From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alex Williamson Subject: Re: [PATCH 4/4] kvm: i386: Add classic PCI device assignment Date: Mon, 03 Sep 2012 21:31:58 -0600 Message-ID: <1346729518.2225.37.camel@ul30vt.home> References: <825e653c9cfe9d8e26185917cbe1f1dd7ae299e2.1346048917.git.jan.kiszka@web.de> <503B62F4.9070500@suse.de> <503B64D9.4000708@siemens.com> <503DD705.8020406@siemens.com> <5044D3DD.2020904@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Peter Maydell , "kvm@vger.kernel.org" , "Michael S. Tsirkin" , Alexey Kardashevskiy , Jan Kiszka , Marcelo Tosatti , "qemu-devel@nongnu.org" , qemu-ppc , Andreas =?ISO-8859-1?Q?F=E4rber?= To: Avi Kivity Return-path: In-Reply-To: <5044D3DD.2020904@redhat.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+gceq-qemu-devel=gmane.org@nongnu.org Sender: qemu-devel-bounces+gceq-qemu-devel=gmane.org@nongnu.org List-Id: kvm.vger.kernel.org On Mon, 2012-09-03 at 18:59 +0300, Avi Kivity wrote: > On 08/29/2012 11:49 AM, Peter Maydell wrote: > > On 29 August 2012 09:47, Jan Kiszka wrote: > >> On 2012-08-28 23:26, Peter Maydell wrote: > >>> Since this is arch-specific we should probably give the > >>> resulting device a more specific name than "pci-assign", > >>> which implies that it is (a) ok for any PCI system and > >>> (b) not something that will be obsolete in the foreseen > >>> future. > >> > >> The name has to be like this for libvirt compatibility. I can provide it > >> as alias, though, calling the device "kvm-pci-assign" by default. Better? > > > > ...is it likely to ever work for non-x86 PCI devices? > > It used to work for ia64 before it died. So it's not architecture > specific by design. Caveat; ia64 is a different architecture, but it was on Intel VT-d chipsets, which therefore supported the IOMMU API and ia64 maps guests the same way. This is the same problem I ran into trying to name the vfio iommu driver for x86. It's not really x86-specific, but it's got to look and smell pretty similar. Thanks, Alex From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:44200) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1T8jry-0001HZ-KD for qemu-devel@nongnu.org; Mon, 03 Sep 2012 23:32:07 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1T8jrx-0002Yu-Fk for qemu-devel@nongnu.org; Mon, 03 Sep 2012 23:32:06 -0400 Message-ID: <1346729518.2225.37.camel@ul30vt.home> From: Alex Williamson Date: Mon, 03 Sep 2012 21:31:58 -0600 In-Reply-To: <5044D3DD.2020904@redhat.com> References: <825e653c9cfe9d8e26185917cbe1f1dd7ae299e2.1346048917.git.jan.kiszka@web.de> <503B62F4.9070500@suse.de> <503B64D9.4000708@siemens.com> <503DD705.8020406@siemens.com> <5044D3DD.2020904@redhat.com> Content-Type: text/plain; charset="UTF-8" Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH 4/4] kvm: i386: Add classic PCI device assignment List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Avi Kivity Cc: Peter Maydell , "kvm@vger.kernel.org" , "Michael S. Tsirkin" , Alexey Kardashevskiy , Jan Kiszka , Marcelo Tosatti , "qemu-devel@nongnu.org" , qemu-ppc , Andreas =?ISO-8859-1?Q?F=E4rber?= On Mon, 2012-09-03 at 18:59 +0300, Avi Kivity wrote: > On 08/29/2012 11:49 AM, Peter Maydell wrote: > > On 29 August 2012 09:47, Jan Kiszka wrote: > >> On 2012-08-28 23:26, Peter Maydell wrote: > >>> Since this is arch-specific we should probably give the > >>> resulting device a more specific name than "pci-assign", > >>> which implies that it is (a) ok for any PCI system and > >>> (b) not something that will be obsolete in the foreseen > >>> future. > >> > >> The name has to be like this for libvirt compatibility. I can provide it > >> as alias, though, calling the device "kvm-pci-assign" by default. Better? > > > > ...is it likely to ever work for non-x86 PCI devices? > > It used to work for ia64 before it died. So it's not architecture > specific by design. Caveat; ia64 is a different architecture, but it was on Intel VT-d chipsets, which therefore supported the IOMMU API and ia64 maps guests the same way. This is the same problem I ran into trying to name the vfio iommu driver for x86. It's not really x86-specific, but it's got to look and smell pretty similar. Thanks, Alex