From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ido Yariv Subject: [PATCH v6] net: tcp: Fix a PTO timing granularity issue Date: Thu, 28 May 2015 08:33:13 -0400 Message-ID: <1432816393-6941-1-git-send-email-ido@wizery.com> References: <063D6719AE5E284EB5DD2968C1650D6D1CB4175B@AcuExch.aculab.com> Cc: Ido Yariv , Ido Yariv To: "David S. Miller" , Alexey Kuznetsov , James Morris , Hideaki YOSHIFUJI , Patrick McHardy , Nandita Dukkipati , Eric Dumazet , David Laight , netdev@vger.kernel.org, linux-kernel@vger.kernel.org Return-path: Received: from mail-qg0-f42.google.com ([209.85.192.42]:36032 "EHLO mail-qg0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753791AbbE1Mdo (ORCPT ); Thu, 28 May 2015 08:33:44 -0400 Received: by qgf2 with SMTP id 2so15336553qgf.3 for ; Thu, 28 May 2015 05:33:43 -0700 (PDT) In-Reply-To: <063D6719AE5E284EB5DD2968C1650D6D1CB4175B@AcuExch.aculab.com> Sender: netdev-owner@vger.kernel.org List-ID: The Tail Loss Probe RFC specifies that the PTO value should be set to max(2 * SRTT, 10ms), where SRTT is the smoothed round-trip time. The PTO value is converted to jiffies, so the timer may expire prematurely. This is especially problematic on systems in which HZ <= 100, so work around this by setting the timeout to at least 2 jiffies on such systems. The 10ms figure was originally selected based on tests performed with the current implementation and HZ = 1000. Thus, leave the behavior on systems with HZ > 100 unchanged. Signed-off-by: Ido Yariv --- include/net/tcp.h | 20 ++++++++++++++++++++ net/ipv4/tcp_output.c | 2 +- 2 files changed, 21 insertions(+), 1 deletion(-) diff --git a/include/net/tcp.h b/include/net/tcp.h index 2bb2bad..2ff0181 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -1751,4 +1751,24 @@ static inline void skb_set_tcp_pure_ack(struct sk_buff *skb) skb->truesize = 2; } +/* Convert msecs to jiffies, ensuring that the return value is at least 2 + * jiffies. + * This can be used when setting tick-based timers to guarantee that they won't + * expire right away. + */ +static inline unsigned long tcp_msecs_to_jiffies_min_2(const unsigned int m) +{ + if (__builtin_constant_p(m)) { + /* The theoretical upper bound of m for 2 jiffies is 2 seconds, + * so compare m with that to avoid potential integer overflows. + */ + if ((m > 2 * MSEC_PER_SEC) || (m * HZ > 2 * MSEC_PER_SEC)) + return msecs_to_jiffies(m); + + return 2; + } + + return max_t(u32, 2, msecs_to_jiffies(m)); +} + #endif /* _TCP_H */ diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 190538a..37694d2 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -2207,7 +2207,7 @@ bool tcp_schedule_loss_probe(struct sock *sk) if (tp->packets_out == 1) timeout = max_t(u32, timeout, (rtt + (rtt >> 1) + TCP_DELACK_MAX)); - timeout = max_t(u32, timeout, msecs_to_jiffies(10)); + timeout = max_t(u32, timeout, tcp_msecs_to_jiffies_min_2(10)); /* If RTO is shorter, just schedule TLP in its place. */ tlp_time_stamp = tcp_time_stamp + timeout; -- 2.1.0