. */
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR),
	    BMU_DIS_RX_CHKSUM | BMU_DIS_RX_RSS_HASH);
#ifdef MSK_JUMBO
	if (sc_if->msk_framesize > (MCLBYTES - ETHER_HDR_LEN)) {
		msk_set_prefetch(sc, sc_if->msk_rxq,
		    sc_if->msk_rdata.msk_jumbo_rx_ring_paddr,
		    MSK_JUMBO_RX_RING_CNT - 1);
		error = msk_init_jumbo_rx_ring(sc_if);
	} else
#endif
	{
		msk_set_prefetch(sc, sc_if->msk_rxq,
		    sc_if->msk_rdata.msk_rx_ring_paddr,
		    MSK_RX_RING_CNT - 1);
		error = msk_init_rx_ring(sc_if);
	}
	if (error != 0) {
		device_printf(sc_if->msk_if_dev,
		    "initialization failed: no memory for Rx buffers\n");
		msk_stop(sc_if);
		return;
	}

	/* Configure interrupt handling. */
	if (sc_if->msk_port == MSK_PORT_A) {
		sc->msk_intrmask |= Y2_IS_PORT_A;
		sc->msk_intrhwemask |= Y2_HWE_L1_MASK;
	} else {
		sc->msk_intrmask |= Y2_IS_PORT_B;
		sc->msk_intrhwemask |= Y2_HWE_L2_MASK;
	}
	CSR_WRITE_4(sc, B0_HWE_IMSK, sc->msk_intrhwemask);
	CSR_READ_4(sc, B0_HWE_IMSK);
	CSR_WRITE_4(sc, B0_IMSK, sc->msk_intrmask);
	CSR_READ_4(sc, B0_IMSK);

	sc_if->msk_link = 0;
	mii_mediachg(mii);

	mskc_set_imtimer(sc);

	ifp->if_flags |= IFF_RUNNING;
	ifp->if_flags &= ~IFF_OACTIVE;

	callout_reset(&sc_if->msk_tick_ch, hz, msk_tick, sc_if);
}

static void
msk_set_rambuffer(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc;
	int ltpp, utpp;

	sc = sc_if->msk_softc;

	/* Setup Rx Queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_rxq, RB_CTRL), RB_RST_CLR);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_START),
	    sc->msk_rxqstart[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_END),
	    sc->msk_rxqend[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_WP),
	    sc->msk_rxqstart[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_RP),
	    sc->msk_rxqstart[sc_if->msk_port] / 8);

	utpp = (sc->msk_rxqend[sc_if->msk_port] + 1 -
	    sc->msk_rxqstart[sc_if->msk_port] - MSK_RB_ULPP) / 8;
	ltpp = (sc->msk_rxqend[sc_if->msk_port] + 1 -
	    sc->msk_rxqstart[sc_if->msk_port] - MSK_RB_LLPP_B) / 8;
	if (sc->msk_rxqsize < MSK_MIN_RXQ_SIZE)
		ltpp += (MSK_RB_LLPP_B - MSK_RB_LLPP_S) / 8;
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_RX_UTPP), utpp);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_rxq, RB_RX_LTPP), ltpp);
	/* Set Rx priority(RB_RX_UTHP/RB_RX_LTHP) thresholds? */

	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_rxq, RB_CTRL), RB_ENA_OP_MD);
	CSR_READ_1(sc, RB_ADDR(sc_if->msk_rxq, RB_CTRL));

	/* Setup Tx Queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL), RB_RST_CLR);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_txq, RB_START),
	    sc->msk_txqstart[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_txq, RB_END),
	    sc->msk_txqend[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_txq, RB_WP),
	    sc->msk_txqstart[sc_if->msk_port] / 8);
	CSR_WRITE_4(sc, RB_ADDR(sc_if->msk_txq, RB_RP),
	    sc->msk_txqstart[sc_if->msk_port] / 8);
	/* Enable Store & Forward for Tx side. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL), RB_ENA_STFWD);
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL), RB_ENA_OP_MD);
	CSR_READ_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL));
}

static void
msk_set_prefetch(struct msk_softc *sc, int qaddr, bus_addr_t addr,
    uint32_t count)
{

	/* Reset the prefetch unit. */
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_CTRL_REG),
	    PREF_UNIT_RST_SET);
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_CTRL_REG),
	    PREF_UNIT_RST_CLR);
	/* Set LE base address. */
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_ADDR_LOW_REG),
	    MSK_ADDR_LO(addr));
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_ADDR_HI_REG),
	    MSK_ADDR_HI(addr));
	/* Set the list last index. */
	CSR_WRITE_2(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_LAST_IDX_REG),
	    count);
	/* Turn on prefetch unit. */
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_CTRL_REG),
	    PREF_UNIT_OP_ON);
	/* Dummy read to ensure write. */
	CSR_READ_4(sc, Y2_PREF_Q_ADDR(qaddr, PREF_UNIT_CTRL_REG));
}

static void
msk_stop(struct msk_if_softc *sc_if)
{
	struct msk_softc *sc = sc_if->msk_softc;
	struct ifnet *ifp = sc_if->msk_ifp;
	struct msk_txdesc *txd;
	struct msk_rxdesc *rxd;
#ifdef MSK_JUMBO
	struct msk_rxdesc *jrxd;
#endif
	uint32_t val;
	int i;

