Re: [patch 3/3] Add tsi108 On Chip Ethernet device driver support

From: Jeff Garzik
Date: Thu Sep 21 2006 - 00:55:46 EST


Zang Roy-r61911 wrote:
+struct tsi108_prv_data {
+ void __iomem *regs; /* Base of normal regs */
+ void __iomem *phyregs; /* Base of register bank used for PHY access */
+
+ int phy; /* Index of PHY for this interface */
+ int irq_num;
+ int id;
+
+ struct timer_list timer;/* Timer that triggers the check phy function */
+ int rxtail; /* Next entry in rxring to read */
+ int rxhead; /* Next entry in rxring to give a new buffer */
+ int rxfree; /* Number of free, allocated RX buffers */
+
+ int rxpending; /* Non-zero if there are still descriptors
+ * to be processed from a previous descriptor
+ * interrupt condition that has been cleared */
+
+ int txtail; /* Next TX descriptor to check status on */
+ int txhead; /* Next TX descriptor to use */

most of these should be unsigned, to prevent bugs.


+ /* Number of free TX descriptors. This could be calculated from
+ * rxhead and rxtail if one descriptor were left unused to disambiguate
+ * full and empty conditions, but it's simpler to just keep track
+ * explicitly. */
+
+ int txfree;
+
+ int phy_ok; /* The PHY is currently powered on. */
+
+ /* PHY status (duplex is 1 for half, 2 for full,
+ * so that the default 0 indicates that neither has
+ * yet been configured). */
+
+ int link_up;
+ int speed;
+ int duplex;
+
+ tx_desc *txring;
+ rx_desc *rxring;
+ struct sk_buff *txskbs[TSI108_TXRING_LEN];
+ struct sk_buff *rxskbs[TSI108_RXRING_LEN];
+
+ dma_addr_t txdma, rxdma;
+
+ /* txlock nests in misclock and phy_lock */
+
+ spinlock_t txlock, misclock;
+
+ /* stats is used to hold the upper bits of each hardware counter,
+ * and tmpstats is used to hold the full values for returning
+ * to the caller of get_stats(). They must be separate in case
+ * an overflow interrupt occurs before the stats are consumed.
+ */
+
+ struct net_device_stats stats;
+ struct net_device_stats tmpstats;
+
+ /* These stats are kept separate in hardware, thus require individual
+ * fields for handling carry. They are combined in get_stats.
+ */
+
+ unsigned long rx_fcs; /* Add to rx_frame_errors */
+ unsigned long rx_short_fcs; /* Add to rx_frame_errors */
+ unsigned long rx_long_fcs; /* Add to rx_frame_errors */
+ unsigned long rx_underruns; /* Add to rx_length_errors */
+ unsigned long rx_overruns; /* Add to rx_length_errors */
+
+ unsigned long tx_coll_abort; /* Add to tx_aborted_errors/collisions */
+ unsigned long tx_pause_drop; /* Add to tx_aborted_errors */
+
+ unsigned long mc_hash[16];
+};
+
+/* Structure for a device driver */
+
+static struct platform_driver tsi_eth_driver = {
+ .probe = tsi108_init_one,
+ .remove = tsi108_ether_remove,
+ .driver = {
+ .name = "tsi-ethernet",
+ },
+};
+
+static void tsi108_timed_checker(unsigned long dev_ptr);
+
+static void dump_eth_one(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ printk("Dumping %s...\n", dev->name);
+ printk("intstat %x intmask %x phy_ok %d"
+ " link %d speed %d duplex %d\n",
+ TSI108_ETH_READ_REG(TSI108_EC_INTSTAT),
+ TSI108_ETH_READ_REG(TSI108_EC_INTMASK), data->phy_ok,
+ data->link_up, data->speed, data->duplex);
+
+ printk("TX: head %d, tail %d, free %d, stat %x, estat %x, err %x\n",
+ data->txhead, data->txtail, data->txfree,
+ TSI108_ETH_READ_REG(TSI108_EC_TXSTAT),
+ TSI108_ETH_READ_REG(TSI108_EC_TXESTAT),
+ TSI108_ETH_READ_REG(TSI108_EC_TXERR));
+
+ printk("RX: head %d, tail %d, free %d, stat %x,"
+ " estat %x, err %x, pending %d\n\n",
+ data->rxhead, data->rxtail, data->rxfree,
+ TSI108_ETH_READ_REG(TSI108_EC_RXSTAT),
+ TSI108_ETH_READ_REG(TSI108_EC_RXESTAT),
+ TSI108_ETH_READ_REG(TSI108_EC_RXERR), data->rxpending);
+}
+
+/* Synchronization is needed between the thread and up/down events.
+ * Note that the PHY is accessed through the same registers for both
+ * interfaces, so this can't be made interface-specific.
+ */
+
+static DEFINE_SPINLOCK(phy_lock);

you should have a chip structure, that contains two structs (one for each interface/port)


+static u16 tsi108_read_mii(struct tsi108_prv_data *data, int reg, int *status)
+{
+ int i;
+ u16 ret;
+
+ TSI108_ETH_WRITE_PHYREG(TSI108_MAC_MII_ADDR,
+ (data->phy << TSI108_MAC_MII_ADDR_PHY) |
+ (reg << TSI108_MAC_MII_ADDR_REG));
+ TSI108_ETH_WRITE_PHYREG(TSI108_MAC_MII_CMD, 0);
+ TSI108_ETH_WRITE_PHYREG(TSI108_MAC_MII_CMD, TSI108_MAC_MII_CMD_READ);
+ for (i = 0; i < 100; i++) {
+ if (!(TSI108_ETH_READ_PHYREG(TSI108_MAC_MII_IND) &
+ (TSI108_MAC_MII_IND_NOTVALID | TSI108_MAC_MII_IND_BUSY)))
+ break;
+ udelay(10);
+ }
+
+ if (i == 100) {
+ if (status)
+ *status = -EBUSY;
+
+ ret = 0xffff;
+ } else {
+ if (status)
+ *status = 0;
+
+ ret = TSI108_ETH_READ_PHYREG(TSI108_MAC_MII_DATAIN);
+ }
+
+ return ret;
+}
+
+static void tsi108_write_mii(struct tsi108_prv_data *data, + int reg, u16 val)
+{
+ TSI108_ETH_WRITE_PHYREG(TSI108_MAC_MII_ADDR,
+ (data->phy << TSI108_MAC_MII_ADDR_PHY) |
+ (reg << TSI108_MAC_MII_ADDR_REG));
+ TSI108_ETH_WRITE_PHYREG(TSI108_MAC_MII_DATAOUT, val);
+ while (TSI108_ETH_READ_PHYREG(TSI108_MAC_MII_IND) &
+ TSI108_MAC_MII_IND_BUSY)
+ cpu_relax();

NAK, potential infinite loop if there is a hardware fault or other highly exceptional event

it should time out after a reasonable period of time.


