Root/
1 | /* |
2 | * linux/kernel/power/snapshot.c |
3 | * |
4 | * This file provides system snapshot/restore functionality for swsusp. |
5 | * |
6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz> |
7 | * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl> |
8 | * |
9 | * This file is released under the GPLv2. |
10 | * |
11 | */ |
12 | |
13 | #include <linux/version.h> |
14 | #include <linux/module.h> |
15 | #include <linux/mm.h> |
16 | #include <linux/suspend.h> |
17 | #include <linux/delay.h> |
18 | #include <linux/bitops.h> |
19 | #include <linux/spinlock.h> |
20 | #include <linux/kernel.h> |
21 | #include <linux/pm.h> |
22 | #include <linux/device.h> |
23 | #include <linux/init.h> |
24 | #include <linux/bootmem.h> |
25 | #include <linux/syscalls.h> |
26 | #include <linux/console.h> |
27 | #include <linux/highmem.h> |
28 | #include <linux/list.h> |
29 | #include <linux/slab.h> |
30 | |
31 | #include <asm/uaccess.h> |
32 | #include <asm/mmu_context.h> |
33 | #include <asm/pgtable.h> |
34 | #include <asm/tlbflush.h> |
35 | #include <asm/io.h> |
36 | |
37 | #include "power.h" |
38 | |
39 | static int swsusp_page_is_free(struct page *); |
40 | static void swsusp_set_page_forbidden(struct page *); |
41 | static void swsusp_unset_page_forbidden(struct page *); |
42 | |
43 | /* |
44 | * Preferred image size in bytes (tunable via /sys/power/image_size). |
45 | * When it is set to N, swsusp will do its best to ensure the image |
46 | * size will not exceed N bytes, but if that is impossible, it will |
47 | * try to create the smallest image possible. |
48 | */ |
49 | unsigned long image_size; |
50 | |
51 | void __init hibernate_image_size_init(void) |
52 | { |
53 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
54 | } |
55 | |
56 | /* List of PBEs needed for restoring the pages that were allocated before |
57 | * the suspend and included in the suspend image, but have also been |
58 | * allocated by the "resume" kernel, so their contents cannot be written |
59 | * directly to their "original" page frames. |
60 | */ |
61 | struct pbe *restore_pblist; |
62 | |
63 | /* Pointer to an auxiliary buffer (1 page) */ |
64 | static void *buffer; |
65 | |
66 | /** |
67 | * @safe_needed - on resume, for storing the PBE list and the image, |
68 | * we can only use memory pages that do not conflict with the pages |
69 | * used before suspend. The unsafe pages have PageNosaveFree set |
70 | * and we count them using unsafe_pages. |
71 | * |
72 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
73 | * so that swsusp_free() can release it. |
74 | */ |
75 | |
76 | #define PG_ANY 0 |
77 | #define PG_SAFE 1 |
78 | #define PG_UNSAFE_CLEAR 1 |
79 | #define PG_UNSAFE_KEEP 0 |
80 | |
81 | static unsigned int allocated_unsafe_pages; |
82 | |
83 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
84 | { |
85 | void *res; |
86 | |
87 | res = (void *)get_zeroed_page(gfp_mask); |
88 | if (safe_needed) |
89 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
90 | /* The page is unsafe, mark it for swsusp_free() */ |
91 | swsusp_set_page_forbidden(virt_to_page(res)); |
92 | allocated_unsafe_pages++; |
93 | res = (void *)get_zeroed_page(gfp_mask); |
94 | } |
95 | if (res) { |
96 | swsusp_set_page_forbidden(virt_to_page(res)); |
97 | swsusp_set_page_free(virt_to_page(res)); |
98 | } |
99 | return res; |
100 | } |
101 | |
102 | unsigned long get_safe_page(gfp_t gfp_mask) |
103 | { |
104 | return (unsigned long)get_image_page(gfp_mask, PG_SAFE); |
105 | } |
106 | |
107 | static struct page *alloc_image_page(gfp_t gfp_mask) |
108 | { |
109 | struct page *page; |
110 | |
111 | page = alloc_page(gfp_mask); |
112 | if (page) { |
113 | swsusp_set_page_forbidden(page); |
114 | swsusp_set_page_free(page); |
115 | } |
116 | return page; |
117 | } |
118 | |
119 | /** |
120 | * free_image_page - free page represented by @addr, allocated with |
121 | * get_image_page (page flags set by it must be cleared) |
122 | */ |
123 | |
124 | static inline void free_image_page(void *addr, int clear_nosave_free) |
125 | { |
126 | struct page *page; |
127 | |
128 | BUG_ON(!virt_addr_valid(addr)); |
129 | |
130 | page = virt_to_page(addr); |
131 | |
132 | swsusp_unset_page_forbidden(page); |
133 | if (clear_nosave_free) |
134 | swsusp_unset_page_free(page); |
135 | |
136 | __free_page(page); |
137 | } |
138 | |
139 | /* struct linked_page is used to build chains of pages */ |
140 | |
141 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) |
142 | |
143 | struct linked_page { |
144 | struct linked_page *next; |
145 | char data[LINKED_PAGE_DATA_SIZE]; |
146 | } __attribute__((packed)); |
147 | |
148 | static inline void |
149 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) |
150 | { |
151 | while (list) { |
152 | struct linked_page *lp = list->next; |
153 | |
154 | free_image_page(list, clear_page_nosave); |
155 | list = lp; |
156 | } |
157 | } |
158 | |
159 | /** |
160 | * struct chain_allocator is used for allocating small objects out of |
161 | * a linked list of pages called 'the chain'. |
162 | * |
163 | * The chain grows each time when there is no room for a new object in |
164 | * the current page. The allocated objects cannot be freed individually. |
165 | * It is only possible to free them all at once, by freeing the entire |
166 | * chain. |
167 | * |
168 | * NOTE: The chain allocator may be inefficient if the allocated objects |
169 | * are not much smaller than PAGE_SIZE. |
170 | */ |
171 | |
172 | struct chain_allocator { |
173 | struct linked_page *chain; /* the chain */ |
174 | unsigned int used_space; /* total size of objects allocated out |
175 | * of the current page |
176 | */ |
177 | gfp_t gfp_mask; /* mask for allocating pages */ |
178 | int safe_needed; /* if set, only "safe" pages are allocated */ |
179 | }; |
180 | |
181 | static void |
182 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) |
183 | { |
184 | ca->chain = NULL; |
185 | ca->used_space = LINKED_PAGE_DATA_SIZE; |
186 | ca->gfp_mask = gfp_mask; |
187 | ca->safe_needed = safe_needed; |
188 | } |
189 | |
190 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) |
191 | { |
192 | void *ret; |
193 | |
194 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { |
195 | struct linked_page *lp; |
196 | |
197 | lp = get_image_page(ca->gfp_mask, ca->safe_needed); |
198 | if (!lp) |
199 | return NULL; |
200 | |
201 | lp->next = ca->chain; |
202 | ca->chain = lp; |
203 | ca->used_space = 0; |
204 | } |
205 | ret = ca->chain->data + ca->used_space; |
206 | ca->used_space += size; |
207 | return ret; |
208 | } |
209 | |
210 | /** |
211 | * Data types related to memory bitmaps. |
212 | * |
213 | * Memory bitmap is a structure consiting of many linked lists of |
214 | * objects. The main list's elements are of type struct zone_bitmap |
215 | * and each of them corresonds to one zone. For each zone bitmap |
216 | * object there is a list of objects of type struct bm_block that |
217 | * represent each blocks of bitmap in which information is stored. |
218 | * |
219 | * struct memory_bitmap contains a pointer to the main list of zone |
220 | * bitmap objects, a struct bm_position used for browsing the bitmap, |
221 | * and a pointer to the list of pages used for allocating all of the |
222 | * zone bitmap objects and bitmap block objects. |
223 | * |
224 | * NOTE: It has to be possible to lay out the bitmap in memory |
225 | * using only allocations of order 0. Additionally, the bitmap is |
226 | * designed to work with arbitrary number of zones (this is over the |
227 | * top for now, but let's avoid making unnecessary assumptions ;-). |
228 | * |
229 | * struct zone_bitmap contains a pointer to a list of bitmap block |
230 | * objects and a pointer to the bitmap block object that has been |
231 | * most recently used for setting bits. Additionally, it contains the |
232 | * pfns that correspond to the start and end of the represented zone. |
233 | * |
234 | * struct bm_block contains a pointer to the memory page in which |
235 | * information is stored (in the form of a block of bitmap) |
236 | * It also contains the pfns that correspond to the start and end of |
237 | * the represented memory area. |
238 | */ |
239 | |
240 | #define BM_END_OF_MAP (~0UL) |
241 | |
242 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
243 | |
244 | struct bm_block { |
245 | struct list_head hook; /* hook into a list of bitmap blocks */ |
246 | unsigned long start_pfn; /* pfn represented by the first bit */ |
247 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ |
248 | unsigned long *data; /* bitmap representing pages */ |
249 | }; |
250 | |
251 | static inline unsigned long bm_block_bits(struct bm_block *bb) |
252 | { |
253 | return bb->end_pfn - bb->start_pfn; |
254 | } |
255 | |
256 | /* strcut bm_position is used for browsing memory bitmaps */ |
257 | |
258 | struct bm_position { |
259 | struct bm_block *block; |
260 | int bit; |
261 | }; |
262 | |
263 | struct memory_bitmap { |
264 | struct list_head blocks; /* list of bitmap blocks */ |
265 | struct linked_page *p_list; /* list of pages used to store zone |
266 | * bitmap objects and bitmap block |
267 | * objects |
268 | */ |
