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 RW |
5 | * |
6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@suse.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> | |
17 | #include <linux/smp_lock.h> | |
25761b6e | 18 | #include <linux/delay.h> |
25761b6e | 19 | #include <linux/bitops.h> |
25761b6e | 20 | #include <linux/spinlock.h> |
25761b6e | 21 | #include <linux/kernel.h> |
25761b6e RW |
22 | #include <linux/pm.h> |
23 | #include <linux/device.h> | |
74dfd666 | 24 | #include <linux/init.h> |
25761b6e RW |
25 | #include <linux/bootmem.h> |
26 | #include <linux/syscalls.h> | |
27 | #include <linux/console.h> | |
28 | #include <linux/highmem.h> | |
25761b6e RW |
29 | |
30 | #include <asm/uaccess.h> | |
31 | #include <asm/mmu_context.h> | |
32 | #include <asm/pgtable.h> | |
33 | #include <asm/tlbflush.h> | |
34 | #include <asm/io.h> | |
35 | ||
25761b6e RW |
36 | #include "power.h" |
37 | ||
74dfd666 RW |
38 | static int swsusp_page_is_free(struct page *); |
39 | static void swsusp_set_page_forbidden(struct page *); | |
40 | static void swsusp_unset_page_forbidden(struct page *); | |
41 | ||
8357376d RW |
42 | /* List of PBEs needed for restoring the pages that were allocated before |
43 | * the suspend and included in the suspend image, but have also been | |
44 | * allocated by the "resume" kernel, so their contents cannot be written | |
45 | * directly to their "original" page frames. | |
46 | */ | |
75534b50 RW |
47 | struct pbe *restore_pblist; |
48 | ||
8357376d | 49 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 50 | static void *buffer; |
7088a5c0 | 51 | |
f6143aa6 RW |
52 | /** |
53 | * @safe_needed - on resume, for storing the PBE list and the image, | |
54 | * we can only use memory pages that do not conflict with the pages | |
8357376d RW |
55 | * used before suspend. The unsafe pages have PageNosaveFree set |
56 | * and we count them using unsafe_pages. | |
f6143aa6 | 57 | * |
8357376d RW |
58 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
59 | * so that swsusp_free() can release it. | |
f6143aa6 RW |
60 | */ |
61 | ||
0bcd888d RW |
62 | #define PG_ANY 0 |
63 | #define PG_SAFE 1 | |
64 | #define PG_UNSAFE_CLEAR 1 | |
65 | #define PG_UNSAFE_KEEP 0 | |
66 | ||
940864dd | 67 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 68 | |
8357376d | 69 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
70 | { |
71 | void *res; | |
72 | ||
73 | res = (void *)get_zeroed_page(gfp_mask); | |
74 | if (safe_needed) | |
7be98234 | 75 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 76 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 77 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 78 | allocated_unsafe_pages++; |
f6143aa6 RW |
79 | res = (void *)get_zeroed_page(gfp_mask); |
80 | } | |
81 | if (res) { | |
7be98234 RW |
82 | swsusp_set_page_forbidden(virt_to_page(res)); |
83 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
84 | } |
85 | return res; | |
86 | } | |
87 | ||
88 | unsigned long get_safe_page(gfp_t gfp_mask) | |
89 | { | |
8357376d RW |
90 | return (unsigned long)get_image_page(gfp_mask, PG_SAFE); |
91 | } | |
92 | ||
5b6d15de RW |
93 | static struct page *alloc_image_page(gfp_t gfp_mask) |
94 | { | |
8357376d RW |
95 | struct page *page; |
96 | ||
97 | page = alloc_page(gfp_mask); | |
98 | if (page) { | |
7be98234 RW |
99 | swsusp_set_page_forbidden(page); |
100 | swsusp_set_page_free(page); | |
8357376d RW |
101 | } |
102 | return page; | |
f6143aa6 RW |
103 | } |
104 | ||
105 | /** | |
106 | * free_image_page - free page represented by @addr, allocated with | |
8357376d | 107 | * get_image_page (page flags set by it must be cleared) |
f6143aa6 RW |
108 | */ |
109 | ||
110 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
111 | { | |
8357376d RW |
112 | struct page *page; |
113 | ||
114 | BUG_ON(!virt_addr_valid(addr)); | |
115 | ||
116 | page = virt_to_page(addr); | |
117 | ||
7be98234 | 118 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 119 | if (clear_nosave_free) |
7be98234 | 120 | swsusp_unset_page_free(page); |
8357376d RW |
121 | |
122 | __free_page(page); | |
f6143aa6 RW |
123 | } |
124 | ||
b788db79 RW |
125 | /* struct linked_page is used to build chains of pages */ |
126 | ||
127 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
128 | ||
129 | struct linked_page { | |
130 | struct linked_page *next; | |
131 | char data[LINKED_PAGE_DATA_SIZE]; | |
132 | } __attribute__((packed)); | |
133 | ||
134 | static inline void | |
135 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) | |
136 | { | |
137 | while (list) { | |
138 | struct linked_page *lp = list->next; | |
139 | ||
140 | free_image_page(list, clear_page_nosave); | |
141 | list = lp; | |
142 | } | |
143 | } | |
144 | ||
145 | /** | |
146 | * struct chain_allocator is used for allocating small objects out of | |
147 | * a linked list of pages called 'the chain'. | |
148 | * | |
149 | * The chain grows each time when there is no room for a new object in | |
150 | * the current page. The allocated objects cannot be freed individually. | |
151 | * It is only possible to free them all at once, by freeing the entire | |
152 | * chain. | |
153 | * | |
154 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
155 | * are not much smaller than PAGE_SIZE. | |
156 | */ | |
157 | ||
158 | struct chain_allocator { | |
159 | struct linked_page *chain; /* the chain */ | |
160 | unsigned int used_space; /* total size of objects allocated out | |
161 | * of the current page | |
162 | */ | |
163 | gfp_t gfp_mask; /* mask for allocating pages */ | |
164 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
165 | }; | |
166 | ||
167 | static void | |
168 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) | |
169 | { | |
170 | ca->chain = NULL; | |
171 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
172 | ca->gfp_mask = gfp_mask; | |
173 | ca->safe_needed = safe_needed; | |
174 | } | |
175 | ||
176 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
177 | { | |
178 | void *ret; | |
179 | ||
180 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
181 | struct linked_page *lp; | |
182 | ||
8357376d | 183 | lp = get_image_page(ca->gfp_mask, ca->safe_needed); |
b788db79 RW |
184 | if (!lp) |
185 | return NULL; | |
186 | ||
187 | lp->next = ca->chain; | |
188 | ca->chain = lp; | |
189 | ca->used_space = 0; | |
190 | } | |
191 | ret = ca->chain->data + ca->used_space; | |
192 | ca->used_space += size; | |
193 | return ret; | |
194 | } | |
195 | ||
196 | static void chain_free(struct chain_allocator *ca, int clear_page_nosave) | |
197 | { | |
198 | free_list_of_pages(ca->chain, clear_page_nosave); | |
199 | memset(ca, 0, sizeof(struct chain_allocator)); | |
200 | } | |
201 | ||
202 | /** | |
203 | * Data types related to memory bitmaps. | |
204 | * | |
205 | * Memory bitmap is a structure consiting of many linked lists of | |
206 | * objects. The main list's elements are of type struct zone_bitmap | |
207 | * and each of them corresonds to one zone. For each zone bitmap | |
208 | * object there is a list of objects of type struct bm_block that | |
209 | * represent each blocks of bit chunks in which information is | |
210 | * stored. | |
211 | * | |
212 | * struct memory_bitmap contains a pointer to the main list of zone | |
213 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
214 | * and a pointer to the list of pages used for allocating all of the | |
215 | * zone bitmap objects and bitmap block objects. | |
216 | * | |
217 | * NOTE: It has to be possible to lay out the bitmap in memory | |
218 | * using only allocations of order 0. Additionally, the bitmap is | |
219 | * designed to work with arbitrary number of zones (this is over the | |
220 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
221 | * | |
222 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
223 | * objects and a pointer to the bitmap block object that has been | |
224 | * most recently used for setting bits. Additionally, it contains the | |
225 | * pfns that correspond to the start and end of the represented zone. | |
226 | * | |
227 | * struct bm_block contains a pointer to the memory page in which | |
228 | * information is stored (in the form of a block of bit chunks | |
229 | * of type unsigned long each). It also contains the pfns that | |
230 | * correspond to the start and end of the represented memory area and | |
231 | * the number of bit chunks in the block. | |
b788db79 RW |
232 | */ |
233 | ||
234 | #define BM_END_OF_MAP (~0UL) | |
235 | ||
