On Thu, Sep 21, 2000 at 11:58:28AM -0700, Chris G. Demetriou wrote:

 > this is kind-of a bummer.  on the mips, for instance, it means that
 > you can't just #define PMAP_PAGEIDLEZERO to use pmap_zero_page(),
 > which is good for some systems on which uncached accesses are Bad but
 > on which is really is pretty much OK to zero pages anyway.

On such a MIPS machine, you can just not define PMAP_PAGEIDLEZERO; take a
look at the code in uvm_page.c.

 > I suppose you can do a more complicated #define, but it's still a bit
 > of a PITA, and kinda makes me think that this was a
 > less-than-completely-thought out change, and could have used
 > additional (any?) discussion...

Well, it was made after a discussion w/ chuq silvers who idenfied the
need to bail out early of slow page zeroing.

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