From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH v2] tcp: avoid cwnd moderation in undo Date: Tue, 22 Mar 2011 19:37:40 -0700 (PDT) Message-ID: <20110322.193740.179920597.davem@davemloft.net> References: <201103141106.19679.carsten@wolffcarsten.de> <1300136223-855-1-git-send-email-ycheng@google.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: ilpo.jarvinen@helsinki.fi, nanditad@google.com, netdev@vger.kernel.org To: ycheng@google.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:46257 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752412Ab1CWChC (ORCPT ); Tue, 22 Mar 2011 22:37:02 -0400 In-Reply-To: <1300136223-855-1-git-send-email-ycheng@google.com> Sender: netdev-owner@vger.kernel.org List-ID: From: Yuchung Cheng Date: Mon, 14 Mar 2011 13:57:03 -0700 > In the current undo logic, cwnd is moderated after it was restored > to the value prior entering fast-recovery. It was moderated first > in tcp_try_undo_recovery then again in tcp_complete_cwr. > > Since the undo indicates recovery was false, these moderations > are not necessary. If the undo is triggered when most of the > outstanding data have been acknowledged, the (restored) cwnd is > falsely pulled down to a small value. > > This patch removes these cwnd moderations if cwnd is undone > a) during fast-recovery > b) by receiving DSACKs past fast-recovery > > Signed-off-by: Yuchung Cheng Applied, thank you. I followed it up with this tidy up: -------------------- tcp: Make undo_ssthresh arg to tcp_undo_cwr() a bool. Signed-off-by: David S. Miller --- net/ipv4/tcp_input.c | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-) diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index e72af7e..bef9f04 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2659,7 +2659,7 @@ static void DBGUNDO(struct sock *sk, const char *msg) #define DBGUNDO(x...) do { } while (0) #endif -static void tcp_undo_cwr(struct sock *sk, const int undo_ssthresh) +static void tcp_undo_cwr(struct sock *sk, const bool undo_ssthresh) { struct tcp_sock *tp = tcp_sk(sk); @@ -2698,7 +2698,7 @@ static int tcp_try_undo_recovery(struct sock *sk) * or our original transmission succeeded. */ DBGUNDO(sk, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans"); - tcp_undo_cwr(sk, 1); + tcp_undo_cwr(sk, true); if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss) mib_idx = LINUX_MIB_TCPLOSSUNDO; else @@ -2725,7 +2725,7 @@ static void tcp_try_undo_dsack(struct sock *sk) if (tp->undo_marker && !tp->undo_retrans) { DBGUNDO(sk, "D-SACK"); - tcp_undo_cwr(sk, 1); + tcp_undo_cwr(sk, true); tp->undo_marker = 0; NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPDSACKUNDO); } @@ -2778,7 +2778,7 @@ static int tcp_try_undo_partial(struct sock *sk, int acked) tcp_update_reordering(sk, tcp_fackets_out(tp) + acked, 1); DBGUNDO(sk, "Hoe"); - tcp_undo_cwr(sk, 0); + tcp_undo_cwr(sk, false); NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPPARTIALUNDO); /* So... Do not make Hoe's retransmit yet. @@ -2807,7 +2807,7 @@ static int tcp_try_undo_loss(struct sock *sk) DBGUNDO(sk, "partial loss"); tp->lost_out = 0; - tcp_undo_cwr(sk, 1); + tcp_undo_cwr(sk, true); NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPLOSSUNDO); inet_csk(sk)->icsk_retransmits = 0; tp->undo_marker = 0; @@ -3496,7 +3496,7 @@ static void tcp_undo_spur_to_response(struct sock *sk, int flag) if (flag & FLAG_ECE) tcp_ratehalving_spur_to_response(sk); else - tcp_undo_cwr(sk, 1); + tcp_undo_cwr(sk, true); } /* F-RTO spurious RTO detection algorithm (RFC4138) -- 1.7.4.1