brian wrote:
> Jason Thorpe (@nas.nasa.gov) posted this:
> 
> > NetBSD/{hp300,mac68k,mvme68k} switchs to
> > UVM virtual memory system
> >
> 
> I have sendmail bugs in netbsd-1.3, which the
> faq suggests upgrading to netbsd-1.3.1. Does
> Jason's post  mean we need to download some
> src  files to compile, or just binaries on top of
> netbsd-1.3.1? Should i wait for UVM before
> installing netbsd-1.3.1?

No, UVM is available in -current only.  UVM is simply a radically
different method of managing virtual memory.  It should be more efficient
and faster.  It will _not_ work under 1.3.1 as it requires a _large_
number of changes here and there to make it work.

> Does UVM mean the kernel will  no longer dump
> into db> for a mandatory reboot and fsck, but
> as a bogus program dump its registers like in
> Windows95.

When the kernel crashes, it will drop into the debugger only if you have
DDB defined in your kernel config file (which the GENERIC kernels do).  If
you don't, it'll do a dump and reboot.

> By the way, is NetBSD-1.3.1 compiled for 68040
> optimization? 

No, how could it be?  You'd be screwing over all the '020 and '030 users.

> I saw a control panel once in MacOS
> that allowed the CPU registers to be altered for 040
> performance, or MUCH slower for 68000-68030
> compiled programs. When i look at the include
> files it appears it is set up for 68020, but many
> makefiles compile with a gcc flag for 68020-68040
> optimization. Is that how the kernel binary are at
> ftp.netbsd.org?

The kernel binaries are compiled with -Dmc68020, which should choose 68020
level compatible instructions.  You can check out the man pages for gcc to
find out more about the options, tho.

> Thanks for any help or directions for help.

If you want, you can certainly compile your own kernel without unneeded
options and with a few added ones.  I have:

makeoptions     COPTS="-O2 -m68040"

at the top of my kernel config file, and I comment out these lines:

#options        M68030
#options        M68020          # Must have 68851 PMMU

Unless you're willing to deal with -current, tho, I'd stick with 1.3.1 (or
1.3.2, which may be out soon).

I hope this helps some.

Later.

-- 
Colin Wood                                 cwood@ichips.intel.com
Component Design Engineer - PMD                 Intel Corporation
-----------------------------------------------------------------
I speak only on my own behalf, not for my employer.