[OpenIndiana-discuss] Power management in OI

[Milan Jurik]

Hi Jim,

On čt, 2012-10-25 at 17:06 +0400, Jim Klimov wrote:
> Hello all,
> 
>    I'm trying to set up power management on my oi_151a5 laptop.
> For one, I found a useful comment in this blog post:
>  
> http://prefetch.net/blog/index.php/2009/07/12/using-the-cpu-power-management-features-in-solaris/
> 
> ...that cpupm (in /etc/power.conf) must include the "poll-mode"
> keyword:
> 
> cpu-threshold  1s
> cpupm          enable poll-mode
> 
>    At least, with this option set, I see reduced core speeds
> with "kstat -m cpu_info -i 0 -s current_clock_Hz" and "powertop"
> much more often than with just "cpupm enable". Bumping the
> threshold to 3s also helped the CPU stay in reduced states
> much longer, because running GNOME seems to bite off about 3%
> user-time and about 3% kernel-time on this laptop, which
> causes the CPU to wake up often needlessly (IMHO).
> 

yes, poll-mode can be more effective on some systems. But not always
what you see with kstat and powertop is what you have in real usage.
powertop and kstat are a bit heavy tools which have impact on the
reality (system does not decrease CPU speed if you are monitoring it).

>    I am not sure HDD powersaving is viable, given there's 1
> disk in the system and ZFS has something to flush every cycle,
> but there were no reportable problems with setting it up with
> http://constantin.glez.de/blog/2010/03/opensolaris-home-server-scripting-2-setting-power-management
> (however, I did not hear spindown-spinup when I set a 1s timeout).
> 
>    Now, I'm having a problem with suspend: when I do the action,
> the system falls out of X11 into text mode (empty screen with
> the cursor bar) and freezes. The box must be hard-rebooted
> afterwards. Logs include:
> 
> 
> Sep  2 12:14:41 nbofh genunix: [ID 535284 kern.notice] System is being 
> suspended
> Sep  2 12:14:42 nbofh genunix: [ID 122848 kern.warning] WARNING: Unable 
> to suspend device display at 1.
> Sep  2 12:14:43 nbofh genunix: [ID 537702 kern.warning] WARNING: Device 
> is busy or does not support suspend/resume.
> Sep  2 12:14:53 nbofh srn: [ID 980641 kern.warning] WARNING: srn_notify: 
> clone 2 did not ack event a03
> Sep  2 12:14:53 nbofh genunix: [ID 121466 kern.warning] WARNING: 
> audiohd#0: Unable to restore value of control record-source
> Sep  2 12:14:54 nbofh genunix: [ID 583038 kern.notice] System has been 
> resumed.
> Sep  2 12:22:52 nbofh genunix: [ID 576964 kern.notice] ^MOpenIndiana 
> Build oi_151a5 64-bit (illumos 13740:836bfdf31fc4)
> Sep  2 12:22:52 nbofh genunix: [ID 107366 kern.notice] SunOS Release 
> 5.11 - Copyright 1983-2010 Oracle and/or its affiliates.
> ...
> 
> So, apparently, some device - the display (generic VESA VGA on the
> Radeon built into the AMD E2 CPU) or one of the audios (hdmi on the
> radeon or separate realtek chip) fail to suspend/resume and may cause
> the freeze.
> 

No surprise, the key problem is lack of support fo Radeon. System cannot
sleep without implementation of it. Without fixing this it makes no
sense to investigate it more. You need to look at system console,
usually serial port, to see what happend after resume. Also, you can try
to suspend and resume with disabled X server.

> On side notes: DPMI works ok to turn off the lights on the LCD screen.
> Suspend-resume worked OK in Win7, though I couldn't get that to 
> hibernate for no apparent reason.
> 
> Thanks for ideas,
> //Jim Klimov

Best regards,

Milan