2 * linux/kernel/power/snapshot.c
4 * This file provides system snapshot/restore functionality for swsusp.
6 * Copyright (C) 1998-2005 Pavel Machek <pavel@suse.cz>
7 * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
9 * This file is released under the GPLv2.
13 #include <linux/version.h>
14 #include <linux/module.h>
16 #include <linux/suspend.h>
17 #include <linux/delay.h>
18 #include <linux/bitops.h>
19 #include <linux/spinlock.h>
20 #include <linux/kernel.h>
22 #include <linux/device.h>
23 #include <linux/init.h>
24 #include <linux/bootmem.h>
25 #include <linux/syscalls.h>
26 #include <linux/console.h>
27 #include <linux/highmem.h>
29 #include <asm/uaccess.h>
30 #include <asm/mmu_context.h>
31 #include <asm/pgtable.h>
32 #include <asm/tlbflush.h>
37 static int swsusp_page_is_free(struct page *);
38 static void swsusp_set_page_forbidden(struct page *);
39 static void swsusp_unset_page_forbidden(struct page *);
41 /* List of PBEs needed for restoring the pages that were allocated before
42 * the suspend and included in the suspend image, but have also been
43 * allocated by the "resume" kernel, so their contents cannot be written
44 * directly to their "original" page frames.
46 struct pbe *restore_pblist;
48 /* Pointer to an auxiliary buffer (1 page) */
52 * @safe_needed - on resume, for storing the PBE list and the image,
53 * we can only use memory pages that do not conflict with the pages
54 * used before suspend. The unsafe pages have PageNosaveFree set
55 * and we count them using unsafe_pages.
57 * Each allocated image page is marked as PageNosave and PageNosaveFree
58 * so that swsusp_free() can release it.
63 #define PG_UNSAFE_CLEAR 1
64 #define PG_UNSAFE_KEEP 0
66 static unsigned int allocated_unsafe_pages;
68 static void *get_image_page(gfp_t gfp_mask, int safe_needed)
72 res = (void *)get_zeroed_page(gfp_mask);
74 while (res && swsusp_page_is_free(virt_to_page(res))) {
75 /* The page is unsafe, mark it for swsusp_free() */
76 swsusp_set_page_forbidden(virt_to_page(res));
77 allocated_unsafe_pages++;
78 res = (void *)get_zeroed_page(gfp_mask);
81 swsusp_set_page_forbidden(virt_to_page(res));
82 swsusp_set_page_free(virt_to_page(res));
87 unsigned long get_safe_page(gfp_t gfp_mask)
89 return (unsigned long)get_image_page(gfp_mask, PG_SAFE);
92 static struct page *alloc_image_page(gfp_t gfp_mask)
96 page = alloc_page(gfp_mask);
98 swsusp_set_page_forbidden(page);
99 swsusp_set_page_free(page);
105 * free_image_page - free page represented by @addr, allocated with
106 * get_image_page (page flags set by it must be cleared)
109 static inline void free_image_page(void *addr, int clear_nosave_free)
113 BUG_ON(!virt_addr_valid(addr));
115 page = virt_to_page(addr);
117 swsusp_unset_page_forbidden(page);
118 if (clear_nosave_free)
119 swsusp_unset_page_free(page);
124 /* struct linked_page is used to build chains of pages */
126 #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *))
129 struct linked_page *next;
130 char data[LINKED_PAGE_DATA_SIZE];
131 } __attribute__((packed));
134 free_list_of_pages(struct linked_page *list, int clear_page_nosave)
137 struct linked_page *lp = list->next;
139 free_image_page(list, clear_page_nosave);
145 * struct chain_allocator is used for allocating small objects out of
146 * a linked list of pages called 'the chain'.
148 * The chain grows each time when there is no room for a new object in
149 * the current page. The allocated objects cannot be freed individually.
150 * It is only possible to free them all at once, by freeing the entire
153 * NOTE: The chain allocator may be inefficient if the allocated objects
154 * are not much smaller than PAGE_SIZE.
157 struct chain_allocator {
158 struct linked_page *chain; /* the chain */
159 unsigned int used_space; /* total size of objects allocated out
160 * of the current page
162 gfp_t gfp_mask; /* mask for allocating pages */
163 int safe_needed; /* if set, only "safe" pages are allocated */
167 chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed)
170 ca->used_space = LINKED_PAGE_DATA_SIZE;
171 ca->gfp_mask = gfp_mask;
172 ca->safe_needed = safe_needed;
175 static void *chain_alloc(struct chain_allocator *ca, unsigned int size)
179 if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) {
180 struct linked_page *lp;
182 lp = get_image_page(ca->gfp_mask, ca->safe_needed);
186 lp->next = ca->chain;
190 ret = ca->chain->data + ca->used_space;
191 ca->used_space += size;
195 static void chain_free(struct chain_allocator *ca, int clear_page_nosave)
197 free_list_of_pages(ca->chain, clear_page_nosave);
198 memset(ca, 0, sizeof(struct chain_allocator));
202 * Data types related to memory bitmaps.
204 * Memory bitmap is a structure consiting of many linked lists of
205 * objects. The main list's elements are of type struct zone_bitmap
206 * and each of them corresonds to one zone. For each zone bitmap
207 * object there is a list of objects of type struct bm_block that
208 * represent each blocks of bit chunks in which information is
211 * struct memory_bitmap contains a pointer to the main list of zone
212 * bitmap objects, a struct bm_position used for browsing the bitmap,
213 * and a pointer to the list of pages used for allocating all of the
214 * zone bitmap objects and bitmap block objects.
216 * NOTE: It has to be possible to lay out the bitmap in memory
217 * using only allocations of order 0. Additionally, the bitmap is
218 * designed to work with arbitrary number of zones (this is over the
219 * top for now, but let's avoid making unnecessary assumptions ;-).
221 * struct zone_bitmap contains a pointer to a list of bitmap block
222 * objects and a pointer to the bitmap block object that has been
223 * most recently used for setting bits. Additionally, it contains the
224 * pfns that correspond to the start and end of the represented zone.
