1 // SPDX-License-Identifier: GPL-2.0-only
3 * linux/kernel/power/snapshot.c
5 * This file provides system snapshot/restore functionality for swsusp.
7 * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz>
8 * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
11 #define pr_fmt(fmt) "PM: hibernation: " fmt
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/memblock.h>
25 #include <linux/nmi.h>
26 #include <linux/syscalls.h>
27 #include <linux/console.h>
28 #include <linux/highmem.h>
29 #include <linux/list.h>
30 #include <linux/slab.h>
31 #include <linux/compiler.h>
32 #include <linux/ktime.h>
33 #include <linux/set_memory.h>
35 #include <linux/uaccess.h>
36 #include <asm/mmu_context.h>
37 #include <asm/tlbflush.h>
42 #if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_ARCH_HAS_SET_MEMORY)
43 static bool hibernate_restore_protection;
44 static bool hibernate_restore_protection_active;
46 void enable_restore_image_protection(void)
48 hibernate_restore_protection = true;
51 static inline void hibernate_restore_protection_begin(void)
53 hibernate_restore_protection_active = hibernate_restore_protection;
56 static inline void hibernate_restore_protection_end(void)
58 hibernate_restore_protection_active = false;
61 static inline void hibernate_restore_protect_page(void *page_address)
63 if (hibernate_restore_protection_active)
64 set_memory_ro((unsigned long)page_address, 1);
67 static inline void hibernate_restore_unprotect_page(void *page_address)
69 if (hibernate_restore_protection_active)
70 set_memory_rw((unsigned long)page_address, 1);
73 static inline void hibernate_restore_protection_begin(void) {}
74 static inline void hibernate_restore_protection_end(void) {}
75 static inline void hibernate_restore_protect_page(void *page_address) {}
76 static inline void hibernate_restore_unprotect_page(void *page_address) {}
77 #endif /* CONFIG_STRICT_KERNEL_RWX && CONFIG_ARCH_HAS_SET_MEMORY */
79 static int swsusp_page_is_free(struct page *);
80 static void swsusp_set_page_forbidden(struct page *);
81 static void swsusp_unset_page_forbidden(struct page *);
84 * Number of bytes to reserve for memory allocations made by device drivers
85 * from their ->freeze() and ->freeze_noirq() callbacks so that they don't
86 * cause image creation to fail (tunable via /sys/power/reserved_size).
88 unsigned long reserved_size;
90 void __init hibernate_reserved_size_init(void)
92 reserved_size = SPARE_PAGES * PAGE_SIZE;
96 * Preferred image size in bytes (tunable via /sys/power/image_size).
97 * When it is set to N, swsusp will do its best to ensure the image
98 * size will not exceed N bytes, but if that is impossible, it will
99 * try to create the smallest image possible.
101 unsigned long image_size;
103 void __init hibernate_image_size_init(void)
105 image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE;
109 * List of PBEs needed for restoring the pages that were allocated before
110 * the suspend and included in the suspend image, but have also been
111 * allocated by the "resume" kernel, so their contents cannot be written
112 * directly to their "original" page frames.
114 struct pbe *restore_pblist;
116 /* struct linked_page is used to build chains of pages */
118 #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *))
121 struct linked_page *next;
122 char data[LINKED_PAGE_DATA_SIZE];
126 * List of "safe" pages (ie. pages that were not used by the image kernel
127 * before hibernation) that may be used as temporary storage for image kernel
130 static struct linked_page *safe_pages_list;
132 /* Pointer to an auxiliary buffer (1 page) */
137 #define PG_UNSAFE_CLEAR 1
138 #define PG_UNSAFE_KEEP 0
140 static unsigned int allocated_unsafe_pages;
143 * get_image_page - Allocate a page for a hibernation image.
144 * @gfp_mask: GFP mask for the allocation.
145 * @safe_needed: Get pages that were not used before hibernation (restore only)
147 * During image restoration, for storing the PBE list and the image data, we can
148 * only use memory pages that do not conflict with the pages used before
149 * hibernation. The "unsafe" pages have PageNosaveFree set and we count them
150 * using allocated_unsafe_pages.
152 * Each allocated image page is marked as PageNosave and PageNosaveFree so that
153 * swsusp_free() can release it.
155 static void *get_image_page(gfp_t gfp_mask, int safe_needed)
159 res = (void *)get_zeroed_page(gfp_mask);
161 while (res && swsusp_page_is_free(virt_to_page(res))) {
162 /* The page is unsafe, mark it for swsusp_free() */
163 swsusp_set_page_forbidden(virt_to_page(res));
164 allocated_unsafe_pages++;
165 res = (void *)get_zeroed_page(gfp_mask);
168 swsusp_set_page_forbidden(virt_to_page(res));
169 swsusp_set_page_free(virt_to_page(res));
174 static void *__get_safe_page(gfp_t gfp_mask)
176 if (safe_pages_list) {
177 void *ret = safe_pages_list;
179 safe_pages_list = safe_pages_list->next;
180 memset(ret, 0, PAGE_SIZE);
183 return get_image_page(gfp_mask, PG_SAFE);
186 unsigned long get_safe_page(gfp_t gfp_mask)
188 return (unsigned long)__get_safe_page(gfp_mask);
191 static struct page *alloc_image_page(gfp_t gfp_mask)
195 page = alloc_page(gfp_mask);
197 swsusp_set_page_forbidden(page);
198 swsusp_set_page_free(page);
203 static void recycle_safe_page(void *page_address)
205 struct linked_page *lp = page_address;
207 lp->next = safe_pages_list;
208 safe_pages_list = lp;
212 * free_image_page - Free a page allocated for hibernation image.
213 * @addr: Address of the page to free.
214 * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page.
216 * The page to free should have been allocated by get_image_page() (page flags
217 * set by it are affected).
219 static inline void free_image_page(void *addr, int clear_nosave_free)
223 BUG_ON(!virt_addr_valid(addr));
225 page = virt_to_page(addr);
227 swsusp_unset_page_forbidden(page);
228 if (clear_nosave_free)
229 swsusp_unset_page_free(page);
234 static inline void free_list_of_pages(struct linked_page *list,
235 int clear_page_nosave)
238 struct linked_page *lp = list->next;
240 free_image_page(list, clear_page_nosave);
246 * struct chain_allocator is used for allocating small objects out of
247 * a linked list of pages called 'the chain'.
249 * The chain grows each time when there is no room for a new object in
250 * the current page. The allocated objects cannot be freed individually.
251 * It is only possible to free them all at once, by freeing the entire
254 * NOTE: The chain allocator may be inefficient if the allocated objects
255 * are not much smaller than PAGE_SIZE.
257 struct chain_allocator {
258 struct linked_page *chain; /* the chain */
259 unsigned int used_space; /* total size of objects allocated out
260 of the current page */
261 gfp_t gfp_mask; /* mask for allocating pages */
262 int safe_needed; /* if set, only "safe" pages are allocated */
265 static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask,
269 ca->used_space = LINKED_PAGE_DATA_SIZE;
270 ca->gfp_mask = gfp_mask;
271 ca->safe_needed = safe_needed;
274 static void *chain_alloc(struct chain_allocator *ca, unsigned int size)
278 if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) {
279 struct linked_page *lp;
281 lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) :
282 get_image_page(ca->gfp_mask, PG_ANY);
286 lp->next = ca->chain;
290 ret = ca->chain->data + ca->used_space;
291 ca->used_space += size;
296 * Data types related to memory bitmaps.
298 * Memory bitmap is a structure consiting of many linked lists of
299 * objects. The main list's elements are of type struct zone_bitmap
300 * and each of them corresonds to one zone. For each zone bitmap
301 * object there is a list of objects of type struct bm_block that
302 * represent each blocks of bitmap in which information is stored.
304 * struct memory_bitmap contains a pointer to the main list of zone
305 * bitmap objects, a struct bm_position used for browsing the bitmap,
306 * and a pointer to the list of pages used for allocating all of the
307 * zone bitmap objects and bitmap block objects.
309 * NOTE: It has to be possible to lay out the bitmap in memory
310 * using only allocations of order 0. Additionally, the bitmap is
311 * designed to work with arbitrary number of zones (this is over the
312 * top for now, but let's avoid making unnecessary assumptions ;-).
