Re: [PATCH] [7/8] Implement true end_pfn_mapped for 32bit

[Andi Kleen <ak@suse.de>]

On Tuesday 12 February 2008 20:39:48 Thomas Gleixner wrote:
> On Mon, 11 Feb 2008, Andi Kleen wrote:
> > Even on 32bit 2MB pages can map more memory than is in the true
> > max_low_pfn if end_pfn is not highmem and not aligned to 2MB. 
> > Add a end_pfn_map similar to x86-64 that accounts for this 
> > fact. This is important for code that really needs to know about
> > all mapping aliases. Needed for followup patches (in this case EFI)
> 
> It's not only necessary for followup patches, it is a question of
> general correctness.

True. ioremap already requires it.

> > @@ -36,7 +36,7 @@
> >  #define max_pfn_mapped		end_pfn_map
> >  #else
> >  #include <asm/page_32.h>
> > -#define max_pfn_mapped		max_low_pfn
> > +#define max_pfn_mapped		end_pfn_map
> 
> We can nuke either max_pfn_mapped or end_pfn_map completely. I don't
> care about which one, but keeping both makes no sense at all.

I didn't want to bundle such a clean up into the bug fix
because my experience is that you usually reject that categorically.

I can send the removal of max_pfn_mapped as a follow up patch
if you apply this one.

-Andi