Commit 0d613a27 authored by Frank Pavlic's avatar Frank Pavlic Committed by Jeff Garzik

[PATCH] s390: lcs performance enhancements

[patch 1/2] s390: lcs performance enhancements

From: Klaus Wacker <kdwacker@de.ibm.com>
	- When flood pinging (with large packet size) an LCS device,
	  about 90 % of all packets are dropped by driver.
	- increased number of lcs IO buffers to 32.
	- use netif_stop_queue/netif_wake_queue in lcs_start_xmit routine
	- don't lock the whole xmit routine but just the piece of code where
	  tx_buffer is touched.
Signed-off-by: default avatarFrank Pavlic <fpavlic@de.ibm.com>

diffstat:
 lcs.c |   31 +++++++++++++++++--------------
 lcs.h |    2 +-
 2 files changed, 18 insertions(+), 15 deletions(-)
Signed-off-by: default avatarJeff Garzik <jgarzik@pobox.com>
parent ca5b0ec8
......@@ -98,9 +98,9 @@ lcs_register_debug_facility(void)
return -ENOMEM;
}
debug_register_view(lcs_dbf_setup, &debug_hex_ascii_view);
debug_set_level(lcs_dbf_setup, 4);
debug_set_level(lcs_dbf_setup, 2);
debug_register_view(lcs_dbf_trace, &debug_hex_ascii_view);
debug_set_level(lcs_dbf_trace, 4);
debug_set_level(lcs_dbf_trace, 2);
return 0;
}
......@@ -1292,9 +1292,8 @@ lcs_set_multicast_list(struct net_device *dev)
LCS_DBF_TEXT(4, trace, "setmulti");
card = (struct lcs_card *) dev->priv;
if (!lcs_set_thread_start_bit(card, LCS_SET_MC_THREAD)) {
if (!lcs_set_thread_start_bit(card, LCS_SET_MC_THREAD))
schedule_work(&card->kernel_thread_starter);
}
}
#endif /* CONFIG_IP_MULTICAST */
......@@ -1459,6 +1458,8 @@ lcs_txbuffer_cb(struct lcs_channel *channel, struct lcs_buffer *buffer)
lcs_release_buffer(channel, buffer);
card = (struct lcs_card *)
((char *) channel - offsetof(struct lcs_card, write));
if (netif_queue_stopped(card->dev))
netif_wake_queue(card->dev);
spin_lock(&card->lock);
card->tx_emitted--;
if (card->tx_emitted <= 0 && card->tx_buffer != NULL)
......@@ -1478,6 +1479,7 @@ __lcs_start_xmit(struct lcs_card *card, struct sk_buff *skb,
struct net_device *dev)
{
struct lcs_header *header;
int rc = 0;
LCS_DBF_TEXT(5, trace, "hardxmit");
if (skb == NULL) {
......@@ -1492,10 +1494,8 @@ __lcs_start_xmit(struct lcs_card *card, struct sk_buff *skb,
card->stats.tx_carrier_errors++;
return 0;
}
if (netif_queue_stopped(dev) ) {
card->stats.tx_dropped++;
return -EBUSY;
}
netif_stop_queue(card->dev);
spin_lock(&card->lock);
if (card->tx_buffer != NULL &&
card->tx_buffer->count + sizeof(struct lcs_header) +
skb->len + sizeof(u16) > LCS_IOBUFFERSIZE)
......@@ -1506,7 +1506,8 @@ __lcs_start_xmit(struct lcs_card *card, struct sk_buff *skb,
card->tx_buffer = lcs_get_buffer(&card->write);
if (card->tx_buffer == NULL) {
card->stats.tx_dropped++;
return -EBUSY;
rc = -EBUSY;
goto out;
}
card->tx_buffer->callback = lcs_txbuffer_cb;
card->tx_buffer->count = 0;
......@@ -1518,13 +1519,18 @@ __lcs_start_xmit(struct lcs_card *card, struct sk_buff *skb,
header->type = card->lan_type;
header->slot = card->portno;
memcpy(header + 1, skb->data, skb->len);
spin_unlock(&card->lock);
card->stats.tx_bytes += skb->len;
card->stats.tx_packets++;
dev_kfree_skb(skb);
if (card->tx_emitted <= 0)
netif_wake_queue(card->dev);
spin_lock(&card->lock);
if (card->tx_emitted <= 0 && card->tx_buffer != NULL)
/* If this is the first tx buffer emit it immediately. */
__lcs_emit_txbuffer(card);
return 0;
out:
spin_unlock(&card->lock);
return rc;
}
static int
......@@ -1535,9 +1541,7 @@ lcs_start_xmit(struct sk_buff *skb, struct net_device *dev)
LCS_DBF_TEXT(5, trace, "pktxmit");
card = (struct lcs_card *) dev->priv;
spin_lock(&card->lock);
rc = __lcs_start_xmit(card, skb, dev);
spin_unlock(&card->lock);
return rc;
}
......@@ -2319,7 +2323,6 @@ __init lcs_init_module(void)
PRINT_ERR("Initialization failed\n");
return rc;
}
return 0;
}
......
......@@ -95,7 +95,7 @@ do { \
*/
#define LCS_ILLEGAL_OFFSET 0xffff
#define LCS_IOBUFFERSIZE 0x5000
#define LCS_NUM_BUFFS 8 /* needs to be power of 2 */
#define LCS_NUM_BUFFS 32 /* needs to be power of 2 */
#define LCS_MAC_LENGTH 6
#define LCS_INVALID_PORT_NO -1
#define LCS_LANCMD_TIMEOUT_DEFAULT 5
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment