/[gentoo]/xml/htdocs/doc/en/power-management-guide.xml
Gentoo

Contents of /xml/htdocs/doc/en/power-management-guide.xml

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.30 - (hide annotations) (download) (as text)
Tue Jul 24 22:32:59 2007 UTC (6 years, 11 months ago) by nightmorph
Branch: MAIN
Changes since 1.29: +4 -7 lines
File MIME type: application/xml
removed outdated hyperlink, bug 186530

1 swift 1.1 <?xml version='1.0' encoding="UTF-8"?>
2     <!DOCTYPE guide SYSTEM "/dtd/guide.dtd">
3 nightmorph 1.30 <!-- $Header: /var/cvsroot/gentoo/xml/htdocs/doc/en/power-management-guide.xml,v 1.29 2007/07/08 17:51:23 nightmorph Exp $ -->
4 so 1.15 <guide link="/doc/en/power-management-guide.xml">
5 swift 1.1 <title>Power Management Guide</title>
6    
7 swift 1.4 <author title="Author">
8 nightmorph 1.17 <mail link="earthwings@gentoo.org">Dennis Nienhüser</mail>
9 swift 1.1 </author>
10 rane 1.19 <author title="Editor">
11     <mail link="chriswhite@gentoo.org">Chris White</mail>
12     </author>
13 nightmorph 1.26 <author title="Editor">
14     <mail link="nightmorph@gentoo.org">Joshua Saddler</mail>
15     </author>
16 swift 1.1
17     <abstract>
18     Power Management is the key to extend battery run time on mobile systems like
19     laptops. This guide assists you setting it up on your laptop.
20     </abstract>
21    
22     <!-- The content of this document is licensed under the CC-BY-SA license -->
23 so 1.15 <!-- See http://creativecommons.org/licenses/by-sa/2.5 -->
24 swift 1.1 <license/>
25    
26 nightmorph 1.30 <version>1.35</version>
27     <date>2007-07-24</date>
28 swift 1.1
29     <chapter>
30     <title>Introduction</title>
31     <section>
32     <body>
33    
34     <p>
35 so 1.15 Capacity and lifetime of laptop batteries have improved much in the last years.
36 swift 1.1 Nevertheless modern processors consume much more energy than older ones and
37     each laptop generation introduces more devices hungry for energy. That's why
38     Power Management is more important than ever. Increasing battery run time
39     doesn't necessarily mean buying another battery. Much can be achieved applying
40     intelligent Power Management policies.
41     </p>
42    
43     </body>
44     </section>
45     <section>
46 rane 1.19 <title>A Quick Overview</title>
47 swift 1.1 <body>
48    
49     <p>
50     Please notice that this guide describes Power Management for <e>laptops</e>.
51     While some sections might also suite for <e>servers</e>, others do not and may
52     even cause harm. Please do not apply anything from this guide to a server
53     unless you really know what you are doing.
54     </p>
55    
56     <p>
57     As this guide has become rather long, here's a short overview helping you to
58     find your way through it.
59     </p>
60    
61     <p>
62 rane 1.19 The <uri link="#doc_chap2">Prerequisites</uri> chapter talks about some
63     requirements that should be met before any of the following device individual
64     sections will work. This includes BIOS settings, kernel configuration and some
65     simplifications in user land. The following three chapters focus on devices
66     that typically consume most energy - processor, display and hard drive. Each
67     can be configured seperately. <uri link="#doc_chap3">CPU Power Management</uri>
68     shows how to adjust the processor's frequency to save a maximum of energy
69 rane 1.21 without losing too much performance. A few different tricks prevent your hard
70 rane 1.19 drive from working unnecessarily often in <uri link="#doc_chap5">Disk Power
71     Management</uri> (decreasing noise level as a nice side effect). Some notes on
72 rane 1.20 graphics cards, Wireless LAN and USB finish the device section in <uri
73     link="#doc_chap6">Power Management For Other Devices</uri> while another
74 rane 1.19 chapter is dedicated to the (rather experimental) <uri link="#doc_chap7">sleep
75 rane 1.20 states</uri>. Last not least <uri link="#doc_chap8">Troubleshooting</uri> lists
76     common pitfalls.
77 swift 1.1 </p>
78    
79     </body>
80     </section>
81     <section>
82 rane 1.19 <title>Power Budget For Each Component</title>
83 swift 1.1 <body>
84    
85 swift 1.2 <figure link="/images/energy-budget.png" short="Which component consumes how
86 swift 1.1 much energy?" caption="Power budget for each component"/>
87    
88     <p>
89     Nearly every component can operate in different states - off, sleep, idle,
90     active to name a few - consuming a different amount of energy. Major parts are
91     consumed by the LCD display, CPU, chipset and hard drives. Often one is able to
92     activate OS-independent Power Management in the BIOS, but an intelligent setup
93     in the operating system adapting to different situations can achieve much more.
94     </p>
95    
96     </body>
97     </section>
98     </chapter>
99    
100     <chapter>
101     <title>Prerequisites</title>
102     <section>
103     <body>
104    
105     <p>
106 nightmorph 1.17 Before discussing the details of making individual devices Power Management
107 rane 1.20 aware, make sure certain requirements are met. After controlling BIOS settings,
108     some kernel options want to be enabled - these are in short ACPI, sleep states
109     and CPU frequency scaling. As power saving most of the time comes along with
110     performance loss or increased latency, it should only be enabled when running
111     on batteries. That's where a new runlevel <e>battery</e> comes in handy.
112 swift 1.1 </p>
113    
114     </body>
115     </section>
116     <section>
117 rane 1.19 <title>The BIOS Part</title>
118 swift 1.1 <body>
119    
120     <p>
121     First have a look into your BIOS Power Management settings. The best way is to
122     combine BIOS and operating system policies, but for the moment it's better to
123     disable most of the BIOS part. This makes sure it doesn't interfere with your
124     policies. Don't forget to re-check BIOS settings after you configured
125     everything else.
126     </p>
127    
128     </body>
129     </section>
130     <section>
131 rane 1.19 <title>Setting USE Flags</title>
132 nightmorph 1.17 <body>
133    
134     <p>
135     Please check that the <c>acpi</c> USE flag is set in
136     <path>/etc/make.conf</path>. Other USE flags that might be interesting for your
137     system are <c>apm</c>, <c>lm_sensors</c>, <c>nforce2</c>, <c>nvidia</c>,
138     <c>pmu</c>. See <path>/usr/portage/profiles/use*.desc</path> for details. If
139     you forgot to set one of these flags, you can recompile affected packages using
140 rane 1.19 the <c>--newuse</c> flag in <c>emerge</c>, see <c>man emerge</c>.
141 nightmorph 1.17 </p>
142    
143     </body>
144     </section>
145     <section>
146 rane 1.19 <title>Configuring The Kernel</title>
147 swift 1.1 <body>
148    
149     <p>
150     ACPI (Advanced Configuration and Power Interface) support in the kernel is
151     still work in progress. Using a recent kernel will make sure you'll get the
152     most out of it.
153     </p>
154    
155     <p>
156 rane 1.20 There are different kernel sources in Portage. I'd recommend using
157 so 1.15 <c>gentoo-sources</c> or <c>suspend2-sources</c>. The latter contains patches
158 rane 1.19 for Software Suspend 2, see the chapter about <uri link="#doc_chap7">sleep
159     states</uri> for more details. When configuring the kernel, activate at least
160     these options:
161 swift 1.1 </p>
162    
163     <pre caption="Minimum kernel setup for Power Management (Kernel 2.6)">
164     Power Management Options ---&gt;
165     [*] Power Management Support
166     [ ] Software Suspend
167    
168     ACPI( Advanced Configuration and Power Interface ) Support ---&gt;
169     [*] ACPI Support
170     [ ] Sleep States
171 so 1.15 [ ] /proc/acpi/sleep (deprecated)
172 neysx 1.12 [*] AC Adapter
173     [*] Battery
174 swift 1.1 &lt;M&gt; Button
175 so 1.15 &lt;M&gt; Video
176     [ ] Generic Hotkey
177 swift 1.1 &lt;M&gt; Fan
178     &lt;M&gt; Processor
179     &lt;M&gt; Thermal Zone
180     &lt; &gt; ASUS/Medion Laptop Extras
181 so 1.15 &lt; &gt; IBM ThinkPad Laptop Extras
182 swift 1.1 &lt; &gt; Toshiba Laptop Extras
183 so 1.15 (0) Disable ACPI for systems before Jan 1st this year
184 swift 1.1 [ ] Debug Statements
185 so 1.15 [*] Power Management Timer Support
186     &lt; &gt; ACPI0004,PNP0A05 and PNP0A06 Container Driver (EXPERIMENTAL)
187 rane 1.20
188 swift 1.1 CPU Frequency Scaling ---&gt;
189     [*] CPU Frequency scaling
190 so 1.15 [ ] Enable CPUfreq debugging
191     &lt; &gt; CPU frequency translation statistics
192     [ ] CPU frequency translation statistics details
193 swift 1.1 Default CPUFreq governor (userspace)
194     &lt;*&gt; 'performance' governor
195     &lt;*&gt; 'powersave' governor
196 swift 1.8 &lt;*&gt; 'ondemand' cpufreq policy governor
197 so 1.15 &lt;*&gt; 'conservative' cpufreq governor
198 swift 1.1 &lt;*&gt; CPU frequency table helpers
199     &lt;M&gt; ACPI Processor P-States driver
200     &lt;*&gt; <i>CPUFreq driver for your processor</i>
201     </pre>
202    
203     <p>
204 so 1.15 Decide yourself whether you want to enable Software Suspend, and Sleep States
205     (see below). If you own an ASUS, Medion, IBM Thinkpad or Toshiba laptop, enable
206 neysx 1.12 the appropriate section.
