From owner-freebsd-questions@FreeBSD.ORG Fri Nov 26 13:05:05 2004 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id BA36916A4CE; Fri, 26 Nov 2004 13:05:05 +0000 (GMT) Received: from lorna.circlesquared.com (host217-45-219-85.in-addr.btopenworld.com [217.45.219.85]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9773743D58; Fri, 26 Nov 2004 13:05:04 +0000 (GMT) (envelope-from peter@circlesquared.com) Received: from [127.0.0.1] (localhost.circlesquared.com [127.0.0.1]) iAQD4XXt058225; Fri, 26 Nov 2004 13:04:42 GMT (envelope-from peter@circlesquared.com) Message-ID: <41A729E1.1060505@circlesquared.com> Date: Fri, 26 Nov 2004 13:04:33 +0000 From: Peter Risdon User-Agent: Mozilla/5.0 (X11; U; FreeBSD i386; en-US; rv:1.7.3) Gecko/20041124 X-Accept-Language: en-us, en MIME-Version: 1.0 To: craig@small-pla.net References: <000001c4d3af$0dd69240$6500a8c0@jupiter> <41A725C2.30204@circlesquared.com> In-Reply-To: <41A725C2.30204@circlesquared.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit cc: freebsd-questions@freebsd.org cc: freebsd-hardware@freebsd.org Subject: Re: WRITE_DMA failures on 5.3 (but NOT on 4.10) X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 26 Nov 2004 13:05:05 -0000 Peter Risdon wrote: > craig wrote: > >> hi, >> >> i wrote about this issue some weeks back, but have still not yet >> adequately >> resolved it. >> (http://lists.freebsd.org/mailman/htdig/freebsd-questions/2004-November/0638 >> >> 07.html) > > [...] > >> to repeat the original problem, when installing 5.3R it fails about >> 12% into >> extracting base into \ >> on the emergency terminal, there is a stream of warnings and failures >> reading: >> >> WARNING : WRITE_DMA UDMA ICRC ERROR > FAILURE : WRITE_DMA status = 51... blah >> this continues until i run out of patience. > > > This is a really major problem that has affected every 5.3 and the more > recent 5.2.1 machines I've operated with largish [1] hard drives. The > novelty of losing several tens of gigs of data any time a drive gets > busy wears off fairly quickly. > >> >> the advice i received was : > > > ... mainly about checking hardware, and this is _not_ the issue. I've > googled extensively on this and, as you did, replaced every hardware > component in the IDE lines, including the disk drives, without affecting > the problem. > > So far as I can make out, there was a change to default settings at some > point (I haven't scoured the CVS repository to find out exactly when) to > enable DMA because some newer drives require this[2]. > No - apologies for wasting bandwidth. I got to this stage of research very late a couple of nights ago and see I should have stopped a few hours earlier. Looking again, this: hw.ata.atapi_dma: 0 in loader.conf might fix the problem with atapi drives but the hw.ata.ata_dma: 1 sysctl setting seems to have been the default in 4.10 too, so that can't be it. I think I might try turning off ata dma in a 5.3 system anyway, and putting a big drive under load to see what happens, but I fear I'm probably back to square one. Peter. -- the circle squared network systems and software http://www.circlesquared.com