+static inline void tsi108_write_tbi(struct tsi108_prv_data *data,
+ int reg, u16 val)
+{
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_MII_ADDR,
+ (0x1e << TSI108_MAC_MII_ADDR_PHY)
+ | (reg << TSI108_MAC_MII_ADDR_REG));
+ TSI108_ETH_WRITE_REG(TSI108_MAC_MII_DATAOUT, val);
+ while (TSI108_ETH_READ_REG(TSI108_MAC_MII_IND) &
+ TSI108_MAC_MII_IND_BUSY)
+ cpu_relax();

ditto


+static void tsi108_check_phy(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u16 sumstat;
+ u32 mac_cfg2_reg, portctrl_reg;
+ u32 fdx_flag = 0, reg_update = 0;
+
+ /* Do a dummy read, as for some reason the first read
+ * after a link becomes up returns link down, even if
+ * it's been a while since the link came up.
+ */
+
+ spin_lock(&phy_lock);
+
+ if (!data->phy_ok)
+ goto out;
+
+ tsi108_read_mii(data, PHY_STAT, NULL);
+
+ if (!(tsi108_read_mii(data, PHY_STAT, NULL) & PHY_STAT_LINKUP)) {

you should use mii.h constants rather than invent your own PHY_xxx



+ if (data->link_up == 1) {
+ netif_stop_queue(dev);
+ data->link_up = 0;
+ printk(KERN_NOTICE "%s : link is down\n", dev->name);
+ netif_carrier_off(dev);

this looks suspiciously similar to mii_check_media() in drivers/net/mii.c


+ goto out;
+ }
+
+ mac_cfg2_reg = TSI108_ETH_READ_REG(TSI108_MAC_CFG2);
+ portctrl_reg = TSI108_ETH_READ_REG(TSI108_EC_PORTCTRL);
+
+ sumstat = tsi108_read_mii(data, PHY_SUM_STAT, NULL);
+
+ switch (sumstat & PHY_SUM_STAT_SPEED_MASK) {
+ case PHY_SUM_STAT_1000T_FD:
+ fdx_flag++;
+ case PHY_SUM_STAT_1000T_HD:
+ if (data->speed != 1000) {
+ mac_cfg2_reg &= ~TSI108_MAC_CFG2_IFACE_MASK;
+ mac_cfg2_reg |= TSI108_MAC_CFG2_GIG;
+ portctrl_reg &= ~TSI108_EC_PORTCTRL_NOGIG;
+ data->speed = 1000;
+ reg_update++;
+ }
+ break;
+ case PHY_SUM_STAT_100TX_FD:
+ fdx_flag++;
+ case PHY_SUM_STAT_100TX_HD:
+ if (data->speed != 100) {
+ mac_cfg2_reg &= ~TSI108_MAC_CFG2_IFACE_MASK;
+ mac_cfg2_reg |= TSI108_MAC_CFG2_NOGIG;
+ portctrl_reg |= TSI108_EC_PORTCTRL_NOGIG;
+ data->speed = 100;
+ reg_update++;
+ }
+ break;
+
+ case PHY_SUM_STAT_10T_FD:
+ fdx_flag++;
+ case PHY_SUM_STAT_10T_HD:
+ if (data->speed != 10) {
+ mac_cfg2_reg &= ~TSI108_MAC_CFG2_IFACE_MASK;
+ mac_cfg2_reg |= TSI108_MAC_CFG2_NOGIG;
+ portctrl_reg |= TSI108_EC_PORTCTRL_NOGIG;
+ data->speed = 10;
+ reg_update++;
+ }
+ break;
+
+ default:
+ if (net_ratelimit())
+ printk(KERN_ERR "PHY reported invalid speed,"
+ KERN_ERR " summary status %x\n",
+ sumstat);
+ goto out;
+ }
+
+ if (fdx_flag || (sumstat & PHY_SUM_STAT_FULLDUPLEX)) {
+ if (data->duplex != 2) {
+ mac_cfg2_reg |= TSI108_MAC_CFG2_FULLDUPLEX;
+ portctrl_reg &= ~TSI108_EC_PORTCTRL_HALFDUPLEX;
+ data->duplex = 2;
+ reg_update++;
+ }
+ } else {
+ if (data->duplex != 1) {
+ mac_cfg2_reg &= ~TSI108_MAC_CFG2_FULLDUPLEX;
+ portctrl_reg |= TSI108_EC_PORTCTRL_HALFDUPLEX;
+ data->duplex = 1;
+ reg_update++;
+ }
+ }
+
+ if (reg_update) {
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG2, mac_cfg2_reg);
+ TSI108_ETH_WRITE_REG(TSI108_EC_PORTCTRL, portctrl_reg);
+
+ }
+
+ if (data->link_up == 0) {
+ /* The manual says it can take 3-4 usecs for the speed change
+ * to take effect.
+ */
+ udelay(5);
+
+ spin_lock(&data->txlock);
+ if (netif_queue_stopped(dev)
+ && is_valid_ether_addr(dev->dev_addr) && data->txfree)
+ netif_wake_queue(dev);
+
+ data->link_up = 1;
+ spin_unlock(&data->txlock);
+ printk("%s : link is up: %dMb %s-duplex\n",
+ dev->name, data->speed,
+ (data->duplex == 2) ? "full" : "half");
+ netif_carrier_on(dev);
+ }
+
+ out:
+ spin_unlock(&phy_lock);
+}
+
+static inline void
+tsi108_stat_carry_one(int carry, int carry_bit, int carry_shift,
+ unsigned long *upper)
+{
+ if (carry & carry_bit)
+ *upper += carry_shift;
+}
+
+static void tsi108_stat_carry(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 carry1, carry2;
+
+ spin_lock_irq(&data->misclock);
+
+ carry1 = TSI108_ETH_READ_REG(TSI108_STAT_CARRY1);
+ carry2 = TSI108_ETH_READ_REG(TSI108_STAT_CARRY2);
+
+ TSI108_ETH_WRITE_REG(TSI108_STAT_CARRY1, carry1);
+ TSI108_ETH_WRITE_REG(TSI108_STAT_CARRY2, carry2);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXBYTES,
+ TSI108_STAT_RXBYTES_CARRY, &data->stats.rx_bytes);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXPKTS,
+ TSI108_STAT_RXPKTS_CARRY,
+ &data->stats.rx_packets);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXFCS,
+ TSI108_STAT_RXFCS_CARRY, &data->rx_fcs);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXMCAST,
+ TSI108_STAT_RXMCAST_CARRY,
+ &data->stats.multicast);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXALIGN,
+ TSI108_STAT_RXALIGN_CARRY,
+ &data->stats.rx_frame_errors);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXLENGTH,
+ TSI108_STAT_RXLENGTH_CARRY,
+ &data->stats.rx_length_errors);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXRUNT,
+ TSI108_STAT_RXRUNT_CARRY, &data->rx_underruns);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXJUMBO,
+ TSI108_STAT_RXJUMBO_CARRY, &data->rx_overruns);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXFRAG,
+ TSI108_STAT_RXFRAG_CARRY, &data->rx_short_fcs);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXJABBER,
+ TSI108_STAT_RXJABBER_CARRY, &data->rx_long_fcs);
+
+ tsi108_stat_carry_one(carry1, TSI108_STAT_CARRY1_RXDROP,
+ TSI108_STAT_RXDROP_CARRY,
+ &data->stats.rx_missed_errors);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXBYTES,
+ TSI108_STAT_TXBYTES_CARRY, &data->stats.tx_bytes);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXPKTS,
+ TSI108_STAT_TXPKTS_CARRY,
+ &data->stats.tx_packets);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXEXDEF,
+ TSI108_STAT_TXEXDEF_CARRY,
+ &data->stats.tx_aborted_errors);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXEXCOL,
+ TSI108_STAT_TXEXCOL_CARRY, &data->tx_coll_abort);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXTCOL,
+ TSI108_STAT_TXTCOL_CARRY,
+ &data->stats.collisions);
+
+ tsi108_stat_carry_one(carry2, TSI108_STAT_CARRY2_TXPAUSE,
+ TSI108_STAT_TXPAUSEDROP_CARRY,
+ &data->tx_pause_drop);
+
+ spin_unlock_irq(&data->misclock);
+}
+
+/* Read a stat counter atomically with respect to carries.
+ * data->misclock must be held.
+ */
+static inline unsigned long
+tsi108_read_stat(struct tsi108_prv_data * data, int reg, int carry_bit,
+ int carry_shift, unsigned long *upper)
+{
+ int carryreg;
+ unsigned long val;
+
+ if (reg < 0xb0)
+ carryreg = TSI108_STAT_CARRY1;
+ else
+ carryreg = TSI108_STAT_CARRY2;
+
+ again:
+ val = TSI108_ETH_READ_REG(reg) | *upper;
+
+ /* Check to see if it overflowed, but the interrupt hasn't
+ * been serviced yet. If so, handle the carry here, and
+ * try again.
+ */
+
+ if (unlikely(TSI108_ETH_READ_REG(carryreg) & carry_bit)) {
+ *upper += carry_shift;
+ TSI108_ETH_WRITE_REG(carryreg, carry_bit);
+ goto again;
+ }
+
+ return val;
+}
+
+static struct net_device_stats *tsi108_get_stats(struct net_device *dev)
+{
+ unsigned long excol;
+
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ spin_lock_irq(&data->misclock);
+
+ data->tmpstats.rx_packets =
+ tsi108_read_stat(data, TSI108_STAT_RXPKTS,
+ TSI108_STAT_CARRY1_RXPKTS,
+ TSI108_STAT_RXPKTS_CARRY, &data->stats.rx_packets);
+
+ data->tmpstats.tx_packets =
+ tsi108_read_stat(data, TSI108_STAT_TXPKTS,
+ TSI108_STAT_CARRY2_TXPKTS,
+ TSI108_STAT_TXPKTS_CARRY, &data->stats.tx_packets);
+
+ data->tmpstats.rx_bytes =
+ tsi108_read_stat(data, TSI108_STAT_RXBYTES,
+ TSI108_STAT_CARRY1_RXBYTES,
+ TSI108_STAT_RXBYTES_CARRY, &data->stats.rx_bytes);
+
+ data->tmpstats.tx_bytes =
+ tsi108_read_stat(data, TSI108_STAT_TXBYTES,
+ TSI108_STAT_CARRY2_TXBYTES,
+ TSI108_STAT_TXBYTES_CARRY, &data->stats.tx_bytes);
+
+ data->tmpstats.multicast =
+ tsi108_read_stat(data, TSI108_STAT_RXMCAST,
+ TSI108_STAT_CARRY1_RXMCAST,
+ TSI108_STAT_RXMCAST_CARRY, &data->stats.multicast);
+
+ excol = tsi108_read_stat(data, TSI108_STAT_TXEXCOL,
+ TSI108_STAT_CARRY2_TXEXCOL,
+ TSI108_STAT_TXEXCOL_CARRY,
+ &data->tx_coll_abort);
+
+ data->tmpstats.collisions =
+ tsi108_read_stat(data, TSI108_STAT_TXTCOL,
+ TSI108_STAT_CARRY2_TXTCOL,
+ TSI108_STAT_TXTCOL_CARRY, &data->stats.collisions);
+
+ data->tmpstats.collisions += excol;
+
+ data->tmpstats.rx_length_errors =
+ tsi108_read_stat(data, TSI108_STAT_RXLENGTH,
+ TSI108_STAT_CARRY1_RXLENGTH,
+ TSI108_STAT_RXLENGTH_CARRY,
+ &data->stats.rx_length_errors);
+
+ data->tmpstats.rx_length_errors +=
+ tsi108_read_stat(data, TSI108_STAT_RXRUNT,
+ TSI108_STAT_CARRY1_RXRUNT,
+ TSI108_STAT_RXRUNT_CARRY, &data->rx_underruns);
+
+ data->tmpstats.rx_length_errors +=
+ tsi108_read_stat(data, TSI108_STAT_RXJUMBO,
+ TSI108_STAT_CARRY1_RXJUMBO,
+ TSI108_STAT_RXJUMBO_CARRY, &data->rx_overruns);
+
+ data->tmpstats.rx_frame_errors =
+ tsi108_read_stat(data, TSI108_STAT_RXALIGN,
+ TSI108_STAT_CARRY1_RXALIGN,
+ TSI108_STAT_RXALIGN_CARRY,
+ &data->stats.rx_frame_errors);
+
+ data->tmpstats.rx_frame_errors +=
+ tsi108_read_stat(data, TSI108_STAT_RXFCS,
+ TSI108_STAT_CARRY1_RXFCS, TSI108_STAT_RXFCS_CARRY,
+ &data->rx_fcs);
+
+ data->tmpstats.rx_frame_errors +=
+ tsi108_read_stat(data, TSI108_STAT_RXFRAG,
+ TSI108_STAT_CARRY1_RXFRAG,
+ TSI108_STAT_RXFRAG_CARRY, &data->rx_short_fcs);
+
+ data->tmpstats.rx_missed_errors =
+ tsi108_read_stat(data, TSI108_STAT_RXDROP,
+ TSI108_STAT_CARRY1_RXDROP,
+ TSI108_STAT_RXDROP_CARRY,
+ &data->stats.rx_missed_errors);
+
+ /* These three are maintained by software. */
+ data->tmpstats.rx_fifo_errors = data->stats.rx_fifo_errors;
+ data->tmpstats.rx_crc_errors = data->stats.rx_crc_errors;
+
+ data->tmpstats.tx_aborted_errors =
+ tsi108_read_stat(data, TSI108_STAT_TXEXDEF,
+ TSI108_STAT_CARRY2_TXEXDEF,
+ TSI108_STAT_TXEXDEF_CARRY,
+ &data->stats.tx_aborted_errors);
+
+ data->tmpstats.tx_aborted_errors +=
+ tsi108_read_stat(data, TSI108_STAT_TXPAUSEDROP,
+ TSI108_STAT_CARRY2_TXPAUSE,
+ TSI108_STAT_TXPAUSEDROP_CARRY,
+ &data->tx_pause_drop);
+
+ data->tmpstats.tx_aborted_errors += excol;
+
+ data->tmpstats.tx_errors = data->tmpstats.tx_aborted_errors;
+ data->tmpstats.rx_errors = data->tmpstats.rx_length_errors +
+ data->tmpstats.rx_crc_errors +
+ data->tmpstats.rx_frame_errors +
+ data->tmpstats.rx_fifo_errors + data->tmpstats.rx_missed_errors;
+
+ spin_unlock_irq(&data->misclock);
+ return &data->tmpstats;
+}
+
+static void tsi108_restart_rx(struct tsi108_prv_data * data, struct net_device *dev)
+{
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXQ_PTRHIGH,
+ TSI108_EC_RXQ_PTRHIGH_VALID);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCTRL, TSI108_EC_RXCTRL_GO
+ | TSI108_EC_RXCTRL_QUEUE0);
+}
+
+static void tsi108_restart_tx(struct tsi108_prv_data * data)
+{
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXQ_PTRHIGH,
+ TSI108_EC_TXQ_PTRHIGH_VALID);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXCTRL, TSI108_EC_TXCTRL_IDLEINT |
+ TSI108_EC_TXCTRL_GO | TSI108_EC_TXCTRL_QUEUE0);
+}
+
+/* txlock must be held by caller, with IRQs disabled, and
+ * with permission to re-enable them when the lock is dropped.
+ */
+static void tsi108_check_for_completed_tx(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ int tx;
+ struct sk_buff *skb;
+ int release = 0;
+
+ while (!data->txfree || data->txhead != data->txtail) {
+ tx = data->txtail;
+
+ if (data->txring[tx].misc & TSI108_TX_OWN)
+ break;
+
+ skb = data->txskbs[tx];
+
+ if (!(data->txring[tx].misc & TSI108_TX_OK))
+ printk("%s: bad tx packet, misc %x\n",
+ dev->name, data->txring[tx].misc);
+
+ data->txtail = (data->txtail + 1) % TSI108_TXRING_LEN;
+ data->txfree++;
+
+ if (data->txring[tx].misc & TSI108_TX_EOF) {
+ dev_kfree_skb_any(skb);
+ release++;
+ }
+ }
+
+ if (release) {
+ if (netif_queue_stopped(dev)
+ && is_valid_ether_addr(dev->dev_addr) && data->link_up)
+ netif_wake_queue(dev);

this is a screwup of mine, copied to many other net drivers.

you don't need to check netif_queue_stopped() before calling netif_wake_queue()


+static int tsi108_send_packet(struct sk_buff * skb, struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ int frags = skb_shinfo(skb)->nr_frags + 1;
+ int i;
+
+ if (!data->phy_ok && net_ratelimit())
+ printk(KERN_ERR "%s: Transmit while PHY is down!\n", dev->name);
+
+ if (!data->link_up) {
+ printk(KERN_ERR "%s: Transmit while link is down!\n",
+ dev->name);
+ netif_stop_queue(dev);
+ return 1;
+ }

use proper NETDEV_TX_xxx return codes


+ if (data->txfree < MAX_SKB_FRAGS + 1) {
+ netif_stop_queue(dev);
+
+ if (net_ratelimit())
+ printk(KERN_ERR "%s: Transmit with full tx ring!\n",
+ dev->name);
+ return 1;
+ }
+
+ if (data->txfree - frags < MAX_SKB_FRAGS + 1) {
+ netif_stop_queue(dev);
+ }
+
+ spin_lock_irq(&data->txlock);
+
+ for (i = 0; i < frags; i++) {
+ int misc = 0;
+ int tx = data->txhead;
+
+ /* This is done to mark every TSI108_TX_INT_FREQ tx buffers with
+ * the interrupt bit. TX descriptor-complete interrupts are
+ * enabled when the queue fills up, and masked when there is
+ * still free space. This way, when saturating the outbound
+ * link, the tx interrupts are kept to a reasonable level. + * When the queue is not full, reclamation of skbs still occurs
+ * as new packets are transmitted, or on a queue-empty
+ * interrupt.
+ */
+
+ if ((tx % TSI108_TX_INT_FREQ == 0) &&
+ ((TSI108_TXRING_LEN - data->txfree) >= TSI108_TX_INT_FREQ))
+ misc = TSI108_TX_INT;
+
+ data->txskbs[tx] = skb;
+
+ if (i == 0) {
+ data->txring[tx].buf0 = virt_to_phys(skb->data);
+ data->txring[tx].len = skb->len - skb->data_len;
+ misc |= TSI108_TX_SOF;
+ } else {
+ skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
+
+ data->txring[tx].buf0 =
+ page_to_phys(frag->page) + frag->page_offset;
+ data->txring[tx].len = frag->size;

use dma mapping, not virt_to_phys

(of course, on some platforms, dma mapping devolves into virt_to_phy under the hood)


+ if (i == frags - 1)
+ misc |= TSI108_TX_EOF;
+
+#ifdef TSI108_PRINT_TX_FRAME
+ {
+ int i;
+ printk("%s: Tx Frame contents (%d)\n", dev->name,
+ skb->len);
+ for (i = 0; i < skb->len; i++)
+ printk(" %2.2x", skb->data[i]);
+ printk(".\n");
+ }
+#endif /* TSI108_PRINT_TX_FRAME */


rather than ifdef, test netif_msg_pktdata() at runtime


+ data->txring[tx].misc = misc | TSI108_TX_OWN;
+
+ data->txhead = (data->txhead + 1) % TSI108_TXRING_LEN;
+ data->txfree--;
+ }
+
+ tsi108_check_for_completed_tx(dev);
+
+ /* This must be done after the check for completed tx descriptors,
+ * so that the tail pointer is correct.
+ */
+
+ if (!(TSI108_ETH_READ_REG(TSI108_EC_TXSTAT) & TSI108_EC_TXSTAT_QUEUE0))
+ tsi108_restart_tx(data);
+
+ spin_unlock_irq(&data->txlock);
+ return 0;
+}
+
+static int tsi108_check_for_completed_rx(struct net_device *dev, int budget)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ int done = 0;
+
+ while (data->rxfree && done != budget) {
+ int rx = data->rxtail;
+ struct sk_buff *skb;
+
+ if (data->rxring[rx].misc & TSI108_RX_OWN)
+ break;
+
+ skb = data->rxskbs[rx];
+ data->rxtail = (data->rxtail + 1) % TSI108_RXRING_LEN;
+ data->rxfree--;
+ done++;
+
+ if (data->rxring[rx].misc & TSI108_RX_BAD) {
+ spin_lock_irq(&data->misclock);
+
+ if (data->rxring[rx].misc & TSI108_RX_CRC)
+ data->stats.rx_crc_errors++;
+ if (data->rxring[rx].misc & TSI108_RX_OVER)
+ data->stats.rx_fifo_errors++;
+
+ spin_unlock_irq(&data->misclock);
+
+ dev_kfree_skb_any(skb);
+ continue;
+ }
+#ifdef TSI108_PRINT_RX_FRAME
+ {
+ int i;
+ printk("%s: Rx Frame contents (%d)\n",
+ dev->name, data->rxring[rx].len);
+ for (i = 0; i < data->rxring[rx].len; i++)
+ printk(" %2.2x", skb->data[i]);
+ printk(".\n");
+ }
+#endif /* TSI108_PRINT_RX_FRAME */

ditto


+ skb->dev = dev;
+ skb_put(skb, data->rxring[rx].len);
+ skb->protocol = eth_type_trans(skb, dev);
+ netif_receive_skb(skb);
+ dev->last_rx = jiffies;
+ }
+
+ return done;
+}
+
+static int tsi108_refill_rx(struct net_device *dev, int budget)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ int done = 0;
+
+ while (data->rxfree != TSI108_RXRING_LEN && done != budget) {
+ int rx = data->rxhead;
+ struct sk_buff *skb;
+
+ data->rxskbs[rx] = skb = dev_alloc_skb(TSI108_RXBUF_SIZE + 2);
+ if (!skb)
+ break;
+
+ skb_reserve(skb, 2); /* Align the data on a 4-byte boundary. */
+
+ data->rxring[rx].buf0 = virt_to_phys(skb->data);

