]>
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 | ||
fe419535 RW |
43 | /* |
44 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
45 | * When it is set to N, swsusp will do its best to ensure the image | |
46 | * size will not exceed N bytes, but if that is impossible, it will | |
47 | * try to create the smallest image possible. | |
48 | */ | |
ac5c24ec RW |
49 | unsigned long image_size; |
50 | ||
51 | void __init hibernate_image_size_init(void) | |
52 | { | |
53 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; | |
54 | } | |
fe419535 | 55 | |
8357376d RW |
56 | /* List of PBEs needed for restoring the pages that were allocated before |
57 | * the suspend and included in the suspend image, but have also been | |
58 | * allocated by the "resume" kernel, so their contents cannot be written | |
59 | * directly to their "original" page frames. | |
60 | */ | |
75534b50 RW |
61 | struct pbe *restore_pblist; |
62 | ||
8357376d | 63 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 64 | static void *buffer; |
7088a5c0 | 65 | |
f6143aa6 RW |
66 | /** |
67 | * @safe_needed - on resume, for storing the PBE list and the image, | |
68 | * we can only use memory pages that do not conflict with the pages | |
8357376d RW |
69 | * used before suspend. The unsafe pages have PageNosaveFree set |
70 | * and we count them using unsafe_pages. | |
f6143aa6 | 71 | * |
8357376d RW |
72 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
73 | * so that swsusp_free() can release it. | |
f6143aa6 RW |
74 | */ |
75 | ||
0bcd888d RW |
76 | #define PG_ANY 0 |
77 | #define PG_SAFE 1 | |
78 | #define PG_UNSAFE_CLEAR 1 | |
79 | #define PG_UNSAFE_KEEP 0 | |
80 | ||
940864dd | 81 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 82 | |
8357376d | 83 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
84 | { |
85 | void *res; | |
86 | ||
87 | res = (void *)get_zeroed_page(gfp_mask); | |
88 | if (safe_needed) | |
7be98234 | 89 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 90 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 91 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 92 | allocated_unsafe_pages++; |
f6143aa6 RW |
93 | res = (void *)get_zeroed_page(gfp_mask); |
94 | } | |
95 | if (res) { | |
7be98234 RW |
96 | swsusp_set_page_forbidden(virt_to_page(res)); |
97 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
98 | } |
99 | return res; | |
100 | } | |
101 | ||
102 | unsigned long get_safe_page(gfp_t gfp_mask) | |
103 | { | |
8357376d RW |
104 | return (unsigned long)get_image_page(gfp_mask, PG_SAFE); |
105 | } | |
106 | ||
5b6d15de RW |
107 | static struct page *alloc_image_page(gfp_t gfp_mask) |
108 | { | |
8357376d RW |
109 | struct page *page; |
110 | ||
111 | page = alloc_page(gfp_mask); | |
112 | if (page) { | |
7be98234 RW |
113 | swsusp_set_page_forbidden(page); |
114 | swsusp_set_page_free(page); | |
8357376d RW |
115 | } |
116 | return page; | |
f6143aa6 RW |
117 | } |
118 | ||
119 | /** | |
120 | * free_image_page - free page represented by @addr, allocated with | |
8357376d | 121 | * get_image_page (page flags set by it must be cleared) |
f6143aa6 RW |
122 | */ |
123 | ||
124 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
125 | { | |
8357376d RW |
126 | struct page *page; |
127 | ||
128 | BUG_ON(!virt_addr_valid(addr)); | |
129 | ||
130 | page = virt_to_page(addr); | |
131 | ||
7be98234 | 132 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 133 | if (clear_nosave_free) |
7be98234 | 134 | swsusp_unset_page_free(page); |
8357376d RW |
135 | |
136 | __free_page(page); | |
f6143aa6 RW |
137 | } |
138 | ||
b788db79 RW |
139 | /* struct linked_page is used to build chains of pages */ |
140 | ||
141 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
142 | ||
143 | struct linked_page { | |
144 | struct linked_page *next; | |
145 | char data[LINKED_PAGE_DATA_SIZE]; | |
146 | } __attribute__((packed)); | |
147 | ||
148 | static inline void | |
149 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) | |
150 | { | |
151 | while (list) { | |
152 | struct linked_page *lp = list->next; | |
153 | ||
154 | free_image_page(list, clear_page_nosave); | |
155 | list = lp; | |
156 | } | |
157 | } | |
158 | ||
159 | /** | |
160 | * struct chain_allocator is used for allocating small objects out of | |
161 | * a linked list of pages called 'the chain'. | |
162 | * | |
163 | * The chain grows each time when there is no room for a new object in | |
164 | * the current page. The allocated objects cannot be freed individually. | |
165 | * It is only possible to free them all at once, by freeing the entire | |
166 | * chain. | |
167 | * | |
168 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
169 | * are not much smaller than PAGE_SIZE. | |
170 | */ | |
171 | ||
172 | struct chain_allocator { | |
173 | struct linked_page *chain; /* the chain */ | |
174 | unsigned int used_space; /* total size of objects allocated out | |
175 | * of the current page | |
176 | */ | |
177 | gfp_t gfp_mask; /* mask for allocating pages */ | |
178 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
179 | }; | |
180 | ||
181 | static void | |
182 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) | |
183 | { | |
184 | ca->chain = NULL; | |
185 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
186 | ca->gfp_mask = gfp_mask; | |
187 | ca->safe_needed = safe_needed; | |
188 | } | |
189 | ||
190 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
191 | { | |
192 | void *ret; | |
193 | ||
194 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
195 | struct linked_page *lp; | |
196 | ||
8357376d | 197 | lp = get_image_page(ca->gfp_mask, ca->safe_needed); |
b788db79 RW |
198 | if (!lp) |
199 | return NULL; | |
200 | ||
201 | lp->next = ca->chain; | |
202 | ca->chain = lp; | |
203 | ca->used_space = 0; | |
204 | } | |
205 | ret = ca->chain->data + ca->used_space; | |
206 | ca->used_space += size; | |
207 | return ret; | |
208 | } | |
209 | ||
b788db79 RW |
210 | /** |
211 | * Data types related to memory bitmaps. | |
212 | * | |
213 | * Memory bitmap is a structure consiting of many linked lists of | |
214 | * objects. The main list's elements are of type struct zone_bitmap | |
215 | * and each of them corresonds to one zone. For each zone bitmap | |
216 | * object there is a list of objects of type struct bm_block that | |
0d83304c | 217 | * represent each blocks of bitmap in which information is stored. |
b788db79 RW |
218 | * |
219 | * struct memory_bitmap contains a pointer to the main list of zone | |
220 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
221 | * and a pointer to the list of pages used for allocating all of the | |
222 | * zone bitmap objects and bitmap block objects. | |
223 | * | |
224 | * NOTE: It has to be possible to lay out the bitmap in memory | |
225 | * using only allocations of order 0. Additionally, the bitmap is | |
226 | * designed to work with arbitrary number of zones (this is over the | |
227 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
228 | * | |
229 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
230 | * objects and a pointer to the bitmap block object that has been | |
231 | * most recently used for setting bits. Additionally, it contains the | |
232 | * pfns that correspond to the start and end of the represented zone. | |
233 | * | |
234 | * struct bm_block contains a pointer to the memory page in which | |
0d83304c AM |
235 | * information is stored (in the form of a block of bitmap) |
236 | * It also contains the pfns that correspond to the start and end of | |
237 | * the represented memory area. | |
b788db79 RW |
238 | */ |
239 | ||
240 | #define BM_END_OF_MAP (~0UL) | |
241 | ||
8de03073 | 242 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
b788db79 RW |
243 | |
244 | struct bm_block { | |
846705de | 245 | struct list_head hook; /* hook into a list of bitmap blocks */ |
b788db79 RW |
246 | unsigned long start_pfn; /* pfn represented by the first bit */ |
247 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ | |
0d83304c | 248 | unsigned long *data; /* bitmap representing pages */ |
b788db79 RW |
249 | }; |
250 | ||
0d83304c AM |
251 | static inline unsigned long bm_block_bits(struct bm_block *bb) |
252 | { | |
253 | return bb->end_pfn - bb->start_pfn; | |
254 | } | |
255 | ||
b788db79 RW |
256 | /* strcut bm_position is used for browsing memory bitmaps */ |
257 | ||
258 | struct bm_position { | |
b788db79 | 259 | struct bm_block *block; |
b788db79 RW |
260 | int bit; |
261 | }; | |
262 | ||
263 | struct memory_bitmap { | |
846705de | 264 | struct list_head blocks; /* list of bitmap blocks */ |
b788db79 RW |
265 | struct linked_page *p_list; /* list of pages used to store zone |
266 | * bitmap objects and bitmap block | |
267 | * objects | |
268 | */ | |
269 | struct bm_position cur; /* most recently used bit position */ | |
270 | }; | |
271 | ||
272 | /* Functions that operate on memory bitmaps */ | |
273 | ||
b788db79 RW |
274 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
275 | { | |
846705de | 276 | bm->cur.block = list_entry(bm->blocks.next, struct bm_block, hook); |
0d83304c | 277 | bm->cur.bit = 0; |
b788db79 RW |
278 | } |
279 | ||
280 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
281 | ||
282 | /** | |
283 | * create_bm_block_list - create a list of block bitmap objects | |
8de03073 | 284 | * @pages - number of pages to track |
846705de RW |
285 | * @list - list to put the allocated blocks into |
286 | * @ca - chain allocator to be used for allocating memory | |
b788db79 | 287 | */ |
846705de RW |
288 | static int create_bm_block_list(unsigned long pages, |
289 | struct list_head *list, | |
290 | struct chain_allocator *ca) | |
b788db79 | 291 | { |
846705de | 292 | unsigned int nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); |
b788db79 RW |
293 | |
294 | while (nr_blocks-- > 0) { | |
295 | struct bm_block *bb; | |
296 | ||
297 | bb = chain_alloc(ca, sizeof(struct bm_block)); | |
298 | if (!bb) | |
846705de RW |
299 | return -ENOMEM; |
300 | list_add(&bb->hook, list); | |
b788db79 | 301 | } |
846705de RW |
302 | |
303 | return 0; | |
b788db79 RW |
304 | } |
305 | ||
846705de RW |
306 | struct mem_extent { |
307 | struct list_head hook; | |
308 | unsigned long start; | |
309 | unsigned long end; | |
310 | }; | |
311 | ||
b788db79 | 312 | /** |
846705de RW |
313 | * free_mem_extents - free a list of memory extents |
314 | * @list - list of extents to empty | |
b788db79 | 315 | */ |
846705de RW |
316 | static void free_mem_extents(struct list_head *list) |
