diff --git a/sys/amd64/conf/GENERIC b/sys/amd64/conf/GENERIC
index 0a2a5198484b..b44d0116a55e 100644
--- a/sys/amd64/conf/GENERIC
+++ b/sys/amd64/conf/GENERIC
@@ -176,6 +176,7 @@ device		ppi		# Parallel port interface device
 device		de		# DEC/Intel DC21x4x (``Tulip'')
 device		em		# Intel PRO/1000 adapter Gigabit Ethernet Card
 device		ixgb		# Intel PRO/10GbE Ethernet Card
+device		nve		# nVidia nForce MCP on-board Ethernet Networking
 device		txp		# 3Com 3cR990 (``Typhoon'')
 device		vx		# 3Com 3c590, 3c595 (``Vortex'')
 
diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64
index 58296b713ad5..baa6d0a686e7 100644
--- a/sys/conf/files.amd64
+++ b/sys/conf/files.amd64
@@ -57,6 +57,16 @@ opt_ah.h			optional	ath_hal			\
 	no-obj no-implicit-rule before-depend				\
 	clean		"opt_ah.h"
 #
+nvenetlib.o			optional	nve pci			\
+	dependency	"$S/contrib/dev/nve/amd64/nvenetlib.o.bz2.uu"	\
+	compile-with	"uudecode $S/contrib/dev/nve/amd64/nvenetlib.o.bz2.uu ; bzip2 -df nvenetlib.o.bz2" \
+ 	no-implicit-rule
+#
+os+%DIKED-nve.h		optional	nve pci			\
+	dependency	"$S/contrib/dev/nve/os.h"			\
+	compile-with	"sed -e 's/^.*#include.*phy\.h.*$$//' $S/contrib/dev/nve/os.h > os+%DIKED-nve.h" \
+	no-implicit-rule no-obj before-depend
+#
 amd64/acpica/OsdEnvironment.c	optional	acpi
 amd64/acpica/acpi_machdep.c	optional	acpi
 amd64/acpica/acpi_wakeup.c	optional	acpi
@@ -140,6 +150,7 @@ dev/kbd/kbd.c			optional	atkbd
 dev/kbd/kbd.c			optional	sc
 dev/kbd/kbd.c			optional	ukbd
 dev/mem/memutil.c		optional	mem
+dev/nve/if_nve.c		optional	nve pci
 dev/ppc/ppc.c			optional	ppc
 dev/sio/sio.c			optional	sio
 dev/sio/sio_isa.c		optional	sio isa
diff --git a/sys/conf/files.i386 b/sys/conf/files.i386
index edaaba8f7e2a..2ae9a4da035f 100644
--- a/sys/conf/files.i386
+++ b/sys/conf/files.i386
@@ -61,6 +61,16 @@ opt_ah.h			optional	ath_hal			\
 	no-obj no-implicit-rule before-depend				\
 	clean		"opt_ah.h"
 #
+nvenetlib.o			optional	nve pci			\
+	dependency	"$S/contrib/dev/nve/i386/nvenetlib.o.bz2.uu"	\
+	compile-with	"uudecode $S/contrib/dev/nve/i386/nvenetlib.o.bz2.uu ; bzip2 -df nvenetlib.o.bz2" \
+	no-implicit-rule
+#
+os+%DIKED-nve.h		optional	nve pci			\
+	dependency	"$S/contrib/dev/nve/os.h"			\
+	compile-with	"sed -e 's/^.*#include.*phy\.h.*$$//' $S/contrib/dev/nve/os.h > os+%DIKED-nve.h" \
+	no-implicit-rule no-obj before-depend
+#
 hptmvraid.o			optional	hptmv			\
 	dependency	"$S/dev/hptmv/i386-elf.raid.o.uu"	\
 	compile-with	"uudecode < $S/dev/hptmv/i386-elf.raid.o.uu" \
@@ -176,6 +186,7 @@ dev/lnc/if_lnc_isa.c		optional lnc isa
 dev/mem/memutil.c		optional mem
 dev/mse/mse.c			optional mse
 dev/mse/mse_isa.c		optional mse isa
+dev/nve/if_nve.c		optional nve pci
 dev/ppc/ppc.c			optional ppc
 dev/ppc/ppc_puc.c		optional ppc puc pci
 dev/random/nehemiah.c		optional random
diff --git a/sys/dev/nve/if_nve.c b/sys/dev/nve/if_nve.c
new file mode 100644
index 000000000000..f65caeff3609
--- /dev/null
+++ b/sys/dev/nve/if_nve.c
@@ -0,0 +1,1700 @@
+/*
+ * Copyright (c) 2003,2004 by Quinton Dolan <q@onthenet.com.au>. 
+ * 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.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND ANY
+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
+ * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ * 
+ * $Id: if_nv.c,v 1.19 2004/08/12 14:00:05 q Exp $
+ */
+
+/*
+ * NVIDIA nForce MCP Networking Adapter driver
+ * 
+ * This is a port of the NVIDIA MCP Linux ethernet driver distributed by NVIDIA
+ * through their web site.
+ * 
+ * All mainstream nForce and nForce2 motherboards are supported. This module
+ * is as stable, sometimes more stable, than the linux version. (Recent
+ * Linux stability issues seem to be related to some issues with newer
+ * distributions using GCC 3.x, however this don't appear to effect FreeBSD
+ * 5.x).
+ * 
+ * In accordance with the NVIDIA distribution license it is necessary to
+ * link this module against the nvlibnet.o binary object included in the
+ * Linux driver source distribution. The binary component is not modified in
+ * any way and is simply linked against a FreeBSD equivalent of the nvnet.c
+ * linux kernel module "wrapper".
+ * 
+ * The Linux driver uses a common code API that is shared between Win32 and
+ * i386 Linux. This abstracts the low level driver functions and uses
+ * callbacks and hooks to access the underlying hardware device. By using
+ * this same API in a FreeBSD kernel module it is possible to support the
+ * hardware without breaching the Linux source distributions licensing
+ * requirements, or obtaining the hardware programming specifications.
+ * 
+ * Although not conventional, it works, and given the relatively small
+ * amount of hardware centric code, it's hopefully no more buggy than its
+ * linux counterpart.
+ *
+ * NVIDIA now suppport the nForce3 AMD64 platform, however I have been
+ * unable to access such a system to verify support. However, the code is
+ * reported to work with little modification when compiled with the AMD64
+ * version of the NVIDIA Linux library. All that should be necessary to make
+ * the driver work is to link it directly into the kernel, instead of as a
+ * module, and apply the docs/amd64.diff patch in this source distribution to
+ * the NVIDIA Linux driver source.
+ *
+ * This driver should work on all versions of FreeBSD since 4.9/5.1 as well
+ * as recent versions of DragonFly.
+ *
+ * Written by Quinton Dolan <q@onthenet.com.au> 
+ * Portions based on existing FreeBSD network drivers. 
+ * NVIDIA API usage derived from distributed NVIDIA NVNET driver source files.
+ * 
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/sockio.h>
+#include <sys/mbuf.h>
+#include <sys/malloc.h>
+#include <sys/kernel.h>
+#include <sys/socket.h>
+#include <sys/sysctl.h>
+#include <sys/queue.h>
+#include <sys/module.h>
+
+#include <net/if.h>
+#include <net/if_arp.h>
+#include <net/ethernet.h>
+#include <net/if_dl.h>
+#include <net/if_media.h>
+#include <net/bpf.h>
+#include <net/if_vlan_var.h>
+
+#include <machine/bus_memio.h>
+#include <machine/bus.h>
+#include <machine/resource.h>
+
+#include <vm/vm.h>		/* for vtophys */
+#include <vm/pmap.h>		/* for vtophys */
+#include <machine/clock.h>	/* for DELAY */
+#include <sys/bus.h>
+#include <sys/rman.h>
+
+#include <dev/pci/pcireg.h>
+#include <dev/pci/pcivar.h>
+#include <dev/mii/mii.h>
+#include <dev/mii/miivar.h>
+#include "miibus_if.h"
+
+/* Include NVIDIA Linux driver header files */
+#define	linux
+#include <contrib/dev/nve/basetype.h>
+#include <contrib/dev/nve/phy.h>
+#include "os+%DIKED-nve.h"
+#include <contrib/dev/nve/drvinfo.h>
+#include <contrib/dev/nve/adapter.h>
+#undef linux
+
+#include <dev/nve/if_nvereg.h>
+
+MODULE_DEPEND(nve, pci, 1, 1, 1);
+MODULE_DEPEND(nve, ether, 1, 1, 1);
+MODULE_DEPEND(nve, miibus, 1, 1, 1);
+
+static int      nve_probe(device_t);
+static int      nve_attach(device_t);
+static int      nve_detach(device_t);
+static void     nve_init(void *);
+static void     nve_stop(struct nve_softc *);
+static void     nve_shutdown(device_t);
+static int      nve_init_rings(struct nve_softc *);
+static void     nve_free_rings(struct nve_softc *);
+
+static void     nve_ifstart(struct ifnet *);
+static int      nve_ioctl(struct ifnet *, u_long, caddr_t);
+static void     nve_intr(void *);
+static void     nve_tick(void *);
+static void     nve_setmulti(struct nve_softc *);
+static void     nve_watchdog(struct ifnet *);
+static void     nve_update_stats(struct nve_softc *);
+
+static int      nve_ifmedia_upd(struct ifnet *);
+static void     nve_ifmedia_sts(struct ifnet *, struct ifmediareq *);
+static int      nve_miibus_readreg(device_t, int, int);
+static void     nve_miibus_writereg(device_t, int, int, int);
+
+static void     nve_dmamap_cb(void *, bus_dma_segment_t *, int, int);
+static void     nve_dmamap_tx_cb(void *, bus_dma_segment_t *, int, bus_size_t, int);
+
+static NV_SINT32 nve_osalloc(PNV_VOID, PMEMORY_BLOCK);
+static NV_SINT32 nve_osfree(PNV_VOID, PMEMORY_BLOCK);
+static NV_SINT32 nve_osallocex(PNV_VOID, PMEMORY_BLOCKEX);
+static NV_SINT32 nve_osfreeex(PNV_VOID, PMEMORY_BLOCKEX);
+static NV_SINT32 nve_osclear(PNV_VOID, PNV_VOID, NV_SINT32);
+static NV_SINT32 nve_osdelay(PNV_VOID, NV_UINT32);
+static NV_SINT32 nve_osallocrxbuf(PNV_VOID, PMEMORY_BLOCK, PNV_VOID *);
+static NV_SINT32 nve_osfreerxbuf(PNV_VOID, PMEMORY_BLOCK, PNV_VOID);
+static NV_SINT32 nve_ospackettx(PNV_VOID, PNV_VOID, NV_UINT32);
+static NV_SINT32 nve_ospacketrx(PNV_VOID, PNV_VOID, NV_UINT32, NV_UINT8 *, NV_UINT8);
+static NV_SINT32 nve_oslinkchg(PNV_VOID, NV_SINT32);
+static NV_SINT32 nve_osalloctimer(PNV_VOID, PNV_VOID *);
+static NV_SINT32 nve_osfreetimer(PNV_VOID, PNV_VOID);
+static NV_SINT32 nve_osinittimer(PNV_VOID, PNV_VOID, PTIMER_FUNC, PNV_VOID);
+static NV_SINT32 nve_ossettimer(PNV_VOID, PNV_VOID, NV_UINT32);
+static NV_SINT32 nve_oscanceltimer(PNV_VOID, PNV_VOID);
+
+static NV_SINT32 nve_ospreprocpkt(PNV_VOID, PNV_VOID, PNV_VOID *, NV_UINT8 *, NV_UINT8);
+static PNV_VOID  nve_ospreprocpktnopq(PNV_VOID, PNV_VOID);
+static NV_SINT32 nve_osindicatepkt(PNV_VOID, PNV_VOID *, NV_UINT32);
+static NV_SINT32 nve_oslockalloc(PNV_VOID, NV_SINT32, PNV_VOID *);
+static NV_SINT32 nve_oslockacquire(PNV_VOID, NV_SINT32, PNV_VOID);
+static NV_SINT32 nve_oslockrelease(PNV_VOID, NV_SINT32, PNV_VOID);
+static PNV_VOID  nve_osreturnbufvirt(PNV_VOID, PNV_VOID);
+
+static device_method_t nve_methods[] = {
+	/* Device interface */
+	DEVMETHOD(device_probe, nve_probe),
+	DEVMETHOD(device_attach, nve_attach),
+	DEVMETHOD(device_detach, nve_detach),
+	DEVMETHOD(device_shutdown, nve_shutdown),
+
+	/* Bus interface */
+	DEVMETHOD(bus_print_child, bus_generic_print_child),
+	DEVMETHOD(bus_driver_added, bus_generic_driver_added),
+
+	/* MII interface */
+	DEVMETHOD(miibus_readreg, nve_miibus_readreg),
+	DEVMETHOD(miibus_writereg, nve_miibus_writereg),
+
+	{0, 0}
+};
+
+static driver_t nve_driver = {
+	"nv",
+	nve_methods,
+	sizeof(struct nve_softc)
+};
+
+static devclass_t nve_devclass;
+
+static int      nve_pollinterval = 0;
+SYSCTL_INT(_hw, OID_AUTO, nve_pollinterval, CTLFLAG_RW,
+	   &nve_pollinterval, 0, "delay between interface polls");
+
+DRIVER_MODULE(nve, pci, nve_driver, nve_devclass, 0, 0);
+DRIVER_MODULE(miibus, nve, miibus_driver, miibus_devclass, 0, 0);
+
+static struct nve_type nve_devs[] = {
+	{NVIDIA_VENDORID, NFORCE_MCPNET1_DEVICEID,
+	"NVIDIA nForce MCP Networking Adapter"},
+	{NVIDIA_VENDORID, NFORCE_MCPNET2_DEVICEID,
+	"NVIDIA nForce MCP2 Networking Adapter"},
+	{NVIDIA_VENDORID, NFORCE_MCPNET3_DEVICEID,
+	"NVIDIA nForce MCP3 Networking Adapter"},
+	{0, 0, NULL}
+};
+
+/* DMA MEM map callback function to get data segment physical address */
+static void
+nve_dmamap_cb(void *arg, bus_dma_segment_t * segs, int nsegs, int error)
+{
+	if (error)
+		return;
+
+	KASSERT(nsegs == 1,
+	    ("Too many DMA segments returned when mapping DMA memory"));
+	*(bus_addr_t *)arg = segs->ds_addr;
+}
+
+/* DMA RX map callback function to get data segment physical address */
+static void
+nve_dmamap_rx_cb(void *arg, bus_dma_segment_t * segs, int nsegs,
+    bus_size_t mapsize, int error)
+{
+	if (error)
+		return;
+	*(bus_addr_t *)arg = segs->ds_addr;
+}
+
+/*
+ * DMA TX buffer callback function to allocate fragment data segment
+ * addresses
+ */
+static void
+nve_dmamap_tx_cb(void *arg, bus_dma_segment_t * segs, int nsegs, bus_size_t mapsize, int error)
+{
+	struct nve_tx_desc *info;
+
+	info = arg;
+	if (error)
+		return;
+	KASSERT(nsegs < NV_MAX_FRAGS,
+	    ("Too many DMA segments returned when mapping mbuf"));
+	info->numfrags = nsegs;
+	bcopy(segs, info->frags, nsegs * sizeof(bus_dma_segment_t));
+}
+
+/* Probe for supported hardware ID's */
+static int
+nve_probe(device_t dev)
+{
+	struct nve_type *t;
+
+	t = nve_devs;
+	/* Check for matching PCI DEVICE ID's */
+	while (t->name != NULL) {
+		if ((pci_get_vendor(dev) == t->vid_id) &&
+		    (pci_get_device(dev) == t->dev_id)) {
+			device_set_desc(dev, t->name);
+			return (0);
+		}
+		t++;
+	}
+
+	return (ENXIO);
+}
+
+/* Attach driver and initialise hardware for use */
+static int
+nve_attach(device_t dev)
+{
+	u_char			eaddr[ETHER_ADDR_LEN];
+	struct nve_softc	*sc;
+	struct ifnet		*ifp;
+	OS_API			*osapi;
+	ADAPTER_OPEN_PARAMS	OpenParams;
+	int			error = 0, i, rid, unit;
+
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_attach - entry\n");
+
+	sc = device_get_softc(dev);
+	unit = device_get_unit(dev);
+
+	/* Allocate mutex */
+	mtx_init(&sc->mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
+	    MTX_DEF | MTX_RECURSE);
+	mtx_init(&sc->osmtx, device_get_nameunit(dev), NULL, MTX_SPIN);
+
+	sc->dev = dev;
+	sc->unit = unit;
+
+	/* Preinitialize data structures */
+	bzero(&OpenParams, sizeof(ADAPTER_OPEN_PARAMS));
+
