1Some warnings, first.
2
3 * BIG FAT WARNING *********************************************************
4 *
5 * If you touch anything on disk between suspend and resume...
6 *				...kiss your data goodbye.
7 *
8 * If you do resume from initrd after your filesystems are mounted...
9 *				...bye bye root partition.
10 *			[this is actually same case as above]
11 *
12 * If you have unsupported (*) devices using DMA, you may have some
13 * problems. If your disk driver does not support suspend... (IDE does),
14 * it may cause some problems, too. If you change kernel command line
15 * between suspend and resume, it may do something wrong. If you change
16 * your hardware while system is suspended... well, it was not good idea;
17 * but it will probably only crash.
18 *
19 * (*) suspend/resume support is needed to make it safe.
20 *
21 * If you have any filesystems on USB devices mounted before software suspend,
22 * they won't be accessible after resume and you may lose data, as though
23 * you have unplugged the USB devices with mounted filesystems on them;
24 * see the FAQ below for details.  (This is not true for more traditional
25 * power states like "standby", which normally don't turn USB off.)
26
27You need to append resume=/dev/your_swap_partition to kernel command
28line. Then you suspend by
29
30echo shutdown > /sys/power/disk; echo disk > /sys/power/state
31
32. If you feel ACPI works pretty well on your system, you might try
33
34echo platform > /sys/power/disk; echo disk > /sys/power/state
35
36. If you would like to write hibernation image to swap and then suspend
37to RAM (provided your platform supports it), you can try
38
39echo suspend > /sys/power/disk; echo disk > /sys/power/state
40
41. If you have SATA disks, you'll need recent kernels with SATA suspend
42support. For suspend and resume to work, make sure your disk drivers
43are built into kernel -- not modules. [There's way to make
44suspend/resume with modular disk drivers, see FAQ, but you probably
45should not do that.]
46
47If you want to limit the suspend image size to N bytes, do
48
49echo N > /sys/power/image_size
50
51before suspend (it is limited to 500 MB by default).
52
53. The resume process checks for the presence of the resume device,
54if found, it then checks the contents for the hibernation image signature.
55If both are found, it resumes the hibernation image.
56
57. The resume process may be triggered in two ways:
58  1) During lateinit:  If resume=/dev/your_swap_partition is specified on
59     the kernel command line, lateinit runs the resume process.  If the
60     resume device has not been probed yet, the resume process fails and
61     bootup continues.
62  2) Manually from an initrd or initramfs:  May be run from
63     the init script by using the /sys/power/resume file.  It is vital
64     that this be done prior to remounting any filesystems (even as
65     read-only) otherwise data may be corrupted.
66
67Article about goals and implementation of Software Suspend for Linux
68~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
69Author: Gábor Kuti
70Last revised: 2003-10-20 by Pavel Machek
71
72Idea and goals to achieve
73
74Nowadays it is common in several laptops that they have a suspend button. It
75saves the state of the machine to a filesystem or to a partition and switches
76to standby mode. Later resuming the machine the saved state is loaded back to
77ram and the machine can continue its work. It has two real benefits. First we
78save ourselves the time machine goes down and later boots up, energy costs
79are real high when running from batteries. The other gain is that we don't have to
80interrupt our programs so processes that are calculating something for a long
81time shouldn't need to be written interruptible.
82
83swsusp saves the state of the machine into active swaps and then reboots or
84powerdowns.  You must explicitly specify the swap partition to resume from with
85``resume='' kernel option. If signature is found it loads and restores saved
86state. If the option ``noresume'' is specified as a boot parameter, it skips
87the resuming.  If the option ``hibernate=nocompress'' is specified as a boot
88parameter, it saves hibernation image without compression.
89
90In the meantime while the system is suspended you should not add/remove any
91of the hardware, write to the filesystems, etc.
