Commit | Line | Data |
---|---|---|
25761b6e | 1 | /* |
96bc7aec | 2 | * linux/kernel/power/snapshot.c |
25761b6e | 3 | * |
8357376d | 4 | * This file provides system snapshot/restore functionality for swsusp. |
25761b6e | 5 | * |
a2531293 | 6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz> |
8357376d | 7 | * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl> |
25761b6e | 8 | * |
8357376d | 9 | * This file is released under the GPLv2. |
25761b6e RW |
10 | * |
11 | */ | |
12 | ||
f577eb30 | 13 | #include <linux/version.h> |
25761b6e RW |
14 | #include <linux/module.h> |
15 | #include <linux/mm.h> | |
16 | #include <linux/suspend.h> | |
25761b6e | 17 | #include <linux/delay.h> |
25761b6e | 18 | #include <linux/bitops.h> |
25761b6e | 19 | #include <linux/spinlock.h> |
25761b6e | 20 | #include <linux/kernel.h> |
25761b6e RW |
21 | #include <linux/pm.h> |
22 | #include <linux/device.h> | |
74dfd666 | 23 | #include <linux/init.h> |
25761b6e RW |
24 | #include <linux/bootmem.h> |
25 | #include <linux/syscalls.h> | |
26 | #include <linux/console.h> | |
27 | #include <linux/highmem.h> | |
846705de | 28 | #include <linux/list.h> |
5a0e3ad6 | 29 | #include <linux/slab.h> |
25761b6e RW |
30 | |
31 | #include <asm/uaccess.h> | |
32 | #include <asm/mmu_context.h> | |
33 | #include <asm/pgtable.h> | |
34 | #include <asm/tlbflush.h> | |
35 | #include <asm/io.h> | |
36 | ||
25761b6e RW |
37 | #include "power.h" |
38 | ||
74dfd666 RW |
39 | static int swsusp_page_is_free(struct page *); |
40 | static void swsusp_set_page_forbidden(struct page *); | |
41 | static void swsusp_unset_page_forbidden(struct page *); | |
42 | ||
ddeb6487 RW |
43 | /* |
44 | * Number of bytes to reserve for memory allocations made by device drivers | |
45 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
46 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
47 | */ | |
48 | unsigned long reserved_size; | |
49 | ||
50 | void __init hibernate_reserved_size_init(void) | |
51 | { | |
52 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
53 | } | |
54 | ||
fe419535 RW |
55 | /* |
56 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
57 | * When it is set to N, swsusp will do its best to ensure the image |
58 | * size will not exceed N bytes, but if that is impossible, it will | |
59 | * try to create the smallest image possible. | |
fe419535 | 60 | */ |
ac5c24ec RW |
61 | unsigned long image_size; |
62 | ||
63 | void __init hibernate_image_size_init(void) | |
64 | { | |
1c1be3a9 | 65 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 66 | } |
fe419535 | 67 | |
8357376d RW |
68 | /* List of PBEs needed for restoring the pages that were allocated before |
69 | * the suspend and included in the suspend image, but have also been | |
70 | * allocated by the "resume" kernel, so their contents cannot be written | |
71 | * directly to their "original" page frames. | |
72 | */ | |
75534b50 RW |
73 | struct pbe *restore_pblist; |
74 | ||
8357376d | 75 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 76 | static void *buffer; |
7088a5c0 | 77 | |
f6143aa6 RW |
78 | /** |
79 | * @safe_needed - on resume, for storing the PBE list and the image, | |
80 | * we can only use memory pages that do not conflict with the pages | |
8357376d RW |
81 | * used before suspend. The unsafe pages have PageNosaveFree set |
82 | * and we count them using unsafe_pages. | |
f6143aa6 | 83 | * |
8357376d RW |
84 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
85 | * so that swsusp_free() can release it. | |
f6143aa6 RW |
86 | */ |
87 | ||
0bcd888d RW |
88 | #define PG_ANY 0 |
89 | #define PG_SAFE 1 | |
90 | #define PG_UNSAFE_CLEAR 1 | |
91 | #define PG_UNSAFE_KEEP 0 | |
92 | ||
940864dd | 93 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 94 | |
8357376d | 95 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
96 | { |
97 | void *res; | |
98 | ||
99 | res = (void *)get_zeroed_page(gfp_mask); | |
100 | if (safe_needed) | |
7be98234 | 101 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 102 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 103 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 104 | allocated_unsafe_pages++; |
f6143aa6 RW |
105 | res = (void *)get_zeroed_page(gfp_mask); |
106 | } | |
107 | if (res) { | |
7be98234 RW |
108 | swsusp_set_page_forbidden(virt_to_page(res)); |
109 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
110 | } |
111 | return res; | |
112 | } | |
113 | ||
114 | unsigned long get_safe_page(gfp_t gfp_mask) | |
115 | { | |
8357376d RW |
116 | return (unsigned long)get_image_page(gfp_mask, PG_SAFE); |
117 | } | |
118 | ||
5b6d15de RW |
119 | static struct page *alloc_image_page(gfp_t gfp_mask) |
120 | { | |
8357376d RW |
121 | struct page *page; |
122 | ||
123 | page = alloc_page(gfp_mask); | |
124 | if (page) { | |
7be98234 RW |
125 | swsusp_set_page_forbidden(page); |
126 | swsusp_set_page_free(page); | |
8357376d RW |
127 | } |
128 | return page; | |
f6143aa6 RW |
129 | } |
130 | ||
131 | /** | |
132 | * free_image_page - free page represented by @addr, allocated with | |
8357376d | 133 | * get_image_page (page flags set by it must be cleared) |
f6143aa6 RW |
134 | */ |
135 | ||
136 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
137 | { | |
8357376d RW |
138 | struct page *page; |
139 | ||
140 | BUG_ON(!virt_addr_valid(addr)); | |
141 | ||
142 | page = virt_to_page(addr); | |
143 | ||
7be98234 | 144 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 145 | if (clear_nosave_free) |
7be98234 | 146 | swsusp_unset_page_free(page); |
8357376d RW |
147 | |
148 | __free_page(page); | |
f6143aa6 RW |
149 | } |
150 | ||
b788db79 RW |
151 | /* struct linked_page is used to build chains of pages */ |
152 | ||
153 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
154 | ||
155 | struct linked_page { | |
156 | struct linked_page *next; | |
157 | char data[LINKED_PAGE_DATA_SIZE]; | |
158 | } __attribute__((packed)); | |
159 | ||
160 | static inline void | |
161 | free_list_of_pages(struct linked_page *list, int clear_page_nosave) | |
162 | { | |
163 | while (list) { | |
164 | struct linked_page *lp = list->next; | |
165 | ||
166 | free_image_page(list, clear_page_nosave); | |
167 | list = lp; | |
168 | } | |
169 | } | |
170 | ||
171 | /** | |
172 | * struct chain_allocator is used for allocating small objects out of | |
173 | * a linked list of pages called 'the chain'. | |
174 | * | |
175 | * The chain grows each time when there is no room for a new object in | |
176 | * the current page. The allocated objects cannot be freed individually. | |
177 | * It is only possible to free them all at once, by freeing the entire | |
178 | * chain. | |
179 | * | |
180 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
181 | * are not much smaller than PAGE_SIZE. | |
182 | */ | |
183 | ||
184 | struct chain_allocator { | |
185 | struct linked_page *chain; /* the chain */ | |
186 | unsigned int used_space; /* total size of objects allocated out | |
187 | * of the current page | |
188 | */ | |
189 | gfp_t gfp_mask; /* mask for allocating pages */ | |
190 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
191 | }; | |
192 | ||
193 | static void | |
194 | chain_init(struct chain_allocator *ca, gfp_t gfp_mask, int safe_needed) | |
195 | { | |
196 | ca->chain = NULL; | |
197 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
198 | ca->gfp_mask = gfp_mask; | |
199 | ca->safe_needed = safe_needed; | |
200 | } | |
201 | ||
202 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
203 | { | |
204 | void *ret; | |
205 | ||
206 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
207 | struct linked_page *lp; | |
208 | ||
8357376d | 209 | lp = get_image_page(ca->gfp_mask, ca->safe_needed); |
b788db79 RW |
210 | if (!lp) |
211 | return NULL; | |
212 | ||
213 | lp->next = ca->chain; | |
214 | ca->chain = lp; | |
215 | ca->used_space = 0; | |
216 | } | |
217 | ret = ca->chain->data + ca->used_space; | |
218 | ca->used_space += size; | |
219 | return ret; | |
220 | } | |
221 | ||
b788db79 RW |
222 | /** |
223 | * Data types related to memory bitmaps. | |
224 | * | |
225 | * Memory bitmap is a structure consiting of many linked lists of | |
226 | * objects. The main list's elements are of type struct zone_bitmap | |
227 | * and each of them corresonds to one zone. For each zone bitmap | |
228 | * object there is a list of objects of type struct bm_block that | |
0d83304c | 229 | * represent each blocks of bitmap in which information is stored. |
b788db79 RW |
230 | * |
231 | * struct memory_bitmap contains a pointer to the main list of zone | |
232 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
233 | * and a pointer to the list of pages used for allocating all of the | |
234 | * zone bitmap objects and bitmap block objects. | |
235 | * | |
236 | * NOTE: It has to be possible to lay out the bitmap in memory | |
237 | * using only allocations of order 0. Additionally, the bitmap is | |
238 | * designed to work with arbitrary number of zones (this is over the | |
239 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
240 | * | |
241 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
242 | * objects and a pointer to the bitmap block object that has been | |
243 | * most recently used for setting bits. Additionally, it contains the | |
244 | * pfns that correspond to the start and end of the represented zone. | |
245 | * | |
246 | * struct bm_block contains a pointer to the memory page in which | |
0d83304c AM |
247 | * information is stored (in the form of a block of bitmap) |
248 | * It also contains the pfns that correspond to the start and end of | |
249 | * the represented memory area. | |
b788db79 RW |
250 | */ |
251 | ||
252 | #define BM_END_OF_MAP (~0UL) | |
253 | ||
8de03073 | 254 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
b788db79 RW |
255 | |
256 | struct bm_block { | |
846705de | 257 | struct list_head hook; /* hook into a list of bitmap blocks */ |
b788db79 RW |
258 | unsigned long start_pfn; /* pfn represented by the first bit */ |
259 | unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ | |
0d83304c | 260 | unsigned long *data; /* bitmap representing pages */ |
b788db79 RW |
261 | }; |
262 | ||
0d83304c AM |
263 | static inline unsigned long bm_block_bits(struct bm_block *bb) |
264 | { | |
265 | return bb->end_pfn - bb->start_pfn; | |
266 | } | |
267 | ||
b788db79 RW |
268 | /* strcut bm_position is used for browsing memory bitmaps */ |
269 | ||
270 | struct bm_position { | |
b788db79 | 271 | struct bm_block *block; |
b788db79 RW |
272 | int bit; |
273 | }; | |
274 | ||
275 | struct memory_bitmap { | |
846705de | 276 | struct list_head blocks; /* list of bitmap blocks */ |
b788db79 RW |
277 | struct linked_page *p_list; /* list of pages used to store zone |
278 | * bitmap objects and bitmap block | |
279 | * objects | |
280 | */ | |
281 | struct bm_position cur; /* most recently used bit position */ | |
282 | }; | |
283 | ||
284 | /* Functions that operate on memory bitmaps */ | |