226 * struct bm_block contains a pointer to the memory page in which
227 * information is stored (in the form of a block of bit chunks
228 * of type unsigned long each). It also contains the pfns that
229 * correspond to the start and end of the represented memory area and
230 * the number of bit chunks in the block.
233 #define BM_END_OF_MAP (~0UL)
235 #define BM_CHUNKS_PER_BLOCK (PAGE_SIZE / sizeof(long))
236 #define BM_BITS_PER_CHUNK (sizeof(long) << 3)
237 #define BM_BITS_PER_BLOCK (PAGE_SIZE << 3)
240 struct bm_block *next; /* next element of the list */
241 unsigned long start_pfn; /* pfn represented by the first bit */
242 unsigned long end_pfn; /* pfn represented by the last bit plus 1 */
243 unsigned int size; /* number of bit chunks */
244 unsigned long *data; /* chunks of bits representing pages */
248 struct zone_bitmap *next; /* next element of the list */
249 unsigned long start_pfn; /* minimal pfn in this zone */
250 unsigned long end_pfn; /* maximal pfn in this zone plus 1 */
251 struct bm_block *bm_blocks; /* list of bitmap blocks */
252 struct bm_block *cur_block; /* recently used bitmap block */
255 /* strcut bm_position is used for browsing memory bitmaps */
258 struct zone_bitmap *zone_bm;
259 struct bm_block *block;
264 struct memory_bitmap {
265 struct zone_bitmap *zone_bm_list; /* list of zone bitmaps */
266 struct linked_page *p_list; /* list of pages used to store zone
267 * bitmap objects and bitmap block
270 struct bm_position cur; /* most recently used bit position */
273 /* Functions that operate on memory bitmaps */
275 static inline void memory_bm_reset_chunk(struct memory_bitmap *bm)
281 static void memory_bm_position_reset(struct memory_bitmap *bm)
283 struct zone_bitmap *zone_bm;
285 zone_bm = bm->zone_bm_list;
286 bm->cur.zone_bm = zone_bm;
287 bm->cur.block = zone_bm->bm_blocks;
288 memory_bm_reset_chunk(bm);
291 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
294 * create_bm_block_list - create a list of block bitmap objects
297 static inline struct bm_block *
298 create_bm_block_list(unsigned int nr_blocks, struct chain_allocator *ca)
300 struct bm_block *bblist = NULL;
302 while (nr_blocks-- > 0) {
305 bb = chain_alloc(ca, sizeof(struct bm_block));
316 * create_zone_bm_list - create a list of zone bitmap objects
319 static inline struct zone_bitmap *
320 create_zone_bm_list(unsigned int nr_zones, struct chain_allocator *ca)
322 struct zone_bitmap *zbmlist = NULL;
324 while (nr_zones-- > 0) {
325 struct zone_bitmap *zbm;
327 zbm = chain_alloc(ca, sizeof(struct zone_bitmap));
338 * memory_bm_create - allocate memory for a memory bitmap
342 memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed)
344 struct chain_allocator ca;
346 struct zone_bitmap *zone_bm;
350 chain_init(&ca, gfp_mask, safe_needed);
352 /* Compute the number of zones */
355 if (populated_zone(zone))
358 /* Allocate the list of zones bitmap objects */
359 zone_bm = create_zone_bm_list(nr, &ca);
360 bm->zone_bm_list = zone_bm;
362 chain_free(&ca, PG_UNSAFE_CLEAR);
366 /* Initialize the zone bitmap objects */
367 for_each_zone(zone) {
370 if (!populated_zone(zone))
373 zone_bm->start_pfn = zone->zone_start_pfn;
374 zone_bm->end_pfn = zone->zone_start_pfn + zone->spanned_pages;
375 /* Allocate the list of bitmap block objects */
376 nr = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
377 bb = create_bm_block_list(nr, &ca);
378 zone_bm->bm_blocks = bb;
379 zone_bm->cur_block = bb;
383 nr = zone->spanned_pages;
384 pfn = zone->zone_start_pfn;
385 /* Initialize the bitmap block objects */
389 ptr = get_image_page(gfp_mask, safe_needed);
395 if (nr >= BM_BITS_PER_BLOCK) {
396 pfn += BM_BITS_PER_BLOCK;
397 bb->size = BM_CHUNKS_PER_BLOCK;
398 nr -= BM_BITS_PER_BLOCK;
400 /* This is executed only once in the loop */
402 bb->size = DIV_ROUND_UP(nr, BM_BITS_PER_CHUNK);
407 zone_bm = zone_bm->next;
409 bm->p_list = ca.chain;
410 memory_bm_position_reset(bm);
414 bm->p_list = ca.chain;
415 memory_bm_free(bm, PG_UNSAFE_CLEAR);
420 * memory_bm_free - free memory occupied by the memory bitmap @bm
423 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free)
425 struct zone_bitmap *zone_bm;
427 /* Free the list of bit blocks for each zone_bitmap object */
428 zone_bm = bm->zone_bm_list;
432 bb = zone_bm->bm_blocks;
435 free_image_page(bb->data, clear_nosave_free);
438 zone_bm = zone_bm->next;
440 free_list_of_pages(bm->p_list, clear_nosave_free);
441 bm->zone_bm_list = NULL;
445 * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds
446 * to given pfn. The cur_zone_bm member of @bm and the cur_block member
447 * of @bm->cur_zone_bm are updated.
