Commit aa6320d3 authored by Herbert Xu's avatar Herbert Xu Committed by David S. Miller

gro: Optimise TCP packet reception

gro: Optimise TCP packet reception

As this function can be called more than half a million times for
10GbE, it's important to optimise it as much as we can.

This patch uses bit ops to logical ops, as well as open coding
memcmp to exploit alignment properties.
Signed-off-by: default avatarHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent a5ad24be
...@@ -2478,9 +2478,9 @@ struct sk_buff **tcp_gro_receive(struct sk_buff **head, struct sk_buff *skb) ...@@ -2478,9 +2478,9 @@ struct sk_buff **tcp_gro_receive(struct sk_buff **head, struct sk_buff *skb)
struct tcphdr *th2; struct tcphdr *th2;
unsigned int thlen; unsigned int thlen;
unsigned int flags; unsigned int flags;
unsigned int total;
unsigned int mss = 1; unsigned int mss = 1;
int flush = 1; int flush = 1;
int i;
th = skb_gro_header(skb, sizeof(*th)); th = skb_gro_header(skb, sizeof(*th));
if (unlikely(!th)) if (unlikely(!th))
...@@ -2504,7 +2504,7 @@ struct sk_buff **tcp_gro_receive(struct sk_buff **head, struct sk_buff *skb) ...@@ -2504,7 +2504,7 @@ struct sk_buff **tcp_gro_receive(struct sk_buff **head, struct sk_buff *skb)
th2 = tcp_hdr(p); th2 = tcp_hdr(p);
if (th->source != th2->source || th->dest != th2->dest) { if ((th->source ^ th2->source) | (th->dest ^ th2->dest)) {
NAPI_GRO_CB(p)->same_flow = 0; NAPI_GRO_CB(p)->same_flow = 0;
continue; continue;
} }
...@@ -2519,14 +2519,15 @@ found: ...@@ -2519,14 +2519,15 @@ found:
flush |= flags & TCP_FLAG_CWR; flush |= flags & TCP_FLAG_CWR;
flush |= (flags ^ tcp_flag_word(th2)) & flush |= (flags ^ tcp_flag_word(th2)) &
~(TCP_FLAG_CWR | TCP_FLAG_FIN | TCP_FLAG_PSH); ~(TCP_FLAG_CWR | TCP_FLAG_FIN | TCP_FLAG_PSH);
flush |= th->ack_seq != th2->ack_seq || th->window != th2->window; flush |= (th->ack_seq ^ th2->ack_seq) | (th->window ^ th2->window);
flush |= memcmp(th + 1, th2 + 1, thlen - sizeof(*th)); for (i = sizeof(*th); !flush && i < thlen; i += 4)
flush |= *(u32 *)((u8 *)th + i) ^
*(u32 *)((u8 *)th2 + i);
total = skb_gro_len(p);
mss = skb_shinfo(p)->gso_size; mss = skb_shinfo(p)->gso_size;
flush |= skb_gro_len(skb) > mss || !skb_gro_len(skb); flush |= (skb_gro_len(skb) > mss) | !skb_gro_len(skb);
flush |= ntohl(th2->seq) + total != ntohl(th->seq); flush |= (ntohl(th2->seq) + skb_gro_len(p)) ^ ntohl(th->seq);
if (flush || skb_gro_receive(head, skb)) { if (flush || skb_gro_receive(head, skb)) {
mss = 1; mss = 1;
......
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