Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / net / ethernet / arc / emac_main.c
diff --git a/kernel/drivers/net/ethernet/arc/emac_main.c b/kernel/drivers/net/ethernet/arc/emac_main.c
new file mode 100644 (file)
index 0000000..abe1eab
--- /dev/null
@@ -0,0 +1,874 @@
+/*
+ * Copyright (C) 2004-2013 Synopsys, Inc. (www.synopsys.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * Driver for the ARC EMAC 10100 (hardware revision 5)
+ *
+ * Contributors:
+ *             Amit Bhor
+ *             Sameer Dhavale
+ *             Vineet Gupta
+ */
+
+#include <linux/crc32.h>
+#include <linux/etherdevice.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/of_mdio.h>
+#include <linux/of_net.h>
+#include <linux/of_platform.h>
+
+#include "emac.h"
+
+
+/**
+ * arc_emac_tx_avail - Return the number of available slots in the tx ring.
+ * @priv: Pointer to ARC EMAC private data structure.
+ *
+ * returns: the number of slots available for transmission in tx the ring.
+ */
+static inline int arc_emac_tx_avail(struct arc_emac_priv *priv)
+{
+       return (priv->txbd_dirty + TX_BD_NUM - priv->txbd_curr - 1) % TX_BD_NUM;
+}
+
+/**
+ * arc_emac_adjust_link - Adjust the PHY link duplex.
+ * @ndev:      Pointer to the net_device structure.
+ *
+ * This function is called to change the duplex setting after auto negotiation
+ * is done by the PHY.
+ */
+static void arc_emac_adjust_link(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       struct phy_device *phy_dev = priv->phy_dev;
+       unsigned int reg, state_changed = 0;
+
+       if (priv->link != phy_dev->link) {
+               priv->link = phy_dev->link;
+               state_changed = 1;
+       }
+
+       if (priv->speed != phy_dev->speed) {
+               priv->speed = phy_dev->speed;
+               state_changed = 1;
+               if (priv->set_mac_speed)
+                       priv->set_mac_speed(priv, priv->speed);
+       }
+
+       if (priv->duplex != phy_dev->duplex) {
+               reg = arc_reg_get(priv, R_CTRL);
+
+               if (DUPLEX_FULL == phy_dev->duplex)
+                       reg |= ENFL_MASK;
+               else
+                       reg &= ~ENFL_MASK;
+
+               arc_reg_set(priv, R_CTRL, reg);
+               priv->duplex = phy_dev->duplex;
+               state_changed = 1;
+       }
+
+       if (state_changed)
+               phy_print_status(phy_dev);
+}
+
+/**
+ * arc_emac_get_settings - Get PHY settings.
+ * @ndev:      Pointer to net_device structure.
+ * @cmd:       Pointer to ethtool_cmd structure.
+ *
+ * This implements ethtool command for getting PHY settings. If PHY could
+ * not be found, the function returns -ENODEV. This function calls the
+ * relevant PHY ethtool API to get the PHY settings.
+ * Issue "ethtool ethX" under linux prompt to execute this function.
+ */
+static int arc_emac_get_settings(struct net_device *ndev,
+                                struct ethtool_cmd *cmd)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       return phy_ethtool_gset(priv->phy_dev, cmd);
+}
+
+/**
+ * arc_emac_set_settings - Set PHY settings as passed in the argument.
+ * @ndev:      Pointer to net_device structure.
+ * @cmd:       Pointer to ethtool_cmd structure.
+ *
+ * This implements ethtool command for setting various PHY settings. If PHY
+ * could not be found, the function returns -ENODEV. This function calls the
+ * relevant PHY ethtool API to set the PHY.
+ * Issue e.g. "ethtool -s ethX speed 1000" under linux prompt to execute this
+ * function.
+ */
+static int arc_emac_set_settings(struct net_device *ndev,
+                                struct ethtool_cmd *cmd)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       return phy_ethtool_sset(priv->phy_dev, cmd);
+}
+
+/**
+ * arc_emac_get_drvinfo - Get EMAC driver information.