269 | struct bm_position cur; /* most recently used bit position */ |
270 | }; |
271 | |
272 | /* Functions that operate on memory bitmaps */ |
273 | |
274 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
275 | { |
276 | bm->cur.block = list_entry(bm->blocks.next, struct bm_block, hook); |
277 | bm->cur.bit = 0; |
278 | } |
279 | |
280 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); |
281 | |
282 | /** |
283 | * create_bm_block_list - create a list of block bitmap objects |
284 | * @pages - number of pages to track |
285 | * @list - list to put the allocated blocks into |
286 | * @ca - chain allocator to be used for allocating memory |
287 | */ |
288 | static int create_bm_block_list(unsigned long pages, |
289 | struct list_head *list, |
290 | struct chain_allocator *ca) |
291 | { |
292 | unsigned int nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); |
293 | |
294 | while (nr_blocks-- > 0) { |
295 | struct bm_block *bb; |
296 | |
297 | bb = chain_alloc(ca, sizeof(struct bm_block)); |
298 | if (!bb) |
299 | return -ENOMEM; |
300 | list_add(&bb->hook, list); |
301 | } |
302 | |
303 | return 0; |
304 | } |
305 | |
306 | struct mem_extent { |
307 | struct list_head hook; |
308 | unsigned long start; |
309 | unsigned long end; |
310 | }; |
311 | |
312 | /** |
313 | * free_mem_extents - free a list of memory extents |
314 | * @list - list of extents to empty |
315 | */ |
316 | static void free_mem_extents(struct list_head *list) |
317 | { |
318 | struct mem_extent *ext, *aux; |
319 | |
320 | list_for_each_entry_safe(ext, aux, list, hook) { |
321 | list_del(&ext->hook); |
322 | kfree(ext); |
323 | } |
324 | } |
325 | |
326 | /** |
327 | * create_mem_extents - create a list of memory extents representing |
328 | * contiguous ranges of PFNs |
329 | * @list - list to put the extents into |
330 | * @gfp_mask - mask to use for memory allocations |
331 | */ |
332 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) |
333 | { |
334 | struct zone *zone; |
335 | |
336 | INIT_LIST_HEAD(list); |
337 | |
338 | for_each_populated_zone(zone) { |
339 | unsigned long zone_start, zone_end; |
340 | struct mem_extent *ext, *cur, *aux; |
341 | |
342 | zone_start = zone->zone_start_pfn; |
343 | zone_end = zone->zone_start_pfn + zone->spanned_pages; |
344 | |
345 | list_for_each_entry(ext, list, hook) |
346 | if (zone_start <= ext->end) |
347 | break; |
348 | |
349 | if (&ext->hook == list || zone_end < ext->start) { |
350 | /* New extent is necessary */ |
351 | struct mem_extent *new_ext; |
352 | |
353 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); |
354 | if (!new_ext) { |
355 | free_mem_extents(list); |
356 | return -ENOMEM; |
357 | } |
358 | new_ext->start = zone_start; |
359 | new_ext->end = zone_end; |
360 | list_add_tail(&new_ext->hook, &ext->hook); |
361 | continue; |
362 | } |
363 | |
364 | /* Merge this zone's range of PFNs with the existing one */ |
365 | if (zone_start < ext->start) |
366 | ext->start = zone_start; |
367 | if (zone_end > ext->end) |
368 | ext->end = zone_end; |
369 | |
370 | /* More merging may be possible */ |
371 | cur = ext; |
372 | list_for_each_entry_safe_continue(cur, aux, list, hook) { |
373 | if (zone_end < cur->start) |
374 | break; |
375 | if (zone_end < cur->end) |
376 | ext->end = cur->end; |
377 | list_del(&cur->hook); |
378 | kfree(cur); |
379 | } |
380 | } |
381 | |
382 | return 0; |
383 | } |
384 | |
385 | /** |
386 | * memory_bm_create - allocate memory for a memory bitmap |
387 | */ |
388 | static int |
389 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) |
390 | { |
391 | struct chain_allocator ca; |
392 | struct list_head mem_extents; |
393 | struct mem_extent *ext; |
394 | int error; |
395 | |
396 | chain_init(&ca, gfp_mask, safe_needed); |
397 | INIT_LIST_HEAD(&bm->blocks); |
398 | |
399 | error = create_mem_extents(&mem_extents, gfp_mask); |
400 | if (error) |
401 | return error; |
402 | |
403 | list_for_each_entry(ext, &mem_extents, hook) { |
404 | struct bm_block *bb; |
405 | unsigned long pfn = ext->start; |
406 | unsigned long pages = ext->end - ext->start; |
407 | |
408 | bb = list_entry(bm->blocks.prev, struct bm_block, hook); |
409 | |
410 | error = create_bm_block_list(pages, bm->blocks.prev, &ca); |
411 | if (error) |
412 | goto Error; |
413 | |
414 | list_for_each_entry_continue(bb, &bm->blocks, hook) { |
415 | bb->data = get_image_page(gfp_mask, safe_needed); |
416 | if (!bb->data) { |
417 | error = -ENOMEM; |
418 | goto Error; |
419 | } |
420 | |
421 | bb->start_pfn = pfn; |
422 | if (pages >= BM_BITS_PER_BLOCK) { |
423 | pfn += BM_BITS_PER_BLOCK; |
424 | pages -= BM_BITS_PER_BLOCK; |
425 | } else { |
426 | /* This is executed only once in the loop */ |
427 | pfn += pages; |
428 | } |
429 | bb->end_pfn = pfn; |
430 | } |
431 | } |
432 | |
433 | bm->p_list = ca.chain; |
434 | memory_bm_position_reset(bm); |
435 | Exit: |
436 | free_mem_extents(&mem_extents); |
437 | return error; |
438 | |
439 | Error: |
440 | bm->p_list = ca.chain; |
441 | memory_bm_free(bm, PG_UNSAFE_CLEAR); |
442 | goto Exit; |
443 | } |
444 | |
445 | /** |
446 | * memory_bm_free - free memory occupied by the memory bitmap @bm |
447 | */ |
448 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
449 | { |
450 | struct bm_block *bb; |
451 | |
452 | list_for_each_entry(bb, &bm->blocks, hook) |
453 | if (bb->data) |
454 | free_image_page(bb->data, clear_nosave_free); |
455 | |
456 | free_list_of_pages(bm->p_list, clear_nosave_free); |
457 | |
458 | INIT_LIST_HEAD(&bm->blocks); |
459 | } |
460 | |
461 | /** |
462 | * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds |
463 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member |
464 | * of @bm->cur_zone_bm are updated. |
465 | */ |
466 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
467 | void **addr, unsigned int *bit_nr) |
468 | { |
469 | struct bm_block *bb; |
470 | |
471 | /* |
472 | * Check if the pfn corresponds to the current bitmap block and find |
473 | * the block where it fits if this is not the case. |
474 | */ |
475 | bb = bm->cur.block; |
476 | if (pfn < bb->start_pfn) |
477 | list_for_each_entry_continue_reverse(bb, &bm->blocks, hook) |
478 | if (pfn >= bb->start_pfn) |
479 | break; |
480 | |
481 | if (pfn >= bb->end_pfn) |
482 | list_for_each_entry_continue(bb, &bm->blocks, hook) |
483 | if (pfn >= bb->start_pfn && pfn < bb->end_pfn) |
484 | break; |
485 | |
486 | if (&bb->hook == &bm->blocks) |
487 | return -EFAULT; |
488 | |
489 | /* The block has been found */ |
490 | bm->cur.block = bb; |
491 | pfn -= bb->start_pfn; |
492 | bm->cur.bit = pfn + 1; |
493 | *bit_nr = pfn; |
494 | *addr = bb->data; |
495 | return 0; |
496 | } |
497 | |
498 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) |
499 | { |
500 | void *addr; |
501 | unsigned int bit; |
502 | int error; |
503 | |
504 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
505 | BUG_ON(error); |
506 | set_bit(bit, addr); |
507 | } |
508 | |
509 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
510 | { |
511 | void *addr; |
512 | unsigned int bit; |
513 | int error; |
514 | |
515 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
516 | if (!error) |
517 | set_bit(bit, addr); |
518 | return error; |
519 | } |
520 | |
521 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
522 | { |
523 | void *addr; |
524 | unsigned int bit; |
525 | int error; |
526 | |
527 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
528 | BUG_ON(error); |
529 | clear_bit(bit, addr); |
530 | } |
531 | |
532 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) |
533 | { |
534 | void *addr; |
535 | unsigned int bit; |
536 | int error; |
537 | |
538 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
539 | BUG_ON(error); |
540 | return test_bit(bit, addr); |
541 | } |
542 | |
543 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
544 | { |
545 | void *addr; |
546 | unsigned int bit; |
547 | |
548 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); |
549 | } |
550 | |
551 | /** |
552 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit |
553 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is |
554 | * returned. |
555 | * |
556 | * It is required to run memory_bm_position_reset() before the first call to |
557 | * this function. |
558 | */ |
559 | |
560 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) |
561 | { |
562 | struct bm_block *bb; |
563 | int bit; |
564 | |
565 | bb = bm->cur.block; |
566 | do { |
567 | bit = bm->cur.bit; |
568 | bit = find_next_bit(bb->data, bm_block_bits(bb), bit); |
569 | if (bit < bm_block_bits(bb)) |
570 | goto Return_pfn; |
571 | |
572 | bb = list_entry(bb->hook.next, struct bm_block, hook); |
573 | bm->cur.block = bb; |
574 | bm->cur.bit = 0; |
575 | } while (&bb->hook != &bm->blocks); |
576 | |
577 | memory_bm_position_reset(bm); |
578 | return BM_END_OF_MAP; |
579 | |
580 | Return_pfn: |
581 | bm->cur.bit = bit + 1; |
582 | return bb->start_pfn + bit; |
583 | } |
584 | |
585 | /** |
586 | * This structure represents a range of page frames the contents of which |
587 | * should not be saved during the suspend. |
588 | */ |
589 | |
590 | struct nosave_region { |
591 | struct list_head list; |
592 | unsigned long start_pfn; |
593 | unsigned long end_pfn; |
