> If I remember correctly, this should fix it:
> 	sys/arch/i386/i386/pmap.c:
> 		Comment out all calls to pmap_zero_page()

Have you verified that the calls to pmap_zero_page you commented out
were being executed at all on your system?

Have you tried booting a kernel with this change on a system with 2G
of memory?  Or on a system with lots of paging, so that
pmap_steal_ptp() is invoked, and plugs an unzeroed page-table-page
into the page directory?

This change sounds extremely dubious; neither of the calls to
pmap_zero_page() in pmap.c has anything to do with uncached idle-page
zeroing which appears to be what's problematic on some Cyrix
processors.

> 	sys/arch/i386/include/pmap.h:
> 		Comment out the definition for PMAP_PAGEIDLEZERO()

This is not a "fix".  this disables a significant performance
optimization new in 1.5.

> All I did was disable the in-kernel debugger, so I'm not seeing the
> panics. I don't know if they're still happening or not; that's why
> solution 1 is probably the best route.

If the kernel debugger is not enabled, "panic" will crash the system.
I'd expect that you'd notice that.

					- Bill