use dma mapping


+ /* Sometimes the hardware sets blen to zero after packet
+ * reception, even though the manual says that it's only ever
+ * modified by the driver.
+ */
+
+ data->rxring[rx].blen = 1536;
+ data->rxring[rx].misc = TSI108_RX_OWN | TSI108_RX_INT;
+
+ data->rxhead = (data->rxhead + 1) % TSI108_RXRING_LEN;
+ data->rxfree++;
+ done++;
+ }
+
+ if (done != 0 && !(TSI108_ETH_READ_REG(TSI108_EC_RXSTAT) &
+ TSI108_EC_RXSTAT_QUEUE0))
+ tsi108_restart_rx(data, dev);
+
+ return done;
+}
+
+static int tsi108_poll(struct net_device *dev, int *budget)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 estat = TSI108_ETH_READ_REG(TSI108_EC_RXESTAT);
+ u32 intstat = TSI108_ETH_READ_REG(TSI108_EC_INTSTAT);
+ int total_budget = min(*budget, dev->quota);
+ int num_received = 0, num_filled = 0, budget_used;
+
+ intstat &= TSI108_INT_RXQUEUE0 | TSI108_INT_RXTHRESH |
+ TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR | TSI108_INT_RXWAIT;
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXESTAT, estat);
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTSTAT, intstat);
+
+ if (data->rxpending || (estat & TSI108_EC_RXESTAT_Q0_DESCINT))
+ num_received = tsi108_check_for_completed_rx(dev, total_budget);
+
+ /* This should normally fill no more slots than the number of
+ * packets received in tsi108_check_for_completed_rx(). The exception
+ * is when we previously ran out of memory for RX SKBs. In that
+ * case, it's helpful to obey the budget, not only so that the
+ * CPU isn't hogged, but so that memory (which may still be low)
+ * is not hogged by one device.
+ *
+ * A work unit is considered to be two SKBs to allow us to catch
+ * up when the ring has shrunk due to out-of-memory but we're
+ * still removing the full budget's worth of packets each time.
+ */
+
+ if (data->rxfree < TSI108_RXRING_LEN)
+ num_filled = tsi108_refill_rx(dev, total_budget * 2);
+
+ if (intstat & TSI108_INT_RXERROR) {
+ u32 err = TSI108_ETH_READ_REG(TSI108_EC_RXERR);
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXERR, err);
+
+ if (err) {
+ if (net_ratelimit())
+ printk(KERN_DEBUG "%s: RX error %x\n",
+ dev->name, err);
+
+ if (!(TSI108_ETH_READ_REG(TSI108_EC_RXSTAT) &
+ TSI108_EC_RXSTAT_QUEUE0))
+ tsi108_restart_rx(data, dev);
+ }
+ }
+
+ if (intstat & TSI108_INT_RXOVERRUN) {
+ spin_lock_irq(&data->misclock);
+ data->stats.rx_fifo_errors++;
+ spin_unlock_irq(&data->misclock);
+ }
+
+ budget_used = max(num_received, num_filled / 2);
+
+ *budget -= budget_used;
+ dev->quota -= budget_used;
+
+ if (budget_used != total_budget) {
+ data->rxpending = 0;
+ netif_rx_complete(dev);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK,
+ TSI108_ETH_READ_REG(TSI108_EC_INTMASK)
+ & ~(TSI108_INT_RXQUEUE0
+ | TSI108_INT_RXTHRESH |
+ TSI108_INT_RXOVERRUN |
+ TSI108_INT_RXERROR |
+ TSI108_INT_RXWAIT));
+
+ /* IRQs are level-triggered, so no need to re-check */
+ return 0;
+ } else {
+ data->rxpending = 1;
+ }
+
+ return 1;
+}
+
+static void tsi108_rx_int(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ /* A race could cause dev to already be scheduled, so it's not an
+ * error if that happens (and interrupts shouldn't be re-masked,
+ * because that can cause harmful races, if poll has already
+ * unmasked them but not cleared LINK_STATE_SCHED). + *
+ * This can happen if this code races with tsi108_poll(), which masks
+ * the interrupts after tsi108_irq_one() read the mask, but before
+ * netif_rx_schedule is called. It could also happen due to calls
+ * from tsi108_check_rxring().
+ */
+
+ if (netif_rx_schedule_prep(dev)) {
+ /* Mask, rather than ack, the receive interrupts. The ack
+ * will happen in tsi108_poll().
+ */
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK,
+ TSI108_ETH_READ_REG(TSI108_EC_INTMASK) |
+ TSI108_INT_RXQUEUE0
+ | TSI108_INT_RXTHRESH |
+ TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR |
+ TSI108_INT_RXWAIT);
+ __netif_rx_schedule(dev);
+ } else {
+ if (!netif_running(dev)) {
+ /* This can happen if an interrupt occurs while the
+ * interface is being brought down, as the START
+ * bit is cleared before the stop function is called.
+ *
+ * In this case, the interrupts must be masked, or
+ * they will continue indefinitely.
+ *
+ * There's a race here if the interface is brought down
+ * and then up in rapid succession, as the device could
+ * be made running after the above check and before
+ * the masking below. This will only happen if the IRQ
+ * thread has a lower priority than the task brining
+ * up the interface. Fixing this race would likely
+ * require changes in generic code.
+ */
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK,
+ TSI108_ETH_READ_REG
+ (TSI108_EC_INTMASK) |
+ TSI108_INT_RXQUEUE0 |
+ TSI108_INT_RXTHRESH |
+ TSI108_INT_RXOVERRUN |
+ TSI108_INT_RXERROR |
+ TSI108_INT_RXWAIT);
+ }
+ }
+}
+
+/* If the RX ring has run out of memory, try periodically
+ * to allocate some more, as otherwise poll would never
+ * get called (apart from the initial end-of-queue condition).
+ *
+ * This is called once per second (by default) from the thread.
+ */
+
+static void tsi108_check_rxring(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ /* A poll is scheduled, as opposed to caling tsi108_refill_rx + * directly, so as to keep the receive path single-threaded
+ * (and thus not needing a lock).
+ */
+
+ if (netif_running(dev) && data->rxfree < TSI108_RXRING_LEN / 4)
+ tsi108_rx_int(dev);
+}
+
+static void tsi108_tx_int(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 estat = TSI108_ETH_READ_REG(TSI108_EC_TXESTAT);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXESTAT, estat);
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTSTAT, TSI108_INT_TXQUEUE0 |
+ TSI108_INT_TXIDLE | TSI108_INT_TXERROR);
+ if (estat & TSI108_EC_TXESTAT_Q0_ERR) {
+ u32 err = TSI108_ETH_READ_REG(TSI108_EC_TXERR);
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXERR, err);
+
+ if (err && net_ratelimit())
+ printk(KERN_ERR "%s: TX error %x\n", dev->name, err);
+ }
+
+ if (estat & (TSI108_EC_TXESTAT_Q0_DESCINT | TSI108_EC_TXESTAT_Q0_EOQ)) {
+ spin_lock(&data->txlock);
+ tsi108_check_for_completed_tx(dev);
+ spin_unlock(&data->txlock);

function names like tsi108_check_for_completed_tx() are just way too long. tsi_complete_tx() would be better.


+static irqreturn_t tsi108_irq_one(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 stat = TSI108_ETH_READ_REG(TSI108_EC_INTSTAT);
+
+ if (!(stat & TSI108_INT_ANY))
+ return IRQ_NONE; /* Not our interrupt */
+
+ stat &= ~TSI108_ETH_READ_REG(TSI108_EC_INTMASK);
+
+ if (stat & (TSI108_INT_TXQUEUE0 | TSI108_INT_TXIDLE |
+ TSI108_INT_TXERROR))
+ tsi108_tx_int(dev);
+ if (stat & (TSI108_INT_RXQUEUE0 | TSI108_INT_RXTHRESH |
+ TSI108_INT_RXWAIT | TSI108_INT_RXOVERRUN |
+ TSI108_INT_RXERROR))
+ tsi108_rx_int(dev);
+
+ if (stat & TSI108_INT_SFN) {
+ if (net_ratelimit())
+ printk(KERN_DEBUG "%s: SFN error\n", dev->name);
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTSTAT, TSI108_INT_SFN);
+ }
+
+ if (stat & TSI108_INT_STATCARRY) {
+ tsi108_stat_carry(dev);
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTSTAT, TSI108_INT_STATCARRY);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t tsi108_irq(int irq, void *dev_id, struct pt_regs *regs)
+{
+
+ return tsi108_irq_one(dev_id);

why create a separate function just for this one line?