+ * @ndev:      Pointer to net_device structure.
+ * @info:      Pointer to ethtool_drvinfo structure.
+ *
+ * This implements ethtool command for getting the driver information.
+ * Issue "ethtool -i ethX" under linux prompt to execute this function.
+ */
+static void arc_emac_get_drvinfo(struct net_device *ndev,
+                                struct ethtool_drvinfo *info)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       strlcpy(info->driver, priv->drv_name, sizeof(info->driver));
+       strlcpy(info->version, priv->drv_version, sizeof(info->version));
+}
+
+static const struct ethtool_ops arc_emac_ethtool_ops = {
+       .get_settings   = arc_emac_get_settings,
+       .set_settings   = arc_emac_set_settings,
+       .get_drvinfo    = arc_emac_get_drvinfo,
+       .get_link       = ethtool_op_get_link,
+};
+
+#define FIRST_OR_LAST_MASK     (FIRST_MASK | LAST_MASK)
+
+/**
+ * arc_emac_tx_clean - clears processed by EMAC Tx BDs.
+ * @ndev:      Pointer to the network device.
+ */
+static void arc_emac_tx_clean(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       struct net_device_stats *stats = &ndev->stats;
+       unsigned int i;
+
+       for (i = 0; i < TX_BD_NUM; i++) {
+               unsigned int *txbd_dirty = &priv->txbd_dirty;
+               struct arc_emac_bd *txbd = &priv->txbd[*txbd_dirty];
+               struct buffer_state *tx_buff = &priv->tx_buff[*txbd_dirty];
+               struct sk_buff *skb = tx_buff->skb;
+               unsigned int info = le32_to_cpu(txbd->info);
+
+               if ((info & FOR_EMAC) || !txbd->data)
+                       break;
+
+               if (unlikely(info & (DROP | DEFR | LTCL | UFLO))) {
+                       stats->tx_errors++;
+                       stats->tx_dropped++;
+
+                       if (info & DEFR)
+                               stats->tx_carrier_errors++;
+
+                       if (info & LTCL)
+                               stats->collisions++;
+
+                       if (info & UFLO)
+                               stats->tx_fifo_errors++;
+               } else if (likely(info & FIRST_OR_LAST_MASK)) {
+                       stats->tx_packets++;
+                       stats->tx_bytes += skb->len;
+               }
+
+               dma_unmap_single(&ndev->dev, dma_unmap_addr(tx_buff, addr),
+                                dma_unmap_len(tx_buff, len), DMA_TO_DEVICE);
+
+               /* return the sk_buff to system */
+               dev_kfree_skb_irq(skb);
+
+               txbd->data = 0;
+               txbd->info = 0;
+
+               *txbd_dirty = (*txbd_dirty + 1) % TX_BD_NUM;
+       }
+
+       /* Ensure that txbd_dirty is visible to tx() before checking
+        * for queue stopped.
+        */
+       smp_mb();
+
+       if (netif_queue_stopped(ndev) && arc_emac_tx_avail(priv))
+               netif_wake_queue(ndev);
+}
+
+/**
+ * arc_emac_rx - processing of Rx packets.
+ * @ndev:      Pointer to the network device.
+ * @budget:    How many BDs to process on 1 call.
+ *
+ * returns:    Number of processed BDs
+ *
+ * Iterate through Rx BDs and deliver received packages to upper layer.
+ */
+static int arc_emac_rx(struct net_device *ndev, int budget)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       unsigned int work_done;
+
+       for (work_done = 0; work_done < budget; work_done++) {
+               unsigned int *last_rx_bd = &priv->last_rx_bd;
+               struct net_device_stats *stats = &ndev->stats;
+               struct buffer_state *rx_buff = &priv->rx_buff[*last_rx_bd];
+               struct arc_emac_bd *rxbd = &priv->rxbd[*last_rx_bd];
+               unsigned int pktlen, info = le32_to_cpu(rxbd->info);
+               struct sk_buff *skb;
+               dma_addr_t addr;
+
+               if (unlikely((info & OWN_MASK) == FOR_EMAC))
+                       break;
+
+               /* Make a note that we saw a packet at this BD.
