Re: Problems with NetBSD 5.0 BETA

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Wed, 7 Jan 2009 14:13:57 +0200
Tero Kivinen <kivinen%iki.fi@localhost> wrote:

> I have been trying to put NetBSD 5.0 beta up, but I have had so my
> problems, that I think that either my hardware is broken, or there is
> something seriously wrong...
> 
> First problem is that when I try to compile the NetBSD I get
> segmentation fault - internal compiler errors on different files when
> compiling. When I continue using "./build.sh -u -x release" it usually
> goes forward and after 3-4 times I get everything compiled. I cvs
> updated the source code 2009-01-05, compiled everything (by trying
> enough times to get it done), installed kernel, rebooted, installed
> userland, rebooted, and tried again and I am still getting those.

That's almost certainly a memory problem -- run memtestplus from pkgsrc.
> 
> Has anybody seen any of that kind of behavior. If not I guess there is
> something wrong with my hardware (it is AMD x2 64 but I needed to
> disable ACPI and SMP to get interrupts working on my adaptec scsi card
> I have attached to it, the machine is booting and running from ATA
> disks using raidframe).
> 
> Another problem which might actually be actual bug, but I do not want
> to submit bug report if it is just my hardware faulting there too is
> following:
> 
> Boot up, mount root (using raidframe, not sure if that is related)
> without softdep, log in, remount root with softdep (mount -o softdep
> -u /), run vi /etc/fstab to add softdep option there, and get instant
> kernel panic. I have multiple crash dumps of this so I can submit bug
> report, but would like to rule out hardware first.
> 
> So if someone has newish (from this year) NetBSD 5.0 running, try
> changing root-fs to not have softdep option on boot, reboot, login,
> remount root-fs to have softdep option using "mount -o softdep -u /",
> run vi /etc/fstab and see if kernel panics. If it does or does not,
> send email to me, so I will know whether to submit bug report or
> not...



                --Steve Bellovin, http://www.cs.columbia.edu/~smb