285 | ||
b788db79 RW |
286 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
287 | { | |
846705de | 288 | bm->cur.block = list_entry(bm->blocks.next, struct bm_block, hook); |
0d83304c | 289 | bm->cur.bit = 0; |
b788db79 RW |
290 | } |
291 | ||
292 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
293 | ||
294 | /** | |
295 | * create_bm_block_list - create a list of block bitmap objects | |
8de03073 | 296 | * @pages - number of pages to track |
846705de RW |
297 | * @list - list to put the allocated blocks into |
298 | * @ca - chain allocator to be used for allocating memory | |
b788db79 | 299 | */ |
846705de RW |
300 | static int create_bm_block_list(unsigned long pages, |
301 | struct list_head *list, | |
302 | struct chain_allocator *ca) | |
b788db79 | 303 | { |
846705de | 304 | unsigned int nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); |
b788db79 RW |
305 | |
306 | while (nr_blocks-- > 0) { | |
307 | struct bm_block *bb; | |
308 | ||
309 | bb = chain_alloc(ca, sizeof(struct bm_block)); | |
310 | if (!bb) | |
846705de RW |
311 | return -ENOMEM; |
312 | list_add(&bb->hook, list); | |
b788db79 | 313 | } |
846705de RW |
314 | |
315 | return 0; | |
b788db79 RW |
316 | } |
317 | ||
846705de RW |
318 | struct mem_extent { |
319 | struct list_head hook; | |
320 | unsigned long start; | |
321 | unsigned long end; | |
322 | }; | |
323 | ||
b788db79 | 324 | /** |
846705de RW |
325 | * free_mem_extents - free a list of memory extents |
326 | * @list - list of extents to empty | |
b788db79 | 327 | */ |
846705de RW |
328 | static void free_mem_extents(struct list_head *list) |
329 | { | |
330 | struct mem_extent *ext, *aux; | |
b788db79 | 331 | |
846705de RW |
332 | list_for_each_entry_safe(ext, aux, list, hook) { |
333 | list_del(&ext->hook); | |
334 | kfree(ext); | |
335 | } | |
336 | } | |
337 | ||
338 | /** | |
339 | * create_mem_extents - create a list of memory extents representing | |
340 | * contiguous ranges of PFNs | |
341 | * @list - list to put the extents into | |
342 | * @gfp_mask - mask to use for memory allocations | |
343 | */ | |
344 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 345 | { |
846705de | 346 | struct zone *zone; |
b788db79 | 347 | |
846705de | 348 | INIT_LIST_HEAD(list); |
b788db79 | 349 | |
ee99c71c | 350 | for_each_populated_zone(zone) { |
846705de RW |
351 | unsigned long zone_start, zone_end; |
352 | struct mem_extent *ext, *cur, *aux; | |
353 | ||
846705de | 354 | zone_start = zone->zone_start_pfn; |
c33bc315 | 355 | zone_end = zone_end_pfn(zone); |
846705de RW |
356 | |
357 | list_for_each_entry(ext, list, hook) | |
358 | if (zone_start <= ext->end) | |
359 | break; | |
b788db79 | 360 | |
846705de RW |
361 | if (&ext->hook == list || zone_end < ext->start) { |
362 | /* New extent is necessary */ | |
363 | struct mem_extent *new_ext; | |
364 | ||
365 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
366 | if (!new_ext) { | |
367 | free_mem_extents(list); | |
368 | return -ENOMEM; | |
369 | } | |
370 | new_ext->start = zone_start; | |
371 | new_ext->end = zone_end; | |
372 | list_add_tail(&new_ext->hook, &ext->hook); | |
373 | continue; | |
374 | } | |
375 | ||
376 | /* Merge this zone's range of PFNs with the existing one */ | |
377 | if (zone_start < ext->start) | |
378 | ext->start = zone_start; | |
379 | if (zone_end > ext->end) | |
380 | ext->end = zone_end; | |
381 | ||
382 | /* More merging may be possible */ | |
383 | cur = ext; | |
384 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
385 | if (zone_end < cur->start) | |
386 | break; | |
387 | if (zone_end < cur->end) | |
388 | ext->end = cur->end; | |
389 | list_del(&cur->hook); | |
390 | kfree(cur); | |
391 | } | |
b788db79 | 392 | } |
846705de RW |
393 | |
394 | return 0; | |
b788db79 RW |
395 | } |
396 | ||
397 | /** | |
398 | * memory_bm_create - allocate memory for a memory bitmap | |
399 | */ | |
b788db79 RW |
400 | static int |
401 | memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed) | |
402 | { | |
403 | struct chain_allocator ca; | |
846705de RW |
404 | struct list_head mem_extents; |
405 | struct mem_extent *ext; | |
406 | int error; | |
b788db79 RW |
407 | |
408 | chain_init(&ca, gfp_mask, safe_needed); | |
846705de | 409 | INIT_LIST_HEAD(&bm->blocks); |
b788db79 | 410 | |
846705de RW |
411 | error = create_mem_extents(&mem_extents, gfp_mask); |
412 | if (error) | |
413 | return error; | |
b788db79 | 414 | |
846705de RW |
415 | list_for_each_entry(ext, &mem_extents, hook) { |
416 | struct bm_block *bb; | |
417 | unsigned long pfn = ext->start; | |
418 | unsigned long pages = ext->end - ext->start; | |
b788db79 | 419 | |
846705de | 420 | bb = list_entry(bm->blocks.prev, struct bm_block, hook); |
b788db79 | 421 | |
846705de RW |
422 | error = create_bm_block_list(pages, bm->blocks.prev, &ca); |
423 | if (error) | |
424 | goto Error; | |
b788db79 | 425 | |
846705de RW |
426 | list_for_each_entry_continue(bb, &bm->blocks, hook) { |
427 | bb->data = get_image_page(gfp_mask, safe_needed); | |
428 | if (!bb->data) { | |
429 | error = -ENOMEM; | |
430 | goto Error; | |
431 | } | |
b788db79 RW |
432 | |
433 | bb->start_pfn = pfn; | |
846705de | 434 | if (pages >= BM_BITS_PER_BLOCK) { |
b788db79 | 435 | pfn += BM_BITS_PER_BLOCK; |
846705de | 436 | pages -= BM_BITS_PER_BLOCK; |
b788db79 RW |
437 | } else { |
438 | /* This is executed only once in the loop */ | |
846705de | 439 | pfn += pages; |
b788db79 RW |
440 | } |
441 | bb->end_pfn = pfn; | |
b788db79 | 442 | } |
b788db79 | 443 | } |
846705de | 444 | |
b788db79 RW |
445 | bm->p_list = ca.chain; |
446 | memory_bm_position_reset(bm); | |
846705de RW |
447 | Exit: |
448 | free_mem_extents(&mem_extents); | |
449 | return error; | |
b788db79 | 450 | |
846705de | 451 | Error: |
b788db79 RW |
452 | bm->p_list = ca.chain; |
453 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 454 | goto Exit; |
b788db79 RW |
455 | } |
456 | ||
457 | /** | |
458 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
459 | */ | |
b788db79 RW |
460 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
461 | { | |
846705de | 462 | struct bm_block *bb; |
b788db79 | 463 | |
846705de RW |
464 | list_for_each_entry(bb, &bm->blocks, hook) |
465 | if (bb->data) | |
466 | free_image_page(bb->data, clear_nosave_free); | |
b788db79 | 467 | |
b788db79 | 468 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de RW |
469 | |
470 | INIT_LIST_HEAD(&bm->blocks); | |
b788db79 RW |
471 | } |
472 | ||
473 | /** | |
74dfd666 | 474 | * memory_bm_find_bit - find the bit in the bitmap @bm that corresponds |
b788db79 RW |
475 | * to given pfn. The cur_zone_bm member of @bm and the cur_block member |
476 | * of @bm->cur_zone_bm are updated. | |
b788db79 | 477 | */ |
a82f7119 | 478 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
74dfd666 | 479 | void **addr, unsigned int *bit_nr) |
b788db79 | 480 | { |
b788db79 RW |
481 | struct bm_block *bb; |
482 | ||
846705de RW |
483 | /* |
484 | * Check if the pfn corresponds to the current bitmap block and find | |
485 | * the block where it fits if this is not the case. | |
486 | */ | |
487 | bb = bm->cur.block; | |
b788db79 | 488 | if (pfn < bb->start_pfn) |
846705de RW |
489 | list_for_each_entry_continue_reverse(bb, &bm->blocks, hook) |
490 | if (pfn >= bb->start_pfn) | |
491 | break; | |
b788db79 | 492 | |
846705de RW |
493 | if (pfn >= bb->end_pfn) |
494 | list_for_each_entry_continue(bb, &bm->blocks, hook) | |
495 | if (pfn >= bb->start_pfn && pfn < bb->end_pfn) | |
496 | break; | |
74dfd666 | 497 | |
846705de RW |
498 | if (&bb->hook == &bm->blocks) |
499 | return -EFAULT; | |
500 | ||
501 | /* The block has been found */ | |
502 | bm->cur.block = bb; | |
b788db79 | 503 | pfn -= bb->start_pfn; |
846705de | 504 | bm->cur.bit = pfn + 1; |
0d83304c AM |
505 | *bit_nr = pfn; |
506 | *addr = bb->data; | |
a82f7119 | 507 | return 0; |
74dfd666 RW |
508 | } |
509 | ||
510 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) | |
511 | { | |
512 | void *addr; | |
513 | unsigned int bit; | |
a82f7119 | 514 | int error; |
74dfd666 | 515 | |
a82f7119 RW |
516 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
517 | BUG_ON(error); | |
74dfd666 RW |
518 | set_bit(bit, addr); |
519 | } | |
520 | ||
a82f7119 RW |
521 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
522 | { | |
523 | void *addr; | |
524 | unsigned int bit; | |
525 | int error; | |
526 | ||
527 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
528 | if (!error) | |
529 | set_bit(bit, addr); | |
530 | return error; | |
531 | } | |
532 | ||
74dfd666 RW |
533 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
534 | { | |
535 | void *addr; | |
536 | unsigned int bit; | |
a82f7119 | 537 | int error; |
74dfd666 | 538 | |
a82f7119 RW |
539 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
540 | BUG_ON(error); | |
74dfd666 RW |
541 | clear_bit(bit, addr); |
542 | } | |
543 | ||
544 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) | |
545 | { | |
546 | void *addr; | |
547 | unsigned int bit; | |
a82f7119 | 548 | int error; |
74dfd666 | 549 | |
a82f7119 RW |
550 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
551 | BUG_ON(error); | |
74dfd666 | 552 | return test_bit(bit, addr); |
b788db79 RW |
553 | } |
554 | ||
69643279 RW |
555 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
556 | { | |
557 | void *addr; | |
558 | unsigned int bit; | |
559 | ||
560 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); | |
561 | } | |
562 | ||
b788db79 RW |
563 | /** |
564 | * memory_bm_next_pfn - find the pfn that corresponds to the next set bit | |
565 | * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is | |
566 | * returned. | |
567 | * | |
568 | * It is required to run memory_bm_position_reset() before the first call to | |
569 | * this function. | |
570 | */ | |
571 | ||
572 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) | |
573 | { | |
b788db79 | 574 | struct bm_block *bb; |
b788db79 RW |
575 | int bit; |
576 | ||
846705de | 577 | bb = bm->cur.block; |
b788db79 | 578 | do { |
846705de RW |
579 | bit = bm->cur.bit; |
580 | bit = find_next_bit(bb->data, bm_block_bits(bb), bit); | |
581 | if (bit < bm_block_bits(bb)) | |
582 | goto Return_pfn; | |
583 | ||
584 | bb = list_entry(bb->hook.next, struct bm_block, hook); | |
585 | bm->cur.block = bb; | |
586 | bm->cur.bit = 0; | |
587 | } while (&bb->hook != &bm->blocks); | |
588 | ||
b788db79 RW |
589 | memory_bm_position_reset(bm); |
590 | return BM_END_OF_MAP; | |
591 | ||
59a49335 | 592 | Return_pfn: |
0d83304c AM |
593 | bm->cur.bit = bit + 1; |
594 | return bb->start_pfn + bit; | |
b788db79 RW |
595 | } |
596 | ||
74dfd666 RW |
597 | /** |
598 | * This structure represents a range of page frames the contents of which | |
599 | * should not be saved during the suspend. | |
600 | */ | |
601 | ||
602 | struct nosave_region { | |
603 | struct list_head list; | |
604 | unsigned long start_pfn; | |
605 | unsigned long end_pfn; | |
606 | }; | |
607 | ||
608 | static LIST_HEAD(nosave_regions); | |