317 | { | |
318 | struct mem_extent *ext, *aux; | |
b788db79 | 319 | |
846705de RW |
320 | list_for_each_entry_safe(ext, aux, list, hook) { |
321 | list_del(&ext->hook); | |
322 | kfree(ext); | |
323 | } | |
324 | } | |
325 | ||
326 | /** | |
327 | * create_mem_extents - create a list of memory extents representing | |
328 | * contiguous ranges of PFNs | |
329 | * @list - list to put the extents into | |
330 | * @gfp_mask - mask to use for memory allocations | |
331 | */ | |
332 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 333 | { |
846705de | 334 | struct zone *zone; |
b788db79 | 335 | |
846705de | 336 | INIT_LIST_HEAD(list); |
b788db79 | 337 | |
ee99c71c | 338 | for_each_populated_zone(zone) { |
846705de RW |
339 | unsigned long zone_start, zone_end; |
340 | struct mem_extent *ext, *cur, *aux; | |
341 | ||
846705de RW |
342 | zone_start = zone->zone_start_pfn; |
343 | zone_end = zone->zone_start_pfn + zone->spanned_pages; | |
344 | ||
345 | list_for_each_entry(ext, list, hook) | |
346 | if (zone_start <= ext->end) | |
347 | break; | |
b788db79 | 348 | |
846705de RW |
349 | if (&ext->hook == list || zone_end < ext->start) { |
350 | /* New extent is necessary */ | |
351 | struct mem_extent *new_ext; | |
352 | ||
353 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
354 | if (!new_ext) { | |
355 | free_mem_extents(list); | |
356 | return -ENOMEM; | |
357 | } | |
358 | new_ext->start = zone_start; | |
359 | new_ext->end = zone_end; | |
360 | list_add_tail(&new_ext->hook, &ext->hook); | |
361 | continue; | |
362 | } | |
363 | ||
364 | /* Merge this zone's range of PFNs with the existing one */ | |
365 | if (zone_start < ext->start) | |
366 | ext->start = zone_start; | |
367 | if (zone_end > ext->end) | |
368 | ext->end = zone_end; | |
369 | ||
370 | /* More merging may be possible */ | |
371 | cur = ext; | |
372 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
373 | if (zone_end < cur->start) | |
374 | break; | |
375 | if (zone_end < cur->end) | |
376 | ext->end = cur->end; | |
377 | list_del(&cur->hook); | |
378 | kfree(cur); | |
379 | } | |
b788db79 | 380 | } |
846705de RW |
381 | |
382 | return 0; | |
b788db79 RW |
383 | } |
384 | ||
385 | /** | |
386 | * memory_bm_create - allocate memory for a memory bitmap | |
387 | */ | |
b788db79 RW |
388 | static int |
389 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) | |
390 | { | |
391 | struct chain_allocator ca; | |
846705de RW |
392 | struct list_head mem_extents; |
393 | struct mem_extent *ext; | |
394 | int error; | |
b788db79 RW |
395 | |
396 | chain_init(&ca, gfp_mask, safe_needed); | |
846705de | 397 | INIT_LIST_HEAD(&bm->blocks); |
b788db79 | 398 | |
846705de RW |
399 | error = create_mem_extents(&mem_extents, gfp_mask); |
400 | if (error) | |
401 | return error; | |
b788db79 | 402 | |
846705de RW |
403 | list_for_each_entry(ext, &mem_extents, hook) { |
404 | struct bm_block *bb; | |
405 | unsigned long pfn = ext->start; | |
406 | unsigned long pages = ext->end - ext->start; | |
b788db79 | 407 | |
846705de | 408 | bb = list_entry(bm->blocks.prev, struct bm_block, hook); |
b788db79 | 409 | |
846705de RW |
410 | error = create_bm_block_list(pages, bm->blocks.prev, &ca); |
411 | if (error) | |
412 | goto Error; | |
b788db79 | 413 | |
846705de RW |
414 | list_for_each_entry_continue(bb, &bm->blocks, hook) { |
415 | bb->data = get_image_page(gfp_mask, safe_needed); | |
416 | if (!bb->data) { | |
417 | error = -ENOMEM; | |
418 | goto Error; | |
419 | } | |
b788db79 RW |
420 | |
421 | bb->start_pfn = pfn; | |
846705de | 422 | if (pages >= BM_BITS_PER_BLOCK) { |
b788db79 | 423 | pfn += BM_BITS_PER_BLOCK; |
846705de | 424 | pages -= BM_BITS_PER_BLOCK; |
b788db79 RW |
425 | } else { |
426 | /* This is executed only once in the loop */ | |
846705de | 427 | pfn += pages; |
b788db79 RW |
428 | } |
429 | bb->end_pfn = pfn; | |
b788db79 | 430 | } |
b788db79 | 431 | } |
846705de | 432 | |
b788db79 RW |
433 | bm->p_list = ca.chain; |
434 | memory_bm_position_reset(bm); | |
846705de RW |
435 | Exit: |
436 | free_mem_extents(&mem_extents); | |
437 | return error; | |
b788db79 | 438 | |
846705de | 439 | Error: |
b788db79 RW |
440 | bm->p_list = ca.chain; |
441 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 442 | goto Exit; |
b788db79 RW |
443 | } |
444 | ||
445 | /** | |
446 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
447 | */ | |
b788db79 RW |
448 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
449 | { | |
846705de | 450 | struct bm_block *bb; |
b788db79 | 451 | |
846705de RW |
452 | list_for_each_entry(bb, &bm->blocks, hook) |
453 | if (bb->data) | |
454 | free_image_page(bb->data, clear_nosave_free); | |
b788db79 | 455 | |
b788db79 | 456 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de RW |
457 | |
458 | INIT_LIST_HEAD(&bm->blocks); | |
b788db79 RW |
459 | } |
460 | ||
461 | /** | |
74dfd666 | 462 | * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds |
b788db79 RW |
463 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member |
464 | * of @bm->cur_zone_bm are updated. | |
b788db79 | 465 | */ |
a82f7119 | 466 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
74dfd666 | 467 | void **addr, unsigned int *bit_nr) |
b788db79 | 468 | { |
b788db79 RW |
469 | struct bm_block *bb; |
470 | ||
846705de RW |
471 | /* |
472 | * Check if the pfn corresponds to the current bitmap block and find | |
473 | * the block where it fits if this is not the case. | |
474 | */ | |
475 | bb = bm->cur.block; | |
b788db79 | 476 | if (pfn < bb->start_pfn) |
846705de RW |
477 | list_for_each_entry_continue_reverse(bb, &bm->blocks, hook) |
478 | if (pfn >= bb->start_pfn) | |
479 | break; | |
b788db79 | 480 | |
846705de RW |
481 | if (pfn >= bb->end_pfn) |
482 | list_for_each_entry_continue(bb, &bm->blocks, hook) | |
483 | if (pfn >= bb->start_pfn && pfn < bb->end_pfn) | |
484 | break; | |
74dfd666 | 485 | |
846705de RW |
486 | if (&bb->hook == &bm->blocks) |
487 | return -EFAULT; | |
488 | ||
489 | /* The block has been found */ | |
490 | bm->cur.block = bb; | |
b788db79 | 491 | pfn -= bb->start_pfn; |
846705de | 492 | bm->cur.bit = pfn + 1; |
0d83304c AM |
493 | *bit_nr = pfn; |
494 | *addr = bb->data; | |
a82f7119 | 495 | return 0; |
74dfd666 RW |
496 | } |
497 | ||
498 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) | |
499 | { | |
500 | void *addr; | |
501 | unsigned int bit; | |
a82f7119 | 502 | int error; |
74dfd666 | 503 | |
a82f7119 RW |
504 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
505 | BUG_ON(error); | |
74dfd666 RW |
506 | set_bit(bit, addr); |
507 | } | |
508 | ||
a82f7119 RW |
509 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
510 | { | |
511 | void *addr; | |
512 | unsigned int bit; | |
513 | int error; | |
514 | ||
515 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
516 | if (!error) | |
517 | set_bit(bit, addr); | |
518 | return error; | |
519 | } | |
520 | ||
74dfd666 RW |
521 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
522 | { | |
523 | void *addr; | |
524 | unsigned int bit; | |
a82f7119 | 525 | int error; |
74dfd666 | 526 | |
a82f7119 RW |
527 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
528 | BUG_ON(error); | |
74dfd666 RW |
529 | clear_bit(bit, addr); |
530 | } | |
531 | ||
532 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) | |
533 | { | |
534 | void *addr; | |
535 | unsigned int bit; | |
a82f7119 | 536 | int error; |
74dfd666 | 537 | |
a82f7119 RW |
538 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
539 | BUG_ON(error); | |
74dfd666 | 540 | return test_bit(bit, addr); |
b788db79 RW |
541 | } |
542 | ||
69643279 RW |
543 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
544 | { | |
545 | void *addr; | |
546 | unsigned int bit; | |
547 | ||
548 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); | |
549 | } | |
550 | ||
b788db79 RW |
551 | /** |
552 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit | |
553 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is | |
554 | * returned. | |
555 | * | |
556 | * It is required to run memory_bm_position_reset() before the first call to | |
557 | * this function. | |
558 | */ | |
559 | ||
560 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) | |
561 | { | |
b788db79 | 562 | struct bm_block *bb; |
b788db79 RW |
563 | int bit; |
564 | ||
846705de | 565 | bb = bm->cur.block; |
b788db79 | 566 | do { |
846705de RW |
567 | bit = bm->cur.bit; |
568 | bit = find_next_bit(bb->data, bm_block_bits(bb), bit); | |
569 | if (bit < bm_block_bits(bb)) | |
570 | goto Return_pfn; | |
571 | ||
572 | bb = list_entry(bb->hook.next, struct bm_block, hook); | |
573 | bm->cur.block = bb; | |
574 | bm->cur.bit = 0; | |
575 | } while (&bb->hook != &bm->blocks); | |
576 | ||
b788db79 RW |
577 | memory_bm_position_reset(bm); |
578 | return BM_END_OF_MAP; | |
579 | ||
59a49335 | 580 | Return_pfn: |
0d83304c AM |
581 | bm->cur.bit = bit + 1; |
582 | return bb->start_pfn + bit; | |
b788db79 RW |
583 | } |
584 | ||
74dfd666 RW |
585 | /** |
586 | * This structure represents a range of page frames the contents of which | |
587 | * should not be saved during the suspend. | |
588 | */ | |
589 | ||
590 | struct nosave_region { | |
591 | struct list_head list; | |
592 | unsigned long start_pfn; | |
593 | unsigned long end_pfn; | |
594 | }; | |
595 | ||
596 | static LIST_HEAD(nosave_regions); | |
597 | ||
598 | /** | |
599 | * register_nosave_region - register a range of page frames the contents | |
600 | * of which should not be saved during the suspend (to be used in the early | |
601 | * initialization code) | |
602 | */ | |
603 | ||
604 | void __init | |
940d67f6 JB |
605 | __register_nosave_region(unsigned long start_pfn, unsigned long end_pfn, |
606 | int use_kmalloc) | |
74dfd666 RW |
607 | { |
608 | struct nosave_region *region; | |
609 | ||
610 | if (start_pfn >= end_pfn) | |
611 | return; | |
612 | ||
613 | if (!list_empty(&nosave_regions)) { | |
614 | /* Try to extend the previous region (they should be sorted) */ | |
615 | region = list_entry(nosave_regions.prev, | |
616 | struct nosave_region, list); | |
617 | if (region->end_pfn == start_pfn) { | |
618 | region->end_pfn = end_pfn; | |
619 | goto Report; | |
620 | } | |
621 | } | |
940d67f6 JB |
622 | if (use_kmalloc) { |
623 | /* during init, this shouldn't fail */ | |
624 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); | |
625 | BUG_ON(!region); | |
626 | } else | |
627 | /* This allocation cannot fail */ | |
3c1596ef | 628 | region = alloc_bootmem(sizeof(struct nosave_region)); |
74dfd666 RW |
629 | region->start_pfn = start_pfn; |
630 | region->end_pfn = end_pfn; | |
631 | list_add_tail(®ion->list, &nosave_regions); | |
632 | Report: | |
23976728 | 633 | printk(KERN_INFO "PM: Registered nosave memory: %016lx - %016lx\n", |
74dfd666 RW |
634 | start_pfn << PAGE_SHIFT, end_pfn << PAGE_SHIFT); |
635 | } | |
636 | ||
637 | /* | |
638 | * Set bits in this map correspond to the page frames the contents of which | |
639 | * should not be saved during the suspend. | |
640 | */ | |
641 | static struct memory_bitmap *forbidden_pages_map; | |
642 | ||
643 | /* Set bits in this map correspond to free page frames. */ | |
644 | static struct memory_bitmap *free_pages_map; | |
645 | ||
646 | /* | |
647 | * Each page frame allocated for creating the image is marked by setting the | |
648 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
649 | */ | |
650 | ||
651 | void swsusp_set_page_free(struct page *page) | |
652 | { | |
653 | if (free_pages_map) | |
654 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
655 | } | |
656 | ||
657 | static int swsusp_page_is_free(struct page *page) | |
658 | { | |
659 | return free_pages_map ? | |
660 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
661 | } | |
662 | ||
663 | void swsusp_unset_page_free(struct page *page) | |
664 | { | |
665 | if (free_pages_map) | |
666 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
667 | } | |
668 | ||
669 | static void swsusp_set_page_forbidden(struct page *page) | |
670 | { | |
671 | if (forbidden_pages_map) | |
672 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
673 | } | |
674 | ||
675 | int swsusp_page_is_forbidden(struct page *page) | |
676 | { | |
677 | return forbidden_pages_map ? | |
678 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
679 | } | |
680 | ||
681 | static void swsusp_unset_page_forbidden(struct page *page) | |
682 | { | |
683 | if (forbidden_pages_map) | |
684 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
685 | } | |
686 | ||
687 | /** | |
688 | * mark_nosave_pages - set bits corresponding to the page frames the | |
689 | * contents of which should not be saved in a given bitmap. | |
690 | */ | |
691 | ||
692 | static void mark_nosave_pages(struct memory_bitmap *bm) | |
693 | { | |
694 | struct nosave_region *region; | |
695 | ||
696 | if (list_empty(&nosave_regions)) | |
697 | return; | |
698 | ||
699 | list_for_each_entry(region, &nosave_regions, list) { | |
700 | unsigned long pfn; | |
701 | ||
23976728 | 702 | pr_debug("PM: Marking nosave pages: %016lx - %016lx\n", |
74dfd666 RW |
703 | region->start_pfn << PAGE_SHIFT, |
704 | region->end_pfn << PAGE_SHIFT); | |
705 | ||
706 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
707 | if (pfn_valid(pfn)) { |
708 | /* | |
709 | * It is safe to ignore the result of | |
710 | * mem_bm_set_bit_check() here, since we won't | |
711 | * touch the PFNs for which the error is | |
712 | * returned anyway. | |
713 | */ | |
714 | mem_bm_set_bit_check(bm, pfn); | |
715 | } | |
74dfd666 RW |
716 | } |
717 | } | |
718 | ||
719 | /** | |
720 | * create_basic_memory_bitmaps - create bitmaps needed for marking page | |
721 | * frames that should not be saved and free page frames. The pointers | |
722 | * forbidden_pages_map and free_pages_map are only modified if everything | |
723 | * goes well, because we don't want the bits to be used before both bitmaps | |
724 | * are set up. | |
725 | */ | |
726 | ||
727 | int create_basic_memory_bitmaps(void) | |
728 | { | |
729 | struct memory_bitmap *bm1, *bm2; | |
730 | int error = 0; | |
731 | ||
732 | BUG_ON(forbidden_pages_map || free_pages_map); | |
733 | ||
0709db60 | 734 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
735 | if (!bm1) |
736 | return -ENOMEM; | |
737 | ||
0709db60 | 738 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
739 | if (error) |
740 | goto Free_first_object; | |
741 | ||
0709db60 | 742 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
743 | if (!bm2) |
744 | goto Free_first_bitmap; | |
745 | ||
0709db60 | 746 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
747 | if (error) |
748 | goto Free_second_object; | |
749 | ||
750 | forbidden_pages_map = bm1; | |
751 | free_pages_map = bm2; | |
752 | mark_nosave_pages(forbidden_pages_map); | |
753 | ||
23976728 | 754 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
755 | |
756 | return 0; | |
757 | ||
758 | Free_second_object: | |
759 | kfree(bm2); | |
760 | Free_first_bitmap: | |
761 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
762 | Free_first_object: | |
763 | kfree(bm1); | |
764 | return -ENOMEM; | |
765 | } | |
766 | ||
767 | /** | |
768 | * free_basic_memory_bitmaps - free memory bitmaps allocated by | |
769 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary | |
770 | * so that the bitmaps themselves are not referred to while they are being | |
771 | * freed. | |
772 | */ | |
773 | ||
774 | void free_basic_memory_bitmaps(void) | |
775 | { | |
776 | struct memory_bitmap *bm1, *bm2; | |
777 | ||
778 | BUG_ON(!(forbidden_pages_map && free_pages_map)); | |
779 | ||
780 | bm1 = forbidden_pages_map; | |
781 | bm2 = free_pages_map; | |
782 | forbidden_pages_map = NULL; | |
783 | free_pages_map = NULL; | |
784 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
785 | kfree(bm1); | |
786 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
787 | kfree(bm2); | |
788 | ||
23976728 | 789 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
790 | } |
791 | ||
b788db79 RW |
792 | /** |
793 | * snapshot_additional_pages - estimate the number of additional pages | |
794 | * be needed for setting up the suspend image data structures for given | |
795 | * zone (usually the returned value is greater than the exact number) | |
796 | */ | |
797 | ||
798 | unsigned int snapshot_additional_pages(struct zone *zone) | |
799 | { | |
800 | unsigned int res; | |
801 | ||
802 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
803 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), PAGE_SIZE); | |
8357376d | 804 | return 2 * res; |
b788db79 RW |
805 | } |
806 | ||
8357376d RW |
807 | #ifdef CONFIG_HIGHMEM |
808 | /** | |
809 | * count_free_highmem_pages - compute the total number of free highmem | |
810 | * pages, system-wide. | |
811 | */ | |
812 | ||
813 | static unsigned int count_free_highmem_pages(void) | |
814 | { | |
815 | struct zone *zone; | |
816 | unsigned int cnt = 0; | |
817 | ||
ee99c71c KM |
818 | for_each_populated_zone(zone) |
819 | if (is_highmem(zone)) | |
d23ad423 | 820 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
821 | |
822 | return cnt; | |
823 | } | |
824 | ||
825 | /** | |
826 | * saveable_highmem_page - Determine whether a highmem page should be | |
827 | * included in the suspend image. | |
828 | * | |
829 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, | |
830 | * and it isn't a part of a free chunk of pages. | |
831 | */ | |
846705de | 832 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
833 | { |
834 | struct page *page; | |
835 | ||
836 | if (!pfn_valid(pfn)) | |
837 | return NULL; | |
838 | ||
839 | page = pfn_to_page(pfn); | |
846705de RW |
840 | if (page_zone(page) != zone) |
841 | return NULL; | |
8357376d RW |
842 | |
843 | BUG_ON(!PageHighMem(page)); | |
844 | ||
7be98234 RW |
845 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
846 | PageReserved(page)) | |
8357376d RW |
847 | return NULL; |
848 | ||
849 | return page; | |
850 | } | |
851 | ||
852 | /** | |
853 | * count_highmem_pages - compute the total number of saveable highmem | |
854 | * pages. | |
855 | */ | |
856 | ||
fe419535 | 857 | static unsigned int count_highmem_pages(void) |
8357376d RW |
858 | { |
859 | struct zone *zone; | |
860 | unsigned int n = 0; | |
861 | ||
98e73dc5 | 862 | for_each_populated_zone(zone) { |
8357376d RW |
863 | unsigned long pfn, max_zone_pfn; |
864 | ||
865 | if (!is_highmem(zone)) | |
866 | continue; | |
867 | ||
868 | mark_free_pages(zone); | |
869 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
870 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
846705de | 871 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
872 | n++; |
873 | } | |
874 | return n; | |
875 | } | |
876 | #else | |
846705de RW |
877 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
878 | { | |
879 | return NULL; | |
880 | } | |
8357376d RW |
881 | #endif /* CONFIG_HIGHMEM */ |
882 | ||
25761b6e | 883 | /** |
8a235efa RW |
884 | * saveable_page - Determine whether a non-highmem page should be included |
885 | * in the suspend image. | |
25761b6e | 886 | * |
8357376d RW |
887 | * We should save the page if it isn't Nosave, and is not in the range |
888 | * of pages statically defined as 'unsaveable', and it isn't a part of | |
889 | * a free chunk of pages. | |
25761b6e | 890 | */ |
846705de | 891 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 892 | { |
de491861 | 893 | struct page *page; |
25761b6e RW |
894 | |
895 | if (!pfn_valid(pfn)) | |
ae83c5ee | 896 | return NULL; |
25761b6e RW |
897 | |
898 | page = pfn_to_page(pfn); | |
846705de RW |
899 | if (page_zone(page) != zone) |
900 | return NULL; | |
ae83c5ee | 901 | |
8357376d RW |
902 | BUG_ON(PageHighMem(page)); |
903 | ||
7be98234 | 904 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 905 | return NULL; |
8357376d | 906 | |
8a235efa RW |
907 | if (PageReserved(page) |
908 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 909 | return NULL; |
25761b6e | 910 | |
ae83c5ee | 911 | return page; |
25761b6e RW |
912 | } |
913 | ||
8357376d RW |
914 | /** |
915 | * count_data_pages - compute the total number of saveable non-highmem | |
916 | * pages. | |
917 | */ | |
918 | ||
fe419535 | 919 | static unsigned int count_data_pages(void) |
25761b6e RW |
920 | { |
921 | struct zone *zone; | |
ae83c5ee | 922 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 923 | unsigned int n = 0; |
25761b6e | 924 | |
98e73dc5 | 925 | for_each_populated_zone(zone) { |
25761b6e RW |
926 | if (is_highmem(zone)) |
927 | continue; | |
8357376d | 928 | |
25761b6e | 929 | mark_free_pages(zone); |
ae83c5ee RW |
930 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
931 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
846705de | 932 | if (saveable_page(zone, pfn)) |
8357376d | 933 | n++; |
25761b6e | 934 | } |
a0f49651 | 935 | return n; |
25761b6e RW |
936 | } |
937 | ||
8357376d RW |
938 | /* This is needed, because copy_page and memcpy are not usable for copying |
939 | * task structs. | |
940 | */ | |
941 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
942 | { |
943 | int n; | |
944 | ||
f623f0db RW |
945 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
946 | *dst++ = *src++; | |
947 | } | |
948 | ||
8a235efa RW |
949 | |
950 | /** | |
951 | * safe_copy_page - check if the page we are going to copy is marked as | |
952 | * present in the kernel page tables (this always is the case if | |
953 | * CONFIG_DEBUG_PAGEALLOC is not set and in that case | |
954 | * kernel_page_present() always returns 'true'). | |
955 | */ | |
956 | static void safe_copy_page(void *dst, struct page *s_page) | |
957 | { | |
958 | if (kernel_page_present(s_page)) { | |
959 | do_copy_page(dst, page_address(s_page)); | |
960 | } else { | |
961 | kernel_map_pages(s_page, 1, 1); | |
962 | do_copy_page(dst, page_address(s_page)); | |
963 | kernel_map_pages(s_page, 1, 0); | |
964 | } | |
965 | } | |
966 | ||
967 | ||
8357376d RW |
968 | #ifdef CONFIG_HIGHMEM |
969 | static inline struct page * | |
970 | page_is_saveable(struct zone *zone, unsigned long pfn) | |
971 | { | |
972 | return is_highmem(zone) ? | |
846705de | 973 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
974 | } |
975 | ||
8a235efa | 976 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
977 | { |
978 | struct page *s_page, *d_page; | |
979 | void *src, *dst; | |
980 | ||
981 | s_page = pfn_to_page(src_pfn); | |
982 | d_page = pfn_to_page(dst_pfn); | |
983 | if (PageHighMem(s_page)) { | |
984 | src = kmap_atomic(s_page, KM_USER0); | |
985 | dst = kmap_atomic(d_page, KM_USER1); | |
986 | do_copy_page(dst, src); | |
987 | kunmap_atomic(src, KM_USER0); | |
988 | kunmap_atomic(dst, KM_USER1); | |
989 | } else { | |
8357376d RW |
990 | if (PageHighMem(d_page)) { |
991 | /* Page pointed to by src may contain some kernel | |
992 | * data modified by kmap_atomic() | |
993 | */ | |
8a235efa | 994 | safe_copy_page(buffer, s_page); |
baa5835d | 995 | dst = kmap_atomic(d_page, KM_USER0); |
8357376d RW |
996 | memcpy(dst, buffer, PAGE_SIZE); |
997 | kunmap_atomic(dst, KM_USER0); | |
998 | } else { | |
8a235efa | 999 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1000 | } |
1001 | } | |
1002 | } | |
1003 | #else | |
846705de | 1004 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1005 | |
8a235efa | 1006 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1007 | { |
8a235efa RW |
1008 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1009 | pfn_to_page(src_pfn)); | |
8357376d RW |
1010 | } |
1011 | #endif /* CONFIG_HIGHMEM */ | |
1012 | ||
b788db79 RW |
1013 | static void |
1014 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) | |
25761b6e RW |
1015 | { |
1016 | struct zone *zone; | |
b788db79 | 1017 | unsigned long pfn; |
25761b6e | 1018 | |
98e73dc5 | 1019 | for_each_populated_zone(zone) { |
b788db79 RW |
1020 | unsigned long max_zone_pfn; |
1021 | ||
25761b6e | 1022 | mark_free_pages(zone); |
ae83c5ee | 1023 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
b788db79 | 1024 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1025 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1026 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1027 | } |
b788db79 RW |
1028 | memory_bm_position_reset(orig_bm); |
1029 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1030 | for(;;) { |
b788db79 | 1031 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1032 | if (unlikely(pfn == BM_END_OF_MAP)) |
1033 | break; | |
1034 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1035 | } | |
25761b6e RW |
1036 | } |
1037 | ||
8357376d RW |
1038 | /* Total number of image pages */ |
1039 | static unsigned int nr_copy_pages; | |
1040 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1041 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1042 | /* |
1043 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1044 | * before suspending devices. | |
1045 | */ | |
1046 | unsigned int alloc_normal, alloc_highmem; | |
1047 | /* | |
1048 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1049 | * hibernation image pages (during restore) | |
1050 | */ | |
1051 | static struct memory_bitmap orig_bm; | |
1052 | /* | |
1053 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1054 | * will contain copies of saveable pages. During restore it is initially used | |
1055 | * for marking hibernation image pages, but then the set bits from it are | |
1056 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1057 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1058 | * this purpose. | |
1059 | */ | |
1060 | static struct memory_bitmap copy_bm; | |
8357376d | 1061 | |
25761b6e | 1062 | /** |
940864dd | 1063 | * swsusp_free - free pages allocated for the suspend. |
cd560bb2 | 1064 | * |
940864dd RW |
1065 | * Suspend pages are alocated before the atomic copy is made, so we |
1066 | * need to release them after the resume. | |
25761b6e RW |
1067 | */ |
1068 | ||
1069 | void swsusp_free(void) | |
1070 | { | |
1071 | struct zone *zone; | |
ae83c5ee | 1072 | unsigned long pfn, max_zone_pfn; |
25761b6e | 1073 | |
98e73dc5 | 1074 | for_each_populated_zone(zone) { |
ae83c5ee RW |
1075 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1076 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1077 | if (pfn_valid(pfn)) { | |
1078 | struct page *page = pfn_to_page(pfn); | |
1079 | ||
7be98234 RW |
1080 | if (swsusp_page_is_forbidden(page) && |
1081 | swsusp_page_is_free(page)) { | |
1082 | swsusp_unset_page_forbidden(page); | |
1083 | swsusp_unset_page_free(page); | |
8357376d | 1084 | __free_page(page); |
25761b6e RW |
1085 | } |
1086 | } | |
1087 | } | |
f577eb30 RW |
1088 | nr_copy_pages = 0; |
1089 | nr_meta_pages = 0; | |
75534b50 | 1090 | restore_pblist = NULL; |
6e1819d6 | 1091 | buffer = NULL; |
64a473cb RW |
1092 | alloc_normal = 0; |
1093 | alloc_highmem = 0; | |
25761b6e RW |
1094 | } |
1095 | ||
4bb33435 RW |
1096 | /* Helper functions used for the shrinking of memory. */ |
1097 | ||
1098 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1099 | ||
fe419535 | 1100 | /** |
4bb33435 RW |
1101 | * preallocate_image_pages - Allocate a number of pages for hibernation image |
1102 | * @nr_pages: Number of page frames to allocate. | |
1103 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1104 | * |
4bb33435 RW |
1105 | * Return value: Number of page frames actually allocated |
1106 | */ | |
1107 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1108 | { | |
1109 | unsigned long nr_alloc = 0; | |
1110 | ||
1111 | while (nr_pages > 0) { | |
64a473cb RW |
1112 | struct page *page; |
1113 | ||
1114 | page = alloc_image_page(mask); | |
1115 | if (!page) | |
4bb33435 | 1116 | break; |
64a473cb RW |
1117 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1118 | if (PageHighMem(page)) | |
1119 | alloc_highmem++; | |
1120 | else | |
1121 | alloc_normal++; | |
4bb33435 RW |
1122 | nr_pages--; |
1123 | nr_alloc++; | |
1124 | } | |
1125 | ||
1126 | return nr_alloc; | |
1127 | } | |
1128 | ||
6715045d RW |
1129 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1130 | unsigned long avail_normal) | |
4bb33435 | 1131 | { |
6715045d RW |
1132 | unsigned long alloc; |
1133 | ||
1134 | if (avail_normal <= alloc_normal) | |
1135 | return 0; | |
1136 | ||
1137 | alloc = avail_normal - alloc_normal; | |
1138 | if (nr_pages < alloc) | |
1139 | alloc = nr_pages; | |
1140 | ||
1141 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1142 | } |
1143 | ||
1144 | #ifdef CONFIG_HIGHMEM | |
1145 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1146 | { | |
1147 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1148 | } | |
1149 | ||
1150 | /** | |
1151 | * __fraction - Compute (an approximation of) x * (multiplier / base) | |
fe419535 | 1152 | */ |
4bb33435 RW |
1153 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1154 | { | |
1155 | x *= multiplier; | |
1156 | do_div(x, base); | |
1157 | return (unsigned long)x; | |
1158 | } | |
fe419535 | 1159 | |
4bb33435 RW |
1160 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
1161 | unsigned long highmem, | |
1162 | unsigned long total) | |
fe419535 | 1163 | { |
4bb33435 RW |
1164 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1165 | ||
1166 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1167 | } |
4bb33435 RW |
1168 | #else /* CONFIG_HIGHMEM */ |
1169 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1170 | { | |
1171 | return 0; | |
1172 | } | |
1173 | ||
1174 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
1175 | unsigned long highmem, | |
1176 | unsigned long total) | |
1177 | { | |
1178 | return 0; | |
1179 | } | |
1180 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1181 | |
4bb33435 | 1182 | /** |
64a473cb RW |
1183 | * free_unnecessary_pages - Release preallocated pages not needed for the image |
1184 | */ | |
1185 | static void free_unnecessary_pages(void) | |
1186 | { | |
6715045d | 1187 | unsigned long save, to_free_normal, to_free_highmem; |
64a473cb | 1188 | |
6715045d RW |
1189 | save = count_data_pages(); |
1190 | if (alloc_normal >= save) { | |
1191 | to_free_normal = alloc_normal - save; | |
1192 | save = 0; | |
1193 | } else { | |
1194 | to_free_normal = 0; | |
1195 | save -= alloc_normal; | |
1196 | } | |
1197 | save += count_highmem_pages(); | |
1198 | if (alloc_highmem >= save) { | |
1199 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1200 | } else { |
1201 | to_free_highmem = 0; | |
6715045d | 1202 | to_free_normal -= save - alloc_highmem; |
64a473cb RW |
1203 | } |
1204 | ||
1205 | memory_bm_position_reset(©_bm); | |
1206 | ||
a9c9b442 | 1207 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1208 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1209 | struct page *page = pfn_to_page(pfn); | |
1210 | ||
1211 | if (PageHighMem(page)) { | |
1212 | if (!to_free_highmem) | |
1213 | continue; | |
1214 | to_free_highmem--; | |
1215 | alloc_highmem--; | |
1216 | } else { | |
1217 | if (!to_free_normal) | |
1218 | continue; | |
1219 | to_free_normal--; | |
1220 | alloc_normal--; | |
1221 | } | |
1222 | memory_bm_clear_bit(©_bm, pfn); | |
1223 | swsusp_unset_page_forbidden(page); | |
1224 | swsusp_unset_page_free(page); | |
1225 | __free_page(page); | |
1226 | } | |
1227 | } | |
1228 | ||
ef4aede3 RW |
1229 | /** |
1230 | * minimum_image_size - Estimate the minimum acceptable size of an image | |
1231 | * @saveable: Number of saveable pages in the system. | |
1232 | * | |
1233 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1234 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1235 | * preallocating memory. | |
1236 | * | |
1237 | * We assume that the minimum image size should be proportional to | |
1238 | * | |
1239 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1240 | * | |
1241 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
1242 | * and (3) inactive anonymouns pages, (4) active and (5) inactive file pages, | |
1243 | * minus mapped file pages. | |
1244 | */ | |
1245 | static unsigned long minimum_image_size(unsigned long saveable) | |
1246 | { | |
1247 | unsigned long size; | |
1248 | ||
1249 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
1250 | + global_page_state(NR_ACTIVE_ANON) | |
1251 | + global_page_state(NR_INACTIVE_ANON) | |
1252 | + global_page_state(NR_ACTIVE_FILE) | |
1253 | + global_page_state(NR_INACTIVE_FILE) | |
1254 | - global_page_state(NR_FILE_MAPPED); | |
1255 | ||
1256 | return saveable <= size ? 0 : saveable - size; | |
1257 | } | |
1258 | ||
64a473cb RW |
1259 | /** |
1260 | * hibernate_preallocate_memory - Preallocate memory for hibernation image | |
4bb33435 RW |
1261 | * |
1262 | * To create a hibernation image it is necessary to make a copy of every page | |
1263 | * frame in use. We also need a number of page frames to be free during | |
1264 | * hibernation for allocations made while saving the image and for device | |
1265 | * drivers, in case they need to allocate memory from their hibernation | |
1266 | * callbacks (these two numbers are given by PAGES_FOR_IO and SPARE_PAGES, | |
1267 | * respectively, both of which are rough estimates). To make this happen, we | |
1268 | * compute the total number of available page frames and allocate at least | |
1269 | * | |
1270 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 + 2 * SPARE_PAGES | |
1271 | * | |
1272 | * of them, which corresponds to the maximum size of a hibernation image. | |
1273 | * | |
1274 | * If image_size is set below the number following from the above formula, | |
1275 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1276 | * pages in the system is below the requested image size or the minimum |
1277 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1278 | */ |
64a473cb | 1279 | int hibernate_preallocate_memory(void) |
fe419535 | 1280 | { |
fe419535 | 1281 | struct zone *zone; |
4bb33435 | 1282 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1283 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
fe419535 | 1284 | struct timeval start, stop; |
64a473cb | 1285 | int error; |
fe419535 | 1286 | |
64a473cb | 1287 | printk(KERN_INFO "PM: Preallocating image memory... "); |
fe419535 | 1288 | do_gettimeofday(&start); |
fe419535 | 1289 | |
64a473cb RW |
1290 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1291 | if (error) | |
1292 | goto err_out; | |
1293 | ||
1294 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1295 | if (error) | |
1296 | goto err_out; | |
1297 | ||
1298 | alloc_normal = 0; | |
1299 | alloc_highmem = 0; | |
1300 | ||
4bb33435 | 1301 | /* Count the number of saveable data pages. */ |
64a473cb | 1302 | save_highmem = count_highmem_pages(); |
4bb33435 | 1303 | saveable = count_data_pages(); |
fe419535 | 1304 | |
4bb33435 RW |
1305 | /* |
1306 | * Compute the total number of page frames we can use (count) and the | |
1307 | * number of pages needed for image metadata (size). | |
1308 | */ | |
1309 | count = saveable; | |
64a473cb RW |
1310 | saveable += save_highmem; |
1311 | highmem = save_highmem; | |
4bb33435 RW |
1312 | size = 0; |
1313 | for_each_populated_zone(zone) { | |
1314 | size += snapshot_additional_pages(zone); | |
1315 | if (is_highmem(zone)) | |
1316 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1317 | else | |
1318 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1319 | } | |
6715045d | 1320 | avail_normal = count; |
4bb33435 RW |
1321 | count += highmem; |
1322 | count -= totalreserve_pages; | |
1323 | ||
1324 | /* Compute the maximum number of saveable pages to leave in memory. */ | |
1325 | max_size = (count - (size + PAGES_FOR_IO)) / 2 - 2 * SPARE_PAGES; | |
266f1a25 | 1326 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1327 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1328 | if (size > max_size) | |
1329 | size = max_size; | |
1330 | /* | |
266f1a25 RW |
1331 | * If the desired number of image pages is at least as large as the |
1332 | * current number of saveable pages in memory, allocate page frames for | |
1333 | * the image and we're done. | |
4bb33435 | 1334 | */ |
64a473cb RW |
1335 | if (size >= saveable) { |
1336 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1337 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1338 | goto out; |
64a473cb | 1339 | } |
4bb33435 | 1340 | |
ef4aede3 RW |
1341 | /* Estimate the minimum size of the image. */ |
1342 | pages = minimum_image_size(saveable); | |
6715045d RW |
1343 | /* |
1344 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1345 | * accommodate an image of the minimum size (unless it's already too | |
1346 | * small, in which case don't preallocate pages from it at all). | |
1347 | */ | |
1348 | if (avail_normal > pages) | |
1349 | avail_normal -= pages; | |
1350 | else | |
1351 | avail_normal = 0; | |
ef4aede3 RW |
1352 | if (size < pages) |
1353 | size = min_t(unsigned long, pages, max_size); | |
1354 | ||
4bb33435 RW |
1355 | /* |
1356 | * Let the memory management subsystem know that we're going to need a | |
1357 | * large number of page frames to allocate and make it free some memory. | |
1358 | * NOTE: If this is not done, performance will be hurt badly in some | |
1359 | * test cases. | |
1360 | */ | |
1361 | shrink_all_memory(saveable - size); | |
1362 | ||
1363 | /* | |
1364 | * The number of saveable pages in memory was too high, so apply some | |
1365 | * pressure to decrease it. First, make room for the largest possible | |
1366 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1367 | * of the image as much as indicated by 'size' using allocations from |
1368 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1369 | */ |
1370 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
1371 | alloc = (count - max_size) - pages_highmem; | |
6715045d RW |
1372 | pages = preallocate_image_memory(alloc, avail_normal); |
1373 | if (pages < alloc) { | |
1374 | /* We have exhausted non-highmem pages, try highmem. */ | |
1375 | alloc -= pages; | |
1376 | pages += pages_highmem; | |
1377 | pages_highmem = preallocate_image_highmem(alloc); | |
1378 | if (pages_highmem < alloc) | |
1379 | goto err_out; | |
1380 | pages += pages_highmem; | |
1381 | /* | |
1382 | * size is the desired number of saveable pages to leave in | |
1383 | * memory, so try to preallocate (all memory - size) pages. | |
1384 | */ | |
1385 | alloc = (count - pages) - size; | |
1386 | pages += preallocate_image_highmem(alloc); | |
1387 | } else { | |
1388 | /* | |
1389 | * There are approximately max_size saveable pages at this point | |
1390 | * and we want to reduce this number down to size. | |
1391 | */ | |
1392 | alloc = max_size - size; | |
1393 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1394 | pages_highmem += size; | |
1395 | alloc -= size; | |
1396 | size = preallocate_image_memory(alloc, avail_normal); | |
1397 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1398 | pages += pages_highmem + size; | |
1399 | } | |
4bb33435 | 1400 | |
64a473cb RW |
1401 | /* |
1402 | * We only need as many page frames for the image as there are saveable | |
1403 | * pages in memory, but we have allocated more. Release the excessive | |
1404 | * ones now. | |
1405 | */ | |
1406 | free_unnecessary_pages(); | |
4bb33435 RW |
1407 | |
1408 | out: | |
fe419535 | 1409 | do_gettimeofday(&stop); |
64a473cb RW |
1410 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
1411 | swsusp_show_speed(&start, &stop, pages, "Allocated"); | |
fe419535 RW |
1412 | |
1413 | return 0; | |
64a473cb RW |
1414 | |
1415 | err_out: | |
1416 | printk(KERN_CONT "\n"); | |
1417 | swsusp_free(); | |
1418 | return -ENOMEM; | |
fe419535 RW |
1419 | } |
1420 | ||
8357376d RW |
1421 | #ifdef CONFIG_HIGHMEM |
1422 | /** | |
1423 | * count_pages_for_highmem - compute the number of non-highmem pages | |
1424 | * that will be necessary for creating copies of highmem pages. | |
1425 | */ | |
1426 | ||
1427 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) | |
1428 | { | |
64a473cb | 1429 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1430 | |
1431 | if (free_highmem >= nr_highmem) | |
1432 | nr_highmem = 0; | |
1433 | else | |
1434 | nr_highmem -= free_highmem; | |
1435 | ||
1436 | return nr_highmem; | |
1437 | } | |
1438 | #else | |
1439 | static unsigned int | |
1440 | count_pages_for_highmem(unsigned int nr_highmem) { return 0; } | |
1441 | #endif /* CONFIG_HIGHMEM */ | |
25761b6e RW |
1442 | |
1443 | /** | |
8357376d RW |
1444 | * enough_free_mem - Make sure we have enough free memory for the |
1445 | * snapshot image. | |
25761b6e RW |
1446 | */ |
1447 | ||
8357376d | 1448 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1449 | { |
e5e2fa78 | 1450 | struct zone *zone; |
64a473cb | 1451 | unsigned int free = alloc_normal; |
e5e2fa78 | 1452 | |
98e73dc5 | 1453 | for_each_populated_zone(zone) |
8357376d | 1454 | if (!is_highmem(zone)) |
d23ad423 | 1455 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1456 | |
8357376d | 1457 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1458 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1459 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1460 | |
64a473cb | 1461 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1462 | } |
1463 | ||
8357376d RW |
1464 | #ifdef CONFIG_HIGHMEM |
1465 | /** | |
1466 | * get_highmem_buffer - if there are some highmem pages in the suspend | |
1467 | * image, we may need the buffer to copy them and/or load their data. | |
1468 | */ | |
1469 | ||
1470 | static inline int get_highmem_buffer(int safe_needed) | |
1471 | { | |
1472 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1473 | return buffer ? 0 : -ENOMEM; | |
1474 | } | |
1475 | ||
1476 | /** | |
1477 | * alloc_highmem_image_pages - allocate some highmem pages for the image. | |
1478 | * Try to allocate as many pages as needed, but if the number of free | |
1479 | * highmem pages is lesser than that, allocate them all. | |
1480 | */ | |
1481 | ||
1482 | static inline unsigned int | |
64a473cb | 1483 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int nr_highmem) |
8357376d RW |
1484 | { |
1485 | unsigned int to_alloc = count_free_highmem_pages(); | |
1486 | ||
1487 | if (to_alloc > nr_highmem) | |
1488 | to_alloc = nr_highmem; | |
1489 | ||
1490 | nr_highmem -= to_alloc; | |
1491 | while (to_alloc-- > 0) { | |
1492 | struct page *page; | |
1493 | ||
1494 | page = alloc_image_page(__GFP_HIGHMEM); | |
1495 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1496 | } | |
1497 | return nr_highmem; | |
1498 | } | |
1499 | #else | |
1500 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1501 | ||
1502 | static inline unsigned int | |
64a473cb | 1503 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int n) { return 0; } |
8357376d RW |
1504 | #endif /* CONFIG_HIGHMEM */ |
1505 | ||
1506 | /** | |
1507 | * swsusp_alloc - allocate memory for the suspend image | |
1508 | * | |
1509 | * We first try to allocate as many highmem pages as there are | |
1510 | * saveable highmem pages in the system. If that fails, we allocate | |
1511 | * non-highmem pages for the copies of the remaining highmem ones. | |
1512 | * | |
1513 | * In this approach it is likely that the copies of highmem pages will | |
1514 | * also be located in the high memory, because of the way in which | |
1515 | * copy_data_pages() works. | |
1516 | */ | |
1517 | ||
b788db79 RW |
1518 | static int |
1519 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, | |
8357376d | 1520 | unsigned int nr_pages, unsigned int nr_highmem) |
054bd4c1 | 1521 | { |
64a473cb | 1522 | int error = 0; |
25761b6e | 1523 | |
8357376d RW |
1524 | if (nr_highmem > 0) { |
1525 | error = get_highmem_buffer(PG_ANY); | |
1526 | if (error) | |
64a473cb RW |
1527 | goto err_out; |
1528 | if (nr_highmem > alloc_highmem) { | |
1529 | nr_highmem -= alloc_highmem; | |
1530 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1531 | } | |
8357376d | 1532 | } |
64a473cb RW |
1533 | if (nr_pages > alloc_normal) { |
1534 | nr_pages -= alloc_normal; | |
1535 | while (nr_pages-- > 0) { | |
1536 | struct page *page; | |
1537 | ||
1538 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1539 | if (!page) | |
1540 | goto err_out; | |
1541 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1542 | } | |
25761b6e | 1543 | } |
64a473cb | 1544 | |
b788db79 | 1545 | return 0; |
25761b6e | 1546 | |
64a473cb | 1547 | err_out: |
b788db79 | 1548 | swsusp_free(); |
64a473cb | 1549 | return error; |
25761b6e RW |
1550 | } |
1551 | ||
2e32a43e | 1552 | asmlinkage int swsusp_save(void) |
25761b6e | 1553 | { |
8357376d | 1554 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1555 | |
07c3bb57 | 1556 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1557 | |
9f8f2172 | 1558 | drain_local_pages(NULL); |
a0f49651 | 1559 | nr_pages = count_data_pages(); |
8357376d | 1560 | nr_highmem = count_highmem_pages(); |
23976728 | 1561 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1562 | |
8357376d | 1563 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1564 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1565 | return -ENOMEM; |
1566 | } | |
1567 | ||
8357376d | 1568 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1569 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1570 | return -ENOMEM; |
8357376d | 1571 | } |
25761b6e RW |
1572 | |
1573 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1574 | * Kill them. | |
1575 | */ | |
9f8f2172 | 1576 | drain_local_pages(NULL); |
b788db79 | 1577 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1578 | |
1579 | /* | |
1580 | * End of critical section. From now on, we can write to memory, | |
1581 | * but we should not touch disk. This specially means we must _not_ | |
1582 | * touch swap space! Except we must write out our image of course. | |
1583 | */ | |
1584 | ||
8357376d | 1585 | nr_pages += nr_highmem; |
a0f49651 | 1586 | nr_copy_pages = nr_pages; |
8357376d | 1587 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 1588 | |
23976728 RW |
1589 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1590 | nr_pages); | |
8357376d | 1591 | |
25761b6e RW |
1592 | return 0; |
1593 | } | |
f577eb30 | 1594 | |
d307c4a8 RW |
1595 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1596 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 1597 | { |
d307c4a8 | 1598 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 1599 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
1600 | return 0; |
1601 | } | |
1602 | ||
1603 | static char *check_image_kernel(struct swsusp_info *info) | |
1604 | { | |
1605 | if (info->version_code != LINUX_VERSION_CODE) | |
1606 | return "kernel version"; | |
1607 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
1608 | return "system type"; | |
1609 | if (strcmp(info->uts.release,init_utsname()->release)) | |
1610 | return "kernel release"; | |
1611 | if (strcmp(info->uts.version,init_utsname()->version)) | |
1612 | return "version"; | |
1613 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
1614 | return "machine"; | |
1615 | return NULL; | |
1616 | } | |
1617 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
1618 | ||
af508b34 RW |
1619 | unsigned long snapshot_get_image_size(void) |
1620 | { | |
1621 | return nr_copy_pages + nr_meta_pages + 1; | |
1622 | } | |
1623 | ||
d307c4a8 RW |
1624 | static int init_header(struct swsusp_info *info) |
1625 | { | |
1626 | memset(info, 0, sizeof(struct swsusp_info)); | |
f577eb30 | 1627 | info->num_physpages = num_physpages; |
f577eb30 | 1628 | info->image_pages = nr_copy_pages; |
af508b34 | 1629 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
1630 | info->size = info->pages; |
1631 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 1632 | return init_header_complete(info); |
f577eb30 RW |
1633 | } |
1634 | ||
1635 | /** | |
940864dd RW |
1636 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1637 | * are stored in the array @buf[] (1 page at a time) | |
f577eb30 RW |
1638 | */ |
1639 | ||
b788db79 | 1640 | static inline void |
940864dd | 1641 | pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1642 | { |
1643 | int j; | |
1644 | ||
b788db79 | 1645 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
1646 | buf[j] = memory_bm_next_pfn(bm); |
1647 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 1648 | break; |
f577eb30 | 1649 | } |
f577eb30 RW |
1650 | } |
1651 | ||
1652 | /** | |
1653 | * snapshot_read_next - used for reading the system memory snapshot. | |
1654 | * | |
1655 | * On the first call to it @handle should point to a zeroed | |
1656 | * snapshot_handle structure. The structure gets updated and a pointer | |
1657 | * to it should be passed to this function every next time. | |
1658 | * | |
f577eb30 RW |
1659 | * On success the function returns a positive number. Then, the caller |
1660 | * is allowed to read up to the returned number of bytes from the memory | |
d3c1b24c | 1661 | * location computed by the data_of() macro. |
f577eb30 RW |
1662 | * |
1663 | * The function returns 0 to indicate the end of data stream condition, | |
1664 | * and a negative number is returned on error. In such cases the | |
1665 | * structure pointed to by @handle is not updated and should not be used | |
1666 | * any more. | |
1667 | */ | |
1668 | ||
d3c1b24c | 1669 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 1670 | { |
fb13a28b | 1671 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1672 | return 0; |
b788db79 | 1673 | |
f577eb30 RW |
1674 | if (!buffer) { |
1675 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 1676 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1677 | if (!buffer) |
1678 | return -ENOMEM; | |
1679 | } | |
d3c1b24c | 1680 | if (!handle->cur) { |
d307c4a8 RW |
1681 | int error; |
1682 | ||
1683 | error = init_header((struct swsusp_info *)buffer); | |
1684 | if (error) | |
1685 | return error; | |
f577eb30 | 1686 | handle->buffer = buffer; |
b788db79 RW |
1687 | memory_bm_position_reset(&orig_bm); |
1688 | memory_bm_position_reset(©_bm); | |
d3c1b24c JS |
1689 | } else if (handle->cur <= nr_meta_pages) { |
1690 | memset(buffer, 0, PAGE_SIZE); | |
1691 | pack_pfns(buffer, &orig_bm); | |
1692 | } else { | |
1693 | struct page *page; | |
b788db79 | 1694 | |
d3c1b24c JS |
1695 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
1696 | if (PageHighMem(page)) { | |
1697 | /* Highmem pages are copied to the buffer, | |
1698 | * because we can't return with a kmapped | |
1699 | * highmem page (we may not be called again). | |
1700 | */ | |
1701 | void *kaddr; | |
8357376d | 1702 | |
d3c1b24c JS |
1703 | kaddr = kmap_atomic(page, KM_USER0); |
1704 | memcpy(buffer, kaddr, PAGE_SIZE); | |
1705 | kunmap_atomic(kaddr, KM_USER0); | |
1706 | handle->buffer = buffer; | |
1707 | } else { | |
1708 | handle->buffer = page_address(page); | |
f577eb30 | 1709 | } |
f577eb30 | 1710 | } |
d3c1b24c JS |
1711 | handle->cur++; |
1712 | return PAGE_SIZE; | |
f577eb30 RW |
1713 | } |
1714 | ||
1715 | /** | |
1716 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
1717 | * the image during resume, because they conflict with the pages that | |
1718 | * had been used before suspend | |
1719 | */ | |
1720 | ||
940864dd | 1721 | static int mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 RW |
1722 | { |
1723 | struct zone *zone; | |
ae83c5ee | 1724 | unsigned long pfn, max_zone_pfn; |
f577eb30 RW |
1725 | |
1726 | /* Clear page flags */ | |
98e73dc5 | 1727 | for_each_populated_zone(zone) { |
ae83c5ee RW |
1728 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1729 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1730 | if (pfn_valid(pfn)) | |
7be98234 | 1731 | swsusp_unset_page_free(pfn_to_page(pfn)); |
f577eb30 RW |
1732 | } |
1733 | ||
940864dd RW |
1734 | /* Mark pages that correspond to the "original" pfns as "unsafe" */ |
1735 | memory_bm_position_reset(bm); | |
1736 | do { | |
1737 | pfn = memory_bm_next_pfn(bm); | |
1738 | if (likely(pfn != BM_END_OF_MAP)) { | |
1739 | if (likely(pfn_valid(pfn))) | |
7be98234 | 1740 | swsusp_set_page_free(pfn_to_page(pfn)); |
940864dd RW |
1741 | else |
1742 | return -EFAULT; | |
1743 | } | |
1744 | } while (pfn != BM_END_OF_MAP); | |
f577eb30 | 1745 | |
940864dd | 1746 | allocated_unsafe_pages = 0; |
968808b8 | 1747 | |
f577eb30 RW |
1748 | return 0; |
1749 | } | |
1750 | ||
940864dd RW |
1751 | static void |
1752 | duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src) | |
f577eb30 | 1753 | { |
940864dd RW |
1754 | unsigned long pfn; |
1755 | ||
1756 | memory_bm_position_reset(src); | |
1757 | pfn = memory_bm_next_pfn(src); | |
1758 | while (pfn != BM_END_OF_MAP) { | |
1759 | memory_bm_set_bit(dst, pfn); | |
1760 | pfn = memory_bm_next_pfn(src); | |
f577eb30 RW |
1761 | } |
1762 | } | |
1763 | ||
d307c4a8 | 1764 | static int check_header(struct swsusp_info *info) |
f577eb30 | 1765 | { |
d307c4a8 | 1766 | char *reason; |
f577eb30 | 1767 | |
d307c4a8 RW |
1768 | reason = check_image_kernel(info); |
1769 | if (!reason && info->num_physpages != num_physpages) | |
f577eb30 | 1770 | reason = "memory size"; |
f577eb30 | 1771 | if (reason) { |
23976728 | 1772 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
1773 | return -EPERM; |
1774 | } | |
1775 | return 0; | |
1776 | } | |
1777 | ||
1778 | /** | |
1779 | * load header - check the image header and copy data from it | |
1780 | */ | |
1781 | ||
940864dd RW |
1782 | static int |
1783 | load_header(struct swsusp_info *info) | |
f577eb30 RW |
1784 | { |
1785 | int error; | |
f577eb30 | 1786 | |
940864dd | 1787 | restore_pblist = NULL; |
f577eb30 RW |
1788 | error = check_header(info); |
1789 | if (!error) { | |
f577eb30 RW |
1790 | nr_copy_pages = info->image_pages; |
1791 | nr_meta_pages = info->pages - info->image_pages - 1; | |
1792 | } | |
1793 | return error; | |
1794 | } | |
1795 | ||
1796 | /** | |
940864dd RW |
1797 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
1798 | * the corresponding bit in the memory bitmap @bm | |
f577eb30 | 1799 | */ |
69643279 | 1800 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1801 | { |
1802 | int j; | |
1803 | ||
940864dd RW |
1804 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1805 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
1806 | break; | |
1807 | ||
69643279 RW |
1808 | if (memory_bm_pfn_present(bm, buf[j])) |
1809 | memory_bm_set_bit(bm, buf[j]); | |
1810 | else | |
1811 | return -EFAULT; | |
f577eb30 | 1812 | } |
69643279 RW |
1813 | |
1814 | return 0; | |
f577eb30 RW |
1815 | } |
1816 | ||
8357376d RW |
1817 | /* List of "safe" pages that may be used to store data loaded from the suspend |
1818 | * image | |
1819 | */ | |
1820 | static struct linked_page *safe_pages_list; | |
1821 | ||
1822 | #ifdef CONFIG_HIGHMEM | |
1823 | /* struct highmem_pbe is used for creating the list of highmem pages that | |
1824 | * should be restored atomically during the resume from disk, because the page | |
1825 | * frames they have occupied before the suspend are in use. | |
1826 | */ | |
1827 | struct highmem_pbe { | |
1828 | struct page *copy_page; /* data is here now */ | |
1829 | struct page *orig_page; /* data was here before the suspend */ | |
1830 | struct highmem_pbe *next; | |
1831 | }; | |
1832 | ||
1833 | /* List of highmem PBEs needed for restoring the highmem pages that were | |
1834 | * allocated before the suspend and included in the suspend image, but have | |
1835 | * also been allocated by the "resume" kernel, so their contents cannot be | |
1836 | * written directly to their "original" page frames. | |
1837 | */ | |
1838 | static struct highmem_pbe *highmem_pblist; | |
1839 | ||
1840 | /** | |
1841 | * count_highmem_image_pages - compute the number of highmem pages in the | |
1842 | * suspend image. The bits in the memory bitmap @bm that correspond to the | |
1843 | * image pages are assumed to be set. | |
1844 | */ | |
1845 | ||
1846 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) | |
1847 | { | |
1848 | unsigned long pfn; | |
1849 | unsigned int cnt = 0; | |
1850 | ||
1851 | memory_bm_position_reset(bm); | |
1852 | pfn = memory_bm_next_pfn(bm); | |
1853 | while (pfn != BM_END_OF_MAP) { | |
1854 | if (PageHighMem(pfn_to_page(pfn))) | |
1855 | cnt++; | |
1856 | ||
1857 | pfn = memory_bm_next_pfn(bm); | |
1858 | } | |
1859 | return cnt; | |
1860 | } | |
1861 | ||
1862 | /** | |
1863 | * prepare_highmem_image - try to allocate as many highmem pages as | |
1864 | * there are highmem image pages (@nr_highmem_p points to the variable | |
1865 | * containing the number of highmem image pages). The pages that are | |
1866 | * "safe" (ie. will not be overwritten when the suspend image is | |
1867 | * restored) have the corresponding bits set in @bm (it must be | |
1868 | * unitialized). | |
1869 | * | |
1870 | * NOTE: This function should not be called if there are no highmem | |
1871 | * image pages. | |
1872 | */ | |
1873 | ||
1874 | static unsigned int safe_highmem_pages; | |
1875 | ||
1876 | static struct memory_bitmap *safe_highmem_bm; | |
1877 | ||
1878 | static int | |
1879 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
1880 | { | |
1881 | unsigned int to_alloc; | |
1882 | ||
1883 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
1884 | return -ENOMEM; | |
1885 | ||
1886 | if (get_highmem_buffer(PG_SAFE)) | |
1887 | return -ENOMEM; | |
1888 | ||
1889 | to_alloc = count_free_highmem_pages(); | |
1890 | if (to_alloc > *nr_highmem_p) | |
1891 | to_alloc = *nr_highmem_p; | |
1892 | else | |
1893 | *nr_highmem_p = to_alloc; | |
1894 | ||
1895 | safe_highmem_pages = 0; | |
1896 | while (to_alloc-- > 0) { | |
1897 | struct page *page; | |
1898 | ||
1899 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 1900 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
1901 | /* The page is "safe", set its bit the bitmap */ |
1902 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1903 | safe_highmem_pages++; | |
1904 | } | |
1905 | /* Mark the page as allocated */ | |
7be98234 RW |
1906 | swsusp_set_page_forbidden(page); |
1907 | swsusp_set_page_free(page); | |
8357376d RW |
1908 | } |
1909 | memory_bm_position_reset(bm); | |
1910 | safe_highmem_bm = bm; | |
1911 | return 0; | |
1912 | } | |
1913 | ||
1914 | /** | |
1915 | * get_highmem_page_buffer - for given highmem image page find the buffer | |
1916 | * that suspend_write_next() should set for its caller to write to. | |
1917 | * | |
1918 | * If the page is to be saved to its "original" page frame or a copy of | |
1919 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
1920 | * the copy of the page is to be made in normal memory, so the address of | |
1921 | * the copy is returned. | |
1922 | * | |
1923 | * If @buffer is returned, the caller of suspend_write_next() will write | |
1924 | * the page's contents to @buffer, so they will have to be copied to the | |
1925 | * right location on the next call to suspend_write_next() and it is done | |
1926 | * with the help of copy_last_highmem_page(). For this purpose, if | |
1927 | * @buffer is returned, @last_highmem page is set to the page to which | |
1928 | * the data will have to be copied from @buffer. | |
1929 | */ | |
1930 | ||
1931 | static struct page *last_highmem_page; | |
1932 | ||
1933 | static void * | |
1934 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
1935 | { | |
1936 | struct highmem_pbe *pbe; | |
1937 | void *kaddr; | |
1938 | ||
7be98234 | 1939 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
8357376d RW |
1940 | /* We have allocated the "original" page frame and we can |
1941 | * use it directly to store the loaded page. | |
1942 | */ | |
1943 | last_highmem_page = page; | |
1944 | return buffer; | |
1945 | } | |
1946 | /* The "original" page frame has not been allocated and we have to | |
1947 | * use a "safe" page frame to store the loaded page. | |
1948 | */ | |
1949 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
1950 | if (!pbe) { | |
1951 | swsusp_free(); | |
69643279 | 1952 | return ERR_PTR(-ENOMEM); |
8357376d RW |
1953 | } |
1954 | pbe->orig_page = page; | |
1955 | if (safe_highmem_pages > 0) { | |
1956 | struct page *tmp; | |
1957 | ||
1958 | /* Copy of the page will be stored in high memory */ | |
1959 | kaddr = buffer; | |
1960 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
1961 | safe_highmem_pages--; | |
1962 | last_highmem_page = tmp; | |
1963 | pbe->copy_page = tmp; | |
1964 | } else { | |
1965 | /* Copy of the page will be stored in normal memory */ | |
1966 | kaddr = safe_pages_list; | |
1967 | safe_pages_list = safe_pages_list->next; | |
1968 | pbe->copy_page = virt_to_page(kaddr); | |
1969 | } | |
1970 | pbe->next = highmem_pblist; | |
1971 | highmem_pblist = pbe; | |
1972 | return kaddr; | |
1973 | } | |
1974 | ||
1975 | /** | |
1976 | * copy_last_highmem_page - copy the contents of a highmem image from | |
1977 | * @buffer, where the caller of snapshot_write_next() has place them, | |
1978 | * to the right location represented by @last_highmem_page . | |
1979 | */ | |
1980 | ||
1981 | static void copy_last_highmem_page(void) | |
1982 | { | |
1983 | if (last_highmem_page) { | |
1984 | void *dst; | |
1985 | ||
1986 | dst = kmap_atomic(last_highmem_page, KM_USER0); | |
1987 | memcpy(dst, buffer, PAGE_SIZE); | |
1988 | kunmap_atomic(dst, KM_USER0); | |
1989 | last_highmem_page = NULL; | |
1990 | } | |
1991 | } | |
1992 | ||
1993 | static inline int last_highmem_page_copied(void) | |
1994 | { | |
1995 | return !last_highmem_page; | |
1996 | } | |
1997 | ||
1998 | static inline void free_highmem_data(void) | |
1999 | { | |
2000 | if (safe_highmem_bm) | |
2001 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2002 | ||
2003 | if (buffer) | |
2004 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2005 | } | |
2006 | #else | |
2007 | static inline int get_safe_write_buffer(void) { return 0; } | |
2008 | ||
2009 | static unsigned int | |
2010 | count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } | |
2011 | ||
2012 | static inline int | |
2013 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
2014 | { | |
2015 | return 0; | |
2016 | } | |
2017 | ||
2018 | static inline void * | |
2019 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
2020 | { | |
69643279 | 2021 | return ERR_PTR(-EINVAL); |
8357376d RW |
2022 | } |
2023 | ||
2024 | static inline void copy_last_highmem_page(void) {} | |
2025 | static inline int last_highmem_page_copied(void) { return 1; } | |
2026 | static inline void free_highmem_data(void) {} | |
2027 | #endif /* CONFIG_HIGHMEM */ | |
2028 | ||
f577eb30 | 2029 | /** |
940864dd RW |
2030 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
2031 | * be overwritten in the process of restoring the system memory state | |
2032 | * from the suspend image ("unsafe" pages) and allocate memory for the | |
2033 | * image. | |
968808b8 | 2034 | * |
940864dd RW |
2035 | * The idea is to allocate a new memory bitmap first and then allocate |
2036 | * as many pages as needed for the image data, but not to assign these | |
2037 | * pages to specific tasks initially. Instead, we just mark them as | |
8357376d RW |
2038 | * allocated and create a lists of "safe" pages that will be used |
2039 | * later. On systems with high memory a list of "safe" highmem pages is | |
2040 | * also created. | |
