Commit | Line | Data |
---|---|---|
25761b6e | 1 | /* |
96bc7aec | 2 | * linux/kernel/power/snapshot.c |
25761b6e | 3 | * |
96bc7aec | 4 | * This file provide system snapshot/restore functionality. |
25761b6e RW |
5 | * |
6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@suse.cz> | |
7 | * | |
8 | * This file is released under the GPLv2, and is based on swsusp.c. | |
9 | * | |
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> | |
25761b6e RW |
24 | #include <linux/bootmem.h> |
25 | #include <linux/syscalls.h> | |
26 | #include <linux/console.h> | |
27 | #include <linux/highmem.h> | |
25761b6e RW |
28 | |
29 | #include <asm/uaccess.h> | |
30 | #include <asm/mmu_context.h> | |
31 | #include <asm/pgtable.h> | |
32 | #include <asm/tlbflush.h> | |
33 | #include <asm/io.h> | |
34 | ||
25761b6e RW |
35 | #include "power.h" |
36 | ||
75534b50 RW |
37 | /* List of PBEs used for creating and restoring the suspend image */ |
38 | struct pbe *restore_pblist; | |
39 | ||
f577eb30 RW |
40 | static unsigned int nr_copy_pages; |
41 | static unsigned int nr_meta_pages; | |
6e1819d6 | 42 | static unsigned long *buffer; |
7088a5c0 | 43 | |
25761b6e | 44 | #ifdef CONFIG_HIGHMEM |
3448097f | 45 | unsigned int count_highmem_pages(void) |
72a97e08 RW |
46 | { |
47 | struct zone *zone; | |
48 | unsigned long zone_pfn; | |
49 | unsigned int n = 0; | |
50 | ||
51 | for_each_zone (zone) | |
52 | if (is_highmem(zone)) { | |
53 | mark_free_pages(zone); | |
54 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; zone_pfn++) { | |
55 | struct page *page; | |
56 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
57 | if (!pfn_valid(pfn)) | |
58 | continue; | |
59 | page = pfn_to_page(pfn); | |
60 | if (PageReserved(page)) | |
61 | continue; | |
62 | if (PageNosaveFree(page)) | |
63 | continue; | |
64 | n++; | |
65 | } | |
66 | } | |
67 | return n; | |
68 | } | |
69 | ||
25761b6e RW |
70 | struct highmem_page { |
71 | char *data; | |
72 | struct page *page; | |
73 | struct highmem_page *next; | |
74 | }; | |
75 | ||
76 | static struct highmem_page *highmem_copy; | |
77 | ||
78 | static int save_highmem_zone(struct zone *zone) | |
79 | { | |
80 | unsigned long zone_pfn; | |
81 | mark_free_pages(zone); | |
82 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
83 | struct page *page; | |
84 | struct highmem_page *save; | |
85 | void *kaddr; | |
86 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
87 | ||
ce6ed29f | 88 | if (!(pfn%10000)) |
25761b6e RW |
89 | printk("."); |
90 | if (!pfn_valid(pfn)) | |
91 | continue; | |
92 | page = pfn_to_page(pfn); | |
93 | /* | |
94 | * This condition results from rvmalloc() sans vmalloc_32() | |
95 | * and architectural memory reservations. This should be | |
96 | * corrected eventually when the cases giving rise to this | |
97 | * are better understood. | |
98 | */ | |
c8adb494 | 99 | if (PageReserved(page)) |
25761b6e | 100 | continue; |
25761b6e RW |
101 | BUG_ON(PageNosave(page)); |
102 | if (PageNosaveFree(page)) | |
103 | continue; | |
104 | save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC); | |
105 | if (!save) | |
106 | return -ENOMEM; | |
107 | save->next = highmem_copy; | |
108 | save->page = page; | |
109 | save->data = (void *) get_zeroed_page(GFP_ATOMIC); | |
110 | if (!save->data) { | |
111 | kfree(save); | |
112 | return -ENOMEM; | |
113 | } | |
114 | kaddr = kmap_atomic(page, KM_USER0); | |
115 | memcpy(save->data, kaddr, PAGE_SIZE); | |
116 | kunmap_atomic(kaddr, KM_USER0); | |
117 | highmem_copy = save; | |
118 | } | |
119 | return 0; | |
120 | } | |
25761b6e | 121 | |
3448097f | 122 | int save_highmem(void) |
25761b6e | 123 | { |
25761b6e RW |
124 | struct zone *zone; |
125 | int res = 0; | |
126 | ||
ce6ed29f | 127 | pr_debug("swsusp: Saving Highmem"); |
e4e4d665 | 128 | drain_local_pages(); |
25761b6e RW |
129 | for_each_zone (zone) { |
130 | if (is_highmem(zone)) | |
131 | res = save_highmem_zone(zone); | |
132 | if (res) | |
133 | return res; | |
134 | } | |
ce6ed29f | 135 | printk("\n"); |
25761b6e RW |
136 | return 0; |
137 | } | |
138 | ||
3448097f | 139 | int restore_highmem(void) |
25761b6e | 140 | { |
25761b6e RW |
141 | printk("swsusp: Restoring Highmem\n"); |
142 | while (highmem_copy) { | |
143 | struct highmem_page *save = highmem_copy; | |
144 | void *kaddr; | |
145 | highmem_copy = save->next; | |
146 | ||
147 | kaddr = kmap_atomic(save->page, KM_USER0); | |
148 | memcpy(kaddr, save->data, PAGE_SIZE); | |
149 | kunmap_atomic(kaddr, KM_USER0); | |
150 | free_page((long) save->data); | |
151 | kfree(save); | |
152 | } | |
25761b6e RW |
153 | return 0; |
154 | } | |
ce4ab001 | 155 | #else |
7bff24e2 AB |
156 | static inline unsigned int count_highmem_pages(void) {return 0;} |
157 | static inline int save_highmem(void) {return 0;} | |
158 | static inline int restore_highmem(void) {return 0;} | |
0fbeb5a4 | 159 | #endif |
25761b6e | 160 | |
f6143aa6 RW |
161 | /** |
162 | * @safe_needed - on resume, for storing the PBE list and the image, | |
163 | * we can only use memory pages that do not conflict with the pages | |
164 | * used before suspend. | |
165 | * | |
166 | * The unsafe pages are marked with the PG_nosave_free flag | |
167 | * and we count them using unsafe_pages | |
168 | */ | |
169 | ||
0bcd888d RW |
170 | #define PG_ANY 0 |
171 | #define PG_SAFE 1 | |
172 | #define PG_UNSAFE_CLEAR 1 | |
173 | #define PG_UNSAFE_KEEP 0 | |
174 | ||
f6143aa6 RW |
175 | static unsigned int unsafe_pages; |
176 | ||
177 | static void *alloc_image_page(gfp_t gfp_mask, int safe_needed) | |
178 | { | |
179 | void *res; | |
180 | ||
181 | res = (void *)get_zeroed_page(gfp_mask); | |
182 | if (safe_needed) | |
183 | while (res && PageNosaveFree(virt_to_page(res))) { | |
184 | /* The page is unsafe, mark it for swsusp_free() */ | |
185 | SetPageNosave(virt_to_page(res)); | |
186 | unsafe_pages++; | |
187 | res = (void *)get_zeroed_page(gfp_mask); | |
188 | } | |
189 | if (res) { | |
190 | SetPageNosave(virt_to_page(res)); | |
191 | SetPageNosaveFree(virt_to_page(res)); | |
192 | } | |
193 | return res; | |
194 | } | |
195 | ||
196 | unsigned long get_safe_page(gfp_t gfp_mask) | |
197 | { | |
0bcd888d | 198 | return (unsigned long)alloc_image_page(gfp_mask, PG_SAFE); |
f6143aa6 RW |
199 | } |
200 | ||
201 | /** | |
202 | * free_image_page - free page represented by @addr, allocated with | |
203 | * alloc_image_page (page flags set by it must be cleared) | |
204 | */ | |
205 | ||
206 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
207 | { | |
208 | ClearPageNosave(virt_to_page(addr)); | |
209 | if (clear_nosave_free) | |
210 | ClearPageNosaveFree(virt_to_page(addr)); | |
211 | free_page((unsigned long)addr); | |
212 | } | |
213 | ||
b788db79 RW |
214 | /* struct linked_page is used to build chains of pages */ |
215 | ||
216 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
217 | ||
218 | struct linked_page { | |
219 | struct linked_page *next; | |
220 | char data[LINKED_PAGE_DATA_SIZE]; | |
221 | } __attribute__((packed)); | |
222 | ||
223 | static inline void | |
224 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) | |
225 | { | |
226 | while (list) { | |
227 | struct linked_page *lp = list->next; | |
228 | ||
229 | free_image_page(list, clear_page_nosave); | |
230 | list = lp; | |
231 | } | |
232 | } | |
233 | ||
234 | /** | |
235 | * struct chain_allocator is used for allocating small objects out of | |
236 | * a linked list of pages called 'the chain'. | |
237 | * | |
238 | * The chain grows each time when there is no room for a new object in | |
239 | * the current page. The allocated objects cannot be freed individually. | |
240 | * It is only possible to free them all at once, by freeing the entire | |
241 | * chain. | |
242 | * | |
243 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
244 | * are not much smaller than PAGE_SIZE. | |
245 | */ | |
246 | ||
247 | struct chain_allocator { | |
248 | struct linked_page *chain; /* the chain */ | |
249 | unsigned int used_space; /* total size of objects allocated out | |
250 | * of the current page | |
251 | */ | |
252 | gfp_t gfp_mask; /* mask for allocating pages */ | |
253 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
254 | }; | |
255 | ||
256 | static void | |
257 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) | |
258 | { | |
259 | ca->chain = NULL; | |
260 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
261 | ca->gfp_mask = gfp_mask; | |
262 | ca->safe_needed = safe_needed; | |
263 | } | |
264 | ||
265 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
266 | { | |
267 | void *ret; | |
268 | ||
269 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
270 | struct linked_page *lp; | |
271 | ||
272 | lp = alloc_image_page(ca->gfp_mask, ca->safe_needed); | |
273 | if (!lp) | |
274 | return NULL; | |
275 | ||
276 | lp->next = ca->chain; | |
277 | ca->chain = lp; | |
278 | ca->used_space = 0; | |
279 | } | |
280 | ret = ca->chain->data + ca->used_space; | |
281 | ca->used_space += size; | |
282 | return ret; | |
283 | } | |
284 | ||
285 | static void chain_free(struct chain_allocator *ca, int clear_page_nosave) | |
286 | { | |
287 | free_list_of_pages(ca->chain, clear_page_nosave); | |
288 | memset(ca, 0, sizeof(struct chain_allocator)); | |
289 | } | |
290 | ||
291 | /** | |
292 | * Data types related to memory bitmaps. | |
293 | * | |
294 | * Memory bitmap is a structure consiting of many linked lists of | |
295 | * objects. The main list's elements are of type struct zone_bitmap | |
296 | * and each of them corresonds to one zone. For each zone bitmap | |
297 | * object there is a list of objects of type struct bm_block that | |
298 | * represent each blocks of bit chunks in which information is | |
299 | * stored. | |
300 | * | |
301 | * struct memory_bitmap contains a pointer to the main list of zone | |
302 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
303 | * and a pointer to the list of pages used for allocating all of the | |
304 | * zone bitmap objects and bitmap block objects. | |
305 | * | |
306 | * NOTE: It has to be possible to lay out the bitmap in memory | |
307 | * using only allocations of order 0. Additionally, the bitmap is | |
308 | * designed to work with arbitrary number of zones (this is over the | |
309 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
310 | * | |
311 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
312 | * objects and a pointer to the bitmap block object that has been | |
313 | * most recently used for setting bits. Additionally, it contains the | |
314 | * pfns that correspond to the start and end of the represented zone. | |
315 | * | |
316 | * struct bm_block contains a pointer to the memory page in which | |
317 | * information is stored (in the form of a block of bit chunks | |
318 | * of type unsigned long each). It also contains the pfns that | |
319 | * correspond to the start and end of the represented memory area and | |
320 | * the number of bit chunks in the block. | |
321 | * | |
322 | * NOTE: Memory bitmaps are used for two types of operations only: | |
323 | * "set a bit" and "find the next bit set". Moreover, the searching | |
324 | * is always carried out after all of the "set a bit" operations | |
325 | * on given bitmap. | |
326 | */ | |
327 | ||
328 | #define BM_END_OF_MAP (~0UL) | |
329 | ||
330 | #define BM_CHUNKS_PER_BLOCK (PAGE_SIZE / sizeof(long)) | |
331 | #define BM_BITS_PER_CHUNK (sizeof(long) << 3) | |
332 | #define BM_BITS_PER_BLOCK (PAGE_SIZE << 3) | |
333 | ||
334 | struct bm_block { | |
335 | struct bm_block *next; /* next element of the list */ | |
336 | unsigned long start_pfn; /* pfn represented by the first bit */ | |
337 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ | |
338 | unsigned int size; /* number of bit chunks */ | |
339 | unsigned long *data; /* chunks of bits representing pages */ | |
340 | }; | |
341 | ||
342 | struct zone_bitmap { | |
343 | struct zone_bitmap *next; /* next element of the list */ | |
344 | unsigned long start_pfn; /* minimal pfn in this zone */ | |
345 | unsigned long end_pfn; /* maximal pfn in this zone plus 1 */ | |
346 | struct bm_block *bm_blocks; /* list of bitmap blocks */ | |
347 | struct bm_block *cur_block; /* recently used bitmap block */ | |
348 | }; | |
349 | ||
350 | /* strcut bm_position is used for browsing memory bitmaps */ | |
351 | ||
352 | struct bm_position { | |
353 | struct zone_bitmap *zone_bm; | |
354 | struct bm_block *block; | |
355 | int chunk; | |
356 | int bit; | |
357 | }; | |
358 | ||
359 | struct memory_bitmap { | |
360 | struct zone_bitmap *zone_bm_list; /* list of zone bitmaps */ | |
361 | struct linked_page *p_list; /* list of pages used to store zone | |
362 | * bitmap objects and bitmap block | |
363 | * objects | |
364 | */ | |
365 | struct bm_position cur; /* most recently used bit position */ | |
366 | }; | |
367 | ||
368 | /* Functions that operate on memory bitmaps */ | |
369 | ||
370 | static inline void memory_bm_reset_chunk(struct memory_bitmap *bm) | |
371 | { | |
372 | bm->cur.chunk = 0; | |
373 | bm->cur.bit = -1; | |
374 | } | |
375 | ||
376 | static void memory_bm_position_reset(struct memory_bitmap *bm) | |
377 | { | |
378 | struct zone_bitmap *zone_bm; | |
379 | ||
380 | zone_bm = bm->zone_bm_list; | |
381 | bm->cur.zone_bm = zone_bm; | |
382 | bm->cur.block = zone_bm->bm_blocks; | |
383 | memory_bm_reset_chunk(bm); | |
384 | } | |
385 | ||
386 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
387 | ||
388 | /** | |
389 | * create_bm_block_list - create a list of block bitmap objects | |
390 | */ | |
391 | ||
392 | static inline struct bm_block * | |
393 | create_bm_block_list(unsigned int nr_blocks, struct chain_allocator *ca) | |
394 | { | |
395 | struct bm_block *bblist = NULL; | |
396 | ||
397 | while (nr_blocks-- > 0) { | |
398 | struct bm_block *bb; | |
399 | ||
400 | bb = chain_alloc(ca, sizeof(struct bm_block)); | |
401 | if (!bb) | |
402 | return NULL; | |
403 | ||
404 | bb->next = bblist; | |
405 | bblist = bb; | |
406 | } | |
407 | return bblist; | |
408 | } | |
409 | ||
410 | /** | |
411 | * create_zone_bm_list - create a list of zone bitmap objects | |
412 | */ | |
413 | ||
414 | static inline struct zone_bitmap * | |
415 | create_zone_bm_list(unsigned int nr_zones, struct chain_allocator *ca) | |
416 | { | |
417 | struct zone_bitmap *zbmlist = NULL; | |
418 | ||
419 | while (nr_zones-- > 0) { | |
420 | struct zone_bitmap *zbm; | |
421 | ||
422 | zbm = chain_alloc(ca, sizeof(struct zone_bitmap)); | |
423 | if (!zbm) | |
424 | return NULL; | |
425 | ||
426 | zbm->next = zbmlist; | |
427 | zbmlist = zbm; | |
428 | } | |
429 | return zbmlist; | |
430 | } | |
431 | ||
432 | /** | |
433 | * memory_bm_create - allocate memory for a memory bitmap | |
434 | */ | |
435 | ||
436 | static int | |
437 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) | |
438 | { | |
439 | struct chain_allocator ca; | |
440 | struct zone *zone; | |
441 | struct zone_bitmap *zone_bm; | |
442 | struct bm_block *bb; | |
443 | unsigned int nr; | |
444 | ||
445 | chain_init(&ca, gfp_mask, safe_needed); | |
446 | ||
447 | /* Compute the number of zones */ | |
448 | nr = 0; | |
449 | for_each_zone (zone) | |
450 | if (populated_zone(zone) && !is_highmem(zone)) | |
451 | nr++; | |
452 | ||
453 | /* Allocate the list of zones bitmap objects */ | |
454 | zone_bm = create_zone_bm_list(nr, &ca); | |
455 | bm->zone_bm_list = zone_bm; | |
456 | if (!zone_bm) { | |
457 | chain_free(&ca, PG_UNSAFE_CLEAR); | |
458 | return -ENOMEM; | |
459 | } | |
460 | ||
461 | /* Initialize the zone bitmap objects */ | |
462 | for_each_zone (zone) { | |
463 | unsigned long pfn; | |
464 | ||
465 | if (!populated_zone(zone) || is_highmem(zone)) | |
466 | continue; | |
467 | ||
468 | zone_bm->start_pfn = zone->zone_start_pfn; | |
469 | zone_bm->end_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
470 | /* Allocate the list of bitmap block objects */ | |
471 | nr = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
472 | bb = create_bm_block_list(nr, &ca); | |
473 | zone_bm->bm_blocks = bb; | |
474 | zone_bm->cur_block = bb; | |
475 | if (!bb) | |
476 | goto Free; | |
477 | ||
478 | nr = zone->spanned_pages; | |
479 | pfn = zone->zone_start_pfn; | |
480 | /* Initialize the bitmap block objects */ | |
481 | while (bb) { | |
482 | unsigned long *ptr; | |
483 | ||
484 | ptr = alloc_image_page(gfp_mask, safe_needed); | |
485 | bb->data = ptr; | |
486 | if (!ptr) | |
487 | goto Free; | |
488 | ||
489 | bb->start_pfn = pfn; | |
490 | if (nr >= BM_BITS_PER_BLOCK) { | |
491 | pfn += BM_BITS_PER_BLOCK; | |
492 | bb->size = BM_CHUNKS_PER_BLOCK; | |
493 | nr -= BM_BITS_PER_BLOCK; | |
494 | } else { | |
495 | /* This is executed only once in the loop */ | |
496 | pfn += nr; | |
497 | bb->size = DIV_ROUND_UP(nr, BM_BITS_PER_CHUNK); | |
498 | } | |
499 | bb->end_pfn = pfn; | |
500 | bb = bb->next; | |
501 | } | |
502 | zone_bm = zone_bm->next; | |
503 | } | |
504 | bm->p_list = ca.chain; | |
505 | memory_bm_position_reset(bm); | |
506 | return 0; | |
507 | ||
508 | Free: | |
509 | bm->p_list = ca.chain; | |
510 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
511 | return -ENOMEM; | |
512 | } | |
513 | ||
514 | /** | |
515 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
516 | */ | |
517 | ||
518 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) | |
519 | { | |
520 | struct zone_bitmap *zone_bm; | |
521 | ||
522 | /* Free the list of bit blocks for each zone_bitmap object */ | |
523 | zone_bm = bm->zone_bm_list; | |
524 | while (zone_bm) { | |
525 | struct bm_block *bb; | |
526 | ||
527 | bb = zone_bm->bm_blocks; | |
528 | while (bb) { | |
529 | if (bb->data) | |
530 | free_image_page(bb->data, clear_nosave_free); | |
531 | bb = bb->next; | |
532 | } | |
533 | zone_bm = zone_bm->next; | |
534 | } | |
535 | free_list_of_pages(bm->p_list, clear_nosave_free); | |
536 | bm->zone_bm_list = NULL; | |
537 | } | |
538 | ||
539 | /** | |
540 | * memory_bm_set_bit - set the bit in the bitmap @bm that corresponds | |
541 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member | |