609 | ||
610 | /** | |
611 | * register_nosave_region - register a range of page frames the contents | |
612 | * of which should not be saved during the suspend (to be used in the early | |
613 | * initialization code) | |
614 | */ | |
615 | ||
616 | void __init | |
940d67f6 JB |
617 | __register_nosave_region(unsigned long start_pfn, unsigned long end_pfn, |
618 | int use_kmalloc) | |
74dfd666 RW |
619 | { |
620 | struct nosave_region *region; | |
621 | ||
622 | if (start_pfn >= end_pfn) | |
623 | return; | |
624 | ||
625 | if (!list_empty(&nosave_regions)) { | |
626 | /* Try to extend the previous region (they should be sorted) */ | |
627 | region = list_entry(nosave_regions.prev, | |
628 | struct nosave_region, list); | |
629 | if (region->end_pfn == start_pfn) { | |
630 | region->end_pfn = end_pfn; | |
631 | goto Report; | |
632 | } | |
633 | } | |
940d67f6 JB |
634 | if (use_kmalloc) { |
635 | /* during init, this shouldn't fail */ | |
636 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); | |
637 | BUG_ON(!region); | |
638 | } else | |
639 | /* This allocation cannot fail */ | |
3c1596ef | 640 | region = alloc_bootmem(sizeof(struct nosave_region)); |
74dfd666 RW |
641 | region->start_pfn = start_pfn; |
642 | region->end_pfn = end_pfn; | |
643 | list_add_tail(®ion->list, &nosave_regions); | |
644 | Report: | |
cd38ca85 BH |
645 | printk(KERN_INFO "PM: Registered nosave memory: [mem %#010llx-%#010llx]\n", |
646 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
647 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
74dfd666 RW |
648 | } |
649 | ||
650 | /* | |
651 | * Set bits in this map correspond to the page frames the contents of which | |
652 | * should not be saved during the suspend. | |
653 | */ | |
654 | static struct memory_bitmap *forbidden_pages_map; | |
655 | ||
656 | /* Set bits in this map correspond to free page frames. */ | |
657 | static struct memory_bitmap *free_pages_map; | |
658 | ||
659 | /* | |
660 | * Each page frame allocated for creating the image is marked by setting the | |
661 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
662 | */ | |
663 | ||
664 | void swsusp_set_page_free(struct page *page) | |
665 | { | |
666 | if (free_pages_map) | |
667 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
668 | } | |
669 | ||
670 | static int swsusp_page_is_free(struct page *page) | |
671 | { | |
672 | return free_pages_map ? | |
673 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
674 | } | |
675 | ||
676 | void swsusp_unset_page_free(struct page *page) | |
677 | { | |
678 | if (free_pages_map) | |
679 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
680 | } | |
681 | ||
682 | static void swsusp_set_page_forbidden(struct page *page) | |
683 | { | |
684 | if (forbidden_pages_map) | |
685 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
686 | } | |
687 | ||
688 | int swsusp_page_is_forbidden(struct page *page) | |
689 | { | |
690 | return forbidden_pages_map ? | |
691 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
692 | } | |
693 | ||
694 | static void swsusp_unset_page_forbidden(struct page *page) | |
695 | { | |
696 | if (forbidden_pages_map) | |
697 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
698 | } | |
699 | ||
700 | /** | |
701 | * mark_nosave_pages - set bits corresponding to the page frames the | |
702 | * contents of which should not be saved in a given bitmap. | |
703 | */ | |
704 | ||
705 | static void mark_nosave_pages(struct memory_bitmap *bm) | |
706 | { | |
707 | struct nosave_region *region; | |
708 | ||
709 | if (list_empty(&nosave_regions)) | |
710 | return; | |
711 | ||
712 | list_for_each_entry(region, &nosave_regions, list) { | |
713 | unsigned long pfn; | |
714 | ||
69f1d475 BH |
715 | pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n", |
716 | (unsigned long long) region->start_pfn << PAGE_SHIFT, | |
717 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
718 | - 1); | |
74dfd666 RW |
719 | |
720 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
721 | if (pfn_valid(pfn)) { |
722 | /* | |
723 | * It is safe to ignore the result of | |
724 | * mem_bm_set_bit_check() here, since we won't | |
725 | * touch the PFNs for which the error is | |
726 | * returned anyway. | |
727 | */ | |
728 | mem_bm_set_bit_check(bm, pfn); | |
729 | } | |
74dfd666 RW |
730 | } |
731 | } | |
732 | ||
733 | /** | |
734 | * create_basic_memory_bitmaps - create bitmaps needed for marking page | |
735 | * frames that should not be saved and free page frames. The pointers | |
736 | * forbidden_pages_map and free_pages_map are only modified if everything | |
737 | * goes well, because we don't want the bits to be used before both bitmaps | |
738 | * are set up. | |
739 | */ | |
740 | ||
741 | int create_basic_memory_bitmaps(void) | |
742 | { | |
743 | struct memory_bitmap *bm1, *bm2; | |
744 | int error = 0; | |
745 | ||
aab17289 RW |
746 | if (forbidden_pages_map && free_pages_map) |
747 | return 0; | |
748 | else | |
749 | BUG_ON(forbidden_pages_map || free_pages_map); | |
74dfd666 | 750 | |
0709db60 | 751 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
752 | if (!bm1) |
753 | return -ENOMEM; | |
754 | ||
0709db60 | 755 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
756 | if (error) |
757 | goto Free_first_object; | |
758 | ||
0709db60 | 759 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
760 | if (!bm2) |
761 | goto Free_first_bitmap; | |
762 | ||
0709db60 | 763 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
764 | if (error) |
765 | goto Free_second_object; | |
766 | ||
767 | forbidden_pages_map = bm1; | |
768 | free_pages_map = bm2; | |
769 | mark_nosave_pages(forbidden_pages_map); | |
770 | ||
23976728 | 771 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
772 | |
773 | return 0; | |
774 | ||
775 | Free_second_object: | |
776 | kfree(bm2); | |
777 | Free_first_bitmap: | |
778 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
779 | Free_first_object: | |
780 | kfree(bm1); | |
781 | return -ENOMEM; | |
782 | } | |
783 | ||
784 | /** | |
785 | * free_basic_memory_bitmaps - free memory bitmaps allocated by | |
786 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary | |
787 | * so that the bitmaps themselves are not referred to while they are being | |
788 | * freed. | |
789 | */ | |
790 | ||
791 | void free_basic_memory_bitmaps(void) | |
792 | { | |
793 | struct memory_bitmap *bm1, *bm2; | |
794 | ||
795 | BUG_ON(!(forbidden_pages_map && free_pages_map)); | |
796 | ||
797 | bm1 = forbidden_pages_map; | |
798 | bm2 = free_pages_map; | |
799 | forbidden_pages_map = NULL; | |
800 | free_pages_map = NULL; | |
801 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
802 | kfree(bm1); | |
803 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
804 | kfree(bm2); | |
805 | ||
23976728 | 806 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
807 | } |
808 | ||
b788db79 RW |
809 | /** |
810 | * snapshot_additional_pages - estimate the number of additional pages | |
811 | * be needed for setting up the suspend image data structures for given | |
812 | * zone (usually the returned value is greater than the exact number) | |
813 | */ | |
814 | ||
815 | unsigned int snapshot_additional_pages(struct zone *zone) | |
816 | { | |
817 | unsigned int res; | |
818 | ||
819 | res = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); | |
160cb5a9 NK |
820 | res += DIV_ROUND_UP(res * sizeof(struct bm_block), |
821 | LINKED_PAGE_DATA_SIZE); | |
8357376d | 822 | return 2 * res; |
b788db79 RW |
823 | } |
824 | ||
8357376d RW |
825 | #ifdef CONFIG_HIGHMEM |
826 | /** | |
827 | * count_free_highmem_pages - compute the total number of free highmem | |
828 | * pages, system-wide. | |
829 | */ | |
830 | ||
831 | static unsigned int count_free_highmem_pages(void) | |
832 | { | |
833 | struct zone *zone; | |
834 | unsigned int cnt = 0; | |
835 | ||
ee99c71c KM |
836 | for_each_populated_zone(zone) |
837 | if (is_highmem(zone)) | |
d23ad423 | 838 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
839 | |
840 | return cnt; | |
841 | } | |
842 | ||
843 | /** | |
844 | * saveable_highmem_page - Determine whether a highmem page should be | |
845 | * included in the suspend image. | |
846 | * | |
847 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, | |
848 | * and it isn't a part of a free chunk of pages. | |
849 | */ | |
846705de | 850 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
851 | { |
852 | struct page *page; | |
853 | ||
854 | if (!pfn_valid(pfn)) | |
855 | return NULL; | |
856 | ||
857 | page = pfn_to_page(pfn); | |
846705de RW |
858 | if (page_zone(page) != zone) |
859 | return NULL; | |
8357376d RW |
860 | |
861 | BUG_ON(!PageHighMem(page)); | |
862 | ||
7be98234 RW |
863 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
864 | PageReserved(page)) | |
8357376d RW |
865 | return NULL; |
866 | ||
c6968e73 SG |
867 | if (page_is_guard(page)) |
868 | return NULL; | |
869 | ||
8357376d RW |
870 | return page; |
871 | } | |
872 | ||
873 | /** | |
874 | * count_highmem_pages - compute the total number of saveable highmem | |
875 | * pages. | |
876 | */ | |
877 | ||
fe419535 | 878 | static unsigned int count_highmem_pages(void) |
8357376d RW |
879 | { |
880 | struct zone *zone; | |
881 | unsigned int n = 0; | |
882 | ||
98e73dc5 | 883 | for_each_populated_zone(zone) { |
8357376d RW |
884 | unsigned long pfn, max_zone_pfn; |
885 | ||
886 | if (!is_highmem(zone)) | |
887 | continue; | |
888 | ||
889 | mark_free_pages(zone); | |
c33bc315 | 890 | max_zone_pfn = zone_end_pfn(zone); |
8357376d | 891 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 892 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
893 | n++; |
894 | } | |
895 | return n; | |
896 | } | |
897 | #else | |
846705de RW |
898 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
899 | { | |
900 | return NULL; | |
901 | } | |
8357376d RW |
902 | #endif /* CONFIG_HIGHMEM */ |
903 | ||
25761b6e | 904 | /** |
8a235efa RW |
905 | * saveable_page - Determine whether a non-highmem page should be included |
906 | * in the suspend image. | |
25761b6e | 907 | * |
8357376d RW |
908 | * We should save the page if it isn't Nosave, and is not in the range |
909 | * of pages statically defined as 'unsaveable', and it isn't a part of | |
910 | * a free chunk of pages. | |
25761b6e | 911 | */ |
846705de | 912 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 913 | { |
de491861 | 914 | struct page *page; |
25761b6e RW |
915 | |
916 | if (!pfn_valid(pfn)) | |
ae83c5ee | 917 | return NULL; |
25761b6e RW |
918 | |
919 | page = pfn_to_page(pfn); | |
846705de RW |
920 | if (page_zone(page) != zone) |
921 | return NULL; | |
ae83c5ee | 922 | |
8357376d RW |
923 | BUG_ON(PageHighMem(page)); |
924 | ||
7be98234 | 925 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 926 | return NULL; |
8357376d | 927 | |
8a235efa RW |
928 | if (PageReserved(page) |
929 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 930 | return NULL; |