+                * So next time, driver starts from this + 1
+                */
+               *last_rx_bd = (*last_rx_bd + 1) % RX_BD_NUM;
+
+               if (unlikely((info & FIRST_OR_LAST_MASK) !=
+                            FIRST_OR_LAST_MASK)) {
+                       /* We pre-allocate buffers of MTU size so incoming
+                        * packets won't be split/chained.
+                        */
+                       if (net_ratelimit())
+                               netdev_err(ndev, "incomplete packet received\n");
+
+                       /* Return ownership to EMAC */
+                       rxbd->info = cpu_to_le32(FOR_EMAC | EMAC_BUFFER_SIZE);
+                       stats->rx_errors++;
+                       stats->rx_length_errors++;
+                       continue;
+               }
+
+               pktlen = info & LEN_MASK;
+               stats->rx_packets++;
+               stats->rx_bytes += pktlen;
+               skb = rx_buff->skb;
+               skb_put(skb, pktlen);
+               skb->dev = ndev;
+               skb->protocol = eth_type_trans(skb, ndev);
+
+               dma_unmap_single(&ndev->dev, dma_unmap_addr(rx_buff, addr),
+                                dma_unmap_len(rx_buff, len), DMA_FROM_DEVICE);
+
+               /* Prepare the BD for next cycle */
+               rx_buff->skb = netdev_alloc_skb_ip_align(ndev,
+                                                        EMAC_BUFFER_SIZE);
+               if (unlikely(!rx_buff->skb)) {
+                       stats->rx_errors++;
+                       /* Because receive_skb is below, increment rx_dropped */
+                       stats->rx_dropped++;
+                       continue;
+               }
+
+               /* receive_skb only if new skb was allocated to avoid holes */
+               netif_receive_skb(skb);
+
+               addr = dma_map_single(&ndev->dev, (void *)rx_buff->skb->data,
+                                     EMAC_BUFFER_SIZE, DMA_FROM_DEVICE);
+               if (dma_mapping_error(&ndev->dev, addr)) {
+                       if (net_ratelimit())
+                               netdev_err(ndev, "cannot dma map\n");
+                       dev_kfree_skb(rx_buff->skb);
+                       stats->rx_errors++;
+                       continue;
+               }
+               dma_unmap_addr_set(rx_buff, addr, addr);
+               dma_unmap_len_set(rx_buff, len, EMAC_BUFFER_SIZE);
+
+               rxbd->data = cpu_to_le32(addr);
+
+               /* Make sure pointer to data buffer is set */
+               wmb();
+
+               /* Return ownership to EMAC */
+               rxbd->info = cpu_to_le32(FOR_EMAC | EMAC_BUFFER_SIZE);
+       }
+
+       return work_done;
+}
+
+/**
+ * arc_emac_poll - NAPI poll handler.
+ * @napi:      Pointer to napi_struct structure.
+ * @budget:    How many BDs to process on 1 call.
+ *
+ * returns:    Number of processed BDs
+ */
+static int arc_emac_poll(struct napi_struct *napi, int budget)
+{
+       struct net_device *ndev = napi->dev;
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       unsigned int work_done;
+
+       arc_emac_tx_clean(ndev);
+
+       work_done = arc_emac_rx(ndev, budget);
+       if (work_done < budget) {
+               napi_complete(napi);
+               arc_reg_or(priv, R_ENABLE, RXINT_MASK | TXINT_MASK);
+       }
+
+       return work_done;
+}
+
+/**
+ * arc_emac_intr - Global interrupt handler for EMAC.
+ * @irq:               irq number.
+ * @dev_instance:      device instance.
+ *
+ * returns: IRQ_HANDLED for all cases.