207 swift 1.8 </p>
208    
209     <p>
210 rane 1.20 The kernel has to know how to enable CPU frequency scaling on your processor.
211     As each type of CPU has a different interface, you've got to choose the right
212 rane 1.19 driver for your processor. Be careful here - enabling <c>Intel Pentium 4 clock
213     modulation</c> on a Pentium M system will lead to strange results for example.
214 swift 1.8 Consult the kernel documentation if you're unsure which one to take.
215 swift 1.1 </p>
216    
217     <p>
218     Compile your kernel, make sure the right modules get loaded at startup and boot
219 rane 1.20 into your new ACPI-enabled kernel. Next run <c>emerge sys-power/acpid</c> to
220     get the acpi daemon. This one informs you about events like switching from AC
221     to battery or closing the lid. Make sure the modules are loaded if you didn't
222     compile them into the kernel and start acpid by executing <c>/etc/init.d/acpid
223     start</c>. Run <c>rc-update add acpid default</c> to load it on startup. You'll
224     soon see how to use it.
225 swift 1.1 </p>
226    
227     <pre caption="Installing acpid">
228 yoswink 1.10 # <i>emerge sys-power/acpid</i>
229 swift 1.1 # <i>/etc/init.d/acpid start</i>
230     # <i>rc-update add acpid default</i>
231     </pre>
232    
233     </body>
234     </section>
235     <section>
236 rane 1.19 <title>Creating A "battery" Runlevel</title>
237 swift 1.1 <body>
238    
239     <p>
240     The default policy will be to enable Power Management only when needed -
241     running on batteries. To make the switch between AC and battery convenient,
242 rane 1.19 create a runlevel <c>battery</c> that holds all the scripts starting and
243 swift 1.1 stopping Power Management.
244     </p>
245    
246     <note>
247     You can safely skip this section if you don't like the idea of having another
248     runlevel. However, skipping this step will make the rest a bit trickier to set
249 rane 1.19 up. The next sections assume a runlevel <c>battery</c> exists.
250 swift 1.1 </note>
251    
252     <pre caption="Creating a battery runlevel">
253     # <i>cd /etc/runlevels</i>
254     # <i>cp -a default battery</i>
255     </pre>
256    
257     <p>
258 rane 1.19 Finished. Your new runlevel <c>battery</c> contains everything like
259     <c>default</c>, but there is no automatic switch between both yet. Time to
260 swift 1.1 change it.
261     </p>
262    
263     </body>
264     </section>
265     <section>
266 rane 1.19 <title>Reacting On ACPI Events</title>
267 swift 1.1 <body>
268    
269     <p>
270     Typical ACPI events are closing the lid, changing the power source or pressing
271 swift 1.8 the sleep button. An important event is changing the power source, which should
272 so 1.15 cause a runlevel switch. A small script will take care of it.
273     </p>
274    
275     <p>
276     First you need a script which changes the runlevel to <c>default</c>
277     respectively <c>battery</c> depending on the power source. The script uses the
278     <c>on_ac_power</c> command from <c>sys-power/powermgmt-base</c> - make sure the
279     package is installed on your system.
280     </p>
281    
282     <pre caption="Installing powermgt-base">
283 rane 1.19 # <i>emerge powermgmt-base</i>
284 so 1.15 </pre>
285    
286     <p>
287 rane 1.20 You are now able to determine the power source by executing <c>on_ac_power
288     &amp;&amp; echo AC available || echo Running on batteries</c> in a shell. The
289     script below is responsible for changing runlevels. Save it as
290 so 1.15 <path>/etc/acpi/actions/pmg_switch_runlevel.sh</path>.
291 swift 1.8 </p>
292    
293 neysx 1.12 <pre caption="/etc/acpi/actions/pmg_switch_runlevel.sh">
294 swift 1.8 #!/bin/bash
295    
296 neysx 1.12 <comment># BEGIN configuration</comment>
297 swift 1.8 RUNLEVEL_AC="default"
298     RUNLEVEL_BATTERY="battery"
299 neysx 1.12 <comment># END configuration</comment>
300 swift 1.8
301 swift 1.1
302 neysx 1.12 if [ ! -d "/etc/runlevels/${RUNLEVEL_AC}" ]
303     then
304 nightmorph 1.17 logger "${0}: Runlevel ${RUNLEVEL_AC} does not exist. Aborting."
305     exit 1
306 neysx 1.12 fi
307 swift 1.8
308 neysx 1.12 if [ ! -d "/etc/runlevels/${RUNLEVEL_BATTERY}" ]
309     then
310 nightmorph 1.17 logger "${0}: Runlevel ${RUNLEVEL_BATTERY} does not exist. Aborting."
311     exit 1
312 neysx 1.12 fi
313 swift 1.1
314 neysx 1.12 if on_ac_power
315     then
316 rane 1.19 if [[ "$(&lt;/var/lib/init.d/softlevel)" != "${RUNLEVEL_AC}" ]]
317 nightmorph 1.17 then
318     logger "Switching to ${RUNLEVEL_AC} runlevel"
319     /sbin/rc ${RUNLEVEL_AC}
320     fi
321 rane 1.19 elif [[ "$(&lt;/var/lib/init.d/softlevel)" != "${RUNLEVEL_BATTERY}" ]]
322 neysx 1.12 then
323 nightmorph 1.17 logger "Switching to ${RUNLEVEL_BATTERY} runlevel"
324     /sbin/rc ${RUNLEVEL_BATTERY}
325 neysx 1.12 fi
326 swift 1.8 </pre>
327    
328 so 1.15 <p>
329     Dont forget to run <c>chmod +x /etc/acpi/actions/pmg_switch_runlevel.sh</c> to
330     make the script executable. The last thing that needs to be done is calling the
331     script whenever the power source changes. That's done by catching ACPI events
332     with the help of <c>acpid</c>. First you need to know which events are
333     generated when the power source changes. The events are called
334 rane 1.19 <c>ac_adapter</c> and <c>battery</c> on most laptops, but it might be different
335 so 1.15 on yours.
336     </p>
337    
338     <pre caption="Determining ACPI events for changing the power source">
339 rane 1.19 # <i>tail -f /var/log/acpid | grep "received event"</i>
340 so 1.15 </pre>
341    
342     <p>
343 rane 1.20 Run the command above and pull the power cable. You should see something like
344     this:
345 so 1.15 </p>
346    
347     <pre caption="Sample output for power source changes">
348     [Tue Sep 20 17:39:06 2005] received event "ac_adapter AC 00000080 00000000"
349     [Tue Sep 20 17:39:06 2005] received event "battery BAT0 00000080 00000001"
350     </pre>
351    
352     <p>
353 rane 1.19 The interesting part is the quoted string after <c>received event</c>. It will
354 so 1.15 be matched by the event line in the files you are going to create below. Don't
355     worry if your system generates multiple events or always the same. As long as
356     any event is generated, runlevel changing will work.
357     </p>
358    
359 swift 1.8 <pre caption="/etc/acpi/events/pmg_ac_adapter">
360 neysx 1.12 <comment># replace "ac_adapter" below with the event generated on your laptop</comment>
361 so 1.15 <comment># For example, ac_adapter.* will match ac_adapter AC 00000080 00000000</comment>
362 swift 1.8 event=ac_adapter.*
363 neysx 1.12 action=/etc/acpi/actions/pmg_switch_runlevel.sh %e
364 swift 1.8 </pre>
365    
366     <pre caption="/etc/acpi/events/pmg_battery">
367 neysx 1.12 <comment># replace "battery" below with the event generated on your laptop</comment>
368 so 1.15 <comment># For example, battery.* will match battery BAT0 00000080 00000001</comment>
369 swift 1.8 event=battery.*
370 neysx 1.12 action=/etc/acpi/actions/pmg_switch_runlevel.sh %e
371 swift 1.8 </pre>
372 swift 1.1
373 swift 1.8 <p>
374 so 1.15 Finally acpid has to be restarted to recognize the changes.
375 swift 1.8 </p>
376 swift 1.1
377 swift 1.8 <pre caption="Finishing runlevel switching with acpid">
378 rane 1.19 # <i>/etc/init.d/acpid restart</i>
379 swift 1.1 </pre>
380    
381     <p>
382     Give it a try: Plug AC in and out and watch syslog for the "Switching to AC
383 rane 1.20 mode" or "Switching to battery mode" messages. See the <uri
384     link="#doc_chap8">Troubleshooting section</uri> if the script is not able to
385     detect the power source correctly.
