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> |
52f5684c | 30 | #include <linux/compiler.h> |
db597605 | 31 | #include <linux/ktime.h> |
25761b6e RW |
32 | |
33 | #include <asm/uaccess.h> | |
34 | #include <asm/mmu_context.h> | |
35 | #include <asm/pgtable.h> | |
36 | #include <asm/tlbflush.h> | |
37 | #include <asm/io.h> | |
38 | ||
25761b6e RW |
39 | #include "power.h" |
40 | ||
74dfd666 RW |
41 | static int swsusp_page_is_free(struct page *); |
42 | static void swsusp_set_page_forbidden(struct page *); | |
43 | static void swsusp_unset_page_forbidden(struct page *); | |
44 | ||
ddeb6487 RW |
45 | /* |
46 | * Number of bytes to reserve for memory allocations made by device drivers | |
47 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
48 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
49 | */ | |
50 | unsigned long reserved_size; | |
51 | ||
52 | void __init hibernate_reserved_size_init(void) | |
53 | { | |
54 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
55 | } | |
56 | ||
fe419535 RW |
57 | /* |
58 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
59 | * When it is set to N, swsusp will do its best to ensure the image |
60 | * size will not exceed N bytes, but if that is impossible, it will | |
61 | * try to create the smallest image possible. | |
fe419535 | 62 | */ |
ac5c24ec RW |
63 | unsigned long image_size; |
64 | ||
65 | void __init hibernate_image_size_init(void) | |
66 | { | |
1c1be3a9 | 67 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 68 | } |
fe419535 | 69 | |
8357376d RW |
70 | /* List of PBEs needed for restoring the pages that were allocated before |
71 | * the suspend and included in the suspend image, but have also been | |
72 | * allocated by the "resume" kernel, so their contents cannot be written | |
73 | * directly to their "original" page frames. | |
74 | */ | |
75534b50 RW |
75 | struct pbe *restore_pblist; |
76 | ||
9c744481 RW |
77 | /* struct linked_page is used to build chains of pages */ |
78 | ||
79 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
80 | ||
81 | struct linked_page { | |
82 | struct linked_page *next; | |
83 | char data[LINKED_PAGE_DATA_SIZE]; | |
84 | } __packed; | |
85 | ||
86 | /* | |
87 | * List of "safe" pages (ie. pages that were not used by the image kernel | |
88 | * before hibernation) that may be used as temporary storage for image kernel | |
89 | * memory contents. | |
90 | */ | |
91 | static struct linked_page *safe_pages_list; | |
92 | ||
8357376d | 93 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 94 | static void *buffer; |
7088a5c0 | 95 | |
f6143aa6 RW |
96 | /** |
97 | * @safe_needed - on resume, for storing the PBE list and the image, | |
98 | * we can only use memory pages that do not conflict with the pages | |
8357376d RW |
99 | * used before suspend. The unsafe pages have PageNosaveFree set |
100 | * and we count them using unsafe_pages. | |
f6143aa6 | 101 | * |
8357376d RW |
102 | * Each allocated image page is marked as PageNosave and PageNosaveFree |
103 | * so that swsusp_free() can release it. | |
f6143aa6 RW |
104 | */ |
105 | ||
0bcd888d RW |
106 | #define PG_ANY 0 |
107 | #define PG_SAFE 1 | |
108 | #define PG_UNSAFE_CLEAR 1 | |
109 | #define PG_UNSAFE_KEEP 0 | |
110 | ||
940864dd | 111 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 112 | |
8357376d | 113 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
114 | { |
115 | void *res; | |
116 | ||
117 | res = (void *)get_zeroed_page(gfp_mask); | |
118 | if (safe_needed) | |
7be98234 | 119 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 120 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 121 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 122 | allocated_unsafe_pages++; |
f6143aa6 RW |
123 | res = (void *)get_zeroed_page(gfp_mask); |
124 | } | |
125 | if (res) { | |
7be98234 RW |
126 | swsusp_set_page_forbidden(virt_to_page(res)); |
127 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
128 | } |
129 | return res; | |
130 | } | |
131 | ||
9c744481 RW |
132 | static void *__get_safe_page(gfp_t gfp_mask) |
133 | { | |
134 | if (safe_pages_list) { | |
135 | void *ret = safe_pages_list; | |
136 | ||
137 | safe_pages_list = safe_pages_list->next; | |
138 | memset(ret, 0, PAGE_SIZE); | |
139 | return ret; | |
140 | } | |
141 | return get_image_page(gfp_mask, PG_SAFE); | |
142 | } | |
143 | ||
f6143aa6 RW |
144 | unsigned long get_safe_page(gfp_t gfp_mask) |
145 | { | |
9c744481 | 146 | return (unsigned long)__get_safe_page(gfp_mask); |
8357376d RW |
147 | } |
148 | ||
5b6d15de RW |
149 | static struct page *alloc_image_page(gfp_t gfp_mask) |
150 | { | |
8357376d RW |
151 | struct page *page; |
152 | ||
153 | page = alloc_page(gfp_mask); | |
154 | if (page) { | |
7be98234 RW |
155 | swsusp_set_page_forbidden(page); |
156 | swsusp_set_page_free(page); | |
8357376d RW |
157 | } |
158 | return page; | |
f6143aa6 RW |
159 | } |
160 | ||
307c5971 RW |
161 | static void recycle_safe_page(void *page_address) |
162 | { | |
163 | struct linked_page *lp = page_address; | |
164 | ||
165 | lp->next = safe_pages_list; | |
166 | safe_pages_list = lp; | |
167 | } | |
168 | ||
f6143aa6 RW |
169 | /** |
170 | * free_image_page - free page represented by @addr, allocated with | |
8357376d | 171 | * get_image_page (page flags set by it must be cleared) |
f6143aa6 RW |
172 | */ |
173 | ||
174 | static inline void free_image_page(void *addr, int clear_nosave_free) | |
175 | { | |
8357376d RW |
176 | struct page *page; |
177 | ||
178 | BUG_ON(!virt_addr_valid(addr)); | |
179 | ||
180 | page = virt_to_page(addr); | |
181 | ||
7be98234 | 182 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 183 | if (clear_nosave_free) |
7be98234 | 184 | swsusp_unset_page_free(page); |
8357376d RW |
185 | |
186 | __free_page(page); | |
f6143aa6 RW |
187 | } |
188 | ||
efd5a852 RW |
189 | static inline void free_list_of_pages(struct linked_page *list, |
190 | int clear_page_nosave) | |
b788db79 RW |
191 | { |
192 | while (list) { | |
193 | struct linked_page *lp = list->next; | |
194 | ||
195 | free_image_page(list, clear_page_nosave); | |
196 | list = lp; | |
197 | } | |
198 | } | |
199 | ||
200 | /** | |
201 | * struct chain_allocator is used for allocating small objects out of | |
202 | * a linked list of pages called 'the chain'. | |
203 | * | |
204 | * The chain grows each time when there is no room for a new object in | |
205 | * the current page. The allocated objects cannot be freed individually. | |
206 | * It is only possible to free them all at once, by freeing the entire | |
207 | * chain. | |
208 | * | |
209 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
210 | * are not much smaller than PAGE_SIZE. | |
211 | */ | |
212 | ||
213 | struct chain_allocator { | |
214 | struct linked_page *chain; /* the chain */ | |
215 | unsigned int used_space; /* total size of objects allocated out | |
216 | * of the current page | |
217 | */ | |
218 | gfp_t gfp_mask; /* mask for allocating pages */ | |
219 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
220 | }; | |
221 | ||
efd5a852 RW |
222 | static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask, |
223 | int safe_needed) | |
b788db79 RW |
224 | { |
225 | ca->chain = NULL; | |
226 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
227 | ca->gfp_mask = gfp_mask; | |
228 | ca->safe_needed = safe_needed; | |
229 | } | |
230 | ||
231 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
232 | { | |
233 | void *ret; | |
234 | ||
235 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
236 | struct linked_page *lp; | |
237 | ||
9c744481 RW |
238 | lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) : |
239 | get_image_page(ca->gfp_mask, PG_ANY); | |
b788db79 RW |
240 | if (!lp) |
241 | return NULL; | |
242 | ||
243 | lp->next = ca->chain; | |
244 | ca->chain = lp; | |
245 | ca->used_space = 0; | |
246 | } | |
247 | ret = ca->chain->data + ca->used_space; | |
248 | ca->used_space += size; | |
249 | return ret; | |
250 | } | |
251 | ||
b788db79 RW |
252 | /** |
253 | * Data types related to memory bitmaps. | |
254 | * | |
255 | * Memory bitmap is a structure consiting of many linked lists of | |
256 | * objects. The main list's elements are of type struct zone_bitmap | |
257 | * and each of them corresonds to one zone. For each zone bitmap | |
258 | * object there is a list of objects of type struct bm_block that | |
0d83304c | 259 | * represent each blocks of bitmap in which information is stored. |
b788db79 RW |
260 | * |
261 | * struct memory_bitmap contains a pointer to the main list of zone | |
262 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
263 | * and a pointer to the list of pages used for allocating all of the | |
264 | * zone bitmap objects and bitmap block objects. | |
265 | * | |
266 | * NOTE: It has to be possible to lay out the bitmap in memory | |
267 | * using only allocations of order 0. Additionally, the bitmap is | |
268 | * designed to work with arbitrary number of zones (this is over the | |
269 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
270 | * | |
271 | * struct zone_bitmap contains a pointer to a list of bitmap block | |
272 | * objects and a pointer to the bitmap block object that has been | |
273 | * most recently used for setting bits. Additionally, it contains the | |
274 | * pfns that correspond to the start and end of the represented zone. | |
275 | * | |
276 | * struct bm_block contains a pointer to the memory page in which | |
0d83304c AM |
277 | * information is stored (in the form of a block of bitmap) |
278 | * It also contains the pfns that correspond to the start and end of | |
279 | * the represented memory area. | |
f469f02d JR |
280 | * |
281 | * The memory bitmap is organized as a radix tree to guarantee fast random | |
282 | * access to the bits. There is one radix tree for each zone (as returned | |
283 | * from create_mem_extents). | |
284 | * | |
285 | * One radix tree is represented by one struct mem_zone_bm_rtree. There are | |
286 | * two linked lists for the nodes of the tree, one for the inner nodes and | |
287 | * one for the leave nodes. The linked leave nodes are used for fast linear | |
288 | * access of the memory bitmap. | |
289 | * | |
290 | * The struct rtree_node represents one node of the radix tree. | |
b788db79 RW |
291 | */ |
292 | ||
293 | #define BM_END_OF_MAP (~0UL) | |
294 | ||
8de03073 | 295 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
f469f02d JR |
296 | #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3) |
297 | #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1) | |
b788db79 | 298 | |
f469f02d JR |
299 | /* |
300 | * struct rtree_node is a wrapper struct to link the nodes | |
301 | * of the rtree together for easy linear iteration over | |
302 | * bits and easy freeing | |
303 | */ | |
304 | struct rtree_node { | |
305 | struct list_head list; | |
306 | unsigned long *data; | |
307 | }; | |
308 | ||
309 | /* | |
310 | * struct mem_zone_bm_rtree represents a bitmap used for one | |
311 | * populated memory zone. | |
312 | */ | |
313 | struct mem_zone_bm_rtree { | |
314 | struct list_head list; /* Link Zones together */ | |
315 | struct list_head nodes; /* Radix Tree inner nodes */ | |
316 | struct list_head leaves; /* Radix Tree leaves */ | |
317 | unsigned long start_pfn; /* Zone start page frame */ | |
318 | unsigned long end_pfn; /* Zone end page frame + 1 */ | |
319 | struct rtree_node *rtree; /* Radix Tree Root */ | |
320 | int levels; /* Number of Radix Tree Levels */ | |
321 | unsigned int blocks; /* Number of Bitmap Blocks */ | |
322 | }; | |
323 | ||
b788db79 RW |
324 | /* strcut bm_position is used for browsing memory bitmaps */ |
325 | ||
326 | struct bm_position { | |
3a20cb17 JR |
327 | struct mem_zone_bm_rtree *zone; |
328 | struct rtree_node *node; | |
329 | unsigned long node_pfn; | |
330 | int node_bit; | |
b788db79 RW |
331 | }; |
332 | ||
333 | struct memory_bitmap { | |
f469f02d | 334 | struct list_head zones; |
b788db79 RW |
335 | struct linked_page *p_list; /* list of pages used to store zone |
336 | * bitmap objects and bitmap block | |
337 | * objects | |
338 | */ | |
