Re: [hybrid]: code review for function mapping pfn to foreign mfn

[Mukesh Rathor <mukesh.rathor@oracle.com>]

On Thu, 26 Apr 2012 20:57:12 +0100
Tim Deegan <tim@xen.org> wrote:

> At 11:18 -0700 on 26 Apr (1335439128), Mukesh Rathor wrote:
> > On Thu, 26 Apr 2012 10:08:47 +0100
> 
> Oh, I see.  This is the domU's L4.  In that case I suspect that what's
> missing is the translation from GFNs to MFNs in the domU. 
> 
> To do that, we may finally have to reintroduce the dreaded 'tell me
> what MFN backs this domU GFN' call, so the dom0 tools can build proper
> pagetables and p2m for the pv domU.  Or have you already taken care of
> that?


The tools get an array of the PV domU mfn's. Next to build pagetables, it
needs to map them. It does via privcmd. I generate needed pfn space (not
virtual address space) via ballooning. Next I need to map each pfn to 
the domU mfns. This is a pv guest being built. So, these are mfns' that
don't need looking up. This mapping is temporary, while the guest is being 
built. For building a hybrid domU, it doesn't need to pin the L4 as the 
hybrid guest doesn't use PV MMU. So, these mfn's that are mapped just 
need temporary page count. 

thanks,
Mukesh