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 | ||
4c0b6c10 RW |
41 | #ifdef CONFIG_DEBUG_RODATA |
42 | static bool hibernate_restore_protection; | |
43 | static bool hibernate_restore_protection_active; | |
44 | ||
45 | void enable_restore_image_protection(void) | |
46 | { | |
47 | hibernate_restore_protection = true; | |
48 | } | |
49 | ||
50 | static inline void hibernate_restore_protection_begin(void) | |
51 | { | |
52 | hibernate_restore_protection_active = hibernate_restore_protection; | |
53 | } | |
54 | ||
55 | static inline void hibernate_restore_protection_end(void) | |
56 | { | |
57 | hibernate_restore_protection_active = false; | |
58 | } | |
59 | ||
60 | static inline void hibernate_restore_protect_page(void *page_address) | |
61 | { | |
62 | if (hibernate_restore_protection_active) | |
63 | set_memory_ro((unsigned long)page_address, 1); | |
64 | } | |
65 | ||
66 | static inline void hibernate_restore_unprotect_page(void *page_address) | |
67 | { | |
68 | if (hibernate_restore_protection_active) | |
69 | set_memory_rw((unsigned long)page_address, 1); | |
70 | } | |
71 | #else | |
72 | static inline void hibernate_restore_protection_begin(void) {} | |
73 | static inline void hibernate_restore_protection_end(void) {} | |
74 | static inline void hibernate_restore_protect_page(void *page_address) {} | |
75 | static inline void hibernate_restore_unprotect_page(void *page_address) {} | |
76 | #endif /* CONFIG_DEBUG_RODATA */ | |
77 | ||
74dfd666 RW |
78 | static int swsusp_page_is_free(struct page *); |
79 | static void swsusp_set_page_forbidden(struct page *); | |
80 | static void swsusp_unset_page_forbidden(struct page *); | |
81 | ||
ddeb6487 RW |
82 | /* |
83 | * Number of bytes to reserve for memory allocations made by device drivers | |
84 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
85 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
86 | */ | |
87 | unsigned long reserved_size; | |
88 | ||
89 | void __init hibernate_reserved_size_init(void) | |
90 | { | |
91 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
92 | } | |
93 | ||
fe419535 RW |
94 | /* |
95 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
96 | * When it is set to N, swsusp will do its best to ensure the image |
97 | * size will not exceed N bytes, but if that is impossible, it will | |
98 | * try to create the smallest image possible. | |
fe419535 | 99 | */ |
ac5c24ec RW |
100 | unsigned long image_size; |
101 | ||
102 | void __init hibernate_image_size_init(void) | |
103 | { | |
1c1be3a9 | 104 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 105 | } |
fe419535 | 106 | |
ef96f639 RW |
107 | /* |
108 | * List of PBEs needed for restoring the pages that were allocated before | |
8357376d RW |
109 | * the suspend and included in the suspend image, but have also been |
110 | * allocated by the "resume" kernel, so their contents cannot be written | |
111 | * directly to their "original" page frames. | |
112 | */ | |
75534b50 RW |
113 | struct pbe *restore_pblist; |
114 | ||
9c744481 RW |
115 | /* struct linked_page is used to build chains of pages */ |
116 | ||
117 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
118 | ||
119 | struct linked_page { | |
120 | struct linked_page *next; | |
121 | char data[LINKED_PAGE_DATA_SIZE]; | |
122 | } __packed; | |
123 | ||
124 | /* | |
125 | * List of "safe" pages (ie. pages that were not used by the image kernel | |
126 | * before hibernation) that may be used as temporary storage for image kernel | |
127 | * memory contents. | |
128 | */ | |
129 | static struct linked_page *safe_pages_list; | |
130 | ||
8357376d | 131 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 132 | static void *buffer; |
7088a5c0 | 133 | |
0bcd888d RW |
134 | #define PG_ANY 0 |
135 | #define PG_SAFE 1 | |
136 | #define PG_UNSAFE_CLEAR 1 | |
137 | #define PG_UNSAFE_KEEP 0 | |
138 | ||
940864dd | 139 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 140 | |
ef96f639 RW |
141 | /** |
142 | * get_image_page - Allocate a page for a hibernation image. | |
143 | * @gfp_mask: GFP mask for the allocation. | |
144 | * @safe_needed: Get pages that were not used before hibernation (restore only) | |
145 | * | |
146 | * During image restoration, for storing the PBE list and the image data, we can | |
147 | * only use memory pages that do not conflict with the pages used before | |
148 | * hibernation. The "unsafe" pages have PageNosaveFree set and we count them | |
149 | * using allocated_unsafe_pages. | |
150 | * | |
151 | * Each allocated image page is marked as PageNosave and PageNosaveFree so that | |
152 | * swsusp_free() can release it. | |
153 | */ | |
8357376d | 154 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
155 | { |
156 | void *res; | |
157 | ||
158 | res = (void *)get_zeroed_page(gfp_mask); | |
159 | if (safe_needed) | |
7be98234 | 160 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 161 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 162 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 163 | allocated_unsafe_pages++; |
f6143aa6 RW |
164 | res = (void *)get_zeroed_page(gfp_mask); |
165 | } | |
166 | if (res) { | |
7be98234 RW |
167 | swsusp_set_page_forbidden(virt_to_page(res)); |
168 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
169 | } |
170 | return res; | |
171 | } | |
172 | ||
9c744481 RW |
173 | static void *__get_safe_page(gfp_t gfp_mask) |
174 | { | |
175 | if (safe_pages_list) { | |
176 | void *ret = safe_pages_list; | |
177 | ||
178 | safe_pages_list = safe_pages_list->next; | |
179 | memset(ret, 0, PAGE_SIZE); | |
180 | return ret; | |
181 | } | |
182 | return get_image_page(gfp_mask, PG_SAFE); | |
183 | } | |
184 | ||
f6143aa6 RW |
185 | unsigned long get_safe_page(gfp_t gfp_mask) |
186 | { | |
9c744481 | 187 | return (unsigned long)__get_safe_page(gfp_mask); |
8357376d RW |
188 | } |
189 | ||
5b6d15de RW |
190 | static struct page *alloc_image_page(gfp_t gfp_mask) |
191 | { | |
8357376d RW |
192 | struct page *page; |
193 | ||
194 | page = alloc_page(gfp_mask); | |
195 | if (page) { | |
7be98234 RW |
196 | swsusp_set_page_forbidden(page); |
197 | swsusp_set_page_free(page); | |
8357376d RW |
198 | } |
199 | return page; | |
f6143aa6 RW |
200 | } |
201 | ||
307c5971 RW |
202 | static void recycle_safe_page(void *page_address) |
203 | { | |
204 | struct linked_page *lp = page_address; | |
205 | ||
206 | lp->next = safe_pages_list; | |
207 | safe_pages_list = lp; | |
208 | } | |
209 | ||
f6143aa6 | 210 | /** |
ef96f639 RW |
211 | * free_image_page - Free a page allocated for hibernation image. |
212 | * @addr: Address of the page to free. | |
213 | * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page. | |
214 | * | |
215 | * The page to free should have been allocated by get_image_page() (page flags | |
216 | * set by it are affected). | |
f6143aa6 | 217 | */ |
f6143aa6 RW |
218 | static inline void free_image_page(void *addr, int clear_nosave_free) |
219 | { | |
8357376d RW |
220 | struct page *page; |
221 | ||
222 | BUG_ON(!virt_addr_valid(addr)); | |
223 | ||
224 | page = virt_to_page(addr); | |
225 | ||
7be98234 | 226 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 227 | if (clear_nosave_free) |
7be98234 | 228 | swsusp_unset_page_free(page); |
8357376d RW |
229 | |
230 | __free_page(page); | |
f6143aa6 RW |
231 | } |
232 | ||
efd5a852 RW |
233 | static inline void free_list_of_pages(struct linked_page *list, |
234 | int clear_page_nosave) | |
b788db79 RW |
235 | { |
236 | while (list) { | |
237 | struct linked_page *lp = list->next; | |
238 | ||
239 | free_image_page(list, clear_page_nosave); | |
240 | list = lp; | |
241 | } | |
242 | } | |
243 | ||
ef96f639 RW |
244 | /* |
245 | * struct chain_allocator is used for allocating small objects out of | |
246 | * a linked list of pages called 'the chain'. | |
247 | * | |
248 | * The chain grows each time when there is no room for a new object in | |
249 | * the current page. The allocated objects cannot be freed individually. | |
250 | * It is only possible to free them all at once, by freeing the entire | |
251 | * chain. | |
252 | * | |
253 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
254 | * are not much smaller than PAGE_SIZE. | |
255 | */ | |
b788db79 RW |
256 | struct chain_allocator { |
257 | struct linked_page *chain; /* the chain */ | |
258 | unsigned int used_space; /* total size of objects allocated out | |
ef96f639 | 259 | of the current page */ |
b788db79 RW |
260 | gfp_t gfp_mask; /* mask for allocating pages */ |
261 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
262 | }; | |
263 | ||
efd5a852 RW |
264 | static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask, |
265 | int safe_needed) | |
b788db79 RW |
266 | { |
267 | ca->chain = NULL; | |
268 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
269 | ca->gfp_mask = gfp_mask; | |
270 | ca->safe_needed = safe_needed; | |
271 | } | |
272 | ||
273 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
274 | { | |
275 | void *ret; | |
276 | ||
277 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
278 | struct linked_page *lp; | |
279 | ||
9c744481 RW |
280 | lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) : |
281 | get_image_page(ca->gfp_mask, PG_ANY); | |
b788db79 RW |
282 | if (!lp) |
283 | return NULL; | |
284 | ||
285 | lp->next = ca->chain; | |
286 | ca->chain = lp; | |
287 | ca->used_space = 0; | |
288 | } | |
289 | ret = ca->chain->data + ca->used_space; | |
290 | ca->used_space += size; | |
291 | return ret; | |
292 | } | |
293 | ||
b788db79 | 294 | /** |
ef96f639 | 295 | * Data types related to memory bitmaps. |
b788db79 | 296 | * |
ef96f639 RW |
297 | * Memory bitmap is a structure consiting of many linked lists of |
298 | * objects. The main list's elements are of type struct zone_bitmap | |
299 | * and each of them corresonds to one zone. For each zone bitmap | |
300 | * object there is a list of objects of type struct bm_block that | |
301 | * represent each blocks of bitmap in which information is stored. | |
b788db79 | 302 | * |
ef96f639 RW |
303 | * struct memory_bitmap contains a pointer to the main list of zone |
304 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
305 | * and a pointer to the list of pages used for allocating all of the | |
306 | * zone bitmap objects and bitmap block objects. | |
b788db79 | 307 | * |
ef96f639 RW |
308 | * NOTE: It has to be possible to lay out the bitmap in memory |
309 | * using only allocations of order 0. Additionally, the bitmap is | |
310 | * designed to work with arbitrary number of zones (this is over the | |
311 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
b788db79 | 312 | * |
ef96f639 RW |
313 | * struct zone_bitmap contains a pointer to a list of bitmap block |
314 | * objects and a pointer to the bitmap block object that has been | |
315 | * most recently used for setting bits. Additionally, it contains the | |
316 | * PFNs that correspond to the start and end of the represented zone. | |
b788db79 | 317 | * |
ef96f639 RW |
318 | * struct bm_block contains a pointer to the memory page in which |
319 | * information is stored (in the form of a block of bitmap) | |
320 | * It also contains the pfns that correspond to the start and end of | |
321 | * the represented memory area. | |
f469f02d | 322 | * |
ef96f639 RW |
323 | * The memory bitmap is organized as a radix tree to guarantee fast random |
324 | * access to the bits. There is one radix tree for each zone (as returned | |
325 | * from create_mem_extents). | |
f469f02d | 326 | * |
ef96f639 RW |
327 | * One radix tree is represented by one struct mem_zone_bm_rtree. There are |
328 | * two linked lists for the nodes of the tree, one for the inner nodes and | |
329 | * one for the leave nodes. The linked leave nodes are used for fast linear | |
330 | * access of the memory bitmap. | |
f469f02d | 331 | * |
ef96f639 | 332 | * The struct rtree_node represents one node of the radix tree. |
b788db79 RW |
333 | */ |
334 | ||
335 | #define BM_END_OF_MAP (~0UL) | |
336 | ||
8de03073 | 337 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
f469f02d JR |
338 | #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3) |
339 | #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1) | |
b788db79 | 340 | |
f469f02d JR |
341 | /* |
342 | * struct rtree_node is a wrapper struct to link the nodes | |
343 | * of the rtree together for easy linear iteration over | |
344 | * bits and easy freeing | |
345 | */ | |
346 | struct rtree_node { | |
347 | struct list_head list; | |
348 | unsigned long *data; | |
349 | }; | |
350 | ||
351 | /* | |
352 | * struct mem_zone_bm_rtree represents a bitmap used for one | |
353 | * populated memory zone. | |
354 | */ | |
355 | struct mem_zone_bm_rtree { | |
356 | struct list_head list; /* Link Zones together */ | |
