xref: /openbmc/linux/kernel/power/snapshot.c (revision f17f06a0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * linux/kernel/power/snapshot.c
4  *
5  * This file provides system snapshot/restore functionality for swsusp.
6  *
7  * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz>
8  * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
9  */
10 
11 #define pr_fmt(fmt) "PM: hibernation: " fmt
12 
13 #include <linux/version.h>
14 #include <linux/module.h>
15 #include <linux/mm.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>
21 #include <linux/pm.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>
34 
35 #include <linux/uaccess.h>
36 #include <asm/mmu_context.h>
37 #include <asm/pgtable.h>
38 #include <asm/tlbflush.h>
39 #include <asm/io.h>
40 
41 #include "power.h"
42 
43 #if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_ARCH_HAS_SET_MEMORY)
44 static bool hibernate_restore_protection;
45 static bool hibernate_restore_protection_active;
46 
47 void enable_restore_image_protection(void)
48 {
49 	hibernate_restore_protection = true;
50 }
51 
52 static inline void hibernate_restore_protection_begin(void)
53 {
54 	hibernate_restore_protection_active = hibernate_restore_protection;
55 }
56 
57 static inline void hibernate_restore_protection_end(void)
58 {
59 	hibernate_restore_protection_active = false;
60 }
61 
62 static inline void hibernate_restore_protect_page(void *page_address)
63 {
64 	if (hibernate_restore_protection_active)
65 		set_memory_ro((unsigned long)page_address, 1);
66 }
67 
68 static inline void hibernate_restore_unprotect_page(void *page_address)
69 {
70 	if (hibernate_restore_protection_active)
71 		set_memory_rw((unsigned long)page_address, 1);
72 }
73 #else
74 static inline void hibernate_restore_protection_begin(void) {}
75 static inline void hibernate_restore_protection_end(void) {}
76 static inline void hibernate_restore_protect_page(void *page_address) {}
77 static inline void hibernate_restore_unprotect_page(void *page_address) {}
78 #endif /* CONFIG_STRICT_KERNEL_RWX  && CONFIG_ARCH_HAS_SET_MEMORY */
79 
80 static int swsusp_page_is_free(struct page *);
81 static void swsusp_set_page_forbidden(struct page *);
82 static void swsusp_unset_page_forbidden(struct page *);
83 
84 /*
85  * Number of bytes to reserve for memory allocations made by device drivers
86  * from their ->freeze() and ->freeze_noirq() callbacks so that they don't
87  * cause image creation to fail (tunable via /sys/power/reserved_size).
88  */
89 unsigned long reserved_size;
90 
91 void __init hibernate_reserved_size_init(void)
92 {
93 	reserved_size = SPARE_PAGES * PAGE_SIZE;
94 }
95 
96 /*
97  * Preferred image size in bytes (tunable via /sys/power/image_size).
98  * When it is set to N, swsusp will do its best to ensure the image
99  * size will not exceed N bytes, but if that is impossible, it will
100  * try to create the smallest image possible.
101  */
102 unsigned long image_size;
103 
104 void __init hibernate_image_size_init(void)
105 {
106 	image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE;
107 }
108 
109 /*
110  * List of PBEs needed for restoring the pages that were allocated before
111  * the suspend and included in the suspend image, but have also been
112  * allocated by the "resume" kernel, so their contents cannot be written
113  * directly to their "original" page frames.
114  */
115 struct pbe *restore_pblist;
116 
117 /* struct linked_page is used to build chains of pages */
118 
119 #define LINKED_PAGE_DATA_SIZE	(PAGE_SIZE - sizeof(void *))
120 
121 struct linked_page {
122 	struct linked_page *next;
123 	char data[LINKED_PAGE_DATA_SIZE];
124 } __packed;
125 
126 /*
127  * List of "safe" pages (ie. pages that were not used by the image kernel
128  * before hibernation) that may be used as temporary storage for image kernel
129  * memory contents.
130  */
131 static struct linked_page *safe_pages_list;
132 
133 /* Pointer to an auxiliary buffer (1 page) */
134 static void *buffer;
135 
136 #define PG_ANY		0
137 #define PG_SAFE		1
138 #define PG_UNSAFE_CLEAR	1
139 #define PG_UNSAFE_KEEP	0
140 
141 static unsigned int allocated_unsafe_pages;
142 
143 /**
144  * get_image_page - Allocate a page for a hibernation image.
145  * @gfp_mask: GFP mask for the allocation.
146  * @safe_needed: Get pages that were not used before hibernation (restore only)
147  *
148  * During image restoration, for storing the PBE list and the image data, we can
149  * only use memory pages that do not conflict with the pages used before
150  * hibernation.  The "unsafe" pages have PageNosaveFree set and we count them
151  * using allocated_unsafe_pages.
152  *
153  * Each allocated image page is marked as PageNosave and PageNosaveFree so that
154  * swsusp_free() can release it.
155  */
156 static void *get_image_page(gfp_t gfp_mask, int safe_needed)
157 {
158 	void *res;
159 
160 	res = (void *)get_zeroed_page(gfp_mask);
161 	if (safe_needed)
162 		while (res && swsusp_page_is_free(virt_to_page(res))) {
163 			/* The page is unsafe, mark it for swsusp_free() */
164 			swsusp_set_page_forbidden(virt_to_page(res));
165 			allocated_unsafe_pages++;
166 			res = (void *)get_zeroed_page(gfp_mask);
167 		}
168 	if (res) {
169 		swsusp_set_page_forbidden(virt_to_page(res));
170 		swsusp_set_page_free(virt_to_page(res));
171 	}
172 	return res;
173 }
174 
175 static void *__get_safe_page(gfp_t gfp_mask)
176 {
177 	if (safe_pages_list) {
178 		void *ret = safe_pages_list;
179 
180 		safe_pages_list = safe_pages_list->next;
181 		memset(ret, 0, PAGE_SIZE);
182 		return ret;
183 	}
184 	return get_image_page(gfp_mask, PG_SAFE);
185 }
186 
187 unsigned long get_safe_page(gfp_t gfp_mask)
188 {
189 	return (unsigned long)__get_safe_page(gfp_mask);
190 }
191 
192 static struct page *alloc_image_page(gfp_t gfp_mask)
193 {
194 	struct page *page;
195 
196 	page = alloc_page(gfp_mask);
197 	if (page) {
198 		swsusp_set_page_forbidden(page);
199 		swsusp_set_page_free(page);
200 	}
201 	return page;
202 }
203 
204 static void recycle_safe_page(void *page_address)
205 {
206 	struct linked_page *lp = page_address;
207 
208 	lp->next = safe_pages_list;
209 	safe_pages_list = lp;
210 }
211 
212 /**
213  * free_image_page - Free a page allocated for hibernation image.
214  * @addr: Address of the page to free.
215  * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page.
216  *
217  * The page to free should have been allocated by get_image_page() (page flags
218  * set by it are affected).
219  */
220 static inline void free_image_page(void *addr, int clear_nosave_free)
221 {
222 	struct page *page;
223 
224 	BUG_ON(!virt_addr_valid(addr));
225 
226 	page = virt_to_page(addr);
227 
228 	swsusp_unset_page_forbidden(page);
229 	if (clear_nosave_free)
230 		swsusp_unset_page_free(page);
231 
232 	__free_page(page);
233 }
234 
235 static inline void free_list_of_pages(struct linked_page *list,
236 				      int clear_page_nosave)
237 {
238 	while (list) {
239 		struct linked_page *lp = list->next;
240 
241 		free_image_page(list, clear_page_nosave);
242 		list = lp;
243 	}
244 }
245 
246 /*
247  * struct chain_allocator is used for allocating small objects out of
248  * a linked list of pages called 'the chain'.
249  *
250  * The chain grows each time when there is no room for a new object in
251  * the current page.  The allocated objects cannot be freed individually.
252  * It is only possible to free them all at once, by freeing the entire
253  * chain.
254  *
255  * NOTE: The chain allocator may be inefficient if the allocated objects
256  * are not much smaller than PAGE_SIZE.
257  */
258 struct chain_allocator {
259 	struct linked_page *chain;	/* the chain */
260 	unsigned int used_space;	/* total size of objects allocated out
261 					   of the current page */
262 	gfp_t gfp_mask;		/* mask for allocating pages */
263 	int safe_needed;	/* if set, only "safe" pages are allocated */
264 };
265 
266 static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask,
267 		       int safe_needed)
268 {
269 	ca->chain = NULL;
270 	ca->used_space = LINKED_PAGE_DATA_SIZE;
271 	ca->gfp_mask = gfp_mask;
272 	ca->safe_needed = safe_needed;
273 }
274 
275 static void *chain_alloc(struct chain_allocator *ca, unsigned int size)
276 {
277 	void *ret;
278 
279 	if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) {
280 		struct linked_page *lp;
281 
282 		lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) :
283 					get_image_page(ca->gfp_mask, PG_ANY);
284 		if (!lp)
285 			return NULL;
286 
287 		lp->next = ca->chain;
288 		ca->chain = lp;
289 		ca->used_space = 0;
290 	}
291 	ret = ca->chain->data + ca->used_space;
292 	ca->used_space += size;
293 	return ret;
294 }
295 
296 /**
297  * Data types related to memory bitmaps.
298  *
299  * Memory bitmap is a structure consiting of many linked lists of
300  * objects.  The main list's elements are of type struct zone_bitmap
301  * and each of them corresonds to one zone.  For each zone bitmap
302  * object there is a list of objects of type struct bm_block that
303  * represent each blocks of bitmap in which information is stored.
304  *
305  * struct memory_bitmap contains a pointer to the main list of zone
306  * bitmap objects, a struct bm_position used for browsing the bitmap,
307  * and a pointer to the list of pages used for allocating all of the
308  * zone bitmap objects and bitmap block objects.
309  *
310  * NOTE: It has to be possible to lay out the bitmap in memory
311  * using only allocations of order 0.  Additionally, the bitmap is
312  * designed to work with arbitrary number of zones (this is over the
313  * top for now, but let's avoid making unnecessary assumptions ;-).
314  *
315  * struct zone_bitmap contains a pointer to a list of bitmap block
316  * objects and a pointer to the bitmap block object that has been
317  * most recently used for setting bits.  Additionally, it contains the
318  * PFNs that correspond to the start and end of the represented zone.
319  *
320  * struct bm_block contains a pointer to the memory page in which
321  * information is stored (in the form of a block of bitmap)
322  * It also contains the pfns that correspond to the start and end of
323  * the represented memory area.
324  *
325  * The memory bitmap is organized as a radix tree to guarantee fast random
326  * access to the bits. There is one radix tree for each zone (as returned
327  * from create_mem_extents).
328  *
329  * One radix tree is represented by one struct mem_zone_bm_rtree. There are
330  * two linked lists for the nodes of the tree, one for the inner nodes and
331  * one for the leave nodes. The linked leave nodes are used for fast linear
332  * access of the memory bitmap.
333  *
334  * The struct rtree_node represents one node of the radix tree.
335  */
336 
337 #define BM_END_OF_MAP	(~0UL)
338 
339 #define BM_BITS_PER_BLOCK	(PAGE_SIZE * BITS_PER_BYTE)
340 #define BM_BLOCK_SHIFT		(PAGE_SHIFT + 3)
341 #define BM_BLOCK_MASK		((1UL << BM_BLOCK_SHIFT) - 1)
342 
343 /*
344  * struct rtree_node is a wrapper struct to link the nodes
345  * of the rtree together for easy linear iteration over
346  * bits and easy freeing
347  */
348 struct rtree_node {
349 	struct list_head list;
350 	unsigned long *data;
351 };
352 
353 /*
354  * struct mem_zone_bm_rtree represents a bitmap used for one
355  * populated memory zone.
356  */
357 struct mem_zone_bm_rtree {
358 	struct list_head list;		/* Link Zones together         */
359 	struct list_head nodes;		/* Radix Tree inner nodes      */
360 	struct list_head leaves;	/* Radix Tree leaves           */
361 	unsigned long start_pfn;	/* Zone start page frame       */
362 	unsigned long end_pfn;		/* Zone end page frame + 1     */
363 	struct rtree_node *rtree;	/* Radix Tree Root             */
364 	int levels;			/* Number of Radix Tree Levels */
365 	unsigned int blocks;		/* Number of Bitmap Blocks     */
366 };
367 
368 /* strcut bm_position is used for browsing memory bitmaps */
369 
370 struct bm_position {
371 	struct mem_zone_bm_rtree *zone;
372 	struct rtree_node *node;
373 	unsigned long node_pfn;
374 	int node_bit;
375 };
376 
377 struct memory_bitmap {
378 	struct list_head zones;
379 	struct linked_page *p_list;	/* list of pages used to store zone
380 					   bitmap objects and bitmap block
381 					   objects */
382 	struct bm_position cur;	/* most recently used bit position */
383 };
384 
385 /* Functions that operate on memory bitmaps */
386 
387 #define BM_ENTRIES_PER_LEVEL	(PAGE_SIZE / sizeof(unsigned long))
388 #if BITS_PER_LONG == 32
389 #define BM_RTREE_LEVEL_SHIFT	(PAGE_SHIFT - 2)
390 #else
391 #define BM_RTREE_LEVEL_SHIFT	(PAGE_SHIFT - 3)
392 #endif
393 #define BM_RTREE_LEVEL_MASK	((1UL << BM_RTREE_LEVEL_SHIFT) - 1)
394 
395 /**
396  * alloc_rtree_node - Allocate a new node and add it to the radix tree.
397  *
398  * This function is used to allocate inner nodes as well as the
399  * leave nodes of the radix tree. It also adds the node to the
400  * corresponding linked list passed in by the *list parameter.
401  */
402 static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed,
403 					   struct chain_allocator *ca,
404 					   struct list_head *list)
405 {
406 	struct rtree_node *node;
407 
408 	node = chain_alloc(ca, sizeof(struct rtree_node));
409 	if (!node)
410 		return NULL;
411 
412 	node->data = get_image_page(gfp_mask, safe_needed);
413 	if (!node->data)
414 		return NULL;
415 
416 	list_add_tail(&node->list, list);
417 
418 	return node;
419 }
420 
421 /**
422  * add_rtree_block - Add a new leave node to the radix tree.
423  *
424  * The leave nodes need to be allocated in order to keep the leaves
425  * linked list in order. This is guaranteed by the zone->blocks
426  * counter.
427  */
428 static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask,
429 			   int safe_needed, struct chain_allocator *ca)
430 {
431 	struct rtree_node *node, *block, **dst;
432 	unsigned int levels_needed, block_nr;
433 	int i;
434 
435 	block_nr = zone->blocks;
436 	levels_needed = 0;
437 
438 	/* How many levels do we need for this block nr? */
439 	while (block_nr) {
440 		levels_needed += 1;
441 		block_nr >>= BM_RTREE_LEVEL_SHIFT;
442 	}
443 
444 	/* Make sure the rtree has enough levels */
445 	for (i = zone->levels; i < levels_needed; i++) {
446 		node = alloc_rtree_node(gfp_mask, safe_needed, ca,
447 					&zone->nodes);
448 		if (!node)
449 			return -ENOMEM;
450 
451 		node->data[0] = (unsigned long)zone->rtree;
452 		zone->rtree = node;
453 		zone->levels += 1;
454 	}
455 
456 	/* Allocate new block */
457 	block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves);
458 	if (!block)
459 		return -ENOMEM;
460 
461 	/* Now walk the rtree to insert the block */
462 	node = zone->rtree;
463 	dst = &zone->rtree;
464 	block_nr = zone->blocks;
465 	for (i = zone->levels; i > 0; i--) {
466 		int index;
467 
468 		if (!node) {
469 			node = alloc_rtree_node(gfp_mask, safe_needed, ca,
470 						&zone->nodes);
471 			if (!node)
472 				return -ENOMEM;
473 			*dst = node;
474 		}
475 
476 		index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
477 		index &= BM_RTREE_LEVEL_MASK;
478 		dst = (struct rtree_node **)&((*dst)->data[index]);
479 		node = *dst;
480 	}
481 
482 	zone->blocks += 1;
483 	*dst = block;
484 
485 	return 0;
486 }
487 
488 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
489 			       int clear_nosave_free);
490 
491 /**
492  * create_zone_bm_rtree - Create a radix tree for one zone.
493  *
494  * Allocated the mem_zone_bm_rtree structure and initializes it.
495  * This function also allocated and builds the radix tree for the
496  * zone.
497  */
498 static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask,
499 						      int safe_needed,
500 						      struct chain_allocator *ca,
501 						      unsigned long start,
502 						      unsigned long end)
503 {
504 	struct mem_zone_bm_rtree *zone;
505 	unsigned int i, nr_blocks;
506 	unsigned long pages;
507 
508 	pages = end - start;
509 	zone  = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree));
510 	if (!zone)
511 		return NULL;
512 
513 	INIT_LIST_HEAD(&zone->nodes);
514 	INIT_LIST_HEAD(&zone->leaves);
515 	zone->start_pfn = start;
516 	zone->end_pfn = end;
517 	nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK);
518 
519 	for (i = 0; i < nr_blocks; i++) {
520 		if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) {
521 			free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR);
522 			return NULL;
523 		}
524 	}
525 
526 	return zone;
527 }
528 
529 /**
530  * free_zone_bm_rtree - Free the memory of the radix tree.
531  *
532  * Free all node pages of the radix tree. The mem_zone_bm_rtree
533  * structure itself is not freed here nor are the rtree_node
534  * structs.
535  */
536 static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone,
537 			       int clear_nosave_free)
538 {
539 	struct rtree_node *node;
540 
541 	list_for_each_entry(node, &zone->nodes, list)
542 		free_image_page(node->data, clear_nosave_free);
543 
544 	list_for_each_entry(node, &zone->leaves, list)
545 		free_image_page(node->data, clear_nosave_free);
546 }
547 
548 static void memory_bm_position_reset(struct memory_bitmap *bm)
549 {
550 	bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree,
551 				  list);
552 	bm->cur.node = list_entry(bm->cur.zone->leaves.next,
553 				  struct rtree_node, list);
554 	bm->cur.node_pfn = 0;
555 	bm->cur.node_bit = 0;
556 }
557 
558 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
559 
560 struct mem_extent {
561 	struct list_head hook;
562 	unsigned long start;
563 	unsigned long end;
564 };
565 
566 /**
567  * free_mem_extents - Free a list of memory extents.
568  * @list: List of extents to free.
569  */
570 static void free_mem_extents(struct list_head *list)
571 {
572 	struct mem_extent *ext, *aux;
573 
574 	list_for_each_entry_safe(ext, aux, list, hook) {
575 		list_del(&ext->hook);
576 		kfree(ext);
577 	}
578 }
579 
580 /**
581  * create_mem_extents - Create a list of memory extents.
582  * @list: List to put the extents into.
583  * @gfp_mask: Mask to use for memory allocations.
584  *
585  * The extents represent contiguous ranges of PFNs.
586  */
587 static int create_mem_extents(struct list_head *list, gfp_t gfp_mask)
588 {
589 	struct zone *zone;
590 
591 	INIT_LIST_HEAD(list);
592 
593 	for_each_populated_zone(zone) {
594 		unsigned long zone_start, zone_end;
595 		struct mem_extent *ext, *cur, *aux;
596 
597 		zone_start = zone->zone_start_pfn;
598 		zone_end = zone_end_pfn(zone);
599 
600 		list_for_each_entry(ext, list, hook)
601 			if (zone_start <= ext->end)
602 				break;
603 
604 		if (&ext->hook == list || zone_end < ext->start) {
605 			/* New extent is necessary */
606 			struct mem_extent *new_ext;
607 
608 			new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask);
609 			if (!new_ext) {
610 				free_mem_extents(list);
611 				return -ENOMEM;
612 			}
613 			new_ext->start = zone_start;
614 			new_ext->end = zone_end;
615 			list_add_tail(&new_ext->hook, &ext->hook);
616 			continue;
617 		}
618 
619 		/* Merge this zone's range of PFNs with the existing one */
620 		if (zone_start < ext->start)
621 			ext->start = zone_start;
622 		if (zone_end > ext->end)
623 			ext->end = zone_end;
624 
625 		/* More merging may be possible */
626 		cur = ext;
627 		list_for_each_entry_safe_continue(cur, aux, list, hook) {
628 			if (zone_end < cur->start)
629 				break;
630 			if (zone_end < cur->end)
631 				ext->end = cur->end;
632 			list_del(&cur->hook);
633 			kfree(cur);
634 		}
635 	}
636 
637 	return 0;
638 }
639 
640 /**
641  * memory_bm_create - Allocate memory for a memory bitmap.
642  */
643 static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask,
644 			    int safe_needed)
645 {
646 	struct chain_allocator ca;
647 	struct list_head mem_extents;
648 	struct mem_extent *ext;
649 	int error;
650 
651 	chain_init(&ca, gfp_mask, safe_needed);
652 	INIT_LIST_HEAD(&bm->zones);
653 
654 	error = create_mem_extents(&mem_extents, gfp_mask);
655 	if (error)
656 		return error;
657 
658 	list_for_each_entry(ext, &mem_extents, hook) {
659 		struct mem_zone_bm_rtree *zone;
660 
661 		zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca,
662 					    ext->start, ext->end);
663 		if (!zone) {
664 			error = -ENOMEM;
665 			goto Error;
666 		}
667 		list_add_tail(&zone->list, &bm->zones);
668 	}
669 
670 	bm->p_list = ca.chain;
671 	memory_bm_position_reset(bm);
672  Exit:
673 	free_mem_extents(&mem_extents);
674 	return error;
675 
676  Error:
677 	bm->p_list = ca.chain;
678 	memory_bm_free(bm, PG_UNSAFE_CLEAR);
679 	goto Exit;
680 }
681 
682 /**
683  * memory_bm_free - Free memory occupied by the memory bitmap.
684  * @bm: Memory bitmap.
685  */
686 static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free)
687 {
688 	struct mem_zone_bm_rtree *zone;
689 
690 	list_for_each_entry(zone, &bm->zones, list)
691 		free_zone_bm_rtree(zone, clear_nosave_free);
692 
693 	free_list_of_pages(bm->p_list, clear_nosave_free);
694 
695 	INIT_LIST_HEAD(&bm->zones);
696 }
697 
698 /**
699  * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap.
700  *
701  * Find the bit in memory bitmap @bm that corresponds to the given PFN.
702  * The cur.zone, cur.block and cur.node_pfn members of @bm are updated.
703  *
704  * Walk the radix tree to find the page containing the bit that represents @pfn
705  * and return the position of the bit in @addr and @bit_nr.
706  */
707 static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
708 			      void **addr, unsigned int *bit_nr)
709 {
710 	struct mem_zone_bm_rtree *curr, *zone;
711 	struct rtree_node *node;
712 	int i, block_nr;
713 
714 	zone = bm->cur.zone;
715 
716 	if (pfn >= zone->start_pfn && pfn < zone->end_pfn)
717 		goto zone_found;
718 
719 	zone = NULL;
720 
721 	/* Find the right zone */
722 	list_for_each_entry(curr, &bm->zones, list) {
723 		if (pfn >= curr->start_pfn && pfn < curr->end_pfn) {
724 			zone = curr;
725 			break;
726 		}
727 	}
728 
729 	if (!zone)
730 		return -EFAULT;
731 
732 zone_found:
733 	/*
734 	 * We have found the zone. Now walk the radix tree to find the leaf node
735 	 * for our PFN.
736 	 */
737 
738 	/*
739 	 * If the zone we wish to scan is the the current zone and the
740 	 * pfn falls into the current node then we do not need to walk
741 	 * the tree.
742 	 */
743 	node = bm->cur.node;
744 	if (zone == bm->cur.zone &&
745 	    ((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn)
746 		goto node_found;
747 
748 	node      = zone->rtree;
749 	block_nr  = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT;
750 
751 	for (i = zone->levels; i > 0; i--) {
752 		int index;
753 
754 		index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT);
755 		index &= BM_RTREE_LEVEL_MASK;
756 		BUG_ON(node->data[index] == 0);
757 		node = (struct rtree_node *)node->data[index];
758 	}
759 
760 node_found:
761 	/* Update last position */
762 	bm->cur.zone = zone;
763 	bm->cur.node = node;
764 	bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK;
765 
766 	/* Set return values */
767 	*addr = node->data;
768 	*bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK;
769 
770 	return 0;
771 }
772 
773 static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn)
774 {
775 	void *addr;
776 	unsigned int bit;
777 	int error;
778 
779 	error = memory_bm_find_bit(bm, pfn, &addr, &bit);
780 	BUG_ON(error);
781 	set_bit(bit, addr);
782 }
783 
784 static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn)
785 {
786 	void *addr;
787 	unsigned int bit;
788 	int error;
789 
790 	error = memory_bm_find_bit(bm, pfn, &addr, &bit);
791 	if (!error)
792 		set_bit(bit, addr);
793 
794 	return error;
795 }
796 
797 static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn)
798 {
799 	void *addr;
800 	unsigned int bit;
801 	int error;
802 
803 	error = memory_bm_find_bit(bm, pfn, &addr, &bit);
804 	BUG_ON(error);
805 	clear_bit(bit, addr);
806 }
807 
808 static void memory_bm_clear_current(struct memory_bitmap *bm)
809 {
810 	int bit;
811 
812 	bit = max(bm->cur.node_bit - 1, 0);
813 	clear_bit(bit, bm->cur.node->data);
814 }
815 
816 static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
817 {
818 	void *addr;
819 	unsigned int bit;
820 	int error;
821 
822 	error = memory_bm_find_bit(bm, pfn, &addr, &bit);
823 	BUG_ON(error);
824 	return test_bit(bit, addr);
825 }
826 
827 static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn)
828 {
829 	void *addr;
830 	unsigned int bit;
831 
832 	return !memory_bm_find_bit(bm, pfn, &addr, &bit);
833 }
834 
835 /*
836  * rtree_next_node - Jump to the next leaf node.
837  *
838  * Set the position to the beginning of the next node in the
839  * memory bitmap. This is either the next node in the current
840  * zone's radix tree or the first node in the radix tree of the
841  * next zone.
842  *
843  * Return true if there is a next node, false otherwise.
844  */
845 static bool rtree_next_node(struct memory_bitmap *bm)
846 {
847 	if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) {
848 		bm->cur.node = list_entry(bm->cur.node->list.next,
849 					  struct rtree_node, list);
850 		bm->cur.node_pfn += BM_BITS_PER_BLOCK;
851 		bm->cur.node_bit  = 0;
852 		touch_softlockup_watchdog();
853 		return true;
854 	}
855 
856 	/* No more nodes, goto next zone */
857 	if (!list_is_last(&bm->cur.zone->list, &bm->zones)) {
858 		bm->cur.zone = list_entry(bm->cur.zone->list.next,
859 				  struct mem_zone_bm_rtree, list);
860 		bm->cur.node = list_entry(bm->cur.zone->leaves.next,
861 					  struct rtree_node, list);
862 		bm->cur.node_pfn = 0;
863 		bm->cur.node_bit = 0;
864 		return true;
865 	}
866 
867 	/* No more zones */
868 	return false;
869 }
870 
871 /**
872  * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap.
873  * @bm: Memory bitmap.
874  *
875  * Starting from the last returned position this function searches for the next
876  * set bit in @bm and returns the PFN represented by it.  If no more bits are
877  * set, BM_END_OF_MAP is returned.
878  *
879  * It is required to run memory_bm_position_reset() before the first call to
880  * this function for the given memory bitmap.
881  */
882 static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
883 {
884 	unsigned long bits, pfn, pages;
885 	int bit;
886 
887 	do {
888 		pages	  = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn;
889 		bits      = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK);
890 		bit	  = find_next_bit(bm->cur.node->data, bits,
891 					  bm->cur.node_bit);
892 		if (bit < bits) {
893 			pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit;
894 			bm->cur.node_bit = bit + 1;
895 			return pfn;
896 		}
897 	} while (rtree_next_node(bm));
898 
899 	return BM_END_OF_MAP;
900 }
901 
902 /*
903  * This structure represents a range of page frames the contents of which
904  * should not be saved during hibernation.
905  */
906 struct nosave_region {
907 	struct list_head list;
908 	unsigned long start_pfn;
909 	unsigned long end_pfn;
910 };
911 
912 static LIST_HEAD(nosave_regions);
913 
914 static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone)
915 {
916 	struct rtree_node *node;
917 
918 	list_for_each_entry(node, &zone->nodes, list)
919 		recycle_safe_page(node->data);
920 
921 	list_for_each_entry(node, &zone->leaves, list)
922 		recycle_safe_page(node->data);
923 }
924 
925 static void memory_bm_recycle(struct memory_bitmap *bm)
926 {
927 	struct mem_zone_bm_rtree *zone;
928 	struct linked_page *p_list;
929 
930 	list_for_each_entry(zone, &bm->zones, list)
931 		recycle_zone_bm_rtree(zone);
932 
933 	p_list = bm->p_list;
934 	while (p_list) {
935 		struct linked_page *lp = p_list;
936 
937 		p_list = lp->next;
938 		recycle_safe_page(lp);
939 	}
940 }
941 
942 /**
943  * register_nosave_region - Register a region of unsaveable memory.
944  *
945  * Register a range of page frames the contents of which should not be saved
946  * during hibernation (to be used in the early initialization code).
947  */
948 void __init __register_nosave_region(unsigned long start_pfn,
949 				     unsigned long end_pfn, int use_kmalloc)
950 {
951 	struct nosave_region *region;
952 
953 	if (start_pfn >= end_pfn)
954 		return;
955 
956 	if (!list_empty(&nosave_regions)) {
957 		/* Try to extend the previous region (they should be sorted) */
958 		region = list_entry(nosave_regions.prev,
959 					struct nosave_region, list);
960 		if (region->end_pfn == start_pfn) {
961 			region->end_pfn = end_pfn;
962 			goto Report;
963 		}
964 	}
965 	if (use_kmalloc) {
966 		/* During init, this shouldn't fail */
967 		region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL);
968 		BUG_ON(!region);
969 	} else {
970 		/* This allocation cannot fail */
971 		region = memblock_alloc(sizeof(struct nosave_region),
972 					SMP_CACHE_BYTES);
973 		if (!region)
974 			panic("%s: Failed to allocate %zu bytes\n", __func__,
975 			      sizeof(struct nosave_region));
976 	}
977 	region->start_pfn = start_pfn;
978 	region->end_pfn = end_pfn;
979 	list_add_tail(&region->list, &nosave_regions);
980  Report:
981 	pr_info("Registered nosave memory: [mem %#010llx-%#010llx]\n",
982 		(unsigned long long) start_pfn << PAGE_SHIFT,
983 		((unsigned long long) end_pfn << PAGE_SHIFT) - 1);
984 }
985 
986 /*
987  * Set bits in this map correspond to the page frames the contents of which
988  * should not be saved during the suspend.
989  */
990 static struct memory_bitmap *forbidden_pages_map;
991 
992 /* Set bits in this map correspond to free page frames. */
993 static struct memory_bitmap *free_pages_map;
994 
995 /*
996  * Each page frame allocated for creating the image is marked by setting the
997  * corresponding bits in forbidden_pages_map and free_pages_map simultaneously
998  */
999 
1000 void swsusp_set_page_free(struct page *page)
1001 {
1002 	if (free_pages_map)
1003 		memory_bm_set_bit(free_pages_map, page_to_pfn(page));
1004 }
1005 
1006 static int swsusp_page_is_free(struct page *page)
1007 {
1008 	return free_pages_map ?
1009 		memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0;
1010 }
1011 
1012 void swsusp_unset_page_free(struct page *page)
1013 {
1014 	if (free_pages_map)
1015 		memory_bm_clear_bit(free_pages_map, page_to_pfn(page));
1016 }
1017 
1018 static void swsusp_set_page_forbidden(struct page *page)
1019 {
1020 	if (forbidden_pages_map)
1021 		memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page));
1022 }
1023 
1024 int swsusp_page_is_forbidden(struct page *page)
1025 {
1026 	return forbidden_pages_map ?
1027 		memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0;
1028 }
1029 
1030 static void swsusp_unset_page_forbidden(struct page *page)
1031 {
1032 	if (forbidden_pages_map)
1033 		memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page));
1034 }
1035 
1036 /**
1037  * mark_nosave_pages - Mark pages that should not be saved.
1038  * @bm: Memory bitmap.
1039  *
1040  * Set the bits in @bm that correspond to the page frames the contents of which
1041  * should not be saved.
1042  */
1043 static void mark_nosave_pages(struct memory_bitmap *bm)
1044 {
1045 	struct nosave_region *region;
1046 
1047 	if (list_empty(&nosave_regions))
1048 		return;
1049 
1050 	list_for_each_entry(region, &nosave_regions, list) {
1051 		unsigned long pfn;
1052 
1053 		pr_debug("Marking nosave pages: [mem %#010llx-%#010llx]\n",
1054 			 (unsigned long long) region->start_pfn << PAGE_SHIFT,
1055 			 ((unsigned long long) region->end_pfn << PAGE_SHIFT)
1056 				- 1);
1057 
1058 		for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++)
1059 			if (pfn_valid(pfn)) {
1060 				/*
1061 				 * It is safe to ignore the result of
1062 				 * mem_bm_set_bit_check() here, since we won't
1063 				 * touch the PFNs for which the error is
1064 				 * returned anyway.
1065 				 */
1066 				mem_bm_set_bit_check(bm, pfn);
1067 			}
1068 	}
1069 }
1070 
1071 /**
1072  * create_basic_memory_bitmaps - Create bitmaps to hold basic page information.
1073  *
1074  * Create bitmaps needed for marking page frames that should not be saved and
1075  * free page frames.  The forbidden_pages_map and free_pages_map pointers are
1076  * only modified if everything goes well, because we don't want the bits to be
1077  * touched before both bitmaps are set up.
1078  */
1079 int create_basic_memory_bitmaps(void)
1080 {
1081 	struct memory_bitmap *bm1, *bm2;
1082 	int error = 0;
1083 
1084 	if (forbidden_pages_map && free_pages_map)
1085 		return 0;
1086 	else
1087 		BUG_ON(forbidden_pages_map || free_pages_map);
1088 
1089 	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1090 	if (!bm1)
1091 		return -ENOMEM;
1092 
1093 	error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
1094 	if (error)
1095 		goto Free_first_object;
1096 
1097 	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
1098 	if (!bm2)
1099 		goto Free_first_bitmap;
1100 
1101 	error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
1102 	if (error)
1103 		goto Free_second_object;
1104 
1105 	forbidden_pages_map = bm1;
1106 	free_pages_map = bm2;
1107 	mark_nosave_pages(forbidden_pages_map);
1108 
1109 	pr_debug("Basic memory bitmaps created\n");
1110 
1111 	return 0;
1112 
1113  Free_second_object:
1114 	kfree(bm2);
1115  Free_first_bitmap:
1116  	memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1117  Free_first_object:
1118 	kfree(bm1);
1119 	return -ENOMEM;
1120 }
1121 
1122 /**
1123  * free_basic_memory_bitmaps - Free memory bitmaps holding basic information.
