From owner-freebsd-hardware@FreeBSD.ORG Thu Dec 13 02:52:53 2007 Return-Path: Delivered-To: freebsd-hardware@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 9D7D716A49A; Thu, 13 Dec 2007 02:52:53 +0000 (UTC) (envelope-from jhb@freebsd.org) Received: from speedfactory.net (mail6.speedfactory.net [66.23.216.219]) by mx1.freebsd.org (Postfix) with ESMTP id 46ADF13C442; Thu, 13 Dec 2007 02:52:52 +0000 (UTC) (envelope-from jhb@freebsd.org) Received: from server.baldwin.cx (unverified [66.23.211.162]) by speedfactory.net (SurgeMail 3.8r) with ESMTP id 224200788-1834499 for multiple; Wed, 12 Dec 2007 21:52:59 -0500 Received: from localhost.corp.yahoo.com (john@localhost [127.0.0.1]) (authenticated bits=0) by server.baldwin.cx (8.13.8/8.13.8) with ESMTP id lBD2qnlx085913; Wed, 12 Dec 2007 21:52:49 -0500 (EST) (envelope-from jhb@freebsd.org) From: John Baldwin To: cokane@freebsd.org Date: Wed, 12 Dec 2007 19:29:46 -0500 User-Agent: KMail/1.9.6 References: <4759730F.6090302@FreeBSD.org> <200712121311.06777.jhb@freebsd.org> <476036E9.9010002@FreeBSD.org> In-Reply-To: <476036E9.9010002@FreeBSD.org> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-15" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200712121929.47314.jhb@freebsd.org> X-Greylist: Sender succeeded SMTP AUTH authentication, not delayed by milter-greylist-2.0.2 (server.baldwin.cx [127.0.0.1]); Wed, 12 Dec 2007 21:52:49 -0500 (EST) X-Virus-Scanned: ClamAV 0.91.2/5110/Wed Dec 12 15:42:31 2007 on server.baldwin.cx X-Virus-Status: Clean X-Spam-Status: No, score=-4.4 required=4.2 tests=ALL_TRUSTED,AWL,BAYES_00 autolearn=ham version=3.1.3 X-Spam-Checker-Version: SpamAssassin 3.1.3 (2006-06-01) on server.baldwin.cx Cc: freebsd-hardware@freebsd.org Subject: Re: Overlapping PCI Memory Locations X-BeenThere: freebsd-hardware@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: General discussion of FreeBSD hardware List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 13 Dec 2007 02:52:53 -0000 On Wednesday 12 December 2007 02:30:49 pm Coleman Kane wrote: > John Baldwin wrote: > > On Friday 07 December 2007 11:21:35 am Coleman Kane wrote: > > > >> Hi, > >> > >> I've got a problem where two components on my system have overlapping > >> PCI memory regions: > >> > >> atapci0: port > >> > > 0x9000-0x9007,0x9008-0x900b,0x9010-0x9017,0x5018-0x501b,0x5020-0x502f mem > > 0xd0609000-0xd06093ff irq 16 at device 18.0 on pci0 > > > >> atapci0: [ITHREAD] > >> > >> pcm0: mem 0xd0608000-0xd060bfff > >> > > irq 16 at device 20.2 on pci0 > > > >> pcm0: hdac_mem_alloc: Unable to allocate memory resource > >> > >> > >> > >> Because of this, I cannot use the sound hardware on this system. In > >> addition, the memory range used by atapci0 is the SATA AHCI space. The > >> ata-chipset.c doesn't currently identify the ATI IXP600 SATA controller > >> (just the paired PATA controller), so I can actually use my drives > >> through the PATA/IDE compatibility registers in the I/O space. However, > >> if I modify ata-chipset.c to add support for the IXP600 SATA controller, > >> I get weird results using ATA_INL(..) calls, which look like something > >> is interfering with the data I *should* be getting from the SATA mem space. > >> > >> In addition, the pcm0 refuses to attach, as above. > >> > >> Also, this is a notebook and has one of those crummy notebook BIOSes > >> that don't allow fiddling with this sort of stuff in BIOS. Is there any > >> facility in the kernel to force these to be remapped (or to perform the > >> mappings ourselves and ignore what BIOS tells us)? > >> > > > > No. You can hack the pci driver to zero out the BAR in either device during > > boot though as a test. > > > What would the kernel do if the BAR were to be zeroed out? PCI's bus_alloc_resource() method will try to assign free resources to the BAR. > I applied the attached fixup to sys/dev/pci/pci.c that allows me to > specify the desired base address from a loader.conf tunable. Is there a > better way? Is there a project to make a better way? There is currently not a better way. Eventually there will be a way to do this as part of improving PCI's resource management in general, but I wouldn't count on it any time soon. -- John Baldwin