314 * struct zone_bitmap contains a pointer to a list of bitmap block
315 * objects and a pointer to the bitmap block object that has been
316 * most recently used for setting bits. Additionally, it contains the
317 * PFNs that correspond to the start and end of the represented zone.
319 * struct bm_block contains a pointer to the memory page in which
320 * information is stored (in the form of a block of bitmap)
321 * It also contains the pfns that correspond to the start and end of
322 * the represented memory area.
324 * The memory bitmap is organized as a radix tree to guarantee fast random
325 * access to the bits. There is one radix tree for each zone (as returned
326 * from create_mem_extents).
328 * One radix tree is represented by one struct mem_zone_bm_rtree. There are
329 * two linked lists for the nodes of the tree, one for the inner nodes and
330 * one for the leave nodes. The linked leave nodes are used for fast linear
331 * access of the memory bitmap.
333 * The struct rtree_node represents one node of the radix tree.
336 #define BM_END_OF_MAP (~0UL)
338 #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE)
339 #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3)
340 #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1)
343 * struct rtree_node is a wrapper struct to link the nodes
344 * of the rtree together for easy linear iteration over
345 * bits and easy freeing
348 struct list_head list;
353 * struct mem_zone_bm_rtree represents a bitmap used for one
354 * populated memory zone.
356 struct mem_zone_bm_rtree {
357 struct list_head list; /* Link Zones together */
358 struct list_head nodes; /* Radix Tree inner nodes */
359 struct list_head leaves; /* Radix Tree leaves */
360 unsigned long start_pfn; /* Zone start page frame */
361 unsigned long end_pfn; /* Zone end page frame + 1 */
362 struct rtree_node *rtree; /* Radix Tree Root */
363 int levels; /* Number of Radix Tree Levels */
364 unsigned int blocks; /* Number of Bitmap Blocks */
367 /* strcut bm_position is used for browsing memory bitmaps */
370 struct mem_zone_bm_rtree *zone;
371 struct rtree_node *node;
372 unsigned long node_pfn;
376 struct memory_bitmap {
377 struct list_head zones;
378 struct linked_page *p_list; /* list of pages used to store zone
379 bitmap objects and bitmap block
381 struct bm_position cur; /* most recently used bit position */
384 /* Functions that operate on memory bitmaps */
386 #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long))
387 #if BITS_PER_LONG == 32
388 #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2)
390 #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3)
392 #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1)
395 * alloc_rtree_node - Allocate a new node and add it to the radix tree.
397 * This function is used to allocate inner nodes as well as the
398 * leave nodes of the radix tree. It also adds the node to the
399 * corresponding linked list passed in by the *list parameter.
401 static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed,
402 struct chain_allocator *ca,
403 struct list_head *list)
405 struct rtree_node *node;
407 node = chain_alloc(ca, sizeof(struct rtree_node));
411 node->data = get_image_page(gfp_mask, safe_needed);
415 list_add_tail(&node->list, list);
421 * add_rtree_block - Add a new leave node to the radix tree.
423 * The leave nodes need to be allocated in order to keep the leaves
424 * linked list in order. This is guaranteed by the zone->blocks
427 static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask,
428 int safe_needed, struct chain_allocator *ca)
430 struct rtree_node *node, *block, **dst;
431 unsigned int levels_needed, block_nr;
434 block_nr = zone->blocks;
437 /* How many levels do we need for this block nr? */
440 block_nr >>= BM_RTREE_LEVEL_SHIFT;
443 /* Make sure the rtree has enough levels */
444 for (i = zone->levels; i < levels_needed; i++) {
445 node = alloc_rtree_node(gfp_mask, safe_needed, ca,
450 node->data[0] = (unsigned long)zone->rtree;
455 /* Allocate new block */
456 block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves);
460 /* Now walk the rtree to insert the block */
463 block_nr = zone->blocks;
464 for (i = zone->levels; i > 0; i--) {
468 node = alloc_rtree_node(gfp_mask, safe_needed, ca,
475 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
476 index &= BM_RTREE_LEVEL_MASK;
477 dst = (struct rtree_node **)&((*dst)->data[index]);
487 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
488 int clear_nosave_free);
491 * create_zone_bm_rtree - Create a radix tree for one zone.
493 * Allocated the mem_zone_bm_rtree structure and initializes it.
494 * This function also allocated and builds the radix tree for the
497 static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask,
499 struct chain_allocator *ca,
503 struct mem_zone_bm_rtree *zone;
504 unsigned int i, nr_blocks;
508 zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree));
512 INIT_LIST_HEAD(&zone->nodes);
513 INIT_LIST_HEAD(&zone->leaves);
514 zone->start_pfn = start;
516 nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK);
518 for (i = 0; i < nr_blocks; i++) {
519 if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) {
520 free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR);
529 * free_zone_bm_rtree - Free the memory of the radix tree.
531 * Free all node pages of the radix tree. The mem_zone_bm_rtree
532 * structure itself is not freed here nor are the rtree_node
535 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
536 int clear_nosave_free)
538 struct rtree_node *node;
540 list_for_each_entry(node, &zone->nodes, list)
541 free_image_page(node->data, clear_nosave_free);
543 list_for_each_entry(node, &zone->leaves, list)
544 free_image_page(node->data, clear_nosave_free);
547 static void memory_bm_position_reset(struct memory_bitmap *bm)
549 bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree,
551 bm->cur.node = list_entry(bm->cur.zone->leaves.next,
552 struct rtree_node, list);
553 bm->cur.node_pfn = 0;
554 bm->cur.node_bit = 0;
557 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
560 struct list_head hook;
566 * free_mem_extents - Free a list of memory extents.
567 * @list: List of extents to free.
569 static void free_mem_extents(struct list_head *list)
571 struct mem_extent *ext, *aux;
573 list_for_each_entry_safe(ext, aux, list, hook) {
574 list_del(&ext->hook);
580 * create_mem_extents - Create a list of memory extents.
581 * @list: List to put the extents into.
582 * @gfp_mask: Mask to use for memory allocations.
584 * The extents represent contiguous ranges of PFNs.
586 static int create_mem_extents(struct list_head *list, gfp_t gfp_mask)
590 INIT_LIST_HEAD(list);
592 for_each_populated_zone(zone) {
593 unsigned long zone_start, zone_end;
594 struct mem_extent *ext, *cur, *aux;
596 zone_start = zone->zone_start_pfn;
597 zone_end = zone_end_pfn(zone);
599 list_for_each_entry(ext, list, hook)
600 if (zone_start <= ext->end)
603 if (&ext->hook == list || zone_end < ext->start) {
604 /* New extent is necessary */
605 struct mem_extent *new_ext;
607 new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask);
609 free_mem_extents(list);
612 new_ext->start = zone_start;
613 new_ext->end = zone_end;
614 list_add_tail(&new_ext->hook, &ext->hook);
618 /* Merge this zone's range of PFNs with the existing one */
619 if (zone_start < ext->start)
620 ext->start = zone_start;
621 if (zone_end > ext->end)
624 /* More merging may be possible */
626 list_for_each_entry_safe_continue(cur, aux, list, hook) {
627 if (zone_end < cur->start)
629 if (zone_end < cur->end)
631 list_del(&cur->hook);
640 * memory_bm_create - Allocate memory for a memory bitmap.
642 static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask,
645 struct chain_allocator ca;
646 struct list_head mem_extents;
647 struct mem_extent *ext;
650 chain_init(&ca, gfp_mask, safe_needed);
651 INIT_LIST_HEAD(&bm->zones);
653 error = create_mem_extents(&mem_extents, gfp_mask);
657 list_for_each_entry(ext, &mem_extents, hook) {
658 struct mem_zone_bm_rtree *zone;
660 zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca,
661 ext->start, ext->end);
666 list_add_tail(&zone->list, &bm->zones);
669 bm->p_list = ca.chain;
670 memory_bm_position_reset(bm);
672 free_mem_extents(&mem_extents);
676 bm->p_list = ca.chain;
677 memory_bm_free(bm, PG_UNSAFE_CLEAR);
682 * memory_bm_free - Free memory occupied by the memory bitmap.