25761b6e | 931 | |
c6968e73 SG |
932 | if (page_is_guard(page)) |
933 | return NULL; | |
934 | ||
ae83c5ee | 935 | return page; |
25761b6e RW |
936 | } |
937 | ||
8357376d RW |
938 | /** |
939 | * count_data_pages - compute the total number of saveable non-highmem | |
940 | * pages. | |
941 | */ | |
942 | ||
fe419535 | 943 | static unsigned int count_data_pages(void) |
25761b6e RW |
944 | { |
945 | struct zone *zone; | |
ae83c5ee | 946 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 947 | unsigned int n = 0; |
25761b6e | 948 | |
98e73dc5 | 949 | for_each_populated_zone(zone) { |
25761b6e RW |
950 | if (is_highmem(zone)) |
951 | continue; | |
8357376d | 952 | |
25761b6e | 953 | mark_free_pages(zone); |
c33bc315 | 954 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee | 955 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 956 | if (saveable_page(zone, pfn)) |
8357376d | 957 | n++; |
25761b6e | 958 | } |
a0f49651 | 959 | return n; |
25761b6e RW |
960 | } |
961 | ||
8357376d RW |
962 | /* This is needed, because copy_page and memcpy are not usable for copying |
963 | * task structs. | |
964 | */ | |
965 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
966 | { |
967 | int n; | |
968 | ||
f623f0db RW |
969 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
970 | *dst++ = *src++; | |
971 | } | |
972 | ||
8a235efa RW |
973 | |
974 | /** | |
975 | * safe_copy_page - check if the page we are going to copy is marked as | |
976 | * present in the kernel page tables (this always is the case if | |
977 | * CONFIG_DEBUG_PAGEALLOC is not set and in that case | |
978 | * kernel_page_present() always returns 'true'). | |
979 | */ | |
980 | static void safe_copy_page(void *dst, struct page *s_page) | |
981 | { | |
982 | if (kernel_page_present(s_page)) { | |
983 | do_copy_page(dst, page_address(s_page)); | |
984 | } else { | |
985 | kernel_map_pages(s_page, 1, 1); | |
986 | do_copy_page(dst, page_address(s_page)); | |
987 | kernel_map_pages(s_page, 1, 0); | |
988 | } | |
989 | } | |
990 | ||
991 | ||
8357376d RW |
992 | #ifdef CONFIG_HIGHMEM |
993 | static inline struct page * | |
994 | page_is_saveable(struct zone *zone, unsigned long pfn) | |
995 | { | |
996 | return is_highmem(zone) ? | |
846705de | 997 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
998 | } |
999 | ||
8a235efa | 1000 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
1001 | { |
1002 | struct page *s_page, *d_page; | |
1003 | void *src, *dst; | |
1004 | ||
1005 | s_page = pfn_to_page(src_pfn); | |
1006 | d_page = pfn_to_page(dst_pfn); | |
1007 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1008 | src = kmap_atomic(s_page); |
1009 | dst = kmap_atomic(d_page); | |
8357376d | 1010 | do_copy_page(dst, src); |
0de9a1e2 CW |
1011 | kunmap_atomic(dst); |
1012 | kunmap_atomic(src); | |
8357376d | 1013 | } else { |
8357376d RW |
1014 | if (PageHighMem(d_page)) { |
1015 | /* Page pointed to by src may contain some kernel | |
1016 | * data modified by kmap_atomic() | |
1017 | */ | |
8a235efa | 1018 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1019 | dst = kmap_atomic(d_page); |
3ecb01df | 1020 | copy_page(dst, buffer); |
0de9a1e2 | 1021 | kunmap_atomic(dst); |
8357376d | 1022 | } else { |
8a235efa | 1023 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1024 | } |
1025 | } | |
1026 | } | |
1027 | #else | |
846705de | 1028 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1029 | |
8a235efa | 1030 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1031 | { |
8a235efa RW |
1032 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1033 | pfn_to_page(src_pfn)); | |
8357376d RW |
1034 | } |
1035 | #endif /* CONFIG_HIGHMEM */ | |
1036 | ||
b788db79 RW |
1037 | static void |
1038 | copy_data_pages(struct memory_bitmap *copy_bm, struct memory_bitmap *orig_bm) | |
25761b6e RW |
1039 | { |
1040 | struct zone *zone; | |
b788db79 | 1041 | unsigned long pfn; |
25761b6e | 1042 | |
98e73dc5 | 1043 | for_each_populated_zone(zone) { |
b788db79 RW |
1044 | unsigned long max_zone_pfn; |
1045 | ||
25761b6e | 1046 | mark_free_pages(zone); |
c33bc315 | 1047 | max_zone_pfn = zone_end_pfn(zone); |
b788db79 | 1048 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1049 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1050 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1051 | } |
b788db79 RW |
1052 | memory_bm_position_reset(orig_bm); |
1053 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1054 | for(;;) { |
b788db79 | 1055 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1056 | if (unlikely(pfn == BM_END_OF_MAP)) |
1057 | break; | |
1058 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1059 | } | |
25761b6e RW |
1060 | } |
1061 | ||
8357376d RW |
1062 | /* Total number of image pages */ |
1063 | static unsigned int nr_copy_pages; | |
1064 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1065 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1066 | /* |
1067 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1068 | * before suspending devices. | |
1069 | */ | |
1070 | unsigned int alloc_normal, alloc_highmem; | |
1071 | /* | |
1072 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1073 | * hibernation image pages (during restore) | |
1074 | */ | |
1075 | static struct memory_bitmap orig_bm; | |
1076 | /* | |
1077 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1078 | * will contain copies of saveable pages. During restore it is initially used | |
1079 | * for marking hibernation image pages, but then the set bits from it are | |
1080 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1081 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1082 | * this purpose. | |
1083 | */ | |
1084 | static struct memory_bitmap copy_bm; | |
8357376d | 1085 | |
25761b6e | 1086 | /** |
940864dd | 1087 | * swsusp_free - free pages allocated for the suspend. |
cd560bb2 | 1088 | * |
940864dd RW |
1089 | * Suspend pages are alocated before the atomic copy is made, so we |
1090 | * need to release them after the resume. | |
25761b6e RW |
1091 | */ |
1092 | ||
1093 | void swsusp_free(void) | |
1094 | { | |
1095 | struct zone *zone; | |
ae83c5ee | 1096 | unsigned long pfn, max_zone_pfn; |
25761b6e | 1097 | |
98e73dc5 | 1098 | for_each_populated_zone(zone) { |
c33bc315 | 1099 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee RW |
1100 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
1101 | if (pfn_valid(pfn)) { | |
1102 | struct page *page = pfn_to_page(pfn); | |
1103 | ||
7be98234 RW |
1104 | if (swsusp_page_is_forbidden(page) && |
1105 | swsusp_page_is_free(page)) { | |
1106 | swsusp_unset_page_forbidden(page); | |
1107 | swsusp_unset_page_free(page); | |
8357376d | 1108 | __free_page(page); |
25761b6e RW |
1109 | } |
1110 | } | |
1111 | } | |
f577eb30 RW |
1112 | nr_copy_pages = 0; |
1113 | nr_meta_pages = 0; | |
75534b50 | 1114 | restore_pblist = NULL; |
6e1819d6 | 1115 | buffer = NULL; |
64a473cb RW |
1116 | alloc_normal = 0; |
1117 | alloc_highmem = 0; | |
25761b6e RW |
1118 | } |
1119 | ||
4bb33435 RW |
1120 | /* Helper functions used for the shrinking of memory. */ |
1121 | ||
1122 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1123 | ||
fe419535 | 1124 | /** |
4bb33435 RW |
1125 | * preallocate_image_pages - Allocate a number of pages for hibernation image |
1126 | * @nr_pages: Number of page frames to allocate. | |
1127 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1128 | * |
4bb33435 RW |
1129 | * Return value: Number of page frames actually allocated |
1130 | */ | |
1131 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1132 | { | |
1133 | unsigned long nr_alloc = 0; | |
1134 | ||
1135 | while (nr_pages > 0) { | |
64a473cb RW |
1136 | struct page *page; |
1137 | ||
1138 | page = alloc_image_page(mask); | |
1139 | if (!page) | |
4bb33435 | 1140 | break; |
64a473cb RW |
1141 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1142 | if (PageHighMem(page)) | |
1143 | alloc_highmem++; | |
1144 | else | |
1145 | alloc_normal++; | |
4bb33435 RW |
1146 | nr_pages--; |
1147 | nr_alloc++; | |
1148 | } | |
1149 | ||
1150 | return nr_alloc; | |
1151 | } | |
1152 | ||
6715045d RW |
1153 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1154 | unsigned long avail_normal) | |
4bb33435 | 1155 | { |
6715045d RW |
1156 | unsigned long alloc; |
1157 | ||
1158 | if (avail_normal <= alloc_normal) | |
1159 | return 0; | |
1160 | ||
1161 | alloc = avail_normal - alloc_normal; | |
1162 | if (nr_pages < alloc) | |
1163 | alloc = nr_pages; | |
1164 | ||
1165 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1166 | } |
1167 | ||
1168 | #ifdef CONFIG_HIGHMEM | |
1169 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1170 | { | |
1171 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1172 | } | |
1173 | ||
1174 | /** | |
1175 | * __fraction - Compute (an approximation of) x * (multiplier / base) | |
fe419535 | 1176 | */ |
4bb33435 RW |
1177 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1178 | { | |
1179 | x *= multiplier; | |
1180 | do_div(x, base); | |
1181 | return (unsigned long)x; | |
1182 | } | |
fe419535 | 1183 | |
4bb33435 RW |
1184 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
1185 | unsigned long highmem, | |
1186 | unsigned long total) | |
fe419535 | 1187 | { |
4bb33435 RW |
1188 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1189 | ||
1190 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1191 | } |
4bb33435 RW |
1192 | #else /* CONFIG_HIGHMEM */ |
1193 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1194 | { | |
1195 | return 0; | |
1196 | } | |
1197 | ||
1198 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
1199 | unsigned long highmem, | |
1200 | unsigned long total) | |
1201 | { | |
1202 | return 0; | |
1203 | } | |
1204 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1205 | |
4bb33435 | 1206 | /** |
64a473cb RW |
1207 | * free_unnecessary_pages - Release preallocated pages not needed for the image |
1208 | */ | |
1209 | static void free_unnecessary_pages(void) | |
1210 | { | |
6715045d | 1211 | unsigned long save, to_free_normal, to_free_highmem; |
64a473cb | 1212 | |
6715045d RW |
1213 | save = count_data_pages(); |
1214 | if (alloc_normal >= save) { | |
1215 | to_free_normal = alloc_normal - save; | |
1216 | save = 0; | |
1217 | } else { | |
1218 | to_free_normal = 0; | |
1219 | save -= alloc_normal; | |
1220 | } | |
1221 | save += count_highmem_pages(); | |
1222 | if (alloc_highmem >= save) { | |
1223 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1224 | } else { |
1225 | to_free_highmem = 0; | |
4d4cf23c RW |
1226 | save -= alloc_highmem; |
1227 | if (to_free_normal > save) | |
1228 | to_free_normal -= save; | |
1229 | else | |
1230 | to_free_normal = 0; | |
64a473cb RW |
1231 | } |
1232 | ||
1233 | memory_bm_position_reset(©_bm); | |
1234 | ||
a9c9b442 | 1235 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1236 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1237 | struct page *page = pfn_to_page(pfn); | |
1238 | ||
1239 | if (PageHighMem(page)) { | |
1240 | if (!to_free_highmem) | |
1241 | continue; | |
1242 | to_free_highmem--; | |
1243 | alloc_highmem--; | |
1244 | } else { | |
1245 | if (!to_free_normal) | |
1246 | continue; | |
1247 | to_free_normal--; | |
1248 | alloc_normal--; | |
1249 | } | |
1250 | memory_bm_clear_bit(©_bm, pfn); | |
1251 | swsusp_unset_page_forbidden(page); | |
1252 | swsusp_unset_page_free(page); | |
1253 | __free_page(page); | |
1254 | } | |
1255 | } | |
1256 | ||
ef4aede3 RW |
1257 | /** |
1258 | * minimum_image_size - Estimate the minimum acceptable size of an image | |
1259 | * @saveable: Number of saveable pages in the system. | |
1260 | * | |
1261 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1262 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1263 | * preallocating memory. | |
1264 | * | |
1265 | * We assume that the minimum image size should be proportional to | |
1266 | * | |
1267 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1268 | * | |
1269 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
1270 | * and (3) inactive anonymouns pages, (4) active and (5) inactive file pages, | |
1271 | * minus mapped file pages. | |
1272 | */ | |
1273 | static unsigned long minimum_image_size(unsigned long saveable) | |
1274 | { | |
1275 | unsigned long size; | |
1276 | ||
1277 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
1278 | + global_page_state(NR_ACTIVE_ANON) | |
1279 | + global_page_state(NR_INACTIVE_ANON) | |
1280 | + global_page_state(NR_ACTIVE_FILE) | |
1281 | + global_page_state(NR_INACTIVE_FILE) | |
1282 | - global_page_state(NR_FILE_MAPPED); | |
1283 | ||
1284 | return saveable <= size ? 0 : saveable - size; | |
1285 | } | |
1286 | ||
64a473cb RW |
1287 | /** |
1288 | * hibernate_preallocate_memory - Preallocate memory for hibernation image | |
4bb33435 RW |
1289 | * |
1290 | * To create a hibernation image it is necessary to make a copy of every page | |
1291 | * frame in use. We also need a number of page frames to be free during | |
1292 | * hibernation for allocations made while saving the image and for device | |
1293 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 RW |
1294 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
1295 | * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through | |
1296 | * /sys/power/reserved_size, respectively). To make this happen, we compute the | |
1297 | * total number of available page frames and allocate at least | |
4bb33435 | 1298 | * |
ddeb6487 RW |
1299 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1300 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1301 | * |
1302 | * of them, which corresponds to the maximum size of a hibernation image. | |
1303 | * | |
1304 | * If image_size is set below the number following from the above formula, | |
1305 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1306 | * pages in the system is below the requested image size or the minimum |
1307 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1308 | */ |
64a473cb | 1309 | int hibernate_preallocate_memory(void) |
fe419535 | 1310 | { |
fe419535 | 1311 | struct zone *zone; |
4bb33435 | 1312 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1313 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
fe419535 | 1314 | struct timeval start, stop; |
64a473cb | 1315 | int error; |
fe419535 | 1316 | |
64a473cb | 1317 | printk(KERN_INFO "PM: Preallocating image memory... "); |
fe419535 | 1318 | do_gettimeofday(&start); |
fe419535 | 1319 | |
64a473cb RW |
1320 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1321 | if (error) | |
1322 | goto err_out; | |
1323 | ||
1324 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1325 | if (error) | |
1326 | goto err_out; | |
1327 | ||
1328 | alloc_normal = 0; | |
1329 | alloc_highmem = 0; | |
1330 | ||
4bb33435 | 1331 | /* Count the number of saveable data pages. */ |
64a473cb | 1332 | save_highmem = count_highmem_pages(); |
4bb33435 | 1333 | saveable = count_data_pages(); |
fe419535 | 1334 | |
4bb33435 RW |
1335 | /* |
1336 | * Compute the total number of page frames we can use (count) and the | |
1337 | * number of pages needed for image metadata (size). | |
1338 | */ | |
1339 | count = saveable; | |
64a473cb RW |
1340 | saveable += save_highmem; |
1341 | highmem = save_highmem; | |
4bb33435 RW |
1342 | size = 0; |
1343 | for_each_populated_zone(zone) { | |
1344 | size += snapshot_additional_pages(zone); | |
1345 | if (is_highmem(zone)) | |
1346 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1347 | else | |
1348 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1349 | } | |
6715045d | 1350 | avail_normal = count; |
4bb33435 RW |
1351 | count += highmem; |
1352 | count -= totalreserve_pages; | |
1353 | ||
85055dd8 MS |
1354 | /* Add number of pages required for page keys (s390 only). */ |
1355 | size += page_key_additional_pages(saveable); | |
1356 | ||
4bb33435 | 1357 | /* Compute the maximum number of saveable pages to leave in memory. */ |
ddeb6487 RW |
1358 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1359 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1360 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1361 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1362 | if (size > max_size) | |
1363 | size = max_size; | |
1364 | /* | |
266f1a25 RW |
1365 | * If the desired number of image pages is at least as large as the |
1366 | * current number of saveable pages in memory, allocate page frames for | |
1367 | * the image and we're done. | |
4bb33435 | 1368 | */ |
64a473cb RW |
1369 | if (size >= saveable) { |
1370 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1371 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1372 | goto out; |
64a473cb | 1373 | } |
4bb33435 | 1374 | |
ef4aede3 RW |
1375 | /* Estimate the minimum size of the image. */ |
1376 | pages = minimum_image_size(saveable); | |
6715045d RW |
1377 | /* |
1378 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1379 | * accommodate an image of the minimum size (unless it's already too | |
1380 | * small, in which case don't preallocate pages from it at all). | |
1381 | */ | |
1382 | if (avail_normal > pages) | |
1383 | avail_normal -= pages; | |
1384 | else | |
1385 | avail_normal = 0; | |
ef4aede3 RW |
1386 | if (size < pages) |
1387 | size = min_t(unsigned long, pages, max_size); | |
1388 | ||
4bb33435 RW |
1389 | /* |
1390 | * Let the memory management subsystem know that we're going to need a | |
1391 | * large number of page frames to allocate and make it free some memory. | |
1392 | * NOTE: If this is not done, performance will be hurt badly in some | |
1393 | * test cases. | |
1394 | */ | |
1395 | shrink_all_memory(saveable - size); | |
1396 | ||
1397 | /* | |
1398 | * The number of saveable pages in memory was too high, so apply some | |
1399 | * pressure to decrease it. First, make room for the largest possible | |
1400 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1401 | * of the image as much as indicated by 'size' using allocations from |
1402 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1403 | */ |
1404 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
1405 | alloc = (count - max_size) - pages_highmem; | |
6715045d RW |
1406 | pages = preallocate_image_memory(alloc, avail_normal); |
1407 | if (pages < alloc) { | |
1408 | /* We have exhausted non-highmem pages, try highmem. */ | |
1409 | alloc -= pages; | |
1410 | pages += pages_highmem; | |
1411 | pages_highmem = preallocate_image_highmem(alloc); | |
1412 | if (pages_highmem < alloc) | |
1413 | goto err_out; | |
1414 | pages += pages_highmem; | |
1415 | /* | |
1416 | * size is the desired number of saveable pages to leave in | |
1417 | * memory, so try to preallocate (all memory - size) pages. | |
1418 | */ | |
1419 | alloc = (count - pages) - size; | |
1420 | pages += preallocate_image_highmem(alloc); | |
1421 | } else { | |
1422 | /* | |
1423 | * There are approximately max_size saveable pages at this point | |
1424 | * and we want to reduce this number down to size. | |
1425 | */ | |
1426 | alloc = max_size - size; | |
1427 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1428 | pages_highmem += size; | |
1429 | alloc -= size; | |
1430 | size = preallocate_image_memory(alloc, avail_normal); | |
1431 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1432 | pages += pages_highmem + size; | |
1433 | } | |
4bb33435 | 1434 | |
64a473cb RW |
1435 | /* |
1436 | * We only need as many page frames for the image as there are saveable | |
1437 | * pages in memory, but we have allocated more. Release the excessive | |
1438 | * ones now. | |
1439 | */ | |
1440 | free_unnecessary_pages(); | |
4bb33435 RW |
1441 | |
1442 | out: | |
fe419535 | 1443 | do_gettimeofday(&stop); |
64a473cb RW |
1444 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
1445 | swsusp_show_speed(&start, &stop, pages, "Allocated"); | |
fe419535 RW |
1446 | |
1447 | return 0; | |
64a473cb RW |
1448 | |
1449 | err_out: | |
1450 | printk(KERN_CONT "\n"); | |
1451 | swsusp_free(); | |
1452 | return -ENOMEM; | |
fe419535 RW |
1453 | } |
1454 | ||
8357376d RW |
1455 | #ifdef CONFIG_HIGHMEM |
1456 | /** | |
1457 | * count_pages_for_highmem - compute the number of non-highmem pages | |
1458 | * that will be necessary for creating copies of highmem pages. | |
1459 | */ | |
1460 | ||
1461 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) | |
1462 | { | |
64a473cb | 1463 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1464 | |
1465 | if (free_highmem >= nr_highmem) | |
1466 | nr_highmem = 0; | |
1467 | else | |
1468 | nr_highmem -= free_highmem; | |
1469 | ||
1470 | return nr_highmem; | |
1471 | } | |
1472 | #else | |
1473 | static unsigned int | |
1474 | count_pages_for_highmem(unsigned int nr_highmem) { return 0; } | |
1475 | #endif /* CONFIG_HIGHMEM */ | |
25761b6e RW |
1476 | |
1477 | /** | |
8357376d RW |
1478 | * enough_free_mem - Make sure we have enough free memory for the |
1479 | * snapshot image. | |
25761b6e RW |
1480 | */ |
1481 | ||
8357376d | 1482 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1483 | { |
e5e2fa78 | 1484 | struct zone *zone; |
64a473cb | 1485 | unsigned int free = alloc_normal; |
e5e2fa78 | 1486 | |
98e73dc5 | 1487 | for_each_populated_zone(zone) |
8357376d | 1488 | if (!is_highmem(zone)) |
d23ad423 | 1489 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1490 | |
8357376d | 1491 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1492 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1493 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1494 | |
64a473cb | 1495 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1496 | } |
1497 | ||
8357376d RW |
1498 | #ifdef CONFIG_HIGHMEM |
1499 | /** | |
1500 | * get_highmem_buffer - if there are some highmem pages in the suspend | |
1501 | * image, we may need the buffer to copy them and/or load their data. | |
1502 | */ | |
1503 | ||
1504 | static inline int get_highmem_buffer(int safe_needed) | |
1505 | { | |
1506 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1507 | return buffer ? 0 : -ENOMEM; | |
1508 | } | |
1509 | ||
1510 | /** | |
1511 | * alloc_highmem_image_pages - allocate some highmem pages for the image. | |
1512 | * Try to allocate as many pages as needed, but if the number of free | |
1513 | * highmem pages is lesser than that, allocate them all. | |
1514 | */ | |
1515 | ||
1516 | static inline unsigned int | |
64a473cb | 1517 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int nr_highmem) |
8357376d RW |
1518 | { |
1519 | unsigned int to_alloc = count_free_highmem_pages(); | |
1520 | ||
1521 | if (to_alloc > nr_highmem) | |
1522 | to_alloc = nr_highmem; | |
1523 | ||
1524 | nr_highmem -= to_alloc; | |
1525 | while (to_alloc-- > 0) { | |
1526 | struct page *page; | |
1527 | ||
1528 | page = alloc_image_page(__GFP_HIGHMEM); | |
1529 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1530 | } | |
1531 | return nr_highmem; | |
1532 | } | |
1533 | #else | |
1534 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1535 | ||
1536 | static inline unsigned int | |
64a473cb | 1537 | alloc_highmem_pages(struct memory_bitmap *bm, unsigned int n) { return 0; } |
8357376d RW |
1538 | #endif /* CONFIG_HIGHMEM */ |
1539 | ||
1540 | /** | |
1541 | * swsusp_alloc - allocate memory for the suspend image | |
1542 | * | |
1543 | * We first try to allocate as many highmem pages as there are | |
1544 | * saveable highmem pages in the system. If that fails, we allocate | |
1545 | * non-highmem pages for the copies of the remaining highmem ones. | |
1546 | * | |
1547 | * In this approach it is likely that the copies of highmem pages will | |
1548 | * also be located in the high memory, because of the way in which | |
1549 | * copy_data_pages() works. | |
1550 | */ | |
1551 | ||
b788db79 RW |
1552 | static int |
1553 | swsusp_alloc(struct memory_bitmap *orig_bm, struct memory_bitmap *copy_bm, | |
8357376d | 1554 | unsigned int nr_pages, unsigned int nr_highmem) |
054bd4c1 | 1555 | { |
8357376d | 1556 | if (nr_highmem > 0) { |
2e725a06 | 1557 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1558 | goto err_out; |
1559 | if (nr_highmem > alloc_highmem) { | |
1560 | nr_highmem -= alloc_highmem; | |
1561 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1562 | } | |
8357376d | 1563 | } |
64a473cb RW |
1564 | if (nr_pages > alloc_normal) { |
1565 | nr_pages -= alloc_normal; | |
1566 | while (nr_pages-- > 0) { | |
1567 | struct page *page; | |
1568 | ||
1569 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1570 | if (!page) | |
1571 | goto err_out; | |
1572 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1573 | } | |
25761b6e | 1574 | } |
64a473cb | 1575 | |
b788db79 | 1576 | return 0; |
25761b6e | 1577 | |
64a473cb | 1578 | err_out: |
b788db79 | 1579 | swsusp_free(); |
2e725a06 | 1580 | return -ENOMEM; |
25761b6e RW |
1581 | } |
1582 | ||
2e32a43e | 1583 | asmlinkage int swsusp_save(void) |
25761b6e | 1584 | { |
8357376d | 1585 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1586 | |
07c3bb57 | 1587 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1588 | |
9f8f2172 | 1589 | drain_local_pages(NULL); |
a0f49651 | 1590 | nr_pages = count_data_pages(); |
8357376d | 1591 | nr_highmem = count_highmem_pages(); |
23976728 | 1592 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1593 | |
8357376d | 1594 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1595 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1596 | return -ENOMEM; |
1597 | } | |
1598 | ||
8357376d | 1599 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1600 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1601 | return -ENOMEM; |
8357376d | 1602 | } |
25761b6e RW |
1603 | |
1604 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1605 | * Kill them. | |
1606 | */ | |
9f8f2172 | 1607 | drain_local_pages(NULL); |
b788db79 | 1608 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1609 | |
1610 | /* | |
1611 | * End of critical section. From now on, we can write to memory, | |
1612 | * but we should not touch disk. This specially means we must _not_ | |
1613 | * touch swap space! Except we must write out our image of course. | |
1614 | */ | |
1615 | ||
8357376d | 1616 | nr_pages += nr_highmem; |
a0f49651 | 1617 | nr_copy_pages = nr_pages; |
8357376d | 1618 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 1619 | |
23976728 RW |
1620 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1621 | nr_pages); | |
8357376d | 1622 | |
25761b6e RW |
1623 | return 0; |
1624 | } | |
f577eb30 | 1625 | |
d307c4a8 RW |
1626 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1627 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 1628 | { |
d307c4a8 | 1629 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 1630 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
1631 | return 0; |
1632 | } | |
1633 | ||
1634 | static char *check_image_kernel(struct swsusp_info *info) | |
1635 | { | |
1636 | if (info->version_code != LINUX_VERSION_CODE) | |
1637 | return "kernel version"; | |
1638 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
1639 | return "system type"; | |
1640 | if (strcmp(info->uts.release,init_utsname()->release)) | |
1641 | return "kernel release"; | |
1642 | if (strcmp(info->uts.version,init_utsname()->version)) | |
1643 | return "version"; | |
1644 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
1645 | return "machine"; | |
1646 | return NULL; | |
1647 | } | |
1648 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
1649 | ||
af508b34 RW |
1650 | unsigned long snapshot_get_image_size(void) |
1651 | { | |
1652 | return nr_copy_pages + nr_meta_pages + 1; | |
1653 | } | |
1654 | ||
d307c4a8 RW |
1655 | static int init_header(struct swsusp_info *info) |
1656 | { | |
1657 | memset(info, 0, sizeof(struct swsusp_info)); | |
0ed5fd13 | 1658 | info->num_physpages = get_num_physpages(); |
f577eb30 | 1659 | info->image_pages = nr_copy_pages; |
af508b34 | 1660 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
1661 | info->size = info->pages; |
1662 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 1663 | return init_header_complete(info); |
f577eb30 RW |
1664 | } |
1665 | ||
1666 | /** | |
940864dd RW |
1667 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1668 | * are stored in the array @buf[] (1 page at a time) | |
f577eb30 RW |
1669 | */ |
1670 | ||
b788db79 | 1671 | static inline void |
940864dd | 1672 | pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1673 | { |
1674 | int j; | |
1675 | ||
b788db79 | 1676 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
1677 | buf[j] = memory_bm_next_pfn(bm); |
1678 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 1679 | break; |
85055dd8 MS |
1680 | /* Save page key for data page (s390 only). */ |
1681 | page_key_read(buf + j); | |
f577eb30 | 1682 | } |
f577eb30 RW |
1683 | } |
1684 | ||
1685 | /** | |
1686 | * snapshot_read_next - used for reading the system memory snapshot. | |
1687 | * | |
1688 | * On the first call to it @handle should point to a zeroed | |
1689 | * snapshot_handle structure. The structure gets updated and a pointer | |
1690 | * to it should be passed to this function every next time. | |
1691 | * | |
f577eb30 RW |
1692 | * On success the function returns a positive number. Then, the caller |
1693 | * is allowed to read up to the returned number of bytes from the memory | |
d3c1b24c | 1694 | * location computed by the data_of() macro. |
f577eb30 RW |
1695 | * |
1696 | * The function returns 0 to indicate the end of data stream condition, | |
1697 | * and a negative number is returned on error. In such cases the | |
1698 | * structure pointed to by @handle is not updated and should not be used | |
1699 | * any more. | |
1700 | */ | |
1701 | ||
d3c1b24c | 1702 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 1703 | { |
fb13a28b | 1704 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 1705 | return 0; |
b788db79 | 1706 | |
f577eb30 RW |
1707 | if (!buffer) { |
1708 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 1709 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
1710 | if (!buffer) |
1711 | return -ENOMEM; | |
1712 | } | |
d3c1b24c | 1713 | if (!handle->cur) { |
d307c4a8 RW |
1714 | int error; |
1715 | ||
1716 | error = init_header((struct swsusp_info *)buffer); | |
1717 | if (error) | |
1718 | return error; | |
f577eb30 | 1719 | handle->buffer = buffer; |
b788db79 RW |
1720 | memory_bm_position_reset(&orig_bm); |
1721 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 1722 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 1723 | clear_page(buffer); |
d3c1b24c JS |
1724 | pack_pfns(buffer, &orig_bm); |
1725 | } else { | |
1726 | struct page *page; | |
b788db79 | 1727 | |
d3c1b24c JS |
1728 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
1729 | if (PageHighMem(page)) { | |
1730 | /* Highmem pages are copied to the buffer, | |
1731 | * because we can't return with a kmapped | |
1732 | * highmem page (we may not be called again). | |
1733 | */ | |
1734 | void *kaddr; | |
8357376d | 1735 | |
0de9a1e2 | 1736 | kaddr = kmap_atomic(page); |
3ecb01df | 1737 | copy_page(buffer, kaddr); |
0de9a1e2 | 1738 | kunmap_atomic(kaddr); |
d3c1b24c JS |
1739 | handle->buffer = buffer; |
1740 | } else { | |
1741 | handle->buffer = page_address(page); | |
f577eb30 | 1742 | } |
f577eb30 | 1743 | } |
d3c1b24c JS |
1744 | handle->cur++; |
1745 | return PAGE_SIZE; | |
f577eb30 RW |
1746 | } |
1747 | ||
1748 | /** | |
1749 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
1750 | * the image during resume, because they conflict with the pages that | |
1751 | * had been used before suspend | |
1752 | */ | |
1753 | ||
940864dd | 1754 | static int mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 RW |
1755 | { |
1756 | struct zone *zone; | |
ae83c5ee | 1757 | unsigned long pfn, max_zone_pfn; |
f577eb30 RW |
1758 | |
1759 | /* Clear page flags */ | |
98e73dc5 | 1760 | for_each_populated_zone(zone) { |
c33bc315 | 1761 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee RW |
1762 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
1763 | if (pfn_valid(pfn)) | |
7be98234 | 1764 | swsusp_unset_page_free(pfn_to_page(pfn)); |
f577eb30 RW |
1765 | } |
1766 | ||
940864dd RW |
1767 | /* Mark pages that correspond to the "original" pfns as "unsafe" */ |
1768 | memory_bm_position_reset(bm); | |
1769 | do { | |
1770 | pfn = memory_bm_next_pfn(bm); | |
1771 | if (likely(pfn != BM_END_OF_MAP)) { | |
1772 | if (likely(pfn_valid(pfn))) | |
7be98234 | 1773 | swsusp_set_page_free(pfn_to_page(pfn)); |
940864dd RW |
1774 | else |
1775 | return -EFAULT; | |
1776 | } | |
1777 | } while (pfn != BM_END_OF_MAP); | |
f577eb30 | 1778 | |
940864dd | 1779 | allocated_unsafe_pages = 0; |
968808b8 | 1780 | |
f577eb30 RW |
1781 | return 0; |
1782 | } | |
1783 | ||
940864dd RW |
1784 | static void |
1785 | duplicate_memory_bitmap(struct memory_bitmap *dst, struct memory_bitmap *src) | |
f577eb30 | 1786 | { |
940864dd RW |
1787 | unsigned long pfn; |
1788 | ||
1789 | memory_bm_position_reset(src); | |
1790 | pfn = memory_bm_next_pfn(src); | |
1791 | while (pfn != BM_END_OF_MAP) { | |
1792 | memory_bm_set_bit(dst, pfn); | |
1793 | pfn = memory_bm_next_pfn(src); | |
f577eb30 RW |
1794 | } |
1795 | } | |
1796 | ||
d307c4a8 | 1797 | static int check_header(struct swsusp_info *info) |
f577eb30 | 1798 | { |
d307c4a8 | 1799 | char *reason; |
f577eb30 | 1800 | |
d307c4a8 | 1801 | reason = check_image_kernel(info); |
0ed5fd13 | 1802 | if (!reason && info->num_physpages != get_num_physpages()) |
f577eb30 | 1803 | reason = "memory size"; |
f577eb30 | 1804 | if (reason) { |
23976728 | 1805 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
1806 | return -EPERM; |
1807 | } | |
1808 | return 0; | |
1809 | } | |
1810 | ||
1811 | /** | |
1812 | * load header - check the image header and copy data from it | |
1813 | */ | |
1814 | ||
940864dd RW |
1815 | static int |
1816 | load_header(struct swsusp_info *info) | |
f577eb30 RW |
1817 | { |
1818 | int error; | |
f577eb30 | 1819 | |
940864dd | 1820 | restore_pblist = NULL; |
f577eb30 RW |
1821 | error = check_header(info); |
1822 | if (!error) { | |
f577eb30 RW |
1823 | nr_copy_pages = info->image_pages; |
1824 | nr_meta_pages = info->pages - info->image_pages - 1; | |
1825 | } | |
1826 | return error; | |
1827 | } | |
1828 | ||
1829 | /** | |
940864dd RW |
1830 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
1831 | * the corresponding bit in the memory bitmap @bm | |
f577eb30 | 1832 | */ |
69643279 | 1833 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1834 | { |
1835 | int j; | |
1836 | ||
940864dd RW |
1837 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
1838 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
1839 | break; | |
1840 | ||
85055dd8 MS |
1841 | /* Extract and buffer page key for data page (s390 only). */ |
1842 | page_key_memorize(buf + j); | |
1843 | ||
69643279 RW |
1844 | if (memory_bm_pfn_present(bm, buf[j])) |
1845 | memory_bm_set_bit(bm, buf[j]); | |
1846 | else | |
1847 | return -EFAULT; | |
f577eb30 | 1848 | } |
69643279 RW |
1849 | |
1850 | return 0; | |
f577eb30 RW |
1851 | } |
1852 | ||
8357376d RW |
1853 | /* List of "safe" pages that may be used to store data loaded from the suspend |
1854 | * image | |
1855 | */ | |
1856 | static struct linked_page *safe_pages_list; | |
1857 | ||
1858 | #ifdef CONFIG_HIGHMEM | |
1859 | /* struct highmem_pbe is used for creating the list of highmem pages that | |
1860 | * should be restored atomically during the resume from disk, because the page | |
1861 | * frames they have occupied before the suspend are in use. | |
1862 | */ | |
1863 | struct highmem_pbe { | |
1864 | struct page *copy_page; /* data is here now */ | |
1865 | struct page *orig_page; /* data was here before the suspend */ | |
1866 | struct highmem_pbe *next; | |
1867 | }; | |
1868 | ||
1869 | /* List of highmem PBEs needed for restoring the highmem pages that were | |
1870 | * allocated before the suspend and included in the suspend image, but have | |
1871 | * also been allocated by the "resume" kernel, so their contents cannot be | |
1872 | * written directly to their "original" page frames. | |
1873 | */ | |
1874 | static struct highmem_pbe *highmem_pblist; | |
1875 | ||
1876 | /** | |
1877 | * count_highmem_image_pages - compute the number of highmem pages in the | |
1878 | * suspend image. The bits in the memory bitmap @bm that correspond to the | |
1879 | * image pages are assumed to be set. | |
1880 | */ | |
1881 | ||
1882 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) | |
1883 | { | |
1884 | unsigned long pfn; | |
1885 | unsigned int cnt = 0; | |
1886 | ||
1887 | memory_bm_position_reset(bm); | |
1888 | pfn = memory_bm_next_pfn(bm); | |
1889 | while (pfn != BM_END_OF_MAP) { | |
1890 | if (PageHighMem(pfn_to_page(pfn))) | |
1891 | cnt++; | |
1892 | ||
1893 | pfn = memory_bm_next_pfn(bm); | |
1894 | } | |
1895 | return cnt; | |
1896 | } | |
1897 | ||
1898 | /** | |
1899 | * prepare_highmem_image - try to allocate as many highmem pages as | |
1900 | * there are highmem image pages (@nr_highmem_p points to the variable | |
1901 | * containing the number of highmem image pages). The pages that are | |
1902 | * "safe" (ie. will not be overwritten when the suspend image is | |
1903 | * restored) have the corresponding bits set in @bm (it must be | |
1904 | * unitialized). | |
1905 | * | |
1906 | * NOTE: This function should not be called if there are no highmem | |
1907 | * image pages. | |
1908 | */ | |
1909 | ||
1910 | static unsigned int safe_highmem_pages; | |
1911 | ||
1912 | static struct memory_bitmap *safe_highmem_bm; | |
1913 | ||
1914 | static int | |
1915 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
1916 | { | |
1917 | unsigned int to_alloc; | |
1918 | ||
1919 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
1920 | return -ENOMEM; | |
1921 | ||
1922 | if (get_highmem_buffer(PG_SAFE)) | |
1923 | return -ENOMEM; | |
1924 | ||
1925 | to_alloc = count_free_highmem_pages(); | |
1926 | if (to_alloc > *nr_highmem_p) | |
1927 | to_alloc = *nr_highmem_p; | |
1928 | else | |
1929 | *nr_highmem_p = to_alloc; | |
1930 | ||
1931 | safe_highmem_pages = 0; | |
1932 | while (to_alloc-- > 0) { | |
1933 | struct page *page; | |
1934 | ||
1935 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 1936 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
1937 | /* The page is "safe", set its bit the bitmap */ |
1938 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
1939 | safe_highmem_pages++; | |
1940 | } | |
1941 | /* Mark the page as allocated */ | |
7be98234 RW |
1942 | swsusp_set_page_forbidden(page); |
1943 | swsusp_set_page_free(page); | |
8357376d RW |
1944 | } |
1945 | memory_bm_position_reset(bm); | |
1946 | safe_highmem_bm = bm; | |
1947 | return 0; | |
1948 | } | |
1949 | ||
1950 | /** | |
1951 | * get_highmem_page_buffer - for given highmem image page find the buffer | |
1952 | * that suspend_write_next() should set for its caller to write to. | |
1953 | * | |
1954 | * If the page is to be saved to its "original" page frame or a copy of | |
1955 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
1956 | * the copy of the page is to be made in normal memory, so the address of | |
1957 | * the copy is returned. | |
1958 | * | |
1959 | * If @buffer is returned, the caller of suspend_write_next() will write | |
1960 | * the page's contents to @buffer, so they will have to be copied to the | |
1961 | * right location on the next call to suspend_write_next() and it is done | |
1962 | * with the help of copy_last_highmem_page(). For this purpose, if | |
1963 | * @buffer is returned, @last_highmem page is set to the page to which | |
1964 | * the data will have to be copied from @buffer. | |
1965 | */ | |
1966 | ||
1967 | static struct page *last_highmem_page; | |
1968 | ||
1969 | static void * | |
1970 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
1971 | { | |
1972 | struct highmem_pbe *pbe; | |
1973 | void *kaddr; | |
1974 | ||
7be98234 | 1975 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
8357376d RW |
1976 | /* We have allocated the "original" page frame and we can |
1977 | * use it directly to store the loaded page. | |
1978 | */ | |
1979 | last_highmem_page = page; | |
1980 | return buffer; | |
1981 | } | |
1982 | /* The "original" page frame has not been allocated and we have to | |
1983 | * use a "safe" page frame to store the loaded page. | |
1984 | */ | |
1985 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
1986 | if (!pbe) { | |
1987 | swsusp_free(); | |
69643279 | 1988 | return ERR_PTR(-ENOMEM); |
8357376d RW |
1989 | } |
1990 | pbe->orig_page = page; | |
1991 | if (safe_highmem_pages > 0) { | |
1992 | struct page *tmp; | |
1993 | ||
1994 | /* Copy of the page will be stored in high memory */ | |
1995 | kaddr = buffer; | |
1996 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
1997 | safe_highmem_pages--; | |
1998 | last_highmem_page = tmp; | |
1999 | pbe->copy_page = tmp; | |
2000 | } else { | |
2001 | /* Copy of the page will be stored in normal memory */ | |
2002 | kaddr = safe_pages_list; | |
2003 | safe_pages_list = safe_pages_list->next; | |
2004 | pbe->copy_page = virt_to_page(kaddr); | |
2005 | } | |
2006 | pbe->next = highmem_pblist; | |
2007 | highmem_pblist = pbe; | |
2008 | return kaddr; | |
2009 | } | |
2010 | ||
2011 | /** | |
2012 | * copy_last_highmem_page - copy the contents of a highmem image from | |
2013 | * @buffer, where the caller of snapshot_write_next() has place them, | |
2014 | * to the right location represented by @last_highmem_page . | |
2015 | */ | |
2016 | ||
2017 | static void copy_last_highmem_page(void) | |
2018 | { | |
2019 | if (last_highmem_page) { | |
2020 | void *dst; | |
2021 | ||
0de9a1e2 | 2022 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2023 | copy_page(dst, buffer); |
0de9a1e2 | 2024 | kunmap_atomic(dst); |
8357376d RW |
2025 | last_highmem_page = NULL; |
2026 | } | |
2027 | } | |
2028 | ||
2029 | static inline int last_highmem_page_copied(void) | |
2030 | { | |
2031 | return !last_highmem_page; | |
2032 | } | |
2033 | ||
2034 | static inline void free_highmem_data(void) | |
2035 | { | |
2036 | if (safe_highmem_bm) | |
2037 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2038 | ||
2039 | if (buffer) | |
2040 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2041 | } | |
2042 | #else | |
2043 | static inline int get_safe_write_buffer(void) { return 0; } | |
2044 | ||
2045 | static unsigned int | |
2046 | count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } | |
2047 | ||
2048 | static inline int | |
2049 | prepare_highmem_image(struct memory_bitmap *bm, unsigned int *nr_highmem_p) | |
2050 | { | |
2051 | return 0; | |
2052 | } | |
2053 | ||
2054 | static inline void * | |
2055 | get_highmem_page_buffer(struct page *page, struct chain_allocator *ca) | |
2056 | { | |
69643279 | 2057 | return ERR_PTR(-EINVAL); |
8357376d RW |
2058 | } |
2059 | ||
2060 | static inline void copy_last_highmem_page(void) {} | |
2061 | static inline int last_highmem_page_copied(void) { return 1; } | |
2062 | static inline void free_highmem_data(void) {} | |
2063 | #endif /* CONFIG_HIGHMEM */ | |
2064 | ||
f577eb30 | 2065 | /** |
940864dd RW |
2066 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
2067 | * be overwritten in the process of restoring the system memory state | |
2068 | * from the suspend image ("unsafe" pages) and allocate memory for the | |
2069 | * image. | |
968808b8 | 2070 | * |
940864dd RW |
2071 | * The idea is to allocate a new memory bitmap first and then allocate |
2072 | * as many pages as needed for the image data, but not to assign these | |
2073 | * pages to specific tasks initially. Instead, we just mark them as | |
8357376d RW |
2074 | * allocated and create a lists of "safe" pages that will be used |
2075 | * later. On systems with high memory a list of "safe" highmem pages is | |
2076 | * also created. | |