357 | struct list_head nodes; /* Radix Tree inner nodes */ | |
358 | struct list_head leaves; /* Radix Tree leaves */ | |
359 | unsigned long start_pfn; /* Zone start page frame */ | |
360 | unsigned long end_pfn; /* Zone end page frame + 1 */ | |
361 | struct rtree_node *rtree; /* Radix Tree Root */ | |
362 | int levels; /* Number of Radix Tree Levels */ | |
363 | unsigned int blocks; /* Number of Bitmap Blocks */ | |
364 | }; | |
365 | ||
b788db79 RW |
366 | /* strcut bm_position is used for browsing memory bitmaps */ |
367 | ||
368 | struct bm_position { | |
3a20cb17 JR |
369 | struct mem_zone_bm_rtree *zone; |
370 | struct rtree_node *node; | |
371 | unsigned long node_pfn; | |
372 | int node_bit; | |
b788db79 RW |
373 | }; |
374 | ||
375 | struct memory_bitmap { | |
f469f02d | 376 | struct list_head zones; |
b788db79 | 377 | struct linked_page *p_list; /* list of pages used to store zone |
ef96f639 RW |
378 | bitmap objects and bitmap block |
379 | objects */ | |
b788db79 RW |
380 | struct bm_position cur; /* most recently used bit position */ |
381 | }; | |
382 | ||
383 | /* Functions that operate on memory bitmaps */ | |
384 | ||
f469f02d JR |
385 | #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long)) |
386 | #if BITS_PER_LONG == 32 | |
387 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2) | |
388 | #else | |
389 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3) | |
390 | #endif | |
391 | #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1) | |
392 | ||
ef96f639 RW |
393 | /** |
394 | * alloc_rtree_node - Allocate a new node and add it to the radix tree. | |
f469f02d | 395 | * |
ef96f639 RW |
396 | * This function is used to allocate inner nodes as well as the |
397 | * leave nodes of the radix tree. It also adds the node to the | |
398 | * corresponding linked list passed in by the *list parameter. | |
f469f02d JR |
399 | */ |
400 | static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed, | |
401 | struct chain_allocator *ca, | |
402 | struct list_head *list) | |
403 | { | |
404 | struct rtree_node *node; | |
405 | ||
406 | node = chain_alloc(ca, sizeof(struct rtree_node)); | |
407 | if (!node) | |
408 | return NULL; | |
409 | ||
410 | node->data = get_image_page(gfp_mask, safe_needed); | |
411 | if (!node->data) | |
412 | return NULL; | |
413 | ||
414 | list_add_tail(&node->list, list); | |
415 | ||
416 | return node; | |
417 | } | |
418 | ||
ef96f639 RW |
419 | /** |
420 | * add_rtree_block - Add a new leave node to the radix tree. | |
f469f02d | 421 | * |
ef96f639 RW |
422 | * The leave nodes need to be allocated in order to keep the leaves |
423 | * linked list in order. This is guaranteed by the zone->blocks | |
424 | * counter. | |
f469f02d JR |
425 | */ |
426 | static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask, | |
427 | int safe_needed, struct chain_allocator *ca) | |
428 | { | |
429 | struct rtree_node *node, *block, **dst; | |
430 | unsigned int levels_needed, block_nr; | |
431 | int i; | |
432 | ||
433 | block_nr = zone->blocks; | |
434 | levels_needed = 0; | |
435 | ||
436 | /* How many levels do we need for this block nr? */ | |
437 | while (block_nr) { | |
438 | levels_needed += 1; | |
439 | block_nr >>= BM_RTREE_LEVEL_SHIFT; | |
440 | } | |
441 | ||
442 | /* Make sure the rtree has enough levels */ | |
443 | for (i = zone->levels; i < levels_needed; i++) { | |
444 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
445 | &zone->nodes); | |
446 | if (!node) | |
447 | return -ENOMEM; | |
448 | ||
449 | node->data[0] = (unsigned long)zone->rtree; | |
450 | zone->rtree = node; | |
451 | zone->levels += 1; | |
452 | } | |
453 | ||
454 | /* Allocate new block */ | |
455 | block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves); | |
456 | if (!block) | |
457 | return -ENOMEM; | |
458 | ||
459 | /* Now walk the rtree to insert the block */ | |
460 | node = zone->rtree; | |
461 | dst = &zone->rtree; | |
462 | block_nr = zone->blocks; | |
463 | for (i = zone->levels; i > 0; i--) { | |
464 | int index; | |
465 | ||
466 | if (!node) { | |
467 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
468 | &zone->nodes); | |
469 | if (!node) | |
470 | return -ENOMEM; | |
471 | *dst = node; | |
472 | } | |
473 | ||
474 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
475 | index &= BM_RTREE_LEVEL_MASK; | |
476 | dst = (struct rtree_node **)&((*dst)->data[index]); | |
477 | node = *dst; | |
478 | } | |
479 | ||
480 | zone->blocks += 1; | |
481 | *dst = block; | |
482 | ||
483 | return 0; | |
484 | } | |
485 | ||
486 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
487 | int clear_nosave_free); | |
488 | ||
ef96f639 RW |
489 | /** |
490 | * create_zone_bm_rtree - Create a radix tree for one zone. | |
f469f02d | 491 | * |
ef96f639 RW |
492 | * Allocated the mem_zone_bm_rtree structure and initializes it. |
493 | * This function also allocated and builds the radix tree for the | |
494 | * zone. | |
f469f02d | 495 | */ |
efd5a852 RW |
496 | static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask, |
497 | int safe_needed, | |
498 | struct chain_allocator *ca, | |
499 | unsigned long start, | |
500 | unsigned long end) | |
f469f02d JR |
501 | { |
502 | struct mem_zone_bm_rtree *zone; | |
503 | unsigned int i, nr_blocks; | |
504 | unsigned long pages; | |
505 | ||
506 | pages = end - start; | |
507 | zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree)); | |
508 | if (!zone) | |
509 | return NULL; | |
510 | ||
511 | INIT_LIST_HEAD(&zone->nodes); | |
512 | INIT_LIST_HEAD(&zone->leaves); | |
513 | zone->start_pfn = start; | |
514 | zone->end_pfn = end; | |
515 | nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); | |
516 | ||
517 | for (i = 0; i < nr_blocks; i++) { | |
518 | if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) { | |
519 | free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR); | |
520 | return NULL; | |
521 | } | |
522 | } | |
523 | ||
524 | return zone; | |
525 | } | |
526 | ||
ef96f639 RW |
527 | /** |
528 | * free_zone_bm_rtree - Free the memory of the radix tree. | |
f469f02d | 529 | * |
ef96f639 RW |
530 | * Free all node pages of the radix tree. The mem_zone_bm_rtree |
531 | * structure itself is not freed here nor are the rtree_node | |
532 | * structs. | |
f469f02d JR |
533 | */ |
534 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
535 | int clear_nosave_free) | |
536 | { | |
537 | struct rtree_node *node; | |
538 | ||
539 | list_for_each_entry(node, &zone->nodes, list) | |
540 | free_image_page(node->data, clear_nosave_free); | |
541 | ||
542 | list_for_each_entry(node, &zone->leaves, list) | |
543 | free_image_page(node->data, clear_nosave_free); | |
544 | } | |
545 | ||
b788db79 RW |
546 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
547 | { | |
3a20cb17 JR |
548 | bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree, |
549 | list); | |
550 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, | |
551 | struct rtree_node, list); | |
552 | bm->cur.node_pfn = 0; | |
553 | bm->cur.node_bit = 0; | |
b788db79 RW |
554 | } |
555 | ||
556 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
557 | ||
846705de RW |
558 | struct mem_extent { |
559 | struct list_head hook; | |
560 | unsigned long start; | |
561 | unsigned long end; | |
562 | }; | |
563 | ||
b788db79 | 564 | /** |
ef96f639 RW |
565 | * free_mem_extents - Free a list of memory extents. |
566 | * @list: List of extents to free. | |
b788db79 | 567 | */ |
846705de RW |
568 | static void free_mem_extents(struct list_head *list) |
569 | { | |
570 | struct mem_extent *ext, *aux; | |
b788db79 | 571 | |
846705de RW |
572 | list_for_each_entry_safe(ext, aux, list, hook) { |
573 | list_del(&ext->hook); | |
574 | kfree(ext); | |
575 | } | |
576 | } | |
577 | ||
578 | /** | |
ef96f639 RW |
579 | * create_mem_extents - Create a list of memory extents. |
580 | * @list: List to put the extents into. | |
581 | * @gfp_mask: Mask to use for memory allocations. | |
582 | * | |
583 | * The extents represent contiguous ranges of PFNs. | |
846705de RW |
584 | */ |
585 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 586 | { |
846705de | 587 | struct zone *zone; |
b788db79 | 588 | |
846705de | 589 | INIT_LIST_HEAD(list); |
b788db79 | 590 | |
ee99c71c | 591 | for_each_populated_zone(zone) { |
846705de RW |
592 | unsigned long zone_start, zone_end; |
593 | struct mem_extent *ext, *cur, *aux; | |
594 | ||
846705de | 595 | zone_start = zone->zone_start_pfn; |
c33bc315 | 596 | zone_end = zone_end_pfn(zone); |
846705de RW |
597 | |
598 | list_for_each_entry(ext, list, hook) | |
599 | if (zone_start <= ext->end) | |
600 | break; | |
b788db79 | 601 | |
846705de RW |
602 | if (&ext->hook == list || zone_end < ext->start) { |
603 | /* New extent is necessary */ | |
604 | struct mem_extent *new_ext; | |
605 | ||
606 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
607 | if (!new_ext) { | |
608 | free_mem_extents(list); | |
609 | return -ENOMEM; | |
610 | } | |
611 | new_ext->start = zone_start; | |
612 | new_ext->end = zone_end; | |
613 | list_add_tail(&new_ext->hook, &ext->hook); | |
614 | continue; | |
615 | } | |
616 | ||
617 | /* Merge this zone's range of PFNs with the existing one */ | |
618 | if (zone_start < ext->start) | |
619 | ext->start = zone_start; | |
620 | if (zone_end > ext->end) | |
621 | ext->end = zone_end; | |
622 | ||
623 | /* More merging may be possible */ | |
624 | cur = ext; | |
625 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
626 | if (zone_end < cur->start) | |
627 | break; | |
628 | if (zone_end < cur->end) | |
629 | ext->end = cur->end; | |
630 | list_del(&cur->hook); | |
631 | kfree(cur); | |
632 | } | |
b788db79 | 633 | } |
846705de RW |
634 | |
635 | return 0; | |
b788db79 RW |
636 | } |
637 | ||
638 | /** | |
ef96f639 RW |
639 | * memory_bm_create - Allocate memory for a memory bitmap. |
640 | */ | |
efd5a852 RW |
641 | static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, |
642 | int safe_needed) | |
b788db79 RW |
643 | { |
644 | struct chain_allocator ca; | |
846705de RW |
645 | struct list_head mem_extents; |
646 | struct mem_extent *ext; | |
647 | int error; | |
b788db79 RW |
648 | |
649 | chain_init(&ca, gfp_mask, safe_needed); | |
f469f02d | 650 | INIT_LIST_HEAD(&bm->zones); |
b788db79 | 651 | |
846705de RW |
652 | error = create_mem_extents(&mem_extents, gfp_mask); |
653 | if (error) | |
654 | return error; | |
b788db79 | 655 | |
846705de | 656 | list_for_each_entry(ext, &mem_extents, hook) { |
f469f02d | 657 | struct mem_zone_bm_rtree *zone; |
f469f02d JR |
658 | |
659 | zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca, | |
660 | ext->start, ext->end); | |
9047eb62 JR |
661 | if (!zone) { |
662 | error = -ENOMEM; | |
f469f02d | 663 | goto Error; |
9047eb62 | 664 | } |
f469f02d | 665 | list_add_tail(&zone->list, &bm->zones); |
b788db79 | 666 | } |
846705de | 667 | |
b788db79 RW |
668 | bm->p_list = ca.chain; |
669 | memory_bm_position_reset(bm); | |
846705de RW |
670 | Exit: |
671 | free_mem_extents(&mem_extents); | |
672 | return error; | |
b788db79 | 673 | |
846705de | 674 | Error: |
b788db79 RW |
675 | bm->p_list = ca.chain; |
676 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 677 | goto Exit; |
b788db79 RW |
678 | } |
679 | ||
680 | /** | |
ef96f639 RW |
681 | * memory_bm_free - Free memory occupied by the memory bitmap. |
682 | * @bm: Memory bitmap. | |
683 | */ | |
b788db79 RW |
684 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
685 | { | |
f469f02d | 686 | struct mem_zone_bm_rtree *zone; |
b788db79 | 687 | |
f469f02d JR |
688 | list_for_each_entry(zone, &bm->zones, list) |
689 | free_zone_bm_rtree(zone, clear_nosave_free); | |
690 | ||
b788db79 | 691 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de | 692 | |
f469f02d | 693 | INIT_LIST_HEAD(&bm->zones); |
b788db79 RW |
694 | } |
695 | ||
696 | /** | |
ef96f639 | 697 | * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap. |
07a33823 | 698 | * |
ef96f639 RW |
699 | * Find the bit in memory bitmap @bm that corresponds to the given PFN. |
700 | * The cur.zone, cur.block and cur.node_pfn members of @bm are updated. | |
701 | * | |
702 | * Walk the radix tree to find the page containing the bit that represents @pfn | |
703 | * and return the position of the bit in @addr and @bit_nr. | |
07a33823 | 704 | */ |
9047eb62 JR |
705 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
706 | void **addr, unsigned int *bit_nr) | |
07a33823 JR |
707 | { |
708 | struct mem_zone_bm_rtree *curr, *zone; | |
709 | struct rtree_node *node; | |
710 | int i, block_nr; | |
711 | ||
3a20cb17 JR |
712 | zone = bm->cur.zone; |
713 | ||
714 | if (pfn >= zone->start_pfn && pfn < zone->end_pfn) | |
715 | goto zone_found; | |
716 | ||
07a33823 JR |
717 | zone = NULL; |
718 | ||
719 | /* Find the right zone */ | |
720 | list_for_each_entry(curr, &bm->zones, list) { | |
721 | if (pfn >= curr->start_pfn && pfn < curr->end_pfn) { | |
722 | zone = curr; | |
723 | break; | |
724 | } | |
725 | } | |
726 | ||
727 | if (!zone) | |
728 | return -EFAULT; | |
729 | ||
3a20cb17 | 730 | zone_found: |
07a33823 | 731 | /* |
ef96f639 RW |
732 | * We have found the zone. Now walk the radix tree to find the leaf node |
733 | * for our PFN. | |
07a33823 | 734 | */ |
3a20cb17 JR |
735 | node = bm->cur.node; |
736 | if (((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn) | |
737 | goto node_found; | |
738 | ||
07a33823 JR |
739 | node = zone->rtree; |
740 | block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT; | |
741 | ||
742 | for (i = zone->levels; i > 0; i--) { | |
743 | int index; | |
744 | ||
745 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
746 | index &= BM_RTREE_LEVEL_MASK; | |
747 | BUG_ON(node->data[index] == 0); | |
748 | node = (struct rtree_node *)node->data[index]; | |
749 | } | |
750 | ||
3a20cb17 JR |
751 | node_found: |
752 | /* Update last position */ | |
753 | bm->cur.zone = zone; | |
754 | bm->cur.node = node; | |
755 | bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK; | |
756 | ||
07a33823 JR |
757 | /* Set return values */ |
758 | *addr = node->data; | |
759 | *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK; | |
760 | ||
761 | return 0; | |
762 | } | |
763 | ||
74dfd666 RW |
764 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) |
765 | { | |
766 | void *addr; | |
767 | unsigned int bit; | |
a82f7119 | 768 | int error; |
74dfd666 | 769 | |
a82f7119 RW |
770 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
771 | BUG_ON(error); | |
74dfd666 RW |
772 | set_bit(bit, addr); |
773 | } | |
774 | ||
a82f7119 RW |
775 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
776 | { | |
777 | void *addr; | |
778 | unsigned int bit; | |
779 | int error; | |
780 | ||
781 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
07a33823 JR |
782 | if (!error) |
783 | set_bit(bit, addr); | |
784 | ||
a82f7119 RW |
785 | return error; |
786 | } | |
787 | ||
74dfd666 RW |
788 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
789 | { | |
790 | void *addr; | |
791 | unsigned int bit; | |
a82f7119 | 792 | int error; |
74dfd666 | 793 | |
a82f7119 RW |
794 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
795 | BUG_ON(error); | |
74dfd666 RW |
796 | clear_bit(bit, addr); |
797 | } | |
798 | ||
fdd64ed5 JR |
799 | static void memory_bm_clear_current(struct memory_bitmap *bm) |
800 | { | |
801 | int bit; | |
802 | ||
803 | bit = max(bm->cur.node_bit - 1, 0); | |
804 | clear_bit(bit, bm->cur.node->data); | |
805 | } | |
806 | ||
74dfd666 RW |
807 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) |
808 | { | |
809 | void *addr; | |
810 | unsigned int bit; | |
9047eb62 | 811 | int error; |
74dfd666 | 812 | |
a82f7119 RW |
813 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
814 | BUG_ON(error); | |
9047eb62 | 815 | return test_bit(bit, addr); |
b788db79 RW |
816 | } |
817 | ||
69643279 RW |
818 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
819 | { | |
820 | void *addr; | |
821 | unsigned int bit; | |
07a33823 | 822 | |
9047eb62 | 823 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); |
b788db79 RW |
824 | } |
825 | ||
3a20cb17 | 826 | /* |
ef96f639 | 827 | * rtree_next_node - Jump to the next leaf node. |
3a20cb17 | 828 | * |
ef96f639 RW |
829 | * Set the position to the beginning of the next node in the |
830 | * memory bitmap. This is either the next node in the current | |
831 | * zone's radix tree or the first node in the radix tree of the | |
832 | * next zone. | |
3a20cb17 | 833 | * |
ef96f639 | 834 | * Return true if there is a next node, false otherwise. |
3a20cb17 JR |
835 | */ |
836 | static bool rtree_next_node(struct memory_bitmap *bm) | |
837 | { | |
924d8696 JM |
838 | if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) { |
839 | bm->cur.node = list_entry(bm->cur.node->list.next, | |
840 | struct rtree_node, list); | |
3a20cb17 JR |
841 | bm->cur.node_pfn += BM_BITS_PER_BLOCK; |
842 | bm->cur.node_bit = 0; | |
0f7d83e8 | 843 | touch_softlockup_watchdog(); |
3a20cb17 JR |
844 | return true; |
845 | } | |
846 | ||
847 | /* No more nodes, goto next zone */ | |
924d8696 JM |
848 | if (!list_is_last(&bm->cur.zone->list, &bm->zones)) { |
849 | bm->cur.zone = list_entry(bm->cur.zone->list.next, | |
3a20cb17 | 850 | struct mem_zone_bm_rtree, list); |
3a20cb17 JR |
851 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, |
852 | struct rtree_node, list); | |
853 | bm->cur.node_pfn = 0; | |
854 | bm->cur.node_bit = 0; | |
855 | return true; | |
856 | } | |
857 | ||
858 | /* No more zones */ | |
859 | return false; | |
860 | } | |
861 | ||
9047eb62 | 862 | /** |
ef96f639 RW |
863 | * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap. |
864 | * @bm: Memory bitmap. | |
3a20cb17 | 865 | * |
ef96f639 RW |
866 | * Starting from the last returned position this function searches for the next |
867 | * set bit in @bm and returns the PFN represented by it. If no more bits are | |
868 | * set, BM_END_OF_MAP is returned. | |
9047eb62 | 869 | * |
ef96f639 RW |
870 | * It is required to run memory_bm_position_reset() before the first call to |
871 | * this function for the given memory bitmap. | |
3a20cb17 | 872 | */ |
9047eb62 | 873 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) |
3a20cb17 JR |
874 | { |
875 | unsigned long bits, pfn, pages; | |
876 | int bit; | |
877 | ||
878 | do { | |
879 | pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn; | |
880 | bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK); | |
881 | bit = find_next_bit(bm->cur.node->data, bits, | |
882 | bm->cur.node_bit); | |
883 | if (bit < bits) { | |
884 | pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit; | |
885 | bm->cur.node_bit = bit + 1; | |
886 | return pfn; | |
887 | } | |
888 | } while (rtree_next_node(bm)); | |
889 | ||
890 | return BM_END_OF_MAP; | |
891 | } | |
892 | ||
ef96f639 RW |
893 | /* |
894 | * This structure represents a range of page frames the contents of which | |
895 | * should not be saved during hibernation. | |
74dfd666 | 896 | */ |
74dfd666 RW |
897 | struct nosave_region { |
898 | struct list_head list; | |
899 | unsigned long start_pfn; | |
900 | unsigned long end_pfn; | |
901 | }; | |
902 | ||
903 | static LIST_HEAD(nosave_regions); | |
904 | ||
307c5971 RW |
905 | static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone) |
906 | { | |
907 | struct rtree_node *node; | |
908 | ||
909 | list_for_each_entry(node, &zone->nodes, list) | |
910 | recycle_safe_page(node->data); | |
911 | ||
912 | list_for_each_entry(node, &zone->leaves, list) | |
913 | recycle_safe_page(node->data); | |
914 | } | |
915 | ||
916 | static void memory_bm_recycle(struct memory_bitmap *bm) | |
917 | { | |
918 | struct mem_zone_bm_rtree *zone; | |
919 | struct linked_page *p_list; | |
920 | ||
921 | list_for_each_entry(zone, &bm->zones, list) | |
922 | recycle_zone_bm_rtree(zone); | |
923 | ||
924 | p_list = bm->p_list; | |
925 | while (p_list) { | |
926 | struct linked_page *lp = p_list; | |
927 | ||
928 | p_list = lp->next; | |
929 | recycle_safe_page(lp); | |
930 | } | |
931 | } | |
932 | ||
74dfd666 | 933 | /** |
ef96f639 RW |
934 | * register_nosave_region - Register a region of unsaveable memory. |
935 | * | |
936 | * Register a range of page frames the contents of which should not be saved | |
937 | * during hibernation (to be used in the early initialization code). | |
74dfd666 | 938 | */ |
efd5a852 RW |
939 | void __init __register_nosave_region(unsigned long start_pfn, |
940 | unsigned long end_pfn, int use_kmalloc) | |
74dfd666 RW |
941 | { |
942 | struct nosave_region *region; | |
943 | ||
944 | if (start_pfn >= end_pfn) | |
945 | return; | |
946 | ||
947 | if (!list_empty(&nosave_regions)) { | |
948 | /* Try to extend the previous region (they should be sorted) */ | |
949 | region = list_entry(nosave_regions.prev, | |
950 | struct nosave_region, list); | |
951 | if (region->end_pfn == start_pfn) { | |
952 | region->end_pfn = end_pfn; | |
953 | goto Report; | |
954 | } | |
955 | } | |
940d67f6 | 956 | if (use_kmalloc) { |
ef96f639 | 957 | /* During init, this shouldn't fail */ |
940d67f6 JB |
958 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); |
959 | BUG_ON(!region); | |
d5f32af3 | 960 | } else { |
940d67f6 | 961 | /* This allocation cannot fail */ |
c2f69cda | 962 | region = memblock_virt_alloc(sizeof(struct nosave_region), 0); |
d5f32af3 | 963 | } |
74dfd666 RW |
964 | region->start_pfn = start_pfn; |
965 | region->end_pfn = end_pfn; | |
966 | list_add_tail(®ion->list, &nosave_regions); | |
967 | Report: | |
cd38ca85 BH |
968 | printk(KERN_INFO "PM: Registered nosave memory: [mem %#010llx-%#010llx]\n", |
969 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
970 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
74dfd666 RW |
971 | } |
972 | ||
973 | /* | |
974 | * Set bits in this map correspond to the page frames the contents of which | |
975 | * should not be saved during the suspend. | |
976 | */ | |
977 | static struct memory_bitmap *forbidden_pages_map; | |
978 | ||
979 | /* Set bits in this map correspond to free page frames. */ | |
980 | static struct memory_bitmap *free_pages_map; | |
981 | ||
982 | /* | |
983 | * Each page frame allocated for creating the image is marked by setting the | |
984 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
985 | */ | |
986 | ||
987 | void swsusp_set_page_free(struct page *page) | |
988 | { | |
989 | if (free_pages_map) | |
990 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
991 | } | |
992 | ||
993 | static int swsusp_page_is_free(struct page *page) | |
994 | { | |
995 | return free_pages_map ? | |
996 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
997 | } | |
998 | ||
999 | void swsusp_unset_page_free(struct page *page) | |
1000 | { | |
1001 | if (free_pages_map) | |
1002 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
1003 | } | |
1004 | ||
1005 | static void swsusp_set_page_forbidden(struct page *page) | |
1006 | { | |
1007 | if (forbidden_pages_map) | |
1008 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
1009 | } | |
1010 | ||
1011 | int swsusp_page_is_forbidden(struct page *page) | |
1012 | { | |
1013 | return forbidden_pages_map ? | |
1014 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
1015 | } | |
1016 | ||
1017 | static void swsusp_unset_page_forbidden(struct page *page) | |
1018 | { | |
1019 | if (forbidden_pages_map) | |
1020 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
1021 | } | |
1022 | ||
1023 | /** | |
ef96f639 RW |
1024 | * mark_nosave_pages - Mark pages that should not be saved. |
1025 | * @bm: Memory bitmap. | |
1026 | * | |
1027 | * Set the bits in @bm that correspond to the page frames the contents of which | |
1028 | * should not be saved. | |
74dfd666 | 1029 | */ |
74dfd666 RW |
1030 | static void mark_nosave_pages(struct memory_bitmap *bm) |
1031 | { | |
1032 | struct nosave_region *region; | |
1033 | ||
1034 | if (list_empty(&nosave_regions)) | |
1035 | return; | |
1036 | ||
1037 | list_for_each_entry(region, &nosave_regions, list) { | |
1038 | unsigned long pfn; | |
1039 | ||
69f1d475 BH |
1040 | pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n", |
1041 | (unsigned long long) region->start_pfn << PAGE_SHIFT, | |
1042 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
1043 | - 1); | |
74dfd666 RW |
1044 | |
1045 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
1046 | if (pfn_valid(pfn)) { |
1047 | /* | |
1048 | * It is safe to ignore the result of | |
1049 | * mem_bm_set_bit_check() here, since we won't | |
1050 | * touch the PFNs for which the error is | |
1051 | * returned anyway. | |
1052 | */ | |
1053 | mem_bm_set_bit_check(bm, pfn); | |
1054 | } | |
74dfd666 RW |
1055 | } |
1056 | } | |
1057 | ||
1058 | /** | |
ef96f639 RW |
1059 | * create_basic_memory_bitmaps - Create bitmaps to hold basic page information. |
1060 | * | |
1061 | * Create bitmaps needed for marking page frames that should not be saved and | |
1062 | * free page frames. The forbidden_pages_map and free_pages_map pointers are | |
1063 | * only modified if everything goes well, because we don't want the bits to be | |
1064 | * touched before both bitmaps are set up. | |
74dfd666 | 1065 | */ |
74dfd666 RW |
1066 | int create_basic_memory_bitmaps(void) |
1067 | { | |
1068 | struct memory_bitmap *bm1, *bm2; | |
1069 | int error = 0; | |
1070 | ||
aab17289 RW |
1071 | if (forbidden_pages_map && free_pages_map) |
1072 | return 0; | |
1073 | else | |
1074 | BUG_ON(forbidden_pages_map || free_pages_map); | |
74dfd666 | 1075 | |
0709db60 | 1076 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1077 | if (!bm1) |
1078 | return -ENOMEM; | |
1079 | ||
0709db60 | 1080 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1081 | if (error) |
