APMD-List:
Archives

  From: Chris Worley <cworley@symbionsys.com>
  To  : <apmd-list@worldvisions.ca>
  Date: Wed, 19 Jul 2000 21:55:51 -0600

Re: Using power switch to go into standby mode

Craig Markwardt wrote:
> 
> Chris Worley writes:
>  > Jont Allen wrote:
>  > > Chris Worley wrote:
>  > > >
>  > > >
>  > > > I looked at apmd_proxy, but I don't see how to specify: "if power off
>  > > > then standby instead".
>  > >
>  > > The response you got from Craig Markwardt is the definative answer.
>  > > He said that "power off" is not supported by the apm kernel module.
>  >
>  > It must be supported in the APM kernel module.  If I compile the kernel
>  > without APM support, then the power switch has no effect (I have to pull
>  > the plug).  Likewise, if I panic a kernel with APM support, the power
>  > switch no longer works.
>... 
> The BIOS may pass some power management messages to the kernel when
> the button is pressed, like, "I'm about to power down!" or "Now
> entering standby mode!" but those actions are determined by the BIOS,
> not Linux.  The hardware is there, everything probably works, but
> there's just no way for the APM BIOS to communicate these kinds of
> events.  As David Brownell mentions, you may be able to configure it
> from the BIOS menu.

The bios has a setting: 

  soft-off by PWRBTN: [delay 4 sec | Instant-Off]

Setting it to "Instant-off", the default, behaves like you describe. 
The kernel gets the power-off message, but can't do anything to stop it.

If I set it to "delay 4 sec", then I get a suspend message instead. 
This brings about other problems: The pegasus USB nic driver goes nuts
(recursive panics), and a "resume" message is sent from the bios about
five seconds after the suspend is sent.

This is true with "apm -s" too.

Standby, behaves similarly, but doesn't cause the pegasus driver to
panic the kernel:  the screen blanks for about 5 seconds, then a resume
message is sent from the bios.

I can effect the "suspend" message coming from the bios, and change it
to a standby. in "check_events" in "apm.c" I changed the "case
APM_SYS_SUSPEND" to:

>			apm_set_power_state(APM_STATE_REJECT);
>			standby();
>			break;

I'm guessing I'm rejecting the suspend request, and changing it to a
standby.  This could just be changing the kernel to think it's a
standby, and not doing "whatever" to the USB drivers that drives them
nuts.

So, now I can make the power button cause a standby... but five seconds
later, a resume is received, and I, again, can't do anything about that
message (I've tried: I can force a standby whenever a resume message is
given, which causes an endless loop of standby/resume and display
off/on).  The resume message needs to be inhibited before it reaches the
kernel.

Any idea where this resume is coming from, and/or how can I stop it?

> The ACPI specification, however, provides lots of ways to program and
> manage special hardware buttons and switches.  Here, it's built right
> into the specification, so there needn't be any monkeying around in
> the BIOS.  Of course, as I mentioned, the Linux ACPI support is still
> rather embryonic.

ACPI doesn't like the I-Opener at all.  I've posted to that group
without reply.  I've given up trying.

> 
> It is likely that the QNX operating system (which originally ran on
> your machine, yes?), was specially modified for the hardware. 

Yes.  That is the behavior I'm trying to duplicate.

Thanks,

Chris