236 | #define BM_CHUNKS_PER_BLOCK (PAGE_SIZE / sizeof(long)) | |
237 | #define BM_BITS_PER_CHUNK (sizeof(long) << 3) | |
238 | #define BM_BITS_PER_BLOCK (PAGE_SIZE << 3) | |
239 | ||
240 | struct bm_block { | |
241 | struct bm_block *next; /* next element of the list */ | |
242 | unsigned long start_pfn; /* pfn represented by the first bit */ | |
243 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ | |
244 | unsigned int size; /* number of bit chunks */ | |
245 | unsigned long *data; /* chunks of bits representing pages */ | |
246 | }; | |
247 | ||
248 | struct zone_bitmap { | |
249 | struct zone_bitmap *next; /* next element of the list */ | |
250 | unsigned long start_pfn; /* minimal pfn in this zone */ | |
251 | unsigned long end_pfn; /* maximal pfn in this zone plus 1 */ | |
252 | struct bm_block *bm_blocks; /* list of bitmap blocks */ | |
253 | struct bm_block *cur_block; /* recently used bitmap block */ | |
254 | }; | |
255 | ||
256 | /* strcut bm_position is used for browsing memory bitmaps */ | |
257 | ||
258 | struct bm_position { | |
259 | struct zone_bitmap *zone_bm; | |
260 | struct bm_block *block; | |
261 | int chunk; | |
262 | int bit; | |
263 | }; | |
264 | ||
265 | struct memory_bitmap { | |
266 | struct zone_bitmap *zone_bm_list; /* list of zone bitmaps */ | |
267 | struct linked_page *p_list; /* list of pages used to store zone | |
268 | * bitmap objects and bitmap block | |
269 | * objects | |
270 | */ | |
271 | struct bm_position cur; /* most recently used bit position */ | |
272 | }; | |
273 | ||
274 | /* Functions that operate on memory bitmaps */ | |
275 | ||
276 | static inline void memory_bm_reset_chunk(struct memory_bitmap *bm) | |
277 | { | |
278 | bm->cur.chunk = 0; | |
279 | bm->cur.bit = -1; | |
280 | } | |
281 | ||
282 | static void memory_bm_position_reset(struct memory_bitmap *bm) | |
283 | { | |
284 | struct zone_bitmap *zone_bm; | |
285 | ||
286 | zone_bm = bm->zone_bm_list; | |
287 | bm->cur.zone_bm = zone_bm; | |
288 | bm->cur.block = zone_bm->bm_blocks; | |
289 | memory_bm_reset_chunk(bm); | |
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 | |
296 | */ | |
297 | ||
298 | static inline struct bm_block * | |
299 | create_bm_block_list(unsigned int nr_blocks, struct chain_allocator *ca) | |
300 | { | |
301 | struct bm_block *bblist = NULL; | |
302 | ||
303 | while (nr_blocks-- > 0) { | |
304 | struct bm_block *bb; | |
305 | ||
306 | bb = chain_alloc(ca, sizeof(struct bm_block)); | |
307 | if (!bb) | |
308 | return NULL; | |
309 | ||
310 | bb->next = bblist; | |
311 | bblist = bb; | |
312 | } | |
313 | return bblist; | |
314 | } | |
315 | ||
316 | /** | |
317 | * create_zone_bm_list - create a list of zone bitmap objects | |
318 | */ | |
319 | ||
320 | static inline struct zone_bitmap * | |
321 | create_zone_bm_list(unsigned int nr_zones, struct chain_allocator *ca) | |
322 | { | |
323 | struct zone_bitmap *zbmlist = NULL; | |
324 | ||
325 | while (nr_zones-- > 0) { | |
326 | struct zone_bitmap *zbm; | |
327 | ||
328 | zbm = chain_alloc(ca, sizeof(struct zone_bitmap)); | |
329 | if (!zbm) | |
330 | return NULL; | |
331 | ||
332 | zbm->next = zbmlist; | |
333 | zbmlist = zbm; | |
334 | } | |
335 | return zbmlist; | |
336 | } | |
337 | ||
338 | /** | |
339 | * memory_bm_create - allocate memory for a memory bitmap | |
340 | */ | |
341 | ||
342 | static int | |
343 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) | |
344 | { | |
345 | struct chain_allocator ca; | |
346 | struct zone *zone; | |
347 | struct zone_bitmap *zone_bm; | |
348 | struct bm_block *bb; | |
349 | unsigned int nr; | |
350 | ||
351 | chain_init(&ca, gfp_mask, safe_needed); | |
352 | ||
353 | /* Compute the number of zones */ | |
354 | nr = 0; | |
8357376d RW |
355 | for_each_zone(zone) |
356 | if (populated_zone(zone)) | |
b788db79 RW |
357 | nr++; |
358 | ||
359 | /* Allocate the list of zones bitmap objects */ | |
360 | zone_bm = create_zone_bm_list(nr, &ca); | |
361 | bm->zone_bm_list = zone_bm; | |
362 | if (!zone_bm) { | |
363 | chain_free(&ca, PG_UNSAFE_CLEAR); | |
364 | return -ENOMEM; | |
365 | } | |
366 | ||
367 | /* Initialize the zone bitmap objects */ | |
8357376d | 368 | for_each_zone(zone) { |
b788db79 RW |
369 | unsigned long pfn; |
370 | ||
8357376d | 371 | if (!populated_zone(zone)) |
b788db79 RW |
372 | continue; |
373 | ||
374 | zone_bm->start_pfn = zone->zone_start_pfn; | |
375 | zone_bm->end_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
376 | /* Allocate the list of bitmap block objects */ | |
377 | nr = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
378 | bb = create_bm_block_list(nr, &ca); | |
379 | zone_bm->bm_blocks = bb; | |
380 | zone_bm->cur_block = bb; | |
381 | if (!bb) | |
382 | goto Free; | |
383 | ||
384 | nr = zone->spanned_pages; | |
385 | pfn = zone->zone_start_pfn; | |
386 | /* Initialize the bitmap block objects */ | |
387 | while (bb) { | |
388 | unsigned long *ptr; | |
389 | ||
8357376d | 390 | ptr = get_image_page(gfp_mask, safe_needed); |
b788db79 RW |
391 | bb->data = ptr; |
392 | if (!ptr) | |
393 | goto Free; | |
394 | ||
395 | bb->start_pfn = pfn; | |
396 | if (nr >= BM_BITS_PER_BLOCK) { | |
397 | pfn += BM_BITS_PER_BLOCK; | |
398 | bb->size = BM_CHUNKS_PER_BLOCK; | |
399 | nr -= BM_BITS_PER_BLOCK; | |
400 | } else { | |
401 | /* This is executed only once in the loop */ | |
402 | pfn += nr; | |
403 | bb->size = DIV_ROUND_UP(nr, BM_BITS_PER_CHUNK); | |
404 | } | |
405 | bb->end_pfn = pfn; | |
406 | bb = bb->next; | |
407 | } | |
408 | zone_bm = zone_bm->next; | |
409 | } | |
410 | bm->p_list = ca.chain; | |
411 | memory_bm_position_reset(bm); | |
412 | return 0; | |
413 | ||
59a49335 | 414 | Free: |
b788db79 RW |
415 | bm->p_list = ca.chain; |
416 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
417 | return -ENOMEM; | |
418 | } | |
419 | ||
420 | /** | |
421 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
422 | */ | |
423 | ||
424 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) | |
425 | { | |
426 | struct zone_bitmap *zone_bm; | |
427 | ||
428 | /* Free the list of bit blocks for each zone_bitmap object */ | |
429 | zone_bm = bm->zone_bm_list; | |
430 | while (zone_bm) { | |
431 | struct bm_block *bb; | |
432 | ||
433 | bb = zone_bm->bm_blocks; | |
434 | while (bb) { | |
435 | if (bb->data) | |
436 | free_image_page(bb->data, clear_nosave_free); | |
437 | bb = bb->next; | |
438 | } | |
439 | zone_bm = zone_bm->next; | |
440 | } | |
441 | free_list_of_pages(bm->p_list, clear_nosave_free); | |
442 | bm->zone_bm_list = NULL; | |
443 | } | |
444 | ||
445 | /** | |
74dfd666 | 446 | * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds |
b788db79 RW |
447 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member |
448 | * of @bm->cur_zone_bm are updated. | |
b788db79 RW |
449 | */ |
450 | ||
74dfd666 RW |
451 | static void memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
452 | void **addr, unsigned int *bit_nr) | |
b788db79 RW |
453 | { |
454 | struct zone_bitmap *zone_bm; | |
455 | struct bm_block *bb; | |
456 | ||
457 | /* Check if the pfn is from the current zone */ | |
458 | zone_bm = bm->cur.zone_bm; | |
459 | if (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) { | |
460 | zone_bm = bm->zone_bm_list; | |
461 | /* We don't assume that the zones are sorted by pfns */ | |
462 | while (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) { | |
463 | zone_bm = zone_bm->next; | |
74dfd666 RW |
464 | |
465 | BUG_ON(!zone_bm); | |
b788db79 RW |
466 | } |
467 | bm->cur.zone_bm = zone_bm; | |
468 | } | |
469 | /* Check if the pfn corresponds to the current bitmap block */ | |
470 | bb = zone_bm->cur_block; | |
471 | if (pfn < bb->start_pfn) | |
472 | bb = zone_bm->bm_blocks; | |
473 | ||
474 | while (pfn >= bb->end_pfn) { | |
475 | bb = bb->next; | |
74dfd666 RW |
476 | |
477 | BUG_ON(!bb); | |
b788db79 RW |
478 | } |
479 | zone_bm->cur_block = bb; | |
480 | pfn -= bb->start_pfn; | |
74dfd666 RW |
481 | *bit_nr = pfn % BM_BITS_PER_CHUNK; |
482 | *addr = bb->data + pfn / BM_BITS_PER_CHUNK; | |
483 | } | |
484 | ||
485 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) | |
486 | { | |
487 | void *addr; | |
488 | unsigned int bit; | |
489 | ||
490 | memory_bm_find_bit(bm, pfn, &addr, &bit); | |
491 | set_bit(bit, addr); | |
492 | } | |
493 | ||
494 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) | |
495 | { | |
496 | void *addr; | |
497 | unsigned int bit; | |
498 | ||
499 | memory_bm_find_bit(bm, pfn, &addr, &bit); | |
500 | clear_bit(bit, addr); | |
501 | } | |
502 | ||
503 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) | |
504 | { | |
505 | void *addr; | |
506 | unsigned int bit; | |
507 | ||
508 | memory_bm_find_bit(bm, pfn, &addr, &bit); | |
509 | return test_bit(bit, addr); | |
b788db79 RW |
510 | } |
511 | ||
512 | /* Two auxiliary functions for memory_bm_next_pfn */ | |
513 | ||
514 | /* Find the first set bit in the given chunk, if there is one */ | |
515 | ||
516 | static inline int next_bit_in_chunk(int bit, unsigned long *chunk_p) | |
517 | { | |
518 | bit++; | |
519 | while (bit < BM_BITS_PER_CHUNK) { | |
520 | if (test_bit(bit, chunk_p)) | |
521 | return bit; | |
522 | ||
523 | bit++; | |
524 | } | |
525 | return -1; | |
526 | } | |
527 | ||
528 | /* Find a chunk containing some bits set in given block of bits */ | |
529 | ||
530 | static inline int next_chunk_in_block(int n, struct bm_block *bb) | |
531 | { | |
532 | n++; | |
533 | while (n < bb->size) { | |
534 | if (bb->data[n]) | |
535 | return n; | |
536 | ||
537 | n++; | |
538 | } | |
539 | return -1; | |
540 | } | |
541 | ||
542 | /** | |
543 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit | |
544 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is | |
545 | * returned. | |
546 | * | |
547 | * It is required to run memory_bm_position_reset() before the first call to | |
548 | * this function. | |
549 | */ | |
550 | ||
551 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) | |
552 | { | |
553 | struct zone_bitmap *zone_bm; | |
554 | struct bm_block *bb; | |
555 | int chunk; | |
556 | int bit; | |
557 | ||
558 | do { | |
559 | bb = bm->cur.block; | |
560 | do { | |
561 | chunk = bm->cur.chunk; | |
562 | bit = bm->cur.bit; | |
563 | do { | |
564 | bit = next_bit_in_chunk(bit, bb->data + chunk); | |
565 | if (bit >= 0) | |
566 | goto Return_pfn; | |
567 | ||
568 | chunk = next_chunk_in_block(chunk, bb); | |
569 | bit = -1; | |
570 | } while (chunk >= 0); | |
571 | bb = bb->next; | |
572 | bm->cur.block = bb; | |
573 | memory_bm_reset_chunk(bm); | |
574 | } while (bb); | |
575 | zone_bm = bm->cur.zone_bm->next; | |
576 | if (zone_bm) { | |
577 | bm->cur.zone_bm = zone_bm; | |
578 | bm->cur.block = zone_bm->bm_blocks; | |
579 | memory_bm_reset_chunk(bm); | |
580 | } | |
581 | } while (zone_bm); | |
582 | memory_bm_position_reset(bm); | |
583 | return BM_END_OF_MAP; | |
584 | ||
59a49335 | 585 | Return_pfn: |
b788db79 RW |
586 | bm->cur.chunk = chunk; |
587 | bm->cur.bit = bit; | |
588 | return bb->start_pfn + chunk * BM_BITS_PER_CHUNK + bit; | |
589 | } | |
590 | ||
74dfd666 RW |
591 | /** |
592 | * This structure represents a range of page frames the contents of which | |
593 | * should not be saved during the suspend. | |
594 | */ | |
595 | ||
596 | struct nosave_region { | |
597 | struct list_head list; | |
598 | unsigned long start_pfn; | |
599 | unsigned long end_pfn; | |
600 | }; | |
601 | ||
602 | static LIST_HEAD(nosave_regions); | |
603 | ||
604 | /** | |
605 | * register_nosave_region - register a range of page frames the contents | |
606 | * of which should not be saved during the suspend (to be used in the early | |
607 | * initialization code) | |
608 | */ | |
609 | ||
610 | void __init | |
611 | register_nosave_region(unsigned long start_pfn, unsigned long end_pfn) | |
612 | { | |
613 | struct nosave_region *region; | |
614 | ||
615 | if (start_pfn >= end_pfn) | |
616 | return; | |
617 | ||
618 | if (!list_empty(&nosave_regions)) { | |
619 | /* Try to extend the previous region (they should be sorted) */ | |
620 | region = list_entry(nosave_regions.prev, | |
621 | struct nosave_region, list); | |
622 | if (region->end_pfn == start_pfn) { | |
623 | region->end_pfn = end_pfn; | |
624 | goto Report; | |
625 | } | |
626 | } | |
627 | /* This allocation cannot fail */ | |
628 | region = alloc_bootmem_low(sizeof(struct nosave_region)); | |
629 | region->start_pfn = start_pfn; | |
630 | region->end_pfn = end_pfn; | |
631 | list_add_tail(®ion->list, &nosave_regions); | |
632 | Report: | |
633 | printk("swsusp: Registered nosave memory region: %016lx - %016lx\n", | |
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 | ||
702 | printk("swsusp: Marking nosave pages: %016lx - %016lx\n", | |
703 | region->start_pfn << PAGE_SHIFT, | |
704 | region->end_pfn << PAGE_SHIFT); | |
705 | ||
706 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
707 | memory_bm_set_bit(bm, pfn); | |
708 | } | |
709 | } | |
710 | ||
711 | /** | |
712 | * create_basic_memory_bitmaps - create bitmaps needed for marking page | |
713 | * frames that should not be saved and free page frames. The pointers | |
714 | * forbidden_pages_map and free_pages_map are only modified if everything | |
715 | * goes well, because we don't want the bits to be used before both bitmaps | |
716 | * are set up. | |
717 | */ | |
718 | ||
719 | int create_basic_memory_bitmaps(void) | |
720 | { | |
721 | struct memory_bitmap *bm1, *bm2; | |
722 | int error = 0; | |
723 | ||
724 | BUG_ON(forbidden_pages_map || free_pages_map); | |
725 | ||
726 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC); | |
727 | if (!bm1) | |
728 | return -ENOMEM; | |
729 | ||
730 | error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY); | |
731 | if (error) | |
732 | goto Free_first_object; | |
733 | ||
734 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC); | |
735 | if (!bm2) | |
736 | goto Free_first_bitmap; | |
737 | ||
738 | error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY); | |
739 | if (error) | |
740 | goto Free_second_object; | |
741 | ||
742 | forbidden_pages_map = bm1; | |
743 | free_pages_map = bm2; | |
744 | mark_nosave_pages(forbidden_pages_map); | |
745 | ||
746 | printk("swsusp: Basic memory bitmaps created\n"); | |
747 | ||
748 | return 0; | |
749 | ||
750 | Free_second_object: | |
751 | kfree(bm2); | |
752 | Free_first_bitmap: | |
753 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
754 | Free_first_object: | |
755 | kfree(bm1); | |
756 | return -ENOMEM; | |
757 | } | |
758 | ||
759 | /** | |
760 | * free_basic_memory_bitmaps - free memory bitmaps allocated by | |
761 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary | |
762 | * so that the bitmaps themselves are not referred to while they are being | |
763 | * freed. | |
764 | */ | |
765 | ||
766 | void free_basic_memory_bitmaps(void) | |
767 | { | |
768 | struct memory_bitmap *bm1, *bm2; | |
769 | ||
770 | BUG_ON(!(forbidden_pages_map && free_pages_map)); | |
771 | ||
772 | bm1 = forbidden_pages_map; | |
773 | bm2 = free_pages_map; | |
774 | forbidden_pages_map = NULL; | |
775 | free_pages_map = NULL; | |
776 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
777 | kfree(bm1); | |
778 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
779 | kfree(bm2); | |
780 | ||
781 | printk("swsusp: Basic memory bitmaps freed\n"); | |
782 | } | |
783 | ||
b788db79 RW |
784 | /** |
785 | * snapshot_additional_pages - estimate the number of additional pages | |
786 | * be needed for setting up the suspend image data structures for given | |
787 | * zone (usually the returned value is greater than the exact number) | |
788 | */ | |
789 | ||
790 | unsigned int snapshot_additional_pages(struct zone *zone) | |
791 | { | |
792 | unsigned int res; | |
793 | ||
794 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
795 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), PAGE_SIZE); | |
8357376d | 796 | return 2 * res; |
b788db79 RW |
797 | } |
798 | ||
8357376d RW |
799 | #ifdef CONFIG_HIGHMEM |
800 | /** | |
801 | * count_free_highmem_pages - compute the total number of free highmem | |
802 | * pages, system-wide. | |
803 | */ | |
804 | ||
805 | static unsigned int count_free_highmem_pages(void) | |
806 | { | |
807 | struct zone *zone; | |
808 | unsigned int cnt = 0; | |
809 | ||
810 | for_each_zone(zone) | |
811 | if (populated_zone(zone) && is_highmem(zone)) | |
d23ad423 | 812 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
813 | |
814 | return cnt; | |
815 | } | |
816 | ||
817 | /** | |
818 | * saveable_highmem_page - Determine whether a highmem page should be | |
819 | * included in the suspend image. | |
820 | * | |
821 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, | |
822 | * and it isn't a part of a free chunk of pages. | |
823 | */ | |
824 | ||
825 | static struct page *saveable_highmem_page(unsigned long pfn) | |
826 | { | |
827 | struct page *page; | |
828 | ||
829 | if (!pfn_valid(pfn)) | |
830 | return NULL; | |
831 | ||
832 | page = pfn_to_page(pfn); | |
833 | ||
834 | BUG_ON(!PageHighMem(page)); | |
835 | ||
7be98234 RW |
836 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
837 | PageReserved(page)) | |
8357376d RW |
838 | return NULL; |
839 | ||
840 | return page; | |
841 | } | |
842 | ||
843 | /** | |
844 | * count_highmem_pages - compute the total number of saveable highmem | |
845 | * pages. | |
846 | */ | |
847 | ||
848 | unsigned int count_highmem_pages(void) | |
849 | { | |
850 | struct zone *zone; | |
851 | unsigned int n = 0; | |
852 | ||
853 | for_each_zone(zone) { | |
854 | unsigned long pfn, max_zone_pfn; | |
855 | ||
856 | if (!is_highmem(zone)) | |
857 | continue; | |
858 | ||
859 | mark_free_pages(zone); | |
860 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
861 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
862 | if (saveable_highmem_page(pfn)) | |
863 | n++; | |
864 | } | |
865 | return n; | |
866 | } | |
867 | #else | |
868 | static inline void *saveable_highmem_page(unsigned long pfn) { return NULL; } | |
869 | static inline unsigned int count_highmem_pages(void) { return 0; } | |
870 | #endif /* CONFIG_HIGHMEM */ | |
871 | ||
25761b6e | 872 | /** |
8357376d RW |
873 | * saveable - Determine whether a non-highmem page should be included in |
874 | * the suspend image. | |
25761b6e | 875 | * |
8357376d RW |
876 | * We should save the page if it isn't Nosave, and is not in the range |
877 | * of pages statically defined as 'unsaveable', and it isn't a part of | |
878 | * a free chunk of pages. | |
25761b6e RW |
879 | */ |
880 | ||
ae83c5ee | 881 | static struct page *saveable_page(unsigned long pfn) |
25761b6e | 882 | { |
de491861 | 883 | struct page *page; |
25761b6e RW |
884 | |
885 | if (!pfn_valid(pfn)) | |
ae83c5ee | 886 | return NULL; |
25761b6e RW |
887 | |
888 | page = pfn_to_page(pfn); | |
ae83c5ee | 889 | |
8357376d RW |
890 | BUG_ON(PageHighMem(page)); |
891 | ||
7be98234 | 892 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 893 | return NULL; |
8357376d | 894 | |
72a97e08 | 895 | if (PageReserved(page) && pfn_is_nosave(pfn)) |
ae83c5ee | 896 | return NULL; |
25761b6e | 897 | |
ae83c5ee | 898 | return page; |
25761b6e RW |
899 | } |
900 | ||
8357376d RW |
901 | /** |
902 | * count_data_pages - compute the total number of saveable non-highmem | |
903 | * pages. | |
904 | */ | |
905 | ||
72a97e08 | 906 | unsigned int count_data_pages(void) |
25761b6e RW |
907 | { |
908 | struct zone *zone; | |
ae83c5ee | 909 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 910 | unsigned int n = 0; |
25761b6e | 911 | |
8357376d | 912 | for_each_zone(zone) { |
25761b6e RW |
913 | if (is_highmem(zone)) |
914 | continue; | |
8357376d | 915 | |
25761b6e | 916 | mark_free_pages(zone); |
ae83c5ee RW |
917 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
918 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
8357376d RW |
919 | if(saveable_page(pfn)) |
920 | n++; | |
25761b6e | 921 | } |
a0f49651 | 922 | return n; |
25761b6e RW |
923 | } |
924 | ||
8357376d RW |
925 | /* This is needed, because copy_page and memcpy are not usable for copying |
926 | * task structs. | |
927 | */ | |
928 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
929 | { |
930 | int n; | |
931 | ||
f623f0db RW |
932 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
933 | *dst++ = *src++; | |
934 | } | |
935 | ||
8357376d RW |
936 | #ifdef CONFIG_HIGHMEM |
937 | static inline struct page * | |
938 | page_is_saveable(struct zone *zone, unsigned long pfn) | |
939 | { | |
940 | return is_highmem(zone) ? | |
941 | saveable_highmem_page(pfn) : saveable_page(pfn); | |
942 | } | |
943 | ||
944 | static inline void | |
945 | copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) | |
946 | { | |
947 | struct page *s_page, *d_page; | |
948 | void *src, *dst; | |
949 | ||
950 | s_page = pfn_to_page(src_pfn); | |
951 | d_page = pfn_to_page(dst_pfn); | |
952 | if (PageHighMem(s_page)) { | |
953 | src = kmap_atomic(s_page, KM_USER0); | |
954 | dst = kmap_atomic(d_page, KM_USER1); | |
955 | do_copy_page(dst, src); | |
956 | kunmap_atomic(src, KM_USER0); | |
957 | kunmap_atomic(dst, KM_USER1); | |
958 | } else { | |
959 | src = page_address(s_page); | |
960 | if (PageHighMem(d_page)) { | |
961 | /* Page pointed to by src may contain some kernel | |
962 | * data modified by kmap_atomic() | |
963 | */ | |
964 | do_copy_page(buffer, src); | |
965 | dst = kmap_atomic(pfn_to_page(dst_pfn), KM_USER0); | |
966 | memcpy(dst, buffer, PAGE_SIZE); | |
967 | kunmap_atomic(dst, KM_USER0); | |
968 | } else { | |
969 | dst = page_address(d_page); | |
970 | do_copy_page(dst, src); | |
971 | } | |
972 | } | |
973 | } | |
974 | #else | |
975 | #define page_is_saveable(zone, pfn) saveable_page(pfn) | |
976 | ||
977 | static inline void | |
978 | copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) | |
979 | { | |
980 | do_copy_page(page_address(pfn_to_page(dst_pfn)), | |
981 | page_address(pfn_to_page(src_pfn))); | |
982 | } | |
983 | #endif /* CONFIG_HIGHMEM */ | |
984 | ||
b788db79 RW |
985 | static void |
986 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) | |
25761b6e RW |
987 | { |
988 | struct zone *zone; | |
b788db79 | 989 | unsigned long pfn; |
25761b6e | 990 | |
8357376d | 991 | for_each_zone(zone) { |
b788db79 RW |
992 | unsigned long max_zone_pfn; |
993 | ||
25761b6e | 994 | mark_free_pages(zone); |
ae83c5ee | 995 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
b788db79 | 996 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 997 | if (page_is_saveable(zone, pfn)) |
b788db79 | 998 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 999 | } |
b788db79 RW |
1000 | memory_bm_position_reset(orig_bm); |
1001 | memory_bm_position_reset(copy_bm); | |
1002 | do { | |
1003 | pfn = memory_bm_next_pfn(orig_bm); | |
8357376d RW |
1004 | if (likely(pfn != BM_END_OF_MAP)) |
1005 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
b788db79 | 1006 | } while (pfn != BM_END_OF_MAP); |
25761b6e RW |
1007 | } |
1008 | ||
8357376d RW |
1009 | /* Total number of image pages */ |
1010 | static unsigned int nr_copy_pages; | |
1011 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1012 | static unsigned int nr_meta_pages; | |
1013 | ||
25761b6e | 1014 | /** |
940864dd | 1015 | * swsusp_free - free pages allocated for the suspend. |
cd560bb2 | 1016 | * |
940864dd RW |
1017 | * Suspend pages are alocated before the atomic copy is made, so we |
1018 | * need to release them after the resume. | |
25761b6e RW |
1019 | */ |
1020 | ||
1021 | void swsusp_free(void) | |
1022 | { | |
1023 | struct zone *zone; | |
ae83c5ee | 1024 | unsigned long pfn, max_zone_pfn; |
25761b6e RW |
1025 | |
1026 | for_each_zone(zone) { | |
ae83c5ee RW |
1027 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1028 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1029 | if (pfn_valid(pfn)) { | |
1030 | struct page *page = pfn_to_page(pfn); | |
1031 | ||
7be98234 RW |
1032 | if (swsusp_page_is_forbidden(page) && |
1033 | swsusp_page_is_free(page)) { | |
1034 | swsusp_unset_page_forbidden(page); | |
1035 | swsusp_unset_page_free(page); | |
8357376d | 1036 | __free_page(page); |
25761b6e RW |
1037 | } |
1038 | } | |
1039 | } | |
f577eb30 RW |
1040 | nr_copy_pages = 0; |
1041 | nr_meta_pages = 0; | |
75534b50 | 1042 | restore_pblist = NULL; |
6e1819d6 | 1043 | buffer = NULL; |
25761b6e RW |
1044 | } |
1045 | ||
8357376d RW |
1046 | #ifdef CONFIG_HIGHMEM |
1047 | /** | |
1048 | * count_pages_for_highmem - compute the number of non-highmem pages | |
1049 | * that will be necessary for creating copies of highmem pages. | |
1050 | */ | |
1051 | ||
1052 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) | |
1053 | { | |
1054 | unsigned int free_highmem = count_free_highmem_pages(); | |
1055 | ||
1056 | if (free_highmem >= nr_highmem) | |
1057 | nr_highmem = 0; | |
1058 | else | |
1059 | nr_highmem -= free_highmem; | |
1060 | ||
1061 | return nr_highmem; | |
1062 | } | |
1063 | #else | |
1064 | static unsigned int | |
1065 | count_pages_for_highmem(unsigned int nr_highmem) { return 0; } | |
1066 | #endif /* CONFIG_HIGHMEM */ | |
25761b6e RW |
1067 | |
1068 | /** | |
8357376d RW |
1069 | * enough_free_mem - Make sure we have enough free memory for the |
1070 | * snapshot image. | |
25761b6e RW |
1071 | */ |
1072 | ||
8357376d | 1073 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1074 | { |
e5e2fa78 | 1075 | struct zone *zone; |
940864dd | 1076 | unsigned int free = 0, meta = 0; |
e5e2fa78 | 1077 | |
8357376d RW |
1078 | for_each_zone(zone) { |
1079 | meta += snapshot_additional_pages(zone); | |
1080 | if (!is_highmem(zone)) | |
