2 * linux/kernel/power/snapshot.c
4 * This file provides system snapshot/restore functionality for swsusp.
6 * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.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>
28 #include <linux/list.h>
29 #include <linux/slab.h>
30 #include <linux/compiler.h>
31 #include <linux/ktime.h>
33 #include <asm/uaccess.h>
34 #include <asm/mmu_context.h>
35 #include <asm/pgtable.h>
36 #include <asm/tlbflush.h>
41 #ifdef CONFIG_DEBUG_RODATA
42 static bool hibernate_restore_protection;
43 static bool hibernate_restore_protection_active;
45 void enable_restore_image_protection(void)
47 hibernate_restore_protection = true;
50 static inline void hibernate_restore_protection_begin(void)
52 hibernate_restore_protection_active = hibernate_restore_protection;
55 static inline void hibernate_restore_protection_end(void)
57 hibernate_restore_protection_active = false;
60 static inline void hibernate_restore_protect_page(void *page_address)
62 if (hibernate_restore_protection_active)
63 set_memory_ro((unsigned long)page_address, 1);
66 static inline void hibernate_restore_unprotect_page(void *page_address)
68 if (hibernate_restore_protection_active)
69 set_memory_rw((unsigned long)page_address, 1);
72 static inline void hibernate_restore_protection_begin(void) {}
73 static inline void hibernate_restore_protection_end(void) {}
74 static inline void hibernate_restore_protect_page(void *page_address) {}
75 static inline void hibernate_restore_unprotect_page(void *page_address) {}
76 #endif /* CONFIG_DEBUG_RODATA */
78 static int swsusp_page_is_free(struct page *);
79 static void swsusp_set_page_forbidden(struct page *);
80 static void swsusp_unset_page_forbidden(struct page *);
83 * Number of bytes to reserve for memory allocations made by device drivers
84 * from their ->freeze() and ->freeze_noirq() callbacks so that they don't
85 * cause image creation to fail (tunable via /sys/power/reserved_size).
87 unsigned long reserved_size;
89 void __init hibernate_reserved_size_init(void)
91 reserved_size = SPARE_PAGES * PAGE_SIZE;
95 * Preferred image size in bytes (tunable via /sys/power/image_size).
96 * When it is set to N, swsusp will do its best to ensure the image
97 * size will not exceed N bytes, but if that is impossible, it will
98 * try to create the smallest image possible.
100 unsigned long image_size;
102 void __init hibernate_image_size_init(void)
104 image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE;
108 * List of PBEs needed for restoring the pages that were allocated before
109 * the suspend and included in the suspend image, but have also been
110 * allocated by the "resume" kernel, so their contents cannot be written
111 * directly to their "original" page frames.
113 struct pbe *restore_pblist;
115 /* struct linked_page is used to build chains of pages */
117 #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *))
120 struct linked_page *next;
121 char data[LINKED_PAGE_DATA_SIZE];
125 * List of "safe" pages (ie. pages that were not used by the image kernel
126 * before hibernation) that may be used as temporary storage for image kernel
129 static struct linked_page *safe_pages_list;
131 /* Pointer to an auxiliary buffer (1 page) */
136 #define PG_UNSAFE_CLEAR 1
137 #define PG_UNSAFE_KEEP 0
139 static unsigned int allocated_unsafe_pages;
142 * get_image_page - Allocate a page for a hibernation image.
143 * @gfp_mask: GFP mask for the allocation.
144 * @safe_needed: Get pages that were not used before hibernation (restore only)
146 * During image restoration, for storing the PBE list and the image data, we can
147 * only use memory pages that do not conflict with the pages used before
148 * hibernation. The "unsafe" pages have PageNosaveFree set and we count them
149 * using allocated_unsafe_pages.
151 * Each allocated image page is marked as PageNosave and PageNosaveFree so that
152 * swsusp_free() can release it.
154 static void *get_image_page(gfp_t gfp_mask, int safe_needed)
158 res = (void *)get_zeroed_page(gfp_mask);
160 while (res && swsusp_page_is_free(virt_to_page(res))) {
161 /* The page is unsafe, mark it for swsusp_free() */
162 swsusp_set_page_forbidden(virt_to_page(res));
163 allocated_unsafe_pages++;
164 res = (void *)get_zeroed_page(gfp_mask);
167 swsusp_set_page_forbidden(virt_to_page(res));
168 swsusp_set_page_free(virt_to_page(res));
173 static void *__get_safe_page(gfp_t gfp_mask)
175 if (safe_pages_list) {
176 void *ret = safe_pages_list;
178 safe_pages_list = safe_pages_list->next;
179 memset(ret, 0, PAGE_SIZE);
182 return get_image_page(gfp_mask, PG_SAFE);
185 unsigned long get_safe_page(gfp_t gfp_mask)
187 return (unsigned long)__get_safe_page(gfp_mask);
190 static struct page *alloc_image_page(gfp_t gfp_mask)
194 page = alloc_page(gfp_mask);
196 swsusp_set_page_forbidden(page);
197 swsusp_set_page_free(page);
202 static void recycle_safe_page(void *page_address)
204 struct linked_page *lp = page_address;
206 lp->next = safe_pages_list;
207 safe_pages_list = lp;
211 * free_image_page - Free a page allocated for hibernation image.
212 * @addr: Address of the page to free.
213 * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page.
215 * The page to free should have been allocated by get_image_page() (page flags
216 * set by it are affected).
218 static inline void free_image_page(void *addr, int clear_nosave_free)
222 BUG_ON(!virt_addr_valid(addr));
224 page = virt_to_page(addr);
226 swsusp_unset_page_forbidden(page);
227 if (clear_nosave_free)
228 swsusp_unset_page_free(page);
233 static inline void free_list_of_pages(struct linked_page *list,
234 int clear_page_nosave)
237 struct linked_page *lp = list->next;
239 free_image_page(list, clear_page_nosave);
245 * struct chain_allocator is used for allocating small objects out of
246 * a linked list of pages called 'the chain'.
248 * The chain grows each time when there is no room for a new object in
249 * the current page. The allocated objects cannot be freed individually.
250 * It is only possible to free them all at once, by freeing the entire
253 * NOTE: The chain allocator may be inefficient if the allocated objects
254 * are not much smaller than PAGE_SIZE.
256 struct chain_allocator {
257 struct linked_page *chain; /* the chain */
258 unsigned int used_space; /* total size of objects allocated out
259 of the current page */
260 gfp_t gfp_mask; /* mask for allocating pages */
261 int safe_needed; /* if set, only "safe" pages are allocated */
264 static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask,
268 ca->used_space = LINKED_PAGE_DATA_SIZE;
269 ca->gfp_mask = gfp_mask;
270 ca->safe_needed = safe_needed;
273 static void *chain_alloc(struct chain_allocator *ca, unsigned int size)
277 if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) {
278 struct linked_page *lp;
280 lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) :
281 get_image_page(ca->gfp_mask, PG_ANY);
285 lp->next = ca->chain;
289 ret = ca->chain->data + ca->used_space;
290 ca->used_space += size;
295 * Data types related to memory bitmaps.
297 * Memory bitmap is a structure consiting of many linked lists of
298 * objects. The main list's elements are of type struct zone_bitmap
299 * and each of them corresonds to one zone. For each zone bitmap
300 * object there is a list of objects of type struct bm_block that
301 * represent each blocks of bitmap in which information is stored.
303 * struct memory_bitmap contains a pointer to the main list of zone
304 * bitmap objects, a struct bm_position used for browsing the bitmap,
305 * and a pointer to the list of pages used for allocating all of the
306 * zone bitmap objects and bitmap block objects.
308 * NOTE: It has to be possible to lay out the bitmap in memory
309 * using only allocations of order 0. Additionally, the bitmap is
310 * designed to work with arbitrary number of zones (this is over the
311 * top for now, but let's avoid making unnecessary assumptions ;-).
313 * struct zone_bitmap contains a pointer to a list of bitmap block
314 * objects and a pointer to the bitmap block object that has been
315 * most recently used for setting bits. Additionally, it contains the
316 * PFNs that correspond to the start and end of the represented zone.
318 * struct bm_block contains a pointer to the memory page in which
319 * information is stored (in the form of a block of bitmap)
320 * It also contains the pfns that correspond to the start and end of
321 * the represented memory area.
323 * The memory bitmap is organized as a radix tree to guarantee fast random
324 * access to the bits. There is one radix tree for each zone (as returned
325 * from create_mem_extents).
327 * One radix tree is represented by one struct mem_zone_bm_rtree. There are
328 * two linked lists for the nodes of the tree, one for the inner nodes and
329 * one for the leave nodes. The linked leave nodes are used for fast linear
330 * access of the memory bitmap.
332 * The struct rtree_node represents one node of the radix tree.
335 #define BM_END_OF_MAP (~0UL)
337 #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE)
338 #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3)
339 #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1)
342 * struct rtree_node is a wrapper struct to link the nodes
343 * of the rtree together for easy linear iteration over
344 * bits and easy freeing
347 struct list_head list;
352 * struct mem_zone_bm_rtree represents a bitmap used for one
353 * populated memory zone.
355 struct mem_zone_bm_rtree {
356 struct list_head list; /* Link Zones together */
357 struct list_head nodes; /* Radix Tree inner nodes */
358 struct list_head leaves; /* Radix Tree leaves */
359 unsigned long start_pfn; /* Zone start page frame */
360 unsigned long end_pfn; /* Zone end page frame + 1 */
361 struct rtree_node *rtree; /* Radix Tree Root */
362 int levels; /* Number of Radix Tree Levels */
363 unsigned int blocks; /* Number of Bitmap Blocks */
366 /* strcut bm_position is used for browsing memory bitmaps */
369 struct mem_zone_bm_rtree *zone;
370 struct rtree_node *node;
371 unsigned long node_pfn;
375 struct memory_bitmap {
376 struct list_head zones;
377 struct linked_page *p_list; /* list of pages used to store zone
378 bitmap objects and bitmap block
380 struct bm_position cur; /* most recently used bit position */
383 /* Functions that operate on memory bitmaps */
385 #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long))
386 #if BITS_PER_LONG == 32
387 #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2)
389 #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3)
391 #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1)
394 * alloc_rtree_node - Allocate a new node and add it to the radix tree.
396 * This function is used to allocate inner nodes as well as the
397 * leave nodes of the radix tree. It also adds the node to the
398 * corresponding linked list passed in by the *list parameter.
400 static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed,
401 struct chain_allocator *ca,
402 struct list_head *list)
404 struct rtree_node *node;
406 node = chain_alloc(ca, sizeof(struct rtree_node));
410 node->data = get_image_page(gfp_mask, safe_needed);
414 list_add_tail(&node->list, list);
420 * add_rtree_block - Add a new leave node to the radix tree.
422 * The leave nodes need to be allocated in order to keep the leaves
423 * linked list in order. This is guaranteed by the zone->blocks
426 static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask,
427 int safe_needed, struct chain_allocator *ca)
429 struct rtree_node *node, *block, **dst;
430 unsigned int levels_needed, block_nr;
433 block_nr = zone->blocks;
436 /* How many levels do we need for this block nr? */
439 block_nr >>= BM_RTREE_LEVEL_SHIFT;
442 /* Make sure the rtree has enough levels */
443 for (i = zone->levels; i < levels_needed; i++) {
444 node = alloc_rtree_node(gfp_mask, safe_needed, ca,
449 node->data[0] = (unsigned long)zone->rtree;
454 /* Allocate new block */
455 block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves);
459 /* Now walk the rtree to insert the block */
462 block_nr = zone->blocks;
463 for (i = zone->levels; i > 0; i--) {
467 node = alloc_rtree_node(gfp_mask, safe_needed, ca,
474 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
475 index &= BM_RTREE_LEVEL_MASK;
476 dst = (struct rtree_node **)&((*dst)->data[index]);
486 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
487 int clear_nosave_free);
490 * create_zone_bm_rtree - Create a radix tree for one zone.
492 * Allocated the mem_zone_bm_rtree structure and initializes it.
493 * This function also allocated and builds the radix tree for the
496 static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask,
498 struct chain_allocator *ca,
502 struct mem_zone_bm_rtree *zone;
503 unsigned int i, nr_blocks;
507 zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree));
511 INIT_LIST_HEAD(&zone->nodes);
512 INIT_LIST_HEAD(&zone->leaves);
513 zone->start_pfn = start;
515 nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK);
517 for (i = 0; i < nr_blocks; i++) {
518 if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) {
519 free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR);
528 * free_zone_bm_rtree - Free the memory of the radix tree.
530 * Free all node pages of the radix tree. The mem_zone_bm_rtree
531 * structure itself is not freed here nor are the rtree_node
534 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
535 int clear_nosave_free)
537 struct rtree_node *node;
539 list_for_each_entry(node, &zone->nodes, list)
540 free_image_page(node->data, clear_nosave_free);
542 list_for_each_entry(node, &zone->leaves, list)
543 free_image_page(node->data, clear_nosave_free);
546 static void memory_bm_position_reset(struct memory_bitmap *bm)
548 bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree,
550 bm->cur.node = list_entry(bm->cur.zone->leaves.next,
551 struct rtree_node, list);
552 bm->cur.node_pfn = 0;
553 bm->cur.node_bit = 0;
556 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
559 struct list_head hook;
565 * free_mem_extents - Free a list of memory extents.
566 * @list: List of extents to free.
568 static void free_mem_extents(struct list_head *list)
570 struct mem_extent *ext, *aux;
572 list_for_each_entry_safe(ext, aux, list, hook) {
573 list_del(&ext->hook);
579 * create_mem_extents - Create a list of memory extents.
580 * @list: List to put the extents into.
581 * @gfp_mask: Mask to use for memory allocations.
583 * The extents represent contiguous ranges of PFNs.
585 static int create_mem_extents(struct list_head *list, gfp_t gfp_mask)
589 INIT_LIST_HEAD(list);
591 for_each_populated_zone(zone) {
592 unsigned long zone_start, zone_end;
593 struct mem_extent *ext, *cur, *aux;
595 zone_start = zone->zone_start_pfn;
596 zone_end = zone_end_pfn(zone);
598 list_for_each_entry(ext, list, hook)
599 if (zone_start <= ext->end)
602 if (&ext->hook == list || zone_end < ext->start) {
603 /* New extent is necessary */
604 struct mem_extent *new_ext;
606 new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask);
608 free_mem_extents(list);
611 new_ext->start = zone_start;
612 new_ext->end = zone_end;
613 list_add_tail(&new_ext->hook, &ext->hook);
617 /* Merge this zone's range of PFNs with the existing one */
618 if (zone_start < ext->start)
619 ext->start = zone_start;
620 if (zone_end > ext->end)
623 /* More merging may be possible */
625 list_for_each_entry_safe_continue(cur, aux, list, hook) {
626 if (zone_end < cur->start)
628 if (zone_end < cur->end)
630 list_del(&cur->hook);
639 * memory_bm_create - Allocate memory for a memory bitmap.
641 static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask,
644 struct chain_allocator ca;
645 struct list_head mem_extents;
646 struct mem_extent *ext;
649 chain_init(&ca, gfp_mask, safe_needed);
650 INIT_LIST_HEAD(&bm->zones);
652 error = create_mem_extents(&mem_extents, gfp_mask);
656 list_for_each_entry(ext, &mem_extents, hook) {
657 struct mem_zone_bm_rtree *zone;
659 zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca,
660 ext->start, ext->end);
665 list_add_tail(&zone->list, &bm->zones);
668 bm->p_list = ca.chain;
669 memory_bm_position_reset(bm);
671 free_mem_extents(&mem_extents);
675 bm->p_list = ca.chain;
676 memory_bm_free(bm, PG_UNSAFE_CLEAR);
681 * memory_bm_free - Free memory occupied by the memory bitmap.
682 * @bm: Memory bitmap.
684 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free)
686 struct mem_zone_bm_rtree *zone;
688 list_for_each_entry(zone, &bm->zones, list)
689 free_zone_bm_rtree(zone, clear_nosave_free);
691 free_list_of_pages(bm->p_list, clear_nosave_free);
693 INIT_LIST_HEAD(&bm->zones);
697 * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap.
699 * Find the bit in memory bitmap @bm that corresponds to the given PFN.
700 * The cur.zone, cur.block and cur.node_pfn members of @bm are updated.
702 * Walk the radix tree to find the page containing the bit that represents @pfn
703 * and return the position of the bit in @addr and @bit_nr.
705 static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
706 void **addr, unsigned int *bit_nr)
708 struct mem_zone_bm_rtree *curr, *zone;
709 struct rtree_node *node;
714 if (pfn >= zone->start_pfn && pfn < zone->end_pfn)
719 /* Find the right zone */
720 list_for_each_entry(curr, &bm->zones, list) {
721 if (pfn >= curr->start_pfn && pfn < curr->end_pfn) {
732 * We have found the zone. Now walk the radix tree to find the leaf node
737 * If the zone we wish to scan is the the current zone and the
738 * pfn falls into the current node then we do not need to walk
742 if (zone == bm->cur.zone &&
743 ((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn)
747 block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT;
749 for (i = zone->levels; i > 0; i--) {
752 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
753 index &= BM_RTREE_LEVEL_MASK;
754 BUG_ON(node->data[index] == 0);
755 node = (struct rtree_node *)node->data[index];
759 /* Update last position */
762 bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK;
764 /* Set return values */
766 *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK;
771 static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn)
777 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
782 static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn)
788 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
795 static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn)
801 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
803 clear_bit(bit, addr);
806 static void memory_bm_clear_current(struct memory_bitmap *bm)
810 bit = max(bm->cur.node_bit - 1, 0);
811 clear_bit(bit, bm->cur.node->data);
814 static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
820 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
822 return test_bit(bit, addr);
825 static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn)
830 return !memory_bm_find_bit(bm, pfn, &addr, &bit);
834 * rtree_next_node - Jump to the next leaf node.
836 * Set the position to the beginning of the next node in the
837 * memory bitmap. This is either the next node in the current
838 * zone's radix tree or the first node in the radix tree of the
841 * Return true if there is a next node, false otherwise.
843 static bool rtree_next_node(struct memory_bitmap *bm)
845 if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) {
846 bm->cur.node = list_entry(bm->cur.node->list.next,
847 struct rtree_node, list);
848 bm->cur.node_pfn += BM_BITS_PER_BLOCK;
849 bm->cur.node_bit = 0;
850 touch_softlockup_watchdog();
854 /* No more nodes, goto next zone */
855 if (!list_is_last(&bm->cur.zone->list, &bm->zones)) {
856 bm->cur.zone = list_entry(bm->cur.zone->list.next,
857 struct mem_zone_bm_rtree, list);
858 bm->cur.node = list_entry(bm->cur.zone->leaves.next,
859 struct rtree_node, list);
860 bm->cur.node_pfn = 0;
861 bm->cur.node_bit = 0;
870 * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap.
871 * @bm: Memory bitmap.
873 * Starting from the last returned position this function searches for the next
874 * set bit in @bm and returns the PFN represented by it. If no more bits are
875 * set, BM_END_OF_MAP is returned.
877 * It is required to run memory_bm_position_reset() before the first call to
878 * this function for the given memory bitmap.
880 static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
882 unsigned long bits, pfn, pages;
886 pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn;
887 bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK);
888 bit = find_next_bit(bm->cur.node->data, bits,
891 pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit;
892 bm->cur.node_bit = bit + 1;
895 } while (rtree_next_node(bm));
897 return BM_END_OF_MAP;
901 * This structure represents a range of page frames the contents of which
902 * should not be saved during hibernation.
904 struct nosave_region {
905 struct list_head list;
906 unsigned long start_pfn;
907 unsigned long end_pfn;
910 static LIST_HEAD(nosave_regions);
912 static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone)
914 struct rtree_node *node;
916 list_for_each_entry(node, &zone->nodes, list)
917 recycle_safe_page(node->data);
919 list_for_each_entry(node, &zone->leaves, list)
920 recycle_safe_page(node->data);
923 static void memory_bm_recycle(struct memory_bitmap *bm)
925 struct mem_zone_bm_rtree *zone;
926 struct linked_page *p_list;
928 list_for_each_entry(zone, &bm->zones, list)
929 recycle_zone_bm_rtree(zone);
933 struct linked_page *lp = p_list;
936 recycle_safe_page(lp);
941 * register_nosave_region - Register a region of unsaveable memory.
943 * Register a range of page frames the contents of which should not be saved
944 * during hibernation (to be used in the early initialization code).
946 void __init __register_nosave_region(unsigned long start_pfn,
947 unsigned long end_pfn, int use_kmalloc)
949 struct nosave_region *region;
951 if (start_pfn >= end_pfn)
954 if (!list_empty(&nosave_regions)) {
955 /* Try to extend the previous region (they should be sorted) */
956 region = list_entry(nosave_regions.prev,
957 struct nosave_region, list);
958 if (region->end_pfn == start_pfn) {
959 region->end_pfn = end_pfn;
964 /* During init, this shouldn't fail */
965 region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL);
968 /* This allocation cannot fail */
969 region = memblock_virt_alloc(sizeof(struct nosave_region), 0);
971 region->start_pfn = start_pfn;
972 region->end_pfn = end_pfn;
973 list_add_tail(®ion->list, &nosave_regions);
975 printk(KERN_INFO "PM: Registered nosave memory: [mem %#010llx-%#010llx]\n",
976 (unsigned long long) start_pfn << PAGE_SHIFT,
977 ((unsigned long long) end_pfn << PAGE_SHIFT) - 1);
981 * Set bits in this map correspond to the page frames the contents of which
982 * should not be saved during the suspend.
984 static struct memory_bitmap *forbidden_pages_map;
986 /* Set bits in this map correspond to free page frames. */
987 static struct memory_bitmap *free_pages_map;
990 * Each page frame allocated for creating the image is marked by setting the
991 * corresponding bits in forbidden_pages_map and free_pages_map simultaneously
994 void swsusp_set_page_free(struct page *page)
997 memory_bm_set_bit(free_pages_map, page_to_pfn(page));
1000 static int swsusp_page_is_free(struct page *page)
1002 return free_pages_map ?
1003 memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0;
1006 void swsusp_unset_page_free(struct page *page)
1009 memory_bm_clear_bit(free_pages_map, page_to_pfn(page));
1012 static void swsusp_set_page_forbidden(struct page *page)
1014 if (forbidden_pages_map)
1015 memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page));
1018 int swsusp_page_is_forbidden(struct page *page)
1020 return forbidden_pages_map ?
1021 memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0;
1024 static void swsusp_unset_page_forbidden(struct page *page)
1026 if (forbidden_pages_map)
1027 memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page));
1031 * mark_nosave_pages - Mark pages that should not be saved.
1032 * @bm: Memory bitmap.
1034 * Set the bits in @bm that correspond to the page frames the contents of which
1035 * should not be saved.
1037 static void mark_nosave_pages(struct memory_bitmap *bm)
1039 struct nosave_region *region;
1041 if (list_empty(&nosave_regions))
1044 list_for_each_entry(region, &nosave_regions, list) {
1047 pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n",
1048 (unsigned long long) region->start_pfn << PAGE_SHIFT,
1049 ((unsigned long long) region->end_pfn << PAGE_SHIFT)
1052 for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++)
1053 if (pfn_valid(pfn)) {
1055 * It is safe to ignore the result of
1056 * mem_bm_set_bit_check() here, since we won't
1057 * touch the PFNs for which the error is
1060 mem_bm_set_bit_check(bm, pfn);
1066 * create_basic_memory_bitmaps - Create bitmaps to hold basic page information.
1068 * Create bitmaps needed for marking page frames that should not be saved and
1069 * free page frames. The forbidden_pages_map and free_pages_map pointers are
1070 * only modified if everything goes well, because we don't want the bits to be
1071 * touched before both bitmaps are set up.
1073 int create_basic_memory_bitmaps(void)
1075 struct memory_bitmap *bm1, *bm2;
1078 if (forbidden_pages_map && free_pages_map)
1081 BUG_ON(forbidden_pages_map || free_pages_map);
1083 bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1087 error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
1089 goto Free_first_object;
1091 bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1093 goto Free_first_bitmap;
1095 error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
1097 goto Free_second_object;
1099 forbidden_pages_map = bm1;
1100 free_pages_map = bm2;
1101 mark_nosave_pages(forbidden_pages_map);
1103 pr_debug("PM: Basic memory bitmaps created\n");
1110 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1117 * free_basic_memory_bitmaps - Free memory bitmaps holding basic information.
1119 * Free memory bitmaps allocated by create_basic_memory_bitmaps(). The
1120 * auxiliary pointers are necessary so that the bitmaps themselves are not
1121 * referred to while they are being freed.
1123 void free_basic_memory_bitmaps(void)
1125 struct memory_bitmap *bm1, *bm2;
1127 if (WARN_ON(!(forbidden_pages_map && free_pages_map)))
1130 bm1 = forbidden_pages_map;
1131 bm2 = free_pages_map;
1132 forbidden_pages_map = NULL;
1133 free_pages_map = NULL;
1134 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1136 memory_bm_free(bm2, PG_UNSAFE_CLEAR);
1139 pr_debug("PM: Basic memory bitmaps freed\n");
1142 void clear_free_pages(void)
1144 #ifdef CONFIG_PAGE_POISONING_ZERO
1145 struct memory_bitmap *bm = free_pages_map;
1148 if (WARN_ON(!(free_pages_map)))
1151 memory_bm_position_reset(bm);
1152 pfn = memory_bm_next_pfn(bm);
1153 while (pfn != BM_END_OF_MAP) {
1155 clear_highpage(pfn_to_page(pfn));
1157 pfn = memory_bm_next_pfn(bm);
1159 memory_bm_position_reset(bm);
1160 pr_info("PM: free pages cleared after restore\n");
1161 #endif /* PAGE_POISONING_ZERO */
1165 * snapshot_additional_pages - Estimate the number of extra pages needed.
1166 * @zone: Memory zone to carry out the computation for.
1168 * Estimate the number of additional pages needed for setting up a hibernation
1169 * image data structures for @zone (usually, the returned value is greater than
1170 * the exact number).
1172 unsigned int snapshot_additional_pages(struct zone *zone)
1174 unsigned int rtree, nodes;
1176 rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
1177 rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node),
1178 LINKED_PAGE_DATA_SIZE);
1180 nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL);
1187 #ifdef CONFIG_HIGHMEM
1189 * count_free_highmem_pages - Compute the total number of free highmem pages.
1191 * The returned number is system-wide.
1193 static unsigned int count_free_highmem_pages(void)
1196 unsigned int cnt = 0;
1198 for_each_populated_zone(zone)
1199 if (is_highmem(zone))
1200 cnt += zone_page_state(zone, NR_FREE_PAGES);
1206 * saveable_highmem_page - Check if a highmem page is saveable.
1208 * Determine whether a highmem page should be included in a hibernation image.
1210 * We should save the page if it isn't Nosave or NosaveFree, or Reserved,
1211 * and it isn't part of a free chunk of pages.
1213 static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn)
1217 if (!pfn_valid(pfn))
1220 page = pfn_to_page(pfn);
1221 if (page_zone(page) != zone)
1224 BUG_ON(!PageHighMem(page));
1226 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) ||
1230 if (page_is_guard(page))
1237 * count_highmem_pages - Compute the total number of saveable highmem pages.
1239 static unsigned int count_highmem_pages(void)
1244 for_each_populated_zone(zone) {
1245 unsigned long pfn, max_zone_pfn;
1247 if (!is_highmem(zone))
1250 mark_free_pages(zone);
1251 max_zone_pfn = zone_end_pfn(zone);
1252 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1253 if (saveable_highmem_page(zone, pfn))
1259 static inline void *saveable_highmem_page(struct zone *z, unsigned long p)
1263 #endif /* CONFIG_HIGHMEM */
1266 * saveable_page - Check if the given page is saveable.
1268 * Determine whether a non-highmem page should be included in a hibernation
1271 * We should save the page if it isn't Nosave, and is not in the range
1272 * of pages statically defined as 'unsaveable', and it isn't part of
1273 * a free chunk of pages.
1275 static struct page *saveable_page(struct zone *zone, unsigned long pfn)
1279 if (!pfn_valid(pfn))
1282 page = pfn_to_page(pfn);
1283 if (page_zone(page) != zone)
1286 BUG_ON(PageHighMem(page));
1288 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
1291 if (PageReserved(page)
1292 && (!kernel_page_present(page) || pfn_is_nosave(pfn)))
1295 if (page_is_guard(page))
1302 * count_data_pages - Compute the total number of saveable non-highmem pages.
1304 static unsigned int count_data_pages(void)
1307 unsigned long pfn, max_zone_pfn;
1310 for_each_populated_zone(zone) {
1311 if (is_highmem(zone))
1314 mark_free_pages(zone);
1315 max_zone_pfn = zone_end_pfn(zone);
1316 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1317 if (saveable_page(zone, pfn))
1324 * This is needed, because copy_page and memcpy are not usable for copying
1327 static inline void do_copy_page(long *dst, long *src)
1331 for (n = PAGE_SIZE / sizeof(long); n; n--)
1336 * safe_copy_page - Copy a page in a safe way.
1338 * Check if the page we are going to copy is marked as present in the kernel
1339 * page tables (this always is the case if CONFIG_DEBUG_PAGEALLOC is not set
1340 * and in that case kernel_page_present() always returns 'true').
1342 static void safe_copy_page(void *dst, struct page *s_page)
1344 if (kernel_page_present(s_page)) {
1345 do_copy_page(dst, page_address(s_page));
1347 kernel_map_pages(s_page, 1, 1);
1348 do_copy_page(dst, page_address(s_page));
1349 kernel_map_pages(s_page, 1, 0);
1353 #ifdef CONFIG_HIGHMEM
1354 static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn)
1356 return is_highmem(zone) ?
1357 saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn);
1360 static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1362 struct page *s_page, *d_page;
1365 s_page = pfn_to_page(src_pfn);
1366 d_page = pfn_to_page(dst_pfn);
1367 if (PageHighMem(s_page)) {
1368 src = kmap_atomic(s_page);
1369 dst = kmap_atomic(d_page);
1370 do_copy_page(dst, src);
1374 if (PageHighMem(d_page)) {
1376 * The page pointed to by src may contain some kernel
1377 * data modified by kmap_atomic()
1379 safe_copy_page(buffer, s_page);
1380 dst = kmap_atomic(d_page);
1381 copy_page(dst, buffer);
1384 safe_copy_page(page_address(d_page), s_page);
1389 #define page_is_saveable(zone, pfn) saveable_page(zone, pfn)
1391 static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1393 safe_copy_page(page_address(pfn_to_page(dst_pfn)),
1394 pfn_to_page(src_pfn));
1396 #endif /* CONFIG_HIGHMEM */
1398 static void copy_data_pages(struct memory_bitmap *copy_bm,
1399 struct memory_bitmap *orig_bm)
1404 for_each_populated_zone(zone) {
1405 unsigned long max_zone_pfn;
1407 mark_free_pages(zone);
1408 max_zone_pfn = zone_end_pfn(zone);
1409 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1410 if (page_is_saveable(zone, pfn))
1411 memory_bm_set_bit(orig_bm, pfn);
1413 memory_bm_position_reset(orig_bm);
1414 memory_bm_position_reset(copy_bm);
1416 pfn = memory_bm_next_pfn(orig_bm);
1417 if (unlikely(pfn == BM_END_OF_MAP))
1419 copy_data_page(memory_bm_next_pfn(copy_bm), pfn);
1423 /* Total number of image pages */
1424 static unsigned int nr_copy_pages;
1425 /* Number of pages needed for saving the original pfns of the image pages */
1426 static unsigned int nr_meta_pages;
1428 * Numbers of normal and highmem page frames allocated for hibernation image
1429 * before suspending devices.
1431 unsigned int alloc_normal, alloc_highmem;
1433 * Memory bitmap used for marking saveable pages (during hibernation) or
1434 * hibernation image pages (during restore)
1436 static struct memory_bitmap orig_bm;
1438 * Memory bitmap used during hibernation for marking allocated page frames that
1439 * will contain copies of saveable pages. During restore it is initially used
1440 * for marking hibernation image pages, but then the set bits from it are
1441 * duplicated in @orig_bm and it is released. On highmem systems it is next
1442 * used for marking "safe" highmem pages, but it has to be reinitialized for
1445 static struct memory_bitmap copy_bm;
1448 * swsusp_free - Free pages allocated for hibernation image.
1450 * Image pages are alocated before snapshot creation, so they need to be
1451 * released after resume.
1453 void swsusp_free(void)
1455 unsigned long fb_pfn, fr_pfn;
1457 if (!forbidden_pages_map || !free_pages_map)
1460 memory_bm_position_reset(forbidden_pages_map);
1461 memory_bm_position_reset(free_pages_map);
1464 fr_pfn = memory_bm_next_pfn(free_pages_map);
1465 fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1468 * Find the next bit set in both bitmaps. This is guaranteed to
1469 * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP.
1472 if (fb_pfn < fr_pfn)
1473 fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1474 if (fr_pfn < fb_pfn)
1475 fr_pfn = memory_bm_next_pfn(free_pages_map);
1476 } while (fb_pfn != fr_pfn);
1478 if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) {
1479 struct page *page = pfn_to_page(fr_pfn);
1481 memory_bm_clear_current(forbidden_pages_map);
1482 memory_bm_clear_current(free_pages_map);
1483 hibernate_restore_unprotect_page(page_address(page));
1491 restore_pblist = NULL;
1495 hibernate_restore_protection_end();
1498 /* Helper functions used for the shrinking of memory. */
1500 #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN)
1503 * preallocate_image_pages - Allocate a number of pages for hibernation image.
1504 * @nr_pages: Number of page frames to allocate.
1505 * @mask: GFP flags to use for the allocation.
1507 * Return value: Number of page frames actually allocated
1509 static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask)
1511 unsigned long nr_alloc = 0;
1513 while (nr_pages > 0) {
1516 page = alloc_image_page(mask);
1519 memory_bm_set_bit(©_bm, page_to_pfn(page));
1520 if (PageHighMem(page))
1531 static unsigned long preallocate_image_memory(unsigned long nr_pages,
1532 unsigned long avail_normal)
1534 unsigned long alloc;
1536 if (avail_normal <= alloc_normal)
1539 alloc = avail_normal - alloc_normal;
1540 if (nr_pages < alloc)
1543 return preallocate_image_pages(alloc, GFP_IMAGE);
1546 #ifdef CONFIG_HIGHMEM
1547 static unsigned long preallocate_image_highmem(unsigned long nr_pages)
1549 return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM);
1553 * __fraction - Compute (an approximation of) x * (multiplier / base).
1555 static unsigned long __fraction(u64 x, u64 multiplier, u64 base)
1559 return (unsigned long)x;
1562 static unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1563 unsigned long highmem,
1564 unsigned long total)
1566 unsigned long alloc = __fraction(nr_pages, highmem, total);
1568 return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM);
1570 #else /* CONFIG_HIGHMEM */
1571 static inline unsigned long preallocate_image_highmem(unsigned long nr_pages)
1576 static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1577 unsigned long highmem,
1578 unsigned long total)
1582 #endif /* CONFIG_HIGHMEM */
1585 * free_unnecessary_pages - Release preallocated pages not needed for the image.
1587 static unsigned long free_unnecessary_pages(void)
1589 unsigned long save, to_free_normal, to_free_highmem, free;
1591 save = count_data_pages();
1592 if (alloc_normal >= save) {
1593 to_free_normal = alloc_normal - save;
1597 save -= alloc_normal;
1599 save += count_highmem_pages();
1600 if (alloc_highmem >= save) {
1601 to_free_highmem = alloc_highmem - save;
1603 to_free_highmem = 0;
1604 save -= alloc_highmem;
1605 if (to_free_normal > save)
1606 to_free_normal -= save;
1610 free = to_free_normal + to_free_highmem;
1612 memory_bm_position_reset(©_bm);
1614 while (to_free_normal > 0 || to_free_highmem > 0) {
1615 unsigned long pfn = memory_bm_next_pfn(©_bm);
1616 struct page *page = pfn_to_page(pfn);
1618 if (PageHighMem(page)) {
1619 if (!to_free_highmem)
1624 if (!to_free_normal)
1629 memory_bm_clear_bit(©_bm, pfn);
1630 swsusp_unset_page_forbidden(page);
1631 swsusp_unset_page_free(page);
1639 * minimum_image_size - Estimate the minimum acceptable size of an image.
1640 * @saveable: Number of saveable pages in the system.
1642 * We want to avoid attempting to free too much memory too hard, so estimate the
1643 * minimum acceptable size of a hibernation image to use as the lower limit for
1644 * preallocating memory.
1646 * We assume that the minimum image size should be proportional to
1648 * [number of saveable pages] - [number of pages that can be freed in theory]
1650 * where the second term is the sum of (1) reclaimable slab pages, (2) active
1651 * and (3) inactive anonymous pages, (4) active and (5) inactive file pages,
1652 * minus mapped file pages.
1654 static unsigned long minimum_image_size(unsigned long saveable)
1658 size = global_page_state(NR_SLAB_RECLAIMABLE)
1659 + global_node_page_state(NR_ACTIVE_ANON)
1660 + global_node_page_state(NR_INACTIVE_ANON)
1661 + global_node_page_state(NR_ACTIVE_FILE)
1662 + global_node_page_state(NR_INACTIVE_FILE)
1663 - global_node_page_state(NR_FILE_MAPPED);
1665 return saveable <= size ? 0 : saveable - size;
1669 * hibernate_preallocate_memory - Preallocate memory for hibernation image.
1671 * To create a hibernation image it is necessary to make a copy of every page
1672 * frame in use. We also need a number of page frames to be free during
1673 * hibernation for allocations made while saving the image and for device
1674 * drivers, in case they need to allocate memory from their hibernation
1675 * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough
1676 * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through
1677 * /sys/power/reserved_size, respectively). To make this happen, we compute the
1678 * total number of available page frames and allocate at least
1680 * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2
1681 * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE)
1683 * of them, which corresponds to the maximum size of a hibernation image.
1685 * If image_size is set below the number following from the above formula,
1686 * the preallocation of memory is continued until the total number of saveable
1687 * pages in the system is below the requested image size or the minimum
1688 * acceptable image size returned by minimum_image_size(), whichever is greater.
1690 int hibernate_preallocate_memory(void)
1693 unsigned long saveable, size, max_size, count, highmem, pages = 0;
1694 unsigned long alloc, save_highmem, pages_highmem, avail_normal;
1695 ktime_t start, stop;
1698 printk(KERN_INFO "PM: Preallocating image memory... ");
1699 start = ktime_get();
1701 error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY);
1705 error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY);
1712 /* Count the number of saveable data pages. */
1713 save_highmem = count_highmem_pages();
1714 saveable = count_data_pages();
1717 * Compute the total number of page frames we can use (count) and the
1718 * number of pages needed for image metadata (size).
1721 saveable += save_highmem;
1722 highmem = save_highmem;
1724 for_each_populated_zone(zone) {
1725 size += snapshot_additional_pages(zone);
1726 if (is_highmem(zone))
1727 highmem += zone_page_state(zone, NR_FREE_PAGES);
1729 count += zone_page_state(zone, NR_FREE_PAGES);
1731 avail_normal = count;
1733 count -= totalreserve_pages;
1735 /* Add number of pages required for page keys (s390 only). */
1736 size += page_key_additional_pages(saveable);
1738 /* Compute the maximum number of saveable pages to leave in memory. */
1739 max_size = (count - (size + PAGES_FOR_IO)) / 2
1740 - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE);
1741 /* Compute the desired number of image pages specified by image_size. */
1742 size = DIV_ROUND_UP(image_size, PAGE_SIZE);
1743 if (size > max_size)
1746 * If the desired number of image pages is at least as large as the
1747 * current number of saveable pages in memory, allocate page frames for
1748 * the image and we're done.
1750 if (size >= saveable) {
1751 pages = preallocate_image_highmem(save_highmem);
1752 pages += preallocate_image_memory(saveable - pages, avail_normal);
1756 /* Estimate the minimum size of the image. */
1757 pages = minimum_image_size(saveable);
1759 * To avoid excessive pressure on the normal zone, leave room in it to
1760 * accommodate an image of the minimum size (unless it's already too
1761 * small, in which case don't preallocate pages from it at all).
1763 if (avail_normal > pages)
1764 avail_normal -= pages;
1768 size = min_t(unsigned long, pages, max_size);
1771 * Let the memory management subsystem know that we're going to need a
1772 * large number of page frames to allocate and make it free some memory.
1773 * NOTE: If this is not done, performance will be hurt badly in some
1776 shrink_all_memory(saveable - size);
1779 * The number of saveable pages in memory was too high, so apply some
1780 * pressure to decrease it. First, make room for the largest possible
1781 * image and fail if that doesn't work. Next, try to decrease the size
1782 * of the image as much as indicated by 'size' using allocations from
1783 * highmem and non-highmem zones separately.
1785 pages_highmem = preallocate_image_highmem(highmem / 2);
1786 alloc = count - max_size;
1787 if (alloc > pages_highmem)
1788 alloc -= pages_highmem;
1791 pages = preallocate_image_memory(alloc, avail_normal);
1792 if (pages < alloc) {
1793 /* We have exhausted non-highmem pages, try highmem. */
1795 pages += pages_highmem;
1796 pages_highmem = preallocate_image_highmem(alloc);
1797 if (pages_highmem < alloc)
1799 pages += pages_highmem;
1801 * size is the desired number of saveable pages to leave in
1802 * memory, so try to preallocate (all memory - size) pages.
