patch-2.4.13 linux/drivers/message/i2o/i2o_lan.c
Next file: linux/drivers/message/i2o/i2o_lan.h
Previous file: linux/drivers/message/i2o/i2o_core.c
Back to the patch index
Back to the overall index
- Lines: 1580
- Date:
Mon Oct 22 08:39:56 2001
- Orig file:
v2.4.12/linux/drivers/message/i2o/i2o_lan.c
- Orig date:
Wed Dec 31 16:00:00 1969
diff -u --recursive --new-file v2.4.12/linux/drivers/message/i2o/i2o_lan.c linux/drivers/message/i2o/i2o_lan.c
@@ -0,0 +1,1579 @@
+/*
+ * drivers/i2o/i2o_lan.c
+ *
+ * I2O LAN CLASS OSM May 26th 2000
+ *
+ * (C) Copyright 1999, 2000 University of Helsinki,
+ * Department of Computer Science
+ *
+ * This code is still under development / test.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ *
+ * Authors: Auvo Häkkinen <Auvo.Hakkinen@cs.Helsinki.FI>
+ * Fixes: Juha Sievänen <Juha.Sievanen@cs.Helsinki.FI>
+ * Taneli Vähäkangas <Taneli.Vahakangas@cs.Helsinki.FI>
+ * Deepak Saxena <deepak@plexity.net>
+ *
+ * Tested: in FDDI environment (using SysKonnect's DDM)
+ * in Gigabit Eth environment (using SysKonnect's DDM)
+ * in Fast Ethernet environment (using Intel 82558 DDM)
+ *
+ * TODO: tests for other LAN classes (Token Ring, Fibre Channel)
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+
+#include <linux/pci.h>
+
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/fddidevice.h>
+#include <linux/trdevice.h>
+#include <linux/fcdevice.h>
+
+#include <linux/skbuff.h>
+#include <linux/if_arp.h>
+#include <linux/slab.h>
+#include <linux/init.h>
+#include <linux/spinlock.h>
+#include <linux/tqueue.h>
+#include <asm/io.h>
+
+#include <linux/errno.h>
+
+#include <linux/i2o.h>
+#include "i2o_lan.h"
+
+//#define DRIVERDEBUG
+#ifdef DRIVERDEBUG
+#define dprintk(s, args...) printk(s, ## args)
+#else
+#define dprintk(s, args...)
+#endif
+
+/* The following module parameters are used as default values
+ * for per interface values located in the net_device private area.
+ * Private values are changed via /proc filesystem.
+ */
+static u32 max_buckets_out = I2O_LAN_MAX_BUCKETS_OUT;
+static u32 bucket_thresh = I2O_LAN_BUCKET_THRESH;
+static u32 rx_copybreak = I2O_LAN_RX_COPYBREAK;
+static u8 tx_batch_mode = I2O_LAN_TX_BATCH_MODE;
+static u32 i2o_event_mask = I2O_LAN_EVENT_MASK;
+
+#define MAX_LAN_CARDS 16
+static struct net_device *i2o_landevs[MAX_LAN_CARDS+1];
+static int unit = -1; /* device unit number */
+
+static void i2o_lan_reply(struct i2o_handler *h, struct i2o_controller *iop, struct i2o_message *m);
+static void i2o_lan_send_post_reply(struct i2o_handler *h, struct i2o_controller *iop, struct i2o_message *m);
+static int i2o_lan_receive_post(struct net_device *dev);
+static void i2o_lan_receive_post_reply(struct i2o_handler *h, struct i2o_controller *iop, struct i2o_message *m);
+static void i2o_lan_release_buckets(struct net_device *dev, u32 *msg);
+
+static int i2o_lan_reset(struct net_device *dev);
+static void i2o_lan_handle_event(struct net_device *dev, u32 *msg);
+
+/* Structures to register handlers for the incoming replies. */
+
+static struct i2o_handler i2o_lan_send_handler = {
+ i2o_lan_send_post_reply, // For send replies
+ NULL,
+ NULL,
+ NULL,
+ "I2O LAN OSM send",
+ -1,
+ I2O_CLASS_LAN
+};
+static int lan_send_context;
+
+static struct i2o_handler i2o_lan_receive_handler = {
+ i2o_lan_receive_post_reply, // For receive replies
+ NULL,
+ NULL,
+ NULL,
+ "I2O LAN OSM receive",
+ -1,
+ I2O_CLASS_LAN
+};
+static int lan_receive_context;
+
+static struct i2o_handler i2o_lan_handler = {
+ i2o_lan_reply, // For other replies
+ NULL,
+ NULL,
+ NULL,
+ "I2O LAN OSM",
+ -1,
+ I2O_CLASS_LAN
+};
+static int lan_context;
+
+DECLARE_TASK_QUEUE(i2o_post_buckets_task);
+struct tq_struct run_i2o_post_buckets_task = {
+ routine: (void (*)(void *)) run_task_queue,
+ data: (void *) 0
+};
+
+/* Functions to handle message failures and transaction errors:
+==============================================================*/
+
+/*
+ * i2o_lan_handle_failure(): Fail bit has been set since IOP's message
+ * layer cannot deliver the request to the target, or the target cannot
+ * process the request.
+ */
+static void i2o_lan_handle_failure(struct net_device *dev, u32 *msg)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+
+ u32 *preserved_msg = (u32*)(iop->mem_offset + msg[7]);
+ u32 *sgl_elem = &preserved_msg[4];
+ struct sk_buff *skb = NULL;
+ u8 le_flag;
+
+ i2o_report_status(KERN_INFO, dev->name, msg);
+
+ /* If PacketSend failed, free sk_buffs reserved by upper layers */
+
+ if (msg[1] >> 24 == LAN_PACKET_SEND) {
+ do {
+ skb = (struct sk_buff *)(sgl_elem[1]);
+ dev_kfree_skb_irq(skb);
+
+ atomic_dec(&priv->tx_out);
+
+ le_flag = *sgl_elem >> 31;
+ sgl_elem +=3;
+ } while (le_flag == 0); /* Last element flag not set */
+
+ if (netif_queue_stopped(dev))
+ netif_wake_queue(dev);
+ }
+
+ /* If ReceivePost failed, free sk_buffs we have reserved */
+
+ if (msg[1] >> 24 == LAN_RECEIVE_POST) {
+ do {
+ skb = (struct sk_buff *)(sgl_elem[1]);
+ dev_kfree_skb_irq(skb);
+
+ atomic_dec(&priv->buckets_out);
+
+ le_flag = *sgl_elem >> 31;
+ sgl_elem +=3;
+ } while (le_flag == 0); /* Last element flag not set */
+ }
+
+ /* Release the preserved msg frame by resubmitting it as a NOP */
+
+ preserved_msg[0] = THREE_WORD_MSG_SIZE | SGL_OFFSET_0;
+ preserved_msg[1] = I2O_CMD_UTIL_NOP << 24 | HOST_TID << 12 | 0;
+ preserved_msg[2] = 0;
+ i2o_post_message(iop, msg[7]);
+}
+/*
+ * i2o_lan_handle_transaction_error(): IOP or DDM has rejected the request
+ * for general cause (format error, bad function code, insufficient resources,
+ * etc.). We get one transaction_error for each failed transaction.
+ */
+static void i2o_lan_handle_transaction_error(struct net_device *dev, u32 *msg)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct sk_buff *skb;
+
+ i2o_report_status(KERN_INFO, dev->name, msg);
+
+ /* If PacketSend was rejected, free sk_buff reserved by upper layers */
+
+ if (msg[1] >> 24 == LAN_PACKET_SEND) {
+ skb = (struct sk_buff *)(msg[3]); // TransactionContext
+ dev_kfree_skb_irq(skb);
+ atomic_dec(&priv->tx_out);
+
+ if (netif_queue_stopped(dev))
+ netif_wake_queue(dev);
+ }
+
+ /* If ReceivePost was rejected, free sk_buff we have reserved */
+
+ if (msg[1] >> 24 == LAN_RECEIVE_POST) {
+ skb = (struct sk_buff *)(msg[3]);
+ dev_kfree_skb_irq(skb);
+ atomic_dec(&priv->buckets_out);
+ }
+}
+
+/*
+ * i2o_lan_handle_status(): Common parts of handling a not succeeded request
+ * (status != SUCCESS).
