Re: [PATCH 04/10] xen/mmu: For 1-1 mapping, automatically set _PAGE_IOMAP.

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Dec 21, 2010 at 02:29:31PM -0800, Jeremy Fitzhardinge wrote:
> On 12/21/2010 01:37 PM, Konrad Rzeszutek Wilk wrote:
> > This means that for PFNs (specifically: those in any E820 gaps
> > or non-RAM E820 regions) that have 1-1 mapping we set the
> > _PAGE_IOMAP flag.
> >
> > Later on we could remove the _PAGE_IOMAP code handling, but
> > for right now lets keep this in to not introduce any bisection
> > failures across this patchset.
> >
> > Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> > ---
> >  arch/x86/xen/mmu.c |    3 +++
> >  1 files changed, 3 insertions(+), 0 deletions(-)
> >
> > diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
> > index 4ba7e4e..bd02e7d 100644
> > --- a/arch/x86/xen/mmu.c
> > +++ b/arch/x86/xen/mmu.c
> > @@ -832,6 +832,9 @@ static pteval_t pte_pfn_to_mfn(pteval_t val)
> >  		pteval_t flags = val & PTE_FLAGS_MASK;
> >  		unsigned long mfn = pfn_to_mfn(pfn);
> >  
> > +		if (mfn == pfn)
> > +			flags |= _PAGE_IOMAP;
> 
> Why?  Does it really make sense to set _PAGE_IOMAP if they just happen
> to be the same value?

Yes. Without this, user applications, such as 'dmidecode' cannot get
data.

But I think with ditching a bunch of the _PAGE_IOMAP in the xen/mmu.c we can
remove this.

I would rather keep this patch as temporary scaffolding and when the
other set of patches is ready for the _PAGE_IOMAP, ditch this one.

> 
>     J