Re: [PATCH 06/11] kvm: Change kvm_iommu_map_pages to map largepages

[Joerg Roedel]

On Fri, Jan 29, 2010 at 10:32:33AM +0100, Joerg Roedel wrote:
> On Thu, Jan 28, 2010 at 08:24:55PM -0200, Marcelo Tosatti wrote:
> > On Thu, Jan 28, 2010 at 12:37:57PM +0100, Joerg Roedel wrote:
> > > +static pfn_t kvm_pin_pages(struct kvm *kvm, struct kvm_memory_slot *slot,
> > > +			   gfn_t gfn, unsigned long size)
> > > +{
> > > +	gfn_t end_gfn;
> > > +	pfn_t pfn;
> > > +
> > > +	pfn     = gfn_to_pfn_memslot(kvm, slot, gfn);
> > 
> > If gfn_to_pfn_memslot returns pfn of bad_page, you might create a
> > large iommu translation for it?
> 
> Right. But that was broken even before this patch. Anyway, I will fix
> it.
> 
> > > +		/* Map into IO address space */
> > > +		r = iommu_map(domain, gfn_to_gpa(gfn), pfn_to_hpa(pfn),
> > > +			      get_order(page_size), flags);
> > > +
> > > +		gfn += page_size >> PAGE_SHIFT;
> > 
> > Should increase gfn after checking for failure, otherwise wrong
> > npages is passed to kvm_iommu_put_pages.
> 
> True. Will fix that too.

Here is the updated patch (also updated in the iommu/largepage branch of
my tree). Does it look ok?