From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [PATCH v2 net-next] tcp: switch rtt estimations to usec resolution Date: Thu, 27 Feb 2014 21:20:25 -0800 Message-ID: <87fvn3dd9i.fsf@tassilo.jf.intel.com> References: <1393137487.2316.48.camel@edumazet-glaptop2.roam.corp.google.com> <1393178122.2316.50.camel@edumazet-glaptop2.roam.corp.google.com> <1393181458.2316.56.camel@edumazet-glaptop2.roam.corp.google.com> Mime-Version: 1.0 Content-Type: text/plain Cc: Julian Anastasov , David Miller , netdev , Yuchung Cheng , Neal Cardwell , Larry Brakmo To: Eric Dumazet Return-path: Received: from mga11.intel.com ([192.55.52.93]:58569 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751342AbaB1FU1 (ORCPT ); Fri, 28 Feb 2014 00:20:27 -0500 In-Reply-To: <1393181458.2316.56.camel@edumazet-glaptop2.roam.corp.google.com> (Eric Dumazet's message of "Sun, 23 Feb 2014 10:50:58 -0800") Sender: netdev-owner@vger.kernel.org List-ID: Eric Dumazet writes: > > - rtt_update = tcp_ack_update_rtt(sk, flag, seq_rtt, sack_rtt); > + if (first_ackt.tv64) { > + ktime_t curt = ktime_get_real(); Sorry I'm late to the party. One problem is that if a system doesn't use TSC for some reason this will be incredibly slow (unfortunately this happens for various reasons). Perhaps we need a ktime_get_real_but_use_jiffies_if_slow() or somesuch. -andi -- ak@linux.intel.com -- Speaking for myself only