summaryrefslogtreecommitdiffstats
path: root/drivers/i2o/i2o_lan.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/i2o/i2o_lan.c')
-rw-r--r--drivers/i2o/i2o_lan.c264
1 files changed, 139 insertions, 125 deletions
diff --git a/drivers/i2o/i2o_lan.c b/drivers/i2o/i2o_lan.c
index 2957424d6..90b252041 100644
--- a/drivers/i2o/i2o_lan.c
+++ b/drivers/i2o/i2o_lan.c
@@ -1,7 +1,7 @@
/*
* drivers/i2o/i2o_lan.c
*
- * I2O LAN CLASS OSM April 3rd 2000
+ * I2O LAN CLASS OSM May 4th 2000
*
* (C) Copyright 1999, 2000 University of Helsinki,
* Department of Computer Science
@@ -22,8 +22,7 @@
* in Gigabit Eth environment (using SysKonnect's DDM)
* in Fast Ethernet environment (using Intel 82558 DDM)
*
- * TODO: check error checking / timeouts
- * code / test for other LAN classes
+ * TODO: tests for other LAN classes (Token Ring, Fibre Channel)
*/
#include <linux/config.h>
@@ -62,8 +61,8 @@
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 tx_batch_mode = I2O_LAN_TX_BATCH_MODE;
-static i2o_event_mask = I2O_LAN_EVENT_MASK;
+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];
@@ -139,8 +138,7 @@ static void i2o_lan_handle_failure(struct net_device *dev, u32 *msg)
struct sk_buff *skb = NULL;
u8 le_flag;
-// To be added to i2o_core.c
-// i2o_report_failure(KERN_INFO, iop, dev->name, msg);
+ i2o_report_status(KERN_INFO, dev->name, msg);
/* If PacketSend failed, free sk_buffs reserved by upper layers */
@@ -190,8 +188,7 @@ 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;
-// To be added to i2o_core.c
-// i2o_report_transaction_error(KERN_INFO, dev->name, msg);
+ i2o_report_status(KERN_INFO, dev->name, msg);
/* If PacketSend was rejected, free sk_buff reserved by upper layers */
@@ -253,17 +250,15 @@ static void i2o_lan_send_post_reply(struct i2o_handler *h,
struct i2o_lan_local *priv = (struct i2o_lan_local *)dev->priv;
u8 trl_count = msg[3] & 0x000000FF;
-#ifdef DRIVERDEBUG
- i2o_report_status(KERN_INFO, dev->name, msg);
-#endif
-
if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
if (i2o_lan_handle_status(dev, msg))
return;
-
- /* Else we get pending transmit request(s) back */
}
+#ifdef DRIVERDEBUG
+ i2o_report_status(KERN_INFO, dev->name, msg);
+#endif
+
/* DDM has handled transmit request(s), free sk_buffs */
while (trl_count) {
@@ -284,7 +279,7 @@ static void i2o_lan_send_post_reply(struct i2o_handler *h,
* 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)
+ struct i2o_controller *iop, struct i2o_message *m)
{
u32 *msg = (u32 *)m;
u8 unit = (u8)(msg[2]>>16); // InitiatorContext
@@ -297,10 +292,6 @@ static void i2o_lan_receive_post_reply(struct i2o_handler *h,
struct sk_buff *skb, *old_skb;
unsigned long flags = 0;
-#ifdef DRIVERDEBUG
- i2o_report_status(KERN_INFO, dev->name, msg);
-#endif
-
if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
if (i2o_lan_handle_status(dev, msg))
return;
@@ -313,44 +304,46 @@ static void i2o_lan_receive_post_reply(struct i2o_handler *h,
return;
}
- /* Which DetailedStatusCodes need special treatment? */
+ /* If other DetailedStatusCodes need special code, add it here */
}
+#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);
-#if 0
-/* Is this enough? If we get erroneous bucket, we can't assume that skb could
- * be reused, can we?
- */
- /* Should we optimise these ifs away from the fast path? -taneli */
- if (packet->flags & 0x0f) {
+ /* 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.\n", dev->name);
- if (packet->flags & 0x0c)
- /* This actually means that the hw is b0rken, since we
- have asked it to not send fragmented packets. */
- printk(KERN_DEBUG "%s: multi-bucket packets not supported!\n", dev->name);
- bucket++;
- if (skb)
- dev_kfree_skb_irq(skb);
- continue;
- }
+ 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);
- if (packet->status & 0xff) {
- /* Silently discard, unless debugging. */
- dprintk(KERN_DEBUG "%s: toasted packet received.\n", dev->name);
bucket++;
- if (skb)
- dev_kfree_skb_irq(skb);
continue;
}
-#endif
+
+ /* 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);
@@ -366,13 +359,17 @@ static void i2o_lan_receive_post_reply(struct i2o_handler *h,
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 "
@@ -387,7 +384,7 @@ static void i2o_lan_receive_post_reply(struct i2o_handler *h,
dev->name, atomic_read(&priv->buckets_out));
#endif
- /* If DDM has already consumed bucket_tresh buckets, post new ones */
+ /* If DDM has already consumed bucket_thresh buckets, post new ones */
if (atomic_read(&priv->buckets_out) <= priv->max_buckets_out - priv->bucket_thresh) {
i2o_post_buckets_task.data = (void *)dev;
@@ -409,10 +406,6 @@ static void i2o_lan_reply(struct i2o_handler *h, struct i2o_controller *iop,
u8 unit = (u8)(msg[2]>>16); // InitiatorContext
struct net_device *dev = i2o_landevs[unit];
-#ifdef DRIVERDEBUG
- i2o_report_status(KERN_INFO, dev->name, msg);
-#endif
-
if ((msg[4] >> 24) != I2O_REPLY_STATUS_SUCCESS) {
if (i2o_lan_handle_status(dev, msg))
return;
@@ -420,6 +413,10 @@ static void i2o_lan_reply(struct i2o_handler *h, struct i2o_controller *iop,
/* This should NOT be reached */
}
+#ifdef DRIVERDEBUG
+ i2o_report_status(KERN_INFO, dev->name, msg);
+#endif
+
switch (msg[1] >> 24) {
case LAN_RESET:
case LAN_SUSPEND:
@@ -465,6 +462,7 @@ 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 {
u8 version_offset;
u8 msg_flags;
@@ -475,7 +473,7 @@ static void i2o_lan_handle_event(struct net_device *dev, u32 *msg)
u32 initiator_context;
u32 transaction_context;
u32 evt_indicator;
- u32 data[(iop->inbound_size - 20) / 4]; /* max */
+ u32 data[max_evt_data_size]; /* max */
} *evt = (struct i2o_reply *)msg;
int evt_data_len = (evt->msg_size - 5) * 4; /* real */
@@ -534,8 +532,8 @@ static void i2o_lan_handle_event(struct net_device *dev, u32 *msg)
case I2O_EVT_IND_FIELD_MODIFIED: {
u16 *work16 = (u16 *)evt->data;
- printk("Group 0x%04x, field %d changed.\n", work16[0],
- work16[1]);
+ printk("Group 0x%04x, field %d changed.\n", work16[0], work16[1]);
+
break;
}
@@ -564,7 +562,7 @@ static void i2o_lan_handle_event(struct net_device *dev, u32 *msg)
break;
default:
- printk("Event Indicator = 0x%08x.\n", evt->evt_indicator);
+ printk("0x%08x. No handler.\n", evt->evt_indicator);
}
/* Note: EventAck necessary only for events that cause the device to
@@ -660,7 +658,7 @@ static int i2o_lan_reset(struct net_device *dev)
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
+ msg[4] = 0; // Keep posted buckets
if (i2o_post_this(iop, msg, sizeof(msg)) < 0)
return -ETIMEDOUT;
@@ -694,43 +692,22 @@ static int i2o_lan_suspend(struct net_device *dev)
}
/*
- * i2o_set_batch_mode(): Set DDM into batch mode.
