From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Jan Beulich" Subject: Re: [PATCH 1/2][VTD] pci mmconfig support to be usedfor ATS Date: Fri, 21 Nov 2008 10:32:39 +0000 Message-ID: <49269C57.76E4.0078.0@novell.com> References: <57C9024A16AD2D4C97DC78E552063EA35C4BCCFE@orsmsx505.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: Content-Disposition: inline List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Keir Fraser , Allen M Kay Cc: "xen-devel@lists.xensource.com" , Weidong Han , Dexuan Cui List-Id: xen-devel@lists.xenproject.org >>> Keir Fraser 20.11.08 21:20 >>> >On 20/11/08 20:10, "Kay, Allen M" wrote: > >>> If this is x86-64 only anyway, why do you need to use fixmap >>> space for this (and limit things arbitrarily to 4 segments) >>> rather than using a 256Mb block from one of the unused virtual >>> address regions in hypervisor space? >>=20 >> Can you point me how to use unused virtual address region to do ioremap >> without using fixmap? Is there a existing code that is already doing = this? > >Pick yourself a nice piece of virtual address space (see the layout in >include/asm-x86/config.h -- 0xffff840000000000 would be fine) and then = use >map_pages_to_xen(). While it can be adjusted later anyway, I'd suggest not using a part of = this largest available range - sooner or later we'll want to use this to extend the 1:1 mapping to 4Tb. Instead, I'd pick another 1Gb range out of the 461Gb reserved region. Jan