From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeremy Fitzhardinge Subject: Re: [RFC Patch] Support for making an E820 PCI hole in toolstack (xl + xm) Date: Tue, 16 Nov 2010 10:01:33 -0800 Message-ID: <4CE2C6FD.8060504@goop.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: 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 Cc: "xen-devel@lists.xensource.com" , Stefano Stabellini , Konrad Rzeszutek Wilk , "bruce.edge@gmail.com" , Ian Campbell , Gianni Tedesco List-Id: xen-devel@lists.xenproject.org On 11/16/2010 02:11 AM, Keir Fraser wrote: > On 16/11/2010 10:02, "Ian Campbell" wrote: > >>> The right answer is probably to refuse save/restore/migrate when devices are >>> passed through. >> Absolutely. >> >> However we are talking about setting up a 1-1 mapping in the P2M region >> corresponding to the PCI hole at guest boot and preserving that until >> such a time as a device is plugged in, which may be after a migration. >> >> I don't think it matters that no device is passed through at the time of >> the migration, in this configuration we still need arrange for the >> relevant P2M entries to be correct after the migration (or at least >> before the device gets plugged in, perhaps we can leave holes and only >> establish the 1-1 p2m on demand in pcifront?). > Leave the hole empty and populate on demand when devices are passed through > would seem sensible. Actually I was originally thinking that the hole would all be INVALID_MFN but then pfn_to_mfn() would translate that to being an identity translation. But that's pretty hacky, and only works if you actually want identity. On-demand population of regions is much cleaner. J