the only drawback to this would be that we couldn't use these functions
on unmanaged memory anymore (at least, not without creating fake page
structures to point to the unmanaged memory).  there could conceivably
be some use for that, though there's nothing in the tree that does now.
if we made this change and then wanted to get that ability back, we could
always make new interfaces for unmanaged versions of these functions.

so I guess I'd say go ahead.

-Chuck


On Tue, Apr 09, 2002 at 07:10:53PM -0700, Jason R Thorpe wrote:
> pmap_copy_page() and pmap_zero_page() currently take paddr_t
> arguments.
> 
> Some platforms (MIPS, ARM, for example) have to make handle virtual
> aliases for the source page of pmap_copy_page(), and thus have to
> look up the vm_page for that page.
> 
> I'd like to change the arguments to these functions to take vm_page *'s
> instead of paddr_t's, to save work on those platforms.  It should have
> no impact on platforms that don't need to handle aliases, since
> VM_PAGE_TO_PHYS() is already used at the call sites.
> 
> Comments?
> 
> -- 
>         -- Jason R. Thorpe <thorpej@wasabisystems.com>