Re: [Xen-devel] [PATCH 08/18] PVH xen: tools changes to create PVH domain

[George Dunlap <george.dunlap@xxxxxxxxxxxxx>]

On 30/08/13 10:53, Ian Campbell wrote:
> On Thu, 2013-08-29 at 18:24 -0700, Mukesh Rathor wrote:
> > On Thu, 29 Aug 2013 12:29:44 +0100
> > Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:
> >
> > > On Thu, 2013-08-29 at 12:13 +0100, George Dunlap wrote:
> > > > On Thu, Aug 1, 2013 at 3:02 AM, Mukesh Rathor
> > > > <mukesh.rathor@xxxxxxxxxx> wrote:
> > > > > > I'm not sure how you are currently signalling to the hypervisor
> > > > > > that a new domain is a PVH domain? I had a look through this
> > > > > > patch and must be being thick because I don't see it.
> > > > > I had a flag set, but it was recommended during RFC to remove it.
> > > > > So, now in xen, a PV with HAP is a PVH guest:
> > > > Why was it recommended to remove it?
> > > >
> > > > "PVH == PV + HAP" is a ridiculous interface, and one which will make
> > > > it hard to import shadow in the future.  In my series I'm planning
> > > > on adding XEN_DOMCTL_CDF_pvh_guest, and using that instead.
> > > These are not stable ABI interfaces, so if someone wants to do PVH
> > > with Shadow then they can just change it.
> > I thought we named PVH for PV with HAP :)
> I thought it was H for HVM myself ;-)
We always talked about an "HVM container", in part to gain back the 
extra protection levels lost when they took away the segmentation limits 
for x86-64.
From a Linux maintenance perspective, autotranslate is of course a big 
win; but there's no reason in principle that we couldn't have used 
shadow pagetables for that.  HAP is a big win in some cases, but a loss 
in others; it is not, as far as I'm concerned, the primary reason for 
introducing this mode.
 -George

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel