openwrt/target/linux/atheros/files/drivers/net/ar2313/ar2313.c

/*
 * ar2313.c: Linux driver for the Atheros AR231x Ethernet device.
 *
 * Copyright (C) 2004 by Sameer Dekate <sdekate@arubanetworks.com>
 * Copyright (C) 2006 Imre Kaloz <kaloz@openwrt.org>
 * Copyright (C) 2006-2007 Felix Fietkau <nbd@openwrt.org>
 *
 * Thanks to Atheros for providing hardware and documentation
 * enabling me to write this driver.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * Additional credits:
 * 	This code is taken from John Taylor's Sibyte driver and then 
 * 	modified for the AR2313.
 */

#include <linux/autoconf.h>
#include <linux/module.h>
#include <linux/version.h>
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/ioport.h>
#include <linux/pci.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/sockios.h>
#include <linux/pkt_sched.h>
#include <linux/compile.h>
#include <linux/mii.h>
#include <linux/ethtool.h>
#include <linux/ctype.h>
#include <linux/platform_device.h>

#include <net/sock.h>
#include <net/ip.h>

#include <asm/system.h>
#include <asm/io.h>
#include <asm/irq.h>
#include <asm/byteorder.h>
#include <asm/uaccess.h>
#include <asm/bootinfo.h>

#define AR2313_MTU                     1692
#define AR2313_PRIOS                   1
#define AR2313_QUEUES                  (2*AR2313_PRIOS)
#define AR2313_DESCR_ENTRIES           64

#undef INDEX_DEBUG
#define DEBUG     0
#define DEBUG_TX  0
#define DEBUG_RX  0
#define DEBUG_INT 0
#define DEBUG_MC  0
#define DEBUG_ERR 1

#ifndef min
#define min(a,b)	(((a)<(b))?(a):(b))
#endif

#ifndef SMP_CACHE_BYTES
#define SMP_CACHE_BYTES	L1_CACHE_BYTES
#endif

#define AR2313_MBOX_SET_BIT  0x8

#define BOARD_IDX_STATIC	0
#define BOARD_IDX_OVERFLOW	-1

#include "dma.h"
#include "ar2313.h"

/*
 * New interrupt handler strategy:
 *
 * An old interrupt handler worked using the traditional method of
 * replacing an skbuff with a new one when a packet arrives. However
 * the rx rings do not need to contain a static number of buffer
 * descriptors, thus it makes sense to move the memory allocation out
 * of the main interrupt handler and do it in a bottom half handler
 * and only allocate new buffers when the number of buffers in the
 * ring is below a certain threshold. In order to avoid starving the
 * NIC under heavy load it is however necessary to force allocation
 * when hitting a minimum threshold. The strategy for alloction is as
 * follows:
 *
 *     RX_LOW_BUF_THRES    - allocate buffers in the bottom half
 *     RX_PANIC_LOW_THRES  - we are very low on buffers, allocate
 *                           the buffers in the interrupt handler
 *     RX_RING_THRES       - maximum number of buffers in the rx ring
 *
 * One advantagous side effect of this allocation approach is that the
 * entire rx processing can be done without holding any spin lock
 * since the rx rings and registers are totally independent of the tx
 * ring and its registers.  This of course includes the kmalloc's of
 * new skb's. Thus start_xmit can run in parallel with rx processing
 * and the memory allocation on SMP systems.
 *
 * Note that running the skb reallocation in a bottom half opens up
 * another can of races which needs to be handled properly. In
 * particular it can happen that the interrupt handler tries to run
 * the reallocation while the bottom half is either running on another
 * CPU or was interrupted on the same CPU. To get around this the
 * driver uses bitops to prevent the reallocation routines from being
 * reentered.
 *
 * TX handling can also be done without holding any spin lock, wheee
 * this is fun! since tx_csm is only written to by the interrupt
 * handler.
 */

/*
 * Threshold values for RX buffer allocation - the low water marks for
 * when to start refilling the rings are set to 75% of the ring
 * sizes. It seems to make sense to refill the rings entirely from the
 * intrrupt handler once it gets below the panic threshold, that way
 * we don't risk that the refilling is moved to another CPU when the
 * one running the interrupt handler just got the slab code hot in its
 * cache.
 */
#define RX_RING_SIZE		AR2313_DESCR_ENTRIES
#define RX_PANIC_THRES	        (RX_RING_SIZE/4)
#define RX_LOW_THRES	        ((3*RX_RING_SIZE)/4)
#define CRC_LEN                 4
#define RX_OFFSET               2

#define AR2313_BUFSIZE		(AR2313_MTU + ETH_HLEN + CRC_LEN + RX_OFFSET)

#ifdef MODULE
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Sameer Dekate <sdekate@arubanetworks.com>, Imre Kaloz <kaloz@openwrt.org>, Felix Fietkau <nbd@openwrt.org>");
MODULE_DESCRIPTION("AR2313 Ethernet driver");
#endif

#define virt_to_phys(x) ((u32)(x) & 0x1fffffff)

// prototypes
static short armiiread(struct net_device *dev, short phy, short reg);
static void armiiwrite(struct net_device *dev, short phy, short reg,
					   short data);
#ifdef TX_TIMEOUT
static void ar2313_tx_timeout(struct net_device *dev);
#endif
static void ar2313_halt(struct net_device *dev);
static void rx_tasklet_func(unsigned long data);
static void ar2313_multicast_list(struct net_device *dev);

#ifndef ERR
#define ERR(fmt, args...) printk("%s: " fmt, __func__, ##args)
#endif


int __init ar2313_probe(struct platform_device *pdev)
{
	struct net_device *dev;
	struct ar2313_private *sp;
	struct resource *res;
	unsigned long ar_eth_base;
	char buf[64];

	dev = alloc_etherdev(sizeof(struct ar2313_private));

	if (dev == NULL) {
		printk(KERN_ERR
			   "ar2313: Unable to allocate net_device structure!\n");
		return -ENOMEM;
	}

