Re: [PATCH 5/5] x86: EFI memory mapping changes according to changes to ioremap and c_p_a

["huang ying" <huang.ying.caritas@gmail.com>]

Jan 31, 2008 8:18 PM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Huang, Ying <ying.huang@intel.com> wrote:
>
> > +                     /* Assume pages are mapped as WB */
> > +                     if (mode == IOR_MODE_UNCACHED)
> > +                             set_memory_uc(md->virt_addr, md->num_pages);
> > +                     /* Assume pages are mapped as unexecutable */
> > +                     if ((__supported_pte_mask & _PAGE_NX) &&
> > +                         xmode == IOR_XMODE_EXEC)
> > +                             set_memory_x(md->virt_addr, md->num_pages);
>
> hm, why is the check for __supported_pte_mask needed? That is already
> listened to in set_memory_x(). [if not then it's an ioremap bug that we
> want to fix.]

OK, I will check it.

> > -static int ioremap_change_attr(unsigned long paddr, unsigned long size,
> > -                            enum ioremap_mode mode,
> > -                            enum ioremap_xmode xmode)
> > +int ioremap_change_attr(unsigned long paddr, unsigned long size,
> > +                     enum ioremap_mode mode,
> > +                     enum ioremap_xmode xmode)
>
> hm, there should be no need to use this raw API.

Oh, you are right. __efi_ioremap() is only used when there is no
identity map. So ioremap_change_attr() need not to be called in
__efi_ioremap().

Best Regards,
Huang Ying