On Sun, Nov 10, 2002 at 01:53:13PM +0900, Toru Nishimura wrote:

 > If the above change broke the MIPS processor whose multi-level caches have
 > different sized cache line, pmap.c should be corrected, not here.
 > I backtrack to pmap_enter/pmap_remove until technically correct pmap.c
 > replaces the current one.

vmapbuf()/vunmapbuf() is machine-dependent for a reason; different
platforms might choose different implementation strategies.  Putting
the logic to handle this in pmap.c would unnecessarily complicate the
pmap routines.

Basically, there is absolutely nothing wrong with using the normal
pmap_enter()/pmap_remove() routines here, and doing so allows the
pmap_k*() routines to be a lot more efficient.

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