92
93Sleep states summary
94====================
95
96There are three different interfaces you can use, /proc/acpi should
97work like this:
98
99In a really perfect world:
100echo 1 > /proc/acpi/sleep       # for standby
101echo 2 > /proc/acpi/sleep       # for suspend to ram
102echo 3 > /proc/acpi/sleep       # for suspend to ram, but with more power conservative
103echo 4 > /proc/acpi/sleep       # for suspend to disk
104echo 5 > /proc/acpi/sleep       # for shutdown unfriendly the system
105
106and perhaps
107echo 4b > /proc/acpi/sleep      # for suspend to disk via s4bios
108
109Frequently Asked Questions
110==========================
111
112Q: well, suspending a server is IMHO a really stupid thing,
113but... (Diego Zuccato):
114
115A: You bought new UPS for your server. How do you install it without
116bringing machine down? Suspend to disk, rearrange power cables,
117resume.
118
119You have your server on UPS. Power died, and UPS is indicating 30
120seconds to failure. What do you do? Suspend to disk.
121
122
123Q: Maybe I'm missing something, but why don't the regular I/O paths work?
124
125A: We do use the regular I/O paths. However we cannot restore the data
126to its original location as we load it. That would create an
127inconsistent kernel state which would certainly result in an oops.
128Instead, we load the image into unused memory and then atomically copy
129it back to it original location. This implies, of course, a maximum
130image size of half the amount of memory.
131
132There are two solutions to this:
133
134* require half of memory to be free during suspend. That way you can
135read "new" data onto free spots, then cli and copy
136
137* assume we had special "polling" ide driver that only uses memory
138between 0-640KB. That way, I'd have to make sure that 0-640KB is free
139during suspending, but otherwise it would work...
140
141suspend2 shares this fundamental limitation, but does not include user
142data and disk caches into "used memory" by saving them in
143advance. That means that the limitation goes away in practice.
144
145Q: Does linux support ACPI S4?
146
147A: Yes. That's what echo platform > /sys/power/disk does.
148
149Q: What is 'suspend2'?
150
151A: suspend2 is 'Software Suspend 2', a forked implementation of
152suspend-to-disk which is available as separate patches for 2.4 and 2.6
153kernels from swsusp.sourceforge.net. It includes support for SMP, 4GB
154highmem and preemption. It also has a extensible architecture that
155allows for arbitrary transformations on the image (compression,
156encryption) and arbitrary backends for writing the image (eg to swap
157or an NFS share[Work In Progress]). Questions regarding suspend2
158should be sent to the mailing list available through the suspend2
159website, and not to the Linux Kernel Mailing List. We are working
160toward merging suspend2 into the mainline kernel.
161
162Q: What is the freezing of tasks and why are we using it?
163
164A: The freezing of tasks is a mechanism by which user space processes and some
165kernel threads are controlled during hibernation or system-wide suspend (on some
166architectures).  See freezing-of-tasks.txt for details.
167
168Q: What is the difference between "platform" and "shutdown"?
169
170A:
171
172shutdown: save state in linux, then tell bios to powerdown
173
174platform: save state in linux, then tell bios to powerdown and blink
175          "suspended led"
176
177"platform" is actually right thing to do where supported, but
178"shutdown" is most reliable (except on ACPI systems).
179
180Q: I do not understand why you have such strong objections to idea of
181selective suspend.
182
183A: Do selective suspend during runtime power management, that's okay. But
184it's useless for suspend-to-disk. (And I do not see how you could use
185it for suspend-to-ram, I hope you do not want that).
186
187Lets see, so you suggest to
188
189* SUSPEND all but swap device and parents
190* Snapshot
191* Write image to disk
192* SUSPEND swap device and parents
193* Powerdown
194
195Oh no, that does not work, if swap device or its parents uses DMA,
196you've corrupted data. You'd have to do
197
198* SUSPEND all but swap device and parents
199* FREEZE swap device and parents
200* Snapshot
201* UNFREEZE swap device and parents
202* Write
203* SUSPEND swap device and parents
204
205Which means that you still need that FREEZE state, and you get more
206complicated code. (And I have not yet introduce details like system
207devices).
208
209Q: There don't seem to be any generally useful behavioral
210distinctions between SUSPEND and FREEZE.
211
212A: Doing SUSPEND when you are asked to do FREEZE is always correct,
213but it may be unnecessarily slow. If you want your driver to stay simple,
214slowness may not matter to you. It can always be fixed later.
215
216For devices like disk it does matter, you do not want to spindown for
217FREEZE.
