Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6

Conflicts:

	drivers/net/wireless/rt2x00/rt2x00dev.c
	net/8021q/vlan_dev.c

37 files changed, 441 insertions, 231 deletions

diff --git a/drivers/net/3c501.c b/drivers/net/3c501.c
index 7d253686ed0..5ba4bab6d43 100644
--- a/drivers/net/3c501.c
+++ b/drivers/net/3c501.c
@@ -485,9 +485,7 @@ static int el_start_xmit(struct sk_buff *skb, struct net_device *dev)
 			printk(KERN_DEBUG "%s: burped during tx load.\n",
 				dev->name);
 		spin_lock_irqsave(&lp->lock, flags);
-	}
-	while (1);
-
+	} while (1);
 }
 
 /**
@@ -612,7 +610,8 @@ static irqreturn_t el_interrupt(int irq, void *dev_id)
 			dev->stats.tx_packets++;
 			if (el_debug > 6)
 				printk(KERN_DEBUG " Tx succeeded %s\n",
-					(txsr & TX_RDY) ? "." : "but tx is busy!");
+					(txsr & TX_RDY) ? "." :
+							"but tx is busy!");
 			/*
 			 *	This is safe the interrupt is atomic WRT itself.
 			 */
@@ -693,7 +692,8 @@ static void el_receive(struct net_device *dev)
 
 	if (pkt_len < 60 || pkt_len > 1536) {
 		if (el_debug)
-			printk(KERN_DEBUG "%s: bogus packet, length=%d\n", dev->name, pkt_len);
+			printk(KERN_DEBUG "%s: bogus packet, length=%d\n",
+						dev->name, pkt_len);
 		dev->stats.rx_over_errors++;
 		return;
 	}
@@ -711,7 +711,8 @@ static void el_receive(struct net_device *dev)
 
 	outw(0x00, GP_LOW);
 	if (skb == NULL) {
-		printk(KERN_INFO "%s: Memory squeeze, dropping packet.\n", dev->name);
+		printk(KERN_INFO "%s: Memory squeeze, dropping packet.\n",
+								dev->name);
 		dev->stats.rx_dropped++;
 		return;
 	} else {
@@ -748,7 +749,8 @@ static void  el_reset(struct net_device *dev)
 	if (el_debug > 2)
 		printk(KERN_INFO "3c501 reset...");
 	outb(AX_RESET, AX_CMD);		/* Reset the chip */
-	outb(AX_LOOP, AX_CMD);		/* Aux control, irq and loopback enabled */
+	/* Aux control, irq and loopback enabled */
+	outb(AX_LOOP, AX_CMD);
 	{
 		int i;
 		for (i = 0; i < 6; i++)	/* Set the station address. */
diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig
index a64bb415f10..fe7b5ec0970 100644
--- a/drivers/net/Kconfig
+++ b/drivers/net/Kconfig
@@ -2366,15 +2366,15 @@ config GELIC_NET
 	  module will be called ps3_gelic.
 
 config GELIC_WIRELESS
-       bool "PS3 Wireless support"
-       depends on GELIC_NET
-       select WIRELESS_EXT
-       help
-        This option adds the support for the wireless feature of PS3.
-        If you have the wireless-less model of PS3 or have no plan to
-        use wireless feature, disabling this option saves memory.  As
-        the driver automatically distinguishes the models, you can
-        safely enable this option even if you have a wireless-less model.
+	bool "PS3 Wireless support"
+	depends on GELIC_NET
+	select WIRELESS_EXT
+	help
+	  This option adds the support for the wireless feature of PS3.
+	  If you have the wireless-less model of PS3 or have no plan to
+	  use wireless feature, disabling this option saves memory.  As
+	  the driver automatically distinguishes the models, you can
+	  safely enable this option even if you have a wireless-less model.
 
 config GIANFAR
 	tristate "Gianfar Ethernet"
diff --git a/drivers/net/ac3200.c b/drivers/net/ac3200.c
index 5136d94923a..b1448637107 100644
--- a/drivers/net/ac3200.c
+++ b/drivers/net/ac3200.c
@@ -369,7 +369,7 @@ MODULE_PARM_DESC(mem, "Memory base address(es)");
 MODULE_DESCRIPTION("Ansel AC3200 EISA ethernet driver");
 MODULE_LICENSE("GPL");
 
-int __init init_module(void)
+static int __init ac3200_module_init(void)
 {
 	struct net_device *dev;
 	int this_dev, found = 0;
@@ -404,8 +404,7 @@ static void cleanup_card(struct net_device *dev)
 	iounmap(ei_status.mem);
 }
 
-void __exit
-cleanup_module(void)
+static void __exit ac3200_module_exit(void)
 {
 	int this_dev;
 
@@ -418,4 +417,6 @@ cleanup_module(void)
 		}
 	}
 }
+module_init(ac3200_module_init);
+module_exit(ac3200_module_exit);
 #endif /* MODULE */
diff --git a/drivers/net/apne.c b/drivers/net/apne.c
index c12cbdf368b..47a8275d396 100644
--- a/drivers/net/apne.c
+++ b/drivers/net/apne.c
@@ -569,7 +569,7 @@ static irqreturn_t apne_interrupt(int irq, void *dev_id)
 #ifdef MODULE
 static struct net_device *apne_dev;
 
-int __init init_module(void)
+static int __init apne_module_init(void)
 {
 	apne_dev = apne_probe(-1);
 	if (IS_ERR(apne_dev))
@@ -577,7 +577,7 @@ int __init init_module(void)
 	return 0;
 }
 
-void __exit cleanup_module(void)
+static void __exit apne_module_exit(void)
 {
 	unregister_netdev(apne_dev);
 
@@ -591,7 +591,8 @@ void __exit cleanup_module(void)
 
 	free_netdev(apne_dev);
 }
-
+module_init(apne_module_init);
+module_exit(apne_module_exit);
 #endif
 
 static int init_pcmcia(void)
diff --git a/drivers/net/appletalk/ltpc.c b/drivers/net/appletalk/ltpc.c
index 6ab2c2d4d67..fef5560bc7a 100644
--- a/drivers/net/appletalk/ltpc.c
+++ b/drivers/net/appletalk/ltpc.c
@@ -1252,7 +1252,7 @@ module_param(irq, int, 0);
 module_param(dma, int, 0);
 
 
-int __init init_module(void)
+static int __init ltpc_module_init(void)
 {
         if(io == 0)
 		printk(KERN_NOTICE
@@ -1263,6 +1263,7 @@ int __init init_module(void)
 		return PTR_ERR(dev_ltpc);
 	return 0;
 }
+module_init(ltpc_module_init);
 #endif
 
 static void __exit ltpc_cleanup(void)
diff --git a/drivers/net/arcnet/capmode.c b/drivers/net/arcnet/capmode.c
index cc4610db639..02cb8f1c114 100644
--- a/drivers/net/arcnet/capmode.c
+++ b/drivers/net/arcnet/capmode.c
@@ -80,17 +80,19 @@ void arcnet_cap_init(void)
 
 #ifdef MODULE
 
-int __init init_module(void)
+static int __init capmode_module_init(void)
 {
 	printk(VERSION);
 	arcnet_cap_init();
 	return 0;
 }
 
-void cleanup_module(void)
+static void __exit capmode_module_exit(void)
 {
 	arcnet_unregister_proto(&capmode_proto);
 }
+module_init(capmode_module_init);
+module_exit(capmode_module_exit);
 
 MODULE_LICENSE("GPL");
 #endif				/* MODULE */
diff --git a/drivers/net/atl1/atl1_main.c b/drivers/net/atl1/atl1_main.c
index 9200ee59d85..129b8b3aa77 100644
--- a/drivers/net/atl1/atl1_main.c
+++ b/drivers/net/atl1/atl1_main.c
@@ -1765,15 +1765,12 @@ static irqreturn_t atl1_intr(int irq, void *data)
 {
 	struct atl1_adapter *adapter = netdev_priv(data);
 	u32 status;
-	u8 update_rx;
 	int max_ints = 10;
 
 	status = adapter->cmb.cmb->int_stats;
 	if (!status)
 		return IRQ_NONE;
 
-	update_rx = 0;
-
 	do {
 		/* clear CMB interrupt status at once */
 		adapter->cmb.cmb->int_stats = 0;
diff --git a/drivers/net/cxgb3/sge.c b/drivers/net/cxgb3/sge.c
index 979f3fc5e76..db586870c5f 100644
--- a/drivers/net/cxgb3/sge.c
+++ b/drivers/net/cxgb3/sge.c
@@ -1107,9 +1107,15 @@ int t3_eth_xmit(struct sk_buff *skb, struct net_device *dev)
 	}
 
 	q->in_use += ndesc;
-	if (unlikely(credits - ndesc < q->stop_thres))
-		if (USE_GTS || !should_restart_tx(q))
-			t3_stop_queue(dev, qs, q);
+	if (unlikely(credits - ndesc < q->stop_thres)) {
+		t3_stop_queue(dev, qs, q);
+
+		if (should_restart_tx(q) &&
+		    test_and_clear_bit(TXQ_ETH, &qs->txq_stopped)) {
+			q->restarts++;
+			netif_wake_queue(dev);
+		}
+	}
 
 	gen = q->gen;
 	q->unacked += ndesc;
diff --git a/drivers/net/enc28j60.c b/drivers/net/enc28j60.c
index 0809a6a5a28..46a90e9ec56 100644
--- a/drivers/net/enc28j60.c
+++ b/drivers/net/enc28j60.c
@@ -900,7 +900,7 @@ static void enc28j60_hw_rx(struct net_device *ndev)
 		if (RSV_GETBIT(rxstat, RSV_LENCHECKERR))
 			ndev->stats.rx_frame_errors++;
 	} else {
-		skb = dev_alloc_skb(len);
+		skb = dev_alloc_skb(len + NET_IP_ALIGN);
 		if (!skb) {
 			if (netif_msg_rx_err(priv))
 				dev_err(&ndev->dev,
@@ -908,6 +908,7 @@ static void enc28j60_hw_rx(struct net_device *ndev)
 			ndev->stats.rx_dropped++;
 		} else {
 			skb->dev = ndev;
+			skb_reserve(skb, NET_IP_ALIGN);
 			/* copy the packet from the receive buffer */
 			enc28j60_mem_read(priv, priv->next_pk_ptr + sizeof(rsv),
 					len, skb_put(skb, len));
diff --git a/drivers/net/epic100.c b/drivers/net/epic100.c
index 0b365b8d947..76118ddd104 100644
--- a/drivers/net/epic100.c
+++ b/drivers/net/epic100.c
@@ -131,8 +131,8 @@ IIIa. Ring buffers
 
 IVb. References
 
-http://www.smsc.com/main/datasheets/83c171.pdf
-http://www.smsc.com/main/datasheets/83c175.pdf
+http://www.smsc.com/main/tools/discontinued/83c171.pdf
+http://www.smsc.com/main/tools/discontinued/83c175.pdf
 http://scyld.com/expert/NWay.html
 http://www.national.com/pf/DP/DP83840A.html
 
@@ -227,7 +227,12 @@ static const u16 media2miictl[16] = {
 	0, 0x0C00, 0x0C00, 0x2000,  0x0100, 0x2100, 0, 0,
 	0, 0, 0, 0,  0, 0, 0, 0 };
 
-/* The EPIC100 Rx and Tx buffer descriptors. */
+/*
+ * The EPIC100 Rx and Tx buffer descriptors.  Note that these
+ * really ARE host-endian; it's not a misannotation.  We tell
+ * the card to byteswap them internally on big-endian hosts -
+ * look for #ifdef CONFIG_BIG_ENDIAN in epic_open().
+ */
 
 struct epic_tx_desc {
 	u32 txstatus;
@@ -418,7 +423,7 @@ static int __devinit epic_init_one (struct pci_dev *pdev,
 