386 swift 1.1 </p>
387    
388     <p>
389     Due to the nature of the event mechanism, your laptop will boot into runlevel
390 rane 1.19 <c>default</c> regardless of the AC/battery state. This is fine when running
391 rane 1.20 from AC, but we'd like to boot into the battery runlevel otherwise. One
392     solution would be to add another entry to the boot loader with the parameter
393 so 1.15 <c>softlevel=battery</c>, but it's likely to forget choosing it. A better way
394 rane 1.20 is faking an ACPI event in the end of the boot process and letting
395     <path>pmg_switch_runlevel.sh</path> script decide whether a runlevel change is
396     necessary. Open <path>/etc/conf.d/local.start</path> in your favourite editor
397     and add these lines:
398 swift 1.1 </p>
399    
400 so 1.15 <pre caption="Runlevel adjustment at boot time by editing local.start">
401 swift 1.1 <comment># Fake acpi event to switch runlevel if running on batteries</comment>
402 neysx 1.12 /etc/acpi/actions/pmg_switch_runlevel.sh "battery/battery"
403 swift 1.1 </pre>
404    
405     <p>
406 rane 1.20 Prepared like this you can activate Power Management policies for individual
407 swift 1.1 devices.
408     </p>
409    
410     </body>
411     </section>
412     </chapter>
413    
414     <chapter>
415     <title>CPU Power Management</title>
416     <section>
417 nightmorph 1.17 <body>
418    
419     <p>
420     Mobile processors can operate at different frequencies. Some allow changing
421     voltage as well. Most of the time your CPU doesn't need to run at full speed
422     and scaling it down will save much energy - often without any performance
423     decrease.
424     </p>
425    
426     </body>
427     </section>
428     <section>
429 rane 1.19 <title>Some Technical Terms</title>
430 swift 1.8 <body>
431    
432     <p>
433     CPU frequency scaling brings up some technical terms that might be unknown to
434     you. Here's a quick introduction.
435     </p>
436    
437     <p>
438 so 1.15 First of all, the kernel has to be able to change the processor's frequency.
439 rane 1.19 The <b>CPUfreq processor driver</b> knows the commands to do it on your CPU.
440 rane 1.20 Thus it's important to choose the right one in your kernel. You should already
441     have done it above. Once the kernel knows how to change frequencies, it has to
442     know which frequency it should set. This is done according to the <b>policy</b>
443     which consists of a <b>CPUfreq policy</b> and a <b>governor</b>. A CPUfreq
444     policy are just two numbers which define a range the frequency has to stay
445     between - minimal and maximal frequency. The governor now decides which of the
446     available frequencies in between minimal and maximal frequency to choose. For
447     example, the <b>powersave governor</b> always chooses the lowest frequency
448     available, the <b>performance governor</b> the highest one. The <b>userspace
449     governor</b> makes no decision but chooses whatever the user (or a program in
450     userspace) wants - which means it reads the frequency from
451 swift 1.8 <path>/sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed</path>.
452     </p>
453    
454     <p>
455     This doesn't sound like dynamic frequency changes yet and in fact it isn't.
456 rane 1.20 Dynamics however can be accomplished with various approaches. For example, the
457     <b>ondemand governor</b> makes its decisions depending on the current CPU load.
458     The same is done by various userland tools like <c>cpudyn</c>, <c>cpufreqd</c>,
459     <c>powernowd</c> and many more. ACPI events can be used to enable or disable
460     dynamic frequency changes depending on power source.
461 swift 1.8 </p>
462    
463     </body>
464     </section>
465     <section>
466 rane 1.19 <title>Setting The Frequency Manually</title>
467 swift 1.1 <body>
468    
469     <p>
470     Decreasing CPU speed and voltage has two advantages: On the one hand less
471     energy is consumed, on the other hand there is thermal improvement as your
472     system doesn't get as hot as running on full speed. The main disadvantage is
473     obviously the loss of performance. Decreasing processor speed is a trade off
474     between performance loss and energy saving.
475     </p>
476    
477     <note>
478     Not every laptop supports frequency scaling. If unsure, have a look at the list
479 rane 1.19 of supported processors in the <uri link="#doc_chap8">Troubleshooting</uri>
480     section to verify yours is supported.
481 swift 1.1 </note>
482    
483     <p>
484 neysx 1.12 It's time to test whether CPU frequency changing works. Let's install another
485     tool which is very handy for debugging purposes: <c>sys-power/cpufrequtils</c>
486 swift 1.1 </p>
487    
488 neysx 1.12 <pre caption="Checking CPU frequency">
489     # <i>emerge cpufrequtils</i>
490     # <i>cpufreq-info</i>
491 swift 1.1 </pre>
492    
493     <p>
494 neysx 1.12 Here is an example output:
495 swift 1.1 </p>
496    
497 neysx 1.12 <pre caption="Sample output from cpufreq-info">
498 so 1.15 cpufrequtils 0.3: cpufreq-info (C) Dominik Brodowski 2004
499 neysx 1.12 Report errors and bugs to linux@brodo.de, please.
500     analyzing CPU 0:
501     driver: centrino
502     CPUs which need to switch frequency at the same time: 0
503     hardware limits: 600 MHz - 1.40 GHz
504     available frequency steps: 600 MHz, 800 MHz, 1000 MHz, 1.20 GHz, 1.40 GHz
505 so 1.15 available cpufreq governors: conservative, ondemand, powersave, userspace, performance
506 neysx 1.12 current policy: frequency should be within 924 MHz and 1.40 GHz.
507     The governor "performance" may decide which speed to use
508     within this range.
509 so 1.15 current CPU frequency is 1.40 GHz.
510 neysx 1.12 </pre>
511    
512 swift 1.1 <p>
513 neysx 1.12 Now play around with <c>cpufreq-set</c> to make sure frequency switching works.
514     Run <c>cpufreq-set -g ondemand</c> for example to activate the ondemand
515     governor and verify the change with <c>cpufreq-info</c>. If it doesn't work as
516 rane 1.20 expected, you might find help in the <uri link="#doc_chap8">Troubleshooting
517     section</uri> in the end of this guide.
518 swift 1.1 </p>
519    
520     </body>
521     </section>
522     <section>
523     <title>Automated frequency adaption</title>
524     <body>
525    
526     <p>
527     The above is quite nice, but not doable in daily life. Better let your system
528 so 1.15 set the appropriate frequency automatically. There are many different
529     approaches to do this. The following table gives a quick overview to help you
530 rane 1.21 decide on one of them. It's roughly separated in three categories <b>kernel</b>
531 rane 1.20 for approaches that only need kernel support, <b>daemon</b> for programs that
532     run in the background and <b>graphical</b> for programs that provide a GUI for
533     easy configuration and changes.
534 swift 1.1 </p>
535    
536     <table>
537     <tr>
538     <th>Name</th>
539 swift 1.8 <th>Category</th>
540     <th>Switch decision</th>
541     <th>Kernel governors</th>
542     <th>Further governors</th>
543     <th>Comments</th>
544     </tr>
545     <tr>
546     <ti>'ondemand' governor</ti>
547     <ti>Kernel</ti>
548     <ti>CPU load</ti>
549     <ti>N.A.</ti>
550     <ti>N.A.</ti>
551     <ti>
552 so 1.15 Chooses maximal frequency on CPU load and slowly steps down when the CPU is
553     idle. Further tuning through files in
554 swift 1.8 <path>/sys/devices/system/cpu/cpu0/cpufreq/ondemand/</path>. Still requires
555     userland tools (programs, scripts) if governor switching or similar is
556     desired.
557     </ti>
558 swift 1.1 </tr>
559     <tr>
560 so 1.15 <ti>'conservative' governor</ti>
561     <ti>Kernel</ti>
562     <ti>CPU load</ti>
563     <ti>N.A.</ti>
564     <ti>N.A.</ti>
565     <ti>
566 neysx 1.16 Unlike the ondemand governor, conversative doesn't jump to maximum
567     frequency when CPU load is high, but increases the frequency step by step.
568     Further tuning through files in
569     <path>/sys/devices/system/cpu/cpu0/cpufreq/ondemand/</path>. Still requires
570     userland tools (programs, scripts) if governor switching or similar is
571     desired.
572 so 1.15 </ti>
573     </tr>
574     <tr>
575 swift 1.1 <ti><uri link="http://mnm.uib.es/~gallir/cpudyn/">cpudyn</uri></ti>
576 swift 1.8 <ti>Daemon</ti>
577     <ti>CPU load</ti>
578 neysx 1.12 <ti>Performance, powersave</ti>
579 swift 1.8 <ti>Dynamic</ti>
580     <ti>
581     Also supports disk standby - notice however that <e>laptop mode</e> in most
582     cases will do a better job.
583     </ti>
584 swift 1.1 </tr>
585     <tr>
586 swift 1.8 <ti><uri link="http://sourceforge.net/projects/cpufreqd/">cpufreqd</uri></ti>
587     <ti>Daemon</ti>
588 nightmorph 1.17 <ti>Battery state, CPU load, temperature, running programs and more</ti>
589 swift 1.8 <ti>All available</ti>
590     <ti>None</ti>
591     <ti>
592 nightmorph 1.17 Sophisticated (but somewhat complicated) setup. Extendible through plugins
593     like sensor monitoring (lm_sensors) or coordinating some NVidia based
594     graphics card memory and core. Cpufreqd is SMP aware and can optionally be
595     controlled manually at runtime.
596 swift 1.8 </ti>
597     </tr>
598     <tr>
599     <ti>
600     <uri link="http://www.deater.net/john/powernowd.html">powernowd</uri>
601     </ti>
602     <ti>Daemon</ti>
603     <ti>CPU load</ti>
604     <ti>None</ti>
605     <ti>Passive, sine, aggressive</ti>
606     <ti>
607     Supports SMP.
608     </ti>
609 swift 1.1 </tr>
610     <tr>
611 nightmorph 1.30 <ti>ncpufreqd</ti>
612 so 1.15 <ti>Daemon</ti>
613     <ti>Temperature</ti>
614     <ti>None</ti>
615     <ti>Powersave, performance</ti>
616     <ti>
617     Toggles the used governor between performance and powersave depending on
618     system temperature. Very useful on laptops with notorious heat problems.
619     </ti>
620     </tr>
621     <tr>
622 swift 1.1 <ti><uri link="http://www.goop.org/~jeremy/speedfreq/">speedfreq</uri></ti>
623 swift 1.8 <ti>Daemon</ti>
624     <ti>CPU load</ti>
625     <ti>None</ti>
626     <ti>Dynamic, powersave, performance, fixed speed</ti>
627     <ti>
628 so 1.15 Easy to configure with a nice client/server interface. Requires a 2.6
629     kernel. Unmaintained, broken and thus removed from Portage. Please switch
630     to cpufreqd if you're still using it.
631 swift 1.8 </ti>
632     </tr>
633     <tr>
634     <ti><uri link="http://cpuspeedy.sourceforge.net/">gtk-cpuspeedy</uri></ti>
635     <ti>Graphical</ti>
636     <ti>None</ti>
637     <ti>None</ti>
638     <ti>None</ti>
639 swift 1.1 <ti>
640 swift 1.8 Gnome application, a graphical tool to set CPU frequency manually. It does
641 neysx 1.12 not offer any automation.
642 swift 1.1 </ti>
643     </tr>
644     <tr>
645 swift 1.8 <ti>klaptopdaemon</ti>
646     <ti>Graphical</ti>
647     <ti>Battery state</ti>
648     <ti>All available</ti>
649     <ti>None</ti>
650     <ti>
651     KDE only, 'ondemand' governor required for dynamic frequency scaling.
652     </ti>
653 swift 1.1 </tr>
654     </table>
655    
656     <p>
657 so 1.15 While adjusting the frequency to the current load looks simple at a first
658     glance, it's not such a trivial task. A bad algorithm can cause switching
659     between two frequencies all the time or wasting energy when setting frequency
660 rane 1.20 to an unnecessary high level.
661 swift 1.1 </p>
662    
663     <p>
664 neysx 1.12 Which one to choose? If you have no idea about it, try <c>cpufreqd</c>:
665 swift 1.1 </p>
666    
667     <pre caption="Installing cpufreqd">
668     # <i>emerge cpufreqd</i>
669     </pre>
670    
671     <p>
672 neysx 1.12 <c>cpufreqd</c> can be configured by editing <path>/etc/cpufreqd.conf</path>.
673     The default one that ships with cpufreqd may look a bit confusing. I recommend
674 nightmorph 1.22 replacing it with the one from former Gentoo developer Henrik Brix Andersen
675     (see below). Please notice that you need cpufreqd-2.0.0 or later. Earlier
676     versions have a different syntax for the config file.
677 swift 1.1 </p>
678    
679 nightmorph 1.17 <pre caption="/etc/cpufreqd.conf (cpufreqd-2.0.0 and later)">
680 swift 1.1 [General]
681     pidfile=/var/run/cpufreqd.pid
682 nightmorph 1.17 poll_interval=3
683     enable_plugins=acpi_ac, acpi_battery
684 nightmorph 1.18 enable_remote=1
685     remote_group=wheel
686 neysx 1.12 verbosity=5
687 nightmorph 1.17 [/General]
688 swift 1.1
689     [Profile]
690 neysx 1.12 name=ondemand
691     minfreq=0%
692     maxfreq=100%
693     policy=ondemand
694 nightmorph 1.17 [/Profile]
695 swift 1.1
696     [Profile]
697 so 1.15 name=conservative
698     minfreq=0%
699     maxfreq=100%
700     policy=conservative
701 nightmorph 1.17 [/Profile]
702 so 1.15
703     [Profile]
704 neysx 1.12 name=powersave
705     minfreq=0%
706     maxfreq=100%
707 swift 1.1 policy=powersave
708 nightmorph 1.17 [/Profile]
709 swift 1.1
710     [Profile]
711 neysx 1.12 name=performance
712     minfreq=0%
713     maxfreq=100%
714     policy=performance
715 nightmorph 1.17 [/Profile]
716 swift 1.1
717     [Rule]
718 neysx 1.12 name=battery
719     ac=off
720 so 1.15 profile=conservative
721 nightmorph 1.17 [/Rule]
722 swift 1.1
723     [Rule]
724 neysx 1.12 name=battery_low
725 swift 1.1 ac=off
726 neysx 1.12 battery_interval=0-10
727     profile=powersave
728 nightmorph 1.17 [/Rule]
729 swift 1.1
730     [Rule]
731 neysx 1.12 name=ac
732     ac=on
733 so 1.15 profile=ondemand
734 nightmorph 1.17 [/Rule]
735 neysx 1.12 </pre>
736    
737     <p>
738 rane 1.19 Now you can start the cpufreqd daemon. Add it to the <c>default</c> and
739     <c>battery</c> runlevel as well.
740 swift 1.1 </p>
741    
742 neysx 1.12 <pre caption="Starting cpufreqd">
743     # <i>rc-update add cpufreqd default battery</i>
744     # <i>rc</i>
745     </pre>
746    
747 nightmorph 1.17 <p>
748     Sometimes it can be desirable to select another policy than the daemon chooses,
749     for example when battery power is low, but you know that AC will be available
750 rane 1.20 soon. In that case you can turn on cpufreqd's manual mode with <c>cpufreqd-set
751     manual</c> and select one of your configured policies (as listed by
752     <c>cpufreqd-get</c>). You can leave manual mode by executing <c>cpufreqd-set
753     dynamic</c>.
754 nightmorph 1.17 </p>
755    
756 neysx 1.12 <warn>
757     Do not run more than one of the above programs at the same time. It may cause
758     confusion like switching between two frequencies all the time.
759     </warn>
760    
761 swift 1.8 </body>
762     </section>
763     <section>
764     <title>Verifying the result</title>
765     <body>
766    
767 swift 1.1 <p>
768     The last thing to check is that your new policies do a good job. An easy way to
769 swift 1.8 do so is monitoring CPU speed while working with your laptop:
770 swift 1.1 </p>
771    
772     <pre caption="Monitoring CPU speed">
773 yoswink 1.13 # <i>watch grep \"cpu MHz\" /proc/cpuinfo</i>
774 swift 1.1 </pre>
775    
776     <p>
777 rane 1.20 If <path>/proc/cpuinfo</path> doesn't get updated (see <uri
778     link="#doc_chap8">Troubleshooting</uri>), monitor the CPU frequency with:
779 swift 1.1 </p>
780    
781     <pre caption="Alternative CPU speed monitoring">
782 neysx 1.12 # <i>watch x86info -mhz</i>
783 swift 1.1 </pre>
784    
785     <p>
786     Depending on your setup, CPU speed should increase on heavy load, decrease on
787 rane 1.19 no activity or just stay at the same level. When using <c>cpufreqd</c> and
788 rane 1.20 verbosity set to 5 or higher in <path>cpufreqd.conf</path> you'll get
789     additional information about what's happening reported to <c>syslog</c>.
790 swift 1.1 </p>
791    
792     </body>
793     </section>
794     </chapter>
795    
796     <chapter>
797     <title>LCD Power Management</title>
798     <section>
799     <body>
800    
801     <p>
802 rane 1.20 As you can see in <uri link="#doc_chap1_fig1">figure 1.1</uri>, the LCD
803     display consumes the biggest part of energy (might not be the case for
804     non-mobile CPU's). Thus it's quite important not only to shut the display off
805     when not needed, but also to reduce it's backlight if possible. Most laptops
806     offer the possibility to control the backlight dimming.
807 swift 1.1 </p>
808    
809 nightmorph 1.17 </body>
810     </section>
811     <section>
812     <title>Standby settings</title>
813     <body>
814    
815 swift 1.1 <p>
816 nightmorph 1.17 The first thing to check is the standby/suspend/off timings of the display. As
817     this depends heavily on your windowmanager, I'll let you figure it out
818 rane 1.20 yourself. Just two common places: Blanking the terminal can be done with
819 nightmorph 1.17 <c>setterm -blank &lt;number-of-minutesM&gt;</c>, <c>setterm -powersave on</c>
820     and <c>setterm -powerdown &lt;number-of-minutesM&gt;</c>. For X.org, modify
821     <path>/etc/X11/xorg.conf</path> similar to this:
822 swift 1.1 </p>
823    
824 so 1.15 <pre caption="LCD suspend settings in X.org and XFree86">
825 swift 1.1 Section "ServerLayout"
826     Identifier [...]
827     [...]
828     Option "BlankTime" "5" <comment># Blank the screen after 5 minutes (Fake)</comment>
829     Option "StandbyTime" "10" <comment># Turn off screen after 10 minutes (DPMS)</comment>
830     Option "SuspendTime" "20" <comment># Full suspend after 20 minutes</comment>
831     Option "OffTime" "30" <comment># Turn off after half an hour</comment>
832     [...]
833     EndSection
834    
835     [...]
836    
837     Section "Monitor"
838     Identifier [...]
839     Option "DPMS" "true"
840     [...]
841     EndSection
842     </pre>
843    
844     <p>
845     This is the same for XFree86 and <path>/etc/X11/XF86Config</path>.
846     </p>
847    
848 nightmorph 1.17 </body>
849     </section>
850     <section>
851     <title>Backlight dimming</title>
852     <body>
853    
854 swift 1.1 <p>
855     Probably more important is the backlight dimming. If you have access to the
856     dimming settings via a tool, write a small script that dims the backlight in
857 rane 1.19 battery mode and place it in your <c>battery</c> runlevel. The following script
858 nightmorph 1.17 should work on most IBM Thinkpads and Toshiba laptops. You've got to enable the
859 rane 1.20 appropriate option in your kernel (IBM Thinkpads only). For Toshiba laptops,
860 nightmorph 1.23 install <c>sys-power/acpitool</c> and skip configuration of <c>ibm_acpi</c> as
861 rane 1.20 described below.
862 neysx 1.12 </p>
863    
864     <warn>
865     Support for setting brightness is marked experimental in ibm-acpi. It accesses
866     hardware directly and may cause severe harm to your system. Please read the
867     <uri link="http://ibm-acpi.sourceforge.net/">ibm-acpi website</uri>
868     </warn>
869    
870     <p>
871     To be able to set the brightness level, the ibm_acpi module has to be loaded
872     with the experimental parameter.
873     </p>
874    
875     <pre caption="automatically loading the ibm_acpi module">
876     <comment>(Please read the warnings above before doing this!)</comment>
877 rane 1.19 # <i>echo "options ibm_acpi experimental=1" >> /etc/modules.d/ibm_acpi</i>
878 nightmorph 1.26 # <i>/sbin/update-modules</i>
879 rane 1.19 # <i>echo ibm_acpi >> /etc/modules.autoload.d/kernel-2.6</i>
880     # <i>modprobe ibm_acpi</i>
881 neysx 1.12 </pre>
882    
883     <p>
884     This should work without error messages and a file
885     <path>/proc/acpi/ibm/brightness</path> should be created after loading the
886 rane 1.20 module. An init script will take care of choosing the brightness according to
887     the power source.
888 neysx 1.12 </p>
889    
890     <pre caption="/etc/conf.d/lcd-brightness">
891     <comment># See /proc/acpi/ibm/brightness for available values</comment>
892 nightmorph 1.17 <comment># Please read /usr/src/linux/Documentation/ibm-acpi.txt</comment>
893 neysx 1.12
894 nightmorph 1.23 <comment># brightness level in ac mode. Default is 7.</comment>
895 neysx 1.12 BRIGHTNESS_AC=7
896    
897     <comment># brightness level in battery mode. Default is 4.</comment>
898     BRIGHTNESS_BATTERY=4
899     </pre>
900    
901     <pre caption="/etc/init.d/lcd-brightness">
902     #!/sbin/runscript
903    
904     set_brightness() {
905     if on_ac_power
906     then
907     LEVEL=${BRIGHTNESS_AC:-7}
908     else
909     LEVEL=${BRIGHTNESS_BATTERY:-4}
910     fi
911    
912     if [ -f /proc/acpi/ibm/brightness ]
913     then
914     ebegin "Setting LCD brightness"
915     echo "level ${LEVEL}" > /proc/acpi/ibm/brightness
916     eend $?
917 nightmorph 1.17 elif [[ -e /usr/bin/acpitool &amp;&amp; -n $(acpitool -T | grep "LCD brightness") ]]
918     then
919     ebegin "Setting LCD brightness"
920     acpitool -l $LEVEL >/dev/null || ewarn "Unable to set lcd brightness"
921     eend $?
922 neysx 1.12 else
923     ewarn "Setting LCD brightness is not supported."
924 nightmorph 1.17 ewarn "For IBM Thinkpads, check that ibm_acpi is loaded into the kernel"
925 nightmorph 1.23 ewarn "For Toshiba laptops, you've got to install sys-power/acpitool"
926 neysx 1.12 fi
927     }
928    
929     start() {
930     set_brightness
931     }
932    
933     stop () {
934     set_brightness
935     }
936     </pre>
937    
938     <p>
939     When done, make sure brightness is adjusted automatically by adding it to the
940     battery runlevel.
941 swift 1.1 </p>
942    
943 neysx 1.12 <pre caption="Enabling automatic brightness adjustment">
944 rane 1.19 # <i>chmod +x /etc/init.d/lcd-brightness</i>
945     # <i>rc-update add lcd-brightness battery</i>
946     # <i>rc</i>
947 neysx 1.12 </pre>
948    
949 swift 1.1 </body>
950     </section>
951     </chapter>
952    
953     <chapter>
954     <title>Disk Power Management</title>
955     <section>
956 nightmorph 1.17 <body>
957 rane 1.19
958 nightmorph 1.17 <p>
959     Hard disks consume less energy in sleep mode. Therefore it makes sense to
960     activate power saving features whenever the hard disk is not used for a certain
961     amount of time. I'll show you two alternative possibilities to do it. First,
962     laptop-mode will save most energy due to several measures which prevent or at
963     least delay write accesses. The drawback is that due to the delayed write
964     accesses a power outage or kernel crash will be more dangerous for data loss.
965     If you don't like this, you have to make sure that there are no processes which
966     write to your hard disk frequently. Afterwards you can enable power saving
967 rane 1.19 features of your hard disk with <c>hdparm</c> as the second alternative.
968 nightmorph 1.17 </p>
969    
970     </body>
971     </section>
972     <section>
973     <title>Increasing idle time - laptop-mode</title>
974 swift 1.1 <body>
975    
976     <p>
977 nightmorph 1.27 Recent 2.6 kernels include the so-called <c>laptop-mode</c>. When activated,
978     dirty buffers are written to disk on read calls or after 10 minutes (instead of
979     30 seconds). This minimizes the time the hard disk needs to be spun up.
980 swift 1.1 </p>
981    
982 nightmorph 1.17 <pre caption="Automated start of laptop-mode">
983     # <i>emerge laptop-mode-tools</i>
984     </pre>
985 swift 1.1
986 nightmorph 1.17 <p>
987     <c>laptop-mode-tools</c> has its configuration file in
988     <path>/etc/laptop-mode/laptop-mode.conf</path>. Adjust it the way you like it,
989     it's well commented. Run <c>rc-update add laptop_mode battery</c> to start it
990     automatically.
991     </p>
992    
993     <p>
994     Recent versions (1.11 and later) of laptop-mode-tools include a new tool
995     <c>lm-profiler</c>. It will monitor your system's disk usage and running
996     network services and suggests to disable unneeded ones. You can either disable
997     them through laptop-mode-tools builtin runlevel support (which will be reverted
998 rane 1.19 by Gentoo's <c>/sbin/rc</c>) or use your <c>default</c>/<c>battery</c>
999 nightmorph 1.17 runlevels (recommended).
1000     </p>
1001    
1002     <pre caption="Sample output from running lm-profiler">
1003 rane 1.19 # <i>lm-profiler</i>
1004 nightmorph 1.17 Profiling session started.
1005     Time remaining: 600 seconds
1006     [4296896.602000] amarokapp
1007     Time remaining: 599 seconds
1008     [4296897.714000] sort
1009     [4296897.970000] mv
1010     Time remaining: 598 seconds
1011     Time remaining: 597 seconds
1012     [4296900.482000] reiserfs/0
1013     </pre>
1014    
1015     <p>
1016     After profiling your system for ten minutes, lm-profiler will present a list of
1017     services which might have caused disk accesses during that time.
1018     </p>
1019    
1020     <pre caption="lm-profiler suggests to disable some services">
1021     Program: "atd"
1022     Reason: standard recommendation (program may not be running)
1023     Init script: /etc/init.d/atd (GUESSED)
1024    
1025 rane 1.19 Do you want to disable this service in battery mode? [y/N]: <i>n</i>
1026 swift 1.1 </pre>
1027    
1028     <p>
1029 nightmorph 1.17 To disable atd as suggested in the example above, you would run <c>rc-update
1030     del atd battery</c>. Be careful not to disable services that are needed for
1031 rane 1.20 your system to run properly - <c>lm-profiler</c> is likely to generate some
1032     false positives. Do not disable a service if you are unsure whether it's
1033     needed.
1034 nightmorph 1.17 </p>
1035    
1036     </body>
1037     </section>
1038     <section>
1039 rane 1.19 <title>Limiting Write Accesses</title>
1040 nightmorph 1.17 <body>
1041    
1042     <p>
1043     If you don't want to use laptop-mode, you must take special care to disable
1044     services that write to your disk frequently - <c>syslogd</c> is a good
1045     candidate, for example. You probably don't want to shut it down completely, but
1046     it's possible to modify the config file so that "unnecessary" things don't get
1047 rane 1.20 logged and thus don't create disk traffic. <c>Cups</c> writes to disk
1048     periodically, so consider shutting it down and only enable it manually when
1049     needed.
1050 nightmorph 1.17 </p>
1051    
1052     <pre caption="Disabling cups in battery mode">
1053     # <i>rc-update del cupsd battery</i>
1054     </pre>
1055    
1056     <p>
1057     You can also use <c>lm-profiler</c> from laptop-mode-tools (see above) to find
1058     services to disable. Once you eliminated all of them, go on with configuring
1059     hdparm.