+	/* Enable bus mastering */
+	pci_enable_busmaster(dev);
+
+	/* Allocate memory mapped address space */
+	rid = NV_RID;
+	sc->res = bus_alloc_resource(dev, SYS_RES_MEMORY, &rid, 0, ~0, 1,
+	    RF_ACTIVE);
+
+	if (sc->res == NULL) {
+		device_printf(dev, "couldn't map memory\n");
+		error = ENXIO;
+		goto fail;
+	}
+	sc->sc_st = rman_get_bustag(sc->res);
+	sc->sc_sh = rman_get_bushandle(sc->res);
+
+	/* Allocate interrupt */
+	rid = 0;
+	sc->irq = bus_alloc_resource(dev, SYS_RES_IRQ, &rid, 0, ~0, 1,
+	    RF_SHAREABLE | RF_ACTIVE);
+
+	if (sc->irq == NULL) {
+		device_printf(dev, "couldn't map interrupt\n");
+		error = ENXIO;
+		goto fail;
+	}
+	/* Allocate DMA tags */
+	error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
+		     BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES * NV_MAX_FRAGS,
+				   NV_MAX_FRAGS, MCLBYTES, 0,
+				   busdma_lock_mutex, &Giant,
+				   &sc->mtag);
+	if (error) {
+		device_printf(dev, "couldn't allocate dma tag\n");
+		goto fail;
+	}
+	error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
+	    BUS_SPACE_MAXADDR, NULL, NULL,
+	    sizeof(struct nve_rx_desc) * RX_RING_SIZE, 1,
+	    sizeof(struct nve_rx_desc) * RX_RING_SIZE, 0,
+	    busdma_lock_mutex, &Giant,
+	    &sc->rtag);
+	if (error) {
+		device_printf(dev, "couldn't allocate dma tag\n");
+		goto fail;
+	}
+	error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
+	    BUS_SPACE_MAXADDR, NULL, NULL,
+	    sizeof(struct nve_tx_desc) * TX_RING_SIZE, 1,
+	    sizeof(struct nve_tx_desc) * TX_RING_SIZE, 0,
+	    busdma_lock_mutex, &Giant,
+	    &sc->ttag);
+	if (error) {
+		device_printf(dev, "couldn't allocate dma tag\n");
+		goto fail;
+	}
+	/* Allocate DMA safe memory and get the DMA addresses. */
+	error = bus_dmamem_alloc(sc->ttag, (void **)&sc->tx_desc,
+	    BUS_DMA_WAITOK, &sc->tmap);
+	if (error) {
+		device_printf(dev, "couldn't allocate dma memory\n");
+		goto fail;
+	}
+	bzero(sc->tx_desc, sizeof(struct nve_tx_desc) * TX_RING_SIZE);
+	error = bus_dmamap_load(sc->ttag, sc->tmap, sc->tx_desc,
+		    sizeof(struct nve_tx_desc) * TX_RING_SIZE, nve_dmamap_cb,
+		    &sc->tx_addr, 0);
+	if (error) {
+		device_printf(dev, "couldn't map dma memory\n");
+		goto fail;
+	}
+	error = bus_dmamem_alloc(sc->rtag, (void **)&sc->rx_desc,
+	    BUS_DMA_WAITOK, &sc->rmap);
+	if (error) {
+		device_printf(dev, "couldn't allocate dma memory\n");
+		goto fail;
+	}
+	bzero(sc->rx_desc, sizeof(struct nve_rx_desc) * RX_RING_SIZE);
+	error = bus_dmamap_load(sc->rtag, sc->rmap, sc->rx_desc,
+	    sizeof(struct nve_rx_desc) * RX_RING_SIZE, nve_dmamap_cb,
+	    &sc->rx_addr, 0);
+	if (error) {
+		device_printf(dev, "couldn't map dma memory\n");
+		goto fail;
+	}
+	/* Initialize rings. */
+	if (nve_init_rings(sc)) {
+		device_printf(dev, "failed to init rings\n");
+		error = ENXIO;
+		goto fail;
+	}
+	/* Setup NVIDIA API callback routines */
+	osapi				= &sc->osapi;
+	osapi->pOSCX			= sc;
+	osapi->pfnAllocMemory		= nve_osalloc;
+	osapi->pfnFreeMemory		= nve_osfree;
+	osapi->pfnAllocMemoryEx		= nve_osallocex;
+	osapi->pfnFreeMemoryEx		= nve_osfreeex;
+	osapi->pfnClearMemory		= nve_osclear;
+	osapi->pfnStallExecution	= nve_osdelay;
+	osapi->pfnAllocReceiveBuffer	= nve_osallocrxbuf;
+	osapi->pfnFreeReceiveBuffer	= nve_osfreerxbuf;
+	osapi->pfnPacketWasSent		= nve_ospackettx;
+	osapi->pfnPacketWasReceived	= nve_ospacketrx;
+	osapi->pfnLinkStateHasChanged	= nve_oslinkchg;
+	osapi->pfnAllocTimer		= nve_osalloctimer;
+	osapi->pfnFreeTimer		= nve_osfreetimer;
+	osapi->pfnInitializeTimer	= nve_osinittimer;
+	osapi->pfnSetTimer		= nve_ossettimer;
+	osapi->pfnCancelTimer		= nve_oscanceltimer;
+	osapi->pfnPreprocessPacket	= nve_ospreprocpkt;
+	osapi->pfnPreprocessPacketNopq	= nve_ospreprocpktnopq;
+	osapi->pfnIndicatePackets	= nve_osindicatepkt;
+	osapi->pfnLockAlloc		= nve_oslockalloc;
+	osapi->pfnLockAcquire		= nve_oslockacquire;
+	osapi->pfnLockRelease		= nve_oslockrelease;
+	osapi->pfnReturnBufferVirtual	= nve_osreturnbufvirt;
+
+	sc->linkup = FALSE;
+	sc->max_frame_size = ETHERMTU + ETHER_HDR_LEN + FCS_LEN;
+
+	/* TODO - We don't support hardware offload yet */
+	sc->hwmode = 1;
+	sc->media = 0;
+
+	/* Set NVIDIA API startup parameters */
+	OpenParams.MaxDpcLoop = 2;
+	OpenParams.MaxRxPkt = RX_RING_SIZE;
+	OpenParams.MaxTxPkt = TX_RING_SIZE;
+	OpenParams.SentPacketStatusSuccess = 1;
+	OpenParams.SentPacketStatusFailure = 0;
+	OpenParams.MaxRxPktToAccumulate = 6;
+	OpenParams.ulPollInterval = nve_pollinterval;
+	OpenParams.SetForcedModeEveryNthRxPacket = 0;
+	OpenParams.SetForcedModeEveryNthTxPacket = 0;
+	OpenParams.RxForcedInterrupt = 0;
+	OpenParams.TxForcedInterrupt = 0;
+	OpenParams.pOSApi = osapi;
+	OpenParams.pvHardwareBaseAddress = rman_get_virtual(sc->res);
+	OpenParams.bASFEnabled = 0;
+	OpenParams.ulDescriptorVersion = sc->hwmode;
+	OpenParams.ulMaxPacketSize = sc->max_frame_size;
+	OpenParams.DeviceId = pci_get_device(dev);
+
+	/* Open NVIDIA Hardware API */
+	error = ADAPTER_Open(&OpenParams, (void **)&(sc->hwapi), &sc->phyaddr);
+	if (error) {
+		device_printf(dev,
+		    "failed to open NVIDIA Hardware API: 0x%x\n", error);
+		goto fail;
+	}
+	
+	/* TODO - Add support for MODE2 hardware offload */ 
+	
+	bzero(&sc->adapterdata, sizeof(sc->adapterdata));
+	
+	sc->adapterdata.ulMediaIF = sc->media;
+	sc->adapterdata.ulModeRegTxReadCompleteEnable = 1;
+	sc->hwapi->pfnSetCommonData(sc->hwapi->pADCX, &sc->adapterdata);
+	
+	/* MAC is loaded backwards into h/w reg */
+	sc->hwapi->pfnGetNodeAddress(sc->hwapi->pADCX, sc->original_mac_addr);
+	for (i = 0; i < 6; i++) {
+		eaddr[i] = sc->original_mac_addr[5 - i];
+	}
+	sc->hwapi->pfnSetNodeAddress(sc->hwapi->pADCX, eaddr);
+	bcopy(eaddr, (char *)&sc->sc_macaddr, ETHER_ADDR_LEN);
+
+	/* Display ethernet address ,... */
+	device_printf(dev, "Ethernet address %6D\n", sc->sc_macaddr, ":");
+
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: do mii_phy_probe\n");
+
+	/* Probe device for MII interface to PHY */
+	if (mii_phy_probe(dev, &sc->miibus, nve_ifmedia_upd, nve_ifmedia_sts)) {
+		device_printf(dev, "MII without any phy!\n");
+		error = ENXIO;
+		goto fail;
+	}
+	/* Setup interface parameters */
+	ifp = &sc->sc_if;
+	ifp->if_softc = sc;
+	if_initname(ifp, "nve", unit);
+	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
+	ifp->if_ioctl = nve_ioctl;
+	ifp->if_output = ether_output;
+	ifp->if_start = nve_ifstart;
+	ifp->if_watchdog = nve_watchdog;
+	ifp->if_timer = 0;
+	ifp->if_init = nve_init;
+	ifp->if_mtu = ETHERMTU;
+	ifp->if_baudrate = IF_Mbps(100);
+	ifp->if_snd.ifq_maxlen = TX_RING_SIZE - 1;
+	ifp->if_capabilities |= IFCAP_VLAN_MTU;
+
+	/* Attach to OS's managers. */
+	ether_ifattach(ifp, sc->sc_macaddr);
+	callout_handle_init(&sc->stat_ch);
+
+	/* Activate our interrupt handler. - attach last to avoid lock */
+	error = bus_setup_intr(sc->dev, sc->irq, INTR_TYPE_NET, nve_intr,
+	    sc, &sc->sc_ih);
+	if (error) {
+		device_printf(sc->dev, "couldn't set up interrupt handler\n");
+		goto fail;
+	}
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_attach - exit\n");
+
+fail:
+	if (error)
+		nve_detach(dev);
+
+	return (error);
+}
+
+/* Detach interface for module unload */
+static int