683 * @bm: Memory bitmap.
685 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free)
687 struct mem_zone_bm_rtree *zone;
689 list_for_each_entry(zone, &bm->zones, list)
690 free_zone_bm_rtree(zone, clear_nosave_free);
692 free_list_of_pages(bm->p_list, clear_nosave_free);
694 INIT_LIST_HEAD(&bm->zones);
698 * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap.
700 * Find the bit in memory bitmap @bm that corresponds to the given PFN.
701 * The cur.zone, cur.block and cur.node_pfn members of @bm are updated.
703 * Walk the radix tree to find the page containing the bit that represents @pfn
704 * and return the position of the bit in @addr and @bit_nr.
706 static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
707 void **addr, unsigned int *bit_nr)
709 struct mem_zone_bm_rtree *curr, *zone;
710 struct rtree_node *node;
715 if (pfn >= zone->start_pfn && pfn < zone->end_pfn)
720 /* Find the right zone */
721 list_for_each_entry(curr, &bm->zones, list) {
722 if (pfn >= curr->start_pfn && pfn < curr->end_pfn) {
733 * We have found the zone. Now walk the radix tree to find the leaf node
738 * If the zone we wish to scan is the current zone and the
739 * pfn falls into the current node then we do not need to walk
743 if (zone == bm->cur.zone &&
744 ((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn)
748 block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT;
750 for (i = zone->levels; i > 0; i--) {
753 index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
754 index &= BM_RTREE_LEVEL_MASK;
755 BUG_ON(node->data[index] == 0);
756 node = (struct rtree_node *)node->data[index];
760 /* Update last position */
763 bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK;
765 /* Set return values */
767 *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK;
772 static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn)
778 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
783 static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn)
789 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
796 static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn)
802 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
804 clear_bit(bit, addr);
807 static void memory_bm_clear_current(struct memory_bitmap *bm)
811 bit = max(bm->cur.node_bit - 1, 0);
812 clear_bit(bit, bm->cur.node->data);
815 static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
821 error = memory_bm_find_bit(bm, pfn, &addr, &bit);
823 return test_bit(bit, addr);
826 static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn)
831 return !memory_bm_find_bit(bm, pfn, &addr, &bit);
835 * rtree_next_node - Jump to the next leaf node.
837 * Set the position to the beginning of the next node in the
838 * memory bitmap. This is either the next node in the current
839 * zone's radix tree or the first node in the radix tree of the
842 * Return true if there is a next node, false otherwise.
844 static bool rtree_next_node(struct memory_bitmap *bm)
846 if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) {
847 bm->cur.node = list_entry(bm->cur.node->list.next,
848 struct rtree_node, list);
849 bm->cur.node_pfn += BM_BITS_PER_BLOCK;
850 bm->cur.node_bit = 0;
851 touch_softlockup_watchdog();
855 /* No more nodes, goto next zone */
856 if (!list_is_last(&bm->cur.zone->list, &bm->zones)) {
857 bm->cur.zone = list_entry(bm->cur.zone->list.next,
858 struct mem_zone_bm_rtree, list);
859 bm->cur.node = list_entry(bm->cur.zone->leaves.next,
860 struct rtree_node, list);
861 bm->cur.node_pfn = 0;
862 bm->cur.node_bit = 0;
871 * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap.
872 * @bm: Memory bitmap.
874 * Starting from the last returned position this function searches for the next
875 * set bit in @bm and returns the PFN represented by it. If no more bits are
876 * set, BM_END_OF_MAP is returned.
878 * It is required to run memory_bm_position_reset() before the first call to
879 * this function for the given memory bitmap.
881 static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
883 unsigned long bits, pfn, pages;
887 pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn;
888 bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK);
889 bit = find_next_bit(bm->cur.node->data, bits,
892 pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit;
893 bm->cur.node_bit = bit + 1;
896 } while (rtree_next_node(bm));
898 return BM_END_OF_MAP;
902 * This structure represents a range of page frames the contents of which
903 * should not be saved during hibernation.
905 struct nosave_region {
906 struct list_head list;
907 unsigned long start_pfn;
908 unsigned long end_pfn;
911 static LIST_HEAD(nosave_regions);
913 static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone)
915 struct rtree_node *node;
917 list_for_each_entry(node, &zone->nodes, list)
918 recycle_safe_page(node->data);
920 list_for_each_entry(node, &zone->leaves, list)
921 recycle_safe_page(node->data);
924 static void memory_bm_recycle(struct memory_bitmap *bm)
926 struct mem_zone_bm_rtree *zone;
927 struct linked_page *p_list;
929 list_for_each_entry(zone, &bm->zones, list)
930 recycle_zone_bm_rtree(zone);
934 struct linked_page *lp = p_list;
937 recycle_safe_page(lp);
942 * register_nosave_region - Register a region of unsaveable memory.
944 * Register a range of page frames the contents of which should not be saved
945 * during hibernation (to be used in the early initialization code).
947 void __init register_nosave_region(unsigned long start_pfn, unsigned long end_pfn)
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;
963 /* This allocation cannot fail */
964 region = memblock_alloc(sizeof(struct nosave_region),
967 panic("%s: Failed to allocate %zu bytes\n", __func__,
968 sizeof(struct nosave_region));
969 region->start_pfn = start_pfn;
970 region->end_pfn = end_pfn;
971 list_add_tail(®ion->list, &nosave_regions);
973 pr_info("Registered nosave memory: [mem %#010llx-%#010llx]\n",
974 (unsigned long long) start_pfn << PAGE_SHIFT,
975 ((unsigned long long) end_pfn << PAGE_SHIFT) - 1);
979 * Set bits in this map correspond to the page frames the contents of which
980 * should not be saved during the suspend.
982 static struct memory_bitmap *forbidden_pages_map;
984 /* Set bits in this map correspond to free page frames. */
985 static struct memory_bitmap *free_pages_map;
988 * Each page frame allocated for creating the image is marked by setting the
989 * corresponding bits in forbidden_pages_map and free_pages_map simultaneously
992 void swsusp_set_page_free(struct page *page)
995 memory_bm_set_bit(free_pages_map, page_to_pfn(page));
998 static int swsusp_page_is_free(struct page *page)
1000 return free_pages_map ?
1001 memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0;
1004 void swsusp_unset_page_free(struct page *page)
1007 memory_bm_clear_bit(free_pages_map, page_to_pfn(page));
1010 static void swsusp_set_page_forbidden(struct page *page)
1012 if (forbidden_pages_map)
1013 memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page));
1016 int swsusp_page_is_forbidden(struct page *page)
1018 return forbidden_pages_map ?
1019 memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0;
1022 static void swsusp_unset_page_forbidden(struct page *page)
1024 if (forbidden_pages_map)
1025 memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page));
1029 * mark_nosave_pages - Mark pages that should not be saved.
1030 * @bm: Memory bitmap.
1032 * Set the bits in @bm that correspond to the page frames the contents of which
1033 * should not be saved.
1035 static void mark_nosave_pages(struct memory_bitmap *bm)
1037 struct nosave_region *region;
1039 if (list_empty(&nosave_regions))
1042 list_for_each_entry(region, &nosave_regions, list) {
1045 pr_debug("Marking nosave pages: [mem %#010llx-%#010llx]\n",
1046 (unsigned long long) region->start_pfn << PAGE_SHIFT,
1047 ((unsigned long long) region->end_pfn << PAGE_SHIFT)
1050 for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++)
1051 if (pfn_valid(pfn)) {
1053 * It is safe to ignore the result of
1054 * mem_bm_set_bit_check() here, since we won't
1055 * touch the PFNs for which the error is
1058 mem_bm_set_bit_check(bm, pfn);
1064 * create_basic_memory_bitmaps - Create bitmaps to hold basic page information.
1066 * Create bitmaps needed for marking page frames that should not be saved and
1067 * free page frames. The forbidden_pages_map and free_pages_map pointers are
1068 * only modified if everything goes well, because we don't want the bits to be
1069 * touched before both bitmaps are set up.