594 | }; |
595 | |
596 | static LIST_HEAD(nosave_regions); |
597 | |
598 | /** |
599 | * register_nosave_region - register a range of page frames the contents |
600 | * of which should not be saved during the suspend (to be used in the early |
601 | * initialization code) |
602 | */ |
603 | |
604 | void __init |
605 | __register_nosave_region(unsigned long start_pfn, unsigned long end_pfn, |
606 | int use_kmalloc) |
607 | { |
608 | struct nosave_region *region; |
609 | |
610 | if (start_pfn >= end_pfn) |
611 | return; |
612 | |
613 | if (!list_empty(&nosave_regions)) { |
614 | /* Try to extend the previous region (they should be sorted) */ |
615 | region = list_entry(nosave_regions.prev, |
616 | struct nosave_region, list); |
617 | if (region->end_pfn == start_pfn) { |
618 | region->end_pfn = end_pfn; |
619 | goto Report; |
620 | } |
621 | } |
622 | if (use_kmalloc) { |
623 | /* during init, this shouldn't fail */ |
624 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); |
625 | BUG_ON(!region); |
626 | } else |
627 | /* This allocation cannot fail */ |
628 | region = alloc_bootmem(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(KERN_INFO "PM: Registered nosave memory: %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 | pr_debug("PM: 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 | if (pfn_valid(pfn)) { |
708 | /* |
709 | * It is safe to ignore the result of |
710 | * mem_bm_set_bit_check() here, since we won't |
711 | * touch the PFNs for which the error is |
712 | * returned anyway. |
713 | */ |
714 | mem_bm_set_bit_check(bm, pfn); |
715 | } |
716 | } |
717 | } |
718 | |
719 | /** |
720 | * create_basic_memory_bitmaps - create bitmaps needed for marking page |
721 | * frames that should not be saved and free page frames. The pointers |
722 | * forbidden_pages_map and free_pages_map are only modified if everything |
723 | * goes well, because we don't want the bits to be used before both bitmaps |
724 | * are set up. |
725 | */ |
726 | |
727 | int create_basic_memory_bitmaps(void) |
728 | { |
729 | struct memory_bitmap *bm1, *bm2; |
730 | int error = 0; |
731 | |
732 | BUG_ON(forbidden_pages_map || free_pages_map); |
733 | |
734 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
735 | if (!bm1) |
736 | return -ENOMEM; |
737 | |
738 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
739 | if (error) |
740 | goto Free_first_object; |
741 | |
742 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
743 | if (!bm2) |
744 | goto Free_first_bitmap; |
745 | |
746 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
747 | if (error) |
748 | goto Free_second_object; |
749 | |
750 | forbidden_pages_map = bm1; |
751 | free_pages_map = bm2; |
752 | mark_nosave_pages(forbidden_pages_map); |
753 | |
754 | pr_debug("PM: Basic memory bitmaps created\n"); |
755 | |
756 | return 0; |
757 | |
758 | Free_second_object: |
759 | kfree(bm2); |
760 | Free_first_bitmap: |
761 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); |
762 | Free_first_object: |
763 | kfree(bm1); |
764 | return -ENOMEM; |
765 | } |
766 | |
767 | /** |
768 | * free_basic_memory_bitmaps - free memory bitmaps allocated by |
769 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary |
770 | * so that the bitmaps themselves are not referred to while they are being |
771 | * freed. |
772 | */ |
773 | |
774 | void free_basic_memory_bitmaps(void) |
775 | { |
776 | struct memory_bitmap *bm1, *bm2; |
777 | |
778 | BUG_ON(!(forbidden_pages_map && free_pages_map)); |
779 | |
780 | bm1 = forbidden_pages_map; |
781 | bm2 = free_pages_map; |
782 | forbidden_pages_map = NULL; |
783 | free_pages_map = NULL; |
784 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); |
785 | kfree(bm1); |
786 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); |
787 | kfree(bm2); |
788 | |
789 | pr_debug("PM: Basic memory bitmaps freed\n"); |
790 | } |
791 | |
792 | /** |
793 | * snapshot_additional_pages - estimate the number of additional pages |
794 | * be needed for setting up the suspend image data structures for given |
795 | * zone (usually the returned value is greater than the exact number) |
796 | */ |
797 | |
798 | unsigned int snapshot_additional_pages(struct zone *zone) |
799 | { |
800 | unsigned int res; |
801 | |
802 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); |
803 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), PAGE_SIZE); |
804 | return 2 * res; |
805 | } |
806 | |
807 | #ifdef CONFIG_HIGHMEM |
808 | /** |
809 | * count_free_highmem_pages - compute the total number of free highmem |
810 | * pages, system-wide. |
811 | */ |
812 | |
813 | static unsigned int count_free_highmem_pages(void) |
814 | { |
815 | struct zone *zone; |
816 | unsigned int cnt = 0; |
817 | |
818 | for_each_populated_zone(zone) |
819 | if (is_highmem(zone)) |
820 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
821 | |
822 | return cnt; |
823 | } |
824 | |
825 | /** |
826 | * saveable_highmem_page - Determine whether a highmem page should be |
827 | * included in the suspend image. |
828 | * |
829 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, |
830 | * and it isn't a part of a free chunk of pages. |
831 | */ |
832 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
833 | { |
834 | struct page *page; |
835 | |
836 | if (!pfn_valid(pfn)) |
837 | return NULL; |
838 | |
839 | page = pfn_to_page(pfn); |
840 | if (page_zone(page) != zone) |
841 | return NULL; |
842 | |
843 | BUG_ON(!PageHighMem(page)); |
844 | |
845 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
846 | PageReserved(page)) |
847 | return NULL; |
848 | |
849 | return page; |
850 | } |
851 | |
852 | /** |
853 | * count_highmem_pages - compute the total number of saveable highmem |
854 | * pages. |
855 | */ |
856 | |
857 | static unsigned int count_highmem_pages(void) |
858 | { |
859 | struct zone *zone; |
860 | unsigned int n = 0; |
861 | |
862 | for_each_populated_zone(zone) { |
863 | unsigned long pfn, max_zone_pfn; |
864 | |
865 | if (!is_highmem(zone)) |
866 | continue; |
867 | |
868 | mark_free_pages(zone); |
869 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
870 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
871 | if (saveable_highmem_page(zone, pfn)) |
872 | n++; |
873 | } |
874 | return n; |
875 | } |
876 | #else |
877 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
878 | { |
879 | return NULL; |
880 | } |
881 | #endif /* CONFIG_HIGHMEM */ |
882 | |
883 | /** |
884 | * saveable_page - Determine whether a non-highmem page should be included |
885 | * in the suspend image. |
886 | * |
887 | * We should save the page if it isn't Nosave, and is not in the range |
888 | * of pages statically defined as 'unsaveable', and it isn't a part of |
889 | * a free chunk of pages. |
890 | */ |
891 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
892 | { |
893 | struct page *page; |
894 | |
895 | if (!pfn_valid(pfn)) |
896 | return NULL; |
897 | |
898 | page = pfn_to_page(pfn); |
899 | if (page_zone(page) != zone) |
900 | return NULL; |
901 | |
902 | BUG_ON(PageHighMem(page)); |
903 | |
904 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
905 | return NULL; |
906 | |
907 | if (PageReserved(page) |
908 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) |
909 | return NULL; |
910 | |
911 | return page; |
912 | } |
913 | |
914 | /** |
915 | * count_data_pages - compute the total number of saveable non-highmem |
916 | * pages. |
917 | */ |
918 | |
919 | static unsigned int count_data_pages(void) |
920 | { |
921 | struct zone *zone; |
922 | unsigned long pfn, max_zone_pfn; |
923 | unsigned int n = 0; |
924 | |
925 | for_each_populated_zone(zone) { |
926 | if (is_highmem(zone)) |
927 | continue; |
928 | |
929 | mark_free_pages(zone); |
930 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
931 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
932 | if (saveable_page(zone, pfn)) |
933 | n++; |
934 | } |
935 | return n; |
936 | } |
937 | |
938 | /* This is needed, because copy_page and memcpy are not usable for copying |
939 | * task structs. |
940 | */ |
941 | static inline void do_copy_page(long *dst, long *src) |
942 | { |
943 | int n; |
944 | |
945 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
946 | *dst++ = *src++; |
947 | } |
948 | |
949 | |
950 | /** |
951 | * safe_copy_page - check if the page we are going to copy is marked as |
952 | * present in the kernel page tables (this always is the case if |
953 | * CONFIG_DEBUG_PAGEALLOC is not set and in that case |
954 | * kernel_page_present() always returns 'true'). |
955 | */ |
956 | static void safe_copy_page(void *dst, struct page *s_page) |
957 | { |
958 | if (kernel_page_present(s_page)) { |
959 | do_copy_page(dst, page_address(s_page)); |
960 | } else { |
961 | kernel_map_pages(s_page, 1, 1); |
962 | do_copy_page(dst, page_address(s_page)); |
963 | kernel_map_pages(s_page, 1, 0); |
964 | } |
965 | } |
966 | |
967 | |
968 | #ifdef CONFIG_HIGHMEM |
969 | static inline struct page * |
970 | page_is_saveable(struct zone *zone, unsigned long pfn) |
971 | { |
972 | return is_highmem(zone) ? |
973 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
974 | } |
975 | |
976 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
977 | { |
978 | struct page *s_page, *d_page; |
979 | void *src, *dst; |
980 | |
981 | s_page = pfn_to_page(src_pfn); |
982 | d_page = pfn_to_page(dst_pfn); |
983 | if (PageHighMem(s_page)) { |
984 | src = kmap_atomic(s_page, KM_USER0); |
985 | dst = kmap_atomic(d_page, KM_USER1); |
986 | do_copy_page(dst, src); |
987 | kunmap_atomic(dst, KM_USER1); |
988 | kunmap_atomic(src, KM_USER0); |
989 | } else { |
990 | if (PageHighMem(d_page)) { |
991 | /* Page pointed to by src may contain some kernel |
992 | * data modified by kmap_atomic() |
993 | */ |
994 | safe_copy_page(buffer, s_page); |
995 | dst = kmap_atomic(d_page, KM_USER0); |
996 | copy_page(dst, buffer); |
997 | kunmap_atomic(dst, KM_USER0); |
998 | } else { |
999 | safe_copy_page(page_address(d_page), s_page); |
1000 | } |
1001 | } |
1002 | } |
1003 | #else |
1004 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
1005 | |
1006 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
1007 | { |
1008 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1009 | pfn_to_page(src_pfn)); |
1010 | } |
1011 | #endif /* CONFIG_HIGHMEM */ |
1012 | |
1013 | static void |
1014 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) |
1015 | { |
1016 | struct zone *zone; |
1017 | unsigned long pfn; |
1018 | |
1019 | for_each_populated_zone(zone) { |
1020 | unsigned long max_zone_pfn; |
1021 | |
1022 | mark_free_pages(zone); |
1023 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1024 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
1025 | if (page_is_saveable(zone, pfn)) |
1026 | memory_bm_set_bit(orig_bm, pfn); |
1027 | } |
1028 | memory_bm_position_reset(orig_bm); |
1029 | memory_bm_position_reset(copy_bm); |
1030 | for(;;) { |
1031 | pfn = memory_bm_next_pfn(orig_bm); |
1032 | if (unlikely(pfn == BM_END_OF_MAP)) |
1033 | break; |
1034 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); |
1035 | } |
1036 | } |
1037 | |
1038 | /* Total number of image pages */ |
1039 | static unsigned int nr_copy_pages; |
1040 | /* Number of pages needed for saving the original pfns of the image pages */ |
1041 | static unsigned int nr_meta_pages; |
1042 | /* |
1043 | * Numbers of normal and highmem page frames allocated for hibernation image |
1044 | * before suspending devices. |
1045 | */ |
1046 | unsigned int alloc_normal, alloc_highmem; |
1047 | /* |
1048 | * Memory bitmap used for marking saveable pages (during hibernation) or |
1049 | * hibernation image pages (during restore) |
1050 | */ |
1051 | static struct memory_bitmap orig_bm; |
1052 | /* |
1053 | * Memory bitmap used during hibernation for marking allocated page frames that |
1054 | * will contain copies of saveable pages. During restore it is initially used |
1055 | * for marking hibernation image pages, but then the set bits from it are |
1056 | * duplicated in @orig_bm and it is released. On highmem systems it is next |
1057 | * used for marking "safe" highmem pages, but it has to be reinitialized for |
1058 | * this purpose. |
1059 | */ |
1060 | static struct memory_bitmap copy_bm; |
1061 | |
1062 | /** |
1063 | * swsusp_free - free pages allocated for the suspend. |
1064 | * |
1065 | * Suspend pages are alocated before the atomic copy is made, so we |
1066 | * need to release them after the resume. |
1067 | */ |
1068 | |
1069 | void swsusp_free(void) |
1070 | { |
1071 | struct zone *zone; |
1072 | unsigned long pfn, max_zone_pfn; |
1073 | |
1074 | for_each_populated_zone(zone) { |
1075 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1076 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
1077 | if (pfn_valid(pfn)) { |
1078 | struct page *page = pfn_to_page(pfn); |
1079 | |
1080 | if (swsusp_page_is_forbidden(page) && |
1081 | swsusp_page_is_free(page)) { |
1082 | swsusp_unset_page_forbidden(page); |
1083 | swsusp_unset_page_free(page); |
1084 | __free_page(page); |
1085 | } |
1086 | } |
1087 | } |
1088 | nr_copy_pages = 0; |
1089 | nr_meta_pages = 0; |
1090 | restore_pblist = NULL; |
1091 | buffer = NULL; |
1092 | alloc_normal = 0; |
1093 | alloc_highmem = 0; |
1094 | } |
1095 | |
1096 | /* Helper functions used for the shrinking of memory. */ |
1097 | |
1098 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) |
1099 | |
1100 | /** |
1101 | * preallocate_image_pages - Allocate a number of pages for hibernation image |
1102 | * @nr_pages: Number of page frames to allocate. |
1103 | * @mask: GFP flags to use for the allocation. |
1104 | * |
1105 | * Return value: Number of page frames actually allocated |
1106 | */ |
1107 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) |
1108 | { |
1109 | unsigned long nr_alloc = 0; |
1110 | |
1111 | while (nr_pages > 0) { |
1112 | struct page *page; |
1113 | |
1114 | page = alloc_image_page(mask); |
1115 | if (!page) |
1116 | break; |
1117 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1118 | if (PageHighMem(page)) |
1119 | alloc_highmem++; |
1120 | else |
1121 | alloc_normal++; |
1122 | nr_pages--; |
1123 | nr_alloc++; |
1124 | } |
1125 | |
1126 | return nr_alloc; |
1127 | } |
1128 | |
1129 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1130 | unsigned long avail_normal) |
1131 | { |
1132 | unsigned long alloc; |
1133 | |
1134 | if (avail_normal <= alloc_normal) |
1135 | return 0; |
1136 | |
1137 | alloc = avail_normal - alloc_normal; |
1138 | if (nr_pages < alloc) |
1139 | alloc = nr_pages; |
1140 | |
1141 | return preallocate_image_pages(alloc, GFP_IMAGE); |
1142 | } |
1143 | |
1144 | #ifdef CONFIG_HIGHMEM |
1145 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) |
1146 | { |
1147 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); |
1148 | } |
1149 | |
1150 | /** |
1151 | * __fraction - Compute (an approximation of) x * (multiplier / base) |
1152 | */ |
1153 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1154 | { |
1155 | x *= multiplier; |
1156 | do_div(x, base); |
1157 | return (unsigned long)x; |
1158 | } |
1159 | |
1160 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
1161 | unsigned long highmem, |
1162 | unsigned long total) |
1163 | { |
1164 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1165 | |
1166 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); |
1167 | } |
1168 | #else /* CONFIG_HIGHMEM */ |
1169 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) |
1170 | { |
1171 | return 0; |
1172 | } |
1173 | |
1174 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
1175 | unsigned long highmem, |
1176 | unsigned long total) |
1177 | { |
1178 | return 0; |
1179 | } |
1180 | #endif /* CONFIG_HIGHMEM */ |
1181 | |
1182 | /** |
1183 | * free_unnecessary_pages - Release preallocated pages not needed for the image |
1184 | */ |
1185 | static void free_unnecessary_pages(void) |
1186 | { |
1187 | unsigned long save, to_free_normal, to_free_highmem; |
1188 | |
1189 | save = count_data_pages(); |
1190 | if (alloc_normal >= save) { |
1191 | to_free_normal = alloc_normal - save; |
1192 | save = 0; |
1193 | } else { |
1194 | to_free_normal = 0; |
1195 | save -= alloc_normal; |
1196 | } |
1197 | save += count_highmem_pages(); |
1198 | if (alloc_highmem >= save) { |
1199 | to_free_highmem = alloc_highmem - save; |
1200 | } else { |
1201 | to_free_highmem = 0; |
1202 | to_free_normal -= save - alloc_highmem; |
1203 | } |
1204 | |
1205 | memory_bm_position_reset(©_bm); |
1206 | |
1207 | while (to_free_normal > 0 || to_free_highmem > 0) { |
1208 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1209 | struct page *page = pfn_to_page(pfn); |
1210 | |
1211 | if (PageHighMem(page)) { |
1212 | if (!to_free_highmem) |
1213 | continue; |
1214 | to_free_highmem--; |
1215 | alloc_highmem--; |
1216 | } else { |
1217 | if (!to_free_normal) |
1218 | continue; |
1219 | to_free_normal--; |
1220 | alloc_normal--; |
1221 | } |
1222 | memory_bm_clear_bit(©_bm, pfn); |
1223 | swsusp_unset_page_forbidden(page); |
1224 | swsusp_unset_page_free(page); |
1225 | __free_page(page); |
1226 | } |
1227 | } |
1228 | |
1229 | /** |
1230 | * minimum_image_size - Estimate the minimum acceptable size of an image |
1231 | * @saveable: Number of saveable pages in the system. |
1232 | * |
1233 | * We want to avoid attempting to free too much memory too hard, so estimate the |
1234 | * minimum acceptable size of a hibernation image to use as the lower limit for |
1235 | * preallocating memory. |
1236 | * |
1237 | * We assume that the minimum image size should be proportional to |
1238 | * |
1239 | * [number of saveable pages] - [number of pages that can be freed in theory] |
1240 | * |
1241 | * where the second term is the sum of (1) reclaimable slab pages, (2) active |
1242 | * and (3) inactive anonymouns pages, (4) active and (5) inactive file pages, |
1243 | * minus mapped file pages. |
1244 | */ |
1245 | static unsigned long minimum_image_size(unsigned long saveable) |
1246 | { |
1247 | unsigned long size; |
1248 | |
1249 | size = global_page_state(NR_SLAB_RECLAIMABLE) |
1250 | + global_page_state(NR_ACTIVE_ANON) |
1251 | + global_page_state(NR_INACTIVE_ANON) |
1252 | + global_page_state(NR_ACTIVE_FILE) |
1253 | + global_page_state(NR_INACTIVE_FILE) |
1254 | - global_page_state(NR_FILE_MAPPED); |
1255 | |
1256 | return saveable <= size ? 0 : saveable - size; |
