lk/linux-kernel-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

[S390] qeth: preallocated qeth header for hiper socket

Browse Source 
For hiper socket devices this patch will economize the reallocation
of the tx skb data segment by allocating separate memory for the qdio
transport information (qeth header).

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This commit is contained in:
Frank Blaschka
2008-08-01 16:39:13 +02:00
committed by Martin Schwidefsky
parent a4b526b3ba
commit 683d718a89
4 changed files with 126 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
drivers/s390/net/qeth_l3_main.c
View File

@@ -2604,6 +2604,7 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
int tx_bytes = skb->len;
enum qeth_large_send_types large_send = QETH_LARGE_SEND_NO;
struct qeth_eddp_context *ctx = NULL;
int data_offset = -1;
if ((card->info.type == QETH_CARD_TYPE_IQD) &&
(skb->protocol != htons(ETH_P_IPV6)) &&
@@ -2624,14 +2625,28 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
card->perf_stats.outbound_start_time = qeth_get_micros();
}
/* create a clone with writeable headroom */
new_skb = skb_realloc_headroom(skb, sizeof(struct qeth_hdr_tso) +
VLAN_HLEN);
if (!new_skb)
goto tx_drop;
if (skb_is_gso(skb))
large_send = card->options.large_send;
if ((card->info.type == QETH_CARD_TYPE_IQD) && (!large_send) &&
(skb_shinfo(skb)->nr_frags == 0)) {
new_skb = skb;
data_offset = ETH_HLEN;
hdr = kmem_cache_alloc(qeth_core_header_cache, GFP_ATOMIC);
if (!hdr)
goto tx_drop;
elements_needed++;
} else {
/* create a clone with writeable headroom */
new_skb = skb_realloc_headroom(skb, sizeof(struct qeth_hdr_tso)
+ VLAN_HLEN);
if (!new_skb)
goto tx_drop;
}
if (card->info.type == QETH_CARD_TYPE_IQD) {
skb_pull(new_skb, ETH_HLEN);
if (data_offset < 0)
skb_pull(new_skb, ETH_HLEN);
} else {
if (new_skb->protocol == htons(ETH_P_IP)) {
if (card->dev->type == ARPHRD_IEEE802_TR)
@@ -2657,9 +2672,6 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
netif_stop_queue(dev);
if (skb_is_gso(new_skb))
large_send = card->options.large_send;
/* fix hardware limitation: as long as we do not have sbal
* chaining we can not send long frag lists so we temporary
* switch to EDDP
@@ -2677,9 +2689,16 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
qeth_tso_fill_header(card, hdr, new_skb);
elements_needed++;
} else {
hdr = (struct qeth_hdr *)skb_push(new_skb,
if (data_offset < 0) {
hdr = (struct qeth_hdr *)skb_push(new_skb,
sizeof(struct qeth_hdr));
qeth_l3_fill_header(card, hdr, new_skb, ipv, cast_type);
qeth_l3_fill_header(card, hdr, new_skb, ipv,
cast_type);
} else {
qeth_l3_fill_header(card, hdr, new_skb, ipv,
cast_type);
hdr->hdr.l3.length = new_skb->len - data_offset;
}
}
if (large_send == QETH_LARGE_SEND_EDDP) {
@@ -2695,8 +2714,11 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
} else {
int elems = qeth_get_elements_no(card, (void *)hdr, new_skb,
elements_needed);
if (!elems)
if (!elems) {
if (data_offset >= 0)
kmem_cache_free(qeth_core_header_cache, hdr);
goto tx_drop;
}
elements_needed += elems;
}
@@ -2709,7 +2731,7 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
elements_needed, ctx);
else
rc = qeth_do_send_packet_fast(card, queue, new_skb, hdr,
elements_needed, ctx);
elements_needed, ctx, data_offset, 0);
if (!rc) {
card->stats.tx_packets++;
@@ -2737,6 +2759,9 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
if (ctx != NULL)
qeth_eddp_put_context(ctx);
if (data_offset >= 0)
kmem_cache_free(qeth_core_header_cache, hdr);
if (rc == -EBUSY) {
if (new_skb != skb)
dev_kfree_skb_any(new_skb);