Re: [PATCH 17/17] PVH xen: PVH dom0 creation....

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

On Fri, 10 May 2013 08:14:55 +0100
"Jan Beulich" <JBeulich@suse.com> wrote:

> >>> On 10.05.13 at 03:53, Mukesh Rathor <mukesh.rathor@oracle.com>
> >>> wrote:
> > On Fri, 26 Apr 2013 08:22:08 +0100
> >> >> > +    /* If the e820 ended under 4GB, we must map the remaining
> >> >> > space upto 4GB */
> >> >> > +    if ( end < GB(4) )
> >> >> > +    {
> >> >> > +        start_pfn = PFN_UP(end);
> >> >> > +        end_pfn = (GB(4)) >> PAGE_SHIFT;
> >> >> > +        nump = end_pfn - start_pfn;
> >> >> > +        rc = domctl_memory_mapping(d, start_pfn, start_pfn,
> >> >> > nump, 1);
> >> >> > +        BUG_ON(rc);
> >> >> > +    }
> >> >> 
> >> >> That's necessary, but not sufficient. Or did I overlook MMIO
> >> >> ranges getting added somewhere else for Dom0, when they sit
> >> >> above the highest E820 covered address?
> >> > 
> >> > construct_dom0() adds the entire range:
> >> > 
> >> >     /* DOM0 is permitted full I/O capabilities. */
> >> >     rc |= ioports_permit_access(dom0, 0, 0xFFFF);
> >> >     rc |= iomem_permit_access(dom0, 0UL, ~0UL);
> >> 
> >> Which does not create any mappings at all - these are just
> >> permissions being granted.
> > 
> > Right. I'm not sure where its happening for dom0.
> 
> So if you don't know where you do this, I have to guess you don't
> do this at all. But you obviously need to. Your main problem is that
> you likely don't want to waste memory on page tables to cover the
> whole (up to 52 bit wide) address space, so I assume you will need
> to add these tables on demand. Yet then again iirc IOMMU faults

Hmm... well, I originally had it where the tables were updated "on
demand" initiated by guest, but then suggestions were to make that
transparent to the guest. I don't really know what the best solution
is, let me investigate/think some more.

thanks,
Mukesh