Bill> > Oh, I think I found the problem with patch (2). I'll send out a 2' in a
Bill> > seperate message.

Bill> Ok, I found one problem with the call to cpu_probe_cache(). While it was
Bill> setting the cache line size right (to 32), it was NOT setting the page
Bill> size right; it was setting it to uvmexp.pagesize, which was not
Bill> necessarily initialized at that point, so the old patch (2) wasn't
Bill> necessarily turning the caching code on.

Bill> Please try this patch 2a.

        +-------- Temporarily set up cache_info
        |
        |     +----syncicache-half-revert to 20020313
        |     |
        |     |     +-- add printf-cacheinfo
        |     |     |
        |     |     |     +-- no cninit
        |     |     |     |
 (1)   (2a)  (3)   (4)   (5)
       YES   YES              works
       YES                    works

20020314-UTC cvs src + (2a) alone makes kbd/screen to work.
---
Makoto Fujiwara,