+nve_detach(device_t dev)
+{
+	struct nve_softc *sc = device_get_softc(dev);
+	struct ifnet *ifp;
+
+	KASSERT(mtx_initialized(&sc->mtx), ("mutex not initialized"));
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: nve_detach - entry\n");
+
+	ifp = &sc->arpcom.ac_if;
+
+	if (device_is_attached(dev)) {
+		nve_stop(sc);
+		ether_ifdetach(ifp);
+	}
+
+	if (sc->miibus)
+		device_delete_child(dev, sc->miibus);
+	bus_generic_detach(dev);
+
+	/* Reload unreversed address back into MAC in original state */
+	if (sc->original_mac_addr)
+		sc->hwapi->pfnSetNodeAddress(sc->hwapi->pADCX,
+		    sc->original_mac_addr);
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: do pfnClose\n");
+	/* Detach from NVIDIA hardware API */
+	if (sc->hwapi->pfnClose)
+		sc->hwapi->pfnClose(sc->hwapi->pADCX, FALSE);
+	/* Release resources */
+	if (sc->sc_ih)
+		bus_teardown_intr(sc->dev, sc->irq, sc->sc_ih);
+	if (sc->irq)
+		bus_release_resource(sc->dev, SYS_RES_IRQ, 0, sc->irq);
+	if (sc->res)
+		bus_release_resource(sc->dev, SYS_RES_MEMORY, NV_RID, sc->res);
+
+	nve_free_rings(sc);
+
+	if (sc->tx_desc) {
+		bus_dmamap_unload(sc->rtag, sc->rmap);
+		bus_dmamem_free(sc->rtag, sc->rx_desc, sc->rmap);
+		bus_dmamap_destroy(sc->rtag, sc->rmap);
+	}
+	if (sc->mtag)
+		bus_dma_tag_destroy(sc->mtag);
+	if (sc->ttag)
+		bus_dma_tag_destroy(sc->ttag);
+	if (sc->rtag)
+		bus_dma_tag_destroy(sc->rtag);
+
+	NVE_UNLOCK(sc);
+	mtx_destroy(&sc->mtx);
+	mtx_destroy(&sc->osmtx);
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: nve_detach - exit\n");
+
+	return (0);
+}
+
+/* Initialise interface and start it "RUNNING" */
+static void
+nve_init(void *xsc)
+{
+	struct nve_softc *sc = xsc;
+	struct ifnet *ifp;
+	int error;
+
+	NVE_LOCK(sc);
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_init - entry (%d)\n", sc->linkup);
+
+	ifp = &sc->sc_if;
+
+	/* Do nothing if already running */
+	if (ifp->if_flags & IFF_RUNNING)
+		goto fail;
+
+	nve_stop(sc);
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: do pfnInit\n");
+
+	/* Setup Hardware interface and allocate memory structures */
+	error = sc->hwapi->pfnInit(sc->hwapi->pADCX, 
+	    0, /* force speed */ 
+	    0, /* force full duplex */
+	    0, /* force mode */
+	    0, /* force async mode */
+	    &sc->linkup);
+
+	if (error) {
+		device_printf(sc->dev,
+		    "failed to start NVIDIA Hardware interface\n");
+		goto fail;
+	}
+	/* Set the MAC address */
+	sc->hwapi->pfnSetNodeAddress(sc->hwapi->pADCX, sc->sc_macaddr);
+	sc->hwapi->pfnEnableInterrupts(sc->hwapi->pADCX);
+	sc->hwapi->pfnStart(sc->hwapi->pADCX);
+
+	/* Setup multicast filter */
+	nve_setmulti(sc);
+	nve_ifmedia_upd(ifp);
+
+	/* Update interface parameters */
+	ifp->if_flags |= IFF_RUNNING;
+	ifp->if_flags &= ~IFF_OACTIVE;
+
+	sc->stat_ch = timeout(nve_tick, sc, hz);
+
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_init - exit\n");
+
+fail:
+	NVE_UNLOCK(sc);
+
+	return;
+}
+
+/* Stop interface activity ie. not "RUNNING" */
+static void
+nve_stop(struct nve_softc *sc)
+{
+	struct ifnet *ifp;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_stop - entry\n");
+
+	ifp = &sc->sc_if;
+	ifp->if_timer = 0;
+
+	/* Cancel tick timer */
+	untimeout(nve_tick, sc, sc->stat_ch);
+
+	/* Stop hardware activity */
+	sc->hwapi->pfnDisableInterrupts(sc->hwapi->pADCX);
+	sc->hwapi->pfnStop(sc->hwapi->pADCX, 0);
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: do pfnDeinit\n");
+	/* Shutdown interface and deallocate memory buffers */
+	if (sc->hwapi->pfnDeinit)
+		sc->hwapi->pfnDeinit(sc->hwapi->pADCX, 0);
+
+	sc->linkup = 0;
+	sc->cur_rx = 0;
+	sc->pending_rxs = 0;
+
+	ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_stop - exit\n");
+
+	NVE_UNLOCK(sc);
+
+	return;
+}
+
+/* Shutdown interface for unload/reboot */
+static void
+nve_shutdown(device_t dev)
+{
+	struct nve_softc *sc;
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: nve_shutdown\n");
+
+	sc = device_get_softc(dev);
+
+	/* Stop hardware activity */
+	nve_stop(sc);
+}
+
+/* Allocate TX ring buffers */
+static int
+nve_init_rings(struct nve_softc *sc)
+{
+	int error, i;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_init_rings - entry\n");
+
+	sc->cur_rx = sc->cur_tx = sc->pending_rxs = sc->pending_txs = 0;
+	/* Initialise RX ring */
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct nve_rx_desc *desc = sc->rx_desc + i;
+		struct nve_map_buffer *buf = &desc->buf;
+
+		buf->mbuf = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
+		if (buf->mbuf == NULL) {
+			device_printf(sc->dev, "couldn't allocate mbuf\n");
+			nve_free_rings(sc);
+			error = ENOBUFS;
+			goto fail;
+		}
+		buf->mbuf->m_len = buf->mbuf->m_pkthdr.len = MCLBYTES;
+		m_adj(buf->mbuf, ETHER_ALIGN);
+
+		error = bus_dmamap_create(sc->mtag, 0, &buf->map);
+		if (error) {
+			device_printf(sc->dev, "couldn't create dma map\n");
+			nve_free_rings(sc);
+			goto fail;
+		}
+		error = bus_dmamap_load_mbuf(sc->mtag, buf->map, buf->mbuf,
+					  nve_dmamap_rx_cb, &desc->paddr, 0);
+		if (error) {
+			device_printf(sc->dev, "couldn't dma map mbuf\n");
+			nve_free_rings(sc);
+			goto fail;
+		}
+		bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_PREREAD);
+
+		desc->buflength = buf->mbuf->m_len;
+		desc->vaddr = mtod(buf->mbuf, caddr_t);
+	}
+	bus_dmamap_sync(sc->rtag, sc->rmap,
+	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
+
+	/* Initialize TX ring */
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		struct nve_tx_desc *desc = sc->tx_desc + i;
+		struct nve_map_buffer *buf = &desc->buf;
+
+		buf->mbuf = NULL;
+
+		error = bus_dmamap_create(sc->mtag, 0, &buf->map);
+		if (error) {
+			device_printf(sc->dev, "couldn't create dma map\n");
+			nve_free_rings(sc);
+			goto fail;
+		}
+	}
+	bus_dmamap_sync(sc->ttag, sc->tmap,
+	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
+
+	DEBUGOUT(NVE_DEBUG_INIT, "nve: nve_init_rings - exit\n");
+
+fail:
+	NVE_UNLOCK(sc);
+
+	return (error);
+}
+
+/* Free the TX ring buffers */
+static void
+nve_free_rings(struct nve_softc *sc)
+{
+	int i;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: nve_free_rings - entry\n");
+
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct nve_rx_desc *desc = sc->rx_desc + i;
+		struct nve_map_buffer *buf = &desc->buf;
+
+		if (buf->mbuf) {
+			bus_dmamap_unload(sc->mtag, buf->map);
+			bus_dmamap_destroy(sc->mtag, buf->map);
+			m_freem(buf->mbuf);
+		}
+		buf->mbuf = NULL;
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		struct nve_tx_desc *desc = sc->tx_desc + i;
+		struct nve_map_buffer *buf = &desc->buf;
+
+		if (buf->mbuf) {
+			bus_dmamap_unload(sc->mtag, buf->map);
+			bus_dmamap_destroy(sc->mtag, buf->map);
+			m_freem(buf->mbuf);
+		}
+		buf->mbuf = NULL;
+	}
+
+	DEBUGOUT(NVE_DEBUG_DEINIT, "nve: nve_free_rings - exit\n");
+
+	NVE_UNLOCK(sc);
+}
+
+/* Main loop for sending packets from OS to interface */
+static void
+nve_ifstart(struct ifnet *ifp)
+{
+	struct nve_softc *sc = ifp->if_softc;
+	struct nve_map_buffer *buf;
+	struct mbuf    *m0, *m;
+	struct nve_tx_desc *desc;
+	ADAPTER_WRITE_DATA txdata;
+	int error, i;
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_ifstart - entry\n");
+