1082 | goto Free_first_object; | |
1083 | ||
0709db60 | 1084 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1085 | if (!bm2) |
1086 | goto Free_first_bitmap; | |
1087 | ||
0709db60 | 1088 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1089 | if (error) |
1090 | goto Free_second_object; | |
1091 | ||
1092 | forbidden_pages_map = bm1; | |
1093 | free_pages_map = bm2; | |
1094 | mark_nosave_pages(forbidden_pages_map); | |
1095 | ||
23976728 | 1096 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
1097 | |
1098 | return 0; | |
1099 | ||
1100 | Free_second_object: | |
1101 | kfree(bm2); | |
1102 | Free_first_bitmap: | |
1103 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1104 | Free_first_object: | |
1105 | kfree(bm1); | |
1106 | return -ENOMEM; | |
1107 | } | |
1108 | ||
1109 | /** | |
ef96f639 RW |
1110 | * free_basic_memory_bitmaps - Free memory bitmaps holding basic information. |
1111 | * | |
1112 | * Free memory bitmaps allocated by create_basic_memory_bitmaps(). The | |
1113 | * auxiliary pointers are necessary so that the bitmaps themselves are not | |
1114 | * referred to while they are being freed. | |
74dfd666 | 1115 | */ |
74dfd666 RW |
1116 | void free_basic_memory_bitmaps(void) |
1117 | { | |
1118 | struct memory_bitmap *bm1, *bm2; | |
1119 | ||
6a0c7cd3 RW |
1120 | if (WARN_ON(!(forbidden_pages_map && free_pages_map))) |
1121 | return; | |
74dfd666 RW |
1122 | |
1123 | bm1 = forbidden_pages_map; | |
1124 | bm2 = free_pages_map; | |
1125 | forbidden_pages_map = NULL; | |
1126 | free_pages_map = NULL; | |
1127 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1128 | kfree(bm1); | |
1129 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
1130 | kfree(bm2); | |
1131 | ||
23976728 | 1132 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
1133 | } |
1134 | ||
b788db79 | 1135 | /** |
ef96f639 RW |
1136 | * snapshot_additional_pages - Estimate the number of extra pages needed. |
1137 | * @zone: Memory zone to carry out the computation for. | |
1138 | * | |
1139 | * Estimate the number of additional pages needed for setting up a hibernation | |
1140 | * image data structures for @zone (usually, the returned value is greater than | |
1141 | * the exact number). | |
b788db79 | 1142 | */ |
b788db79 RW |
1143 | unsigned int snapshot_additional_pages(struct zone *zone) |
1144 | { | |
f469f02d | 1145 | unsigned int rtree, nodes; |
b788db79 | 1146 | |
f469f02d JR |
1147 | rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); |
1148 | rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node), | |
1149 | LINKED_PAGE_DATA_SIZE); | |
1150 | while (nodes > 1) { | |
1151 | nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL); | |
1152 | rtree += nodes; | |
1153 | } | |
1154 | ||
9047eb62 | 1155 | return 2 * rtree; |
b788db79 RW |
1156 | } |
1157 | ||
8357376d RW |
1158 | #ifdef CONFIG_HIGHMEM |
1159 | /** | |
ef96f639 RW |
1160 | * count_free_highmem_pages - Compute the total number of free highmem pages. |
1161 | * | |
1162 | * The returned number is system-wide. | |
8357376d | 1163 | */ |
8357376d RW |
1164 | static unsigned int count_free_highmem_pages(void) |
1165 | { | |
1166 | struct zone *zone; | |
1167 | unsigned int cnt = 0; | |
1168 | ||
ee99c71c KM |
1169 | for_each_populated_zone(zone) |
1170 | if (is_highmem(zone)) | |
d23ad423 | 1171 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
1172 | |
1173 | return cnt; | |
1174 | } | |
1175 | ||
1176 | /** | |
ef96f639 RW |
1177 | * saveable_highmem_page - Check if a highmem page is saveable. |
1178 | * | |
1179 | * Determine whether a highmem page should be included in a hibernation image. | |
8357376d | 1180 | * |
ef96f639 RW |
1181 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, |
1182 | * and it isn't part of a free chunk of pages. | |
8357376d | 1183 | */ |
846705de | 1184 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
1185 | { |
1186 | struct page *page; | |
1187 | ||
1188 | if (!pfn_valid(pfn)) | |
1189 | return NULL; | |
1190 | ||
1191 | page = pfn_to_page(pfn); | |
846705de RW |
1192 | if (page_zone(page) != zone) |
1193 | return NULL; | |
8357376d RW |
1194 | |
1195 | BUG_ON(!PageHighMem(page)); | |
1196 | ||
7be98234 RW |
1197 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
1198 | PageReserved(page)) | |
8357376d RW |
1199 | return NULL; |
1200 | ||
c6968e73 SG |
1201 | if (page_is_guard(page)) |
1202 | return NULL; | |
1203 | ||
8357376d RW |
1204 | return page; |
1205 | } | |
1206 | ||
1207 | /** | |
ef96f639 | 1208 | * count_highmem_pages - Compute the total number of saveable highmem pages. |
8357376d | 1209 | */ |
fe419535 | 1210 | static unsigned int count_highmem_pages(void) |
8357376d RW |
1211 | { |
1212 | struct zone *zone; | |
1213 | unsigned int n = 0; | |
1214 | ||
98e73dc5 | 1215 | for_each_populated_zone(zone) { |
8357376d RW |
1216 | unsigned long pfn, max_zone_pfn; |
1217 | ||
1218 | if (!is_highmem(zone)) | |
1219 | continue; | |
1220 | ||
1221 | mark_free_pages(zone); | |
c33bc315 | 1222 | max_zone_pfn = zone_end_pfn(zone); |
8357376d | 1223 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1224 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
1225 | n++; |
1226 | } | |
1227 | return n; | |
1228 | } | |
1229 | #else | |
846705de RW |
1230 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
1231 | { | |
1232 | return NULL; | |
1233 | } | |
8357376d RW |
1234 | #endif /* CONFIG_HIGHMEM */ |
1235 | ||
25761b6e | 1236 | /** |
ef96f639 RW |
1237 | * saveable_page - Check if the given page is saveable. |
1238 | * | |
1239 | * Determine whether a non-highmem page should be included in a hibernation | |
1240 | * image. | |
25761b6e | 1241 | * |
ef96f639 RW |
1242 | * We should save the page if it isn't Nosave, and is not in the range |
1243 | * of pages statically defined as 'unsaveable', and it isn't part of | |
1244 | * a free chunk of pages. | |
25761b6e | 1245 | */ |
846705de | 1246 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 1247 | { |
de491861 | 1248 | struct page *page; |
25761b6e RW |
1249 | |
1250 | if (!pfn_valid(pfn)) | |
ae83c5ee | 1251 | return NULL; |
25761b6e RW |
1252 | |
1253 | page = pfn_to_page(pfn); | |
846705de RW |
1254 | if (page_zone(page) != zone) |
1255 | return NULL; | |
ae83c5ee | 1256 | |
8357376d RW |
1257 | BUG_ON(PageHighMem(page)); |
1258 | ||
7be98234 | 1259 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 1260 | return NULL; |
8357376d | 1261 | |
8a235efa RW |
1262 | if (PageReserved(page) |
1263 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 1264 | return NULL; |
25761b6e | 1265 | |
c6968e73 SG |
1266 | if (page_is_guard(page)) |
1267 | return NULL; | |
1268 | ||
ae83c5ee | 1269 | return page; |
25761b6e RW |
1270 | } |
1271 | ||
8357376d | 1272 | /** |
ef96f639 | 1273 | * count_data_pages - Compute the total number of saveable non-highmem pages. |
8357376d | 1274 | */ |
fe419535 | 1275 | static unsigned int count_data_pages(void) |
25761b6e RW |
1276 | { |
1277 | struct zone *zone; | |
ae83c5ee | 1278 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 1279 | unsigned int n = 0; |
25761b6e | 1280 | |
98e73dc5 | 1281 | for_each_populated_zone(zone) { |
25761b6e RW |
1282 | if (is_highmem(zone)) |
1283 | continue; | |
8357376d | 1284 | |
25761b6e | 1285 | mark_free_pages(zone); |
c33bc315 | 1286 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee | 1287 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1288 | if (saveable_page(zone, pfn)) |
8357376d | 1289 | n++; |
25761b6e | 1290 | } |
a0f49651 | 1291 | return n; |
25761b6e RW |
1292 | } |
1293 | ||
ef96f639 RW |
1294 | /* |
1295 | * This is needed, because copy_page and memcpy are not usable for copying | |
8357376d RW |
1296 | * task structs. |
1297 | */ | |
1298 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
1299 | { |
1300 | int n; | |
1301 | ||
f623f0db RW |
1302 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
1303 | *dst++ = *src++; | |
1304 | } | |
1305 | ||
8a235efa | 1306 | /** |
ef96f639 RW |
1307 | * safe_copy_page - Copy a page in a safe way. |
1308 | * | |
1309 | * Check if the page we are going to copy is marked as present in the kernel | |
1310 | * page tables (this always is the case if CONFIG_DEBUG_PAGEALLOC is not set | |
1311 | * and in that case kernel_page_present() always returns 'true'). | |
8a235efa RW |
1312 | */ |
1313 | static void safe_copy_page(void *dst, struct page *s_page) | |
1314 | { | |
1315 | if (kernel_page_present(s_page)) { | |
1316 | do_copy_page(dst, page_address(s_page)); | |
1317 | } else { | |
1318 | kernel_map_pages(s_page, 1, 1); | |
1319 | do_copy_page(dst, page_address(s_page)); | |
1320 | kernel_map_pages(s_page, 1, 0); | |
1321 | } | |
1322 | } | |
1323 | ||
8357376d | 1324 | #ifdef CONFIG_HIGHMEM |
efd5a852 | 1325 | static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn) |
8357376d RW |
1326 | { |
1327 | return is_highmem(zone) ? | |
846705de | 1328 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
1329 | } |
1330 | ||
8a235efa | 1331 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
1332 | { |
1333 | struct page *s_page, *d_page; | |
1334 | void *src, *dst; | |
1335 | ||
1336 | s_page = pfn_to_page(src_pfn); | |
1337 | d_page = pfn_to_page(dst_pfn); | |
1338 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1339 | src = kmap_atomic(s_page); |
1340 | dst = kmap_atomic(d_page); | |
8357376d | 1341 | do_copy_page(dst, src); |
0de9a1e2 CW |
1342 | kunmap_atomic(dst); |
1343 | kunmap_atomic(src); | |
8357376d | 1344 | } else { |
8357376d | 1345 | if (PageHighMem(d_page)) { |
ef96f639 RW |
1346 | /* |
1347 | * The page pointed to by src may contain some kernel | |
8357376d RW |
1348 | * data modified by kmap_atomic() |
1349 | */ | |
8a235efa | 1350 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1351 | dst = kmap_atomic(d_page); |
3ecb01df | 1352 | copy_page(dst, buffer); |
0de9a1e2 | 1353 | kunmap_atomic(dst); |
8357376d | 1354 | } else { |
8a235efa | 1355 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1356 | } |
1357 | } | |
1358 | } | |
1359 | #else | |
846705de | 1360 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1361 | |
8a235efa | 1362 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1363 | { |
8a235efa RW |
1364 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1365 | pfn_to_page(src_pfn)); | |
8357376d RW |
1366 | } |
1367 | #endif /* CONFIG_HIGHMEM */ | |
1368 | ||
efd5a852 RW |
1369 | static void copy_data_pages(struct memory_bitmap *copy_bm, |
1370 | struct memory_bitmap *orig_bm) | |
25761b6e RW |
1371 | { |
1372 | struct zone *zone; | |
b788db79 | 1373 | unsigned long pfn; |
25761b6e | 1374 | |
98e73dc5 | 1375 | for_each_populated_zone(zone) { |
b788db79 RW |
1376 | unsigned long max_zone_pfn; |
1377 | ||
25761b6e | 1378 | mark_free_pages(zone); |
c33bc315 | 1379 | max_zone_pfn = zone_end_pfn(zone); |
b788db79 | 1380 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1381 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1382 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1383 | } |
b788db79 RW |
1384 | memory_bm_position_reset(orig_bm); |
1385 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1386 | for(;;) { |
b788db79 | 1387 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1388 | if (unlikely(pfn == BM_END_OF_MAP)) |
1389 | break; | |
1390 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1391 | } | |
25761b6e RW |
1392 | } |
1393 | ||
8357376d RW |
1394 | /* Total number of image pages */ |
1395 | static unsigned int nr_copy_pages; | |
1396 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1397 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1398 | /* |
1399 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1400 | * before suspending devices. | |
1401 | */ | |
1402 | unsigned int alloc_normal, alloc_highmem; | |
1403 | /* | |
1404 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1405 | * hibernation image pages (during restore) | |
1406 | */ | |
1407 | static struct memory_bitmap orig_bm; | |
1408 | /* | |
1409 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1410 | * will contain copies of saveable pages. During restore it is initially used | |
1411 | * for marking hibernation image pages, but then the set bits from it are | |
1412 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1413 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1414 | * this purpose. | |
1415 | */ | |
1416 | static struct memory_bitmap copy_bm; | |
8357376d | 1417 | |
25761b6e | 1418 | /** |
ef96f639 | 1419 | * swsusp_free - Free pages allocated for hibernation image. |
cd560bb2 | 1420 | * |
ef96f639 RW |
1421 | * Image pages are alocated before snapshot creation, so they need to be |
1422 | * released after resume. | |
25761b6e | 1423 | */ |
25761b6e RW |
1424 | void swsusp_free(void) |
