Merge heisenberg.transvirtual.com:/tmp/linus-2.5

into heisenberg.transvirtual.com:/tmp/fbdev-2.5

Documentation/networking/dl2k.txt

 
     D-Link DL2000-based Gigabit Ethernet Adapter Installation
     for Linux
-    Jan 29, 2002
+    May 23, 2002
 
 Contents
 ========
@@ -202,7 +202,7 @@ media=media_type		- Specifies the media type the NIC operates at.
 				  1000mbps_fd and 1000mbps_hd types are only
 				  available for fiber adapter.
 
-vlan=[0|1]			- Specifies the VLAN ID. If vlan=0, the
+vlan=n				- Specifies the VLAN ID. If vlan=0, the
 				  Virtual Local Area Network (VLAN) function is
 				  disable.
 
@@ -211,24 +211,34 @@ jumbo=[0|1]			- Specifies the jumbo frame support. If jumbo=1,
 				  function is disabled.
 				  Jumbo frame usually improve the performance
 				  int gigabit.
+				  This feature need jumbo frame compatible 
+				  remote.
 				  
-rx_coalesce=n			- Rx frame count each interrupt.
-rx_timeout=n			- Rx DMA wait time for an interrupt. Proper 
-				  values of rx_coalesce and rx_timeout bring 
-				  a conspicuous performance in the fast machine.
-				  Ex. rx_coalesce=5 and rx_timeout=750 
-
-tx_coalesce=n			- Tx transmit count each TxComp interrupt.
-				  Setting value larger than 1 will improve 
-				  performance, but this is possible to lower 
-				  stability in slow UP machines. By default, 
-				  tx_coalesce=1. (dl2k)
+rx_coalesce=m			- Number of rx frame handled each interrupt.
+rx_timeout=n			- Rx DMA wait time for an interrupt. 
+				  If set rx_coalesce > 0, hardware only assert 
+				  an interrupt for m frames. Hardware won't 
+				  assert rx interrupt until m frames received or
+				  reach timeout of n * 640 nano seconds. 
+				  Set proper rx_coalesce and rx_timeout can 
+				  reduce congestion collapse and overload which
+				  has been a bottlenect for high speed network.
 				  
-tx_flow=[1|0]			- Specifies the Tx flow control. If tx_flow=1, 
-				  the Tx flow control enable.
+				  For example, rx_coalesce=10 rx_timeout=800.
+				  that is, hardware assert only 1 interrupt 
+				  for 10 frames received or timeout of 512 us. 
+
+tx_coalesce=n			- Number of tx frame handled each interrupt.
+				  Set n > 1 can reduce the interrupts 
+				  congestion usually lower performance of
+				  high speed network card. Default is 16.
 				  
-rx_flow=[1|0]			- Specifies the Rx flow control. If rx_flow=1, 
-				  the Rx flow control enable.
+tx_flow=[1|0]			- Specifies the Tx flow control. If tx_flow=0, 
+				  the Tx flow control disable else driver
+				  autodetect.
+rx_flow=[1|0]			- Specifies the Rx flow control. If rx_flow=0, 
+				  the Rx flow control enable else driver
+				  autodetect.
 
 
 Configuration Script Sample

Makefile

@@ -607,7 +607,7 @@ clean:	archclean
 
 mrproper: clean archmrproper
 	@echo 'Making mrproper'
-	@find . \( -size 0 -o -name .depend -o -name .\*.cmd \) \
+	@find . \( -name .depend -o -name .\*.cmd \) \
 		   -type f -print | xargs rm -f
 	@rm -f $(MRPROPER_FILES)
 	@rm -rf $(MRPROPER_DIRS)

drivers/net/8139cp.c

@@ -100,13 +100,16 @@ MODULE_PARM_DESC (multicast_filter_limit, "8139cp maximum number of filtered mul
 #define CP_DEF_MSG_ENABLE	(NETIF_MSG_DRV		| \
 				 NETIF_MSG_PROBE 	| \
 				 NETIF_MSG_LINK)
+#define CP_NUM_STATS		14	/* struct cp_dma_stats, plus one */
+#define CP_STATS_SIZE		64	/* size in bytes of DMA stats block */
 #define CP_REGS_SIZE		(0xff + 1)
 #define CP_REGS_VER		1		/* version 1 */
 #define CP_RX_RING_SIZE		64
 #define CP_TX_RING_SIZE		64
 #define CP_RING_BYTES		\
 		((sizeof(struct cp_desc) * CP_RX_RING_SIZE) +	\
-		(sizeof(struct cp_desc) * CP_TX_RING_SIZE))
+		 (sizeof(struct cp_desc) * CP_TX_RING_SIZE) +	\
+		 CP_STATS_SIZE)
 #define NEXT_TX(N)		(((N) + 1) & (CP_TX_RING_SIZE - 1))
 #define NEXT_RX(N)		(((N) + 1) & (CP_RX_RING_SIZE - 1))
 #define TX_BUFFS_AVAIL(CP)					\
@@ -135,6 +138,7 @@ enum {
 	/* NIC register offsets */
 	MAC0		= 0x00,	/* Ethernet hardware address. */
 	MAR0		= 0x08,	/* Multicast filter. */
+	StatsAddr	= 0x10,	/* 64-bit start addr of 64-byte DMA stats blk */
 	TxRingAddr	= 0x20, /* 64-bit start addr of Tx ring */
 	HiTxRingAddr	= 0x28, /* 64-bit start addr of high priority Tx ring */
 	Cmd		= 0x37, /* Command register */
@@ -197,6 +201,9 @@ enum {
 	RxErrLong	= (1 << 21), /* Rx error, packet > 4096 bytes */
 	RxErrFIFO	= (1 << 22), /* Rx error, FIFO overflowed, pkt bad */
 
+	/* StatsAddr register */
+	DumpStats	= (1 << 3),  /* Begin stats dump */
+
 	/* RxConfig register */
 	RxCfgFIFOShift	= 13,	     /* Shift, to get Rx FIFO thresh value */
 	RxCfgDMAShift	= 8,	     /* Shift, to get Rx Max DMA value */
@@ -284,6 +291,22 @@ struct ring_info {
 	unsigned		frag;
 };
 
+struct cp_dma_stats {
+	u64			tx_ok;
+	u64			rx_ok;
+	u64			tx_err;
+	u32			rx_err;
+	u16			rx_fifo;
+	u16			frame_align;
+	u32			tx_ok_1col;
+	u32			tx_ok_mcol;
+	u64			rx_ok_phys;
+	u64			rx_ok_bcast;
+	u32			rx_ok_mcast;
+	u16			tx_abort;
+	u16			tx_underrun;
+} __attribute__((packed));
+
 struct cp_extra_stats {
 	unsigned long		rx_frags;
 };
@@ -312,6 +335,8 @@ struct cp_private {
 
 	struct net_device_stats net_stats;
 	struct cp_extra_stats	cp_stats;
+	struct cp_dma_stats	*nic_stats;
+	dma_addr_t		nic_stats_dma;
 
 	struct pci_dev		*pdev;
 	u32			rx_config;
@@ -373,6 +398,26 @@ static struct pci_device_id cp_pci_tbl[] __devinitdata = {
 };
 MODULE_DEVICE_TABLE(pci, cp_pci_tbl);
 
+static struct {
+	const char str[ETH_GSTRING_LEN];
+} ethtool_stats_keys[] = {
+	{ "tx_ok" },
+	{ "rx_ok" },
+	{ "tx_err" },
+	{ "rx_err" },
+	{ "rx_fifo" },
+	{ "frame_align" },
+	{ "tx_ok_1col" },
+	{ "tx_ok_mcol" },
+	{ "rx_ok_phys" },
+	{ "rx_ok_bcast" },
+	{ "rx_ok_mcast" },
+	{ "tx_abort" },
+	{ "tx_underrun" },
+	{ "rx_frags" },
+};
+
+
 static inline void cp_set_rxbufsize (struct cp_private *cp)
 {
 	unsigned int mtu = cp->dev->mtu;
@@ -949,9 +994,9 @@ static void cp_init_hw (struct cp_private *cp)
 	cpw32_f(HiTxRingAddr + 4, 0);
 
