2 * linux/kernel/power/swsusp.c
4 * This file is to realize architecture-independent
5 * machine suspend feature using pretty near only high-level routines
7 * Copyright (C) 1998-2001 Gabor Kuti <seasons@fornax.hu>
8 * Copyright (C) 1998,2001-2004 Pavel Machek <pavel@suse.cz>
10 * This file is released under the GPLv2.
12 * I'd like to thank the following people for their work:
14 * Pavel Machek <pavel@ucw.cz>:
15 * Modifications, defectiveness pointing, being with me at the very beginning,
16 * suspend to swap space, stop all tasks. Port to 2.4.18-ac and 2.5.17.
18 * Steve Doddi <dirk@loth.demon.co.uk>:
19 * Support the possibility of hardware state restoring.
21 * Raph <grey.havens@earthling.net>:
22 * Support for preserving states of network devices and virtual console
23 * (including X and svgatextmode)
25 * Kurt Garloff <garloff@suse.de>:
26 * Straightened the critical function in order to prevent compilers from
27 * playing tricks with local variables.
29 * Andreas Mohr <a.mohr@mailto.de>
31 * Alex Badea <vampire@go.ro>:
34 * Andreas Steinmetz <ast@domdv.de>:
35 * Added encrypted suspend option
37 * More state savers are welcome. Especially for the scsi layer...
39 * For TODOs,FIXMEs also look in Documentation/power/swsusp.txt
42 #include <linux/module.h>
44 #include <linux/suspend.h>
45 #include <linux/smp_lock.h>
46 #include <linux/file.h>
47 #include <linux/utsname.h>
48 #include <linux/version.h>
49 #include <linux/delay.h>
50 #include <linux/reboot.h>
51 #include <linux/bitops.h>
52 #include <linux/vt_kern.h>
53 #include <linux/kbd_kern.h>
54 #include <linux/keyboard.h>
55 #include <linux/spinlock.h>
56 #include <linux/genhd.h>
57 #include <linux/kernel.h>
58 #include <linux/major.h>
59 #include <linux/swap.h>
61 #include <linux/device.h>
62 #include <linux/buffer_head.h>
63 #include <linux/swapops.h>
64 #include <linux/bootmem.h>
65 #include <linux/syscalls.h>
66 #include <linux/console.h>
67 #include <linux/highmem.h>
68 #include <linux/bio.h>
69 #include <linux/mount.h>
71 #include <asm/uaccess.h>
72 #include <asm/mmu_context.h>
73 #include <asm/pgtable.h>
74 #include <asm/tlbflush.h>
77 #include <linux/random.h>
78 #include <linux/crypto.h>
79 #include <asm/scatterlist.h>
87 /* References to section boundaries */
88 extern const void __nosave_begin, __nosave_end;
90 /* Variables to be preserved over suspend */
91 static int nr_copy_pages_check;
93 extern char resume_file[];
95 /* Local variables that should not be affected by save */
96 static unsigned int nr_copy_pages __nosavedata = 0;
98 /* Suspend pagedir is allocated before final copy, therefore it
99 must be freed after resume
101 Warning: this is evil. There are actually two pagedirs at time of
102 resume. One is "pagedir_save", which is empty frame allocated at
103 time of suspend, that must be freed. Second is "pagedir_nosave",
104 allocated at time of resume, that travels through memory not to
105 collide with anything.
107 Warning: this is even more evil than it seems. Pagedirs this file
108 talks about are completely different from page directories used by
111 suspend_pagedir_t *pagedir_nosave __nosavedata = NULL;
112 static suspend_pagedir_t *pagedir_save;
114 #define SWSUSP_SIG "S1SUSPEND"
116 static struct swsusp_header {
117 char reserved[PAGE_SIZE - 20 - MAXKEY - MAXIV - sizeof(swp_entry_t)];
118 u8 key_iv[MAXKEY+MAXIV];
119 swp_entry_t swsusp_info;
122 } __attribute__((packed, aligned(PAGE_SIZE))) swsusp_header;
124 static struct swsusp_info swsusp_info;
127 * XXX: We try to keep some more pages free so that I/O operations succeed
128 * without paging. Might this be more?
