Plan 9 on Sun IPC + cgthree?

Plan 9 on Sun IPC + cgthree?

[forsyth]

it's probably safe to assume the ROM has loaded the kernel
(it usually moans if not), so your problem most likely is that
the kernel is failing after it is started by the ROM but before
it has mapped the screen.  it's probably crashing while mapping
or accessing the screen.

the banks of memory needn't be the same.  on your machine,
it will use the 16 from the first bank and 4 from the second.
it will ignore the third bank.

the initial hex numbers are printed by the ROM, so they
don't say much about plan 9's state once started.

Plan 9 on Sun IPC + cgthree?

[Andrew]

dhog wrote:
> Speaking of hacks to the ss confinit(), here is a patch to enable
> the use of a serial console on a sparcstation which has a monitor.

Thanks for the suggestions so far. Unfortunately none of them seem to 
have impressed my IPC very much...
Various combinations of hardcoded cgthree and/or selection according 
to slot number always resulted in the same "nothing". Using the 
serial console patch does the same, only on the terminal, not the 
screen.

I guess there must be more to it than just the screen problem. Any 
hints as to where to look? (only main.c or also mmu.c/trap.c etc.?)
Which main stages are there before the first output on the screen?

Andrew.

P.S. Brian L. Stuart apparently got Plan 9 to run on a bwtwo-only 
IPC. I've got no problems with removing the cgthree for the time 
being, however just yanking it out and plugging the monitor (Sony 
GDM-1962B) into the bwtwo results in no picture. Are there any nvram 
incantations I should be aware of? Would it help to wait for a full 
moon?

P.P.S.
Type  help  for more information
ok help
There's no help for the weary.

I might expect this kind of sarcastic wit from Plan 9, but that's my 
IPC's Boot-ROM...

Plan 9 on Sun IPC + cgthree?

[David]

Speaking of hacks to the ss confinit(), here is a patch to enable
the use of a serial console on a sparcstation which has a monitor.
The comment explains what you have to do to take advantage of it.
You probably also need to say ``setenv input-device ttya'' and
``setenv output-device ttya'' to make the boot monitor use the
serial console...

diff /sys/src/9/ss/main.c ...
321a322
> 	int montype;
332a334
> 	montype = ((uchar*)va)[0x1f];
340,343c342,345
< 	 * If we find a frame buffer, we always use it as a console
< 	 * rather than the attached terminal, if any.  This means
< 	 * if you have a frame buffer you'd better have a builtin
< 	 * keyboard, too.
---
> 	 * If we find a frame buffer, we use it as a console
> 	 * rather than the attached terminal (if any), except if
> 	 * the console is set to ttya in the nv-ram.  Type ``10 1f eeprom!''
> 	 * in new command mode to set this...
344a347
> 	if (montype != 0x10)		/* ttya */

Plan 9 on Sun IPC + cgthree?

[Brian]

>I found that although my cgthree is in a lower slot than the
>onboard bwtwo the kernel hardcodes bwtwo for SLC and IPC systems,
>rather than scanning the slots. Compiled a new kernel with quick
>hack: case IPC hardcode my cgthree instead. With this kernel the
>numbers go up to 7ec00 but otherwise no change.
>My first thought was that it was switching to the bwtwo after
>reaching the 7ea00, but the new kernel doesn't help. Any ideas
>where else to look?

I found exactly the same behavior with the original kernel,
but as it turned out, I did have a bwtwo only machine I
could try it on and it worked there.  The kernel hack I
did was to allow the IPC to be picked up by the same code
as the SPARC 2 and let the kernel figure it out.  It seems
to work fine.

In particular, I just commented out the case 0x52: label
in the switch statement and let the default: case pick it
up.  That has booted correctly on an IPC with a cgthree.

Hope this helps,
Brian L. Stuart
Math/CS Dept, Rhodes College, Memphis, TN
stuartb@acm.org
http://www.mathcs.rhodes.edu/~stuart/

Plan 9 on Sun IPC + cgthree?

[Andrew]

Recently I was virtually given a Sun IPC, with 24MB RAM,
cgthree graphics and Sony 19" monitor. I think it would make
a really nice Plan 9 terminal.

I've read the docs ("should work, but..."), also John Kennedy's
"There is Nothing so Simple" guide. Booting the 9ss that is
on the distribution CD, the system counts up the hex numbers to
7ea00 and then freezes (no L1-A, nothing)

I found that although my cgthree is in a lower slot than the
onboard bwtwo the kernel hardcodes bwtwo for SLC and IPC systems,
rather than scanning the slots. Compiled a new kernel with quick
hack: case IPC hardcode my cgthree instead. With this kernel the
numbers go up to 7ec00 but otherwise no change.
My first thought was that it was switching to the bwtwo after
reaching the 7ea00, but the new kernel doesn't help. Any ideas
where else to look?

I've read that only two banks of memory are supported. Might this
inherently mean the banks have to be the same? The 24MB is 16MB in
bank 0 and 4MB each in banks 1 and 2.

I assume the hex numbers come from the ROM receiving the kernel
via tftp. Can I calculate which value should be reached given a
9ss kernel on my PC?
I have booted a Sun3/60 sucessfully with my setup, so for now I
am blaming the Sparc.

Anyone ever had an IPC running Plan 9?

Andrew.

P.S.
SBus slot 0 le esp dma
SBus slot 1
SBus slot 2 cgthree
SBus slot 3 bwtwo