1071 int create_basic_memory_bitmaps(void)
1073 struct memory_bitmap *bm1, *bm2;
1076 if (forbidden_pages_map && free_pages_map)
1079 BUG_ON(forbidden_pages_map || free_pages_map);
1081 bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1085 error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
1087 goto Free_first_object;
1089 bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1091 goto Free_first_bitmap;
1093 error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
1095 goto Free_second_object;
1097 forbidden_pages_map = bm1;
1098 free_pages_map = bm2;
1099 mark_nosave_pages(forbidden_pages_map);
1101 pr_debug("Basic memory bitmaps created\n");
1108 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1115 * free_basic_memory_bitmaps - Free memory bitmaps holding basic information.
1117 * Free memory bitmaps allocated by create_basic_memory_bitmaps(). The
1118 * auxiliary pointers are necessary so that the bitmaps themselves are not
1119 * referred to while they are being freed.
1121 void free_basic_memory_bitmaps(void)
1123 struct memory_bitmap *bm1, *bm2;
1125 if (WARN_ON(!(forbidden_pages_map && free_pages_map)))
1128 bm1 = forbidden_pages_map;
1129 bm2 = free_pages_map;
1130 forbidden_pages_map = NULL;
1131 free_pages_map = NULL;
1132 memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1134 memory_bm_free(bm2, PG_UNSAFE_CLEAR);
1137 pr_debug("Basic memory bitmaps freed\n");
1140 void clear_free_pages(void)
1142 struct memory_bitmap *bm = free_pages_map;
1145 if (WARN_ON(!(free_pages_map)))
1148 if (IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) || want_init_on_free()) {
1149 memory_bm_position_reset(bm);
1150 pfn = memory_bm_next_pfn(bm);
1151 while (pfn != BM_END_OF_MAP) {
1153 clear_highpage(pfn_to_page(pfn));
1155 pfn = memory_bm_next_pfn(bm);
1157 memory_bm_position_reset(bm);
1158 pr_info("free pages cleared after restore\n");
1163 * snapshot_additional_pages - Estimate the number of extra pages needed.
1164 * @zone: Memory zone to carry out the computation for.
1166 * Estimate the number of additional pages needed for setting up a hibernation
1167 * image data structures for @zone (usually, the returned value is greater than
1168 * the exact number).
1170 unsigned int snapshot_additional_pages(struct zone *zone)
1172 unsigned int rtree, nodes;
1174 rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
1175 rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node),
1176 LINKED_PAGE_DATA_SIZE);
1178 nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL);
1185 #ifdef CONFIG_HIGHMEM
1187 * count_free_highmem_pages - Compute the total number of free highmem pages.
1189 * The returned number is system-wide.
1191 static unsigned int count_free_highmem_pages(void)
1194 unsigned int cnt = 0;
1196 for_each_populated_zone(zone)
1197 if (is_highmem(zone))
1198 cnt += zone_page_state(zone, NR_FREE_PAGES);
1204 * saveable_highmem_page - Check if a highmem page is saveable.
1206 * Determine whether a highmem page should be included in a hibernation image.
1208 * We should save the page if it isn't Nosave or NosaveFree, or Reserved,
1209 * and it isn't part of a free chunk of pages.
1211 static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn)
1215 if (!pfn_valid(pfn))
1218 page = pfn_to_online_page(pfn);
1219 if (!page || page_zone(page) != zone)
1222 BUG_ON(!PageHighMem(page));
1224 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
1227 if (PageReserved(page) || PageOffline(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_online_page(pfn);
1283 if (!page || page_zone(page) != zone)
1286 BUG_ON(PageHighMem(page));
1288 if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
1291 if (PageOffline(page))
1294 if (PageReserved(page)
1295 && (!kernel_page_present(page) || pfn_is_nosave(pfn)))
1298 if (page_is_guard(page))
1305 * count_data_pages - Compute the total number of saveable non-highmem pages.
1307 static unsigned int count_data_pages(void)
1310 unsigned long pfn, max_zone_pfn;
1313 for_each_populated_zone(zone) {
1314 if (is_highmem(zone))
1317 mark_free_pages(zone);
1318 max_zone_pfn = zone_end_pfn(zone);
1319 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1320 if (saveable_page(zone, pfn))
1327 * This is needed, because copy_page and memcpy are not usable for copying
1330 static inline void do_copy_page(long *dst, long *src)
1334 for (n = PAGE_SIZE / sizeof(long); n; n--)
1339 * safe_copy_page - Copy a page in a safe way.
1341 * Check if the page we are going to copy is marked as present in the kernel
1342 * page tables. This always is the case if CONFIG_DEBUG_PAGEALLOC or
1343 * CONFIG_ARCH_HAS_SET_DIRECT_MAP is not set. In that case kernel_page_present()
1344 * always returns 'true'.
1346 static void safe_copy_page(void *dst, struct page *s_page)
1348 if (kernel_page_present(s_page)) {
1349 do_copy_page(dst, page_address(s_page));
1351 kernel_map_pages(s_page, 1, 1);
1352 do_copy_page(dst, page_address(s_page));
1353 kernel_map_pages(s_page, 1, 0);
1357 #ifdef CONFIG_HIGHMEM
1358 static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn)
1360 return is_highmem(zone) ?
1361 saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn);
1364 static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1366 struct page *s_page, *d_page;
1369 s_page = pfn_to_page(src_pfn);
1370 d_page = pfn_to_page(dst_pfn);
1371 if (PageHighMem(s_page)) {
1372 src = kmap_atomic(s_page);
1373 dst = kmap_atomic(d_page);
1374 do_copy_page(dst, src);
1378 if (PageHighMem(d_page)) {
1380 * The page pointed to by src may contain some kernel
1381 * data modified by kmap_atomic()
1383 safe_copy_page(buffer, s_page);
1384 dst = kmap_atomic(d_page);
1385 copy_page(dst, buffer);
1388 safe_copy_page(page_address(d_page), s_page);
1393 #define page_is_saveable(zone, pfn) saveable_page(zone, pfn)
1395 static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1397 safe_copy_page(page_address(pfn_to_page(dst_pfn)),
1398 pfn_to_page(src_pfn));
1400 #endif /* CONFIG_HIGHMEM */
1402 static void copy_data_pages(struct memory_bitmap *copy_bm,
1403 struct memory_bitmap *orig_bm)
1408 for_each_populated_zone(zone) {
1409 unsigned long max_zone_pfn;
1411 mark_free_pages(zone);
1412 max_zone_pfn = zone_end_pfn(zone);
1413 for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1414 if (page_is_saveable(zone, pfn))
1415 memory_bm_set_bit(orig_bm, pfn);
1417 memory_bm_position_reset(orig_bm);
1418 memory_bm_position_reset(copy_bm);
1420 pfn = memory_bm_next_pfn(orig_bm);
1421 if (unlikely(pfn == BM_END_OF_MAP))
1423 copy_data_page(memory_bm_next_pfn(copy_bm), pfn);
1427 /* Total number of image pages */
1428 static unsigned int nr_copy_pages;
1429 /* Number of pages needed for saving the original pfns of the image pages */
1430 static unsigned int nr_meta_pages;
1432 * Numbers of normal and highmem page frames allocated for hibernation image
1433 * before suspending devices.
1435 static unsigned int alloc_normal, alloc_highmem;
1437 * Memory bitmap used for marking saveable pages (during hibernation) or
1438 * hibernation image pages (during restore)
1440 static struct memory_bitmap orig_bm;
1442 * Memory bitmap used during hibernation for marking allocated page frames that
1443 * will contain copies of saveable pages. During restore it is initially used
1444 * for marking hibernation image pages, but then the set bits from it are
1445 * duplicated in @orig_bm and it is released. On highmem systems it is next
1446 * used for marking "safe" highmem pages, but it has to be reinitialized for
1449 static struct memory_bitmap copy_bm;
1452 * swsusp_free - Free pages allocated for hibernation image.
1454 * Image pages are alocated before snapshot creation, so they need to be
1455 * released after resume.
1457 void swsusp_free(void)
1459 unsigned long fb_pfn, fr_pfn;
1461 if (!forbidden_pages_map || !free_pages_map)
1464 memory_bm_position_reset(forbidden_pages_map);
1465 memory_bm_position_reset(free_pages_map);
1468 fr_pfn = memory_bm_next_pfn(free_pages_map);
1469 fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1472 * Find the next bit set in both bitmaps. This is guaranteed to
1473 * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP.