1425 | { | |
fdd64ed5 | 1426 | unsigned long fb_pfn, fr_pfn; |
6efde38f | 1427 | |
fdd64ed5 JR |
1428 | if (!forbidden_pages_map || !free_pages_map) |
1429 | goto out; | |
1430 | ||
1431 | memory_bm_position_reset(forbidden_pages_map); | |
1432 | memory_bm_position_reset(free_pages_map); | |
1433 | ||
1434 | loop: | |
1435 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1436 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1437 | ||
1438 | /* | |
1439 | * Find the next bit set in both bitmaps. This is guaranteed to | |
1440 | * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP. | |
1441 | */ | |
1442 | do { | |
1443 | if (fb_pfn < fr_pfn) | |
1444 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1445 | if (fr_pfn < fb_pfn) | |
1446 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1447 | } while (fb_pfn != fr_pfn); | |
1448 | ||
1449 | if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) { | |
1450 | struct page *page = pfn_to_page(fr_pfn); | |
1451 | ||
1452 | memory_bm_clear_current(forbidden_pages_map); | |
1453 | memory_bm_clear_current(free_pages_map); | |
4c0b6c10 | 1454 | hibernate_restore_unprotect_page(page_address(page)); |
fdd64ed5 JR |
1455 | __free_page(page); |
1456 | goto loop; | |
25761b6e | 1457 | } |
fdd64ed5 JR |
1458 | |
1459 | out: | |
f577eb30 RW |
1460 | nr_copy_pages = 0; |
1461 | nr_meta_pages = 0; | |
75534b50 | 1462 | restore_pblist = NULL; |
6e1819d6 | 1463 | buffer = NULL; |
64a473cb RW |
1464 | alloc_normal = 0; |
1465 | alloc_highmem = 0; | |
4c0b6c10 | 1466 | hibernate_restore_protection_end(); |
25761b6e RW |
1467 | } |
1468 | ||
4bb33435 RW |
1469 | /* Helper functions used for the shrinking of memory. */ |
1470 | ||
1471 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1472 | ||
fe419535 | 1473 | /** |
ef96f639 | 1474 | * preallocate_image_pages - Allocate a number of pages for hibernation image. |
4bb33435 RW |
1475 | * @nr_pages: Number of page frames to allocate. |
1476 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1477 | * |
4bb33435 RW |
1478 | * Return value: Number of page frames actually allocated |
1479 | */ | |
1480 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1481 | { | |
1482 | unsigned long nr_alloc = 0; | |
1483 | ||
1484 | while (nr_pages > 0) { | |
64a473cb RW |
1485 | struct page *page; |
1486 | ||
1487 | page = alloc_image_page(mask); | |
1488 | if (!page) | |
4bb33435 | 1489 | break; |
64a473cb RW |
1490 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1491 | if (PageHighMem(page)) | |
1492 | alloc_highmem++; | |
1493 | else | |
1494 | alloc_normal++; | |
4bb33435 RW |
1495 | nr_pages--; |
1496 | nr_alloc++; | |
1497 | } | |
1498 | ||
1499 | return nr_alloc; | |
1500 | } | |
1501 | ||
6715045d RW |
1502 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1503 | unsigned long avail_normal) | |
4bb33435 | 1504 | { |
6715045d RW |
1505 | unsigned long alloc; |
1506 | ||
1507 | if (avail_normal <= alloc_normal) | |
1508 | return 0; | |
1509 | ||
1510 | alloc = avail_normal - alloc_normal; | |
1511 | if (nr_pages < alloc) | |
1512 | alloc = nr_pages; | |
1513 | ||
1514 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1515 | } |
1516 | ||
1517 | #ifdef CONFIG_HIGHMEM | |
1518 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1519 | { | |
1520 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1521 | } | |
1522 | ||
1523 | /** | |
ef96f639 | 1524 | * __fraction - Compute (an approximation of) x * (multiplier / base). |
fe419535 | 1525 | */ |
4bb33435 RW |
1526 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1527 | { | |
1528 | x *= multiplier; | |
1529 | do_div(x, base); | |
1530 | return (unsigned long)x; | |
1531 | } | |
fe419535 | 1532 | |
4bb33435 | 1533 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
efd5a852 RW |
1534 | unsigned long highmem, |
1535 | unsigned long total) | |
fe419535 | 1536 | { |
4bb33435 RW |
1537 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1538 | ||
1539 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1540 | } |
4bb33435 RW |
1541 | #else /* CONFIG_HIGHMEM */ |
1542 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1543 | { | |
1544 | return 0; | |
1545 | } | |
1546 | ||
1547 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
efd5a852 RW |
1548 | unsigned long highmem, |
1549 | unsigned long total) | |
4bb33435 RW |
1550 | { |
1551 | return 0; | |
1552 | } | |
1553 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1554 | |
4bb33435 | 1555 | /** |
ef96f639 | 1556 | * free_unnecessary_pages - Release preallocated pages not needed for the image. |
64a473cb | 1557 | */ |
a64fc82c | 1558 | static unsigned long free_unnecessary_pages(void) |
64a473cb | 1559 | { |
a64fc82c | 1560 | unsigned long save, to_free_normal, to_free_highmem, free; |
64a473cb | 1561 | |
6715045d RW |
1562 | save = count_data_pages(); |
1563 | if (alloc_normal >= save) { | |
1564 | to_free_normal = alloc_normal - save; | |
1565 | save = 0; | |
1566 | } else { | |
1567 | to_free_normal = 0; | |
1568 | save -= alloc_normal; | |
1569 | } | |
1570 | save += count_highmem_pages(); | |
1571 | if (alloc_highmem >= save) { | |
1572 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1573 | } else { |
1574 | to_free_highmem = 0; | |
4d4cf23c RW |
1575 | save -= alloc_highmem; |
1576 | if (to_free_normal > save) | |
1577 | to_free_normal -= save; | |
1578 | else | |
1579 | to_free_normal = 0; | |
64a473cb | 1580 | } |
a64fc82c | 1581 | free = to_free_normal + to_free_highmem; |
64a473cb RW |
1582 | |
1583 | memory_bm_position_reset(©_bm); | |
1584 | ||
a9c9b442 | 1585 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1586 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1587 | struct page *page = pfn_to_page(pfn); | |
1588 | ||
1589 | if (PageHighMem(page)) { | |
1590 | if (!to_free_highmem) | |
1591 | continue; | |
1592 | to_free_highmem--; | |
1593 | alloc_highmem--; | |
1594 | } else { | |
1595 | if (!to_free_normal) | |
1596 | continue; | |
1597 | to_free_normal--; | |
1598 | alloc_normal--; | |
1599 | } | |
1600 | memory_bm_clear_bit(©_bm, pfn); | |
1601 | swsusp_unset_page_forbidden(page); | |
1602 | swsusp_unset_page_free(page); | |
1603 | __free_page(page); | |
1604 | } | |
a64fc82c WK |
1605 | |
1606 | return free; | |
64a473cb RW |
1607 | } |
1608 | ||
ef4aede3 | 1609 | /** |
ef96f639 | 1610 | * minimum_image_size - Estimate the minimum acceptable size of an image. |
ef4aede3 RW |
1611 | * @saveable: Number of saveable pages in the system. |
1612 | * | |
1613 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1614 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1615 | * preallocating memory. | |
1616 | * | |
1617 | * We assume that the minimum image size should be proportional to | |
1618 | * | |
1619 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1620 | * | |
1621 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
4d434820 | 1622 | * and (3) inactive anonymous pages, (4) active and (5) inactive file pages, |
ef4aede3 RW |
1623 | * minus mapped file pages. |
1624 | */ | |
1625 | static unsigned long minimum_image_size(unsigned long saveable) | |
1626 | { | |
1627 | unsigned long size; | |
1628 | ||
1629 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
599d0c95 MG |
1630 | + global_node_page_state(NR_ACTIVE_ANON) |
1631 | + global_node_page_state(NR_INACTIVE_ANON) | |
1632 | + global_node_page_state(NR_ACTIVE_FILE) | |
1633 | + global_node_page_state(NR_INACTIVE_FILE) | |
1634 | - global_node_page_state(NR_FILE_MAPPED); | |
ef4aede3 RW |
1635 | |
1636 | return saveable <= size ? 0 : saveable - size; | |
1637 | } | |
1638 | ||
64a473cb | 1639 | /** |
ef96f639 | 1640 | * hibernate_preallocate_memory - Preallocate memory for hibernation image. |
4bb33435 RW |
1641 | * |
1642 | * To create a hibernation image it is necessary to make a copy of every page | |
1643 | * frame in use. We also need a number of page frames to be free during | |
1644 | * hibernation for allocations made while saving the image and for device | |
1645 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 RW |
1646 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
1647 | * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through | |
1648 | * /sys/power/reserved_size, respectively). To make this happen, we compute the | |
1649 | * total number of available page frames and allocate at least | |
4bb33435 | 1650 | * |
ddeb6487 RW |
1651 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1652 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1653 | * |
1654 | * of them, which corresponds to the maximum size of a hibernation image. | |
1655 | * | |
1656 | * If image_size is set below the number following from the above formula, | |
1657 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1658 | * pages in the system is below the requested image size or the minimum |
1659 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1660 | */ |
64a473cb | 1661 | int hibernate_preallocate_memory(void) |
fe419535 | 1662 | { |
fe419535 | 1663 | struct zone *zone; |
4bb33435 | 1664 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1665 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
db597605 | 1666 | ktime_t start, stop; |
64a473cb | 1667 | int error; |
fe419535 | 1668 | |
64a473cb | 1669 | printk(KERN_INFO "PM: Preallocating image memory... "); |
db597605 | 1670 | start = ktime_get(); |
fe419535 | 1671 | |
64a473cb RW |
1672 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1673 | if (error) | |
1674 | goto err_out; | |
1675 | ||
1676 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1677 | if (error) | |
1678 | goto err_out; | |
1679 | ||
1680 | alloc_normal = 0; | |
1681 | alloc_highmem = 0; | |
1682 | ||
4bb33435 | 1683 | /* Count the number of saveable data pages. */ |
64a473cb | 1684 | save_highmem = count_highmem_pages(); |
4bb33435 | 1685 | saveable = count_data_pages(); |
fe419535 | 1686 | |
4bb33435 RW |
1687 | /* |
1688 | * Compute the total number of page frames we can use (count) and the | |
1689 | * number of pages needed for image metadata (size). | |
1690 | */ | |
1691 | count = saveable; | |
64a473cb RW |
1692 | saveable += save_highmem; |
1693 | highmem = save_highmem; | |
4bb33435 RW |
1694 | size = 0; |
1695 | for_each_populated_zone(zone) { | |
1696 | size += snapshot_additional_pages(zone); | |
1697 | if (is_highmem(zone)) | |
1698 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1699 | else | |
1700 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1701 | } | |
6715045d | 1702 | avail_normal = count; |
4bb33435 RW |
1703 | count += highmem; |
1704 | count -= totalreserve_pages; | |
1705 | ||
85055dd8 MS |
1706 | /* Add number of pages required for page keys (s390 only). */ |
1707 | size += page_key_additional_pages(saveable); | |
1708 | ||
4bb33435 | 1709 | /* Compute the maximum number of saveable pages to leave in memory. */ |
ddeb6487 RW |
1710 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1711 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1712 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1713 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1714 | if (size > max_size) | |
1715 | size = max_size; | |
1716 | /* | |
266f1a25 RW |
1717 | * If the desired number of image pages is at least as large as the |
1718 | * current number of saveable pages in memory, allocate page frames for | |
1719 | * the image and we're done. | |
4bb33435 | 1720 | */ |
64a473cb RW |
1721 | if (size >= saveable) { |
1722 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1723 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1724 | goto out; |
64a473cb | 1725 | } |
4bb33435 | 1726 | |
ef4aede3 RW |
1727 | /* Estimate the minimum size of the image. */ |
1728 | pages = minimum_image_size(saveable); | |
6715045d RW |
1729 | /* |
1730 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1731 | * accommodate an image of the minimum size (unless it's already too | |
1732 | * small, in which case don't preallocate pages from it at all). | |
1733 | */ | |
1734 | if (avail_normal > pages) | |
1735 | avail_normal -= pages; | |
1736 | else | |
1737 | avail_normal = 0; | |
ef4aede3 RW |
1738 | if (size < pages) |
1739 | size = min_t(unsigned long, pages, max_size); | |
1740 | ||
4bb33435 RW |
1741 | /* |
1742 | * Let the memory management subsystem know that we're going to need a | |
1743 | * large number of page frames to allocate and make it free some memory. | |
1744 | * NOTE: If this is not done, performance will be hurt badly in some | |
1745 | * test cases. | |
1746 | */ | |
1747 | shrink_all_memory(saveable - size); | |
1748 | ||
1749 | /* | |
1750 | * The number of saveable pages in memory was too high, so apply some | |
1751 | * pressure to decrease it. First, make room for the largest possible | |