 	/* Note: the '175 does not have a serial EEPROM. */
 	for (i = 0; i < 3; i++)
-		((u16 *)dev->dev_addr)[i] = le16_to_cpu(inw(ioaddr + LAN0 + i*4));
+		((__le16 *)dev->dev_addr)[i] = cpu_to_le16(inw(ioaddr + LAN0 + i*4));
 
 	if (debug > 2) {
 		dev_printk(KERN_DEBUG, &pdev->dev, "EEPROM contents:\n");
@@ -682,7 +687,8 @@ static int epic_open(struct net_device *dev)
 	if (ep->chip_flags & MII_PWRDWN)
 		outl((inl(ioaddr + NVCTL) & ~0x003C) | 0x4800, ioaddr + NVCTL);
 
-#if defined(__powerpc__) || defined(__sparc__)		/* Big endian */
+	/* Tell the chip to byteswap descriptors on big-endian hosts */
+#ifdef CONFIG_BIG_ENDIAN
 	outl(0x4432 | (RX_FIFO_THRESH<<8), ioaddr + GENCTL);
 	inl(ioaddr + GENCTL);
 	outl(0x0432 | (RX_FIFO_THRESH<<8), ioaddr + GENCTL);
@@ -695,7 +701,7 @@ static int epic_open(struct net_device *dev)
 	udelay(20); /* Looks like EPII needs that if you want reliable RX init. FIXME: pci posting bug? */
 
 	for (i = 0; i < 3; i++)
-		outl(cpu_to_le16(((u16*)dev->dev_addr)[i]), ioaddr + LAN0 + i*4);
+		outl(le16_to_cpu(((__le16*)dev->dev_addr)[i]), ioaddr + LAN0 + i*4);
 
 	ep->tx_threshold = TX_FIFO_THRESH;
 	outl(ep->tx_threshold, ioaddr + TxThresh);
@@ -798,7 +804,7 @@ static void epic_restart(struct net_device *dev)
 	for (i = 16; i > 0; i--)
 		outl(0x0008, ioaddr + TEST1);
 
-#if defined(__powerpc__) || defined(__sparc__)		/* Big endian */
+#ifdef CONFIG_BIG_ENDIAN
 	outl(0x0432 | (RX_FIFO_THRESH<<8), ioaddr + GENCTL);
 #else
 	outl(0x0412 | (RX_FIFO_THRESH<<8), ioaddr + GENCTL);
@@ -808,7 +814,7 @@ static void epic_restart(struct net_device *dev)
 		outl((inl(ioaddr + NVCTL) & ~0x003C) | 0x4800, ioaddr + NVCTL);
 
 	for (i = 0; i < 3; i++)
-		outl(cpu_to_le16(((u16*)dev->dev_addr)[i]), ioaddr + LAN0 + i*4);
+		outl(le16_to_cpu(((__le16*)dev->dev_addr)[i]), ioaddr + LAN0 + i*4);
 
 	ep->tx_threshold = TX_FIFO_THRESH;
 	outl(ep->tx_threshold, ioaddr + TxThresh);
@@ -919,7 +925,7 @@ static void epic_init_ring(struct net_device *dev)
 	/* Initialize all Rx descriptors. */
 	for (i = 0; i < RX_RING_SIZE; i++) {
 		ep->rx_ring[i].rxstatus = 0;
-		ep->rx_ring[i].buflength = cpu_to_le32(ep->rx_buf_sz);
+		ep->rx_ring[i].buflength = ep->rx_buf_sz;
 		ep->rx_ring[i].next = ep->rx_ring_dma +
 				      (i+1)*sizeof(struct epic_rx_desc);
 		ep->rx_skbuff[i] = NULL;
@@ -936,7 +942,7 @@ static void epic_init_ring(struct net_device *dev)
 		skb_reserve(skb, 2);	/* 16 byte align the IP header. */
 		ep->rx_ring[i].bufaddr = pci_map_single(ep->pci_dev,
 			skb->data, ep->rx_buf_sz, PCI_DMA_FROMDEVICE);
-		ep->rx_ring[i].rxstatus = cpu_to_le32(DescOwn);
+		ep->rx_ring[i].rxstatus = DescOwn;
 	}
 	ep->dirty_rx = (unsigned int)(i - RX_RING_SIZE);
 
@@ -974,20 +980,20 @@ static int epic_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	ep->tx_ring[entry].bufaddr = pci_map_single(ep->pci_dev, skb->data,
 		 			            skb->len, PCI_DMA_TODEVICE);
 	if (free_count < TX_QUEUE_LEN/2) {/* Typical path */
-		ctrl_word = cpu_to_le32(0x100000); /* No interrupt */
+		ctrl_word = 0x100000; /* No interrupt */
 	} else if (free_count == TX_QUEUE_LEN/2) {
-		ctrl_word = cpu_to_le32(0x140000); /* Tx-done intr. */
+		ctrl_word = 0x140000; /* Tx-done intr. */
 	} else if (free_count < TX_QUEUE_LEN - 1) {
-		ctrl_word = cpu_to_le32(0x100000); /* No Tx-done intr. */
+		ctrl_word = 0x100000; /* No Tx-done intr. */
 	} else {
 		/* Leave room for an additional entry. */
-		ctrl_word = cpu_to_le32(0x140000); /* Tx-done intr. */
+		ctrl_word = 0x140000; /* Tx-done intr. */
 		ep->tx_full = 1;
 	}
-	ep->tx_ring[entry].buflength = ctrl_word | cpu_to_le32(skb->len);
+	ep->tx_ring[entry].buflength = ctrl_word | skb->len;
 	ep->tx_ring[entry].txstatus =
 		((skb->len >= ETH_ZLEN ? skb->len : ETH_ZLEN) << 16)
-		| cpu_to_le32(DescOwn);
+			    | DescOwn;
 
 	ep->cur_tx++;
 	if (ep->tx_full)
@@ -1041,7 +1047,7 @@ static void epic_tx(struct net_device *dev, struct epic_private *ep)
 	for (dirty_tx = ep->dirty_tx; cur_tx - dirty_tx > 0; dirty_tx++) {
 		struct sk_buff *skb;
 		int entry = dirty_tx % TX_RING_SIZE;
-		int txstatus = le32_to_cpu(ep->tx_ring[entry].txstatus);
+		int txstatus = ep->tx_ring[entry].txstatus;
 
 		if (txstatus & DescOwn)
 			break;	/* It still hasn't been Txed */
@@ -1163,8 +1169,8 @@ static int epic_rx(struct net_device *dev, int budget)
 		rx_work_limit = budget;
 
 	/* If we own the next entry, it's a new packet. Send it up. */
-	while ((ep->rx_ring[entry].rxstatus & cpu_to_le32(DescOwn)) == 0) {
-		int status = le32_to_cpu(ep->rx_ring[entry].rxstatus);
+	while ((ep->rx_ring[entry].rxstatus & DescOwn) == 0) {
+		int status = ep->rx_ring[entry].rxstatus;
 
 		if (debug > 4)
 			printk(KERN_DEBUG "  epic_rx() status was %8.8x.\n", status);
@@ -1238,7 +1244,8 @@ static int epic_rx(struct net_device *dev, int budget)
 				skb->data, ep->rx_buf_sz, PCI_DMA_FROMDEVICE);
 			work_done++;
 		}
-		ep->rx_ring[entry].rxstatus = cpu_to_le32(DescOwn);
+		/* AV: shouldn't we add a barrier here? */
+		ep->rx_ring[entry].rxstatus = DescOwn;
 	}
 	return work_done;
 }
diff --git a/drivers/net/fec.c b/drivers/net/fec.c
index 0fbf1bbbaee..d7a3ea88edd 100644
--- a/drivers/net/fec.c
+++ b/drivers/net/fec.c
@@ -1253,7 +1253,7 @@ static void __inline__ fec_request_intrs(struct net_device *dev)
 
 	/* Setup interrupt handlers. */
 	for (idp = id; idp->name; idp++) {
-		if (request_irq(idp->irq, idp->handler, 0, idp->name, dev) != 0)
+		if (request_irq(idp->irq, idp->handler, IRQF_DISABLED, idp->name, dev) != 0)
 			printk("FEC: Could not allocate %s IRQ(%d)!\n", idp->name, idp->irq);
 	}
 
@@ -1382,7 +1382,7 @@ static void __inline__ fec_request_intrs(struct net_device *dev)
 
 	/* Setup interrupt handlers. */
 	for (idp = id; idp->name; idp++) {
-		if (request_irq(b+idp->irq, fec_enet_interrupt, 0, idp->name, dev) != 0)
+		if (request_irq(b+idp->irq, fec_enet_interrupt, IRQF_DISABLED, idp->name, dev) != 0)
 			printk("FEC: Could not allocate %s IRQ(%d)!\n", idp->name, b+idp->irq);
 	}
 
@@ -1553,7 +1553,7 @@ static void __inline__ fec_request_intrs(struct net_device *dev)
 
 	/* Setup interrupt handlers. */
 	for (idp = id; idp->name; idp++) {
-		if (request_irq(b+idp->irq,fec_enet_interrupt,0,idp->name,dev)!=0)
+		if (request_irq(b+idp->irq, fec_enet_interrupt, IRQF_DISABLED, idp->name,dev) != 0)
 			printk("FEC: Could not allocate %s IRQ(%d)!\n", idp->name, b+idp->irq);
 	}
 
@@ -1680,7 +1680,7 @@ static void __inline__ fec_request_intrs(struct net_device *dev)
 
 	/* Setup interrupt handlers. */
 	for (idp = id; idp->name; idp++) {
-		if (request_irq(b+idp->irq,fec_enet_interrupt,0,idp->name,dev)!=0)
+		if (request_irq(b+idp->irq, fec_enet_interrupt, IRQF_DISABLED, idp->name,dev) != 0)
 			printk("FEC: Could not allocate %s IRQ(%d)!\n",
 				idp->name, b+idp->irq);
 	}
diff --git a/drivers/net/forcedeth.c b/drivers/net/forcedeth.c
index 801b4d9cd97..6f7e3fde9e7 100644
--- a/drivers/net/forcedeth.c
+++ b/drivers/net/forcedeth.c
@@ -184,6 +184,7 @@
 #define DEV_HAS_PAUSEFRAME_TX_V1   0x08000  /* device supports tx pause frames version 1 */
 #define DEV_HAS_PAUSEFRAME_TX_V2   0x10000  /* device supports tx pause frames version 2 */
 #define DEV_HAS_PAUSEFRAME_TX_V3   0x20000  /* device supports tx pause frames version 3 */
+#define DEV_NEED_TX_LIMIT          0x40000  /* device needs to limit tx */
 
 enum {
 	NvRegIrqStatus = 0x000,
@@ -635,6 +636,8 @@ union ring_type {
 #define NV_RESTART_TX         0x1
 #define NV_RESTART_RX         0x2
 
+#define NV_TX_LIMIT_COUNT     16
+
 /* statistics */
 struct nv_ethtool_str {
 	char name[ETH_GSTRING_LEN];
@@ -743,6 +746,8 @@ struct nv_skb_map {
 	struct sk_buff *skb;
 	dma_addr_t dma;
 	unsigned int dma_len;
+	struct ring_desc_ex *first_tx_desc;
+	struct nv_skb_map *next_tx_ctx;
 };
 
 /*
@@ -827,6 +832,10 @@ struct fe_priv {
 	union ring_type tx_ring;
 	u32 tx_flags;
 	int tx_ring_size;
+	int tx_limit;
+	u32 tx_pkts_in_progress;
+	struct nv_skb_map *tx_change_owner;
+	struct nv_skb_map *tx_end_flip;
 	int tx_stop;
 
 	/* vlan fields */
@@ -1707,6 +1716,9 @@ static void nv_init_tx(struct net_device *dev)
 		np->last_tx.ex = &np->tx_ring.ex[np->tx_ring_size-1];
 	np->get_tx_ctx = np->put_tx_ctx = np->first_tx_ctx = np->tx_skb;
 	np->last_tx_ctx = &np->tx_skb[np->tx_ring_size-1];
+	np->tx_pkts_in_progress = 0;
+	np->tx_change_owner = NULL;
+	np->tx_end_flip = NULL;
 
 	for (i = 0; i < np->tx_ring_size; i++) {
 		if (np->desc_ver == DESC_VER_1 || np->desc_ver == DESC_VER_2) {
@@ -1720,6 +1732,9 @@ static void nv_init_tx(struct net_device *dev)
 		}
 		np->tx_skb[i].skb = NULL;
 		np->tx_skb[i].dma = 0;
+		np->tx_skb[i].dma_len = 0;
+		np->tx_skb[i].first_tx_desc = NULL;
+		np->tx_skb[i].next_tx_ctx = NULL;
 	}
 }
 
@@ -1771,7 +1786,14 @@ static void nv_drain_tx(struct net_device *dev)
 		}
 		if (nv_release_txskb(dev, &np->tx_skb[i]))
 			dev->stats.tx_dropped++;
+		np->tx_skb[i].dma = 0;
+		np->tx_skb[i].dma_len = 0;
+		np->tx_skb[i].first_tx_desc = NULL;
+		np->tx_skb[i].next_tx_ctx = NULL;
 	}
+	np->tx_pkts_in_progress = 0;
+	np->tx_change_owner = NULL;
+	np->tx_end_flip = NULL;
 }
 
 static void nv_drain_rx(struct net_device *dev)
@@ -1948,6 +1970,7 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
 	struct ring_desc_ex* start_tx;
 	struct ring_desc_ex* prev_tx;
 	struct nv_skb_map* prev_tx_ctx;
+	struct nv_skb_map* start_tx_ctx;
 
 	/* add fragments to entries count */
 	for (i = 0; i < fragments; i++) {
@@ -1965,6 +1988,7 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
 	}
 
 	start_tx = put_tx = np->put_tx.ex;
+	start_tx_ctx = np->put_tx_ctx;
 
 	/* setup the header buffer */
 	do {
@@ -2037,6 +2061,26 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
 
 	spin_lock_irq(&np->lock);
 
+	if (np->tx_limit) {
+		/* Limit the number of outstanding tx. Setup all fragments, but
+		 * do not set the VALID bit on the first descriptor. Save a pointer
+		 * to that descriptor and also for next skb_map element.
+		 */
+
+		if (np->tx_pkts_in_progress == NV_TX_LIMIT_COUNT) {
+			if (!np->tx_change_owner)
+				np->tx_change_owner = start_tx_ctx;
+
+			/* remove VALID bit */
+			tx_flags &= ~NV_TX2_VALID;
+			start_tx_ctx->first_tx_desc = start_tx;
+			start_tx_ctx->next_tx_ctx = np->put_tx_ctx;
+			np->tx_end_flip = np->put_tx_ctx;
+		} else {
+			np->tx_pkts_in_progress++;
+		}
+	}
+
 	/* set tx flags */
 	start_tx->flaglen |= cpu_to_le32(tx_flags | tx_flags_extra);
 	np->put_tx.ex = put_tx;
@@ -2060,6 +2104,25 @@ static int nv_start_xmit_optimized(struct sk_buff *skb, struct net_device *dev)
 	return NETDEV_TX_OK;
 }
 
+static inline void nv_tx_flip_ownership(struct net_device *dev)
+{
+	struct fe_priv *np = netdev_priv(dev);
+
+	np->tx_pkts_in_progress--;
+	if (np->tx_change_owner) {
+		__le32 flaglen = le32_to_cpu(np->tx_change_owner->first_tx_desc->flaglen);
+		flaglen |= NV_TX2_VALID;
+		np->tx_change_owner->first_tx_desc->flaglen = cpu_to_le32(flaglen);
+		np->tx_pkts_in_progress++;
+
+		np->tx_change_owner = np->tx_change_owner->next_tx_ctx;
+		if (np->tx_change_owner == np->tx_end_flip)
+			np->tx_change_owner = NULL;
+
+		writel(NVREG_TXRXCTL_KICK|np->txrxctl_bits, get_hwbase(dev) + NvRegTxRxControl);
+	}
+}
+
 /*
  * nv_tx_done: check for completed packets, release the skbs.
  *
@@ -2147,6 +2210,10 @@ static void nv_tx_done_optimized(struct net_device *dev, int limit)
 				dev->stats.tx_packets++;
 			dev_kfree_skb_any(np->get_tx_ctx->skb);
 			np->get_tx_ctx->skb = NULL;
+
+			if (np->tx_limit) {
+				nv_tx_flip_ownership(dev);
+			}
 		}
 		if (unlikely(np->get_tx.ex++ == np->last_tx.ex))
 			np->get_tx.ex = np->first_tx.ex;
@@ -5333,6 +5400,21 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
 		np->need_linktimer = 0;
 	}
 
+	/* Limit the number of tx's outstanding for hw bug */
+	if (id->driver_data & DEV_NEED_TX_LIMIT) {
+		np->tx_limit = 1;
+		if ((id->device == PCI_DEVICE_ID_NVIDIA_NVENET_32 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_33 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_34 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_35 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_36 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_37 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_38 ||
+		     id->device == PCI_DEVICE_ID_NVIDIA_NVENET_39) &&
+		    pci_dev->revision >= 0xA2)
+			np->tx_limit = 0;
+	}
+
 	/* clear phy state and temporarily halt phy interrupts */
 	writel(0, base + NvRegMIIMask);
 	phystate = readl(base + NvRegAdapterControl);
@@ -5563,19 +5645,19 @@ static struct pci_device_id pci_tbl[] = {
 	},
 	{	/* CK804 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_8),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1|DEV_NEED_TX_LIMIT,
 	},
 	{	/* CK804 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_9),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP04 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_10),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP04 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_11),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_STATISTICS_V1|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP51 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_12),
@@ -5587,11 +5669,11 @@ static struct pci_device_id pci_tbl[] = {
 	},
 	{	/* MCP55 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_14),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_VLAN|DEV_HAS_MSI|DEV_HAS_MSI_X|DEV_HAS_POWER_CNTRL|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_VLAN|DEV_HAS_MSI|DEV_HAS_MSI_X|DEV_HAS_POWER_CNTRL|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP55 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_15),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_VLAN|DEV_HAS_MSI|DEV_HAS_MSI_X|DEV_HAS_POWER_CNTRL|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_CHECKSUM|DEV_HAS_HIGH_DMA|DEV_HAS_VLAN|DEV_HAS_MSI|DEV_HAS_MSI_X|DEV_HAS_POWER_CNTRL|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP61 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_16),
@@ -5611,19 +5693,19 @@ static struct pci_device_id pci_tbl[] = {
 	},
 	{	/* MCP65 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_20),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_HIGH_DMA|DEV_HAS_POWER_CNTRL|DEV_HAS_MSI|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT|DEV_HAS_CORRECT_MACADDR,
+		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_HIGH_DMA|DEV_HAS_POWER_CNTRL|DEV_HAS_MSI|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT|DEV_HAS_CORRECT_MACADDR|DEV_NEED_TX_LIMIT|DEV_NEED_TX_LIMIT,
 	},
 	{	/* MCP65 Ethernet Controller */
 		PCI_DEVICE(PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NVENET_21),
-		.driver_data = DEV_NEED_TIMERIRQ|DEV_NEED_LINKTIMER|DEV_HAS_LARGEDESC|DEV_HAS_HIGH_DMA|DEV_HAS_POWER_CNTRL|DEV_HAS_MSI|DEV_HAS_PAUSEFRAME_TX_V1|DEV_HAS_STATISTICS_V2|DEV_HAS_TEST_EXTENDED|DEV_HAS_MGMT_UNIT|DEV_HAS_CORRECT_MACADDR,