Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | From kernel/suspend.c: |
| 2 | |
| 3 | * BIG FAT WARNING ********************************************************* |
| 4 | * |
| 5 | * If you have unsupported (*) devices using DMA... |
| 6 | * ...say goodbye to your data. |
| 7 | * |
| 8 | * If you touch anything on disk between suspend and resume... |
| 9 | * ...kiss your data goodbye. |
| 10 | * |
| 11 | * If your disk driver does not support suspend... (IDE does) |
| 12 | * ...you'd better find out how to get along |
| 13 | * without your data. |
| 14 | * |
| 15 | * If you change kernel command line between suspend and resume... |
| 16 | * ...prepare for nasty fsck or worse. |
| 17 | * |
| 18 | * If you change your hardware while system is suspended... |
| 19 | * ...well, it was not good idea. |
| 20 | * |
| 21 | * (*) suspend/resume support is needed to make it safe. |
| 22 | |
| 23 | You need to append resume=/dev/your_swap_partition to kernel command |
| 24 | line. Then you suspend by |
| 25 | |
| 26 | echo shutdown > /sys/power/disk; echo disk > /sys/power/state |
| 27 | |
| 28 | . If you feel ACPI works pretty well on your system, you might try |
| 29 | |
| 30 | echo platform > /sys/power/disk; echo disk > /sys/power/state |
| 31 | |
| 32 | |
| 33 | |
| 34 | Article about goals and implementation of Software Suspend for Linux |
| 35 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 36 | Author: Gábor Kuti |
| 37 | Last revised: 2003-10-20 by Pavel Machek |
| 38 | |
| 39 | Idea and goals to achieve |
| 40 | |
| 41 | Nowadays it is common in several laptops that they have a suspend button. It |
| 42 | saves the state of the machine to a filesystem or to a partition and switches |
| 43 | to standby mode. Later resuming the machine the saved state is loaded back to |
| 44 | ram and the machine can continue its work. It has two real benefits. First we |
| 45 | save ourselves the time machine goes down and later boots up, energy costs |
| 46 | are real high when running from batteries. The other gain is that we don't have to |
| 47 | interrupt our programs so processes that are calculating something for a long |
| 48 | time shouldn't need to be written interruptible. |
| 49 | |
| 50 | swsusp saves the state of the machine into active swaps and then reboots or |
| 51 | powerdowns. You must explicitly specify the swap partition to resume from with |
| 52 | ``resume='' kernel option. If signature is found it loads and restores saved |
| 53 | state. If the option ``noresume'' is specified as a boot parameter, it skips |
| 54 | the resuming. |
| 55 | |
| 56 | In the meantime while the system is suspended you should not add/remove any |
| 57 | of the hardware, write to the filesystems, etc. |
| 58 | |
| 59 | Sleep states summary |
| 60 | ==================== |
| 61 | |
| 62 | There are three different interfaces you can use, /proc/acpi should |
| 63 | work like this: |
| 64 | |
| 65 | In a really perfect world: |
| 66 | echo 1 > /proc/acpi/sleep # for standby |
| 67 | echo 2 > /proc/acpi/sleep # for suspend to ram |
| 68 | echo 3 > /proc/acpi/sleep # for suspend to ram, but with more power conservative |
| 69 | echo 4 > /proc/acpi/sleep # for suspend to disk |
| 70 | echo 5 > /proc/acpi/sleep # for shutdown unfriendly the system |
| 71 | |
| 72 | and perhaps |
| 73 | echo 4b > /proc/acpi/sleep # for suspend to disk via s4bios |
| 74 | |
| 75 | Frequently Asked Questions |
| 76 | ========================== |
| 77 | |
| 78 | Q: well, suspending a server is IMHO a really stupid thing, |
| 79 | but... (Diego Zuccato): |
| 80 | |
| 81 | A: You bought new UPS for your server. How do you install it without |
| 82 | bringing machine down? Suspend to disk, rearrange power cables, |
| 83 | resume. |
| 84 | |
| 85 | You have your server on UPS. Power died, and UPS is indicating 30 |
| 86 | seconds to failure. What do you do? Suspend to disk. |
| 87 | |
| 88 | Ethernet card in your server died. You want to replace it. Your |
| 89 | server is not hotplug capable. What do you do? Suspend to disk, |
| 90 | replace ethernet card, resume. If you are fast your users will not |
| 91 | even see broken connections. |
| 92 | |
| 93 | |
| 94 | Q: Maybe I'm missing something, but why don't the regular I/O paths work? |
| 95 | |
| 96 | A: We do use the regular I/O paths. However we cannot restore the data |
| 97 | to its original location as we load it. That would create an |
| 98 | inconsistent kernel state which would certainly result in an oops. |
| 99 | Instead, we load the image into unused memory and then atomically copy |
| 100 | it back to it original location. This implies, of course, a maximum |
| 101 | image size of half the amount of memory. |
| 102 | |
| 103 | There are two solutions to this: |
| 104 | |
| 105 | * require half of memory to be free during suspend. That way you can |
| 106 | read "new" data onto free spots, then cli and copy |
| 107 | |
| 108 | * assume we had special "polling" ide driver that only uses memory |
| 109 | between 0-640KB. That way, I'd have to make sure that 0-640KB is free |
| 110 | during suspending, but otherwise it would work... |
| 111 | |
| 112 | suspend2 shares this fundamental limitation, but does not include user |
