Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / net / cris / eth_v10.c
diff --git a/kernel/drivers/net/cris/eth_v10.c b/kernel/drivers/net/cris/eth_v10.c
new file mode 100644 (file)
index 0000000..64c016a
--- /dev/null
@@ -0,0 +1,1746 @@
+/*
+ * e100net.c: A network driver for the ETRAX 100LX network controller.
+ *
+ * Copyright (c) 1998-2002 Axis Communications AB.
+ *
+ * The outline of this driver comes from skeleton.c.
+ *
+ */
+
+
+#include <linux/module.h>
+
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/types.h>
+#include <linux/fcntl.h>
+#include <linux/interrupt.h>
+#include <linux/ptrace.h>
+#include <linux/ioport.h>
+#include <linux/in.h>
+#include <linux/string.h>
+#include <linux/spinlock.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/bitops.h>
+
+#include <linux/if.h>
+#include <linux/mii.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/ethtool.h>
+
+#include <arch/svinto.h>/* DMA and register descriptions */
+#include <asm/io.h>         /* CRIS_LED_* I/O functions */
+#include <asm/irq.h>
+#include <asm/dma.h>
+#include <asm/ethernet.h>
+#include <asm/cache.h>
+#include <arch/io_interface_mux.h>
+
+//#define ETHDEBUG
+#define D(x)
+
+/*
+ * The name of the card. Is used for messages and in the requests for
+ * io regions, irqs and dma channels
+ */
+
+static const char* cardname = "ETRAX 100LX built-in ethernet controller";
+
+/* A default ethernet address. Highlevel SW will set the real one later */
+
+static struct sockaddr default_mac = {
+       0,
+       { 0x00, 0x40, 0x8C, 0xCD, 0x00, 0x00 }
+};
+
+/* Information that need to be kept for each board. */
+struct net_local {
+       struct mii_if_info mii_if;
+
+       /* Tx control lock.  This protects the transmit buffer ring
+        * state along with the "tx full" state of the driver.  This
+        * means all netif_queue flow control actions are protected
+        * by this lock as well.
+        */
+       spinlock_t lock;
+
+       spinlock_t led_lock; /* Protect LED state */
+       spinlock_t transceiver_lock; /* Protect transceiver state. */
+};
+
+typedef struct etrax_eth_descr
+{
+       etrax_dma_descr descr;
+       struct sk_buff* skb;
+} etrax_eth_descr;
+
+/* Some transceivers requires special handling */
+struct transceiver_ops
+{
+       unsigned int oui;
+       void (*check_speed)(struct net_device* dev);
+       void (*check_duplex)(struct net_device* dev);
+};
+
+/* Duplex settings */
+enum duplex
+{
+       half,
+       full,
+       autoneg
+};
+
+/* Dma descriptors etc. */
+
+#define MAX_MEDIA_DATA_SIZE 1522
+
+#define MIN_PACKET_LEN      46
+#define ETHER_HEAD_LEN      14
+
+/*
+** MDIO constants.
+*/
+#define MDIO_START                          0x1
+#define MDIO_READ                           0x2
+#define MDIO_WRITE                          0x1
+#define MDIO_PREAMBLE              0xfffffffful
+
+/* Broadcom specific */
+#define MDIO_AUX_CTRL_STATUS_REG           0x18
+#define MDIO_BC_FULL_DUPLEX_IND             0x1
+#define MDIO_BC_SPEED                       0x2
+
+/* TDK specific */
+#define MDIO_TDK_DIAGNOSTIC_REG              18
+#define MDIO_TDK_DIAGNOSTIC_RATE          0x400
+#define MDIO_TDK_DIAGNOSTIC_DPLX          0x800
+
+/*Intel LXT972A specific*/
+#define MDIO_INT_STATUS_REG_2                  0x0011
+#define MDIO_INT_FULL_DUPLEX_IND       (1 << 9)
+#define MDIO_INT_SPEED                (1 << 14)
+
+/* Network flash constants */
+#define NET_FLASH_TIME                  (HZ/50) /* 20 ms */
+#define NET_FLASH_PAUSE                (HZ/100) /* 10 ms */
+#define NET_LINK_UP_CHECK_INTERVAL       (2*HZ) /* 2 s   */
+#define NET_DUPLEX_CHECK_INTERVAL        (2*HZ) /* 2 s   */
+
+#define NO_NETWORK_ACTIVITY 0
+#define NETWORK_ACTIVITY    1
+
+#define NBR_OF_RX_DESC     32
+#define NBR_OF_TX_DESC     16
+
+/* Large packets are sent directly to upper layers while small packets are */
+/* copied (to reduce memory waste). The following constant decides the breakpoint */
+#define RX_COPYBREAK 256
+
+/* Due to a chip bug we need to flush the cache when descriptors are returned */
+/* to the DMA. To decrease performance impact we return descriptors in chunks. */
+/* The following constant determines the number of descriptors to return. */
+#define RX_QUEUE_THRESHOLD  NBR_OF_RX_DESC/2
+
+#define GET_BIT(bit,val)   (((val) >> (bit)) & 0x01)
+
+/* Define some macros to access ETRAX 100 registers */
+#define SETF(var, reg, field, val) var = (var & ~IO_MASK_(reg##_, field##_)) | \
+                                         IO_FIELD_(reg##_, field##_, val)
+#define SETS(var, reg, field, val) var = (var & ~IO_MASK_(reg##_, field##_)) | \
+                                         IO_STATE_(reg##_, field##_, _##val)
+
+static etrax_eth_descr *myNextRxDesc;  /* Points to the next descriptor to
+                                          to be processed */
+static etrax_eth_descr *myLastRxDesc;  /* The last processed descriptor */
+
+static etrax_eth_descr RxDescList[NBR_OF_RX_DESC] __attribute__ ((aligned(32)));
+
+static etrax_eth_descr* myFirstTxDesc; /* First packet not yet sent */
+static etrax_eth_descr* myLastTxDesc;  /* End of send queue */
+static etrax_eth_descr* myNextTxDesc;  /* Next descriptor to use */
+static etrax_eth_descr TxDescList[NBR_OF_TX_DESC] __attribute__ ((aligned(32)));
+
+static unsigned int network_rec_config_shadow = 0;
+
+static unsigned int network_tr_ctrl_shadow = 0;
+
+/* Network speed indication. */
+static DEFINE_TIMER(speed_timer, NULL, 0, 0);
+static DEFINE_TIMER(clear_led_timer, NULL, 0, 0);
+static int current_speed; /* Speed read from transceiver */
+static int current_speed_selection; /* Speed selected by user */
+static unsigned long led_next_time;
+static int led_active;
+static int rx_queue_len;
+
+/* Duplex */
+static DEFINE_TIMER(duplex_timer, NULL, 0, 0);
+static int full_duplex;
+static enum duplex current_duplex;
+
+/* Index to functions, as function prototypes. */
+
+static int etrax_ethernet_init(void);
+
+static int e100_open(struct net_device *dev);
+static int e100_set_mac_address(struct net_device *dev, void *addr);
+static int e100_send_packet(struct sk_buff *skb, struct net_device *dev);
+static irqreturn_t e100rxtx_interrupt(int irq, void *dev_id);
+static irqreturn_t e100nw_interrupt(int irq, void *dev_id);
+static void e100_rx(struct net_device *dev);
+static int e100_close(struct net_device *dev);
+static int e100_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
+static int e100_set_config(struct net_device* dev, struct ifmap* map);
+static void e100_tx_timeout(struct net_device *dev);
+static struct net_device_stats *e100_get_stats(struct net_device *dev);
+static void set_multicast_list(struct net_device *dev);
+static void e100_hardware_send_packet(struct net_local* np, char *buf, int length);
+static void update_rx_stats(struct net_device_stats *);
+static void update_tx_stats(struct net_device_stats *);
+static int e100_probe_transceiver(struct net_device* dev);
+
+static void e100_check_speed(unsigned long priv);
+static void e100_set_speed(struct net_device* dev, unsigned long speed);
+static void e100_check_duplex(unsigned long priv);
+static void e100_set_duplex(struct net_device* dev, enum duplex);
+static void e100_negotiate(struct net_device* dev);
+
+static int e100_get_mdio_reg(struct net_device *dev, int phy_id, int location);
+static void e100_set_mdio_reg(struct net_device *dev, int phy_id, int location, int value);
+
+static void e100_send_mdio_cmd(unsigned short cmd, int write_cmd);
+static void e100_send_mdio_bit(unsigned char bit);
+static unsigned char e100_receive_mdio_bit(void);
+static void e100_reset_transceiver(struct net_device* net);
+
+static void e100_clear_network_leds(unsigned long dummy);
+static void e100_set_network_leds(int active);
+
+static const struct ethtool_ops e100_ethtool_ops;
+#if defined(CONFIG_ETRAX_NO_PHY)
+static void dummy_check_speed(struct net_device* dev);
+static void dummy_check_duplex(struct net_device* dev);
+#else
+static void broadcom_check_speed(struct net_device* dev);
+static void broadcom_check_duplex(struct net_device* dev);
+static void tdk_check_speed(struct net_device* dev);
+static void tdk_check_duplex(struct net_device* dev);
+static void intel_check_speed(struct net_device* dev);
+static void intel_check_duplex(struct net_device* dev);
+static void generic_check_speed(struct net_device* dev);
+static void generic_check_duplex(struct net_device* dev);
+#endif
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void e100_netpoll(struct net_device* dev);
+#endif
+
+static int autoneg_normal = 1;
+
+struct transceiver_ops transceivers[] =
+{
+#if defined(CONFIG_ETRAX_NO_PHY)
+       {0x0000, dummy_check_speed, dummy_check_duplex}        /* Dummy */
+#else
+       {0x1018, broadcom_check_speed, broadcom_check_duplex},  /* Broadcom */
+       {0xC039, tdk_check_speed, tdk_check_duplex},            /* TDK 2120 */
+       {0x039C, tdk_check_speed, tdk_check_duplex},            /* TDK 2120C */
+        {0x04de, intel_check_speed, intel_check_duplex},       /* Intel LXT972A*/
+       {0x0000, generic_check_speed, generic_check_duplex}     /* Generic, must be last */
+#endif
+};
+
+struct transceiver_ops* transceiver = &transceivers[0];
+
+static const struct net_device_ops e100_netdev_ops = {
+       .ndo_open               = e100_open,
+       .ndo_stop               = e100_close,
+       .ndo_start_xmit         = e100_send_packet,
+       .ndo_tx_timeout         = e100_tx_timeout,
+       .ndo_get_stats          = e100_get_stats,
+       .ndo_set_rx_mode        = set_multicast_list,
+       .ndo_do_ioctl           = e100_ioctl,
+       .ndo_set_mac_address    = e100_set_mac_address,
+       .ndo_validate_addr      = eth_validate_addr,
+       .ndo_change_mtu         = eth_change_mtu,
+       .ndo_set_config         = e100_set_config,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+       .ndo_poll_controller    = e100_netpoll,
+#endif
+};
+
+#define tx_done(dev) (*R_DMA_CH0_CMD == 0)
+
+/*
+ * Check for a network adaptor of this type, and return '0' if one exists.
+ * If dev->base_addr == 0, probe all likely locations.
+ * If dev->base_addr == 1, always return failure.
+ * If dev->base_addr == 2, allocate space for the device and return success
+ * (detachable devices only).
+ */
+
+static int __init
+etrax_ethernet_init(void)
+{
+       struct net_device *dev;
+        struct net_local* np;
+       int i, err;
+
+       printk(KERN_INFO
+              "ETRAX 100LX 10/100MBit ethernet v2.0 (c) 1998-2007 Axis Communications AB\n");
+
+       if (cris_request_io_interface(if_eth, cardname)) {
+               printk(KERN_CRIT "etrax_ethernet_init failed to get IO interface\n");
+               return -EBUSY;
+       }
+
+       dev = alloc_etherdev(sizeof(struct net_local));
+       if (!dev)
+               return -ENOMEM;
+
+       np = netdev_priv(dev);
+
+       /* we do our own locking */
+       dev->features |= NETIF_F_LLTX;
+
+       dev->base_addr = (unsigned int)R_NETWORK_SA_0; /* just to have something to show */
+
+       /* now setup our etrax specific stuff */
+
+       dev->irq = NETWORK_DMA_RX_IRQ_NBR; /* we really use DMATX as well... */
+       dev->dma = NETWORK_RX_DMA_NBR;
+
+       /* fill in our handlers so the network layer can talk to us in the future */
+
+       dev->ethtool_ops        = &e100_ethtool_ops;
+       dev->netdev_ops         = &e100_netdev_ops;
+
+       spin_lock_init(&np->lock);
+       spin_lock_init(&np->led_lock);
+       spin_lock_init(&np->transceiver_lock);
+
+       /* Initialise the list of Etrax DMA-descriptors */
+
+       /* Initialise receive descriptors */
+
+       for (i = 0; i < NBR_OF_RX_DESC; i++) {
+               /* Allocate two extra cachelines to make sure that buffer used
+                * by DMA does not share cacheline with any other data (to
+                * avoid cache bug)
+                */
+               RxDescList[i].skb = dev_alloc_skb(MAX_MEDIA_DATA_SIZE + 2 * L1_CACHE_BYTES);
+               if (!RxDescList[i].skb)
+                       return -ENOMEM;
+               RxDescList[i].descr.ctrl   = 0;
+               RxDescList[i].descr.sw_len = MAX_MEDIA_DATA_SIZE;
+               RxDescList[i].descr.next   = virt_to_phys(&RxDescList[i + 1]);
+               RxDescList[i].descr.buf    = L1_CACHE_ALIGN(virt_to_phys(RxDescList[i].skb->data));
+               RxDescList[i].descr.status = 0;
+               RxDescList[i].descr.hw_len = 0;
+               prepare_rx_descriptor(&RxDescList[i].descr);
+       }
+
+       RxDescList[NBR_OF_RX_DESC - 1].descr.ctrl   = d_eol;
+       RxDescList[NBR_OF_RX_DESC - 1].descr.next   = virt_to_phys(&RxDescList[0]);
+       rx_queue_len = 0;
+
+       /* Initialize transmit descriptors */
+       for (i = 0; i < NBR_OF_TX_DESC; i++) {
+               TxDescList[i].descr.ctrl   = 0;
+               TxDescList[i].descr.sw_len = 0;
+               TxDescList[i].descr.next   = virt_to_phys(&TxDescList[i + 1].descr);
+               TxDescList[i].descr.buf    = 0;
+               TxDescList[i].descr.status = 0;
+               TxDescList[i].descr.hw_len = 0;
+               TxDescList[i].skb = 0;
+       }
+
+       TxDescList[NBR_OF_TX_DESC - 1].descr.ctrl   = d_eol;
+       TxDescList[NBR_OF_TX_DESC - 1].descr.next   = virt_to_phys(&TxDescList[0].descr);
+
+       /* Initialise initial pointers */
+
+       myNextRxDesc  = &RxDescList[0];
+       myLastRxDesc  = &RxDescList[NBR_OF_RX_DESC - 1];
+       myFirstTxDesc = &TxDescList[0];
+       myNextTxDesc  = &TxDescList[0];
+       myLastTxDesc  = &TxDescList[NBR_OF_TX_DESC - 1];
+
+       /* Register device */
+       err = register_netdev(dev);
+       if (err) {
+               free_netdev(dev);
+               return err;
+       }
+
+       /* set the default MAC address */
+
+       e100_set_mac_address(dev, &default_mac);
+
+       /* Initialize speed indicator stuff. */
+
+       current_speed = 10;
+       current_speed_selection = 0; /* Auto */
+       speed_timer.expires = jiffies + NET_LINK_UP_CHECK_INTERVAL;
+       speed_timer.data = (unsigned long)dev;
+       speed_timer.function = e100_check_speed;
+
+       clear_led_timer.function = e100_clear_network_leds;
+       clear_led_timer.data = (unsigned long)dev;
+
+       full_duplex = 0;
+       current_duplex = autoneg;
+       duplex_timer.expires = jiffies + NET_DUPLEX_CHECK_INTERVAL;
+        duplex_timer.data = (unsigned long)dev;
+       duplex_timer.function = e100_check_duplex;
+
+        /* Initialize mii interface */
+       np->mii_if.phy_id_mask = 0x1f;
+       np->mii_if.reg_num_mask = 0x1f;
+       np->mii_if.dev = dev;
+       np->mii_if.mdio_read = e100_get_mdio_reg;
+       np->mii_if.mdio_write = e100_set_mdio_reg;
+
+       /* Initialize group address registers to make sure that no */
+       /* unwanted addresses are matched */
+       *R_NETWORK_GA_0 = 0x00000000;
+       *R_NETWORK_GA_1 = 0x00000000;
+
+       /* Initialize next time the led can flash */
+       led_next_time = jiffies;
+       return 0;
+}
+
+/* set MAC address of the interface. called from the core after a
+ * SIOCSIFADDR ioctl, and from the bootup above.