542 | * of @bm->cur_zone_bm are updated. | |
543 | * | |
544 | * If the bit cannot be set, the function returns -EINVAL . | |
545 | */ | |
546 | ||
547 | static int | |
548 | memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) | |
549 | { | |
550 | struct zone_bitmap *zone_bm; | |
551 | struct bm_block *bb; | |
552 | ||
553 | /* Check if the pfn is from the current zone */ | |
554 | zone_bm = bm->cur.zone_bm; | |
555 | if (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) { | |
556 | zone_bm = bm->zone_bm_list; | |
557 | /* We don't assume that the zones are sorted by pfns */ | |
558 | while (pfn < zone_bm->start_pfn || pfn >= zone_bm->end_pfn) { | |
559 | zone_bm = zone_bm->next; | |
560 | if (unlikely(!zone_bm)) | |
561 | return -EINVAL; | |
562 | } | |
563 | bm->cur.zone_bm = zone_bm; | |
564 | } | |
565 | /* Check if the pfn corresponds to the current bitmap block */ | |
566 | bb = zone_bm->cur_block; | |
567 | if (pfn < bb->start_pfn) | |
568 | bb = zone_bm->bm_blocks; | |
569 | ||
570 | while (pfn >= bb->end_pfn) { | |
571 | bb = bb->next; | |
572 | if (unlikely(!bb)) | |
573 | return -EINVAL; | |
574 | } | |
575 | zone_bm->cur_block = bb; | |
576 | pfn -= bb->start_pfn; | |
577 | set_bit(pfn % BM_BITS_PER_CHUNK, bb->data + pfn / BM_BITS_PER_CHUNK); | |
578 | return 0; | |
579 | } | |
580 | ||
581 | /* Two auxiliary functions for memory_bm_next_pfn */ | |
582 | ||
583 | /* Find the first set bit in the given chunk, if there is one */ | |
584 | ||
585 | static inline int next_bit_in_chunk(int bit, unsigned long *chunk_p) | |
586 | { | |
587 | bit++; | |
588 | while (bit < BM_BITS_PER_CHUNK) { | |
589 | if (test_bit(bit, chunk_p)) | |
590 | return bit; | |
591 | ||
592 | bit++; | |
593 | } | |
594 | return -1; | |
595 | } | |
596 | ||
597 | /* Find a chunk containing some bits set in given block of bits */ | |
598 | ||
599 | static inline int next_chunk_in_block(int n, struct bm_block *bb) | |
600 | { | |
601 | n++; | |
602 | while (n < bb->size) { | |
603 | if (bb->data[n]) | |
604 | return n; | |
605 | ||
606 | n++; | |
607 | } | |
608 | return -1; | |
609 | } | |
610 | ||
611 | /** | |
612 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit | |
613 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is | |
614 | * returned. | |
615 | * | |
616 | * It is required to run memory_bm_position_reset() before the first call to | |
617 | * this function. | |
618 | */ | |
619 | ||
620 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) | |
621 | { | |
622 | struct zone_bitmap *zone_bm; | |
623 | struct bm_block *bb; | |
624 | int chunk; | |
625 | int bit; | |
626 | ||
627 | do { | |
628 | bb = bm->cur.block; | |
629 | do { | |
630 | chunk = bm->cur.chunk; | |
631 | bit = bm->cur.bit; | |
632 | do { | |
633 | bit = next_bit_in_chunk(bit, bb->data + chunk); | |
634 | if (bit >= 0) | |
635 | goto Return_pfn; | |
636 | ||
637 | chunk = next_chunk_in_block(chunk, bb); | |
638 | bit = -1; | |
639 | } while (chunk >= 0); | |
640 | bb = bb->next; | |
641 | bm->cur.block = bb; | |
642 | memory_bm_reset_chunk(bm); | |
643 | } while (bb); | |
644 | zone_bm = bm->cur.zone_bm->next; | |
645 | if (zone_bm) { | |
646 | bm->cur.zone_bm = zone_bm; | |
647 | bm->cur.block = zone_bm->bm_blocks; | |
648 | memory_bm_reset_chunk(bm); | |
649 | } | |
650 | } while (zone_bm); | |
651 | memory_bm_position_reset(bm); | |
652 | return BM_END_OF_MAP; | |
653 | ||
654 | Return_pfn: | |
655 | bm->cur.chunk = chunk; | |
656 | bm->cur.bit = bit; | |
657 | return bb->start_pfn + chunk * BM_BITS_PER_CHUNK + bit; | |
658 | } | |
659 | ||
660 | /** | |
661 | * snapshot_additional_pages - estimate the number of additional pages | |
662 | * be needed for setting up the suspend image data structures for given | |
663 | * zone (usually the returned value is greater than the exact number) | |
664 | */ | |
665 | ||
666 | unsigned int snapshot_additional_pages(struct zone *zone) | |
667 | { | |
668 | unsigned int res; | |
669 | ||
670 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
671 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), PAGE_SIZE); | |
672 | return res; | |
673 | } | |
674 | ||
f6143aa6 RW |
675 | /** |
676 | * pfn_is_nosave - check if given pfn is in the 'nosave' section | |
677 | */ | |
678 | ||
ae83c5ee | 679 | static inline int pfn_is_nosave(unsigned long pfn) |
25761b6e RW |
680 | { |
681 | unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT; | |
682 | unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT; | |
683 | return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn); | |
684 | } | |
685 | ||
686 | /** | |
687 | * saveable - Determine whether a page should be cloned or not. | |
688 | * @pfn: The page | |
689 | * | |
ae83c5ee RW |
690 | * We save a page if it isn't Nosave, and is not in the range of pages |
691 | * statically defined as 'unsaveable', and it | |
692 | * isn't a part of a free chunk of pages. | |
25761b6e RW |
693 | */ |
694 | ||
ae83c5ee | 695 | static struct page *saveable_page(unsigned long pfn) |
25761b6e | 696 | { |
de491861 | 697 | struct page *page; |
25761b6e RW |
698 | |
699 | if (!pfn_valid(pfn)) | |
ae83c5ee | 700 | return NULL; |
25761b6e RW |
701 | |
702 | page = pfn_to_page(pfn); | |
ae83c5ee | 703 | |
25761b6e | 704 | if (PageNosave(page)) |
ae83c5ee | 705 | return NULL; |
72a97e08 | 706 | if (PageReserved(page) && pfn_is_nosave(pfn)) |
ae83c5ee | 707 | return NULL; |
25761b6e | 708 | if (PageNosaveFree(page)) |
ae83c5ee | 709 | return NULL; |
25761b6e | 710 | |
ae83c5ee | 711 | return page; |
25761b6e RW |
712 | } |
713 | ||
72a97e08 | 714 | unsigned int count_data_pages(void) |
25761b6e RW |
715 | { |
716 | struct zone *zone; | |
ae83c5ee | 717 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 718 | unsigned int n = 0; |
25761b6e | 719 | |
25761b6e RW |
720 | for_each_zone (zone) { |
721 | if (is_highmem(zone)) | |
722 | continue; | |
723 | mark_free_pages(zone); | |
ae83c5ee RW |
724 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
