From owner-freebsd-current@FreeBSD.ORG Mon Dec 10 04:44:29 2007 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 0857416A418 for ; Mon, 10 Dec 2007 04:44:29 +0000 (UTC) (envelope-from bofh@terranova.net) Received: from tog.net (tog.net [216.89.226.5]) by mx1.freebsd.org (Postfix) with ESMTP id CFAF413C4D3 for ; Mon, 10 Dec 2007 04:44:28 +0000 (UTC) (envelope-from bofh@terranova.net) Received: from [216.89.228.170] (host-216-89-228-170.wireless.terranova.net [216.89.228.170]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by tog.net (Postfix) with ESMTP id 1B21929B5FA; Sun, 9 Dec 2007 23:44:27 -0500 (EST) Message-ID: <475CC426.3060808@terranova.net> Date: Sun, 09 Dec 2007 23:44:22 -0500 From: Travis Mikalson Organization: TerraNovaNet Internet Services User-Agent: Thunderbird 1.5.0.13 (Windows/20070809) MIME-Version: 1.0 To: =?ISO-8859-1?Q?S=F8ren_Schmidt?= , freebsd-current@freebsd.org References: <73807.10710.qm@web63912.mail.re1.yahoo.com> <200711280842.09340.jhb@freebsd.org> <474D726A.8080807@deepcore.dk> <200711280938.38545.jhb@freebsd.org> <474E5B69.7070406@yandex.ru> <474E65D6.4040403@deepcore.dk> <474E69AE.7000105@yandex.ru> <475978E1.2090507@deepcore.dk> <475C6C3E.6070004@deepcore.dk> In-Reply-To: <475C6C3E.6070004@deepcore.dk> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit Cc: Subject: Re: Any successful installs on a Broadcom HT1000 chipset? X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 10 Dec 2007 04:44:29 -0000 Søren Schmidt wrote: > > OK, so I found at least one problem with the HT1000, its sometimes > doesn't handle 128 sector transfers correctly, its DMA engine looses its > wits. > The result is that it just messes up completely and barfs data all over > the place on the next read operation. > > Anyhow the attached patch seems to work around the problem by limitting > transfer to a max of 126 sectors in one go, at least I have move several > hundred Gb's around now on a zfs volume without problems, which > certainly wasn't possible before. > > Let me know how this turns out for you! > > Patch is against a clean RELENG_7 as last time, should fit both RELENG_6 > and -current... You definitely found the trouble area. This patch allows me to boot and use my Tyan on-board SATA controller without any errors or memory corruption. As an aside, this patch breaks dumping. The disk device (ad6) complains about an oversized DMA and fails to dump on panic now. (No, your patch hasn't created any new panics or anything, btw...) So this is a successful workaround for the HT1000 on-board SATA, and it looks like this additional workaround you just posted is pretty universal and would also limit a Marvell PCI-X SATA controller's DMA size. Do you suppose there will be some way to make this workaround more "global" so that in the presence of an HT1000 chipset, everything (not just ata) will avoid trying large DMA transfers and thus triggering the bug? I don't suppose you think ServerWorks/Broadcom would care to help comment on what might be a more proper global fix-all for this issue... I think it could really save some people with this hardware some trouble if an HT1000 workaround made it into 7.0-RELEASE. Without your fix, if someone afflicted by this HT1000 problem booted up 7.0 it would start corrupting their existing data and there could be some major crying at some point down the road :) Thanks for your work, good job finding it! -- TerraNovaNet Internet Services - Key Largo, FL Voice: (305)453-4011 x101 Fax: (305)451-5991 http://www.terranova.net/ ---------------------------------------------- Life's not fair, but the root password helps.