+ *
+ * ARC EMAC has only 1 interrupt line, and depending on bits raised in
+ * STATUS register we may tell what is a reason for interrupt to fire.
+ */
+static irqreturn_t arc_emac_intr(int irq, void *dev_instance)
+{
+       struct net_device *ndev = dev_instance;
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       struct net_device_stats *stats = &ndev->stats;
+       unsigned int status;
+
+       status = arc_reg_get(priv, R_STATUS);
+       status &= ~MDIO_MASK;
+
+       /* Reset all flags except "MDIO complete" */
+       arc_reg_set(priv, R_STATUS, status);
+
+       if (status & (RXINT_MASK | TXINT_MASK)) {
+               if (likely(napi_schedule_prep(&priv->napi))) {
+                       arc_reg_clr(priv, R_ENABLE, RXINT_MASK | TXINT_MASK);
+                       __napi_schedule(&priv->napi);
+               }
+       }
+
+       if (status & ERR_MASK) {
+               /* MSER/RXCR/RXFR/RXFL interrupt fires on corresponding
+                * 8-bit error counter overrun.
+                */
+
+               if (status & MSER_MASK) {
+                       stats->rx_missed_errors += 0x100;
+                       stats->rx_errors += 0x100;
+               }
+
+               if (status & RXCR_MASK) {
+                       stats->rx_crc_errors += 0x100;
+                       stats->rx_errors += 0x100;
+               }
+
+               if (status & RXFR_MASK) {
+                       stats->rx_frame_errors += 0x100;
+                       stats->rx_errors += 0x100;
+               }
+
+               if (status & RXFL_MASK) {
+                       stats->rx_over_errors += 0x100;
+                       stats->rx_errors += 0x100;
+               }
+       }
+
+       return IRQ_HANDLED;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void arc_emac_poll_controller(struct net_device *dev)
+{
+       disable_irq(dev->irq);
+       arc_emac_intr(dev->irq, dev);
+       enable_irq(dev->irq);
+}
+#endif
+
+/**
+ * arc_emac_open - Open the network device.
+ * @ndev:      Pointer to the network device.
+ *
+ * returns: 0, on success or non-zero error value on failure.
+ *
+ * This function sets the MAC address, requests and enables an IRQ
+ * for the EMAC device and starts the Tx queue.
+ * It also connects to the phy device.
+ */
+static int arc_emac_open(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       struct phy_device *phy_dev = priv->phy_dev;
+       int i;
+
+       phy_dev->autoneg = AUTONEG_ENABLE;
+       phy_dev->speed = 0;
+       phy_dev->duplex = 0;
+       phy_dev->advertising &= phy_dev->supported;
+
+       priv->last_rx_bd = 0;
+
+       /* Allocate and set buffers for Rx BD's */
+       for (i = 0; i < RX_BD_NUM; i++) {
+               dma_addr_t addr;
+               unsigned int *last_rx_bd = &priv->last_rx_bd;
+               struct arc_emac_bd *rxbd = &priv->rxbd[*last_rx_bd];
+               struct buffer_state *rx_buff = &priv->rx_buff[*last_rx_bd];
+
+               rx_buff->skb = netdev_alloc_skb_ip_align(ndev,
+                                                        EMAC_BUFFER_SIZE);
+               if (unlikely(!rx_buff->skb))
+                       return -ENOMEM;
+
+               addr = dma_map_single(&ndev->dev, (void *)rx_buff->skb->data,
+                                     EMAC_BUFFER_SIZE, DMA_FROM_DEVICE);
+               if (dma_mapping_error(&ndev->dev, addr)) {
+                       netdev_err(ndev, "cannot dma map\n");
+                       dev_kfree_skb(rx_buff->skb);
+                       return -ENOMEM;
+               }
+               dma_unmap_addr_set(rx_buff, addr, addr);
+               dma_unmap_len_set(rx_buff, len, EMAC_BUFFER_SIZE);
+
+               rxbd->data = cpu_to_le32(addr);
+
+               /* Make sure pointer to data buffer is set */
+               wmb();
+
+               /* Return ownership to EMAC */
+               rxbd->info = cpu_to_le32(FOR_EMAC | EMAC_BUFFER_SIZE);
+
+               *last_rx_bd = (*last_rx_bd + 1) % RX_BD_NUM;
+       }
+
+       /* Clean Tx BD's */
+       memset(priv->txbd, 0, TX_RING_SZ);
+
+       /* Initialize logical address filter */
+       arc_reg_set(priv, R_LAFL, 0);
+       arc_reg_set(priv, R_LAFH, 0);
+
+       /* Set BD ring pointers for device side */
+       arc_reg_set(priv, R_RX_RING, (unsigned int)priv->rxbd_dma);
+       arc_reg_set(priv, R_TX_RING, (unsigned int)priv->txbd_dma);
+
+       /* Enable interrupts */
+       arc_reg_set(priv, R_ENABLE, RXINT_MASK | TXINT_MASK | ERR_MASK);
+
+       /* Set CONTROL */
+       arc_reg_set(priv, R_CTRL,
+                    (RX_BD_NUM << 24) |        /* RX BD table length */
+                    (TX_BD_NUM << 16) |        /* TX BD table length */
+                    TXRN_MASK | RXRN_MASK);
+
+       napi_enable(&priv->napi);
+
+       /* Enable EMAC */
+       arc_reg_or(priv, R_CTRL, EN_MASK);
+
+       phy_start_aneg(priv->phy_dev);
+
+       netif_start_queue(ndev);
+
+       return 0;
+}
+
+/**
+ * arc_emac_set_rx_mode - Change the receive filtering mode.
+ * @ndev:      Pointer to the network device.
+ *
+ * This function enables/disables promiscuous or all-multicast mode
+ * and updates the multicast filtering list of the network device.
+ */
+static void arc_emac_set_rx_mode(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       if (ndev->flags & IFF_PROMISC) {
+               arc_reg_or(priv, R_CTRL, PROM_MASK);
+       } else {
+               arc_reg_clr(priv, R_CTRL, PROM_MASK);
+
+               if (ndev->flags & IFF_ALLMULTI) {
+                       arc_reg_set(priv, R_LAFL, ~0);
+                       arc_reg_set(priv, R_LAFH, ~0);
+               } else {
+                       struct netdev_hw_addr *ha;
+                       unsigned int filter[2] = { 0, 0 };
+                       int bit;
+
+                       netdev_for_each_mc_addr(ha, ndev) {
+                               bit = ether_crc_le(ETH_ALEN, ha->addr) >> 26;
+                               filter[bit >> 5] |= 1 << (bit & 31);
+                       }
+
+                       arc_reg_set(priv, R_LAFL, filter[0]);
+                       arc_reg_set(priv, R_LAFH, filter[1]);
+               }
+       }
+}
+
+/**
+ * arc_emac_stop - Close the network device.
+ * @ndev:      Pointer to the network device.
+ *
+ * This function stops the Tx queue, disables interrupts and frees the IRQ for
+ * the EMAC device.
+ * It also disconnects the PHY device associated with the EMAC device.
+ */
+static int arc_emac_stop(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       napi_disable(&priv->napi);
+       netif_stop_queue(ndev);
+
+       /* Disable interrupts */
+       arc_reg_clr(priv, R_ENABLE, RXINT_MASK | TXINT_MASK | ERR_MASK);
+
+       /* Disable EMAC */
+       arc_reg_clr(priv, R_CTRL, EN_MASK);
+
+       return 0;
+}
+
+/**
+ * arc_emac_stats - Get system network statistics.
+ * @ndev:      Pointer to net_device structure.
+ *
+ * Returns the address of the device statistics structure.
+ * Statistics are updated in interrupt handler.
