From owner-freebsd-hackers@FreeBSD.ORG Mon Sep 29 12:22:34 2008 Return-Path: Delivered-To: freebsd-hackers@FreeBSD.ORG Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 255611065687; Mon, 29 Sep 2008 12:22:34 +0000 (UTC) (envelope-from rwatson@FreeBSD.org) Received: from cyrus.watson.org (cyrus.watson.org [65.122.17.42]) by mx1.freebsd.org (Postfix) with ESMTP id 0055B8FC22; Mon, 29 Sep 2008 12:22:34 +0000 (UTC) (envelope-from rwatson@FreeBSD.org) Received: from fledge.watson.org (fledge.watson.org [65.122.17.41]) by cyrus.watson.org (Postfix) with ESMTP id 8A3B546B8C; Mon, 29 Sep 2008 08:22:33 -0400 (EDT) Date: Mon, 29 Sep 2008 13:22:33 +0100 (BST) From: Robert Watson X-X-Sender: robert@fledge.watson.org To: freebsd-hackers@FreeBSD.ORG, freebsd-stable@FreeBSD.ORG, danny@cs.huji.ac.il, Jeremy Chadwick , kris@FreeBSD.org In-Reply-To: <200809291158.m8TBwped076763@lurza.secnetix.de> Message-ID: References: <200809291158.m8TBwped076763@lurza.secnetix.de> User-Agent: Alpine 1.10 (BSF 962 2008-03-14) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: Subject: Re: bad NFS/UDP performance X-BeenThere: freebsd-hackers@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Technical Discussions relating to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 29 Sep 2008 12:22:34 -0000 On Mon, 29 Sep 2008, Oliver Fromme wrote: > Danny Braniss wrote: > > Grr, there goes binary search theory out of the window, > > So far I have managed to pinpoint the day that the changes affect the > > throughput: > > 18/08/08 00:00:00 19/08/08 00:00:00 > > (I assume cvs's date is GMT). > > now would be a good time for some help, specially how to undo changes, my > > knowledge of csup/cvs are close to zero. > > So you've nailed to down to this 24-hour window: > > http://www.secnetix.de/olli/FreeBSD/svnews/?day=2008-08-18&p=/stable/7 > > I'm afraid that r181822 by rwatson is the most likely candidate that might > be causing the regression. If we can confirm that it was that specific change, then I can create a patch to restore exclusive locking for UDP and we can see if it was the general move to rwlocking, or specifically the read-locking of certain data structures. Perhaps what we've done is moved contention from a mutex to a sleep lock, reducing the efficiency of handling contention? Adding Kris to the CC line because he often has useful insights on this sort of thing. Robert N M Watson Computer Laboratory University of Cambridge