Commit | Line | Data |
---|---|---|
25761b6e | 1 | /* |
96bc7aec | 2 | * linux/kernel/power/snapshot.c |
25761b6e | 3 | * |
8357376d | 4 | * This file provides system snapshot/restore functionality for swsusp. |
25761b6e | 5 | * |
a2531293 | 6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz> |
8357376d | 7 | * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl> |
25761b6e | 8 | * |
8357376d | 9 | * This file is released under the GPLv2. |
25761b6e RW |
10 | * |
11 | */ | |
12 | ||
f577eb30 | 13 | #include <linux/version.h> |
25761b6e RW |
14 | #include <linux/module.h> |
15 | #include <linux/mm.h> | |
16 | #include <linux/suspend.h> | |
25761b6e | 17 | #include <linux/delay.h> |
25761b6e | 18 | #include <linux/bitops.h> |
25761b6e | 19 | #include <linux/spinlock.h> |
25761b6e | 20 | #include <linux/kernel.h> |
25761b6e RW |
21 | #include <linux/pm.h> |
22 | #include <linux/device.h> | |
74dfd666 | 23 | #include <linux/init.h> |
25761b6e RW |
24 | #include <linux/bootmem.h> |
25 | #include <linux/syscalls.h> | |
26 | #include <linux/console.h> | |
27 | #include <linux/highmem.h> | |
846705de | 28 | #include <linux/list.h> |
5a0e3ad6 | 29 | #include <linux/slab.h> |
25761b6e RW |
30 | |
31 | #include <asm/uaccess.h> | |
32 | #include <asm/mmu_context.h> | |
33 | #include <asm/pgtable.h> | |
34 | #include <asm/tlbflush.h> | |
35 | #include <asm/io.h> | |
36 | ||
25761b6e RW |
37 | #include "power.h" |
38 | ||
74dfd666 RW |
39 | static int swsusp_page_is_free(struct page *); |
40 | static void swsusp_set_page_forbidden(struct page *); | |
41 | static void swsusp_unset_page_forbidden(struct page *); | |
42 | ||
ddeb6487 RW |
43 | /* |
44 | * Number of bytes to reserve for memory allocations made by device drivers | |
45 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
46 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
47 | */ | |
48 | unsigned long reserved_size; | |
49 | ||
50 | void __init hibernate_reserved_size_init(void) | |
51 | { | |
52 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
53 | } | |
54 | ||
fe419535 RW |
55 | /* |
56 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
57 | * When it is set to N, swsusp will do its best to ensure the image |
58 | * size will not exceed N bytes, but if that is impossible, it will | |
59 | * try to create the smallest image possible. | |
fe419535 | 60 | */ |
ac5c24ec RW |
61 | unsigned long image_size; |
62 | ||
63 | void __init hibernate_image_size_init(void) | |
64 | { | |
1c1be3a9 | 65 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 66 | } |
fe419535 | 67 | |
8357376d RW |
68 | /* List of PBEs needed for restoring the pages that were allocated before |
69 | * the suspend and included in the suspend image, but have also been | |
70 | * allocated by the "resume" kernel, so their contents cannot be written | |
71 | * directly to their "original" page frames. | |
72 | */ | |
75534b50 RW |
73 | struct pbe *restore_pblist; |
74 | ||
8357376d | 75 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 76 | static void *buffer; |
7088a5c0 | 77 | |
f6143aa6 RW |
78 | /** |
79 | * @safe_needed - on resume, for storing the PBE list and the image, | |
80 | * we can only use memory pages that do not conflict with the pages | |
8357376d RW |
81 | * used before suspend. The unsafe pages have PageNosaveFree set |
82 | * and we count them using unsafe_pages. | |
f6143aa6 | 83 | * |
8357376d RW |
84 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
85 | * so that swsusp_free() can release it. | |
f6143aa6 RW |
86 | */ |
87 | ||
0bcd888d RW |
88 | #define PG_ANY 0 |
89 | #define PG_SAFE 1 | |
90 | #define PG_UNSAFE_CLEAR 1 | |
91 | #define PG_UNSAFE_KEEP 0 | |
92 | ||
940864dd | 93 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 94 | |
8357376d | 95 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
96 | { |
97 | void *res; | |
98 | ||
99 | res = (void *)get_zeroed_page(gfp_mask); | |
100 | if (safe_needed) | |
7be98234 | 101 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 102 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 103 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 104 | allocated_unsafe_pages++; |
f6143aa6 RW |
105 | res = (void *)get_zeroed_page(gfp_mask); |
106 | } | |
107 | if (res) { | |
7be98234 RW |
108 | swsusp_set_page_forbidden(virt_to_page(res)); |
109 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
110 | } |
111 | return res; | |
112 | } | |
113 | ||
114 | unsigned long get_safe_page(gfp_t gfp_mask) | |
115 | { | |
8357376d RW |
116 | return (unsigned long)get_image_page(gfp_mask, PG_SAFE); |
117 | } | |
118 | ||
5b6d15de RW |
119 | static struct page *alloc_image_page(gfp_t gfp_mask) |
120 | { | |
8357376d RW |
121 | struct page *page; |
122 | ||
123 | page = alloc_page(gfp_mask); | |
124 | if (page) { | |
7be98234 RW |
125 | swsusp_set_page_forbidden(page); |
126 | swsusp_set_page_free(page); | |
8357376d RW |
127 | } |
128 | return page; | |
f6143aa6 RW |
129 | } |
130 | ||
131 | /** | |
132 | * free_image_page - free page represented by @addr, allocated with | |
8357376d | 133 | * get_image_page (page flags set by it must be cleared) |
f6143aa6 RW |
134 | */ |
135 | ||
136 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
137 | { | |
8357376d RW |
138 | struct page *page; |
139 | ||
140 | BUG_ON(!virt_addr_valid(addr)); | |
141 | ||
142 | page = virt_to_page(addr); | |
143 | ||
7be98234 | 144 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 145 | if (clear_nosave_free) |
7be98234 | 146 | swsusp_unset_page_free(page); |
8357376d RW |
147 | |
148 | __free_page(page); | |
f6143aa6 RW |
149 | } |
150 | ||
b788db79 RW |
151 | /* struct linked_page is used to build chains of pages */ |
152 | ||
153 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
154 | ||
155 | struct linked_page { | |
156 | struct linked_page *next; | |
157 | char data[LINKED_PAGE_DATA_SIZE]; | |
158 | } __attribute__((packed)); | |
159 | ||
160 | static inline void | |
161 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) | |
162 | { | |
163 | while (list) { | |
164 | struct linked_page *lp = list->next; | |
165 | ||
166 | free_image_page(list, clear_page_nosave); | |
167 | list = lp; | |
168 | } | |
169 | } | |
170 | ||
171 | /** | |
172 | * struct chain_allocator is used for allocating small objects out of | |
173 | * a linked list of pages called 'the chain'. | |
174 | * | |
175 | * The chain grows each time when there is no room for a new object in | |
176 | * the current page. The allocated objects cannot be freed individually. | |
177 | * It is only possible to free them all at once, by freeing the entire | |
178 | * chain. | |
179 | * | |
180 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
181 | * are not much smaller than PAGE_SIZE. | |
182 | */ | |
183 | ||
184 | struct chain_allocator { | |
185 | struct linked_page *chain; /* the chain */ | |
186 | unsigned int used_space; /* total size of objects allocated out | |
187 | * of the current page | |
188 | */ | |
189 | gfp_t gfp_mask; /* mask for allocating pages */ | |
190 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
191 | }; | |
192 | ||
193 | static void | |
194 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) | |
195 | { | |
196 | ca->chain = NULL; | |
197 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
198 | ca->gfp_mask = gfp_mask; | |
199 | ca->safe_needed = safe_needed; | |
200 | } | |
201 | ||
202 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
203 | { | |
204 | void *ret; | |
205 | ||
206 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
207 | struct linked_page *lp; | |
208 | ||
8357376d | 209 | lp = get_image_page(ca->gfp_mask, ca->safe_needed); |
b788db79 RW |
210 | if (!lp) |
211 | return NULL; | |
212 | ||
213 | lp->next = ca->chain; | |
214 | ca->chain = lp; | |
215 | ca->used_space = 0; | |
216 | } | |
217 | ret = ca->chain->data + ca->used_space; | |
218 | ca->used_space += size; | |
219 | return ret; | |
220 | } | |
221 | ||
b788db79 RW |
222 | /** |
223 | * Data types related to memory bitmaps. | |
224 | * | |
225 | * Memory bitmap is a structure consiting of many linked lists of | |
226 | * objects. The main list's elements are of type struct zone_bitmap | |
227 | * and each of them corresonds to one zone. For each zone bitmap | |
228 | * object there is a list of objects of type struct bm_block that | |
0d83304c | 229 | * represent each blocks of bitmap in which information is stored. |
b788db79 RW |
230 | * |
231 | * struct memory_bitmap contains a pointer to the main list of zone | |
232 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
233 | * and a pointer to the list of pages used for allocating all of the | |
234 | * zone bitmap objects and bitmap block objects. | |
235 | * | |
236 | * NOTE: It has to be possible to lay out the bitmap in memory | |
237 | * using only allocations of order 0. Additionally, the bitmap is | |
238 | * designed to work with arbitrary number of zones (this is over the | |
239 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
240 | * | |
241 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
242 | * objects and a pointer to the bitmap block object that has been | |
243 | * most recently used for setting bits. Additionally, it contains the | |
244 | * pfns that correspond to the start and end of the represented zone. | |
245 | * | |
246 | * struct bm_block contains a pointer to the memory page in which | |
0d83304c AM |
247 | * information is stored (in the form of a block of bitmap) |
248 | * It also contains the pfns that correspond to the start and end of | |
249 | * the represented memory area. | |
b788db79 RW |
250 | */ |
251 | ||
252 | #define BM_END_OF_MAP (~0UL) | |
253 | ||
8de03073 | 254 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
b788db79 RW |
255 | |
256 | struct bm_block { | |
846705de | 257 | struct list_head hook; /* hook into a list of bitmap blocks */ |
b788db79 RW |
258 | unsigned long start_pfn; /* pfn represented by the first bit */ |
259 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ | |
0d83304c | 260 | unsigned long *data; /* bitmap representing pages */ |
b788db79 RW |
261 | }; |
262 | ||
0d83304c AM |
263 | static inline unsigned long bm_block_bits(struct bm_block *bb) |
264 | { | |
265 | return bb->end_pfn - bb->start_pfn; | |
266 | } | |
267 | ||
b788db79 RW |
268 | /* strcut bm_position is used for browsing memory bitmaps */ |
269 | ||
270 | struct bm_position { | |
b788db79 | 271 | struct bm_block *block; |
b788db79 RW |
272 | int bit; |
273 | }; | |
274 | ||
275 | struct memory_bitmap { | |
846705de | 276 | struct list_head blocks; /* list of bitmap blocks */ |
b788db79 RW |
277 | struct linked_page *p_list; /* list of pages used to store zone |
278 | * bitmap objects and bitmap block | |
279 | * objects | |
280 | */ | |
281 | struct bm_position cur; /* most recently used bit position */ | |
282 | }; | |
283 | ||
284 | /* Functions that operate on memory bitmaps */ | |
285 | ||
b788db79 RW |
286 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
287 | { | |
846705de | 288 | bm->cur.block = list_entry(bm->blocks.next, struct bm_block, hook); |
0d83304c | 289 | bm->cur.bit = 0; |
b788db79 RW |
290 | } |
291 | ||
292 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
293 | ||
294 | /** | |
295 | * create_bm_block_list - create a list of block bitmap objects | |
8de03073 | 296 | * @pages - number of pages to track |
846705de RW |
297 | * @list - list to put the allocated blocks into |
298 | * @ca - chain allocator to be used for allocating memory | |
b788db79 | 299 | */ |
846705de RW |
300 | static int create_bm_block_list(unsigned long pages, |
301 | struct list_head *list, | |
302 | struct chain_allocator *ca) | |
b788db79 | 303 | { |
846705de | 304 | unsigned int nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); |
b788db79 RW |
305 | |
306 | while (nr_blocks-- > 0) { | |
307 | struct bm_block *bb; | |
308 | ||
309 | bb = chain_alloc(ca, sizeof(struct bm_block)); | |
310 | if (!bb) | |
846705de RW |
311 | return -ENOMEM; |
312 | list_add(&bb->hook, list); | |
b788db79 | 313 | } |
846705de RW |
314 | |
315 | return 0; | |
b788db79 RW |
316 | } |
317 | ||
846705de RW |
318 | struct mem_extent { |
319 | struct list_head hook; | |
320 | unsigned long start; | |
321 | unsigned long end; | |
322 | }; | |
323 | ||
b788db79 | 324 | /** |
846705de RW |
325 | * free_mem_extents - free a list of memory extents |
326 | * @list - list of extents to empty | |
b788db79 | 327 | */ |
846705de RW |
328 | static void free_mem_extents(struct list_head *list) |
329 | { | |
330 | struct mem_extent *ext, *aux; | |
b788db79 | 331 | |
846705de RW |
332 | list_for_each_entry_safe(ext, aux, list, hook) { |
333 | list_del(&ext->hook); | |
334 | kfree(ext); | |
335 | } | |
336 | } | |
337 | ||
338 | /** | |
339 | * create_mem_extents - create a list of memory extents representing | |
340 | * contiguous ranges of PFNs | |
341 | * @list - list to put the extents into | |
342 | * @gfp_mask - mask to use for memory allocations | |
343 | */ | |
344 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 345 | { |
846705de | 346 | struct zone *zone; |
b788db79 | 347 | |
846705de | 348 | INIT_LIST_HEAD(list); |
b788db79 | 349 | |
ee99c71c | 350 | for_each_populated_zone(zone) { |
846705de RW |
351 | unsigned long zone_start, zone_end; |
352 | struct mem_extent *ext, *cur, *aux; | |
353 | ||
846705de RW |
354 | zone_start = zone->zone_start_pfn; |
355 | zone_end = zone->zone_start_pfn + zone->spanned_pages; | |
356 | ||
357 | list_for_each_entry(ext, list, hook) | |
358 | if (zone_start <= ext->end) | |
359 | break; | |
b788db79 | 360 | |
846705de RW |
361 | if (&ext->hook == list || zone_end < ext->start) { |
362 | /* New extent is necessary */ | |
363 | struct mem_extent *new_ext; | |
364 | ||
365 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
366 | if (!new_ext) { | |
367 | free_mem_extents(list); | |
368 | return -ENOMEM; | |
369 | } | |
370 | new_ext->start = zone_start; | |
371 | new_ext->end = zone_end; | |
372 | list_add_tail(&new_ext->hook, &ext->hook); | |
373 | continue; | |
374 | } | |
375 | ||
376 | /* Merge this zone's range of PFNs with the existing one */ | |
377 | if (zone_start < ext->start) | |
378 | ext->start = zone_start; | |
379 | if (zone_end > ext->end) | |
380 | ext->end = zone_end; | |
381 | ||
382 | /* More merging may be possible */ | |
383 | cur = ext; | |
384 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
385 | if (zone_end < cur->start) | |
386 | break; | |
387 | if (zone_end < cur->end) | |
388 | ext->end = cur->end; | |
389 | list_del(&cur->hook); | |
390 | kfree(cur); | |
391 | } | |
b788db79 | 392 | } |
846705de RW |
393 | |
394 | return 0; | |
b788db79 RW |
395 | } |
396 | ||
397 | /** | |
398 | * memory_bm_create - allocate memory for a memory bitmap | |
399 | */ | |
b788db79 RW |
400 | static int |
401 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) | |
402 | { | |
403 | struct chain_allocator ca; | |
846705de RW |
404 | struct list_head mem_extents; |
405 | struct mem_extent *ext; | |
406 | int error; | |
b788db79 RW |
407 | |
408 | chain_init(&ca, gfp_mask, safe_needed); | |
846705de | 409 | INIT_LIST_HEAD(&bm->blocks); |
b788db79 | 410 | |
846705de RW |
411 | error = create_mem_extents(&mem_extents, gfp_mask); |
412 | if (error) | |
413 | return error; | |
b788db79 | 414 | |
846705de RW |
415 | list_for_each_entry(ext, &mem_extents, hook) { |
416 | struct bm_block *bb; | |
417 | unsigned long pfn = ext->start; | |
418 | unsigned long pages = ext->end - ext->start; | |
b788db79 | 419 | |
846705de | 420 | bb = list_entry(bm->blocks.prev, struct bm_block, hook); |
b788db79 | 421 | |
846705de RW |
422 | error = create_bm_block_list(pages, bm->blocks.prev, &ca); |
423 | if (error) | |
424 | goto Error; | |
b788db79 | 425 | |
846705de RW |
426 | list_for_each_entry_continue(bb, &bm->blocks, hook) { |
427 | bb->data = get_image_page(gfp_mask, safe_needed); | |
428 | if (!bb->data) { | |
429 | error = -ENOMEM; | |
430 | goto Error; | |
431 | } | |
b788db79 RW |
432 | |
433 | bb->start_pfn = pfn; | |
846705de | 434 | if (pages >= BM_BITS_PER_BLOCK) { |
b788db79 | 435 | pfn += BM_BITS_PER_BLOCK; |
846705de | 436 | pages -= BM_BITS_PER_BLOCK; |
b788db79 RW |
437 | } else { |
438 | /* This is executed only once in the loop */ | |
846705de | 439 | pfn += pages; |
b788db79 RW |
440 | } |
441 | bb->end_pfn = pfn; | |
b788db79 | 442 | } |
b788db79 | 443 | } |
846705de | 444 | |
b788db79 RW |
445 | bm->p_list = ca.chain; |
446 | memory_bm_position_reset(bm); | |
846705de RW |
447 | Exit: |
448 | free_mem_extents(&mem_extents); | |
449 | return error; | |
b788db79 | 450 | |
846705de | 451 | Error: |
b788db79 RW |
452 | bm->p_list = ca.chain; |
453 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 454 | goto Exit; |
b788db79 RW |
455 | } |
456 | ||
457 | /** | |
458 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
459 | */ | |
b788db79 RW |
460 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
461 | { | |
846705de | 462 | struct bm_block *bb; |
b788db79 | 463 | |
846705de RW |
464 | list_for_each_entry(bb, &bm->blocks, hook) |
465 | if (bb->data) | |
466 | free_image_page(bb->data, clear_nosave_free); | |
b788db79 | 467 | |
b788db79 | 468 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de RW |
469 | |
470 | INIT_LIST_HEAD(&bm->blocks); | |
b788db79 RW |
471 | } |
472 | ||
473 | /** | |
74dfd666 | 474 | * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds |
b788db79 RW |
475 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member |
476 | * of @bm->cur_zone_bm are updated. | |
b788db79 | 477 | */ |
a82f7119 | 478 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
74dfd666 | 479 | void **addr, unsigned int *bit_nr) |
b788db79 | 480 | { |
b788db79 RW |
481 | struct bm_block *bb; |
482 | ||
846705de RW |
483 | /* |
484 | * Check if the pfn corresponds to the current bitmap block and find | |
485 | * the block where it fits if this is not the case. | |
486 | */ | |
487 | bb = bm->cur.block; | |
b788db79 | 488 | if (pfn < bb->start_pfn) |
846705de RW |
489 | list_for_each_entry_continue_reverse(bb, &bm->blocks, hook) |
490 | if (pfn >= bb->start_pfn) | |
491 | break; | |
b788db79 | 492 | |
846705de RW |
493 | if (pfn >= bb->end_pfn) |
494 | list_for_each_entry_continue(bb, &bm->blocks, hook) | |
495 | if (pfn >= bb->start_pfn && pfn < bb->end_pfn) | |
496 | break; | |
74dfd666 | 497 | |
846705de RW |
498 | if (&bb->hook == &bm->blocks) |
499 | return -EFAULT; | |
500 | ||
501 | /* The block has been found */ | |
502 | bm->cur.block = bb; | |
b788db79 | 503 | pfn -= bb->start_pfn; |
846705de | 504 | bm->cur.bit = pfn + 1; |
0d83304c AM |
505 | *bit_nr = pfn; |
506 | *addr = bb->data; | |
a82f7119 | 507 | return 0; |
74dfd666 RW |
508 | } |
509 | ||
510 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) | |
511 | { | |
512 | void *addr; | |
513 | unsigned int bit; | |
a82f7119 | 514 | int error; |
74dfd666 | 515 | |
a82f7119 RW |
516 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
517 | BUG_ON(error); | |
74dfd666 RW |
518 | set_bit(bit, addr); |
519 | } | |
520 | ||
a82f7119 RW |
521 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
522 | { | |
523 | void *addr; | |
524 | unsigned int bit; | |
525 | int error; | |
526 | ||
527 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
528 | if (!error) | |
529 | set_bit(bit, addr); | |
530 | return error; | |
531 | } | |
532 | ||
74dfd666 RW |
533 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
534 | { | |
535 | void *addr; | |
536 | unsigned int bit; | |
a82f7119 | 537 | int error; |
74dfd666 | 538 | |
a82f7119 RW |
539 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
540 | BUG_ON(error); | |
74dfd666 RW |
541 | clear_bit(bit, addr); |
542 | } | |
543 | ||
544 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) | |
545 | { | |
546 | void *addr; | |
547 | unsigned int bit; | |
a82f7119 | 548 | int error; |
74dfd666 | 549 | |
a82f7119 RW |
550 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
551 | BUG_ON(error); | |
74dfd666 | 552 | return test_bit(bit, addr); |
b788db79 RW |
553 | } |
554 | ||
69643279 RW |
555 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
556 | { | |
557 | void *addr; | |
558 | unsigned int bit; | |
559 | ||
560 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); | |
561 | } | |
562 | ||
b788db79 RW |
563 | /** |
564 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit | |
565 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is | |
566 | * returned. | |
567 | * | |
568 | * It is required to run memory_bm_position_reset() before the first call to | |
569 | * this function. | |
570 | */ | |
571 | ||
572 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) | |
573 | { | |
b788db79 | 574 | struct bm_block *bb; |
b788db79 RW |
575 | int bit; |
576 | ||
846705de | 577 | bb = bm->cur.block; |
b788db79 | 578 | do { |
846705de RW |
579 | bit = bm->cur.bit; |
580 | bit = find_next_bit(bb->data, bm_block_bits(bb), bit); | |
581 | if (bit < bm_block_bits(bb)) | |
582 | goto Return_pfn; | |
583 | ||
584 | bb = list_entry(bb->hook.next, struct bm_block, hook); | |
585 | bm->cur.block = bb; | |
586 | bm->cur.bit = 0; | |
587 | } while (&bb->hook != &bm->blocks); | |
588 | ||
b788db79 RW |
589 | memory_bm_position_reset(bm); |
590 | return BM_END_OF_MAP; | |
591 | ||
59a49335 | 592 | Return_pfn: |
0d83304c AM |
593 | bm->cur.bit = bit + 1; |
594 | return bb->start_pfn + bit; | |
b788db79 RW |
595 | } |
596 | ||
74dfd666 RW |
597 | /** |
598 | * This structure represents a range of page frames the contents of which | |
599 | * should not be saved during the suspend. | |
600 | */ | |
601 | ||
602 | struct nosave_region { | |
603 | struct list_head list; | |
604 | unsigned long start_pfn; | |
605 | unsigned long end_pfn; | |
606 | }; | |
607 | ||
608 | static LIST_HEAD(nosave_regions); | |
609 | ||
610 | /** | |
611 | * register_nosave_region - register a range of page frames the contents | |
612 | * of which should not be saved during the suspend (to be used in the early | |
613 | * initialization code) | |
614 | */ | |
615 | ||
616 | void __init | |
940d67f6 JB |
617 | __register_nosave_region(unsigned long start_pfn, unsigned long end_pfn, |
618 | int use_kmalloc) | |
74dfd666 RW |
619 | { |
620 | struct nosave_region *region; | |
621 | ||
622 | if (start_pfn >= end_pfn) | |
623 | return; | |
624 | ||
625 | if (!list_empty(&nosave_regions)) { | |
626 | /* Try to extend the previous region (they should be sorted) */ | |
627 | region = list_entry(nosave_regions.prev, | |
628 | struct nosave_region, list); | |
629 | if (region->end_pfn == start_pfn) { | |
630 | region->end_pfn = end_pfn; | |
631 | goto Report; | |
632 | } | |
633 | } | |
940d67f6 JB |
634 | if (use_kmalloc) { |
635 | /* during init, this shouldn't fail */ | |
636 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); | |
637 | BUG_ON(!region); | |
638 | } else | |
639 | /* This allocation cannot fail */ | |
3c1596ef | 640 | region = alloc_bootmem(sizeof(struct nosave_region)); |
74dfd666 RW |
641 | region->start_pfn = start_pfn; |
642 | region->end_pfn = end_pfn; | |
643 | list_add_tail(®ion->list, &nosave_regions); | |
644 | Report: | |
23976728 | 645 | printk(KERN_INFO "PM: Registered nosave memory: %016lx - %016lx\n", |
74dfd666 RW |
646 | start_pfn << PAGE_SHIFT, end_pfn << PAGE_SHIFT); |
647 | } | |
648 | ||
649 | /* | |
650 | * Set bits in this map correspond to the page frames the contents of which | |
651 | * should not be saved during the suspend. | |
652 | */ | |
653 | static struct memory_bitmap *forbidden_pages_map; | |
654 | ||
655 | /* Set bits in this map correspond to free page frames. */ | |
656 | static struct memory_bitmap *free_pages_map; | |
657 | ||
658 | /* | |
659 | * Each page frame allocated for creating the image is marked by setting the | |
660 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
661 | */ | |
662 | ||
663 | void swsusp_set_page_free(struct page *page) | |
664 | { | |
665 | if (free_pages_map) | |
666 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
667 | } | |
668 | ||
669 | static int swsusp_page_is_free(struct page *page) | |
670 | { | |
671 | return free_pages_map ? | |
672 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
673 | } | |
674 | ||
675 | void swsusp_unset_page_free(struct page *page) | |
676 | { | |
677 | if (free_pages_map) | |
678 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
679 | } | |
680 | ||
681 | static void swsusp_set_page_forbidden(struct page *page) | |
682 | { | |
683 | if (forbidden_pages_map) | |
684 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
685 | } | |
686 | ||
687 | int swsusp_page_is_forbidden(struct page *page) | |
688 | { | |
689 | return forbidden_pages_map ? | |
690 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
691 | } | |
692 | ||
693 | static void swsusp_unset_page_forbidden(struct page *page) | |
694 | { | |
695 | if (forbidden_pages_map) | |
696 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
697 | } | |
698 | ||
699 | /** | |
700 | * mark_nosave_pages - set bits corresponding to the page frames the | |
701 | * contents of which should not be saved in a given bitmap. | |
702 | */ | |
703 | ||
704 | static void mark_nosave_pages(struct memory_bitmap *bm) | |
705 | { | |
706 | struct nosave_region *region; | |
707 | ||
708 | if (list_empty(&nosave_regions)) | |
709 | return; | |
710 | ||
711 | list_for_each_entry(region, &nosave_regions, list) { | |
712 | unsigned long pfn; | |
713 | ||
69f1d475 BH |
714 | pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n", |
715 | (unsigned long long) region->start_pfn << PAGE_SHIFT, | |
716 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
717 | - 1); | |
74dfd666 RW |
718 | |
719 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
720 | if (pfn_valid(pfn)) { |
721 | /* | |
722 | * It is safe to ignore the result of | |
723 | * mem_bm_set_bit_check() here, since we won't | |
724 | * touch the PFNs for which the error is | |
725 | * returned anyway. | |
726 | */ | |
727 | mem_bm_set_bit_check(bm, pfn); | |
728 | } | |
74dfd666 RW |
729 | } |
730 | } | |
731 | ||
732 | /** | |
733 | * create_basic_memory_bitmaps - create bitmaps needed for marking page | |
734 | * frames that should not be saved and free page frames. The pointers | |
735 | * forbidden_pages_map and free_pages_map are only modified if everything | |
736 | * goes well, because we don't want the bits to be used before both bitmaps | |
737 | * are set up. | |
738 | */ | |
739 | ||
740 | int create_basic_memory_bitmaps(void) | |
741 | { | |
742 | struct memory_bitmap *bm1, *bm2; | |
743 | int error = 0; | |
744 | ||
745 | BUG_ON(forbidden_pages_map || free_pages_map); | |
746 | ||
0709db60 | 747 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
748 | if (!bm1) |
749 | return -ENOMEM; | |
750 | ||
0709db60 | 751 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
752 | if (error) |
753 | goto Free_first_object; | |
754 | ||
0709db60 | 755 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
756 | if (!bm2) |
757 | goto Free_first_bitmap; | |
758 | ||
0709db60 | 759 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
760 | if (error) |
761 | goto Free_second_object; | |
762 | ||
763 | forbidden_pages_map = bm1; | |
764 | free_pages_map = bm2; | |
765 | mark_nosave_pages(forbidden_pages_map); | |
766 | ||
23976728 | 767 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
768 | |
769 | return 0; | |
770 | ||
771 | Free_second_object: | |
772 | kfree(bm2); | |
773 | Free_first_bitmap: | |
774 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
775 | Free_first_object: | |
776 | kfree(bm1); | |
777 | return -ENOMEM; | |
778 | } | |
779 | ||
780 | /** | |
781 | * free_basic_memory_bitmaps - free memory bitmaps allocated by | |
782 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary | |
783 | * so that the bitmaps themselves are not referred to while they are being | |
784 | * freed. | |
785 | */ | |
786 | ||
787 | void free_basic_memory_bitmaps(void) | |
788 | { | |
789 | struct memory_bitmap *bm1, *bm2; | |
790 | ||
791 | BUG_ON(!(forbidden_pages_map && free_pages_map)); | |
792 | ||
793 | bm1 = forbidden_pages_map; | |
794 | bm2 = free_pages_map; | |
795 | forbidden_pages_map = NULL; | |
796 | free_pages_map = NULL; | |
797 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
798 | kfree(bm1); | |
799 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
800 | kfree(bm2); | |
801 | ||
23976728 | 802 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
803 | } |
804 | ||
b788db79 RW |
805 | /** |
806 | * snapshot_additional_pages - estimate the number of additional pages | |
807 | * be needed for setting up the suspend image data structures for given | |
808 | * zone (usually the returned value is greater than the exact number) | |
809 | */ | |
810 | ||
811 | unsigned int snapshot_additional_pages(struct zone *zone) | |
812 | { | |
813 | unsigned int res; | |
814 | ||
815 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
160cb5a9 NK |
816 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), |
817 | LINKED_PAGE_DATA_SIZE); | |
8357376d | 818 | return 2 * res; |
b788db79 RW |
819 | } |
820 | ||
8357376d RW |
821 | #ifdef CONFIG_HIGHMEM |
822 | /** | |
823 | * count_free_highmem_pages - compute the total number of free highmem | |
824 | * pages, system-wide. | |
825 | */ | |
826 | ||
827 | static unsigned int count_free_highmem_pages(void) | |
828 | { | |
829 | struct zone *zone; | |
830 | unsigned int cnt = 0; | |
831 | ||
ee99c71c KM |
832 | for_each_populated_zone(zone) |
833 | if (is_highmem(zone)) | |
d23ad423 | 834 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
835 | |
836 | return cnt; | |
837 | } | |
838 | ||
839 | /** | |
840 | * saveable_highmem_page - Determine whether a highmem page should be | |
841 | * included in the suspend image. | |
842 | * | |
843 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, | |
844 | * and it isn't a part of a free chunk of pages. | |
845 | */ | |
846705de | 846 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
847 | { |
848 | struct page *page; | |
849 | ||
850 | if (!pfn_valid(pfn)) | |
851 | return NULL; | |
852 | ||
853 | page = pfn_to_page(pfn); | |
846705de RW |
854 | if (page_zone(page) != zone) |
855 | return NULL; | |
8357376d RW |
856 | |
857 | BUG_ON(!PageHighMem(page)); | |
858 | ||
7be98234 RW |
859 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
860 | PageReserved(page)) | |
8357376d RW |
861 | return NULL; |
862 | ||
c6968e73 SG |
863 | if (page_is_guard(page)) |
864 | return NULL; | |
865 | ||
8357376d RW |
866 | return page; |
867 | } | |
868 | ||
869 | /** | |
870 | * count_highmem_pages - compute the total number of saveable highmem | |
871 | * pages. | |
872 | */ | |
873 | ||
fe419535 | 874 | static unsigned int count_highmem_pages(void) |
8357376d RW |
875 | { |
876 | struct zone *zone; | |
877 | unsigned int n = 0; | |
878 | ||
98e73dc5 | 879 | for_each_populated_zone(zone) { |
8357376d RW |
880 | unsigned long pfn, max_zone_pfn; |
881 | ||
882 | if (!is_highmem(zone)) | |
883 | continue; | |
884 | ||
885 | mark_free_pages(zone); | |
886 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
887 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
846705de | 888 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
889 | n++; |
890 | } | |
891 | return n; | |
892 | } | |
893 | #else | |
846705de RW |
894 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
895 | { | |
896 | return NULL; | |
897 | } | |
8357376d RW |
898 | #endif /* CONFIG_HIGHMEM */ |
899 | ||
25761b6e | 900 | /** |
8a235efa RW |
901 | * saveable_page - Determine whether a non-highmem page should be included |
902 | * in the suspend image. | |
25761b6e | 903 | * |
8357376d RW |
904 | * We should save the page if it isn't Nosave, and is not in the range |
905 | * of pages statically defined as 'unsaveable', and it isn't a part of | |
906 | * a free chunk of pages. | |
25761b6e | 907 | */ |
846705de | 908 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 909 | { |
de491861 | 910 | struct page *page; |
25761b6e RW |
911 | |
912 | if (!pfn_valid(pfn)) | |
ae83c5ee | 913 | return NULL; |
25761b6e RW |
914 | |
915 | page = pfn_to_page(pfn); | |
846705de RW |
916 | if (page_zone(page) != zone) |
917 | return NULL; | |
ae83c5ee | 918 | |
8357376d RW |
919 | BUG_ON(PageHighMem(page)); |
920 | ||
7be98234 | 921 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 922 | return NULL; |
8357376d | 923 | |
8a235efa RW |
924 | if (PageReserved(page) |
925 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 926 | return NULL; |
25761b6e | 927 | |
c6968e73 SG |
928 | if (page_is_guard(page)) |
929 | return NULL; | |
930 | ||
ae83c5ee | 931 | return page; |
25761b6e RW |
932 | } |
933 | ||
8357376d RW |
934 | /** |
935 | * count_data_pages - compute the total number of saveable non-highmem | |
936 | * pages. | |
937 | */ | |
938 | ||
fe419535 | 939 | static unsigned int count_data_pages(void) |
25761b6e RW |
940 | { |
941 | struct zone *zone; | |
ae83c5ee | 942 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 943 | unsigned int n = 0; |
25761b6e | 944 | |
98e73dc5 | 945 | for_each_populated_zone(zone) { |
25761b6e RW |
946 | if (is_highmem(zone)) |
947 | continue; | |
8357376d | 948 | |
25761b6e | 949 | mark_free_pages(zone); |
ae83c5ee RW |
950 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
951 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
846705de | 952 | if (saveable_page(zone, pfn)) |
8357376d | 953 | n++; |
25761b6e | 954 | } |
a0f49651 | 955 | return n; |
25761b6e RW |
956 | } |
957 | ||
8357376d RW |
958 | /* This is needed, because copy_page and memcpy are not usable for copying |
959 | * task structs. | |
960 | */ | |
961 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
962 | { |
963 | int n; | |
964 | ||
f623f0db RW |
965 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
966 | *dst++ = *src++; | |
967 | } | |
968 | ||
8a235efa RW |
969 | |
970 | /** | |
971 | * safe_copy_page - check if the page we are going to copy is marked as | |
972 | * present in the kernel page tables (this always is the case if | |
973 | * CONFIG_DEBUG_PAGEALLOC is not set and in that case | |
974 | * kernel_page_present() always returns 'true'). | |
975 | */ | |
976 | static void safe_copy_page(void *dst, struct page *s_page) | |
977 | { | |
978 | if (kernel_page_present(s_page)) { | |
979 | do_copy_page(dst, page_address(s_page)); | |
980 | } else { | |
981 | kernel_map_pages(s_page, 1, 1); | |
982 | do_copy_page(dst, page_address(s_page)); | |
983 | kernel_map_pages(s_page, 1, 0); | |
984 | } | |
985 | } | |
986 | ||
987 | ||
8357376d RW |
988 | #ifdef CONFIG_HIGHMEM |
989 | static inline struct page * | |
990 | page_is_saveable(struct zone *zone, unsigned long pfn) | |
991 | { | |
992 | return is_highmem(zone) ? | |
846705de | 993 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
994 | } |
995 | ||
8a235efa | 996 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
997 | { |
998 | struct page *s_page, *d_page; | |
999 | void *src, *dst; | |
1000 | ||
1001 | s_page = pfn_to_page(src_pfn); | |
1002 | d_page = pfn_to_page(dst_pfn); | |
1003 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1004 | src = kmap_atomic(s_page); |
1005 | dst = kmap_atomic(d_page); | |
8357376d | 1006 | do_copy_page(dst, src); |
0de9a1e2 CW |
1007 | kunmap_atomic(dst); |
1008 | kunmap_atomic(src); | |
8357376d | 1009 | } else { |
8357376d RW |
1010 | if (PageHighMem(d_page)) { |
1011 | /* Page pointed to by src may contain some kernel | |
1012 | * data modified by kmap_atomic() | |
1013 | */ | |
8a235efa | 1014 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1015 | dst = kmap_atomic(d_page); |
3ecb01df | 1016 | copy_page(dst, buffer); |
0de9a1e2 | 1017 | kunmap_atomic(dst); |
8357376d | 1018 | } else { |
8a235efa | 1019 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1020 | } |
1021 | } | |
1022 | } | |
1023 | #else | |
846705de | 1024 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1025 | |
8a235efa | 1026 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1027 | { |
8a235efa RW |
1028 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1029 | pfn_to_page(src_pfn)); | |
8357376d RW |
1030 | } |
1031 | #endif /* CONFIG_HIGHMEM */ | |
1032 | ||
b788db79 RW |
1033 | static void |
1034 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) | |
25761b6e RW |
1035 | { |
1036 | struct zone *zone; | |
b788db79 | 1037 | unsigned long pfn; |
25761b6e | 1038 | |
98e73dc5 | 1039 | for_each_populated_zone(zone) { |
b788db79 RW |
1040 | unsigned long max_zone_pfn; |
1041 | ||
25761b6e | 1042 | mark_free_pages(zone); |
ae83c5ee | 1043 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
b788db79 | 1044 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1045 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1046 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1047 | } |
b788db79 RW |
1048 | memory_bm_position_reset(orig_bm); |
1049 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1050 | for(;;) { |
b788db79 | 1051 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1052 | if (unlikely(pfn == BM_END_OF_MAP)) |
1053 | break; | |
1054 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1055 | } | |
25761b6e RW |
1056 | } |
1057 | ||
8357376d RW |
1058 | /* Total number of image pages */ |
1059 | static unsigned int nr_copy_pages; | |
1060 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1061 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1062 | /* |
1063 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1064 | * before suspending devices. | |
1065 | */ | |
1066 | unsigned int alloc_normal, alloc_highmem; | |
1067 | /* | |
1068 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1069 | * hibernation image pages (during restore) | |
1070 | */ | |
1071 | static struct memory_bitmap orig_bm; | |
1072 | /* | |
1073 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1074 | * will contain copies of saveable pages. During restore it is initially used | |
1075 | * for marking hibernation image pages, but then the set bits from it are | |
1076 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1077 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1078 | * this purpose. | |
1079 | */ | |
1080 | static struct memory_bitmap copy_bm; | |
8357376d | 1081 | |
25761b6e | 1082 | /** |
940864dd | 1083 | * swsusp_free - free pages allocated for the suspend. |
cd560bb2 | 1084 | * |
940864dd RW |
1085 | * Suspend pages are alocated before the atomic copy is made, so we |
1086 | * need to release them after the resume. | |
25761b6e RW |
1087 | */ |
1088 | ||
1089 | void swsusp_free(void) | |
1090 | { | |
1091 | struct zone *zone; | |
ae83c5ee | 1092 | unsigned long pfn, max_zone_pfn; |
25761b6e | 1093 | |
98e73dc5 | 1094 | for_each_populated_zone(zone) { |
ae83c5ee RW |
1095 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1096 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1097 | if (pfn_valid(pfn)) { | |
1098 | struct page *page = pfn_to_page(pfn); | |
1099 | ||
7be98234 RW |
1100 | if (swsusp_page_is_forbidden(page) && |
1101 | swsusp_page_is_free(page)) { | |
1102 | swsusp_unset_page_forbidden(page); | |
1103 | swsusp_unset_page_free(page); | |
8357376d | 1104 | __free_page(page); |
25761b6e RW |
1105 | } |
1106 | } | |
1107 | } | |
f577eb30 RW |
1108 | nr_copy_pages = 0; |
1109 | nr_meta_pages = 0; | |
75534b50 | 1110 | restore_pblist = NULL; |
6e1819d6 | 1111 | buffer = NULL; |
64a473cb RW |
1112 | alloc_normal = 0; |
1113 | alloc_highmem = 0; | |
25761b6e RW |
1114 | } |
1115 | ||
4bb33435 RW |
1116 | /* Helper functions used for the shrinking of memory. */ |
1117 | ||
1118 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1119 | ||
fe419535 | 1120 | /** |
4bb33435 RW |
1121 | * preallocate_image_pages - Allocate a number of pages for hibernation image |
1122 | * @nr_pages: Number of page frames to allocate. | |
1123 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1124 | * |
4bb33435 RW |
1125 | * Return value: Number of page frames actually allocated |
1126 | */ | |
1127 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1128 | { | |
1129 | unsigned long nr_alloc = 0; | |
1130 | ||
1131 | while (nr_pages > 0) { | |
64a473cb RW |
1132 | struct page *page; |
1133 | ||
1134 | page = alloc_image_page(mask); | |
1135 | if (!page) | |
4bb33435 | 1136 | break; |
64a473cb RW |
1137 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1138 | if (PageHighMem(page)) | |
1139 | alloc_highmem++; | |
1140 | else | |
1141 | alloc_normal++; | |
4bb33435 RW |
1142 | nr_pages--; |
1143 | nr_alloc++; | |
1144 | } | |
1145 | ||
1146 | return nr_alloc; | |
1147 | } | |
1148 | ||
6715045d RW |
1149 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1150 | unsigned long avail_normal) | |
4bb33435 | 1151 | { |
6715045d RW |
1152 | unsigned long alloc; |
1153 | ||
1154 | if (avail_normal <= alloc_normal) | |
1155 | return 0; | |
1156 | ||
1157 | alloc = avail_normal - alloc_normal; | |
1158 | if (nr_pages < alloc) | |
1159 | alloc = nr_pages; | |
1160 | ||
1161 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1162 | } |
1163 | ||
1164 | #ifdef CONFIG_HIGHMEM | |
1165 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1166 | { | |
1167 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1168 | } | |
1169 | ||
1170 | /** | |
1171 | * __fraction - Compute (an approximation of) x * (multiplier / base) | |
fe419535 | 1172 | */ |
4bb33435 RW |
1173 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1174 | { | |
1175 | x *= multiplier; | |
1176 | do_div(x, base); | |
1177 | return (unsigned long)x; | |
1178 | } | |
fe419535 | 1179 | |
4bb33435 RW |
1180 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
1181 | unsigned long highmem, | |
1182 | unsigned long total) | |
fe419535 | 1183 | { |
4bb33435 RW |
1184 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1185 | ||
1186 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1187 | } |
4bb33435 RW |
1188 | #else /* CONFIG_HIGHMEM */ |
1189 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1190 | { | |
1191 | return 0; | |
1192 | } | |
1193 | ||
1194 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
1195 | unsigned long highmem, | |
1196 | unsigned long total) | |
1197 | { | |
1198 | return 0; | |
1199 | } | |
1200 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1201 | |
4bb33435 | 1202 | /** |
64a473cb RW |
1203 | * free_unnecessary_pages - Release preallocated pages not needed for the image |
1204 | */ | |
1205 | static void free_unnecessary_pages(void) | |
1206 | { | |
6715045d | 1207 | unsigned long save, to_free_normal, to_free_highmem; |
64a473cb | 1208 | |
6715045d RW |
1209 | save = count_data_pages(); |
1210 | if (alloc_normal >= save) { | |
1211 | to_free_normal = alloc_normal - save; | |
1212 | save = 0; | |
1213 | } else { | |
1214 | to_free_normal = 0; | |
1215 | save -= alloc_normal; | |
1216 | } | |
1217 | save += count_highmem_pages(); | |
1218 | if (alloc_highmem >= save) { | |
1219 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1220 | } else { |
1221 | to_free_highmem = 0; | |
4d4cf23c RW |
1222 | save -= alloc_highmem; |
1223 | if (to_free_normal > save) | |
1224 | to_free_normal -= save; | |
1225 | else | |
1226 | to_free_normal = 0; | |
64a473cb RW |
1227 | } |
1228 | ||
1229 | memory_bm_position_reset(©_bm); | |
1230 | ||
a9c9b442 | 1231 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1232 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1233 | struct page *page = pfn_to_page(pfn); | |
1234 | ||
1235 | if (PageHighMem(page)) { | |
1236 | if (!to_free_highmem) | |