 	cpw32_f(RxRingAddr, cp->ring_dma);
-	cpw32_f(RxRingAddr + 4, 0);
+	cpw32_f(RxRingAddr + 4, 0);		/* FIXME: 64-bit PCI */
 	cpw32_f(TxRingAddr, cp->ring_dma + (sizeof(struct cp_desc) * CP_RX_RING_SIZE));
-	cpw32_f(TxRingAddr + 4, 0);
+	cpw32_f(TxRingAddr + 4, 0);		/* FIXME: 64-bit PCI */
 
 	cpw16(MultiIntr, 0);
 
@@ -1010,10 +1055,19 @@ static int cp_init_rings (struct cp_private *cp)
 
 static int cp_alloc_rings (struct cp_private *cp)
 {
-	cp->rx_ring = pci_alloc_consistent(cp->pdev, CP_RING_BYTES, &cp->ring_dma);
-	if (!cp->rx_ring)
+	void *mem;
+
+	mem = pci_alloc_consistent(cp->pdev, CP_RING_BYTES, &cp->ring_dma);
+	if (!mem)
 		return -ENOMEM;
+
+	cp->rx_ring = mem;
 	cp->tx_ring = &cp->rx_ring[CP_RX_RING_SIZE];
+
+	mem += (CP_RING_BYTES - CP_STATS_SIZE);
+	cp->nic_stats = mem;
+	cp->nic_stats_dma = cp->ring_dma + (CP_RING_BYTES - CP_STATS_SIZE);
+
 	return cp_init_rings(cp);
 }
 
@@ -1052,6 +1106,7 @@ static void cp_free_rings (struct cp_private *cp)
 	pci_free_consistent(cp->pdev, CP_RING_BYTES, cp->rx_ring, cp->ring_dma);
 	cp->rx_ring = NULL;
 	cp->tx_ring = NULL;
+	cp->nic_stats = NULL;
 }
 
 static int cp_open (struct net_device *dev)
@@ -1181,6 +1236,7 @@ static int cp_ethtool_ioctl (struct cp_private *cp, void *useraddr)
 		strcpy (info.version, DRV_VERSION);
 		strcpy (info.bus_info, cp->pdev->slot_name);
 		info.regdump_len = CP_REGS_SIZE;
+		info.n_stats = CP_NUM_STATS;
 		if (copy_to_user (useraddr, &info, sizeof (info)))
 			return -EFAULT;
 		return 0;
@@ -1359,6 +1415,83 @@ static int cp_ethtool_ioctl (struct cp_private *cp, void *useraddr)
 		return 0;
 	}
 
+	/* get string list(s) */
+	case ETHTOOL_GSTRINGS: {
+		struct ethtool_gstrings estr = { ETHTOOL_GSTRINGS };
+
+		if (copy_from_user(&estr, useraddr, sizeof(estr)))
+			return -EFAULT;
+		if (estr.string_set != ETH_SS_STATS)
+			return -EINVAL;
+
+		estr.len = CP_NUM_STATS;
+		if (copy_to_user(useraddr, &estr, sizeof(estr)))
+			return -EFAULT;
+		if (copy_to_user(useraddr + sizeof(estr),
+				 &ethtool_stats_keys,
+				 sizeof(ethtool_stats_keys)))
+			return -EFAULT;
+		return 0;
+	}
+
+	/* get NIC-specific statistics */
+	case ETHTOOL_GSTATS: {
+		struct ethtool_stats estats = { ETHTOOL_GSTATS };
+		u64 *tmp_stats;
+		unsigned int work = 100;
+		const unsigned int sz = sizeof(u64) * CP_NUM_STATS;
+		int i;
+
+		/* begin NIC statistics dump */
+		cpw32(StatsAddr + 4, 0); /* FIXME: 64-bit PCI */
+		cpw32(StatsAddr, cp->nic_stats_dma | DumpStats);
+		cpr32(StatsAddr);
+
+		estats.n_stats = CP_NUM_STATS;
+		if (copy_to_user(useraddr, &estats, sizeof(estats)))
+			return -EFAULT;
+
+		while (work-- > 0) {
+			if ((cpr32(StatsAddr) & DumpStats) == 0)
+				break;
+			cpu_relax();
+		}
+
+		if (cpr32(StatsAddr) & DumpStats)
+			return -EIO;
+
+		tmp_stats = kmalloc(sz, GFP_KERNEL);
+		if (!tmp_stats)
+			return -ENOMEM;
+		memset(tmp_stats, 0, sz);
+
+		i = 0;
+		tmp_stats[i++] = le64_to_cpu(cp->nic_stats->tx_ok);
+		tmp_stats[i++] = le64_to_cpu(cp->nic_stats->rx_ok);
+		tmp_stats[i++] = le64_to_cpu(cp->nic_stats->tx_err);
+		tmp_stats[i++] = le32_to_cpu(cp->nic_stats->rx_err);
+		tmp_stats[i++] = le16_to_cpu(cp->nic_stats->rx_fifo);
+		tmp_stats[i++] = le16_to_cpu(cp->nic_stats->frame_align);
+		tmp_stats[i++] = le32_to_cpu(cp->nic_stats->tx_ok_1col);
+		tmp_stats[i++] = le32_to_cpu(cp->nic_stats->tx_ok_mcol);
+		tmp_stats[i++] = le64_to_cpu(cp->nic_stats->rx_ok_phys);
+		tmp_stats[i++] = le64_to_cpu(cp->nic_stats->rx_ok_bcast);
+		tmp_stats[i++] = le32_to_cpu(cp->nic_stats->rx_ok_mcast);
+		tmp_stats[i++] = le16_to_cpu(cp->nic_stats->tx_abort);
+		tmp_stats[i++] = le16_to_cpu(cp->nic_stats->tx_underrun);
+		tmp_stats[i++] = cp->cp_stats.rx_frags;
+		if (i != CP_NUM_STATS)
+			BUG();
+
+		i = copy_to_user(useraddr + sizeof(estats),
+				 tmp_stats, sz);
+		kfree(tmp_stats);
+
+		if (i)
+			return -EFAULT;
+		return 0;
+	}
+
 	default:
 		break;
 	}

drivers/net/dl2k.c

 /*  D-Link DL2000-based Gigabit Ethernet Adapter Linux driver */
 /*
-    Copyright (c) 2001,2002 by D-Link Corporation
+    Copyright (c) 2001, 2002 by D-Link Corporation
     Written by Edward Peng.<edward_peng@dlink.com.tw>
     Created 03-May-2001, base on Linux' sundance.c.
 
@@ -33,12 +33,13 @@
     1.11	2002/05/23	Added ISR schedule scheme.
     				Fixed miscount of rx frame error for DGE-550SX.
     				Fixed VLAN bug.
+    1.12	2002/06/13	Lock tx_coalesce=1 on 10/100Mbps mode.
  */
 
 #include "dl2k.h"
 
 static char version[] __devinitdata =
-    KERN_INFO "D-Link DL2000-based linux driver v1.11 2002/05/23\n";
+    KERN_INFO "D-Link DL2000-based linux driver v1.12 2002/06/13\n";
 
 #define MAX_UNITS 8
 static int mtu[MAX_UNITS];
@@ -138,15 +139,15 @@ rio_probe1 (struct pci_dev *pdev, const struct pci_device_id *ent)
 	}
 	SET_MODULE_OWNER (dev);
 
-#ifdef USE_IO_OPS
-	ioaddr = pci_resource_start (pdev, 0);
-#else
+#ifdef MEM_MAPPING
 	ioaddr = pci_resource_start (pdev, 1);
 	ioaddr = (long) ioremap (ioaddr, RIO_IO_SIZE);
 	if (!ioaddr) {
 		err = -ENOMEM;
 		goto err_out_dev;
 	}
+#else
+	ioaddr = pci_resource_start (pdev, 0);
 #endif
 	dev->base_addr = ioaddr;
 	dev->irq = irq;
@@ -158,6 +159,7 @@ rio_probe1 (struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	/* Parse manual configuration */
 	np->an_enable = 1;
+	np->tx_coalesce = 1;
 	if (card_idx < MAX_UNITS) {
 		if (media[card_idx] != NULL) {
 			np->an_enable = 0;
@@ -213,7 +215,7 @@ rio_probe1 (struct pci_dev *pdev, const struct pci_device_id *ent)
 
 		if (tx_coalesce < 1)
 			tx_coalesce = 1;
-		if (tx_coalesce > TX_RING_SIZE-1)
+		else if (tx_coalesce > TX_RING_SIZE-1)
 			tx_coalesce = TX_RING_SIZE - 1;
 	}
 	dev->open = &rio_open;
@@ -303,7 +305,7 @@ rio_probe1 (struct pci_dev *pdev, const struct pci_device_id *ent)
       err_out_unmap_tx:
 	pci_free_consistent (pdev, TX_TOTAL_SIZE, np->tx_ring, np->tx_ring_dma);
       err_out_iounmap:
-#ifndef USE_IO_OPS
+#ifdef MEM_MAPPING
 	iounmap ((void *) ioaddr);
 
       err_out_dev:
@@ -361,7 +363,7 @@ parse_eeprom (struct net_device *dev)
 	}
 
 	/* Check CRC */
-	crc = ~ether_crc_le(256 - 4, sromdata);
+	crc = ~ether_crc_le (256 - 4, sromdata);
 	if (psrom->crc != crc) {
 		printk (KERN_ERR "%s: EEPROM data CRC error.\n", dev->name);
 		return -1;
@@ -636,7 +638,7 @@ start_xmit (struct sk_buff *skb, struct net_device *dev)
 
 	/* DL2K bug: DMA fails to get next descriptor ptr in 10Mbps mode
 	 * Work around: Always use 1 descriptor in 10Mbps mode */
-	if (entry % tx_coalesce == 0 || np->speed == 10)
+	if (entry % np->tx_coalesce == 0 || np->speed == 10)
 		txdesc->status = cpu_to_le64 (entry | tfc_vlan_tag |
 					      WordAlignDisable | 
 					      TxDMAIndicate |
@@ -936,6 +938,10 @@ rio_error (struct net_device *dev, int int_status)
 				mii_get_media_pcs (dev);
 			else
 				mii_get_media (dev);
+			if (np->speed == 1000)
+				np->tx_coalesce = tx_coalesce;
+			else 
+				np->tx_coalesce = 1;
 			macctrl = 0;
 			macctrl |= (np->vlan) ? AutoVLANuntagging : 0;
 			macctrl |= (np->full_duplex) ? DuplexSelect : 0;
@@ -1671,7 +1677,7 @@ rio_remove1 (struct pci_dev *pdev)
 				     np->rx_ring_dma);
 		pci_free_consistent (pdev, TX_TOTAL_SIZE, np->tx_ring,
 				     np->tx_ring_dma);
-#ifndef USE_IO_OPS
+#ifdef MEM_MAPPING
 		iounmap ((char *) (dev->base_addr));
 #endif
 		kfree (dev);

drivers/net/dl2k.h

@@ -33,15 +33,15 @@
 #include <linux/delay.h>
 #include <linux/spinlock.h>
 #include <linux/time.h>
-#define TX_RING_SIZE	128
-#define TX_QUEUE_LEN	120	/* Limit ring entries actually used.  */
-#define RX_RING_SIZE 	128
+#define TX_RING_SIZE	256
+#define TX_QUEUE_LEN	(TX_RING_SIZE - 1) /* Limit ring entries actually used.*/
+#define RX_RING_SIZE 	256
 #define TX_TOTAL_SIZE	TX_RING_SIZE*sizeof(struct netdev_desc)
 #define RX_TOTAL_SIZE	RX_RING_SIZE*sizeof(struct netdev_desc)
 
 /* This driver was written to use PCI memory space, however x86-oriented
    hardware often uses I/O space accesses. */
-#ifdef USE_IO_OPS
+#ifndef MEM_MAPPING
 #undef readb
 #undef readw
 #undef readl
@@ -658,6 +658,7 @@ struct netdev_private {
 	unsigned int chip_id;		/* PCI table chip id */
 	unsigned int rx_coalesce; 	/* Maximum frames each RxDMAComplete intr */
 	unsigned int rx_timeout; 	/* Wait time between RxDMAComplete intr */
+	unsigned int tx_coalesce;	/* Maximum frames each tx interrupt */
 	unsigned int full_duplex:1;	/* Full-duplex operation requested. */
 	unsigned int an_enable:2;	/* Auto-Negotiated Enable */
 	unsigned int jumbo:1;		/* Jumbo frame enable */
@@ -681,10 +682,10 @@ struct netdev_private {
 };
 
 /* The station address location in the EEPROM. */
-#ifdef USE_IO_OPS
-#define PCI_IOTYPE (PCI_USES_MASTER | PCI_USES_IO  | PCI_ADDR0)
-#else
+#ifdef MEM_MAPPING
 #define PCI_IOTYPE (PCI_USES_MASTER | PCI_USES_MEM | PCI_ADDR1)
+#else
+#define PCI_IOTYPE (PCI_USES_MASTER | PCI_USES_IO  | PCI_ADDR0)
 #endif
 /* The struct pci_device_id consist of:
         vendor, device          Vendor and device ID to match (or PCI_ANY_ID)

include/linux/ethtool.h

@@ -39,7 +39,8 @@ struct ethtool_drvinfo {
 	char	bus_info[ETHTOOL_BUSINFO_LEN];	/* Bus info for this IF. */
 				/* For PCI devices, use pci_dev->slot_name. */
 	char	reserved1[32];
-	char	reserved2[20];
+	char	reserved2[16];
+	u32	n_stats;	/* number of u64's from ETHTOOL_GSTATS */
 	u32	testinfo_len;
 	u32	eedump_len;	/* Size of data from ETHTOOL_GEEPROM (bytes) */
 	u32	regdump_len;	/* Size of data from ETHTOOL_GREGS (bytes) */
@@ -242,6 +243,13 @@ struct ethtool_test {
 	u64	data[0];
 };
 
+/* for dumping NIC-specific statistics */
+struct ethtool_stats {
+	u32	cmd;		/* ETHTOOL_GSTATS */
+	u32	n_stats;	/* number of u64's being returned */
+	u64	data[0];
+};
+
 /* CMDs currently supported */
 #define ETHTOOL_GSET		0x00000001 /* Get settings. */
 #define ETHTOOL_SSET		0x00000002 /* Set settings, privileged. */
@@ -272,6 +280,7 @@ struct ethtool_test {
 #define ETHTOOL_TEST		0x0000001a /* execute NIC self-test, priv. */
 #define ETHTOOL_GSTRINGS	0x0000001b /* get specified string set */
 #define ETHTOOL_PHYS_ID		0x0000001c /* identify the NIC */
+#define ETHTOOL_GSTATS		0x0000001d /* get NIC-specific statistics */
 
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET

include/linux/smp.h

@@ -86,6 +86,7 @@ extern volatile int smp_msg_id;
 #define smp_call_function(func,info,retry,wait)	({ 0; })
 static inline void smp_send_reschedule(int cpu) { }
 static inline void smp_send_reschedule_all(void) { }
+#define cpu_online_map				1
 #define cpu_online(cpu)				1
 #define num_online_cpus()			1
 #define __per_cpu_data

kernel/ksyms.c

@@ -459,6 +459,7 @@ EXPORT_SYMBOL(iomem_resource);
 
 /* process management */
 EXPORT_SYMBOL(complete_and_exit);
+EXPORT_SYMBOL(default_wake_function);
 EXPORT_SYMBOL(__wake_up);
 #if CONFIG_SMP
 EXPORT_SYMBOL_GPL(__wake_up_sync); /* internal use only */