339 | struct bm_position cur; /* most recently used bit position */ | |
340 | }; | |
341 | ||
342 | /* Functions that operate on memory bitmaps */ | |
343 | ||
f469f02d JR |
344 | #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long)) |
345 | #if BITS_PER_LONG == 32 | |
346 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2) | |
347 | #else | |
348 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3) | |
349 | #endif | |
350 | #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1) | |
351 | ||
352 | /* | |
353 | * alloc_rtree_node - Allocate a new node and add it to the radix tree. | |
354 | * | |
355 | * This function is used to allocate inner nodes as well as the | |
356 | * leave nodes of the radix tree. It also adds the node to the | |
357 | * corresponding linked list passed in by the *list parameter. | |
358 | */ | |
359 | static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed, | |
360 | struct chain_allocator *ca, | |
361 | struct list_head *list) | |
362 | { | |
363 | struct rtree_node *node; | |
364 | ||
365 | node = chain_alloc(ca, sizeof(struct rtree_node)); | |
366 | if (!node) | |
367 | return NULL; | |
368 | ||
369 | node->data = get_image_page(gfp_mask, safe_needed); | |
370 | if (!node->data) | |
371 | return NULL; | |
372 | ||
373 | list_add_tail(&node->list, list); | |
374 | ||
375 | return node; | |
376 | } | |
377 | ||
378 | /* | |
379 | * add_rtree_block - Add a new leave node to the radix tree | |
380 | * | |
381 | * The leave nodes need to be allocated in order to keep the leaves | |
382 | * linked list in order. This is guaranteed by the zone->blocks | |
383 | * counter. | |
384 | */ | |
385 | static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask, | |
386 | int safe_needed, struct chain_allocator *ca) | |
387 | { | |
388 | struct rtree_node *node, *block, **dst; | |
389 | unsigned int levels_needed, block_nr; | |
390 | int i; | |
391 | ||
392 | block_nr = zone->blocks; | |
393 | levels_needed = 0; | |
394 | ||
395 | /* How many levels do we need for this block nr? */ | |
396 | while (block_nr) { | |
397 | levels_needed += 1; | |
398 | block_nr >>= BM_RTREE_LEVEL_SHIFT; | |
399 | } | |
400 | ||
401 | /* Make sure the rtree has enough levels */ | |
402 | for (i = zone->levels; i < levels_needed; i++) { | |
403 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
404 | &zone->nodes); | |
405 | if (!node) | |
406 | return -ENOMEM; | |
407 | ||
408 | node->data[0] = (unsigned long)zone->rtree; | |
409 | zone->rtree = node; | |
410 | zone->levels += 1; | |
411 | } | |
412 | ||
413 | /* Allocate new block */ | |
414 | block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves); | |
415 | if (!block) | |
416 | return -ENOMEM; | |
417 | ||
418 | /* Now walk the rtree to insert the block */ | |
419 | node = zone->rtree; | |
420 | dst = &zone->rtree; | |
421 | block_nr = zone->blocks; | |
422 | for (i = zone->levels; i > 0; i--) { | |
423 | int index; | |
424 | ||
425 | if (!node) { | |
426 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
427 | &zone->nodes); | |
428 | if (!node) | |
429 | return -ENOMEM; | |
430 | *dst = node; | |
431 | } | |
432 | ||
433 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
434 | index &= BM_RTREE_LEVEL_MASK; | |
435 | dst = (struct rtree_node **)&((*dst)->data[index]); | |
436 | node = *dst; | |
437 | } | |
438 | ||
439 | zone->blocks += 1; | |
440 | *dst = block; | |
441 | ||
442 | return 0; | |
443 | } | |
444 | ||
445 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
446 | int clear_nosave_free); | |
447 | ||
448 | /* | |
449 | * create_zone_bm_rtree - create a radix tree for one zone | |
450 | * | |
451 | * Allocated the mem_zone_bm_rtree structure and initializes it. | |
452 | * This function also allocated and builds the radix tree for the | |
453 | * zone. | |
454 | */ | |
efd5a852 RW |
455 | static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask, |
456 | int safe_needed, | |
457 | struct chain_allocator *ca, | |
458 | unsigned long start, | |
459 | unsigned long end) | |
f469f02d JR |
460 | { |
461 | struct mem_zone_bm_rtree *zone; | |
462 | unsigned int i, nr_blocks; | |
463 | unsigned long pages; | |
464 | ||
465 | pages = end - start; | |
466 | zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree)); | |
467 | if (!zone) | |
468 | return NULL; | |
469 | ||
470 | INIT_LIST_HEAD(&zone->nodes); | |
471 | INIT_LIST_HEAD(&zone->leaves); | |
472 | zone->start_pfn = start; | |
473 | zone->end_pfn = end; | |
474 | nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); | |
475 | ||
476 | for (i = 0; i < nr_blocks; i++) { | |
477 | if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) { | |
478 | free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR); | |
479 | return NULL; | |
480 | } | |
481 | } | |
482 | ||
483 | return zone; | |
484 | } | |
485 | ||
486 | /* | |
487 | * free_zone_bm_rtree - Free the memory of the radix tree | |
488 | * | |
489 | * Free all node pages of the radix tree. The mem_zone_bm_rtree | |
490 | * structure itself is not freed here nor are the rtree_node | |
491 | * structs. | |
492 | */ | |
493 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
494 | int clear_nosave_free) | |
495 | { | |
496 | struct rtree_node *node; | |
497 | ||
498 | list_for_each_entry(node, &zone->nodes, list) | |
499 | free_image_page(node->data, clear_nosave_free); | |
500 | ||
501 | list_for_each_entry(node, &zone->leaves, list) | |
502 | free_image_page(node->data, clear_nosave_free); | |
503 | } | |
504 | ||
b788db79 RW |
505 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
506 | { | |
3a20cb17 JR |
507 | bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree, |
508 | list); | |
509 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, | |
510 | struct rtree_node, list); | |
511 | bm->cur.node_pfn = 0; | |
512 | bm->cur.node_bit = 0; | |
b788db79 RW |
513 | } |
514 | ||
515 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
516 | ||
846705de RW |
517 | struct mem_extent { |
518 | struct list_head hook; | |
519 | unsigned long start; | |
520 | unsigned long end; | |
521 | }; | |
522 | ||
b788db79 | 523 | /** |
846705de RW |
524 | * free_mem_extents - free a list of memory extents |
525 | * @list - list of extents to empty | |
b788db79 | 526 | */ |
846705de RW |
527 | static void free_mem_extents(struct list_head *list) |
528 | { | |
529 | struct mem_extent *ext, *aux; | |
b788db79 | 530 | |
846705de RW |
531 | list_for_each_entry_safe(ext, aux, list, hook) { |
532 | list_del(&ext->hook); | |
533 | kfree(ext); | |
534 | } | |
535 | } | |
536 | ||
537 | /** | |
538 | * create_mem_extents - create a list of memory extents representing | |
539 | * contiguous ranges of PFNs | |
540 | * @list - list to put the extents into | |
541 | * @gfp_mask - mask to use for memory allocations | |
542 | */ | |
543 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 544 | { |
846705de | 545 | struct zone *zone; |
b788db79 | 546 | |
846705de | 547 | INIT_LIST_HEAD(list); |
b788db79 | 548 | |
ee99c71c | 549 | for_each_populated_zone(zone) { |
846705de RW |
550 | unsigned long zone_start, zone_end; |
551 | struct mem_extent *ext, *cur, *aux; | |
552 | ||
846705de | 553 | zone_start = zone->zone_start_pfn; |
c33bc315 | 554 | zone_end = zone_end_pfn(zone); |
846705de RW |
555 | |
556 | list_for_each_entry(ext, list, hook) | |
557 | if (zone_start <= ext->end) | |
558 | break; | |
b788db79 | 559 | |
846705de RW |
560 | if (&ext->hook == list || zone_end < ext->start) { |
561 | /* New extent is necessary */ | |
562 | struct mem_extent *new_ext; | |
563 | ||
564 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
565 | if (!new_ext) { | |
566 | free_mem_extents(list); | |
567 | return -ENOMEM; | |
568 | } | |
569 | new_ext->start = zone_start; | |
570 | new_ext->end = zone_end; | |
571 | list_add_tail(&new_ext->hook, &ext->hook); | |
572 | continue; | |
573 | } | |
574 | ||
575 | /* Merge this zone's range of PFNs with the existing one */ | |
576 | if (zone_start < ext->start) | |
577 | ext->start = zone_start; | |
578 | if (zone_end > ext->end) | |
579 | ext->end = zone_end; | |
580 | ||
581 | /* More merging may be possible */ | |
582 | cur = ext; | |
583 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
584 | if (zone_end < cur->start) | |
585 | break; | |
586 | if (zone_end < cur->end) | |
587 | ext->end = cur->end; | |
588 | list_del(&cur->hook); | |
589 | kfree(cur); | |
590 | } | |
b788db79 | 591 | } |
846705de RW |
592 | |
593 | return 0; | |
b788db79 RW |
594 | } |
595 | ||
596 | /** | |
597 | * memory_bm_create - allocate memory for a memory bitmap | |
598 | */ | |
efd5a852 RW |
599 | static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, |
600 | int safe_needed) | |
b788db79 RW |
601 | { |
602 | struct chain_allocator ca; | |
846705de RW |
603 | struct list_head mem_extents; |
604 | struct mem_extent *ext; | |
605 | int error; | |
b788db79 RW |
606 | |
607 | chain_init(&ca, gfp_mask, safe_needed); | |
f469f02d | 608 | INIT_LIST_HEAD(&bm->zones); |
b788db79 | 609 | |
846705de RW |
610 | error = create_mem_extents(&mem_extents, gfp_mask); |
611 | if (error) | |
612 | return error; | |
b788db79 | 613 | |
846705de | 614 | list_for_each_entry(ext, &mem_extents, hook) { |
f469f02d | 615 | struct mem_zone_bm_rtree *zone; |
f469f02d JR |
616 | |
617 | zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca, | |
618 | ext->start, ext->end); | |
9047eb62 JR |
619 | if (!zone) { |
620 | error = -ENOMEM; | |
f469f02d | 621 | goto Error; |
9047eb62 | 622 | } |
f469f02d | 623 | list_add_tail(&zone->list, &bm->zones); |
b788db79 | 624 | } |
846705de | 625 | |
b788db79 RW |
626 | bm->p_list = ca.chain; |
627 | memory_bm_position_reset(bm); | |
846705de RW |
628 | Exit: |
629 | free_mem_extents(&mem_extents); | |
630 | return error; | |
b788db79 | 631 | |
846705de | 632 | Error: |
b788db79 RW |
633 | bm->p_list = ca.chain; |
634 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 635 | goto Exit; |
b788db79 RW |
636 | } |
637 | ||
638 | /** | |
639 | * memory_bm_free - free memory occupied by the memory bitmap @bm | |
640 | */ | |
b788db79 RW |
641 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
642 | { | |
f469f02d | 643 | struct mem_zone_bm_rtree *zone; |
b788db79 | 644 | |
f469f02d JR |
645 | list_for_each_entry(zone, &bm->zones, list) |
646 | free_zone_bm_rtree(zone, clear_nosave_free); | |
647 | ||
b788db79 | 648 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de | 649 | |
f469f02d | 650 | INIT_LIST_HEAD(&bm->zones); |
b788db79 RW |
651 | } |
652 | ||
653 | /** | |
9047eb62 JR |
654 | * memory_bm_find_bit - Find the bit for pfn in the memory |
655 | * bitmap | |
07a33823 | 656 | * |
9047eb62 JR |
657 | * Find the bit in the bitmap @bm that corresponds to given pfn. |
658 | * The cur.zone, cur.block and cur.node_pfn member of @bm are | |
659 | * updated. | |
660 | * It walks the radix tree to find the page which contains the bit for | |
07a33823 JR |
661 | * pfn and returns the bit position in **addr and *bit_nr. |
662 | */ | |
9047eb62 JR |
663 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
664 | void **addr, unsigned int *bit_nr) | |
07a33823 JR |
665 | { |
666 | struct mem_zone_bm_rtree *curr, *zone; | |
667 | struct rtree_node *node; | |
668 | int i, block_nr; | |
669 | ||
3a20cb17 JR |
670 | zone = bm->cur.zone; |
671 | ||
672 | if (pfn >= zone->start_pfn && pfn < zone->end_pfn) | |
673 | goto zone_found; | |
674 | ||
07a33823 JR |
675 | zone = NULL; |
676 | ||
677 | /* Find the right zone */ | |
678 | list_for_each_entry(curr, &bm->zones, list) { | |
679 | if (pfn >= curr->start_pfn && pfn < curr->end_pfn) { | |
680 | zone = curr; | |
681 | break; | |
682 | } | |
683 | } | |
684 | ||
685 | if (!zone) | |
686 | return -EFAULT; | |
687 | ||
3a20cb17 | 688 | zone_found: |
07a33823 JR |
689 | /* |
690 | * We have a zone. Now walk the radix tree to find the leave | |
691 | * node for our pfn. | |
692 | */ | |
3a20cb17 JR |
693 | |
694 | node = bm->cur.node; | |
695 | if (((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn) | |
696 | goto node_found; | |
697 | ||
07a33823 JR |
698 | node = zone->rtree; |
