Re: [Xen-devel] [PATCH 3 of 7] Use a reserved pfn in the guest address space to store mem event ringsg

[Tim Deegan <tim@xxxxxxx>]

At 09:44 -0800 on 01 Mar (1330595065), Andres Lagar-Cavilla wrote:
> > At 08:30 -0800 on 01 Mar (1330590637), Andres Lagar-Cavilla wrote:
> >> > At 07:47 +0000 on 01 Mar (1330588073), Ian Campbell wrote:
> >> >> > > Is there no scope for making these pages owned by the domain but
> >> not
> >> >> > > actually part of the P2M? We can cope with that for other types
> >> of
> >> >> magic
> >> >> > > page, can't we?
> >> >
> >> > It would need a new operation to map the ring into the tool that uses
> >> > it; normal map-foreign-page ops need a GFN.
> >>
> >> Actually, confirmed: we can call xc_domain_decrease_reservation on the
> >> ring after it's mapped by the helper. Guest won't get at it. Nothing
> >> breaks.
> >
> > But that would only work if:
> >  - the helper always attaches before the guest gets to run; and
> 
> The helper will ignore whatever contents there were on the page. And if
> the guest is out there poking in e820 reserved ranges, then the guest has
> it coming. Note that we've narrowed the window of "vulnerability".
> 
> >  - you never need to restart the helper.
> 
> The helper can re-populate the pfn every time it starts. That's
> contemplated in the current patch series. In fact it can do it
> "atomically" by pausing the guest.

OK, that seems good enough. 

Tim.

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