+ */
+static int i2o_lan_handle_status(struct net_device *dev, u32 *msg)
+{
+ /* Fail bit set? */
+
+ if (msg[0] & MSG_FAIL) {
+ i2o_lan_handle_failure(dev, msg);
+ return -1;
+ }
+
+ /* Message rejected for general cause? */
+
+ if ((msg[4]>>24) == I2O_REPLY_STATUS_TRANSACTION_ERROR) {
+ i2o_lan_handle_transaction_error(dev, msg);
+ return -1;
+ }
+
+ /* Else have to handle it in the callback function */
+
+ return 0;
+}
+
+/* Callback functions called from the interrupt routine:
+=======================================================*/
+
+/*
+ * i2o_lan_send_post_reply(): Callback function to handle PostSend replies.
+ */
+static void i2o_lan_send_post_reply(struct i2o_handler *h,
+ struct i2o_controller *iop, struct i2o_message *m)
+{
+ u32 *msg = (u32 *)m;
+ u8 unit = (u8)(msg[2]>>16); // InitiatorContext
+ struct net_device *dev = i2o_landevs[unit];
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ u8 trl_count = msg[3] & 0x000000FF;
+
+ if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
+ if (i2o_lan_handle_status(dev, msg))
+ return;
+ }
+
+#ifdef DRIVERDEBUG
+ i2o_report_status(KERN_INFO, dev->name, msg);
+#endif
+
+ /* DDM has handled transmit request(s), free sk_buffs.
+ * We get similar single transaction reply also in error cases
+ * (except if msg failure or transaction error).
+ */
+ while (trl_count) {
+ dev_kfree_skb_irq((struct sk_buff *)msg[4 + trl_count]);
+ dprintk(KERN_INFO "%s: tx skb freed (trl_count=%d).\n",
+ dev->name, trl_count);
+ atomic_dec(&priv->tx_out);
+ trl_count--;
+ }
+
+ /* If priv->tx_out had reached tx_max_out, the queue was stopped */
+
+ if (netif_queue_stopped(dev))
+ netif_wake_queue(dev);
+}
+
+/*
+ * i2o_lan_receive_post_reply(): Callback function to process incoming packets.
+ */
+static void i2o_lan_receive_post_reply(struct i2o_handler *h,
+ struct i2o_controller *iop, struct i2o_message *m)
+{
+ u32 *msg = (u32 *)m;
+ u8 unit = (u8)(msg[2]>>16); // InitiatorContext
+ struct net_device *dev = i2o_landevs[unit];
+
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_bucket_descriptor *bucket = (struct i2o_bucket_descriptor *)&msg[6];
+ struct i2o_packet_info *packet;
+ u8 trl_count = msg[3] & 0x000000FF;
+ struct sk_buff *skb, *old_skb;
+ unsigned long flags = 0;
+
+ if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
+ if (i2o_lan_handle_status(dev, msg))
+ return;
+
+ i2o_lan_release_buckets(dev, msg);
+ return;
+ }
+
+#ifdef DRIVERDEBUG
+ i2o_report_status(KERN_INFO, dev->name, msg);
+#endif
+
+ /* Else we are receiving incoming post. */
+
+ while (trl_count--) {
+ skb = (struct sk_buff *)bucket->context;
+ packet = (struct i2o_packet_info *)bucket->packet_info;
+ atomic_dec(&priv->buckets_out);
+
+ /* Sanity checks: Any weird characteristics in bucket? */
+
+ if (packet->flags & 0x0f || ! packet->flags & 0x40) {
+ if (packet->flags & 0x01)
+ printk(KERN_WARNING "%s: packet with errors, error code=0x%02x.\n",
+ dev->name, packet->status & 0xff);
+
+ /* The following shouldn't happen, unless parameters in
+ * LAN_OPERATION group are changed during the run time.
+ */
+ if (packet->flags & 0x0c)
+ printk(KERN_DEBUG "%s: multi-bucket packets not supported!\n",
+ dev->name);
+
+ if (! packet->flags & 0x40)
+ printk(KERN_DEBUG "%s: multiple packets in a bucket not supported!\n",
+ dev->name);
+
+ dev_kfree_skb_irq(skb);
+
+ bucket++;
+ continue;
+ }
+
+ /* Copy short packet to a new skb */
+
+ if (packet->len < priv->rx_copybreak) {
+ old_skb = skb;
+ skb = (struct sk_buff *)dev_alloc_skb(packet->len+2);
+ if (skb == NULL) {
+ printk(KERN_ERR "%s: Can't allocate skb.\n", dev->name);
+ return;
+ }
+ skb_reserve(skb, 2);
+ memcpy(skb_put(skb, packet->len), old_skb->data, packet->len);
+
+ spin_lock_irqsave(&priv->fbl_lock, flags);
+ if (priv->i2o_fbl_tail < I2O_LAN_MAX_BUCKETS_OUT)
+ priv->i2o_fbl[++priv->i2o_fbl_tail] = old_skb;
+ else
+ dev_kfree_skb_irq(old_skb);
+
+ spin_unlock_irqrestore(&priv->fbl_lock, flags);
+ } else
+ skb_put(skb, packet->len);
+
+ /* Deliver to upper layers */
+
+ skb->dev = dev;
+ skb->protocol = priv->type_trans(skb, dev);
+ netif_rx(skb);
+
+ dev->last_rx = jiffies;
+
+ dprintk(KERN_INFO "%s: Incoming packet (%d bytes) delivered "
+ "to upper level.\n", dev->name, packet->len);
+
+ bucket++; // to next Packet Descriptor Block
+ }
+
+#ifdef DRIVERDEBUG
+ if (msg[5] == 0)
+ printk(KERN_INFO "%s: DDM out of buckets (priv->count = %d)!\n",
+ dev->name, atomic_read(&priv->buckets_out));
+#endif
+
+ /* If DDM has already consumed bucket_thresh buckets, post new ones */
+
+ if (atomic_read(&priv->buckets_out) <= priv->max_buckets_out - priv->bucket_thresh) {
+ run_i2o_post_buckets_task.data = (void *)dev;
+ queue_task(&run_i2o_post_buckets_task, &tq_immediate);
+ mark_bh(IMMEDIATE_BH);
+ }
+
+ return;
+}
+
+/*
+ * i2o_lan_reply(): Callback function to handle other incoming messages
+ * except SendPost and ReceivePost.
+ */
+static void i2o_lan_reply(struct i2o_handler *h, struct i2o_controller *iop,
+ struct i2o_message *m)
+{
+ u32 *msg = (u32 *)m;
+ u8 unit = (u8)(msg[2]>>16); // InitiatorContext
+ struct net_device *dev = i2o_landevs[unit];
+
+ if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
+ if (i2o_lan_handle_status(dev, msg))
+ return;
+
+ /* In other error cases just report and continue */
+
+ i2o_report_status(KERN_INFO, dev->name, msg);
+ }
+
+#ifdef DRIVERDEBUG
+ i2o_report_status(KERN_INFO, dev->name, msg);
+#endif
+ switch (msg[1] >> 24) {
+ case LAN_RESET:
+ case LAN_SUSPEND:
+ /* default reply without payload */
+ break;
+
+ case I2O_CMD_UTIL_EVT_REGISTER:
+ case I2O_CMD_UTIL_EVT_ACK:
+ i2o_lan_handle_event(dev, msg);
+ break;
+
+ case I2O_CMD_UTIL_PARAMS_SET:
+ /* default reply, results in ReplyPayload (not examined) */
+ switch (msg[3] >> 16) {
+ case 1: dprintk(KERN_INFO "%s: Reply to set MAC filter mask.\n",
+ dev->name);
+ break;
+ case 2: dprintk(KERN_INFO "%s: Reply to set MAC table.\n",
+ dev->name);
+ break;
+ default: printk(KERN_WARNING "%s: Bad group 0x%04X\n",
+ dev->name,msg[3] >> 16);
+ }
+ break;
+
+ default:
+ printk(KERN_ERR "%s: No handler for the reply.\n",
+ dev->name);
+ i2o_report_status(KERN_INFO, dev->name, msg);
+ }
+}
+
+/* Functions used by the above callback functions:
+=================================================*/
+/*
+ * i2o_lan_release_buckets(): Free unused buckets (sk_buffs).
+ */
+static void i2o_lan_release_buckets(struct net_device *dev, u32 *msg)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ u8 trl_elem_size = (u8)(msg[3]>>8 & 0x000000FF);
+ u8 trl_count = (u8)(msg[3] & 0x000000FF);
+ u32 *pskb = &msg[6];
+
+ while (trl_count--) {
+ dprintk(KERN_DEBUG "%s: Releasing unused rx skb %p (trl_count=%d).\n",
+ dev->name, (struct sk_buff*)(*pskb),trl_count+1);
+ dev_kfree_skb_irq((struct sk_buff *)(*pskb));
+ pskb += 1 + trl_elem_size;
+ atomic_dec(&priv->buckets_out);
+ }
+}
+
+/*
+ * i2o_lan_event_reply(): Handle events.
+ */
+static void i2o_lan_handle_event(struct net_device *dev, u32 *msg)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 max_evt_data_size =iop->status_block->inbound_frame_size-5;
+ struct i2o_reply {
+ u32 header[4];
+ u32 evt_indicator;
+ u32 data[max_evt_data_size];
+ } *evt = (struct i2o_reply *)msg;
+ int evt_data_len = ((msg[0]>>16) - 5) * 4; /* real size*/
+
+ printk(KERN_INFO "%s: I2O event - ", dev->name);
+
+ if (msg[1]>>24 == I2O_CMD_UTIL_EVT_ACK) {
+ printk("Event acknowledgement reply.\n");
+ return;
+ }
+
+ /* Else evt->function == I2O_CMD_UTIL_EVT_REGISTER) */
+
+ switch (evt->evt_indicator) {
+ case I2O_EVT_IND_STATE_CHANGE: {
+ struct state_data {
+ u16 status;
+ u8 state;
+ u8 data;
+ } *evt_data = (struct state_data *)(evt->data[0]);
+
+ printk("State chance 0x%08x.\n", evt->data[0]);
+
+ /* If the DDM is in error state, recovery may be
+ * possible if status = Transmit or Receive Control
+ * Unit Inoperable.
+ */
+ if (evt_data->state==0x05 && evt_data->status==0x0003)
+ i2o_lan_reset(dev);
+ break;
+ }
+
+ case I2O_EVT_IND_FIELD_MODIFIED: {
+ u16 *work16 = (u16 *)evt->data;
+ printk("Group 0x%04x, field %d changed.\n", work16[0], work16[1]);
+ break;
+ }
+
+ case I2O_EVT_IND_VENDOR_EVT: {
+ int i;
+ printk("Vendor event:\n");
+ for (i = 0; i < evt_data_len / 4; i++)
+ printk(" 0x%08x\n", evt->data[i]);
+ break;
+ }
+
+ case I2O_EVT_IND_DEVICE_RESET:
+ /* Spec 2.0 p. 6-121:
+ * The event of _DEVICE_RESET should also be responded
+ */
+ printk("Device reset.\n");
+ if (i2o_event_ack(iop, msg) < 0)
+ printk("%s: Event Acknowledge timeout.\n", dev->name);
+ break;
+
+#if 0
+ case I2O_EVT_IND_EVT_MASK_MODIFIED:
+ printk("Event mask modified, 0x%08x.\n", evt->data[0]);
+ break;
+
+ case I2O_EVT_IND_GENERAL_WARNING:
+ printk("General warning 0x%04x.\n", evt->data[0]);
+ break;
+
+ case I2O_EVT_IND_CONFIGURATION_FLAG:
+ printk("Configuration requested.\n");
+ break;
+
+ case I2O_EVT_IND_CAPABILITY_CHANGE:
+ printk("Capability change 0x%04x.\n", evt->data[0]);
+ break;
+
+ case I2O_EVT_IND_DEVICE_STATE:
+ printk("Device state changed 0x%08x.\n", evt->data[0]);
+ break;
+#endif
+ case I2O_LAN_EVT_LINK_DOWN:
+ netif_carrier_off(dev);
+ printk("Link to the physical device is lost.\n");
+ break;
+
+ case I2O_LAN_EVT_LINK_UP:
+ netif_carrier_on(dev);
+ printk("Link to the physical device is (re)established.\n");
+ break;
+
+ case I2O_LAN_EVT_MEDIA_CHANGE:
+ printk("Media change.\n");
+ break;
+ default:
+ printk("0x%08x. No handler.\n", evt->evt_indicator);
+ }
+}
+
+/*
+ * i2o_lan_receive_post(): Post buckets to receive packets.
+ */
+static int i2o_lan_receive_post(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ struct sk_buff *skb;
+ u32 m, *msg;
+ u32 bucket_len = (dev->mtu + dev->hard_header_len);
+ u32 total = priv->max_buckets_out - atomic_read(&priv->buckets_out);
+ u32 bucket_count;
+ u32 *sgl_elem;
+ unsigned long flags;
+
+ /* Send (total/bucket_count) separate I2O requests */
+
+ while (total) {
+ m = I2O_POST_READ32(iop);
+ if (m == 0xFFFFFFFF)
+ return -ETIMEDOUT;
+ msg = (u32 *)(iop->mem_offset + m);
+
+ bucket_count = (total >= priv->sgl_max) ? priv->sgl_max : total;
+ total -= bucket_count;
+ atomic_add(bucket_count, &priv->buckets_out);
+
+ dprintk(KERN_INFO "%s: Sending %d buckets (size %d) to LAN DDM.\n",
+ dev->name, bucket_count, bucket_len);
+
+ /* Fill in the header */
+
+ __raw_writel(I2O_MESSAGE_SIZE(4 + 3 * bucket_count) | SGL_OFFSET_4, msg);
+ __raw_writel(LAN_RECEIVE_POST<<24 | HOST_TID<<12 | i2o_dev->lct_data.tid, msg+1);
+ __raw_writel(priv->unit << 16 | lan_receive_context, msg+2);
+ __raw_writel(bucket_count, msg+3);
+ sgl_elem = &msg[4];
+
+ /* Fill in the payload - contains bucket_count SGL elements */
+
+ while (bucket_count--) {
+ spin_lock_irqsave(&priv->fbl_lock, flags);
+ if (priv->i2o_fbl_tail >= 0)
+ skb = priv->i2o_fbl[priv->i2o_fbl_tail--];
+ else {
+ skb = dev_alloc_skb(bucket_len + 2);
+ if (skb == NULL) {
+ spin_unlock_irqrestore(&priv->fbl_lock, flags);
+ return -ENOMEM;
+ }
+ skb_reserve(skb, 2);
+ }
+ spin_unlock_irqrestore(&priv->fbl_lock, flags);
+
+ __raw_writel(0x51000000 | bucket_len, sgl_elem);
+ __raw_writel((u32)skb, sgl_elem+1);
+ __raw_writel(virt_to_bus(skb->data), sgl_elem+2);
+ sgl_elem += 3;
+ }
+
+ /* set LE flag and post */
+ __raw_writel(__raw_readl(sgl_elem-3) | 0x80000000, (sgl_elem-3));
+ i2o_post_message(iop, m);
+ }
+
+ return 0;
+}
+
+/* Functions called from the network stack, and functions called by them:
+========================================================================*/
+
+/*
+ * i2o_lan_reset(): Reset the LAN adapter into the operational state and
+ * restore it to full operation.
+ */
+static int i2o_lan_reset(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 msg[5];
+
+ dprintk(KERN_INFO "%s: LAN RESET MESSAGE.\n", dev->name);
+ msg[0] = FIVE_WORD_MSG_SIZE | SGL_OFFSET_0;
+ msg[1] = LAN_RESET<<24 | HOST_TID<<12 | i2o_dev->lct_data.tid;
+ msg[2] = priv->unit << 16 | lan_context; // InitiatorContext
+ msg[3] = 0; // TransactionContext
+ msg[4] = 0; // Keep posted buckets
+
+ if (i2o_post_this(iop, msg, sizeof(msg)) < 0)
+ return -ETIMEDOUT;
+
+ return 0;
+}
+
+/*
+ * i2o_lan_suspend(): Put LAN adapter into a safe, non-active state.
+ * IOP replies to any LAN class message with status error_no_data_transfer
+ * / suspended.
+ */
+static int i2o_lan_suspend(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 msg[5];
+
+ dprintk(KERN_INFO "%s: LAN SUSPEND MESSAGE.\n", dev->name);
+ msg[0] = FIVE_WORD_MSG_SIZE | SGL_OFFSET_0;
+ msg[1] = LAN_SUSPEND<<24 | HOST_TID<<12 | i2o_dev->lct_data.tid;
+ msg[2] = priv->unit << 16 | lan_context; // InitiatorContext
+ msg[3] = 0; // TransactionContext
+ msg[4] = 1 << 16; // return posted buckets
+
+ if (i2o_post_this(iop, msg, sizeof(msg)) < 0)
+ return -ETIMEDOUT;
+
+ return 0;
+}
+
+/*
+ * i2o_set_ddm_parameters:
+ * These settings are done to ensure proper initial values for DDM.
+ * They can be changed via proc file system or vai configuration utility.
+ */
+static void i2o_set_ddm_parameters(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 val;
+
+ /*
+ * When PacketOrphanlimit is set to the maximum packet length,
+ * the packets will never be split into two separate buckets
+ */
+ val = dev->mtu + dev->hard_header_len;
+ if (i2o_set_scalar(iop, i2o_dev->lct_data.tid, 0x0004, 2, &val, sizeof(val)) < 0)
+ printk(KERN_WARNING "%s: Unable to set PacketOrphanLimit.\n",
+ dev->name);
+ else
+ dprintk(KERN_INFO "%s: PacketOrphanLimit set to %d.\n",
+ dev->name, val);
+
+ /* When RxMaxPacketsBucket = 1, DDM puts only one packet into bucket */
+
+ val = 1;
+ if (i2o_set_scalar(iop, i2o_dev->lct_data.tid, 0x0008, 4, &val, sizeof(val)) <0)
+ printk(KERN_WARNING "%s: Unable to set RxMaxPacketsBucket.\n",
+ dev->name);
+ else
+ dprintk(KERN_INFO "%s: RxMaxPacketsBucket set to %d.\n",
+ dev->name, val);
+ return;
+}
+
+/* Functions called from the network stack:
+==========================================*/
+
+/*
+ * i2o_lan_open(): Open the device to send/receive packets via
+ * the network device.
+ */
+static int i2o_lan_open(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 mc_addr_group[64];
+
+ MOD_INC_USE_COUNT;
+
+ if (i2o_claim_device(i2o_dev, &i2o_lan_handler)) {
+ printk(KERN_WARNING "%s: Unable to claim the I2O LAN device.\n", dev->name);
+ MOD_DEC_USE_COUNT;
+ return -EAGAIN;
+ }
+ dprintk(KERN_INFO "%s: I2O LAN device (tid=%d) claimed by LAN OSM.\n",
+ dev->name, i2o_dev->lct_data.tid);
+
+ if (i2o_event_register(iop, i2o_dev->lct_data.tid,
+ priv->unit << 16 | lan_context, 0, priv->i2o_event_mask) < 0)
+ printk(KERN_WARNING "%s: Unable to set the event mask.\n", dev->name);
+
+ i2o_lan_reset(dev);
+
+ /* Get the max number of multicast addresses */
+
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0001, -1,
+ &mc_addr_group, sizeof(mc_addr_group)) < 0 ) {
+ printk(KERN_WARNING "%s: Unable to query LAN_MAC_ADDRESS group.\n", dev->name);
+ MOD_DEC_USE_COUNT;
+ return -EAGAIN;
+ }
+ priv->max_size_mc_table = mc_addr_group[8];
+
+ /* Malloc space for free bucket list to resuse reveive post buckets */
+
+ priv->i2o_fbl = kmalloc(priv->max_buckets_out * sizeof(struct sk_buff *),
+ GFP_KERNEL);
+ if (priv->i2o_fbl == NULL) {
+ MOD_DEC_USE_COUNT;
+ return -ENOMEM;
+ }
+ priv->i2o_fbl_tail = -1;
+ priv->send_active = 0;
+
+ i2o_set_ddm_parameters(dev);
+ i2o_lan_receive_post(dev);
+
+ netif_start_queue(dev);
+
+ return 0;
+}
+
+/*
+ * i2o_lan_close(): End the transfering.
+ */
+static int i2o_lan_close(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ int ret = 0;
+
+ netif_stop_queue(dev);
+ i2o_lan_suspend(dev);
+
+ if (i2o_event_register(iop, i2o_dev->lct_data.tid,
+ priv->unit << 16 | lan_context, 0, 0) < 0)
+ printk(KERN_WARNING "%s: Unable to clear the event mask.\n",
+ dev->name);
+
+ while (priv->i2o_fbl_tail >= 0)
+ dev_kfree_skb(priv->i2o_fbl[priv->i2o_fbl_tail--]);
+
+ kfree(priv->i2o_fbl);
+
+ if (i2o_release_device(i2o_dev, &i2o_lan_handler)) {
+ printk(KERN_WARNING "%s: Unable to unclaim I2O LAN device "
+ "(tid=%d).\n", dev->name, i2o_dev->lct_data.tid);
+ ret = -EBUSY;
+ }
+
+ MOD_DEC_USE_COUNT;
+
+ return ret;
+}
+
+/*
+ * i2o_lan_tx_timeout(): Tx timeout handler.
+ */
+static void i2o_lan_tx_timeout(struct net_device *dev)
+{
+ if (!netif_queue_stopped(dev))
+ netif_start_queue(dev);
+}
+
+/*
+ * i2o_lan_batch_send(): Send packets in batch.
+ * Both i2o_lan_sdu_send and i2o_lan_packet_send use this.
+ */
+static void i2o_lan_batch_send(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_controller *iop = priv->i2o_dev->controller;
+
+ spin_lock_irq(&priv->tx_lock);
+ if (priv->tx_count != 0) {
+ dev->trans_start = jiffies;
+ i2o_post_message(iop, priv->m);
+ dprintk(KERN_DEBUG "%s: %d packets sent.\n", dev->name, priv->tx_count);
+ priv->tx_count = 0;
+ }
+ priv->send_active = 0;
+ spin_unlock_irq(&priv->tx_lock);
+ MOD_DEC_USE_COUNT;
+}
+
+#ifdef CONFIG_NET_FC
+/*
+ * i2o_lan_sdu_send(): Send a packet, MAC header added by the DDM.
+ * Must be supported by Fibre Channel, optional for Ethernet/802.3,
+ * Token Ring, FDDI
+ */
+static int i2o_lan_sdu_send(struct sk_buff *skb, struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ int tickssofar = jiffies - dev->trans_start;
+ u32 m, *msg;
+ u32 *sgl_elem;
+
+ spin_lock_irq(&priv->tx_lock);
+
+ priv->tx_count++;
+ atomic_inc(&priv->tx_out);
+
+ /*
+ * If tx_batch_mode = 0x00 forced to immediate mode
+ * If tx_batch_mode = 0x01 forced to batch mode
+ * If tx_batch_mode = 0x10 switch automatically, current mode immediate
+ * If tx_batch_mode = 0x11 switch automatically, current mode batch
+ * If gap between two packets is > 0 ticks, switch to immediate
+ */
+ if (priv->tx_batch_mode >> 1) // switch automatically
+ priv->tx_batch_mode = tickssofar ? 0x02 : 0x03;
+
+ if (priv->tx_count == 1) {
+ m = I2O_POST_READ32(iop);
+ if (m == 0xFFFFFFFF) {
+ spin_unlock_irq(&priv->tx_lock);
+ return 1;
+ }
+ msg = (u32 *)(iop->mem_offset + m);
+ priv->m = m;
+
+ __raw_writel(NINE_WORD_MSG_SIZE | 1<<12 | SGL_OFFSET_4, msg);
+ __raw_writel(LAN_PACKET_SEND<<24 | HOST_TID<<12 | i2o_dev->lct_data.tid, msg+1);
+ __raw_writel(priv->unit << 16 | lan_send_context, msg+2); // InitiatorContext
+ __raw_writel(1 << 30 | 1 << 3, msg+3); // TransmitControlWord
+
+ __raw_writel(0xD7000000 | skb->len, msg+4); // MAC hdr included
+ __raw_writel((u32)skb, msg+5); // TransactionContext
+ __raw_writel(virt_to_bus(skb->data), msg+6);
+ __raw_writel((u32)skb->mac.raw, msg+7);
+ __raw_writel((u32)skb->mac.raw+4, msg+8);
+
+ if ((priv->tx_batch_mode & 0x01) && !priv->send_active) {
+ priv->send_active = 1;
+ MOD_INC_USE_COUNT;
+ if (schedule_task(&priv->i2o_batch_send_task) == 0)
+ MOD_DEC_USE_COUNT;
+ }
+ } else { /* Add new SGL element to the previous message frame */
+
+ msg = (u32 *)(iop->mem_offset + priv->m);
+ sgl_elem = &msg[priv->tx_count * 5 + 1];
+
+ __raw_writel(I2O_MESSAGE_SIZE((__raw_readl(msg)>>16) + 5) | 1<<12 | SGL_OFFSET_4, msg);
+ __raw_writel(__raw_readl(sgl_elem-5) & 0x7FFFFFFF, sgl_elem-5); /* clear LE flag */
+ __raw_writel(0xD5000000 | skb->len, sgl_elem);
+ __raw_writel((u32)skb, sgl_elem+1);
+ __raw_writel(virt_to_bus(skb->data), sgl_elem+2);
+ __raw_writel((u32)(skb->mac.raw), sgl_elem+3);
+ __raw_writel((u32)(skb->mac.raw)+1, sgl_elem+4);
+ }
+
+ /* If tx not in batch mode or frame is full, send immediatelly */
+
+ if (!(priv->tx_batch_mode & 0x01) || priv->tx_count == priv->sgl_max) {
+ dev->trans_start = jiffies;
+ i2o_post_message(iop, priv->m);
+ dprintk(KERN_DEBUG "%s: %d packets sent.\n", dev->name, priv->tx_count);
+ priv->tx_count = 0;
+ }
+
+ /* If DDMs TxMaxPktOut reached, stop queueing layer to send more */
+
+ if (atomic_read(&priv->tx_out) >= priv->tx_max_out)
+ netif_stop_queue(dev);
+
+ spin_unlock_irq(&priv->tx_lock);
+ return 0;
+}
+#endif /* CONFIG_NET_FC */
+
+/*
+ * i2o_lan_packet_send(): Send a packet as is, including the MAC header.
+ *
+ * Must be supported by Ethernet/802.3, Token Ring, FDDI, optional for
+ * Fibre Channel
+ */
+static int i2o_lan_packet_send(struct sk_buff *skb, struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ int tickssofar = jiffies - dev->trans_start;
+ u32 m, *msg;
+ u32 *sgl_elem;
+
+ spin_lock_irq(&priv->tx_lock);
+
+ priv->tx_count++;
+ atomic_inc(&priv->tx_out);
+
+ /*
+ * If tx_batch_mode = 0x00 forced to immediate mode
+ * If tx_batch_mode = 0x01 forced to batch mode
+ * If tx_batch_mode = 0x10 switch automatically, current mode immediate
+ * If tx_batch_mode = 0x11 switch automatically, current mode batch
+ * If gap between two packets is > 0 ticks, switch to immediate
+ */
+ if (priv->tx_batch_mode >> 1) // switch automatically
+ priv->tx_batch_mode = tickssofar ? 0x02 : 0x03;
+
+ if (priv->tx_count == 1) {
+ m = I2O_POST_READ32(iop);
+ if (m == 0xFFFFFFFF) {
+ spin_unlock_irq(&priv->tx_lock);
+ return 1;
+ }
+ msg = (u32 *)(iop->mem_offset + m);
+ priv->m = m;
+
+ __raw_writel(SEVEN_WORD_MSG_SIZE | 1<<12 | SGL_OFFSET_4, msg);
+ __raw_writel(LAN_PACKET_SEND<<24 | HOST_TID<<12 | i2o_dev->lct_data.tid, msg+1);
+ __raw_writel(priv->unit << 16 | lan_send_context, msg+2); // InitiatorContext
+ __raw_writel(1 << 30 | 1 << 3, msg+3); // TransmitControlWord
+ // bit 30: reply as soon as transmission attempt is complete
+ // bit 3: Suppress CRC generation
+ __raw_writel(0xD5000000 | skb->len, msg+4); // MAC hdr included
+ __raw_writel((u32)skb, msg+5); // TransactionContext
+ __raw_writel(virt_to_bus(skb->data), msg+6);
+
+ if ((priv->tx_batch_mode & 0x01) && !priv->send_active) {
+ priv->send_active = 1;
+ MOD_INC_USE_COUNT;
+ if (schedule_task(&priv->i2o_batch_send_task) == 0)
+ MOD_DEC_USE_COUNT;
+ }
+ } else { /* Add new SGL element to the previous message frame */
+
+ msg = (u32 *)(iop->mem_offset + priv->m);
+ sgl_elem = &msg[priv->tx_count * 3 + 1];
+
+ __raw_writel(I2O_MESSAGE_SIZE((__raw_readl(msg)>>16) + 3) | 1<<12 | SGL_OFFSET_4, msg);
+ __raw_writel(__raw_readl(sgl_elem-3) & 0x7FFFFFFF, sgl_elem-3); /* clear LE flag */
+ __raw_writel(0xD5000000 | skb->len, sgl_elem);
+ __raw_writel((u32)skb, sgl_elem+1);
+ __raw_writel(virt_to_bus(skb->data), sgl_elem+2);
+ }
+
+ /* If tx is in immediate mode or frame is full, send now */
+
+ if (!(priv->tx_batch_mode & 0x01) || priv->tx_count == priv->sgl_max) {
+ dev->trans_start = jiffies;
+ i2o_post_message(iop, priv->m);
+ dprintk(KERN_DEBUG "%s: %d packets sent.\n", dev->name, priv->tx_count);
+ priv->tx_count = 0;
+ }
+
+ /* If DDMs TxMaxPktOut reached, stop queueing layer to send more */
+
+ if (atomic_read(&priv->tx_out) >= priv->tx_max_out)
+ netif_stop_queue(dev);
+
+ spin_unlock_irq(&priv->tx_lock);
+ return 0;
+}
+
+/*
+ * i2o_lan_get_stats(): Fill in the statistics.
+ */
+static struct net_device_stats *i2o_lan_get_stats(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u64 val64[16];
+ u64 supported_group[4] = { 0, 0, 0, 0 };
+
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0100, -1, val64,
+ sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_HISTORICAL_STATS.\n", dev->name);
+ else {
+ dprintk(KERN_DEBUG "%s: LAN_HISTORICAL_STATS queried.\n", dev->name);
+ priv->stats.tx_packets = val64[0];
+ priv->stats.tx_bytes = val64[1];
+ priv->stats.rx_packets = val64[2];
+ priv->stats.rx_bytes = val64[3];
+ priv->stats.tx_errors = val64[4];
+ priv->stats.rx_errors = val64[5];
+ priv->stats.rx_dropped = val64[6];
+ }
+
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0180, -1,
+ &supported_group, sizeof(supported_group)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_SUPPORTED_OPTIONAL_HISTORICAL_STATS.\n", dev->name);
+
+ if (supported_group[2]) {
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0183, -1,
+ val64, sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_OPTIONAL_RX_HISTORICAL_STATS.\n", dev->name);
+ else {
+ dprintk(KERN_DEBUG "%s: LAN_OPTIONAL_RX_HISTORICAL_STATS queried.\n", dev->name);
+ priv->stats.multicast = val64[4];
+ priv->stats.rx_length_errors = val64[10];
+ priv->stats.rx_crc_errors = val64[0];
+ }
+ }
+
+ if (i2o_dev->lct_data.sub_class == I2O_LAN_ETHERNET) {
+ u64 supported_stats = 0;
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0200, -1,
+ val64, sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_802_3_HISTORICAL_STATS.\n", dev->name);
+ else {
+ dprintk(KERN_DEBUG "%s: LAN_802_3_HISTORICAL_STATS queried.\n", dev->name);
+ priv->stats.transmit_collision = val64[1] + val64[2];
+ priv->stats.rx_frame_errors = val64[0];
+ priv->stats.tx_carrier_errors = val64[6];
+ }
+
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0280, -1,
+ &supported_stats, sizeof(supported_stats)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_SUPPORTED_802_3_HISTORICAL_STATS.\n", dev->name);
+
+ if (supported_stats != 0) {
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0281, -1,
+ val64, sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_OPTIONAL_802_3_HISTORICAL_STATS.\n", dev->name);
+ else {
+ dprintk(KERN_DEBUG "%s: LAN_OPTIONAL_802_3_HISTORICAL_STATS queried.\n", dev->name);
+ if (supported_stats & 0x1)
+ priv->stats.rx_over_errors = val64[0];
+ if (supported_stats & 0x4)
+ priv->stats.tx_heartbeat_errors = val64[2];
+ }
+ }
+ }
+
+#ifdef CONFIG_TR
+ if (i2o_dev->lct_data.sub_class == I2O_LAN_TR) {
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0300, -1,
+ val64, sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_802_5_HISTORICAL_STATS.\n", dev->name);
+ else {
+ struct tr_statistics *stats =
+ (struct tr_statistics *)&priv->stats;
+ dprintk(KERN_DEBUG "%s: LAN_802_5_HISTORICAL_STATS queried.\n", dev->name);
+
+ stats->line_errors = val64[0];
+ stats->internal_errors = val64[7];
+ stats->burst_errors = val64[4];
+ stats->A_C_errors = val64[2];
+ stats->abort_delimiters = val64[3];
+ stats->lost_frames = val64[1];
+ /* stats->recv_congest_count = ?; FIXME ??*/
+ stats->frame_copied_errors = val64[5];
+ stats->frequency_errors = val64[6];
+ stats->token_errors = val64[9];
+ }
+ /* Token Ring optional stats not yet defined */
+ }
+#endif
+
+#ifdef CONFIG_FDDI
+ if (i2o_dev->lct_data.sub_class == I2O_LAN_FDDI) {
+ if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0400, -1,
+ val64, sizeof(val64)) < 0)
+ printk(KERN_INFO "%s: Unable to query LAN_FDDI_HISTORICAL_STATS.\n", dev->name);
+ else {
+ dprintk(KERN_DEBUG "%s: LAN_FDDI_HISTORICAL_STATS queried.\n", dev->name);
+ priv->stats.smt_cf_state = val64[0];
+ memcpy(priv->stats.mac_upstream_nbr, &val64[1], FDDI_K_ALEN);
+ memcpy(priv->stats.mac_downstream_nbr, &val64[2], FDDI_K_ALEN);
+ priv->stats.mac_error_cts = val64[3];
+ priv->stats.mac_lost_cts = val64[4];
+ priv->stats.mac_rmt_state = val64[5];
+ memcpy(priv->stats.port_lct_fail_cts, &val64[6], 8);
+ memcpy(priv->stats.port_lem_reject_cts, &val64[7], 8);
+ memcpy(priv->stats.port_lem_cts, &val64[8], 8);
+ memcpy(priv->stats.port_pcm_state, &val64[9], 8);
+ }
+ /* FDDI optional stats not yet defined */
+ }
+#endif
+
+#ifdef CONFIG_NET_FC
+ /* Fibre Channel Statistics not yet defined in 1.53 nor 2.0 */
+#endif
+
+ return (struct net_device_stats *)&priv->stats;
+}
+
+/*
+ * i2o_lan_set_mc_filter(): Post a request to set multicast filter.
+ */
+int i2o_lan_set_mc_filter(struct net_device *dev, u32 filter_mask)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ u32 msg[10];
+
+ msg[0] = TEN_WORD_MSG_SIZE | SGL_OFFSET_5;
+ msg[1] = I2O_CMD_UTIL_PARAMS_SET << 24 | HOST_TID << 12 | i2o_dev->lct_data.tid;
+ msg[2] = priv->unit << 16 | lan_context;
+ msg[3] = 0x0001 << 16 | 3 ; // TransactionContext: group&field
+ msg[4] = 0;
+ msg[5] = 0xCC000000 | 16; // Immediate data SGL
+ msg[6] = 1; // OperationCount
+ msg[7] = 0x0001<<16 | I2O_PARAMS_FIELD_SET; // Group, Operation
+ msg[8] = 3 << 16 | 1; // FieldIndex, FieldCount
+ msg[9] = filter_mask; // Value
+
+ return i2o_post_this(iop, msg, sizeof(msg));
+}
+
+/*
+ * i2o_lan_set_mc_table(): Post a request to set LAN_MULTICAST_MAC_ADDRESS table.
+ */
+int i2o_lan_set_mc_table(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ struct i2o_controller *iop = i2o_dev->controller;
+ struct dev_mc_list *mc;
+ u32 msg[10 + 2 * dev->mc_count];
+ u8 *work8 = (u8 *)(msg + 10);
+
+ msg[0] = I2O_MESSAGE_SIZE(10 + 2 * dev->mc_count) | SGL_OFFSET_5;
+ msg[1] = I2O_CMD_UTIL_PARAMS_SET << 24 | HOST_TID << 12 | i2o_dev->lct_data.tid;
+ msg[2] = priv->unit << 16 | lan_context; // InitiatorContext
+ msg[3] = 0x0002 << 16 | (u16)-1; // TransactionContext
+ msg[4] = 0; // OperationFlags
+ msg[5] = 0xCC000000 | (16 + 8 * dev->mc_count); // Immediate data SGL
+ msg[6] = 2; // OperationCount
+ msg[7] = 0x0002 << 16 | I2O_PARAMS_TABLE_CLEAR; // Group, Operation
+ msg[8] = 0x0002 << 16 | I2O_PARAMS_ROW_ADD; // Group, Operation
+ msg[9] = dev->mc_count << 16 | (u16)-1; // RowCount, FieldCount
+
+ for (mc = dev->mc_list; mc ; mc = mc->next, work8 += 8) {
+ memset(work8, 0, 8);
+ memcpy(work8, mc->dmi_addr, mc->dmi_addrlen); // Values
+ }
+
+ return i2o_post_this(iop, msg, sizeof(msg));
+}
+
+/*
+ * i2o_lan_set_multicast_list(): Enable a network device to receive packets
+ * not send to the protocol address.