	SET_MODULE_OWNER(dev);
	platform_set_drvdata(pdev, dev);

	sp = dev->priv;
	sp->dev = dev;
	sp->cfg = pdev->dev.platform_data;

	sprintf(buf, "eth%d_membase", pdev->id);
	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, buf);
	if (!res)
		return -ENODEV;

	sp->link = 0;
	ar_eth_base = res->start;
	sp->phy = sp->cfg->phy;

	sprintf(buf, "eth%d_irq", pdev->id);
	dev->irq = platform_get_irq_byname(pdev, buf);

	spin_lock_init(&sp->lock);

	/* initialize func pointers */
	dev->open = &ar2313_open;
	dev->stop = &ar2313_close;
	dev->hard_start_xmit = &ar2313_start_xmit;

	dev->get_stats = &ar2313_get_stats;
	dev->set_multicast_list = &ar2313_multicast_list;
#ifdef TX_TIMEOUT
	dev->tx_timeout = ar2313_tx_timeout;
	dev->watchdog_timeo = AR2313_TX_TIMEOUT;
#endif
	dev->do_ioctl = &ar2313_ioctl;

	// SAMEER: do we need this?
	dev->features |= NETIF_F_SG | NETIF_F_HIGHDMA;

	tasklet_init(&sp->rx_tasklet, rx_tasklet_func, (unsigned long) dev);
	tasklet_disable(&sp->rx_tasklet);

	sp->eth_regs =
		ioremap_nocache(virt_to_phys(ar_eth_base), sizeof(*sp->eth_regs));
	if (!sp->eth_regs) {
		printk("Can't remap eth registers\n");
		return (-ENXIO);
	}

	/* 
	 * When there's only one MAC, PHY regs are typically on ENET0, 
	 * even though the MAC might be on ENET1.
	 * Needto remap PHY regs separately in this case
	 */
	if (virt_to_phys(ar_eth_base) == virt_to_phys(sp->phy_regs))
		sp->phy_regs = sp->eth_regs;
	else {
		sp->phy_regs =
			ioremap_nocache(virt_to_phys(sp->cfg->phy_base),
							sizeof(*sp->phy_regs));
		if (!sp->phy_regs) {
			printk("Can't remap phy registers\n");
			return (-ENXIO);
		}
	}

	sp->dma_regs =
		ioremap_nocache(virt_to_phys(ar_eth_base + 0x1000),
						sizeof(*sp->dma_regs));
	dev->base_addr = (unsigned int) sp->dma_regs;
	if (!sp->dma_regs) {
		printk("Can't remap DMA registers\n");
		return (-ENXIO);
	}

	sp->int_regs = ioremap_nocache(virt_to_phys(sp->cfg->reset_base), 4);
	if (!sp->int_regs) {
		printk("Can't remap INTERRUPT registers\n");
		return (-ENXIO);
	}

	strncpy(sp->name, "Atheros AR231x", sizeof(sp->name) - 1);
	sp->name[sizeof(sp->name) - 1] = '\0';
	memcpy(dev->dev_addr, sp->cfg->macaddr, 6);
	sp->board_idx = BOARD_IDX_STATIC;

	if (ar2313_init(dev)) {
		/* 
		 * ar2313_init() calls ar2313_init_cleanup() on error.
		 */
		kfree(dev);
		return -ENODEV;
	}

	if (register_netdev(dev)) {
		printk("%s: register_netdev failed\n", __func__);
		return -1;
	}

	printk("%s: %s: %02x:%02x:%02x:%02x:%02x:%02x, irq %d\n",
		   dev->name, sp->name,
		   dev->dev_addr[0], dev->dev_addr[1], dev->dev_addr[2],
		   dev->dev_addr[3], dev->dev_addr[4], dev->dev_addr[5], dev->irq);

	/* start link poll timer */
	ar2313_setup_timer(dev);

	return 0;
}

#if 0
static void ar2313_dump_regs(struct net_device *dev)
{
	unsigned int *ptr, i;
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;

	ptr = (unsigned int *) sp->eth_regs;
	for (i = 0; i < (sizeof(ETHERNET_STRUCT) / sizeof(unsigned int));
		 i++, ptr++) {
		printk("ENET: %08x = %08x\n", (int) ptr, *ptr);
	}

	ptr = (unsigned int *) sp->dma_regs;
	for (i = 0; i < (sizeof(DMA) / sizeof(unsigned int)); i++, ptr++) {
		printk("DMA: %08x = %08x\n", (int) ptr, *ptr);
	}

	ptr = (unsigned int *) sp->int_regs;
	for (i = 0; i < (sizeof(INTERRUPT) / sizeof(unsigned int)); i++, ptr++) {
		printk("INT: %08x = %08x\n", (int) ptr, *ptr);
	}

	for (i = 0; i < AR2313_DESCR_ENTRIES; i++) {
		ar2313_descr_t *td = &sp->tx_ring[i];
		printk("Tx desc %2d: %08x %08x %08x %08x\n", i,
			   td->status, td->devcs, td->addr, td->descr);
	}
}
#endif

#ifdef TX_TIMEOUT
static void ar2313_tx_timeout(struct net_device *dev)
{
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;
	unsigned long flags;

#if DEBUG_TX
	printk("Tx timeout\n");
#endif
	spin_lock_irqsave(&sp->lock, flags);
	ar2313_restart(dev);
	spin_unlock_irqrestore(&sp->lock, flags);
}
#endif

#if DEBUG_MC
static void printMcList(struct net_device *dev)
{
	struct dev_mc_list *list = dev->mc_list;
	int num = 0, i;
	while (list) {
		printk("%d MC ADDR ", num);
		for (i = 0; i < list->dmi_addrlen; i++) {
			printk(":%02x", list->dmi_addr[i]);
		}
		list = list->next;
		printk("\n");
	}
}
#endif

/*
 * Set or clear the multicast filter for this adaptor.
 * THIS IS ABSOLUTE CRAP, disabled
 */
static void ar2313_multicast_list(struct net_device *dev)
{
	/* 
	 * Always listen to broadcasts and 
	 * treat IFF bits independently 
	 */
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;
	unsigned int recognise;

	recognise = sp->eth_regs->mac_control;

	if (dev->flags & IFF_PROMISC) {	/* set promiscuous mode */
		recognise |= MAC_CONTROL_PR;
	} else {
		recognise &= ~MAC_CONTROL_PR;
	}

	if ((dev->flags & IFF_ALLMULTI) || (dev->mc_count > 15)) {
#if DEBUG_MC
		printMcList(dev);
		printk("%s: all MULTICAST mc_count %d\n", __FUNCTION__,
			   dev->mc_count);
#endif
		recognise |= MAC_CONTROL_PM;	/* all multicast */
	} else if (dev->mc_count > 0) {
#if DEBUG_MC
		printMcList(dev);
		printk("%s: mc_count %d\n", __FUNCTION__, dev->mc_count);
#endif
		recognise |= MAC_CONTROL_PM;	/* for the time being */
	}
#if DEBUG_MC
	printk("%s: setting %08x to %08x\n", __FUNCTION__, (int) sp->eth_regs,
		   recognise);
#endif

	sp->eth_regs->mac_control = recognise;
}

static void rx_tasklet_cleanup(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;

	/* 
	 * Tasklet may be scheduled. Need to get it removed from the list
	 * since we're about to free the struct.
	 */

	sp->unloading = 1;
	tasklet_enable(&sp->rx_tasklet);
	tasklet_kill(&sp->rx_tasklet);
}

static int __exit ar2313_remove(struct platform_device *pdev)
{
	struct net_device *dev = platform_get_drvdata(pdev);
	rx_tasklet_cleanup(dev);
	ar2313_init_cleanup(dev);
	unregister_netdev(dev);
	kfree(dev);
	return 0;
}


/*
 * Restart the AR2313 ethernet controller. 
 */
static int ar2313_restart(struct net_device *dev)
{
	/* disable interrupts */
	disable_irq(dev->irq);

	/* stop mac */
	ar2313_halt(dev);

	/* initialize */
	ar2313_init(dev);

	/* enable interrupts */
	enable_irq(dev->irq);

	return 0;
}

static struct platform_driver ar2313_driver = {
	.driver.name = "ar531x-eth",
	.probe = ar2313_probe,
	.remove = ar2313_remove,
};

int __init ar2313_module_init(void)
{
	return platform_driver_register(&ar2313_driver);
}

void __exit ar2313_module_cleanup(void)
{
	platform_driver_unregister(&ar2313_driver);
}

module_init(ar2313_module_init);
module_exit(ar2313_module_cleanup);


static void ar2313_free_descriptors(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	if (sp->rx_ring != NULL) {
		kfree((void *) KSEG0ADDR(sp->rx_ring));
		sp->rx_ring = NULL;
		sp->tx_ring = NULL;
	}
}


static int ar2313_allocate_descriptors(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	int size;
	int j;
	ar2313_descr_t *space;

	if (sp->rx_ring != NULL) {
		printk("%s: already done.\n", __FUNCTION__);
		return 0;
	}

	size =
		(sizeof(ar2313_descr_t) * (AR2313_DESCR_ENTRIES * AR2313_QUEUES));
	space = kmalloc(size, GFP_KERNEL);
	if (space == NULL)
		return 1;

	/* invalidate caches */
	dma_cache_inv((unsigned int) space, size);

	/* now convert pointer to KSEG1 */
	space = (ar2313_descr_t *) KSEG1ADDR(space);

	memset((void *) space, 0, size);

	sp->rx_ring = space;
	space += AR2313_DESCR_ENTRIES;

	sp->tx_ring = space;
	space += AR2313_DESCR_ENTRIES;

	/* Initialize the transmit Descriptors */
	for (j = 0; j < AR2313_DESCR_ENTRIES; j++) {
		ar2313_descr_t *td = &sp->tx_ring[j];
		td->status = 0;
		td->devcs = DMA_TX1_CHAINED;
		td->addr = 0;
		td->descr =
			virt_to_phys(&sp->
						 tx_ring[(j + 1) & (AR2313_DESCR_ENTRIES - 1)]);
	}

	return 0;
}


/*
 * Generic cleanup handling data allocated during init. Used when the
 * module is unloaded or if an error occurs during initialization
 */
static void ar2313_init_cleanup(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	struct sk_buff *skb;
	int j;

	ar2313_free_descriptors(dev);

	if (sp->eth_regs)
		iounmap((void *) sp->eth_regs);
	if (sp->dma_regs)
		iounmap((void *) sp->dma_regs);

	if (sp->rx_skb) {
		for (j = 0; j < AR2313_DESCR_ENTRIES; j++) {
			skb = sp->rx_skb[j];
			if (skb) {
				sp->rx_skb[j] = NULL;
				dev_kfree_skb(skb);
			}
		}
		kfree(sp->rx_skb);
		sp->rx_skb = NULL;
	}

	if (sp->tx_skb) {
		for (j = 0; j < AR2313_DESCR_ENTRIES; j++) {
			skb = sp->tx_skb[j];
			if (skb) {
				sp->tx_skb[j] = NULL;
				dev_kfree_skb(skb);
			}
		}
		kfree(sp->tx_skb);
		sp->tx_skb = NULL;
	}
}

static int ar2313_setup_timer(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;

	init_timer(&sp->link_timer);

	sp->link_timer.function = ar2313_link_timer_fn;
	sp->link_timer.data = (int) dev;
	sp->link_timer.expires = jiffies + HZ;

	add_timer(&sp->link_timer);
	return 0;

}

static void ar2313_link_timer_fn(unsigned long data)
{
	struct net_device *dev = (struct net_device *) data;
	struct ar2313_private *sp = dev->priv;

	// see if the link status changed
	// This was needed to make sure we set the PHY to the
	// autonegotiated value of half or full duplex.
	ar2313_check_link(dev);

	// Loop faster when we don't have link. 
	// This was needed to speed up the AP bootstrap time.
	if (sp->link == 0) {
		mod_timer(&sp->link_timer, jiffies + HZ / 2);
	} else {
		mod_timer(&sp->link_timer, jiffies + LINK_TIMER);
	}
}

static void ar2313_check_link(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	u16 phyData;

	phyData = armiiread(dev, sp->phy, MII_BMSR);
	if (sp->phyData != phyData) {
		if (phyData & BMSR_LSTATUS) {
			/* link is present, ready link partner ability to deterine
			   duplexity */
			int duplex = 0;
			u16 reg;

			sp->link = 1;
			reg = armiiread(dev, sp->phy, MII_BMCR);
			if (reg & BMCR_ANENABLE) {
				/* auto neg enabled */
				reg = armiiread(dev, sp->phy, MII_LPA);
				duplex = (reg & (LPA_100FULL | LPA_10FULL)) ? 1 : 0;
			} else {
				/* no auto neg, just read duplex config */
				duplex = (reg & BMCR_FULLDPLX) ? 1 : 0;
			}

			printk(KERN_INFO "%s: Configuring MAC for %s duplex\n",
				   dev->name, (duplex) ? "full" : "half");

			if (duplex) {
				/* full duplex */
				sp->eth_regs->mac_control =
					((sp->eth_regs->
					  mac_control | MAC_CONTROL_F) & ~MAC_CONTROL_DRO);
			} else {
				/* half duplex */
				sp->eth_regs->mac_control =
					((sp->eth_regs->
					  mac_control | MAC_CONTROL_DRO) & ~MAC_CONTROL_F);
			}
		} else {
			/* no link */
			sp->link = 0;
		}
		sp->phyData = phyData;
	}
}

static int ar2313_reset_reg(struct net_device *dev)
{
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;
	unsigned int ethsal, ethsah;
	unsigned int flags;

	*sp->int_regs |= sp->cfg->reset_mac;
	mdelay(10);
	*sp->int_regs &= ~sp->cfg->reset_mac;
	mdelay(10);
	*sp->int_regs |= sp->cfg->reset_phy;
	mdelay(10);
	*sp->int_regs &= ~sp->cfg->reset_phy;
	mdelay(10);

	sp->dma_regs->bus_mode = (DMA_BUS_MODE_SWR);
	mdelay(10);
	sp->dma_regs->bus_mode =
		((32 << DMA_BUS_MODE_PBL_SHIFT) | DMA_BUS_MODE_BLE);

	/* enable interrupts */
	sp->dma_regs->intr_ena = (DMA_STATUS_AIS |
							  DMA_STATUS_NIS |
							  DMA_STATUS_RI |
							  DMA_STATUS_TI | DMA_STATUS_FBE);
	sp->dma_regs->xmt_base = virt_to_phys(sp->tx_ring);
	sp->dma_regs->rcv_base = virt_to_phys(sp->rx_ring);
	sp->dma_regs->control =
		(DMA_CONTROL_SR | DMA_CONTROL_ST | DMA_CONTROL_SF);

	sp->eth_regs->flow_control = (FLOW_CONTROL_FCE);
	sp->eth_regs->vlan_tag = (0x8100);

	/* Enable Ethernet Interface */
	flags = (MAC_CONTROL_TE |	/* transmit enable */
			 MAC_CONTROL_PM |	/* pass mcast */
			 MAC_CONTROL_F |	/* full duplex */
			 MAC_CONTROL_HBD);	/* heart beat disabled */

	if (dev->flags & IFF_PROMISC) {	/* set promiscuous mode */
		flags |= MAC_CONTROL_PR;
	}
	sp->eth_regs->mac_control = flags;

	/* Set all Ethernet station address registers to their initial values */
	ethsah = ((((u_int) (dev->dev_addr[5]) << 8) & (u_int) 0x0000FF00) |
			  (((u_int) (dev->dev_addr[4]) << 0) & (u_int) 0x000000FF));

	ethsal = ((((u_int) (dev->dev_addr[3]) << 24) & (u_int) 0xFF000000) |
			  (((u_int) (dev->dev_addr[2]) << 16) & (u_int) 0x00FF0000) |
			  (((u_int) (dev->dev_addr[1]) << 8) & (u_int) 0x0000FF00) |
			  (((u_int) (dev->dev_addr[0]) << 0) & (u_int) 0x000000FF));

	sp->eth_regs->mac_addr[0] = ethsah;
	sp->eth_regs->mac_addr[1] = ethsal;

	mdelay(10);

	return (0);
}


static int ar2313_init(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	int ecode = 0;

	/* 
	 * Allocate descriptors
	 */
	if (ar2313_allocate_descriptors(dev)) {
		printk("%s: %s: ar2313_allocate_descriptors failed\n",
			   dev->name, __FUNCTION__);
		ecode = -EAGAIN;
		goto init_error;
	}

	/* 
	 * Get the memory for the skb rings.
	 */
	if (sp->rx_skb == NULL) {
		sp->rx_skb =
			kmalloc(sizeof(struct sk_buff *) * AR2313_DESCR_ENTRIES,
					GFP_KERNEL);
		if (!(sp->rx_skb)) {
			printk("%s: %s: rx_skb kmalloc failed\n",
				   dev->name, __FUNCTION__);
			ecode = -EAGAIN;
			goto init_error;
		}
	}
	memset(sp->rx_skb, 0, sizeof(struct sk_buff *) * AR2313_DESCR_ENTRIES);

	if (sp->tx_skb == NULL) {
		sp->tx_skb =
			kmalloc(sizeof(struct sk_buff *) * AR2313_DESCR_ENTRIES,
					GFP_KERNEL);
		if (!(sp->tx_skb)) {
			printk("%s: %s: tx_skb kmalloc failed\n",
				   dev->name, __FUNCTION__);
			ecode = -EAGAIN;
			goto init_error;
		}
	}
	memset(sp->tx_skb, 0, sizeof(struct sk_buff *) * AR2313_DESCR_ENTRIES);

