Re: Which NetBSD on a VS3176?

[Johnny Billquist <bqt%update.uu.se@localhost>]

On 2014-05-20 07:36, Holm Tiffe wrote:
> Johnny Billquist wrote:
>
> [..]
> > > >   On which disk do you want to install NetBSD?
> > > >
> > > >
> > > >                       +-----------------------------------+
> > > >                       | Available disks                   |
> > > >                       |                                   |
> > > >                       |uid 0, pid 6, command sysinst, on /: file system
> > > >                       full
> > > >                       | b: sd1 (4.0G, IBM DCAS-34330)     |
> > > > /: write failepid 6 (sysinst): user write of 16388@0x1a2000 at 93080
> > > > failed: 28
> > > > d, file system is full
> > > > [1]   Illegal instruction     sysinst
> > > > #
> > > >
> > > > The sd0 is a Fujitsu M2954S-512, that one with the +5V on the case (!),
> > > > I'll try a smaller disk next, have n IBM0663 (?) here and a quantum, have
> > > > to look if they work at all....
> > > >
> > > > Regards,
> > > >
> > > > Holm
> > > > --
> > >
> > > Ok, I've tried to install on the IBM0663 disk (displayed as 995MB, sd0)
> > > this makes no difference at all:
> >
> > Did I miss something? Why did you expect any different result?
> > You're getting an error about the root file system being full, and this
> > on the installation system, where the root file system is a ram disk.
> > Makes no difference what harddrive you hook up.
> >
> >          Johnny
>
>
> The Filesystem gets full since sysinst is core dumping with an illegal
> instruction fault. After that I have a prompt, stty sane recovers the
> terminal, I can remove the core and do commands be entering them in the
> shell. Disklabel isn't really helpful then, it displays a "564" ??

You are first getting the file system full error. You then get sysinst 
crashing, and that is what triggers the writing of the core file, or 
else the screen dump you sent isn't telling the full story.

> I've tried the different disk size since the MV3100 M76 has a problem
> with 12BIT SCSI cmds and some wrape around of sizes over 4 Gig in the
> Bootroms.

Right. Which is only a problem in the booting, as that is the only time 
the device drivers from ROM is used. They are not used on a running 
system, and you have no such size limitations on disks on a running system.
Oh, and the size limit is actually only just over 1G unless I remember 
wrong.

> Again, that is NOT simh, that's on the iron and the miniroot had enough
> space to work and install NetBSD to the disk. It is just sysinst that bails
> out.

Yeah. sysinst can be a problem of its own.

> BTW: Miniroot is called miniroot because it should be so small to fit in
> the memory of the smallest possible machine for installation purposes..,
> just in Case someone doesn't know..
> SCNR

That's your interpretation. Others might have different opinions...

> So it seems to be exactly as I wrote before, You have a running System but
> no one is able to install it on the VS3100 "the normal way".
> Don't try to find the Bug using simh and the KA655, I've tried this, on
> this "architecture" the sysinst is working flawlessly and doesn't core dump.
>
> Last year I've tried to install over network so the network was up and
> running when sysinst core dumped. Now I'm booting from CD. Lets see if I
> can get up the network and copy the core dump for debugging purposes.

I did a network install on a 3600 only a couple of months ago, but that 
machine had 24 megs if I remember right.

> I tought sysinst may be compiled with the wrong CPU definition causing that
> "illegal instruction" core dump, but the ka655 used in simh is also a CVAX.
> It is possible that simh lets running instructions that should'nt,
> emulating to much fp a CVAX? Really don't know..

No. Such style of CPU differences are not the issue.

        Johnny