699 | block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT; | |
700 | ||
701 | for (i = zone->levels; i > 0; i--) { | |
702 | int index; | |
703 | ||
704 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
705 | index &= BM_RTREE_LEVEL_MASK; | |
706 | BUG_ON(node->data[index] == 0); | |
707 | node = (struct rtree_node *)node->data[index]; | |
708 | } | |
709 | ||
3a20cb17 JR |
710 | node_found: |
711 | /* Update last position */ | |
712 | bm->cur.zone = zone; | |
713 | bm->cur.node = node; | |
714 | bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK; | |
715 | ||
07a33823 JR |
716 | /* Set return values */ |
717 | *addr = node->data; | |
718 | *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK; | |
719 | ||
720 | return 0; | |
721 | } | |
722 | ||
74dfd666 RW |
723 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) |
724 | { | |
725 | void *addr; | |
726 | unsigned int bit; | |
a82f7119 | 727 | int error; |
74dfd666 | 728 | |
a82f7119 RW |
729 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
730 | BUG_ON(error); | |
74dfd666 RW |
731 | set_bit(bit, addr); |
732 | } | |
733 | ||
a82f7119 RW |
734 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
735 | { | |
736 | void *addr; | |
737 | unsigned int bit; | |
738 | int error; | |
739 | ||
740 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
07a33823 JR |
741 | if (!error) |
742 | set_bit(bit, addr); | |
743 | ||
a82f7119 RW |
744 | return error; |
745 | } | |
746 | ||
74dfd666 RW |
747 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
748 | { | |
749 | void *addr; | |
750 | unsigned int bit; | |
a82f7119 | 751 | int error; |
74dfd666 | 752 | |
a82f7119 RW |
753 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
754 | BUG_ON(error); | |
74dfd666 RW |
755 | clear_bit(bit, addr); |
756 | } | |
757 | ||
fdd64ed5 JR |
758 | static void memory_bm_clear_current(struct memory_bitmap *bm) |
759 | { | |
760 | int bit; | |
761 | ||
762 | bit = max(bm->cur.node_bit - 1, 0); | |
763 | clear_bit(bit, bm->cur.node->data); | |
764 | } | |
765 | ||
74dfd666 RW |
766 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) |
767 | { | |
768 | void *addr; | |
769 | unsigned int bit; | |
9047eb62 | 770 | int error; |
74dfd666 | 771 | |
a82f7119 RW |
772 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
773 | BUG_ON(error); | |
9047eb62 | 774 | return test_bit(bit, addr); |
b788db79 RW |
775 | } |
776 | ||
69643279 RW |
777 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
778 | { | |
779 | void *addr; | |
780 | unsigned int bit; | |
07a33823 | 781 | |
9047eb62 | 782 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); |
b788db79 RW |
783 | } |
784 | ||
3a20cb17 JR |
785 | /* |
786 | * rtree_next_node - Jumps to the next leave node | |
787 | * | |
788 | * Sets the position to the beginning of the next node in the | |
789 | * memory bitmap. This is either the next node in the current | |
790 | * zone's radix tree or the first node in the radix tree of the | |
791 | * next zone. | |
792 | * | |
793 | * Returns true if there is a next node, false otherwise. | |
794 | */ | |
795 | static bool rtree_next_node(struct memory_bitmap *bm) | |
796 | { | |
797 | bm->cur.node = list_entry(bm->cur.node->list.next, | |
798 | struct rtree_node, list); | |
799 | if (&bm->cur.node->list != &bm->cur.zone->leaves) { | |
800 | bm->cur.node_pfn += BM_BITS_PER_BLOCK; | |
801 | bm->cur.node_bit = 0; | |
0f7d83e8 | 802 | touch_softlockup_watchdog(); |
3a20cb17 JR |
803 | return true; |
804 | } | |
805 | ||
806 | /* No more nodes, goto next zone */ | |
807 | bm->cur.zone = list_entry(bm->cur.zone->list.next, | |
808 | struct mem_zone_bm_rtree, list); | |
809 | if (&bm->cur.zone->list != &bm->zones) { | |
810 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, | |
811 | struct rtree_node, list); | |
812 | bm->cur.node_pfn = 0; | |
813 | bm->cur.node_bit = 0; | |
814 | return true; | |
815 | } | |
816 | ||
817 | /* No more zones */ | |
818 | return false; | |
819 | } | |
820 | ||
9047eb62 JR |
821 | /** |
822 | * memory_bm_rtree_next_pfn - Find the next set bit in the bitmap @bm | |
3a20cb17 JR |
823 | * |
824 | * Starting from the last returned position this function searches | |
825 | * for the next set bit in the memory bitmap and returns its | |
826 | * number. If no more bit is set BM_END_OF_MAP is returned. | |
9047eb62 JR |
827 | * |
828 | * It is required to run memory_bm_position_reset() before the | |
829 | * first call to this function. | |
3a20cb17 | 830 | */ |
9047eb62 | 831 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) |
3a20cb17 JR |
832 | { |
833 | unsigned long bits, pfn, pages; | |
834 | int bit; | |
835 | ||
836 | do { | |
837 | pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn; | |
838 | bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK); | |
839 | bit = find_next_bit(bm->cur.node->data, bits, | |
840 | bm->cur.node_bit); | |
841 | if (bit < bits) { | |
842 | pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit; | |
843 | bm->cur.node_bit = bit + 1; | |
844 | return pfn; | |
845 | } | |
846 | } while (rtree_next_node(bm)); | |
847 | ||
848 | return BM_END_OF_MAP; | |
849 | } | |
850 | ||
74dfd666 RW |
851 | /** |
852 | * This structure represents a range of page frames the contents of which | |
853 | * should not be saved during the suspend. | |
854 | */ | |
855 | ||
856 | struct nosave_region { | |
857 | struct list_head list; | |
858 | unsigned long start_pfn; | |
859 | unsigned long end_pfn; | |
860 | }; | |
861 | ||
862 | static LIST_HEAD(nosave_regions); | |
863 | ||
307c5971 RW |
864 | static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone) |
865 | { | |
866 | struct rtree_node *node; | |
867 | ||
868 | list_for_each_entry(node, &zone->nodes, list) | |
869 | recycle_safe_page(node->data); | |
870 | ||
871 | list_for_each_entry(node, &zone->leaves, list) | |
872 | recycle_safe_page(node->data); | |
873 | } | |
874 | ||
875 | static void memory_bm_recycle(struct memory_bitmap *bm) | |
876 | { | |
877 | struct mem_zone_bm_rtree *zone; | |
878 | struct linked_page *p_list; | |
879 | ||
880 | list_for_each_entry(zone, &bm->zones, list) | |
881 | recycle_zone_bm_rtree(zone); | |
882 | ||
883 | p_list = bm->p_list; | |
884 | while (p_list) { | |
885 | struct linked_page *lp = p_list; | |
886 | ||
887 | p_list = lp->next; | |
888 | recycle_safe_page(lp); | |
889 | } | |
890 | } | |
891 | ||
74dfd666 RW |
892 | /** |
893 | * register_nosave_region - register a range of page frames the contents | |
894 | * of which should not be saved during the suspend (to be used in the early | |
895 | * initialization code) | |
896 | */ | |
897 | ||
efd5a852 RW |
898 | void __init __register_nosave_region(unsigned long start_pfn, |
899 | unsigned long end_pfn, int use_kmalloc) | |
74dfd666 RW |
900 | { |
901 | struct nosave_region *region; | |
902 | ||
903 | if (start_pfn >= end_pfn) | |
904 | return; | |
905 | ||
906 | if (!list_empty(&nosave_regions)) { | |
907 | /* Try to extend the previous region (they should be sorted) */ | |
908 | region = list_entry(nosave_regions.prev, | |
909 | struct nosave_region, list); | |
910 | if (region->end_pfn == start_pfn) { | |
911 | region->end_pfn = end_pfn; | |
912 | goto Report; | |
913 | } | |
914 | } | |
940d67f6 JB |
915 | if (use_kmalloc) { |
916 | /* during init, this shouldn't fail */ | |
917 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); | |
918 | BUG_ON(!region); | |
919 | } else | |
920 | /* This allocation cannot fail */ | |
c2f69cda | 921 | region = memblock_virt_alloc(sizeof(struct nosave_region), 0); |
74dfd666 RW |
922 | region->start_pfn = start_pfn; |
923 | region->end_pfn = end_pfn; | |
924 | list_add_tail(®ion->list, &nosave_regions); | |
925 | Report: | |
cd38ca85 BH |
926 | printk(KERN_INFO "PM: Registered nosave memory: [mem %#010llx-%#010llx]\n", |
927 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
928 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
74dfd666 RW |
929 | } |
930 | ||
931 | /* | |
932 | * Set bits in this map correspond to the page frames the contents of which | |
933 | * should not be saved during the suspend. | |
934 | */ | |
935 | static struct memory_bitmap *forbidden_pages_map; | |
936 | ||
937 | /* Set bits in this map correspond to free page frames. */ | |
938 | static struct memory_bitmap *free_pages_map; | |
939 | ||
940 | /* | |
941 | * Each page frame allocated for creating the image is marked by setting the | |
942 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
943 | */ | |
944 | ||
945 | void swsusp_set_page_free(struct page *page) | |
946 | { | |
947 | if (free_pages_map) | |
948 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
949 | } | |
950 | ||
951 | static int swsusp_page_is_free(struct page *page) | |
952 | { | |
953 | return free_pages_map ? | |
954 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
955 | } | |
956 | ||
957 | void swsusp_unset_page_free(struct page *page) | |
958 | { | |
959 | if (free_pages_map) | |
960 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
961 | } | |
962 | ||
963 | static void swsusp_set_page_forbidden(struct page *page) | |
964 | { | |
965 | if (forbidden_pages_map) | |
966 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
967 | } | |
968 | ||
969 | int swsusp_page_is_forbidden(struct page *page) | |
970 | { | |
971 | return forbidden_pages_map ? | |
972 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
973 | } | |
974 | ||
975 | static void swsusp_unset_page_forbidden(struct page *page) | |
976 | { | |
977 | if (forbidden_pages_map) | |
978 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
979 | } | |
980 | ||
981 | /** | |
982 | * mark_nosave_pages - set bits corresponding to the page frames the | |
983 | * contents of which should not be saved in a given bitmap. | |
984 | */ | |
985 | ||
986 | static void mark_nosave_pages(struct memory_bitmap *bm) | |
987 | { | |
988 | struct nosave_region *region; | |
989 | ||
990 | if (list_empty(&nosave_regions)) | |
991 | return; | |
992 | ||
993 | list_for_each_entry(region, &nosave_regions, list) { | |
994 | unsigned long pfn; | |
995 | ||
69f1d475 BH |
996 | pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n", |
997 | (unsigned long long) region->start_pfn << PAGE_SHIFT, | |
998 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
999 | - 1); | |
74dfd666 RW |
1000 | |
1001 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
1002 | if (pfn_valid(pfn)) { |
1003 | /* | |
1004 | * It is safe to ignore the result of | |
1005 | * mem_bm_set_bit_check() here, since we won't | |
1006 | * touch the PFNs for which the error is | |
1007 | * returned anyway. | |
1008 | */ | |
1009 | mem_bm_set_bit_check(bm, pfn); | |
1010 | } | |
74dfd666 RW |
1011 | } |
1012 | } | |
1013 | ||
1014 | /** | |
1015 | * create_basic_memory_bitmaps - create bitmaps needed for marking page | |
1016 | * frames that should not be saved and free page frames. The pointers | |
1017 | * forbidden_pages_map and free_pages_map are only modified if everything | |
1018 | * goes well, because we don't want the bits to be used before both bitmaps | |
1019 | * are set up. | |
1020 | */ | |
1021 | ||
1022 | int create_basic_memory_bitmaps(void) | |
1023 | { | |
1024 | struct memory_bitmap *bm1, *bm2; | |
1025 | int error = 0; | |
1026 | ||
aab17289 RW |
1027 | if (forbidden_pages_map && free_pages_map) |
1028 | return 0; | |
1029 | else | |
1030 | BUG_ON(forbidden_pages_map || free_pages_map); | |
74dfd666 | 1031 | |
0709db60 | 1032 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1033 | if (!bm1) |
1034 | return -ENOMEM; | |
1035 | ||
0709db60 | 1036 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1037 | if (error) |
1038 | goto Free_first_object; | |
1039 | ||
0709db60 | 1040 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1041 | if (!bm2) |
1042 | goto Free_first_bitmap; | |
1043 | ||
0709db60 | 1044 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1045 | if (error) |
1046 | goto Free_second_object; | |
1047 | ||
1048 | forbidden_pages_map = bm1; | |
1049 | free_pages_map = bm2; | |
1050 | mark_nosave_pages(forbidden_pages_map); | |
1051 | ||
