Re: Installing 8.1 on a Beige G3

[Michael <macallan%netbsd.org@localhost>]

Hello,

On Sun, 21 Jul 2019 21:35:00 +0200
fixerb@national.shitposting.agency wrote:

> On 2019-07-19 20:39, Michael wrote:
> > I had the same problem with my beige G3, the IDE controllers in this
> > thing are notoriously sensitive - try different ( newer! ) IDE cables,
> > reseat all the connectors, cross fingers.
> > This worked on mine at least.  
> 
> Thanks very much Michael, I did some swapping and while I'm still
> getting some errors here and there, it appears to be working (e.g
> sysinst got the disklabel written).

Progress.
Things I'd try:
- make sure the master and slave devices are jumpered correctly ( some
  drives have a 'slave present' jumper )
- try with only the boot drive present on the 1st IDE bus
- try the other connector on the cable
- make sure the cable isn't backwards ( if it has 3 connectors, the
  ones that are closer together are for drives )
- try another cable

As I said, the thing is PICKY.

IIRC bad cables caused all sorts of DMA errors on mine while MacOS
didn't have any trouble at all ( probably doesn't use DMA... )
First I thought our DMA code is buggy on heathrow chips, so I disabled
it and most errors went away. That turned out to be wrong though.

The G3 also has a SCSI controller, if you have a drive sitting around
that may work as a workaround.

> But as usual, as one problem gets resolved, more pop up.
> Writing is agonizingly slow. The disklabel alone took ~20mins and
> after that I let mkfs work for 3 hours (I left the geometry on
> defaults) and it hasn't finished. During all that there are device
> timeout errors popping out now and then. Fsbn numbers are
> changing so I presume it actually writes something, but again, it's
> terribly slow.

I think this is still very much the same problem.
> 
> Should I just let it run for a week, or is there something I can do
> about it?

That is not how things should go.
Did the kernel complain about DMA errors or read/write errors?
If there are read errors and the block numbers are random then you
likely have DMA errors.

have fun
Michael