+static void tsi108_stop_ethernet(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ /* Disable all TX and RX queues ... */
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXCTRL, 0);
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCTRL, 0);
+
+ /* ...and wait for them to become idle */
+ while (TSI108_ETH_READ_REG(TSI108_EC_TXSTAT) &
+ TSI108_EC_TXSTAT_ACTIVE) + cpu_relax();
+ while (TSI108_ETH_READ_REG(TSI108_EC_RXSTAT) &
+ TSI108_EC_RXSTAT_ACTIVE)
+ cpu_relax();

infinite loops with no timeout


+static void tsi108_reset_ether(struct tsi108_prv_data * data)
+{
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG1, TSI108_MAC_CFG1_SOFTRST);
+ udelay(100);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG1, 0);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_PORTCTRL, TSI108_EC_PORTCTRL_STATRST);
+ udelay(100);
+ TSI108_ETH_WRITE_REG(TSI108_EC_PORTCTRL,
+ TSI108_ETH_READ_REG(TSI108_EC_PORTCTRL) &
+ ~TSI108_EC_PORTCTRL_STATRST);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXCFG, TSI108_EC_TXCFG_RST);
+ udelay(100);
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXCFG,
+ TSI108_ETH_READ_REG(TSI108_EC_TXCFG) &
+ ~TSI108_EC_TXCFG_RST);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCFG, TSI108_EC_RXCFG_RST);
+ udelay(100);
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCFG,
+ TSI108_ETH_READ_REG(TSI108_EC_RXCFG) &
+ ~TSI108_EC_RXCFG_RST);
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_MII_MGMT_CFG,
+ TSI108_ETH_READ_REG(TSI108_MAC_MII_MGMT_CFG) |
+ TSI108_MAC_MII_MGMT_RST);
+ udelay(100);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_MII_MGMT_CFG,
+ TSI108_ETH_READ_REG(TSI108_MAC_MII_MGMT_CFG) &
+ ~(TSI108_MAC_MII_MGMT_RST |
+ TSI108_MAC_MII_MGMT_CLK));
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_MII_MGMT_CFG,
+ TSI108_ETH_READ_REG(TSI108_MAC_MII_MGMT_CFG) |
+ TSI108_MAC_MII_MGMT_CLK);

bus write posting bugs?


+static int tsi108_get_mac(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 word1 = TSI108_ETH_READ_REG(TSI108_MAC_ADDR1);
+ u32 word2 = TSI108_ETH_READ_REG(TSI108_MAC_ADDR2);
+
+ /* Note that the octets are reversed from what the manual says,
+ * producing an even weirder ordering...
+ */
+ if (word2 == 0 && word1 == 0) {
+ dev->dev_addr[0] = 0x00;
+ dev->dev_addr[1] = 0x06;
+ dev->dev_addr[2] = 0xd2;
+ dev->dev_addr[3] = 0x00;
+ dev->dev_addr[4] = 0x00;
+ if (0x8 == data->phy)
+ dev->dev_addr[5] = 0x01;
+ else
+ dev->dev_addr[5] = 0x02;
+
+ word2 = (dev->dev_addr[0] << 16) | (dev->dev_addr[1] << 24);
+
+ word1 = (dev->dev_addr[2] << 0) | (dev->dev_addr[3] << 8) |
+ (dev->dev_addr[4] << 16) | (dev->dev_addr[5] << 24);
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_ADDR1, word1);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_ADDR2, word2);
+ } else {
+ dev->dev_addr[0] = (word2 >> 16) & 0xff;
+ dev->dev_addr[1] = (word2 >> 24) & 0xff;
+ dev->dev_addr[2] = (word1 >> 0) & 0xff;
+ dev->dev_addr[3] = (word1 >> 8) & 0xff;
+ dev->dev_addr[4] = (word1 >> 16) & 0xff;
+ dev->dev_addr[5] = (word1 >> 24) & 0xff;
+ }
+
+ if (!is_valid_ether_addr(dev->dev_addr)) {
+ printk("KERN_ERR: word1: %08x, word2: %08x\n", word1, word2);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static int tsi108_set_mac(struct net_device *dev, void *addr)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 word1, word2;
+ int i;
+
+ if (!is_valid_ether_addr(addr))
+ return -EINVAL;
+
+ for (i = 0; i < 6; i++)
+ /* +2 is for the offset of the HW addr type */
+ dev->dev_addr[i] = ((unsigned char *)addr)[i + 2];
+
+ word2 = (dev->dev_addr[0] << 16) | (dev->dev_addr[1] << 24);
+
+ word1 = (dev->dev_addr[2] << 0) | (dev->dev_addr[3] << 8) |
+ (dev->dev_addr[4] << 16) | (dev->dev_addr[5] << 24);
+
+ spin_lock_irq(&data->misclock);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_ADDR1, word1);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_ADDR2, word2);
+ spin_lock(&data->txlock);
+
+ if (netif_queue_stopped(dev) && data->txfree && data->link_up)
+ netif_wake_queue(dev);

no need to test netif_queue_stopped()


+ spin_unlock(&data->txlock);
+ spin_unlock_irq(&data->misclock);
+ return 0;
+}
+
+/* Protected by dev->xmit_lock. */
+static void tsi108_set_rx_mode(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 rxcfg = TSI108_ETH_READ_REG(TSI108_EC_RXCFG);
+
+ if (dev->flags & IFF_PROMISC) {
+ rxcfg &= ~(TSI108_EC_RXCFG_UC_HASH | TSI108_EC_RXCFG_MC_HASH);
+ rxcfg |= TSI108_EC_RXCFG_UFE | TSI108_EC_RXCFG_MFE;
+ goto out;
+ }
+
+ rxcfg &= ~(TSI108_EC_RXCFG_UFE | TSI108_EC_RXCFG_MFE);
+
+ if (dev->mc_count) {

need to test for ALLMULTI


+ int i;
+ struct dev_mc_list *mc = dev->mc_list;
+ rxcfg |= TSI108_EC_RXCFG_MFE | TSI108_EC_RXCFG_MC_HASH;
+
+ memset(data->mc_hash, 0, sizeof(data->mc_hash));
+
+ while (mc) {
+ u32 hash, crc;
+
+ if (mc->dmi_addrlen == 6) {
+ crc = ether_crc(6, mc->dmi_addr);
+ hash = crc >> 23;
+
+ __set_bit(hash, &data->mc_hash[0]);
+ } else {
+ printk(KERN_ERR
+ "%s: got multicast address of length %d "
+ "instead of 6.\n", dev->name,
+ mc->dmi_addrlen);
+ }
+
+ mc = mc->next;
+ }
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_HASHADDR,
+ TSI108_EC_HASHADDR_AUTOINC |
+ TSI108_EC_HASHADDR_MCAST);
+
+ for (i = 0; i < 16; i++) {
+ /* The manual says that the hardware may drop
+ * back-to-back writes to the data register.
+ */
+ udelay(1);
+ TSI108_ETH_WRITE_REG(TSI108_EC_HASHDATA,
+ data->mc_hash[i]);
+ }
+ }
+
+ out:
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCFG, rxcfg);
+}
+
+static void tsi108_init_phy(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ u32 i = 0;
+ u16 phyVal = 0;
+
+ spin_lock_irq(&phy_lock);
+
+ tsi108_write_mii(data, PHY_CTRL, PHY_CTRL_RESET);
+ while (tsi108_read_mii(data, PHY_CTRL, NULL) & PHY_CTRL_RESET)
+ cpu_relax();

infinite loop with no timeout

+#if (TSI108_PHY_TYPE == PHY_BCM54XX) /* Broadcom BCM54xx PHY */
+ tsi108_write_mii(data, 0x09, 0x0300);
+ tsi108_write_mii(data, 0x10, 0x1020);
+ tsi108_write_mii(data, 0x1c, 0x8c00);
+#endif
+
+ tsi108_write_mii(data,
+ PHY_CTRL,
+ PHY_CTRL_AUTONEG_EN | PHY_CTRL_AUTONEG_START);
+ while (tsi108_read_mii(data, PHY_CTRL, NULL) & PHY_CTRL_AUTONEG_START)
+ cpu_relax();