1804 alloc = (count - pages) - size;
1805 pages += preallocate_image_highmem(alloc);
1808 * There are approximately max_size saveable pages at this point
1809 * and we want to reduce this number down to size.
1811 alloc = max_size - size;
1812 size = preallocate_highmem_fraction(alloc, highmem, count);
1813 pages_highmem += size;
1815 size = preallocate_image_memory(alloc, avail_normal);
1816 pages_highmem += preallocate_image_highmem(alloc - size);
1817 pages += pages_highmem + size;
1821 * We only need as many page frames for the image as there are saveable
1822 * pages in memory, but we have allocated more. Release the excessive
1825 pages -= free_unnecessary_pages();
1829 printk(KERN_CONT "done (allocated %lu pages)\n", pages);
1830 swsusp_show_speed(start, stop, pages, "Allocated");
1835 printk(KERN_CONT "\n");
1840 #ifdef CONFIG_HIGHMEM
1842 * count_pages_for_highmem - Count non-highmem pages needed for copying highmem.
1844 * Compute the number of non-highmem pages that will be necessary for creating
1845 * copies of highmem pages.
1847 static unsigned int count_pages_for_highmem(unsigned int nr_highmem)
1849 unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem;
1851 if (free_highmem >= nr_highmem)
1854 nr_highmem -= free_highmem;
1859 static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; }
1860 #endif /* CONFIG_HIGHMEM */
1863 * enough_free_mem - Check if there is enough free memory for the image.
1865 static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem)
1868 unsigned int free = alloc_normal;
1870 for_each_populated_zone(zone)
1871 if (!is_highmem(zone))
1872 free += zone_page_state(zone, NR_FREE_PAGES);
1874 nr_pages += count_pages_for_highmem(nr_highmem);
1875 pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n",
1876 nr_pages, PAGES_FOR_IO, free);
1878 return free > nr_pages + PAGES_FOR_IO;
1881 #ifdef CONFIG_HIGHMEM
1883 * get_highmem_buffer - Allocate a buffer for highmem pages.
1885 * If there are some highmem pages in the hibernation image, we may need a
1886 * buffer to copy them and/or load their data.
1888 static inline int get_highmem_buffer(int safe_needed)
1890 buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed);
1891 return buffer ? 0 : -ENOMEM;
1895 * alloc_highmem_image_pages - Allocate some highmem pages for the image.
1897 * Try to allocate as many pages as needed, but if the number of free highmem
1898 * pages is less than that, allocate them all.
1900 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1901 unsigned int nr_highmem)
1903 unsigned int to_alloc = count_free_highmem_pages();
1905 if (to_alloc > nr_highmem)
1906 to_alloc = nr_highmem;
1908 nr_highmem -= to_alloc;
1909 while (to_alloc-- > 0) {
1912 page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM);
1913 memory_bm_set_bit(bm, page_to_pfn(page));
1918 static inline int get_highmem_buffer(int safe_needed) { return 0; }
1920 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1921 unsigned int n) { return 0; }
1922 #endif /* CONFIG_HIGHMEM */
1925 * swsusp_alloc - Allocate memory for hibernation image.
1927 * We first try to allocate as many highmem pages as there are
1928 * saveable highmem pages in the system. If that fails, we allocate
1929 * non-highmem pages for the copies of the remaining highmem ones.
1931 * In this approach it is likely that the copies of highmem pages will
1932 * also be located in the high memory, because of the way in which
1933 * copy_data_pages() works.
1935 static int swsusp_alloc(struct memory_bitmap *orig_bm,
1936 struct memory_bitmap *copy_bm,
1937 unsigned int nr_pages, unsigned int nr_highmem)
1939 if (nr_highmem > 0) {
1940 if (get_highmem_buffer(PG_ANY))
1942 if (nr_highmem > alloc_highmem) {
1943 nr_highmem -= alloc_highmem;
1944 nr_pages += alloc_highmem_pages(copy_bm, nr_highmem);
1947 if (nr_pages > alloc_normal) {
1948 nr_pages -= alloc_normal;
1949 while (nr_pages-- > 0) {
1952 page = alloc_image_page(GFP_ATOMIC | __GFP_COLD);
1955 memory_bm_set_bit(copy_bm, page_to_pfn(page));
1966 asmlinkage __visible int swsusp_save(void)
1968 unsigned int nr_pages, nr_highmem;
1970 printk(KERN_INFO "PM: Creating hibernation image:\n");
1972 drain_local_pages(NULL);
1973 nr_pages = count_data_pages();
1974 nr_highmem = count_highmem_pages();
1975 printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem);
1977 if (!enough_free_mem(nr_pages, nr_highmem)) {
1978 printk(KERN_ERR "PM: Not enough free memory\n");
1982 if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) {
1983 printk(KERN_ERR "PM: Memory allocation failed\n");
1988 * During allocating of suspend pagedir, new cold pages may appear.
1991 drain_local_pages(NULL);
1992 copy_data_pages(©_bm, &orig_bm);
1995 * End of critical section. From now on, we can write to memory,
1996 * but we should not touch disk. This specially means we must _not_
1997 * touch swap space! Except we must write out our image of course.
2000 nr_pages += nr_highmem;
2001 nr_copy_pages = nr_pages;
2002 nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE);
2004 printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n",
2010 #ifndef CONFIG_ARCH_HIBERNATION_HEADER
2011 static int init_header_complete(struct swsusp_info *info)
2013 memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname));
2014 info->version_code = LINUX_VERSION_CODE;
2018 static char *check_image_kernel(struct swsusp_info *info)
2020 if (info->version_code != LINUX_VERSION_CODE)
2021 return "kernel version";
2022 if (strcmp(info->uts.sysname,init_utsname()->sysname))
2023 return "system type";
2024 if (strcmp(info->uts.release,init_utsname()->release))
2025 return "kernel release";
2026 if (strcmp(info->uts.version,init_utsname()->version))
2028 if (strcmp(info->uts.machine,init_utsname()->machine))
2032 #endif /* CONFIG_ARCH_HIBERNATION_HEADER */
2034 unsigned long snapshot_get_image_size(void)
2036 return nr_copy_pages + nr_meta_pages + 1;
2039 static int init_header(struct swsusp_info *info)
2041 memset(info, 0, sizeof(struct swsusp_info));
2042 info->num_physpages = get_num_physpages();
2043 info->image_pages = nr_copy_pages;
2044 info->pages = snapshot_get_image_size();
2045 info->size = info->pages;
2046 info->size <<= PAGE_SHIFT;
2047 return init_header_complete(info);
2051 * pack_pfns - Prepare PFNs for saving.
2052 * @bm: Memory bitmap.
2053 * @buf: Memory buffer to store the PFNs in.
2055 * PFNs corresponding to set bits in @bm are stored in the area of memory
2056 * pointed to by @buf (1 page at a time).
2058 static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
2062 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2063 buf[j] = memory_bm_next_pfn(bm);
2064 if (unlikely(buf[j] == BM_END_OF_MAP))
2066 /* Save page key for data page (s390 only). */
2067 page_key_read(buf + j);
2072 * snapshot_read_next - Get the address to read the next image page from.
2073 * @handle: Snapshot handle to be used for the reading.
2075 * On the first call, @handle should point to a zeroed snapshot_handle
2076 * structure. The structure gets populated then and a pointer to it should be
2077 * passed to this function every next time.
2079 * On success, the function returns a positive number. Then, the caller
2080 * is allowed to read up to the returned number of bytes from the memory
2081 * location computed by the data_of() macro.
2083 * The function returns 0 to indicate the end of the data stream condition,
2084 * and negative numbers are returned on errors. If that happens, the structure
2085 * pointed to by @handle is not updated and should not be used any more.
2087 int snapshot_read_next(struct snapshot_handle *handle)
2089 if (handle->cur > nr_meta_pages + nr_copy_pages)
2093 /* This makes the buffer be freed by swsusp_free() */
2094 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2101 error = init_header((struct swsusp_info *)buffer);
2104 handle->buffer = buffer;
2105 memory_bm_position_reset(&orig_bm);
2106 memory_bm_position_reset(©_bm);
2107 } else if (handle->cur <= nr_meta_pages) {
2109 pack_pfns(buffer, &orig_bm);
2113 page = pfn_to_page(memory_bm_next_pfn(©_bm));
2114 if (PageHighMem(page)) {
2116 * Highmem pages are copied to the buffer,
2117 * because we can't return with a kmapped
2118 * highmem page (we may not be called again).