| 113 | data and disk caches into "used memory" by saving them in |
| 114 | advance. That means that the limitation goes away in practice. |
| 115 | |
| 116 | Q: Does linux support ACPI S4? |
| 117 | |
| 118 | A: Yes. That's what echo platform > /sys/power/disk does. |
| 119 | |
| 120 | Q: My machine doesn't work with ACPI. How can I use swsusp than ? |
| 121 | |
| 122 | A: Do a reboot() syscall with right parameters. Warning: glibc gets in |
| 123 | its way, so check with strace: |
| 124 | |
| 125 | reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, 0xd000fce2) |
| 126 | |
| 127 | (Thanks to Peter Osterlund:) |
| 128 | |
| 129 | #include <unistd.h> |
| 130 | #include <syscall.h> |
| 131 | |
| 132 | #define LINUX_REBOOT_MAGIC1 0xfee1dead |
| 133 | #define LINUX_REBOOT_MAGIC2 672274793 |
| 134 | #define LINUX_REBOOT_CMD_SW_SUSPEND 0xD000FCE2 |
| 135 | |
| 136 | int main() |
| 137 | { |
| 138 | syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, |
| 139 | LINUX_REBOOT_CMD_SW_SUSPEND, 0); |
| 140 | return 0; |
| 141 | } |
| 142 | |
| 143 | Also /sys/ interface should be still present. |
| 144 | |
| 145 | Q: What is 'suspend2'? |
| 146 | |
| 147 | A: suspend2 is 'Software Suspend 2', a forked implementation of |
| 148 | suspend-to-disk which is available as separate patches for 2.4 and 2.6 |
| 149 | kernels from swsusp.sourceforge.net. It includes support for SMP, 4GB |
| 150 | highmem and preemption. It also has a extensible architecture that |
| 151 | allows for arbitrary transformations on the image (compression, |
| 152 | encryption) and arbitrary backends for writing the image (eg to swap |
| 153 | or an NFS share[Work In Progress]). Questions regarding suspend2 |
| 154 | should be sent to the mailing list available through the suspend2 |
| 155 | website, and not to the Linux Kernel Mailing List. We are working |
| 156 | toward merging suspend2 into the mainline kernel. |
| 157 | |
| 158 | Q: A kernel thread must voluntarily freeze itself (call 'refrigerator'). |
| 159 | I found some kernel threads that don't do it, and they don't freeze |
| 160 | so the system can't sleep. Is this a known behavior? |
| 161 | |
| 162 | A: All such kernel threads need to be fixed, one by one. Select the |
| 163 | place where the thread is safe to be frozen (no kernel semaphores |
| 164 | should be held at that point and it must be safe to sleep there), and |
| 165 | add: |
| 166 | |
| 167 | if (current->flags & PF_FREEZE) |
| 168 | refrigerator(PF_FREEZE); |
| 169 | |
| 170 | If the thread is needed for writing the image to storage, you should |
| 171 | instead set the PF_NOFREEZE process flag when creating the thread. |
| 172 | |
| 173 | |
| 174 | Q: What is the difference between between "platform", "shutdown" and |
| 175 | "firmware" in /sys/power/disk? |
| 176 | |
| 177 | A: |
| 178 | |
| 179 | shutdown: save state in linux, then tell bios to powerdown |
| 180 | |
| 181 | platform: save state in linux, then tell bios to powerdown and blink |
| 182 | "suspended led" |
| 183 | |
| 184 | firmware: tell bios to save state itself [needs BIOS-specific suspend |
| 185 | partition, and has very little to do with swsusp] |
| 186 | |
| 187 | "platform" is actually right thing to do, but "shutdown" is most |
| 188 | reliable. |
| 189 | |
| 190 | Q: I do not understand why you have such strong objections to idea of |
| 191 | selective suspend. |
| 192 | |
| 193 | A: Do selective suspend during runtime power managment, that's okay. But |
| 194 | its useless for suspend-to-disk. (And I do not see how you could use |
| 195 | it for suspend-to-ram, I hope you do not want that). |
| 196 | |
| 197 | Lets see, so you suggest to |
| 198 | |
| 199 | * SUSPEND all but swap device and parents |
| 200 | * Snapshot |
| 201 | * Write image to disk |
| 202 | * SUSPEND swap device and parents |
| 203 | * Powerdown |
| 204 | |
| 205 | Oh no, that does not work, if swap device or its parents uses DMA, |
| 206 | you've corrupted data. You'd have to do |
| 207 | |
| 208 | * SUSPEND all but swap device and parents |
| 209 | * FREEZE swap device and parents |
| 210 | * Snapshot |
| 211 | * UNFREEZE swap device and parents |
| 212 | * Write |
| 213 | * SUSPEND swap device and parents |
| 214 | |
| 215 | Which means that you still need that FREEZE state, and you get more |
| 216 | complicated code. (And I have not yet introduce details like system |
| 217 | devices). |
| 218 | |
| 219 | Q: There don't seem to be any generally useful behavioral |
| 220 | distinctions between SUSPEND and FREEZE. |
| 221 | |
| 222 | A: Doing SUSPEND when you are asked to do FREEZE is always correct, |
| 223 | but it may be unneccessarily slow. If you want USB to stay simple, |
| 224 | slowness may not matter to you. It can always be fixed later. |
| 225 | |
| 226 | For devices like disk it does matter, you do not want to spindown for |
| 227 | FREEZE. |
| 228 | |
| 229 | Q: After resuming, system is paging heavilly, leading to very bad interactivity. |
| 230 | |
| 231 | A: Try running |
| 232 | |
| 233 | cat `cat /proc/[0-9]*/maps | grep / | sed 's:.* /:/:' | sort -u` > /dev/null |
| 234 | |
| 235 | after resume. swapoff -a; swapon -a may also be usefull. |