450 static void memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
451 void **addr, unsigned int *bit_nr)
453 struct zone_bitmap *zone_bm;
456 /* Check if the pfn is from the current zone */
457 zone_bm = bm->cur.zone_bm;
458 if (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) {
459 zone_bm = bm->zone_bm_list;
460 /* We don't assume that the zones are sorted by pfns */
461 while (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) {
462 zone_bm = zone_bm->next;
466 bm->cur.zone_bm = zone_bm;
468 /* Check if the pfn corresponds to the current bitmap block */
469 bb = zone_bm->cur_block;
470 if (pfn < bb->start_pfn)
471 bb = zone_bm->bm_blocks;
473 while (pfn >= bb->end_pfn) {
478 zone_bm->cur_block = bb;
479 pfn -= bb->start_pfn;
480 *bit_nr = pfn % BM_BITS_PER_CHUNK;
481 *addr = bb->data + pfn / BM_BITS_PER_CHUNK;
484 static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn)
489 memory_bm_find_bit(bm, pfn, &addr, &bit);
493 static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn)
498 memory_bm_find_bit(bm, pfn, &addr, &bit);
499 clear_bit(bit, addr);
502 static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
507 memory_bm_find_bit(bm, pfn, &addr, &bit);
508 return test_bit(bit, addr);
511 /* Two auxiliary functions for memory_bm_next_pfn */
513 /* Find the first set bit in the given chunk, if there is one */
515 static inline int next_bit_in_chunk(int bit, unsigned long *chunk_p)
518 while (bit < BM_BITS_PER_CHUNK) {
519 if (test_bit(bit, chunk_p))
527 /* Find a chunk containing some bits set in given block of bits */
529 static inline int next_chunk_in_block(int n, struct bm_block *bb)
532 while (n < bb->size) {
542 * memory_bm_next_pfn - find the pfn that corresponds to the next set bit
543 * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is
546 * It is required to run memory_bm_position_reset() before the first call to
550 static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
552 struct zone_bitmap *zone_bm;
560 chunk = bm->cur.chunk;
563 bit = next_bit_in_chunk(bit, bb->data + chunk);
567 chunk = next_chunk_in_block(chunk, bb);
569 } while (chunk >= 0);
572 memory_bm_reset_chunk(bm);
574 zone_bm = bm->cur.zone_bm->next;
576 bm->cur.zone_bm = zone_bm;
577 bm->cur.block = zone_bm->bm_blocks;
578 memory_bm_reset_chunk(bm);
581 memory_bm_position_reset(bm);
582 return BM_END_OF_MAP;
585 bm->cur.chunk = chunk;
587 return bb->start_pfn + chunk * BM_BITS_PER_CHUNK + bit;
591 * This structure represents a range of page frames the contents of which
592 * should not be saved during the suspend.
595 struct nosave_region {
596 struct list_head list;
597 unsigned long start_pfn;
598 unsigned long end_pfn;
601 static LIST_HEAD(nosave_regions);
604 * register_nosave_region - register a range of page frames the contents
605 * of which should not be saved during the suspend (to be used in the early
606 * initialization code)
610 __register_nosave_region(unsigned long start_pfn, unsigned long end_pfn,
613 struct nosave_region *region;
615 if (start_pfn >= end_pfn)
618 if (!list_empty(&nosave_regions)) {
619 /* Try to extend the previous region (they should be sorted) */
620 region = list_entry(nosave_regions.prev,
621 struct nosave_region, list);
622 if (region->end_pfn == start_pfn) {
623 region->end_pfn = end_pfn;
628 /* during init, this shouldn't fail */
629 region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL);
632 /* This allocation cannot fail */
633 region = alloc_bootmem_low(sizeof(struct nosave_region));
634 region->start_pfn = start_pfn;
635 region->end_pfn = end_pfn;
636 list_add_tail(®ion->list, &nosave_regions);
638 printk(KERN_INFO "PM: Registered nosave memory: %016lx - %016lx\n",
639 start_pfn << PAGE_SHIFT, end_pfn << PAGE_SHIFT);
643 * Set bits in this map correspond to the page frames the contents of which
644 * should not be saved during the suspend.
646 static struct memory_bitmap *forbidden_pages_map;
648 /* Set bits in this map correspond to free page frames. */
649 static struct memory_bitmap *free_pages_map;
652 * Each page frame allocated for creating the image is marked by setting the
653 * corresponding bits in forbidden_pages_map and free_pages_map simultaneously
656 void swsusp_set_page_free(struct page *page)
659 memory_bm_set_bit(free_pages_map, page_to_pfn(page));
662 static int swsusp_page_is_free(struct page *page)
664 return free_pages_map ?
665 memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0;
668 void swsusp_unset_page_free(struct page *page)
671 memory_bm_clear_bit(free_pages_map, page_to_pfn(page));
674 static void swsusp_set_page_forbidden(struct page *page)
676 if (forbidden_pages_map)
677 memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page));
680 int swsusp_page_is_forbidden(struct page *page)
682 return forbidden_pages_map ?
683 memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0;
686 static void swsusp_unset_page_forbidden(struct page *page)
688 if (forbidden_pages_map)
689 memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page));
693 * mark_nosave_pages - set bits corresponding to the page frames the
694 * contents of which should not be saved in a given bitmap.
697 static void mark_nosave_pages(struct memory_bitmap *bm)
699 struct nosave_region *region;
701 if (list_empty(&nosave_regions))
704 list_for_each_entry(region, &nosave_regions, list) {
707 pr_debug("PM: Marking nosave pages: %016lx - %016lx\n",
708 region->start_pfn << PAGE_SHIFT,
709 region->end_pfn << PAGE_SHIFT);
711 for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++)
713 memory_bm_set_bit(bm, pfn);
718 * create_basic_memory_bitmaps - create bitmaps needed for marking page
719 * frames that should not be saved and free page frames. The pointers
720 * forbidden_pages_map and free_pages_map are only modified if everything
721 * goes well, because we don't want the bits to be used before both bitmaps
725 int create_basic_memory_bitmaps(void)
727 struct memory_bitmap *bm1, *bm2;
730 BUG_ON(forbidden_pages_map || free_pages_map);
732 bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
736 error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
738 goto Free_first_object;
740 bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
742 goto Free_first_bitmap;
744 error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
746 goto Free_second_object;
748 forbidden_pages_map = bm1;
749 free_pages_map = bm2;
750 mark_nosave_pages(forbidden_pages_map);
752 pr_debug("PM: Basic memory bitmaps created\n");
759 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
766 * free_basic_memory_bitmaps - free memory bitmaps allocated by
767 * create_basic_memory_bitmaps(). The auxiliary pointers are necessary
768 * so that the bitmaps themselves are not referred to while they are being
772 void free_basic_memory_bitmaps(void)
774 struct memory_bitmap *bm1, *bm2;
776 BUG_ON(!(forbidden_pages_map && free_pages_map));
778 bm1 = forbidden_pages_map;
779 bm2 = free_pages_map;
780 forbidden_pages_map = NULL;
781 free_pages_map = NULL;
782 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
784 memory_bm_free(bm2, PG_UNSAFE_CLEAR);
787 pr_debug("PM: Basic memory bitmaps freed\n");
791 * snapshot_additional_pages - estimate the number of additional pages
792 * be needed for setting up the suspend image data structures for given
793 * zone (usually the returned value is greater than the exact number)
796 unsigned int snapshot_additional_pages(struct zone *zone)
800 res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
801 res += DIV_ROUND_UP(res * sizeof(struct bm_block), PAGE_SIZE);
805 #ifdef CONFIG_HIGHMEM
807 * count_free_highmem_pages - compute the total number of free highmem
808 * pages, system-wide.
811 static unsigned int count_free_highmem_pages(void)
814 unsigned int cnt = 0;
817 if (populated_zone(zone) && is_highmem(zone))
818 cnt += zone_page_state(zone, NR_FREE_PAGES);
824 * saveable_highmem_page - Determine whether a highmem page should be
825 * included in the suspend image.
827 * We should save the page if it isn't Nosave or NosaveFree, or Reserved,
828 * and it isn't a part of a free chunk of pages.
831 static struct page *saveable_highmem_page(unsigned long pfn)
838 page = pfn_to_page(pfn);
840 BUG_ON(!PageHighMem(page));
842 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) ||
850 * count_highmem_pages - compute the total number of saveable highmem
854 unsigned int count_highmem_pages(void)
859 for_each_zone(zone) {
860 unsigned long pfn, max_zone_pfn;
862 if (!is_highmem(zone))
865 mark_free_pages(zone);
866 max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
867 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
868 if (saveable_highmem_page(pfn))
874 static inline void *saveable_highmem_page(unsigned long pfn) { return NULL; }
875 #endif /* CONFIG_HIGHMEM */
878 * saveable_page - Determine whether a non-highmem page should be included
879 * in the suspend image.
881 * We should save the page if it isn't Nosave, and is not in the range
882 * of pages statically defined as 'unsaveable', and it isn't a part of
883 * a free chunk of pages.
886 static struct page *saveable_page(unsigned long pfn)
893 page = pfn_to_page(pfn);
895 BUG_ON(PageHighMem(page));
897 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
900 if (PageReserved(page)
901 && (!kernel_page_present(page) || pfn_is_nosave(pfn)))
908 * count_data_pages - compute the total number of saveable non-highmem
912 unsigned int count_data_pages(void)
915 unsigned long pfn, max_zone_pfn;
918 for_each_zone(zone) {
919 if (is_highmem(zone))
922 mark_free_pages(zone);
923 max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
924 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
925 if(saveable_page(pfn))
931 /* This is needed, because copy_page and memcpy are not usable for copying
934 static inline void do_copy_page(long *dst, long *src)
938 for (n = PAGE_SIZE / sizeof(long); n; n--)
944 * safe_copy_page - check if the page we are going to copy is marked as
945 * present in the kernel page tables (this always is the case if
946 * CONFIG_DEBUG_PAGEALLOC is not set and in that case
947 * kernel_page_present() always returns 'true').
949 static void safe_copy_page(void *dst, struct page *s_page)
951 if (kernel_page_present(s_page)) {
952 do_copy_page(dst, page_address(s_page));
954 kernel_map_pages(s_page, 1, 1);
955 do_copy_page(dst, page_address(s_page));
956 kernel_map_pages(s_page, 1, 0);
961 #ifdef CONFIG_HIGHMEM
962 static inline struct page *
963 page_is_saveable(struct zone *zone, unsigned long pfn)
965 return is_highmem(zone) ?
966 saveable_highmem_page(pfn) : saveable_page(pfn);
969 static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
971 struct page *s_page, *d_page;
974 s_page = pfn_to_page(src_pfn);
975 d_page = pfn_to_page(dst_pfn);
976 if (PageHighMem(s_page)) {
977 src = kmap_atomic(s_page, KM_USER0);
978 dst = kmap_atomic(d_page, KM_USER1);
979 do_copy_page(dst, src);
980 kunmap_atomic(src, KM_USER0);
981 kunmap_atomic(dst, KM_USER1);
983 if (PageHighMem(d_page)) {
984 /* Page pointed to by src may contain some kernel
985 * data modified by kmap_atomic()
987 safe_copy_page(buffer, s_page);
988 dst = kmap_atomic(pfn_to_page(dst_pfn), KM_USER0);
989 memcpy(dst, buffer, PAGE_SIZE);
990 kunmap_atomic(dst, KM_USER0);
992 safe_copy_page(page_address(d_page), s_page);
997 #define page_is_saveable(zone, pfn) saveable_page(pfn)
999 static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1001 safe_copy_page(page_address(pfn_to_page(dst_pfn)),
1002 pfn_to_page(src_pfn));
1004 #endif /* CONFIG_HIGHMEM */
1007 copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm)
1012 for_each_zone(zone) {
1013 unsigned long max_zone_pfn;
1015 mark_free_pages(zone);
1016 max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
1017 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1018 if (page_is_saveable(zone, pfn))
1019 memory_bm_set_bit(orig_bm, pfn);
1021 memory_bm_position_reset(orig_bm);
1022 memory_bm_position_reset(copy_bm);
1024 pfn = memory_bm_next_pfn(orig_bm);
1025 if (unlikely(pfn == BM_END_OF_MAP))
1027 copy_data_page(memory_bm_next_pfn(copy_bm), pfn);
1031 /* Total number of image pages */
1032 static unsigned int nr_copy_pages;
1033 /* Number of pages needed for saving the original pfns of the image pages */
1034 static unsigned int nr_meta_pages;
1037 * swsusp_free - free pages allocated for the suspend.
1039 * Suspend pages are alocated before the atomic copy is made, so we
1040 * need to release them after the resume.
1043 void swsusp_free(void)
1046 unsigned long pfn, max_zone_pfn;
1048 for_each_zone(zone) {
1049 max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
1050 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1051 if (pfn_valid(pfn)) {
1052 struct page *page = pfn_to_page(pfn);
1054 if (swsusp_page_is_forbidden(page) &&
1055 swsusp_page_is_free(page)) {
1056 swsusp_unset_page_forbidden(page);
1057 swsusp_unset_page_free(page);
1064 restore_pblist = NULL;
1068 #ifdef CONFIG_HIGHMEM
1070 * count_pages_for_highmem - compute the number of non-highmem pages
1071 * that will be necessary for creating copies of highmem pages.
