cns3xxx: merge ethernet driver changes from gateworks

git-svn-id: svn://svn.openwrt.org/openwrt/trunk@33489 3c298f89-4303-0410-b956-a3cf2f4a3e73

1 files changed, 258 insertions, 230 deletions

diff --git a/target/linux/cns3xxx/patches-3.3/051-cns3xxx_gigabit.patch b/target/linux/cns3xxx/patches-3.3/051-cns3xxx_gigabit.patch
index 5eb40c5a1f..2f54a7299f 100644
--- a/target/linux/cns3xxx/patches-3.3/051-cns3xxx_gigabit.patch
+++ b/target/linux/cns3xxx/patches-3.3/051-cns3xxx_gigabit.patch
@@ -1,6 +1,6 @@
 --- /dev/null
 +++ b/drivers/net/ethernet/cavium/cns3xxx_eth.c
-@@ -0,0 +1,1270 @@
+@@ -0,0 +1,1298 @@
 +/*
 + * Cavium CNS3xxx Gigabit driver for Linux
 + *
@@ -36,7 +36,8 @@
 +#define RX_POOL_ALLOC_SIZE (sizeof(struct rx_desc) * RX_DESCS)
 +#define TX_POOL_ALLOC_SIZE (sizeof(struct tx_desc) * TX_DESCS)
 +#define REGS_SIZE 336
-+#define MAX_MRU	9500
++#define MAX_MRU (1536 + SKB_DMA_REALIGN)
++#define CNS3XXX_MAX_MTU (1536)
 +
 +#define NAPI_WEIGHT 64
 +
@@ -57,7 +58,12 @@
 +#define TCP_CHECKSUM 0x00010000
 +
 +/* Port Config Defines */
++#define PORT_BP_ENABLE 0x00020000
 +#define PORT_DISABLE 0x00040000
++#define PORT_LEARN_DIS 0x00080000
++#define PORT_BLOCK_STATE 0x00100000
++#define PORT_BLOCK_MODE 0x00200000
++
 +#define PROMISC_OFFSET 29
 +
 +/* Global Config Defines */
@@ -79,6 +85,14 @@
 +#define QUEUE_THRESHOLD 0x000000f0
 +#define CLR_FS_STATE 0x80000000
 +
++/* Interrupt Status Defines */
++#define MAC0_STATUS_CHANGE 0x00004000
++#define MAC1_STATUS_CHANGE 0x00008000
++#define MAC2_STATUS_CHANGE 0x00010000
++#define MAC0_RX_ERROR 0x00100000
++#define MAC1_RX_ERROR 0x00200000
++#define MAC2_RX_ERROR 0x00400000
++
 +struct tx_desc
 +{
 +	u32 sdp; /* segment data pointer */
@@ -188,6 +202,7 @@
 +	u8 alignment[16]; /* for 32 byte alignment */
 +};
 +
++
 +struct switch_regs {
 +	u32 phy_control;
 +	u32 phy_auto_addr;
@@ -233,6 +248,8 @@
 +	u32 fs_dma_ctrl1;
 +	u32 fs_desc_ptr1;
 +	u32 fs_desc_base_addr1;
++	u32 __res7[109];
++	u32 mac_counter0[13];
 +};
 +
 +struct _tx_ring {
@@ -240,6 +257,7 @@
 +	dma_addr_t phys_addr;
 +	struct tx_desc *cur_addr;
 +	struct sk_buff *buff_tab[TX_DESCS];
++	unsigned int phys_tab[TX_DESCS];
 +	u32 free_index;
 +	u32 count_index;
 +	u32 cur_index;
@@ -252,6 +270,7 @@
 +	dma_addr_t phys_addr;
 +	struct rx_desc *cur_addr;
 +	struct sk_buff *buff_tab[RX_DESCS];
++	unsigned int phys_tab[RX_DESCS];
 +	u32 cur_index;
 +	u32 alloc_index;
 +	int alloc_count;
@@ -264,7 +283,6 @@
 +	struct cns3xxx_plat_info *plat;
 +	struct _tx_ring *tx_ring;
 +	struct _rx_ring *rx_ring;
-+	u32 mtu;
 +};
 +
 +struct port {
@@ -273,16 +291,14 @@
 +	struct sw *sw;
 +	int id;			/* logical port ID */
 +	int speed, duplex;
-+	u32 mtu;
 +};
 +
 +static spinlock_t mdio_lock;
-+static spinlock_t tx_lock;
-+static spinlock_t stat_lock;
++static DEFINE_SPINLOCK(tx_lock);
 +static struct switch_regs __iomem *mdio_regs; /* mdio command and status only */
 +struct mii_bus *mdio_bus;
 +static int ports_open;
-+static struct port *switch_port_tab[3];
++static struct port *switch_port_tab[4];
 +static struct dma_pool *rx_dma_pool;
 +static struct dma_pool *tx_dma_pool;
 +struct net_device *napi_dev;
@@ -380,6 +396,16 @@
 +	mdiobus_free(mdio_bus);
 +}
 +
++static void enable_tx_dma(struct sw *sw)
++{
++	__raw_writel(0x1, &sw->regs->ts_dma_ctrl0);
++}
++
++static void enable_rx_dma(struct sw *sw)
++{
++	__raw_writel(0x1, &sw->regs->fs_dma_ctrl0);
++}
++
 +static void cns3xxx_adjust_link(struct net_device *dev)
 +{
 +	struct port *port = netdev_priv(dev);
@@ -414,30 +440,85 @@
 +	return (IRQ_HANDLED);
 +}
 +
++irqreturn_t eth_stat_irq(int irq, void *pdev)
++{
++	struct net_device *dev = pdev;
++	struct sw *sw = netdev_priv(dev);
++	u32 cfg;
++	u32 stat = __raw_readl(&sw->regs->intr_stat);
++	__raw_writel(0xffffffff, &sw->regs->intr_stat);
++
++	if (stat & MAC2_RX_ERROR)
++		switch_port_tab[3]->netdev->stats.rx_dropped++;
++	if (stat & MAC1_RX_ERROR)
++		switch_port_tab[1]->netdev->stats.rx_dropped++;
++	if (stat & MAC0_RX_ERROR)
++		switch_port_tab[0]->netdev->stats.rx_dropped++;
++
++	if (stat & MAC0_STATUS_CHANGE) {
++		cfg = __raw_readl(&sw->regs->mac_cfg[0]);
++		switch_port_tab[0]->phydev->link = (cfg & 0x1);
++		switch_port_tab[0]->phydev->duplex = ((cfg >> 4) & 0x1);
++		if (((cfg >> 2) & 0x3) == 2)
++			switch_port_tab[0]->phydev->speed = 1000;
++		else if (((cfg >> 2) & 0x3) == 1)
++			switch_port_tab[0]->phydev->speed = 100;
++		else
++			switch_port_tab[0]->phydev->speed = 10;
++		cns3xxx_adjust_link(switch_port_tab[0]->netdev);
++	}
++
++	if (stat & MAC1_STATUS_CHANGE) {
++		cfg = __raw_readl(&sw->regs->mac_cfg[1]);
++		switch_port_tab[1]->phydev->link = (cfg & 0x1);
++		switch_port_tab[1]->phydev->duplex = ((cfg >> 4) & 0x1);
++		if (((cfg >> 2) & 0x3) == 2)
++			switch_port_tab[1]->phydev->speed = 1000;
++		else if (((cfg >> 2) & 0x3) == 1)
++			switch_port_tab[1]->phydev->speed = 100;
++		else
++			switch_port_tab[1]->phydev->speed = 10;
++		cns3xxx_adjust_link(switch_port_tab[1]->netdev);
++	}
++
++	if (stat & MAC2_STATUS_CHANGE) {
++		cfg = __raw_readl(&sw->regs->mac_cfg[3]);
++		switch_port_tab[3]->phydev->link = (cfg & 0x1);
++		switch_port_tab[3]->phydev->duplex = ((cfg >> 4) & 0x1);
++		if (((cfg >> 2) & 0x3) == 2)
++			switch_port_tab[3]->phydev->speed = 1000;
++		else if (((cfg >> 2) & 0x3) == 1)
++			switch_port_tab[3]->phydev->speed = 100;
++		else
++			switch_port_tab[3]->phydev->speed = 10;
++		cns3xxx_adjust_link(switch_port_tab[3]->netdev);
++	}
++
++	return (IRQ_HANDLED);
++}
++
++
 +static void cns3xxx_alloc_rx_buf(struct sw *sw, int received)
 +{