1060     </p>
1061    
1062     </body>
1063     </section>
1064     <section>
1065     <title>hdparm</title>
1066     <body>
1067    
1068     <p>
1069 nightmorph 1.28 The second possibility is using <c>hdparm</c>. Skip this if
1070     you are using laptop-mode. Otherwise, edit <path>/etc/conf.d/hdparm</path> and
1071     add the following values to your drive entries. This example assumes your hard
1072     drive is called <b>hda</b>:
1073 swift 1.1 </p>
1074    
1075 nightmorph 1.28 <pre caption="Using /etc/conf.d/hdparm for disk standby">
1076     hda_args="-q -S12"
1077     </pre>
1078 neysx 1.12
1079 nightmorph 1.28 <p>
1080     This will activate power management for your hard drive. If you ever want to
1081     deactivate power management, you can edit <path>/etc/conf.d/hdparm</path> and
1082     change the values to <c>-q -S0</c>, or just run <c>hdparm -q -S0 /dev/hda</c>.
1083     </p>
1084 swift 1.1
1085     <p>
1086 nightmorph 1.28 See <c>man hdparm</c> for the options. Though you can always start <c>hdparm</c>
1087     manually when you are on battery power by running <c>/etc/init.d/hdparm
1088     start</c>, it's much easier to automate its startup and shutdown. To do so, add
1089     <c>hdparm</c> to the battery runlevel so that it will automatically enable power
1090     management.
1091 swift 1.1 </p>
1092    
1093     <pre caption="Automate disk standby settings">
1094 nightmorph 1.28 # <i>rc-update add hdparm battery</i>
1095 swift 1.1 </pre>
1096    
1097     <impo>
1098     Be careful with sleep/spin down settings of your hard drive. Setting it to
1099     small values might wear out your drive and lose warranty.
1100     </impo>
1101    
1102     </body>
1103     </section>
1104     <section>
1105     <title>Other tricks</title>
1106     <body>
1107    
1108     <p>
1109     Another possibility is to deactivate swap in battery mode. Before writing a
1110     swapon/swapoff switcher, make sure there is enough RAM and swap isn't used
1111     heavily, otherwise you'll be in big problems.
1112     </p>
1113    
1114     <p>
1115     If you don't want to use laptop-mode, it's still possible to minimize disk
1116 rane 1.19 access by mounting certain directories as <c>tmpfs</c> - write accesses are not
1117 swift 1.1 stored on a disk, but in main memory and get lost with unmounting. Often it's
1118     useful to mount <path>/tmp</path> like this - you don't have to pay special
1119     attention as it gets cleared on every reboot regardless whether it was mounted
1120     on disk or in RAM. Just make sure you have enough RAM and no program (like a
1121     download client or compress utility) needs extraordinary much space in
1122     <path>/tmp</path>. To activate this, enable tmpfs support in your kernel and
1123     add a line to <path>/etc/fstab</path> like this:
1124     </p>
1125    
1126     <pre caption="Editing /etc/fstab to make /tmp even more volatile">
1127     none /tmp tmpfs size=32m 0 0
1128     </pre>
1129    
1130     <warn>
1131     Pay attention to the size parameter and modify it for your system. If you're
1132 rane 1.21 unsure, don't try this at all, it can become a performance bottleneck easily. In
1133 swift 1.1 case you want to mount <path>/var/log</path> like this, make sure to merge the
1134     log files to disk before unmounting. They are essential. Don't attempt to mount
1135 rane 1.19 <path>/var/tmp</path> like this. Portage uses it for compiling...
1136 swift 1.1 </warn>
1137    
1138     </body>
1139     </section>
1140     </chapter>
1141    
1142     <chapter>
1143 rane 1.19 <title>Power Management For Other Devices</title>
1144 swift 1.1 <section>
1145 rane 1.19 <title>Graphics Cards</title>
1146 so 1.15 <body>
1147    
1148     <p>
1149     In case you own an ATI graphics card supporting PowerPlay (dynamic clock
1150 nightmorph 1.25 scaling for the graphics processing unit GPU), you can activate this
1151 rane 1.20 feature in X.org. Open <path>/etc/X11/xorg.conf</path> and add (or enable) the
1152     <c>DynamicClocks</c> option in the Device section. Please notice that this
1153     feature will lead to crashes on some systems.
1154 so 1.15 </p>
1155    
1156     <pre caption="Enabling ATI PowerPlay support in X.org">
1157     Section "Device"
1158     [...]
1159     Option "DynamicClocks" "on"
1160     EndSection
1161     </pre>
1162    
1163     </body>
1164     </section>
1165     <section>
1166 swift 1.1 <title>Wireless Power Management</title>
1167     <body>
1168    
1169     <p>
1170 nightmorph 1.17 Wireless LAN cards consume quite a bit of energy. Put them in Power Management
1171 nightmorph 1.28 mode just like your hard drives.
1172 swift 1.1 </p>
1173    
1174 nightmorph 1.17 <note>
1175     This script assumes your wireless interface is called <c>wlan0</c>; replace
1176     this with the actual name of your interface.
1177     </note>
1178    
1179 nightmorph 1.28 <p>
1180 nightmorph 1.29 Add the following option to <path>/etc/conf.d/net</path> to automatically enable
1181 nightmorph 1.28 power management for your wireless card:
1182     </p>
1183 swift 1.1
1184 nightmorph 1.28 <pre caption="Automated WLAN Power Management">
1185     iwconfig_wlan0="power on"
1186 swift 1.1 </pre>
1187    
1188     <p>
1189 nightmorph 1.28 See <c>man iwconfig</c> for details and more options like the period between
1190     wakeups or timeout settings. If your driver and access point support changing
1191     the beacon time, this is a good starting point to save even more energy.
1192 swift 1.1 </p>
1193    
1194     </body>
1195     </section>
1196     <section>
1197     <title>USB Power Management</title>
1198     <body>
1199    
1200     <p>
1201     There are two problems with USB devices regarding energy consumption: First,
1202     devices like USB mice, digital cameras or USB sticks consume energy while
1203     plugged in. You cannot avoid this (nevertheless remove them in case they're not
1204     needed). Second, when there are USB devices plugged in, the USB host controller
1205     periodically accesses the bus which in turn prevents the CPU from going into
1206 rane 1.20 sleep mode. The kernel offers an experimental option to enable suspension of
1207 so 1.15 USB devices through driver calls or one of the <path>power/state</path> files
1208     in <path>/sys</path>.
1209 swift 1.1 </p>
1210    
1211 so 1.15 <pre caption="Enabling USB suspend support in the kernel">
1212     Device Drivers
1213     USB support
1214     [*] Support for Host-side USB
1215     [*] USB suspend/resume (EXPERIMENTAL)
1216     </pre>
1217    
1218 swift 1.1 </body>
1219     </section>
1220     </chapter>
1221    
1222     <chapter>
1223 rane 1.19 <title>Sleep States: sleep, standby, and suspend to disk</title>
1224 swift 1.1 <section>
1225     <body>
1226    
1227     <p>
1228     ACPI defines different sleep states. The more important ones are
1229     </p>
1230    
1231 rane 1.20 <ul>
1232 swift 1.1 <li>S1 aka Standby</li>
1233     <li>S3 aka Suspend to RAM aka Sleep</li>
1234     <li>S4 aka Suspend to Disk aka Hibernate</li>
1235     </ul>
1236    
1237     <p>
1238     They can be called whenever the system is not in use, but a shutdown is not
1239     wanted due to the long boot time.
1240     </p>
1241    
1242     </body>
1243     </section>
1244     <section>
1245 so 1.15 <title>Sleep (S3)</title>
1246 swift 1.1 <body>
1247    
1248     <p>
1249 so 1.15 The ACPI support for these sleep states is marked experimental for good reason.
1250     APM sleep states seem to be more stable, however you can't use APM and ACPI
1251     together.
1252     </p>
1253    
1254     <pre caption="Kernel configuration for the various suspend types">
1255     Power Management Options ---&gt;
1256     [*] Power Management support
1257     ACPI (Advanced Configuration and Power Interface) Support ---&gt;
1258     [*] ACPI Support
1259     [*] Sleep States
1260     </pre>
1261    
1262     <p>
1263 nightmorph 1.17 Once your kernel is properly configured, you can use the
1264 so 1.15 <c>hibernate-script</c> to activate suspend or sleep mode. Let's install that
1265     first.
1266     </p>
1267    
1268     <pre caption="Installing the hibernate-script">
1269 rane 1.19 # <i>emerge hibernate-script</i>
1270 so 1.15 </pre>
1271    
1272     <p>
1273 nightmorph 1.25 Some configuration has to be done in <path>/etc/hibernate</path>. The default
1274 nightmorph 1.24 package introduces a few configuration files for each sleep state. Options that
1275     are common to all suspend methods are placed in <path>common.conf</path>; make
1276     sure this file is properly set up for your system.
1277 so 1.15 </p>
1278    
1279     <p>
1280 nightmorph 1.24 To configure sleep, edit <path>sysfs-ram.conf</path> in
1281     <path>/etc/hibernate</path>. <c>UseSysfsPowerState mem</c> is already setup
1282 nightmorph 1.25 correctly, but if you need to make further changes to this particular sleep
1283     state (or any other sleep state) you should add them to
1284 nightmorph 1.24 <path>/etc/hibernate/hibernate.conf</path>. The comments and option names will
1285     guide you. If you use nfs or samba shares over the network, make sure to
1286     shutdown the appropriate init scripts to avoid timeouts.