+ * 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_batch_mode(struct net_device *dev)
+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;
- /* Set defaults LAN_BATCH_CONTROL attributes */
- /* May be changed via /proc or Configuration Utility */
-
- val = 0x00000000; // enable batch mode, toggle automatically
- if (i2o_set_scalar(iop, i2o_dev->lct_data.tid, 0x0003, 0, &val, sizeof(val)) <0)
- printk(KERN_WARNING "%s: Unable to enter I2O LAN batch mode.\n",
- dev->name);
- else
- dprintk(KERN_INFO "%s: I2O LAN batch mode enabled.\n", dev->name);
-
- /* Set LAN_OPERATION attributes */
-
-#ifdef DRIVERDEBUG
-/* Added for testing: this will be removed */
- val = 0x00000003; // 1 = UserFlags
- if (i2o_set_scalar(iop, i2o_dev->lct_data.tid, 0x0004, 1, &val, sizeof(val)) < 0)
- printk(KERN_WARNING "%s: Can't enable ErrorReporting & BadPacketHandling.\n",
- dev->name);
- else
- dprintk(KERN_INFO "%s: ErrorReporting enabled, "
- "BadPacketHandling enabled.\n", dev->name);
-#endif /* DRIVERDEBUG */
-
/*
- * When PacketOrphanlimit is same as the maximum packet length,
+ * 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; // 2 = PacketOrphanLimit
+ 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);
@@ -738,6 +715,15 @@ static void i2o_set_batch_mode(struct net_device *dev)
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;
}
@@ -779,7 +765,7 @@ static int i2o_lan_open(struct net_device *dev)
priv->i2o_fbl_tail = -1;
priv->send_active = 0;
- i2o_set_batch_mode(dev);
+ i2o_set_ddm_parameters(dev);
i2o_lan_receive_post(dev);
netif_start_queue(dev);
@@ -795,9 +781,9 @@ 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,
@@ -805,19 +791,20 @@ static int i2o_lan_close(struct net_device *dev)
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);
- return -EBUSY;
+ ret = -EBUSY;
}
- while (priv->i2o_fbl_tail >= 0)
- dev_kfree_skb(priv->i2o_fbl[priv->i2o_fbl_tail--]);
- kfree(priv->i2o_fbl);
-
MOD_DEC_USE_COUNT;
- return 0;
+ return ret;
}
/*
@@ -829,12 +816,12 @@ static void i2o_lan_tx_timeout(struct net_device *dev)
netif_start_queue(dev);
}
-#define batching(x, cond) ( (x)->tx_batch_mode==1 || ((x)->tx_batch_mode==2 && (cond)) )
-
/*
- * Batch send packets. Both i2o_lan_sdu_send and i2o_lan_packet_send
- * use this. I'm still not pleased. If you come up with
- * something better, please tell me. -taneli
+ * i2o_lan_batch_send(): Send packets in batch.
+ * Both i2o_lan_sdu_send and i2o_lan_packet_send use this.
+ *
+ * This is a coarse first approximation for the tx_batching.
+ * If you come up with something better, please tell me. -taneli
*/
static void i2o_lan_batch_send(struct net_device *dev)
{
@@ -848,20 +835,16 @@ static void i2o_lan_batch_send(struct net_device *dev)
dprintk(KERN_DEBUG "%s: %d packets sent.\n", dev->name, priv->tx_count);
priv->tx_count = 0;
}
- spin_unlock_irq(&priv->tx_lock);
-
priv->send_active = 0;
+ spin_unlock_irq(&priv->tx_lock);
}
+#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
*/
-
-/*
- * This is a coarse first approximation. Needs testing. Any takers? -taneli
- */
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;
@@ -876,6 +859,16 @@ static int i2o_lan_sdu_send(struct sk_buff *skb, struct net_device *dev)
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 > 2 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) {
@@ -888,14 +881,15 @@ static int i2o_lan_sdu_send(struct sk_buff *skb, struct net_device *dev)
__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 << 3, msg+3); // TransmitControlWord
+ __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 (batching(priv, !tickssofar) && !priv->send_active) {
+
+ if ((priv->tx_batch_mode & 0x01) && !priv->send_active) {
priv->send_active = 1;
queue_task(&priv->i2o_batch_send_task, &tq_scheduler);
}
@@ -915,7 +909,7 @@ static int i2o_lan_sdu_send(struct sk_buff *skb, struct net_device *dev)
/* If tx not in batch mode or frame is full, send immediatelly */
- if (!batching(priv, !tickssofar) || priv->tx_count == priv->sgl_max) {
+ 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);
@@ -930,6 +924,7 @@ static int i2o_lan_sdu_send(struct sk_buff *skb, struct net_device *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.