+ */
+static void i2o_lan_set_multicast_list(struct net_device *dev)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ u32 filter_mask;
+
+ if (dev->flags & IFF_PROMISC) {
+ filter_mask = 0x00000002;
+ dprintk(KERN_INFO "%s: Enabling promiscuous mode...\n", dev->name);
+ } else if ((dev->flags & IFF_ALLMULTI) || dev->mc_count > priv->max_size_mc_table) {
+ filter_mask = 0x00000004;
+ dprintk(KERN_INFO "%s: Enabling all multicast mode...\n", dev->name);
+ } else if (dev->mc_count) {
+ filter_mask = 0x00000000;
+ dprintk(KERN_INFO "%s: Enabling multicast mode...\n", dev->name);
+ if (i2o_lan_set_mc_table(dev) < 0)
+ printk(KERN_WARNING "%s: Unable to send MAC table.\n", dev->name);
+ } else {
+ filter_mask = 0x00000300; // Broadcast, Multicast disabled
+ dprintk(KERN_INFO "%s: Enabling unicast mode...\n", dev->name);
+ }
+
+ /* Finally copy new FilterMask to DDM */
+
+ if (i2o_lan_set_mc_filter(dev, filter_mask) < 0)
+ printk(KERN_WARNING "%s: Unable to send MAC FilterMask.\n", dev->name);
+}
+
+/*
+ * i2o_lan_change_mtu(): Change maximum transfer unit size.
+ */
+static int i2o_lan_change_mtu(struct net_device *dev, int new_mtu)
+{
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+ u32 max_pkt_size;
+
+ if (i2o_query_scalar(i2o_dev->controller, i2o_dev->lct_data.tid,
+ 0x0000, 6, &max_pkt_size, 4) < 0)
+ return -EFAULT;
+
+ if (new_mtu < 68 || new_mtu > 9000 || new_mtu > max_pkt_size)
+ return -EINVAL;
+
+ dev->mtu = new_mtu;
+
+ i2o_lan_suspend(dev); // to SUSPENDED state, return buckets
+
+ while (priv->i2o_fbl_tail >= 0) // free buffered buckets
+ dev_kfree_skb(priv->i2o_fbl[priv->i2o_fbl_tail--]);
+
+ i2o_lan_reset(dev); // to OPERATIONAL state
+ i2o_set_ddm_parameters(dev); // reset some parameters
+ i2o_lan_receive_post(dev); // post new buckets (new size)
+
+ return 0;
+}
+
+/* Functions to initialize I2O LAN OSM:
+======================================*/
+
+/*
+ * i2o_lan_register_device(): Register LAN class device to kernel.
+ */
+struct net_device *i2o_lan_register_device(struct i2o_device *i2o_dev)
+{
+ struct net_device *dev = NULL;
+ struct i2o_lan_local *priv = NULL;
+ u8 hw_addr[8];
+ u32 tx_max_out = 0;
+ unsigned short (*type_trans)(struct sk_buff *, struct net_device *);
+ void (*unregister_dev)(struct net_device *dev);
+
+ switch (i2o_dev->lct_data.sub_class) {
+ case I2O_LAN_ETHERNET:
+ dev = init_etherdev(NULL, sizeof(struct i2o_lan_local));
+ if (dev == NULL)
+ return NULL;
+ type_trans = eth_type_trans;
+ unregister_dev = unregister_netdev;
+ break;
+
+#ifdef CONFIG_ANYLAN
+ case I2O_LAN_100VG:
+ printk(KERN_ERR "i2o_lan: 100base VG not yet supported.\n");
+ return NULL;
+ break;
+#endif
+
+#ifdef CONFIG_TR
+ case I2O_LAN_TR:
+ dev = init_trdev(NULL, sizeof(struct i2o_lan_local));
+ if (dev==NULL)
+ return NULL;
+ type_trans = tr_type_trans;
+ unregister_dev = unregister_trdev;
+ break;
+#endif
+
+#ifdef CONFIG_FDDI
+ case I2O_LAN_FDDI:
+ {
+ int size = sizeof(struct net_device) + sizeof(struct i2o_lan_local);
+
+ dev = (struct net_device *) kmalloc(size, GFP_KERNEL);
+ if (dev == NULL)
+ return NULL;
+ memset((char *)dev, 0, size);
+ dev->priv = (void *)(dev + 1);
+
+ if (dev_alloc_name(dev, "fddi%d") < 0) {
+ printk(KERN_WARNING "i2o_lan: Too many FDDI devices.\n");
+ kfree(dev);
+ return NULL;
+ }
+ type_trans = fddi_type_trans;
+ unregister_dev = (void *)unregister_netdevice;
+
+ fddi_setup(dev);
+ register_netdev(dev);
+ }
+ break;
+#endif
+
+#ifdef CONFIG_NET_FC
+ case I2O_LAN_FIBRE_CHANNEL:
+ dev = init_fcdev(NULL, sizeof(struct i2o_lan_local));
+ if (dev == NULL)
+ return NULL;
+ type_trans = NULL;
+/* FIXME: Move fc_type_trans() from drivers/net/fc/iph5526.c to net/802/fc.c
+ * and export it in include/linux/fcdevice.h
+ * type_trans = fc_type_trans;
+ */
+ unregister_dev = (void *)unregister_fcdev;
+ break;
+#endif
+
+ case I2O_LAN_UNKNOWN:
+ default:
+ printk(KERN_ERR "i2o_lan: LAN type 0x%04x not supported.\n",
+ i2o_dev->lct_data.sub_class);
+ return NULL;
+ }
+
+ priv = (struct i2o_lan_local *)dev->priv;
+ priv->i2o_dev = i2o_dev;
+ priv->type_trans = type_trans;
+ priv->sgl_max = (i2o_dev->controller->status_block->inbound_frame_size - 4) / 3;
+ atomic_set(&priv->buckets_out, 0);
+
+ /* Set default values for user configurable parameters */
+ /* Private values are changed via /proc file system */
+
+ priv->max_buckets_out = max_buckets_out;
+ priv->bucket_thresh = bucket_thresh;
+ priv->rx_copybreak = rx_copybreak;
+ priv->tx_batch_mode = tx_batch_mode & 0x03;
+ priv->i2o_event_mask = i2o_event_mask;
+
+ priv->tx_lock = SPIN_LOCK_UNLOCKED;
+ priv->fbl_lock = SPIN_LOCK_UNLOCKED;
+
+ unit++;
+ i2o_landevs[unit] = dev;
+ priv->unit = unit;
+
+ if (i2o_query_scalar(i2o_dev->controller, i2o_dev->lct_data.tid,
+ 0x0001, 0, &hw_addr, sizeof(hw_addr)) < 0) {
+ printk(KERN_ERR "%s: Unable to query hardware address.\n", dev->name);
+ unit--;
+ unregister_dev(dev);
+ kfree(dev);
+ return NULL;
+ }
+ dprintk(KERN_DEBUG "%s: hwaddr = %02X:%02X:%02X:%02X:%02X:%02X\n",
+ dev->name, hw_addr[0], hw_addr[1], hw_addr[2], hw_addr[3],
+ hw_addr[4], hw_addr[5]);
+
+ dev->addr_len = 6;
+ memcpy(dev->dev_addr, hw_addr, 6);
+
