Power-button event after resume from S3

[Felix Kuehling <fkuehlin@ati.com>]

Hi,

I am investigating a problem where acpid receives a power-button event
immediately after resuming from S3. I looked at the mailing list
archives and found that this problem is not entirely uncommon. The
advice users got so far all looked like workarounds for the real
problem. The same goes for
http://bugzilla.kernel.org/show_bug.cgi?id=3525, which further declares
it as a bug in acpid. However, I'm not sure it is really a bug in acpid.

This is my understanding:

The only ways for acpid to distinguish power-up and power-down are hacks
that somehow remember that the system is going into suspend and to
ignore the next power button event. However, there is no guarantee that
there will be such an event after resuming, so you need some kind of a
time-out. All that doesn't feel very robust to me and that's why I'm
calling it a hack.

The kernel on the other hand has more information. It could reliably
distinguish power-up and power-down events. The first SCI interrupt
after resume always seems to be the power-up event, no time-out
required. Thus it could generate different events for power-up and
power-down in /proc/acpi/event. That would make it much easier and
cleaner for acpid to distinguish them and take the appropriate action.

Finally my assumption is that it is correct behaviour for the ACPI
hardware to report the power-button to the OS on resuming from S3. I
came to that conclusion after reading parts of the ACPI 2.0 spec. Please
confirm.

Are there plans for a real fix for this problem, either in the kernel or
in acpid?

Thanks,
  Felix

P.S.: Please include the CC-list on replies.