ditto


+ /* Set G/MII mode and receive clock select in TBI control #2. The
+ * second port won't work if this isn't done, even though we don't
+ * use TBI mode.
+ */
+
+ tsi108_write_tbi(data, 0x11, 0x30);
+
+ /* FIXME: It seems to take more than 2 back-to-back reads to the
+ * PHY_STAT register before the link up status bit is set.
+ */
+
+ data->link_up = 1;
+
+ while (!((phyVal = tsi108_read_mii(data, PHY_STAT, NULL)) &
+ PHY_STAT_LINKUP)) {
+ if (i++ > (MII_READ_DELAY / 10)) {
+ data->link_up = 0;
+ break;
+ }
+ spin_unlock_irq(&phy_lock);
+ msleep(10);
+ spin_lock_irq(&phy_lock);
+ }
+
+ printk(KERN_DEBUG "PHY_STAT reg contains %08x\n", phyVal);
+ data->phy_ok = 1;
+ spin_unlock_irq(&phy_lock);
+}
+
+static void tsi108_kill_phy(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ spin_lock_irq(&phy_lock);
+ tsi108_write_mii(data, PHY_CTRL, PHY_CTRL_POWERDOWN);
+ data->phy_ok = 0;
+ spin_unlock_irq(&phy_lock);
+}
+
+static int tsi108_open(struct net_device *dev)
+{
+ int i;
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ unsigned int rxring_size = TSI108_RXRING_LEN * sizeof(rx_desc);
+ unsigned int txring_size = TSI108_TXRING_LEN * sizeof(tx_desc);
+
+ printk(KERN_DEBUG "Inside tsi108_open()!\n");

kill debugging printk, or use netif_msg_xxx properly


+ i = request_irq(data->irq_num, tsi108_irq, 0, dev->name, dev);
+ if (i != 0) {
+ printk(KERN_ERR "tsi108_eth%d: Could not allocate IRQ%d.\n",
+ data->id, data->irq_num);
+ return i;
+ } else {
+ dev->irq = data->irq_num;
+ printk(KERN_NOTICE
+ "tsi108_open : Port %d Assigned IRQ %d to %s\n",
+ data->id, dev->irq, dev->name);
+ }
+
+ data->rxring = pci_alloc_consistent(NULL, rxring_size, &data->rxdma);
+
+ if (!data->rxring) {
+ printk(KERN_DEBUG
+ "TSI108_ETH: failed to allocate memory for rxring!\n");
+ return -ENOMEM;
+ } else {
+ memset(data->rxring, 0, rxring_size);
+ }
+
+ data->txring = pci_alloc_consistent(NULL, txring_size, &data->txdma);

since this is a platform driver, you should be using dma_alloc_coherent(), passing in the struct device*


+ if (!data->txring) {
+ printk(KERN_DEBUG
+ "TSI108_ETH: failed to allocate memory for txring!\n");
+ pci_free_consistent(0, rxring_size, data->rxring, data->rxdma);

ditto


+ return -ENOMEM;
+ } else {
+ memset(data->txring, 0, txring_size);
+ }
+
+ for (i = 0; i < TSI108_RXRING_LEN; i++) {
+ data->rxring[i].next0 = data->rxdma + (i + 1) * sizeof(rx_desc);
+ data->rxring[i].blen = TSI108_RXBUF_SIZE;
+ data->rxring[i].vlan = 0;
+ }
+
+ data->rxring[TSI108_RXRING_LEN - 1].next0 = data->rxdma;
+
+ data->rxtail = 0;
+ data->rxhead = 0;
+
+ for (i = 0; i < TSI108_RXRING_LEN; i++) {
+ struct sk_buff *skb = dev_alloc_skb(TSI108_RXBUF_SIZE + NET_IP_ALIGN);
+
+ if (!skb) {
+ /* Bah. No memory for now, but maybe we'll get
+ * some more later.
+ * For now, we'll live with the smaller ring.
+ */
+ printk(KERN_WARNING
+ "%s: Could only allocate %d receive skb(s).\n",
+ dev->name, i);
+ data->rxhead = i;
+ break;
+ }
+
+ data->rxskbs[i] = skb;
+ /* Align the payload on a 4-byte boundary */
+ skb_reserve(skb, 2);
+ data->rxskbs[i] = skb;
+ data->rxring[i].buf0 = virt_to_phys(data->rxskbs[i]->data);

use dma mapping


+ data->rxring[i].misc = TSI108_RX_OWN | TSI108_RX_INT;
+ }
+
+ data->rxfree = i;
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXQ_PTRLOW, data->rxdma);
+
+ for (i = 0; i < TSI108_TXRING_LEN; i++) {
+ data->txring[i].next0 = data->txdma + (i + 1) * sizeof(tx_desc);
+ data->txring[i].misc = 0;
+ }
+
+ data->txring[TSI108_TXRING_LEN - 1].next0 = data->txdma;
+ data->txtail = 0;
+ data->txhead = 0;
+ data->txfree = TSI108_TXRING_LEN;
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXQ_PTRLOW, data->txdma);
+ tsi108_init_phy(dev);
+
+ setup_timer(&data->timer, tsi108_timed_checker, (unsigned long)dev);
+ mod_timer(&data->timer, jiffies + 1);
+
+ tsi108_restart_rx(data, dev);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTSTAT, ~0);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK,
+ ~(TSI108_INT_TXQUEUE0 | TSI108_INT_RXERROR |
+ TSI108_INT_RXTHRESH | TSI108_INT_RXQUEUE0 |
+ TSI108_INT_RXOVERRUN | TSI108_INT_RXWAIT |
+ TSI108_INT_SFN | TSI108_INT_STATCARRY));
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG1,
+ TSI108_MAC_CFG1_RXEN | TSI108_MAC_CFG1_TXEN);
+ netif_start_queue(dev);
+ return 0;
+}
+
+static int tsi108_close(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ netif_stop_queue(dev);
+
+ del_timer_sync(&data->timer);
+
+ printk(KERN_DEBUG "Inside tsi108_ifdown!\n");

kill debugging printk, or use netif_msg_xxx properly


+ tsi108_stop_ethernet(dev);
+ tsi108_kill_phy(dev);
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK, ~0);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG1, 0);
+
+ /* Check for any pending TX packets, and drop them. */
+
+ while (!data->txfree || data->txhead != data->txtail) {
+ int tx = data->txtail;
+ struct sk_buff *skb;
+ skb = data->txskbs[tx];
+ data->txtail = (data->txtail + 1) % TSI108_TXRING_LEN;
+ data->txfree++;
+ dev_kfree_skb(skb);
+ }
+
+ synchronize_irq(data->irq_num);
+ free_irq(data->irq_num, dev);
+
+ /* Discard the RX ring. */
+
+ while (data->rxfree) {
+ int rx = data->rxtail;
+ struct sk_buff *skb;
+
+ skb = data->rxskbs[rx];
+ data->rxtail = (data->rxtail + 1) % TSI108_RXRING_LEN;
+ data->rxfree--;
+ dev_kfree_skb(skb);
+ }
+
+ pci_free_consistent(0,
+ TSI108_RXRING_LEN * sizeof(rx_desc),
+ data->rxring, data->rxdma);
+ pci_free_consistent(0,
+ TSI108_TXRING_LEN * sizeof(tx_desc),
+ data->txring, data->txdma);