	ASSERT_SERIALIZED(ifp->if_serializer);

	callout_stop(&sc_if->msk_tick_ch);
	ifp->if_timer = 0;

	/* Disable interrupts. */
	if (sc_if->msk_port == MSK_PORT_A) {
		sc->msk_intrmask &= ~Y2_IS_PORT_A;
		sc->msk_intrhwemask &= ~Y2_HWE_L1_MASK;
	} else {
		sc->msk_intrmask &= ~Y2_IS_PORT_B;
		sc->msk_intrhwemask &= ~Y2_HWE_L2_MASK;
	}
	CSR_WRITE_4(sc, B0_HWE_IMSK, sc->msk_intrhwemask);
	CSR_READ_4(sc, B0_HWE_IMSK);
	CSR_WRITE_4(sc, B0_IMSK, sc->msk_intrmask);
	CSR_READ_4(sc, B0_IMSK);

	/* Disable Tx/Rx MAC. */
	val = GMAC_READ_2(sc, sc_if->msk_port, GM_GP_CTRL);
	val &= ~(GM_GPCR_RX_ENA | GM_GPCR_TX_ENA);
	GMAC_WRITE_2(sc, sc_if->msk_port, GM_GP_CTRL, val);
	/* Read again to ensure writing. */
	GMAC_READ_2(sc, sc_if->msk_port, GM_GP_CTRL);

	/* Stop Tx BMU. */
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR), BMU_STOP);
	val = CSR_READ_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR));
	for (i = 0; i < MSK_TIMEOUT; i++) {
		if ((val & (BMU_STOP | BMU_IDLE)) == 0) {
			CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR),
			    BMU_STOP);
			val = CSR_READ_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR));
		} else
			break;
		DELAY(1);
	}
	if (i == MSK_TIMEOUT)
		device_printf(sc_if->msk_if_dev, "Tx BMU stop failed\n");
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL),
	    RB_RST_SET | RB_DIS_OP_MD);

	/* Disable all GMAC interrupt. */
	CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, GMAC_IRQ_MSK), 0);
	/* Disable PHY interrupt. */
	msk_phy_writereg(sc_if, PHY_ADDR_MARV, PHY_MARV_INT_MASK, 0);

	/* Disable the RAM Interface Arbiter. */
	CSR_WRITE_1(sc, MR_ADDR(sc_if->msk_port, TXA_CTRL), TXA_DIS_ARB);

	/* Reset the PCI FIFO of the async Tx queue */
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_txq, Q_CSR),
	    BMU_RST_SET | BMU_FIFO_RST);

	/* Reset the Tx prefetch units. */
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(sc_if->msk_txq, PREF_UNIT_CTRL_REG),
	    PREF_UNIT_RST_SET);

	/* Reset the RAM Buffer async Tx queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_txq, RB_CTRL), RB_RST_SET);

	/* Reset Tx MAC FIFO. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, TX_GMF_CTRL_T), GMF_RST_SET);
	/* Set Pause Off. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, GMAC_CTRL), GMC_PAUSE_OFF);

	/*
	 * The Rx Stop command will not work for Yukon-2 if the BMU does not
	 * reach the end of packet and since we can't make sure that we have
	 * incoming data, we must reset the BMU while it is not during a DMA
	 * transfer. Since it is possible that the Rx path is still active,
	 * the Rx RAM buffer will be stopped first, so any possible incoming
	 * data will not trigger a DMA. After the RAM buffer is stopped, the
	 * BMU is polled until any DMA in progress is ended and only then it
	 * will be reset.
	 */

	/* Disable the RAM Buffer receive queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_rxq, RB_CTRL), RB_DIS_OP_MD);
	for (i = 0; i < MSK_TIMEOUT; i++) {
		if (CSR_READ_1(sc, RB_ADDR(sc_if->msk_rxq, Q_RSL)) ==
		    CSR_READ_1(sc, RB_ADDR(sc_if->msk_rxq, Q_RL)))
			break;
		DELAY(1);
	}
	if (i == MSK_TIMEOUT)
		device_printf(sc_if->msk_if_dev, "Rx BMU stop failed\n");
	CSR_WRITE_4(sc, Q_ADDR(sc_if->msk_rxq, Q_CSR),
	    BMU_RST_SET | BMU_FIFO_RST);
	/* Reset the Rx prefetch unit. */
	CSR_WRITE_4(sc, Y2_PREF_Q_ADDR(sc_if->msk_rxq, PREF_UNIT_CTRL_REG),
	    PREF_UNIT_RST_SET);
	/* Reset the RAM Buffer receive queue. */
	CSR_WRITE_1(sc, RB_ADDR(sc_if->msk_rxq, RB_CTRL), RB_RST_SET);
	/* Reset Rx MAC FIFO. */
	CSR_WRITE_4(sc, MR_ADDR(sc_if->msk_port, RX_GMF_CTRL_T), GMF_RST_SET);