+	/* If link is down/busy or queue is empty do nothing */
+	if (ifp->if_flags & IFF_OACTIVE || ifp->if_snd.ifq_head == NULL)
+		return;
+
+	/* Transmit queued packets until sent or TX ring is full */
+	while (sc->pending_txs < TX_RING_SIZE) {
+		desc = sc->tx_desc + sc->cur_tx;
+		buf = &desc->buf;
+
+		/* Get next packet to send. */
+		IF_DEQUEUE(&ifp->if_snd, m0);
+
+		/* If nothing to send, return. */
+		if (m0 == NULL)
+			return;
+
+		/* Map MBUF for DMA access */
+		error = bus_dmamap_load_mbuf(sc->mtag, buf->map, m0,
+		    nve_dmamap_tx_cb, desc, BUS_DMA_NOWAIT);
+
+		if (error && error != EFBIG) {
+			m_freem(m0);
+			sc->tx_errors++;
+			continue;
+		}
+		/*
+		 * Packet has too many fragments - defrag into new mbuf
+		 * cluster
+		 */
+		if (error) {
+			m = m_defrag(m0, M_DONTWAIT);
+			if (m == NULL) {
+				m_freem(m0);
+				sc->tx_errors++;
+				continue;
+			}
+			m_freem(m0);
+			m0 = m;
+
+			error = bus_dmamap_load_mbuf(sc->mtag, buf->map, m,
+			    nve_dmamap_tx_cb, desc, BUS_DMA_NOWAIT);
+			if (error) {
+				m_freem(m);
+				sc->tx_errors++;
+				continue;
+			}
+		}
+		/* Do sync on DMA bounce buffer */
+		bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_PREWRITE);
+
+		buf->mbuf = m0;
+		txdata.ulNumberOfElements = desc->numfrags;
+		txdata.pvID = (PVOID)desc;
+
+		/* Put fragments into API element list */
+		txdata.ulTotalLength = buf->mbuf->m_len;
+		for (i = 0; i < desc->numfrags; i++) {
+			txdata.sElement[i].ulLength =
+			    (ulong)desc->frags[i].ds_len;
+			txdata.sElement[i].pPhysical =
+			    (PVOID)desc->frags[i].ds_addr;
+		}
+
+		/* Send packet to Nvidia API for transmission */
+		error = sc->hwapi->pfnWrite(sc->hwapi->pADCX, &txdata);
+
+		switch (error) {
+		case ADAPTERERR_NONE:
+			/* Packet was queued in API TX queue successfully */
+			sc->pending_txs++;
+			sc->cur_tx = (sc->cur_tx + 1) % TX_RING_SIZE;
+			break;
+
+		case ADAPTERERR_TRANSMIT_QUEUE_FULL:
+			/* The API TX queue is full - requeue the packet */
+			device_printf(sc->dev,
+			    "nve_ifstart: transmit queue is full\n");
+			ifp->if_flags |= IFF_OACTIVE;
+			bus_dmamap_unload(sc->mtag, buf->map);
+			IF_PREPEND(&ifp->if_snd, buf->mbuf);
+			buf->mbuf = NULL;
+			return;
+
+		default:
+			/* The API failed to queue/send the packet so dump it */
+			device_printf(sc->dev, "nve_ifstart: transmit error\n");
+			bus_dmamap_unload(sc->mtag, buf->map);
+			m_freem(buf->mbuf);
+			buf->mbuf = NULL;
+			sc->tx_errors++;
+			return;
+		}
+		/* Set watchdog timer. */
+		ifp->if_timer = 8;
+
+		/* Copy packet to BPF tap */
+		BPF_MTAP(ifp, m0);
+	}
+	ifp->if_flags |= IFF_OACTIVE;
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_ifstart - exit\n");
+}
+
+/* Handle IOCTL events */
+static int
+nve_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
+{
+	struct nve_softc *sc = ifp->if_softc;
+	struct ifreq *ifr = (struct ifreq *) data;
+	struct mii_data *mii;
+	int error = 0;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_IOCTL, "nve: nve_ioctl - entry\n");
+
+	switch (command) {
+	case SIOCSIFMTU:
+		/* Set MTU size */
+		if (ifp->if_mtu == ifr->ifr_mtu)
+			break;
+		if (ifr->ifr_mtu + ifp->if_hdrlen <= MAX_PACKET_SIZE_1518) {
+			ifp->if_mtu = ifr->ifr_mtu;
+			nve_stop(sc);
+			nve_init(sc);
+		} else
+			error = EINVAL;
+		break;
+
+	case SIOCSIFFLAGS:
+		/* Setup interface flags */
+		if (ifp->if_flags & IFF_UP) {
+			if ((ifp->if_flags & IFF_RUNNING) == 0) {
+				nve_init(sc);
+				break;
+			}
+		} else {
+			if (ifp->if_flags & IFF_RUNNING) {
+				nve_stop(sc);
+				break;
+			}
+		}
+		/* Handle IFF_PROMISC and IFF_ALLMULTI flags. */
+		nve_setmulti(sc);
+		break;
+
+	case SIOCADDMULTI:
+	case SIOCDELMULTI:
+		/* Setup multicast filter */
+		if (ifp->if_flags & IFF_RUNNING) {
+			nve_setmulti(sc);
+		}
+		break;
+
+	case SIOCGIFMEDIA:
+	case SIOCSIFMEDIA:
+		/* Get/Set interface media parameters */
+		mii = device_get_softc(sc->miibus);
+		error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
+		break;
+
+	default:
+		/* Everything else we forward to generic ether ioctl */
+		error = ether_ioctl(ifp, (int)command, data);
+		break;
+	}
+
+	DEBUGOUT(NVE_DEBUG_IOCTL, "nve: nve_ioctl - exit\n");
+
+	NVE_UNLOCK(sc);
+
+	return (error);
+}
+
+/* Interrupt service routine */
+static void
+nve_intr(void *arg)
+{
+	struct nve_softc *sc = arg;
+	struct ifnet *ifp = &sc->sc_if;
+
+	DEBUGOUT(NVE_DEBUG_INTERRUPT, "nve: nve_intr - entry\n");
+
+	if (!ifp->if_flags & IFF_UP) {
+		nve_stop(sc);
+		return;
+	}
+	/* Handle interrupt event */
+	if (sc->hwapi->pfnQueryInterrupt(sc->hwapi->pADCX)) {
+		sc->hwapi->pfnHandleInterrupt(sc->hwapi->pADCX);
+		sc->hwapi->pfnEnableInterrupts(sc->hwapi->pADCX);
+	}
+	if (ifp->if_snd.ifq_head != NULL)
+		nve_ifstart(ifp);
+
+	/* If no pending packets we don't need a timeout */
+	if (sc->pending_txs == 0)
+		sc->sc_if.if_timer = 0;
+
+	DEBUGOUT(NVE_DEBUG_INTERRUPT, "nve: nve_intr - exit\n");
+
+	return;
+}
+
+/* Setup multicast filters */
+static void
+nve_setmulti(struct nve_softc *sc)
+{
+	struct ifnet *ifp;
+	struct ifmultiaddr *ifma;
+	PACKET_FILTER hwfilter;
+	int i;
+	u_int8_t andaddr[6], oraddr[6];
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_setmulti - entry\n");
+
+	ifp = &sc->sc_if;
+
+	/* Initialize filter */
+	hwfilter.ulFilterFlags = 0;
+	for (i = 0; i < 6; i++) {
+		hwfilter.acMulticastAddress[i] = 0;
+		hwfilter.acMulticastMask[i] = 0;
+	}
+
+	if (ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI)) {
+		/* Accept all packets */
+		hwfilter.ulFilterFlags |= ACCEPT_ALL_PACKETS;
+		sc->hwapi->pfnSetPacketFilter(sc->hwapi->pADCX, &hwfilter);
+		NVE_UNLOCK(sc);
+		return;
+	}
+	/* Setup multicast filter */
+	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
+		u_char *addrp;
+
+		if (ifma->ifma_addr->sa_family != AF_LINK)
+			continue;
+
+		addrp = LLADDR((struct sockaddr_dl *) ifma->ifma_addr);
+		for (i = 0; i < 6; i++) {
+			u_int8_t mcaddr = addrp[i];
+			andaddr[i] &= mcaddr;
+			oraddr[i] |= mcaddr;
+		}
+	}
+	for (i = 0; i < 6; i++) {
+		hwfilter.acMulticastAddress[i] = andaddr[i] & oraddr[i];
+		hwfilter.acMulticastMask[i] = andaddr[i] | (~oraddr[i]);
+	}
+
+	/* Send filter to NVIDIA API */
+	sc->hwapi->pfnSetPacketFilter(sc->hwapi->pADCX, &hwfilter);
+
+	NVE_UNLOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_RUNNING, "nve: nve_setmulti - exit\n");
+
+	return;
+}
+
+/* Change the current media/mediaopts */
+static int
+nve_ifmedia_upd(struct ifnet *ifp)
+{
+	struct nve_softc *sc = ifp->if_softc;
+	struct mii_data *mii;
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_ifmedia_upd\n");
+
+	mii = device_get_softc(sc->miibus);
+
+	if (mii->mii_instance) {
+		struct mii_softc *miisc;
+		for (miisc = LIST_FIRST(&mii->mii_phys); miisc != NULL;
+		    miisc = LIST_NEXT(miisc, mii_list)) {
+			mii_phy_reset(miisc);
+		}
+	}
+	mii_mediachg(mii);
+
+	return (0);
+}
+
+/* Update current miibus PHY status of media */
+static void
+nve_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
+{
+	struct nve_softc *sc;
+	struct mii_data *mii;
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_ifmedia_sts\n");
+
+	sc = ifp->if_softc;
+	mii = device_get_softc(sc->miibus);
+	mii_pollstat(mii);
+
+	ifmr->ifm_active = mii->mii_media_active;
+	ifmr->ifm_status = mii->mii_media_status;
+
+	return;
+}
+
+/* miibus tick timer - maintain link status */
+static void
+nve_tick(void *xsc)
+{
+	struct nve_softc *sc = xsc;
+	struct mii_data *mii;
+	struct ifnet *ifp;
+
+	NVE_LOCK(sc);
+
+	ifp = &sc->sc_if;
+	nve_update_stats(sc);
+
+	mii = device_get_softc(sc->miibus);
+	mii_tick(mii);
+
+	if (mii->mii_media_status & IFM_ACTIVE &&
+	    IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
+		if (ifp->if_snd.ifq_head != NULL)
+			nve_ifstart(ifp);
+	}
+	sc->stat_ch = timeout(nve_tick, sc, hz);
+
+	NVE_UNLOCK(sc);
+
+	return;
+}
+
+/* Update ifnet data structure with collected interface stats from API */
+static void
+nve_update_stats(struct nve_softc *sc)
+{
+	struct ifnet *ifp = &sc->sc_if;
+	ADAPTER_STATS stats;
+
+	NVE_LOCK(sc);
+
+	if (sc->hwapi) {
+		sc->hwapi->pfnGetStatistics(sc->hwapi->pADCX, &stats);
+
+		ifp->if_ipackets = stats.ulSuccessfulReceptions;
+		ifp->if_ierrors = stats.ulMissedFrames +
+			stats.ulFailedReceptions +
+			stats.ulCRCErrors +
+			stats.ulFramingErrors +
+			stats.ulOverFlowErrors;
+
+		ifp->if_opackets = stats.ulSuccessfulTransmissions;
+		ifp->if_oerrors = sc->tx_errors +
+			stats.ulFailedTransmissions +
+			stats.ulRetryErrors +
+			stats.ulUnderflowErrors +
+			stats.ulLossOfCarrierErrors +
+			stats.ulLateCollisionErrors;
+
+		ifp->if_collisions = stats.ulLateCollisionErrors;
+	}
+	NVE_UNLOCK(sc);
+
+	return;
+}
+
+/* miibus Read PHY register wrapper - calls Nvidia API entry point */
+static int
+nve_miibus_readreg(device_t dev, int phy, int reg)
+{
+	struct nve_softc *sc = device_get_softc(dev);
+	ULONG data;
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_miibus_readreg - entry\n");
+
+	ADAPTER_ReadPhy(sc->hwapi->pADCX, phy, reg, &data);
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_miibus_readreg - exit\n");
+
+	return (data);
+}
+
+/* miibus Write PHY register wrapper - calls Nvidia API entry point */
+static void
+nve_miibus_writereg(device_t dev, int phy, int reg, int data)
+{
+	struct nve_softc *sc = device_get_softc(dev);
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_miibus_writereg - entry\n");
+
+	ADAPTER_WritePhy(sc->hwapi->pADCX, phy, reg, (ulong)data);
+
+	DEBUGOUT(NVE_DEBUG_MII, "nve: nve_miibus_writereg - exit\n");
+
+	return;
+}
+
+/* Watchdog timer to prevent PHY lockups */
+static void
+nve_watchdog(struct ifnet *ifp)
+{
+	struct nve_softc *sc = ifp->if_softc;
+
+	device_printf(sc->dev, "device timeout (%d)\n", sc->pending_txs);
+
+	sc->tx_errors++;
+
+	nve_stop(sc);
+	ifp->if_flags &= ~IFF_RUNNING;
+	nve_init(sc);
+
+	if (ifp->if_snd.ifq_head != NULL)
+		nve_ifstart(ifp);
+
+	return;
+}
+
+/* --- Start of NVOSAPI interface --- */
+
+/* Allocate DMA enabled general use memory for API */
+static NV_SINT32
+nve_osalloc(PNV_VOID ctx, PMEMORY_BLOCK mem)
+{
+	struct nve_softc *sc;
+	bus_addr_t mem_physical;
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osalloc - %d\n", mem->uiLength);
+
+	sc = (struct nve_softc *)ctx;
+
+	mem->pLogical = (PVOID)contigmalloc(mem->uiLength, M_DEVBUF,
+	    M_NOWAIT | M_ZERO, 0, ~0, PAGE_SIZE, 0);
+
+	if (!mem->pLogical) {
+		device_printf(sc->dev, "memory allocation failed\n");
+		return (0);
+	}
+	memset(mem->pLogical, 0, (ulong)mem->uiLength);
+	mem_physical = vtophys(mem->pLogical);
+	mem->pPhysical = (PVOID)mem_physical;
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osalloc 0x%x/0x%x - %d\n",
+	    (uint)mem->pLogical, (uint)mem->pPhysical, (uint)mem->uiLength);
+
+	return (1);
+}
+
+/* Free allocated memory */
+static NV_SINT32
+nve_osfree(PNV_VOID ctx, PMEMORY_BLOCK mem)
+{
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osfree - 0x%x - %d\n",
+	    (uint)mem->pLogical, (uint) mem->uiLength);
+
+	contigfree(mem->pLogical, PAGE_SIZE, M_DEVBUF);
+	return (1);
+}
+
+/* Copied directly from nvnet.c */
+static NV_SINT32
+nve_osallocex(PNV_VOID ctx, PMEMORY_BLOCKEX mem_block_ex)
+{
+	MEMORY_BLOCK mem_block;
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osallocex\n");
+
+	mem_block_ex->pLogical = NULL;
+	mem_block_ex->uiLengthOrig = mem_block_ex->uiLength;
+
+	if ((mem_block_ex->AllocFlags & ALLOC_MEMORY_ALIGNED) &&
+	    (mem_block_ex->AlignmentSize > 1)) {
+		DEBUGOUT(NVE_DEBUG_API, "     aligning on %d\n",
+		    mem_block_ex->AlignmentSize);
+		mem_block_ex->uiLengthOrig += mem_block_ex->AlignmentSize;
+	}
+	mem_block.uiLength = mem_block_ex->uiLengthOrig;
+
+	if (nve_osalloc(ctx, &mem_block) == 0) {
+		return (0);
+	}
+	mem_block_ex->pLogicalOrig = mem_block.pLogical;
+	mem_block_ex->pPhysicalOrigLow = (unsigned long)mem_block.pPhysical;
+	mem_block_ex->pPhysicalOrigHigh = 0;
+
+	mem_block_ex->pPhysical = mem_block.pPhysical;
+	mem_block_ex->pLogical = mem_block.pLogical;
+
+	if (mem_block_ex->uiLength != mem_block_ex->uiLengthOrig) {
+		unsigned int offset;
+		offset = mem_block_ex->pPhysicalOrigLow &
+		    (mem_block_ex->AlignmentSize - 1);
+
+		if (offset) {
+			mem_block_ex->pPhysical =
+			    (PVOID)((ulong)mem_block_ex->pPhysical +
+			    mem_block_ex->AlignmentSize - offset);
+			mem_block_ex->pLogical =
+			    (PVOID)((ulong)mem_block_ex->pLogical +
+			    mem_block_ex->AlignmentSize - offset);
+		} /* if (offset) */
+	} /* if (mem_block_ex->uiLength != *mem_block_ex->uiLengthOrig) */
+	return (1);
+}
+
+/* Copied directly from nvnet.c */
+static NV_SINT32
+nve_osfreeex(PNV_VOID ctx, PMEMORY_BLOCKEX mem_block_ex)
+{
+	MEMORY_BLOCK mem_block;
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osfreeex\n");
+
+	mem_block.pLogical = mem_block_ex->pLogicalOrig;
+	mem_block.pPhysical = (PVOID)((ulong)mem_block_ex->pPhysicalOrigLow);
+	mem_block.uiLength = mem_block_ex->uiLengthOrig;
+
+	return (nve_osfree(ctx, &mem_block));
+}
+
+/* Clear memory region */
+static NV_SINT32
+nve_osclear(PNV_VOID ctx, PNV_VOID mem, NV_SINT32 length)
+{
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osclear\n");
+	memset(mem, 0, length);
+	return (1);
+}
+
+/* Sleep for a tick */
+static NV_SINT32
+nve_osdelay(PNV_VOID ctx, NV_UINT32 usec)
+{
+	DELAY(usec);
+	return (1);
+}
+
+/* Allocate memory for rx buffer */
+static NV_SINT32
+nve_osallocrxbuf(PNV_VOID ctx, PMEMORY_BLOCK mem, PNV_VOID *id)
+{
+	struct nve_softc *sc = ctx;
+	struct nve_rx_desc *desc;
+	struct nve_map_buffer *buf;
+	int error;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osallocrxbuf\n");
+
+	if (sc->pending_rxs == RX_RING_SIZE) {
+		device_printf(sc->dev, "rx ring buffer is full\n");
+		goto fail;
+	}
+	desc = sc->rx_desc + sc->cur_rx;
+	buf = &desc->buf;
+
+	if (buf->mbuf == NULL) {
+		buf->mbuf = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
+		if (buf->mbuf == NULL) {
+			device_printf(sc->dev, "failed to allocate memory\n");
+			goto fail;
+		}
+		buf->mbuf->m_len = buf->mbuf->m_pkthdr.len = MCLBYTES;
+		m_adj(buf->mbuf, ETHER_ALIGN);
+
+		error = bus_dmamap_load_mbuf(sc->mtag, buf->map, buf->mbuf,
+		    nve_dmamap_rx_cb, &desc->paddr, 0);
+		if (error) {
+			device_printf(sc->dev, "failed to dmamap mbuf\n");
+			m_freem(buf->mbuf);
+			buf->mbuf = NULL;
+			goto fail;
+		}
+		bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_PREREAD);