23976728 | 1052 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
1053 | |
1054 | return 0; | |
1055 | ||
1056 | Free_second_object: | |
1057 | kfree(bm2); | |
1058 | Free_first_bitmap: | |
1059 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1060 | Free_first_object: | |
1061 | kfree(bm1); | |
1062 | return -ENOMEM; | |
1063 | } | |
1064 | ||
1065 | /** | |
1066 | * free_basic_memory_bitmaps - free memory bitmaps allocated by | |
1067 | * create_basic_memory_bitmaps(). The auxiliary pointers are necessary | |
1068 | * so that the bitmaps themselves are not referred to while they are being | |
1069 | * freed. | |
1070 | */ | |
1071 | ||
1072 | void free_basic_memory_bitmaps(void) | |
1073 | { | |
1074 | struct memory_bitmap *bm1, *bm2; | |
1075 | ||
6a0c7cd3 RW |
1076 | if (WARN_ON(!(forbidden_pages_map && free_pages_map))) |
1077 | return; | |
74dfd666 RW |
1078 | |
1079 | bm1 = forbidden_pages_map; | |
1080 | bm2 = free_pages_map; | |
1081 | forbidden_pages_map = NULL; | |
1082 | free_pages_map = NULL; | |
1083 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1084 | kfree(bm1); | |
1085 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
1086 | kfree(bm2); | |
1087 | ||
23976728 | 1088 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
1089 | } |
1090 | ||
b788db79 RW |
1091 | /** |
1092 | * snapshot_additional_pages - estimate the number of additional pages | |
1093 | * be needed for setting up the suspend image data structures for given | |
1094 | * zone (usually the returned value is greater than the exact number) | |
1095 | */ | |
1096 | ||
1097 | unsigned int snapshot_additional_pages(struct zone *zone) | |
1098 | { | |
f469f02d | 1099 | unsigned int rtree, nodes; |
b788db79 | 1100 | |
f469f02d JR |
1101 | rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); |
1102 | rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node), | |
1103 | LINKED_PAGE_DATA_SIZE); | |
1104 | while (nodes > 1) { | |
1105 | nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL); | |
1106 | rtree += nodes; | |
1107 | } | |
1108 | ||
9047eb62 | 1109 | return 2 * rtree; |
b788db79 RW |
1110 | } |
1111 | ||
8357376d RW |
1112 | #ifdef CONFIG_HIGHMEM |
1113 | /** | |
1114 | * count_free_highmem_pages - compute the total number of free highmem | |
1115 | * pages, system-wide. | |
1116 | */ | |
1117 | ||
1118 | static unsigned int count_free_highmem_pages(void) | |
1119 | { | |
1120 | struct zone *zone; | |
1121 | unsigned int cnt = 0; | |
1122 | ||
ee99c71c KM |
1123 | for_each_populated_zone(zone) |
1124 | if (is_highmem(zone)) | |
d23ad423 | 1125 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
1126 | |
1127 | return cnt; | |
1128 | } | |
1129 | ||
1130 | /** | |
1131 | * saveable_highmem_page - Determine whether a highmem page should be | |
1132 | * included in the suspend image. | |
1133 | * | |
1134 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, | |
1135 | * and it isn't a part of a free chunk of pages. | |
1136 | */ | |
846705de | 1137 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
1138 | { |
1139 | struct page *page; | |
1140 | ||
1141 | if (!pfn_valid(pfn)) | |
1142 | return NULL; | |
1143 | ||
1144 | page = pfn_to_page(pfn); | |
846705de RW |
1145 | if (page_zone(page) != zone) |
1146 | return NULL; | |
8357376d RW |
1147 | |
1148 | BUG_ON(!PageHighMem(page)); | |
1149 | ||
7be98234 RW |
1150 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
1151 | PageReserved(page)) | |
8357376d RW |
1152 | return NULL; |
1153 | ||
c6968e73 SG |
1154 | if (page_is_guard(page)) |
1155 | return NULL; | |
1156 | ||
8357376d RW |
1157 | return page; |
1158 | } | |
1159 | ||
1160 | /** | |
1161 | * count_highmem_pages - compute the total number of saveable highmem | |
1162 | * pages. | |
1163 | */ | |
1164 | ||
fe419535 | 1165 | static unsigned int count_highmem_pages(void) |
8357376d RW |
1166 | { |
1167 | struct zone *zone; | |
1168 | unsigned int n = 0; | |
1169 | ||
98e73dc5 | 1170 | for_each_populated_zone(zone) { |
8357376d RW |
1171 | unsigned long pfn, max_zone_pfn; |
1172 | ||
1173 | if (!is_highmem(zone)) | |
1174 | continue; | |
1175 | ||
1176 | mark_free_pages(zone); | |
c33bc315 | 1177 | max_zone_pfn = zone_end_pfn(zone); |
8357376d | 1178 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1179 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
1180 | n++; |
1181 | } | |
1182 | return n; | |
1183 | } | |
1184 | #else | |
846705de RW |
1185 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
1186 | { | |
1187 | return NULL; | |
1188 | } | |
8357376d RW |
1189 | #endif /* CONFIG_HIGHMEM */ |
1190 | ||
25761b6e | 1191 | /** |
8a235efa RW |
1192 | * saveable_page - Determine whether a non-highmem page should be included |
1193 | * in the suspend image. | |
25761b6e | 1194 | * |
8357376d RW |
1195 | * We should save the page if it isn't Nosave, and is not in the range |
1196 | * of pages statically defined as 'unsaveable', and it isn't a part of | |
1197 | * a free chunk of pages. | |
25761b6e | 1198 | */ |
846705de | 1199 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 1200 | { |
de491861 | 1201 | struct page *page; |
25761b6e RW |
1202 | |
1203 | if (!pfn_valid(pfn)) | |
ae83c5ee | 1204 | return NULL; |
25761b6e RW |
1205 | |
1206 | page = pfn_to_page(pfn); | |
846705de RW |
1207 | if (page_zone(page) != zone) |
1208 | return NULL; | |
ae83c5ee | 1209 | |
8357376d RW |
1210 | BUG_ON(PageHighMem(page)); |
1211 | ||
7be98234 | 1212 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 1213 | return NULL; |
8357376d | 1214 | |
8a235efa RW |
1215 | if (PageReserved(page) |
1216 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 1217 | return NULL; |
25761b6e | 1218 | |
c6968e73 SG |
1219 | if (page_is_guard(page)) |
1220 | return NULL; | |
1221 | ||
ae83c5ee | 1222 | return page; |
25761b6e RW |
1223 | } |
1224 | ||
8357376d RW |
1225 | /** |
1226 | * count_data_pages - compute the total number of saveable non-highmem | |
1227 | * pages. | |
1228 | */ | |
1229 | ||
fe419535 | 1230 | static unsigned int count_data_pages(void) |
25761b6e RW |
1231 | { |
1232 | struct zone *zone; | |
ae83c5ee | 1233 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 1234 | unsigned int n = 0; |
25761b6e | 1235 | |
98e73dc5 | 1236 | for_each_populated_zone(zone) { |
25761b6e RW |
1237 | if (is_highmem(zone)) |
1238 | continue; | |
8357376d | 1239 | |
25761b6e | 1240 | mark_free_pages(zone); |
c33bc315 | 1241 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee | 1242 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1243 | if (saveable_page(zone, pfn)) |
8357376d | 1244 | n++; |
25761b6e | 1245 | } |
a0f49651 | 1246 | return n; |
25761b6e RW |
1247 | } |
1248 | ||
8357376d RW |
1249 | /* This is needed, because copy_page and memcpy are not usable for copying |
1250 | * task structs. | |
1251 | */ | |
1252 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
1253 | { |
1254 | int n; | |
1255 | ||
f623f0db RW |
1256 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
1257 | *dst++ = *src++; | |
1258 | } | |
1259 | ||
8a235efa RW |
1260 | |
1261 | /** | |
1262 | * safe_copy_page - check if the page we are going to copy is marked as | |
1263 | * present in the kernel page tables (this always is the case if | |
1264 | * CONFIG_DEBUG_PAGEALLOC is not set and in that case | |
1265 | * kernel_page_present() always returns 'true'). | |
1266 | */ | |
1267 | static void safe_copy_page(void *dst, struct page *s_page) | |
1268 | { | |
1269 | if (kernel_page_present(s_page)) { | |
1270 | do_copy_page(dst, page_address(s_page)); | |
1271 | } else { | |
1272 | kernel_map_pages(s_page, 1, 1); | |
1273 | do_copy_page(dst, page_address(s_page)); | |
1274 | kernel_map_pages(s_page, 1, 0); | |
1275 | } | |
1276 | } | |
1277 | ||
1278 | ||
8357376d | 1279 | #ifdef CONFIG_HIGHMEM |
efd5a852 | 1280 | static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn) |
8357376d RW |
1281 | { |
1282 | return is_highmem(zone) ? | |
846705de | 1283 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
1284 | } |
1285 | ||
8a235efa | 1286 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
1287 | { |
1288 | struct page *s_page, *d_page; | |
1289 | void *src, *dst; | |
1290 | ||
1291 | s_page = pfn_to_page(src_pfn); | |
1292 | d_page = pfn_to_page(dst_pfn); | |
1293 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1294 | src = kmap_atomic(s_page); |
1295 | dst = kmap_atomic(d_page); | |
8357376d | 1296 | do_copy_page(dst, src); |
0de9a1e2 CW |
1297 | kunmap_atomic(dst); |
1298 | kunmap_atomic(src); | |
8357376d | 1299 | } else { |
8357376d RW |
1300 | if (PageHighMem(d_page)) { |
1301 | /* Page pointed to by src may contain some kernel | |
1302 | * data modified by kmap_atomic() | |
1303 | */ | |
8a235efa | 1304 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1305 | dst = kmap_atomic(d_page); |
3ecb01df | 1306 | copy_page(dst, buffer); |
0de9a1e2 | 1307 | kunmap_atomic(dst); |
8357376d | 1308 | } else { |
8a235efa | 1309 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1310 | } |
1311 | } | |
1312 | } | |
1313 | #else | |
846705de | 1314 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1315 | |
8a235efa | 1316 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1317 | { |
8a235efa RW |
1318 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1319 | pfn_to_page(src_pfn)); | |
8357376d RW |
1320 | } |
1321 | #endif /* CONFIG_HIGHMEM */ | |
1322 | ||
efd5a852 RW |
1323 | static void copy_data_pages(struct memory_bitmap *copy_bm, |
1324 | struct memory_bitmap *orig_bm) | |
25761b6e RW |
1325 | { |
1326 | struct zone *zone; | |
b788db79 | 1327 | unsigned long pfn; |
25761b6e | 1328 | |
98e73dc5 | 1329 | for_each_populated_zone(zone) { |
b788db79 RW |
1330 | unsigned long max_zone_pfn; |
1331 | ||
25761b6e | 1332 | mark_free_pages(zone); |
c33bc315 | 1333 | max_zone_pfn = zone_end_pfn(zone); |
b788db79 | 1334 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1335 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1336 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1337 | } |
b788db79 RW |
1338 | memory_bm_position_reset(orig_bm); |
1339 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1340 | for(;;) { |
b788db79 | 1341 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1342 | if (unlikely(pfn == BM_END_OF_MAP)) |
1343 | break; | |
1344 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1345 | } | |
25761b6e RW |
1346 | } |
1347 | ||
8357376d RW |
1348 | /* Total number of image pages */ |
1349 | static unsigned int nr_copy_pages; | |
1350 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1351 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1352 | /* |
1353 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1354 | * before suspending devices. | |
1355 | */ | |
1356 | unsigned int alloc_normal, alloc_highmem; | |
1357 | /* | |
1358 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1359 | * hibernation image pages (during restore) | |
1360 | */ | |
1361 | static struct memory_bitmap orig_bm; | |
1362 | /* | |
1363 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1364 | * will contain copies of saveable pages. During restore it is initially used | |
1365 | * for marking hibernation image pages, but then the set bits from it are | |
1366 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1367 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1368 | * this purpose. | |
1369 | */ | |
1370 | static struct memory_bitmap copy_bm; | |
8357376d | 1371 | |
25761b6e | 1372 | /** |
940864dd | 1373 | * swsusp_free - free pages allocated for the suspend. |
cd560bb2 | 1374 | * |
940864dd RW |
1375 | * Suspend pages are alocated before the atomic copy is made, so we |
1376 | * need to release them after the resume. | |
25761b6e RW |
1377 | */ |
1378 | ||
1379 | void swsusp_free(void) | |
1380 | { | |
fdd64ed5 | 1381 | unsigned long fb_pfn, fr_pfn; |
6efde38f | 1382 | |
fdd64ed5 JR |
1383 | if (!forbidden_pages_map || !free_pages_map) |
1384 | goto out; | |
1385 | ||
1386 | memory_bm_position_reset(forbidden_pages_map); | |
1387 | memory_bm_position_reset(free_pages_map); | |
1388 | ||
1389 | loop: | |
