From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gerd Hoffmann Subject: Re: pv_ops & gntdev? Date: Wed, 25 Feb 2009 19:37:51 +0100 Message-ID: <49A58FFF.3050604@redhat.com> References: <49A44030.2070709@redhat.com> <49A4640E.1000807@goop.org> <49A470DD.2000008@redhat.com> <49A517F6.30005@redhat.com> <49A58506.2020407@goop.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <49A58506.2020407@goop.org> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Jeremy Fitzhardinge Cc: Xen Development Mailing List List-Id: xen-devel@lists.xenproject.org Jeremy Fitzhardinge wrote: > Gerd Hoffmann wrote: >> Gerd Hoffmann wrote: >> >>>> Do you want to bring it over? >>>> >>> Well. I hoped someone did that already ... >>> >> >> Ideas how to handle the grant unmap case? >> I suspect the vm_ops->zap_pte() approach isn't going to fly for mainline >> merge ... > > Yes, my plan is: > > 1. when installing a grant in a pte page, allocate a shadow page and > hang it off page->private; also set a page flag on the pgd page > flags (steal something spare) pgd or pte page flags? or both? > 2. store the grant handle in the slot corresponding to the pte ok. > 3. when clearing the pte, check to see if there's a grant handle and > zap it That needs a new paravirt hook I think ... > 4. when unpinning the pagetable, check to see if there's the "grant > mapping" flag in the pgd, and do a grant-zap pass before unpinning ok. Prelimary patches by chance? cheers, Gerd