[ On Thu, November 20, 1997 at 18:35:55 (-0500), Gordon W. Ross wrote: ]
> Subject: Re: mapped device panic (fix)
>
> You know, you can use the serial console and still run a getty on
> the keyboard/display (or better, an xdm on /dev/kd).  It's a lot
> easier to work on the serial console for debugging, expecially if
> the "terminal" is an xterm (with logging) on another machine that
> runs tip or something similar.  You can cut and past stuff into
> and out of the remote console xterm.  That's how I do it.

We tried that with an NCD19c terminal's serial port but couldn't get the
cabling right or some such thing -- need to check with a breakout box....
 
> OK, so maybe it is NFS-root or X related.

Probably NFS-root or NFS-swap related, though oddly only after running
the X server is the problem manifest, so perhaps it's an interaction
between the pmap code for device access and something else in the VM.

Does the NFS swap use the same read/write size as the root mount,
i.e. as controlled by NFS_BOOT_RWSIZE?

I should admit that one of the reasons I've been doing this work on
trying out NetBSD/sun3 so vigorously of late is that just the other day
something on my network changed such that I'm no longer able to boot
SunOS on my workstations.  The problem seems to be that NFS 8k reads
fail (the stations boot and load /sbin/init, but then hang and tcpdump
shows the last traffic as the seemingly successful load of /sbin/init).

It turned out that I could boot a netbsd kernel off the same server and
root filesystem as the SunOS was coming from, and even get it far enough
to have it load SunOS init and sh and all worked OK, so I re-built a
NetBSD server with 1.3A to see if I could get my workstation running
from that.

Unfortunately I've been unable to identify the cause of the SunOS boot
failure to date and nothing I change or revert seems to make any
difference.

So, if NetBSD's NFS swap block size is 8KB, then there may very well be
a local problem on my network that could be causing the problems I see
since installing the pmap-88.c patch.

> BTW, does the problem machine have swap space?
> (i.e. what does "swapctl -l" show you?)

Yes, I have a 32MB file exported and mounted as per the diskless(8)
instructions, and swapctl seemed to show the right things.  The problem
machine is dead just now, but the /etc/exports contains:

20:32 [30] $ ls -l /export/swap/very
-rw-------  1 root  wheel  33554432 Nov 20 15:17 /export/swap/very
21:33 [31] $ grep swap /etc/exports
/export/swap/very -maproot=root very.weird.com

-- 
							Greg A. Woods

+1 416 443-1734      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>