f577eb30 RW |
2077 | */ |
2078 | ||
940864dd RW |
2079 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2080 | ||
940864dd RW |
2081 | static int |
2082 | prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) | |
f577eb30 | 2083 | { |
8357376d | 2084 | unsigned int nr_pages, nr_highmem; |
940864dd RW |
2085 | struct linked_page *sp_list, *lp; |
2086 | int error; | |
f577eb30 | 2087 | |
8357376d RW |
2088 | /* If there is no highmem, the buffer will not be necessary */ |
2089 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2090 | buffer = NULL; | |
2091 | ||
2092 | nr_highmem = count_highmem_image_pages(bm); | |
940864dd RW |
2093 | error = mark_unsafe_pages(bm); |
2094 | if (error) | |
2095 | goto Free; | |
2096 | ||
2097 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2098 | if (error) | |
2099 | goto Free; | |
2100 | ||
2101 | duplicate_memory_bitmap(new_bm, bm); | |
2102 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2103 | if (nr_highmem > 0) { |
2104 | error = prepare_highmem_image(bm, &nr_highmem); | |
2105 | if (error) | |
2106 | goto Free; | |
2107 | } | |
940864dd RW |
2108 | /* Reserve some safe pages for potential later use. |
2109 | * | |
2110 | * NOTE: This way we make sure there will be enough safe pages for the | |
2111 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2112 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
2113 | */ | |
2114 | sp_list = NULL; | |
2115 | /* nr_copy_pages cannot be lesser than allocated_unsafe_pages */ | |
8357376d | 2116 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2117 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2118 | while (nr_pages > 0) { | |
8357376d | 2119 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2120 | if (!lp) { |
f577eb30 | 2121 | error = -ENOMEM; |
940864dd RW |
2122 | goto Free; |
2123 | } | |
2124 | lp->next = sp_list; | |
2125 | sp_list = lp; | |
2126 | nr_pages--; | |
f577eb30 | 2127 | } |
940864dd RW |
2128 | /* Preallocate memory for the image */ |
2129 | safe_pages_list = NULL; | |
8357376d | 2130 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2131 | while (nr_pages > 0) { |
2132 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2133 | if (!lp) { | |
2134 | error = -ENOMEM; | |
2135 | goto Free; | |
2136 | } | |
7be98234 | 2137 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2138 | /* The page is "safe", add it to the list */ |
2139 | lp->next = safe_pages_list; | |
2140 | safe_pages_list = lp; | |
968808b8 | 2141 | } |
940864dd | 2142 | /* Mark the page as allocated */ |
7be98234 RW |
2143 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2144 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2145 | nr_pages--; |
968808b8 | 2146 | } |
940864dd RW |
2147 | /* Free the reserved safe pages so that chain_alloc() can use them */ |
2148 | while (sp_list) { | |
2149 | lp = sp_list->next; | |
2150 | free_image_page(sp_list, PG_UNSAFE_CLEAR); | |
2151 | sp_list = lp; | |
f577eb30 | 2152 | } |
940864dd RW |
2153 | return 0; |
2154 | ||
59a49335 | 2155 | Free: |
940864dd | 2156 | swsusp_free(); |
f577eb30 RW |
2157 | return error; |
2158 | } | |
2159 | ||
940864dd RW |
2160 | /** |
2161 | * get_buffer - compute the address that snapshot_write_next() should | |
2162 | * set for its caller to write to. | |
2163 | */ | |
2164 | ||
2165 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) | |
968808b8 | 2166 | { |
940864dd | 2167 | struct pbe *pbe; |
69643279 RW |
2168 | struct page *page; |
2169 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2170 | |
69643279 RW |
2171 | if (pfn == BM_END_OF_MAP) |
2172 | return ERR_PTR(-EFAULT); | |
2173 | ||
2174 | page = pfn_to_page(pfn); | |
8357376d RW |
2175 | if (PageHighMem(page)) |
2176 | return get_highmem_page_buffer(page, ca); | |
2177 | ||
7be98234 | 2178 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
940864dd RW |
2179 | /* We have allocated the "original" page frame and we can |
2180 | * use it directly to store the loaded page. | |
968808b8 | 2181 | */ |
940864dd RW |
2182 | return page_address(page); |
2183 | ||
2184 | /* The "original" page frame has not been allocated and we have to | |
2185 | * use a "safe" page frame to store the loaded page. | |
968808b8 | 2186 | */ |
940864dd RW |
2187 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2188 | if (!pbe) { | |
2189 | swsusp_free(); | |
69643279 | 2190 | return ERR_PTR(-ENOMEM); |
940864dd | 2191 | } |
8357376d RW |
2192 | pbe->orig_address = page_address(page); |
2193 | pbe->address = safe_pages_list; | |
940864dd RW |
2194 | safe_pages_list = safe_pages_list->next; |
2195 | pbe->next = restore_pblist; | |
2196 | restore_pblist = pbe; | |
8357376d | 2197 | return pbe->address; |
968808b8 RW |
2198 | } |
2199 | ||
f577eb30 RW |
2200 | /** |
2201 | * snapshot_write_next - used for writing the system memory snapshot. | |
2202 | * | |
2203 | * On the first call to it @handle should point to a zeroed | |
2204 | * snapshot_handle structure. The structure gets updated and a pointer | |
2205 | * to it should be passed to this function every next time. | |
2206 | * | |
f577eb30 RW |
2207 | * On success the function returns a positive number. Then, the caller |
2208 | * is allowed to write up to the returned number of bytes to the memory | |
d3c1b24c | 2209 | * location computed by the data_of() macro. |
f577eb30 RW |
2210 | * |
2211 | * The function returns 0 to indicate the "end of file" condition, | |
2212 | * and a negative number is returned on error. In such cases the | |
2213 | * structure pointed to by @handle is not updated and should not be used | |
2214 | * any more. | |
2215 | */ | |
2216 | ||
d3c1b24c | 2217 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2218 | { |
940864dd | 2219 | static struct chain_allocator ca; |
f577eb30 RW |
2220 | int error = 0; |
2221 | ||
940864dd | 2222 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2223 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2224 | return 0; |
940864dd | 2225 | |
d3c1b24c JS |
2226 | handle->sync_read = 1; |
2227 | ||
2228 | if (!handle->cur) { | |
8357376d RW |
2229 | if (!buffer) |
2230 | /* This makes the buffer be freed by swsusp_free() */ | |
2231 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2232 | ||
f577eb30 RW |
2233 | if (!buffer) |
2234 | return -ENOMEM; | |
8357376d | 2235 | |
f577eb30 | 2236 | handle->buffer = buffer; |
d3c1b24c JS |
2237 | } else if (handle->cur == 1) { |
2238 | error = load_header(buffer); | |
2239 | if (error) | |
2240 | return error; | |
940864dd | 2241 | |
d3c1b24c JS |
2242 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2243 | if (error) | |
2244 | return error; | |
2245 | ||
85055dd8 MS |
2246 | /* Allocate buffer for page keys. */ |
2247 | error = page_key_alloc(nr_copy_pages); | |
2248 | if (error) | |
2249 | return error; | |
2250 | ||
d3c1b24c JS |
2251 | } else if (handle->cur <= nr_meta_pages + 1) { |
2252 | error = unpack_orig_pfns(buffer, ©_bm); | |
2253 | if (error) | |
2254 | return error; | |
940864dd | 2255 | |
d3c1b24c JS |
2256 | if (handle->cur == nr_meta_pages + 1) { |
2257 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2258 | if (error) |
2259 | return error; | |
2260 | ||
d3c1b24c JS |
2261 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2262 | memory_bm_position_reset(&orig_bm); | |
2263 | restore_pblist = NULL; | |
940864dd | 2264 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2265 | handle->sync_read = 0; |
69643279 RW |
2266 | if (IS_ERR(handle->buffer)) |
2267 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2268 | } |
f577eb30 | 2269 | } else { |
d3c1b24c | 2270 | copy_last_highmem_page(); |
85055dd8 MS |
2271 | /* Restore page key for data page (s390 only). */ |
2272 | page_key_write(handle->buffer); | |
d3c1b24c JS |
2273 | handle->buffer = get_buffer(&orig_bm, &ca); |
2274 | if (IS_ERR(handle->buffer)) | |
2275 | return PTR_ERR(handle->buffer); | |
2276 | if (handle->buffer != buffer) | |
2277 | handle->sync_read = 0; | |
f577eb30 | 2278 | } |
d3c1b24c JS |
2279 | handle->cur++; |
2280 | return PAGE_SIZE; | |
f577eb30 RW |
2281 | } |
2282 | ||
8357376d RW |
2283 | /** |
2284 | * snapshot_write_finalize - must be called after the last call to | |
2285 | * snapshot_write_next() in case the last page in the image happens | |
2286 | * to be a highmem page and its contents should be stored in the | |
2287 | * highmem. Additionally, it releases the memory that will not be | |
2288 | * used any more. | |
2289 | */ | |
2290 | ||
2291 | void snapshot_write_finalize(struct snapshot_handle *handle) | |
2292 | { | |
2293 | copy_last_highmem_page(); | |
85055dd8 MS |
2294 | /* Restore page key for data page (s390 only). */ |
2295 | page_key_write(handle->buffer); | |
2296 | page_key_free(); | |
8357376d | 2297 | /* Free only if we have loaded the image entirely */ |
d3c1b24c | 2298 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
8357376d RW |
2299 | memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR); |
2300 | free_highmem_data(); | |
2301 | } | |
2302 | } | |
2303 | ||
f577eb30 RW |
2304 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2305 | { | |
8357376d | 2306 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2307 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2308 | } | |
2309 | ||
8357376d RW |
2310 | #ifdef CONFIG_HIGHMEM |
2311 | /* Assumes that @buf is ready and points to a "safe" page */ | |
2312 | static inline void | |
2313 | swap_two_pages_data(struct page *p1, struct page *p2, void *buf) | |
940864dd | 2314 | { |
8357376d RW |
2315 | void *kaddr1, *kaddr2; |
2316 | ||
0de9a1e2 CW |
2317 | kaddr1 = kmap_atomic(p1); |
2318 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2319 | copy_page(buf, kaddr1); |
2320 | copy_page(kaddr1, kaddr2); | |
2321 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2322 | kunmap_atomic(kaddr2); |
2323 | kunmap_atomic(kaddr1); | |
8357376d RW |
2324 | } |
2325 | ||
2326 | /** | |
2327 | * restore_highmem - for each highmem page that was allocated before | |
2328 | * the suspend and included in the suspend image, and also has been | |
2329 | * allocated by the "resume" kernel swap its current (ie. "before | |
2330 | * resume") contents with the previous (ie. "before suspend") one. | |
2331 | * | |
2332 | * If the resume eventually fails, we can call this function once | |
2333 | * again and restore the "before resume" highmem state. | |
2334 | */ | |
2335 | ||
2336 | int restore_highmem(void) | |
2337 | { | |
2338 | struct highmem_pbe *pbe = highmem_pblist; | |
2339 | void *buf; | |
2340 | ||
2341 | if (!pbe) | |
2342 | return 0; | |
2343 | ||
2344 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2345 | if (!buf) | |
2346 | return -ENOMEM; | |
2347 | ||
2348 | while (pbe) { | |
2349 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2350 | pbe = pbe->next; | |
2351 | } | |
2352 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2353 | return 0; | |
f577eb30 | 2354 | } |
8357376d | 2355 | #endif /* CONFIG_HIGHMEM */ |