	/* Free Rx and Tx mbufs still in the queues. */
	for (i = 0; i < MSK_RX_RING_CNT; i++) {
		rxd = &sc_if->msk_cdata.msk_rxdesc[i];
		if (rxd->rx_m != NULL) {
			bus_dmamap_sync(sc_if->msk_cdata.msk_rx_tag,
			    rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
			bus_dmamap_unload(sc_if->msk_cdata.msk_rx_tag,
			    rxd->rx_dmamap);
			m_freem(rxd->rx_m);
			rxd->rx_m = NULL;
		}
	}
#ifdef MSK_JUMBO
	for (i = 0; i < MSK_JUMBO_RX_RING_CNT; i++) {
		jrxd = &sc_if->msk_cdata.msk_jumbo_rxdesc[i];
		if (jrxd->rx_m != NULL) {
			bus_dmamap_sync(sc_if->msk_cdata.msk_jumbo_rx_tag,
			    jrxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
			bus_dmamap_unload(sc_if->msk_cdata.msk_jumbo_rx_tag,
			    jrxd->rx_dmamap);
			m_freem(jrxd->rx_m);
			jrxd->rx_m = NULL;
		}
	}
#endif
	for (i = 0; i < MSK_TX_RING_CNT; i++) {
		txd = &sc_if->msk_cdata.msk_txdesc[i];
		if (txd->tx_m != NULL) {
			bus_dmamap_sync(sc_if->msk_cdata.msk_tx_tag,
			    txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
			bus_dmamap_unload(sc_if->msk_cdata.msk_tx_tag,
			    txd->tx_dmamap);
			m_freem(txd->tx_m);
			txd->tx_m = NULL;
		}
	}

	/*
	 * Mark the interface down.
	 */
	ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
	sc_if->msk_link = 0;
}

static int
mskc_sysctl_proc_limit(SYSCTL_HANDLER_ARGS)
{
	return sysctl_int_range(oidp, arg1, arg2, req,
				MSK_PROC_MIN, MSK_PROC_MAX);
}

static int
mskc_sysctl_intr_rate(SYSCTL_HANDLER_ARGS)
{
	struct msk_softc *sc = arg1;
	struct lwkt_serialize *serializer = &sc->msk_serializer;
	int error = 0, v;

	lwkt_serialize_enter(serializer);

	v = sc->msk_intr_rate;
	error = sysctl_handle_int(oidp, &v, 0, req);
	if (error || req->newptr == NULL)
		goto back;
	if (v < 0) {
		error = EINVAL;
		goto back;
	}

	if (sc->msk_intr_rate != v) {
		int flag = 0, i;

		sc->msk_intr_rate = v;
		for (i = 0; i < 2; ++i) {
			if (sc->msk_if[i] != NULL) {
				flag |= sc->msk_if[i]->
					arpcom.ac_if.if_flags & IFF_RUNNING;
			}
		}
		if (flag)
			mskc_set_imtimer(sc);
	}
back:
	lwkt_serialize_exit(serializer);
	return error;
}

static int
msk_dmamem_create(device_t dev, bus_size_t size, bus_dma_tag_t *dtag,
		  void **addr, bus_addr_t *paddr, bus_dmamap_t *dmap)
{
	struct msk_if_softc *sc_if = device_get_softc(dev);
	bus_dmamem_t dmem;
	int error;

	error = bus_dmamem_coherent(sc_if->msk_cdata.msk_parent_tag,
			MSK_RING_ALIGN, 0,
			BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR,
			size, BUS_DMA_WAITOK | BUS_DMA_ZERO, &dmem);
	if (error) {
		device_printf(dev, "can't create coherent DMA memory\n");
		return error;
	}

	*dtag = dmem.dmem_tag;
	*dmap = dmem.dmem_map;
	*addr = dmem.dmem_addr;
	*paddr = dmem.dmem_busaddr;

	return 0;
}

static void
msk_dmamem_destroy(bus_dma_tag_t dtag, void *addr, bus_dmamap_t dmap)
{
	if (dtag != NULL) {
		bus_dmamap_unload(dtag, dmap);
		bus_dmamem_free(dtag, addr, dmap);
		bus_dma_tag_destroy(dtag);
	}
}

static void
mskc_set_imtimer(struct msk_softc *sc)
{
	if (sc->msk_intr_rate > 0) {
		/*
		 * XXX myk(4) seems to use 125MHz for EC/FE/XL
		 *     and 78.125MHz for rest of chip types
		 */
		CSR_WRITE_4(sc, B2_IRQM_INI,
			    MSK_USECS(sc, 1000000 / sc->msk_intr_rate));
		CSR_WRITE_4(sc, B2_IRQM_MSK, sc->msk_intrmask);
		CSR_WRITE_4(sc, B2_IRQM_CTRL, TIM_START);
	} else {
		CSR_WRITE_4(sc, B2_IRQM_CTRL, TIM_STOP);
	}
}