Re: how to turn off devices that monitor sensors

[Martin Husemann <martin%duskware.de@localhost>]

On Sun, Jun 12, 2022 at 06:18:26PM +0900, Henry wrote:
> In my post to the port-amd64 list with subject "/sbin/poweroff not
> working (NetBSD-current)", I reported that `/sbin/shutdown -p` fails
> to poweroff the machine.  The machine freezes with the last messages
> to the console:
> acpi0: entering state S5
> acpitz0: workqueue busy: updates stopped
> coretemp0: workqueue busy: updates stopped
> coretemp1: workqueue busy: updates stopped

I see - but this is only a secondary effect from ACPI failing to power
off the system. The kernel expects everything to stop, but some timers
are still running and trigger later workqueue actions that should
never happen when the machine is switched off ;-) The code dealing
with the queued results is not running any more and so the queue
stays busy and updates stop.

My guess:

 - this is  a bug in the acpitz(4) and coretemp(4) drivers that don't
   deal with shutdown properly (but usually harmless, the devices
   can not detach so if everything else works, this bugs are never
   triggered)
 - these bugs are not the cause for the failing power off (but that
   is plain speculation)

> I was thinking maybe if I turned those devices off before executing
> poweroff, maybe it would complete and the machine would turn off.
> 
> Wishful thinking?

I guess it is. The method nia pointed out is not stopping those devices
before shutdown, but disabling them totally at boot time. If that makes
a difference for the shutdown, it is at least a helpful data point
for further diagnostics.

Martin