	/* 
	 * Set tx_csm before we start receiving interrupts, otherwise
	 * the interrupt handler might think it is supposed to process
	 * tx ints before we are up and running, which may cause a null
	 * pointer access in the int handler.
	 */
	sp->rx_skbprd = 0;
	sp->cur_rx = 0;
	sp->tx_prd = 0;
	sp->tx_csm = 0;

	/* 
	 * Zero the stats before starting the interface
	 */
	memset(&sp->stats, 0, sizeof(sp->stats));

	/* 
	 * We load the ring here as there seem to be no way to tell the
	 * firmware to wipe the ring without re-initializing it.
	 */
	ar2313_load_rx_ring(dev, RX_RING_SIZE);

	/* 
	 * Init hardware
	 */
	ar2313_reset_reg(dev);

	/* 
	 * Get the IRQ
	 */
	ecode =
		request_irq(dev->irq, &ar2313_interrupt,
					IRQF_SHARED | IRQF_DISABLED | IRQF_SAMPLE_RANDOM,
					dev->name, dev);
	if (ecode) {
		printk(KERN_WARNING "%s: %s: Requested IRQ %d is busy\n",
			   dev->name, __FUNCTION__, dev->irq);
		goto init_error;
	}


	tasklet_enable(&sp->rx_tasklet);

	return 0;

  init_error:
	ar2313_init_cleanup(dev);
	return ecode;
}

/*
 * Load the rx ring.
 *
 * Loading rings is safe without holding the spin lock since this is
 * done only before the device is enabled, thus no interrupts are
 * generated and by the interrupt handler/tasklet handler.
 */
static void ar2313_load_rx_ring(struct net_device *dev, int nr_bufs)
{

	struct ar2313_private *sp = ((struct net_device *) dev)->priv;
	short i, idx;

	idx = sp->rx_skbprd;

	for (i = 0; i < nr_bufs; i++) {
		struct sk_buff *skb;
		ar2313_descr_t *rd;

		if (sp->rx_skb[idx]) {
#if DEBUG_RX
			printk(KERN_INFO "ar2313 rx refill full\n");
#endif							/* DEBUG */
			break;
		}
		// partha: create additional room for the second GRE fragment
		skb = alloc_skb(AR2313_BUFSIZE + 128, GFP_ATOMIC);
		if (!skb) {
			printk("\n\n\n\n %s: No memory in system\n\n\n\n",
				   __FUNCTION__);
			break;
		}
		// partha: create additional room in the front for tx pkt capture
		skb_reserve(skb, 32);

		/* 
		 * Make sure IP header starts on a fresh cache line.
		 */
		skb->dev = dev;
		skb_reserve(skb, RX_OFFSET);
		sp->rx_skb[idx] = skb;

		rd = (ar2313_descr_t *) & sp->rx_ring[idx];

		/* initialize dma descriptor */
		rd->devcs = ((AR2313_BUFSIZE << DMA_RX1_BSIZE_SHIFT) |
					 DMA_RX1_CHAINED);
		rd->addr = virt_to_phys(skb->data);
		rd->descr =
			virt_to_phys(&sp->
						 rx_ring[(idx + 1) & (AR2313_DESCR_ENTRIES - 1)]);
		rd->status = DMA_RX_OWN;

		idx = DSC_NEXT(idx);
	}

	if (!i) {
#if DEBUG_ERR
		printk(KERN_INFO
			   "Out of memory when allocating standard receive buffers\n");
#endif							/* DEBUG */
	} else {
		sp->rx_skbprd = idx;
	}

	return;
}

#define AR2313_MAX_PKTS_PER_CALL        64

static int ar2313_rx_int(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	struct sk_buff *skb, *skb_new;
	ar2313_descr_t *rxdesc;
	unsigned int status;
	u32 idx;
	int pkts = 0;
	int rval;

	idx = sp->cur_rx;

	/* process at most the entire ring and then wait for another interrupt 
	 */
	while (1) {

		rxdesc = &sp->rx_ring[idx];
		status = rxdesc->status;
		if (status & DMA_RX_OWN) {
			/* SiByte owns descriptor or descr not yet filled in */
			rval = 0;
			break;
		}

		if (++pkts > AR2313_MAX_PKTS_PER_CALL) {
			rval = 1;
			break;
		}
#if DEBUG_RX
		printk("index %d\n", idx);
		printk("RX status %08x\n", rxdesc->status);
		printk("RX devcs  %08x\n", rxdesc->devcs);
		printk("RX addr   %08x\n", rxdesc->addr);
		printk("RX descr  %08x\n", rxdesc->descr);
#endif

		if ((status & (DMA_RX_ERROR | DMA_RX_ERR_LENGTH)) &&
			(!(status & DMA_RX_LONG))) {
#if DEBUG_RX
			printk("%s: rx ERROR %08x\n", __FUNCTION__, status);
#endif
			sp->stats.rx_errors++;
			sp->stats.rx_dropped++;

			/* add statistics counters */
			if (status & DMA_RX_ERR_CRC)
				sp->stats.rx_crc_errors++;
			if (status & DMA_RX_ERR_COL)
				sp->stats.rx_over_errors++;
			if (status & DMA_RX_ERR_LENGTH)
				sp->stats.rx_length_errors++;
			if (status & DMA_RX_ERR_RUNT)
				sp->stats.rx_over_errors++;
			if (status & DMA_RX_ERR_DESC)
				sp->stats.rx_over_errors++;