f577eb30 RW |
2041 | */ |
2042 | ||
940864dd RW |
2043 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2044 | ||
940864dd RW |
2045 | static int |
2046 | prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) | |
f577eb30 | 2047 | { |
8357376d | 2048 | unsigned int nr_pages, nr_highmem; |
940864dd RW |
2049 | struct linked_page *sp_list, *lp; |
2050 | int error; | |
f577eb30 | 2051 | |
8357376d RW |
2052 | /* If there is no highmem, the buffer will not be necessary */ |
2053 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2054 | buffer = NULL; | |
2055 | ||
2056 | nr_highmem = count_highmem_image_pages(bm); | |
940864dd RW |
2057 | error = mark_unsafe_pages(bm); |
2058 | if (error) | |
2059 | goto Free; | |
2060 | ||
2061 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2062 | if (error) | |
2063 | goto Free; | |
2064 | ||
2065 | duplicate_memory_bitmap(new_bm, bm); | |
2066 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2067 | if (nr_highmem > 0) { |
2068 | error = prepare_highmem_image(bm, &nr_highmem); | |
2069 | if (error) | |
2070 | goto Free; | |
2071 | } | |
940864dd RW |
2072 | /* Reserve some safe pages for potential later use. |
2073 | * | |
2074 | * NOTE: This way we make sure there will be enough safe pages for the | |
2075 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2076 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
2077 | */ | |
2078 | sp_list = NULL; | |
2079 | /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */ | |
8357376d | 2080 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2081 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2082 | while (nr_pages > 0) { | |
8357376d | 2083 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2084 | if (!lp) { |
f577eb30 | 2085 | error = -ENOMEM; |
940864dd RW |
2086 | goto Free; |
2087 | } | |
2088 | lp->next = sp_list; | |
2089 | sp_list = lp; | |
2090 | nr_pages--; | |
f577eb30 | 2091 | } |
940864dd RW |
2092 | /* Preallocate memory for the image */ |
2093 | safe_pages_list = NULL; | |
8357376d | 2094 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2095 | while (nr_pages > 0) { |
2096 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2097 | if (!lp) { | |
2098 | error = -ENOMEM; | |
2099 | goto Free; | |
2100 | } | |
7be98234 | 2101 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2102 | /* The page is "safe", add it to the list */ |
2103 | lp->next = safe_pages_list; | |
2104 | safe_pages_list = lp; | |
968808b8 | 2105 | } |
940864dd | 2106 | /* Mark the page as allocated */ |
7be98234 RW |
2107 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2108 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2109 | nr_pages--; |
968808b8 | 2110 | } |
940864dd RW |
2111 | /* Free the reserved safe pages so that chain_alloc() can use them */ |
2112 | while (sp_list) { | |
2113 | lp = sp_list->next; | |
2114 | free_image_page(sp_list, PG_UNSAFE_CLEAR); | |
2115 | sp_list = lp; | |
f577eb30 | 2116 | } |
940864dd RW |
2117 | return 0; |
2118 | ||
59a49335 | 2119 | Free: |
940864dd | 2120 | swsusp_free(); |
f577eb30 RW |
2121 | return error; |
2122 | } | |
2123 | ||
940864dd RW |
2124 | /** |
2125 | * get_buffer - compute the address that snapshot_write_next() should | |
2126 | * set for its caller to write to. | |
2127 | */ | |
2128 | ||
2129 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) | |
968808b8 | 2130 | { |
940864dd | 2131 | struct pbe *pbe; |
69643279 RW |
2132 | struct page *page; |
2133 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2134 | |
69643279 RW |
2135 | if (pfn == BM_END_OF_MAP) |
2136 | return ERR_PTR(-EFAULT); | |
2137 | ||
2138 | page = pfn_to_page(pfn); | |
8357376d RW |
2139 | if (PageHighMem(page)) |
2140 | return get_highmem_page_buffer(page, ca); | |
2141 | ||
7be98234 | 2142 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
940864dd RW |
2143 | /* We have allocated the "original" page frame and we can |
2144 | * use it directly to store the loaded page. | |
968808b8 | 2145 | */ |
940864dd RW |
2146 | return page_address(page); |
2147 | ||
2148 | /* The "original" page frame has not been allocated and we have to | |
2149 | * use a "safe" page frame to store the loaded page. | |
968808b8 | 2150 | */ |
940864dd RW |
2151 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2152 | if (!pbe) { | |
2153 | swsusp_free(); | |
69643279 | 2154 | return ERR_PTR(-ENOMEM); |
940864dd | 2155 | } |
8357376d RW |
2156 | pbe->orig_address = page_address(page); |
2157 | pbe->address = safe_pages_list; | |
940864dd RW |
2158 | safe_pages_list = safe_pages_list->next; |
2159 | pbe->next = restore_pblist; | |
2160 | restore_pblist = pbe; | |
8357376d | 2161 | return pbe->address; |
968808b8 RW |
2162 | } |
2163 | ||
f577eb30 RW |
2164 | /** |
2165 | * snapshot_write_next - used for writing the system memory snapshot. | |
2166 | * | |
2167 | * On the first call to it @handle should point to a zeroed | |
2168 | * snapshot_handle structure. The structure gets updated and a pointer | |
2169 | * to it should be passed to this function every next time. | |
2170 | * | |
f577eb30 RW |
2171 | * On success the function returns a positive number. Then, the caller |
2172 | * is allowed to write up to the returned number of bytes to the memory | |
d3c1b24c | 2173 | * location computed by the data_of() macro. |
f577eb30 RW |
2174 | * |
2175 | * The function returns 0 to indicate the "end of file" condition, | |
2176 | * and a negative number is returned on error. In such cases the | |
2177 | * structure pointed to by @handle is not updated and should not be used | |
2178 | * any more. | |
2179 | */ | |
2180 | ||
d3c1b24c | 2181 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2182 | { |
940864dd | 2183 | static struct chain_allocator ca; |
f577eb30 RW |
2184 | int error = 0; |
2185 | ||
940864dd | 2186 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2187 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2188 | return 0; |
940864dd | 2189 | |
d3c1b24c JS |
2190 | handle->sync_read = 1; |
2191 | ||
2192 | if (!handle->cur) { | |
8357376d RW |
2193 | if (!buffer) |
2194 | /* This makes the buffer be freed by swsusp_free() */ | |
2195 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2196 | ||
f577eb30 RW |
2197 | if (!buffer) |
2198 | return -ENOMEM; | |
8357376d | 2199 | |
f577eb30 | 2200 | handle->buffer = buffer; |
d3c1b24c JS |
2201 | } else if (handle->cur == 1) { |
2202 | error = load_header(buffer); | |
2203 | if (error) | |
2204 | return error; | |
940864dd | 2205 | |
d3c1b24c JS |
2206 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2207 | if (error) | |
2208 | return error; | |
2209 | ||
2210 | } else if (handle->cur <= nr_meta_pages + 1) { | |
2211 | error = unpack_orig_pfns(buffer, ©_bm); | |
2212 | if (error) | |
2213 | return error; | |
940864dd | 2214 | |
d3c1b24c JS |
2215 | if (handle->cur == nr_meta_pages + 1) { |
2216 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2217 | if (error) |
2218 | return error; | |
2219 | ||
d3c1b24c JS |
2220 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2221 | memory_bm_position_reset(&orig_bm); | |
2222 | restore_pblist = NULL; | |
940864dd | 2223 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2224 | handle->sync_read = 0; |
69643279 RW |
2225 | if (IS_ERR(handle->buffer)) |
2226 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2227 | } |
f577eb30 | 2228 | } else { |
d3c1b24c JS |
2229 | copy_last_highmem_page(); |
2230 | handle->buffer = get_buffer(&orig_bm, &ca); | |
2231 | if (IS_ERR(handle->buffer)) | |
2232 | return PTR_ERR(handle->buffer); | |
2233 | if (handle->buffer != buffer) | |
2234 | handle->sync_read = 0; | |
f577eb30 | 2235 | } |
d3c1b24c JS |
2236 | handle->cur++; |
2237 | return PAGE_SIZE; | |
f577eb30 RW |
2238 | } |
2239 | ||
8357376d RW |
2240 | /** |
2241 | * snapshot_write_finalize - must be called after the last call to | |
2242 | * snapshot_write_next() in case the last page in the image happens | |
2243 | * to be a highmem page and its contents should be stored in the | |
2244 | * highmem. Additionally, it releases the memory that will not be | |
2245 | * used any more. | |
2246 | */ | |
2247 | ||
2248 | void snapshot_write_finalize(struct snapshot_handle *handle) | |
2249 | { | |
2250 | copy_last_highmem_page(); | |
2251 | /* Free only if we have loaded the image entirely */ | |
d3c1b24c | 2252 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
8357376d RW |
2253 | memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR); |
2254 | free_highmem_data(); | |
2255 | } | |
2256 | } | |
2257 | ||
f577eb30 RW |
2258 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2259 | { | |
8357376d | 2260 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2261 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2262 | } | |
2263 | ||
8357376d RW |
2264 | #ifdef CONFIG_HIGHMEM |
2265 | /* Assumes that @buf is ready and points to a "safe" page */ | |
2266 | static inline void | |
2267 | swap_two_pages_data(struct page *p1, struct page *p2, void *buf) | |
940864dd | 2268 | { |
8357376d RW |
2269 | void *kaddr1, *kaddr2; |
2270 | ||
2271 | kaddr1 = kmap_atomic(p1, KM_USER0); | |
2272 | kaddr2 = kmap_atomic(p2, KM_USER1); | |
2273 | memcpy(buf, kaddr1, PAGE_SIZE); | |
2274 | memcpy(kaddr1, kaddr2, PAGE_SIZE); | |
2275 | memcpy(kaddr2, buf, PAGE_SIZE); | |
2276 | kunmap_atomic(kaddr1, KM_USER0); | |
2277 | kunmap_atomic(kaddr2, KM_USER1); | |
2278 | } | |
2279 | ||
2280 | /** | |
2281 | * restore_highmem - for each highmem page that was allocated before | |
2282 | * the suspend and included in the suspend image, and also has been | |
2283 | * allocated by the "resume" kernel swap its current (ie. "before | |
2284 | * resume") contents with the previous (ie. "before suspend") one. | |
2285 | * | |
2286 | * If the resume eventually fails, we can call this function once | |
2287 | * again and restore the "before resume" highmem state. | |
2288 | */ | |
2289 | ||
2290 | int restore_highmem(void) | |
2291 | { | |
2292 | struct highmem_pbe *pbe = highmem_pblist; | |
2293 | void *buf; | |
2294 | ||
2295 | if (!pbe) | |
2296 | return 0; | |
2297 | ||
2298 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2299 | if (!buf) | |
2300 | return -ENOMEM; | |
2301 | ||
2302 | while (pbe) { | |
2303 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2304 | pbe = pbe->next; | |
2305 | } | |
2306 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2307 | return 0; | |
f577eb30 | 2308 | } |
8357376d | 2309 | #endif /* CONFIG_HIGHMEM */ |