From owner-freebsd-stable Wed Jan 8 6:35:10 2003 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 9B32637B401 for ; Wed, 8 Jan 2003 06:35:08 -0800 (PST) Received: from herbelot.dyndns.org (herbelot.net1.nerim.net [62.212.117.213]) by mx1.FreeBSD.org (Postfix) with ESMTP id 3330A43EE1 for ; Wed, 8 Jan 2003 06:35:07 -0800 (PST) (envelope-from thierry@herbelot.com) Received: from diversion.herbelot.nom (diversion.herbelot.nom [192.168.2.6]) by herbelot.dyndns.org (8.12.6/8.12.6) with ESMTP id h08EVP0f020392; Wed, 8 Jan 2003 15:31:25 +0100 (CET) Content-Type: text/plain; charset="iso-8859-1" From: Thierry Herbelot To: Mike Tancsa , Janet Sullivan , freebsd-stable@FreeBSD.ORG Subject: Re: ATA corruption in DMA mode Date: Wed, 8 Jan 2003 15:34:56 +0100 User-Agent: KMail/1.4.3 References: <3E1C2EA6.18C86871@vzavenue.net> <5.2.0.9.0.20030108091446.05ae8730@marble.sentex.ca> In-Reply-To: <5.2.0.9.0.20030108091446.05ae8730@marble.sentex.ca> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-Id: <200301081534.56606.thierry@herbelot.com> Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Le Wednesday 08 January 2003 15:15, Mike Tancsa a écrit : > Can you boot with a 4.6 kernel now and test it again to confirm the problem > really does go away ? > > ---Mike > > At 07:05 AM 08/01/2003 -0700, Janet Sullivan wrote: > >P.S. I've seen absolutely no ata error messages (command timeout, > >etc.). > > > >Janet Sullivan wrote: > > > If I copy large (500 meg) files between directories with DMA on, the > > > copy becomes corrupt. If I turn DMA off and use PIO mode, the copy is [SNIP] The DMA may be more aggressive in its memory timings It took me some time to find out why disk copies of large files (typically CD-ROM ISO image files) went corrupt : I had two semi-broken SDRAM SIMMs. I could single out which SIMMs were bad with the "canonical" memory test : a string of "make buildworld", with different combinations of SIMMs (I even had to try : does this SIMM work when alone one a motherboard ?) HtH TfH PS : obviously, a broken SIMM should be broken for any release of FreeBSD To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message