+ */
+static struct net_device_stats *arc_emac_stats(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       struct net_device_stats *stats = &ndev->stats;
+       unsigned long miss, rxerr;
+       u8 rxcrc, rxfram, rxoflow;
+
+       rxerr = arc_reg_get(priv, R_RXERR);
+       miss = arc_reg_get(priv, R_MISS);
+
+       rxcrc = rxerr;
+       rxfram = rxerr >> 8;
+       rxoflow = rxerr >> 16;
+
+       stats->rx_errors += miss;
+       stats->rx_errors += rxcrc + rxfram + rxoflow;
+
+       stats->rx_over_errors += rxoflow;
+       stats->rx_frame_errors += rxfram;
+       stats->rx_crc_errors += rxcrc;
+       stats->rx_missed_errors += miss;
+
+       return stats;
+}
+
+/**
+ * arc_emac_tx - Starts the data transmission.
+ * @skb:       sk_buff pointer that contains data to be Transmitted.
+ * @ndev:      Pointer to net_device structure.
+ *
+ * returns: NETDEV_TX_OK, on success
+ *             NETDEV_TX_BUSY, if any of the descriptors are not free.
+ *
+ * This function is invoked from upper layers to initiate transmission.
+ */
+static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       unsigned int len, *txbd_curr = &priv->txbd_curr;
+       struct net_device_stats *stats = &ndev->stats;
+       __le32 *info = &priv->txbd[*txbd_curr].info;
+       dma_addr_t addr;
+
+       if (skb_padto(skb, ETH_ZLEN))
+               return NETDEV_TX_OK;
+
+       len = max_t(unsigned int, ETH_ZLEN, skb->len);
+
+       if (unlikely(!arc_emac_tx_avail(priv))) {
+               netif_stop_queue(ndev);
+               netdev_err(ndev, "BUG! Tx Ring full when queue awake!\n");
+               return NETDEV_TX_BUSY;
+       }
+
+       addr = dma_map_single(&ndev->dev, (void *)skb->data, len,
+                             DMA_TO_DEVICE);
+
+       if (unlikely(dma_mapping_error(&ndev->dev, addr))) {
+               stats->tx_dropped++;
+               stats->tx_errors++;
+               dev_kfree_skb(skb);
+               return NETDEV_TX_OK;
+       }
+       dma_unmap_addr_set(&priv->tx_buff[*txbd_curr], addr, addr);
+       dma_unmap_len_set(&priv->tx_buff[*txbd_curr], len, len);
+
+       priv->tx_buff[*txbd_curr].skb = skb;
+       priv->txbd[*txbd_curr].data = cpu_to_le32(addr);
+
+       /* Make sure pointer to data buffer is set */
+       wmb();
+
+       skb_tx_timestamp(skb);
+
+       *info = cpu_to_le32(FOR_EMAC | FIRST_OR_LAST_MASK | len);
+
+       /* Increment index to point to the next BD */
+       *txbd_curr = (*txbd_curr + 1) % TX_BD_NUM;
+
+       /* Ensure that tx_clean() sees the new txbd_curr before
+        * checking the queue status. This prevents an unneeded wake
+        * of the queue in tx_clean().
+        */
+       smp_mb();
+
+       if (!arc_emac_tx_avail(priv)) {
+               netif_stop_queue(ndev);
+               /* Refresh tx_dirty */
+               smp_mb();
+               if (arc_emac_tx_avail(priv))
+                       netif_start_queue(ndev);
+       }
+
+       arc_reg_set(priv, R_STATUS, TXPL_MASK);
+
+       return NETDEV_TX_OK;
+}
+
+static void arc_emac_set_address_internal(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+       unsigned int addr_low, addr_hi;
+
+       addr_low = le32_to_cpu(*(__le32 *) &ndev->dev_addr[0]);
+       addr_hi = le16_to_cpu(*(__le16 *) &ndev->dev_addr[4]);
+
+       arc_reg_set(priv, R_ADDRL, addr_low);
+       arc_reg_set(priv, R_ADDRH, addr_hi);
+}
+
+/**
+ * arc_emac_set_address - Set the MAC address for this device.
+ * @ndev:      Pointer to net_device structure.
+ * @p:         6 byte Address to be written as MAC address.