1237 | continue; | |
1238 | to_free_highmem--; | |
1239 | alloc_highmem--; | |
1240 | } else { | |
1241 | if (!to_free_normal) | |
1242 | continue; | |
1243 | to_free_normal--; | |
1244 | alloc_normal--; | |
1245 | } | |
1246 | memory_bm_clear_bit(©_bm, pfn); | |
1247 | swsusp_unset_page_forbidden(page); | |
1248 | swsusp_unset_page_free(page); | |
1249 | __free_page(page); | |
1250 | } | |
1251 | } | |
1252 | ||
ef4aede3 RW |
1253 | /** |
1254 | * minimum_image_size - Estimate the minimum acceptable size of an image | |
1255 | * @saveable: Number of saveable pages in the system. | |
1256 | * | |
1257 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1258 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1259 | * preallocating memory. | |
1260 | * | |
1261 | * We assume that the minimum image size should be proportional to | |
1262 | * | |
1263 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1264 | * | |
1265 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
1266 | * and (3) inactive anonymouns pages, (4) active and (5) inactive file pages, | |
1267 | * minus mapped file pages. | |
1268 | */ | |
1269 | static unsigned long minimum_image_size(unsigned long saveable) | |
1270 | { | |
1271 | unsigned long size; | |
1272 | ||
1273 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
1274 | + global_page_state(NR_ACTIVE_ANON) | |
1275 | + global_page_state(NR_INACTIVE_ANON) | |
1276 | + global_page_state(NR_ACTIVE_FILE) | |
1277 | + global_page_state(NR_INACTIVE_FILE) | |
1278 | - global_page_state(NR_FILE_MAPPED); | |
1279 | ||
1280 | return saveable <= size ? 0 : saveable - size; | |
1281 | } | |
1282 | ||
64a473cb RW |
1283 | /** |
1284 | * hibernate_preallocate_memory - Preallocate memory for hibernation image | |
4bb33435 RW |
1285 | * |
1286 | * To create a hibernation image it is necessary to make a copy of every page | |
1287 | * frame in use. We also need a number of page frames to be free during | |
1288 | * hibernation for allocations made while saving the image and for device | |
1289 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 RW |
1290 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
1291 | * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through | |
1292 | * /sys/power/reserved_size, respectively). To make this happen, we compute the | |
1293 | * total number of available page frames and allocate at least | |
4bb33435 | 1294 | * |
ddeb6487 RW |
1295 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1296 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1297 | * |
1298 | * of them, which corresponds to the maximum size of a hibernation image. | |
1299 | * | |
1300 | * If image_size is set below the number following from the above formula, | |
1301 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1302 | * pages in the system is below the requested image size or the minimum |
1303 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1304 | */ |
64a473cb | 1305 | int hibernate_preallocate_memory(void) |
fe419535 | 1306 | { |
fe419535 | 1307 | struct zone *zone; |
4bb33435 | 1308 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1309 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
fe419535 | 1310 | struct timeval start, stop; |
64a473cb | 1311 | int error; |
fe419535 | 1312 | |
64a473cb | 1313 | printk(KERN_INFO "PM: Preallocating image memory... "); |
fe419535 | 1314 | do_gettimeofday(&start); |
fe419535 | 1315 | |
64a473cb RW |
1316 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1317 | if (error) | |
1318 | goto err_out; | |
1319 | ||
1320 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1321 | if (error) | |
1322 | goto err_out; | |
1323 | ||
1324 | alloc_normal = 0; | |
1325 | alloc_highmem = 0; | |
1326 | ||
4bb33435 | 1327 | /* Count the number of saveable data pages. */ |
64a473cb | 1328 | save_highmem = count_highmem_pages(); |
4bb33435 | 1329 | saveable = count_data_pages(); |
fe419535 | 1330 | |
4bb33435 RW |
1331 | /* |
1332 | * Compute the total number of page frames we can use (count) and the | |
1333 | * number of pages needed for image metadata (size). | |
1334 | */ | |
1335 | count = saveable; | |
64a473cb RW |
1336 | saveable += save_highmem; |
1337 | highmem = save_highmem; | |
4bb33435 RW |
1338 | size = 0; |
1339 | for_each_populated_zone(zone) { | |
1340 | size += snapshot_additional_pages(zone); | |
1341 | if (is_highmem(zone)) | |
1342 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1343 | else | |
1344 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1345 | } | |
6715045d | 1346 | avail_normal = count; |
4bb33435 RW |
1347 | count += highmem; |
1348 | count -= totalreserve_pages; | |
1349 | ||
85055dd8 MS |
1350 | /* Add number of pages required for page keys (s390 only). */ |
1351 | size += page_key_additional_pages(saveable); | |
1352 | ||
4bb33435 | 1353 | /* Compute the maximum number of saveable pages to leave in memory. */ |
ddeb6487 RW |
1354 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1355 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1356 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1357 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1358 | if (size > max_size) | |
1359 | size = max_size; | |
1360 | /* | |
266f1a25 RW |
1361 | * If the desired number of image pages is at least as large as the |
1362 | * current number of saveable pages in memory, allocate page frames for | |
1363 | * the image and we're done. | |
4bb33435 | 1364 | */ |
64a473cb RW |
1365 | if (size >= saveable) { |
1366 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1367 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1368 | goto out; |
64a473cb | 1369 | } |
4bb33435 | 1370 | |
ef4aede3 RW |
1371 | /* Estimate the minimum size of the image. */ |
1372 | pages = minimum_image_size(saveable); | |
6715045d RW |
1373 | /* |
1374 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1375 | * accommodate an image of the minimum size (unless it's already too | |
1376 | * small, in which case don't preallocate pages from it at all). | |
1377 | */ | |
1378 | if (avail_normal > pages) | |
1379 | avail_normal -= pages; | |
1380 | else | |
1381 | avail_normal = 0; | |
ef4aede3 RW |
1382 | if (size < pages) |
1383 | size = min_t(unsigned long, pages, max_size); | |
1384 | ||
4bb33435 RW |
1385 | /* |
1386 | * Let the memory management subsystem know that we're going to need a | |
1387 | * large number of page frames to allocate and make it free some memory. | |
1388 | * NOTE: If this is not done, performance will be hurt badly in some | |
1389 | * test cases. | |
1390 | */ | |
1391 | shrink_all_memory(saveable - size); | |
1392 | ||
1393 | /* | |
1394 | * The number of saveable pages in memory was too high, so apply some | |
1395 | * pressure to decrease it. First, make room for the largest possible | |
1396 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1397 | * of the image as much as indicated by 'size' using allocations from |
1398 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1399 | */ |
1400 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
1401 | alloc = (count - max_size) - pages_highmem; | |
6715045d RW |
1402 | pages = preallocate_image_memory(alloc, avail_normal); |
1403 | if (pages < alloc) { | |
1404 | /* We have exhausted non-highmem pages, try highmem. */ | |
1405 | alloc -= pages; | |
1406 | pages += pages_highmem; | |
1407 | pages_highmem = preallocate_image_highmem(alloc); | |
1408 | if (pages_highmem < alloc) | |
1409 | goto err_out; | |
1410 | pages += pages_highmem; | |
1411 | /* | |
1412 | * size is the desired number of saveable pages to leave in | |
1413 | * memory, so try to preallocate (all memory - size) pages. | |
1414 | */ | |
1415 | alloc = (count - pages) - size; | |
1416 | pages += preallocate_image_highmem(alloc); | |
1417 | } else { | |
1418 | /* | |
1419 | * There are approximately max_size saveable pages at this point | |
1420 | * and we want to reduce this number down to size. | |
1421 | */ | |
1422 | alloc = max_size - size; | |
1423 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1424 | pages_highmem += size; | |
1425 | alloc -= size; | |
1426 | size = preallocate_image_memory(alloc, avail_normal); | |
1427 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1428 | pages += pages_highmem + size; | |
1429 | } | |
4bb33435 | 1430 | |
64a473cb RW |
1431 | /* |
1432 | * We only need as many page frames for the image as there are saveable | |
1433 | * pages in memory, but we have allocated more. Release the excessive | |
1434 | * ones now. | |
1435 | */ | |
1436 | free_unnecessary_pages(); | |
4bb33435 RW |
1437 | |
1438 | out: | |
fe419535 | 1439 | do_gettimeofday(&stop); |
64a473cb RW |
1440 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
1441 | swsusp_show_speed(&start, &stop, pages, "Allocated"); | |
fe419535 RW |
1442 | |
1443 | return 0; | |
64a473cb RW |
1444 | |
1445 | err_out: | |
1446 | printk(KERN_CONT "\n"); | |
1447 | swsusp_free(); | |
1448 | return -ENOMEM; | |
fe419535 RW |
1449 | } |
1450 | ||
8357376d RW |
1451 | #ifdef CONFIG_HIGHMEM |
1452 | /** | |
1453 | * count_pages_for_highmem - compute the number of non-highmem pages | |
1454 | * that will be necessary for creating copies of highmem pages. | |
1455 | */ | |
1456 | ||
1457 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) | |
1458 | { | |
64a473cb | 1459 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1460 | |
1461 | if (free_highmem >= nr_highmem) | |
1462 | nr_highmem = 0; | |
1463 | else | |
1464 | nr_highmem -= free_highmem; | |
1465 | ||
1466 | return nr_highmem; | |
1467 | } | |
1468 | #else | |
1469 | static unsigned int | |
1470 | count_pages_for_highmem(unsigned int nr_highmem) { return 0; } | |
1471 | #endif /* CONFIG_HIGHMEM */ | |
25761b6e RW |
1472 | |
1473 | /** | |
8357376d RW |
1474 | * enough_free_mem - Make sure we have enough free memory for the |
1475 | * snapshot image. | |
25761b6e RW |
1476 | */ |
1477 | ||
8357376d | 1478 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1479 | { |
e5e2fa78 | 1480 | struct zone *zone; |
64a473cb | 1481 | unsigned int free = alloc_normal; |
e5e2fa78 | 1482 | |
98e73dc5 | 1483 | for_each_populated_zone(zone) |
8357376d | 1484 | if (!is_highmem(zone)) |
d23ad423 | 1485 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1486 | |
8357376d | 1487 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1488 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1489 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1490 | |
64a473cb | 1491 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1492 | } |
1493 | ||
8357376d RW |
1494 | #ifdef CONFIG_HIGHMEM |
1495 | /** | |
1496 | * get_highmem_buffer - if there are some highmem pages in the suspend | |
1497 | * image, we may need the buffer to copy them and/or load their data. | |
1498 | */ | |
1499 | ||
1500 | static inline int get_highmem_buffer(int safe_needed) | |
1501 | { | |
1502 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1503 | return buffer ? 0 : -ENOMEM; | |
1504 | } | |
1505 | ||
1506 | /** | |
1507 | * alloc_highmem_image_pages - allocate some highmem pages for the image. | |
1508 | * Try to allocate as many pages as needed, but if the number of free | |
1509 | * highmem pages is lesser than that, allocate them all. | |
1510 | */ | |
1511 | ||
1512 | static inline unsigned int | |
64a473cb | 1513 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int nr_highmem) |
8357376d RW |
1514 | { |
1515 | unsigned int to_alloc = count_free_highmem_pages(); | |
1516 | ||
1517 | if (to_alloc > nr_highmem) | |
1518 | to_alloc = nr_highmem; | |
1519 | ||
1520 | nr_highmem -= to_alloc; | |
1521 | while (to_alloc-- > 0) { | |
1522 | struct page *page; | |
1523 | ||
1524 | page = alloc_image_page(__GFP_HIGHMEM); | |
1525 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1526 | } | |
1527 | return nr_highmem; | |
1528 | } | |
1529 | #else | |
1530 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1531 | ||
1532 | static inline unsigned int | |
64a473cb | 1533 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int n) { return 0; } |
8357376d RW |
1534 | #endif /* CONFIG_HIGHMEM */ |
1535 | ||
1536 | /** | |
1537 | * swsusp_alloc - allocate memory for the suspend image | |
1538 | * | |
1539 | * We first try to allocate as many highmem pages as there are | |
1540 | * saveable highmem pages in the system. If that fails, we allocate | |
1541 | * non-highmem pages for the copies of the remaining highmem ones. | |
1542 | * | |
1543 | * In this approach it is likely that the copies of highmem pages will | |
1544 | * also be located in the high memory, because of the way in which | |
1545 | * copy_data_pages() works. | |
1546 | */ | |
1547 | ||
b788db79 RW |
1548 | static int |
1549 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, | |
8357376d | 1550 | unsigned int nr_pages, unsigned int nr_highmem) |
054bd4c1 | 1551 | { |
8357376d | 1552 | if (nr_highmem > 0) { |
2e725a06 | 1553 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1554 | goto err_out; |
1555 | if (nr_highmem > alloc_highmem) { | |
1556 | nr_highmem -= alloc_highmem; | |
1557 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1558 | } | |
8357376d | 1559 | } |
64a473cb RW |
1560 | if (nr_pages > alloc_normal) { |
1561 | nr_pages -= alloc_normal; | |
1562 | while (nr_pages-- > 0) { | |
1563 | struct page *page; | |
1564 | ||
1565 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1566 | if (!page) | |
1567 | goto err_out; | |
1568 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1569 | } | |
25761b6e | 1570 | } |
64a473cb | 1571 | |
b788db79 | 1572 | return 0; |
25761b6e | 1573 | |
64a473cb | 1574 | err_out: |
b788db79 | 1575 | swsusp_free(); |
2e725a06 | 1576 | return -ENOMEM; |
25761b6e RW |
1577 | } |
1578 | ||
2e32a43e | 1579 | asmlinkage int swsusp_save(void) |
25761b6e | 1580 | { |
8357376d | 1581 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1582 | |
07c3bb57 | 1583 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1584 | |
9f8f2172 | 1585 | drain_local_pages(NULL); |
a0f49651 | 1586 | nr_pages = count_data_pages(); |
8357376d | 1587 | nr_highmem = count_highmem_pages(); |
23976728 | 1588 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1589 | |
8357376d | 1590 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1591 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1592 | return -ENOMEM; |
1593 | } | |
1594 | ||
8357376d | 1595 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1596 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1597 | return -ENOMEM; |
8357376d | 1598 | } |
25761b6e RW |
1599 | |
1600 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1601 | * Kill them. | |
1602 | */ | |
9f8f2172 | 1603 | drain_local_pages(NULL); |
b788db79 | 1604 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1605 | |
1606 | /* | |
1607 | * End of critical section. From now on, we can write to memory, | |
1608 | * but we should not touch disk. This specially means we must _not_ | |
1609 | * touch swap space! Except we must write out our image of course. | |
1610 | */ | |
1611 | ||
8357376d | 1612 | nr_pages += nr_highmem; |
a0f49651 | 1613 | nr_copy_pages = nr_pages; |
8357376d | 1614 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 1615 | |
23976728 RW |
1616 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1617 | nr_pages); | |
8357376d | 1618 | |
25761b6e RW |
1619 | return 0; |
1620 | } | |
f577eb30 | 1621 | |
d307c4a8 RW |
1622 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1623 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 1624 | { |
d307c4a8 | 1625 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 1626 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
1627 | return 0; |
1628 | } | |
1629 | ||
1630 | static char *check_image_kernel(struct swsusp_info *info) | |
1631 | { | |
1632 | if (info->version_code != LINUX_VERSION_CODE) | |
1633 | return "kernel version"; | |
1634 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
1635 | return "system type"; | |
1636 | if (strcmp(info->uts.release,init_utsname()->release)) | |
1637 | return "kernel release"; | |
1638 | if (strcmp(info->uts.version,init_utsname()->version)) | |
1639 | return "version"; | |
1640 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
1641 | return "machine"; | |
1642 | return NULL; | |
1643 | } | |
1644 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
1645 | ||
af508b34 RW |
1646 | unsigned long snapshot_get_image_size(void) |
1647 | { | |
1648 | return nr_copy_pages + nr_meta_pages + 1; | |
1649 | } | |
1650 | ||
d307c4a8 RW |
1651 | static int init_header(struct swsusp_info *info) |
1652 | { | |
1653 | memset(info, 0, sizeof(struct swsusp_info)); | |
f577eb30 | 1654 | info->num_physpages = num_physpages; |
f577eb30 | 1655 | info->image_pages = nr_copy_pages; |
af508b34 | 1656 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
1657 | info->size = info->pages; |
1658 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 1659 | return init_header_complete(info); |
f577eb30 RW |
1660 | } |
1661 | ||
1662 | /** | |
940864dd RW |
1663 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1664 | * are stored in the array @buf[] (1 page at a time) | |
f577eb30 RW |
1665 | */ |
1666 | ||
b788db79 | 1667 | static inline void |
940864dd | 1668 | pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1669 | { |
1670 | int j; | |
1671 | ||
b788db79 | 1672 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
1673 | buf[j] = memory_bm_next_pfn(bm); |
1674 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 1675 | break; |
85055dd8 MS |
1676 | /* Save page key for data page (s390 only). */ |
1677 | page_key_read(buf + j); | |
f577eb30 | 1678 | } |
f577eb30 RW |
1679 | } |
1680 | ||
1681 | /** | |
1682 | * snapshot_read_next - used for reading the system memory snapshot. | |
1683 | * | |
1684 | * On the first call to it @handle should point to a zeroed | |
1685 | * snapshot_handle structure. The structure gets updated and a pointer | |
1686 | * to it should be passed to this function every next time. | |
1687 | * | |
f577eb30 RW |
1688 | * On success the function returns a positive number. Then, the caller |
1689 | * is allowed to read up to the returned number of bytes from the memory | |
d3c1b24c | 1690 | * location computed by the data_of() macro. |
f577eb30 RW |
1691 | * |
1692 | * The function returns 0 to indicate the end of data stream condition, | |
1693 | * and a negative number is returned on error. In such cases the | |
1694 | * structure pointed to by @handle is not updated and should not be used | |
1695 | * any more. | |
1696 | */ | |
1697 | ||
d3c1b24c | 1698 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 1699 | { |
fb13a28b | 1700 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1701 | return 0; |
b788db79 | 1702 | |
f577eb30 RW |
1703 | if (!buffer) { |
1704 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 1705 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1706 | if (!buffer) |
1707 | return -ENOMEM; | |
1708 | } | |
d3c1b24c | 1709 | if (!handle->cur) { |
d307c4a8 RW |
1710 | int error; |
1711 | ||
1712 | error = init_header((struct swsusp_info *)buffer); | |
1713 | if (error) | |
1714 | return error; | |
f577eb30 | 1715 | handle->buffer = buffer; |
b788db79 RW |
1716 | memory_bm_position_reset(&orig_bm); |
1717 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 1718 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 1719 | clear_page(buffer); |
d3c1b24c JS |
1720 | pack_pfns(buffer, &orig_bm); |
1721 | } else { | |
1722 | struct page *page; | |
b788db79 | 1723 | |
d3c1b24c JS |
1724 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
1725 | if (PageHighMem(page)) { | |
1726 | /* Highmem pages are copied to the buffer, | |
1727 | * because we can't return with a kmapped | |
1728 | * highmem page (we may not be called again). | |
1729 | */ | |
1730 | void *kaddr; | |
8357376d | 1731 | |
0de9a1e2 | 1732 | kaddr = kmap_atomic(page); |
3ecb01df | 1733 | copy_page(buffer, kaddr); |
0de9a1e2 | 1734 | kunmap_atomic(kaddr); |
d3c1b24c JS |
1735 | handle->buffer = buffer; |
1736 | } else { | |
1737 | handle->buffer = page_address(page); | |
f577eb30 | 1738 | } |
f577eb30 | 1739 | } |
d3c1b24c JS |
1740 | handle->cur++; |
1741 | return PAGE_SIZE; | |
f577eb30 RW |
1742 | } |
1743 | ||
1744 | /** | |
1745 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
1746 | * the image during resume, because they conflict with the pages that | |
1747 | * had been used before suspend | |
1748 | */ | |
1749 | ||
940864dd | 1750 | static int mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 RW |
1751 | { |
1752 | struct zone *zone; | |
ae83c5ee | 1753 | unsigned long pfn, max_zone_pfn; |
f577eb30 RW |
1754 | |
1755 | /* Clear page flags */ | |
98e73dc5 | 1756 | for_each_populated_zone(zone) { |
ae83c5ee RW |
1757 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1758 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1759 | if (pfn_valid(pfn)) | |
7be98234 | 1760 | swsusp_unset_page_free(pfn_to_page(pfn)); |
f577eb30 RW |
1761 | } |
1762 | ||
940864dd RW |
1763 | /* Mark pages that correspond to the "original" pfns as "unsafe" */ |
1764 | memory_bm_position_reset(bm); | |
1765 | do { | |
1766 | pfn = memory_bm_next_pfn(bm); | |
1767 | if (likely(pfn != BM_END_OF_MAP)) { | |
1768 | if (likely(pfn_valid(pfn))) | |
7be98234 | 1769 | swsusp_set_page_free(pfn_to_page(pfn)); |
940864dd RW |
1770 | else |
1771 | return -EFAULT; | |
1772 | } | |
1773 | } while (pfn != BM_END_OF_MAP); | |
f577eb30 | 1774 | |
940864dd | 1775 | allocated_unsafe_pages = 0; |
968808b8 | 1776 | |
f577eb30 RW |
1777 | return 0; |
1778 | } | |
1779 | ||
940864dd RW |
1780 | static void |
1781 | duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src) | |
f577eb30 | 1782 | { |
940864dd RW |
1783 | unsigned long pfn; |
1784 | ||
1785 | memory_bm_position_reset(src); | |
1786 | pfn = memory_bm_next_pfn(src); | |
1787 | while (pfn != BM_END_OF_MAP) { | |
1788 | memory_bm_set_bit(dst, pfn); | |
1789 | pfn = memory_bm_next_pfn(src); | |
f577eb30 RW |
1790 | } |
1791 | } | |
1792 | ||
d307c4a8 | 1793 | static int check_header(struct swsusp_info *info) |
f577eb30 | 1794 | { |
d307c4a8 | 1795 | char *reason; |
f577eb30 | 1796 | |
d307c4a8 RW |
1797 | reason = check_image_kernel(info); |
1798 | if (!reason && info->num_physpages != num_physpages) | |
f577eb30 | 1799 | reason = "memory size"; |
f577eb30 | 1800 | if (reason) { |
23976728 | 1801 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
1802 | return -EPERM; |
1803 | } | |
1804 | return 0; | |
1805 | } | |
1806 | ||
1807 | /** | |
1808 | * load header - check the image header and copy data from it | |
1809 | */ | |
1810 | ||
940864dd RW |
1811 | static int |
1812 | load_header(struct swsusp_info *info) | |
f577eb30 RW |
1813 | { |
1814 | int error; | |
f577eb30 | 1815 | |
940864dd | 1816 | restore_pblist = NULL; |
f577eb30 RW |
1817 | error = check_header(info); |
1818 | if (!error) { | |
f577eb30 RW |
1819 | nr_copy_pages = info->image_pages; |
1820 | nr_meta_pages = info->pages - info->image_pages - 1; | |
1821 | } | |
1822 | return error; | |
1823 | } | |
1824 | ||
1825 | /** | |
940864dd RW |
1826 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
1827 | * the corresponding bit in the memory bitmap @bm | |
f577eb30 | 1828 | */ |
69643279 | 1829 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1830 | { |
1831 | int j; | |
1832 | ||
940864dd RW |
1833 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1834 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
1835 | break; | |
1836 | ||
85055dd8 MS |
1837 | /* Extract and buffer page key for data page (s390 only). */ |
1838 | page_key_memorize(buf + j); | |
1839 | ||
69643279 RW |
1840 | if (memory_bm_pfn_present(bm, buf[j])) |
1841 | memory_bm_set_bit(bm, buf[j]); | |
1842 | else | |
1843 | return -EFAULT; | |
f577eb30 | 1844 | } |
69643279 RW |
1845 | |
1846 | return 0; | |
f577eb30 RW |
1847 | } |
1848 | ||
8357376d RW |
1849 | /* List of "safe" pages that may be used to store data loaded from the suspend |
1850 | * image | |
1851 | */ | |
1852 | static struct linked_page *safe_pages_list; | |
1853 | ||
1854 | #ifdef CONFIG_HIGHMEM | |
1855 | /* struct highmem_pbe is used for creating the list of highmem pages that | |
1856 | * should be restored atomically during the resume from disk, because the page | |
1857 | * frames they have occupied before the suspend are in use. | |
1858 | */ | |
1859 | struct highmem_pbe { | |
1860 | struct page *copy_page; /* data is here now */ | |
1861 | struct page *orig_page; /* data was here before the suspend */ | |
1862 | struct highmem_pbe *next; | |
1863 | }; | |
1864 | ||
1865 | /* List of highmem PBEs needed for restoring the highmem pages that were | |
1866 | * allocated before the suspend and included in the suspend image, but have | |
1867 | * also been allocated by the "resume" kernel, so their contents cannot be | |
1868 | * written directly to their "original" page frames. | |
1869 | */ | |
1870 | static struct highmem_pbe *highmem_pblist; | |
1871 | ||
1872 | /** | |
1873 | * count_highmem_image_pages - compute the number of highmem pages in the | |
1874 | * suspend image. The bits in the memory bitmap @bm that correspond to the | |
1875 | * image pages are assumed to be set. | |
1876 | */ | |
1877 | ||
1878 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) | |
1879 | { | |
1880 | unsigned long pfn; | |
1881 | unsigned int cnt = 0; | |
1882 | ||
1883 | memory_bm_position_reset(bm); | |
1884 | pfn = memory_bm_next_pfn(bm); | |
1885 | while (pfn != BM_END_OF_MAP) { | |
1886 | if (PageHighMem(pfn_to_page(pfn))) | |
1887 | cnt++; | |
1888 | ||
1889 | pfn = memory_bm_next_pfn(bm); | |
1890 | } | |
1891 | return cnt; | |
1892 | } | |
1893 | ||
1894 | /** | |
1895 | * prepare_highmem_image - try to allocate as many highmem pages as | |
1896 | * there are highmem image pages (@nr_highmem_p points to the variable | |
1897 | * containing the number of highmem image pages). The pages that are | |
1898 | * "safe" (ie. will not be overwritten when the suspend image is | |
1899 | * restored) have the corresponding bits set in @bm (it must be | |
1900 | * unitialized). | |
1901 | * | |
1902 | * NOTE: This function should not be called if there are no highmem | |
1903 | * image pages. | |
1904 | */ | |
1905 | ||
1906 | static unsigned int safe_highmem_pages; | |
1907 | ||
1908 | static struct memory_bitmap *safe_highmem_bm; | |
1909 | ||
1910 | static int | |
1911 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
1912 | { | |
1913 | unsigned int to_alloc; | |
1914 | ||
1915 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
1916 | return -ENOMEM; | |
1917 | ||
1918 | if (get_highmem_buffer(PG_SAFE)) | |
1919 | return -ENOMEM; | |
1920 | ||
1921 | to_alloc = count_free_highmem_pages(); | |
1922 | if (to_alloc > *nr_highmem_p) | |
1923 | to_alloc = *nr_highmem_p; | |
1924 | else | |
1925 | *nr_highmem_p = to_alloc; | |
1926 | ||
1927 | safe_highmem_pages = 0; | |
1928 | while (to_alloc-- > 0) { | |
1929 | struct page *page; | |
1930 | ||
1931 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 1932 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
1933 | /* The page is "safe", set its bit the bitmap */ |
1934 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1935 | safe_highmem_pages++; | |
1936 | } | |
1937 | /* Mark the page as allocated */ | |
7be98234 RW |
1938 | swsusp_set_page_forbidden(page); |
1939 | swsusp_set_page_free(page); | |
8357376d RW |
1940 | } |
1941 | memory_bm_position_reset(bm); | |
1942 | safe_highmem_bm = bm; | |
1943 | return 0; | |
1944 | } | |
1945 | ||
1946 | /** | |
1947 | * get_highmem_page_buffer - for given highmem image page find the buffer | |
1948 | * that suspend_write_next() should set for its caller to write to. | |
1949 | * | |
1950 | * If the page is to be saved to its "original" page frame or a copy of | |
1951 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
1952 | * the copy of the page is to be made in normal memory, so the address of | |
1953 | * the copy is returned. | |
1954 | * | |
1955 | * If @buffer is returned, the caller of suspend_write_next() will write | |
1956 | * the page's contents to @buffer, so they will have to be copied to the | |
1957 | * right location on the next call to suspend_write_next() and it is done | |
1958 | * with the help of copy_last_highmem_page(). For this purpose, if | |
1959 | * @buffer is returned, @last_highmem page is set to the page to which | |
1960 | * the data will have to be copied from @buffer. | |
1961 | */ | |
1962 | ||
1963 | static struct page *last_highmem_page; | |
1964 | ||
1965 | static void * | |
1966 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
1967 | { | |
1968 | struct highmem_pbe *pbe; | |
1969 | void *kaddr; | |
1970 | ||
7be98234 | 1971 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
8357376d RW |
1972 | /* We have allocated the "original" page frame and we can |
1973 | * use it directly to store the loaded page. | |
1974 | */ | |
1975 | last_highmem_page = page; | |
1976 | return buffer; | |
1977 | } | |
1978 | /* The "original" page frame has not been allocated and we have to | |
1979 | * use a "safe" page frame to store the loaded page. | |
1980 | */ | |
1981 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
1982 | if (!pbe) { | |
1983 | swsusp_free(); | |
69643279 | 1984 | return ERR_PTR(-ENOMEM); |
8357376d RW |
1985 | } |
1986 | pbe->orig_page = page; | |
1987 | if (safe_highmem_pages > 0) { | |
1988 | struct page *tmp; | |
1989 | ||
1990 | /* Copy of the page will be stored in high memory */ | |
1991 | kaddr = buffer; | |
1992 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
1993 | safe_highmem_pages--; | |
1994 | last_highmem_page = tmp; | |
1995 | pbe->copy_page = tmp; | |
1996 | } else { | |
1997 | /* Copy of the page will be stored in normal memory */ | |
1998 | kaddr = safe_pages_list; | |
1999 | safe_pages_list = safe_pages_list->next; | |
2000 | pbe->copy_page = virt_to_page(kaddr); | |
2001 | } | |
2002 | pbe->next = highmem_pblist; | |
2003 | highmem_pblist = pbe; | |
2004 | return kaddr; | |
2005 | } | |
2006 | ||
2007 | /** | |
2008 | * copy_last_highmem_page - copy the contents of a highmem image from | |
2009 | * @buffer, where the caller of snapshot_write_next() has place them, | |
2010 | * to the right location represented by @last_highmem_page . | |
2011 | */ | |
2012 | ||
2013 | static void copy_last_highmem_page(void) | |
2014 | { | |
2015 | if (last_highmem_page) { | |
2016 | void *dst; | |
2017 | ||
0de9a1e2 | 2018 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2019 | copy_page(dst, buffer); |
0de9a1e2 | 2020 | kunmap_atomic(dst); |
8357376d RW |
2021 | last_highmem_page = NULL; |
2022 | } | |
2023 | } | |
2024 | ||
2025 | static inline int last_highmem_page_copied(void) | |
2026 | { | |
2027 | return !last_highmem_page; | |
2028 | } | |
2029 | ||
2030 | static inline void free_highmem_data(void) | |
2031 | { | |
2032 | if (safe_highmem_bm) | |
2033 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2034 | ||
2035 | if (buffer) | |
2036 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2037 | } | |
2038 | #else | |
2039 | static inline int get_safe_write_buffer(void) { return 0; } | |
2040 | ||
2041 | static unsigned int | |
2042 | count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } | |
2043 | ||
2044 | static inline int | |
2045 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
2046 | { | |
2047 | return 0; | |
2048 | } | |
2049 | ||
2050 | static inline void * | |
2051 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
2052 | { | |
69643279 | 2053 | return ERR_PTR(-EINVAL); |
8357376d RW |
2054 | } |
2055 | ||
2056 | static inline void copy_last_highmem_page(void) {} | |
2057 | static inline int last_highmem_page_copied(void) { return 1; } | |
2058 | static inline void free_highmem_data(void) {} | |
2059 | #endif /* CONFIG_HIGHMEM */ | |
2060 | ||
f577eb30 | 2061 | /** |
940864dd RW |
2062 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
2063 | * be overwritten in the process of restoring the system memory state | |
2064 | * from the suspend image ("unsafe" pages) and allocate memory for the | |
2065 | * image. | |
968808b8 | 2066 | * |
940864dd RW |
2067 | * The idea is to allocate a new memory bitmap first and then allocate |
2068 | * as many pages as needed for the image data, but not to assign these | |
2069 | * pages to specific tasks initially. Instead, we just mark them as | |
8357376d RW |
2070 | * allocated and create a lists of "safe" pages that will be used |
2071 | * later. On systems with high memory a list of "safe" highmem pages is | |
2072 | * also created. | |
f577eb30 RW |
2073 | */ |
2074 | ||
940864dd RW |
