On: Thu, 24 Jul 1997 02:31:07 -0700,
   STEVE LUMOS  <slumos@nevada.edu> writes:

 >We just upgraded to the latest snapshot, but can't build a kernel yet
 >(using source supped on the 22nd I think), because I can't find
 >information on what the config line is supposed to look like.

Uh, like ``GENERIC''.... 

options(4) describes MI options, the Red Demon Book has more detail
about kernel internals and (sort-of) what the options mean, and there
are config files tailored for each model of DECstation and for
diskless boot in /usr/src/sys/arch/pmax/conf.

Was there more you wanted to know?


 >I know I have to add a swap line to /etc/fstab (done), I know I need to
 >call swapctl in /etc/rc (not done yet). 

Swapon -a also works, as long as you have the fstab line and "options
COMPAT_12".  It is deprecated, though.


 > What I don't know is what the
 >current syntax of the config line is.  The config file for GENERIC says:
 >
 ># Note that this configuration is unlikely to work, yet...
 >config          netbsd  root on ? type ?

Oops.  Actually this comment is in all the config files and is
now well-and-truly out of date. The "root on? type ?" config *does*
work...


 >and the NEWCONF config is out of date.

Actually, if you get a current sup, NEWCONF is *gone*.  NEWCONF was
the first ``new config'' file I used when smashing the 4.4bsd/pmax
kernel code to use the Torek config.

  Should I just use what GENERIC
 >uses, or am I supposed to do "root on rz0a type ffs".  What about dumps?
 >Are we currently transitioning?

You can do 'root on rz0a type ffs'  if you want to boot only off rz0a.
You can do  'root on rz?a type ffs' if you want to boot off any disk.
There is no transition there.

Dumps, AFAIK, have never worked, not even on 4.4bsd/pmax and I've
never added code to make dumps work.  Officially that's waiting on
using the MI scsi code.


 >The other question I have is about debugging.  I saw something recently
 >here about DDB.  Does that mean DDB is now working on PMAX?

Yes, it is. No symbols though, only hex addresses, so it's painful
unless you have an adjacent window on which to use nm to map addresses
to names and vice-versa.  Everything else (breakpoints, stepping,
stack traceback, etc) seems to work.

 >Also, has
 >anyone successfully built a kernel with the KGDB options uncommented?

Nope.  There is no support for that.  I didn't find any 4.4bsd/pmax
code for KGDB support (though I may have looked in the wrong place).
So far I've never found time to write a serial remote-gdb stub...
if you wanted to write one that would be  really, really neat.