 +	struct _rx_ring *rx_ring = sw->rx_ring;
 +	unsigned int i = rx_ring->alloc_index;
-+	struct rx_desc *desc;
++	struct rx_desc *desc = &(rx_ring)->desc[i];
 +	struct sk_buff *skb;
-+	u32 mtu = sw->mtu;
-+
-+	rx_ring->alloc_count += received;
++	unsigned int phys;
 +
-+	for (received = rx_ring->alloc_count; received > 0; received--) {
-+		desc = &(rx_ring)->desc[i];
-+
-+		if ((skb = dev_alloc_skb(mtu))) {
++	for (received += rx_ring->alloc_count; received > 0; received--) {
++		if ((skb = dev_alloc_skb(MAX_MRU))) {
 +			if (SKB_DMA_REALIGN)
 +				skb_reserve(skb, SKB_DMA_REALIGN);
 +			skb_reserve(skb, NET_IP_ALIGN);
-+			desc->sdp = dma_map_single(NULL, skb->data,
-+				    mtu, DMA_FROM_DEVICE);
-+			if (dma_mapping_error(NULL, desc->sdp)) {
++			phys = dma_map_single(NULL, skb->data,
++				    CNS3XXX_MAX_MTU, DMA_FROM_DEVICE);
++			if (dma_mapping_error(NULL, phys)) {
 +				dev_kfree_skb(skb);
 +				/* Failed to map, better luck next time */
 +				goto out;;
 +			}
++			desc->sdp = phys;
 +		} else {
 +			/* Failed to allocate skb, try again next time */
 +			goto out;
@@ -445,13 +526,16 @@
 +
 +		/* put the new buffer on RX-free queue */
 +		rx_ring->buff_tab[i] = skb;
-+
-+		if (++i == RX_DESCS) {
++		rx_ring->phys_tab[i] = phys;
++		if (i == RX_DESCS - 1) {
 +			i = 0;
 +			desc->config0 = END_OF_RING | FIRST_SEGMENT |
-+					LAST_SEGMENT | mtu;
++					LAST_SEGMENT | CNS3XXX_MAX_MTU;
++			desc = &(rx_ring)->desc[i];
 +		} else {
-+			desc->config0 = FIRST_SEGMENT | LAST_SEGMENT | mtu;
++			desc->config0 = FIRST_SEGMENT | LAST_SEGMENT | CNS3XXX_MAX_MTU;
++			i++;
++			desc++;
 +		}
 +	}
 +out:
@@ -459,58 +543,14 @@
 +	rx_ring->alloc_index = i;
 +}
 +
-+static void update_tx_stats(struct sw *sw)
-+{
-+	struct _tx_ring *tx_ring = sw->tx_ring;
-+	struct tx_desc *desc;
-+	struct tx_desc *next_desc;
-+	struct sk_buff *skb;
-+	int i;
-+	int index;
-+	int num_count;
-+
-+	spin_lock_bh(&stat_lock);
-+
-+	num_count = tx_ring->num_count;
-+
-+	if (!num_count) {
-+		spin_unlock_bh(&stat_lock);
-+		return;
-+	}
-+
-+	index = tx_ring->count_index;
-+	desc = &(tx_ring)->desc[index];
-+	for (i = 0; i < num_count; i++) {
-+		skb = tx_ring->buff_tab[index];
-+		if (desc->cown) {
-+			tx_ring->buff_tab[index] = 0;
-+			if (unlikely(++index == TX_DESCS)) index = 0;
-+			next_desc = &(tx_ring)->desc[index];
-+			prefetch(next_desc + 4);
-+			if (likely(skb)) {
-+				skb->dev->stats.tx_packets++;
-+				skb->dev->stats.tx_bytes += skb->len;
-+				dev_kfree_skb_any(skb);
-+			}
-+			desc = next_desc;
-+		} else {
-+			break;
-+		}
-+	}
-+	tx_ring->num_count -= i;
-+	tx_ring->count_index = index;
-+
-+	spin_unlock_bh(&stat_lock);
-+}
-+
 +static void clear_tx_desc(struct sw *sw)
 +{
 +	struct _tx_ring *tx_ring = sw->tx_ring;
 +	struct tx_desc *desc;
-+	struct tx_desc *next_desc;
 +	int i;
 +	int index;
-+	int num_used = tx_ring->num_used - tx_ring->num_count;
++	int num_used = tx_ring->num_used;
++	struct sk_buff *skb;
 +
 +	if (num_used < (TX_DESCS >> 1))
 +		return;
@@ -519,14 +559,18 @@
 +	desc = &(tx_ring)->desc[index];
 +	for (i = 0; i < num_used; i++) {
 +		if (desc->cown) {
-+			if (unlikely(++index == TX_DESCS)) index = 0;
-+			next_desc = &(tx_ring)->desc[index];
-+			prefetch(next_desc);
-+			prefetch(next_desc + 4);
-+			if (likely(desc->sdp))
-+				dma_unmap_single(NULL, desc->sdp,
-+					desc->sdl, DMA_TO_DEVICE);
-+			desc = next_desc;
++			skb = tx_ring->buff_tab[index];
++			tx_ring->buff_tab[index] = 0;
++			if (skb)
++				dev_kfree_skb_any(skb);
++			dma_unmap_single(NULL, tx_ring->phys_tab[index],
++				desc->sdl, DMA_TO_DEVICE);
++			if (++index == TX_DESCS) {
++				index = 0;
++				desc = &(tx_ring)->desc[index];
++			} else {
++				desc++;
++			}
 +		} else {
 +			break;
 +		}
@@ -543,9 +587,7 @@
 +	int received = 0;
 +	unsigned int length;
 +	unsigned int i = rx_ring->cur_index;
-+	struct rx_desc *next_desc;
 +	struct rx_desc *desc = &(rx_ring)->desc[i];
-+	int port_id;
 +
 +	while (desc->cown) {
 +		struct sk_buff *skb;
@@ -555,19 +597,11 @@
 +
 +		skb = rx_ring->buff_tab[i];
 +
-+		if (++i == RX_DESCS) i = 0;
-+		next_desc = &(rx_ring)->desc[i];
-+		prefetch(next_desc);
-+
-+		port_id = desc->sp;
-+		if (port_id == 4)
-+			dev = switch_port_tab[2]->netdev;
-+		else
-+			dev = switch_port_tab[port_id]->netdev;
++		dev = switch_port_tab[desc->sp]->netdev;
 +
 +		length = desc->sdl;
 +		/* process received frame */
-+		dma_unmap_single(&dev->dev, desc->sdp,
++		dma_unmap_single(&dev->dev, rx_ring->phys_tab[i],
 +				 length, DMA_FROM_DEVICE);
 +
 +		skb_put(skb, length);
@@ -578,6 +612,7 @@
 +		dev->stats.rx_packets++;
 +		dev->stats.rx_bytes += length;
 +
++		/* RX Hardware checksum offload */
 +		switch (desc->prot) {
 +			case 1:
 +			case 2:
@@ -598,10 +633,17 @@
 +		napi_gro_receive(napi, skb);
 +
 +		received++;
-+		desc = next_desc;
++
++		if (++i == RX_DESCS) {
++			i = 0;
++			desc = &(rx_ring)->desc[i];
++		} else {
++			desc++;
++		}
 +	}
 +
 +	cns3xxx_alloc_rx_buf(sw, received);
++
 +	rx_ring->cur_index = i;
 +
 +	if (received != budget) {
@@ -609,6 +651,8 @@
 +		enable_irq(IRQ_CNS3XXX_SW_R0RXC);
 +	}
 +
++	enable_rx_dma(sw);
++
 +	return received;
 +}
 +
@@ -619,63 +663,132 @@
 +	struct _tx_ring *tx_ring = sw->tx_ring;
 +	struct tx_desc *tx_desc;
 +	int index;
-+	int len = skb->len;
++	int len;
 +	char pmap = (1 << port->id);
++	unsigned int phys;
++	int nr_frags = skb_shinfo(skb)->nr_frags;
++	struct skb_frag_struct *frag;
++	unsigned int i;
 +
 +	if (pmap == 8)
 +		pmap = (1 << 4);
 +
-+	if (unlikely(len > sw->mtu)) {
++	if (skb->len > CNS3XXX_MAX_MTU) {
 +		dev_kfree_skb(skb);
 +		dev->stats.tx_errors++;