+ */
+
+static int
+e100_set_mac_address(struct net_device *dev, void *p)
+{
+       struct net_local *np = netdev_priv(dev);
+       struct sockaddr *addr = p;
+
+       spin_lock(&np->lock); /* preemption protection */
+
+       /* remember it */
+
+       memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
+
+       /* Write it to the hardware.
+        * Note the way the address is wrapped:
+        * *R_NETWORK_SA_0 = a0_0 | (a0_1 << 8) | (a0_2 << 16) | (a0_3 << 24);
+        * *R_NETWORK_SA_1 = a0_4 | (a0_5 << 8);
+        */
+
+       *R_NETWORK_SA_0 = dev->dev_addr[0] | (dev->dev_addr[1] << 8) |
+               (dev->dev_addr[2] << 16) | (dev->dev_addr[3] << 24);
+       *R_NETWORK_SA_1 = dev->dev_addr[4] | (dev->dev_addr[5] << 8);
+       *R_NETWORK_SA_2 = 0;
+
+       /* show it in the log as well */
+
+       printk(KERN_INFO "%s: changed MAC to %pM\n", dev->name, dev->dev_addr);
+
+       spin_unlock(&np->lock);
+
+       return 0;
+}
+
+/*
+ * Open/initialize the board. This is called (in the current kernel)
+ * sometime after booting when the 'ifconfig' program is run.
+ *
+ * This routine should set everything up anew at each open, even
+ * registers that "should" only need to be set once at boot, so that
+ * there is non-reboot way to recover if something goes wrong.
+ */
+
+static int
+e100_open(struct net_device *dev)
+{
+       unsigned long flags;
+
+       /* enable the MDIO output pin */
+
+       *R_NETWORK_MGM_CTRL = IO_STATE(R_NETWORK_MGM_CTRL, mdoe, enable);
+
+       *R_IRQ_MASK0_CLR =
+               IO_STATE(R_IRQ_MASK0_CLR, overrun, clr) |
+               IO_STATE(R_IRQ_MASK0_CLR, underrun, clr) |
+               IO_STATE(R_IRQ_MASK0_CLR, excessive_col, clr);
+
+       /* clear dma0 and 1 eop and descr irq masks */
+       *R_IRQ_MASK2_CLR =
+               IO_STATE(R_IRQ_MASK2_CLR, dma0_descr, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma0_eop, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma1_descr, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma1_eop, clr);
+
+       /* Reset and wait for the DMA channels */
+
+       RESET_DMA(NETWORK_TX_DMA_NBR);
+       RESET_DMA(NETWORK_RX_DMA_NBR);
+       WAIT_DMA(NETWORK_TX_DMA_NBR);
+       WAIT_DMA(NETWORK_RX_DMA_NBR);
+
+       /* Initialise the etrax network controller */
+
+       /* allocate the irq corresponding to the receiving DMA */
+
+       if (request_irq(NETWORK_DMA_RX_IRQ_NBR, e100rxtx_interrupt, 0, cardname,
+                       (void *)dev)) {
+               goto grace_exit0;
+       }
+
+       /* allocate the irq corresponding to the transmitting DMA */
+
+       if (request_irq(NETWORK_DMA_TX_IRQ_NBR, e100rxtx_interrupt, 0,
+                       cardname, (void *)dev)) {
+               goto grace_exit1;
+       }
+
+       /* allocate the irq corresponding to the network errors etc */
+
+       if (request_irq(NETWORK_STATUS_IRQ_NBR, e100nw_interrupt, 0,
+                       cardname, (void *)dev)) {
+               goto grace_exit2;
+       }
+
+       /*
+        * Always allocate the DMA channels after the IRQ,
+        * and clean up on failure.
+        */
+
+       if (cris_request_dma(NETWORK_TX_DMA_NBR,
+                            cardname,
+                            DMA_VERBOSE_ON_ERROR,
+                            dma_eth)) {
+               goto grace_exit3;
+        }
+
+       if (cris_request_dma(NETWORK_RX_DMA_NBR,
+                            cardname,
+                            DMA_VERBOSE_ON_ERROR,
+                            dma_eth)) {
+               goto grace_exit4;
+        }
+
+       /* give the HW an idea of what MAC address we want */
+
+       *R_NETWORK_SA_0 = dev->dev_addr[0] | (dev->dev_addr[1] << 8) |
+               (dev->dev_addr[2] << 16) | (dev->dev_addr[3] << 24);
+       *R_NETWORK_SA_1 = dev->dev_addr[4] | (dev->dev_addr[5] << 8);
+       *R_NETWORK_SA_2 = 0;
+
+#if 0
+       /* use promiscuous mode for testing */
+       *R_NETWORK_GA_0 = 0xffffffff;
+       *R_NETWORK_GA_1 = 0xffffffff;
+
+       *R_NETWORK_REC_CONFIG = 0xd; /* broadcast rec, individ. rec, ma0 enabled */
+#else
+       SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, max_size, size1522);
+       SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, broadcast, receive);
+       SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, ma0, enable);
+       SETF(network_rec_config_shadow, R_NETWORK_REC_CONFIG, duplex, full_duplex);
+       *R_NETWORK_REC_CONFIG = network_rec_config_shadow;
+#endif
+
+       *R_NETWORK_GEN_CONFIG =
+               IO_STATE(R_NETWORK_GEN_CONFIG, phy,    mii_clk) |
+               IO_STATE(R_NETWORK_GEN_CONFIG, enable, on);
+
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, clr_error, clr);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, delay, none);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, cancel, dont);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, cd, enable);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, retry, enable);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, pad, enable);
+       SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, crc, enable);
+       *R_NETWORK_TR_CTRL = network_tr_ctrl_shadow;
+
+       local_irq_save(flags);
+
+       /* enable the irq's for ethernet DMA */
+
+       *R_IRQ_MASK2_SET =
+               IO_STATE(R_IRQ_MASK2_SET, dma0_eop, set) |
+               IO_STATE(R_IRQ_MASK2_SET, dma1_eop, set);
+
+       *R_IRQ_MASK0_SET =
+               IO_STATE(R_IRQ_MASK0_SET, overrun,       set) |
+               IO_STATE(R_IRQ_MASK0_SET, underrun,      set) |
+               IO_STATE(R_IRQ_MASK0_SET, excessive_col, set);
+
+       /* make sure the irqs are cleared */
+
+       *R_DMA_CH0_CLR_INTR = IO_STATE(R_DMA_CH0_CLR_INTR, clr_eop, do);
+       *R_DMA_CH1_CLR_INTR = IO_STATE(R_DMA_CH1_CLR_INTR, clr_eop, do);
+
+       /* make sure the rec and transmit error counters are cleared */
+
+       (void)*R_REC_COUNTERS;  /* dummy read */
+       (void)*R_TR_COUNTERS;   /* dummy read */
+
+       /* start the receiving DMA channel so we can receive packets from now on */
+
+       *R_DMA_CH1_FIRST = virt_to_phys(myNextRxDesc);
+       *R_DMA_CH1_CMD = IO_STATE(R_DMA_CH1_CMD, cmd, start);
+
+       /* Set up transmit DMA channel so it can be restarted later */
+
+       *R_DMA_CH0_FIRST = 0;
+       *R_DMA_CH0_DESCR = virt_to_phys(myLastTxDesc);
+       netif_start_queue(dev);
+
+       local_irq_restore(flags);
+
+       /* Probe for transceiver */
+       if (e100_probe_transceiver(dev))
+               goto grace_exit5;
+
+       /* Start duplex/speed timers */
+       add_timer(&speed_timer);
+       add_timer(&duplex_timer);
+
+       /* We are now ready to accept transmit requeusts from
+        * the queueing layer of the networking.