130 #define PAGES_FOR_IO 512
136 /* We memorize in swapfile_used what swap devices are used for suspension */
137 #define SWAPFILE_UNUSED 0
138 #define SWAPFILE_SUSPEND 1 /* This is the suspending device */
139 #define SWAPFILE_IGNORED 2 /* Those are other swap devices ignored for suspension */
141 static unsigned short swapfile_used[MAX_SWAPFILES];
142 static unsigned short root_swap;
144 static int write_page(unsigned long addr, swp_entry_t * loc);
145 static int bio_read_page(pgoff_t page_off, void * page);
147 static u8 key_iv[MAXKEY+MAXIV];
149 #ifdef CONFIG_SWSUSP_ENCRYPT
151 static int crypto_init(int mode, void **mem)
156 struct crypto_tfm *tfm;
158 modemsg = mode ? "suspend not possible" : "resume not possible";
160 tfm = crypto_alloc_tfm(CIPHER, CRYPTO_TFM_MODE_CBC);
162 printk(KERN_ERR "swsusp: no tfm, %s\n", modemsg);
167 if(MAXKEY < crypto_tfm_alg_min_keysize(tfm)) {
168 printk(KERN_ERR "swsusp: key buffer too small, %s\n", modemsg);
174 get_random_bytes(key_iv, MAXKEY+MAXIV);
176 len = crypto_tfm_alg_max_keysize(tfm);
180 if (crypto_cipher_setkey(tfm, key_iv, len)) {
181 printk(KERN_ERR "swsusp: key setup failure, %s\n", modemsg);
182 error = -EKEYREJECTED;
186 len = crypto_tfm_alg_ivsize(tfm);
189 printk(KERN_ERR "swsusp: iv buffer too small, %s\n", modemsg);
194 crypto_cipher_set_iv(tfm, key_iv+MAXKEY, len);
200 fail: crypto_free_tfm(tfm);
204 static __inline__ void crypto_exit(void *mem)
206 crypto_free_tfm((struct crypto_tfm *)mem);
209 static __inline__ int crypto_write(struct pbe *p, void *mem)
212 struct scatterlist src, dst;
214 src.page = virt_to_page(p->address);
216 src.length = PAGE_SIZE;
217 dst.page = virt_to_page((void *)&swsusp_header);
219 dst.length = PAGE_SIZE;
221 error = crypto_cipher_encrypt((struct crypto_tfm *)mem, &dst, &src,
225 error = write_page((unsigned long)&swsusp_header,
230 static __inline__ int crypto_read(struct pbe *p, void *mem)
233 struct scatterlist src, dst;
235 error = bio_read_page(swp_offset(p->swap_address), (void *)p->address);
238 src.length = PAGE_SIZE;
240 dst.length = PAGE_SIZE;
241 src.page = dst.page = virt_to_page((void *)p->address);
243 error = crypto_cipher_decrypt((struct crypto_tfm *)mem, &dst,
249 static __inline__ int crypto_init(int mode, void *mem)
254 static __inline__ void crypto_exit(void *mem)
258 static __inline__ int crypto_write(struct pbe *p, void *mem)
260 return write_page(p->address, &(p->swap_address));
263 static __inline__ int crypto_read(struct pbe *p, void *mem)
265 return bio_read_page(swp_offset(p->swap_address), (void *)p->address);
269 static int mark_swapfiles(swp_entry_t prev)
273 rw_swap_page_sync(READ,
274 swp_entry(root_swap, 0),
275 virt_to_page((unsigned long)&swsusp_header));
276 if (!memcmp("SWAP-SPACE",swsusp_header.sig, 10) ||
277 !memcmp("SWAPSPACE2",swsusp_header.sig, 10)) {
278 memcpy(swsusp_header.orig_sig,swsusp_header.sig, 10);
279 memcpy(swsusp_header.sig,SWSUSP_SIG, 10);
280 memcpy(swsusp_header.key_iv, key_iv, MAXKEY+MAXIV);
281 swsusp_header.swsusp_info = prev;
282 error = rw_swap_page_sync(WRITE,
283 swp_entry(root_swap, 0),
284 virt_to_page((unsigned long)
287 pr_debug("swsusp: Partition is not swap space.\n");
294 * Check whether the swap device is the specified resume
295 * device, irrespective of whether they are specified by
298 * (Thus, device inode aliasing is allowed. You can say /dev/hda4
299 * instead of /dev/ide/host0/bus0/target0/lun0/part4 [if using devfs]
300 * and they'll be considered the same device. This is *necessary* for
301 * devfs, since the resume code can only recognize the form /dev/hda4,
302 * but the suspend code would see the long name.)
304 static int is_resume_device(const struct swap_info_struct *swap_info)
306 struct file *file = swap_info->swap_file;
307 struct inode *inode = file->f_dentry->d_inode;
309 return S_ISBLK(inode->i_mode) &&
310 swsusp_resume_device == MKDEV(imajor(inode), iminor(inode));
313 static int swsusp_swap_check(void) /* This is called before saving image */
317 len=strlen(resume_file);
320 spin_lock(&swap_lock);
321 for (i=0; i<MAX_SWAPFILES; i++) {
322 if (!(swap_info[i].flags & SWP_WRITEOK)) {
323 swapfile_used[i]=SWAPFILE_UNUSED;
326 printk(KERN_WARNING "resume= option should be used to set suspend device" );
327 if (root_swap == 0xFFFF) {
328 swapfile_used[i] = SWAPFILE_SUSPEND;
331 swapfile_used[i] = SWAPFILE_IGNORED;
333 /* we ignore all swap devices that are not the resume_file */
334 if (is_resume_device(&swap_info[i])) {
335 swapfile_used[i] = SWAPFILE_SUSPEND;
338 swapfile_used[i] = SWAPFILE_IGNORED;
343 spin_unlock(&swap_lock);
344 return (root_swap != 0xffff) ? 0 : -ENODEV;
348 * This is called after saving image so modification
349 * will be lost after resume... and that's what we want.
350 * we make the device unusable. A new call to
351 * lock_swapdevices can unlock the devices.
353 static void lock_swapdevices(void)
357 spin_lock(&swap_lock);
358 for (i = 0; i< MAX_SWAPFILES; i++)
359 if (swapfile_used[i] == SWAPFILE_IGNORED) {
360 swap_info[i].flags ^= SWP_WRITEOK;
362 spin_unlock(&swap_lock);
366 * write_swap_page - Write one page to a fresh swap location.
367 * @addr: Address we're writing.
368 * @loc: Place to store the entry we used.
370 * Allocate a new swap entry and 'sync' it. Note we discard -EIO
371 * errors. That is an artifact left over from swsusp. It did not
372 * check the return of rw_swap_page_sync() at all, since most pages
373 * written back to swap would return -EIO.
374 * This is a partial improvement, since we will at least return other
375 * errors, though we need to eventually fix the damn code.
377 static int write_page(unsigned long addr, swp_entry_t * loc)
382 entry = get_swap_page();
383 if (swp_offset(entry) &&
384 swapfile_used[swp_type(entry)] == SWAPFILE_SUSPEND) {
385 error = rw_swap_page_sync(WRITE, entry,
397 * data_free - Free the swap entries used by the saved image.
399 * Walk the list of used swap entries and free each one.
400 * This is only used for cleanup when suspend fails.
402 static void data_free(void)
407 for (i = 0; i < nr_copy_pages; i++) {
408 entry = (pagedir_nosave + i)->swap_address;
413 (pagedir_nosave + i)->swap_address = (swp_entry_t){0};
418 * data_write - Write saved image to swap.
420 * Walk the list of pages in the image and sync each one to swap.
422 static int data_write(void)
424 int error = 0, i = 0;
425 unsigned int mod = nr_copy_pages / 100;
429 if ((error = crypto_init(1, &tfm)))
435 printk( "Writing data to swap (%d pages)... ", nr_copy_pages );
436 for_each_pbe (p, pagedir_nosave) {
438 printk( "\b\b\b\b%3d%%", i / mod );
439 if ((error = crypto_write(p, tfm))) {
445 printk("\b\b\b\bdone\n");
450 static void dump_info(void)
452 pr_debug(" swsusp: Version: %u\n",swsusp_info.version_code);
453 pr_debug(" swsusp: Num Pages: %ld\n",swsusp_info.num_physpages);
454 pr_debug(" swsusp: UTS Sys: %s\n",swsusp_info.uts.sysname);
455 pr_debug(" swsusp: UTS Node: %s\n",swsusp_info.uts.nodename);
456 pr_debug(" swsusp: UTS Release: %s\n",swsusp_info.uts.release);
457 pr_debug(" swsusp: UTS Version: %s\n",swsusp_info.uts.version);
458 pr_debug(" swsusp: UTS Machine: %s\n",swsusp_info.uts.machine);
459 pr_debug(" swsusp: UTS Domain: %s\n",swsusp_info.uts.domainname);
460 pr_debug(" swsusp: CPUs: %d\n",swsusp_info.cpus);
461 pr_debug(" swsusp: Image: %ld Pages\n",swsusp_info.image_pages);
462 pr_debug(" swsusp: Pagedir: %ld Pages\n",swsusp_info.pagedir_pages);
465 static void init_header(void)
467 memset(&swsusp_info, 0, sizeof(swsusp_info));
468 swsusp_info.version_code = LINUX_VERSION_CODE;
469 swsusp_info.num_physpages = num_physpages;
470 memcpy(&swsusp_info.uts, &system_utsname, sizeof(system_utsname));
472 swsusp_info.suspend_pagedir = pagedir_nosave;
473 swsusp_info.cpus = num_online_cpus();
474 swsusp_info.image_pages = nr_copy_pages;
477 static int close_swap(void)
483 error = write_page((unsigned long)&swsusp_info, &entry);
486 error = mark_swapfiles(entry);
493 * free_pagedir_entries - Free pages used by the page directory.
495 * This is used during suspend for error recovery.
498 static void free_pagedir_entries(void)
502 for (i = 0; i < swsusp_info.pagedir_pages; i++)
503 swap_free(swsusp_info.pagedir[i]);
508 * write_pagedir - Write the array of pages holding the page directory.
509 * @last: Last swap entry we write (needed for header).
512 static int write_pagedir(void)
518 printk( "Writing pagedir...");
519 for_each_pb_page (pbe, pagedir_nosave) {
520 if ((error = write_page((unsigned long)pbe, &swsusp_info.pagedir[n++])))
524 swsusp_info.pagedir_pages = n;
525 printk("done (%u pages)\n", n);
530 * write_suspend_image - Write entire image and metadata.
533 static int write_suspend_image(void)
538 if ((error = data_write()))
541 if ((error = write_pagedir()))
544 if ((error = close_swap()))
547 memset(key_iv, 0, MAXKEY+MAXIV);
550 free_pagedir_entries();
557 #ifdef CONFIG_HIGHMEM
558 struct highmem_page {
561 struct highmem_page *next;
564 static struct highmem_page *highmem_copy;
566 static int save_highmem_zone(struct zone *zone)
568 unsigned long zone_pfn;
569 mark_free_pages(zone);
570 for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
572 struct highmem_page *save;
574 unsigned long pfn = zone_pfn + zone->zone_start_pfn;
580 page = pfn_to_page(pfn);
582 * This condition results from rvmalloc() sans vmalloc_32()
583 * and architectural memory reservations. This should be
584 * corrected eventually when the cases giving rise to this
585 * are better understood.
587 if (PageReserved(page)) {
588 printk("highmem reserved page?!\n");
591 BUG_ON(PageNosave(page));
592 if (PageNosaveFree(page))
594 save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC);
597 save->next = highmem_copy;
599 save->data = (void *) get_zeroed_page(GFP_ATOMIC);
604 kaddr = kmap_atomic(page, KM_USER0);
605 memcpy(save->data, kaddr, PAGE_SIZE);
606 kunmap_atomic(kaddr, KM_USER0);
611 #endif /* CONFIG_HIGHMEM */
614 static int save_highmem(void)
616 #ifdef CONFIG_HIGHMEM
620 pr_debug("swsusp: Saving Highmem\n");
621 for_each_zone (zone) {
622 if (is_highmem(zone))
623 res = save_highmem_zone(zone);
631 static int restore_highmem(void)
633 #ifdef CONFIG_HIGHMEM
634 printk("swsusp: Restoring Highmem\n");
635 while (highmem_copy) {
636 struct highmem_page *save = highmem_copy;
638 highmem_copy = save->next;
640 kaddr = kmap_atomic(save->page, KM_USER0);
641 memcpy(kaddr, save->data, PAGE_SIZE);
642 kunmap_atomic(kaddr, KM_USER0);
643 free_page((long) save->data);
651 static int pfn_is_nosave(unsigned long pfn)
653 unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT;
654 unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT;
655 return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn);
659 * saveable - Determine whether a page should be cloned or not.
662 * We save a page if it's Reserved, and not in the range of pages
663 * statically defined as 'unsaveable', or if it isn't reserved, and
664 * isn't part of a free chunk of pages.
667 static int saveable(struct zone * zone, unsigned long * zone_pfn)
669 unsigned long pfn = *zone_pfn + zone->zone_start_pfn;
675 page = pfn_to_page(pfn);
676 BUG_ON(PageReserved(page) && PageNosave(page));
677 if (PageNosave(page))
679 if (PageReserved(page) && pfn_is_nosave(pfn)) {
680 pr_debug("[nosave pfn 0x%lx]", pfn);
683 if (PageNosaveFree(page))
689 static void count_data_pages(void)
692 unsigned long zone_pfn;
696 for_each_zone (zone) {
697 if (is_highmem(zone))
699 mark_free_pages(zone);
700 for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
701 nr_copy_pages += saveable(zone, &zone_pfn);
706 static void copy_data_pages(void)
709 unsigned long zone_pfn;
710 struct pbe * pbe = pagedir_nosave;
712 pr_debug("copy_data_pages(): pages to copy: %d\n", nr_copy_pages);
713 for_each_zone (zone) {
714 if (is_highmem(zone))
716 mark_free_pages(zone);
717 for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
718 if (saveable(zone, &zone_pfn)) {
720 page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
722 pbe->orig_address = (long) page_address(page);
723 /* copy_page is not usable for copying task structs. */
724 memcpy((void *)pbe->address, (void *)pbe->orig_address, PAGE_SIZE);
734 * calc_nr - Determine the number of pages needed for a pbe list.
737 static int calc_nr(int nr_copy)
739 return nr_copy + (nr_copy+PBES_PER_PAGE-2)/(PBES_PER_PAGE-1);
743 * free_pagedir - free pages allocated with alloc_pagedir()
746 static inline void free_pagedir(struct pbe *pblist)
751 pbe = (pblist + PB_PAGE_SKIP)->next;
752 free_page((unsigned long)pblist);
758 * fill_pb_page - Create a list of PBEs on a given memory page
761 static inline void fill_pb_page(struct pbe *pbpage)
766 pbpage += PB_PAGE_SKIP;
769 while (++p < pbpage);
773 * create_pbe_list - Create a list of PBEs on top of a given chain
774 * of memory pages allocated with alloc_pagedir()
777 static void create_pbe_list(struct pbe *pblist, unsigned nr_pages)
779 struct pbe *pbpage, *p;
780 unsigned num = PBES_PER_PAGE;
782 for_each_pb_page (pbpage, pblist) {
786 fill_pb_page(pbpage);
787 num += PBES_PER_PAGE;
790 for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++)
794 pr_debug("create_pbe_list(): initialized %d PBEs\n", num);
798 * alloc_pagedir - Allocate the page directory.
800 * First, determine exactly how many pages we need and
803 * We arrange the pages in a chain: each page is an array of PBES_PER_PAGE
804 * struct pbe elements (pbes) and the last element in the page points
807 * On each page we set up a list of struct_pbe elements.
810 static struct pbe * alloc_pagedir(unsigned nr_pages)
813 struct pbe *pblist, *pbe;
818 pr_debug("alloc_pagedir(): nr_pages = %d\n", nr_pages);
819 pblist = (struct pbe *)get_zeroed_page(GFP_ATOMIC | __GFP_COLD);
820 for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages;
821 pbe = pbe->next, num += PBES_PER_PAGE) {
823 pbe->next = (struct pbe *)get_zeroed_page(GFP_ATOMIC | __GFP_COLD);
825 if (!pbe) { /* get_zeroed_page() failed */
826 free_pagedir(pblist);
833 * free_image_pages - Free pages allocated for snapshot
836 static void free_image_pages(void)
840 for_each_pbe (p, pagedir_save) {
842 ClearPageNosave(virt_to_page(p->address));
843 free_page(p->address);
850 * alloc_image_pages - Allocate pages for the snapshot.
853 static int alloc_image_pages(void)
857 for_each_pbe (p, pagedir_save) {
858 p->address = get_zeroed_page(GFP_ATOMIC | __GFP_COLD);
861 SetPageNosave(virt_to_page(p->address));
866 void swsusp_free(void)
868 BUG_ON(PageNosave(virt_to_page(pagedir_save)));
869 BUG_ON(PageNosaveFree(virt_to_page(pagedir_save)));
871 free_pagedir(pagedir_save);
876 * enough_free_mem - Make sure we enough free memory to snapshot.
878 * Returns TRUE or FALSE after checking the number of available
882 static int enough_free_mem(void)
884 if (nr_free_pages() < (nr_copy_pages + PAGES_FOR_IO)) {
885 pr_debug("swsusp: Not enough free pages: Have %d\n",
894 * enough_swap - Make sure we have enough swap to save the image.
896 * Returns TRUE or FALSE after checking the total amount of swap
899 * FIXME: si_swapinfo(&i) returns all swap devices information.
900 * We should only consider resume_device.
903 static int enough_swap(void)
908 if (i.freeswap < (nr_copy_pages + PAGES_FOR_IO)) {
909 pr_debug("swsusp: Not enough swap. Need %ld\n",i.freeswap);
915 static int swsusp_alloc(void)
919 pagedir_nosave = NULL;
920 nr_copy_pages = calc_nr(nr_copy_pages);
922 pr_debug("suspend: (pages needed: %d + %d free: %d)\n",
923 nr_copy_pages, PAGES_FOR_IO, nr_free_pages());
925 if (!enough_free_mem())
931 if (!(pagedir_save = alloc_pagedir(nr_copy_pages))) {
932 printk(KERN_ERR "suspend: Allocating pagedir failed.\n");
935 create_pbe_list(pagedir_save, nr_copy_pages);
936 pagedir_nosave = pagedir_save;
937 if ((error = alloc_image_pages())) {
938 printk(KERN_ERR "suspend: Allocating image pages failed.\n");
943 nr_copy_pages_check = nr_copy_pages;
947 static int suspend_prepare_image(void)
951 pr_debug("swsusp: critical section: \n");
952 if (save_highmem()) {
953 printk(KERN_CRIT "Suspend machine: Not enough free pages for highmem\n");
960 printk("swsusp: Need to copy %u pages\n", nr_copy_pages);
962 error = swsusp_alloc();
966 /* During allocating of suspend pagedir, new cold pages may appear.
