Hi,

> this happens because the machine runs out of memory and the ppc pmap
> doesn't know how to reclaim another active pvo entry.
> try the attached patch.

>> panic: pmap_pvo_enter: failed

With the patch you made, I can run a stress test which essentially does a 
"make fetch" in every subdirectory of a pkgsrc tree and the system doesn't 
crash. With a kernel compiled with maxusers set to 512, 2 gigs of memory, 
and 2 gigs of swap, the system seems to be able to handle large amounts of 
memory use much better than before.

load averages: 932.34, 807.40, 254.447                             17:47:00
7172 processes:1498 runnable, 5665 sleeping, 8 zombie, 1 on processor
CPU states:  0.4% user, 32.7% nice, 60.3% system,  6.7% interrupt,  0.0% idle
Memory: 1159M Act, 566M Inact, 44M Wired, 8920K Exec, 388M File, 7940K Free
Swap: 2048M Total, 1617M Used, 431M Free

I see that you've committed the patch to current; have you asked releng to 
pull this up into 2.0?

Thanks!
John Klos