On Friday, May 30, 2003, at 07:45  PM, Stephan Uphoff wrote:

> pmap_map_ptes(pmap) normally locks a pmap.
> However if the pmap argument is pmap_kernel() it does not lock it.

Hm, indeed.

In any case, I think it's still safe to assume that two different CPUs 
won't be changing mappings in the same range of address space.  
Something has to *own* that address space, and presumably the correct 
synchronization will be happening at that layer, as well.

         -- Jason R. Thorpe <thorpej@wasabisystems.com>