+        */
+       netif_carrier_on(dev);
+
+       return 0;
+
+grace_exit5:
+       cris_free_dma(NETWORK_RX_DMA_NBR, cardname);
+grace_exit4:
+       cris_free_dma(NETWORK_TX_DMA_NBR, cardname);
+grace_exit3:
+       free_irq(NETWORK_STATUS_IRQ_NBR, (void *)dev);
+grace_exit2:
+       free_irq(NETWORK_DMA_TX_IRQ_NBR, (void *)dev);
+grace_exit1:
+       free_irq(NETWORK_DMA_RX_IRQ_NBR, (void *)dev);
+grace_exit0:
+       return -EAGAIN;
+}
+
+#if defined(CONFIG_ETRAX_NO_PHY)
+static void
+dummy_check_speed(struct net_device* dev)
+{
+       current_speed = 100;
+}
+#else
+static void
+generic_check_speed(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_ADVERTISE);
+       if ((data & ADVERTISE_100FULL) ||
+           (data & ADVERTISE_100HALF))
+               current_speed = 100;
+       else
+               current_speed = 10;
+}
+
+static void
+tdk_check_speed(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_TDK_DIAGNOSTIC_REG);
+       current_speed = (data & MDIO_TDK_DIAGNOSTIC_RATE ? 100 : 10);
+}
+
+static void
+broadcom_check_speed(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_AUX_CTRL_STATUS_REG);
+       current_speed = (data & MDIO_BC_SPEED ? 100 : 10);
+}
+
+static void
+intel_check_speed(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_INT_STATUS_REG_2);
+       current_speed = (data & MDIO_INT_SPEED ? 100 : 10);
+}
+#endif
+static void
+e100_check_speed(unsigned long priv)
+{
+       struct net_device* dev = (struct net_device*)priv;
+       struct net_local *np = netdev_priv(dev);
+       static int led_initiated = 0;
+       unsigned long data;
+       int old_speed = current_speed;
+
+       spin_lock(&np->transceiver_lock);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_BMSR);
+       if (!(data & BMSR_LSTATUS)) {
+               current_speed = 0;
+       } else {
+               transceiver->check_speed(dev);
+       }
+
+       spin_lock(&np->led_lock);
+       if ((old_speed != current_speed) || !led_initiated) {
+               led_initiated = 1;
+               e100_set_network_leds(NO_NETWORK_ACTIVITY);
+               if (current_speed)
+                       netif_carrier_on(dev);
+               else
+                       netif_carrier_off(dev);
+       }
+       spin_unlock(&np->led_lock);
+
+       /* Reinitialize the timer. */
+       speed_timer.expires = jiffies + NET_LINK_UP_CHECK_INTERVAL;
+       add_timer(&speed_timer);
+
+       spin_unlock(&np->transceiver_lock);
+}
+
+static void
+e100_negotiate(struct net_device* dev)
+{
+       struct net_local *np = netdev_priv(dev);
+       unsigned short data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                               MII_ADVERTISE);
+
+       /* Discard old speed and duplex settings */
+       data &= ~(ADVERTISE_100HALF | ADVERTISE_100FULL |
+                 ADVERTISE_10HALF | ADVERTISE_10FULL);
+
+       switch (current_speed_selection) {
+               case 10:
+                       if (current_duplex == full)
+                               data |= ADVERTISE_10FULL;
+                       else if (current_duplex == half)
+                               data |= ADVERTISE_10HALF;
+                       else
+                               data |= ADVERTISE_10HALF | ADVERTISE_10FULL;
+                       break;
+
+               case 100:
+                        if (current_duplex == full)
+                               data |= ADVERTISE_100FULL;
+                       else if (current_duplex == half)
+                               data |= ADVERTISE_100HALF;
+                       else
+                               data |= ADVERTISE_100HALF | ADVERTISE_100FULL;
+                       break;
+
+               case 0: /* Auto */
+                        if (current_duplex == full)
+                               data |= ADVERTISE_100FULL | ADVERTISE_10FULL;
+                       else if (current_duplex == half)
+                               data |= ADVERTISE_100HALF | ADVERTISE_10HALF;
+                       else
+                               data |= ADVERTISE_10HALF | ADVERTISE_10FULL |
+                                 ADVERTISE_100HALF | ADVERTISE_100FULL;
+                       break;
+
+               default: /* assume autoneg speed and duplex */
+                       data |= ADVERTISE_10HALF | ADVERTISE_10FULL |
+                                 ADVERTISE_100HALF | ADVERTISE_100FULL;
+                       break;
+       }
+
+       e100_set_mdio_reg(dev, np->mii_if.phy_id, MII_ADVERTISE, data);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_BMCR);
+       if (autoneg_normal) {
+               /* Renegotiate with link partner */
+               data |= BMCR_ANENABLE | BMCR_ANRESTART;
+       } else {
+               /* Don't negotiate speed or duplex */
+               data &= ~(BMCR_ANENABLE | BMCR_ANRESTART);
+
+               /* Set speed and duplex static */
+               if (current_speed_selection == 10)
+                       data &= ~BMCR_SPEED100;
+               else
+                       data |= BMCR_SPEED100;
+
+               if (current_duplex != full)
+                       data &= ~BMCR_FULLDPLX;
+               else
+                       data |= BMCR_FULLDPLX;
+       }
+       e100_set_mdio_reg(dev, np->mii_if.phy_id, MII_BMCR, data);
+}
+
+static void
+e100_set_speed(struct net_device* dev, unsigned long speed)
+{
+       struct net_local *np = netdev_priv(dev);
+
+       spin_lock(&np->transceiver_lock);
+       if (speed != current_speed_selection) {
+               current_speed_selection = speed;
+               e100_negotiate(dev);
+       }
+       spin_unlock(&np->transceiver_lock);
+}
+
+static void
+e100_check_duplex(unsigned long priv)
+{
+       struct net_device *dev = (struct net_device *)priv;
+       struct net_local *np = netdev_priv(dev);
+       int old_duplex;
+
+       spin_lock(&np->transceiver_lock);
+       old_duplex = full_duplex;
+       transceiver->check_duplex(dev);
+       if (old_duplex != full_duplex) {
+               /* Duplex changed */
+               SETF(network_rec_config_shadow, R_NETWORK_REC_CONFIG, duplex, full_duplex);
+               *R_NETWORK_REC_CONFIG = network_rec_config_shadow;
+       }
+
+       /* Reinitialize the timer. */
+       duplex_timer.expires = jiffies + NET_DUPLEX_CHECK_INTERVAL;
+       add_timer(&duplex_timer);
+       np->mii_if.full_duplex = full_duplex;
+       spin_unlock(&np->transceiver_lock);
+}
+#if defined(CONFIG_ETRAX_NO_PHY)
+static void
+dummy_check_duplex(struct net_device* dev)
+{
+       full_duplex = 1;
+}
+#else
+static void
+generic_check_duplex(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_ADVERTISE);
+       if ((data & ADVERTISE_10FULL) ||
+           (data & ADVERTISE_100FULL))
+               full_duplex = 1;
+       else
+               full_duplex = 0;
+}
+
+static void
+tdk_check_duplex(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_TDK_DIAGNOSTIC_REG);
+       full_duplex = (data & MDIO_TDK_DIAGNOSTIC_DPLX) ? 1 : 0;
+}
+
+static void
+broadcom_check_duplex(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_AUX_CTRL_STATUS_REG);
+       full_duplex = (data & MDIO_BC_FULL_DUPLEX_IND) ? 1 : 0;
+}
+
+static void
+intel_check_duplex(struct net_device* dev)
+{
+       unsigned long data;
+       struct net_local *np = netdev_priv(dev);
