Subject: Re: Summarisation on uvII/VS3100 booting
To: NetBSD Mailing list <email@example.com>
From: Chuck McManis <firstname.lastname@example.org>
Date: 06/20/1999 20:28:51
At 08:01 PM 6/20/99 +0000, NetBSD Mailing list wrote:
>o Booting from tape doesn't work for both platforms.
Actually I've done this on my Uvax II with a TK70, I don't have a TK50 to
try it on.
>The summary of all this is:
> - You simply MUST netboot the uvII to create a system from scratch.
> If you don't have a system to netboot from, you're out of luck.
> Once the system is built as mentioned above (in the install comment),
> you can boot and go just fine. NetBSD/vax STABILITY becomes the
> issue now.
This is, by far, the most reliable means to get to a running system.
Now it is very stable so far given two conditions:
1) You never swap.
2) You use NFS to mount all your disks.
I avoided problem 1 by configuring up a system with 64MB of RAM. This
clearly is only a stop gap until the real bug(s) are found.
By clearing up #1 I am able to focus on #2. Now for me, the issue is that
with MSCP disks mounted on my Sigma RQD11-E controller (not -S like Bruce
is using), after a period of time the kernel ceases to be able to talk to
the disk controller, and one by one processes get stuck in "diskwait" and
I've been whacking on some of the ddb code to get better visibility into
what is going on during this deadlock. The last time I got it nailed the
system was stuck with everyone waiting for a page to be released that was
dirty (delayed write was active). I don't get enough time to play with it
though, and I'm a bit concerned that perhaps the MSCP implementation in the
Sigma card isn't broken in the same way the DEC MSCP is broken. So to check
that I'm bringing up a second uVAX with a DEC MSCP board (KFQSA) and a
couple of DSSI disks. That project is waiting on the arrival of BA123 and
some misc parts (like a DSSI terminator). (not all from the same source.)
There was also the problem that I cut my VM teeth on the SunOS VM system
and the Mach VM system used in NetBSD is significantly different.
Anyway, I'll keep the list informed of my progress as I make it. And send
Ragge source code as I generate it. Of course he just rewrote the MSCP
driver so I may be stuck doing a major resync again ;-)