[9fans] kbd fail

[9fans] kbd fail

[Richard Miller]

The change to /sys/src/9/pc/kbd.c made on 11 Jun breaks my keyboard
(an IBM "compact" ps/2 keyboard which has no LEDs).  Symptom:
bootstrap hangs at "root is from" prompt.  If I change the new
setleds() function to return immediately, the keyboard works fine.

Re: [9fans] kbd fail

[erik quanstrom]

On Mon Jul 19 12:44:08 EDT 2010, 9fans@hamnavoe.com wrote:
> The change to /sys/src/9/pc/kbd.c made on 11 Jun breaks my keyboard
> (an IBM "compact" ps/2 keyboard which has no LEDs).  Symptom:
> bootstrap hangs at "root is from" prompt.  If I change the new
> setleds() function to return immediately, the keyboard works fine.

without this change, a large number of compact keyboards
don't work.  :-)

perhaps adding

	outb(Cmd, 0x20);
	intb(Data);

at the end of setleds fixes your problem?

- erik

Re: [9fans] kbd fail

[Richard Miller]

> perhaps adding
>
> 	outb(Cmd, 0x20);
> 	intb(Data);
>
> at the end of setleds fixes your problem?

No, it doesn't.

Re: [9fans] kbd fail

[Steve Simon]

Sorry, this one was my bad. I tested it on ebver machine I had to
hand and had no problems.

I'am sure geoff will roll it back - he had concerns but I was confident,
oh well.

-Steve

Re: [9fans] kbd fail

[Richard Miller]

> I'am sure geoff will roll it back

If it's useful I don't mind helping with testing to get it right.

Could the problem be something to do with not reading back the
acknowledge responses after sending the command and parameter to
the keyboard in setleds?

Re: [9fans] kbd fail

[erik quanstrom]

On Mon Jul 19 14:49:52 EDT 2010, 9fans@hamnavoe.com wrote:
> > I'am sure geoff will roll it back
>
> If it's useful I don't mind helping with testing to get it right.
>
> Could the problem be something to do with not reading back the
> acknowledge responses after sending the command and parameter to
> the keyboard in setleds?

the original code came from me, though i wasn't the original
author.  while steve is gratious to assume the blame,
it's probablly my fault.  the problem the code is trying to solve
was that a fair number of compact keyboards, such as the one
that steve has, need numlock in the proper state to be able to
type many of the keys.  but setting numlock must be done from
the host.

since setting the leds in a ps/2 keyboard has big side effects,
it may be that your ledless keyboard takes exception to our
tricks.  or that somehow we are leaving the keyboard in a
potentially bad state.  is it possible that the keyboard is
just sucking its thumb, rather than the machine being hung?

i don't have my keyboard controller documentation handy,
but i'll review the code and the documentation when i get
home.

- erik

Re: [9fans] kbd fail

[Richard Miller]

> is it possible that the keyboard is
> just sucking its thumb, rather than the machine being hung?

Yes, it is almost certainly the keyboard (or controller?) not
sending anything to the pc, which is effectively hanging the
bootstrap procedure.