1074 static unsigned int count_pages_for_highmem(unsigned int nr_highmem)
1076 unsigned int free_highmem = count_free_highmem_pages();
1078 if (free_highmem >= nr_highmem)
1081 nr_highmem -= free_highmem;
1087 count_pages_for_highmem(unsigned int nr_highmem) { return 0; }
1088 #endif /* CONFIG_HIGHMEM */
1091 * enough_free_mem - Make sure we have enough free memory for the
1095 static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem)
1098 unsigned int free = 0, meta = 0;
1100 for_each_zone(zone) {
1101 meta += snapshot_additional_pages(zone);
1102 if (!is_highmem(zone))
1103 free += zone_page_state(zone, NR_FREE_PAGES);
1106 nr_pages += count_pages_for_highmem(nr_highmem);
1107 pr_debug("PM: Normal pages needed: %u + %u + %u, available pages: %u\n",
1108 nr_pages, PAGES_FOR_IO, meta, free);
1110 return free > nr_pages + PAGES_FOR_IO + meta;
1113 #ifdef CONFIG_HIGHMEM
1115 * get_highmem_buffer - if there are some highmem pages in the suspend
1116 * image, we may need the buffer to copy them and/or load their data.
1119 static inline int get_highmem_buffer(int safe_needed)
1121 buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed);
1122 return buffer ? 0 : -ENOMEM;
1126 * alloc_highmem_image_pages - allocate some highmem pages for the image.
1127 * Try to allocate as many pages as needed, but if the number of free
1128 * highmem pages is lesser than that, allocate them all.
1131 static inline unsigned int
1132 alloc_highmem_image_pages(struct memory_bitmap *bm, unsigned int nr_highmem)
1134 unsigned int to_alloc = count_free_highmem_pages();
1136 if (to_alloc > nr_highmem)
1137 to_alloc = nr_highmem;
1139 nr_highmem -= to_alloc;
1140 while (to_alloc-- > 0) {
1143 page = alloc_image_page(__GFP_HIGHMEM);
1144 memory_bm_set_bit(bm, page_to_pfn(page));
1149 static inline int get_highmem_buffer(int safe_needed) { return 0; }
1151 static inline unsigned int
1152 alloc_highmem_image_pages(struct memory_bitmap *bm, unsigned int n) { return 0; }
1153 #endif /* CONFIG_HIGHMEM */
1156 * swsusp_alloc - allocate memory for the suspend image
1158 * We first try to allocate as many highmem pages as there are
1159 * saveable highmem pages in the system. If that fails, we allocate
1160 * non-highmem pages for the copies of the remaining highmem ones.
1162 * In this approach it is likely that the copies of highmem pages will
1163 * also be located in the high memory, because of the way in which
1164 * copy_data_pages() works.
1168 swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm,
1169 unsigned int nr_pages, unsigned int nr_highmem)
1173 error = memory_bm_create(orig_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY);
1177 error = memory_bm_create(copy_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY);
1181 if (nr_highmem > 0) {
1182 error = get_highmem_buffer(PG_ANY);
1186 nr_pages += alloc_highmem_image_pages(copy_bm, nr_highmem);
1188 while (nr_pages-- > 0) {
1189 struct page *page = alloc_image_page(GFP_ATOMIC | __GFP_COLD);
1194 memory_bm_set_bit(copy_bm, page_to_pfn(page));
1203 /* Memory bitmap used for marking saveable pages (during suspend) or the
1204 * suspend image pages (during resume)
1206 static struct memory_bitmap orig_bm;
1207 /* Memory bitmap used on suspend for marking allocated pages that will contain
1208 * the copies of saveable pages. During resume it is initially used for
1209 * marking the suspend image pages, but then its set bits are duplicated in
1210 * @orig_bm and it is released. Next, on systems with high memory, it may be
1211 * used for marking "safe" highmem pages, but it has to be reinitialized for
1214 static struct memory_bitmap copy_bm;
1216 asmlinkage int swsusp_save(void)
1218 unsigned int nr_pages, nr_highmem;
1220 printk(KERN_INFO "PM: Creating hibernation image: \n");
1222 drain_local_pages(NULL);
1223 nr_pages = count_data_pages();
1224 nr_highmem = count_highmem_pages();
1225 printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem);
1227 if (!enough_free_mem(nr_pages, nr_highmem)) {
1228 printk(KERN_ERR "PM: Not enough free memory\n");
1232 if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) {
1233 printk(KERN_ERR "PM: Memory allocation failed\n");
1237 /* During allocating of suspend pagedir, new cold pages may appear.
1240 drain_local_pages(NULL);
1241 copy_data_pages(©_bm, &orig_bm);
1244 * End of critical section. From now on, we can write to memory,
1245 * but we should not touch disk. This specially means we must _not_
1246 * touch swap space! Except we must write out our image of course.
1249 nr_pages += nr_highmem;
1250 nr_copy_pages = nr_pages;
1251 nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE);
1253 printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n",
1259 #ifndef CONFIG_ARCH_HIBERNATION_HEADER
1260 static int init_header_complete(struct swsusp_info *info)
1262 memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname));
1263 info->version_code = LINUX_VERSION_CODE;
1267 static char *check_image_kernel(struct swsusp_info *info)
1269 if (info->version_code != LINUX_VERSION_CODE)
1270 return "kernel version";
1271 if (strcmp(info->uts.sysname,init_utsname()->sysname))
1272 return "system type";
1273 if (strcmp(info->uts.release,init_utsname()->release))
1274 return "kernel release";
1275 if (strcmp(info->uts.version,init_utsname()->version))
1277 if (strcmp(info->uts.machine,init_utsname()->machine))
1281 #endif /* CONFIG_ARCH_HIBERNATION_HEADER */
1283 unsigned long snapshot_get_image_size(void)
1285 return nr_copy_pages + nr_meta_pages + 1;
1288 static int init_header(struct swsusp_info *info)
1290 memset(info, 0, sizeof(struct swsusp_info));
1291 info->num_physpages = num_physpages;
1292 info->image_pages = nr_copy_pages;
1293 info->pages = snapshot_get_image_size();
1294 info->size = info->pages;
1295 info->size <<= PAGE_SHIFT;
1296 return init_header_complete(info);
1300 * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm
1301 * are stored in the array @buf[] (1 page at a time)
1305 pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
1309 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
1310 buf[j] = memory_bm_next_pfn(bm);
1311 if (unlikely(buf[j] == BM_END_OF_MAP))
1317 * snapshot_read_next - used for reading the system memory snapshot.
1319 * On the first call to it @handle should point to a zeroed
1320 * snapshot_handle structure. The structure gets updated and a pointer
1321 * to it should be passed to this function every next time.
1323 * The @count parameter should contain the number of bytes the caller
1324 * wants to read from the snapshot. It must not be zero.
1326 * On success the function returns a positive number. Then, the caller
1327 * is allowed to read up to the returned number of bytes from the memory
1328 * location computed by the data_of() macro. The number returned
1329 * may be smaller than @count, but this only happens if the read would
1330 * cross a page boundary otherwise.
1332 * The function returns 0 to indicate the end of data stream condition,
1333 * and a negative number is returned on error. In such cases the
1334 * structure pointed to by @handle is not updated and should not be used
1338 int snapshot_read_next(struct snapshot_handle *handle, size_t count)
1340 if (handle->cur > nr_meta_pages + nr_copy_pages)
1344 /* This makes the buffer be freed by swsusp_free() */
1345 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
1349 if (!handle->offset) {
1352 error = init_header((struct swsusp_info *)buffer);
1355 handle->buffer = buffer;
1356 memory_bm_position_reset(&orig_bm);
1357 memory_bm_position_reset(©_bm);
1359 if (handle->prev < handle->cur) {
1360 if (handle->cur <= nr_meta_pages) {
1361 memset(buffer, 0, PAGE_SIZE);
1362 pack_pfns(buffer, &orig_bm);
1366 page = pfn_to_page(memory_bm_next_pfn(©_bm));
1367 if (PageHighMem(page)) {
1368 /* Highmem pages are copied to the buffer,
1369 * because we can't return with a kmapped
1370 * highmem page (we may not be called again).
1374 kaddr = kmap_atomic(page, KM_USER0);
1375 memcpy(buffer, kaddr, PAGE_SIZE);
1376 kunmap_atomic(kaddr, KM_USER0);
1377 handle->buffer = buffer;
1379 handle->buffer = page_address(page);
1382 handle->prev = handle->cur;
1384 handle->buf_offset = handle->cur_offset;
1385 if (handle->cur_offset + count >= PAGE_SIZE) {
1386 count = PAGE_SIZE - handle->cur_offset;
1387 handle->cur_offset = 0;
1390 handle->cur_offset += count;
1392 handle->offset += count;
1397 * mark_unsafe_pages - mark the pages that cannot be used for storing
1398 * the image during resume, because they conflict with the pages that
1399 * had been used before suspend
1402 static int mark_unsafe_pages(struct memory_bitmap *bm)
1405 unsigned long pfn, max_zone_pfn;
1407 /* Clear page flags */
1408 for_each_zone(zone) {
1409 max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
1410 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1412 swsusp_unset_page_free(pfn_to_page(pfn));
1415 /* Mark pages that correspond to the "original" pfns as "unsafe" */
1416 memory_bm_position_reset(bm);
1418 pfn = memory_bm_next_pfn(bm);
1419 if (likely(pfn != BM_END_OF_MAP)) {
1420 if (likely(pfn_valid(pfn)))
1421 swsusp_set_page_free(pfn_to_page(pfn));
1425 } while (pfn != BM_END_OF_MAP);
1427 allocated_unsafe_pages = 0;
1433 duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src)
1437 memory_bm_position_reset(src);
1438 pfn = memory_bm_next_pfn(src);
1439 while (pfn != BM_END_OF_MAP) {
1440 memory_bm_set_bit(dst, pfn);
1441 pfn = memory_bm_next_pfn(src);
1445 static int check_header(struct swsusp_info *info)
1449 reason = check_image_kernel(info);
1450 if (!reason && info->num_physpages != num_physpages)
1451 reason = "memory size";
1453 printk(KERN_ERR "PM: Image mismatch: %s\n", reason);
1460 * load header - check the image header and copy data from it
1464 load_header(struct swsusp_info *info)
1468 restore_pblist = NULL;
1469 error = check_header(info);
1471 nr_copy_pages = info->image_pages;
1472 nr_meta_pages = info->pages - info->image_pages - 1;
1478 * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set
1479 * the corresponding bit in the memory bitmap @bm
1483 unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
1487 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
1488 if (unlikely(buf[j] == BM_END_OF_MAP))
1491 memory_bm_set_bit(bm, buf[j]);
1495 /* List of "safe" pages that may be used to store data loaded from the suspend
1498 static struct linked_page *safe_pages_list;
1500 #ifdef CONFIG_HIGHMEM
1501 /* struct highmem_pbe is used for creating the list of highmem pages that
1502 * should be restored atomically during the resume from disk, because the page
1503 * frames they have occupied before the suspend are in use.
1505 struct highmem_pbe {
1506 struct page *copy_page; /* data is here now */
1507 struct page *orig_page; /* data was here before the suspend */
1508 struct highmem_pbe *next;
1511 /* List of highmem PBEs needed for restoring the highmem pages that were
1512 * allocated before the suspend and included in the suspend image, but have
1513 * also been allocated by the "resume" kernel, so their contents cannot be
1514 * written directly to their "original" page frames.
1516 static struct highmem_pbe *highmem_pblist;
1519 * count_highmem_image_pages - compute the number of highmem pages in the
1520 * suspend image. The bits in the memory bitmap @bm that correspond to the
1521 * image pages are assumed to be set.