1257 | } |
1258 | |
1259 | /** |
1260 | * hibernate_preallocate_memory - Preallocate memory for hibernation image |
1261 | * |
1262 | * To create a hibernation image it is necessary to make a copy of every page |
1263 | * frame in use. We also need a number of page frames to be free during |
1264 | * hibernation for allocations made while saving the image and for device |
1265 | * drivers, in case they need to allocate memory from their hibernation |
1266 | * callbacks (these two numbers are given by PAGES_FOR_IO and SPARE_PAGES, |
1267 | * respectively, both of which are rough estimates). To make this happen, we |
1268 | * compute the total number of available page frames and allocate at least |
1269 | * |
1270 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 + 2 * SPARE_PAGES |
1271 | * |
1272 | * of them, which corresponds to the maximum size of a hibernation image. |
1273 | * |
1274 | * If image_size is set below the number following from the above formula, |
1275 | * the preallocation of memory is continued until the total number of saveable |
1276 | * pages in the system is below the requested image size or the minimum |
1277 | * acceptable image size returned by minimum_image_size(), whichever is greater. |
1278 | */ |
1279 | int hibernate_preallocate_memory(void) |
1280 | { |
1281 | struct zone *zone; |
1282 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
1283 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
1284 | struct timeval start, stop; |
1285 | int error; |
1286 | |
1287 | printk(KERN_INFO "PM: Preallocating image memory... "); |
1288 | do_gettimeofday(&start); |
1289 | |
1290 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1291 | if (error) |
1292 | goto err_out; |
1293 | |
1294 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); |
1295 | if (error) |
1296 | goto err_out; |
1297 | |
1298 | alloc_normal = 0; |
1299 | alloc_highmem = 0; |
1300 | |
1301 | /* Count the number of saveable data pages. */ |
1302 | save_highmem = count_highmem_pages(); |
1303 | saveable = count_data_pages(); |
1304 | |
1305 | /* |
1306 | * Compute the total number of page frames we can use (count) and the |
1307 | * number of pages needed for image metadata (size). |
1308 | */ |
1309 | count = saveable; |
1310 | saveable += save_highmem; |
1311 | highmem = save_highmem; |
1312 | size = 0; |
1313 | for_each_populated_zone(zone) { |
1314 | size += snapshot_additional_pages(zone); |
1315 | if (is_highmem(zone)) |
1316 | highmem += zone_page_state(zone, NR_FREE_PAGES); |
1317 | else |
1318 | count += zone_page_state(zone, NR_FREE_PAGES); |
1319 | } |
1320 | avail_normal = count; |
1321 | count += highmem; |
1322 | count -= totalreserve_pages; |
1323 | |
1324 | /* Compute the maximum number of saveable pages to leave in memory. */ |
1325 | max_size = (count - (size + PAGES_FOR_IO)) / 2 - 2 * SPARE_PAGES; |
1326 | /* Compute the desired number of image pages specified by image_size. */ |
1327 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1328 | if (size > max_size) |
1329 | size = max_size; |
1330 | /* |
1331 | * If the desired number of image pages is at least as large as the |
1332 | * current number of saveable pages in memory, allocate page frames for |
1333 | * the image and we're done. |
1334 | */ |
1335 | if (size >= saveable) { |
1336 | pages = preallocate_image_highmem(save_highmem); |
1337 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
1338 | goto out; |
1339 | } |
1340 | |
1341 | /* Estimate the minimum size of the image. */ |
1342 | pages = minimum_image_size(saveable); |
1343 | /* |
1344 | * To avoid excessive pressure on the normal zone, leave room in it to |
1345 | * accommodate an image of the minimum size (unless it's already too |
1346 | * small, in which case don't preallocate pages from it at all). |
1347 | */ |
1348 | if (avail_normal > pages) |
1349 | avail_normal -= pages; |
1350 | else |
1351 | avail_normal = 0; |
1352 | if (size < pages) |
1353 | size = min_t(unsigned long, pages, max_size); |
1354 | |
1355 | /* |
1356 | * Let the memory management subsystem know that we're going to need a |
1357 | * large number of page frames to allocate and make it free some memory. |
1358 | * NOTE: If this is not done, performance will be hurt badly in some |
1359 | * test cases. |
1360 | */ |
1361 | shrink_all_memory(saveable - size); |
1362 | |
1363 | /* |
1364 | * The number of saveable pages in memory was too high, so apply some |
1365 | * pressure to decrease it. First, make room for the largest possible |
1366 | * image and fail if that doesn't work. Next, try to decrease the size |
1367 | * of the image as much as indicated by 'size' using allocations from |
1368 | * highmem and non-highmem zones separately. |
1369 | */ |
1370 | pages_highmem = preallocate_image_highmem(highmem / 2); |
1371 | alloc = (count - max_size) - pages_highmem; |
1372 | pages = preallocate_image_memory(alloc, avail_normal); |
1373 | if (pages < alloc) { |
1374 | /* We have exhausted non-highmem pages, try highmem. */ |
1375 | alloc -= pages; |
1376 | pages += pages_highmem; |
1377 | pages_highmem = preallocate_image_highmem(alloc); |
1378 | if (pages_highmem < alloc) |
1379 | goto err_out; |
1380 | pages += pages_highmem; |
1381 | /* |
1382 | * size is the desired number of saveable pages to leave in |
1383 | * memory, so try to preallocate (all memory - size) pages. |
1384 | */ |
1385 | alloc = (count - pages) - size; |
1386 | pages += preallocate_image_highmem(alloc); |
1387 | } else { |
1388 | /* |
1389 | * There are approximately max_size saveable pages at this point |
1390 | * and we want to reduce this number down to size. |
1391 | */ |
1392 | alloc = max_size - size; |
1393 | size = preallocate_highmem_fraction(alloc, highmem, count); |
1394 | pages_highmem += size; |
1395 | alloc -= size; |
1396 | size = preallocate_image_memory(alloc, avail_normal); |
1397 | pages_highmem += preallocate_image_highmem(alloc - size); |
1398 | pages += pages_highmem + size; |
1399 | } |
1400 | |
1401 | /* |
1402 | * We only need as many page frames for the image as there are saveable |
1403 | * pages in memory, but we have allocated more. Release the excessive |
1404 | * ones now. |
1405 | */ |
1406 | free_unnecessary_pages(); |
1407 | |
1408 | out: |
1409 | do_gettimeofday(&stop); |
1410 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
1411 | swsusp_show_speed(&start, &stop, pages, "Allocated"); |
1412 | |
1413 | return 0; |
1414 | |
1415 | err_out: |
1416 | printk(KERN_CONT "\n"); |
1417 | swsusp_free(); |
1418 | return -ENOMEM; |
1419 | } |
1420 | |
1421 | #ifdef CONFIG_HIGHMEM |
1422 | /** |
1423 | * count_pages_for_highmem - compute the number of non-highmem pages |
1424 | * that will be necessary for creating copies of highmem pages. |
1425 | */ |
1426 | |
1427 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) |
1428 | { |
1429 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
1430 | |
1431 | if (free_highmem >= nr_highmem) |
1432 | nr_highmem = 0; |
1433 | else |
1434 | nr_highmem -= free_highmem; |
1435 | |
1436 | return nr_highmem; |
1437 | } |
1438 | #else |
1439 | static unsigned int |
1440 | count_pages_for_highmem(unsigned int nr_highmem) { return 0; } |
1441 | #endif /* CONFIG_HIGHMEM */ |
1442 | |
1443 | /** |
1444 | * enough_free_mem - Make sure we have enough free memory for the |
1445 | * snapshot image. |
1446 | */ |
1447 | |
1448 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
1449 | { |
1450 | struct zone *zone; |
1451 | unsigned int free = alloc_normal; |
1452 | |
1453 | for_each_populated_zone(zone) |
1454 | if (!is_highmem(zone)) |
1455 | free += zone_page_state(zone, NR_FREE_PAGES); |
1456 | |
1457 | nr_pages += count_pages_for_highmem(nr_highmem); |
1458 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1459 | nr_pages, PAGES_FOR_IO, free); |
1460 | |
1461 | return free > nr_pages + PAGES_FOR_IO; |
1462 | } |
1463 | |
1464 | #ifdef CONFIG_HIGHMEM |
1465 | /** |
1466 | * get_highmem_buffer - if there are some highmem pages in the suspend |
1467 | * image, we may need the buffer to copy them and/or load their data. |
1468 | */ |
1469 | |
1470 | static inline int get_highmem_buffer(int safe_needed) |
1471 | { |
1472 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); |
1473 | return buffer ? 0 : -ENOMEM; |
1474 | } |
1475 | |
1476 | /** |
1477 | * alloc_highmem_image_pages - allocate some highmem pages for the image. |
1478 | * Try to allocate as many pages as needed, but if the number of free |
1479 | * highmem pages is lesser than that, allocate them all. |
1480 | */ |
1481 | |
1482 | static inline unsigned int |
1483 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int nr_highmem) |
1484 | { |
1485 | unsigned int to_alloc = count_free_highmem_pages(); |
1486 | |
1487 | if (to_alloc > nr_highmem) |
1488 | to_alloc = nr_highmem; |
1489 | |
1490 | nr_highmem -= to_alloc; |
1491 | while (to_alloc-- > 0) { |
1492 | struct page *page; |
1493 | |
1494 | page = alloc_image_page(__GFP_HIGHMEM); |
1495 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1496 | } |
1497 | return nr_highmem; |
1498 | } |
1499 | #else |