+ if (i2o_query_scalar(i2o_dev->controller, i2o_dev->lct_data.tid,
+ 0x0007, 2, &tx_max_out, sizeof(tx_max_out)) < 0) {
+ printk(KERN_ERR "%s: Unable to query max TX queue.\n", dev->name);
+ unit--;
+ unregister_dev(dev);
+ kfree(dev);
+ return NULL;
+ }
+ dprintk(KERN_INFO "%s: Max TX Outstanding = %d.\n", dev->name, tx_max_out);
+ priv->tx_max_out = tx_max_out;
+ atomic_set(&priv->tx_out, 0);
+ priv->tx_count = 0;
+
+ INIT_LIST_HEAD(&priv->i2o_batch_send_task.list);
+ priv->i2o_batch_send_task.sync = 0;
+ priv->i2o_batch_send_task.routine = (void *)i2o_lan_batch_send;
+ priv->i2o_batch_send_task.data = (void *)dev;
+
+ dev->open = i2o_lan_open;
+ dev->stop = i2o_lan_close;
+ dev->get_stats = i2o_lan_get_stats;
+ dev->set_multicast_list = i2o_lan_set_multicast_list;
+ dev->tx_timeout = i2o_lan_tx_timeout;
+ dev->watchdog_timeo = I2O_LAN_TX_TIMEOUT;
+
+#ifdef CONFIG_NET_FC
+ if (i2o_dev->lct_data.sub_class == I2O_LAN_FIBRE_CHANNEL)
+ dev->hard_start_xmit = i2o_lan_sdu_send;
+ else
+#endif
+ dev->hard_start_xmit = i2o_lan_packet_send;
+
+ if (i2o_dev->lct_data.sub_class == I2O_LAN_ETHERNET)
+ dev->change_mtu = i2o_lan_change_mtu;
+
+ return dev;
+}
+
+#ifdef MODULE
+#define i2o_lan_init init_module
+#endif
+
+int __init i2o_lan_init(void)
+{
+ struct net_device *dev;
+ int i;
+
+ printk(KERN_INFO "I2O LAN OSM (C) 1999 University of Helsinki.\n");
+
+ /* Module params are used as global defaults for private values */
+
+ if (max_buckets_out > I2O_LAN_MAX_BUCKETS_OUT)
+ max_buckets_out = I2O_LAN_MAX_BUCKETS_OUT;
+ if (bucket_thresh > max_buckets_out)
+ bucket_thresh = max_buckets_out;
+
+ /* Install handlers for incoming replies */
+
+ if (i2o_install_handler(&i2o_lan_send_handler) < 0) {
+ printk(KERN_ERR "i2o_lan: Unable to register I2O LAN OSM.\n");
+ return -EINVAL;
+ }
+ lan_send_context = i2o_lan_send_handler.context;
+
+ if (i2o_install_handler(&i2o_lan_receive_handler) < 0) {
+ printk(KERN_ERR "i2o_lan: Unable to register I2O LAN OSM.\n");
+ return -EINVAL;
+ }
+ lan_receive_context = i2o_lan_receive_handler.context;
+
+ if (i2o_install_handler(&i2o_lan_handler) < 0) {
+ printk(KERN_ERR "i2o_lan: Unable to register I2O LAN OSM.\n");
+ return -EINVAL;
+ }
+ lan_context = i2o_lan_handler.context;
+
+ for(i=0; i <= MAX_LAN_CARDS; i++)
+ i2o_landevs[i] = NULL;
+
+ for (i=0; i < MAX_I2O_CONTROLLERS; i++) {
+ struct i2o_controller *iop = i2o_find_controller(i);
+ struct i2o_device *i2o_dev;
+
+ if (iop==NULL)
+ continue;
+
+ for (i2o_dev=iop->devices;i2o_dev != NULL;i2o_dev=i2o_dev->next) {
+
+ if (i2o_dev->lct_data.class_id != I2O_CLASS_LAN)
+ continue;
+
+ /* Make sure device not already claimed by an ISM */
+ if (i2o_dev->lct_data.user_tid != 0xFFF)
+ continue;
+
+ if (unit == MAX_LAN_CARDS) {
+ i2o_unlock_controller(iop);
+ printk(KERN_WARNING "i2o_lan: Too many I2O LAN devices.\n");
+ return -EINVAL;
+ }
+
+ dev = i2o_lan_register_device(i2o_dev);
+ if (dev == NULL) {
+ printk(KERN_ERR "i2o_lan: Unable to register I2O LAN device 0x%04x.\n",
+ i2o_dev->lct_data.sub_class);
+ continue;
+ }
+
+ printk(KERN_INFO "%s: I2O LAN device registered, "
+ "subclass = 0x%04x, unit = %d, tid = %d.\n",
+ dev->name, i2o_dev->lct_data.sub_class,
+ ((struct i2o_lan_local *)dev->priv)->unit,
+ i2o_dev->lct_data.tid);
+ }
+
+ i2o_unlock_controller(iop);
+ }
+
+ dprintk(KERN_INFO "%d I2O LAN devices found and registered.\n", unit+1);
+
+ return 0;
+}
+
+#ifdef MODULE
+
+void cleanup_module(void)
+{
+ int i;
+
+ for (i = 0; i <= unit; i++) {
+ struct net_device *dev = i2o_landevs[i];
+ struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
+ struct i2o_device *i2o_dev = priv->i2o_dev;
+
+ switch (i2o_dev->lct_data.sub_class) {
+ case I2O_LAN_ETHERNET:
+ unregister_netdev(dev);
+ break;
+#ifdef CONFIG_FDDI
+ case I2O_LAN_FDDI:
+ unregister_netdevice(dev);
+ break;
+#endif
+#ifdef CONFIG_TR
+ case I2O_LAN_TR:
+ unregister_trdev(dev);
+ break;
+#endif
+#ifdef CONFIG_NET_FC
+ case I2O_LAN_FIBRE_CHANNEL:
+ unregister_fcdev(dev);
+ break;
+#endif
+ default:
+ printk(KERN_WARNING "%s: Spurious I2O LAN subclass 0x%08x.\n",
+ dev->name, i2o_dev->lct_data.sub_class);
+ }
+
+ dprintk(KERN_INFO "%s: I2O LAN device unregistered.\n",
+ dev->name);
+ kfree(dev);
+ }
+
+ i2o_remove_handler(&i2o_lan_handler);
+ i2o_remove_handler(&i2o_lan_send_handler);
+ i2o_remove_handler(&i2o_lan_receive_handler);
+}
+
+EXPORT_NO_SYMBOLS;
+
+MODULE_AUTHOR("University of Helsinki, Department of Computer Science");
+MODULE_DESCRIPTION("I2O Lan OSM");
+MODULE_LICENSE("GPL");
+
+
+MODULE_PARM(max_buckets_out, "1-" __MODULE_STRING(I2O_LAN_MAX_BUCKETS_OUT) "i");
+MODULE_PARM_DESC(max_buckets_out, "Total number of buckets to post (1-)");
+MODULE_PARM(bucket_thresh, "1-" __MODULE_STRING(I2O_LAN_MAX_BUCKETS_OUT) "i");
+MODULE_PARM_DESC(bucket_thresh, "Bucket post threshold (1-)");
+MODULE_PARM(rx_copybreak, "1-" "i");
+MODULE_PARM_DESC(rx_copybreak, "Copy breakpoint for copy only small frames (1-)");
+MODULE_PARM(tx_batch_mode, "0-2" "i");
+MODULE_PARM_DESC(tx_batch_mode, "0=Send immediatelly, 1=Send in batches, 2=Switch automatically");
+
+#endif
FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen (who was at: slshen@lbl.gov)