dma_free_coherent


+ return 0;
+}
+
+static void tsi108_init_mac(struct net_device *dev)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG2, TSI108_MAC_CFG2_DFLT_PREAMBLE |
+ TSI108_MAC_CFG2_PADCRC);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXTHRESH,
+ (192 << TSI108_EC_TXTHRESH_STARTFILL) |
+ (192 << TSI108_EC_TXTHRESH_STOPFILL));
+
+ TSI108_ETH_WRITE_REG(TSI108_STAT_CARRYMASK1,
+ ~(TSI108_STAT_CARRY1_RXBYTES |
+ TSI108_STAT_CARRY1_RXPKTS |
+ TSI108_STAT_CARRY1_RXFCS |
+ TSI108_STAT_CARRY1_RXMCAST |
+ TSI108_STAT_CARRY1_RXALIGN |
+ TSI108_STAT_CARRY1_RXLENGTH |
+ TSI108_STAT_CARRY1_RXRUNT |
+ TSI108_STAT_CARRY1_RXJUMBO |
+ TSI108_STAT_CARRY1_RXFRAG |
+ TSI108_STAT_CARRY1_RXJABBER |
+ TSI108_STAT_CARRY1_RXDROP));
+
+ TSI108_ETH_WRITE_REG(TSI108_STAT_CARRYMASK2,
+ ~(TSI108_STAT_CARRY2_TXBYTES |
+ TSI108_STAT_CARRY2_TXPKTS |
+ TSI108_STAT_CARRY2_TXEXDEF |
+ TSI108_STAT_CARRY2_TXEXCOL |
+ TSI108_STAT_CARRY2_TXTCOL |
+ TSI108_STAT_CARRY2_TXPAUSE));
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_PORTCTRL, TSI108_EC_PORTCTRL_STATEN);
+ TSI108_ETH_WRITE_REG(TSI108_MAC_CFG1, 0);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXCFG,
+ TSI108_EC_RXCFG_SE | TSI108_EC_RXCFG_BFE);
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXQ_CFG, TSI108_EC_TXQ_CFG_DESC_INT |
+ TSI108_EC_TXQ_CFG_EOQ_OWN_INT |
+ TSI108_EC_TXQ_CFG_WSWP | (TSI108_PBM_PORT <<
+ TSI108_EC_TXQ_CFG_SFNPORT));
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXQ_CFG, TSI108_EC_RXQ_CFG_DESC_INT |
+ TSI108_EC_RXQ_CFG_EOQ_OWN_INT |
+ TSI108_EC_RXQ_CFG_WSWP | (TSI108_PBM_PORT <<
+ TSI108_EC_RXQ_CFG_SFNPORT));
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_TXQ_BUFCFG,
+ TSI108_EC_TXQ_BUFCFG_BURST256 |
+ TSI108_EC_TXQ_BUFCFG_BSWP | (TSI108_PBM_PORT <<
+ TSI108_EC_TXQ_BUFCFG_SFNPORT));
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_RXQ_BUFCFG,
+ TSI108_EC_RXQ_BUFCFG_BURST256 |
+ TSI108_EC_RXQ_BUFCFG_BSWP | (TSI108_PBM_PORT <<
+ TSI108_EC_RXQ_BUFCFG_SFNPORT));
+
+ TSI108_ETH_WRITE_REG(TSI108_EC_INTMASK, ~0);
+}
+
+static int tsi108_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
+{
+ struct tsi108_prv_data *data = netdev_priv(dev);
+ struct mii_ioctl_data *mii_data =
+ (struct mii_ioctl_data *)&rq->ifr_data;
+ int ret;
+
+ switch (cmd) {
+ case SIOCGMIIPHY:
+ mii_data->phy_id = data->phy;
+ ret = 0;
+ break;
+
+ case SIOCGMIIREG:
+ spin_lock_irq(&phy_lock);
+ mii_data->val_out =
+ tsi108_read_mii(data, mii_data->reg_num, &ret);
+ spin_unlock_irq(&phy_lock);
+ break;

use generic_mii_ioctl from drivers/net/mii.c


+ default:
+ ret = -EOPNOTSUPP;
+ }
+
+ return ret;
+}
+
+static int
+tsi108_init_one(struct platform_device *pdev)
+{
+ struct net_device *dev = NULL;
+ struct tsi108_prv_data *data = NULL;
+ hw_info *einfo;
+ int ret;
+
+ einfo = pdev->dev.platform_data;
+
+ if (NULL == einfo) {
+ printk(KERN_ERR "tsi-eth %d: Missing additional data!\n",
+ pdev->id);
+ return -ENODEV;
+ }
+
+ /* Create an ethernet device instance */
+
+ dev = alloc_etherdev(sizeof(struct tsi108_prv_data));
+ if (!dev) {
+ printk("tsi108_eth: Could not allocate a device structure\n");
+ return -ENOMEM;
+ }
+
+ printk("tsi108_eth%d: probe...\n", pdev->id);
+ data = netdev_priv(dev);
+
+ pr_debug("tsi108_eth%d:regs:phyresgs:phy:irq_num=0x%x:0x%x:0x%x:0x%x\n",
+ pdev->id, einfo->regs, einfo->phyregs,
+ einfo->phy, einfo->irq_num);
+
+ data->regs = ioremap(einfo->regs, 0x400);
+ data->phyregs = ioremap(einfo->phyregs, 0x400);

check for failure (==NULL)


+ data->phy = einfo->phy;
+ data->irq_num = einfo->irq_num;
+ data->id = pdev->id;
+ dev->open = tsi108_open;
+ dev->stop = tsi108_close;
+ dev->hard_start_xmit = tsi108_send_packet;
+ dev->set_mac_address = tsi108_set_mac;
+ dev->set_multicast_list = tsi108_set_rx_mode;
+ dev->get_stats = tsi108_get_stats;
+ dev->poll = tsi108_poll;
+ dev->do_ioctl = tsi108_ioctl;
+ dev->weight = 64; /* 64 is more suitable for GigE interface - klai */
+
+ /* Apparently, the Linux networking code won't use scatter-gather
+ * if the hardware doesn't do checksums. However, it's faster
+ * to checksum in place and use SG, as (among other reasons)
+ * the cache won't be dirtied (which then has to be flushed
+ * before DMA). The checksumming is done by the driver (via
+ * a new function skb_csum_dev() in net/core/skbuff.c).
+ */
+
+ dev->features = NETIF_F_HIGHDMA;
+ SET_MODULE_OWNER(dev);
+
+ spin_lock_init(&data->txlock);
+ spin_lock_init(&data->misclock);
+
+ tsi108_reset_ether(data);
+ tsi108_kill_phy(dev);
+
+ if (tsi108_get_mac(dev) != 0)
+ printk(KERN_ERR "%s: Invalid MAC address. Please correct.\n",
+ dev->name);

handle failure


+ tsi108_init_mac(dev);
+ ret = register_netdev(dev);
+ if (ret < 0) {
+ free_netdev(dev);
+ return ret;

leak: need to iounmap on error


+ }
+
+ printk(KERN_INFO "%s: Tsi108 Gigabit Ethernet, MAC: "
+ "%02x:%02x:%02x:%02x:%02x:%02x\n", dev->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]);
+#ifdef DEBUG
+ dump_eth_one(dev);
+#endif
+
+ return 0;
+}
+
+/* There's no way to either get interrupts from the PHY when
+ * something changes, or to have the Tsi108 automatically communicate
+ * with the PHY to reconfigure itself.
+ *
+ * Thus, we have to do it using a timer. + */
+
+static void tsi108_timed_checker(unsigned long dev_ptr)
+{
+ struct net_device *dev = (struct net_device *)dev_ptr;
+ struct tsi108_prv_data *data = netdev_priv(dev);
+
+ tsi108_check_phy(dev);
+ tsi108_check_rxring(dev);
+ mod_timer(&data->timer, jiffies + CHECK_PHY_INTERVAL);
+}
+
+static int tsi108_ether_init(void)
+{
+ int ret;
+ ret = platform_driver_register (&tsi_eth_driver);
+ if (ret < 0){
+ printk("tsi108_ether_init: error initializing ethernet "
+ "device\n");
+ return ret;
+ }
+ return 0;
+}
+
+static int tsi108_ether_remove(struct platform_device *pdev)
+{
+ struct net_device *dev = platform_get_drvdata(pdev);
+ struct tsi108_prv_data *priv = netdev_priv(dev);
+
+ unregister_netdev(dev);
+ tsi108_stop_ethernet(dev);
+ platform_set_drvdata(pdev, NULL);
+ iounmap((void __iomem *)priv->regs);
+ iounmap((void __iomem *)priv->phyregs);

delete casts

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/