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id,
+                                MDIO_INT_STATUS_REG_2);
+       full_duplex = (data & MDIO_INT_FULL_DUPLEX_IND) ? 1 : 0;
+}
+#endif
+static void
+e100_set_duplex(struct net_device* dev, enum duplex new_duplex)
+{
+       struct net_local *np = netdev_priv(dev);
+
+       spin_lock(&np->transceiver_lock);
+       if (new_duplex != current_duplex) {
+               current_duplex = new_duplex;
+               e100_negotiate(dev);
+       }
+       spin_unlock(&np->transceiver_lock);
+}
+
+static int
+e100_probe_transceiver(struct net_device* dev)
+{
+       int ret = 0;
+
+#if !defined(CONFIG_ETRAX_NO_PHY)
+       unsigned int phyid_high;
+       unsigned int phyid_low;
+       unsigned int oui;
+       struct transceiver_ops* ops = NULL;
+       struct net_local *np = netdev_priv(dev);
+
+       spin_lock(&np->transceiver_lock);
+
+       /* Probe MDIO physical address */
+       for (np->mii_if.phy_id = 0; np->mii_if.phy_id <= 31;
+            np->mii_if.phy_id++) {
+               if (e100_get_mdio_reg(dev,
+                                     np->mii_if.phy_id, MII_BMSR) != 0xffff)
+                       break;
+       }
+       if (np->mii_if.phy_id == 32) {
+               ret = -ENODEV;
+               goto out;
+       }
+
+       /* Get manufacturer */
+       phyid_high = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_PHYSID1);
+       phyid_low = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_PHYSID2);
+       oui = (phyid_high << 6) | (phyid_low >> 10);
+
+       for (ops = &transceivers[0]; ops->oui; ops++) {
+               if (ops->oui == oui)
+                       break;
+       }
+       transceiver = ops;
+out:
+       spin_unlock(&np->transceiver_lock);
+#endif
+       return ret;
+}
+
+static int
+e100_get_mdio_reg(struct net_device *dev, int phy_id, int location)
+{
+       unsigned short cmd;    /* Data to be sent on MDIO port */
+       int data;   /* Data read from MDIO */
+       int bitCounter;
+
+       /* Start of frame, OP Code, Physical Address, Register Address */
+       cmd = (MDIO_START << 14) | (MDIO_READ << 12) | (phy_id << 7) |
+               (location << 2);
+
+       e100_send_mdio_cmd(cmd, 0);
+
+       data = 0;
+
+       /* Data... */
+       for (bitCounter=15; bitCounter>=0 ; bitCounter--) {
+               data |= (e100_receive_mdio_bit() << bitCounter);
+       }
+
+       return data;
+}
+
+static void
+e100_set_mdio_reg(struct net_device *dev, int phy_id, int location, int value)
+{
+       int bitCounter;
+       unsigned short cmd;
+
+       cmd = (MDIO_START << 14) | (MDIO_WRITE << 12) | (phy_id << 7) |
+             (location << 2);
+
+       e100_send_mdio_cmd(cmd, 1);
+
+       /* Data... */
+       for (bitCounter=15; bitCounter>=0 ; bitCounter--) {
+               e100_send_mdio_bit(GET_BIT(bitCounter, value));
+       }
+
+}
+
+static void
+e100_send_mdio_cmd(unsigned short cmd, int write_cmd)
+{
+       int bitCounter;
+       unsigned char data = 0x2;
+
+       /* Preamble */
+       for (bitCounter = 31; bitCounter>= 0; bitCounter--)
+               e100_send_mdio_bit(GET_BIT(bitCounter, MDIO_PREAMBLE));
+
+       for (bitCounter = 15; bitCounter >= 2; bitCounter--)
+               e100_send_mdio_bit(GET_BIT(bitCounter, cmd));
+
+       /* Turnaround */
+       for (bitCounter = 1; bitCounter >= 0 ; bitCounter--)
+               if (write_cmd)
+                       e100_send_mdio_bit(GET_BIT(bitCounter, data));
+               else
+                       e100_receive_mdio_bit();
+}
+
+static void
+e100_send_mdio_bit(unsigned char bit)
+{
+       *R_NETWORK_MGM_CTRL =
+               IO_STATE(R_NETWORK_MGM_CTRL, mdoe, enable) |
+               IO_FIELD(R_NETWORK_MGM_CTRL, mdio, bit);
+       udelay(1);
+       *R_NETWORK_MGM_CTRL =
+               IO_STATE(R_NETWORK_MGM_CTRL, mdoe, enable) |
+               IO_MASK(R_NETWORK_MGM_CTRL, mdck) |
+               IO_FIELD(R_NETWORK_MGM_CTRL, mdio, bit);
+       udelay(1);
+}
+
+static unsigned char
+e100_receive_mdio_bit(void)
+{
+       unsigned char bit;
+       *R_NETWORK_MGM_CTRL = 0;
+       bit = IO_EXTRACT(R_NETWORK_STAT, mdio, *R_NETWORK_STAT);
+       udelay(1);
+       *R_NETWORK_MGM_CTRL = IO_MASK(R_NETWORK_MGM_CTRL, mdck);
+       udelay(1);
+       return bit;
+}
+
+static void
+e100_reset_transceiver(struct net_device* dev)
+{
+       struct net_local *np = netdev_priv(dev);
+       unsigned short cmd;
+       unsigned short data;
+       int bitCounter;
+
+       data = e100_get_mdio_reg(dev, np->mii_if.phy_id, MII_BMCR);
+
+       cmd = (MDIO_START << 14) | (MDIO_WRITE << 12) | (np->mii_if.phy_id << 7) | (MII_BMCR << 2);
+
+       e100_send_mdio_cmd(cmd, 1);
+
+       data |= 0x8000;
+
+       for (bitCounter = 15; bitCounter >= 0 ; bitCounter--) {
+               e100_send_mdio_bit(GET_BIT(bitCounter, data));
+       }
+}
+
+/* Called by upper layers if they decide it took too long to complete
+ * sending a packet - we need to reset and stuff.
+ */
+
+static void
+e100_tx_timeout(struct net_device *dev)
+{
+       struct net_local *np = netdev_priv(dev);
+       unsigned long flags;
+
+       spin_lock_irqsave(&np->lock, flags);
+
+       printk(KERN_WARNING "%s: transmit timed out, %s?\n", dev->name,
+              tx_done(dev) ? "IRQ problem" : "network cable problem");
+
+       /* remember we got an error */
+
+       dev->stats.tx_errors++;
+
+       /* reset the TX DMA in case it has hung on something */
+
+       RESET_DMA(NETWORK_TX_DMA_NBR);
+       WAIT_DMA(NETWORK_TX_DMA_NBR);
+
+       /* Reset the transceiver. */
+
+       e100_reset_transceiver(dev);
+
+       /* and get rid of the packets that never got an interrupt */
+       while (myFirstTxDesc != myNextTxDesc) {
+               dev_kfree_skb(myFirstTxDesc->skb);
+               myFirstTxDesc->skb = 0;
+               myFirstTxDesc = phys_to_virt(myFirstTxDesc->descr.next);
+       }
+
+       /* Set up transmit DMA channel so it can be restarted later */
+       *R_DMA_CH0_FIRST = 0;
+       *R_DMA_CH0_DESCR = virt_to_phys(myLastTxDesc);
+
+       /* tell the upper layers we're ok again */
+
+       netif_wake_queue(dev);
+       spin_unlock_irqrestore(&np->lock, flags);
+}
+
+
+/* This will only be invoked if the driver is _not_ in XOFF state.
+ * What this means is that we need not check it, and that this
+ * invariant will hold if we make sure that the netif_*_queue()
+ * calls are done at the proper times.
+ */
+
+static int
+e100_send_packet(struct sk_buff *skb, struct net_device *dev)
+{
+       struct net_local *np = netdev_priv(dev);
+       unsigned char *buf = skb->data;
+       unsigned long flags;
+
+#ifdef ETHDEBUG
+       printk("send packet len %d\n", length);
+#endif
+       spin_lock_irqsave(&np->lock, flags);  /* protect from tx_interrupt and ourself */
+
+       myNextTxDesc->skb = skb;
+
+       dev->trans_start = jiffies; /* NETIF_F_LLTX driver :( */
+
+       e100_hardware_send_packet(np, buf, skb->len);
+
+       myNextTxDesc = phys_to_virt(myNextTxDesc->descr.next);
+
+       /* Stop queue if full */
+       if (myNextTxDesc == myFirstTxDesc) {
+               netif_stop_queue(dev);
+       }
+
+       spin_unlock_irqrestore(&np->lock, flags);
+
+       return NETDEV_TX_OK;
+}
+
+/*
+ * The typical workload of the driver:
+ *   Handle the network interface interrupts.