1752 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1753 | * of the image as much as indicated by 'size' using allocations from |
1754 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1755 | */ |
1756 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
fd432b9f AL |
1757 | alloc = count - max_size; |
1758 | if (alloc > pages_highmem) | |
1759 | alloc -= pages_highmem; | |
1760 | else | |
1761 | alloc = 0; | |
6715045d RW |
1762 | pages = preallocate_image_memory(alloc, avail_normal); |
1763 | if (pages < alloc) { | |
1764 | /* We have exhausted non-highmem pages, try highmem. */ | |
1765 | alloc -= pages; | |
1766 | pages += pages_highmem; | |
1767 | pages_highmem = preallocate_image_highmem(alloc); | |
1768 | if (pages_highmem < alloc) | |
1769 | goto err_out; | |
1770 | pages += pages_highmem; | |
1771 | /* | |
1772 | * size is the desired number of saveable pages to leave in | |
1773 | * memory, so try to preallocate (all memory - size) pages. | |
1774 | */ | |
1775 | alloc = (count - pages) - size; | |
1776 | pages += preallocate_image_highmem(alloc); | |
1777 | } else { | |
1778 | /* | |
1779 | * There are approximately max_size saveable pages at this point | |
1780 | * and we want to reduce this number down to size. | |
1781 | */ | |
1782 | alloc = max_size - size; | |
1783 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1784 | pages_highmem += size; | |
1785 | alloc -= size; | |
1786 | size = preallocate_image_memory(alloc, avail_normal); | |
1787 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1788 | pages += pages_highmem + size; | |
1789 | } | |
4bb33435 | 1790 | |
64a473cb RW |
1791 | /* |
1792 | * We only need as many page frames for the image as there are saveable | |
1793 | * pages in memory, but we have allocated more. Release the excessive | |
1794 | * ones now. | |
1795 | */ | |
a64fc82c | 1796 | pages -= free_unnecessary_pages(); |
4bb33435 RW |
1797 | |
1798 | out: | |
db597605 | 1799 | stop = ktime_get(); |
64a473cb | 1800 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
db597605 | 1801 | swsusp_show_speed(start, stop, pages, "Allocated"); |
fe419535 RW |
1802 | |
1803 | return 0; | |
64a473cb RW |
1804 | |
1805 | err_out: | |
1806 | printk(KERN_CONT "\n"); | |
1807 | swsusp_free(); | |
1808 | return -ENOMEM; | |
fe419535 RW |
1809 | } |
1810 | ||
8357376d RW |
1811 | #ifdef CONFIG_HIGHMEM |
1812 | /** | |
ef96f639 RW |
1813 | * count_pages_for_highmem - Count non-highmem pages needed for copying highmem. |
1814 | * | |
1815 | * Compute the number of non-highmem pages that will be necessary for creating | |
1816 | * copies of highmem pages. | |
1817 | */ | |
8357376d RW |
1818 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) |
1819 | { | |
64a473cb | 1820 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1821 | |
1822 | if (free_highmem >= nr_highmem) | |
1823 | nr_highmem = 0; | |
1824 | else | |
1825 | nr_highmem -= free_highmem; | |
1826 | ||
1827 | return nr_highmem; | |
1828 | } | |
1829 | #else | |
efd5a852 | 1830 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; } |
8357376d | 1831 | #endif /* CONFIG_HIGHMEM */ |
25761b6e RW |
1832 | |
1833 | /** | |
ef96f639 | 1834 | * enough_free_mem - Check if there is enough free memory for the image. |
25761b6e | 1835 | */ |
8357376d | 1836 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1837 | { |
e5e2fa78 | 1838 | struct zone *zone; |
64a473cb | 1839 | unsigned int free = alloc_normal; |
e5e2fa78 | 1840 | |
98e73dc5 | 1841 | for_each_populated_zone(zone) |
8357376d | 1842 | if (!is_highmem(zone)) |
d23ad423 | 1843 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1844 | |
8357376d | 1845 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1846 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1847 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1848 | |
64a473cb | 1849 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1850 | } |
1851 | ||
8357376d RW |
1852 | #ifdef CONFIG_HIGHMEM |
1853 | /** | |
ef96f639 RW |
1854 | * get_highmem_buffer - Allocate a buffer for highmem pages. |
1855 | * | |
1856 | * If there are some highmem pages in the hibernation image, we may need a | |
1857 | * buffer to copy them and/or load their data. | |
8357376d | 1858 | */ |
8357376d RW |
1859 | static inline int get_highmem_buffer(int safe_needed) |
1860 | { | |
1861 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1862 | return buffer ? 0 : -ENOMEM; | |
1863 | } | |
1864 | ||
1865 | /** | |
ef96f639 RW |
1866 | * alloc_highmem_image_pages - Allocate some highmem pages for the image. |
1867 | * | |
1868 | * Try to allocate as many pages as needed, but if the number of free highmem | |
1869 | * pages is less than that, allocate them all. | |
8357376d | 1870 | */ |
efd5a852 RW |
1871 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1872 | unsigned int nr_highmem) | |
8357376d RW |
1873 | { |
1874 | unsigned int to_alloc = count_free_highmem_pages(); | |
1875 | ||
1876 | if (to_alloc > nr_highmem) | |
1877 | to_alloc = nr_highmem; | |
1878 | ||
1879 | nr_highmem -= to_alloc; | |
1880 | while (to_alloc-- > 0) { | |
1881 | struct page *page; | |
1882 | ||
d0164adc | 1883 | page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM); |
8357376d RW |
1884 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1885 | } | |
1886 | return nr_highmem; | |
1887 | } | |
1888 | #else | |
1889 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1890 | ||
efd5a852 RW |
1891 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1892 | unsigned int n) { return 0; } | |
8357376d RW |
1893 | #endif /* CONFIG_HIGHMEM */ |
1894 | ||
1895 | /** | |
ef96f639 | 1896 | * swsusp_alloc - Allocate memory for hibernation image. |
8357376d | 1897 | * |
ef96f639 RW |
1898 | * We first try to allocate as many highmem pages as there are |
1899 | * saveable highmem pages in the system. If that fails, we allocate | |
1900 | * non-highmem pages for the copies of the remaining highmem ones. | |
8357376d | 1901 | * |
ef96f639 RW |
1902 | * In this approach it is likely that the copies of highmem pages will |
1903 | * also be located in the high memory, because of the way in which | |
1904 | * copy_data_pages() works. | |
8357376d | 1905 | */ |
efd5a852 RW |
1906 | static int swsusp_alloc(struct memory_bitmap *orig_bm, |
1907 | struct memory_bitmap *copy_bm, | |
1908 | unsigned int nr_pages, unsigned int nr_highmem) | |
054bd4c1 | 1909 | { |
8357376d | 1910 | if (nr_highmem > 0) { |
2e725a06 | 1911 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1912 | goto err_out; |
1913 | if (nr_highmem > alloc_highmem) { | |
1914 | nr_highmem -= alloc_highmem; | |
1915 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1916 | } | |
8357376d | 1917 | } |
64a473cb RW |
1918 | if (nr_pages > alloc_normal) { |
1919 | nr_pages -= alloc_normal; | |
1920 | while (nr_pages-- > 0) { | |
1921 | struct page *page; | |
1922 | ||
1923 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1924 | if (!page) | |
1925 | goto err_out; | |
1926 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1927 | } | |
25761b6e | 1928 | } |
64a473cb | 1929 | |
b788db79 | 1930 | return 0; |
25761b6e | 1931 | |
64a473cb | 1932 | err_out: |
b788db79 | 1933 | swsusp_free(); |
2e725a06 | 1934 | return -ENOMEM; |
25761b6e RW |
1935 | } |
1936 | ||
722a9f92 | 1937 | asmlinkage __visible int swsusp_save(void) |
25761b6e | 1938 | { |
8357376d | 1939 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1940 | |
07c3bb57 | 1941 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1942 | |
9f8f2172 | 1943 | drain_local_pages(NULL); |
a0f49651 | 1944 | nr_pages = count_data_pages(); |
8357376d | 1945 | nr_highmem = count_highmem_pages(); |
23976728 | 1946 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1947 | |
8357376d | 1948 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1949 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1950 | return -ENOMEM; |
1951 | } | |
1952 | ||
8357376d | 1953 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1954 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1955 | return -ENOMEM; |
8357376d | 1956 | } |
25761b6e | 1957 | |
ef96f639 RW |
1958 | /* |
1959 | * During allocating of suspend pagedir, new cold pages may appear. | |
25761b6e RW |
1960 | * Kill them. |
1961 | */ | |
9f8f2172 | 1962 | drain_local_pages(NULL); |
b788db79 | 1963 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1964 | |
1965 | /* | |
1966 | * End of critical section. From now on, we can write to memory, | |
1967 | * but we should not touch disk. This specially means we must _not_ | |
1968 | * touch swap space! Except we must write out our image of course. | |
1969 | */ | |
1970 | ||
8357376d | 1971 | nr_pages += nr_highmem; |
a0f49651 | 1972 | nr_copy_pages = nr_pages; |
8357376d | 1973 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 1974 | |
23976728 RW |
1975 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
1976 | nr_pages); | |
8357376d | 1977 | |
25761b6e RW |
1978 | return 0; |
1979 | } | |
f577eb30 | 1980 | |
d307c4a8 RW |
1981 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
1982 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 1983 | { |
d307c4a8 | 1984 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 1985 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
1986 | return 0; |
1987 | } | |
1988 | ||
1989 | static char *check_image_kernel(struct swsusp_info *info) | |
1990 | { | |
1991 | if (info->version_code != LINUX_VERSION_CODE) | |
1992 | return "kernel version"; | |
1993 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
1994 | return "system type"; | |
1995 | if (strcmp(info->uts.release,init_utsname()->release)) | |
1996 | return "kernel release"; | |
1997 | if (strcmp(info->uts.version,init_utsname()->version)) | |
1998 | return "version"; | |
1999 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
2000 | return "machine"; | |
2001 | return NULL; | |
2002 | } | |
2003 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
2004 | ||
af508b34 RW |
2005 | unsigned long snapshot_get_image_size(void) |
2006 | { | |
2007 | return nr_copy_pages + nr_meta_pages + 1; | |
2008 | } | |
2009 | ||
d307c4a8 RW |
2010 | static int init_header(struct swsusp_info *info) |
2011 | { | |
2012 | memset(info, 0, sizeof(struct swsusp_info)); | |
0ed5fd13 | 2013 | info->num_physpages = get_num_physpages(); |
f577eb30 | 2014 | info->image_pages = nr_copy_pages; |
af508b34 | 2015 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
2016 | info->size = info->pages; |
2017 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 2018 | return init_header_complete(info); |
f577eb30 RW |
2019 | } |
2020 | ||
2021 | /** | |
ef96f639 RW |
2022 | * pack_pfns - Prepare PFNs for saving. |
2023 | * @bm: Memory bitmap. | |
2024 | * @buf: Memory buffer to store the PFNs in. | |
2025 | * | |
2026 | * PFNs corresponding to set bits in @bm are stored in the area of memory | |
2027 | * pointed to by @buf (1 page at a time). | |
f577eb30 | 2028 | */ |
efd5a852 | 2029 | static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2030 | { |
2031 | int j; | |
2032 | ||
b788db79 | 2033 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
2034 | buf[j] = memory_bm_next_pfn(bm); |
2035 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 2036 | break; |
85055dd8 MS |
2037 | /* Save page key for data page (s390 only). */ |
2038 | page_key_read(buf + j); | |
f577eb30 | 2039 | } |
f577eb30 RW |
2040 | } |
2041 | ||
2042 | /** | |
ef96f639 RW |
2043 | * snapshot_read_next - Get the address to read the next image page from. |
2044 | * @handle: Snapshot handle to be used for the reading. | |
f577eb30 | 2045 | * |
ef96f639 RW |
2046 | * On the first call, @handle should point to a zeroed snapshot_handle |
2047 | * structure. The structure gets populated then and a pointer to it should be | |
2048 | * passed to this function every next time. | |
f577eb30 | 2049 | * |
ef96f639 RW |
2050 | * On success, the function returns a positive number. Then, the caller |
2051 | * is allowed to read up to the returned number of bytes from the memory | |
2052 | * location computed by the data_of() macro. | |
f577eb30 | 2053 | * |
ef96f639 RW |
2054 | * The function returns 0 to indicate the end of the data stream condition, |
2055 | * and negative numbers are returned on errors. If that happens, the structure | |
2056 | * pointed to by @handle is not updated and should not be used any more. | |
f577eb30 | 2057 | */ |
d3c1b24c | 2058 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 2059 | { |
fb13a28b | 2060 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2061 | return 0; |
b788db79 | 2062 | |
f577eb30 RW |
2063 | if (!buffer) { |
2064 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 2065 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
2066 | if (!buffer) |
2067 | return -ENOMEM; | |
2068 | } | |
d3c1b24c | 2069 | if (!handle->cur) { |