1500 | static inline int get_highmem_buffer(int safe_needed) { return 0; } |
1501 | |
1502 | static inline unsigned int |
1503 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int n) { return 0; } |
1504 | #endif /* CONFIG_HIGHMEM */ |
1505 | |
1506 | /** |
1507 | * swsusp_alloc - allocate memory for the suspend image |
1508 | * |
1509 | * We first try to allocate as many highmem pages as there are |
1510 | * saveable highmem pages in the system. If that fails, we allocate |
1511 | * non-highmem pages for the copies of the remaining highmem ones. |
1512 | * |
1513 | * In this approach it is likely that the copies of highmem pages will |
1514 | * also be located in the high memory, because of the way in which |
1515 | * copy_data_pages() works. |
1516 | */ |
1517 | |
1518 | static int |
1519 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, |
1520 | unsigned int nr_pages, unsigned int nr_highmem) |
1521 | { |
1522 | if (nr_highmem > 0) { |
1523 | if (get_highmem_buffer(PG_ANY)) |
1524 | goto err_out; |
1525 | if (nr_highmem > alloc_highmem) { |
1526 | nr_highmem -= alloc_highmem; |
1527 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); |
1528 | } |
1529 | } |
1530 | if (nr_pages > alloc_normal) { |
1531 | nr_pages -= alloc_normal; |
1532 | while (nr_pages-- > 0) { |
1533 | struct page *page; |
1534 | |
1535 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); |
1536 | if (!page) |
1537 | goto err_out; |
1538 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); |
1539 | } |
1540 | } |
1541 | |
1542 | return 0; |
1543 | |
1544 | err_out: |
1545 | swsusp_free(); |
1546 | return -ENOMEM; |
1547 | } |
1548 | |
1549 | asmlinkage int swsusp_save(void) |
1550 | { |
1551 | unsigned int nr_pages, nr_highmem; |
1552 | |
1553 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
1554 | |
1555 | drain_local_pages(NULL); |
1556 | nr_pages = count_data_pages(); |
1557 | nr_highmem = count_highmem_pages(); |
1558 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
1559 | |
1560 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
1561 | printk(KERN_ERR "PM: Not enough free memory\n"); |
1562 | return -ENOMEM; |
1563 | } |
1564 | |
1565 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
1566 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
1567 | return -ENOMEM; |
1568 | } |
1569 | |
1570 | /* During allocating of suspend pagedir, new cold pages may appear. |
1571 | * Kill them. |
1572 | */ |
1573 | drain_local_pages(NULL); |
1574 | copy_data_pages(©_bm, &orig_bm); |
1575 | |
1576 | /* |
1577 | * End of critical section. From now on, we can write to memory, |
1578 | * but we should not touch disk. This specially means we must _not_ |
1579 | * touch swap space! Except we must write out our image of course. |
1580 | */ |
1581 | |
1582 | nr_pages += nr_highmem; |
1583 | nr_copy_pages = nr_pages; |
1584 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
1585 | |
1586 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1587 | nr_pages); |
1588 | |
1589 | return 0; |
1590 | } |
1591 | |
1592 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1593 | static int init_header_complete(struct swsusp_info *info) |
1594 | { |
1595 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
1596 | info->version_code = LINUX_VERSION_CODE; |
1597 | return 0; |
1598 | } |
1599 | |
1600 | static char *check_image_kernel(struct swsusp_info *info) |
1601 | { |
1602 | if (info->version_code != LINUX_VERSION_CODE) |
1603 | return "kernel version"; |
1604 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) |
1605 | return "system type"; |
1606 | if (strcmp(info->uts.release,init_utsname()->release)) |
1607 | return "kernel release"; |
1608 | if (strcmp(info->uts.version,init_utsname()->version)) |
1609 | return "version"; |
1610 | if (strcmp(info->uts.machine,init_utsname()->machine)) |
1611 | return "machine"; |
1612 | return NULL; |
1613 | } |
1614 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ |
1615 | |
1616 | unsigned long snapshot_get_image_size(void) |
1617 | { |
1618 | return nr_copy_pages + nr_meta_pages + 1; |
1619 | } |
1620 | |
1621 | static int init_header(struct swsusp_info *info) |
1622 | { |
1623 | memset(info, 0, sizeof(struct swsusp_info)); |
1624 | info->num_physpages = num_physpages; |
1625 | info->image_pages = nr_copy_pages; |
1626 | info->pages = snapshot_get_image_size(); |
1627 | info->size = info->pages; |
1628 | info->size <<= PAGE_SHIFT; |
1629 | return init_header_complete(info); |
1630 | } |
1631 | |
1632 | /** |
1633 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1634 | * are stored in the array @buf[] (1 page at a time) |
1635 | */ |
1636 | |
1637 | static inline void |
1638 | pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
1639 | { |
1640 | int j; |
1641 | |
1642 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1643 | buf[j] = memory_bm_next_pfn(bm); |
1644 | if (unlikely(buf[j] == BM_END_OF_MAP)) |
1645 | break; |
1646 | } |
1647 | } |
1648 | |
1649 | /** |
1650 | * snapshot_read_next - used for reading the system memory snapshot. |
1651 | * |
1652 | * On the first call to it @handle should point to a zeroed |
1653 | * snapshot_handle structure. The structure gets updated and a pointer |
1654 | * to it should be passed to this function every next time. |
1655 | * |
1656 | * On success the function returns a positive number. Then, the caller |
1657 | * is allowed to read up to the returned number of bytes from the memory |
1658 | * location computed by the data_of() macro. |
1659 | * |
1660 | * The function returns 0 to indicate the end of data stream condition, |
1661 | * and a negative number is returned on error. In such cases the |
1662 | * structure pointed to by @handle is not updated and should not be used |
1663 | * any more. |
1664 | */ |
1665 | |
1666 | int snapshot_read_next(struct snapshot_handle *handle) |
1667 | { |
1668 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
1669 | return 0; |
1670 | |
1671 | if (!buffer) { |
1672 | /* This makes the buffer be freed by swsusp_free() */ |
1673 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
1674 | if (!buffer) |
1675 | return -ENOMEM; |
1676 | } |
1677 | if (!handle->cur) { |
1678 | int error; |
1679 | |
1680 | error = init_header((struct swsusp_info *)buffer); |
1681 | if (error) |
1682 | return error; |
1683 | handle->buffer = buffer; |
1684 | memory_bm_position_reset(&orig_bm); |
1685 | memory_bm_position_reset(©_bm); |
1686 | } else if (handle->cur <= nr_meta_pages) { |
1687 | clear_page(buffer); |
1688 | pack_pfns(buffer, &orig_bm); |
1689 | } else { |
1690 | struct page *page; |
1691 | |
1692 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
1693 | if (PageHighMem(page)) { |
1694 | /* Highmem pages are copied to the buffer, |
1695 | * because we can't return with a kmapped |
1696 | * highmem page (we may not be called again). |
1697 | */ |
1698 | void *kaddr; |
1699 | |
1700 | kaddr = kmap_atomic(page, KM_USER0); |
1701 | copy_page(buffer, kaddr); |
1702 | kunmap_atomic(kaddr, KM_USER0); |
1703 | handle->buffer = buffer; |
1704 | } else { |
1705 | handle->buffer = page_address(page); |
1706 | } |
1707 | } |
1708 | handle->cur++; |
1709 | return PAGE_SIZE; |
1710 | } |
1711 | |
1712 | /** |
1713 | * mark_unsafe_pages - mark the pages that cannot be used for storing |
1714 | * the image during resume, because they conflict with the pages that |
1715 | * had been used before suspend |
1716 | */ |
1717 | |
1718 | static int mark_unsafe_pages(struct memory_bitmap *bm) |
1719 | { |
1720 | struct zone *zone; |
1721 | unsigned long pfn, max_zone_pfn; |
1722 | |
1723 | /* Clear page flags */ |
1724 | for_each_populated_zone(zone) { |
1725 | max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages; |
1726 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
1727 | if (pfn_valid(pfn)) |
1728 | swsusp_unset_page_free(pfn_to_page(pfn)); |
1729 | } |
1730 | |
1731 | /* Mark pages that correspond to the "original" pfns as "unsafe" */ |
1732 | memory_bm_position_reset(bm); |
1733 | do { |
1734 | pfn = memory_bm_next_pfn(bm); |
1735 | if (likely(pfn != BM_END_OF_MAP)) { |
1736 | if (likely(pfn_valid(pfn))) |
1737 | swsusp_set_page_free(pfn_to_page(pfn)); |
1738 | else |
1739 | return -EFAULT; |
1740 | } |
1741 | } while (pfn != BM_END_OF_MAP); |
1742 | |
1743 | allocated_unsafe_pages = 0; |
1744 | |
1745 | return 0; |
1746 | } |
1747 | |
1748 | static void |
1749 | duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src) |
1750 | { |
1751 | unsigned long pfn; |
1752 | |
1753 | memory_bm_position_reset(src); |
1754 | pfn = memory_bm_next_pfn(src); |
1755 | while (pfn != BM_END_OF_MAP) { |
1756 | memory_bm_set_bit(dst, pfn); |
1757 | pfn = memory_bm_next_pfn(src); |
1758 | } |
1759 | } |
1760 | |
1761 | static int check_header(struct swsusp_info *info) |
1762 | { |
1763 | char *reason; |
1764 | |
1765 | reason = check_image_kernel(info); |
1766 | if (!reason && info->num_physpages != num_physpages) |
1767 | reason = "memory size"; |
1768 | if (reason) { |
1769 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
1770 | return -EPERM; |
1771 | } |
1772 | return 0; |
1773 | } |
1774 | |
1775 | /** |
1776 | * load header - check the image header and copy data from it |