+ */
+
+static irqreturn_t
+e100rxtx_interrupt(int irq, void *dev_id)
+{
+       struct net_device *dev = (struct net_device *)dev_id;
+       unsigned long irqbits;
+
+       /*
+        * Note that both rx and tx interrupts are blocked at this point,
+        * regardless of which got us here.
+        */
+
+       irqbits = *R_IRQ_MASK2_RD;
+
+       /* Handle received packets */
+       if (irqbits & IO_STATE(R_IRQ_MASK2_RD, dma1_eop, active)) {
+               /* acknowledge the eop interrupt */
+
+               *R_DMA_CH1_CLR_INTR = IO_STATE(R_DMA_CH1_CLR_INTR, clr_eop, do);
+
+               /* check if one or more complete packets were indeed received */
+
+               while ((*R_DMA_CH1_FIRST != virt_to_phys(myNextRxDesc)) &&
+                      (myNextRxDesc != myLastRxDesc)) {
+                       /* Take out the buffer and give it to the OS, then
+                        * allocate a new buffer to put a packet in.
+                        */
+                       e100_rx(dev);
+                       dev->stats.rx_packets++;
+                       /* restart/continue on the channel, for safety */
+                       *R_DMA_CH1_CMD = IO_STATE(R_DMA_CH1_CMD, cmd, restart);
+                       /* clear dma channel 1 eop/descr irq bits */
+                       *R_DMA_CH1_CLR_INTR =
+                               IO_STATE(R_DMA_CH1_CLR_INTR, clr_eop, do) |
+                               IO_STATE(R_DMA_CH1_CLR_INTR, clr_descr, do);
+
+                       /* now, we might have gotten another packet
+                          so we have to loop back and check if so */
+               }
+       }
+
+       /* Report any packets that have been sent */
+       while (virt_to_phys(myFirstTxDesc) != *R_DMA_CH0_FIRST &&
+              (netif_queue_stopped(dev) || myFirstTxDesc != myNextTxDesc)) {
+               dev->stats.tx_bytes += myFirstTxDesc->skb->len;
+               dev->stats.tx_packets++;
+
+               /* dma is ready with the transmission of the data in tx_skb, so now
+                  we can release the skb memory */
+               dev_kfree_skb_irq(myFirstTxDesc->skb);
+               myFirstTxDesc->skb = 0;
+               myFirstTxDesc = phys_to_virt(myFirstTxDesc->descr.next);
+                /* Wake up queue. */
+               netif_wake_queue(dev);
+       }
+
+       if (irqbits & IO_STATE(R_IRQ_MASK2_RD, dma0_eop, active)) {
+               /* acknowledge the eop interrupt. */
+               *R_DMA_CH0_CLR_INTR = IO_STATE(R_DMA_CH0_CLR_INTR, clr_eop, do);
+       }
+
+       return IRQ_HANDLED;
+}
+
+static irqreturn_t
+e100nw_interrupt(int irq, void *dev_id)
+{
+       struct net_device *dev = (struct net_device *)dev_id;
+       unsigned long irqbits = *R_IRQ_MASK0_RD;
+
+       /* check for underrun irq */
+       if (irqbits & IO_STATE(R_IRQ_MASK0_RD, underrun, active)) {
+               SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, clr_error, clr);
+               *R_NETWORK_TR_CTRL = network_tr_ctrl_shadow;
+               SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, clr_error, nop);
+               dev->stats.tx_errors++;
+               D(printk("ethernet receiver underrun!\n"));
+       }
+
+       /* check for overrun irq */
+       if (irqbits & IO_STATE(R_IRQ_MASK0_RD, overrun, active)) {
+               update_rx_stats(&dev->stats); /* this will ack the irq */
+               D(printk("ethernet receiver overrun!\n"));
+       }
+       /* check for excessive collision irq */
+       if (irqbits & IO_STATE(R_IRQ_MASK0_RD, excessive_col, active)) {
+               SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, clr_error, clr);
+               *R_NETWORK_TR_CTRL = network_tr_ctrl_shadow;
+               SETS(network_tr_ctrl_shadow, R_NETWORK_TR_CTRL, clr_error, nop);
+               dev->stats.tx_errors++;
+               D(printk("ethernet excessive collisions!\n"));
+       }
+       return IRQ_HANDLED;
+}
+
+/* We have a good packet(s), get it/them out of the buffers. */
+static void
+e100_rx(struct net_device *dev)
+{
+       struct sk_buff *skb;
+       int length = 0;
+       struct net_local *np = netdev_priv(dev);
+       unsigned char *skb_data_ptr;
+#ifdef ETHDEBUG
+       int i;
+#endif
+       etrax_eth_descr *prevRxDesc;  /* The descriptor right before myNextRxDesc */
+       spin_lock(&np->led_lock);
+       if (!led_active && time_after(jiffies, led_next_time)) {
+               /* light the network leds depending on the current speed. */
+               e100_set_network_leds(NETWORK_ACTIVITY);
+
+               /* Set the earliest time we may clear the LED */
+               led_next_time = jiffies + NET_FLASH_TIME;
+               led_active = 1;
+               mod_timer(&clear_led_timer, jiffies + HZ/10);
+       }
+       spin_unlock(&np->led_lock);
+
+       length = myNextRxDesc->descr.hw_len - 4;
+       dev->stats.rx_bytes += length;
+
+#ifdef ETHDEBUG
+       printk("Got a packet of length %d:\n", length);
+       /* dump the first bytes in the packet */
+       skb_data_ptr = (unsigned char *)phys_to_virt(myNextRxDesc->descr.buf);
+       for (i = 0; i < 8; i++) {
+               printk("%d: %.2x %.2x %.2x %.2x %.2x %.2x %.2x %.2x\n", i * 8,
+                      skb_data_ptr[0],skb_data_ptr[1],skb_data_ptr[2],skb_data_ptr[3],
+                      skb_data_ptr[4],skb_data_ptr[5],skb_data_ptr[6],skb_data_ptr[7]);
+               skb_data_ptr += 8;
+       }
+#endif
+
+       if (length < RX_COPYBREAK) {
+               /* Small packet, copy data */
+               skb = dev_alloc_skb(length - ETHER_HEAD_LEN);
+               if (!skb) {
+                       dev->stats.rx_errors++;
+                       printk(KERN_NOTICE "%s: Memory squeeze, dropping packet.\n", dev->name);
+                       goto update_nextrxdesc;
+               }
+
+               skb_put(skb, length - ETHER_HEAD_LEN);        /* allocate room for the packet body */
+               skb_data_ptr = skb_push(skb, ETHER_HEAD_LEN); /* allocate room for the header */
+
+#ifdef ETHDEBUG
+               printk("head = 0x%x, data = 0x%x, tail = 0x%x, end = 0x%x\n",
+                      skb->head, skb->data, skb_tail_pointer(skb),
+                      skb_end_pointer(skb));
+               printk("copying packet to 0x%x.\n", skb_data_ptr);
+#endif
+
+               memcpy(skb_data_ptr, phys_to_virt(myNextRxDesc->descr.buf), length);
+       }
+       else {
+               /* Large packet, send directly to upper layers and allocate new
+                * memory (aligned to cache line boundary to avoid bug).
+                * Before sending the skb to upper layers we must make sure
+                * that skb->data points to the aligned start of the packet.
