Re: [PATCH] fix ppc ioremap prototype

[Matt Porter <mporter@kernel.crashing.org>]

On Thu, Sep 04, 2003 at 07:30:09AM -0700, David S. Miller wrote:
> On Thu, 4 Sep 2003 07:36:50 -0700
> Matt Porter <mporter@kernel.crashing.org> wrote:
> 
> > Ok, now the other part of making PCI devices work is to support
> > mmap.
> 
> You get the pci device in the arch PCI mmap() routines, what
> more do you need?  Grep for HAVE_PCI_MMAP in the source tree
> and how sparc64 implements that.

That will work for the PCI case.  It does force us to maintain a
remap_page_range64() in arch/ppc, duplicating code, and creating a
maintainer headache.  I don't think we can hack set_pte() (so as
to call the standard remap_page_range()) to trap and add the upper
bits because there is not enough context to make the right decision.
We've got to have the remap_page_range64() for PPC4xx-specific OCP
drivers too.

A remap_page_range() that even took a paddr_t would be helpful if
not some resource based remapper.

I suppose we could even use the HAVE_PCI_MMAP approach to make
mem.c usable with a PPC44x specific hack.  Of course, a
remap_page_range() with a paddr_t arg would also help here. 

-Matt