		} else {
			/* alloc new buffer. */
			skb_new = dev_alloc_skb(AR2313_BUFSIZE + RX_OFFSET + 128);
			if (skb_new != NULL) {

				skb = sp->rx_skb[idx];
				/* set skb */
				skb_put(skb,
						((status >> DMA_RX_LEN_SHIFT) & 0x3fff) - CRC_LEN);

				sp->stats.rx_bytes += skb->len;
				skb->protocol = eth_type_trans(skb, dev);
				/* pass the packet to upper layers */
				netif_rx(skb);

				skb_new->dev = dev;
				/* 16 bit align */
				skb_reserve(skb_new, RX_OFFSET + 32);
				/* reset descriptor's curr_addr */
				rxdesc->addr = virt_to_phys(skb_new->data);

				sp->stats.rx_packets++;
				sp->rx_skb[idx] = skb_new;
			} else {
				sp->stats.rx_dropped++;
			}
		}

		rxdesc->devcs = ((AR2313_BUFSIZE << DMA_RX1_BSIZE_SHIFT) |
						 DMA_RX1_CHAINED);
		rxdesc->status = DMA_RX_OWN;

		idx = DSC_NEXT(idx);
	}

	sp->cur_rx = idx;

	return rval;
}


static void ar2313_tx_int(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	u32 idx;
	struct sk_buff *skb;
	ar2313_descr_t *txdesc;
	unsigned int status = 0;

	idx = sp->tx_csm;

	while (idx != sp->tx_prd) {

		txdesc = &sp->tx_ring[idx];

#if DEBUG_TX
		printk
			("%s: TXINT: csm=%d idx=%d prd=%d status=%x devcs=%x addr=%08x descr=%x\n",
			 dev->name, sp->tx_csm, idx, sp->tx_prd, txdesc->status,
			 txdesc->devcs, txdesc->addr, txdesc->descr);
#endif							/* DEBUG */

		if ((status = txdesc->status) & DMA_TX_OWN) {
			/* ar2313 dma still owns descr */
			break;
		}
		/* done with this descriptor */
		dma_unmap_single(NULL, txdesc->addr,
						 txdesc->devcs & DMA_TX1_BSIZE_MASK,
						 DMA_TO_DEVICE);
		txdesc->status = 0;

		if (status & DMA_TX_ERROR) {
			sp->stats.tx_errors++;
			sp->stats.tx_dropped++;
			if (status & DMA_TX_ERR_UNDER)
				sp->stats.tx_fifo_errors++;
			if (status & DMA_TX_ERR_HB)
				sp->stats.tx_heartbeat_errors++;
			if (status & (DMA_TX_ERR_LOSS | DMA_TX_ERR_LINK))
				sp->stats.tx_carrier_errors++;
			if (status & (DMA_TX_ERR_LATE |
						  DMA_TX_ERR_COL |
						  DMA_TX_ERR_JABBER | DMA_TX_ERR_DEFER))
				sp->stats.tx_aborted_errors++;
		} else {
			/* transmit OK */
			sp->stats.tx_packets++;
		}

		skb = sp->tx_skb[idx];
		sp->tx_skb[idx] = NULL;
		idx = DSC_NEXT(idx);
		sp->stats.tx_bytes += skb->len;
		dev_kfree_skb_irq(skb);
	}

	sp->tx_csm = idx;

	return;
}


static void rx_tasklet_func(unsigned long data)
{
	struct net_device *dev = (struct net_device *) data;
	struct ar2313_private *sp = dev->priv;

	if (sp->unloading) {
		return;
	}

	if (ar2313_rx_int(dev)) {
		tasklet_hi_schedule(&sp->rx_tasklet);
	} else {
		unsigned long flags;
		spin_lock_irqsave(&sp->lock, flags);
		sp->dma_regs->intr_ena |= DMA_STATUS_RI;
		spin_unlock_irqrestore(&sp->lock, flags);
	}
}

static void rx_schedule(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;

	sp->dma_regs->intr_ena &= ~DMA_STATUS_RI;

	tasklet_hi_schedule(&sp->rx_tasklet);
}

static irqreturn_t ar2313_interrupt(int irq, void *dev_id)
{
	struct net_device *dev = (struct net_device *) dev_id;
	struct ar2313_private *sp = dev->priv;
	unsigned int status, enabled;

	/* clear interrupt */
	/* 
	 * Don't clear RI bit if currently disabled.
	 */
	status = sp->dma_regs->status;
	enabled = sp->dma_regs->intr_ena;
	sp->dma_regs->status = status & enabled;

	if (status & DMA_STATUS_NIS) {
		/* normal status */
		/* 
		 * Don't schedule rx processing if interrupt
		 * is already disabled.
		 */
		if (status & enabled & DMA_STATUS_RI) {
			/* receive interrupt */
			rx_schedule(dev);
		}
		if (status & DMA_STATUS_TI) {
			/* transmit interrupt */
			ar2313_tx_int(dev);
		}
	}

	if (status & DMA_STATUS_AIS) {
#if DEBUG_INT
		printk("%s: AIS set %08x & %x\n", __FUNCTION__,
			   status, (DMA_STATUS_FBE | DMA_STATUS_TPS));
#endif
		/* abnormal status */
		if (status & (DMA_STATUS_FBE | DMA_STATUS_TPS)) {
			ar2313_restart(dev);
		}
	}
	return IRQ_HANDLED;
}


static int ar2313_open(struct net_device *dev)
{
	struct ar2313_private *sp;

	sp = dev->priv;

	dev->mtu = 1500;
	netif_start_queue(dev);

	sp->eth_regs->mac_control |= MAC_CONTROL_RE;

	return 0;
}

static void ar2313_halt(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	int j;

	tasklet_disable(&sp->rx_tasklet);

	/* kill the MAC */
	sp->eth_regs->mac_control &= ~(MAC_CONTROL_RE |	/* disable Receives */
								   MAC_CONTROL_TE);	/* disable Transmits */
	/* stop dma */
	sp->dma_regs->control = 0;
	sp->dma_regs->bus_mode = DMA_BUS_MODE_SWR;