+                */
+               int align;
+               struct sk_buff *new_skb = dev_alloc_skb(MAX_MEDIA_DATA_SIZE + 2 * L1_CACHE_BYTES);
+               if (!new_skb) {
+                       dev->stats.rx_errors++;
+                       printk(KERN_NOTICE "%s: Memory squeeze, dropping packet.\n", dev->name);
+                       goto update_nextrxdesc;
+               }
+               skb = myNextRxDesc->skb;
+               align = (int)phys_to_virt(myNextRxDesc->descr.buf) - (int)skb->data;
+               skb_put(skb, length + align);
+               skb_pull(skb, align); /* Remove alignment bytes */
+               myNextRxDesc->skb = new_skb;
+               myNextRxDesc->descr.buf = L1_CACHE_ALIGN(virt_to_phys(myNextRxDesc->skb->data));
+       }
+
+       skb->protocol = eth_type_trans(skb, dev);
+
+       /* Send the packet to the upper layers */
+       netif_rx(skb);
+
+  update_nextrxdesc:
+       /* Prepare for next packet */
+       myNextRxDesc->descr.status = 0;
+       prevRxDesc = myNextRxDesc;
+       myNextRxDesc = phys_to_virt(myNextRxDesc->descr.next);
+
+       rx_queue_len++;
+
+       /* Check if descriptors should be returned */
+       if (rx_queue_len == RX_QUEUE_THRESHOLD) {
+               flush_etrax_cache();
+               prevRxDesc->descr.ctrl |= d_eol;
+               myLastRxDesc->descr.ctrl &= ~d_eol;
+               myLastRxDesc = prevRxDesc;
+               rx_queue_len = 0;
+       }
+}
+
+/* The inverse routine to net_open(). */
+static int
+e100_close(struct net_device *dev)
+{
+       printk(KERN_INFO "Closing %s.\n", dev->name);
+
+       netif_stop_queue(dev);
+
+       *R_IRQ_MASK0_CLR =
+               IO_STATE(R_IRQ_MASK0_CLR, overrun, clr) |
+               IO_STATE(R_IRQ_MASK0_CLR, underrun, clr) |
+               IO_STATE(R_IRQ_MASK0_CLR, excessive_col, clr);
+
+       *R_IRQ_MASK2_CLR =
+               IO_STATE(R_IRQ_MASK2_CLR, dma0_descr, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma0_eop, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma1_descr, clr) |
+               IO_STATE(R_IRQ_MASK2_CLR, dma1_eop, clr);
+
+       /* Stop the receiver and the transmitter */
+
+       RESET_DMA(NETWORK_TX_DMA_NBR);
+       RESET_DMA(NETWORK_RX_DMA_NBR);
+
+       /* Flush the Tx and disable Rx here. */
+
+       free_irq(NETWORK_DMA_RX_IRQ_NBR, (void *)dev);
+       free_irq(NETWORK_DMA_TX_IRQ_NBR, (void *)dev);
+       free_irq(NETWORK_STATUS_IRQ_NBR, (void *)dev);
+
+       cris_free_dma(NETWORK_TX_DMA_NBR, cardname);
+       cris_free_dma(NETWORK_RX_DMA_NBR, cardname);
+
+       /* Update the statistics here. */
+
+       update_rx_stats(&dev->stats);
+       update_tx_stats(&dev->stats);
+
+       /* Stop speed/duplex timers */
+       del_timer(&speed_timer);
+       del_timer(&duplex_timer);
+
+       return 0;
+}
+
+static int
+e100_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+       struct mii_ioctl_data *data = if_mii(ifr);
+       struct net_local *np = netdev_priv(dev);
+       int rc = 0;
+        int old_autoneg;
+
+       spin_lock(&np->lock); /* Preempt protection */
+       switch (cmd) {
+               /* The ioctls below should be considered obsolete but are */
+               /* still present for compatibility with old scripts/apps  */
+               case SET_ETH_SPEED_10:                  /* 10 Mbps */
+                       e100_set_speed(dev, 10);
+                       break;
+               case SET_ETH_SPEED_100:                /* 100 Mbps */
+                       e100_set_speed(dev, 100);
+                       break;
+               case SET_ETH_SPEED_AUTO:        /* Auto-negotiate speed */
+                       e100_set_speed(dev, 0);
+                       break;
+               case SET_ETH_DUPLEX_HALF:       /* Half duplex */
+                       e100_set_duplex(dev, half);
+                       break;
+               case SET_ETH_DUPLEX_FULL:       /* Full duplex */
+                       e100_set_duplex(dev, full);
+                       break;
+               case SET_ETH_DUPLEX_AUTO:       /* Auto-negotiate duplex */
+                       e100_set_duplex(dev, autoneg);
+                       break;
+               case SET_ETH_AUTONEG:
+                       old_autoneg = autoneg_normal;
+                       autoneg_normal = *(int*)data;
+                       if (autoneg_normal != old_autoneg)
+                               e100_negotiate(dev);
+                       break;
+               default:
+                       rc = generic_mii_ioctl(&np->mii_if, if_mii(ifr),
+                                               cmd, NULL);
+                       break;
+       }
+       spin_unlock(&np->lock);
+       return rc;
+}
+
+static int e100_get_settings(struct net_device *dev,
+                            struct ethtool_cmd *cmd)
+{
+       struct net_local *np = netdev_priv(dev);
+       int err;
+
+       spin_lock_irq(&np->lock);
+       err = mii_ethtool_gset(&np->mii_if, cmd);
+       spin_unlock_irq(&np->lock);
+
+       /* The PHY may support 1000baseT, but the Etrax100 does not.  */
+       cmd->supported &= ~(SUPPORTED_1000baseT_Half
+                           | SUPPORTED_1000baseT_Full);
+       return err;
+}
+
+static int e100_set_settings(struct net_device *dev,
+                            struct ethtool_cmd *ecmd)
+{
+       if (ecmd->autoneg == AUTONEG_ENABLE) {
+               e100_set_duplex(dev, autoneg);
+               e100_set_speed(dev, 0);
+       } else {
+               e100_set_duplex(dev, ecmd->duplex == DUPLEX_HALF ? half : full);
+               e100_set_speed(dev, ecmd->speed == SPEED_10 ? 10: 100);
+       }
+
+       return 0;
+}
+
+static void e100_get_drvinfo(struct net_device *dev,
+                            struct ethtool_drvinfo *info)
+{
+       strlcpy(info->driver, "ETRAX 100LX", sizeof(info->driver));
+       strlcpy(info->version, "$Revision: 1.31 $", sizeof(info->version));
+       strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
+       strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
+}
+
+static int e100_nway_reset(struct net_device *dev)
+{
+       if (current_duplex == autoneg && current_speed_selection == 0)
+               e100_negotiate(dev);
+       return 0;
+}
+
+static const struct ethtool_ops e100_ethtool_ops = {
+       .get_settings   = e100_get_settings,
+       .set_settings   = e100_set_settings,
+       .get_drvinfo    = e100_get_drvinfo,
+       .nway_reset     = e100_nway_reset,
+       .get_link       = ethtool_op_get_link,
+};
+
+static int
+e100_set_config(struct net_device *dev, struct ifmap *map)
+{
+       struct net_local *np = netdev_priv(dev);
+
+       spin_lock(&np->lock); /* Preempt protection */
+
+       switch(map->port) {
+               case IF_PORT_UNKNOWN:
+                       /* Use autoneg */
+                       e100_set_speed(dev, 0);
+                       e100_set_duplex(dev, autoneg);
+                       break;
+               case IF_PORT_10BASET:
+                       e100_set_speed(dev, 10);
+                       e100_set_duplex(dev, autoneg);
+                       break;
+               case IF_PORT_100BASET:
+               case IF_PORT_100BASETX:
+                       e100_set_speed(dev, 100);
+                       e100_set_duplex(dev, autoneg);
+                       break;
+               case IF_PORT_100BASEFX:
+               case IF_PORT_10BASE2:
+               case IF_PORT_AUI:
+                       spin_unlock(&np->lock);
+                       return -EOPNOTSUPP;
+               default:
+                       printk(KERN_ERR "%s: Invalid media selected", dev->name);
+                       spin_unlock(&np->lock);
+                       return -EINVAL;
+       }
+       spin_unlock(&np->lock);
+       return 0;
+}
+
+static void
+update_rx_stats(struct net_device_stats *es)
+{
+       unsigned long r = *R_REC_COUNTERS;
+       /* update stats relevant to reception errors */
+       es->rx_fifo_errors += IO_EXTRACT(R_REC_COUNTERS, congestion, r);
+       es->rx_crc_errors += IO_EXTRACT(R_REC_COUNTERS, crc_error, r);
+       es->rx_frame_errors += IO_EXTRACT(R_REC_COUNTERS, alignment_error, r);
+       es->rx_length_errors += IO_EXTRACT(R_REC_COUNTERS, oversize, r);
+}
+
+static void
+update_tx_stats(struct net_device_stats *es)
+{
+       unsigned long r = *R_TR_COUNTERS;
+       /* update stats relevant to transmission errors */
+       es->collisions +=
+               IO_EXTRACT(R_TR_COUNTERS, single_col, r) +
+               IO_EXTRACT(R_TR_COUNTERS, multiple_col, r);
+}
+
+/*
+ * Get the current statistics.