1390 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1391 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1392 | ||
1393 | /* | |
1394 | * Find the next bit set in both bitmaps. This is guaranteed to | |
1395 | * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP. | |
1396 | */ | |
1397 | do { | |
1398 | if (fb_pfn < fr_pfn) | |
1399 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1400 | if (fr_pfn < fb_pfn) | |
1401 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1402 | } while (fb_pfn != fr_pfn); | |
1403 | ||
1404 | if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) { | |
1405 | struct page *page = pfn_to_page(fr_pfn); | |
1406 | ||
1407 | memory_bm_clear_current(forbidden_pages_map); | |
1408 | memory_bm_clear_current(free_pages_map); | |
1409 | __free_page(page); | |
1410 | goto loop; | |
25761b6e | 1411 | } |
fdd64ed5 JR |
1412 | |
1413 | out: | |
f577eb30 RW |
1414 | nr_copy_pages = 0; |
1415 | nr_meta_pages = 0; | |
75534b50 | 1416 | restore_pblist = NULL; |
6e1819d6 | 1417 | buffer = NULL; |
64a473cb RW |
1418 | alloc_normal = 0; |
1419 | alloc_highmem = 0; | |
25761b6e RW |
1420 | } |
1421 | ||
4bb33435 RW |
1422 | /* Helper functions used for the shrinking of memory. */ |
1423 | ||
1424 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1425 | ||
fe419535 | 1426 | /** |
4bb33435 RW |
1427 | * preallocate_image_pages - Allocate a number of pages for hibernation image |
1428 | * @nr_pages: Number of page frames to allocate. | |
1429 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1430 | * |
4bb33435 RW |
1431 | * Return value: Number of page frames actually allocated |
1432 | */ | |
1433 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1434 | { | |
1435 | unsigned long nr_alloc = 0; | |
1436 | ||
1437 | while (nr_pages > 0) { | |
64a473cb RW |
1438 | struct page *page; |
1439 | ||
1440 | page = alloc_image_page(mask); | |
1441 | if (!page) | |
4bb33435 | 1442 | break; |
64a473cb RW |
1443 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1444 | if (PageHighMem(page)) | |
1445 | alloc_highmem++; | |
1446 | else | |
1447 | alloc_normal++; | |
4bb33435 RW |
1448 | nr_pages--; |
1449 | nr_alloc++; | |
1450 | } | |
1451 | ||
1452 | return nr_alloc; | |
1453 | } | |
1454 | ||
6715045d RW |
1455 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1456 | unsigned long avail_normal) | |
4bb33435 | 1457 | { |
6715045d RW |
1458 | unsigned long alloc; |
1459 | ||
1460 | if (avail_normal <= alloc_normal) | |
1461 | return 0; | |
1462 | ||
1463 | alloc = avail_normal - alloc_normal; | |
1464 | if (nr_pages < alloc) | |
1465 | alloc = nr_pages; | |
1466 | ||
1467 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1468 | } |
1469 | ||
1470 | #ifdef CONFIG_HIGHMEM | |
1471 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1472 | { | |
1473 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1474 | } | |
1475 | ||
1476 | /** | |
1477 | * __fraction - Compute (an approximation of) x * (multiplier / base) | |
fe419535 | 1478 | */ |
4bb33435 RW |
1479 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1480 | { | |
1481 | x *= multiplier; | |
1482 | do_div(x, base); | |
1483 | return (unsigned long)x; | |
1484 | } | |
fe419535 | 1485 | |
4bb33435 | 1486 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
efd5a852 RW |
1487 | unsigned long highmem, |
1488 | unsigned long total) | |
fe419535 | 1489 | { |
4bb33435 RW |
1490 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1491 | ||
1492 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1493 | } |
4bb33435 RW |
1494 | #else /* CONFIG_HIGHMEM */ |
1495 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1496 | { | |
1497 | return 0; | |
1498 | } | |
1499 | ||
1500 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
efd5a852 RW |
1501 | unsigned long highmem, |
1502 | unsigned long total) | |
4bb33435 RW |
1503 | { |
1504 | return 0; | |
1505 | } | |
1506 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1507 | |
4bb33435 | 1508 | /** |
64a473cb RW |
1509 | * free_unnecessary_pages - Release preallocated pages not needed for the image |
1510 | */ | |
a64fc82c | 1511 | static unsigned long free_unnecessary_pages(void) |
64a473cb | 1512 | { |
a64fc82c | 1513 | unsigned long save, to_free_normal, to_free_highmem, free; |
64a473cb | 1514 | |
6715045d RW |
1515 | save = count_data_pages(); |
1516 | if (alloc_normal >= save) { | |
1517 | to_free_normal = alloc_normal - save; | |
1518 | save = 0; | |
1519 | } else { | |
1520 | to_free_normal = 0; | |
1521 | save -= alloc_normal; | |
1522 | } | |
1523 | save += count_highmem_pages(); | |
1524 | if (alloc_highmem >= save) { | |
1525 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1526 | } else { |
1527 | to_free_highmem = 0; | |
4d4cf23c RW |
1528 | save -= alloc_highmem; |
1529 | if (to_free_normal > save) | |
1530 | to_free_normal -= save; | |
1531 | else | |
1532 | to_free_normal = 0; | |
64a473cb | 1533 | } |
a64fc82c | 1534 | free = to_free_normal + to_free_highmem; |
64a473cb RW |
1535 | |
1536 | memory_bm_position_reset(©_bm); | |
1537 | ||
a9c9b442 | 1538 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1539 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1540 | struct page *page = pfn_to_page(pfn); | |
1541 | ||
1542 | if (PageHighMem(page)) { | |
1543 | if (!to_free_highmem) | |
1544 | continue; | |
1545 | to_free_highmem--; | |
1546 | alloc_highmem--; | |
1547 | } else { | |
1548 | if (!to_free_normal) | |
1549 | continue; | |
1550 | to_free_normal--; | |
1551 | alloc_normal--; | |
1552 | } | |
1553 | memory_bm_clear_bit(©_bm, pfn); | |
1554 | swsusp_unset_page_forbidden(page); | |
1555 | swsusp_unset_page_free(page); | |
1556 | __free_page(page); | |
1557 | } | |
a64fc82c WK |
1558 | |
1559 | return free; | |
64a473cb RW |
1560 | } |
1561 | ||
ef4aede3 RW |
1562 | /** |
1563 | * minimum_image_size - Estimate the minimum acceptable size of an image | |
1564 | * @saveable: Number of saveable pages in the system. | |
1565 | * | |
1566 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1567 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1568 | * preallocating memory. | |
1569 | * | |
1570 | * We assume that the minimum image size should be proportional to | |
1571 | * | |
1572 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1573 | * | |
1574 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
4d434820 | 1575 | * and (3) inactive anonymous pages, (4) active and (5) inactive file pages, |
ef4aede3 RW |
1576 | * minus mapped file pages. |
1577 | */ | |
1578 | static unsigned long minimum_image_size(unsigned long saveable) | |
1579 | { | |
1580 | unsigned long size; | |
1581 | ||
1582 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
1583 | + global_page_state(NR_ACTIVE_ANON) | |
1584 | + global_page_state(NR_INACTIVE_ANON) | |
1585 | + global_page_state(NR_ACTIVE_FILE) | |
1586 | + global_page_state(NR_INACTIVE_FILE) | |
1587 | - global_page_state(NR_FILE_MAPPED); | |
1588 | ||
1589 | return saveable <= size ? 0 : saveable - size; | |
1590 | } | |
1591 | ||
64a473cb RW |
1592 | /** |
1593 | * hibernate_preallocate_memory - Preallocate memory for hibernation image | |
4bb33435 RW |
1594 | * |
1595 | * To create a hibernation image it is necessary to make a copy of every page | |
1596 | * frame in use. We also need a number of page frames to be free during | |
1597 | * hibernation for allocations made while saving the image and for device | |
1598 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 RW |
1599 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
1600 | * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through | |
1601 | * /sys/power/reserved_size, respectively). To make this happen, we compute the | |
1602 | * total number of available page frames and allocate at least | |
4bb33435 | 1603 | * |
ddeb6487 RW |
1604 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1605 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1606 | * |
1607 | * of them, which corresponds to the maximum size of a hibernation image. | |
1608 | * | |
1609 | * If image_size is set below the number following from the above formula, | |
1610 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1611 | * pages in the system is below the requested image size or the minimum |
1612 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1613 | */ |
64a473cb | 1614 | int hibernate_preallocate_memory(void) |
fe419535 | 1615 | { |
fe419535 | 1616 | struct zone *zone; |
4bb33435 | 1617 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1618 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
db597605 | 1619 | ktime_t start, stop; |
64a473cb | 1620 | int error; |
fe419535 | 1621 | |
64a473cb | 1622 | printk(KERN_INFO "PM: Preallocating image memory... "); |
db597605 | 1623 | start = ktime_get(); |
fe419535 | 1624 | |
64a473cb RW |
1625 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1626 | if (error) | |
1627 | goto err_out; | |
1628 | ||
1629 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1630 | if (error) | |
1631 | goto err_out; | |
1632 | ||
1633 | alloc_normal = 0; | |
1634 | alloc_highmem = 0; | |
1635 | ||
4bb33435 | 1636 | /* Count the number of saveable data pages. */ |
64a473cb | 1637 | save_highmem = count_highmem_pages(); |
4bb33435 | 1638 | saveable = count_data_pages(); |
fe419535 | 1639 | |
4bb33435 RW |
1640 | /* |
1641 | * Compute the total number of page frames we can use (count) and the | |
1642 | * number of pages needed for image metadata (size). | |
1643 | */ | |
1644 | count = saveable; | |
64a473cb RW |
1645 | saveable += save_highmem; |
1646 | highmem = save_highmem; | |
4bb33435 RW |
1647 | size = 0; |
1648 | for_each_populated_zone(zone) { | |
1649 | size += snapshot_additional_pages(zone); | |
1650 | if (is_highmem(zone)) | |
1651 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1652 | else | |
1653 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1654 | } | |
6715045d | 1655 | avail_normal = count; |
4bb33435 RW |
1656 | count += highmem; |
1657 | count -= totalreserve_pages; | |
1658 | ||
85055dd8 MS |
1659 | /* Add number of pages required for page keys (s390 only). */ |
1660 | size += page_key_additional_pages(saveable); | |
1661 | ||
4bb33435 | 1662 | /* Compute the maximum number of saveable pages to leave in memory. */ |
ddeb6487 RW |
1663 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1664 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1665 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1666 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1667 | if (size > max_size) | |
1668 | size = max_size; | |
1669 | /* | |
266f1a25 RW |
1670 | * If the desired number of image pages is at least as large as the |
1671 | * current number of saveable pages in memory, allocate page frames for | |
1672 | * the image and we're done. | |
4bb33435 | 1673 | */ |
64a473cb RW |
1674 | if (size >= saveable) { |
1675 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1676 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1677 | goto out; |
64a473cb | 1678 | } |
4bb33435 | 1679 | |
ef4aede3 RW |
1680 | /* Estimate the minimum size of the image. */ |
1681 | pages = minimum_image_size(saveable); | |
6715045d RW |
1682 | /* |
1683 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1684 | * accommodate an image of the minimum size (unless it's already too | |
1685 | * small, in which case don't preallocate pages from it at all). | |
1686 | */ | |
1687 | if (avail_normal > pages) | |
1688 | avail_normal -= pages; | |
1689 | else | |
1690 | avail_normal = 0; | |
ef4aede3 RW |
1691 | if (size < pages) |
1692 | size = min_t(unsigned long, pages, max_size); | |
1693 | ||
4bb33435 RW |
1694 | /* |
1695 | * Let the memory management subsystem know that we're going to need a | |
1696 | * large number of page frames to allocate and make it free some memory. | |
1697 | * NOTE: If this is not done, performance will be hurt badly in some | |
1698 | * test cases. | |
1699 | */ | |
1700 | shrink_all_memory(saveable - size); | |
1701 | ||
1702 | /* | |
1703 | * The number of saveable pages in memory was too high, so apply some | |
