Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S269702AbTGJX3D (ORCPT ); Thu, 10 Jul 2003 19:29:03 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S269703AbTGJX3C (ORCPT ); Thu, 10 Jul 2003 19:29:02 -0400 Received: from mion.elka.pw.edu.pl ([194.29.160.35]:54182 "EHLO mion.elka.pw.edu.pl") by vger.kernel.org with ESMTP id S269702AbTGJX2t (ORCPT ); Thu, 10 Jul 2003 19:28:49 -0400 Date: Fri, 11 Jul 2003 01:43:06 +0200 (MET DST) From: Bartlomiej Zolnierkiewicz To: Steven Dake cc: Samuel Flory , Chad Kitching , , Subject: Re: IDE/Promise 20276 FastTrack RAID Doesn't work in 2.4.21, patchattached to fix In-Reply-To: <3F0DF5B8.9040304@mvista.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2342 Lines: 60 On Thu, 10 Jul 2003, Steven Dake wrote: > Samuel Flory wrote: > > > Steven Dake wrote: > > > >> Even with special fasttrack feature enabled, my disk devices on the > >> PDC20276 is not found. There is code in pci-setup.c which blocks > >> other PDC controllers, why not the 20276? Is that code for some > >> other purpose, or orthagonal to the force option? > > > > > > The comments would seem to indicate that this is only needed if you > > have a second controller. Which leads me to wonder what if I have 3 > > or 4 pdc controllers. > > Hmm thats not how I read the code. My system has a standard IDE device Hmmm... read the code again :-). > as part of the chipset, and then also has a fasttrack controller. The > fasttrack controller comes in 2nd, (hence making it the 2nd controller > and making it marked disabled). I think your right about the 3rd/4th > controller though, what happens to those ! > -steve > > > > > for (port = 0; port <= 1; ++port) { > > ide_pci_enablebit_t *e = &(d->enablebits[port]); > > > > /* > > * If this is a Promise FakeRaid controller, > > * the 2nd controller will be marked as > > * disabled while it is actually there and enabled > > * by the bios for raid purposes. > > * Skip the normal "is it enabled" test for those. > > */ > > if (((d->vendor == PCI_VENDOR_ID_PROMISE) && > > ((d->device == PCI_DEVICE_ID_PROMISE_20262) || > > (d->device == PCI_DEVICE_ID_PROMISE_20265))) && > > (secondpdc++==1) && (port==1)) > > goto controller_ok; I think this test in reality does something different then comment states. For first port of PDC20262/65 this test increases secondpdc variable (so it is 1 after test). For second port this test is true (its PDC20262/65 && secondpdc == 1 && port == 1) so we don't test whether 2nd port (not controller!) of 1st controller is enabled. Or I am reading it wrong? -- Bartlomiej - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/