From mboxrd@z Thu Jan 1 00:00:00 1970 From: shemminger@oldman.hamilton.local Subject: [PATCH 3/3] TCP: htcp - use measured rtt Date: Wed, 25 Jul 2007 11:44:38 +0100 Message-ID: <20070725104826.760544771@oldman.hamilton.local> References: <20070725104434.998037763@oldman.hamilton.local> Cc: netdev@vger.kernel.org To: David Miller Return-path: Received: from smtp2.linux-foundation.org ([207.189.120.14]:44186 "EHLO smtp2.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758919AbXGYLBr (ORCPT ); Wed, 25 Jul 2007 07:01:47 -0400 Content-Disposition: inline; filename=htcp-fix.patch Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Change HTCP to use measured RTT rather than smooth RTT. Srtt is computed using the TCP receive timestamp options, so it is vulnerable to hostile receivers. To avoid any problems this might cause use the measured RTT instead. Signed-off-by: Stephen Hemminger --- a/net/ipv4/tcp_htcp.c 2007-07-19 08:26:40.000000000 +0100 +++ b/net/ipv4/tcp_htcp.c 2007-07-19 08:28:07.000000000 +0100 @@ -76,12 +76,11 @@ static u32 htcp_cwnd_undo(struct sock *s return max(tp->snd_cwnd, (tp->snd_ssthresh << 7) / ca->beta); } -static inline void measure_rtt(struct sock *sk) +static inline void measure_rtt(struct sock *sk, u32 srtt) { const struct inet_connection_sock *icsk = inet_csk(sk); const struct tcp_sock *tp = tcp_sk(sk); struct htcp *ca = inet_csk_ca(sk); - u32 srtt = tp->srtt >> 3; /* keep track of minimum RTT seen so far, minRTT is zero at first */ if (ca->minRTT > srtt || !ca->minRTT) @@ -108,6 +107,9 @@ static void measure_achieved_throughput( if (icsk->icsk_ca_state == TCP_CA_Open) ca->pkts_acked = pkts_acked; + if (rtt > 0) + measure_rtt(sk, usecs_to_jiffies(rtt)); + if (!use_bandwidth_switch) return; @@ -237,8 +239,6 @@ static void htcp_cong_avoid(struct sock if (tp->snd_cwnd <= tp->snd_ssthresh) tcp_slow_start(tp); else { - measure_rtt(sk); - /* In dangerous area, increase slowly. * In theory this is tp->snd_cwnd += alpha / tp->snd_cwnd */ --