On Mon, May 29, 2000 at 09:15:30PM +0300, Jukka Marin wrote:

 > kernel: page fault trap, code=0
 > Stopped at pmap_zero_page_uncached+0x48: repe stosl %es:(%edi)
 > db> t
 > pmap_zero_page_uncached(1388000,0,7fffffff,c34e87c8,c34e87c8) at
 > pmap_zero_page_uncached+0x48
 > uvm_pageidlezero(e000ffe7,c34e87c8,0,7fffffff,c019eec0) at uvm_pageidlezero+0x8c

Sigh!  You are not the first person to report wackiness like this.

This is somewhat puzzling to me.  I have not seen this problem on
ANY system I have, and they all are used quite heavily (NFS included).

I suppose it's possible that the Cyrix doesn't work properly with PG_N,
but that would cause e.g. all PCI memory space access to fail, too (since
it's mapped PG_N, also).

If you look at pmap_zero_page_uncached(), it's basically a mirror of
pmap_zero_page(), with the exception of setting PG_N (not-cached) on
non-i386 CPUs.

Can you please reproduce it again, and e-mail me the following:

db> t
db> print $ecx
db> print $edi

Thanks.

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