[9fans] meaning of "*nomp=" has reversed

[9fans] meaning of "*nomp=" has reversed

[Dave Eckhardt]

I have an odd machine which doesn't work right in the brave new APIC
world (for example, VESA mode can draw windows and menus but not paint
the cursor...?).

Anyway, last year I figured out that adding "*nomp=" to plan9.ini would
work around whatever it was.  I just pulled, and was surprised to see
the world revert to a sad state.  Eventually I noticed the kernel
chatting about the LAPIC during boot and investigated.

Now archmp.c:identify() doesn't only check for the existence of a *nomp
line but also feeds the right hand side to strtol().  It appears that if
there is no right hand side, strtol() returns 0 and the effect is as if
there were no *nomp line--which is a little counterintuitive.

Looking around it seems like *nomce is like *nomp, but *noetherprobe,
*nousbprobe, *nobios, *pcibios, *nopcirouting, *norealmode, etc., all are
the "old way" (presence is all that matters, not the right hand side).

Isn't there an argument for all of these working the same way?

Dave Eckhardt

Re: [9fans] meaning of "*nomp=" has reversed

[erik quanstrom]

On Fri Aug 10 18:44:19 EDT 2007, davide+p9@cs.cmu.edu wrote:
> I have an odd machine which doesn't work right in the brave new APIC
> world (for example, VESA mode can draw windows and menus but not paint
> the cursor...?).

you wouldn't happen to have usb legacy support enabled in bios, would you?

(unfortunately apic is no longer the brave new world.  in fact, it's getting
a bit harder to get newer machines to work based on mp tables alone.
msi is getting to be assumed.)

> Looking around it seems like *nomce is like *nomp, but *noetherprobe,
> *nousbprobe, *nobios, *pcibios, *nopcirouting, *norealmode, etc., all are
> the "old way" (presence is all that matters, not the right hand side).
> 
> Isn't there an argument for all of these working the same way?

there sure is.

- erik

[9fans] Weird symptoms unless APIC support is disabled

[Dave Eckhardt]

>> I have an odd machine which doesn't work right in the brave new APIC
>> world (for example, VESA mode can draw windows and menus but not paint
>> the cursor...?).

> you wouldn't happen to have usb legacy support enabled in bios, would you?

That's a good suggestion.  I saw that and disabled it, and also tried
*nousbprobe on the theory that these days usbstart happens automatically
whereas it didn't before, and I'm not using USB for anything on this
machine.  No dice.

Further details:

1. 9load says "pcirouting: 8086/24C1 at pin 1, irq 10".  I think that's
the IDE controller.  Is that a complaint, a warning, ...?  Since one
of the three USB controllers was ending up on IRQ 10, I was hopeful
that skipping them might help, but it had no effect.

2. While rio comes up in VESA mode and can paint, the mouse cursor is
invisible and keyboard input doesn't work--with one very interesting
exception.  ^T^Tr doesn't reboot the machine, but Control-Alt-Delete
does ("cpu0: exiting")--and this isn't happening somehow below Plan 9
because the key I'm using for Control is the one labelled "Caps Lock"
on the physical keyboard.  What could be responsible for that?

3. Turning off "USB Boot" in the BIOS doesn't help either.

This isn't a show-stopper for me as long as *nomp continues to work,
but it sure is odd.

Dave Eckhardt

Re: [9fans] meaning of "*nomp=" has reversed

[Dave Eckhardt]

>> Looking around it seems like *nomce is like *nomp, but *noetherprobe,
>> *nousbprobe, *nobios, *pcibios, *nopcirouting, *norealmode, etc., all
>> are the "old way" (presence is all that matters, not the right hand
>> side).
>> 
>> Isn't there an argument for all of these working the same way?

> there sure is.

Who has opinions on which way is better?

Dave Eckhardt

Re: [9fans] meaning of "*nomp=" has reversed

[ron minnich]

On 8/11/07, Dave Eckhardt <davide+p9@cs.cmu.edu> wrote:

> Who has opinions on which way is better?

everything takes an equal sign. Value is set by rval.

Somebody writes a simple program to tell you what your plan9.ini is
saying, since it seems to be ambiguous.

ron