1124  *
1125  * Free memory bitmaps allocated by create_basic_memory_bitmaps().  The
1126  * auxiliary pointers are necessary so that the bitmaps themselves are not
1127  * referred to while they are being freed.
1128  */
1129 void free_basic_memory_bitmaps(void)
1130 {
1131 	struct memory_bitmap *bm1, *bm2;
1132 
1133 	if (WARN_ON(!(forbidden_pages_map && free_pages_map)))
1134 		return;
1135 
1136 	bm1 = forbidden_pages_map;
1137 	bm2 = free_pages_map;
1138 	forbidden_pages_map = NULL;
1139 	free_pages_map = NULL;
1140 	memory_bm_free(bm1, PG_UNSAFE_CLEAR);
1141 	kfree(bm1);
1142 	memory_bm_free(bm2, PG_UNSAFE_CLEAR);
1143 	kfree(bm2);
1144 
1145 	pr_debug("Basic memory bitmaps freed\n");
1146 }
1147 
1148 void clear_free_pages(void)
1149 {
1150 	struct memory_bitmap *bm = free_pages_map;
1151 	unsigned long pfn;
1152 
1153 	if (WARN_ON(!(free_pages_map)))
1154 		return;
1155 
1156 	if (IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) || want_init_on_free()) {
1157 		memory_bm_position_reset(bm);
1158 		pfn = memory_bm_next_pfn(bm);
1159 		while (pfn != BM_END_OF_MAP) {
1160 			if (pfn_valid(pfn))
1161 				clear_highpage(pfn_to_page(pfn));
1162 
1163 			pfn = memory_bm_next_pfn(bm);
1164 		}
1165 		memory_bm_position_reset(bm);
1166 		pr_info("free pages cleared after restore\n");
1167 	}
1168 }
1169 
1170 /**
1171  * snapshot_additional_pages - Estimate the number of extra pages needed.
1172  * @zone: Memory zone to carry out the computation for.
1173  *
1174  * Estimate the number of additional pages needed for setting up a hibernation
1175  * image data structures for @zone (usually, the returned value is greater than
1176  * the exact number).
1177  */
1178 unsigned int snapshot_additional_pages(struct zone *zone)
1179 {
1180 	unsigned int rtree, nodes;
1181 
1182 	rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK);
1183 	rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node),
1184 			      LINKED_PAGE_DATA_SIZE);
1185 	while (nodes > 1) {
1186 		nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL);
1187 		rtree += nodes;
1188 	}
1189 
1190 	return 2 * rtree;
1191 }
1192 
1193 #ifdef CONFIG_HIGHMEM
1194 /**
1195  * count_free_highmem_pages - Compute the total number of free highmem pages.
1196  *
1197  * The returned number is system-wide.
1198  */
1199 static unsigned int count_free_highmem_pages(void)
1200 {
1201 	struct zone *zone;
1202 	unsigned int cnt = 0;
1203 
1204 	for_each_populated_zone(zone)
1205 		if (is_highmem(zone))
1206 			cnt += zone_page_state(zone, NR_FREE_PAGES);
1207 
1208 	return cnt;
1209 }
1210 
1211 /**
1212  * saveable_highmem_page - Check if a highmem page is saveable.
1213  *
1214  * Determine whether a highmem page should be included in a hibernation image.
1215  *
1216  * We should save the page if it isn't Nosave or NosaveFree, or Reserved,
1217  * and it isn't part of a free chunk of pages.
1218  */
1219 static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn)
1220 {
1221 	struct page *page;
1222 
1223 	if (!pfn_valid(pfn))
1224 		return NULL;
1225 
1226 	page = pfn_to_online_page(pfn);
1227 	if (!page || page_zone(page) != zone)
1228 		return NULL;
1229 
1230 	BUG_ON(!PageHighMem(page));
1231 
1232 	if (swsusp_page_is_forbidden(page) ||  swsusp_page_is_free(page))
1233 		return NULL;
1234 
1235 	if (PageReserved(page) || PageOffline(page))
1236 		return NULL;
1237 
1238 	if (page_is_guard(page))
1239 		return NULL;
1240 
1241 	return page;
1242 }
1243 
1244 /**
1245  * count_highmem_pages - Compute the total number of saveable highmem pages.
1246  */
1247 static unsigned int count_highmem_pages(void)
1248 {
1249 	struct zone *zone;
1250 	unsigned int n = 0;
1251 
1252 	for_each_populated_zone(zone) {
1253 		unsigned long pfn, max_zone_pfn;
1254 
1255 		if (!is_highmem(zone))
1256 			continue;
1257 
1258 		mark_free_pages(zone);
1259 		max_zone_pfn = zone_end_pfn(zone);
1260 		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1261 			if (saveable_highmem_page(zone, pfn))
1262 				n++;
1263 	}
1264 	return n;
1265 }
1266 #else
1267 static inline void *saveable_highmem_page(struct zone *z, unsigned long p)
1268 {
1269 	return NULL;
1270 }
1271 #endif /* CONFIG_HIGHMEM */
1272 
1273 /**
1274  * saveable_page - Check if the given page is saveable.
1275  *
1276  * Determine whether a non-highmem page should be included in a hibernation
1277  * image.
1278  *
1279  * We should save the page if it isn't Nosave, and is not in the range
1280  * of pages statically defined as 'unsaveable', and it isn't part of
1281  * a free chunk of pages.
1282  */
1283 static struct page *saveable_page(struct zone *zone, unsigned long pfn)
1284 {
1285 	struct page *page;
1286 
1287 	if (!pfn_valid(pfn))
1288 		return NULL;
1289 
1290 	page = pfn_to_online_page(pfn);
1291 	if (!page || page_zone(page) != zone)
1292 		return NULL;
1293 
1294 	BUG_ON(PageHighMem(page));
1295 
1296 	if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page))
1297 		return NULL;
1298 
1299 	if (PageOffline(page))
1300 		return NULL;
1301 
1302 	if (PageReserved(page)
1303 	    && (!kernel_page_present(page) || pfn_is_nosave(pfn)))
1304 		return NULL;
1305 
1306 	if (page_is_guard(page))
1307 		return NULL;
1308 
1309 	return page;
1310 }
1311 
1312 /**
1313  * count_data_pages - Compute the total number of saveable non-highmem pages.
1314  */
1315 static unsigned int count_data_pages(void)
1316 {
1317 	struct zone *zone;
1318 	unsigned long pfn, max_zone_pfn;
1319 	unsigned int n = 0;
1320 
1321 	for_each_populated_zone(zone) {
1322 		if (is_highmem(zone))
1323 			continue;
1324 
1325 		mark_free_pages(zone);
1326 		max_zone_pfn = zone_end_pfn(zone);
1327 		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1328 			if (saveable_page(zone, pfn))
1329 				n++;
1330 	}
1331 	return n;
1332 }
1333 
1334 /*
1335  * This is needed, because copy_page and memcpy are not usable for copying
1336  * task structs.
1337  */
1338 static inline void do_copy_page(long *dst, long *src)
1339 {
1340 	int n;
1341 
1342 	for (n = PAGE_SIZE / sizeof(long); n; n--)
1343 		*dst++ = *src++;
1344 }
1345 
1346 /**
1347  * safe_copy_page - Copy a page in a safe way.
1348  *
1349  * Check if the page we are going to copy is marked as present in the kernel
1350  * page tables. This always is the case if CONFIG_DEBUG_PAGEALLOC or
1351  * CONFIG_ARCH_HAS_SET_DIRECT_MAP is not set. In that case kernel_page_present()
1352  * always returns 'true'.
1353  */
1354 static void safe_copy_page(void *dst, struct page *s_page)
1355 {
1356 	if (kernel_page_present(s_page)) {
1357 		do_copy_page(dst, page_address(s_page));
1358 	} else {
1359 		kernel_map_pages(s_page, 1, 1);
1360 		do_copy_page(dst, page_address(s_page));
1361 		kernel_map_pages(s_page, 1, 0);
1362 	}
1363 }
1364 
1365 #ifdef CONFIG_HIGHMEM
1366 static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn)
1367 {
1368 	return is_highmem(zone) ?
1369 		saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn);
1370 }
1371 
1372 static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1373 {
1374 	struct page *s_page, *d_page;
1375 	void *src, *dst;
1376 
1377 	s_page = pfn_to_page(src_pfn);
1378 	d_page = pfn_to_page(dst_pfn);
1379 	if (PageHighMem(s_page)) {
1380 		src = kmap_atomic(s_page);
1381 		dst = kmap_atomic(d_page);
1382 		do_copy_page(dst, src);
1383 		kunmap_atomic(dst);
1384 		kunmap_atomic(src);
1385 	} else {
1386 		if (PageHighMem(d_page)) {
1387 			/*
1388 			 * The page pointed to by src may contain some kernel
1389 			 * data modified by kmap_atomic()
1390 			 */
1391 			safe_copy_page(buffer, s_page);
1392 			dst = kmap_atomic(d_page);
1393 			copy_page(dst, buffer);
1394 			kunmap_atomic(dst);
1395 		} else {
1396 			safe_copy_page(page_address(d_page), s_page);
1397 		}
1398 	}
1399 }
1400 #else
1401 #define page_is_saveable(zone, pfn)	saveable_page(zone, pfn)
1402 
1403 static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn)
1404 {
1405 	safe_copy_page(page_address(pfn_to_page(dst_pfn)),
1406 				pfn_to_page(src_pfn));
1407 }
1408 #endif /* CONFIG_HIGHMEM */
1409 
1410 static void copy_data_pages(struct memory_bitmap *copy_bm,
1411 			    struct memory_bitmap *orig_bm)
1412 {
1413 	struct zone *zone;
1414 	unsigned long pfn;
1415 
1416 	for_each_populated_zone(zone) {
1417 		unsigned long max_zone_pfn;
1418 
1419 		mark_free_pages(zone);
1420 		max_zone_pfn = zone_end_pfn(zone);
1421 		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
1422 			if (page_is_saveable(zone, pfn))
1423 				memory_bm_set_bit(orig_bm, pfn);
1424 	}
1425 	memory_bm_position_reset(orig_bm);
1426 	memory_bm_position_reset(copy_bm);
1427 	for(;;) {
1428 		pfn = memory_bm_next_pfn(orig_bm);
1429 		if (unlikely(pfn == BM_END_OF_MAP))
1430 			break;
1431 		copy_data_page(memory_bm_next_pfn(copy_bm), pfn);
1432 	}
1433 }
1434 
1435 /* Total number of image pages */
1436 static unsigned int nr_copy_pages;
1437 /* Number of pages needed for saving the original pfns of the image pages */
1438 static unsigned int nr_meta_pages;
1439 /*
1440  * Numbers of normal and highmem page frames allocated for hibernation image
1441  * before suspending devices.
1442  */
1443 static unsigned int alloc_normal, alloc_highmem;
1444 /*
1445  * Memory bitmap used for marking saveable pages (during hibernation) or
1446  * hibernation image pages (during restore)
1447  */
1448 static struct memory_bitmap orig_bm;
1449 /*
1450  * Memory bitmap used during hibernation for marking allocated page frames that
1451  * will contain copies of saveable pages.  During restore it is initially used
1452  * for marking hibernation image pages, but then the set bits from it are
1453  * duplicated in @orig_bm and it is released.  On highmem systems it is next
1454  * used for marking "safe" highmem pages, but it has to be reinitialized for
1455  * this purpose.