218
219Q: After resuming, system is paging heavily, leading to very bad interactivity.
220
221A: Try running
222
223cat /proc/[0-9]*/maps | grep / | sed 's:.* /:/:' | sort -u | while read file
224do
225  test -f "$file" && cat "$file" > /dev/null
226done
227
228after resume. swapoff -a; swapon -a may also be useful.
229
230Q: What happens to devices during swsusp? They seem to be resumed
231during system suspend?
232
233A: That's correct. We need to resume them if we want to write image to
234disk. Whole sequence goes like
235
236      Suspend part
237      ~~~~~~~~~~~~
238      running system, user asks for suspend-to-disk
239
240      user processes are stopped
241
242      suspend(PMSG_FREEZE): devices are frozen so that they don't interfere
243      		      with state snapshot
244
245      state snapshot: copy of whole used memory is taken with interrupts disabled
246
247      resume(): devices are woken up so that we can write image to swap
248
249      write image to swap
250
251      suspend(PMSG_SUSPEND): suspend devices so that we can power off
252
253      turn the power off
254
255      Resume part
256      ~~~~~~~~~~~
257      (is actually pretty similar)
258
259      running system, user asks for suspend-to-disk
260
261      user processes are stopped (in common case there are none, but with resume-from-initrd, no one knows)
262
263      read image from disk
264
265      suspend(PMSG_FREEZE): devices are frozen so that they don't interfere
266      		      with image restoration
267
268      image restoration: rewrite memory with image
269
270      resume(): devices are woken up so that system can continue
271
272      thaw all user processes
273
274Q: What is this 'Encrypt suspend image' for?
275
276A: First of all: it is not a replacement for dm-crypt encrypted swap.
277It cannot protect your computer while it is suspended. Instead it does
278protect from leaking sensitive data after resume from suspend.
279
280Think of the following: you suspend while an application is running
281that keeps sensitive data in memory. The application itself prevents
282the data from being swapped out. Suspend, however, must write these
283data to swap to be able to resume later on. Without suspend encryption
284your sensitive data are then stored in plaintext on disk.  This means
285that after resume your sensitive data are accessible to all
286applications having direct access to the swap device which was used
287for suspend. If you don't need swap after resume these data can remain
288on disk virtually forever. Thus it can happen that your system gets
289broken in weeks later and sensitive data which you thought were
290encrypted and protected are retrieved and stolen from the swap device.
291To prevent this situation you should use 'Encrypt suspend image'.
292
293During suspend a temporary key is created and this key is used to
294encrypt the data written to disk. When, during resume, the data was
295read back into memory the temporary key is destroyed which simply
296means that all data written to disk during suspend are then
297inaccessible so they can't be stolen later on.  The only thing that
298you must then take care of is that you call 'mkswap' for the swap
299partition used for suspend as early as possible during regular
300boot. This asserts that any temporary key from an oopsed suspend or
301from a failed or aborted resume is erased from the swap device.
302
303As a rule of thumb use encrypted swap to protect your data while your
304system is shut down or suspended. Additionally use the encrypted
305suspend image to prevent sensitive data from being stolen after
306resume.
307
308Q: Can I suspend to a swap file?
309
310A: Generally, yes, you can.  However, it requires you to use the "resume=" and
311"resume_offset=" kernel command line parameters, so the resume from a swap file
312cannot be initiated from an initrd or initramfs image.  See
313swsusp-and-swap-files.txt for details.
314
315Q: Is there a maximum system RAM size that is supported by swsusp?
316
317A: It should work okay with highmem.
318
319Q: Does swsusp (to disk) use only one swap partition or can it use
320multiple swap partitions (aggregate them into one logical space)?
321
322A: Only one swap partition, sorry.
323
324Q: If my application(s) causes lots of memory & swap space to be used
325(over half of the total system RAM), is it correct that it is likely
326to be useless to try to suspend to disk while that app is running?
327
328A: No, it should work okay, as long as your app does not mlock()
329it. Just prepare big enough swap partition.
330
331Q: What information is useful for debugging suspend-to-disk problems?
