A while ago, I posted a question about panics on a 64Mb i486,
related to pmap stuff in the kernel. Charles Hannum adviced
me to increas NKPDE in pmap.h from 7 -> 14, and indeed this worked.

Last week, I installed a 32Mb machine. It gave some problems with
a shellscript that went nuts, and kept forking. It should have
resulted in 'proc table full', this failing the fork(), but
in stead it crashed. But, sure enough, increasing the NKPDE
once again seemed to work.

Is there anyone else out there who has had this problem?

Related problem: the 32Mb machine mentioned above, is meant to
run a MUD (yes, one of those damn multi-user adventures). It
crashed sometimes, it was hard to find out why. One of the things
that we found, was a panic under heavy network usage (26 connections
to localhost, constantly busy): 'mb_map too small'. Indeed, increasing
VM_MBUF_SIZE helped here. When you switch on GATEWAY, it is already
set bigger. Is running out of mbufs a known problem with lots of
TCP/IP connections (I vaguely remember some discussions about it) ?
Is there a dynamoc way to fix it, or are you stuck with setting
the # of mbufs to a large number.

Thanks for any info,

- Frank

------------------------------------------------------------------------------