1456  */
1457 static struct memory_bitmap copy_bm;
1458 
1459 /**
1460  * swsusp_free - Free pages allocated for hibernation image.
1461  *
1462  * Image pages are alocated before snapshot creation, so they need to be
1463  * released after resume.
1464  */
1465 void swsusp_free(void)
1466 {
1467 	unsigned long fb_pfn, fr_pfn;
1468 
1469 	if (!forbidden_pages_map || !free_pages_map)
1470 		goto out;
1471 
1472 	memory_bm_position_reset(forbidden_pages_map);
1473 	memory_bm_position_reset(free_pages_map);
1474 
1475 loop:
1476 	fr_pfn = memory_bm_next_pfn(free_pages_map);
1477 	fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1478 
1479 	/*
1480 	 * Find the next bit set in both bitmaps. This is guaranteed to
1481 	 * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP.
1482 	 */
1483 	do {
1484 		if (fb_pfn < fr_pfn)
1485 			fb_pfn = memory_bm_next_pfn(forbidden_pages_map);
1486 		if (fr_pfn < fb_pfn)
1487 			fr_pfn = memory_bm_next_pfn(free_pages_map);
1488 	} while (fb_pfn != fr_pfn);
1489 
1490 	if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) {
1491 		struct page *page = pfn_to_page(fr_pfn);
1492 
1493 		memory_bm_clear_current(forbidden_pages_map);
1494 		memory_bm_clear_current(free_pages_map);
1495 		hibernate_restore_unprotect_page(page_address(page));
1496 		__free_page(page);
1497 		goto loop;
1498 	}
1499 
1500 out:
1501 	nr_copy_pages = 0;
1502 	nr_meta_pages = 0;
1503 	restore_pblist = NULL;
1504 	buffer = NULL;
1505 	alloc_normal = 0;
1506 	alloc_highmem = 0;
1507 	hibernate_restore_protection_end();
1508 }
1509 
1510 /* Helper functions used for the shrinking of memory. */
1511 
1512 #define GFP_IMAGE	(GFP_KERNEL | __GFP_NOWARN)
1513 
1514 /**
1515  * preallocate_image_pages - Allocate a number of pages for hibernation image.
1516  * @nr_pages: Number of page frames to allocate.
1517  * @mask: GFP flags to use for the allocation.
1518  *
1519  * Return value: Number of page frames actually allocated
1520  */
1521 static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask)
1522 {
1523 	unsigned long nr_alloc = 0;
1524 
1525 	while (nr_pages > 0) {
1526 		struct page *page;
1527 
1528 		page = alloc_image_page(mask);
1529 		if (!page)
1530 			break;
1531 		memory_bm_set_bit(&copy_bm, page_to_pfn(page));
1532 		if (PageHighMem(page))
1533 			alloc_highmem++;
1534 		else
1535 			alloc_normal++;
1536 		nr_pages--;
1537 		nr_alloc++;
1538 	}
1539 
1540 	return nr_alloc;
1541 }
1542 
1543 static unsigned long preallocate_image_memory(unsigned long nr_pages,
1544 					      unsigned long avail_normal)
1545 {
1546 	unsigned long alloc;
1547 
1548 	if (avail_normal <= alloc_normal)
1549 		return 0;
1550 
1551 	alloc = avail_normal - alloc_normal;
1552 	if (nr_pages < alloc)
1553 		alloc = nr_pages;
1554 
1555 	return preallocate_image_pages(alloc, GFP_IMAGE);
1556 }
1557 
1558 #ifdef CONFIG_HIGHMEM
1559 static unsigned long preallocate_image_highmem(unsigned long nr_pages)
1560 {
1561 	return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM);
1562 }
1563 
1564 /**
1565  *  __fraction - Compute (an approximation of) x * (multiplier / base).
1566  */
1567 static unsigned long __fraction(u64 x, u64 multiplier, u64 base)
1568 {
1569 	return div64_u64(x * multiplier, base);
1570 }
1571 
1572 static unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1573 						  unsigned long highmem,
1574 						  unsigned long total)
1575 {
1576 	unsigned long alloc = __fraction(nr_pages, highmem, total);
1577 
1578 	return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM);
1579 }
1580 #else /* CONFIG_HIGHMEM */
1581 static inline unsigned long preallocate_image_highmem(unsigned long nr_pages)
1582 {
1583 	return 0;
1584 }
1585 
1586 static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages,
1587 							 unsigned long highmem,
1588 							 unsigned long total)
1589 {
1590 	return 0;
1591 }
1592 #endif /* CONFIG_HIGHMEM */
1593 
1594 /**
1595  * free_unnecessary_pages - Release preallocated pages not needed for the image.
1596  */
1597 static unsigned long free_unnecessary_pages(void)
1598 {
1599 	unsigned long save, to_free_normal, to_free_highmem, free;
1600 
1601 	save = count_data_pages();
1602 	if (alloc_normal >= save) {
1603 		to_free_normal = alloc_normal - save;
1604 		save = 0;
1605 	} else {
1606 		to_free_normal = 0;
1607 		save -= alloc_normal;
1608 	}
1609 	save += count_highmem_pages();
1610 	if (alloc_highmem >= save) {
1611 		to_free_highmem = alloc_highmem - save;
1612 	} else {
1613 		to_free_highmem = 0;
1614 		save -= alloc_highmem;
1615 		if (to_free_normal > save)
1616 			to_free_normal -= save;
1617 		else
1618 			to_free_normal = 0;
1619 	}
1620 	free = to_free_normal + to_free_highmem;
1621 
1622 	memory_bm_position_reset(&copy_bm);
1623 
1624 	while (to_free_normal > 0 || to_free_highmem > 0) {
1625 		unsigned long pfn = memory_bm_next_pfn(&copy_bm);
1626 		struct page *page = pfn_to_page(pfn);
1627 
1628 		if (PageHighMem(page)) {
1629 			if (!to_free_highmem)
1630 				continue;
1631 			to_free_highmem--;
1632 			alloc_highmem--;
1633 		} else {
1634 			if (!to_free_normal)
1635 				continue;
1636 			to_free_normal--;
1637 			alloc_normal--;
1638 		}
1639 		memory_bm_clear_bit(&copy_bm, pfn);
1640 		swsusp_unset_page_forbidden(page);
1641 		swsusp_unset_page_free(page);
1642 		__free_page(page);
1643 	}
1644 
1645 	return free;
1646 }
1647 
1648 /**
1649  * minimum_image_size - Estimate the minimum acceptable size of an image.
1650  * @saveable: Number of saveable pages in the system.
1651  *
1652  * We want to avoid attempting to free too much memory too hard, so estimate the
1653  * minimum acceptable size of a hibernation image to use as the lower limit for
1654  * preallocating memory.
1655  *
1656  * We assume that the minimum image size should be proportional to
1657  *
1658  * [number of saveable pages] - [number of pages that can be freed in theory]
1659  *
1660  * where the second term is the sum of (1) reclaimable slab pages, (2) active
1661  * and (3) inactive anonymous pages, (4) active and (5) inactive file pages.
1662  */
1663 static unsigned long minimum_image_size(unsigned long saveable)
1664 {
1665 	unsigned long size;
1666 
1667 	size = global_node_page_state(NR_SLAB_RECLAIMABLE)
1668 		+ global_node_page_state(NR_ACTIVE_ANON)
1669 		+ global_node_page_state(NR_INACTIVE_ANON)
1670 		+ global_node_page_state(NR_ACTIVE_FILE)
1671 		+ global_node_page_state(NR_INACTIVE_FILE);
1672 
1673 	return saveable <= size ? 0 : saveable - size;
1674 }
1675 
1676 /**
1677  * hibernate_preallocate_memory - Preallocate memory for hibernation image.
1678  *
1679  * To create a hibernation image it is necessary to make a copy of every page
1680  * frame in use.  We also need a number of page frames to be free during
1681  * hibernation for allocations made while saving the image and for device
1682  * drivers, in case they need to allocate memory from their hibernation
1683  * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough
1684  * estimate) and reserved_size divided by PAGE_SIZE (which is tunable through
1685  * /sys/power/reserved_size, respectively).  To make this happen, we compute the
1686  * total number of available page frames and allocate at least
1687  *
1688  * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2
1689  *  + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE)
1690  *
1691  * of them, which corresponds to the maximum size of a hibernation image.
1692  *
1693  * If image_size is set below the number following from the above formula,
1694  * the preallocation of memory is continued until the total number of saveable
1695  * pages in the system is below the requested image size or the minimum
1696  * acceptable image size returned by minimum_image_size(), whichever is greater.
1697  */
1698 int hibernate_preallocate_memory(void)
1699 {
1700 	struct zone *zone;
1701 	unsigned long saveable, size, max_size, count, highmem, pages = 0;
1702 	unsigned long alloc, save_highmem, pages_highmem, avail_normal;
1703 	ktime_t start, stop;
1704 	int error;
1705 
1706 	pr_info("Preallocating image memory\n");
1707 	start = ktime_get();
1708 
1709 	error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY);
1710 	if (error) {
1711 		pr_err("Cannot allocate original bitmap\n");
1712 		goto err_out;
1713 	}
1714 
1715 	error = memory_bm_create(&copy_bm, GFP_IMAGE, PG_ANY);
1716 	if (error) {
1717 		pr_err("Cannot allocate copy bitmap\n");
1718 		goto err_out;
1719 	}
1720 
1721 	alloc_normal = 0;
1722 	alloc_highmem = 0;
1723 
1724 	/* Count the number of saveable data pages. */
1725 	save_highmem = count_highmem_pages();
1726 	saveable = count_data_pages();
1727 
1728 	/*
1729 	 * Compute the total number of page frames we can use (count) and the
1730 	 * number of pages needed for image metadata (size).
1731 	 */
1732 	count = saveable;
1733 	saveable += save_highmem;
1734 	highmem = save_highmem;
1735 	size = 0;
1736 	for_each_populated_zone(zone) {
1737 		size += snapshot_additional_pages(zone);
1738 		if (is_highmem(zone))
1739 			highmem += zone_page_state(zone, NR_FREE_PAGES);
1740 		else
1741 			count += zone_page_state(zone, NR_FREE_PAGES);
1742 	}
1743 	avail_normal = count;
1744 	count += highmem;
1745 	count -= totalreserve_pages;
1746 
1747 	/* Add number of pages required for page keys (s390 only). */
1748 	size += page_key_additional_pages(saveable);
1749 
1750 	/* Compute the maximum number of saveable pages to leave in memory. */
1751 	max_size = (count - (size + PAGES_FOR_IO)) / 2
1752 			- 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE);
1753 	/* Compute the desired number of image pages specified by image_size. */
1754 	size = DIV_ROUND_UP(image_size, PAGE_SIZE);
1755 	if (size > max_size)
1756 		size = max_size;
1757 	/*
1758 	 * If the desired number of image pages is at least as large as the
1759 	 * current number of saveable pages in memory, allocate page frames for
1760 	 * the image and we're done.
1761 	 */
1762 	if (size >= saveable) {
1763 		pages = preallocate_image_highmem(save_highmem);
1764 		pages += preallocate_image_memory(saveable - pages, avail_normal);
1765 		goto out;
1766 	}
1767 
1768 	/* Estimate the minimum size of the image. */
1769 	pages = minimum_image_size(saveable);
1770 	/*
1771 	 * To avoid excessive pressure on the normal zone, leave room in it to
1772 	 * accommodate an image of the minimum size (unless it's already too
1773 	 * small, in which case don't preallocate pages from it at all).
1774 	 */
1775 	if (avail_normal > pages)
1776 		avail_normal -= pages;
1777 	else
1778 		avail_normal = 0;
1779 	if (size < pages)
1780 		size = min_t(unsigned long, pages, max_size);
1781 
1782 	/*
1783 	 * Let the memory management subsystem know that we're going to need a
1784 	 * large number of page frames to allocate and make it free some memory.
