* [PATCH] tcp: properly increase rcv_ssthresh for ofo packets
@ 2013-09-06 17:35 Eric Dumazet
2013-09-06 18:48 ` David Miller
0 siblings, 1 reply; 2+ messages in thread
From: Eric Dumazet @ 2013-09-06 17:35 UTC (permalink / raw)
To: David Miller; +Cc: netdev, Neal Cardwell
From: Eric Dumazet <edumazet@google.com>
TCP receive window handling is multi staged.
A socket has a memory budget, static or dynamic, in sk_rcvbuf.
Because we do not really know how this memory budget translates to
a TCP window (payload), TCP announces a small initial window
(about 20 MSS).
When a packet is received, we increase TCP rcv_win depending
on the payload/truesize ratio of this packet. Good citizen
packets give a hint that it's reasonable to have rcv_win = sk_rcvbuf/2
This heuristic takes place in tcp_grow_window()
Problem is : We currently call tcp_grow_window() only for in-order
packets.
This means that reorders or packet losses stop proper grow of
rcv_win, and senders are unable to benefit from fast recovery,
or proper reordering level detection.
Really, a packet being stored in OFO queue is not a bad citizen.
It should be part of the game as in-order packets.
In our traces, we very often see sender is limited by linux small
receive windows, even if linux hosts use autotuning (DRS) and should
allow rcv_win to grow to ~3MB.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
---
net/ipv4/tcp_input.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 1969e16..28708d3 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -4141,6 +4141,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
if (!tcp_try_coalesce(sk, skb1, skb, &fragstolen)) {
__skb_queue_after(&tp->out_of_order_queue, skb1, skb);
} else {
+ tcp_grow_window(sk, skb);
kfree_skb_partial(skb, fragstolen);
skb = NULL;
}
@@ -4216,8 +4217,10 @@ add_sack:
if (tcp_is_sack(tp))
tcp_sack_new_ofo_skb(sk, seq, end_seq);
end:
- if (skb)
+ if (skb) {
+ tcp_grow_window(sk, skb);
skb_set_owner_r(skb, sk);
+ }
}
static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int hdrlen,
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] tcp: properly increase rcv_ssthresh for ofo packets
2013-09-06 17:35 [PATCH] tcp: properly increase rcv_ssthresh for ofo packets Eric Dumazet
@ 2013-09-06 18:48 ` David Miller
0 siblings, 0 replies; 2+ messages in thread
From: David Miller @ 2013-09-06 18:48 UTC (permalink / raw)
To: eric.dumazet; +Cc: netdev, ncardwell
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Fri, 06 Sep 2013 10:35:58 -0700
> From: Eric Dumazet <edumazet@google.com>
>
> TCP receive window handling is multi staged.
>
> A socket has a memory budget, static or dynamic, in sk_rcvbuf.
>
> Because we do not really know how this memory budget translates to
> a TCP window (payload), TCP announces a small initial window
> (about 20 MSS).
>
> When a packet is received, we increase TCP rcv_win depending
> on the payload/truesize ratio of this packet. Good citizen
> packets give a hint that it's reasonable to have rcv_win = sk_rcvbuf/2
>
> This heuristic takes place in tcp_grow_window()
>
> Problem is : We currently call tcp_grow_window() only for in-order
> packets.
>
> This means that reorders or packet losses stop proper grow of
> rcv_win, and senders are unable to benefit from fast recovery,
> or proper reordering level detection.
>
> Really, a packet being stored in OFO queue is not a bad citizen.
> It should be part of the game as in-order packets.
>
> In our traces, we very often see sender is limited by linux small
> receive windows, even if linux hosts use autotuning (DRS) and should
> allow rcv_win to grow to ~3MB.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Acked-by: Neal Cardwell <ncardwell@google.com>
Applied.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2013-09-06 18:48 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-09-06 17:35 [PATCH] tcp: properly increase rcv_ssthresh for ofo packets Eric Dumazet
2013-09-06 18:48 ` David Miller
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.