gro: Optimise Ethernet header comparison
This patch optimises the Ethernet header comparison to use 2-byte
and 4-byte xors instead of memcmp. In order to facilitate this,
the actual comparison is now carried out by the callers of the
shared dev_gro_receive function.
This has a significant impact when receiving 1500B packets through
10GbE.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/net/core/dev.c b/net/core/dev.c
index ae0b669..1e27a67 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -215,13 +215,6 @@
return &net->dev_index_head[ifindex & ((1 << NETDEV_HASHBITS) - 1)];
}
-static inline void *skb_gro_mac_header(struct sk_buff *skb)
-{
- return skb_mac_header(skb) < skb->data ? skb_mac_header(skb) :
- page_address(skb_shinfo(skb)->frags[0].page) +
- skb_shinfo(skb)->frags[0].page_offset;
-}
-
/* Device list insertion */
static int list_netdevice(struct net_device *dev)
{
@@ -2415,29 +2408,16 @@
rcu_read_lock();
list_for_each_entry_rcu(ptype, head, list) {
- struct sk_buff *p;
- void *mac;
-
if (ptype->type != type || ptype->dev || !ptype->gro_receive)
continue;
skb_set_network_header(skb, skb_gro_offset(skb));
- mac = skb_gro_mac_header(skb);
mac_len = skb->network_header - skb->mac_header;
skb->mac_len = mac_len;
NAPI_GRO_CB(skb)->same_flow = 0;
NAPI_GRO_CB(skb)->flush = 0;
NAPI_GRO_CB(skb)->free = 0;
- for (p = napi->gro_list; p; p = p->next) {
- if (!NAPI_GRO_CB(p)->same_flow)
- continue;
-
- if (p->mac_len != mac_len ||
- memcmp(skb_mac_header(p), mac, mac_len))
- NAPI_GRO_CB(p)->same_flow = 0;
- }
-
pp = ptype->gro_receive(&napi->gro_list, skb);
break;
}
@@ -2492,7 +2472,8 @@
struct sk_buff *p;
for (p = napi->gro_list; p; p = p->next) {
- NAPI_GRO_CB(p)->same_flow = 1;
+ NAPI_GRO_CB(p)->same_flow = !compare_ether_header(
+ skb_mac_header(p), skb_gro_mac_header(skb));
NAPI_GRO_CB(p)->flush = 0;
}