973 * End of critical section. From now on, we can write to memory,
974 * but we should not touch disk. This specially means we must _not_
975 * touch swap space! Except we must write out our image of course.
978 printk("swsusp: critical section/: done (%d pages copied)\n", nr_copy_pages );
983 /* It is important _NOT_ to umount filesystems at this point. We want
984 * them synced (in case something goes wrong) but we DO not want to mark
985 * filesystem clean: it is not. (And it does not matter, if we resume
986 * correctly, we'll mark system clean, anyway.)
988 int swsusp_write(void)
993 error = write_suspend_image();
994 /* This will unlock ignored swap devices since writing is finished */
1001 extern asmlinkage int swsusp_arch_suspend(void);
1002 extern asmlinkage int swsusp_arch_resume(void);
1005 asmlinkage int swsusp_save(void)
1007 return suspend_prepare_image();
1010 int swsusp_suspend(void)
1013 if ((error = arch_prepare_suspend()))
1015 local_irq_disable();
1016 /* At this point, device_suspend() has been called, but *not*
1017 * device_power_down(). We *must* device_power_down() now.
1018 * Otherwise, drivers for some devices (e.g. interrupt controllers)
1019 * become desynchronized with the actual state of the hardware
1020 * at resume time, and evil weirdness ensues.
1022 if ((error = device_power_down(PMSG_FREEZE))) {
1023 printk(KERN_ERR "Some devices failed to power down, aborting suspend\n");
1028 if ((error = swsusp_swap_check())) {
1029 printk(KERN_ERR "swsusp: cannot find swap device, try swapon -a.\n");
1035 save_processor_state();
1036 if ((error = swsusp_arch_suspend()))
1037 printk(KERN_ERR "Error %d suspending\n", error);
1038 /* Restore control flow magically appears here */
1039 restore_processor_state();
1040 BUG_ON (nr_copy_pages_check != nr_copy_pages);
1047 int swsusp_resume(void)
1050 local_irq_disable();
1051 if (device_power_down(PMSG_FREEZE))
1052 printk(KERN_ERR "Some devices failed to power down, very bad\n");
1053 /* We'll ignore saved state, but this gets preempt count (etc) right */
1054 save_processor_state();
1055 error = swsusp_arch_resume();
1056 /* Code below is only ever reached in case of failure. Otherwise
1057 * execution continues at place where swsusp_arch_suspend was called
1060 restore_processor_state();
1068 * On resume, for storing the PBE list and the image,
1069 * we can only use memory pages that do not conflict with the pages
1070 * which had been used before suspend.
1072 * We don't know which pages are usable until we allocate them.
1074 * Allocated but unusable (ie eaten) memory pages are linked together
1075 * to create a list, so that we can free them easily
1077 * We could have used a type other than (void *)
1078 * for this purpose, but ...
1080 static void **eaten_memory = NULL;
1082 static inline void eat_page(void *page)
1087 eaten_memory = page;
1091 static unsigned long get_usable_page(unsigned gfp_mask)
1095 m = get_zeroed_page(gfp_mask);
1096 while (!PageNosaveFree(virt_to_page(m))) {
1097 eat_page((void *)m);
1098 m = get_zeroed_page(gfp_mask);
1105 static void free_eaten_memory(void)
1113 m = (unsigned long)c;
1118 eaten_memory = NULL;
1119 pr_debug("swsusp: %d unused pages freed\n", i);
1123 * check_pagedir - We ensure here that pages that the PBEs point to
1124 * won't collide with pages where we're going to restore from the loaded
1128 static int check_pagedir(struct pbe *pblist)
1132 /* This is necessary, so that we can free allocated pages
1133 * in case of failure
1135 for_each_pbe (p, pblist)
1138 for_each_pbe (p, pblist) {
1139 p->address = get_usable_page(GFP_ATOMIC);
1147 * swsusp_pagedir_relocate - It is possible, that some memory pages
1148 * occupied by the list of PBEs collide with pages where we're going to
1149 * restore from the loaded pages later. We relocate them here.
