standby and hibernate questions
David Mathog
Question 1: what exactly does xfce4's "suspend" option do when it is
invoked in the logout menu? That works properly, but I have not been
able to get the system to suspend the same way using command lines, for
instance, with pm-suspend or writing to /sys/power/state.
Question 2: is it possible to enable the system to wake on a WOL packet
received while in this suspended state?
Question 3: does hibernate work?
Background:
Mandriva 2008.1, desktop system.
I was experimenting today to see what happens when this machine is
placed into a (very) low power state. Initially I tried:
echo -n "$X" > /sys/power/state
where X was standby, mem, or disk. That worked very poorly, indicating
that more was needed. (How poorly, let's just say that I strongly urge
you not to repeat that experiment!)
Then I ran Gnome, KDE, and finally xfce4 looking for power controls.
xfce4 had a "suspend" option in the logout menu. That worked great.
The system dropped down from 70 watts (running but idle) to 5 watts in
about 1 second, and it came back up again also in a second or so
following a press of the power button.
Afterwards there was a bunch of stuff in /var/log/pm-suspend.log,
indicating that pm-utils was involved somehow. Clearly it went through
the steps in /usr/share/pm-utils/sleep.d on shutting down and starting
up again. However, running the command "pm-suspend" in a shell did not
do the same thing. What script or command did xfce4's suspend use???
---
Also, in terms of being able to restart the system, it would be nice if
it could be done remotely (for cluster work, not for desktops so much).
Even though these machines have WOL working, sending such a packet to
the machine while it was in the suspend mode did nothing. Is there some
acpi configuration that would enable the WOL packet to start the system
from suspend?
---
Lastly, xfce4's "hibernate" option appeared to do a save to disk, but
then locked the machine. Moreover, the power useage increased from 70
watts to 107 watts. That increase is consistent with the Athlon64 power
saving modes being disabled. The power button was ignored. To reboot I
had to remove the power cord, wait a bit, then plug it back in. After
that the boot came straight into linux (skipping the normal grub menu,
which in any case defaults to windows). Following that, subsequent
reboots were normal. So "hibernate" seems to have rewritten grub, then
not actually managed to get the system into the hibernate state. Does
this work for anybody on a desktop machine?
Thanks,
David Mathog
David Mathog
method is this:
1. lspci | grep -i ethernet
find the PCI entry that corresponds to your ethernet controller.
2. cat /proc/acpi/wakeup
find the corresponding PCI entry. It will say "disabled". In my case
it was MMAC.
3. echo MMAC > /proc/acpi/wakeup
3.5 ps -ef > before.txt
4. In xfce4 logout with suspend and the system goes to the low power state.
5. from another machine: ether-wake MAC_OF_TARGET_MACHINE
6. The suspended machine wakes.
6.5 ps -ef > after.txt #wait a bit before doing this, everything must
restart
At this point I compared before.txt and after.txt and all of the
processes were accounted for. ethtool showed WOL still enabled, and
/proc/acpi/wakeup showed MMAC still enabled.
Here things go wrong, it turns out that the machine will only go into
the proper suspend state ONCE.
7. (as in (4) above) in xfce4 logout with suspend.
This time the system sort of looks like it is trying to suspend, but it
fails. A bit of text flashes briefly across the screen, which then
blanks, but the fans keep going on the machine, the power light is on
steady, and it is locked. It ignores not only WOL but also the power
button. This looks more or less like what happens if it is ordered to
"hibernate", which enters a similar state even the first time.
Anybody have a guess why it would suspend once, but not then a second
time? The before/after test showed that everything was present, but
perhaps everything was not quite in the proper state?
Thanks,
David Mathog
David Mathog
> 7. (as in (4) above) in xfce4 logout with suspend.
>
> This time the system sort of looks like it is trying to suspend, but it
> fails. A bit of text flashes briefly across the screen, which then
> blanks, but the fans keep going on the machine, the power light is on
> steady, and it is locked. It ignores not only WOL but also the power
> button. This looks more or less like what happens if it is ordered to
> "hibernate", which enters a similar state even the first time.
>
> Anybody have a guess why it would suspend once, but not then a second
> time? The before/after test showed that everything was present, but
> perhaps everything was not quite in the proper state?
1. Hardware failure is ruled out (ASUS A8N5X motherboard). The same
system has Windows XP SP3 on it. That OS can cycle repeatedly
->standby
(power button) recovers
->standby
etc.
2. It isn't the version of pm-utils. pm-utils-1.2.3.tar.gz was
downloaded, built, and installed in /usr/local. After that:
/usr/local/sbin/pm-suspend
suspended properly once, following recovery, it would not suspend
correctly a second time.
3. The issue, whatever it is, is happening after the "quirks" have all
been run through: /var/log/pm-suspend.log shows all the normal steps
down through "done running suspend hooks". After that the system goes
into some odd state that isn't suspend. Power usage is high and the
power light is on steady, the system fans are still spinning, but video
is off (monitor goes to standby). In true suspend the power indicator
flashes, the fans are off, and power usage is low.
4. The loaded modules were compared with "lsmod | sort" before and
after the first suspend/resume cycle. The only difference was the
presence of modules sit and tunnel4 afterwards. However, doing
rmmod sit
rmmod tunnel4
pm-suspend
locked up as before.
5. Unset and reset MMAC enabled in /proc/acpi/wakeup; reset the WOL
with ethtool; service acpid restart. It made no difference.
I am pretty much out of ideas. There seems to be something not quite
"normal" about the system state after it recovers from a suspend, such
that it cannot be suspended, but what?
Thanks,
David Mathog