2075 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2076 | ||
940864dd RW |
2077 | static int |
2078 | prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) | |
f577eb30 | 2079 | { |
8357376d | 2080 | unsigned int nr_pages, nr_highmem; |
940864dd RW |
2081 | struct linked_page *sp_list, *lp; |
2082 | int error; | |
f577eb30 | 2083 | |
8357376d RW |
2084 | /* If there is no highmem, the buffer will not be necessary */ |
2085 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2086 | buffer = NULL; | |
2087 | ||
2088 | nr_highmem = count_highmem_image_pages(bm); | |
940864dd RW |
2089 | error = mark_unsafe_pages(bm); |
2090 | if (error) | |
2091 | goto Free; | |
2092 | ||
2093 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2094 | if (error) | |
2095 | goto Free; | |
2096 | ||
2097 | duplicate_memory_bitmap(new_bm, bm); | |
2098 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2099 | if (nr_highmem > 0) { |
2100 | error = prepare_highmem_image(bm, &nr_highmem); | |
2101 | if (error) | |
2102 | goto Free; | |
2103 | } | |
940864dd RW |
2104 | /* Reserve some safe pages for potential later use. |
2105 | * | |
2106 | * NOTE: This way we make sure there will be enough safe pages for the | |
2107 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2108 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
2109 | */ | |
2110 | sp_list = NULL; | |
2111 | /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */ | |
8357376d | 2112 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2113 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2114 | while (nr_pages > 0) { | |
8357376d | 2115 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2116 | if (!lp) { |
f577eb30 | 2117 | error = -ENOMEM; |
940864dd RW |
2118 | goto Free; |
2119 | } | |
2120 | lp->next = sp_list; | |
2121 | sp_list = lp; | |
2122 | nr_pages--; | |
f577eb30 | 2123 | } |
940864dd RW |
2124 | /* Preallocate memory for the image */ |
2125 | safe_pages_list = NULL; | |
8357376d | 2126 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2127 | while (nr_pages > 0) { |
2128 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2129 | if (!lp) { | |
2130 | error = -ENOMEM; | |
2131 | goto Free; | |
2132 | } | |
7be98234 | 2133 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2134 | /* The page is "safe", add it to the list */ |
2135 | lp->next = safe_pages_list; | |
2136 | safe_pages_list = lp; | |
968808b8 | 2137 | } |
940864dd | 2138 | /* Mark the page as allocated */ |
7be98234 RW |
2139 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2140 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2141 | nr_pages--; |
968808b8 | 2142 | } |
940864dd RW |
2143 | /* Free the reserved safe pages so that chain_alloc() can use them */ |
2144 | while (sp_list) { | |
2145 | lp = sp_list->next; | |
2146 | free_image_page(sp_list, PG_UNSAFE_CLEAR); | |
2147 | sp_list = lp; | |
f577eb30 | 2148 | } |
940864dd RW |
2149 | return 0; |
2150 | ||
59a49335 | 2151 | Free: |
940864dd | 2152 | swsusp_free(); |
f577eb30 RW |
2153 | return error; |
2154 | } | |
2155 | ||
940864dd RW |
2156 | /** |
2157 | * get_buffer - compute the address that snapshot_write_next() should | |
2158 | * set for its caller to write to. | |
2159 | */ | |
2160 | ||
2161 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) | |
968808b8 | 2162 | { |
940864dd | 2163 | struct pbe *pbe; |
69643279 RW |
2164 | struct page *page; |
2165 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2166 | |
69643279 RW |
2167 | if (pfn == BM_END_OF_MAP) |
2168 | return ERR_PTR(-EFAULT); | |
2169 | ||
2170 | page = pfn_to_page(pfn); | |
8357376d RW |
2171 | if (PageHighMem(page)) |
2172 | return get_highmem_page_buffer(page, ca); | |
2173 | ||
7be98234 | 2174 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
940864dd RW |
2175 | /* We have allocated the "original" page frame and we can |
2176 | * use it directly to store the loaded page. | |
968808b8 | 2177 | */ |
940864dd RW |
2178 | return page_address(page); |
2179 | ||
2180 | /* The "original" page frame has not been allocated and we have to | |
2181 | * use a "safe" page frame to store the loaded page. | |
968808b8 | 2182 | */ |
940864dd RW |
2183 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2184 | if (!pbe) { | |
2185 | swsusp_free(); | |
69643279 | 2186 | return ERR_PTR(-ENOMEM); |
940864dd | 2187 | } |
8357376d RW |
2188 | pbe->orig_address = page_address(page); |
2189 | pbe->address = safe_pages_list; | |
940864dd RW |
2190 | safe_pages_list = safe_pages_list->next; |
2191 | pbe->next = restore_pblist; | |
2192 | restore_pblist = pbe; | |
8357376d | 2193 | return pbe->address; |
968808b8 RW |
2194 | } |
2195 | ||
f577eb30 RW |
2196 | /** |
2197 | * snapshot_write_next - used for writing the system memory snapshot. | |
2198 | * | |
2199 | * On the first call to it @handle should point to a zeroed | |
2200 | * snapshot_handle structure. The structure gets updated and a pointer | |
2201 | * to it should be passed to this function every next time. | |
2202 | * | |
f577eb30 RW |
2203 | * On success the function returns a positive number. Then, the caller |
2204 | * is allowed to write up to the returned number of bytes to the memory | |
d3c1b24c | 2205 | * location computed by the data_of() macro. |
f577eb30 RW |
2206 | * |
2207 | * The function returns 0 to indicate the "end of file" condition, | |
2208 | * and a negative number is returned on error. In such cases the | |
2209 | * structure pointed to by @handle is not updated and should not be used | |
2210 | * any more. | |
2211 | */ | |
2212 | ||
d3c1b24c | 2213 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2214 | { |
940864dd | 2215 | static struct chain_allocator ca; |
f577eb30 RW |
2216 | int error = 0; |
2217 | ||
940864dd | 2218 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2219 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2220 | return 0; |
940864dd | 2221 | |
d3c1b24c JS |
2222 | handle->sync_read = 1; |
2223 | ||
2224 | if (!handle->cur) { | |
8357376d RW |
2225 | if (!buffer) |
2226 | /* This makes the buffer be freed by swsusp_free() */ | |
2227 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2228 | ||
f577eb30 RW |
2229 | if (!buffer) |
2230 | return -ENOMEM; | |
8357376d | 2231 | |
f577eb30 | 2232 | handle->buffer = buffer; |
d3c1b24c JS |
2233 | } else if (handle->cur == 1) { |
2234 | error = load_header(buffer); | |
2235 | if (error) | |
2236 | return error; | |
940864dd | 2237 | |
d3c1b24c JS |
2238 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2239 | if (error) | |
2240 | return error; | |
2241 | ||
85055dd8 MS |
2242 | /* Allocate buffer for page keys. */ |
2243 | error = page_key_alloc(nr_copy_pages); | |
2244 | if (error) | |
2245 | return error; | |
2246 | ||
d3c1b24c JS |
2247 | } else if (handle->cur <= nr_meta_pages + 1) { |
2248 | error = unpack_orig_pfns(buffer, ©_bm); | |
2249 | if (error) | |
2250 | return error; | |
940864dd | 2251 | |
d3c1b24c JS |
2252 | if (handle->cur == nr_meta_pages + 1) { |
2253 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2254 | if (error) |
2255 | return error; | |
2256 | ||
d3c1b24c JS |
2257 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2258 | memory_bm_position_reset(&orig_bm); | |
2259 | restore_pblist = NULL; | |
940864dd | 2260 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2261 | handle->sync_read = 0; |
69643279 RW |
2262 | if (IS_ERR(handle->buffer)) |
2263 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2264 | } |
f577eb30 | 2265 | } else { |
d3c1b24c | 2266 | copy_last_highmem_page(); |
85055dd8 MS |
2267 | /* Restore page key for data page (s390 only). */ |
2268 | page_key_write(handle->buffer); | |
d3c1b24c JS |
2269 | handle->buffer = get_buffer(&orig_bm, &ca); |
2270 | if (IS_ERR(handle->buffer)) | |
2271 | return PTR_ERR(handle->buffer); | |
2272 | if (handle->buffer != buffer) | |
2273 | handle->sync_read = 0; | |
f577eb30 | 2274 | } |
d3c1b24c JS |
2275 | handle->cur++; |
2276 | return PAGE_SIZE; | |
f577eb30 RW |
2277 | } |
2278 | ||
8357376d RW |
2279 | /** |
2280 | * snapshot_write_finalize - must be called after the last call to | |
2281 | * snapshot_write_next() in case the last page in the image happens | |
2282 | * to be a highmem page and its contents should be stored in the | |
2283 | * highmem. Additionally, it releases the memory that will not be | |
2284 | * used any more. | |
2285 | */ | |
2286 | ||
2287 | void snapshot_write_finalize(struct snapshot_handle *handle) | |
2288 | { | |
2289 | copy_last_highmem_page(); | |
85055dd8 MS |
2290 | /* Restore page key for data page (s390 only). */ |
2291 | page_key_write(handle->buffer); | |
2292 | page_key_free(); | |
8357376d | 2293 | /* Free only if we have loaded the image entirely */ |
d3c1b24c | 2294 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
8357376d RW |
2295 | memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR); |
2296 | free_highmem_data(); | |
2297 | } | |
2298 | } | |
2299 | ||
f577eb30 RW |
2300 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2301 | { | |
8357376d | 2302 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2303 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2304 | } | |
2305 | ||
8357376d RW |
2306 | #ifdef CONFIG_HIGHMEM |
2307 | /* Assumes that @buf is ready and points to a "safe" page */ | |
2308 | static inline void | |
2309 | swap_two_pages_data(struct page *p1, struct page *p2, void *buf) | |
940864dd | 2310 | { |
8357376d RW |
2311 | void *kaddr1, *kaddr2; |
2312 | ||
0de9a1e2 CW |
2313 | kaddr1 = kmap_atomic(p1); |
2314 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2315 | copy_page(buf, kaddr1); |
2316 | copy_page(kaddr1, kaddr2); | |
2317 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2318 | kunmap_atomic(kaddr2); |
2319 | kunmap_atomic(kaddr1); | |
8357376d RW |
2320 | } |
2321 | ||
2322 | /** | |
2323 | * restore_highmem - for each highmem page that was allocated before | |
2324 | * the suspend and included in the suspend image, and also has been | |
2325 | * allocated by the "resume" kernel swap its current (ie. "before | |
2326 | * resume") contents with the previous (ie. "before suspend") one. | |
2327 | * | |
2328 | * If the resume eventually fails, we can call this function once | |
2329 | * again and restore the "before resume" highmem state. | |
2330 | */ | |
2331 | ||
2332 | int restore_highmem(void) | |
2333 | { | |
2334 | struct highmem_pbe *pbe = highmem_pblist; | |
2335 | void *buf; | |
2336 | ||
2337 | if (!pbe) | |
2338 | return 0; | |
2339 | ||
2340 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2341 | if (!buf) | |
2342 | return -ENOMEM; | |
2343 | ||
2344 | while (pbe) { | |
2345 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2346 | pbe = pbe->next; | |
2347 | } | |
2348 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2349 | return 0; | |
f577eb30 | 2350 | } |
8357376d | 2351 | #endif /* CONFIG_HIGHMEM */ |