1152 static struct pbe * swsusp_pagedir_relocate(struct pbe *pblist)
1155 unsigned long zone_pfn;
1156 struct pbe *pbpage, *tail, *p;
1158 int rel = 0, error = 0;
1160 if (!pblist) /* a sanity check */
1163 pr_debug("swsusp: Relocating pagedir (%lu pages to check)\n",
1164 swsusp_info.pagedir_pages);
1166 /* Set page flags */
1168 for_each_zone (zone) {
1169 for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
1170 SetPageNosaveFree(pfn_to_page(zone_pfn +
1171 zone->zone_start_pfn));
1174 /* Clear orig addresses */
1176 for_each_pbe (p, pblist)
1177 ClearPageNosaveFree(virt_to_page(p->orig_address));
1179 tail = pblist + PB_PAGE_SKIP;
1181 /* Relocate colliding pages */
1183 for_each_pb_page (pbpage, pblist) {
1184 if (!PageNosaveFree(virt_to_page((unsigned long)pbpage))) {
1185 m = (void *)get_usable_page(GFP_ATOMIC | __GFP_COLD);
1190 memcpy(m, (void *)pbpage, PAGE_SIZE);
1191 if (pbpage == pblist)
1192 pblist = (struct pbe *)m;
1194 tail->next = (struct pbe *)m;
1196 eat_page((void *)pbpage);
1197 pbpage = (struct pbe *)m;
1199 /* We have to link the PBEs again */
1201 for (p = pbpage; p < pbpage + PB_PAGE_SKIP; p++)
1202 if (p->next) /* needed to save the end */
1207 tail = pbpage + PB_PAGE_SKIP;
1211 printk("\nswsusp: Out of memory\n\n");
1212 free_pagedir(pblist);
1213 free_eaten_memory();
1217 printk("swsusp: Relocated %d pages\n", rel);
1223 * Using bio to read from swap.
1224 * This code requires a bit more work than just using buffer heads
1225 * but, it is the recommended way for 2.5/2.6.
1226 * The following are to signal the beginning and end of I/O. Bios
1227 * finish asynchronously, while we want them to happen synchronously.
1228 * A simple atomic_t, and a wait loop take care of this problem.
1231 static atomic_t io_done = ATOMIC_INIT(0);
1233 static int end_io(struct bio * bio, unsigned int num, int err)
1235 if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
1236 panic("I/O error reading memory image");
1237 atomic_set(&io_done, 0);
1241 static struct block_device * resume_bdev;
1244 * submit - submit BIO request.
1245 * @rw: READ or WRITE.
1246 * @off physical offset of page.
1247 * @page: page we're reading or writing.
1249 * Straight from the textbook - allocate and initialize the bio.
1250 * If we're writing, make sure the page is marked as dirty.
1251 * Then submit it and wait.
1254 static int submit(int rw, pgoff_t page_off, void * page)
1259 bio = bio_alloc(GFP_ATOMIC, 1);
1262 bio->bi_sector = page_off * (PAGE_SIZE >> 9);
1264 bio->bi_bdev = resume_bdev;
1265 bio->bi_end_io = end_io;
1267 if (bio_add_page(bio, virt_to_page(page), PAGE_SIZE, 0) < PAGE_SIZE) {
1268 printk("swsusp: ERROR: adding page to bio at %ld\n",page_off);
1274 bio_set_pages_dirty(bio);
1276 atomic_set(&io_done, 1);
1277 submit_bio(rw | (1 << BIO_RW_SYNC), bio);
1278 while (atomic_read(&io_done))
1286 static int bio_read_page(pgoff_t page_off, void * page)
1288 return submit(READ, page_off, page);
1291 static int bio_write_page(pgoff_t page_off, void * page)
1293 return submit(WRITE, page_off, page);
1297 * Sanity check if this image makes sense with this kernel/swap context
1298 * I really don't think that it's foolproof but more than nothing..