1287 so 1.15 </p>
1288    
1289 nightmorph 1.24 <note>
1290     For more information on setting up sleep states, read <c>man
1291     hibernate.conf</c>.
1292     </note>
1293    
1294 so 1.15 <p>
1295     Ready? Now is the last chance to backup any data you want to keep after
1296     executing the next command. Notice that you probably have to hit a special key
1297 rane 1.19 like <c>Fn</c> to resume from sleep.
1298 so 1.15 </p>
1299    
1300     <pre caption="Calling sleep">
1301 rane 1.19 # <i>hibernate-ram</i>
1302 so 1.15 </pre>
1303    
1304     <p>
1305 rane 1.20 If you're still reading, it seems to work. You can also setup standby (S1) in a
1306 nightmorph 1.24 similar way by editing <path>sysfs-ram.conf</path> and changing
1307     "UseSysfsPowerState mem" to "UseSysfsPowerState standby". S3 and S4 are the more
1308     interesting sleep states due to greater energy savings however.
1309 so 1.15 </p>
1310    
1311     </body>
1312     </section>
1313     <section>
1314     <title>Hibernate (S4)</title>
1315     <body>
1316    
1317     <p>
1318     This section introduces hibernation, where a snapshot of the running system is
1319     written to disk before powering off. On resume, the snapshot is loaded and you
1320     can go on working at exactly the point you called hibernate before.
1321 swift 1.1 </p>
1322    
1323     <warn>
1324 so 1.15 Don't exchange non hot-pluggable hardware when suspended. Don't attempt to load
1325     a snapshot with a different kernel image than the one it was created with.
1326     Shutdown any NFS or samba server/client before hibernating.
1327 swift 1.1 </warn>
1328    
1329     <p>
1330 so 1.15 There are two different implementations for S4. The original one is swsusp,
1331 rane 1.20 then there is the newer suspend2 with a nicer interface (including fbsplash
1332     support). A <uri link="http://suspend2.net/features.html#compare"> feature
1333     comparison</uri> is available at the <uri link="http://suspend2.net"> suspend2
1334     Homepage</uri>. There used to be Suspend-to-Disk (pmdisk), a fork of swsusp,
1335     but it has been merged back.
1336 swift 1.1 </p>
1337    
1338     <p>
1339 so 1.15 Suspend2 is not included in the mainline kernel yet, therefore you either have
1340 rane 1.20 to patch your kernel sources with the patches provided by <uri
1341     link="http://suspend2.net">suspend2.net</uri> or use
1342 so 1.15 <c>sys-kernel/suspend2-sources</c>.
1343 swift 1.1 </p>
1344    
1345     <p>
1346 so 1.15 The kernel part for both swusp and suspend2 is as follows:
1347 swift 1.1 </p>
1348    
1349     <pre caption="Kernel configuration for the various suspend types">
1350     Power Management Options ---&gt;
1351     <comment>(hibernate with swsusp)</comment>
1352 so 1.15 [*] Software Suspend
1353     <comment>(replace /dev/SWAP with your swap partition)</comment>
1354     (/dev/SWAP) Default resume partition
1355 rane 1.20
1356 so 1.15 <comment>(hibernate with suspend2)</comment>
1357 swift 1.1 Software Suspend 2
1358     --- Image Storage (you need at least one writer)
1359 so 1.15 [*] File Writer
1360 swift 1.1 [*] Swap Writer
1361 so 1.15 --- General Options
1362 swift 1.1 [*] LZF image compression
1363 so 1.15 <comment>(replace /dev/SWAP with your swap partition)</comment>
1364     (swap:/dev/SWAP) Default resume device name
1365     [ ] Allow Keep Image Mode
1366 swift 1.1 </pre>
1367    
1368     <p>
1369 so 1.15 The configuration for swsusp is rather easy. If you didn't store the location
1370     of your swap partition in the kernel config, you can also pass it as a
1371     parameter with the <c>resume=/dev/SWAP</c> directive. If booting is not
1372     possible due to a broken image, use the <c>noresume</c> kernel parameter. The
1373 rane 1.20 <c>hibernate-cleanup</c> init script invalidates swsusp images during the boot
1374     process.
1375 swift 1.1 </p>
1376    
1377 so 1.15 <pre caption="Invalidating swsusp images during the boot process">
1378 rane 1.19 # <i>rc-update add hibernate-cleanup boot</i>
1379 so 1.15 </pre>
1380    
1381 swift 1.1 <p>
1382 so 1.15 To activate hibernate with swsusp, use the hibernate script and set
1383 nightmorph 1.24 <c>UseSysfsPowerState disk</c> in <path>/etc/hibernate/sysfs-disk</path>.
1384 swift 1.1 </p>
1385    
1386     <warn>
1387     Backup your data before doing this. Run <c>sync</c> before executing one of the
1388     commands to have cached data written to disk. First try it outside of X, then
1389     with X running, but not logged in.
1390     </warn>
1391 rane 1.20
1392 swift 1.1 <p>
1393     If you experience kernel panics due to uhci or similar, try to compile USB
1394     support as module and unload the modules before sending your laptop to sleep
1395 nightmorph 1.24 mode. There are configuration options for this in <path>common.conf</path>
1396 swift 1.1 </p>
1397    
1398 so 1.15 <pre caption="Hibernating with swsusp">
1399 nightmorph 1.24 # <i>nano -w /etc/hibernate/common.conf</i>
1400 so 1.15 <comment>(Make sure you have a backup of your data)</comment>
1401 rane 1.19 # <i>hibernate</i>
1402 so 1.15 </pre>
1403    
1404 swift 1.1 <p>
1405 so 1.15 The following section discusses the setup of suspend2 including fbsplash
1406     support for a nice graphical progress bar during suspend and resume.
1407     </p>
1408    
1409     <p>
1410 rane 1.20 The first part of the configuration is similar to the configuration of swsusp.
1411     In case you didn't store the location of your swap partition in the kernel
1412     config, you have to pass it as a kernel parameter with the
1413 so 1.15 <c>resume2=swap:/dev/SWAP</c> directive. If booting is not possible due to a
1414     broken image, append the <c>noresume2</c> parameter. Additionally, the
1415     <c>hibernate-cleanup</c> init script invalidates suspend2 images during the
1416     boot process.
1417     </p>
1418    
1419     <pre caption="Invalidating suspend2 images during the boot process">
1420 rane 1.19 # <i>rc-update add hibernate-cleanup boot</i>
1421 so 1.15 </pre>
1422    
1423 rane 1.20 <p>
1424 nightmorph 1.24 Now edit <path>/etc/hibernate/suspend2.conf</path>, enable the <c>suspend2</c>
1425     options you need. Do not enable the <c>fbsplash</c> options in
1426     <c>common.conf</c> just yet.
1427 swift 1.1 </p>
1428    
1429 so 1.15 <pre caption="Hibernating with suspend2">
1430 nightmorph 1.24 # <i>nano -w /etc/hibernate/suspend2.conf</i>
1431 so 1.15 <comment>(Make sure you have a backup of your data)</comment>
1432 rane 1.19 # <i>hibernate</i>
1433 swift 1.8 </pre>
1434    
1435 so 1.15 <p>
1436 rane 1.20 Please configure <c>fbsplash</c> now if you didn't do already. To enable
1437     fbsplash support during hibernation, the <c>sys-apps/suspend2-userui</c>
1438     package is needed. Additionally, you've got to enable the <c>fbsplash</c> USE
1439     flag.
1440 so 1.15 </p>
1441    
1442     <pre caption="Installing suspend2-userui">
1443 rane 1.19 # <i>echo "sys-apps/suspend2-userui fbsplash" >> /etc/portage/package.use</i>
1444 nightmorph 1.24 <comment>(It may be marked ~arch, so first it must be keyworded)</comment>
1445     # <i>echo "sys-apps/suspend2-userui" >> /etc/portage/package.keywords</i>
1446 rane 1.19 # <i>emerge suspend2-userui</i>
1447 so 1.15 </pre>
1448    
1449     <p>
1450     The ebuild tells you to make a symlink to the theme you want to use. For
1451     example, to use the <c>livecd-2005.1</c> theme, run the following command:
1452     </p>
1453    
1454     <pre caption="Using the livecd-2005.1 theme during hibernation">
1455 rane 1.19 # <i>ln -sfn /etc/splash/livecd-2005.1 /etc/splash/suspend2</i>
1456 so 1.15 </pre>
1457    
1458     <p>
1459     If you don't want a black screen in the first part of the resume process, you
1460     have to add the <c>suspend2ui_fbsplash</c> tool to your initrd image. Assuming
1461     you created the initrd image with <c>splash_geninitramfs</c> and saved it as
1462 rane 1.20 <path>/boot/fbsplash-emergence-1024x768</path>, here's how to do that.