d23ad423 | 1081 | free += zone_page_state(zone, NR_FREE_PAGES); |
8357376d | 1082 | } |
940864dd | 1083 | |
8357376d RW |
1084 | nr_pages += count_pages_for_highmem(nr_highmem); |
1085 | pr_debug("swsusp: Normal pages needed: %u + %u + %u, available pages: %u\n", | |
940864dd RW |
1086 | nr_pages, PAGES_FOR_IO, meta, free); |
1087 | ||
1088 | return free > nr_pages + PAGES_FOR_IO + meta; | |
25761b6e RW |
1089 | } |
1090 | ||
8357376d RW |
1091 | #ifdef CONFIG_HIGHMEM |
1092 | /** | |
1093 | * get_highmem_buffer - if there are some highmem pages in the suspend | |
1094 | * image, we may need the buffer to copy them and/or load their data. | |
1095 | */ | |
1096 | ||
1097 | static inline int get_highmem_buffer(int safe_needed) | |
1098 | { | |
1099 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1100 | return buffer ? 0 : -ENOMEM; | |
1101 | } | |
1102 | ||
1103 | /** | |
1104 | * alloc_highmem_image_pages - allocate some highmem pages for the image. | |
1105 | * Try to allocate as many pages as needed, but if the number of free | |
1106 | * highmem pages is lesser than that, allocate them all. | |
1107 | */ | |
1108 | ||
1109 | static inline unsigned int | |
1110 | alloc_highmem_image_pages(struct memory_bitmap *bm, unsigned int nr_highmem) | |
1111 | { | |
1112 | unsigned int to_alloc = count_free_highmem_pages(); | |
1113 | ||
1114 | if (to_alloc > nr_highmem) | |
1115 | to_alloc = nr_highmem; | |
1116 | ||
1117 | nr_highmem -= to_alloc; | |
1118 | while (to_alloc-- > 0) { | |
1119 | struct page *page; | |
1120 | ||
1121 | page = alloc_image_page(__GFP_HIGHMEM); | |
1122 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1123 | } | |
1124 | return nr_highmem; | |
1125 | } | |
1126 | #else | |
1127 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1128 | ||
1129 | static inline unsigned int | |
1130 | alloc_highmem_image_pages(struct memory_bitmap *bm, unsigned int n) { return 0; } | |
1131 | #endif /* CONFIG_HIGHMEM */ | |
1132 | ||
1133 | /** | |
1134 | * swsusp_alloc - allocate memory for the suspend image | |
1135 | * | |
1136 | * We first try to allocate as many highmem pages as there are | |
1137 | * saveable highmem pages in the system. If that fails, we allocate | |
1138 | * non-highmem pages for the copies of the remaining highmem ones. | |
1139 | * | |
1140 | * In this approach it is likely that the copies of highmem pages will | |
1141 | * also be located in the high memory, because of the way in which | |
1142 | * copy_data_pages() works. | |
1143 | */ | |
1144 | ||
b788db79 RW |
1145 | static int |
1146 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, | |
8357376d | 1147 | unsigned int nr_pages, unsigned int nr_highmem) |
054bd4c1 | 1148 | { |
b788db79 | 1149 | int error; |
054bd4c1 | 1150 | |
b788db79 RW |
1151 | error = memory_bm_create(orig_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY); |
1152 | if (error) | |
1153 | goto Free; | |
25761b6e | 1154 | |
b788db79 RW |
1155 | error = memory_bm_create(copy_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY); |
1156 | if (error) | |
1157 | goto Free; | |
25761b6e | 1158 | |
8357376d RW |
1159 | if (nr_highmem > 0) { |
1160 | error = get_highmem_buffer(PG_ANY); | |
1161 | if (error) | |
1162 | goto Free; | |
1163 | ||
1164 | nr_pages += alloc_highmem_image_pages(copy_bm, nr_highmem); | |
1165 | } | |
b788db79 | 1166 | while (nr_pages-- > 0) { |
8357376d RW |
1167 | struct page *page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); |
1168 | ||
b788db79 RW |
1169 | if (!page) |
1170 | goto Free; | |
25761b6e | 1171 | |
b788db79 | 1172 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); |
25761b6e | 1173 | } |
b788db79 | 1174 | return 0; |
25761b6e | 1175 | |
59a49335 | 1176 | Free: |
b788db79 RW |
1177 | swsusp_free(); |
1178 | return -ENOMEM; | |
25761b6e RW |
1179 | } |
1180 | ||
8357376d RW |
1181 | /* Memory bitmap used for marking saveable pages (during suspend) or the |
1182 | * suspend image pages (during resume) | |
1183 | */ | |
b788db79 | 1184 | static struct memory_bitmap orig_bm; |
8357376d RW |
1185 | /* Memory bitmap used on suspend for marking allocated pages that will contain |
1186 | * the copies of saveable pages. During resume it is initially used for | |
1187 | * marking the suspend image pages, but then its set bits are duplicated in | |
1188 | * @orig_bm and it is released. Next, on systems with high memory, it may be | |
1189 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1190 | * this purpose. | |
b788db79 RW |
1191 | */ |
1192 | static struct memory_bitmap copy_bm; | |
1193 | ||
2e32a43e | 1194 | asmlinkage int swsusp_save(void) |
25761b6e | 1195 | { |
8357376d | 1196 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1197 | |
8357376d | 1198 | printk("swsusp: critical section: \n"); |
25761b6e RW |
1199 | |
1200 | drain_local_pages(); | |
a0f49651 | 1201 | nr_pages = count_data_pages(); |
8357376d RW |
1202 | nr_highmem = count_highmem_pages(); |
1203 | printk("swsusp: Need to copy %u pages\n", nr_pages + nr_highmem); | |
25761b6e | 1204 | |
8357376d | 1205 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
25761b6e RW |
1206 | printk(KERN_ERR "swsusp: Not enough free memory\n"); |
1207 | return -ENOMEM; | |
1208 | } | |
1209 | ||
8357376d RW |
1210 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
1211 | printk(KERN_ERR "swsusp: Memory allocation failed\n"); | |
a0f49651 | 1212 | return -ENOMEM; |
8357376d | 1213 | } |
25761b6e RW |
1214 | |
1215 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1216 | * Kill them. | |
1217 | */ | |
1218 | drain_local_pages(); | |
b788db79 | 1219 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1220 | |
1221 | /* | |
1222 | * End of critical section. From now on, we can write to memory, | |
1223 | * but we should not touch disk. This specially means we must _not_ | |
1224 | * touch swap space! Except we must write out our image of course. | |
1225 | */ | |
1226 | ||
8357376d | 1227 | nr_pages += nr_highmem; |
a0f49651 | 1228 | nr_copy_pages = nr_pages; |
8357376d | 1229 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 RW |
1230 | |
1231 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); | |
8357376d | 1232 | |
25761b6e RW |
1233 | return 0; |
1234 | } | |
f577eb30 RW |
1235 | |
1236 | static void init_header(struct swsusp_info *info) | |
1237 | { | |
1238 | memset(info, 0, sizeof(struct swsusp_info)); | |
1239 | info->version_code = LINUX_VERSION_CODE; | |
1240 | info->num_physpages = num_physpages; | |
96b644bd | 1241 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 RW |
1242 | info->cpus = num_online_cpus(); |
1243 | info->image_pages = nr_copy_pages; | |
1244 | info->pages = nr_copy_pages + nr_meta_pages + 1; | |
6e1819d6 RW |
1245 | info->size = info->pages; |
1246 | info->size <<= PAGE_SHIFT; | |
f577eb30 RW |
1247 | } |
1248 | ||
1249 | /** | |
940864dd RW |
1250 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1251 | * are stored in the array @buf[] (1 page at a time) | |
f577eb30 RW |
1252 | */ |
1253 | ||
b788db79 | 1254 | static inline void |
940864dd | 1255 | pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1256 | { |
1257 | int j; | |
1258 | ||
b788db79 | 1259 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
1260 | buf[j] = memory_bm_next_pfn(bm); |
1261 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 1262 | break; |
f577eb30 | 1263 | } |
f577eb30 RW |
1264 | } |
1265 | ||
1266 | /** | |
1267 | * snapshot_read_next - used for reading the system memory snapshot. | |
1268 | * | |
1269 | * On the first call to it @handle should point to a zeroed | |
1270 | * snapshot_handle structure. The structure gets updated and a pointer | |
1271 | * to it should be passed to this function every next time. | |
1272 | * | |
1273 | * The @count parameter should contain the number of bytes the caller | |
1274 | * wants to read from the snapshot. It must not be zero. | |
1275 | * | |
1276 | * On success the function returns a positive number. Then, the caller | |
1277 | * is allowed to read up to the returned number of bytes from the memory | |
1278 | * location computed by the data_of() macro. The number returned | |
1279 | * may be smaller than @count, but this only happens if the read would | |
1280 | * cross a page boundary otherwise. | |
1281 | * | |
1282 | * The function returns 0 to indicate the end of data stream condition, | |
1283 | * and a negative number is returned on error. In such cases the | |
1284 | * structure pointed to by @handle is not updated and should not be used | |
1285 | * any more. | |
1286 | */ | |