1777 | */ |
1778 | |
1779 | static int |
1780 | load_header(struct swsusp_info *info) |
1781 | { |
1782 | int error; |
1783 | |
1784 | restore_pblist = NULL; |
1785 | error = check_header(info); |
1786 | if (!error) { |
1787 | nr_copy_pages = info->image_pages; |
1788 | nr_meta_pages = info->pages - info->image_pages - 1; |
1789 | } |
1790 | return error; |
1791 | } |
1792 | |
1793 | /** |
1794 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
1795 | * the corresponding bit in the memory bitmap @bm |
1796 | */ |
1797 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
1798 | { |
1799 | int j; |
1800 | |
1801 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1802 | if (unlikely(buf[j] == BM_END_OF_MAP)) |
1803 | break; |
1804 | |
1805 | if (memory_bm_pfn_present(bm, buf[j])) |
1806 | memory_bm_set_bit(bm, buf[j]); |
1807 | else |
1808 | return -EFAULT; |
1809 | } |
1810 | |
1811 | return 0; |
1812 | } |
1813 | |
1814 | /* List of "safe" pages that may be used to store data loaded from the suspend |
1815 | * image |
1816 | */ |
1817 | static struct linked_page *safe_pages_list; |
1818 | |
1819 | #ifdef CONFIG_HIGHMEM |
1820 | /* struct highmem_pbe is used for creating the list of highmem pages that |
1821 | * should be restored atomically during the resume from disk, because the page |
1822 | * frames they have occupied before the suspend are in use. |
1823 | */ |
1824 | struct highmem_pbe { |
1825 | struct page *copy_page; /* data is here now */ |
1826 | struct page *orig_page; /* data was here before the suspend */ |
1827 | struct highmem_pbe *next; |
1828 | }; |
1829 | |
1830 | /* List of highmem PBEs needed for restoring the highmem pages that were |
1831 | * allocated before the suspend and included in the suspend image, but have |
1832 | * also been allocated by the "resume" kernel, so their contents cannot be |
1833 | * written directly to their "original" page frames. |
1834 | */ |
1835 | static struct highmem_pbe *highmem_pblist; |
1836 | |
1837 | /** |
1838 | * count_highmem_image_pages - compute the number of highmem pages in the |
1839 | * suspend image. The bits in the memory bitmap @bm that correspond to the |
1840 | * image pages are assumed to be set. |
1841 | */ |
1842 | |
1843 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) |
1844 | { |
1845 | unsigned long pfn; |
1846 | unsigned int cnt = 0; |
1847 | |
1848 | memory_bm_position_reset(bm); |
1849 | pfn = memory_bm_next_pfn(bm); |
1850 | while (pfn != BM_END_OF_MAP) { |
1851 | if (PageHighMem(pfn_to_page(pfn))) |
1852 | cnt++; |
1853 | |
1854 | pfn = memory_bm_next_pfn(bm); |
1855 | } |
1856 | return cnt; |
1857 | } |
1858 | |
1859 | /** |
1860 | * prepare_highmem_image - try to allocate as many highmem pages as |
1861 | * there are highmem image pages (@nr_highmem_p points to the variable |
1862 | * containing the number of highmem image pages). The pages that are |
1863 | * "safe" (ie. will not be overwritten when the suspend image is |
1864 | * restored) have the corresponding bits set in @bm (it must be |
1865 | * unitialized). |
1866 | * |
1867 | * NOTE: This function should not be called if there are no highmem |
1868 | * image pages. |
1869 | */ |
1870 | |
1871 | static unsigned int safe_highmem_pages; |
1872 | |
1873 | static struct memory_bitmap *safe_highmem_bm; |
1874 | |
1875 | static int |
1876 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) |
1877 | { |
1878 | unsigned int to_alloc; |
1879 | |
1880 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) |
1881 | return -ENOMEM; |
1882 | |
1883 | if (get_highmem_buffer(PG_SAFE)) |
1884 | return -ENOMEM; |
1885 | |
1886 | to_alloc = count_free_highmem_pages(); |
1887 | if (to_alloc > *nr_highmem_p) |
1888 | to_alloc = *nr_highmem_p; |
1889 | else |
1890 | *nr_highmem_p = to_alloc; |
1891 | |
1892 | safe_highmem_pages = 0; |
1893 | while (to_alloc-- > 0) { |
1894 | struct page *page; |
1895 | |
1896 | page = alloc_page(__GFP_HIGHMEM); |
1897 | if (!swsusp_page_is_free(page)) { |
1898 | /* The page is "safe", set its bit the bitmap */ |
1899 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1900 | safe_highmem_pages++; |
1901 | } |
1902 | /* Mark the page as allocated */ |
1903 | swsusp_set_page_forbidden(page); |
1904 | swsusp_set_page_free(page); |
1905 | } |
1906 | memory_bm_position_reset(bm); |
1907 | safe_highmem_bm = bm; |
1908 | return 0; |
1909 | } |
1910 | |
1911 | /** |
1912 | * get_highmem_page_buffer - for given highmem image page find the buffer |
1913 | * that suspend_write_next() should set for its caller to write to. |
1914 | * |
1915 | * If the page is to be saved to its "original" page frame or a copy of |
1916 | * the page is to be made in the highmem, @buffer is returned. Otherwise, |
1917 | * the copy of the page is to be made in normal memory, so the address of |
1918 | * the copy is returned. |
1919 | * |
1920 | * If @buffer is returned, the caller of suspend_write_next() will write |
1921 | * the page's contents to @buffer, so they will have to be copied to the |
1922 | * right location on the next call to suspend_write_next() and it is done |
1923 | * with the help of copy_last_highmem_page(). For this purpose, if |
1924 | * @buffer is returned, @last_highmem page is set to the page to which |
1925 | * the data will have to be copied from @buffer. |
1926 | */ |
1927 | |
1928 | static struct page *last_highmem_page; |
1929 | |
1930 | static void * |
1931 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) |
1932 | { |
1933 | struct highmem_pbe *pbe; |
1934 | void *kaddr; |
1935 | |
1936 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
1937 | /* We have allocated the "original" page frame and we can |
1938 | * use it directly to store the loaded page. |
1939 | */ |
1940 | last_highmem_page = page; |
1941 | return buffer; |
1942 | } |
1943 | /* The "original" page frame has not been allocated and we have to |
1944 | * use a "safe" page frame to store the loaded page. |
1945 | */ |
1946 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); |
1947 | if (!pbe) { |
1948 | swsusp_free(); |
1949 | return ERR_PTR(-ENOMEM); |
1950 | } |
1951 | pbe->orig_page = page; |
1952 | if (safe_highmem_pages > 0) { |
1953 | struct page *tmp; |
1954 | |
1955 | /* Copy of the page will be stored in high memory */ |
1956 | kaddr = buffer; |
1957 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); |
1958 | safe_highmem_pages--; |
1959 | last_highmem_page = tmp; |
1960 | pbe->copy_page = tmp; |
1961 | } else { |
1962 | /* Copy of the page will be stored in normal memory */ |
1963 | kaddr = safe_pages_list; |
1964 | safe_pages_list = safe_pages_list->next; |
1965 | pbe->copy_page = virt_to_page(kaddr); |
1966 | } |
1967 | pbe->next = highmem_pblist; |
1968 | highmem_pblist = pbe; |
1969 | return kaddr; |
1970 | } |
1971 | |
1972 | /** |
1973 | * copy_last_highmem_page - copy the contents of a highmem image from |
1974 | * @buffer, where the caller of snapshot_write_next() has place them, |
1975 | * to the right location represented by @last_highmem_page . |
1976 | */ |
1977 | |
1978 | static void copy_last_highmem_page(void) |
1979 | { |
1980 | if (last_highmem_page) { |
1981 | void *dst; |
1982 | |
1983 | dst = kmap_atomic(last_highmem_page, KM_USER0); |
1984 | copy_page(dst, buffer); |
1985 | kunmap_atomic(dst, KM_USER0); |
1986 | last_highmem_page = NULL; |
1987 | } |
1988 | } |
1989 | |
1990 | static inline int last_highmem_page_copied(void) |
1991 | { |
1992 | return !last_highmem_page; |
1993 | } |
1994 | |
1995 | static inline void free_highmem_data(void) |
1996 | { |
1997 | if (safe_highmem_bm) |
1998 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); |
1999 | |
2000 | if (buffer) |
2001 | free_image_page(buffer, PG_UNSAFE_CLEAR); |
2002 | } |
2003 | #else |
2004 | static inline int get_safe_write_buffer(void) { return 0; } |
2005 | |
2006 | static unsigned int |
2007 | count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } |
2008 | |
2009 | static inline int |
2010 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) |
2011 | { |
2012 | return 0; |
2013 | } |
2014 | |
2015 | static inline void * |
2016 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) |
2017 | { |
2018 | return ERR_PTR(-EINVAL); |
2019 | } |
2020 | |
2021 | static inline void copy_last_highmem_page(void) {} |
2022 | static inline int last_highmem_page_copied(void) { return 1; } |
2023 | static inline void free_highmem_data(void) {} |
2024 | #endif /* CONFIG_HIGHMEM */ |
2025 | |
2026 | /** |
2027 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
2028 | * be overwritten in the process of restoring the system memory state |
2029 | * from the suspend image ("unsafe" pages) and allocate memory for the |
2030 | * image. |
2031 | * |
2032 | * The idea is to allocate a new memory bitmap first and then allocate |
2033 | * as many pages as needed for the image data, but not to assign these |
2034 | * pages to specific tasks initially. Instead, we just mark them as |
2035 | * allocated and create a lists of "safe" pages that will be used |
2036 | * later. On systems with high memory a list of "safe" highmem pages is |
2037 | * also created. |
2038 | */ |
2039 | |