 +		return NETDEV_TX_OK;
 +	}
 +
-+	update_tx_stats(sw);
-+
-+	spin_lock_bh(&tx_lock);
++	spin_lock(&tx_lock);
 +
-+	clear_tx_desc(sw);
-+
-+	if (unlikely(tx_ring->num_used == TX_DESCS)) {
-+		spin_unlock_bh(&tx_lock);
-+		return NETDEV_TX_BUSY;
++	if ((tx_ring->num_used + nr_frags) >= TX_DESCS) {
++		clear_tx_desc(sw);
++		if ((tx_ring->num_used + nr_frags) >= TX_DESCS) {
++			spin_unlock(&tx_lock);
++			return NETDEV_TX_BUSY;
++		}
 +	}
 +
 +	index = tx_ring->cur_index;
++	tx_ring->cur_index = ((tx_ring->cur_index + nr_frags + 1) % TX_DESCS);
 +
-+	if (unlikely(++tx_ring->cur_index == TX_DESCS))
-+		tx_ring->cur_index = 0;
++	spin_unlock(&tx_lock);
 +
-+	tx_ring->num_used++;
-+	tx_ring->num_count++;
++	if (!nr_frags) {
++		tx_desc = &(tx_ring)->desc[index];
 +
-+	spin_unlock_bh(&tx_lock);
++		len = skb->len;
 +
-+	tx_desc = &(tx_ring)->desc[index];
++		phys = dma_map_single(NULL, skb->data, len,
++				      	DMA_TO_DEVICE);
 +
-+	tx_desc->sdp = dma_map_single(NULL, skb->data, len,
-+				      DMA_TO_DEVICE);
++		tx_desc->sdp = phys;
++		tx_desc->pmap = pmap;
++		tx_ring->phys_tab[index] = phys;
 +
-+	if (dma_mapping_error(NULL, tx_desc->sdp)) {
-+		dev_kfree_skb(skb);
-+		dev->stats.tx_errors++;
-+		return NETDEV_TX_OK;
-+	}
++		tx_ring->buff_tab[index] = skb;
++		if (index == TX_DESCS - 1) {
++			tx_desc->config0 = END_OF_RING | FIRST_SEGMENT | LAST_SEGMENT |
++				   	FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
++				   	TCP_CHECKSUM | len;
++		} else {
++			tx_desc->config0 = FIRST_SEGMENT | LAST_SEGMENT |
++				   	FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
++				   	TCP_CHECKSUM | len;
++		}
++	} else {
++		unsigned int config;
 +
-+	tx_desc->pmap = pmap;
-+	tx_ring->buff_tab[index] = skb;
++		index = ((index + nr_frags) % TX_DESCS);
++		tx_desc = &(tx_ring)->desc[index];
 +
-+	if (index == TX_DESCS - 1) {
-+		tx_desc->config0 = END_OF_RING | FIRST_SEGMENT | LAST_SEGMENT |
-+				   FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
-+				   TCP_CHECKSUM | len;
-+	} else {
-+		tx_desc->config0 = FIRST_SEGMENT | LAST_SEGMENT |
-+				   FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
-+				   TCP_CHECKSUM | len;
++		/* fragments */
++		for (i = nr_frags; i > 0; i--) {
++			void *addr;
++
++			frag = &skb_shinfo(skb)->frags[i-1];
++			len = frag->size;
++
++			addr = page_address(skb_frag_page(frag)) +
++			       frag->page_offset;
++			phys = dma_map_single(NULL, addr, len, DMA_TO_DEVICE);
++
++			tx_desc->sdp = phys;
++
++			tx_desc->pmap = pmap;
++			tx_ring->phys_tab[index] = phys;
++
++			config = FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
++				TCP_CHECKSUM | len;
++			if (i == nr_frags) {
++				config |= LAST_SEGMENT;
++				tx_ring->buff_tab[index] = skb;
++			}
++			if (index == TX_DESCS - 1)
++				config |= END_OF_RING;
++			tx_desc->config0 = config;
++
++			if (index == 0) {
++				index = TX_DESCS - 1;
++				tx_desc = &(tx_ring)->desc[index];
++			} else {
++				index--;
++				tx_desc--;
++			}
++		}
++
++		/* header */
++		len = skb->len - skb->data_len;
++
++		phys = dma_map_single(NULL, skb->data, len,
++				      	DMA_TO_DEVICE);
++
++		tx_desc->sdp = phys;
++		tx_desc->pmap = pmap;
++		tx_ring->phys_tab[index] = phys;
++
++		if (index == TX_DESCS - 1) {
++			tx_desc->config0 = END_OF_RING | FIRST_SEGMENT |
++				   	FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
++				   	TCP_CHECKSUM | len;
++		} else {
++			tx_desc->config0 = FIRST_SEGMENT |
++				   	FORCE_ROUTE | IP_CHECKSUM | UDP_CHECKSUM |
++				   	TCP_CHECKSUM | len;
++		}
 +	}
 +
++	mb();
++
++	spin_lock(&tx_lock);
++	tx_ring->num_used += nr_frags + 1;
++	spin_unlock(&tx_lock);
++
++	dev->stats.tx_packets++;
++	dev->stats.tx_bytes += skb->len;
++
++	enable_tx_dma(sw);
++
 +	return NETDEV_TX_OK;
 +}
 +
@@ -750,23 +863,24 @@
 +	for (i = 0; i < RX_DESCS; i++) {
 +		struct rx_desc *desc = &(rx_ring)->desc[i];
 +		struct sk_buff *skb;
-+		if (!(skb = dev_alloc_skb(sw->mtu)))
++		if (!(skb = dev_alloc_skb(MAX_MRU)))
 +			return -ENOMEM;
 +		if (SKB_DMA_REALIGN)
 +			skb_reserve(skb, SKB_DMA_REALIGN);
 +		skb_reserve(skb, NET_IP_ALIGN);
-+		desc->sdl = sw->mtu;
++		desc->sdl = CNS3XXX_MAX_MTU;
 +		if (i == (RX_DESCS - 1))
 +			desc->eor = 1;
 +		desc->fsd = 1;
 +		desc->lsd = 1;
 +
 +		desc->sdp = dma_map_single(NULL, skb->data,
-+					    sw->mtu, DMA_FROM_DEVICE);
++					    CNS3XXX_MAX_MTU, DMA_FROM_DEVICE);
 +		if (dma_mapping_error(NULL, desc->sdp)) {
 +			return -EIO;
 +		}
 +		rx_ring->buff_tab[i] = skb;
++		rx_ring->phys_tab[i] = desc->sdp;
 +		desc->cown = 0;
 +	}
 +	__raw_writel(rx_ring->phys_addr, &sw->regs->fs_desc_ptr0);
@@ -807,7 +921,7 @@
 +			if (skb) {
 +				dma_unmap_single(NULL,
 +						 desc->sdp,
-+						 sw->mtu, DMA_FROM_DEVICE);
++						 CNS3XXX_MAX_MTU, DMA_FROM_DEVICE);
 +				dev_kfree_skb(skb);
 +			}
 +		}
@@ -847,9 +961,12 @@
 +
 +	if (!ports_open) {
 +		request_irq(IRQ_CNS3XXX_SW_R0RXC, eth_rx_irq, IRQF_SHARED, "gig_switch", napi_dev);
++		request_irq(IRQ_CNS3XXX_SW_STATUS, eth_stat_irq, IRQF_SHARED, "gig_stat", napi_dev);
 +		napi_enable(&sw->napi);
 +		netif_start_queue(napi_dev);
-+		//enable_irq(IRQ_CNS3XXX_SW_R0RXC);
++
++ 		__raw_writel(~(MAC0_STATUS_CHANGE | MAC1_STATUS_CHANGE | MAC2_STATUS_CHANGE |
++ 			       MAC0_RX_ERROR | MAC1_RX_ERROR | MAC2_RX_ERROR), &sw->regs->intr_mask);
 +
 +		temp = __raw_readl(&sw->regs->mac_cfg[2]);
 +		temp &= ~(PORT_DISABLE);
@@ -859,7 +976,7 @@
 +		temp &= ~(TS_SUSPEND | FS_SUSPEND);
 +		__raw_writel(temp, &sw->regs->dma_auto_poll_cfg);
 +
-+		__raw_writel((TS_POLL_EN | FS_POLL_EN), &sw->regs->dma_auto_poll_cfg);
++		enable_rx_dma(sw);
 +	}
 +	temp = __raw_readl(&sw->regs->mac_cfg[port->id]);
 +	temp &= ~(PORT_DISABLE);
@@ -890,6 +1007,8 @@
 +	if (!ports_open) {
 +		disable_irq(IRQ_CNS3XXX_SW_R0RXC);
 +		free_irq(IRQ_CNS3XXX_SW_R0RXC, napi_dev);
++		disable_irq(IRQ_CNS3XXX_SW_STATUS);
++		free_irq(IRQ_CNS3XXX_SW_STATUS, napi_dev);
 +		napi_disable(&sw->napi);
 +		netif_stop_queue(napi_dev);
 +		temp = __raw_readl(&sw->regs->mac_cfg[2]);
@@ -979,96 +1098,13 @@
 +	return 0;
 +}
 +
-+static int cns3xxx_change_mtu(struct net_device *netdev, int new_mtu)
-+{
-+	struct port *port = netdev_priv(netdev);
-+	struct sw *sw = port->sw;
-+	u32 temp;
-+	int i;
-+	struct _rx_ring *rx_ring = sw->rx_ring;
-+	struct rx_desc *desc;
-+	struct sk_buff *skb;
-+
-+	if (new_mtu > MAX_MRU)
-+		return -EINVAL;
-+
-+	netdev->mtu = new_mtu;
-+
-+	new_mtu += 36 + SKB_DMA_REALIGN;
-+	port->mtu = new_mtu;
-+
-+	new_mtu = 0;
-+	for (i = 0; i < 3; i++) {
-+		if (switch_port_tab[i]) {
-+			if (switch_port_tab[i]->mtu > new_mtu)
-+				new_mtu = switch_port_tab[i]->mtu;
-+		}
-+	}
-+
-+
-+	if (new_mtu == sw->mtu)
-+		return 0;
-+
-+	disable_irq(IRQ_CNS3XXX_SW_R0RXC);
-+
-+	sw->mtu = new_mtu;
-+
-+	/* Disable DMA */
-+	__raw_writel(TS_SUSPEND | FS_SUSPEND, &sw->regs->dma_auto_poll_cfg);
-+
-+	for (i = 0; i < RX_DESCS; i++) {
-+		desc = &(rx_ring)->desc[i];
-+		/* Check if we own it, if we do, it will get set correctly
-+		 * when it is re-used */
-+		if (!desc->cown) {
-+			skb = rx_ring->buff_tab[i];
-+			dma_unmap_single(NULL, desc->sdp, desc->sdl,
-+					 DMA_FROM_DEVICE);
-+			dev_kfree_skb(skb);
-+
-+			if ((skb = dev_alloc_skb(new_mtu))) {
-+				if (SKB_DMA_REALIGN)
-+					skb_reserve(skb, SKB_DMA_REALIGN);
-+				skb_reserve(skb, NET_IP_ALIGN);
-+				desc->sdp = dma_map_single(NULL, skb->data,
-+					    new_mtu, DMA_FROM_DEVICE);
-+				if (dma_mapping_error(NULL, desc->sdp)) {
-+					dev_kfree_skb(skb);
-+					skb = NULL;
-+				}
-+			}
-+
-+			/* put the new buffer on RX-free queue */
-+			rx_ring->buff_tab[i] = skb;
-+
-+			if (i == RX_DESCS - 1)
-+				desc->config0 = END_OF_RING | FIRST_SEGMENT |
-+						LAST_SEGMENT | new_mtu;
-+			else
-+				desc->config0 = FIRST_SEGMENT |
-+						LAST_SEGMENT | new_mtu;
-+		}
-+	}
-+
-+	/* Re-ENABLE DMA */
-+	temp = __raw_readl(&sw->regs->dma_auto_poll_cfg);
-+	temp &= ~(TS_SUSPEND | FS_SUSPEND);
-+	__raw_writel(temp, &sw->regs->dma_auto_poll_cfg);
-+
-+	__raw_writel((TS_POLL_EN | FS_POLL_EN), &sw->regs->dma_auto_poll_cfg);
-+
-+	enable_irq(IRQ_CNS3XXX_SW_R0RXC);
-+
-+	return 0;
-+}
-+
 +static const struct net_device_ops cns3xxx_netdev_ops = {
 +	.ndo_open = eth_open,
 +	.ndo_stop = eth_close,
 +	.ndo_start_xmit = eth_xmit,
 +	.ndo_set_rx_mode = eth_rx_mode,
 +	.ndo_do_ioctl = eth_ioctl,
-+	.ndo_change_mtu = cns3xxx_change_mtu,