+ * This may be called with the card open or closed.
+ */
+static struct net_device_stats *
+e100_get_stats(struct net_device *dev)
+{
+       struct net_local *lp = netdev_priv(dev);
+       unsigned long flags;
+
+       spin_lock_irqsave(&lp->lock, flags);
+
+       update_rx_stats(&dev->stats);
+       update_tx_stats(&dev->stats);
+
+       spin_unlock_irqrestore(&lp->lock, flags);
+       return &dev->stats;
+}
+
+/*
+ * Set or clear the multicast filter for this adaptor.
+ * num_addrs == -1     Promiscuous mode, receive all packets
+ * num_addrs == 0      Normal mode, clear multicast list
+ * num_addrs > 0       Multicast mode, receive normal and MC packets,
+ *                     and do best-effort filtering.
+ */
+static void
+set_multicast_list(struct net_device *dev)
+{
+       struct net_local *lp = netdev_priv(dev);
+       int num_addr = netdev_mc_count(dev);
+       unsigned long int lo_bits;
+       unsigned long int hi_bits;
+
+       spin_lock(&lp->lock);
+       if (dev->flags & IFF_PROMISC) {
+               /* promiscuous mode */
+               lo_bits = 0xfffffffful;
+               hi_bits = 0xfffffffful;
+
+               /* Enable individual receive */
+               SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, individual, receive);
+               *R_NETWORK_REC_CONFIG = network_rec_config_shadow;
+       } else if (dev->flags & IFF_ALLMULTI) {
+               /* enable all multicasts */
+               lo_bits = 0xfffffffful;
+               hi_bits = 0xfffffffful;
+
+               /* Disable individual receive */
+               SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, individual, discard);
+               *R_NETWORK_REC_CONFIG =  network_rec_config_shadow;
+       } else if (num_addr == 0) {
+               /* Normal, clear the mc list */
+               lo_bits = 0x00000000ul;
+               hi_bits = 0x00000000ul;
+
+               /* Disable individual receive */
+               SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, individual, discard);
+               *R_NETWORK_REC_CONFIG =  network_rec_config_shadow;
+       } else {
+               /* MC mode, receive normal and MC packets */
+               char hash_ix;
+               struct netdev_hw_addr *ha;
+               char *baddr;
+
+               lo_bits = 0x00000000ul;
+               hi_bits = 0x00000000ul;
+               netdev_for_each_mc_addr(ha, dev) {
+                       /* Calculate the hash index for the GA registers */
+
+                       hash_ix = 0;
+                       baddr = ha->addr;
+                       hash_ix ^= (*baddr) & 0x3f;
+                       hash_ix ^= ((*baddr) >> 6) & 0x03;
+                       ++baddr;
+                       hash_ix ^= ((*baddr) << 2) & 0x03c;
+                       hash_ix ^= ((*baddr) >> 4) & 0xf;
+                       ++baddr;
+                       hash_ix ^= ((*baddr) << 4) & 0x30;
+                       hash_ix ^= ((*baddr) >> 2) & 0x3f;
+                       ++baddr;
+                       hash_ix ^= (*baddr) & 0x3f;
+                       hash_ix ^= ((*baddr) >> 6) & 0x03;
+                       ++baddr;
+                       hash_ix ^= ((*baddr) << 2) & 0x03c;
+                       hash_ix ^= ((*baddr) >> 4) & 0xf;
+                       ++baddr;
+                       hash_ix ^= ((*baddr) << 4) & 0x30;
+                       hash_ix ^= ((*baddr) >> 2) & 0x3f;
+
+                       hash_ix &= 0x3f;
+
+                       if (hash_ix >= 32) {
+                               hi_bits |= (1 << (hash_ix-32));
+                       } else {
+                               lo_bits |= (1 << hash_ix);
+                       }
+               }
+               /* Disable individual receive */
+               SETS(network_rec_config_shadow, R_NETWORK_REC_CONFIG, individual, discard);
+               *R_NETWORK_REC_CONFIG = network_rec_config_shadow;
+       }
+       *R_NETWORK_GA_0 = lo_bits;
+       *R_NETWORK_GA_1 = hi_bits;
+       spin_unlock(&lp->lock);
+}
+
+void
+e100_hardware_send_packet(struct net_local *np, char *buf, int length)
+{
+       D(printk("e100 send pack, buf 0x%x len %d\n", buf, length));
+
+       spin_lock(&np->led_lock);
+       if (!led_active && time_after(jiffies, led_next_time)) {
+               /* light the network leds depending on the current speed. */
+               e100_set_network_leds(NETWORK_ACTIVITY);
+
+               /* Set the earliest time we may clear the LED */
+               led_next_time = jiffies + NET_FLASH_TIME;
+               led_active = 1;
+               mod_timer(&clear_led_timer, jiffies + HZ/10);
+       }
+       spin_unlock(&np->led_lock);
+
+       /* configure the tx dma descriptor */
+       myNextTxDesc->descr.sw_len = length;
+       myNextTxDesc->descr.ctrl = d_eop | d_eol | d_wait;
+       myNextTxDesc->descr.buf = virt_to_phys(buf);
+
+        /* Move end of list */
+        myLastTxDesc->descr.ctrl &= ~d_eol;
+        myLastTxDesc = myNextTxDesc;
+
+       /* Restart DMA channel */
+       *R_DMA_CH0_CMD = IO_STATE(R_DMA_CH0_CMD, cmd, restart);
+}
+
+static void
+e100_clear_network_leds(unsigned long dummy)
+{
+       struct net_device *dev = (struct net_device *)dummy;
+       struct net_local *np = netdev_priv(dev);
+
+       spin_lock(&np->led_lock);
+
+       if (led_active && time_after(jiffies, led_next_time)) {
+               e100_set_network_leds(NO_NETWORK_ACTIVITY);
+
+               /* Set the earliest time we may set the LED */
+               led_next_time = jiffies + NET_FLASH_PAUSE;
+               led_active = 0;
+       }
+
+       spin_unlock(&np->led_lock);
+}
+
+static void
+e100_set_network_leds(int active)
+{
+#if defined(CONFIG_ETRAX_NETWORK_LED_ON_WHEN_LINK)
+       int light_leds = (active == NO_NETWORK_ACTIVITY);
+#elif defined(CONFIG_ETRAX_NETWORK_LED_ON_WHEN_ACTIVITY)
+       int light_leds = (active == NETWORK_ACTIVITY);
+#else
+#error "Define either CONFIG_ETRAX_NETWORK_LED_ON_WHEN_LINK or CONFIG_ETRAX_NETWORK_LED_ON_WHEN_ACTIVITY"
+#endif
+
+       if (!current_speed) {
+               /* Make LED red, link is down */
+               CRIS_LED_NETWORK_SET(CRIS_LED_OFF);
+       } else if (light_leds) {
+               if (current_speed == 10) {
+                       CRIS_LED_NETWORK_SET(CRIS_LED_ORANGE);
+               } else {
+                       CRIS_LED_NETWORK_SET(CRIS_LED_GREEN);
+               }
+       } else {
+               CRIS_LED_NETWORK_SET(CRIS_LED_OFF);
+       }
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void
+e100_netpoll(struct net_device* netdev)
+{
+       e100rxtx_interrupt(NETWORK_DMA_TX_IRQ_NBR, netdev);
+}
+#endif
+
+static int
+etrax_init_module(void)
+{
+       return etrax_ethernet_init();
+}
+
+static int __init
+e100_boot_setup(char* str)
+{
+       struct sockaddr sa = {0};
+       int i;
+
+       /* Parse the colon separated Ethernet station address */
+       for (i = 0; i <  ETH_ALEN; i++) {
+               unsigned int tmp;
+               if (sscanf(str + 3*i, "%2x", &tmp) != 1) {
+                       printk(KERN_WARNING "Malformed station address");
+                       return 0;
+               }
+               sa.sa_data[i] = (char)tmp;
+       }
+
+       default_mac = sa;
+       return 1;
+}
+
+__setup("etrax100_eth=", e100_boot_setup);
+
+module_init(etrax_init_module);