	/* place phy and MAC in reset */
	*sp->int_regs |= (sp->cfg->reset_mac | sp->cfg->reset_phy);

	/* free buffers on tx ring */
	for (j = 0; j < AR2313_DESCR_ENTRIES; j++) {
		struct sk_buff *skb;
		ar2313_descr_t *txdesc;

		txdesc = &sp->tx_ring[j];
		txdesc->descr = 0;

		skb = sp->tx_skb[j];
		if (skb) {
			dev_kfree_skb(skb);
			sp->tx_skb[j] = NULL;
		}
	}
}

/*
 * close should do nothing. Here's why. It's called when
 * 'ifconfig bond0 down' is run. If it calls free_irq then
 * the irq is gone forever ! When bond0 is made 'up' again,
 * the ar2313_open () does not call request_irq (). Worse,
 * the call to ar2313_halt() generates a WDOG reset due to
 * the write to 'sp->int_regs' and the box reboots.
 * Commenting this out is good since it allows the
 * system to resume when bond0 is made up again.
 */
static int ar2313_close(struct net_device *dev)
{
#if 0
	/* 
	 * Disable interrupts
	 */
	disable_irq(dev->irq);

	/* 
	 * Without (or before) releasing irq and stopping hardware, this
	 * is an absolute non-sense, by the way. It will be reset instantly
	 * by the first irq.
	 */
	netif_stop_queue(dev);

	/* stop the MAC and DMA engines */
	ar2313_halt(dev);

	/* release the interrupt */
	free_irq(dev->irq, dev);

#endif
	return 0;
}

static int ar2313_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	ar2313_descr_t *td;
	u32 idx;

	idx = sp->tx_prd;
	td = &sp->tx_ring[idx];

	if (td->status & DMA_TX_OWN) {
#if DEBUG_TX
		printk("%s: No space left to Tx\n", __FUNCTION__);
#endif
		/* free skbuf and lie to the caller that we sent it out */
		sp->stats.tx_dropped++;
		dev_kfree_skb(skb);

		/* restart transmitter in case locked */
		sp->dma_regs->xmt_poll = 0;
		return 0;
	}

	/* Setup the transmit descriptor. */
	td->devcs = ((skb->len << DMA_TX1_BSIZE_SHIFT) |
				 (DMA_TX1_LS | DMA_TX1_IC | DMA_TX1_CHAINED));
	td->addr = dma_map_single(NULL, skb->data, skb->len, DMA_TO_DEVICE);
	td->status = DMA_TX_OWN;

	/* kick transmitter last */
	sp->dma_regs->xmt_poll = 0;

#if DEBUG_TX
	printk("index %d\n", idx);
	printk("TX status %08x\n", td->status);
	printk("TX devcs  %08x\n", td->devcs);
	printk("TX addr   %08x\n", td->addr);
	printk("TX descr  %08x\n", td->descr);
#endif

	sp->tx_skb[idx] = skb;
	idx = DSC_NEXT(idx);
	sp->tx_prd = idx;

	return 0;
}

static int netdev_get_ecmd(struct net_device *dev,
						   struct ethtool_cmd *ecmd)
{
	struct ar2313_private *np = dev->priv;
	u32 tmp;

	ecmd->supported =
		(SUPPORTED_10baseT_Half | SUPPORTED_10baseT_Full |
		 SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full |
		 SUPPORTED_Autoneg | SUPPORTED_TP | SUPPORTED_MII);

	ecmd->port = PORT_TP;
	/* only supports internal transceiver */
	ecmd->transceiver = XCVR_INTERNAL;
	/* not sure what this is for */
	ecmd->phy_address = 1;

	ecmd->advertising = ADVERTISED_MII;
	tmp = armiiread(dev, np->phy, MII_ADVERTISE);
	if (tmp & ADVERTISE_10HALF)
		ecmd->advertising |= ADVERTISED_10baseT_Half;
	if (tmp & ADVERTISE_10FULL)
		ecmd->advertising |= ADVERTISED_10baseT_Full;
	if (tmp & ADVERTISE_100HALF)
		ecmd->advertising |= ADVERTISED_100baseT_Half;
	if (tmp & ADVERTISE_100FULL)
		ecmd->advertising |= ADVERTISED_100baseT_Full;

	tmp = armiiread(dev, np->phy, MII_BMCR);
	if (tmp & BMCR_ANENABLE) {
		ecmd->advertising |= ADVERTISED_Autoneg;
		ecmd->autoneg = AUTONEG_ENABLE;
	} else {
		ecmd->autoneg = AUTONEG_DISABLE;
	}

	if (ecmd->autoneg == AUTONEG_ENABLE) {
		tmp = armiiread(dev, np->phy, MII_LPA);
		if (tmp & (LPA_100FULL | LPA_10FULL)) {
			ecmd->duplex = DUPLEX_FULL;
		} else {
			ecmd->duplex = DUPLEX_HALF;
		}
		if (tmp & (LPA_100FULL | LPA_100HALF)) {
			ecmd->speed = SPEED_100;
		} else {
			ecmd->speed = SPEED_10;
		}
	} else {
		if (tmp & BMCR_FULLDPLX) {
			ecmd->duplex = DUPLEX_FULL;
		} else {
			ecmd->duplex = DUPLEX_HALF;
		}
		if (tmp & BMCR_SPEED100) {
			ecmd->speed = SPEED_100;
		} else {
			ecmd->speed = SPEED_10;
		}
	}

	/* ignore maxtxpkt, maxrxpkt for now */

	return 0;
}

static int netdev_set_ecmd(struct net_device *dev,
						   struct ethtool_cmd *ecmd)
{
	struct ar2313_private *np = dev->priv;
	u32 tmp;

	if (ecmd->speed != SPEED_10 && ecmd->speed != SPEED_100)
		return -EINVAL;
	if (ecmd->duplex != DUPLEX_HALF && ecmd->duplex != DUPLEX_FULL)
		return -EINVAL;
	if (ecmd->port != PORT_TP)
		return -EINVAL;
	if (ecmd->transceiver != XCVR_INTERNAL)
		return -EINVAL;
	if (ecmd->autoneg != AUTONEG_DISABLE
		&& ecmd->autoneg != AUTONEG_ENABLE)
		return -EINVAL;
	/* ignore phy_address, maxtxpkt, maxrxpkt for now */