d307c4a8 RW |
2070 | int error; |
2071 | ||
2072 | error = init_header((struct swsusp_info *)buffer); | |
2073 | if (error) | |
2074 | return error; | |
f577eb30 | 2075 | handle->buffer = buffer; |
b788db79 RW |
2076 | memory_bm_position_reset(&orig_bm); |
2077 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 2078 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 2079 | clear_page(buffer); |
d3c1b24c JS |
2080 | pack_pfns(buffer, &orig_bm); |
2081 | } else { | |
2082 | struct page *page; | |
b788db79 | 2083 | |
d3c1b24c JS |
2084 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
2085 | if (PageHighMem(page)) { | |
ef96f639 RW |
2086 | /* |
2087 | * Highmem pages are copied to the buffer, | |
d3c1b24c JS |
2088 | * because we can't return with a kmapped |
2089 | * highmem page (we may not be called again). | |
2090 | */ | |
2091 | void *kaddr; | |
8357376d | 2092 | |
0de9a1e2 | 2093 | kaddr = kmap_atomic(page); |
3ecb01df | 2094 | copy_page(buffer, kaddr); |
0de9a1e2 | 2095 | kunmap_atomic(kaddr); |
d3c1b24c JS |
2096 | handle->buffer = buffer; |
2097 | } else { | |
2098 | handle->buffer = page_address(page); | |
f577eb30 | 2099 | } |
f577eb30 | 2100 | } |
d3c1b24c JS |
2101 | handle->cur++; |
2102 | return PAGE_SIZE; | |
f577eb30 RW |
2103 | } |
2104 | ||
6dbecfd3 RW |
2105 | static void duplicate_memory_bitmap(struct memory_bitmap *dst, |
2106 | struct memory_bitmap *src) | |
2107 | { | |
2108 | unsigned long pfn; | |
2109 | ||
2110 | memory_bm_position_reset(src); | |
2111 | pfn = memory_bm_next_pfn(src); | |
2112 | while (pfn != BM_END_OF_MAP) { | |
2113 | memory_bm_set_bit(dst, pfn); | |
2114 | pfn = memory_bm_next_pfn(src); | |
2115 | } | |
2116 | } | |
2117 | ||
f577eb30 | 2118 | /** |
ef96f639 RW |
2119 | * mark_unsafe_pages - Mark pages that were used before hibernation. |
2120 | * | |
2121 | * Mark the pages that cannot be used for storing the image during restoration, | |
2122 | * because they conflict with the pages that had been used before hibernation. | |
f577eb30 | 2123 | */ |
6dbecfd3 | 2124 | static void mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 | 2125 | { |
6dbecfd3 | 2126 | unsigned long pfn; |
f577eb30 | 2127 | |
6dbecfd3 RW |
2128 | /* Clear the "free"/"unsafe" bit for all PFNs */ |
2129 | memory_bm_position_reset(free_pages_map); | |
2130 | pfn = memory_bm_next_pfn(free_pages_map); | |
2131 | while (pfn != BM_END_OF_MAP) { | |
2132 | memory_bm_clear_current(free_pages_map); | |
2133 | pfn = memory_bm_next_pfn(free_pages_map); | |
f577eb30 RW |
2134 | } |
2135 | ||
6dbecfd3 RW |
2136 | /* Mark pages that correspond to the "original" PFNs as "unsafe" */ |
2137 | duplicate_memory_bitmap(free_pages_map, bm); | |
f577eb30 | 2138 | |
940864dd | 2139 | allocated_unsafe_pages = 0; |
f577eb30 RW |
2140 | } |
2141 | ||
d307c4a8 | 2142 | static int check_header(struct swsusp_info *info) |
f577eb30 | 2143 | { |
d307c4a8 | 2144 | char *reason; |
f577eb30 | 2145 | |
d307c4a8 | 2146 | reason = check_image_kernel(info); |
0ed5fd13 | 2147 | if (!reason && info->num_physpages != get_num_physpages()) |
f577eb30 | 2148 | reason = "memory size"; |
f577eb30 | 2149 | if (reason) { |
23976728 | 2150 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
2151 | return -EPERM; |
2152 | } | |
2153 | return 0; | |
2154 | } | |
2155 | ||
2156 | /** | |
ef96f639 | 2157 | * load header - Check the image header and copy the data from it. |
f577eb30 | 2158 | */ |
efd5a852 | 2159 | static int load_header(struct swsusp_info *info) |
f577eb30 RW |
2160 | { |
2161 | int error; | |
f577eb30 | 2162 | |
940864dd | 2163 | restore_pblist = NULL; |
f577eb30 RW |
2164 | error = check_header(info); |
2165 | if (!error) { | |
f577eb30 RW |
2166 | nr_copy_pages = info->image_pages; |
2167 | nr_meta_pages = info->pages - info->image_pages - 1; | |
2168 | } | |
2169 | return error; | |
2170 | } | |
2171 | ||
2172 | /** | |
ef96f639 RW |
2173 | * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap. |
2174 | * @bm: Memory bitmap. | |
2175 | * @buf: Area of memory containing the PFNs. | |
2176 | * | |
2177 | * For each element of the array pointed to by @buf (1 page at a time), set the | |
2178 | * corresponding bit in @bm. | |
f577eb30 | 2179 | */ |
69643279 | 2180 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2181 | { |
2182 | int j; | |
2183 | ||
940864dd RW |
2184 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
2185 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
2186 | break; | |
2187 | ||
85055dd8 MS |
2188 | /* Extract and buffer page key for data page (s390 only). */ |
2189 | page_key_memorize(buf + j); | |
2190 | ||
6dbecfd3 | 2191 | if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j])) |
69643279 RW |
2192 | memory_bm_set_bit(bm, buf[j]); |
2193 | else | |
2194 | return -EFAULT; | |
f577eb30 | 2195 | } |
69643279 RW |
2196 | |
2197 | return 0; | |
f577eb30 RW |
2198 | } |
2199 | ||
8357376d | 2200 | #ifdef CONFIG_HIGHMEM |
ef96f639 RW |
2201 | /* |
2202 | * struct highmem_pbe is used for creating the list of highmem pages that | |
8357376d RW |
2203 | * should be restored atomically during the resume from disk, because the page |
2204 | * frames they have occupied before the suspend are in use. | |
2205 | */ | |
2206 | struct highmem_pbe { | |
2207 | struct page *copy_page; /* data is here now */ | |
2208 | struct page *orig_page; /* data was here before the suspend */ | |
2209 | struct highmem_pbe *next; | |
2210 | }; | |
2211 | ||
ef96f639 RW |
2212 | /* |
2213 | * List of highmem PBEs needed for restoring the highmem pages that were | |
8357376d RW |
2214 | * allocated before the suspend and included in the suspend image, but have |
2215 | * also been allocated by the "resume" kernel, so their contents cannot be | |
2216 | * written directly to their "original" page frames. | |
2217 | */ | |
2218 | static struct highmem_pbe *highmem_pblist; | |
2219 | ||
2220 | /** | |
ef96f639 RW |
2221 | * count_highmem_image_pages - Compute the number of highmem pages in the image. |
2222 | * @bm: Memory bitmap. | |
2223 | * | |
2224 | * The bits in @bm that correspond to image pages are assumed to be set. | |
8357376d | 2225 | */ |
8357376d RW |
2226 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) |
2227 | { | |
2228 | unsigned long pfn; | |
2229 | unsigned int cnt = 0; | |
2230 | ||
2231 | memory_bm_position_reset(bm); | |
2232 | pfn = memory_bm_next_pfn(bm); | |
2233 | while (pfn != BM_END_OF_MAP) { | |
2234 | if (PageHighMem(pfn_to_page(pfn))) | |
2235 | cnt++; | |
2236 | ||
2237 | pfn = memory_bm_next_pfn(bm); | |
2238 | } | |
2239 | return cnt; | |
2240 | } | |
2241 | ||
8357376d RW |
2242 | static unsigned int safe_highmem_pages; |
2243 | ||
2244 | static struct memory_bitmap *safe_highmem_bm; | |
2245 | ||
ef96f639 RW |
2246 | /** |
2247 | * prepare_highmem_image - Allocate memory for loading highmem data from image. | |
2248 | * @bm: Pointer to an uninitialized memory bitmap structure. | |
2249 | * @nr_highmem_p: Pointer to the number of highmem image pages. | |
2250 | * | |
2251 | * Try to allocate as many highmem pages as there are highmem image pages | |
2252 | * (@nr_highmem_p points to the variable containing the number of highmem image | |
2253 | * pages). The pages that are "safe" (ie. will not be overwritten when the | |
2254 | * hibernation image is restored entirely) have the corresponding bits set in | |
2255 | * @bm (it must be unitialized). | |
2256 | * | |
2257 | * NOTE: This function should not be called if there are no highmem image pages. | |
2258 | */ | |
efd5a852 RW |
2259 | static int prepare_highmem_image(struct memory_bitmap *bm, |
2260 | unsigned int *nr_highmem_p) | |
8357376d RW |
2261 | { |
2262 | unsigned int to_alloc; | |
2263 | ||
2264 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
2265 | return -ENOMEM; | |
2266 | ||
2267 | if (get_highmem_buffer(PG_SAFE)) | |
2268 | return -ENOMEM; | |
2269 | ||
2270 | to_alloc = count_free_highmem_pages(); | |
2271 | if (to_alloc > *nr_highmem_p) | |
2272 | to_alloc = *nr_highmem_p; | |
2273 | else | |
2274 | *nr_highmem_p = to_alloc; | |
2275 | ||
2276 | safe_highmem_pages = 0; | |
2277 | while (to_alloc-- > 0) { | |
2278 | struct page *page; | |
2279 | ||
2280 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 2281 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
2282 | /* The page is "safe", set its bit the bitmap */ |
2283 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
2284 | safe_highmem_pages++; | |
2285 | } | |
2286 | /* Mark the page as allocated */ | |
7be98234 RW |
2287 | swsusp_set_page_forbidden(page); |
2288 | swsusp_set_page_free(page); | |
8357376d RW |
2289 | } |
2290 | memory_bm_position_reset(bm); | |
2291 | safe_highmem_bm = bm; | |
2292 | return 0; | |
2293 | } | |
2294 | ||
ef96f639 RW |
2295 | static struct page *last_highmem_page; |
2296 | ||
8357376d | 2297 | /** |
ef96f639 RW |
2298 | * get_highmem_page_buffer - Prepare a buffer to store a highmem image page. |
2299 | * | |
2300 | * For a given highmem image page get a buffer that suspend_write_next() should | |
2301 | * return to its caller to write to. | |
8357376d | 2302 | * |
ef96f639 RW |
2303 | * If the page is to be saved to its "original" page frame or a copy of |
2304 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
2305 | * the copy of the page is to be made in normal memory, so the address of | |
2306 | * the copy is returned. | |
8357376d | 2307 | * |
ef96f639 RW |
2308 | * If @buffer is returned, the caller of suspend_write_next() will write |
2309 | * the page's contents to @buffer, so they will have to be copied to the | |
2310 | * right location on the next call to suspend_write_next() and it is done | |
2311 | * with the help of copy_last_highmem_page(). For this purpose, if | |
2312 | * @buffer is returned, @last_highmem_page is set to the page to which | |
2313 | * the data will have to be copied from @buffer. | |
8357376d | 2314 | */ |
efd5a852 RW |
2315 | static void *get_highmem_page_buffer(struct page *page, |
2316 | struct chain_allocator *ca) | |
8357376d RW |
2317 | { |
2318 | struct highmem_pbe *pbe; | |
2319 | void *kaddr; | |
2320 | ||
7be98234 | 2321 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
ef96f639 RW |
2322 | /* |
2323 | * We have allocated the "original" page frame and we can | |
8357376d RW |
2324 | * use it directly to store the loaded page. |
2325 | */ | |
2326 | last_highmem_page = page; | |
2327 | return buffer; | |
2328 | } | |
ef96f639 RW |
2329 | /* |
2330 | * The "original" page frame has not been allocated and we have to | |
8357376d RW |
2331 | * use a "safe" page frame to store the loaded page. |
2332 | */ | |
2333 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
2334 | if (!pbe) { | |
2335 | swsusp_free(); | |
69643279 | 2336 | return ERR_PTR(-ENOMEM); |
8357376d RW |
2337 | } |
2338 | pbe->orig_page = page; | |
2339 | if (safe_highmem_pages > 0) { | |
2340 | struct page *tmp; | |
2341 | ||
2342 | /* Copy of the page will be stored in high memory */ | |
2343 | kaddr = buffer; | |
2344 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
2345 | safe_highmem_pages--; | |
2346 | last_highmem_page = tmp; | |
2347 | pbe->copy_page = tmp; | |
2348 | } else { | |
2349 | /* Copy of the page will be stored in normal memory */ | |
2350 | kaddr = safe_pages_list; | |
2351 | safe_pages_list = safe_pages_list->next; | |
2352 | pbe->copy_page = virt_to_page(kaddr); | |
2353 | } | |
2354 | pbe->next = highmem_pblist; | |
2355 | highmem_pblist = pbe; | |
2356 | return kaddr; | |
2357 | } | |
2358 | ||
2359 | /** | |
ef96f639 RW |
2360 | * copy_last_highmem_page - Copy most the most recent highmem image page. |
2361 | * | |
2362 | * Copy the contents of a highmem image from @buffer, where the caller of | |
2363 | * snapshot_write_next() has stored them, to the right location represented by | |
2364 | * @last_highmem_page . | |
8357376d | 2365 | */ |
8357376d RW |
2366 | static void copy_last_highmem_page(void) |
2367 | { | |
2368 | if (last_highmem_page) { | |
2369 | void *dst; | |
2370 | ||
0de9a1e2 | 2371 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2372 | copy_page(dst, buffer); |
0de9a1e2 | 2373 | kunmap_atomic(dst); |
8357376d RW |
2374 | last_highmem_page = NULL; |
2375 | } | |
2376 | } | |
2377 | ||
2378 | static inline int last_highmem_page_copied(void) | |
2379 | { | |
2380 | return !last_highmem_page; | |
2381 | } | |
2382 | ||
2383 | static inline void free_highmem_data(void) | |
2384 | { | |
2385 | if (safe_highmem_bm) | |
2386 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2387 | ||
2388 | if (buffer) | |
2389 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2390 | } | |
2391 | #else | |
efd5a852 | 2392 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } |
8357376d | 2393 | |
efd5a852 RW |