1476 if (fb_pfn < fr_pfn)
1477 fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1478 if (fr_pfn < fb_pfn)
1479 fr_pfn = memory_bm_next_pfn(free_pages_map);
1480 } while (fb_pfn != fr_pfn);
1482 if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) {
1483 struct page *page = pfn_to_page(fr_pfn);
1485 memory_bm_clear_current(forbidden_pages_map);
1486 memory_bm_clear_current(free_pages_map);
1487 hibernate_restore_unprotect_page(page_address(page));
1495 restore_pblist = NULL;
1499 hibernate_restore_protection_end();
1502 /* Helper functions used for the shrinking of memory. */
1504 #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN)
1507 * preallocate_image_pages - Allocate a number of pages for hibernation image.
1508 * @nr_pages: Number of page frames to allocate.
1509 * @mask: GFP flags to use for the allocation.
1511 * Return value: Number of page frames actually allocated
1513 static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask)
1515 unsigned long nr_alloc = 0;
1517 while (nr_pages > 0) {
1520 page = alloc_image_page(mask);
1523 memory_bm_set_bit(©_bm, page_to_pfn(page));
1524 if (PageHighMem(page))
1535 static unsigned long preallocate_image_memory(unsigned long nr_pages,
1536 unsigned long avail_normal)
1538 unsigned long alloc;
1540 if (avail_normal <= alloc_normal)
1543 alloc = avail_normal - alloc_normal;
1544 if (nr_pages < alloc)
1547 return preallocate_image_pages(alloc, GFP_IMAGE);
1550 #ifdef CONFIG_HIGHMEM
1551 static unsigned long preallocate_image_highmem(unsigned long nr_pages)
1553 return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM);
1557 * __fraction - Compute (an approximation of) x * (multiplier / base).
1559 static unsigned long __fraction(u64 x, u64 multiplier, u64 base)
1561 return div64_u64(x * multiplier, base);
1564 static unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1565 unsigned long highmem,
1566 unsigned long total)
1568 unsigned long alloc = __fraction(nr_pages, highmem, total);
1570 return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM);
1572 #else /* CONFIG_HIGHMEM */
1573 static inline unsigned long preallocate_image_highmem(unsigned long nr_pages)
1578 static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1579 unsigned long highmem,
1580 unsigned long total)
1584 #endif /* CONFIG_HIGHMEM */
1587 * free_unnecessary_pages - Release preallocated pages not needed for the image.
1589 static unsigned long free_unnecessary_pages(void)
1591 unsigned long save, to_free_normal, to_free_highmem, free;
1593 save = count_data_pages();
1594 if (alloc_normal >= save) {
1595 to_free_normal = alloc_normal - save;
1599 save -= alloc_normal;
1601 save += count_highmem_pages();
1602 if (alloc_highmem >= save) {
1603 to_free_highmem = alloc_highmem - save;
1605 to_free_highmem = 0;
1606 save -= alloc_highmem;
1607 if (to_free_normal > save)
1608 to_free_normal -= save;
1612 free = to_free_normal + to_free_highmem;
1614 memory_bm_position_reset(©_bm);
1616 while (to_free_normal > 0 || to_free_highmem > 0) {
1617 unsigned long pfn = memory_bm_next_pfn(©_bm);
1618 struct page *page = pfn_to_page(pfn);
1620 if (PageHighMem(page)) {
1621 if (!to_free_highmem)
1626 if (!to_free_normal)
1631 memory_bm_clear_bit(©_bm, pfn);
1632 swsusp_unset_page_forbidden(page);
1633 swsusp_unset_page_free(page);
1641 * minimum_image_size - Estimate the minimum acceptable size of an image.
1642 * @saveable: Number of saveable pages in the system.
1644 * We want to avoid attempting to free too much memory too hard, so estimate the
1645 * minimum acceptable size of a hibernation image to use as the lower limit for
1646 * preallocating memory.
1648 * We assume that the minimum image size should be proportional to
1650 * [number of saveable pages] - [number of pages that can be freed in theory]
1652 * where the second term is the sum of (1) reclaimable slab pages, (2) active
1653 * and (3) inactive anonymous pages, (4) active and (5) inactive file pages.
1655 static unsigned long minimum_image_size(unsigned long saveable)
1659 size = global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B)
1660 + global_node_page_state(NR_ACTIVE_ANON)
1661 + global_node_page_state(NR_INACTIVE_ANON)
1662 + global_node_page_state(NR_ACTIVE_FILE)
1663 + global_node_page_state(NR_INACTIVE_FILE);
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 reserved_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 pr_info("Preallocating image memory\n");
1699 start = ktime_get();
1701 error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY);
1703 pr_err("Cannot allocate original bitmap\n");
1707 error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY);
1709 pr_err("Cannot allocate copy bitmap\n");
1716 /* Count the number of saveable data pages. */
1717 save_highmem = count_highmem_pages();
1718 saveable = count_data_pages();
1721 * Compute the total number of page frames we can use (count) and the
1722 * number of pages needed for image metadata (size).
1725 saveable += save_highmem;
1726 highmem = save_highmem;
1728 for_each_populated_zone(zone) {
1729 size += snapshot_additional_pages(zone);
1730 if (is_highmem(zone))
1731 highmem += zone_page_state(zone, NR_FREE_PAGES);
1733 count += zone_page_state(zone, NR_FREE_PAGES);
1735 avail_normal = count;
1737 count -= totalreserve_pages;
1739 /* Compute the maximum number of saveable pages to leave in memory. */
1740 max_size = (count - (size + PAGES_FOR_IO)) / 2
1741 - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE);
1742 /* Compute the desired number of image pages specified by image_size. */
1743 size = DIV_ROUND_UP(image_size, PAGE_SIZE);
1744 if (size > max_size)
1747 * If the desired number of image pages is at least as large as the
1748 * current number of saveable pages in memory, allocate page frames for
1749 * the image and we're done.
1751 if (size >= saveable) {
1752 pages = preallocate_image_highmem(save_highmem);
1753 pages += preallocate_image_memory(saveable - pages, avail_normal);
1757 /* Estimate the minimum size of the image. */
1758 pages = minimum_image_size(saveable);
1760 * To avoid excessive pressure on the normal zone, leave room in it to
1761 * accommodate an image of the minimum size (unless it's already too
1762 * small, in which case don't preallocate pages from it at all).
1764 if (avail_normal > pages)
1765 avail_normal -= pages;
1769 size = min_t(unsigned long, pages, max_size);
1772 * Let the memory management subsystem know that we're going to need a
1773 * large number of page frames to allocate and make it free some memory.
1774 * NOTE: If this is not done, performance will be hurt badly in some
1777 shrink_all_memory(saveable - size);
1780 * The number of saveable pages in memory was too high, so apply some
1781 * pressure to decrease it. First, make room for the largest possible
1782 * image and fail if that doesn't work. Next, try to decrease the size
1783 * of the image as much as indicated by 'size' using allocations from
1784 * highmem and non-highmem zones separately.
1786 pages_highmem = preallocate_image_highmem(highmem / 2);
1787 alloc = count - max_size;
1788 if (alloc > pages_highmem)
1789 alloc -= pages_highmem;
1792 pages = preallocate_image_memory(alloc, avail_normal);
1793 if (pages < alloc) {
1794 /* We have exhausted non-highmem pages, try highmem. */
1796 pages += pages_highmem;
1797 pages_highmem = preallocate_image_highmem(alloc);
1798 if (pages_highmem < alloc) {
1799 pr_err("Image allocation is %lu pages short\n",
1800 alloc - pages_highmem);
1803 pages += pages_highmem;
1805 * size is the desired number of saveable pages to leave in
1806 * memory, so try to preallocate (all memory - size) pages.
1808 alloc = (count - pages) - size;
1809 pages += preallocate_image_highmem(alloc);
1812 * There are approximately max_size saveable pages at this point
1813 * and we want to reduce this number down to size.