332
333A: Well, last messages on the screen are always useful. If something
334is broken, it is usually some kernel driver, therefore trying with as
335little as possible modules loaded helps a lot. I also prefer people to
336suspend from console, preferably without X running. Booting with
337init=/bin/bash, then swapon and starting suspend sequence manually
338usually does the trick. Then it is good idea to try with latest
339vanilla kernel.
340
341Q: How can distributions ship a swsusp-supporting kernel with modular
342disk drivers (especially SATA)?
343
344A: Well, it can be done, load the drivers, then do echo into
345/sys/power/resume file from initrd. Be sure not to mount
346anything, not even read-only mount, or you are going to lose your
347data.
348
349Q: How do I make suspend more verbose?
350
351A: If you want to see any non-error kernel messages on the virtual
352terminal the kernel switches to during suspend, you have to set the
353kernel console loglevel to at least 4 (KERN_WARNING), for example by
354doing
355
356	# save the old loglevel
357	read LOGLEVEL DUMMY < /proc/sys/kernel/printk
358	# set the loglevel so we see the progress bar.
359	# if the level is higher than needed, we leave it alone.
360	if [ $LOGLEVEL -lt 5 ]; then
361	        echo 5 > /proc/sys/kernel/printk
362		fi
363
364        IMG_SZ=0
365        read IMG_SZ < /sys/power/image_size
366        echo -n disk > /sys/power/state
367        RET=$?
368        #
369        # the logic here is:
370        # if image_size > 0 (without kernel support, IMG_SZ will be zero),
371        # then try again with image_size set to zero.
372	if [ $RET -ne 0 -a $IMG_SZ -ne 0 ]; then # try again with minimal image size
373                echo 0 > /sys/power/image_size
374                echo -n disk > /sys/power/state
375                RET=$?
376        fi
377
378	# restore previous loglevel
379	echo $LOGLEVEL > /proc/sys/kernel/printk
380	exit $RET
381
382Q: Is this true that if I have a mounted filesystem on a USB device and
383I suspend to disk, I can lose data unless the filesystem has been mounted
384with "sync"?
385
386A: That's right ... if you disconnect that device, you may lose data.
387In fact, even with "-o sync" you can lose data if your programs have
388information in buffers they haven't written out to a disk you disconnect,
389or if you disconnect before the device finished saving data you wrote.
390
391Software suspend normally powers down USB controllers, which is equivalent
392to disconnecting all USB devices attached to your system.
393
394Your system might well support low-power modes for its USB controllers
395while the system is asleep, maintaining the connection, using true sleep
396modes like "suspend-to-RAM" or "standby".  (Don't write "disk" to the
397/sys/power/state file; write "standby" or "mem".)  We've not seen any
398hardware that can use these modes through software suspend, although in
399theory some systems might support "platform" modes that won't break the
400USB connections.
401
402Remember that it's always a bad idea to unplug a disk drive containing a
403mounted filesystem.  That's true even when your system is asleep!  The
404safest thing is to unmount all filesystems on removable media (such USB,
405Firewire, CompactFlash, MMC, external SATA, or even IDE hotplug bays)
406before suspending; then remount them after resuming.
407
408There is a work-around for this problem.  For more information, see
409Documentation/usb/persist.txt.
410
411Q: Can I suspend-to-disk using a swap partition under LVM?
412
413A: No. You can suspend successfully, but you'll not be able to
414resume. uswsusp should be able to work with LVM. See suspend.sf.net.
415
416Q: I upgraded the kernel from 2.6.15 to 2.6.16. Both kernels were
417compiled with the similar configuration files. Anyway I found that
418suspend to disk (and resume) is much slower on 2.6.16 compared to
4192.6.15. Any idea for why that might happen or how can I speed it up?
420
421A: This is because the size of the suspend image is now greater than
422for 2.6.15 (by saving more data we can get more responsive system
423after resume).
424
425There's the /sys/power/image_size knob that controls the size of the
426image.  If you set it to 0 (eg. by echo 0 > /sys/power/image_size as
427root), the 2.6.15 behavior should be restored.  If it is still too
428slow, take a look at suspend.sf.net -- userland suspend is faster and
429supports LZF compression to speed it up further.
430