++	.ndo_change_mtu = eth_change_mtu,
 +	.ndo_set_mac_address = eth_set_mac,
 +	.ndo_validate_addr = eth_validate_addr,
 +};
@@ -1085,12 +1121,10 @@
 +	int err;
 +	u32 temp;
 +
-+	spin_lock_init(&tx_lock);
-+	spin_lock_init(&stat_lock);
-+
 +	if (!(napi_dev = alloc_etherdev(sizeof(struct sw))))
 +		return -ENOMEM;
 +	strcpy(napi_dev->name, "switch%d");
++	napi_dev->features = NETIF_F_IP_CSUM | NETIF_F_SG;
 +
 +	SET_NETDEV_DEV(napi_dev, &pdev->dev);
 +	sw = netdev_priv(napi_dev);
@@ -1103,11 +1137,9 @@
 +		goto err_free;
 +	}
 +
-+	sw->mtu = 1536 + SKB_DMA_REALIGN;
-+
 +	for (i = 0; i < 4; i++) {
 +		temp = __raw_readl(&sw->regs->mac_cfg[i]);
-+		temp |= (PORT_DISABLE) | 0x80000000;
++		temp |= (PORT_DISABLE);
 +		__raw_writel(temp, &sw->regs->mac_cfg[i]);
 +	}
 +
@@ -1118,7 +1150,7 @@
 +	temp |= NIC_MODE | VLAN_UNAWARE;
 +	__raw_writel(temp, &sw->regs->vlan_cfg);
 +
-+	__raw_writel(UNKNOWN_VLAN_TO_CPU | ACCEPT_CRC_PACKET |
++	__raw_writel(UNKNOWN_VLAN_TO_CPU |
 +		     CRC_STRIPPING, &sw->regs->mac_glob_cfg);
 +
 +	if (!(sw->rx_ring = kmalloc(sizeof(struct _rx_ring), GFP_KERNEL))) {
@@ -1151,7 +1183,6 @@
 +			goto free_ports;
 +		}
 +
-+		//SET_NETDEV_DEV(dev, &pdev->dev);
 +		port = netdev_priv(dev);
 +		port->netdev = dev;
 +		if (i == 2)
@@ -1159,36 +1190,33 @@
 +		else
 +			port->id = i;
 +		port->sw = sw;
-+		port->mtu = sw->mtu;
 +
 +		temp = __raw_readl(&sw->regs->mac_cfg[port->id]);
-+		temp |= (PORT_DISABLE);
++		temp |= (PORT_DISABLE | PORT_BLOCK_STATE | PORT_LEARN_DIS);
 +		__raw_writel(temp, &sw->regs->mac_cfg[port->id]);
 +
 +		dev->netdev_ops = &cns3xxx_netdev_ops;
 +		dev->ethtool_ops = &cns3xxx_ethtool_ops;
 +		dev->tx_queue_len = 1000;
-+		dev->features = NETIF_F_HW_CSUM;
-+
-+		dev->vlan_features = NETIF_F_HW_CSUM;
++		dev->features = NETIF_F_IP_CSUM | NETIF_F_SG;
 +
-+		switch_port_tab[i] = port;
++		switch_port_tab[port->id] = port;
 +		memcpy(dev->dev_addr, &plat->hwaddr[i], ETH_ALEN);
 +
 +		snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT, "0", plat->phy[i]);
 +		port->phydev = phy_connect(dev, phy_id, &cns3xxx_adjust_link, 0,
 +			PHY_INTERFACE_MODE_RGMII);
 +		if ((err = IS_ERR(port->phydev))) {
-+			switch_port_tab[i] = 0;
++			switch_port_tab[port->id] = 0;
 +			free_netdev(dev);
 +			goto free_ports;
 +		}
 +
-+		port->phydev->irq = PHY_POLL;
++		port->phydev->irq = PHY_IGNORE_INTERRUPT;
 +
 +		if ((err = register_netdev(dev))) {
 +			phy_disconnect(port->phydev);
-+			switch_port_tab[i] = 0;
++			switch_port_tab[port->id] = 0;
 +			free_netdev(dev);
 +			goto free_ports;
 +		}
@@ -1228,7 +1256,7 @@
 +	int i;
 +	destroy_rings(sw);
 +
-+	for (i = 2; i >= 0; i--) {
++	for (i = 3; i >= 0; i--) {
 +		if (switch_port_tab[i]) {
 +			struct port *port = switch_port_tab[i];
 +			struct net_device *dev = port->netdev;