1704 | * pressure to decrease it. First, make room for the largest possible | |
1705 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1706 | * of the image as much as indicated by 'size' using allocations from |
1707 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1708 | */ |
1709 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
fd432b9f AL |
1710 | alloc = count - max_size; |
1711 | if (alloc > pages_highmem) | |
1712 | alloc -= pages_highmem; | |
1713 | else | |
1714 | alloc = 0; | |
6715045d RW |
1715 | pages = preallocate_image_memory(alloc, avail_normal); |
1716 | if (pages < alloc) { | |
1717 | /* We have exhausted non-highmem pages, try highmem. */ | |
1718 | alloc -= pages; | |
1719 | pages += pages_highmem; | |
1720 | pages_highmem = preallocate_image_highmem(alloc); | |
1721 | if (pages_highmem < alloc) | |
1722 | goto err_out; | |
1723 | pages += pages_highmem; | |
1724 | /* | |
1725 | * size is the desired number of saveable pages to leave in | |
1726 | * memory, so try to preallocate (all memory - size) pages. | |
1727 | */ | |
1728 | alloc = (count - pages) - size; | |
1729 | pages += preallocate_image_highmem(alloc); | |
1730 | } else { | |
1731 | /* | |
1732 | * There are approximately max_size saveable pages at this point | |
1733 | * and we want to reduce this number down to size. | |
1734 | */ | |
1735 | alloc = max_size - size; | |
1736 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1737 | pages_highmem += size; | |
1738 | alloc -= size; | |
1739 | size = preallocate_image_memory(alloc, avail_normal); | |
1740 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1741 | pages += pages_highmem + size; | |
1742 | } | |
4bb33435 | 1743 | |
64a473cb RW |
1744 | /* |
1745 | * We only need as many page frames for the image as there are saveable | |
1746 | * pages in memory, but we have allocated more. Release the excessive | |
1747 | * ones now. | |
1748 | */ | |
a64fc82c | 1749 | pages -= free_unnecessary_pages(); |
4bb33435 RW |
1750 | |
1751 | out: | |
db597605 | 1752 | stop = ktime_get(); |
64a473cb | 1753 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
db597605 | 1754 | swsusp_show_speed(start, stop, pages, "Allocated"); |
fe419535 RW |
1755 | |
1756 | return 0; | |
64a473cb RW |
1757 | |
1758 | err_out: | |
1759 | printk(KERN_CONT "\n"); | |
1760 | swsusp_free(); | |
1761 | return -ENOMEM; | |
fe419535 RW |
1762 | } |
1763 | ||
8357376d RW |
1764 | #ifdef CONFIG_HIGHMEM |
1765 | /** | |
1766 | * count_pages_for_highmem - compute the number of non-highmem pages | |
1767 | * that will be necessary for creating copies of highmem pages. | |
1768 | */ | |
1769 | ||
1770 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) | |
1771 | { | |
64a473cb | 1772 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1773 | |
1774 | if (free_highmem >= nr_highmem) | |
1775 | nr_highmem = 0; | |
1776 | else | |
1777 | nr_highmem -= free_highmem; | |
1778 | ||
1779 | return nr_highmem; | |
1780 | } | |
1781 | #else | |
efd5a852 | 1782 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; } |
8357376d | 1783 | #endif /* CONFIG_HIGHMEM */ |
25761b6e RW |
1784 | |
1785 | /** | |
8357376d RW |
1786 | * enough_free_mem - Make sure we have enough free memory for the |
1787 | * snapshot image. | |
25761b6e RW |
1788 | */ |
1789 | ||
8357376d | 1790 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1791 | { |
e5e2fa78 | 1792 | struct zone *zone; |
64a473cb | 1793 | unsigned int free = alloc_normal; |
e5e2fa78 | 1794 | |
98e73dc5 | 1795 | for_each_populated_zone(zone) |
8357376d | 1796 | if (!is_highmem(zone)) |
d23ad423 | 1797 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1798 | |
8357376d | 1799 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1800 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1801 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1802 | |
64a473cb | 1803 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1804 | } |
1805 | ||
8357376d RW |
1806 | #ifdef CONFIG_HIGHMEM |
1807 | /** | |
1808 | * get_highmem_buffer - if there are some highmem pages in the suspend | |
1809 | * image, we may need the buffer to copy them and/or load their data. | |
1810 | */ | |
1811 | ||
1812 | static inline int get_highmem_buffer(int safe_needed) | |
1813 | { | |
1814 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1815 | return buffer ? 0 : -ENOMEM; | |
1816 | } | |
1817 | ||
1818 | /** | |
1819 | * alloc_highmem_image_pages - allocate some highmem pages for the image. | |
1820 | * Try to allocate as many pages as needed, but if the number of free | |
1821 | * highmem pages is lesser than that, allocate them all. | |
1822 | */ | |
1823 | ||
efd5a852 RW |
1824 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1825 | unsigned int nr_highmem) | |
8357376d RW |
1826 | { |
1827 | unsigned int to_alloc = count_free_highmem_pages(); | |
1828 | ||
1829 | if (to_alloc > nr_highmem) | |
1830 | to_alloc = nr_highmem; | |
1831 | ||
1832 | nr_highmem -= to_alloc; | |
1833 | while (to_alloc-- > 0) { | |
1834 | struct page *page; | |
1835 | ||
d0164adc | 1836 | page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM); |
8357376d RW |
1837 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1838 | } | |
1839 | return nr_highmem; | |
1840 | } | |
1841 | #else | |
1842 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1843 | ||
efd5a852 RW |
1844 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1845 | unsigned int n) { return 0; } | |
8357376d RW |
1846 | #endif /* CONFIG_HIGHMEM */ |
1847 | ||
1848 | /** | |
1849 | * swsusp_alloc - allocate memory for the suspend image | |
1850 | * | |
1851 | * We first try to allocate as many highmem pages as there are | |
1852 | * saveable highmem pages in the system. If that fails, we allocate | |
1853 | * non-highmem pages for the copies of the remaining highmem ones. | |
1854 | * | |
1855 | * In this approach it is likely that the copies of highmem pages will | |
1856 | * also be located in the high memory, because of the way in which | |
1857 | * copy_data_pages() works. | |
1858 | */ | |
1859 | ||
efd5a852 RW |
1860 | static int swsusp_alloc(struct memory_bitmap *orig_bm, |
1861 | struct memory_bitmap *copy_bm, | |
1862 | unsigned int nr_pages, unsigned int nr_highmem) | |
054bd4c1 | 1863 | { |
8357376d | 1864 | if (nr_highmem > 0) { |
2e725a06 | 1865 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1866 | goto err_out; |
1867 | if (nr_highmem > alloc_highmem) { | |
1868 | nr_highmem -= alloc_highmem; | |
1869 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1870 | } | |
8357376d | 1871 | } |
64a473cb RW |
1872 | if (nr_pages > alloc_normal) { |
1873 | nr_pages -= alloc_normal; | |
1874 | while (nr_pages-- > 0) { | |
1875 | struct page *page; | |
1876 | ||
1877 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1878 | if (!page) | |
1879 | goto err_out; | |
1880 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1881 | } | |
25761b6e | 1882 | } |
64a473cb | 1883 | |
b788db79 | 1884 | return 0; |
25761b6e | 1885 | |
64a473cb | 1886 | err_out: |
b788db79 | 1887 | swsusp_free(); |
2e725a06 | 1888 | return -ENOMEM; |
25761b6e RW |
1889 | } |
1890 | ||
722a9f92 | 1891 | asmlinkage __visible int swsusp_save(void) |
25761b6e | 1892 | { |
8357376d | 1893 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1894 | |
07c3bb57 | 1895 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1896 | |
9f8f2172 | 1897 | drain_local_pages(NULL); |
a0f49651 | 1898 | nr_pages = count_data_pages(); |
8357376d | 1899 | nr_highmem = count_highmem_pages(); |
23976728 | 1900 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1901 | |
8357376d | 1902 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1903 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1904 | return -ENOMEM; |
1905 | } | |
1906 | ||
8357376d | 1907 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1908 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1909 | return -ENOMEM; |
8357376d | 1910 | } |
25761b6e RW |
1911 | |
1912 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1913 | * Kill them. | |
1914 | */ | |
9f8f2172 | 1915 | drain_local_pages(NULL); |
b788db79 | 1916 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1917 | |
1918 | /* | |
1919 | * End of critical section. From now on, we can write to memory, | |
1920 | * but we should not touch disk. This specially means we must _not_ | |
1921 | * touch swap space! Except we must write out our image of course. | |
1922 | */ | |
1923 | ||
8357376d | 1924 | nr_pages += nr_highmem; |
a0f49651 | 1925 | nr_copy_pages = nr_pages; |
8357376d | 1926 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 1927 | |
23976728 RW |
1928 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1929 | nr_pages); | |
8357376d | 1930 | |
25761b6e RW |
1931 | return 0; |
1932 | } | |
f577eb30 | 1933 | |
d307c4a8 RW |
1934 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1935 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 1936 | { |
d307c4a8 | 1937 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 1938 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
1939 | return 0; |
1940 | } | |
1941 | ||
1942 | static char *check_image_kernel(struct swsusp_info *info) | |
1943 | { | |
1944 | if (info->version_code != LINUX_VERSION_CODE) | |
1945 | return "kernel version"; | |
1946 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
1947 | return "system type"; | |
1948 | if (strcmp(info->uts.release,init_utsname()->release)) | |
1949 | return "kernel release"; | |
1950 | if (strcmp(info->uts.version,init_utsname()->version)) | |
1951 | return "version"; | |
1952 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
1953 | return "machine"; | |
1954 | return NULL; | |
1955 | } | |
1956 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
1957 | ||
af508b34 RW |
1958 | unsigned long snapshot_get_image_size(void) |
1959 | { | |
1960 | return nr_copy_pages + nr_meta_pages + 1; | |
1961 | } | |
1962 | ||
d307c4a8 RW |
1963 | static int init_header(struct swsusp_info *info) |
1964 | { | |
1965 | memset(info, 0, sizeof(struct swsusp_info)); | |
0ed5fd13 | 1966 | info->num_physpages = get_num_physpages(); |
f577eb30 | 1967 | info->image_pages = nr_copy_pages; |
af508b34 | 1968 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
1969 | info->size = info->pages; |
1970 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 1971 | return init_header_complete(info); |
f577eb30 RW |
1972 | } |
1973 | ||
1974 | /** | |
940864dd RW |
1975 | * pack_pfns - pfns corresponding to the set bits found in the bitmap @bm |
1976 | * are stored in the array @buf[] (1 page at a time) | |
f577eb30 RW |
1977 | */ |
1978 | ||
efd5a852 | 1979 | static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
1980 | { |
1981 | int j; | |
1982 | ||
b788db79 | 1983 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
1984 | buf[j] = memory_bm_next_pfn(bm); |
1985 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 1986 | break; |
85055dd8 MS |
1987 | /* Save page key for data page (s390 only). */ |
1988 | page_key_read(buf + j); | |
f577eb30 | 1989 | } |
f577eb30 RW |
1990 | } |
1991 | ||
1992 | /** | |
1993 | * snapshot_read_next - used for reading the system memory snapshot. | |
1994 | * | |
1995 | * On the first call to it @handle should point to a zeroed | |
1996 | * snapshot_handle structure. The structure gets updated and a pointer | |
1997 | * to it should be passed to this function every next time. | |
1998 | * | |
f577eb30 RW |
1999 | * On success the function returns a positive number. Then, the caller |
2000 | * is allowed to read up to the returned number of bytes from the memory | |
d3c1b24c | 2001 | * location computed by the data_of() macro. |
f577eb30 RW |
2002 | * |
2003 | * The function returns 0 to indicate the end of data stream condition, | |
2004 | * and a negative number is returned on error. In such cases the | |
2005 | * structure pointed to by @handle is not updated and should not be used | |
2006 | * any more. | |
2007 | */ | |
2008 | ||
d3c1b24c | 2009 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 2010 | { |
fb13a28b | 2011 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2012 | return 0; |