+		desc->buflength = buf->mbuf->m_len;
+		desc->vaddr = mtod(buf->mbuf, caddr_t);
+	}
+	sc->pending_rxs++;
+	sc->cur_rx = (sc->cur_rx + 1) % RX_RING_SIZE;
+
+	mem->pLogical = (void *)desc->vaddr;
+	mem->pPhysical = (void *)desc->paddr;
+	mem->uiLength = desc->buflength;
+	*id = (void *)desc;
+
+	NVE_UNLOCK(sc);
+	return (1);
+	
+fail:
+	NVE_UNLOCK(sc);
+	return (0);
+}
+
+/* Free the rx buffer */
+static NV_SINT32
+nve_osfreerxbuf(PNV_VOID ctx, PMEMORY_BLOCK mem, PNV_VOID id)
+{
+	struct nve_softc *sc = ctx;
+	struct nve_rx_desc *desc;
+	struct nve_map_buffer *buf;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_osfreerxbuf\n");
+
+	desc = (struct nve_rx_desc *) id;
+	buf = &desc->buf;
+
+	if (buf->mbuf) {
+		bus_dmamap_unload(sc->mtag, buf->map);
+		bus_dmamap_destroy(sc->mtag, buf->map);
+		m_freem(buf->mbuf);
+	}
+	sc->pending_rxs--;
+	buf->mbuf = NULL;
+
+	NVE_UNLOCK(sc);
+
+	return (1);
+}
+
+/* This gets called by the Nvidia API after our TX packet has been sent */
+static NV_SINT32
+nve_ospackettx(PNV_VOID ctx, PNV_VOID id, NV_UINT32 success)
+{
+	struct nve_softc *sc = ctx;
+	struct nve_map_buffer *buf;
+	struct nve_tx_desc *desc = (struct nve_tx_desc *) id;
+	struct ifnet *ifp;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_ospackettx\n");
+
+	ifp = &sc->sc_if;
+	buf = &desc->buf;
+	sc->pending_txs--;
+
+	/* Unload and free mbuf cluster */
+	if (buf->mbuf == NULL)
+		goto fail;
+
+	bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_POSTWRITE);
+	bus_dmamap_unload(sc->mtag, buf->map);
+	m_freem(buf->mbuf);
+	buf->mbuf = NULL;
+
+	/* Send more packets if we have them */
+	if (sc->pending_txs < TX_RING_SIZE)
+		sc->sc_if.if_flags &= ~IFF_OACTIVE;
+
+	if (ifp->if_snd.ifq_head != NULL && sc->pending_txs < TX_RING_SIZE)
+		nve_ifstart(ifp);
+
+fail:
+	NVE_UNLOCK(sc);
+
+	return (1);
+}
+
+/* This gets called by the Nvidia API when a new packet has been received */
+/* XXX What is newbuf used for? XXX */
+static NV_SINT32
+nve_ospacketrx(PNV_VOID ctx, PNV_VOID data, NV_UINT32 success, NV_UINT8 *newbuf,
+    NV_UINT8 priority)
+{
+	struct nve_softc *sc = ctx;
+	struct ifnet *ifp;
+	struct nve_rx_desc *desc;
+	struct nve_map_buffer *buf;
+	ADAPTER_READ_DATA *readdata;
+
+	NVE_LOCK(sc);
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_ospacketrx\n");
+
+	ifp = &sc->sc_if;
+
+	readdata = (ADAPTER_READ_DATA *) data;
+	desc = readdata->pvID;
+	buf = &desc->buf;
+	bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_POSTREAD);
+
+	if (success) {
+		/* Sync DMA bounce buffer. */
+		bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_POSTREAD);
+
+		/* First mbuf in packet holds the ethernet and packet headers */
+		buf->mbuf->m_pkthdr.rcvif = ifp;
+		buf->mbuf->m_pkthdr.len = buf->mbuf->m_len =
+		    readdata->ulTotalLength;
+
+		bus_dmamap_unload(sc->mtag, buf->map);
+
+		/* Give mbuf to OS. */
+		(*ifp->if_input) (ifp, buf->mbuf);
+		if (readdata->ulFilterMatch & ADREADFL_MULTICAST_MATCH)
+			ifp->if_imcasts++;
+
+		/* Blat the mbuf pointer, kernel will free the mbuf cluster */
+		buf->mbuf = NULL;
+	} else {
+		bus_dmamap_sync(sc->mtag, buf->map, BUS_DMASYNC_POSTREAD);
+		bus_dmamap_unload(sc->mtag, buf->map);
+		m_freem(buf->mbuf);
+		buf->mbuf = NULL;
+	}
+
+	sc->cur_rx = desc - sc->rx_desc;
+	sc->pending_rxs--;
+
+	NVE_UNLOCK(sc);
+
+	return (1);
+}
+
+/* This gets called by NVIDIA API when the PHY link state changes */
+static NV_SINT32
+nve_oslinkchg(PNV_VOID ctx, NV_SINT32 enabled)
+{
+	struct nve_softc *sc = (struct nve_softc *)ctx;
+	struct ifnet *ifp;
+
+	DEBUGOUT(NVE_DEBUG_API, "nve: nve_oslinkchg\n");
+
+	ifp = &sc->sc_if;
+
+	if (enabled)
+		ifp->if_flags |= IFF_UP;
+	else
+		ifp->if_flags &= ~IFF_UP;
+
+	return (1);
+}
+
+/* Setup a watchdog timer */
+static NV_SINT32
+nve_osalloctimer(PNV_VOID ctx, PNV_VOID *timer)
+{
+	struct nve_softc *sc = (struct nve_softc *)ctx;
+
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_osalloctimer\n");
+
+	callout_handle_init(&sc->ostimer);
+	*timer = &sc->ostimer;
+
+	return (1);
+}
+
+/* Free the timer */
+static NV_SINT32
+nve_osfreetimer(PNV_VOID ctx, PNV_VOID timer)
+{
+
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_osfreetimer\n");
+
+	return (1);
+}
+
+/* Setup timer parameters */
+static NV_SINT32
+nve_osinittimer(PNV_VOID ctx, PNV_VOID timer, PTIMER_FUNC func, PNV_VOID parameters)
+{
+	struct nve_softc *sc = (struct nve_softc *)ctx;
+
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_osinittimer\n");
+
+	sc->ostimer_func = func;
+	sc->ostimer_params = parameters;
+
+	return (1);
+}
+
+/* Set the timer to go off */
+static NV_SINT32
+nve_ossettimer(PNV_VOID ctx, PNV_VOID timer, NV_UINT32 delay)
+{
+	struct nve_softc *sc = ctx;
+
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_ossettimer\n");
+
+	*(struct callout_handle *)timer = timeout(sc->ostimer_func,
+	    sc->ostimer_params, delay);
+
+	return (1);
+}
+
+/* Cancel the timer */
+static NV_SINT32
+nve_oscanceltimer(PNV_VOID ctx, PNV_VOID timer)
+{
+	struct nve_softc *sc = ctx;
+
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_oscanceltimer\n");
+
+	untimeout(sc->ostimer_func, sc->ostimer_params,
+	    *(struct callout_handle *)timer);
+
+	return (1);
+}
+
+static NV_SINT32
+nve_ospreprocpkt(PNV_VOID ctx, PNV_VOID readdata, PNV_VOID *id,
+    NV_UINT8 *newbuffer, NV_UINT8 priority)
+{
+
+	/* Not implemented */
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_ospreprocpkt\n");
+
+	return (1);
+}
+
+static PNV_VOID
+nve_ospreprocpktnopq(PNV_VOID ctx, PNV_VOID readdata)
+{
+
+	/* Not implemented */
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_ospreprocpkt\n");
+
+	return (NULL);
+}
+
+static NV_SINT32
+nve_osindicatepkt(PNV_VOID ctx, PNV_VOID *id, NV_UINT32 pktno)
+{
+
+	/* Not implemented */
+	DEBUGOUT(NVE_DEBUG_BROKEN, "nve: nve_osindicatepkt\n");
+
+	return (1);
+}
+
+/* Allocate mutex context (already done in nve_attach) */
+static NV_SINT32
+nve_oslockalloc(PNV_VOID ctx, NV_SINT32 type, PNV_VOID *pLock)
+{
+	struct nve_softc *sc = (struct nve_softc *)ctx;
+
+	DEBUGOUT(NVE_DEBUG_LOCK, "nve: nve_oslockalloc\n");
+
+	*pLock = (void **)sc;
+
+	return (1);
+}
+
+/* Obtain a spin lock */
+static NV_SINT32
+nve_oslockacquire(PNV_VOID ctx, NV_SINT32 type, PNV_VOID lock)
+{
+
+	DEBUGOUT(NVE_DEBUG_LOCK, "nve: nve_oslockacquire\n");
+
+	NVE_OSLOCK((struct nve_softc *)lock);
+
+	return (1);
+}
+
+/* Release lock */
+static NV_SINT32
+nve_oslockrelease(PNV_VOID ctx, NV_SINT32 type, PNV_VOID lock)
+{
+
+	DEBUGOUT(NVE_DEBUG_LOCK, "nve: nve_oslockrelease\n");
+
+	NVE_OSUNLOCK((struct nve_softc *)lock);
+
+	return (1);
+}
+
+/* I have no idea what this is for */
+static PNV_VOID
+nve_osreturnbufvirt(PNV_VOID ctx, PNV_VOID readdata)
+{
+
+	/* Not implemented */
+	DEBUGOUT(NVE_DEBUG_LOCK, "nve: nve_osreturnbufvirt\n");
+	panic("nve: nve_osreturnbufvirtual not implemented\n");
+
+	return (NULL);
+}
+
+/* --- End on NVOSAPI interface --- */
diff --git a/sys/dev/nve/if_nvereg.h b/sys/dev/nve/if_nvereg.h
new file mode 100644
index 000000000000..66c3d1e6616a
--- /dev/null
+++ b/sys/dev/nve/if_nvereg.h
@@ -0,0 +1,175 @@
+/*
+ * Copyright (c) 2003 by Quinton Dolan <q@onthenet.com.au>.
