powerdown restarts

[Geoffrey Leach]

On 07/02/2012 11:00:36 AM, Bill Davidsen wrote:
> Mateusz Marzantowicz wrote:
> > On 30.06.2012 23:32, Aaron Konstam wrote:
> >> On Sat, 2012-06-30 at 08:36 -0700, Geoffrey Leach wrote:
> >>> On 06/30/2012 03:51:45 AM, Mateusz Marzantowicz wrote:
> >>>> On 29.06.2012 22:37, Geoffrey Leach wrote:
> >>>>> This problem has been submitted to Bugzilla (836657), but I
> thought
> >>>> I'd
> >>>>> ask here to see if there are any fixes lurking.
> >>>>>
> >>>>> System is running 3.4.3-1.fc17.x86_64. When I systemctl 
> poweroff
> >>>> the
> >>>>
> >>>>> kernel reboots instead of powering off. Under Windows 7, power
> off
> >>>>> works as expected. All packages are up-to-date.
> >>>>>
> >>>>> Any ideas?
> >>>> What about shutdown -h ? Does it work as expected? Then try halt
> and
> >>>> poweroff commands.
> >>> It's my understanding that poweroff is a backwards-compatibility
> >>> implementation of systemctl poweroff, which I have tried to no
> avail. I
> >>> should have mentioned that. It appears that halt is the same.
> shutdown
> >>> if a link to systemctl. Bottom line is that I would not expect 
> any
> of
> >>> these to be any different, but I live in hope and will report 
> back
> if
> >>> there's any difference.
> >>>
> >>> I should also mention that systemctl poweroff works fine on my
> laptop
> >>> running the 32-bit version of Fedora 16.
> >>>
> >>> One point, FWIW. Power off is essential for my application. 
> Merely
> >>> halt-ing is no better than just leaving the system running.
> >>>
> >>> Thanks. I don't wish to seem ungrateful -:)
> >>>
> >>>
> >> I disagree with the other posters. There is a magic related to
> shutdown
> >> poweroff and halt. If you look at the man pages you will find that
> >> shutdown and poweroff have different options. It is clear that 
> when
> >> systemctl is called under a different name it checks the name and
> >> potentially reacts differently. For example poweroff by itself 
> will
> >> shutdown the machine . systemctl called by itself will not.
> >
> > If you check source code with is more reliable then any man page
> could
> > ever be, you will find that there really is nothing magical. Please
> see
> > file src/systemctl/systemctl.c in systemd source tree. Commands 
> like
> > halt, shutdown and power off call the reboot() function. I can 
> agree
> > that argument to reboot() may change between this calls but it's
> still
> > the same function they're calling.
> >
> Source code shows what the source code says, testing shows what it
> does.
> 
> I have several systems which either reboot or drop to some zombie 
> mode
> on 
> shutdown from the WM (GNOME3, XFCE, Cinnamon) and on all of them
> "shutdowen -h" doesn't power off (as the man page says is optional)
> while "shutdown -P" does.
> 
> So there is a problem, and while I generally agree that no matter how
> you get 
> the the system call it will do the same thing, clearly some user
> interfaces do 
> not call powerdown as part of shutdown.
> 
> Don't know if that makes it a system issue or a user inteerface 
> issue,
> but 
> hopefully that tip will give people a way to really power the damn
> thing off.
> 
> Oddly, hibernate does power down on all those systems, although as
> usual they 
> don't reboot cleanly.

Alas, while shutdown -P says "The system is going down for power-
off", it lies. poweroff -h fails to power off as well. Both reboot.

Presumably poweroff (by either route) eventually communicates with the 
BIOS. Does anyone know if this is correct?