725 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
726 | n += !!saveable_page(pfn); | |
25761b6e | 727 | } |
a0f49651 | 728 | return n; |
25761b6e RW |
729 | } |
730 | ||
f623f0db RW |
731 | static inline void copy_data_page(long *dst, long *src) |
732 | { | |
733 | int n; | |
734 | ||
735 | /* copy_page and memcpy are not usable for copying task structs. */ | |
736 | for (n = PAGE_SIZE / sizeof(long); n; n--) | |
737 | *dst++ = *src++; | |
738 | } | |
739 | ||
b788db79 RW |
740 | static void |
741 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) | |
25761b6e RW |
742 | { |
743 | struct zone *zone; | |
b788db79 | 744 | unsigned long pfn; |
25761b6e | 745 | |
25761b6e | 746 | for_each_zone (zone) { |
b788db79 RW |
747 | unsigned long max_zone_pfn; |
748 | ||
25761b6e RW |
749 | if (is_highmem(zone)) |
750 | continue; | |
b788db79 | 751 | |
25761b6e | 752 | mark_free_pages(zone); |
ae83c5ee | 753 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
b788db79 RW |
754 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
755 | if (saveable_page(pfn)) | |
756 | memory_bm_set_bit(orig_bm, pfn); | |
25761b6e | 757 | } |
b788db79 RW |
758 | memory_bm_position_reset(orig_bm); |
759 | memory_bm_position_reset(copy_bm); | |
760 | do { | |
761 | pfn = memory_bm_next_pfn(orig_bm); | |
762 | if (likely(pfn != BM_END_OF_MAP)) { | |
763 | struct page *page; | |
764 | void *src; | |
765 | ||
766 | page = pfn_to_page(pfn); | |
767 | src = page_address(page); | |
768 | page = pfn_to_page(memory_bm_next_pfn(copy_bm)); | |
769 | copy_data_page(page_address(page), src); | |
770 | } | |
771 | } while (pfn != BM_END_OF_MAP); | |
25761b6e RW |
772 | } |
773 | ||
25761b6e RW |
774 | /** |
775 | * free_pagedir - free pages allocated with alloc_pagedir() | |
776 | */ | |
777 | ||
4a3b98a4 | 778 | static void free_pagedir(struct pbe *pblist, int clear_nosave_free) |
25761b6e RW |
779 | { |
780 | struct pbe *pbe; | |
781 | ||
782 | while (pblist) { | |
783 | pbe = (pblist + PB_PAGE_SKIP)->next; | |
f6143aa6 | 784 | free_image_page(pblist, clear_nosave_free); |
25761b6e RW |
785 | pblist = pbe; |
786 | } | |
787 | } | |
788 | ||
789 | /** | |
790 | * fill_pb_page - Create a list of PBEs on a given memory page | |
791 | */ | |
792 | ||
cd560bb2 | 793 | static inline void fill_pb_page(struct pbe *pbpage, unsigned int n) |
25761b6e RW |
794 | { |
795 | struct pbe *p; | |
796 | ||
797 | p = pbpage; | |
cd560bb2 | 798 | pbpage += n - 1; |
25761b6e RW |
799 | do |
800 | p->next = p + 1; | |
801 | while (++p < pbpage); | |
802 | } | |
803 | ||
804 | /** | |
805 | * create_pbe_list - Create a list of PBEs on top of a given chain | |
806 | * of memory pages allocated with alloc_pagedir() | |
cd560bb2 RW |
807 | * |
808 | * This function assumes that pages allocated by alloc_image_page() will | |
809 | * always be zeroed. | |
25761b6e RW |
810 | */ |
811 | ||
7088a5c0 | 812 | static inline void create_pbe_list(struct pbe *pblist, unsigned int nr_pages) |
25761b6e | 813 | { |
cd560bb2 | 814 | struct pbe *pbpage; |
dc19d507 | 815 | unsigned int num = PBES_PER_PAGE; |
25761b6e RW |
816 | |
817 | for_each_pb_page (pbpage, pblist) { | |
818 | if (num >= nr_pages) | |
819 | break; | |
820 | ||
cd560bb2 | 821 | fill_pb_page(pbpage, PBES_PER_PAGE); |
25761b6e RW |
822 | num += PBES_PER_PAGE; |
823 | } | |
824 | if (pbpage) { | |
cd560bb2 RW |
825 | num -= PBES_PER_PAGE; |
826 | fill_pb_page(pbpage, nr_pages - num); | |
25761b6e | 827 | } |
25761b6e RW |
828 | } |
829 | ||
25761b6e RW |
830 | /** |
831 | * alloc_pagedir - Allocate the page directory. | |
832 | * | |
833 | * First, determine exactly how many pages we need and | |
834 | * allocate them. | |
835 | * | |
836 | * We arrange the pages in a chain: each page is an array of PBES_PER_PAGE | |
837 | * struct pbe elements (pbes) and the last element in the page points | |
838 | * to the next page. | |
839 | * | |
840 | * On each page we set up a list of struct_pbe elements. | |
841 | */ | |
842 | ||
7bff24e2 AB |
843 | static struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, |
844 | int safe_needed) | |
25761b6e | 845 | { |
dc19d507 | 846 | unsigned int num; |
25761b6e RW |
847 | struct pbe *pblist, *pbe; |
848 | ||
849 | if (!nr_pages) | |
850 | return NULL; | |
851 | ||
054bd4c1 | 852 | pblist = alloc_image_page(gfp_mask, safe_needed); |
cd560bb2 RW |
853 | pbe = pblist; |
854 | for (num = PBES_PER_PAGE; num < nr_pages; num += PBES_PER_PAGE) { | |
855 | if (!pbe) { | |
0bcd888d | 856 | free_pagedir(pblist, PG_UNSAFE_CLEAR); |
cd560bb2 RW |
857 | return NULL; |
858 | } | |
25761b6e | 859 | pbe += PB_PAGE_SKIP; |
054bd4c1 | 860 | pbe->next = alloc_image_page(gfp_mask, safe_needed); |
cd560bb2 | 861 | pbe = pbe->next; |
25761b6e | 862 | } |
cd560bb2 | 863 | create_pbe_list(pblist, nr_pages); |
25761b6e RW |
864 | return pblist; |
865 | } | |
866 | ||
867 | /** | |
868 | * Free pages we allocated for suspend. Suspend pages are alocated | |
869 | * before atomic copy, so we need to free them after resume. | |
870 | */ | |
871 | ||
872 | void swsusp_free(void) | |
873 | { | |
874 | struct zone *zone; | |
ae83c5ee | 875 | unsigned long pfn, max_zone_pfn; |
25761b6e RW |
876 | |
877 | for_each_zone(zone) { | |
ae83c5ee RW |
878 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
879 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
880 | if (pfn_valid(pfn)) { | |
881 | struct page *page = pfn_to_page(pfn); | |
882 | ||
25761b6e RW |
883 | if (PageNosave(page) && PageNosaveFree(page)) { |
884 | ClearPageNosave(page); | |
885 | ClearPageNosaveFree(page); | |
886 | free_page((long) page_address(page)); | |
887 | } | |
888 | } | |
889 | } | |
f577eb30 RW |
890 | nr_copy_pages = 0; |
891 | nr_meta_pages = 0; | |
75534b50 | 892 | restore_pblist = NULL; |
6e1819d6 | 893 | buffer = NULL; |
25761b6e RW |
894 | } |
895 | ||
896 | ||
897 | /** | |
898 | * enough_free_mem - Make sure we enough free memory to snapshot. | |
899 | * | |
900 | * Returns TRUE or FALSE after checking the number of available | |
901 | * free pages. | |
902 | */ | |
903 | ||
dc19d507 | 904 | static int enough_free_mem(unsigned int nr_pages) |
25761b6e | 905 | { |
e5e2fa78 RW |
906 | struct zone *zone; |
907 | unsigned int n = 0; | |
908 | ||
909 | for_each_zone (zone) | |
910 | if (!is_highmem(zone)) | |
911 | n += zone->free_pages; | |
912 | pr_debug("swsusp: available memory: %u pages\n", n); | |
913 | return n > (nr_pages + PAGES_FOR_IO + | |
a0f49651 | 914 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE); |
25761b6e RW |
915 | } |
916 | ||
b788db79 RW |
917 | static int |
918 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, | |
919 | unsigned int nr_pages) | |
054bd4c1 | 920 | { |
b788db79 | 921 | int error; |
054bd4c1 | 922 | |
b788db79 RW |
923 | error = memory_bm_create(orig_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY); |
924 | if (error) | |
925 | goto Free; | |
25761b6e | 926 | |
b788db79 RW |
927 | error = memory_bm_create(copy_bm, GFP_ATOMIC | __GFP_COLD, PG_ANY); |
928 | if (error) | |
929 | goto Free; | |
25761b6e | 930 | |
b788db79 RW |
931 | while (nr_pages-- > 0) { |
932 | struct page *page = alloc_page(GFP_ATOMIC | __GFP_COLD); | |
933 | if (!page) | |
934 | goto Free; | |
25761b6e | 935 | |
b788db79 RW |
936 | SetPageNosave(page); |
937 | SetPageNosaveFree(page); | |
938 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
25761b6e | 939 | } |
b788db79 | 940 | return 0; |
25761b6e | 941 | |
b788db79 RW |
942 | Free: |
943 | swsusp_free(); | |
944 | return -ENOMEM; | |
25761b6e RW |
945 | } |
946 | ||
b788db79 RW |
947 | /* Memory bitmap used for marking saveable pages */ |
948 | static struct memory_bitmap orig_bm; | |
949 | /* Memory bitmap used for marking allocated pages that will contain the copies | |
950 | * of saveable pages | |
951 | */ | |
952 | static struct memory_bitmap copy_bm; | |
953 | ||
2e32a43e | 954 | asmlinkage int swsusp_save(void) |
25761b6e | 955 | { |
dc19d507 | 956 | unsigned int nr_pages; |
25761b6e RW |
957 | |
958 | pr_debug("swsusp: critical section: \n"); | |
25761b6e RW |
959 | |
960 | drain_local_pages(); | |
a0f49651 RW |
961 | nr_pages = count_data_pages(); |
962 | printk("swsusp: Need to copy %u pages\n", nr_pages); | |
25761b6e RW |
963 | |
964 | pr_debug("swsusp: pages needed: %u + %lu + %u, free: %u\n", | |
a0f49651 RW |
965 | nr_pages, |
966 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE, | |
25761b6e RW |
967 | PAGES_FOR_IO, nr_free_pages()); |
968 | ||
a0f49651 | 969 | if (!enough_free_mem(nr_pages)) { |
25761b6e RW |
970 | printk(KERN_ERR "swsusp: Not enough free memory\n"); |
971 | return -ENOMEM; | |
972 | } | |
973 | ||
b788db79 | 974 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages)) |
a0f49651 | 975 | return -ENOMEM; |
25761b6e RW |
976 | |
977 | /* During allocating of suspend pagedir, new cold pages may appear. | |
978 | * Kill them. | |
979 | */ | |
980 | drain_local_pages(); | |
b788db79 | 981 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
982 | |
983 | /* | |
984 | * End of critical section. From now on, we can write to memory, | |
985 | * but we should not touch disk. This specially means we must _not_ | |
986 | * touch swap space! Except we must write out our image of course. | |
987 | */ | |
988 | ||
a0f49651 | 989 | nr_copy_pages = nr_pages; |
f577eb30 | 990 | nr_meta_pages = (nr_pages * sizeof(long) + PAGE_SIZE - 1) >> PAGE_SHIFT; |
a0f49651 RW |
991 | |
992 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); | |
25761b6e RW |
993 | return 0; |
994 | } | |
f577eb30 RW |
995 | |
996 | static void init_header(struct swsusp_info *info) | |
997 | { | |
998 | memset(info, 0, sizeof(struct swsusp_info)); | |
999 | info->version_code = LINUX_VERSION_CODE; | |
1000 | info->num_physpages = num_physpages; | |
1001 | memcpy(&info->uts, &system_utsname, sizeof(system_utsname)); | |
1002 | info->cpus = num_online_cpus(); | |
1003 | info->image_pages = nr_copy_pages; | |
1004 | info->pages = nr_copy_pages + nr_meta_pages + 1; | |
6e1819d6 RW |
1005 | info->size = info->pages; |
1006 | info->size <<= PAGE_SHIFT; | |
f577eb30 RW |
1007 | } |
1008 | ||
1009 | /** | |
b788db79 RW |
1010 | * pack_addresses - the addresses corresponding to pfns found in the |
1011 | * bitmap @bm are stored in the array @buf[] (1 page) | |
f577eb30 RW |
1012 | */ |
1013 | ||
b788db79 RW |
1014 | static inline void |
1015 | pack_addresses(unsigned long *buf, struct memory_bitmap *bm) | |
f577eb30 RW |
1016 | { |
1017 | int j; | |
1018 | ||
b788db79 RW |
1019 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1020 | unsigned long pfn = memory_bm_next_pfn(bm); | |
1021 | ||
1022 | if (unlikely(pfn == BM_END_OF_MAP)) | |
1023 | break; | |
1024 | ||
1025 | buf[j] = (unsigned long)page_address(pfn_to_page(pfn)); | |
f577eb30 | 1026 | } |
f577eb30 RW |
1027 | } |
1028 | ||
1029 | /** | |
1030 | * snapshot_read_next - used for reading the system memory snapshot. | |
1031 | * | |
1032 | * On the first call to it @handle should point to a zeroed | |
1033 | * snapshot_handle structure. The structure gets updated and a pointer | |
1034 | * to it should be passed to this function every next time. | |
1035 | * | |
1036 | * The @count parameter should contain the number of bytes the caller | |
1037 | * wants to read from the snapshot. It must not be zero. | |
1038 | * | |
1039 | * On success the function returns a positive number. Then, the caller | |
1040 | * is allowed to read up to the returned number of bytes from the memory | |
1041 | * location computed by the data_of() macro. The number returned | |
1042 | * may be smaller than @count, but this only happens if the read would | |
1043 | * cross a page boundary otherwise. | |
1044 | * | |
1045 | * The function returns 0 to indicate the end of data stream condition, | |
1046 | * and a negative number is returned on error. In such cases the | |
1047 | * structure pointed to by @handle is not updated and should not be used | |
1048 | * any more. | |
1049 | */ | |
1050 | ||
1051 | int snapshot_read_next(struct snapshot_handle *handle, size_t count) | |
1052 | { | |
fb13a28b | 1053 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1054 | return 0; |
b788db79 | 1055 | |
f577eb30 RW |
1056 | if (!buffer) { |
1057 | /* This makes the buffer be freed by swsusp_free() */ | |
0bcd888d | 1058 | buffer = alloc_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1059 | if (!buffer) |
1060 | return -ENOMEM; | |
1061 | } | |
1062 | if (!handle->offset) { | |
1063 | init_header((struct swsusp_info *)buffer); | |
1064 | handle->buffer = buffer; | |
b788db79 RW |
1065 | memory_bm_position_reset(&orig_bm); |
1066 | memory_bm_position_reset(©_bm); | |
f577eb30 | 1067 | } |
fb13a28b RW |
1068 | if (handle->prev < handle->cur) { |
1069 | if (handle->cur <= nr_meta_pages) { | |
b788db79 RW |
1070 | memset(buffer, 0, PAGE_SIZE); |
1071 | pack_addresses(buffer, &orig_bm); | |
f577eb30 | 1072 | } else { |
b788db79 RW |
1073 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1074 | ||
1075 | handle->buffer = page_address(pfn_to_page(pfn)); | |
f577eb30 | 1076 | } |
fb13a28b | 1077 | handle->prev = handle->cur; |
f577eb30 | 1078 | } |
fb13a28b RW |
1079 | handle->buf_offset = handle->cur_offset; |
1080 | if (handle->cur_offset + count >= PAGE_SIZE) { | |
1081 | count = PAGE_SIZE - handle->cur_offset; | |
1082 | handle->cur_offset = 0; | |
1083 | handle->cur++; | |
f577eb30 | 1084 | } else { |
fb13a28b | 1085 | handle->cur_offset += count; |
f577eb30 RW |
1086 | } |
1087 | handle->offset += count; | |
1088 | return count; | |
1089 | } | |
1090 | ||
1091 | /** | |
1092 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
1093 | * the image during resume, because they conflict with the pages that | |
1094 | * had been used before suspend | |
1095 | */ | |
1096 | ||
1097 | static int mark_unsafe_pages(struct pbe *pblist) | |
1098 | { | |
1099 | struct zone *zone; | |
ae83c5ee | 1100 | unsigned long pfn, max_zone_pfn; |
f577eb30 RW |
1101 | struct pbe *p; |
1102 | ||
1103 | if (!pblist) /* a sanity check */ | |
1104 | return -EINVAL; | |
1105 | ||
1106 | /* Clear page flags */ | |
1107 | for_each_zone (zone) { | |
ae83c5ee RW |
1108 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1109 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) | |
1110 | if (pfn_valid(pfn)) | |
1111 | ClearPageNosaveFree(pfn_to_page(pfn)); | |
f577eb30 RW |
1112 | } |
1113 | ||
1114 | /* Mark orig addresses */ | |
1115 | for_each_pbe (p, pblist) { | |
1116 | if (virt_addr_valid(p->orig_address)) | |
1117 | SetPageNosaveFree(virt_to_page(p->orig_address)); | |
1118 | else | |
1119 | return -EFAULT; | |
1120 | } | |
1121 | ||
968808b8 RW |
1122 | unsafe_pages = 0; |
1123 | ||
f577eb30 RW |
1124 | return 0; |
1125 | } | |
1126 | ||
1127 | static void copy_page_backup_list(struct pbe *dst, struct pbe *src) | |
1128 | { | |
1129 | /* We assume both lists contain the same number of elements */ | |
1130 | while (src) { | |
1131 | dst->orig_address = src->orig_address; | |
1132 | dst = dst->next; | |
1133 | src = src->next; | |
1134 | } | |
1135 | } | |
1136 | ||
1137 | static int check_header(struct swsusp_info *info) | |
1138 | { | |
1139 | char *reason = NULL; | |
1140 | ||
1141 | if (info->version_code != LINUX_VERSION_CODE) | |
1142 | reason = "kernel version"; | |
1143 | if (info->num_physpages != num_physpages) | |
1144 | reason = "memory size"; | |
1145 | if (strcmp(info->uts.sysname,system_utsname.sysname)) | |
1146 | reason = "system type"; | |
1147 | if (strcmp(info->uts.release,system_utsname.release)) | |
1148 | reason = "kernel release"; | |
1149 | if (strcmp(info->uts.version,system_utsname.version)) | |
1150 | reason = "version"; | |
1151 | if (strcmp(info->uts.machine,system_utsname.machine)) | |
1152 | reason = "machine"; | |
1153 | if (reason) { | |
1154 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | |
1155 | return -EPERM; | |
1156 | } | |
1157 | return 0; | |
1158 | } | |
1159 | ||
1160 | /** | |
1161 | * load header - check the image header and copy data from it | |
1162 | */ | |
1163 | ||
1164 | static int load_header(struct snapshot_handle *handle, | |
1165 | struct swsusp_info *info) | |
1166 | { | |
1167 | int error; | |
1168 | struct pbe *pblist; | |
1169 | ||
1170 | error = check_header(info); | |
1171 | if (!error) { | |
0bcd888d | 1172 | pblist = alloc_pagedir(info->image_pages, GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1173 | if (!pblist) |
1174 | return -ENOMEM; | |
75534b50 | 1175 | restore_pblist = pblist; |
f577eb30 RW |
1176 | handle->pbe = pblist; |
1177 | nr_copy_pages = info->image_pages; | |
1178 | nr_meta_pages = info->pages - info->image_pages - 1; | |
1179 | } | |
1180 | return error; | |
1181 | } | |
1182 | ||
1183 | /** | |
1184 | * unpack_orig_addresses - copy the elements of @buf[] (1 page) to | |
1185 | * the PBEs in the list starting at @pbe | |
1186 | */ | |
1187 | ||
1188 | static inline struct pbe *unpack_orig_addresses(unsigned long *buf, | |
1189 | struct pbe *pbe) | |
1190 | { | |
1191 | int j; | |
1192 | ||
1193 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | |
1194 | pbe->orig_address = buf[j]; | |
1195 | pbe = pbe->next; | |
1196 | } | |
1197 | return pbe; | |
1198 | } | |
1199 | ||
1200 | /** | |
968808b8 | 1201 | * prepare_image - use metadata contained in the PBE list |
75534b50 | 1202 | * pointed to by restore_pblist to mark the pages that will |
f577eb30 | 1203 | * be overwritten in the process of restoring the system |
968808b8 RW |
1204 | * memory state from the image ("unsafe" pages) and allocate |
1205 | * memory for the image | |
1206 | * | |
1207 | * The idea is to allocate the PBE list first and then | |
1208 | * allocate as many pages as it's needed for the image data, | |
1209 | * but not to assign these pages to the PBEs initially. | |
1210 | * Instead, we just mark them as allocated and create a list | |
1211 | * of "safe" which will be used later | |
f577eb30 RW |
1212 | */ |
1213 | ||
b788db79 | 1214 | static struct linked_page *safe_pages; |
968808b8 RW |
1215 | |
1216 | static int prepare_image(struct snapshot_handle *handle) | |
f577eb30 RW |
1217 | { |
1218 | int error = 0; | |
968808b8 RW |
1219 | unsigned int nr_pages = nr_copy_pages; |
1220 | struct pbe *p, *pblist = NULL; | |
f577eb30 | 1221 | |
75534b50 | 1222 | p = restore_pblist; |
f577eb30 RW |
1223 | error = mark_unsafe_pages(p); |
1224 | if (!error) { | |
0bcd888d | 1225 | pblist = alloc_pagedir(nr_pages, GFP_ATOMIC, PG_SAFE); |
f577eb30 RW |
1226 | if (pblist) |
1227 | copy_page_backup_list(pblist, p); | |
0bcd888d | 1228 | free_pagedir(p, PG_UNSAFE_KEEP); |
f577eb30 RW |
1229 | if (!pblist) |
1230 | error = -ENOMEM; | |
1231 | } | |
968808b8 RW |
1232 | safe_pages = NULL; |
1233 | if (!error && nr_pages > unsafe_pages) { | |
1234 | nr_pages -= unsafe_pages; | |
1235 | while (nr_pages--) { | |
b788db79 | 1236 | struct linked_page *ptr; |
968808b8 | 1237 | |
b788db79 | 1238 | ptr = (void *)get_zeroed_page(GFP_ATOMIC); |
968808b8 RW |
1239 | if (!ptr) { |
1240 | error = -ENOMEM; | |
1241 | break; | |
1242 | } | |
1243 | if (!PageNosaveFree(virt_to_page(ptr))) { | |
1244 | /* The page is "safe", add it to the list */ | |
1245 | ptr->next = safe_pages; | |
1246 | safe_pages = ptr; | |
1247 | } | |
1248 | /* Mark the page as allocated */ | |
1249 | SetPageNosave(virt_to_page(ptr)); | |
1250 | SetPageNosaveFree(virt_to_page(ptr)); | |
1251 | } | |
1252 | } | |
f577eb30 | 1253 | if (!error) { |
75534b50 | 1254 | restore_pblist = pblist; |
f577eb30 | 1255 | } else { |
f577eb30 | 1256 | handle->pbe = NULL; |
968808b8 | 1257 | swsusp_free(); |
f577eb30 RW |
1258 | } |
1259 | return error; | |
1260 | } | |
1261 | ||
968808b8 RW |
1262 | static void *get_buffer(struct snapshot_handle *handle) |
1263 | { | |
1264 | struct pbe *pbe = handle->pbe, *last = handle->last_pbe; | |
1265 | struct page *page = virt_to_page(pbe->orig_address); | |
1266 | ||
1267 | if (PageNosave(page) && PageNosaveFree(page)) { | |
1268 | /* | |
1269 | * We have allocated the "original" page frame and we can | |
1270 | * use it directly to store the read page | |
1271 | */ | |
1272 | pbe->address = 0; | |
1273 | if (last && last->next) | |
1274 | last->next = NULL; | |
1275 | return (void *)pbe->orig_address; | |
1276 | } | |
1277 | /* | |
1278 | * The "original" page frame has not been allocated and we have to | |
1279 | * use a "safe" page frame to store the read page | |
1280 | */ | |
1281 | pbe->address = (unsigned long)safe_pages; | |
1282 | safe_pages = safe_pages->next; | |
1283 | if (last) | |
1284 | last->next = pbe; | |
1285 | handle->last_pbe = pbe; | |
1286 | return (void *)pbe->address; | |
1287 | } | |
1288 | ||
f577eb30 RW |
1289 | /** |
1290 | * snapshot_write_next - used for writing the system memory snapshot. | |
1291 | * | |
1292 | * On the first call to it @handle should point to a zeroed | |
1293 | * snapshot_handle structure. The structure gets updated and a pointer | |
1294 | * to it should be passed to this function every next time. | |
1295 | * | |
1296 | * The @count parameter should contain the number of bytes the caller | |
1297 | * wants to write to the image. It must not be zero. | |
1298 | * | |
1299 | * On success the function returns a positive number. Then, the caller | |
1300 | * is allowed to write up to the returned number of bytes to the memory | |
1301 | * location computed by the data_of() macro. The number returned | |
1302 | * may be smaller than @count, but this only happens if the write would | |
1303 | * cross a page boundary otherwise. | |
1304 | * | |
1305 | * The function returns 0 to indicate the "end of file" condition, | |
1306 | * and a negative number is returned on error. In such cases the | |
1307 | * structure pointed to by @handle is not updated and should not be used | |
1308 | * any more. | |
1309 | */ | |
1310 | ||
1311 | int snapshot_write_next(struct snapshot_handle *handle, size_t count) | |
1312 | { | |
f577eb30 RW |
1313 | int error = 0; |
1314 | ||
fb13a28b | 1315 | if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 RW |
1316 | return 0; |
1317 | if (!buffer) { | |
1318 | /* This makes the buffer be freed by swsusp_free() */ | |
0bcd888d | 1319 | buffer = alloc_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1320 | if (!buffer) |
1321 | return -ENOMEM; | |
1322 | } | |
1323 | if (!handle->offset) | |
1324 | handle->buffer = buffer; | |
546e0d27 | 1325 | handle->sync_read = 1; |
fb13a28b | 1326 | if (handle->prev < handle->cur) { |
f577eb30 | 1327 | if (!handle->prev) { |
546e0d27 AM |
1328 | error = load_header(handle, |
1329 | (struct swsusp_info *)buffer); | |
f577eb30 RW |
1330 | if (error) |
1331 | return error; | |
1332 | } else if (handle->prev <= nr_meta_pages) { | |
546e0d27 AM |
1333 | handle->pbe = unpack_orig_addresses(buffer, |
1334 | handle->pbe); | |
f577eb30 | 1335 | if (!handle->pbe) { |
968808b8 | 1336 | error = prepare_image(handle); |
f577eb30 RW |
1337 | if (error) |
1338 | return error; | |
75534b50 | 1339 | handle->pbe = restore_pblist; |
968808b8 RW |
1340 | handle->last_pbe = NULL; |
1341 | handle->buffer = get_buffer(handle); | |
546e0d27 | 1342 | handle->sync_read = 0; |
f577eb30 RW |
1343 | } |
1344 | } else { | |
1345 | handle->pbe = handle->pbe->next; | |
968808b8 | 1346 | handle->buffer = get_buffer(handle); |
546e0d27 | 1347 | handle->sync_read = 0; |
f577eb30 | 1348 | } |
fb13a28b | 1349 | handle->prev = handle->cur; |
f577eb30 | 1350 | } |
fb13a28b RW |
1351 | handle->buf_offset = handle->cur_offset; |
1352 | if (handle->cur_offset + count >= PAGE_SIZE) { | |
1353 | count = PAGE_SIZE - handle->cur_offset; | |
1354 | handle->cur_offset = 0; | |
1355 | handle->cur++; | |
f577eb30 | 1356 | } else { |
fb13a28b | 1357 | handle->cur_offset += count; |
f577eb30 RW |
1358 | } |
1359 | handle->offset += count; | |
1360 | return count; | |
1361 | } | |
1362 | ||
1363 | int snapshot_image_loaded(struct snapshot_handle *handle) | |
1364 | { | |
1365 | return !(!handle->pbe || handle->pbe->next || !nr_copy_pages || | |
fb13a28b | 1366 | handle->cur <= nr_meta_pages + nr_copy_pages); |
f577eb30 | 1367 | } |