From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH] Bound TSO defer time (resend) Date: Mon, 16 Oct 2006 22:35:13 -0700 (PDT) Message-ID: <20061016.223513.35356292.davem@davemloft.net> References: <20061016202035.6d55b96e@localhost.localdomain> <45345999.4000300@psc.edu> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: shemminger@osdl.org, netdev@vger.kernel.org Return-path: Received: from dsl027-180-168.sfo1.dsl.speakeasy.net ([216.27.180.168]:16833 "EHLO sunset.davemloft.net") by vger.kernel.org with ESMTP id S1423115AbWJQFfM (ORCPT ); Tue, 17 Oct 2006 01:35:12 -0400 To: jheffner@psc.edu In-Reply-To: <45345999.4000300@psc.edu> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org From: John Heffner Date: Tue, 17 Oct 2006 00:18:33 -0400 > Stephen Hemminger wrote: > > On Mon, 16 Oct 2006 20:53:20 -0400 (EDT) > > John Heffner wrote: > > >> This patch limits the amount of time you will defer sending a TSO segment > >> to less than two clock ticks, or the time between two acks, whichever is > >> longer. > > > > > Okay, but doing any timing on clock ticks makes the behavior dependent > > on the value of HZ which doesn't seem desirable. Should this be based > > on RTT or a real-time values? > > It would be nice to use a high res clock so you don't depend on HZ, but > this is still expensive on most SMP arch's as I understand it. Right so we do need to use a jiffies based solution. Since HZ is variable, I have a feeling that the thing to do here is pick some timeout in msec. Then replace the "2 clock ticks" with some msec_to_jiffies() calls, bottoming out at 1 jiffie. How does that sound?