From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kiszka Subject: Re: [SeaBIOS] Graphics card pass-through working with two pass pci-initialization Date: Tue, 07 Jun 2011 10:05:59 +0200 Message-ID: <4DEDDBE7.6090708@siemens.com> References: <4DDCF9ED.5050208@redhat.com> <4DDDDA8D.6080807@web.de> <4DDDF62C.2010603@redhat.com> <4DDF3768.5010002@web.de> <4DDF867A.9040700@siemens.com> <4DDFFE1E.1060900@web.de> <4DE3486F.10502@redhat.com> <20110531022922.GB10673@valinux.co.jp> <4DE5EA84.7040400@redhat.com> <20110601142029.GC3554@valinux.co.jp> <20110601223405.GB18369@morn.localdomain> <4DEC7412.80506@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: "Kevin O'Connor" , Isaku Yamahata , Alex Williamson , "seabios@seabios.org" , "kvm@vger.kernel.org" To: Gerd Hoffmann Return-path: Received: from goliath.siemens.de ([192.35.17.28]:15849 "EHLO goliath.siemens.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752045Ab1FGIGV (ORCPT ); Tue, 7 Jun 2011 04:06:21 -0400 In-Reply-To: <4DEC7412.80506@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On 2011-06-06 08:30, Gerd Hoffmann wrote: > Hi, > >> As Jan points out though, is a dynamic PCI region really needed? >> Those that need a large PCI region are also likely to need a large >> amount of memory. Maybe the space for PCI should just be increased. > > Just changing it will not work as it will break live migration. Changing logic in the BIOS won't break migration (the active BIOS is included in the migration of RAM, current mappings are part of the device states). Changing the 4G mapping in qemu's hw/pc_piix.c would break it and needs to be coupled to the machine version. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux