diff options
author | Eric Dumazet <edumazet@google.com> | 2012-05-18 17:51:44 -0400 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2012-05-19 04:02:12 -0400 |
commit | a34a101e1e6365638b02ea83a38d7a4cb228dc04 (patch) | |
tree | 6757af0793572021c9284bccdae87a8a737e62ab /net/ipv6 | |
parent | a1e8b307986ab27b7608f107aec71d3569650f46 (diff) |
ipv6: disable GSO on sockets hitting dst_allfrag
If the allfrag feature has been set on a host route (due to an ICMPv6
Packet Too Big received indicating a MTU of less than 1280), we hit a
very slow behavior in TCP stack, because all big packets are dropped and
only a retransmit timer is able to push one MSS frame every 200 ms.
One way to handle this is to disable GSO on the socket the first time a
super packet is dropped. Adding a specific dst_allfrag() in the fast
path is probably overkill since the dst_allfrag() case almost never
happen.
Result on netperf TCP_STREAM, one flow :
Before : 60 kbit/sec
After : 1.6 Gbit/sec
Reported-by: Tore Anderson <tore@fud.no>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Tore Anderson <tore@fud.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv6')
-rw-r--r-- | net/ipv6/ip6_output.c | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c index 3dc633f39372..d99fdc699625 100644 --- a/net/ipv6/ip6_output.c +++ b/net/ipv6/ip6_output.c | |||
@@ -643,7 +643,10 @@ int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *)) | |||
643 | /* We must not fragment if the socket is set to force MTU discovery | 643 | /* We must not fragment if the socket is set to force MTU discovery |
644 | * or if the skb it not generated by a local socket. | 644 | * or if the skb it not generated by a local socket. |
645 | */ | 645 | */ |
646 | if (!skb->local_df && skb->len > mtu) { | 646 | if (unlikely(!skb->local_df && skb->len > mtu)) { |
647 | if (skb->sk && dst_allfrag(skb_dst(skb))) | ||
648 | sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK); | ||
649 | |||
647 | skb->dev = skb_dst(skb)->dev; | 650 | skb->dev = skb_dst(skb)->dev; |
648 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); | 651 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
649 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), | 652 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |