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