1463 so 1.15 </p>
1464    
1465     <pre caption="Adding suspend2ui_fbsplash to an initrd image">
1466 rane 1.19 # <i>mount /boot</i>
1467     # <i>mkdir ~/initrd.d</i>
1468     # <i>cp /boot/fbsplash-emergence-1024x768 ~/initrd.d/</i>
1469     # <i>cd ~/initrd.d</i>
1470     # <i>gunzip -c fbsplash-emergence-1024x768 | cpio -idm --quiet -H newc</i>
1471     # <i>rm fbsplash-emergence-1024x768</i>
1472     # <i>cp /usr/sbin/suspend2ui_fbsplash sbin/</i>
1473     # <i>find . | cpio --quiet --dereference -o -H newc | gzip -9 > /boot/fbsplash-suspend2-emergence-1024x768</i>
1474 so 1.15 </pre>
1475    
1476     <p>
1477 nightmorph 1.24 Afterwards adjust <path>grub.conf</path> (or <path>lilo.conf</path>) so that
1478     your suspend2 kernel uses
1479     <path>/boot/fbsplash-suspend2-emergence-1024x768</path> as initrd image. You can
1480     now test a dry run to see if everything is setup correctly.
1481 so 1.15 </p>
1482    
1483     <pre caption="Test run for fbsplash hibernation">
1484 rane 1.19 # <i>suspend2ui_fbsplash -t</i>
1485 so 1.15 </pre>
1486    
1487     <p>
1488 nightmorph 1.24 Afterwards open <path>/etc/hibernate/common.conf</path> and activate the
1489     fbsplash options. Execute <c>hibernate</c> and enjoy.
1490 so 1.15 </p>
1491    
1492 swift 1.1 </body>
1493     </section>
1494     </chapter>
1495    
1496     <chapter>
1497     <title>Troubleshooting</title>
1498     <section>
1499     <body>
1500    
1501     <p>
1502     <e>Q:</e> I'm trying to change the CPU frequency, but
1503     <path>/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor</path> does not
1504     exist.
1505     </p>
1506    
1507     <p>
1508     <e>A:</e> Make sure your processor supports CPU frequency scaling and you chose
1509     the right CPUFreq driver for your processor. Here is a list of processors that
1510     are supported by cpufreq (kernel 2.6.7): ARM Integrator, ARM-SA1100,
1511     ARM-SA1110, AMD Elan - SC400, SC410, AMD mobile K6-2+, AMD mobile K6-3+, AMD
1512     mobile Duron, AMD mobile Athlon, AMD Opteron, AMD Athlon 64, Cyrix Media GXm,
1513     Intel mobile PIII and Intel mobile PIII-M on certain chipsets, Intel Pentium 4,
1514     Intel Xeon, Intel Pentium M (Centrino), National Semiconductors Geode GX,
1515     Transmeta Crusoe, VIA Cyrix 3 / C3, UltraSPARC-III, SuperH SH-3, SH-4, several
1516     "PowerBook" and "iBook2" and various processors on some ACPI 2.0-compatible
1517     systems (only if "ACPI Processor Performance States" are available to the
1518     ACPI/BIOS interface).
1519     </p>
1520    
1521     <p>
1522     <e>Q:</e> My laptop supports frequency scaling, but
1523     <path>/sys/devices/system/cpu/cpu0/cpufreq/</path> is empty.
1524     </p>
1525    
1526     <p>
1527     <e>A:</e> Look for ACPI related error messages with <c>dmesg | grep ACPI</c>.
1528     Try to update the BIOS, especially if a broken DSDT is reported. You can also
1529     try to fix it yourself (which is beyond the scope of this guide).
1530     </p>
1531    
1532     <p>
1533 rane 1.19 <e>Q:</e> My laptop supports frequency scaling, but according to
1534     <path>/proc/cpuinfo</path> the speed never changes.
1535 swift 1.1 </p>
1536    
1537     <p>
1538 swift 1.8 <e>A:</e> Probably you have activated symmetric multiprocessing support
1539     (CONFIG_SMP) in your kernel. Deactivate it and it should work. Some older
1540 rane 1.20 kernels had a bug causing this. In that case, run <c>emerge x86info</c>, update
1541     your kernel as asked and check the current frequency with
1542 swift 1.8 <c>x86info -mhz</c>.
1543 swift 1.1 </p>
1544    
1545     <p>
1546     <e>Q:</e> I can change the CPU frequency, but the range is not as wide as in
1547     another OS.
1548     </p>
1549    
1550     <p>
1551     <e>A:</e> You can combine frequency scaling with ACPI throttling to get a lower
1552     minimum frequency. Notice that throttling doesn't save much energy and is
1553     mainly used for thermal management (keeping your laptop cool and quiet). You
1554     can read the current throttling state with <c>cat
1555     /proc/acpi/processor/CPU/throttling</c> and change it with <c>echo -n "0:x" >
1556     /proc/acpi/processor/CPU/limit</c>, where x is one of the Tx states listed in
1557     <path>/proc/acpi/processor/CPU/throttling</path>.
1558     </p>
1559    
1560     <p>
1561 swift 1.8 <e>Q:</e> When configuring the kernel, powersave, performance and userspace
1562     governors show up, but that ondemand thing is missing. Where do I get it?
1563     </p>
1564    
1565     <p>
1566     <e>A:</e> The ondemand governor is only included in recent kernel sources. Try
1567     updating them.
1568     </p>
1569    
1570     <p>
1571 swift 1.1 <e>Q:</e> Battery life time seems to be worse than before.
1572     </p>
1573    
1574     <p>
1575     <e>A:</e> Check your BIOS settings. Maybe you forgot to re-enable some of the
1576     settings.
1577     </p>
1578    
1579     <p>
1580     <e>Q:</e> My battery is charged, but KDE reports there would be 0% left and
1581     immediately shuts down.
1582     </p>
1583    
1584     <p>
1585     <e>A:</e> Check that battery support is compiled into your kernel. If you use
1586     it as a module, make sure the module is loaded.
1587     </p>
1588    
1589     <p>
1590 rane 1.20 <e>Q:</e> My system logger reports things like "logger: ACPI group battery /
1591     action battery is not defined".
1592 nightmorph 1.17 </p>
1593    
1594     <p>
1595 rane 1.20 <e>A:</e> This message is generated by the <path>/etc/acpi/default.sh</path>
1596     script that is shipped with acpid. You can safely ignore it. If you like to get
1597     rid of it, you can comment the appropriate line in
1598     <path>/etc/acpi/default.sh</path> as shown below:
1599 nightmorph 1.17 </p>
1600    
1601     <pre caption="Disabling warnings about unknown acpi events">
1602     *) # logger "ACPI action $action is not defined"
1603     </pre>
1604    
1605     <p>
1606 swift 1.1 <e>Q:</e> I have a Dell Inspiron 51XX and I don't get any ACPI events.
1607     </p>
1608    
1609     <p>
1610     <e>A:</e> This seems to be a kernel bug. Read on <uri
1611     link="http://bugme.osdl.org/show_bug.cgi?id=1752">here</uri>.
1612     </p>
1613    
1614     <p>
1615 rane 1.20 <e>Q:</e> I activated the <c>DynamicClocks</c> option in <path>xorg.conf</path>
1616     and now X.org crashes / the screen stays black / my laptop doesn't shutdown
1617 so 1.15 properly.
1618     </p>
1619    
1620     <p>
1621 rane 1.20 <e>A:</e> This happens on some systems. You have to disable
1622     <c>DynamicClocks</c>.
1623 so 1.15 </p>
1624    
1625     <p>
1626     <e>Q:</e> I want to use suspend2, but it tells me my swap partition is too
1627     small. Resizing is not an option.
1628     </p>
1629    
1630     <p>
1631 rane 1.20 <e>A:</e> If there is enough free space on your system, you can use the
1632 so 1.15 filewriter instead of the swapwriter. The <c>hibernate-script</c> supports it
1633     as well. More information can be found in
1634     <path>/usr/src/linux/Documentation/power/suspend2.txt</path>.
1635     </p>
1636    
1637     <p>
1638 swift 1.1 <e>Q:</e> I just bought a brand new battery, but it only lasts for some
1639     minutes! What am I doing wrong?
1640     </p>
1641    
1642     <p>
1643     <e>A:</e> First follow your manufacturer's advice on how to charge the battery
1644 rane 1.20 correctly.
1645 swift 1.1 </p>
1646    
1647     <p>
1648     <e>Q:</e> The above didn't help. What should I do then?
1649     </p>
1650    
1651     <p>
1652     <e>A:</e> Some batteries sold as "new" are in fact old ones. Try the following:
1653     </p>
1654    
1655     <pre caption="Querying battery state">
1656     $ <i>grep capacity /proc/acpi/battery/BAT0/info</i>
1657     design capacity: 47520 mWh
1658     last full capacity: 41830 mWh
1659     </pre>
1660    
1661     <p>
1662     If the "last full capacity" differs significantly from the design capacity,
1663     your battery is probably broken. Try to claim your warranty.
1664     </p>
1665    
1666 swift 1.8 <p>
1667     <e>Q:</e> My problem is not listed above. Where should I go next?
1668     </p>
1669    
1670     <p>
1671 nightmorph 1.17 <e>A:</e> Don't fear to contact me, <mail link="earthwings@gentoo.org">Dennis
1672 rane 1.20 Nienhüser</mail>, directly. The <uri link="http://forums.gentoo.org">Gentoo
1673     Forums</uri> are a good place to get help as well. If you prefer IRC, try the
1674     <c>#gentoo-laptop</c> channel at <uri
1675     link="irc://irc.freenode.net">irc.freenode.net</uri>.
1676 swift 1.8 </p>
1677    
1678 swift 1.1 </body>
1679     </section>
1680     </chapter>
1681     </guide>

  ViewVC Help
Powered by ViewVC 1.1.20