[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Fwd: Macintosh Quadra 950 Not Booting



On Fri, Aug 7, 2015 at 8:31 PM, Finn Thain <fthain@telegraphics.com.au> wrote:

On Fri, 7 Aug 2015, Greg Andrzejewski wrote:

> On Fri, Aug 7, 2015 at 12:43 AM, Finn Thain <fthain@telegraphics.com.au>
> wrote:
>
> >
> > On Thu, 6 Aug 2015, Greg Andrzejewski wrote:
> >
> > > Greetings,
> > >
> > > After finally getting my Mac SE/30 working again, I set about trying
> > > to get a modern version of Linux installed on the little fellow.
> > > Early experiments with 3.14 kernels were successful
> >
> > BTW, are you using mac_scsi on the SE/30?
> >
>
> I've just been grabbing kernel images from the Debian ports archive, so
> I think so.

If convenient sometime, can you send me the dmesg output (or
/var/log/kern.log) from your SE/30?

Sure, I'll grab it next time it's booted up.
 

> 3.14 kernels would panic with a bus error on bootup unless I disabled
> pseudo DMA, but I'm thinking that might be a hardware issue with my
> machine.

Probably not a hardware issue -- mac_scsi always did that.

Good to know.

> I had to solder a few jumper wires on the motherboard to repair damage
> from leaking capacitors around the video/audio circuitry. Never really
> had any problems with SCSI in MacOS, but I cleaned a lot of capacitor
> goo from around the NCR chip, so sometimes I wonder. Linux gets very
> upset if I have anything attached to the external SCSI port

Also likely to be a mac_scsi issue if MacOS works OK with that setup.

> and MacOS bombed out yesterday copying files to a zip disk, so it's
> probably something I'm going to need to address in the future.
>
> FWIW, the pseudo DMA problems have gone away for me in the 4.0 kernels,
> by which I mean I no longer have to pass mac5380=-1,-1,-1,-1,0 on the
> command line. I've been meaning to check if there was a fix or if pseudo
> DMA is just disabled by default now.

Right. I disabled PDMA by default as part of a patch that went into v3.19
because I've never seen it work properly on any Mac. (Some LC and Classic
models will need v4.0 or later because of commit 2e874d178c9e.)

But PIO is horribly slow. I have a bunch of unfinished mac_scsi patches
that I hope to return to. The PDMA bug is one that I'd really like to fix.

Yeah, the SE/30 needs all the help it can get. It's about 20 minutes from bootup to login prompt.

> > If you would like, I can send you a current kernel binary that should
> > boot any Mac (that is, SE/30 or Quadra 950).
> >
>
> Yeah, that'd be great. I think the most recent kernel I was able to boot
> was 2.6.29 or something that I found online.

I've uploaded a kernel binary here,

https://www.telegraphics.com.au/~fthain/gregta/vmlinux-4.1.3-mac
https://www.telegraphics.com.au/~fthain/gregta/linux-m68k-image-4.1.3-mac.tar.gz

SHA1
48b6c18792479a6d9886151c0457675df36b8678  vmlinux-4.1.3-mac
9140852171bdf0a9a5e9b8d0cd28530dca5697eb  linux-m68k-image-4.1.3-mac.tar.gz

I'm afraid I haven't tested it, and I don't have much reason to believe
that it will work better than gunzip'd vmlinuz-4.0.0-2-m68k. The tar file
has some modules but you probably won't need them.

Tried booting both compressed and decompressed kernel images, same behavior with Penguin. Can't say if anything comes across the serial port as I left my cable at work, still hooked to one of the Quadras I've stashed under my desk. I tried the new 4.1 kernel from the ports repository this afternoon, same story with Penguin, but EMILE seems to get a little further. The screen doesn't clear, but no death chimes either. Jumping in to MacsBug on reboot, I found this in __log_buf:

 Displaying memory from 3240d0
  003240D0  0000 0000 0000 0000  004C 003B 0000 0033  .........L.;...3
  003240E0  4255 473A 2073 6368  6564 756C 696E 6720  BUG: scheduling
  003240F0  7768 696C 6520 6174  6F6D 6963 3A20 286E  while atomic: (n
  00324100  756C 6C29 2F2D 3136  3232 3532 3337 3736  ull)/-1622523776
  00324110  2F30 7834 3038 3032  3666 3200 0000 0000  /0x408026f2.....
  00324120  0000 0000 0022 0012  0000 0014 4D6F 6475  ....."......Modu
  00324130  6C65 7320 6C69 6E6B  6564 2069 6E3A 0000  les linked in:..
  00324140  0000 0000 0000 0058  0048 0000 0034 4350  .......X.H...4CP
  00324150  553A 2030 2050 4944  3A20 3133 3232 3638  U: 0 PID: 132268
  00324160  3630 3138 2043 6F6D  6D3A 2020 4E6F 7420  6018 Comm:  Not
  00324170  7461 696E 7465 6420  342E 312E 302D 312D  tainted 4.1.0-1-
  00324180  6D36 386B 2023 3120  4465 6269 616E 2034  m68k #1 Debian 4
  00324190  2E31 2E33 2D31 0000  0000 0000 0000 0000  .1.3-1..........

Signs of life! If the scheduler's starting, we're way past the early init stuff, right? Just to be sure, I powered down, verified that region of memory was trashed, and tried booting again. Same message.

I was confused at first, because I wasn't seeing all the earlyprintk stuff (ABCDEFG...) in the buffer. Then it dawned on me to check if any of that stuff goes actually through the log buffer or just directly to the console/serial port, and it appears to be the latter. Maybe it needs to, but if that were easy/a-good-idea I'd have to imagine it'd be doing it already. Guess I could give it a shot, but time's probably better spent figuring out what's going on with the display and serial port.



>
> Just tried with decompressed kernel from the debian ports repository and
> extensions disabled. Same behavior :-(

Well, I was not expecting that. I would double-check the obvious, if you
didn't already. That is, check that you selected the right kernel binary
and confirm no gunzipping in the log as Penguin boots it.

--

Right, Penguin correctly notices the lack of compression and loads the ELF directly.

-Greg

Reply to: