FreeBSD 6.1-STABLE: Unexplained power off

[Ian Smith]

On Tue, 15 Aug 2006, Christian Walther wrote:

 > This is just a wild, uneducated guess, because I'm not a long FreeBSD
 > user, but from my point of view this error could really be related to
 > ACPI/APM, as already has been suggested.

It smells a bit that way to me too.  I've just read the whole thread,
but going back to the original post's kernel conf, android had APM and
apm_saver in there, but the dmesg confirmed an ACPI boot, complete with
a complaint by apm_saver refusing to load because APM wasn't loaded.  As
it never is if ACPI is loaded, as I understand it.  (caveat: 5.5-STABLE)

android also mentioned trying to do things with APM settings in BIOS.  I
suspect APM should be _disabled_ in BIOS, and ACPI enabled, with ACPI
power (etc) management used instead .. someone correct me if I'm wrong;
I'm really unsure how much APM functional emulation remains in ACPI?

 > Maybe the machine is trying to go to suspend, but fails while doing
 > so, which in the end would mean that it can't recover from the
 > suspend, but has to reboot completely, resulting in dirty file
 > systems. It wouldn't reach the suspend state correctly, which could
 > leave everything depending on ACPI/APM in a undefined state, including
 > the hardware. This would explain why the machine has to be turned off
 > properly by pressing toe power button for such a long time.

Maybe.  If APM is enabled in BIOS, but not loaded, could spell trouble.

 > I'd try to use the machine without ACPI/APM enabled. If possible,
 > compile a new kernel without it being enabled. This might not be
 > possible because you're on a SMP-system, thou, but you might want to
 > check your configuration files for suspend or hibernation -- and turn
 > them of.

Well it won't likely work with _neither_ enabled, and I suspect you're
right about SMP needing ACPI.  android suggested failure to boot with
neither enabled, which sounds likely.  What's in /boot/loader.conf?

Cheers, Ian

 > With ACPI/APM turned on, leave the machine idle for some time and see
 > if it shows the same behaviour. When it shuts down cleanly it's likely
 > that suspend/hibernation fails due to the high load introduced by the
 > build process.
 > 
 > I've seen and experienced similar problems on other platforms, such as
 > OS X and (sorry) Linux.