+ *
+ * This function copies the HW address from the sockaddr structure to the
+ * net_device structure and updates the address in HW.
+ *
+ * returns:    -EBUSY if the net device is busy or 0 if the address is set
+ *             successfully.
+ */
+static int arc_emac_set_address(struct net_device *ndev, void *p)
+{
+       struct sockaddr *addr = p;
+
+       if (netif_running(ndev))
+               return -EBUSY;
+
+       if (!is_valid_ether_addr(addr->sa_data))
+               return -EADDRNOTAVAIL;
+
+       memcpy(ndev->dev_addr, addr->sa_data, ndev->addr_len);
+
+       arc_emac_set_address_internal(ndev);
+
+       return 0;
+}
+
+static const struct net_device_ops arc_emac_netdev_ops = {
+       .ndo_open               = arc_emac_open,
+       .ndo_stop               = arc_emac_stop,
+       .ndo_start_xmit         = arc_emac_tx,
+       .ndo_set_mac_address    = arc_emac_set_address,
+       .ndo_get_stats          = arc_emac_stats,
+       .ndo_set_rx_mode        = arc_emac_set_rx_mode,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+       .ndo_poll_controller    = arc_emac_poll_controller,
+#endif
+};
+
+int arc_emac_probe(struct net_device *ndev, int interface)
+{
+       struct device *dev = ndev->dev.parent;
+       struct resource res_regs;
+       struct device_node *phy_node;
+       struct arc_emac_priv *priv;
+       const char *mac_addr;
+       unsigned int id, clock_frequency, irq;
+       int err;
+
+
+       /* Get PHY from device tree */
+       phy_node = of_parse_phandle(dev->of_node, "phy", 0);
+       if (!phy_node) {
+               dev_err(dev, "failed to retrieve phy description from device tree\n");
+               return -ENODEV;
+       }
+
+       /* Get EMAC registers base address from device tree */
+       err = of_address_to_resource(dev->of_node, 0, &res_regs);
+       if (err) {
+               dev_err(dev, "failed to retrieve registers base from device tree\n");
+               return -ENODEV;
+       }
+
+       /* Get IRQ from device tree */
+       irq = irq_of_parse_and_map(dev->of_node, 0);
+       if (!irq) {
+               dev_err(dev, "failed to retrieve <irq> value from device tree\n");
+               return -ENODEV;
+       }
+
+
+       ndev->netdev_ops = &arc_emac_netdev_ops;
+       ndev->ethtool_ops = &arc_emac_ethtool_ops;
+       ndev->watchdog_timeo = TX_TIMEOUT;
+       /* FIXME :: no multicast support yet */
+       ndev->flags &= ~IFF_MULTICAST;
+
+       priv = netdev_priv(ndev);
+       priv->dev = dev;
+
+       priv->regs = devm_ioremap_resource(dev, &res_regs);
+       if (IS_ERR(priv->regs)) {
+               return PTR_ERR(priv->regs);
+       }
+       dev_dbg(dev, "Registers base address is 0x%p\n", priv->regs);
+
+       if (priv->clk) {
+               err = clk_prepare_enable(priv->clk);
+               if (err) {
+                       dev_err(dev, "failed to enable clock\n");
+                       return err;
+               }
+
+               clock_frequency = clk_get_rate(priv->clk);
+       } else {
+               /* Get CPU clock frequency from device tree */
+               if (of_property_read_u32(dev->of_node, "clock-frequency",
+                                        &clock_frequency)) {
+                       dev_err(dev, "failed to retrieve <clock-frequency> from device tree\n");
+                       return -EINVAL;
+               }
+       }
+
+       id = arc_reg_get(priv, R_ID);
+
+       /* Check for EMAC revision 5 or 7, magic number */
+       if (!(id == 0x0005fd02 || id == 0x0007fd02)) {
+               dev_err(dev, "ARC EMAC not detected, id=0x%x\n", id);
+               err = -ENODEV;
+               goto out_clken;
+       }
+       dev_info(dev, "ARC EMAC detected with id: 0x%x\n", id);
+
+       /* Set poll rate so that it polls every 1 ms */
+       arc_reg_set(priv, R_POLLRATE, clock_frequency / 1000000);
+
+       ndev->irq = irq;
+       dev_info(dev, "IRQ is %d\n", ndev->irq);
+
+       /* Register interrupt handler for device */
+       err = devm_request_irq(dev, ndev->irq, arc_emac_intr, 0,
+                              ndev->name, ndev);
+       if (err) {
+               dev_err(dev, "could not allocate IRQ\n");
+               goto out_clken;
+       }
+
+       /* Get MAC address from device tree */
+       mac_addr = of_get_mac_address(dev->of_node);
+
+       if (mac_addr)
+               memcpy(ndev->dev_addr, mac_addr, ETH_ALEN);
+       else
+               eth_hw_addr_random(ndev);
+
+       arc_emac_set_address_internal(ndev);
+       dev_info(dev, "MAC address is now %pM\n", ndev->dev_addr);
+
+       /* Do 1 allocation instead of 2 separate ones for Rx and Tx BD rings */
+       priv->rxbd = dmam_alloc_coherent(dev, RX_RING_SZ + TX_RING_SZ,
+                                        &priv->rxbd_dma, GFP_KERNEL);
+
+       if (!priv->rxbd) {
+               dev_err(dev, "failed to allocate data buffers\n");
+               err = -ENOMEM;
+               goto out_clken;
+       }
+
+       priv->txbd = priv->rxbd + RX_BD_NUM;
+
+       priv->txbd_dma = priv->rxbd_dma + RX_RING_SZ;
+       dev_dbg(dev, "EMAC Device addr: Rx Ring [0x%x], Tx Ring[%x]\n",
+               (unsigned int)priv->rxbd_dma, (unsigned int)priv->txbd_dma);
+
+       err = arc_mdio_probe(priv);
+       if (err) {
+               dev_err(dev, "failed to probe MII bus\n");
+               goto out_clken;
+       }
+
+       priv->phy_dev = of_phy_connect(ndev, phy_node, arc_emac_adjust_link, 0,
+                                      interface);
+       if (!priv->phy_dev) {
+               dev_err(dev, "of_phy_connect() failed\n");
+               err = -ENODEV;
+               goto out_mdio;
+       }
+
+       dev_info(dev, "connected to %s phy with id 0x%x\n",
+                priv->phy_dev->drv->name, priv->phy_dev->phy_id);
+
+       netif_napi_add(ndev, &priv->napi, arc_emac_poll, ARC_EMAC_NAPI_WEIGHT);
+
+       err = register_netdev(ndev);
+       if (err) {
+               dev_err(dev, "failed to register network device\n");
+               goto out_netif_api;
+       }
+
+       return 0;
+
+out_netif_api:
+       netif_napi_del(&priv->napi);
+       phy_disconnect(priv->phy_dev);
+       priv->phy_dev = NULL;
+out_mdio:
+       arc_mdio_remove(priv);
+out_clken:
+       if (priv->clk)
+               clk_disable_unprepare(priv->clk);
+       return err;
+}
+EXPORT_SYMBOL_GPL(arc_emac_probe);
+
+int arc_emac_remove(struct net_device *ndev)
+{
+       struct arc_emac_priv *priv = netdev_priv(ndev);
+
+       phy_disconnect(priv->phy_dev);
+       priv->phy_dev = NULL;
+       arc_mdio_remove(priv);
+       unregister_netdev(ndev);
+       netif_napi_del(&priv->napi);
+
+       if (!IS_ERR(priv->clk)) {
+               clk_disable_unprepare(priv->clk);
+       }
+
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(arc_emac_remove);
+
+MODULE_AUTHOR("Alexey Brodkin <abrodkin@synopsys.com>");
+MODULE_DESCRIPTION("ARC EMAC driver");
+MODULE_LICENSE("GPL");