2122 kaddr = kmap_atomic(page);
2123 copy_page(buffer, kaddr);
2124 kunmap_atomic(kaddr);
2125 handle->buffer = buffer;
2127 handle->buffer = page_address(page);
2134 static void duplicate_memory_bitmap(struct memory_bitmap *dst,
2135 struct memory_bitmap *src)
2139 memory_bm_position_reset(src);
2140 pfn = memory_bm_next_pfn(src);
2141 while (pfn != BM_END_OF_MAP) {
2142 memory_bm_set_bit(dst, pfn);
2143 pfn = memory_bm_next_pfn(src);
2148 * mark_unsafe_pages - Mark pages that were used before hibernation.
2150 * Mark the pages that cannot be used for storing the image during restoration,
2151 * because they conflict with the pages that had been used before hibernation.
2153 static void mark_unsafe_pages(struct memory_bitmap *bm)
2157 /* Clear the "free"/"unsafe" bit for all PFNs */
2158 memory_bm_position_reset(free_pages_map);
2159 pfn = memory_bm_next_pfn(free_pages_map);
2160 while (pfn != BM_END_OF_MAP) {
2161 memory_bm_clear_current(free_pages_map);
2162 pfn = memory_bm_next_pfn(free_pages_map);
2165 /* Mark pages that correspond to the "original" PFNs as "unsafe" */
2166 duplicate_memory_bitmap(free_pages_map, bm);
2168 allocated_unsafe_pages = 0;
2171 static int check_header(struct swsusp_info *info)
2175 reason = check_image_kernel(info);
2176 if (!reason && info->num_physpages != get_num_physpages())
2177 reason = "memory size";
2179 printk(KERN_ERR "PM: Image mismatch: %s\n", reason);
2186 * load header - Check the image header and copy the data from it.
2188 static int load_header(struct swsusp_info *info)
2192 restore_pblist = NULL;
2193 error = check_header(info);
2195 nr_copy_pages = info->image_pages;
2196 nr_meta_pages = info->pages - info->image_pages - 1;
2202 * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap.
2203 * @bm: Memory bitmap.
2204 * @buf: Area of memory containing the PFNs.
2206 * For each element of the array pointed to by @buf (1 page at a time), set the
2207 * corresponding bit in @bm.
2209 static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
2213 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2214 if (unlikely(buf[j] == BM_END_OF_MAP))
2217 /* Extract and buffer page key for data page (s390 only). */
2218 page_key_memorize(buf + j);
2220 if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j]))
2221 memory_bm_set_bit(bm, buf[j]);
2229 #ifdef CONFIG_HIGHMEM
2231 * struct highmem_pbe is used for creating the list of highmem pages that
2232 * should be restored atomically during the resume from disk, because the page
2233 * frames they have occupied before the suspend are in use.
2235 struct highmem_pbe {
2236 struct page *copy_page; /* data is here now */
2237 struct page *orig_page; /* data was here before the suspend */
2238 struct highmem_pbe *next;
2242 * List of highmem PBEs needed for restoring the highmem pages that were
2243 * allocated before the suspend and included in the suspend image, but have
2244 * also been allocated by the "resume" kernel, so their contents cannot be
2245 * written directly to their "original" page frames.
2247 static struct highmem_pbe *highmem_pblist;
2250 * count_highmem_image_pages - Compute the number of highmem pages in the image.
2251 * @bm: Memory bitmap.
2253 * The bits in @bm that correspond to image pages are assumed to be set.
2255 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm)
2258 unsigned int cnt = 0;
2260 memory_bm_position_reset(bm);
2261 pfn = memory_bm_next_pfn(bm);
2262 while (pfn != BM_END_OF_MAP) {
2263 if (PageHighMem(pfn_to_page(pfn)))
2266 pfn = memory_bm_next_pfn(bm);
2271 static unsigned int safe_highmem_pages;
2273 static struct memory_bitmap *safe_highmem_bm;
2276 * prepare_highmem_image - Allocate memory for loading highmem data from image.
2277 * @bm: Pointer to an uninitialized memory bitmap structure.
2278 * @nr_highmem_p: Pointer to the number of highmem image pages.
2280 * Try to allocate as many highmem pages as there are highmem image pages
2281 * (@nr_highmem_p points to the variable containing the number of highmem image
2282 * pages). The pages that are "safe" (ie. will not be overwritten when the
2283 * hibernation image is restored entirely) have the corresponding bits set in
2284 * @bm (it must be unitialized).
2286 * NOTE: This function should not be called if there are no highmem image pages.
2288 static int prepare_highmem_image(struct memory_bitmap *bm,
2289 unsigned int *nr_highmem_p)
2291 unsigned int to_alloc;
2293 if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE))
2296 if (get_highmem_buffer(PG_SAFE))
2299 to_alloc = count_free_highmem_pages();
2300 if (to_alloc > *nr_highmem_p)
2301 to_alloc = *nr_highmem_p;
2303 *nr_highmem_p = to_alloc;
2305 safe_highmem_pages = 0;
2306 while (to_alloc-- > 0) {
2309 page = alloc_page(__GFP_HIGHMEM);
2310 if (!swsusp_page_is_free(page)) {
2311 /* The page is "safe", set its bit the bitmap */
2312 memory_bm_set_bit(bm, page_to_pfn(page));
2313 safe_highmem_pages++;
2315 /* Mark the page as allocated */
2316 swsusp_set_page_forbidden(page);
2317 swsusp_set_page_free(page);
2319 memory_bm_position_reset(bm);
2320 safe_highmem_bm = bm;
2324 static struct page *last_highmem_page;
2327 * get_highmem_page_buffer - Prepare a buffer to store a highmem image page.
2329 * For a given highmem image page get a buffer that suspend_write_next() should
2330 * return to its caller to write to.
2332 * If the page is to be saved to its "original" page frame or a copy of
2333 * the page is to be made in the highmem, @buffer is returned. Otherwise,
2334 * the copy of the page is to be made in normal memory, so the address of
2335 * the copy is returned.
2337 * If @buffer is returned, the caller of suspend_write_next() will write
2338 * the page's contents to @buffer, so they will have to be copied to the
2339 * right location on the next call to suspend_write_next() and it is done
2340 * with the help of copy_last_highmem_page(). For this purpose, if
2341 * @buffer is returned, @last_highmem_page is set to the page to which
2342 * the data will have to be copied from @buffer.
2344 static void *get_highmem_page_buffer(struct page *page,
2345 struct chain_allocator *ca)
2347 struct highmem_pbe *pbe;
2350 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) {
2352 * We have allocated the "original" page frame and we can
2353 * use it directly to store the loaded page.
2355 last_highmem_page = page;
2359 * The "original" page frame has not been allocated and we have to
2360 * use a "safe" page frame to store the loaded page.
2362 pbe = chain_alloc(ca, sizeof(struct highmem_pbe));
2365 return ERR_PTR(-ENOMEM);
2367 pbe->orig_page = page;
2368 if (safe_highmem_pages > 0) {
2371 /* Copy of the page will be stored in high memory */
2373 tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm));
2374 safe_highmem_pages--;
2375 last_highmem_page = tmp;
2376 pbe->copy_page = tmp;
2378 /* Copy of the page will be stored in normal memory */
2379 kaddr = safe_pages_list;
2380 safe_pages_list = safe_pages_list->next;
2381 pbe->copy_page = virt_to_page(kaddr);
2383 pbe->next = highmem_pblist;
2384 highmem_pblist = pbe;
2389 * copy_last_highmem_page - Copy most the most recent highmem image page.
2391 * Copy the contents of a highmem image from @buffer, where the caller of
2392 * snapshot_write_next() has stored them, to the right location represented by
2393 * @last_highmem_page .
2395 static void copy_last_highmem_page(void)
2397 if (last_highmem_page) {
2400 dst = kmap_atomic(last_highmem_page);
2401 copy_page(dst, buffer);
2403 last_highmem_page = NULL;
2407 static inline int last_highmem_page_copied(void)
2409 return !last_highmem_page;
2412 static inline void free_highmem_data(void)
2414 if (safe_highmem_bm)
2415 memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR);
2418 free_image_page(buffer, PG_UNSAFE_CLEAR);
2421 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; }
2423 static inline int prepare_highmem_image(struct memory_bitmap *bm,
2424 unsigned int *nr_highmem_p) { return 0; }
2426 static inline void *get_highmem_page_buffer(struct page *page,
2427 struct chain_allocator *ca)
2429 return ERR_PTR(-EINVAL);
2432 static inline void copy_last_highmem_page(void) {}
2433 static inline int last_highmem_page_copied(void) { return 1; }
2434 static inline void free_highmem_data(void) {}
2435 #endif /* CONFIG_HIGHMEM */
2437 #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe))
2440 * prepare_image - Make room for loading hibernation image.
