> 
> 
> >This just doesn't seem right to me, any of it.  If the user chooses to
> >override certain values like bufpages, nbuf, nkpde, or SHMMAX, we ought to
> >adjust the maximum sizes of the larger structures to accomodate him.  Is there
>
One comment, there is absolutely NO reason to map SYSVSHM into kva space.
We are in the process of a total SYSVSHM rewrite, but for now, I have
made some changes to our code derived from NetBSD to eliminate the
unnecessary mapping.  Those changes aren't very dependent on the FreeBSD
VM code, and can be ported fairly easily.  This could certainly help certain
NetBSD users, if a fix (similar to FreeBSD's) for this problem is adopted.

Note that we have been carefully managing our KVA usage, because large
servers can run into trouble.  As soon as we make a mistake in KVA space
management, we get negative responses fairly quickly.  Of course, I am
running only with 80MB connected by 28.8 modem, so I don't see the problems.
Large internet servers, with large #'s of mbufs, buffers, and processes
use lots of kva space, and are very susceptable to KVA space limitations.

John
dyson@freebsd.org