1815 alloc = max_size - size;
1816 size = preallocate_highmem_fraction(alloc, highmem, count);
1817 pages_highmem += size;
1819 size = preallocate_image_memory(alloc, avail_normal);
1820 pages_highmem += preallocate_image_highmem(alloc - size);
1821 pages += pages_highmem + size;
1825 * We only need as many page frames for the image as there are saveable
1826 * pages in memory, but we have allocated more. Release the excessive
1829 pages -= free_unnecessary_pages();
1833 pr_info("Allocated %lu pages for snapshot\n", pages);
1834 swsusp_show_speed(start, stop, pages, "Allocated");
1843 #ifdef CONFIG_HIGHMEM
1845 * count_pages_for_highmem - Count non-highmem pages needed for copying highmem.
1847 * Compute the number of non-highmem pages that will be necessary for creating
1848 * copies of highmem pages.
1850 static unsigned int count_pages_for_highmem(unsigned int nr_highmem)
1852 unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem;
1854 if (free_highmem >= nr_highmem)
1857 nr_highmem -= free_highmem;
1862 static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; }
1863 #endif /* CONFIG_HIGHMEM */
1866 * enough_free_mem - Check if there is enough free memory for the image.
1868 static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem)
1871 unsigned int free = alloc_normal;
1873 for_each_populated_zone(zone)
1874 if (!is_highmem(zone))
1875 free += zone_page_state(zone, NR_FREE_PAGES);
1877 nr_pages += count_pages_for_highmem(nr_highmem);
1878 pr_debug("Normal pages needed: %u + %u, available pages: %u\n",
1879 nr_pages, PAGES_FOR_IO, free);
1881 return free > nr_pages + PAGES_FOR_IO;
1884 #ifdef CONFIG_HIGHMEM
1886 * get_highmem_buffer - Allocate a buffer for highmem pages.
1888 * If there are some highmem pages in the hibernation image, we may need a
1889 * buffer to copy them and/or load their data.
1891 static inline int get_highmem_buffer(int safe_needed)
1893 buffer = get_image_page(GFP_ATOMIC, safe_needed);
1894 return buffer ? 0 : -ENOMEM;
1898 * alloc_highmem_image_pages - Allocate some highmem pages for the image.
1900 * Try to allocate as many pages as needed, but if the number of free highmem
1901 * pages is less than that, allocate them all.
1903 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1904 unsigned int nr_highmem)
1906 unsigned int to_alloc = count_free_highmem_pages();
1908 if (to_alloc > nr_highmem)
1909 to_alloc = nr_highmem;
1911 nr_highmem -= to_alloc;
1912 while (to_alloc-- > 0) {
1915 page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM);
1916 memory_bm_set_bit(bm, page_to_pfn(page));
1921 static inline int get_highmem_buffer(int safe_needed) { return 0; }
1923 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1924 unsigned int n) { return 0; }
1925 #endif /* CONFIG_HIGHMEM */
1928 * swsusp_alloc - Allocate memory for hibernation image.
1930 * We first try to allocate as many highmem pages as there are
1931 * saveable highmem pages in the system. If that fails, we allocate
1932 * non-highmem pages for the copies of the remaining highmem ones.
1934 * In this approach it is likely that the copies of highmem pages will
1935 * also be located in the high memory, because of the way in which
1936 * copy_data_pages() works.
1938 static int swsusp_alloc(struct memory_bitmap *copy_bm,
1939 unsigned int nr_pages, unsigned int nr_highmem)
1941 if (nr_highmem > 0) {
1942 if (get_highmem_buffer(PG_ANY))
1944 if (nr_highmem > alloc_highmem) {
1945 nr_highmem -= alloc_highmem;
1946 nr_pages += alloc_highmem_pages(copy_bm, nr_highmem);
1949 if (nr_pages > alloc_normal) {
1950 nr_pages -= alloc_normal;
1951 while (nr_pages-- > 0) {
1954 page = alloc_image_page(GFP_ATOMIC);
1957 memory_bm_set_bit(copy_bm, page_to_pfn(page));
1968 asmlinkage __visible int swsusp_save(void)
1970 unsigned int nr_pages, nr_highmem;
1972 pr_info("Creating image:\n");
1974 drain_local_pages(NULL);
1975 nr_pages = count_data_pages();
1976 nr_highmem = count_highmem_pages();
1977 pr_info("Need to copy %u pages\n", nr_pages + nr_highmem);
1979 if (!enough_free_mem(nr_pages, nr_highmem)) {
1980 pr_err("Not enough free memory\n");
1984 if (swsusp_alloc(©_bm, nr_pages, nr_highmem)) {
1985 pr_err("Memory allocation failed\n");
1990 * During allocating of suspend pagedir, new cold pages may appear.
1993 drain_local_pages(NULL);
1994 copy_data_pages(©_bm, &orig_bm);
1997 * End of critical section. From now on, we can write to memory,
1998 * but we should not touch disk. This specially means we must _not_
1999 * touch swap space! Except we must write out our image of course.
2002 nr_pages += nr_highmem;
2003 nr_copy_pages = nr_pages;
2004 nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE);
2006 pr_info("Image created (%d pages copied)\n", nr_pages);
2011 #ifndef CONFIG_ARCH_HIBERNATION_HEADER
2012 static int init_header_complete(struct swsusp_info *info)
2014 memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname));
2015 info->version_code = LINUX_VERSION_CODE;
2019 static const char *check_image_kernel(struct swsusp_info *info)
2021 if (info->version_code != LINUX_VERSION_CODE)
2022 return "kernel version";
2023 if (strcmp(info->uts.sysname,init_utsname()->sysname))
2024 return "system type";
2025 if (strcmp(info->uts.release,init_utsname()->release))
2026 return "kernel release";
2027 if (strcmp(info->uts.version,init_utsname()->version))
2029 if (strcmp(info->uts.machine,init_utsname()->machine))
2033 #endif /* CONFIG_ARCH_HIBERNATION_HEADER */
2035 unsigned long snapshot_get_image_size(void)
2037 return nr_copy_pages + nr_meta_pages + 1;
2040 static int init_header(struct swsusp_info *info)
2042 memset(info, 0, sizeof(struct swsusp_info));
2043 info->num_physpages = get_num_physpages();
2044 info->image_pages = nr_copy_pages;
2045 info->pages = snapshot_get_image_size();
2046 info->size = info->pages;
2047 info->size <<= PAGE_SHIFT;
2048 return init_header_complete(info);
2052 * pack_pfns - Prepare PFNs for saving.
2053 * @bm: Memory bitmap.
2054 * @buf: Memory buffer to store the PFNs in.
2056 * PFNs corresponding to set bits in @bm are stored in the area of memory
2057 * pointed to by @buf (1 page at a time).
2059 static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
2063 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2064 buf[j] = memory_bm_next_pfn(bm);
2065 if (unlikely(buf[j] == BM_END_OF_MAP))
2071 * snapshot_read_next - Get the address to read the next image page from.
2072 * @handle: Snapshot handle to be used for the reading.
2074 * On the first call, @handle should point to a zeroed snapshot_handle
2075 * structure. The structure gets populated then and a pointer to it should be
2076 * passed to this function every next time.
2078 * On success, the function returns a positive number. Then, the caller
2079 * is allowed to read up to the returned number of bytes from the memory
2080 * location computed by the data_of() macro.
2082 * The function returns 0 to indicate the end of the data stream condition,
2083 * and negative numbers are returned on errors. If that happens, the structure
2084 * pointed to by @handle is not updated and should not be used any more.
2086 int snapshot_read_next(struct snapshot_handle *handle)
2088 if (handle->cur > nr_meta_pages + nr_copy_pages)
2092 /* This makes the buffer be freed by swsusp_free() */
2093 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2100 error = init_header((struct swsusp_info *)buffer);
2103 handle->buffer = buffer;
2104 memory_bm_position_reset(&orig_bm);
2105 memory_bm_position_reset(©_bm);
2106 } else if (handle->cur <= nr_meta_pages) {
2108 pack_pfns(buffer, &orig_bm);
2112 page = pfn_to_page(memory_bm_next_pfn(©_bm));
2113 if (PageHighMem(page)) {
2115 * Highmem pages are copied to the buffer,
2116 * because we can't return with a kmapped
2117 * highmem page (we may not be called again).
2121 kaddr = kmap_atomic(page);
2122 copy_page(buffer, kaddr);
2123 kunmap_atomic(kaddr);
2124 handle->buffer = buffer;
2126 handle->buffer = page_address(page);
2133 static void duplicate_memory_bitmap(struct memory_bitmap *dst,
2134 struct memory_bitmap *src)
2138 memory_bm_position_reset(src);
2139 pfn = memory_bm_next_pfn(src);
2140 while (pfn != BM_END_OF_MAP) {
2141 memory_bm_set_bit(dst, pfn);
2142 pfn = memory_bm_next_pfn(src);
2147 * mark_unsafe_pages - Mark pages that were used before hibernation.
2149 * Mark the pages that cannot be used for storing the image during restoration,
2150 * because they conflict with the pages that had been used before hibernation.
2152 static void mark_unsafe_pages(struct memory_bitmap *bm)
2156 /* Clear the "free"/"unsafe" bit for all PFNs */
2157 memory_bm_position_reset(free_pages_map);
2158 pfn = memory_bm_next_pfn(free_pages_map);
2159 while (pfn != BM_END_OF_MAP) {
2160 memory_bm_clear_current(free_pages_map);
2161 pfn = memory_bm_next_pfn(free_pages_map);
2164 /* Mark pages that correspond to the "original" PFNs as "unsafe" */
2165 duplicate_memory_bitmap(free_pages_map, bm);
2167 allocated_unsafe_pages = 0;
2170 static int check_header(struct swsusp_info *info)
2174 reason = check_image_kernel(info);
2175 if (!reason && info->num_physpages != get_num_physpages())
2176 reason = "memory size";
2178 pr_err("Image mismatch: %s\n", reason);
2185 * load header - Check the image header and copy the data from it.
2187 static int load_header(struct swsusp_info *info)
2191 restore_pblist = NULL;
2192 error = check_header(info);
2194 nr_copy_pages = info->image_pages;
2195 nr_meta_pages = info->pages - info->image_pages - 1;
2201 * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap.
2202 * @bm: Memory bitmap.
2203 * @buf: Area of memory containing the PFNs.
2205 * For each element of the array pointed to by @buf (1 page at a time), set the
2206 * corresponding bit in @bm.
2208 static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
2212 for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2213 if (unlikely(buf[j] == BM_END_OF_MAP))
2216 if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j]))
2217 memory_bm_set_bit(bm, buf[j]);
2225 #ifdef CONFIG_HIGHMEM
2227 * struct highmem_pbe is used for creating the list of highmem pages that
2228 * should be restored atomically during the resume from disk, because the page
2229 * frames they have occupied before the suspend are in use.
2231 struct highmem_pbe {
2232 struct page *copy_page; /* data is here now */
2233 struct page *orig_page; /* data was here before the suspend */
2234 struct highmem_pbe *next;
2238 * List of highmem PBEs needed for restoring the highmem pages that were
2239 * allocated before the suspend and included in the suspend image, but have
2240 * also been allocated by the "resume" kernel, so their contents cannot be
2241 * written directly to their "original" page frames.
2243 static struct highmem_pbe *highmem_pblist;
2246 * count_highmem_image_pages - Compute the number of highmem pages in the image.
2247 * @bm: Memory bitmap.
2249 * The bits in @bm that correspond to image pages are assumed to be set.
2251 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm)
2254 unsigned int cnt = 0;
2256 memory_bm_position_reset(bm);
2257 pfn = memory_bm_next_pfn(bm);
2258 while (pfn != BM_END_OF_MAP) {
2259 if (PageHighMem(pfn_to_page(pfn)))
2262 pfn = memory_bm_next_pfn(bm);
2267 static unsigned int safe_highmem_pages;
2269 static struct memory_bitmap *safe_highmem_bm;
2272 * prepare_highmem_image - Allocate memory for loading highmem data from image.
2273 * @bm: Pointer to an uninitialized memory bitmap structure.
2274 * @nr_highmem_p: Pointer to the number of highmem image pages.
2276 * Try to allocate as many highmem pages as there are highmem image pages
2277 * (@nr_highmem_p points to the variable containing the number of highmem image
2278 * pages). The pages that are "safe" (ie. will not be overwritten when the
2279 * hibernation image is restored entirely) have the corresponding bits set in
2280 * @bm (it must be unitialized).
2282 * NOTE: This function should not be called if there are no highmem image pages.
2284 static int prepare_highmem_image(struct memory_bitmap *bm,
2285 unsigned int *nr_highmem_p)
2287 unsigned int to_alloc;
2289 if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE))
2292 if (get_highmem_buffer(PG_SAFE))
2295 to_alloc = count_free_highmem_pages();
2296 if (to_alloc > *nr_highmem_p)
2297 to_alloc = *nr_highmem_p;
2299 *nr_highmem_p = to_alloc;
2301 safe_highmem_pages = 0;
2302 while (to_alloc-- > 0) {
2305 page = alloc_page(__GFP_HIGHMEM);
2306 if (!swsusp_page_is_free(page)) {
2307 /* The page is "safe", set its bit the bitmap */
2308 memory_bm_set_bit(bm, page_to_pfn(page));
2309 safe_highmem_pages++;
2311 /* Mark the page as allocated */
2312 swsusp_set_page_forbidden(page);
2313 swsusp_set_page_free(page);
2315 memory_bm_position_reset(bm);
2316 safe_highmem_bm = bm;
2320 static struct page *last_highmem_page;
2323 * get_highmem_page_buffer - Prepare a buffer to store a highmem image page.
2325 * For a given highmem image page get a buffer that suspend_write_next() should
2326 * return to its caller to write to.
2328 * If the page is to be saved to its "original" page frame or a copy of
2329 * the page is to be made in the highmem, @buffer is returned. Otherwise,
2330 * the copy of the page is to be made in normal memory, so the address of
2331 * the copy is returned.
2333 * If @buffer is returned, the caller of suspend_write_next() will write
2334 * the page's contents to @buffer, so they will have to be copied to the
2335 * right location on the next call to suspend_write_next() and it is done
2336 * with the help of copy_last_highmem_page(). For this purpose, if
2337 * @buffer is returned, @last_highmem_page is set to the page to which
2338 * the data will have to be copied from @buffer.
2340 static void *get_highmem_page_buffer(struct page *page,
2341 struct chain_allocator *ca)
2343 struct highmem_pbe *pbe;
2346 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) {
2348 * We have allocated the "original" page frame and we can
2349 * use it directly to store the loaded page.
2351 last_highmem_page = page;
2355 * The "original" page frame has not been allocated and we have to
2356 * use a "safe" page frame to store the loaded page.
2358 pbe = chain_alloc(ca, sizeof(struct highmem_pbe));
2361 return ERR_PTR(-ENOMEM);
2363 pbe->orig_page = page;
2364 if (safe_highmem_pages > 0) {
2367 /* Copy of the page will be stored in high memory */
2369 tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm));
2370 safe_highmem_pages--;
2371 last_highmem_page = tmp;
2372 pbe->copy_page = tmp;
2374 /* Copy of the page will be stored in normal memory */
2375 kaddr = __get_safe_page(ca->gfp_mask);
2377 return ERR_PTR(-ENOMEM);
2378 pbe->copy_page = virt_to_page(kaddr);
2380 pbe->next = highmem_pblist;
2381 highmem_pblist = pbe;
2386 * copy_last_highmem_page - Copy most the most recent highmem image page.
2388 * Copy the contents of a highmem image from @buffer, where the caller of
2389 * snapshot_write_next() has stored them, to the right location represented by
2390 * @last_highmem_page .
2392 static void copy_last_highmem_page(void)
2394 if (last_highmem_page) {
2397 dst = kmap_atomic(last_highmem_page);
2398 copy_page(dst, buffer);
2400 last_highmem_page = NULL;
2404 static inline int last_highmem_page_copied(void)
2406 return !last_highmem_page;
2409 static inline void free_highmem_data(void)
2411 if (safe_highmem_bm)
2412 memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR);
2415 free_image_page(buffer, PG_UNSAFE_CLEAR);
2418 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; }
2420 static inline int prepare_highmem_image(struct memory_bitmap *bm,
2421 unsigned int *nr_highmem_p) { return 0; }
2423 static inline void *get_highmem_page_buffer(struct page *page,
2424 struct chain_allocator *ca)
2426 return ERR_PTR(-EINVAL);
2429 static inline void copy_last_highmem_page(void) {}
2430 static inline int last_highmem_page_copied(void) { return 1; }
2431 static inline void free_highmem_data(void) {}
2432 #endif /* CONFIG_HIGHMEM */
2434 #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe))
2437 * prepare_image - Make room for loading hibernation image.
2438 * @new_bm: Unitialized memory bitmap structure.
2439 * @bm: Memory bitmap with unsafe pages marked.
2441 * Use @bm to mark the pages that will be overwritten in the process of
2442 * restoring the system memory state from the suspend image ("unsafe" pages)
2443 * and allocate memory for the image.
2445 * The idea is to allocate a new memory bitmap first and then allocate
2446 * as many pages as needed for image data, but without specifying what those
2447 * pages will be used for just yet. Instead, we mark them all as allocated and
2448 * create a lists of "safe" pages to be used later. On systems with high
2449 * memory a list of "safe" highmem pages is created too.
2451 static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm)
2453 unsigned int nr_pages, nr_highmem;
2454 struct linked_page *lp;
2457 /* If there is no highmem, the buffer will not be necessary */
2458 free_image_page(buffer, PG_UNSAFE_CLEAR);
2461 nr_highmem = count_highmem_image_pages(bm);
2462 mark_unsafe_pages(bm);
2464 error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE);
2468 duplicate_memory_bitmap(new_bm, bm);
2469 memory_bm_free(bm, PG_UNSAFE_KEEP);
2470 if (nr_highmem > 0) {
2471 error = prepare_highmem_image(bm, &nr_highmem);
2476 * Reserve some safe pages for potential later use.
2478 * NOTE: This way we make sure there will be enough safe pages for the
2479 * chain_alloc() in get_buffer(). It is a bit wasteful, but
2480 * nr_copy_pages cannot be greater than 50% of the memory anyway.
2482 * nr_copy_pages cannot be less than allocated_unsafe_pages too.
2484 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2485 nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE);
2486 while (nr_pages > 0) {
2487 lp = get_image_page(GFP_ATOMIC, PG_SAFE);
2492 lp->next = safe_pages_list;
2493 safe_pages_list = lp;
2496 /* Preallocate memory for the image */
2497 nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2498 while (nr_pages > 0) {
2499 lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC);
2504 if (!swsusp_page_is_free(virt_to_page(lp))) {
2505 /* The page is "safe", add it to the list */
2506 lp->next = safe_pages_list;
2507 safe_pages_list = lp;
2509 /* Mark the page as allocated */
2510 swsusp_set_page_forbidden(virt_to_page(lp));
2511 swsusp_set_page_free(virt_to_page(lp));
2522 * get_buffer - Get the address to store the next image data page.
2524 * Get the address that snapshot_write_next() should return to its caller to
2527 static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca)
2531 unsigned long pfn = memory_bm_next_pfn(bm);
2533 if (pfn == BM_END_OF_MAP)
2534 return ERR_PTR(-EFAULT);
2536 page = pfn_to_page(pfn);
2537 if (PageHighMem(page))
2538 return get_highmem_page_buffer(page, ca);
2540 if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page))
2542 * We have allocated the "original" page frame and we can
2543 * use it directly to store the loaded page.
2545 return page_address(page);
2548 * The "original" page frame has not been allocated and we have to
2549 * use a "safe" page frame to store the loaded page.
2551 pbe = chain_alloc(ca, sizeof(struct pbe));
2554 return ERR_PTR(-ENOMEM);
2556 pbe->orig_address = page_address(page);
2557 pbe->address = __get_safe_page(ca->gfp_mask);
2559 return ERR_PTR(-ENOMEM);
2560 pbe->next = restore_pblist;
2561 restore_pblist = pbe;
2562 return pbe->address;
2566 * snapshot_write_next - Get the address to store the next image page.
2567 * @handle: Snapshot handle structure to guide the writing.
2569 * On the first call, @handle should point to a zeroed snapshot_handle
2570 * structure. The structure gets populated then and a pointer to it should be
2571 * passed to this function every next time.
2573 * On success, the function returns a positive number. Then, the caller
2574 * is allowed to write up to the returned number of bytes to the memory
2575 * location computed by the data_of() macro.
2577 * The function returns 0 to indicate the "end of file" condition. Negative
2578 * numbers are returned on errors, in which cases the structure pointed to by
2579 * @handle is not updated and should not be used any more.
2581 int snapshot_write_next(struct snapshot_handle *handle)
2583 static struct chain_allocator ca;
2586 /* Check if we have already loaded the entire image */
2587 if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages)
2592 /* This makes the buffer be freed by swsusp_free() */
2593 buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2598 handle->buffer = buffer;
2599 } else if (handle->cur == 1) {
2600 error = load_header(buffer);
2604 safe_pages_list = NULL;
2606 error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY);
2610 hibernate_restore_protection_begin();
2611 } else if (handle->cur <= nr_meta_pages + 1) {
2612 error = unpack_orig_pfns(buffer, ©_bm);
2616 if (handle->cur == nr_meta_pages + 1) {
2617 error = prepare_image(&orig_bm, ©_bm);
2621 chain_init(&ca, GFP_ATOMIC, PG_SAFE);
2622 memory_bm_position_reset(&orig_bm);
2623 restore_pblist = NULL;
2624 handle->buffer = get_buffer(&orig_bm, &ca);
2625 if (IS_ERR(handle->buffer))
2626 return PTR_ERR(handle->buffer);
2629 copy_last_highmem_page();
2630 hibernate_restore_protect_page(handle->buffer);
2631 handle->buffer = get_buffer(&orig_bm, &ca);
2632 if (IS_ERR(handle->buffer))
2633 return PTR_ERR(handle->buffer);
2635 handle->sync_read = (handle->buffer == buffer);
2641 * snapshot_write_finalize - Complete the loading of a hibernation image.
2643 * Must be called after the last call to snapshot_write_next() in case the last
2644 * page in the image happens to be a highmem page and its contents should be
2645 * stored in highmem. Additionally, it recycles bitmap memory that's not
2646 * necessary any more.
2648 void snapshot_write_finalize(struct snapshot_handle *handle)
2650 copy_last_highmem_page();
2651 hibernate_restore_protect_page(handle->buffer);
2652 /* Do that only if we have loaded the image entirely */
2653 if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) {
2654 memory_bm_recycle(&orig_bm);
2655 free_highmem_data();
2659 int snapshot_image_loaded(struct snapshot_handle *handle)
2661 return !(!nr_copy_pages || !last_highmem_page_copied() ||
2662 handle->cur <= nr_meta_pages + nr_copy_pages);
2665 #ifdef CONFIG_HIGHMEM
2666 /* Assumes that @buf is ready and points to a "safe" page */
2667 static inline void swap_two_pages_data(struct page *p1, struct page *p2,
2670 void *kaddr1, *kaddr2;
2672 kaddr1 = kmap_atomic(p1);
2673 kaddr2 = kmap_atomic(p2);
2674 copy_page(buf, kaddr1);
2675 copy_page(kaddr1, kaddr2);
2676 copy_page(kaddr2, buf);
2677 kunmap_atomic(kaddr2);
2678 kunmap_atomic(kaddr1);
2682 * restore_highmem - Put highmem image pages into their original locations.
2684 * For each highmem page that was in use before hibernation and is included in
2685 * the image, and also has been allocated by the "restore" kernel, swap its
2686 * current contents with the previous (ie. "before hibernation") ones.
2688 * If the restore eventually fails, we can call this function once again and
2689 * restore the highmem state as seen by the restore kernel.
2691 int restore_highmem(void)
2693 struct highmem_pbe *pbe = highmem_pblist;
2699 buf = get_image_page(GFP_ATOMIC, PG_SAFE);
2704 swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf);
2707 free_image_page(buf, PG_UNSAFE_CLEAR);
2710 #endif /* CONFIG_HIGHMEM */