2441 * @new_bm: Unitialized memory bitmap structure.
2442 * @bm: Memory bitmap with unsafe pages marked.
2444 * Use @bm to mark the pages that will be overwritten in the process of
2445 * restoring the system memory state from the suspend image ("unsafe" pages)
2446 * and allocate memory for the image.
2448 * The idea is to allocate a new memory bitmap first and then allocate
2449 * as many pages as needed for image data, but without specifying what those
2450 * pages will be used for just yet. Instead, we mark them all as allocated and
2451 * create a lists of "safe" pages to be used later. On systems with high
2452 * memory a list of "safe" highmem pages is created too.
2454 static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm)
2456 unsigned int nr_pages, nr_highmem;
2457 struct linked_page *lp;
2460 /* If there is no highmem, the buffer will not be necessary */
2461 free_image_page(buffer, PG_UNSAFE_CLEAR);
2464 nr_highmem = count_highmem_image_pages(bm);
2465 mark_unsafe_pages(bm);
2467 error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE);
2471 duplicate_memory_bitmap(new_bm, bm);
2472 memory_bm_free(bm, PG_UNSAFE_KEEP);
2473 if (nr_highmem > 0) {
2474 error = prepare_highmem_image(bm, &nr_highmem);
2479 * Reserve some safe pages for potential later use.
2481 * NOTE: This way we make sure there will be enough safe pages for the
2482 * chain_alloc() in get_buffer(). It is a bit wasteful, but
2483 * nr_copy_pages cannot be greater than 50% of the memory anyway.
2485 * nr_copy_pages cannot be less than allocated_unsafe_pages too.
2487 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2488 nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE);
2489 while (nr_pages > 0) {
2490 lp = get_image_page(GFP_ATOMIC, PG_SAFE);
2495 lp->next = safe_pages_list;
2496 safe_pages_list = lp;
2499 /* Preallocate memory for the image */
2500 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2501 while (nr_pages > 0) {
2502 lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC);
2507 if (!swsusp_page_is_free(virt_to_page(lp))) {
2508 /* The page is "safe", add it to the list */
2509 lp->next = safe_pages_list;
2510 safe_pages_list = lp;
2512 /* Mark the page as allocated */
2513 swsusp_set_page_forbidden(virt_to_page(lp));
2514 swsusp_set_page_free(virt_to_page(lp));
2525 * get_buffer - Get the address to store the next image data page.
2527 * Get the address that snapshot_write_next() should return to its caller to
2530 static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca)
2534 unsigned long pfn = memory_bm_next_pfn(bm);
2536 if (pfn == BM_END_OF_MAP)
2537 return ERR_PTR(-EFAULT);
2539 page = pfn_to_page(pfn);
2540 if (PageHighMem(page))
2541 return get_highmem_page_buffer(page, ca);
2543 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page))
2545 * We have allocated the "original" page frame and we can
2546 * use it directly to store the loaded page.
2548 return page_address(page);
2551 * The "original" page frame has not been allocated and we have to
2552 * use a "safe" page frame to store the loaded page.
2554 pbe = chain_alloc(ca, sizeof(struct pbe));
2557 return ERR_PTR(-ENOMEM);
2559 pbe->orig_address = page_address(page);
2560 pbe->address = safe_pages_list;
2561 safe_pages_list = safe_pages_list->next;
2562 pbe->next = restore_pblist;
2563 restore_pblist = pbe;
2564 return pbe->address;
2568 * snapshot_write_next - Get the address to store the next image page.
2569 * @handle: Snapshot handle structure to guide the writing.
2571 * On the first call, @handle should point to a zeroed snapshot_handle
2572 * structure. The structure gets populated then and a pointer to it should be
2573 * passed to this function every next time.
2575 * On success, the function returns a positive number. Then, the caller
2576 * is allowed to write up to the returned number of bytes to the memory
2577 * location computed by the data_of() macro.
2579 * The function returns 0 to indicate the "end of file" condition. Negative
2580 * numbers are returned on errors, in which cases the structure pointed to by
2581 * @handle is not updated and should not be used any more.
2583 int snapshot_write_next(struct snapshot_handle *handle)
2585 static struct chain_allocator ca;
2588 /* Check if we have already loaded the entire image */
2589 if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages)
2592 handle->sync_read = 1;
2596 /* This makes the buffer be freed by swsusp_free() */
2597 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2602 handle->buffer = buffer;
2603 } else if (handle->cur == 1) {
2604 error = load_header(buffer);
2608 safe_pages_list = NULL;
2610 error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY);
2614 /* Allocate buffer for page keys. */
2615 error = page_key_alloc(nr_copy_pages);
2619 hibernate_restore_protection_begin();
2620 } else if (handle->cur <= nr_meta_pages + 1) {
2621 error = unpack_orig_pfns(buffer, ©_bm);
2625 if (handle->cur == nr_meta_pages + 1) {
2626 error = prepare_image(&orig_bm, ©_bm);
2630 chain_init(&ca, GFP_ATOMIC, PG_SAFE);
2631 memory_bm_position_reset(&orig_bm);
2632 restore_pblist = NULL;
2633 handle->buffer = get_buffer(&orig_bm, &ca);
2634 handle->sync_read = 0;
2635 if (IS_ERR(handle->buffer))
2636 return PTR_ERR(handle->buffer);
2639 copy_last_highmem_page();
2640 /* Restore page key for data page (s390 only). */
2641 page_key_write(handle->buffer);
2642 hibernate_restore_protect_page(handle->buffer);
2643 handle->buffer = get_buffer(&orig_bm, &ca);
2644 if (IS_ERR(handle->buffer))
2645 return PTR_ERR(handle->buffer);
2646 if (handle->buffer != buffer)
2647 handle->sync_read = 0;
2654 * snapshot_write_finalize - Complete the loading of a hibernation image.
2656 * Must be called after the last call to snapshot_write_next() in case the last
2657 * page in the image happens to be a highmem page and its contents should be
2658 * stored in highmem. Additionally, it recycles bitmap memory that's not
2659 * necessary any more.
2661 void snapshot_write_finalize(struct snapshot_handle *handle)
2663 copy_last_highmem_page();
2664 /* Restore page key for data page (s390 only). */
2665 page_key_write(handle->buffer);
2667 hibernate_restore_protect_page(handle->buffer);
2668 /* Do that only if we have loaded the image entirely */
2669 if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) {
2670 memory_bm_recycle(&orig_bm);
2671 free_highmem_data();
2675 int snapshot_image_loaded(struct snapshot_handle *handle)
2677 return !(!nr_copy_pages || !last_highmem_page_copied() ||
2678 handle->cur <= nr_meta_pages + nr_copy_pages);
2681 #ifdef CONFIG_HIGHMEM
2682 /* Assumes that @buf is ready and points to a "safe" page */
2683 static inline void swap_two_pages_data(struct page *p1, struct page *p2,
2686 void *kaddr1, *kaddr2;
2688 kaddr1 = kmap_atomic(p1);
2689 kaddr2 = kmap_atomic(p2);
2690 copy_page(buf, kaddr1);
2691 copy_page(kaddr1, kaddr2);
2692 copy_page(kaddr2, buf);
2693 kunmap_atomic(kaddr2);
2694 kunmap_atomic(kaddr1);
2698 * restore_highmem - Put highmem image pages into their original locations.
2700 * For each highmem page that was in use before hibernation and is included in
2701 * the image, and also has been allocated by the "restore" kernel, swap its
2702 * current contents with the previous (ie. "before hibernation") ones.
2704 * If the restore eventually fails, we can call this function once again and
2705 * restore the highmem state as seen by the restore kernel.
2707 int restore_highmem(void)
2709 struct highmem_pbe *pbe = highmem_pblist;
2715 buf = get_image_page(GFP_ATOMIC, PG_SAFE);
2720 swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf);
2723 free_image_page(buf, PG_UNSAFE_CLEAR);
2726 #endif /* CONFIG_HIGHMEM */