1301 static const char * sanity_check(void)
1304 if (swsusp_info.version_code != LINUX_VERSION_CODE)
1305 return "kernel version";
1306 if (swsusp_info.num_physpages != num_physpages)
1307 return "memory size";
1308 if (strcmp(swsusp_info.uts.sysname,system_utsname.sysname))
1309 return "system type";
1310 if (strcmp(swsusp_info.uts.release,system_utsname.release))
1311 return "kernel release";
1312 if (strcmp(swsusp_info.uts.version,system_utsname.version))
1314 if (strcmp(swsusp_info.uts.machine,system_utsname.machine))
1317 /* We can't use number of online CPUs when we use hotplug to remove them ;-))) */
1318 if (swsusp_info.cpus != num_possible_cpus())
1319 return "number of cpus";
1325 static int check_header(void)
1327 const char * reason = NULL;
1330 if ((error = bio_read_page(swp_offset(swsusp_header.swsusp_info), &swsusp_info)))
1333 /* Is this same machine? */
1334 if ((reason = sanity_check())) {
1335 printk(KERN_ERR "swsusp: Resume mismatch: %s\n",reason);
1338 nr_copy_pages = swsusp_info.image_pages;
1342 static int check_sig(void)
1346 memset(&swsusp_header, 0, sizeof(swsusp_header));
1347 if ((error = bio_read_page(0, &swsusp_header)))
1349 if (!memcmp(SWSUSP_SIG, swsusp_header.sig, 10)) {
1350 memcpy(swsusp_header.sig, swsusp_header.orig_sig, 10);
1351 memcpy(key_iv, swsusp_header.key_iv, MAXKEY+MAXIV);
1352 memset(swsusp_header.key_iv, 0, MAXKEY+MAXIV);
1355 * Reset swap signature now.
1357 error = bio_write_page(0, &swsusp_header);
1362 pr_debug("swsusp: Signature found, resuming\n");
1367 * data_read - Read image pages from swap.
1369 * You do not need to check for overlaps, check_pagedir()
1373 static int data_read(struct pbe *pblist)
1378 int mod = swsusp_info.image_pages / 100;
1381 if ((error = crypto_init(0, &tfm)))
1387 printk("swsusp: Reading image data (%lu pages): ",
1388 swsusp_info.image_pages);
1390 for_each_pbe (p, pblist) {
1392 printk("\b\b\b\b%3d%%", i / mod);
1394 if ((error = crypto_read(p, tfm))) {
1401 printk("\b\b\b\bdone\n");
1407 * read_pagedir - Read page backup list pages from swap
1410 static int read_pagedir(struct pbe *pblist)
1412 struct pbe *pbpage, *p;
1419 printk("swsusp: Reading pagedir (%lu pages)\n",
1420 swsusp_info.pagedir_pages);
1422 for_each_pb_page (pbpage, pblist) {
1423 unsigned long offset = swp_offset(swsusp_info.pagedir[i++]);
1427 p = (pbpage + PB_PAGE_SKIP)->next;
1428 error = bio_read_page(offset, (void *)pbpage);
1429 (pbpage + PB_PAGE_SKIP)->next = p;
1436 free_page((unsigned long)pblist);
1438 BUG_ON(i != swsusp_info.pagedir_pages);
1444 static int check_suspend_image(void)
1448 if ((error = check_sig()))
1451 if ((error = check_header()))
1457 static int read_suspend_image(void)
1462 if (!(p = alloc_pagedir(nr_copy_pages)))
1465 if ((error = read_pagedir(p)))
1468 create_pbe_list(p, nr_copy_pages);
1470 if (!(pagedir_nosave = swsusp_pagedir_relocate(p)))
1473 /* Allocate memory for the image and read the data from swap */
1475 error = check_pagedir(pagedir_nosave);
1476 free_eaten_memory();
1478 error = data_read(pagedir_nosave);
1480 if (error) { /* We fail cleanly */
1481 for_each_pbe (p, pagedir_nosave)
1483 free_page(p->address);
1486 free_pagedir(pagedir_nosave);
1492 * swsusp_check - Check for saved image in swap
1495 int swsusp_check(void)
1499 resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ);
1500 if (!IS_ERR(resume_bdev)) {
1501 set_blocksize(resume_bdev, PAGE_SIZE);
1502 error = check_suspend_image();
1504 blkdev_put(resume_bdev);
1506 error = PTR_ERR(resume_bdev);
1509 pr_debug("swsusp: resume file found\n");
1511 pr_debug("swsusp: Error %d check for resume file\n", error);
1516 * swsusp_read - Read saved image from swap.
1519 int swsusp_read(void)
1523 if (IS_ERR(resume_bdev)) {
1524 pr_debug("swsusp: block device not initialised\n");
1525 return PTR_ERR(resume_bdev);
1528 error = read_suspend_image();
1529 blkdev_put(resume_bdev);
1530 memset(key_iv, 0, MAXKEY+MAXIV);
1533 pr_debug("swsusp: Reading resume file was successful\n");
1535 pr_debug("swsusp: Error %d resuming\n", error);
1540 * swsusp_close - close swap device.
1543 void swsusp_close(void)
1545 if (IS_ERR(resume_bdev)) {
1546 pr_debug("swsusp: block device not initialised\n");
1550 blkdev_put(resume_bdev);