1524 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm)
1527 unsigned int cnt = 0;
1529 memory_bm_position_reset(bm);
1530 pfn = memory_bm_next_pfn(bm);
1531 while (pfn != BM_END_OF_MAP) {
1532 if (PageHighMem(pfn_to_page(pfn)))
1535 pfn = memory_bm_next_pfn(bm);
1541 * prepare_highmem_image - try to allocate as many highmem pages as
1542 * there are highmem image pages (@nr_highmem_p points to the variable
1543 * containing the number of highmem image pages). The pages that are
1544 * "safe" (ie. will not be overwritten when the suspend image is
1545 * restored) have the corresponding bits set in @bm (it must be
1548 * NOTE: This function should not be called if there are no highmem
1552 static unsigned int safe_highmem_pages;
1554 static struct memory_bitmap *safe_highmem_bm;
1557 prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p)
1559 unsigned int to_alloc;
1561 if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE))
1564 if (get_highmem_buffer(PG_SAFE))
1567 to_alloc = count_free_highmem_pages();
1568 if (to_alloc > *nr_highmem_p)
1569 to_alloc = *nr_highmem_p;
1571 *nr_highmem_p = to_alloc;
1573 safe_highmem_pages = 0;
1574 while (to_alloc-- > 0) {
1577 page = alloc_page(__GFP_HIGHMEM);
1578 if (!swsusp_page_is_free(page)) {
1579 /* The page is "safe", set its bit the bitmap */
1580 memory_bm_set_bit(bm, page_to_pfn(page));
1581 safe_highmem_pages++;
1583 /* Mark the page as allocated */
1584 swsusp_set_page_forbidden(page);
1585 swsusp_set_page_free(page);
1587 memory_bm_position_reset(bm);
1588 safe_highmem_bm = bm;
1593 * get_highmem_page_buffer - for given highmem image page find the buffer
1594 * that suspend_write_next() should set for its caller to write to.
1596 * If the page is to be saved to its "original" page frame or a copy of
1597 * the page is to be made in the highmem, @buffer is returned. Otherwise,
1598 * the copy of the page is to be made in normal memory, so the address of
1599 * the copy is returned.
1601 * If @buffer is returned, the caller of suspend_write_next() will write
1602 * the page's contents to @buffer, so they will have to be copied to the
1603 * right location on the next call to suspend_write_next() and it is done
1604 * with the help of copy_last_highmem_page(). For this purpose, if
1605 * @buffer is returned, @last_highmem page is set to the page to which
1606 * the data will have to be copied from @buffer.
1609 static struct page *last_highmem_page;
1612 get_highmem_page_buffer(struct page *page, struct chain_allocator *ca)
1614 struct highmem_pbe *pbe;
1617 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) {
1618 /* We have allocated the "original" page frame and we can
1619 * use it directly to store the loaded page.
1621 last_highmem_page = page;
1624 /* The "original" page frame has not been allocated and we have to
1625 * use a "safe" page frame to store the loaded page.
1627 pbe = chain_alloc(ca, sizeof(struct highmem_pbe));
1632 pbe->orig_page = page;
1633 if (safe_highmem_pages > 0) {
1636 /* Copy of the page will be stored in high memory */
1638 tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm));
1639 safe_highmem_pages--;
1640 last_highmem_page = tmp;
1641 pbe->copy_page = tmp;
1643 /* Copy of the page will be stored in normal memory */
1644 kaddr = safe_pages_list;
1645 safe_pages_list = safe_pages_list->next;
1646 pbe->copy_page = virt_to_page(kaddr);
1648 pbe->next = highmem_pblist;
1649 highmem_pblist = pbe;
1654 * copy_last_highmem_page - copy the contents of a highmem image from
1655 * @buffer, where the caller of snapshot_write_next() has place them,
1656 * to the right location represented by @last_highmem_page .
1659 static void copy_last_highmem_page(void)
1661 if (last_highmem_page) {
1664 dst = kmap_atomic(last_highmem_page, KM_USER0);
1665 memcpy(dst, buffer, PAGE_SIZE);
1666 kunmap_atomic(dst, KM_USER0);
1667 last_highmem_page = NULL;
1671 static inline int last_highmem_page_copied(void)
1673 return !last_highmem_page;
1676 static inline void free_highmem_data(void)
1678 if (safe_highmem_bm)
1679 memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR);
1682 free_image_page(buffer, PG_UNSAFE_CLEAR);
1685 static inline int get_safe_write_buffer(void) { return 0; }
1688 count_highmem_image_pages(struct memory_bitmap *bm) { return 0; }
1691 prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p)
1696 static inline void *
1697 get_highmem_page_buffer(struct page *page, struct chain_allocator *ca)
1702 static inline void copy_last_highmem_page(void) {}
1703 static inline int last_highmem_page_copied(void) { return 1; }
1704 static inline void free_highmem_data(void) {}
1705 #endif /* CONFIG_HIGHMEM */
1708 * prepare_image - use the memory bitmap @bm to mark the pages that will
1709 * be overwritten in the process of restoring the system memory state
1710 * from the suspend image ("unsafe" pages) and allocate memory for the
1713 * The idea is to allocate a new memory bitmap first and then allocate
1714 * as many pages as needed for the image data, but not to assign these
1715 * pages to specific tasks initially. Instead, we just mark them as
1716 * allocated and create a lists of "safe" pages that will be used
1717 * later. On systems with high memory a list of "safe" highmem pages is
1721 #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe))
1724 prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm)
1726 unsigned int nr_pages, nr_highmem;
1727 struct linked_page *sp_list, *lp;
1730 /* If there is no highmem, the buffer will not be necessary */
1731 free_image_page(buffer, PG_UNSAFE_CLEAR);
1734 nr_highmem = count_highmem_image_pages(bm);
1735 error = mark_unsafe_pages(bm);
1739 error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE);
1743 duplicate_memory_bitmap(new_bm, bm);
1744 memory_bm_free(bm, PG_UNSAFE_KEEP);
1745 if (nr_highmem > 0) {
1746 error = prepare_highmem_image(bm, &nr_highmem);
1750 /* Reserve some safe pages for potential later use.
1752 * NOTE: This way we make sure there will be enough safe pages for the
1753 * chain_alloc() in get_buffer(). It is a bit wasteful, but
1754 * nr_copy_pages cannot be greater than 50% of the memory anyway.
