Noticed that when BUFCACHE kernel option is set `too high', the kernel
fails to boot and panics immediately after printing the amount ot total memory
with a message like 'cannot allocate memory for kernel buffers'. 'High' is
somewhere between 100MB and 200MB (128MB?), 20% of 1GB does the trick at
least, while 20% of 0.5GB is fine. Why is that? Is there some hardware
opposed limit involved here? This is on 1.6 (but I noticed on earlier
-current too once). 

It is not a big problem, but it was somewhat annoying to discover that
a quick upgrade from 0.5 to 1 GB of our server (anoncvs.no.netbsd.org)
took a bit longer than expected (sorry to those that were affected by
this unannounced 30 min. downtime) as the kernel failed to boot. I
wonder whether a simple fix might cause a bit more elegant behaviour
like maxing out at the limit for BUFCACHE but continue booting.

Feico.