2040 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2041 | |
2042 | static int |
2043 | prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) |
2044 | { |
2045 | unsigned int nr_pages, nr_highmem; |
2046 | struct linked_page *sp_list, *lp; |
2047 | int error; |
2048 | |
2049 | /* If there is no highmem, the buffer will not be necessary */ |
2050 | free_image_page(buffer, PG_UNSAFE_CLEAR); |
2051 | buffer = NULL; |
2052 | |
2053 | nr_highmem = count_highmem_image_pages(bm); |
2054 | error = mark_unsafe_pages(bm); |
2055 | if (error) |
2056 | goto Free; |
2057 | |
2058 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); |
2059 | if (error) |
2060 | goto Free; |
2061 | |
2062 | duplicate_memory_bitmap(new_bm, bm); |
2063 | memory_bm_free(bm, PG_UNSAFE_KEEP); |
2064 | if (nr_highmem > 0) { |
2065 | error = prepare_highmem_image(bm, &nr_highmem); |
2066 | if (error) |
2067 | goto Free; |
2068 | } |
2069 | /* Reserve some safe pages for potential later use. |
2070 | * |
2071 | * NOTE: This way we make sure there will be enough safe pages for the |
2072 | * chain_alloc() in get_buffer(). It is a bit wasteful, but |
2073 | * nr_copy_pages cannot be greater than 50% of the memory anyway. |
2074 | */ |
2075 | sp_list = NULL; |
2076 | /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */ |
2077 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
2078 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2079 | while (nr_pages > 0) { |
2080 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
2081 | if (!lp) { |
2082 | error = -ENOMEM; |
2083 | goto Free; |
2084 | } |
2085 | lp->next = sp_list; |
2086 | sp_list = lp; |
2087 | nr_pages--; |
2088 | } |
2089 | /* Preallocate memory for the image */ |
2090 | safe_pages_list = NULL; |
2091 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
2092 | while (nr_pages > 0) { |
2093 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); |
2094 | if (!lp) { |
2095 | error = -ENOMEM; |
2096 | goto Free; |
2097 | } |
2098 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
2099 | /* The page is "safe", add it to the list */ |
2100 | lp->next = safe_pages_list; |
2101 | safe_pages_list = lp; |
2102 | } |
2103 | /* Mark the page as allocated */ |
2104 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2105 | swsusp_set_page_free(virt_to_page(lp)); |
2106 | nr_pages--; |
2107 | } |
2108 | /* Free the reserved safe pages so that chain_alloc() can use them */ |
2109 | while (sp_list) { |
2110 | lp = sp_list->next; |
2111 | free_image_page(sp_list, PG_UNSAFE_CLEAR); |
2112 | sp_list = lp; |
2113 | } |
2114 | return 0; |
2115 | |
2116 | Free: |
2117 | swsusp_free(); |
2118 | return error; |
2119 | } |
2120 | |
2121 | /** |
2122 | * get_buffer - compute the address that snapshot_write_next() should |
2123 | * set for its caller to write to. |
2124 | */ |
2125 | |
2126 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) |
2127 | { |
2128 | struct pbe *pbe; |
2129 | struct page *page; |
2130 | unsigned long pfn = memory_bm_next_pfn(bm); |
2131 | |
2132 | if (pfn == BM_END_OF_MAP) |
2133 | return ERR_PTR(-EFAULT); |
2134 | |
2135 | page = pfn_to_page(pfn); |
2136 | if (PageHighMem(page)) |
2137 | return get_highmem_page_buffer(page, ca); |
2138 | |
2139 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
2140 | /* We have allocated the "original" page frame and we can |
2141 | * use it directly to store the loaded page. |
2142 | */ |
2143 | return page_address(page); |
2144 | |
2145 | /* The "original" page frame has not been allocated and we have to |
2146 | * use a "safe" page frame to store the loaded page. |
2147 | */ |
2148 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2149 | if (!pbe) { |
2150 | swsusp_free(); |
2151 | return ERR_PTR(-ENOMEM); |
2152 | } |
2153 | pbe->orig_address = page_address(page); |
2154 | pbe->address = safe_pages_list; |
2155 | safe_pages_list = safe_pages_list->next; |
2156 | pbe->next = restore_pblist; |
2157 | restore_pblist = pbe; |
2158 | return pbe->address; |
2159 | } |
2160 | |
2161 | /** |
2162 | * snapshot_write_next - used for writing the system memory snapshot. |
2163 | * |
2164 | * On the first call to it @handle should point to a zeroed |
2165 | * snapshot_handle structure. The structure gets updated and a pointer |
2166 | * to it should be passed to this function every next time. |
2167 | * |
2168 | * On success the function returns a positive number. Then, the caller |
2169 | * is allowed to write up to the returned number of bytes to the memory |
2170 | * location computed by the data_of() macro. |
2171 | * |
2172 | * The function returns 0 to indicate the "end of file" condition, |
2173 | * and a negative number is returned on error. In such cases the |
2174 | * structure pointed to by @handle is not updated and should not be used |
2175 | * any more. |
2176 | */ |
2177 | |
2178 | int snapshot_write_next(struct snapshot_handle *handle) |
2179 | { |
2180 | static struct chain_allocator ca; |
2181 | int error = 0; |
2182 | |
2183 | /* Check if we have already loaded the entire image */ |
2184 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
2185 | return 0; |
2186 | |
2187 | handle->sync_read = 1; |
2188 | |
2189 | if (!handle->cur) { |
2190 | if (!buffer) |
2191 | /* This makes the buffer be freed by swsusp_free() */ |
2192 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
2193 | |
2194 | if (!buffer) |
2195 | return -ENOMEM; |
2196 | |
2197 | handle->buffer = buffer; |
2198 | } else if (handle->cur == 1) { |
2199 | error = load_header(buffer); |
2200 | if (error) |
2201 | return error; |
2202 | |
2203 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2204 | if (error) |
2205 | return error; |
2206 | |
2207 | } else if (handle->cur <= nr_meta_pages + 1) { |
2208 | error = unpack_orig_pfns(buffer, ©_bm); |
2209 | if (error) |
2210 | return error; |
2211 | |
2212 | if (handle->cur == nr_meta_pages + 1) { |
2213 | error = prepare_image(&orig_bm, ©_bm); |
2214 | if (error) |
2215 | return error; |
2216 | |
2217 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2218 | memory_bm_position_reset(&orig_bm); |
2219 | restore_pblist = NULL; |
2220 | handle->buffer = get_buffer(&orig_bm, &ca); |
2221 | handle->sync_read = 0; |
2222 | if (IS_ERR(handle->buffer)) |
2223 | return PTR_ERR(handle->buffer); |
2224 | } |
2225 | } else { |
2226 | copy_last_highmem_page(); |
2227 | handle->buffer = get_buffer(&orig_bm, &ca); |
2228 | if (IS_ERR(handle->buffer)) |
2229 | return PTR_ERR(handle->buffer); |
2230 | if (handle->buffer != buffer) |
2231 | handle->sync_read = 0; |
2232 | } |
2233 | handle->cur++; |
2234 | return PAGE_SIZE; |
2235 | } |
2236 | |
2237 | /** |
2238 | * snapshot_write_finalize - must be called after the last call to |
2239 | * snapshot_write_next() in case the last page in the image happens |
2240 | * to be a highmem page and its contents should be stored in the |
2241 | * highmem. Additionally, it releases the memory that will not be |
2242 | * used any more. |
2243 | */ |
2244 | |
2245 | void snapshot_write_finalize(struct snapshot_handle *handle) |
2246 | { |
2247 | copy_last_highmem_page(); |
2248 | /* Free only if we have loaded the image entirely */ |
2249 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
2250 | memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR); |
2251 | free_highmem_data(); |
2252 | } |
2253 | } |
2254 | |
2255 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2256 | { |
2257 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
2258 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2259 | } |
2260 | |
2261 | #ifdef CONFIG_HIGHMEM |
2262 | /* Assumes that @buf is ready and points to a "safe" page */ |
2263 | static inline void |
2264 | swap_two_pages_data(struct page *p1, struct page *p2, void *buf) |
2265 | { |
2266 | void *kaddr1, *kaddr2; |
2267 | |
2268 | kaddr1 = kmap_atomic(p1, KM_USER0); |
2269 | kaddr2 = kmap_atomic(p2, KM_USER1); |
2270 | copy_page(buf, kaddr1); |
2271 | copy_page(kaddr1, kaddr2); |
2272 | copy_page(kaddr2, buf); |
2273 | kunmap_atomic(kaddr2, KM_USER1); |
2274 | kunmap_atomic(kaddr1, KM_USER0); |
2275 | } |
2276 | |
2277 | /** |
2278 | * restore_highmem - for each highmem page that was allocated before |
2279 | * the suspend and included in the suspend image, and also has been |
2280 | * allocated by the "resume" kernel swap its current (ie. "before |
2281 | * resume") contents with the previous (ie. "before suspend") one. |
2282 | * |
2283 | * If the resume eventually fails, we can call this function once |
2284 | * again and restore the "before resume" highmem state. |
2285 | */ |
2286 | |
2287 | int restore_highmem(void) |
2288 | { |
2289 | struct highmem_pbe *pbe = highmem_pblist; |
2290 | void *buf; |
2291 | |
2292 | if (!pbe) |
2293 | return 0; |
2294 | |
2295 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); |
2296 | if (!buf) |
2297 | return -ENOMEM; |
2298 | |
2299 | while (pbe) { |
2300 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); |
2301 | pbe = pbe->next; |
2302 | } |
2303 | free_image_page(buf, PG_UNSAFE_CLEAR); |
2304 | return 0; |
2305 | } |
2306 | #endif /* CONFIG_HIGHMEM */ |
2307 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9