+ * 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.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS `AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 
+ * SUCH DAMAGE.
+ *
+ * $Id: if_nvreg.h,v 1.6 2004/08/12 14:00:05 q Exp $
+ * $FreeBSD$
+ */
+ 
+#ifndef _IF_NVEREG_H_
+#define _IF_NVEREG_H_
+
+#ifndef NVIDIA_VENDORID
+#define	NVIDIA_VENDORID 0x10DE
+#endif
+
+#define	NFORCE_MCPNET1_DEVICEID 0x01C3
+#define	NFORCE_MCPNET2_DEVICEID 0x0066
+#define	NFORCE_MCPNET3_DEVICEID 0x00D6
+
+#define	NV_RID		0x10
+
+#define	TX_RING_SIZE	64
+#define	RX_RING_SIZE	64
+#define	NV_MAX_FRAGS	63
+
+#define	FCS_LEN 4
+
+#define	NVE_DEBUG		0x0000
+#define	NVE_DEBUG_INIT		0x0001
+#define	NVE_DEBUG_RUNNING	0x0002
+#define	NVE_DEBUG_DEINIT 	0x0004
+#define	NVE_DEBUG_IOCTL		0x0008
+#define	NVE_DEBUG_INTERRUPT	0x0010
+#define	NVE_DEBUG_API		0x0020
+#define	NVE_DEBUG_LOCK		0x0040
+#define	NVE_DEBUG_BROKEN	0x0080
+#define	NVE_DEBUG_MII		0x0100
+#define	NVE_DEBUG_ALL		0xFFFF
+
+#if NVE_DEBUG
+#define	DEBUGOUT(level, fmt, args...) if (NVE_DEBUG & level) \
+    printf(fmt, ## args)
+#else
+#define	DEBUGOUT(level, fmt, args...)
+#endif
+
+typedef unsigned long	ulong;
+
+struct nve_map_buffer {
+	struct mbuf *mbuf;	/* mbuf receiving packet */
+	bus_dmamap_t map;	/* DMA map */	
+};
+
+struct nve_dma_info {
+	bus_dma_tag_t tag;
+	struct nve_map_buffer buf;
+	u_int16_t buflength;
+	caddr_t vaddr;		/* Virtual memory address */
+	bus_addr_t paddr;	/* DMA physical address */
+};
+
+struct nve_rx_desc {
+	struct nve_rx_desc *next;
+	struct nve_map_buffer buf;
+	u_int16_t buflength;
+	caddr_t vaddr;
+	bus_addr_t paddr;
+};
+
+struct nve_tx_desc {
+	/* Don't add anything above this structure */
+	TX_INFO_ADAP TxInfoAdap;
+	struct nve_tx_desc *next;
+	struct nve_map_buffer buf;
+	u_int16_t buflength;
+	u_int32_t numfrags;
+	bus_dma_segment_t frags[NV_MAX_FRAGS + 1];
+};
+
+struct nve_softc {
+	struct arpcom arpcom;	/* interface info */
+	struct resource *res;
+	struct resource *irq;
+
+	ADAPTER_API *hwapi;
+	OS_API osapi;
+		
+	device_t miibus;
+	device_t dev;
+	struct callout_handle stat_ch;
+
+	u_int32_t unit;
+	void *sc_ih;
+	bus_space_tag_t sc_st;
+	bus_space_handle_t sc_sh;
+	bus_dma_tag_t mtag;
+	bus_dma_tag_t rtag;
+	bus_dmamap_t rmap;
+	bus_dma_tag_t ttag;
+	bus_dmamap_t tmap;
+
+	struct nve_rx_desc *rx_desc;
+	struct nve_tx_desc *tx_desc;
+	bus_addr_t rx_addr;
+	bus_addr_t tx_addr;
+	u_int16_t rx_ring_full;
+	u_int16_t tx_ring_full;
+	u_int32_t cur_rx;
+	u_int32_t cur_tx;
+	u_int32_t pending_rxs;
+	u_int32_t pending_txs;
+
+	u_int32_t flags;
+	u_int32_t miicfg;
+	struct mtx mtx;
+	struct mtx osmtx;
+
+	/* Stuff for dealing with the NVIDIA OS API */
+	struct callout_handle ostimer;
+	PTIMER_FUNC ostimer_func;
+	void *ostimer_params;
+	int linkup;
+	ulong tx_errors;
+	NV_UINT32 hwmode;
+	NV_UINT32 max_frame_size;
+	NV_UINT32 phyaddr;
+	NV_UINT32 media;
+	CMNDATA_OS_ADAPTER adapterdata;
+	unsigned char original_mac_addr[6];
+};
+
+struct nve_type {
+	u_int16_t	vid_id;
+	u_int16_t	dev_id;
+	char		*name;
+};
+
+#define sc_if arpcom.ac_if
+#define sc_macaddr arpcom.ac_enaddr
+
+#define NVE_LOCK(_sc)		mtx_lock(&(_sc)->mtx)
+#define NVE_UNLOCK(_sc)		mtx_unlock(&(_sc)->mtx)
+#define NVE_OSLOCK(_sc)		mtx_lock_spin(&(_sc)->osmtx)
+#define NVE_OSUNLOCK(_sc)	mtx_unlock_spin(&(_sc)->osmtx)
+
+#define IF_Kbps(x) ((x) * 1000)			/* kilobits/sec. */
+#define IF_Mbps(x) (IF_Kbps((x) * 1000))	/* megabits/sec. */
+#define ETHER_ALIGN 2
+
+extern int ADAPTER_ReadPhy (PVOID pContext, ULONG ulPhyAddr, ULONG ulReg, ULONG *pulVal);
+extern int ADAPTER_WritePhy (PVOID pContext, ULONG ulPhyAddr, ULONG ulReg, ULONG ulVal);
+extern int ADAPTER_Init (PVOID pContext, USHORT usForcedSpeed, UCHAR ucForceDpx, UCHAR ucForceMode, UINT *puiLinkState);
+
+#endif	/* _IF_NVEREG_H_ */
diff --git a/sys/i386/conf/GENERIC b/sys/i386/conf/GENERIC
index 18cffb76d703..ee36c6e4cd6d 100644
--- a/sys/i386/conf/GENERIC
+++ b/sys/i386/conf/GENERIC
@@ -195,6 +195,7 @@ device		ppi		# Parallel port interface device
 device		de		# DEC/Intel DC21x4x (``Tulip'')
 device		em		# Intel PRO/1000 adapter Gigabit Ethernet Card
 device		ixgb		# Intel PRO/10GbE Ethernet Card
+device		nve		# nVidia nForce MCP on-board Ethernet Networking
 device		txp		# 3Com 3cR990 (``Typhoon'')
 device		vx		# 3Com 3c590, 3c595 (``Vortex'')
 
diff --git a/sys/modules/Makefile b/sys/modules/Makefile
index 666150a92d8b..13e67f926e94 100644
--- a/sys/modules/Makefile
+++ b/sys/modules/Makefile
@@ -154,6 +154,7 @@ SUBDIR=	${_3dfx} \
 	nfsclient \
 	nfsserver \
 	nge \
+	${_nve} \
 	nmdm \
 	${_nsp} \
 	ntfs \
@@ -331,6 +332,7 @@ _ncp=		ncp
 _ncv=		ncv
 _ndis=		ndis
 _nsp=		nsp
+_nve=		nve
 _nwfs=		nwfs
 _oltr=		oltr
 _pccard=	pccard
@@ -416,6 +418,7 @@ _ips=		ips
 #_lnc=		lnc
 _mly=		mly
 _ndis=		ndis
+_nve=		nve
 _safe=		safe
 _scsi_low=	scsi_low
 _smbfs=		smbfs
diff --git a/sys/modules/nve/Makefile b/sys/modules/nve/Makefile
new file mode 100644
index 000000000000..5a055fc864bc
--- /dev/null
+++ b/sys/modules/nve/Makefile
@@ -0,0 +1,15 @@
+# $FreeBSD$
+
+.PATH: ${.CURDIR}/../../dev/nve
+
+KMOD=	if_nve
+SRCS=	if_nve.c if_nvereg.h miidevs.h \
+	device_if.h bus_if.h pci_if.h miibus_if.h
+OBJS+=	nvenetlib.o
+
+CLEANFILES+=	nvenetlib.o
+nvenetlib.o: ${.CURDIR}/../../contrib/dev/nve/${MACHINE}/${.TARGET}.bz2.uu
+	uudecode < ${.CURDIR}/../../contrib/dev/nve/${MACHINE}/${.TARGET}.bz2.uu
+	bzip2 -d ${.TARGET}.bz2
+
+.include <bsd.kmod.mk>