Re: Atari TT

[Michael Schmitz <schmitz@mail.biophys.uni-duesseldorf.de>]

Thorsten,

> >I don't know - I've never checked out that branch. Time to check what's
> >actually in there. 
> 
> Just in case: my top is 86c0899bb70ee8fa168a7a0711506d616161786e plus
> Alan’s patch (m68k-queue is regularily rebased against torvalds).

I'll look out for that. 
 
> >Looking at the git web interface for m68k-queue, the current state still is
> >to use the old atari_ethernec (hacked copy of ancient ne.c driver). So the 
> >error you will be getting is precisely what I've been seeing while using that
> >version - kernel locks up on the first timer-faked interrupt in. 
> 
> Oh the joy. But good to know it’s a *known* problem.

Not entirely sure - see below.

> 
> >The new driver may or may not fare better in that regard; I'll test that
> >ASAP. If it does the same thing, fixing this will be _much_ harder because I
> >cannot touch any code in ne.c. 
> 
> Mh. Is your “new driver” somewhere in Geert’s branches?

It appears not - I last sent something off (or so I believed) last August
but that may have been lost in my mail setup. 

> >The whole point may be sort of moot anyway - has anyone managed to boot a TT
> >sucessfully all the way, with console screen output,  to the point where it 
> >fails to mount a root filesystem with your nonet kernel? That may require
> 
> Erm… gemini8 booted it, and he says he gets a readonly filesystem,
> but we weren’t finished debugging that, as he’s got no experience
> using GNU/Linux. The kernel may actually have *succeeded* booting.

That sounds like it's booted alright - 'ext filesystem mounted readonly' or
similar is one of the last messages before startinh init, or one of the first messages
from user space. 
 
> >use of the -s bootstrap flag, as I've said before. If that works out, we'll
> 
> The -s flag is required for video output, at the very least, yes.

Yep - so does the net kernel show any console output with the -s boot flag?
If yes, it's a known bug. If not, it's something new and sinister. I'd be
surprised if the TT behaved substantially different from the Falcon there. 

> >also have console output at the time the network driver is loaded. That
> >would allow some kind of debugging (at one or two removed). Without any kind
> 
> Eh, I still have CONFIG_MODULES=n ;-) but copying a .ko onto the

s/loaded/probed - putting a module on the system won't do you any good
without modules support (unless you're extremely good at kernel hacking).

> filesystem using MiNT is probably possible. It’s just the several
> hundred kilometres between the box and me…

Several hundred - bah. How's 20000 for you?  :-)

Anyway - I'll see what I can get with my current tree. 

Cheers,

	Michael