[PATCH] ARM: fix highmem with VIPT cache and DMA

[Russell King - ARM Linux]

On Fri, Mar 26, 2010 at 11:51:58AM -0400, Nicolas Pitre wrote:
> On Fri, 26 Mar 2010, Catalin Marinas wrote:
> 
> > On Thu, 2010-03-25 at 21:02 +0000, Nicolas Pitre wrote:
> > > --- a/arch/arm/include/asm/highmem.h
> > > +++ b/arch/arm/include/asm/highmem.h
> > > @@ -11,7 +11,11 @@
> > > 
> > >  #define kmap_prot              PAGE_KERNEL
> > > 
> > > -#define flush_cache_kmaps()    flush_cache_all()
> > > +#define flush_cache_kmaps() \
> > > +       do { \
> > > +               if (cache_is_vivt()) \
> > > +                       flush_cache_all(); \
> > > +       } while (0)
> > 
> > Do the aliasing VIPT caches need flushing as well?
> 
> No idea.  Highmem is not supported with aliasing VIPT at the moment 
> anyway -- see commit 3f973e2216.  I don't have hardware with aliasing 
> VIPT cache either.

I don't think we'll ever support aliasing VIPT caches with highmem -
it'd quadruple the amount of kmap space that's required for things like
KM_USER*.