Re: [Xen-devel] [V6 PATCH 6.2/7] pvh dom0: Add and remove foreign pages

[Tim Deegan <tim@xxxxxxx>]

At 18:34 -0800 on 17 Dec (1387301666), Mukesh Rathor wrote:
> On Tue, 17 Dec 2013 11:10:46 +0100
> Tim Deegan <tim@xxxxxxx> wrote:
> 
> > At 15:27 -0800 on 16 Dec (1387204042), Mukesh Rathor wrote:
> > > On Mon, 16 Dec 2013 08:40:41 +0000
> > > "Jan Beulich" <JBeulich@xxxxxxxx> wrote:
> > > 
> > > > >>> On 14.12.13 at 03:48, Mukesh Rathor <mukesh.rathor@xxxxxxxxxx>
> > > > >>> wrote:
> > > > >> Also, Jan may have an opinion about whether a teardown
> > > > >> operation that has to walk each p2m entry would have to be made
> > > > >> preemptible.  I'm not sure where we draw the line on such
> > > > >> things.
> > > > > 
> > > > > Since at present teardown cleanup of foreign is not really that
> > > > > important as its only applicable to dom0, let me submit another
> > > > > patch for it on Mon with few ideas. That would also keep this
> > > > > patch size reasonable, and keep you from having to look at the
> > > > > same code over and over. 
> > > > > 
> > > > > So, please take a look at the version below with above two
> > > > > fixes. If you approve it, i can resubmit the entire series
> > > > > rebased to latest with your ack on Monday, and the series can
> > > > > go in while we resolve the p2m teardown.
> > > > 
> > > > Going through the patch again, I'm not seeing any loop being
> > > > added. Am I missing something here?
> > > 
> > > Yes. Since the destruction of p2m leaking foreign pages only applies
> > > to control domain being destroyed, i don't think it is that critical
> > > that part get into 4.4.
> > 
> > Nak.  We can't let the tools break the refcounting.
> 
> BTW, it appears that currently this cleanup doesn't happen on PV. If on a 
> PV control domain and the tool stack has mapped foreign pages, and the 
> PV control domain dies, the refcnt will be messed up. Right?

Not at all.  PV domains' foreign mappings are refcounted through their
pagetables, which are unwound on domain destruction -- otherwise even
their local mappings would mess up the refcounts!

Tim.

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