On Thu, 19 Sep 2002, Markus W Kilbinger wrote:

>     Bill> Please also try "printf() debugging".
>
> How/where to enter?

Oh, "printf() debugging" is adding printf's to the code to see where it
is. :-)

You need to be able to recompile source to make it work.

I used that to find that if I use the "bad" ofwboot, mbinit() failes when
setting low-water marks on the pools it sets up. No idea why.

>     Bill> I'm seeing a similar but slightly different error on my 2.01
>     Bill> system with -current kernels. I get the using so many
>     Bill> buffers containing so much memory line, then things die.
>
>     Bill> In my case, the interesting thing is I get this error with
>     Bill> ofwboot 1.5, but NOT with ofwboot 1.2! I'd been using an old
>     Bill> version, and so I wasn't seeing the problems.
>
> That means, with ofwboot 1.2 you can boot your g3!? Should I try that
> one with your modification?

Problem is I can't recreate it. What we can probably do is take the
ofwboot I have and binary edit the offending instructions to be nops.

>     Bill> Hmmm.... You aren't seeing the copyright. Not sure what to
>     Bill> suggest as that's printed right after the console's inited.
>
> I took a NetBSD 1.6 kernel so far... Should I try a -current kernel
> instead?

If you want to. I suspect that part of what's going on is I'm using a
custom kernel. I think the problem's ofwboot is loading the kernel wrong,
in a subtle manner. So adding/removing options will change failure modes.

The weird thing is I think it's a toolchain problem; I tried checking out
ofwboot code from the past, and it kept failing in the same way. Even
ofwboot from June, 2001. I really doubt no one's tested ofwboot since
then, so I think it was a change elsewhere (in the 1.6 toolchain) that is
causing the problem. Since I'm using the 1.6 toolchain to make all of the
test ofwboots, I keep encountering the same error.

Take care,

Bill