Subject: Re: NetBSD-current resets
To: Mason Loring Bliss <firstname.lastname@example.org>
From: Guenther Grau <Guenther.Grau@marconicomms.com>
Date: 03/09/2000 21:15:14
Mason Loring Bliss wrote:
> On Thu, Mar 09, 2000 at 07:56:58PM +0100, Guenther Grau wrote:
> > I still don't buy this completely, yet.
> That's fine. I have nothing invested in the idea softdep being buggy... I
> just want to see it all work. :) I'll happily be a guinea pig.
Cool! Great attitude :-)
> > Are you sure that your mainboard can cope with 128MB-DIMMs.
> Yeah. When I disabled softdep, it ran fine with the newer 128 meg stick.
> I then threw in the original (presumed bad) 128 meg stick and it worked
> fine. I think I even swapped positions with no problems. The only difference
> between it working and not working was my booting single-user and turning
> off softdep everywhere where I had it turned on.
It might be that softdep is excercising the memory a lot, whereas
softdep, memory doesn't get stressed that much.
> My mainboard is, incidentally, a relatively recent DFI. I'll nail down the
Hmm strange, recent mainboards shouldn't have any problems with 128MB
> > Is it possible to put 4 32MB sticks in there (from work? :-)?
> That would be tough, but I might be able to arrange it... I have to borrow
> it from unused machines, and we're about out of those! Heh.
You might borrow them during weekend.
> > If this system is not stable, then we should go bug hunting in the softdep
> > code.
> Since someone suggested using COMPAT_43 as a workaround to get ifconfig and
> dhcp and stuff working, my system is now rock solid once again, so it's time
> to shake it up. :) My kernel right now has softdep compiled it - I'm just not
> using it anywhere. However, since I have a silly large amount of RAM, I've
> just started (yesterday) using an mfs for /tmp, so I've got a partition free
> on which I can inflict softdep.
> > Sorry, but I forgot: What happens when it's unstable?
> > Will the kernel crash? Have you got a backtrace?
> I *believe* it was unable to successfully access a device for purposes of
> dumping. I'll have a far more detailed report for you this evening, once
> I get home.
No hurry, I am going home now (and I'm probably not the
right person to debug this anyway :-( It's just some
bad memories (pun intended ;-) that popped up when you
were talking about 128MB sticks and I also wonder why everybody
else is running softdep fine in their machines,
most of which probably have at least 128MB or more.
Maybe we can make a poll:
Who is running softdep with 128MB or more in their machine?