1785 	 * NOTE: If this is not done, performance will be hurt badly in some
1786 	 * test cases.
1787 	 */
1788 	shrink_all_memory(saveable - size);
1789 
1790 	/*
1791 	 * The number of saveable pages in memory was too high, so apply some
1792 	 * pressure to decrease it.  First, make room for the largest possible
1793 	 * image and fail if that doesn't work.  Next, try to decrease the size
1794 	 * of the image as much as indicated by 'size' using allocations from
1795 	 * highmem and non-highmem zones separately.
1796 	 */
1797 	pages_highmem = preallocate_image_highmem(highmem / 2);
1798 	alloc = count - max_size;
1799 	if (alloc > pages_highmem)
1800 		alloc -= pages_highmem;
1801 	else
1802 		alloc = 0;
1803 	pages = preallocate_image_memory(alloc, avail_normal);
1804 	if (pages < alloc) {
1805 		/* We have exhausted non-highmem pages, try highmem. */
1806 		alloc -= pages;
1807 		pages += pages_highmem;
1808 		pages_highmem = preallocate_image_highmem(alloc);
1809 		if (pages_highmem < alloc) {
1810 			pr_err("Image allocation is %lu pages short\n",
1811 				alloc - pages_highmem);
1812 			goto err_out;
1813 		}
1814 		pages += pages_highmem;
1815 		/*
1816 		 * size is the desired number of saveable pages to leave in
1817 		 * memory, so try to preallocate (all memory - size) pages.
1818 		 */
1819 		alloc = (count - pages) - size;
1820 		pages += preallocate_image_highmem(alloc);
1821 	} else {
1822 		/*
1823 		 * There are approximately max_size saveable pages at this point
1824 		 * and we want to reduce this number down to size.
1825 		 */
1826 		alloc = max_size - size;
1827 		size = preallocate_highmem_fraction(alloc, highmem, count);
1828 		pages_highmem += size;
1829 		alloc -= size;
1830 		size = preallocate_image_memory(alloc, avail_normal);
1831 		pages_highmem += preallocate_image_highmem(alloc - size);
1832 		pages += pages_highmem + size;
1833 	}
1834 
1835 	/*
1836 	 * We only need as many page frames for the image as there are saveable
1837 	 * pages in memory, but we have allocated more.  Release the excessive
1838 	 * ones now.
1839 	 */
1840 	pages -= free_unnecessary_pages();
1841 
1842  out:
1843 	stop = ktime_get();
1844 	pr_info("Allocated %lu pages for snapshot\n", pages);
1845 	swsusp_show_speed(start, stop, pages, "Allocated");
1846 
1847 	return 0;
1848 
1849  err_out:
1850 	swsusp_free();
1851 	return -ENOMEM;
1852 }
1853 
1854 #ifdef CONFIG_HIGHMEM
1855 /**
1856  * count_pages_for_highmem - Count non-highmem pages needed for copying highmem.
1857  *
1858  * Compute the number of non-highmem pages that will be necessary for creating
1859  * copies of highmem pages.
1860  */
1861 static unsigned int count_pages_for_highmem(unsigned int nr_highmem)
1862 {
1863 	unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem;
1864 
1865 	if (free_highmem >= nr_highmem)
1866 		nr_highmem = 0;
1867 	else
1868 		nr_highmem -= free_highmem;
1869 
1870 	return nr_highmem;
1871 }
1872 #else
1873 static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; }
1874 #endif /* CONFIG_HIGHMEM */
1875 
1876 /**
1877  * enough_free_mem - Check if there is enough free memory for the image.
1878  */
1879 static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem)
1880 {
1881 	struct zone *zone;
1882 	unsigned int free = alloc_normal;
1883 
1884 	for_each_populated_zone(zone)
1885 		if (!is_highmem(zone))
1886 			free += zone_page_state(zone, NR_FREE_PAGES);
1887 
1888 	nr_pages += count_pages_for_highmem(nr_highmem);
1889 	pr_debug("Normal pages needed: %u + %u, available pages: %u\n",
1890 		 nr_pages, PAGES_FOR_IO, free);
1891 
1892 	return free > nr_pages + PAGES_FOR_IO;
1893 }
1894 
1895 #ifdef CONFIG_HIGHMEM
1896 /**
1897  * get_highmem_buffer - Allocate a buffer for highmem pages.
1898  *
1899  * If there are some highmem pages in the hibernation image, we may need a
1900  * buffer to copy them and/or load their data.
1901  */
1902 static inline int get_highmem_buffer(int safe_needed)
1903 {
1904 	buffer = get_image_page(GFP_ATOMIC, safe_needed);
1905 	return buffer ? 0 : -ENOMEM;
1906 }
1907 
1908 /**
1909  * alloc_highmem_image_pages - Allocate some highmem pages for the image.
1910  *
1911  * Try to allocate as many pages as needed, but if the number of free highmem
1912  * pages is less than that, allocate them all.
1913  */
1914 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1915 					       unsigned int nr_highmem)
1916 {
1917 	unsigned int to_alloc = count_free_highmem_pages();
1918 
1919 	if (to_alloc > nr_highmem)
1920 		to_alloc = nr_highmem;
1921 
1922 	nr_highmem -= to_alloc;
1923 	while (to_alloc-- > 0) {
1924 		struct page *page;
1925 
1926 		page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM);
1927 		memory_bm_set_bit(bm, page_to_pfn(page));
1928 	}
1929 	return nr_highmem;
1930 }
1931 #else
1932 static inline int get_highmem_buffer(int safe_needed) { return 0; }
1933 
1934 static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm,
1935 					       unsigned int n) { return 0; }
1936 #endif /* CONFIG_HIGHMEM */
1937 
1938 /**
1939  * swsusp_alloc - Allocate memory for hibernation image.
1940  *
1941  * We first try to allocate as many highmem pages as there are
1942  * saveable highmem pages in the system.  If that fails, we allocate
1943  * non-highmem pages for the copies of the remaining highmem ones.
1944  *
1945  * In this approach it is likely that the copies of highmem pages will
1946  * also be located in the high memory, because of the way in which
1947  * copy_data_pages() works.
1948  */
1949 static int swsusp_alloc(struct memory_bitmap *copy_bm,
1950 			unsigned int nr_pages, unsigned int nr_highmem)
1951 {
1952 	if (nr_highmem > 0) {
1953 		if (get_highmem_buffer(PG_ANY))
1954 			goto err_out;
1955 		if (nr_highmem > alloc_highmem) {
1956 			nr_highmem -= alloc_highmem;
1957 			nr_pages += alloc_highmem_pages(copy_bm, nr_highmem);
1958 		}
1959 	}
1960 	if (nr_pages > alloc_normal) {
1961 		nr_pages -= alloc_normal;
1962 		while (nr_pages-- > 0) {
1963 			struct page *page;
1964 
1965 			page = alloc_image_page(GFP_ATOMIC);
1966 			if (!page)
1967 				goto err_out;
1968 			memory_bm_set_bit(copy_bm, page_to_pfn(page));
1969 		}
1970 	}
1971 
1972 	return 0;
1973 
1974  err_out:
1975 	swsusp_free();
1976 	return -ENOMEM;
1977 }
1978 
1979 asmlinkage __visible int swsusp_save(void)
1980 {
1981 	unsigned int nr_pages, nr_highmem;
1982 
1983 	pr_info("Creating image:\n");
1984 
1985 	drain_local_pages(NULL);
1986 	nr_pages = count_data_pages();
1987 	nr_highmem = count_highmem_pages();
1988 	pr_info("Need to copy %u pages\n", nr_pages + nr_highmem);
1989 
1990 	if (!enough_free_mem(nr_pages, nr_highmem)) {
1991 		pr_err("Not enough free memory\n");
1992 		return -ENOMEM;
1993 	}
1994 
1995 	if (swsusp_alloc(&copy_bm, nr_pages, nr_highmem)) {
1996 		pr_err("Memory allocation failed\n");
1997 		return -ENOMEM;
1998 	}
1999 
2000 	/*
2001 	 * During allocating of suspend pagedir, new cold pages may appear.
2002 	 * Kill them.
2003 	 */
2004 	drain_local_pages(NULL);
2005 	copy_data_pages(&copy_bm, &orig_bm);
2006 
2007 	/*
2008 	 * End of critical section. From now on, we can write to memory,
2009 	 * but we should not touch disk. This specially means we must _not_
2010 	 * touch swap space! Except we must write out our image of course.
2011 	 */
2012 
2013 	nr_pages += nr_highmem;
2014 	nr_copy_pages = nr_pages;
2015 	nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE);
2016 
2017 	pr_info("Image created (%d pages copied)\n", nr_pages);
2018 
2019 	return 0;
2020 }
2021 
2022 #ifndef CONFIG_ARCH_HIBERNATION_HEADER
2023 static int init_header_complete(struct swsusp_info *info)
2024 {
2025 	memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname));
2026 	info->version_code = LINUX_VERSION_CODE;
2027 	return 0;
2028 }
2029 
2030 static char *check_image_kernel(struct swsusp_info *info)
2031 {
2032 	if (info->version_code != LINUX_VERSION_CODE)
2033 		return "kernel version";
2034 	if (strcmp(info->uts.sysname,init_utsname()->sysname))
2035 		return "system type";
2036 	if (strcmp(info->uts.release,init_utsname()->release))
2037 		return "kernel release";
2038 	if (strcmp(info->uts.version,init_utsname()->version))
2039 		return "version";
2040 	if (strcmp(info->uts.machine,init_utsname()->machine))
2041 		return "machine";
2042 	return NULL;
2043 }
2044 #endif /* CONFIG_ARCH_HIBERNATION_HEADER */
2045 
2046 unsigned long snapshot_get_image_size(void)
2047 {
2048 	return nr_copy_pages + nr_meta_pages + 1;
2049 }
2050 
2051 static int init_header(struct swsusp_info *info)
2052 {
2053 	memset(info, 0, sizeof(struct swsusp_info));
2054 	info->num_physpages = get_num_physpages();
2055 	info->image_pages = nr_copy_pages;
2056 	info->pages = snapshot_get_image_size();
2057 	info->size = info->pages;
2058 	info->size <<= PAGE_SHIFT;
2059 	return init_header_complete(info);
2060 }
2061 
2062 /**
2063  * pack_pfns - Prepare PFNs for saving.
2064  * @bm: Memory bitmap.
2065  * @buf: Memory buffer to store the PFNs in.
2066  *
2067  * PFNs corresponding to set bits in @bm are stored in the area of memory
2068  * pointed to by @buf (1 page at a time).
2069  */
2070 static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
2071 {
2072 	int j;
2073 
2074 	for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2075 		buf[j] = memory_bm_next_pfn(bm);
2076 		if (unlikely(buf[j] == BM_END_OF_MAP))
2077 			break;
2078 		/* Save page key for data page (s390 only). */
2079 		page_key_read(buf + j);
2080 	}
2081 }
2082 
2083 /**
2084  * snapshot_read_next - Get the address to read the next image page from.
2085  * @handle: Snapshot handle to be used for the reading.
2086  *
2087  * On the first call, @handle should point to a zeroed snapshot_handle
2088  * structure.  The structure gets populated then and a pointer to it should be
2089  * passed to this function every next time.
2090  *
2091  * On success, the function returns a positive number.  Then, the caller
2092  * is allowed to read up to the returned number of bytes from the memory
2093  * location computed by the data_of() macro.
2094  *
2095  * The function returns 0 to indicate the end of the data stream condition,
2096  * and negative numbers are returned on errors.  If that happens, the structure
2097  * pointed to by @handle is not updated and should not be used any more.
2098  */
2099 int snapshot_read_next(struct snapshot_handle *handle)
2100 {
2101 	if (handle->cur > nr_meta_pages + nr_copy_pages)
2102 		return 0;
2103 
2104 	if (!buffer) {
2105 		/* This makes the buffer be freed by swsusp_free() */
2106 		buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2107 		if (!buffer)
2108 			return -ENOMEM;
2109 	}
2110 	if (!handle->cur) {
2111 		int error;
2112 
2113 		error = init_header((struct swsusp_info *)buffer);
2114 		if (error)
2115 			return error;
2116 		handle->buffer = buffer;
2117 		memory_bm_position_reset(&orig_bm);
2118 		memory_bm_position_reset(&copy_bm);
2119 	} else if (handle->cur <= nr_meta_pages) {
2120 		clear_page(buffer);
2121 		pack_pfns(buffer, &orig_bm);
2122 	} else {
2123 		struct page *page;
2124 
2125 		page = pfn_to_page(memory_bm_next_pfn(&copy_bm));
2126 		if (PageHighMem(page)) {
2127 			/*
2128 			 * Highmem pages are copied to the buffer,
2129 			 * because we can't return with a kmapped
2130 			 * highmem page (we may not be called again).
2131 			 */
2132 			void *kaddr;
2133 
2134 			kaddr = kmap_atomic(page);
2135 			copy_page(buffer, kaddr);
2136 			kunmap_atomic(kaddr);
2137 			handle->buffer = buffer;
2138 		} else {
2139 			handle->buffer = page_address(page);
2140 		}
2141 	}
2142 	handle->cur++;
2143 	return PAGE_SIZE;
2144 }
2145 
2146 static void duplicate_memory_bitmap(struct memory_bitmap *dst,
2147 				    struct memory_bitmap *src)
2148 {
2149 	unsigned long pfn;
2150 
2151 	memory_bm_position_reset(src);
2152 	pfn = memory_bm_next_pfn(src);
2153 	while (pfn != BM_END_OF_MAP) {
2154 		memory_bm_set_bit(dst, pfn);
2155 		pfn = memory_bm_next_pfn(src);
2156 	}
2157 }
2158 
2159 /**
2160  * mark_unsafe_pages - Mark pages that were used before hibernation.
2161  *
2162  * Mark the pages that cannot be used for storing the image during restoration,
2163  * because they conflict with the pages that had been used before hibernation.
2164  */
2165 static void mark_unsafe_pages(struct memory_bitmap *bm)
2166 {
2167 	unsigned long pfn;
2168 
2169 	/* Clear the "free"/"unsafe" bit for all PFNs */
2170 	memory_bm_position_reset(free_pages_map);
2171 	pfn = memory_bm_next_pfn(free_pages_map);
2172 	while (pfn != BM_END_OF_MAP) {
2173 		memory_bm_clear_current(free_pages_map);
2174 		pfn = memory_bm_next_pfn(free_pages_map);
2175 	}
2176 
2177 	/* Mark pages that correspond to the "original" PFNs as "unsafe" */
2178 	duplicate_memory_bitmap(free_pages_map, bm);
2179 
2180 	allocated_unsafe_pages = 0;
2181 }
2182 
2183 static int check_header(struct swsusp_info *info)
2184 {
2185 	char *reason;
2186 
2187 	reason = check_image_kernel(info);
2188 	if (!reason && info->num_physpages != get_num_physpages())
2189 		reason = "memory size";
2190 	if (reason) {
2191 		pr_err("Image mismatch: %s\n", reason);
2192 		return -EPERM;
2193 	}
2194 	return 0;
2195 }
2196 
2197 /**
2198  * load header - Check the image header and copy the data from it.
2199  */
2200 static int load_header(struct swsusp_info *info)
2201 {
2202 	int error;
2203 
2204 	restore_pblist = NULL;
2205 	error = check_header(info);
2206 	if (!error) {
2207 		nr_copy_pages = info->image_pages;
2208 		nr_meta_pages = info->pages - info->image_pages - 1;
2209 	}
2210 	return error;
2211 }
2212 
2213 /**
2214  * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap.
2215  * @bm: Memory bitmap.
2216  * @buf: Area of memory containing the PFNs.
2217  *
2218  * For each element of the array pointed to by @buf (1 page at a time), set the
2219  * corresponding bit in @bm.
2220  */
2221 static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
2222 {
2223 	int j;
2224 
2225 	for (j = 0; j < PAGE_SIZE / sizeof(long); j++) {
2226 		if (unlikely(buf[j] == BM_END_OF_MAP))
2227 			break;
2228 
2229 		/* Extract and buffer page key for data page (s390 only). */
2230 		page_key_memorize(buf + j);
2231 
2232 		if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j]))
2233 			memory_bm_set_bit(bm, buf[j]);
2234 		else
2235 			return -EFAULT;
2236 	}
2237 
2238 	return 0;
2239 }
2240 
2241 #ifdef CONFIG_HIGHMEM
2242 /*
2243  * struct highmem_pbe is used for creating the list of highmem pages that
2244  * should be restored atomically during the resume from disk, because the page
2245  * frames they have occupied before the suspend are in use.
2246  */
2247 struct highmem_pbe {
2248 	struct page *copy_page;	/* data is here now */
2249 	struct page *orig_page;	/* data was here before the suspend */
2250 	struct highmem_pbe *next;
2251 };
2252 
2253 /*
2254  * List of highmem PBEs needed for restoring the highmem pages that were
2255  * allocated before the suspend and included in the suspend image, but have
2256  * also been allocated by the "resume" kernel, so their contents cannot be
2257  * written directly to their "original" page frames.
2258  */
2259 static struct highmem_pbe *highmem_pblist;
2260 
2261 /**
2262  * count_highmem_image_pages - Compute the number of highmem pages in the image.
2263  * @bm: Memory bitmap.
2264  *
2265  * The bits in @bm that correspond to image pages are assumed to be set.
2266  */
2267 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm)
2268 {
2269 	unsigned long pfn;
2270 	unsigned int cnt = 0;
2271 
2272 	memory_bm_position_reset(bm);
2273 	pfn = memory_bm_next_pfn(bm);
2274 	while (pfn != BM_END_OF_MAP) {
2275 		if (PageHighMem(pfn_to_page(pfn)))
2276 			cnt++;
2277 
2278 		pfn = memory_bm_next_pfn(bm);
2279 	}
2280 	return cnt;
2281 }
2282 
2283 static unsigned int safe_highmem_pages;
2284 
2285 static struct memory_bitmap *safe_highmem_bm;
2286 
2287 /**
2288  * prepare_highmem_image - Allocate memory for loading highmem data from image.
2289  * @bm: Pointer to an uninitialized memory bitmap structure.
2290  * @nr_highmem_p: Pointer to the number of highmem image pages.
2291  *
2292  * Try to allocate as many highmem pages as there are highmem image pages
2293  * (@nr_highmem_p points to the variable containing the number of highmem image
2294  * pages).  The pages that are "safe" (ie. will not be overwritten when the
2295  * hibernation image is restored entirely) have the corresponding bits set in
2296  * @bm (it must be unitialized).
2297  *
2298  * NOTE: This function should not be called if there are no highmem image pages.
2299  */
2300 static int prepare_highmem_image(struct memory_bitmap *bm,
2301 				 unsigned int *nr_highmem_p)
2302 {
2303 	unsigned int to_alloc;
2304 
2305 	if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE))
2306 		return -ENOMEM;
2307 
2308 	if (get_highmem_buffer(PG_SAFE))
2309 		return -ENOMEM;
2310 
2311 	to_alloc = count_free_highmem_pages();
2312 	if (to_alloc > *nr_highmem_p)
2313 		to_alloc = *nr_highmem_p;
2314 	else
2315 		*nr_highmem_p = to_alloc;
2316 
2317 	safe_highmem_pages = 0;
2318 	while (to_alloc-- > 0) {
2319 		struct page *page;
2320 
2321 		page = alloc_page(__GFP_HIGHMEM);
2322 		if (!swsusp_page_is_free(page)) {
2323 			/* The page is "safe", set its bit the bitmap */
2324 			memory_bm_set_bit(bm, page_to_pfn(page));
2325 			safe_highmem_pages++;
2326 		}
2327 		/* Mark the page as allocated */
2328 		swsusp_set_page_forbidden(page);
2329 		swsusp_set_page_free(page);
2330 	}
2331 	memory_bm_position_reset(bm);
2332 	safe_highmem_bm = bm;
2333 	return 0;
2334 }
2335 
2336 static struct page *last_highmem_page;
2337 
2338 /**
2339  * get_highmem_page_buffer - Prepare a buffer to store a highmem image page.
2340  *
2341  * For a given highmem image page get a buffer that suspend_write_next() should
2342  * return to its caller to write to.
2343  *
2344  * If the page is to be saved to its "original" page frame or a copy of
2345  * the page is to be made in the highmem, @buffer is returned.  Otherwise,
2346  * the copy of the page is to be made in normal memory, so the address of
2347  * the copy is returned.
2348  *
2349  * If @buffer is returned, the caller of suspend_write_next() will write
2350  * the page's contents to @buffer, so they will have to be copied to the
2351  * right location on the next call to suspend_write_next() and it is done
2352  * with the help of copy_last_highmem_page().  For this purpose, if
2353  * @buffer is returned, @last_highmem_page is set to the page to which
2354  * the data will have to be copied from @buffer.
2355  */
2356 static void *get_highmem_page_buffer(struct page *page,
2357 				     struct chain_allocator *ca)
2358 {
2359 	struct highmem_pbe *pbe;
2360 	void *kaddr;
2361 
2362 	if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) {
2363 		/*
2364 		 * We have allocated the "original" page frame and we can
2365 		 * use it directly to store the loaded page.
2366 		 */
2367 		last_highmem_page = page;
2368 		return buffer;
2369 	}
2370 	/*
2371 	 * The "original" page frame has not been allocated and we have to
2372 	 * use a "safe" page frame to store the loaded page.
2373 	 */
2374 	pbe = chain_alloc(ca, sizeof(struct highmem_pbe));
2375 	if (!pbe) {
2376 		swsusp_free();
2377 		return ERR_PTR(-ENOMEM);
2378 	}
2379 	pbe->orig_page = page;
2380 	if (safe_highmem_pages > 0) {
2381 		struct page *tmp;
2382 
2383 		/* Copy of the page will be stored in high memory */
2384 		kaddr = buffer;
2385 		tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm));
2386 		safe_highmem_pages--;
2387 		last_highmem_page = tmp;
2388 		pbe->copy_page = tmp;
2389 	} else {
2390 		/* Copy of the page will be stored in normal memory */
2391 		kaddr = safe_pages_list;
2392 		safe_pages_list = safe_pages_list->next;
2393 		pbe->copy_page = virt_to_page(kaddr);
2394 	}
2395 	pbe->next = highmem_pblist;
2396 	highmem_pblist = pbe;
2397 	return kaddr;
2398 }
2399 
2400 /**
2401  * copy_last_highmem_page - Copy most the most recent highmem image page.
2402  *
2403  * Copy the contents of a highmem image from @buffer, where the caller of
2404  * snapshot_write_next() has stored them, to the right location represented by
2405  * @last_highmem_page .