	/* WHEW! now lets bang some bits */

	tmp = armiiread(dev, np->phy, MII_BMCR);
	if (ecmd->autoneg == AUTONEG_ENABLE) {
		/* turn on autonegotiation */
		tmp |= BMCR_ANENABLE;
		printk("%s: Enabling auto-neg\n", dev->name);
	} else {
		/* turn off auto negotiation, set speed and duplexity */
		tmp &= ~(BMCR_ANENABLE | BMCR_SPEED100 | BMCR_FULLDPLX);
		if (ecmd->speed == SPEED_100)
			tmp |= BMCR_SPEED100;
		if (ecmd->duplex == DUPLEX_FULL)
			tmp |= BMCR_FULLDPLX;
		printk("%s: Hard coding %d/%s\n", dev->name,
			   (ecmd->speed == SPEED_100) ? 100 : 10,
			   (ecmd->duplex == DUPLEX_FULL) ? "full" : "half");
	}
	armiiwrite(dev, np->phy, MII_BMCR, tmp);
	np->phyData = 0;
	return 0;
}

static int netdev_ethtool_ioctl(struct net_device *dev, void *useraddr)
{
	struct ar2313_private *np = dev->priv;
	u32 cmd;

	if (get_user(cmd, (u32 *) useraddr))
		return -EFAULT;

	switch (cmd) {
		/* get settings */
	case ETHTOOL_GSET:{
			struct ethtool_cmd ecmd = { ETHTOOL_GSET };
			spin_lock_irq(&np->lock);
			netdev_get_ecmd(dev, &ecmd);
			spin_unlock_irq(&np->lock);
			if (copy_to_user(useraddr, &ecmd, sizeof(ecmd)))
				return -EFAULT;
			return 0;
		}
		/* set settings */
	case ETHTOOL_SSET:{
			struct ethtool_cmd ecmd;
			int r;
			if (copy_from_user(&ecmd, useraddr, sizeof(ecmd)))
				return -EFAULT;
			spin_lock_irq(&np->lock);
			r = netdev_set_ecmd(dev, &ecmd);
			spin_unlock_irq(&np->lock);
			return r;
		}
		/* restart autonegotiation */
	case ETHTOOL_NWAY_RST:{
			int tmp;
			int r = -EINVAL;
			/* if autoneg is off, it's an error */
			tmp = armiiread(dev, np->phy, MII_BMCR);
			if (tmp & BMCR_ANENABLE) {
				tmp |= (BMCR_ANRESTART);
				armiiwrite(dev, np->phy, MII_BMCR, tmp);
				r = 0;
			}
			return r;
		}
		/* get link status */
	case ETHTOOL_GLINK:{
			struct ethtool_value edata = { ETHTOOL_GLINK };
			edata.data =
				(armiiread(dev, np->phy, MII_BMSR) & BMSR_LSTATUS) ? 1 : 0;
			if (copy_to_user(useraddr, &edata, sizeof(edata)))
				return -EFAULT;
			return 0;
		}
	}

	return -EOPNOTSUPP;
}

static int ar2313_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
{
	struct mii_ioctl_data *data = (struct mii_ioctl_data *) &ifr->ifr_data;

	switch (cmd) {

	case SIOCETHTOOL:
		return netdev_ethtool_ioctl(dev, (void *) ifr->ifr_data);

	case SIOCGMIIPHY:			/* Get address of MII PHY in use. */
		data->phy_id = 1;
		/* Fall Through */

	case SIOCGMIIREG:			/* Read MII PHY register. */
		data->val_out = armiiread(dev, data->phy_id & 0x1f,
								  data->reg_num & 0x1f);
		return 0;
	case SIOCSMIIREG:			/* Write MII PHY register. */
		if (!capable(CAP_NET_ADMIN))
			return -EPERM;
		armiiwrite(dev, data->phy_id & 0x1f,
				   data->reg_num & 0x1f, data->val_in);
		return 0;

	case SIOCSIFHWADDR:
		if (copy_from_user
			(dev->dev_addr, ifr->ifr_data, sizeof(dev->dev_addr)))
			return -EFAULT;
		return 0;

	case SIOCGIFHWADDR:
		if (copy_to_user
			(ifr->ifr_data, dev->dev_addr, sizeof(dev->dev_addr)))
			return -EFAULT;
		return 0;

	default:
		break;
	}

	return -EOPNOTSUPP;
}

static struct net_device_stats *ar2313_get_stats(struct net_device *dev)
{
	struct ar2313_private *sp = dev->priv;
	return &sp->stats;
}


#define MII_ADDR(phy, reg) \
	((reg << MII_ADDR_REG_SHIFT) | (phy << MII_ADDR_PHY_SHIFT))

static short armiiread(struct net_device *dev, short phy, short reg)
{
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;
	volatile ETHERNET_STRUCT *ethernet = sp->phy_regs;

	ethernet->mii_addr = MII_ADDR(phy, reg);
	while (ethernet->mii_addr & MII_ADDR_BUSY);
	return (ethernet->mii_data >> MII_DATA_SHIFT);
}

static void
armiiwrite(struct net_device *dev, short phy, short reg, short data)
{
	struct ar2313_private *sp = (struct ar2313_private *) dev->priv;
	volatile ETHERNET_STRUCT *ethernet = sp->phy_regs;

	while (ethernet->mii_addr & MII_ADDR_BUSY);
	ethernet->mii_data = data << MII_DATA_SHIFT;
	ethernet->mii_addr = MII_ADDR(phy, reg) | MII_ADDR_WRITE;
}