1287 | ||
1288 | int snapshot_read_next(struct snapshot_handle *handle, size_t count) | |
1289 | { | |
fb13a28b | 1290 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1291 | return 0; |
b788db79 | 1292 | |
f577eb30 RW |
1293 | if (!buffer) { |
1294 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 1295 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1296 | if (!buffer) |
1297 | return -ENOMEM; | |
1298 | } | |
1299 | if (!handle->offset) { | |
1300 | init_header((struct swsusp_info *)buffer); | |
1301 | handle->buffer = buffer; | |
b788db79 RW |
1302 | memory_bm_position_reset(&orig_bm); |
1303 | memory_bm_position_reset(©_bm); | |
f577eb30 | 1304 | } |
fb13a28b RW |
1305 | if (handle->prev < handle->cur) { |
1306 | if (handle->cur <= nr_meta_pages) { | |
b788db79 | 1307 | memset(buffer, 0, PAGE_SIZE); |
940864dd | 1308 | pack_pfns(buffer, &orig_bm); |
f577eb30 | 1309 | } else { |
8357376d | 1310 | struct page *page; |
b788db79 | 1311 | |
8357376d RW |
1312 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
1313 | if (PageHighMem(page)) { | |
1314 | /* Highmem pages are copied to the buffer, | |
1315 | * because we can't return with a kmapped | |
1316 | * highmem page (we may not be called again). | |
1317 | */ | |
1318 | void *kaddr; | |
1319 | ||
1320 | kaddr = kmap_atomic(page, KM_USER0); | |
1321 | memcpy(buffer, kaddr, PAGE_SIZE); | |
1322 | kunmap_atomic(kaddr, KM_USER0); | |
1323 | handle->buffer = buffer; | |
1324 | } else { | |
1325 | handle->buffer = page_address(page); | |
1326 | } | |
f577eb30 | 1327 | } |
fb13a28b | 1328 | handle->prev = handle->cur; |
f577eb30 | 1329 | } |
fb13a28b RW |
1330 | handle->buf_offset = handle->cur_offset; |
1331 | if (handle->cur_offset + count >= PAGE_SIZE) { | |
1332 | count = PAGE_SIZE - handle->cur_offset; | |
1333 | handle->cur_offset = 0; | |
1334 | handle->cur++; | |
f577eb30 | 1335 | } else { |
fb13a28b | 1336 | handle->cur_offset += count; |
f577eb30 RW |
1337 | } |
1338 | handle->offset += count; | |
1339 | return count; | |
1340 | } | |
1341 | ||
1342 | /** | |
1343 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
1344 | * the image during resume, because they conflict with the pages that | |
1345 | * had been used before suspend | |
1346 | */ | |
1347 | ||
940864dd | 1348 | static int mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 RW |
1349 | { |
1350 | struct zone *zone; | |
ae83c5ee | 1351 | unsigned long pfn, max_zone_pfn; |
f577eb30 RW |
1352 | |
1353 | /* Clear page flags */ | |
8357376d | 1354 | for_each_zone(zone) { |
ae83c5ee RW |
1355 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1356 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1357 | if (pfn_valid(pfn)) | |
7be98234 | 1358 | swsusp_unset_page_free(pfn_to_page(pfn)); |
f577eb30 RW |
1359 | } |
1360 | ||
940864dd RW |
1361 | /* Mark pages that correspond to the "original" pfns as "unsafe" */ |
1362 | memory_bm_position_reset(bm); | |
1363 | do { | |
1364 | pfn = memory_bm_next_pfn(bm); | |
1365 | if (likely(pfn != BM_END_OF_MAP)) { | |
1366 | if (likely(pfn_valid(pfn))) | |
7be98234 | 1367 | swsusp_set_page_free(pfn_to_page(pfn)); |
940864dd RW |
1368 | else |
1369 | return -EFAULT; | |
1370 | } | |
1371 | } while (pfn != BM_END_OF_MAP); | |
f577eb30 | 1372 | |
940864dd | 1373 | allocated_unsafe_pages = 0; |
968808b8 | 1374 | |
f577eb30 RW |
1375 | return 0; |
1376 | } | |
1377 | ||
940864dd RW |
1378 | static void |
1379 | duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src) | |
f577eb30 | 1380 | { |
940864dd RW |
1381 | unsigned long pfn; |
1382 | ||
1383 | memory_bm_position_reset(src); | |
1384 | pfn = memory_bm_next_pfn(src); | |
1385 | while (pfn != BM_END_OF_MAP) { | |
1386 | memory_bm_set_bit(dst, pfn); | |
1387 | pfn = memory_bm_next_pfn(src); | |
f577eb30 RW |
1388 | } |
1389 | } | |
1390 | ||
940864dd | 1391 | static inline int check_header(struct swsusp_info *info) |
f577eb30 RW |
1392 | { |
1393 | char *reason = NULL; | |
1394 | ||
1395 | if (info->version_code != LINUX_VERSION_CODE) | |
1396 | reason = "kernel version"; | |
1397 | if (info->num_physpages != num_physpages) | |
1398 | reason = "memory size"; | |
96b644bd | 1399 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) |
f577eb30 | 1400 | reason = "system type"; |
96b644bd | 1401 | if (strcmp(info->uts.release,init_utsname()->release)) |
f577eb30 | 1402 | reason = "kernel release"; |
96b644bd | 1403 | if (strcmp(info->uts.version,init_utsname()->version)) |
f577eb30 | 1404 | reason = "version"; |
96b644bd | 1405 | if (strcmp(info->uts.machine,init_utsname()->machine)) |
f577eb30 RW |
1406 | reason = "machine"; |
1407 | if (reason) { | |
1408 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | |
1409 | return -EPERM; | |
1410 | } | |
1411 | return 0; | |
1412 | } | |
1413 | ||
1414 | /** | |
1415 | * load header - check the image header and copy data from it | |
1416 | */ | |
1417 | ||
940864dd RW |
1418 | static int |
1419 | load_header(struct swsusp_info *info) | |
f577eb30 RW |
1420 | { |
1421 | int error; | |
f577eb30 | 1422 | |
940864dd | 1423 | restore_pblist = NULL; |
f577eb30 RW |
1424 | error = check_header(info); |
1425 | if (!error) { | |
f577eb30 RW |
1426 | nr_copy_pages = info->image_pages; |
1427 | nr_meta_pages = info->pages - info->image_pages - 1; | |
1428 | } | |
1429 | return error; | |
1430 | } | |
1431 | ||
1432 | /** | |
940864dd RW |
1433 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
1434 | * the corresponding bit in the memory bitmap @bm | |
f577eb30 RW |
1435 | */ |
1436 | ||
940864dd RW |
1437 | static inline void |
1438 | unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) | |
f577eb30 RW |
1439 | { |
1440 | int j; | |
1441 | ||
940864dd RW |
1442 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1443 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
1444 | break; | |
1445 | ||
1446 | memory_bm_set_bit(bm, buf[j]); | |
f577eb30 | 1447 | } |
f577eb30 RW |
1448 | } |
1449 | ||
8357376d RW |
1450 | /* List of "safe" pages that may be used to store data loaded from the suspend |
1451 | * image | |
1452 | */ | |
1453 | static struct linked_page *safe_pages_list; | |
1454 | ||
1455 | #ifdef CONFIG_HIGHMEM | |
1456 | /* struct highmem_pbe is used for creating the list of highmem pages that | |
1457 | * should be restored atomically during the resume from disk, because the page | |
1458 | * frames they have occupied before the suspend are in use. | |
1459 | */ | |
1460 | struct highmem_pbe { | |
1461 | struct page *copy_page; /* data is here now */ | |
1462 | struct page *orig_page; /* data was here before the suspend */ | |
1463 | struct highmem_pbe *next; | |
1464 | }; | |
1465 | ||
1466 | /* List of highmem PBEs needed for restoring the highmem pages that were | |
1467 | * allocated before the suspend and included in the suspend image, but have | |
1468 | * also been allocated by the "resume" kernel, so their contents cannot be | |
1469 | * written directly to their "original" page frames. | |
1470 | */ | |
1471 | static struct highmem_pbe *highmem_pblist; | |
1472 | ||
1473 | /** | |
1474 | * count_highmem_image_pages - compute the number of highmem pages in the | |
1475 | * suspend image. The bits in the memory bitmap @bm that correspond to the | |
1476 | * image pages are assumed to be set. | |
1477 | */ | |
1478 | ||
1479 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) | |
1480 | { | |
1481 | unsigned long pfn; | |
1482 | unsigned int cnt = 0; | |
1483 | ||
1484 | memory_bm_position_reset(bm); | |
1485 | pfn = memory_bm_next_pfn(bm); | |
1486 | while (pfn != BM_END_OF_MAP) { | |
1487 | if (PageHighMem(pfn_to_page(pfn))) | |
1488 | cnt++; | |
1489 | ||
1490 | pfn = memory_bm_next_pfn(bm); | |
1491 | } | |
1492 | return cnt; | |
1493 | } | |
1494 | ||
1495 | /** | |
1496 | * prepare_highmem_image - try to allocate as many highmem pages as | |
1497 | * there are highmem image pages (@nr_highmem_p points to the variable | |
1498 | * containing the number of highmem image pages). The pages that are | |
1499 | * "safe" (ie. will not be overwritten when the suspend image is | |
1500 | * restored) have the corresponding bits set in @bm (it must be | |
1501 | * unitialized). | |
1502 | * | |
1503 | * NOTE: This function should not be called if there are no highmem | |
1504 | * image pages. | |
1505 | */ | |
1506 | ||
1507 | static unsigned int safe_highmem_pages; | |
1508 | ||
1509 | static struct memory_bitmap *safe_highmem_bm; | |
1510 | ||
1511 | static int | |
1512 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
1513 | { | |
1514 | unsigned int to_alloc; | |
1515 | ||
1516 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
1517 | return -ENOMEM; | |
1518 | ||
1519 | if (get_highmem_buffer(PG_SAFE)) | |
1520 | return -ENOMEM; | |
1521 | ||
1522 | to_alloc = count_free_highmem_pages(); | |
1523 | if (to_alloc > *nr_highmem_p) | |
1524 | to_alloc = *nr_highmem_p; | |
1525 | else | |
1526 | *nr_highmem_p = to_alloc; | |
1527 | ||
1528 | safe_highmem_pages = 0; | |
1529 | while (to_alloc-- > 0) { | |
1530 | struct page *page; | |
1531 | ||
1532 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 1533 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
1534 | /* The page is "safe", set its bit the bitmap */ |
1535 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1536 | safe_highmem_pages++; | |
1537 | } | |
1538 | /* Mark the page as allocated */ | |
7be98234 RW |
1539 | swsusp_set_page_forbidden(page); |
1540 | swsusp_set_page_free(page); | |
8357376d RW |
1541 | } |
1542 | memory_bm_position_reset(bm); | |
1543 | safe_highmem_bm = bm; | |
1544 | return 0; | |
1545 | } | |
1546 | ||
1547 | /** | |
1548 | * get_highmem_page_buffer - for given highmem image page find the buffer | |
1549 | * that suspend_write_next() should set for its caller to write to. | |
1550 | * | |
1551 | * If the page is to be saved to its "original" page frame or a copy of | |
1552 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
1553 | * the copy of the page is to be made in normal memory, so the address of | |
1554 | * the copy is returned. | |
1555 | * | |
1556 | * If @buffer is returned, the caller of suspend_write_next() will write | |
1557 | * the page's contents to @buffer, so they will have to be copied to the | |
1558 | * right location on the next call to suspend_write_next() and it is done | |
1559 | * with the help of copy_last_highmem_page(). For this purpose, if | |
1560 | * @buffer is returned, @last_highmem page is set to the page to which | |
1561 | * the data will have to be copied from @buffer. | |
1562 | */ | |
1563 | ||
1564 | static struct page *last_highmem_page; | |
1565 | ||
1566 | static void * | |
1567 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
1568 | { | |
1569 | struct highmem_pbe *pbe; | |
1570 | void *kaddr; | |
1571 | ||
7be98234 | 1572 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
8357376d RW |
1573 | /* We have allocated the "original" page frame and we can |
1574 | * use it directly to store the loaded page. | |
1575 | */ | |
1576 | last_highmem_page = page; | |
1577 | return buffer; | |
1578 | } | |
1579 | /* The "original" page frame has not been allocated and we have to | |
1580 | * use a "safe" page frame to store the loaded page. | |
1581 | */ | |
1582 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
1583 | if (!pbe) { | |
1584 | swsusp_free(); | |
1585 | return NULL; | |
1586 | } | |
1587 | pbe->orig_page = page; | |
1588 | if (safe_highmem_pages > 0) { | |
1589 | struct page *tmp; | |
1590 | ||
1591 | /* Copy of the page will be stored in high memory */ | |
1592 | kaddr = buffer; | |
1593 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
1594 | safe_highmem_pages--; | |
1595 | last_highmem_page = tmp; | |
1596 | pbe->copy_page = tmp; | |
1597 | } else { | |
1598 | /* Copy of the page will be stored in normal memory */ | |
1599 | kaddr = safe_pages_list; | |
1600 | safe_pages_list = safe_pages_list->next; | |
1601 | pbe->copy_page = virt_to_page(kaddr); | |
1602 | } | |
1603 | pbe->next = highmem_pblist; | |
1604 | highmem_pblist = pbe; | |
1605 | return kaddr; | |
1606 | } | |
1607 | ||
1608 | /** | |
1609 | * copy_last_highmem_page - copy the contents of a highmem image from | |
1610 | * @buffer, where the caller of snapshot_write_next() has place them, | |
1611 | * to the right location represented by @last_highmem_page . | |
1612 | */ | |
1613 | ||
1614 | static void copy_last_highmem_page(void) | |
1615 | { | |
1616 | if (last_highmem_page) { | |
1617 | void *dst; | |
1618 | ||
1619 | dst = kmap_atomic(last_highmem_page, KM_USER0); | |
1620 | memcpy(dst, buffer, PAGE_SIZE); | |
1621 | kunmap_atomic(dst, KM_USER0); | |
1622 | last_highmem_page = NULL; | |
1623 | } | |
1624 | } | |
1625 | ||
1626 | static inline int last_highmem_page_copied(void) | |
1627 | { | |
1628 | return !last_highmem_page; | |
1629 | } | |
1630 | ||
1631 | static inline void free_highmem_data(void) | |
1632 | { | |
1633 | if (safe_highmem_bm) | |
1634 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
1635 | ||
1636 | if (buffer) | |
1637 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
1638 | } | |
1639 | #else | |
1640 | static inline int get_safe_write_buffer(void) { return 0; } | |
1641 | ||
1642 | static unsigned int | |
1643 | count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } | |
1644 | ||
1645 | static inline int | |
1646 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
1647 | { | |
1648 | return 0; | |
1649 | } | |
1650 | ||
1651 | static inline void * | |
1652 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
1653 | { | |
1654 | return NULL; | |
1655 | } | |
1656 | ||
1657 | static inline void copy_last_highmem_page(void) {} | |
1658 | static inline int last_highmem_page_copied(void) { return 1; } | |
1659 | static inline void free_highmem_data(void) {} | |
1660 | #endif /* CONFIG_HIGHMEM */ | |
1661 | ||
f577eb30 | 1662 | /** |
940864dd RW |
1663 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
1664 | * be overwritten in the process of restoring the system memory state | |
1665 | * from the suspend image ("unsafe" pages) and allocate memory for the | |
1666 | * image. | |
968808b8 | 1667 | * |
940864dd RW |
1668 | * The idea is to allocate a new memory bitmap first and then allocate |
1669 | * as many pages as needed for the image data, but not to assign these | |
1670 | * pages to specific tasks initially. Instead, we just mark them as | |
8357376d RW |
1671 | * allocated and create a lists of "safe" pages that will be used |
1672 | * later. On systems with high memory a list of "safe" highmem pages is | |
1673 | * also created. | |
f577eb30 RW |
1674 | */ |
1675 | ||
940864dd RW |
1676 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
1677 | ||
940864dd RW |
1678 | static int |
1679 | prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) | |
f577eb30 | 1680 | { |
8357376d | 1681 | unsigned int nr_pages, nr_highmem; |
940864dd RW |
1682 | struct linked_page *sp_list, *lp; |
1683 | int error; | |
f577eb30 | 1684 | |
8357376d RW |
1685 | /* If there is no highmem, the buffer will not be necessary */ |
1686 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
1687 | buffer = NULL; | |
1688 | ||
1689 | nr_highmem = count_highmem_image_pages(bm); | |
940864dd RW |
1690 | error = mark_unsafe_pages(bm); |
1691 | if (error) | |
1692 | goto Free; | |
1693 | ||
1694 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
1695 | if (error) | |
1696 | goto Free; | |
1697 | ||
1698 | duplicate_memory_bitmap(new_bm, bm); | |
1699 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
1700 | if (nr_highmem > 0) { |
1701 | error = prepare_highmem_image(bm, &nr_highmem); | |
1702 | if (error) | |
1703 | goto Free; | |
1704 | } | |
940864dd RW |
1705 | /* Reserve some safe pages for potential later use. |
1706 | * | |
1707 | * NOTE: This way we make sure there will be enough safe pages for the | |
1708 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
1709 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
1710 | */ | |
1711 | sp_list = NULL; | |
1712 | /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */ | |
8357376d | 1713 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
1714 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
1715 | while (nr_pages > 0) { | |
8357376d | 1716 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 1717 | if (!lp) { |
f577eb30 | 1718 | error = -ENOMEM; |
940864dd RW |
1719 | goto Free; |
1720 | } | |
1721 | lp->next = sp_list; | |
1722 | sp_list = lp; | |
1723 | nr_pages--; | |
f577eb30 | 1724 | } |
940864dd RW |
1725 | /* Preallocate memory for the image */ |
1726 | safe_pages_list = NULL; | |
8357376d | 1727 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
1728 | while (nr_pages > 0) { |
1729 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
1730 | if (!lp) { | |
1731 | error = -ENOMEM; | |
1732 | goto Free; | |
1733 | } | |
7be98234 | 1734 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
1735 | /* The page is "safe", add it to the list */ |
1736 | lp->next = safe_pages_list; | |
1737 | safe_pages_list = lp; | |
968808b8 | 1738 | } |
940864dd | 1739 | /* Mark the page as allocated */ |
7be98234 RW |
1740 | swsusp_set_page_forbidden(virt_to_page(lp)); |
1741 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 1742 | nr_pages--; |
968808b8 | 1743 | } |
940864dd RW |
1744 | /* Free the reserved safe pages so that chain_alloc() can use them */ |
1745 | while (sp_list) { | |
1746 | lp = sp_list->next; | |
1747 | free_image_page(sp_list, PG_UNSAFE_CLEAR); | |
1748 | sp_list = lp; | |
f577eb30 | 1749 | } |
940864dd RW |
1750 | return 0; |
1751 | ||
59a49335 | 1752 | Free: |
940864dd | 1753 | swsusp_free(); |
f577eb30 RW |
1754 | return error; |
1755 | } | |
1756 | ||
940864dd RW |
1757 | /** |
1758 | * get_buffer - compute the address that snapshot_write_next() should | |
1759 | * set for its caller to write to. | |
1760 | */ | |
1761 | ||
1762 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) | |
968808b8 | 1763 | { |
940864dd RW |
1764 | struct pbe *pbe; |
1765 | struct page *page = pfn_to_page(memory_bm_next_pfn(bm)); | |
968808b8 | 1766 | |
8357376d RW |
1767 | if (PageHighMem(page)) |
1768 | return get_highmem_page_buffer(page, ca); | |
1769 | ||
7be98234 | 1770 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
940864dd RW |
1771 | /* We have allocated the "original" page frame and we can |
1772 | * use it directly to store the loaded page. | |
968808b8 | 1773 | */ |
940864dd RW |
1774 | return page_address(page); |
1775 | ||
1776 | /* The "original" page frame has not been allocated and we have to | |
1777 | * use a "safe" page frame to store the loaded page. | |
968808b8 | 1778 | */ |
940864dd RW |
1779 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
1780 | if (!pbe) { | |
1781 | swsusp_free(); | |
1782 | return NULL; | |
1783 | } | |
8357376d RW |
1784 | pbe->orig_address = page_address(page); |
1785 | pbe->address = safe_pages_list; | |
940864dd RW |
1786 | safe_pages_list = safe_pages_list->next; |
1787 | pbe->next = restore_pblist; | |
1788 | restore_pblist = pbe; | |
8357376d | 1789 | return pbe->address; |
968808b8 RW |
1790 | } |
1791 | ||
f577eb30 RW |
1792 | /** |
1793 | * snapshot_write_next - used for writing the system memory snapshot. | |
1794 | * | |
1795 | * On the first call to it @handle should point to a zeroed | |
1796 | * snapshot_handle structure. The structure gets updated and a pointer | |
1797 | * to it should be passed to this function every next time. | |
1798 | * | |
1799 | * The @count parameter should contain the number of bytes the caller | |
1800 | * wants to write to the image. It must not be zero. | |
1801 | * | |
1802 | * On success the function returns a positive number. Then, the caller | |
1803 | * is allowed to write up to the returned number of bytes to the memory | |
1804 | * location computed by the data_of() macro. The number returned | |
1805 | * may be smaller than @count, but this only happens if the write would | |
1806 | * cross a page boundary otherwise. | |
1807 | * | |
1808 | * The function returns 0 to indicate the "end of file" condition, | |
1809 | * and a negative number is returned on error. In such cases the | |
1810 | * structure pointed to by @handle is not updated and should not be used | |
1811 | * any more. | |
1812 | */ | |
1813 | ||
1814 | int snapshot_write_next(struct snapshot_handle *handle, size_t count) | |
1815 | { | |
940864dd | 1816 | static struct chain_allocator ca; |
f577eb30 RW |
1817 | int error = 0; |
1818 | ||
940864dd | 1819 | /* Check if we have already loaded the entire image */ |
fb13a28b | 1820 | if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1821 | return 0; |
940864dd | 1822 | |
8357376d RW |
1823 | if (handle->offset == 0) { |
1824 | if (!buffer) | |
1825 | /* This makes the buffer be freed by swsusp_free() */ | |
1826 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
1827 | ||
f577eb30 RW |
1828 | if (!buffer) |
1829 | return -ENOMEM; | |
8357376d | 1830 | |
f577eb30 | 1831 | handle->buffer = buffer; |
8357376d | 1832 | } |
546e0d27 | 1833 | handle->sync_read = 1; |
fb13a28b | 1834 | if (handle->prev < handle->cur) { |
940864dd RW |
1835 | if (handle->prev == 0) { |
1836 | error = load_header(buffer); | |
1837 | if (error) | |
1838 | return error; | |
1839 | ||
1840 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); | |
f577eb30 RW |
1841 | if (error) |
1842 | return error; | |
940864dd | 1843 | |
f577eb30 | 1844 | } else if (handle->prev <= nr_meta_pages) { |
940864dd RW |
1845 | unpack_orig_pfns(buffer, ©_bm); |
1846 | if (handle->prev == nr_meta_pages) { | |
1847 | error = prepare_image(&orig_bm, ©_bm); | |
f577eb30 RW |
1848 | if (error) |
1849 | return error; | |
940864dd RW |
1850 | |
1851 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); | |
1852 | memory_bm_position_reset(&orig_bm); | |
1853 | restore_pblist = NULL; | |
1854 | handle->buffer = get_buffer(&orig_bm, &ca); | |
546e0d27 | 1855 | handle->sync_read = 0; |
940864dd RW |
1856 | if (!handle->buffer) |
1857 | return -ENOMEM; | |
f577eb30 RW |
1858 | } |
1859 | } else { | |
8357376d | 1860 | copy_last_highmem_page(); |
940864dd | 1861 | handle->buffer = get_buffer(&orig_bm, &ca); |
8357376d RW |
1862 | if (handle->buffer != buffer) |
1863 | handle->sync_read = 0; | |
f577eb30 | 1864 | } |
fb13a28b | 1865 | handle->prev = handle->cur; |
f577eb30 | 1866 | } |
fb13a28b RW |
1867 | handle->buf_offset = handle->cur_offset; |
1868 | if (handle->cur_offset + count >= PAGE_SIZE) { | |
1869 | count = PAGE_SIZE - handle->cur_offset; | |
1870 | handle->cur_offset = 0; | |
1871 | handle->cur++; | |
f577eb30 | 1872 | } else { |
fb13a28b | 1873 | handle->cur_offset += count; |
f577eb30 RW |
1874 | } |
1875 | handle->offset += count; | |
1876 | return count; | |
1877 | } | |
1878 | ||
8357376d RW |
1879 | /** |
1880 | * snapshot_write_finalize - must be called after the last call to | |
1881 | * snapshot_write_next() in case the last page in the image happens | |
1882 | * to be a highmem page and its contents should be stored in the | |
1883 | * highmem. Additionally, it releases the memory that will not be | |
1884 | * used any more. | |
1885 | */ | |
1886 | ||
1887 | void snapshot_write_finalize(struct snapshot_handle *handle) | |
1888 | { | |
1889 | copy_last_highmem_page(); | |
1890 | /* Free only if we have loaded the image entirely */ | |
1891 | if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages) { | |
1892 | memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR); | |
1893 | free_highmem_data(); | |
1894 | } | |
1895 | } | |
1896 | ||
f577eb30 RW |
1897 | int snapshot_image_loaded(struct snapshot_handle *handle) |
1898 | { | |
8357376d | 1899 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
1900 | handle->cur <= nr_meta_pages + nr_copy_pages); |
1901 | } | |
1902 | ||
8357376d RW |
1903 | #ifdef CONFIG_HIGHMEM |
1904 | /* Assumes that @buf is ready and points to a "safe" page */ | |
1905 | static inline void | |
1906 | swap_two_pages_data(struct page *p1, struct page *p2, void *buf) | |
940864dd | 1907 | { |
8357376d RW |
1908 | void *kaddr1, *kaddr2; |
1909 | ||
1910 | kaddr1 = kmap_atomic(p1, KM_USER0); | |
1911 | kaddr2 = kmap_atomic(p2, KM_USER1); | |
1912 | memcpy(buf, kaddr1, PAGE_SIZE); | |
1913 | memcpy(kaddr1, kaddr2, PAGE_SIZE); | |
1914 | memcpy(kaddr2, buf, PAGE_SIZE); | |
1915 | kunmap_atomic(kaddr1, KM_USER0); | |
1916 | kunmap_atomic(kaddr2, KM_USER1); | |
1917 | } | |
1918 | ||
1919 | /** | |
1920 | * restore_highmem - for each highmem page that was allocated before | |
1921 | * the suspend and included in the suspend image, and also has been | |
1922 | * allocated by the "resume" kernel swap its current (ie. "before | |
1923 | * resume") contents with the previous (ie. "before suspend") one. | |
1924 | * | |
1925 | * If the resume eventually fails, we can call this function once | |
1926 | * again and restore the "before resume" highmem state. | |
1927 | */ | |
1928 | ||
1929 | int restore_highmem(void) | |
1930 | { | |
1931 | struct highmem_pbe *pbe = highmem_pblist; | |
1932 | void *buf; | |
1933 | ||
1934 | if (!pbe) | |
1935 | return 0; | |
1936 | ||
1937 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
1938 | if (!buf) | |
1939 | return -ENOMEM; | |
1940 | ||
1941 | while (pbe) { | |
1942 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
1943 | pbe = pbe->next; | |
1944 | } | |
1945 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
1946 | return 0; | |
f577eb30 | 1947 | } |
8357376d | 1948 | #endif /* CONFIG_HIGHMEM */ |