b788db79 | 2013 | |
f577eb30 RW |
2014 | if (!buffer) { |
2015 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 2016 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
2017 | if (!buffer) |
2018 | return -ENOMEM; | |
2019 | } | |
d3c1b24c | 2020 | if (!handle->cur) { |
d307c4a8 RW |
2021 | int error; |
2022 | ||
2023 | error = init_header((struct swsusp_info *)buffer); | |
2024 | if (error) | |
2025 | return error; | |
f577eb30 | 2026 | handle->buffer = buffer; |
b788db79 RW |
2027 | memory_bm_position_reset(&orig_bm); |
2028 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 2029 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 2030 | clear_page(buffer); |
d3c1b24c JS |
2031 | pack_pfns(buffer, &orig_bm); |
2032 | } else { | |
2033 | struct page *page; | |
b788db79 | 2034 | |
d3c1b24c JS |
2035 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
2036 | if (PageHighMem(page)) { | |
2037 | /* Highmem pages are copied to the buffer, | |
2038 | * because we can't return with a kmapped | |
2039 | * highmem page (we may not be called again). | |
2040 | */ | |
2041 | void *kaddr; | |
8357376d | 2042 | |
0de9a1e2 | 2043 | kaddr = kmap_atomic(page); |
3ecb01df | 2044 | copy_page(buffer, kaddr); |
0de9a1e2 | 2045 | kunmap_atomic(kaddr); |
d3c1b24c JS |
2046 | handle->buffer = buffer; |
2047 | } else { | |
2048 | handle->buffer = page_address(page); | |
f577eb30 | 2049 | } |
f577eb30 | 2050 | } |
d3c1b24c JS |
2051 | handle->cur++; |
2052 | return PAGE_SIZE; | |
f577eb30 RW |
2053 | } |
2054 | ||
6dbecfd3 RW |
2055 | static void duplicate_memory_bitmap(struct memory_bitmap *dst, |
2056 | struct memory_bitmap *src) | |
2057 | { | |
2058 | unsigned long pfn; | |
2059 | ||
2060 | memory_bm_position_reset(src); | |
2061 | pfn = memory_bm_next_pfn(src); | |
2062 | while (pfn != BM_END_OF_MAP) { | |
2063 | memory_bm_set_bit(dst, pfn); | |
2064 | pfn = memory_bm_next_pfn(src); | |
2065 | } | |
2066 | } | |
2067 | ||
f577eb30 RW |
2068 | /** |
2069 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
2070 | * the image during resume, because they conflict with the pages that | |
2071 | * had been used before suspend | |
2072 | */ | |
2073 | ||
6dbecfd3 | 2074 | static void mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 | 2075 | { |
6dbecfd3 | 2076 | unsigned long pfn; |
f577eb30 | 2077 | |
6dbecfd3 RW |
2078 | /* Clear the "free"/"unsafe" bit for all PFNs */ |
2079 | memory_bm_position_reset(free_pages_map); | |
2080 | pfn = memory_bm_next_pfn(free_pages_map); | |
2081 | while (pfn != BM_END_OF_MAP) { | |
2082 | memory_bm_clear_current(free_pages_map); | |
2083 | pfn = memory_bm_next_pfn(free_pages_map); | |
f577eb30 RW |
2084 | } |
2085 | ||
6dbecfd3 RW |
2086 | /* Mark pages that correspond to the "original" PFNs as "unsafe" */ |
2087 | duplicate_memory_bitmap(free_pages_map, bm); | |
f577eb30 | 2088 | |
940864dd | 2089 | allocated_unsafe_pages = 0; |
f577eb30 RW |
2090 | } |
2091 | ||
d307c4a8 | 2092 | static int check_header(struct swsusp_info *info) |
f577eb30 | 2093 | { |
d307c4a8 | 2094 | char *reason; |
f577eb30 | 2095 | |
d307c4a8 | 2096 | reason = check_image_kernel(info); |
0ed5fd13 | 2097 | if (!reason && info->num_physpages != get_num_physpages()) |
f577eb30 | 2098 | reason = "memory size"; |
f577eb30 | 2099 | if (reason) { |
23976728 | 2100 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
2101 | return -EPERM; |
2102 | } | |
2103 | return 0; | |
2104 | } | |
2105 | ||
2106 | /** | |
2107 | * load header - check the image header and copy data from it | |
2108 | */ | |
2109 | ||
efd5a852 | 2110 | static int load_header(struct swsusp_info *info) |
f577eb30 RW |
2111 | { |
2112 | int error; | |
f577eb30 | 2113 | |
940864dd | 2114 | restore_pblist = NULL; |
f577eb30 RW |
2115 | error = check_header(info); |
2116 | if (!error) { | |
f577eb30 RW |
2117 | nr_copy_pages = info->image_pages; |
2118 | nr_meta_pages = info->pages - info->image_pages - 1; | |
2119 | } | |
2120 | return error; | |
2121 | } | |
2122 | ||
2123 | /** | |
940864dd RW |
2124 | * unpack_orig_pfns - for each element of @buf[] (1 page at a time) set |
2125 | * the corresponding bit in the memory bitmap @bm | |
f577eb30 | 2126 | */ |
69643279 | 2127 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2128 | { |
2129 | int j; | |
2130 | ||
940864dd RW |
2131 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
2132 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
2133 | break; | |
2134 | ||
85055dd8 MS |
2135 | /* Extract and buffer page key for data page (s390 only). */ |
2136 | page_key_memorize(buf + j); | |
2137 | ||
6dbecfd3 | 2138 | if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j])) |
69643279 RW |
2139 | memory_bm_set_bit(bm, buf[j]); |
2140 | else | |
2141 | return -EFAULT; | |
f577eb30 | 2142 | } |
69643279 RW |
2143 | |
2144 | return 0; | |
f577eb30 RW |
2145 | } |
2146 | ||
8357376d RW |
2147 | #ifdef CONFIG_HIGHMEM |
2148 | /* struct highmem_pbe is used for creating the list of highmem pages that | |
2149 | * should be restored atomically during the resume from disk, because the page | |
2150 | * frames they have occupied before the suspend are in use. | |
2151 | */ | |
2152 | struct highmem_pbe { | |
2153 | struct page *copy_page; /* data is here now */ | |
2154 | struct page *orig_page; /* data was here before the suspend */ | |
2155 | struct highmem_pbe *next; | |
2156 | }; | |
2157 | ||
2158 | /* List of highmem PBEs needed for restoring the highmem pages that were | |
2159 | * allocated before the suspend and included in the suspend image, but have | |
2160 | * also been allocated by the "resume" kernel, so their contents cannot be | |
2161 | * written directly to their "original" page frames. | |
2162 | */ | |
2163 | static struct highmem_pbe *highmem_pblist; | |
2164 | ||
2165 | /** | |
2166 | * count_highmem_image_pages - compute the number of highmem pages in the | |
2167 | * suspend image. The bits in the memory bitmap @bm that correspond to the | |
2168 | * image pages are assumed to be set. | |
2169 | */ | |
2170 | ||
2171 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) | |
2172 | { | |
2173 | unsigned long pfn; | |
2174 | unsigned int cnt = 0; | |
2175 | ||
2176 | memory_bm_position_reset(bm); | |
2177 | pfn = memory_bm_next_pfn(bm); | |
2178 | while (pfn != BM_END_OF_MAP) { | |
2179 | if (PageHighMem(pfn_to_page(pfn))) | |
2180 | cnt++; | |
2181 | ||
2182 | pfn = memory_bm_next_pfn(bm); | |
2183 | } | |
2184 | return cnt; | |
2185 | } | |
2186 | ||
2187 | /** | |
2188 | * prepare_highmem_image - try to allocate as many highmem pages as | |
2189 | * there are highmem image pages (@nr_highmem_p points to the variable | |
2190 | * containing the number of highmem image pages). The pages that are | |
2191 | * "safe" (ie. will not be overwritten when the suspend image is | |
2192 | * restored) have the corresponding bits set in @bm (it must be | |
2193 | * unitialized). | |
2194 | * | |
2195 | * NOTE: This function should not be called if there are no highmem | |
2196 | * image pages. | |
2197 | */ | |
2198 | ||
2199 | static unsigned int safe_highmem_pages; | |
2200 | ||
2201 | static struct memory_bitmap *safe_highmem_bm; | |
2202 | ||
efd5a852 RW |
2203 | static int prepare_highmem_image(struct memory_bitmap *bm, |
2204 | unsigned int *nr_highmem_p) | |
8357376d RW |
2205 | { |
2206 | unsigned int to_alloc; | |
2207 | ||
2208 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
2209 | return -ENOMEM; | |
2210 | ||
2211 | if (get_highmem_buffer(PG_SAFE)) | |
2212 | return -ENOMEM; | |
2213 | ||
2214 | to_alloc = count_free_highmem_pages(); | |
2215 | if (to_alloc > *nr_highmem_p) | |
2216 | to_alloc = *nr_highmem_p; | |
2217 | else | |
2218 | *nr_highmem_p = to_alloc; | |
2219 | ||
2220 | safe_highmem_pages = 0; | |
2221 | while (to_alloc-- > 0) { | |
2222 | struct page *page; | |
2223 | ||
2224 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 2225 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
2226 | /* The page is "safe", set its bit the bitmap */ |
2227 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
2228 | safe_highmem_pages++; | |
2229 | } | |
2230 | /* Mark the page as allocated */ | |
7be98234 RW |
2231 | swsusp_set_page_forbidden(page); |
2232 | swsusp_set_page_free(page); | |
8357376d RW |
2233 | } |
2234 | memory_bm_position_reset(bm); | |
2235 | safe_highmem_bm = bm; | |
2236 | return 0; | |
2237 | } | |
2238 | ||
2239 | /** | |
2240 | * get_highmem_page_buffer - for given highmem image page find the buffer | |
2241 | * that suspend_write_next() should set for its caller to write to. | |
2242 | * | |
2243 | * If the page is to be saved to its "original" page frame or a copy of | |
2244 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
2245 | * the copy of the page is to be made in normal memory, so the address of | |
2246 | * the copy is returned. | |
2247 | * | |
2248 | * If @buffer is returned, the caller of suspend_write_next() will write | |
2249 | * the page's contents to @buffer, so they will have to be copied to the | |
2250 | * right location on the next call to suspend_write_next() and it is done | |
2251 | * with the help of copy_last_highmem_page(). For this purpose, if | |
2252 | * @buffer is returned, @last_highmem page is set to the page to which | |
2253 | * the data will have to be copied from @buffer. | |
2254 | */ | |
2255 | ||
2256 | static struct page *last_highmem_page; | |
2257 | ||
efd5a852 RW |
2258 | static void *get_highmem_page_buffer(struct page *page, |
2259 | struct chain_allocator *ca) | |
8357376d RW |
2260 | { |
2261 | struct highmem_pbe *pbe; | |
2262 | void *kaddr; | |
2263 | ||
7be98234 | 2264 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
8357376d RW |
2265 | /* We have allocated the "original" page frame and we can |
2266 | * use it directly to store the loaded page. | |
2267 | */ | |
2268 | last_highmem_page = page; | |
2269 | return buffer; | |
2270 | } | |
2271 | /* The "original" page frame has not been allocated and we have to | |
2272 | * use a "safe" page frame to store the loaded page. | |
2273 | */ | |
2274 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
2275 | if (!pbe) { | |
2276 | swsusp_free(); | |
69643279 | 2277 | return ERR_PTR(-ENOMEM); |
8357376d RW |
2278 | } |
2279 | pbe->orig_page = page; | |
2280 | if (safe_highmem_pages > 0) { | |
2281 | struct page *tmp; | |
2282 | ||
2283 | /* Copy of the page will be stored in high memory */ | |
2284 | kaddr = buffer; | |
2285 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
2286 | safe_highmem_pages--; | |
2287 | last_highmem_page = tmp; | |
2288 | pbe->copy_page = tmp; | |
2289 | } else { | |
2290 | /* Copy of the page will be stored in normal memory */ | |
2291 | kaddr = safe_pages_list; | |
2292 | safe_pages_list = safe_pages_list->next; | |
2293 | pbe->copy_page = virt_to_page(kaddr); | |
2294 | } | |
2295 | pbe->next = highmem_pblist; | |
2296 | highmem_pblist = pbe; | |
2297 | return kaddr; | |
2298 | } | |
2299 | ||
2300 | /** | |
2301 | * copy_last_highmem_page - copy the contents of a highmem image from | |
2302 | * @buffer, where the caller of snapshot_write_next() has place them, | |
2303 | * to the right location represented by @last_highmem_page . | |
2304 | */ | |
2305 | ||
2306 | static void copy_last_highmem_page(void) | |
2307 | { | |
2308 | if (last_highmem_page) { | |
2309 | void *dst; | |
2310 | ||
0de9a1e2 | 2311 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2312 | copy_page(dst, buffer); |
0de9a1e2 | 2313 | kunmap_atomic(dst); |
8357376d RW |
2314 | last_highmem_page = NULL; |
2315 | } | |
2316 | } | |
2317 | ||
2318 | static inline int last_highmem_page_copied(void) | |
2319 | { | |
2320 | return !last_highmem_page; | |
2321 | } | |
2322 | ||
2323 | static inline void free_highmem_data(void) | |
2324 | { | |
2325 | if (safe_highmem_bm) | |
2326 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2327 | ||
2328 | if (buffer) | |
2329 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2330 | } | |
2331 | #else | |
efd5a852 | 2332 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } |
8357376d | 2333 | |
efd5a852 RW |
2334 | static inline int prepare_highmem_image(struct memory_bitmap *bm, |
2335 | unsigned int *nr_highmem_p) { return 0; } | |
8357376d | 2336 | |
efd5a852 RW |
2337 | static inline void *get_highmem_page_buffer(struct page *page, |
2338 | struct chain_allocator *ca) | |
8357376d | 2339 | { |
69643279 | 2340 | return ERR_PTR(-EINVAL); |
8357376d RW |
2341 | } |
2342 | ||
2343 | static inline void copy_last_highmem_page(void) {} | |
2344 | static inline int last_highmem_page_copied(void) { return 1; } | |
2345 | static inline void free_highmem_data(void) {} | |
2346 | #endif /* CONFIG_HIGHMEM */ | |
2347 | ||
f577eb30 | 2348 | /** |
940864dd RW |
2349 | * prepare_image - use the memory bitmap @bm to mark the pages that will |
2350 | * be overwritten in the process of restoring the system memory state | |
2351 | * from the suspend image ("unsafe" pages) and allocate memory for the | |
2352 | * image. | |
968808b8 | 2353 | * |
940864dd RW |
2354 | * The idea is to allocate a new memory bitmap first and then allocate |
2355 | * as many pages as needed for the image data, but not to assign these | |
2356 | * pages to specific tasks initially. Instead, we just mark them as | |
8357376d RW |
2357 | * allocated and create a lists of "safe" pages that will be used |
2358 | * later. On systems with high memory a list of "safe" highmem pages is | |
2359 | * also created. | |
f577eb30 RW |
2360 | */ |
2361 | ||
940864dd RW |
2362 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2363 | ||
efd5a852 | 2364 | static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) |
f577eb30 | 2365 | { |
8357376d | 2366 | unsigned int nr_pages, nr_highmem; |
9c744481 | 2367 | struct linked_page *lp; |
940864dd | 2368 | int error; |
f577eb30 | 2369 | |
8357376d RW |
2370 | /* If there is no highmem, the buffer will not be necessary */ |
2371 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2372 | buffer = NULL; | |
2373 | ||
2374 | nr_highmem = count_highmem_image_pages(bm); | |
6dbecfd3 | 2375 | mark_unsafe_pages(bm); |
940864dd RW |
2376 | |
2377 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2378 | if (error) | |
2379 | goto Free; | |
2380 | ||
2381 | duplicate_memory_bitmap(new_bm, bm); | |
2382 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2383 | if (nr_highmem > 0) { |
2384 | error = prepare_highmem_image(bm, &nr_highmem); | |
2385 | if (error) | |
2386 | goto Free; | |
2387 | } | |
940864dd RW |
2388 | /* Reserve some safe pages for potential later use. |
2389 | * | |
2390 | * NOTE: This way we make sure there will be enough safe pages for the | |
2391 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2392 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
9c744481 RW |
2393 | * |
2394 | * nr_copy_pages cannot be less than allocated_unsafe_pages too. | |
940864dd | 2395 | */ |
8357376d | 2396 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2397 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2398 | while (nr_pages > 0) { | |
8357376d | 2399 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2400 | if (!lp) { |
f577eb30 | 2401 | error = -ENOMEM; |
940864dd RW |
2402 | goto Free; |
2403 | } | |
9c744481 RW |
2404 | lp->next = safe_pages_list; |
2405 | safe_pages_list = lp; | |
940864dd | 2406 | nr_pages--; |
f577eb30 | 2407 | } |
940864dd | 2408 | /* Preallocate memory for the image */ |
8357376d | 2409 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2410 | while (nr_pages > 0) { |
2411 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2412 | if (!lp) { | |
2413 | error = -ENOMEM; | |
2414 | goto Free; | |
2415 | } | |
7be98234 | 2416 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2417 | /* The page is "safe", add it to the list */ |
2418 | lp->next = safe_pages_list; | |
2419 | safe_pages_list = lp; | |
968808b8 | 2420 | } |
940864dd | 2421 | /* Mark the page as allocated */ |
7be98234 RW |
2422 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2423 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2424 | nr_pages--; |
968808b8 | 2425 | } |
940864dd RW |
2426 | return 0; |
2427 | ||
59a49335 | 2428 | Free: |
940864dd | 2429 | swsusp_free(); |
f577eb30 RW |
2430 | return error; |
2431 | } | |
2432 | ||
940864dd RW |
2433 | /** |
2434 | * get_buffer - compute the address that snapshot_write_next() should | |
2435 | * set for its caller to write to. | |
2436 | */ | |
2437 | ||
2438 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) | |
968808b8 | 2439 | { |
940864dd | 2440 | struct pbe *pbe; |
69643279 RW |
2441 | struct page *page; |
2442 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2443 | |
69643279 RW |
2444 | if (pfn == BM_END_OF_MAP) |
2445 | return ERR_PTR(-EFAULT); | |
2446 | ||
2447 | page = pfn_to_page(pfn); | |
8357376d RW |
2448 | if (PageHighMem(page)) |
2449 | return get_highmem_page_buffer(page, ca); | |
2450 | ||
7be98234 | 2451 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
940864dd RW |
2452 | /* We have allocated the "original" page frame and we can |
2453 | * use it directly to store the loaded page. | |
968808b8 | 2454 | */ |
940864dd RW |
2455 | return page_address(page); |
2456 | ||
2457 | /* The "original" page frame has not been allocated and we have to | |
2458 | * use a "safe" page frame to store the loaded page. | |
968808b8 | 2459 | */ |
940864dd RW |
2460 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2461 | if (!pbe) { | |
2462 | swsusp_free(); | |
69643279 | 2463 | return ERR_PTR(-ENOMEM); |
940864dd | 2464 | } |
8357376d RW |
2465 | pbe->orig_address = page_address(page); |
2466 | pbe->address = safe_pages_list; | |
940864dd RW |
2467 | safe_pages_list = safe_pages_list->next; |
2468 | pbe->next = restore_pblist; | |
2469 | restore_pblist = pbe; | |
8357376d | 2470 | return pbe->address; |
968808b8 RW |
2471 | } |
2472 | ||
f577eb30 RW |
2473 | /** |
2474 | * snapshot_write_next - used for writing the system memory snapshot. | |
2475 | * | |
2476 | * On the first call to it @handle should point to a zeroed | |
2477 | * snapshot_handle structure. The structure gets updated and a pointer | |
2478 | * to it should be passed to this function every next time. | |
2479 | * | |
f577eb30 RW |
2480 | * On success the function returns a positive number. Then, the caller |
2481 | * is allowed to write up to the returned number of bytes to the memory | |
d3c1b24c | 2482 | * location computed by the data_of() macro. |
f577eb30 RW |
2483 | * |
2484 | * The function returns 0 to indicate the "end of file" condition, | |
2485 | * and a negative number is returned on error. In such cases the | |
2486 | * structure pointed to by @handle is not updated and should not be used | |
2487 | * any more. | |
2488 | */ | |
2489 | ||
d3c1b24c | 2490 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2491 | { |
940864dd | 2492 | static struct chain_allocator ca; |
f577eb30 RW |
2493 | int error = 0; |
2494 | ||
940864dd | 2495 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2496 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2497 | return 0; |
940864dd | 2498 | |
d3c1b24c JS |
2499 | handle->sync_read = 1; |
2500 | ||
2501 | if (!handle->cur) { | |
8357376d RW |
2502 | if (!buffer) |
2503 | /* This makes the buffer be freed by swsusp_free() */ | |
2504 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2505 | ||
f577eb30 RW |
2506 | if (!buffer) |
2507 | return -ENOMEM; | |
8357376d | 2508 | |
f577eb30 | 2509 | handle->buffer = buffer; |
d3c1b24c JS |
2510 | } else if (handle->cur == 1) { |
2511 | error = load_header(buffer); | |
2512 | if (error) | |
2513 | return error; | |
940864dd | 2514 | |
9c744481 RW |
2515 | safe_pages_list = NULL; |
2516 | ||
d3c1b24c JS |
2517 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2518 | if (error) | |
2519 | return error; | |
2520 | ||
85055dd8 MS |
2521 | /* Allocate buffer for page keys. */ |
2522 | error = page_key_alloc(nr_copy_pages); | |
2523 | if (error) | |
2524 | return error; | |
2525 | ||
d3c1b24c JS |
2526 | } else if (handle->cur <= nr_meta_pages + 1) { |
2527 | error = unpack_orig_pfns(buffer, ©_bm); | |
2528 | if (error) | |
2529 | return error; | |
940864dd | 2530 | |
d3c1b24c JS |
2531 | if (handle->cur == nr_meta_pages + 1) { |
2532 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2533 | if (error) |
2534 | return error; | |
2535 | ||
d3c1b24c JS |
2536 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2537 | memory_bm_position_reset(&orig_bm); | |
2538 | restore_pblist = NULL; | |
940864dd | 2539 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2540 | handle->sync_read = 0; |
69643279 RW |
2541 | if (IS_ERR(handle->buffer)) |
2542 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2543 | } |
f577eb30 | 2544 | } else { |
d3c1b24c | 2545 | copy_last_highmem_page(); |
85055dd8 MS |
2546 | /* Restore page key for data page (s390 only). */ |
2547 | page_key_write(handle->buffer); | |
d3c1b24c JS |
2548 | handle->buffer = get_buffer(&orig_bm, &ca); |
2549 | if (IS_ERR(handle->buffer)) | |
2550 | return PTR_ERR(handle->buffer); | |
2551 | if (handle->buffer != buffer) | |
2552 | handle->sync_read = 0; | |
f577eb30 | 2553 | } |
d3c1b24c JS |
2554 | handle->cur++; |
2555 | return PAGE_SIZE; | |
f577eb30 RW |
2556 | } |
2557 | ||
8357376d RW |
2558 | /** |
2559 | * snapshot_write_finalize - must be called after the last call to | |
2560 | * snapshot_write_next() in case the last page in the image happens | |
2561 | * to be a highmem page and its contents should be stored in the | |
2562 | * highmem. Additionally, it releases the memory that will not be | |
2563 | * used any more. | |
2564 | */ | |
2565 | ||
2566 | void snapshot_write_finalize(struct snapshot_handle *handle) | |
2567 | { | |
2568 | copy_last_highmem_page(); | |
85055dd8 MS |
2569 | /* Restore page key for data page (s390 only). */ |
2570 | page_key_write(handle->buffer); | |
2571 | page_key_free(); | |
307c5971 | 2572 | /* Do that only if we have loaded the image entirely */ |
d3c1b24c | 2573 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
307c5971 | 2574 | memory_bm_recycle(&orig_bm); |
8357376d RW |
2575 | free_highmem_data(); |
2576 | } | |
2577 | } | |
2578 | ||
f577eb30 RW |
2579 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2580 | { | |
8357376d | 2581 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2582 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2583 | } | |
2584 | ||
8357376d RW |
2585 | #ifdef CONFIG_HIGHMEM |
2586 | /* Assumes that @buf is ready and points to a "safe" page */ | |
efd5a852 RW |
2587 | static inline void swap_two_pages_data(struct page *p1, struct page *p2, |
2588 | void *buf) | |
940864dd | 2589 | { |
8357376d RW |
2590 | void *kaddr1, *kaddr2; |
2591 | ||
0de9a1e2 CW |
2592 | kaddr1 = kmap_atomic(p1); |
2593 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2594 | copy_page(buf, kaddr1); |
2595 | copy_page(kaddr1, kaddr2); | |
2596 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2597 | kunmap_atomic(kaddr2); |
2598 | kunmap_atomic(kaddr1); | |
8357376d RW |
2599 | } |
2600 | ||
2601 | /** | |
2602 | * restore_highmem - for each highmem page that was allocated before | |
2603 | * the suspend and included in the suspend image, and also has been | |
2604 | * allocated by the "resume" kernel swap its current (ie. "before | |
2605 | * resume") contents with the previous (ie. "before suspend") one. | |
2606 | * | |
2607 | * If the resume eventually fails, we can call this function once | |
2608 | * again and restore the "before resume" highmem state. | |
2609 | */ | |
2610 | ||
2611 | int restore_highmem(void) | |
2612 | { | |
2613 | struct highmem_pbe *pbe = highmem_pblist; | |
2614 | void *buf; | |
2615 | ||
2616 | if (!pbe) | |
2617 | return 0; | |
2618 | ||
2619 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2620 | if (!buf) | |
2621 | return -ENOMEM; | |
2622 | ||
2623 | while (pbe) { | |
2624 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2625 | pbe = pbe->next; | |
2626 | } | |
2627 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2628 | return 0; | |
f577eb30 | 2629 | } |
8357376d | 2630 | #endif /* CONFIG_HIGHMEM */ |