From mboxrd@z Thu Jan 1 00:00:00 1970 From: Benjamin Herrenschmidt Subject: Re: Promise IDE controller crashes 2.4.22 Date: Mon, 01 Dec 2003 10:31:15 +1100 Sender: linux-ide-owner@vger.kernel.org Message-ID: <1070235074.658.58.camel@gaston> References: <20031129.140802.28395.1074931@webmail05.lax.untd.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from pentafluge.infradead.org ([213.86.99.235]:10709 "EHLO pentafluge.infradead.org") by vger.kernel.org with ESMTP id S262119AbTK3Xbz (ORCPT ); Sun, 30 Nov 2003 18:31:55 -0500 In-Reply-To: <20031129.140802.28395.1074931@webmail05.lax.untd.com> List-Id: linux-ide@vger.kernel.org To: beolach@juno.com Cc: jgoerzen@complete.org, linux-ide@vger.kernel.org On Sun, 2003-11-30 at 09:07, beolach@juno.com wrote: > I recently had a very similar problem to this on my computer with a > Promise Ultra133TX2 (PDC20269) addon card. In my case I had a Western > Digital drive attached to the PDC20269 that would have major filesystem > corruption when DMA was enabled during the boot process. Interestingly, > if I disabled DMA during boot, and then enabled it after boot with > hdparm, the drive worked OK. Also, if I moved the drive to my > mainboard's IDE controller, it would work. It seemed to be a > combination of my drive & IDE controller, because the IDE controller > works fine with other drives, and the WD hard drive worked fine with > another IDE controller. I replaced the drive (20 GB) with a 120 GB > Maxtor drive, and haven't had any problem since. Did you have something on the "other" channel ? Andre seems to imply that the Promise chip may trigger interesting bugs if you do things like SET_FEATURE on one channel while another is doing DMA, kind of things that may well happen during boot Ben.