@@ -951,6 +946,16 @@ static int i2o_lan_packet_send(struct sk_buff *skb, struct net_device *dev)
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) {
@@ -963,12 +968,14 @@ static int i2o_lan_packet_send(struct sk_buff *skb, struct net_device *dev)
__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 << 3, msg+3); // TransmitControlWord
-
+ __raw_writel(1 << 30 | 1 << 3, msg+3); // TransmitControlWord
+ // bit 30: reply as soon as transmission attempt is complete
+ // bit 3: Supress 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 (batching(priv, !tickssofar) && !priv->send_active) {
+
+ if ((priv->tx_batch_mode & 0x01) && !priv->send_active) {
priv->send_active = 1;
queue_task(&priv->i2o_batch_send_task, &tq_scheduler);
}
@@ -984,9 +991,9 @@ static int i2o_lan_packet_send(struct sk_buff *skb, struct net_device *dev)
__raw_writel(virt_to_bus(skb->data), sgl_elem+2);
}
- /* If tx not in batch mode or frame is full, send immediatelly */
+ /* If tx is in immediate mode or frame is full, send now */
- if (!batching(priv, !tickssofar) || priv->tx_count == priv->sgl_max) {
+ 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);
@@ -1141,12 +1148,14 @@ static void i2o_lan_set_mc_list(struct net_device *dev)
u32 max_size_mc_table;
u32 mc_addr_group[64];
-// This isn't safe yet. Needs to be async.
+// This isn't safe yet in SMP. Needs to be async.
+// Seems to work in uniprocessor environment.
+
return;
// read_lock_bh(&dev_mc_lock);
-// spin_lock(&dev->xmit_lock);
-// dev->xmit_lock_owner = smp_processor_id();
+ spin_lock(&dev->xmit_lock);
+ dev->xmit_lock_owner = smp_processor_id();
if (i2o_query_scalar(iop, i2o_dev->lct_data.tid, 0x0001, -1,
&mc_addr_group, sizeof(mc_addr_group)) < 0 ) {
@@ -1212,15 +1221,13 @@ static struct tq_struct i2o_lan_set_mc_list_task = {
* Queue routine i2o_lan_set_mc_list() to be called later.
* Needs to be async.
*/
-
static void i2o_lan_set_multicast_list(struct net_device *dev)
{
- if (!in_interrupt()) {
+ if (in_interrupt()) {
i2o_lan_set_mc_list_task.data = (void *)dev;
queue_task(&i2o_lan_set_mc_list_task, &tq_scheduler);
- } else {
+ } else
i2o_lan_set_mc_list(dev);
- }
}
/*
@@ -1236,10 +1243,20 @@ static int i2o_lan_change_mtu(struct net_device *dev, int new_mtu)
0x0000, 6, &max_pkt_size, 4) < 0)
return -EFAULT;
- if (new_mtu < 68 || max_pkt_size < new_mtu)
+ 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;
}
@@ -1287,15 +1304,13 @@ struct net_device *i2o_lan_register_device(struct i2o_device *i2o_dev)
#ifdef CONFIG_FDDI
case I2O_LAN_FDDI:
{
- int size = sizeof(struct net_device) + sizeof(struct i2o_lan_local)
- + sizeof("fddi%d ");
+ 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);
- dev->name = (char *)(dev + 1) + sizeof(struct i2o_lan_local);
if (dev_alloc_name(dev, "fddi%d") < 0) {
printk(KERN_WARNING "i2o_lan: Too many FDDI devices.\n");
@@ -1335,7 +1350,7 @@ struct net_device *i2o_lan_register_device(struct i2o_device *i2o_dev)
priv = (struct i2o_lan_local *)dev->priv;
priv->i2o_dev = i2o_dev;
priv->type_trans = type_trans;
- priv->sgl_max = (i2o_dev->controller->inbound_size - 16) / 12;
+ 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 */
@@ -1344,7 +1359,7 @@ struct net_device *i2o_lan_register_device(struct i2o_device *i2o_dev)
priv->max_buckets_out = max_buckets_out;
priv->bucket_thresh = bucket_thresh;
priv->rx_copybreak = rx_copybreak;
- priv->tx_batch_mode = tx_batch_mode;
+ priv->tx_batch_mode = tx_batch_mode & 0x03;
priv->i2o_event_mask = i2o_event_mask;
priv->tx_lock = SPIN_LOCK_UNLOCKED;
@@ -1362,7 +1377,6 @@ struct net_device *i2o_lan_register_device(struct i2o_device *i2o_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]);
@@ -1417,9 +1431,9 @@ int __init i2o_lan_init(void)
struct net_device *dev;
int i;
- printk(KERN_INFO "I2O LAN OSM (c) 1999 University of Helsinki.\n");
+ printk(KERN_INFO "I2O LAN OSM (C) 1999 University of Helsinki.\n");
- /* Module params used as global defaults for private values */
+ /* 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;
@@ -1549,7 +1563,7 @@ 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, "0-1" "i");
-MODULE_PARM_DESC(tx_batch_mode, "0=Use immediate mode send, 1=Use batch mode send");
+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