1757 /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */
1758 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
1759 nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE);
1760 while (nr_pages > 0) {
1761 lp = get_image_page(GFP_ATOMIC, PG_SAFE);
1770 /* Preallocate memory for the image */
1771 safe_pages_list = NULL;
1772 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
1773 while (nr_pages > 0) {
1774 lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC);
1779 if (!swsusp_page_is_free(virt_to_page(lp))) {
1780 /* The page is "safe", add it to the list */
1781 lp->next = safe_pages_list;
1782 safe_pages_list = lp;
1784 /* Mark the page as allocated */
1785 swsusp_set_page_forbidden(virt_to_page(lp));
1786 swsusp_set_page_free(virt_to_page(lp));
1789 /* Free the reserved safe pages so that chain_alloc() can use them */
1792 free_image_page(sp_list, PG_UNSAFE_CLEAR);
1803 * get_buffer - compute the address that snapshot_write_next() should
1804 * set for its caller to write to.
1807 static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca)
1810 struct page *page = pfn_to_page(memory_bm_next_pfn(bm));
1812 if (PageHighMem(page))
1813 return get_highmem_page_buffer(page, ca);
1815 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page))
1816 /* We have allocated the "original" page frame and we can
1817 * use it directly to store the loaded page.
1819 return page_address(page);
1821 /* The "original" page frame has not been allocated and we have to
1822 * use a "safe" page frame to store the loaded page.
1824 pbe = chain_alloc(ca, sizeof(struct pbe));
1829 pbe->orig_address = page_address(page);
1830 pbe->address = safe_pages_list;
1831 safe_pages_list = safe_pages_list->next;
1832 pbe->next = restore_pblist;
1833 restore_pblist = pbe;
1834 return pbe->address;
1838 * snapshot_write_next - used for writing the system memory snapshot.
1840 * On the first call to it @handle should point to a zeroed
1841 * snapshot_handle structure. The structure gets updated and a pointer
1842 * to it should be passed to this function every next time.
1844 * The @count parameter should contain the number of bytes the caller
1845 * wants to write to the image. It must not be zero.
1847 * On success the function returns a positive number. Then, the caller
1848 * is allowed to write up to the returned number of bytes to the memory
1849 * location computed by the data_of() macro. The number returned
1850 * may be smaller than @count, but this only happens if the write would
1851 * cross a page boundary otherwise.
1853 * The function returns 0 to indicate the "end of file" condition,
1854 * and a negative number is returned on error. In such cases the
1855 * structure pointed to by @handle is not updated and should not be used
1859 int snapshot_write_next(struct snapshot_handle *handle, size_t count)
1861 static struct chain_allocator ca;
1864 /* Check if we have already loaded the entire image */
1865 if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages)
1868 if (handle->offset == 0) {
1870 /* This makes the buffer be freed by swsusp_free() */
1871 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
1876 handle->buffer = buffer;
1878 handle->sync_read = 1;
1879 if (handle->prev < handle->cur) {
1880 if (handle->prev == 0) {
1881 error = load_header(buffer);
1885 error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY);
1889 } else if (handle->prev <= nr_meta_pages) {
1890 unpack_orig_pfns(buffer, ©_bm);
1891 if (handle->prev == nr_meta_pages) {
1892 error = prepare_image(&orig_bm, ©_bm);
1896 chain_init(&ca, GFP_ATOMIC, PG_SAFE);
1897 memory_bm_position_reset(&orig_bm);
1898 restore_pblist = NULL;
1899 handle->buffer = get_buffer(&orig_bm, &ca);
1900 handle->sync_read = 0;
1901 if (!handle->buffer)
1905 copy_last_highmem_page();
1906 handle->buffer = get_buffer(&orig_bm, &ca);
1907 if (handle->buffer != buffer)
1908 handle->sync_read = 0;
1910 handle->prev = handle->cur;
1912 handle->buf_offset = handle->cur_offset;
1913 if (handle->cur_offset + count >= PAGE_SIZE) {
1914 count = PAGE_SIZE - handle->cur_offset;
1915 handle->cur_offset = 0;
1918 handle->cur_offset += count;
1920 handle->offset += count;
1925 * snapshot_write_finalize - must be called after the last call to
1926 * snapshot_write_next() in case the last page in the image happens
1927 * to be a highmem page and its contents should be stored in the
1928 * highmem. Additionally, it releases the memory that will not be
1932 void snapshot_write_finalize(struct snapshot_handle *handle)
1934 copy_last_highmem_page();
1935 /* Free only if we have loaded the image entirely */
1936 if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages) {
1937 memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR);
1938 free_highmem_data();
1942 int snapshot_image_loaded(struct snapshot_handle *handle)
1944 return !(!nr_copy_pages || !last_highmem_page_copied() ||
1945 handle->cur <= nr_meta_pages + nr_copy_pages);
1948 #ifdef CONFIG_HIGHMEM
1949 /* Assumes that @buf is ready and points to a "safe" page */
1951 swap_two_pages_data(struct page *p1, struct page *p2, void *buf)
1953 void *kaddr1, *kaddr2;
1955 kaddr1 = kmap_atomic(p1, KM_USER0);
1956 kaddr2 = kmap_atomic(p2, KM_USER1);
1957 memcpy(buf, kaddr1, PAGE_SIZE);
1958 memcpy(kaddr1, kaddr2, PAGE_SIZE);
1959 memcpy(kaddr2, buf, PAGE_SIZE);
1960 kunmap_atomic(kaddr1, KM_USER0);
1961 kunmap_atomic(kaddr2, KM_USER1);
1965 * restore_highmem - for each highmem page that was allocated before
1966 * the suspend and included in the suspend image, and also has been
1967 * allocated by the "resume" kernel swap its current (ie. "before
1968 * resume") contents with the previous (ie. "before suspend") one.
1970 * If the resume eventually fails, we can call this function once
1971 * again and restore the "before resume" highmem state.
1974 int restore_highmem(void)
1976 struct highmem_pbe *pbe = highmem_pblist;
1982 buf = get_image_page(GFP_ATOMIC, PG_SAFE);
1987 swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf);
1990 free_image_page(buf, PG_UNSAFE_CLEAR);
1993 #endif /* CONFIG_HIGHMEM */