2394 | static inline int prepare_highmem_image(struct memory_bitmap *bm, |
2395 | unsigned int *nr_highmem_p) { return 0; } | |
8357376d | 2396 | |
efd5a852 RW |
2397 | static inline void *get_highmem_page_buffer(struct page *page, |
2398 | struct chain_allocator *ca) | |
8357376d | 2399 | { |
69643279 | 2400 | return ERR_PTR(-EINVAL); |
8357376d RW |
2401 | } |
2402 | ||
2403 | static inline void copy_last_highmem_page(void) {} | |
2404 | static inline int last_highmem_page_copied(void) { return 1; } | |
2405 | static inline void free_highmem_data(void) {} | |
2406 | #endif /* CONFIG_HIGHMEM */ | |
2407 | ||
ef96f639 RW |
2408 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2409 | ||
f577eb30 | 2410 | /** |
ef96f639 RW |
2411 | * prepare_image - Make room for loading hibernation image. |
2412 | * @new_bm: Unitialized memory bitmap structure. | |
2413 | * @bm: Memory bitmap with unsafe pages marked. | |
2414 | * | |
2415 | * Use @bm to mark the pages that will be overwritten in the process of | |
2416 | * restoring the system memory state from the suspend image ("unsafe" pages) | |
2417 | * and allocate memory for the image. | |
968808b8 | 2418 | * |
ef96f639 RW |
2419 | * The idea is to allocate a new memory bitmap first and then allocate |
2420 | * as many pages as needed for image data, but without specifying what those | |
2421 | * pages will be used for just yet. Instead, we mark them all as allocated and | |
2422 | * create a lists of "safe" pages to be used later. On systems with high | |
2423 | * memory a list of "safe" highmem pages is created too. | |
f577eb30 | 2424 | */ |
efd5a852 | 2425 | static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) |
f577eb30 | 2426 | { |
8357376d | 2427 | unsigned int nr_pages, nr_highmem; |
9c744481 | 2428 | struct linked_page *lp; |
940864dd | 2429 | int error; |
f577eb30 | 2430 | |
8357376d RW |
2431 | /* If there is no highmem, the buffer will not be necessary */ |
2432 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2433 | buffer = NULL; | |
2434 | ||
2435 | nr_highmem = count_highmem_image_pages(bm); | |
6dbecfd3 | 2436 | mark_unsafe_pages(bm); |
940864dd RW |
2437 | |
2438 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2439 | if (error) | |
2440 | goto Free; | |
2441 | ||
2442 | duplicate_memory_bitmap(new_bm, bm); | |
2443 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2444 | if (nr_highmem > 0) { |
2445 | error = prepare_highmem_image(bm, &nr_highmem); | |
2446 | if (error) | |
2447 | goto Free; | |
2448 | } | |
ef96f639 RW |
2449 | /* |
2450 | * Reserve some safe pages for potential later use. | |
940864dd RW |
2451 | * |
2452 | * NOTE: This way we make sure there will be enough safe pages for the | |
2453 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2454 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
9c744481 RW |
2455 | * |
2456 | * nr_copy_pages cannot be less than allocated_unsafe_pages too. | |
940864dd | 2457 | */ |
8357376d | 2458 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2459 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2460 | while (nr_pages > 0) { | |
8357376d | 2461 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2462 | if (!lp) { |
f577eb30 | 2463 | error = -ENOMEM; |
940864dd RW |
2464 | goto Free; |
2465 | } | |
9c744481 RW |
2466 | lp->next = safe_pages_list; |
2467 | safe_pages_list = lp; | |
940864dd | 2468 | nr_pages--; |
f577eb30 | 2469 | } |
940864dd | 2470 | /* Preallocate memory for the image */ |
8357376d | 2471 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2472 | while (nr_pages > 0) { |
2473 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2474 | if (!lp) { | |
2475 | error = -ENOMEM; | |
2476 | goto Free; | |
2477 | } | |
7be98234 | 2478 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2479 | /* The page is "safe", add it to the list */ |
2480 | lp->next = safe_pages_list; | |
2481 | safe_pages_list = lp; | |
968808b8 | 2482 | } |
940864dd | 2483 | /* Mark the page as allocated */ |
7be98234 RW |
2484 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2485 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2486 | nr_pages--; |
968808b8 | 2487 | } |
940864dd RW |
2488 | return 0; |
2489 | ||
59a49335 | 2490 | Free: |
940864dd | 2491 | swsusp_free(); |
f577eb30 RW |
2492 | return error; |
2493 | } | |
2494 | ||
940864dd | 2495 | /** |
ef96f639 RW |
2496 | * get_buffer - Get the address to store the next image data page. |
2497 | * | |
2498 | * Get the address that snapshot_write_next() should return to its caller to | |
2499 | * write to. | |
940864dd | 2500 | */ |
940864dd | 2501 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) |
968808b8 | 2502 | { |
940864dd | 2503 | struct pbe *pbe; |
69643279 RW |
2504 | struct page *page; |
2505 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2506 | |
69643279 RW |
2507 | if (pfn == BM_END_OF_MAP) |
2508 | return ERR_PTR(-EFAULT); | |
2509 | ||
2510 | page = pfn_to_page(pfn); | |
8357376d RW |
2511 | if (PageHighMem(page)) |
2512 | return get_highmem_page_buffer(page, ca); | |
2513 | ||
7be98234 | 2514 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
ef96f639 RW |
2515 | /* |
2516 | * We have allocated the "original" page frame and we can | |
940864dd | 2517 | * use it directly to store the loaded page. |
968808b8 | 2518 | */ |
940864dd RW |
2519 | return page_address(page); |
2520 | ||
ef96f639 RW |
2521 | /* |
2522 | * The "original" page frame has not been allocated and we have to | |
940864dd | 2523 | * use a "safe" page frame to store the loaded page. |
968808b8 | 2524 | */ |
940864dd RW |
2525 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2526 | if (!pbe) { | |
2527 | swsusp_free(); | |
69643279 | 2528 | return ERR_PTR(-ENOMEM); |
940864dd | 2529 | } |
8357376d RW |
2530 | pbe->orig_address = page_address(page); |
2531 | pbe->address = safe_pages_list; | |
940864dd RW |
2532 | safe_pages_list = safe_pages_list->next; |
2533 | pbe->next = restore_pblist; | |
2534 | restore_pblist = pbe; | |
8357376d | 2535 | return pbe->address; |
968808b8 RW |
2536 | } |
2537 | ||
f577eb30 | 2538 | /** |
ef96f639 RW |
2539 | * snapshot_write_next - Get the address to store the next image page. |
2540 | * @handle: Snapshot handle structure to guide the writing. | |
f577eb30 | 2541 | * |
ef96f639 RW |
2542 | * On the first call, @handle should point to a zeroed snapshot_handle |
2543 | * structure. The structure gets populated then and a pointer to it should be | |
2544 | * passed to this function every next time. | |
f577eb30 | 2545 | * |
ef96f639 RW |
2546 | * On success, the function returns a positive number. Then, the caller |
2547 | * is allowed to write up to the returned number of bytes to the memory | |
2548 | * location computed by the data_of() macro. | |
f577eb30 | 2549 | * |
ef96f639 RW |
2550 | * The function returns 0 to indicate the "end of file" condition. Negative |
2551 | * numbers are returned on errors, in which cases the structure pointed to by | |
2552 | * @handle is not updated and should not be used any more. | |
f577eb30 | 2553 | */ |
d3c1b24c | 2554 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2555 | { |
940864dd | 2556 | static struct chain_allocator ca; |
f577eb30 RW |
2557 | int error = 0; |
2558 | ||
940864dd | 2559 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2560 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2561 | return 0; |
940864dd | 2562 | |
d3c1b24c JS |
2563 | handle->sync_read = 1; |
2564 | ||
2565 | if (!handle->cur) { | |
8357376d RW |
2566 | if (!buffer) |
2567 | /* This makes the buffer be freed by swsusp_free() */ | |
2568 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2569 | ||
f577eb30 RW |
2570 | if (!buffer) |
2571 | return -ENOMEM; | |
8357376d | 2572 | |
f577eb30 | 2573 | handle->buffer = buffer; |
d3c1b24c JS |
2574 | } else if (handle->cur == 1) { |
2575 | error = load_header(buffer); | |
2576 | if (error) | |
2577 | return error; | |
940864dd | 2578 | |
9c744481 RW |
2579 | safe_pages_list = NULL; |
2580 | ||
d3c1b24c JS |
2581 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2582 | if (error) | |
2583 | return error; | |
2584 | ||
85055dd8 MS |
2585 | /* Allocate buffer for page keys. */ |
2586 | error = page_key_alloc(nr_copy_pages); | |
2587 | if (error) | |
2588 | return error; | |
2589 | ||
4c0b6c10 | 2590 | hibernate_restore_protection_begin(); |
d3c1b24c JS |
2591 | } else if (handle->cur <= nr_meta_pages + 1) { |
2592 | error = unpack_orig_pfns(buffer, ©_bm); | |
2593 | if (error) | |
2594 | return error; | |
940864dd | 2595 | |
d3c1b24c JS |
2596 | if (handle->cur == nr_meta_pages + 1) { |
2597 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2598 | if (error) |
2599 | return error; | |
2600 | ||
d3c1b24c JS |
2601 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2602 | memory_bm_position_reset(&orig_bm); | |
2603 | restore_pblist = NULL; | |
940864dd | 2604 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2605 | handle->sync_read = 0; |
69643279 RW |
2606 | if (IS_ERR(handle->buffer)) |
2607 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2608 | } |
f577eb30 | 2609 | } else { |
d3c1b24c | 2610 | copy_last_highmem_page(); |
85055dd8 MS |
2611 | /* Restore page key for data page (s390 only). */ |
2612 | page_key_write(handle->buffer); | |
4c0b6c10 | 2613 | hibernate_restore_protect_page(handle->buffer); |
d3c1b24c JS |
2614 | handle->buffer = get_buffer(&orig_bm, &ca); |
2615 | if (IS_ERR(handle->buffer)) | |
2616 | return PTR_ERR(handle->buffer); | |
2617 | if (handle->buffer != buffer) | |
2618 | handle->sync_read = 0; | |
f577eb30 | 2619 | } |
d3c1b24c JS |
2620 | handle->cur++; |
2621 | return PAGE_SIZE; | |
f577eb30 RW |
2622 | } |
2623 | ||
8357376d | 2624 | /** |
ef96f639 RW |
2625 | * snapshot_write_finalize - Complete the loading of a hibernation image. |
2626 | * | |
2627 | * Must be called after the last call to snapshot_write_next() in case the last | |
2628 | * page in the image happens to be a highmem page and its contents should be | |
2629 | * stored in highmem. Additionally, it recycles bitmap memory that's not | |
2630 | * necessary any more. | |
8357376d | 2631 | */ |
8357376d RW |
2632 | void snapshot_write_finalize(struct snapshot_handle *handle) |
2633 | { | |
2634 | copy_last_highmem_page(); | |
85055dd8 MS |
2635 | /* Restore page key for data page (s390 only). */ |
2636 | page_key_write(handle->buffer); | |
2637 | page_key_free(); | |
4c0b6c10 | 2638 | hibernate_restore_protect_page(handle->buffer); |
307c5971 | 2639 | /* Do that only if we have loaded the image entirely */ |
d3c1b24c | 2640 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
307c5971 | 2641 | memory_bm_recycle(&orig_bm); |
8357376d RW |
2642 | free_highmem_data(); |
2643 | } | |
2644 | } | |
2645 | ||
f577eb30 RW |
2646 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2647 | { | |
8357376d | 2648 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2649 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2650 | } | |
2651 | ||
8357376d RW |
2652 | #ifdef CONFIG_HIGHMEM |
2653 | /* Assumes that @buf is ready and points to a "safe" page */ | |
efd5a852 RW |
2654 | static inline void swap_two_pages_data(struct page *p1, struct page *p2, |
2655 | void *buf) | |
940864dd | 2656 | { |
8357376d RW |
2657 | void *kaddr1, *kaddr2; |
2658 | ||
0de9a1e2 CW |
2659 | kaddr1 = kmap_atomic(p1); |
2660 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2661 | copy_page(buf, kaddr1); |
2662 | copy_page(kaddr1, kaddr2); | |
2663 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2664 | kunmap_atomic(kaddr2); |
2665 | kunmap_atomic(kaddr1); | |
8357376d RW |
2666 | } |
2667 | ||
2668 | /** | |
ef96f639 RW |
2669 | * restore_highmem - Put highmem image pages into their original locations. |
2670 | * | |
2671 | * For each highmem page that was in use before hibernation and is included in | |
2672 | * the image, and also has been allocated by the "restore" kernel, swap its | |
2673 | * current contents with the previous (ie. "before hibernation") ones. | |
8357376d | 2674 | * |
ef96f639 RW |
2675 | * If the restore eventually fails, we can call this function once again and |
2676 | * restore the highmem state as seen by the restore kernel. | |
8357376d | 2677 | */ |
8357376d RW |
2678 | int restore_highmem(void) |
2679 | { | |
2680 | struct highmem_pbe *pbe = highmem_pblist; | |
2681 | void *buf; | |
2682 | ||
2683 | if (!pbe) | |
2684 | return 0; | |
2685 | ||
2686 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2687 | if (!buf) | |
2688 | return -ENOMEM; | |
2689 | ||
2690 | while (pbe) { | |
2691 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2692 | pbe = pbe->next; | |
2693 | } | |
2694 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2695 | return 0; | |
f577eb30 | 2696 | } |
8357376d | 2697 | #endif /* CONFIG_HIGHMEM */ |