2406  */
2407 static void copy_last_highmem_page(void)
2408 {
2409 	if (last_highmem_page) {
2410 		void *dst;
2411 
2412 		dst = kmap_atomic(last_highmem_page);
2413 		copy_page(dst, buffer);
2414 		kunmap_atomic(dst);
2415 		last_highmem_page = NULL;
2416 	}
2417 }
2418 
2419 static inline int last_highmem_page_copied(void)
2420 {
2421 	return !last_highmem_page;
2422 }
2423 
2424 static inline void free_highmem_data(void)
2425 {
2426 	if (safe_highmem_bm)
2427 		memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR);
2428 
2429 	if (buffer)
2430 		free_image_page(buffer, PG_UNSAFE_CLEAR);
2431 }
2432 #else
2433 static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; }
2434 
2435 static inline int prepare_highmem_image(struct memory_bitmap *bm,
2436 					unsigned int *nr_highmem_p) { return 0; }
2437 
2438 static inline void *get_highmem_page_buffer(struct page *page,
2439 					    struct chain_allocator *ca)
2440 {
2441 	return ERR_PTR(-EINVAL);
2442 }
2443 
2444 static inline void copy_last_highmem_page(void) {}
2445 static inline int last_highmem_page_copied(void) { return 1; }
2446 static inline void free_highmem_data(void) {}
2447 #endif /* CONFIG_HIGHMEM */
2448 
2449 #define PBES_PER_LINKED_PAGE	(LINKED_PAGE_DATA_SIZE / sizeof(struct pbe))
2450 
2451 /**
2452  * prepare_image - Make room for loading hibernation image.
2453  * @new_bm: Unitialized memory bitmap structure.
2454  * @bm: Memory bitmap with unsafe pages marked.
2455  *
2456  * Use @bm to mark the pages that will be overwritten in the process of
2457  * restoring the system memory state from the suspend image ("unsafe" pages)
2458  * and allocate memory for the image.
2459  *
2460  * The idea is to allocate a new memory bitmap first and then allocate
2461  * as many pages as needed for image data, but without specifying what those
2462  * pages will be used for just yet.  Instead, we mark them all as allocated and
2463  * create a lists of "safe" pages to be used later.  On systems with high
2464  * memory a list of "safe" highmem pages is created too.
2465  */
2466 static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm)
2467 {
2468 	unsigned int nr_pages, nr_highmem;
2469 	struct linked_page *lp;
2470 	int error;
2471 
2472 	/* If there is no highmem, the buffer will not be necessary */
2473 	free_image_page(buffer, PG_UNSAFE_CLEAR);
2474 	buffer = NULL;
2475 
2476 	nr_highmem = count_highmem_image_pages(bm);
2477 	mark_unsafe_pages(bm);
2478 
2479 	error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE);
2480 	if (error)
2481 		goto Free;
2482 
2483 	duplicate_memory_bitmap(new_bm, bm);
2484 	memory_bm_free(bm, PG_UNSAFE_KEEP);
2485 	if (nr_highmem > 0) {
2486 		error = prepare_highmem_image(bm, &nr_highmem);
2487 		if (error)
2488 			goto Free;
2489 	}
2490 	/*
2491 	 * Reserve some safe pages for potential later use.
2492 	 *
2493 	 * NOTE: This way we make sure there will be enough safe pages for the
2494 	 * chain_alloc() in get_buffer().  It is a bit wasteful, but
2495 	 * nr_copy_pages cannot be greater than 50% of the memory anyway.
2496 	 *
2497 	 * nr_copy_pages cannot be less than allocated_unsafe_pages too.
2498 	 */
2499 	nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2500 	nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE);
2501 	while (nr_pages > 0) {
2502 		lp = get_image_page(GFP_ATOMIC, PG_SAFE);
2503 		if (!lp) {
2504 			error = -ENOMEM;
2505 			goto Free;
2506 		}
2507 		lp->next = safe_pages_list;
2508 		safe_pages_list = lp;
2509 		nr_pages--;
2510 	}
2511 	/* Preallocate memory for the image */
2512 	nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages;
2513 	while (nr_pages > 0) {
2514 		lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC);
2515 		if (!lp) {
2516 			error = -ENOMEM;
2517 			goto Free;
2518 		}
2519 		if (!swsusp_page_is_free(virt_to_page(lp))) {
2520 			/* The page is "safe", add it to the list */
2521 			lp->next = safe_pages_list;
2522 			safe_pages_list = lp;
2523 		}
2524 		/* Mark the page as allocated */
2525 		swsusp_set_page_forbidden(virt_to_page(lp));
2526 		swsusp_set_page_free(virt_to_page(lp));
2527 		nr_pages--;
2528 	}
2529 	return 0;
2530 
2531  Free:
2532 	swsusp_free();
2533 	return error;
2534 }
2535 
2536 /**
2537  * get_buffer - Get the address to store the next image data page.
2538  *
2539  * Get the address that snapshot_write_next() should return to its caller to
2540  * write to.
2541  */
2542 static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca)
2543 {
2544 	struct pbe *pbe;
2545 	struct page *page;
2546 	unsigned long pfn = memory_bm_next_pfn(bm);
2547 
2548 	if (pfn == BM_END_OF_MAP)
2549 		return ERR_PTR(-EFAULT);
2550 
2551 	page = pfn_to_page(pfn);
2552 	if (PageHighMem(page))
2553 		return get_highmem_page_buffer(page, ca);
2554 
2555 	if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page))
2556 		/*
2557 		 * We have allocated the "original" page frame and we can
2558 		 * use it directly to store the loaded page.
2559 		 */
2560 		return page_address(page);
2561 
2562 	/*
2563 	 * The "original" page frame has not been allocated and we have to
2564 	 * use a "safe" page frame to store the loaded page.
2565 	 */
2566 	pbe = chain_alloc(ca, sizeof(struct pbe));
2567 	if (!pbe) {
2568 		swsusp_free();
2569 		return ERR_PTR(-ENOMEM);
2570 	}
2571 	pbe->orig_address = page_address(page);
2572 	pbe->address = safe_pages_list;
2573 	safe_pages_list = safe_pages_list->next;
2574 	pbe->next = restore_pblist;
2575 	restore_pblist = pbe;
2576 	return pbe->address;
2577 }
2578 
2579 /**
2580  * snapshot_write_next - Get the address to store the next image page.
2581  * @handle: Snapshot handle structure to guide the writing.
2582  *
2583  * On the first call, @handle should point to a zeroed snapshot_handle
2584  * structure.  The structure gets populated then and a pointer to it should be
2585  * passed to this function every next time.
2586  *
2587  * On success, the function returns a positive number.  Then, the caller
2588  * is allowed to write up to the returned number of bytes to the memory
2589  * location computed by the data_of() macro.
2590  *
2591  * The function returns 0 to indicate the "end of file" condition.  Negative
2592  * numbers are returned on errors, in which cases the structure pointed to by
2593  * @handle is not updated and should not be used any more.
2594  */
2595 int snapshot_write_next(struct snapshot_handle *handle)
2596 {
2597 	static struct chain_allocator ca;
2598 	int error = 0;
2599 
2600 	/* Check if we have already loaded the entire image */
2601 	if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages)
2602 		return 0;
2603 
2604 	handle->sync_read = 1;
2605 
2606 	if (!handle->cur) {
2607 		if (!buffer)
2608 			/* This makes the buffer be freed by swsusp_free() */
2609 			buffer = get_image_page(GFP_ATOMIC, PG_ANY);
2610 
2611 		if (!buffer)
2612 			return -ENOMEM;
2613 
2614 		handle->buffer = buffer;
2615 	} else if (handle->cur == 1) {
2616 		error = load_header(buffer);
2617 		if (error)
2618 			return error;
2619 
2620 		safe_pages_list = NULL;
2621 
2622 		error = memory_bm_create(&copy_bm, GFP_ATOMIC, PG_ANY);
2623 		if (error)
2624 			return error;
2625 
2626 		/* Allocate buffer for page keys. */
2627 		error = page_key_alloc(nr_copy_pages);
2628 		if (error)
2629 			return error;
2630 
2631 		hibernate_restore_protection_begin();
2632 	} else if (handle->cur <= nr_meta_pages + 1) {
2633 		error = unpack_orig_pfns(buffer, &copy_bm);
2634 		if (error)
2635 			return error;
2636 
2637 		if (handle->cur == nr_meta_pages + 1) {
2638 			error = prepare_image(&orig_bm, &copy_bm);
2639 			if (error)
2640 				return error;
2641 
2642 			chain_init(&ca, GFP_ATOMIC, PG_SAFE);
2643 			memory_bm_position_reset(&orig_bm);
2644 			restore_pblist = NULL;
2645 			handle->buffer = get_buffer(&orig_bm, &ca);
2646 			handle->sync_read = 0;
2647 			if (IS_ERR(handle->buffer))
2648 				return PTR_ERR(handle->buffer);
2649 		}
2650 	} else {
2651 		copy_last_highmem_page();
2652 		/* Restore page key for data page (s390 only). */
2653 		page_key_write(handle->buffer);
2654 		hibernate_restore_protect_page(handle->buffer);
2655 		handle->buffer = get_buffer(&orig_bm, &ca);
2656 		if (IS_ERR(handle->buffer))
2657 			return PTR_ERR(handle->buffer);
2658 		if (handle->buffer != buffer)
2659 			handle->sync_read = 0;
2660 	}
2661 	handle->cur++;
2662 	return PAGE_SIZE;
2663 }
2664 
2665 /**
2666  * snapshot_write_finalize - Complete the loading of a hibernation image.
2667  *
2668  * Must be called after the last call to snapshot_write_next() in case the last
2669  * page in the image happens to be a highmem page and its contents should be
2670  * stored in highmem.  Additionally, it recycles bitmap memory that's not
2671  * necessary any more.
2672  */
2673 void snapshot_write_finalize(struct snapshot_handle *handle)
2674 {
2675 	copy_last_highmem_page();
2676 	/* Restore page key for data page (s390 only). */
2677 	page_key_write(handle->buffer);
2678 	page_key_free();
2679 	hibernate_restore_protect_page(handle->buffer);
2680 	/* Do that only if we have loaded the image entirely */
2681 	if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) {
2682 		memory_bm_recycle(&orig_bm);
2683 		free_highmem_data();
2684 	}
2685 }
2686 
2687 int snapshot_image_loaded(struct snapshot_handle *handle)
2688 {
2689 	return !(!nr_copy_pages || !last_highmem_page_copied() ||
2690 			handle->cur <= nr_meta_pages + nr_copy_pages);
2691 }
2692 
2693 #ifdef CONFIG_HIGHMEM
2694 /* Assumes that @buf is ready and points to a "safe" page */
2695 static inline void swap_two_pages_data(struct page *p1, struct page *p2,
2696 				       void *buf)
2697 {
2698 	void *kaddr1, *kaddr2;
2699 
2700 	kaddr1 = kmap_atomic(p1);
2701 	kaddr2 = kmap_atomic(p2);
2702 	copy_page(buf, kaddr1);
2703 	copy_page(kaddr1, kaddr2);
2704 	copy_page(kaddr2, buf);
2705 	kunmap_atomic(kaddr2);
2706 	kunmap_atomic(kaddr1);
2707 }
2708 
2709 /**
2710  * restore_highmem - Put highmem image pages into their original locations.
2711  *
2712  * For each highmem page that was in use before hibernation and is included in
2713  * the image, and also has been allocated by the "restore" kernel, swap its
2714  * current contents with the previous (ie. "before hibernation") ones.
2715  *
2716  * If the restore eventually fails, we can call this function once again and
2717  * restore the highmem state as seen by the restore kernel.
2718  */
2719 int restore_highmem(void)
2720 {
2721 	struct highmem_pbe *pbe = highmem_pblist;
2722 	void *buf;
2723 
2724 	if (!pbe)
2725 		return 0;
2726 
2727 	buf = get_image_page(GFP_ATOMIC, PG_SAFE);
2728 	if (!buf)
2729 		return -ENOMEM;
2730 
2731 	while (pbe) {
2732 		swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf);
2733 		pbe = pbe->next;
2734 	}
2735 	free_image_page(buf, PG_UNSAFE_CLEAR);
2736 	return 0;
2737 }
2738 #endif /* CONFIG_HIGHMEM */
2739