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 | 24 | #include <linux/bootmem.h> |
38b8d208 | 25 | #include <linux/nmi.h> |
25761b6e RW |
26 | #include <linux/syscalls.h> |
27 | #include <linux/console.h> | |
28 | #include <linux/highmem.h> | |
846705de | 29 | #include <linux/list.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
52f5684c | 31 | #include <linux/compiler.h> |
db597605 | 32 | #include <linux/ktime.h> |
25761b6e | 33 | |
7c0f6ba6 | 34 | #include <linux/uaccess.h> |
25761b6e RW |
35 | #include <asm/mmu_context.h> |
36 | #include <asm/pgtable.h> | |
37 | #include <asm/tlbflush.h> | |
38 | #include <asm/io.h> | |
50327ddf LA |
39 | #ifdef CONFIG_STRICT_KERNEL_RWX |
40 | #include <asm/set_memory.h> | |
41 | #endif | |
25761b6e | 42 | |
25761b6e RW |
43 | #include "power.h" |
44 | ||
0f5bf6d0 | 45 | #ifdef CONFIG_STRICT_KERNEL_RWX |
4c0b6c10 RW |
46 | static bool hibernate_restore_protection; |
47 | static bool hibernate_restore_protection_active; | |
48 | ||
49 | void enable_restore_image_protection(void) | |
50 | { | |
51 | hibernate_restore_protection = true; | |
52 | } | |
53 | ||
54 | static inline void hibernate_restore_protection_begin(void) | |
55 | { | |
56 | hibernate_restore_protection_active = hibernate_restore_protection; | |
57 | } | |
58 | ||
59 | static inline void hibernate_restore_protection_end(void) | |
60 | { | |
61 | hibernate_restore_protection_active = false; | |
62 | } | |
63 | ||
64 | static inline void hibernate_restore_protect_page(void *page_address) | |
65 | { | |
66 | if (hibernate_restore_protection_active) | |
67 | set_memory_ro((unsigned long)page_address, 1); | |
68 | } | |
69 | ||
70 | static inline void hibernate_restore_unprotect_page(void *page_address) | |
71 | { | |
72 | if (hibernate_restore_protection_active) | |
73 | set_memory_rw((unsigned long)page_address, 1); | |
74 | } | |
75 | #else | |
76 | static inline void hibernate_restore_protection_begin(void) {} | |
77 | static inline void hibernate_restore_protection_end(void) {} | |
78 | static inline void hibernate_restore_protect_page(void *page_address) {} | |
79 | static inline void hibernate_restore_unprotect_page(void *page_address) {} | |
0f5bf6d0 | 80 | #endif /* CONFIG_STRICT_KERNEL_RWX */ |
4c0b6c10 | 81 | |
74dfd666 RW |
82 | static int swsusp_page_is_free(struct page *); |
83 | static void swsusp_set_page_forbidden(struct page *); | |
84 | static void swsusp_unset_page_forbidden(struct page *); | |
85 | ||
ddeb6487 RW |
86 | /* |
87 | * Number of bytes to reserve for memory allocations made by device drivers | |
88 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
89 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
90 | */ | |
91 | unsigned long reserved_size; | |
92 | ||
93 | void __init hibernate_reserved_size_init(void) | |
94 | { | |
95 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
96 | } | |
97 | ||
fe419535 RW |
98 | /* |
99 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
100 | * When it is set to N, swsusp will do its best to ensure the image |
101 | * size will not exceed N bytes, but if that is impossible, it will | |
102 | * try to create the smallest image possible. | |
fe419535 | 103 | */ |
ac5c24ec RW |
104 | unsigned long image_size; |
105 | ||
106 | void __init hibernate_image_size_init(void) | |
107 | { | |
1c1be3a9 | 108 | image_size = ((totalram_pages * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 109 | } |
fe419535 | 110 | |
ef96f639 RW |
111 | /* |
112 | * List of PBEs needed for restoring the pages that were allocated before | |
8357376d RW |
113 | * the suspend and included in the suspend image, but have also been |
114 | * allocated by the "resume" kernel, so their contents cannot be written | |
115 | * directly to their "original" page frames. | |
116 | */ | |
75534b50 RW |
117 | struct pbe *restore_pblist; |
118 | ||
9c744481 RW |
119 | /* struct linked_page is used to build chains of pages */ |
120 | ||
121 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
122 | ||
123 | struct linked_page { | |
124 | struct linked_page *next; | |
125 | char data[LINKED_PAGE_DATA_SIZE]; | |
126 | } __packed; | |
127 | ||
128 | /* | |
129 | * List of "safe" pages (ie. pages that were not used by the image kernel | |
130 | * before hibernation) that may be used as temporary storage for image kernel | |
131 | * memory contents. | |
132 | */ | |
133 | static struct linked_page *safe_pages_list; | |
134 | ||
8357376d | 135 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 136 | static void *buffer; |
7088a5c0 | 137 | |
0bcd888d RW |
138 | #define PG_ANY 0 |
139 | #define PG_SAFE 1 | |
140 | #define PG_UNSAFE_CLEAR 1 | |
141 | #define PG_UNSAFE_KEEP 0 | |
142 | ||
940864dd | 143 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 144 | |
ef96f639 RW |
145 | /** |
146 | * get_image_page - Allocate a page for a hibernation image. | |
147 | * @gfp_mask: GFP mask for the allocation. | |
148 | * @safe_needed: Get pages that were not used before hibernation (restore only) | |
149 | * | |
150 | * During image restoration, for storing the PBE list and the image data, we can | |
151 | * only use memory pages that do not conflict with the pages used before | |
152 | * hibernation. The "unsafe" pages have PageNosaveFree set and we count them | |
153 | * using allocated_unsafe_pages. | |
154 | * | |
155 | * Each allocated image page is marked as PageNosave and PageNosaveFree so that | |
156 | * swsusp_free() can release it. | |
157 | */ | |
8357376d | 158 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
159 | { |
160 | void *res; | |
161 | ||
162 | res = (void *)get_zeroed_page(gfp_mask); | |
163 | if (safe_needed) | |
7be98234 | 164 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 165 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 166 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 167 | allocated_unsafe_pages++; |
f6143aa6 RW |
168 | res = (void *)get_zeroed_page(gfp_mask); |
169 | } | |
170 | if (res) { | |
7be98234 RW |
171 | swsusp_set_page_forbidden(virt_to_page(res)); |
172 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
173 | } |
174 | return res; | |
175 | } | |
176 | ||
9c744481 RW |
177 | static void *__get_safe_page(gfp_t gfp_mask) |
178 | { | |
179 | if (safe_pages_list) { | |
180 | void *ret = safe_pages_list; | |
181 | ||
182 | safe_pages_list = safe_pages_list->next; | |
183 | memset(ret, 0, PAGE_SIZE); | |
184 | return ret; | |
185 | } | |
186 | return get_image_page(gfp_mask, PG_SAFE); | |
187 | } | |
188 | ||
f6143aa6 RW |
189 | unsigned long get_safe_page(gfp_t gfp_mask) |
190 | { | |
9c744481 | 191 | return (unsigned long)__get_safe_page(gfp_mask); |
8357376d RW |
192 | } |
193 | ||
5b6d15de RW |
194 | static struct page *alloc_image_page(gfp_t gfp_mask) |
195 | { | |
8357376d RW |
196 | struct page *page; |
197 | ||
198 | page = alloc_page(gfp_mask); | |
199 | if (page) { | |
7be98234 RW |
200 | swsusp_set_page_forbidden(page); |
201 | swsusp_set_page_free(page); | |
8357376d RW |
202 | } |
203 | return page; | |
f6143aa6 RW |
204 | } |
205 | ||
307c5971 RW |
206 | static void recycle_safe_page(void *page_address) |
207 | { | |
208 | struct linked_page *lp = page_address; | |
209 | ||
210 | lp->next = safe_pages_list; | |
211 | safe_pages_list = lp; | |
212 | } | |
213 | ||
f6143aa6 | 214 | /** |
ef96f639 RW |
215 | * free_image_page - Free a page allocated for hibernation image. |
216 | * @addr: Address of the page to free. | |
217 | * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page. | |
218 | * | |
219 | * The page to free should have been allocated by get_image_page() (page flags | |
220 | * set by it are affected). | |
f6143aa6 | 221 | */ |
f6143aa6 RW |
222 | static inline void free_image_page(void *addr, int clear_nosave_free) |
223 | { | |
8357376d RW |
224 | struct page *page; |
225 | ||
226 | BUG_ON(!virt_addr_valid(addr)); | |
227 | ||
228 | page = virt_to_page(addr); | |
229 | ||
7be98234 | 230 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 231 | if (clear_nosave_free) |
7be98234 | 232 | swsusp_unset_page_free(page); |
8357376d RW |
233 | |
234 | __free_page(page); | |
f6143aa6 RW |
235 | } |
236 | ||
efd5a852 RW |
237 | static inline void free_list_of_pages(struct linked_page *list, |
238 | int clear_page_nosave) | |
b788db79 RW |
239 | { |
240 | while (list) { | |
241 | struct linked_page *lp = list->next; | |
242 | ||
243 | free_image_page(list, clear_page_nosave); | |
244 | list = lp; | |
245 | } | |
246 | } | |
247 | ||
ef96f639 RW |
248 | /* |
249 | * struct chain_allocator is used for allocating small objects out of | |
250 | * a linked list of pages called 'the chain'. | |
251 | * | |
252 | * The chain grows each time when there is no room for a new object in | |
253 | * the current page. The allocated objects cannot be freed individually. | |
254 | * It is only possible to free them all at once, by freeing the entire | |
255 | * chain. | |
256 | * | |
257 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
258 | * are not much smaller than PAGE_SIZE. | |
259 | */ | |
b788db79 RW |
260 | struct chain_allocator { |
261 | struct linked_page *chain; /* the chain */ | |
262 | unsigned int used_space; /* total size of objects allocated out | |
ef96f639 | 263 | of the current page */ |
b788db79 RW |
264 | gfp_t gfp_mask; /* mask for allocating pages */ |
265 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
266 | }; | |
267 | ||
efd5a852 RW |
268 | static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask, |
269 | int safe_needed) | |
b788db79 RW |
270 | { |
271 | ca->chain = NULL; | |
272 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
273 | ca->gfp_mask = gfp_mask; | |
274 | ca->safe_needed = safe_needed; | |
275 | } | |
276 | ||
277 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
278 | { | |
279 | void *ret; | |
280 | ||
281 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
282 | struct linked_page *lp; | |
283 | ||
9c744481 RW |
284 | lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) : |
285 | get_image_page(ca->gfp_mask, PG_ANY); | |
b788db79 RW |
286 | if (!lp) |
287 | return NULL; | |
288 | ||
289 | lp->next = ca->chain; | |
290 | ca->chain = lp; | |
291 | ca->used_space = 0; | |
292 | } | |
293 | ret = ca->chain->data + ca->used_space; | |
294 | ca->used_space += size; | |
295 | return ret; | |
296 | } | |
297 | ||
b788db79 | 298 | /** |
ef96f639 | 299 | * Data types related to memory bitmaps. |
b788db79 | 300 | * |
ef96f639 RW |
301 | * Memory bitmap is a structure consiting of many linked lists of |
302 | * objects. The main list's elements are of type struct zone_bitmap | |
303 | * and each of them corresonds to one zone. For each zone bitmap | |
304 | * object there is a list of objects of type struct bm_block that | |
305 | * represent each blocks of bitmap in which information is stored. | |
b788db79 | 306 | * |
ef96f639 RW |
307 | * struct memory_bitmap contains a pointer to the main list of zone |
308 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
309 | * and a pointer to the list of pages used for allocating all of the | |
310 | * zone bitmap objects and bitmap block objects. | |
b788db79 | 311 | * |
ef96f639 RW |
312 | * NOTE: It has to be possible to lay out the bitmap in memory |
313 | * using only allocations of order 0. Additionally, the bitmap is | |
314 | * designed to work with arbitrary number of zones (this is over the | |
315 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
b788db79 | 316 | * |
ef96f639 RW |
317 | * struct zone_bitmap contains a pointer to a list of bitmap block |
318 | * objects and a pointer to the bitmap block object that has been | |
319 | * most recently used for setting bits. Additionally, it contains the | |
320 | * PFNs that correspond to the start and end of the represented zone. | |
b788db79 | 321 | * |
ef96f639 RW |
322 | * struct bm_block contains a pointer to the memory page in which |
323 | * information is stored (in the form of a block of bitmap) | |
324 | * It also contains the pfns that correspond to the start and end of | |
325 | * the represented memory area. | |
f469f02d | 326 | * |
ef96f639 RW |
327 | * The memory bitmap is organized as a radix tree to guarantee fast random |
328 | * access to the bits. There is one radix tree for each zone (as returned | |
329 | * from create_mem_extents). | |
f469f02d | 330 | * |
ef96f639 RW |
331 | * One radix tree is represented by one struct mem_zone_bm_rtree. There are |
332 | * two linked lists for the nodes of the tree, one for the inner nodes and | |
333 | * one for the leave nodes. The linked leave nodes are used for fast linear | |
334 | * access of the memory bitmap. | |
f469f02d | 335 | * |
ef96f639 | 336 | * The struct rtree_node represents one node of the radix tree. |
b788db79 RW |
337 | */ |
338 | ||
339 | #define BM_END_OF_MAP (~0UL) | |
340 | ||
8de03073 | 341 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
f469f02d JR |
342 | #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3) |
343 | #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1) | |
b788db79 | 344 | |
f469f02d JR |
345 | /* |
346 | * struct rtree_node is a wrapper struct to link the nodes | |
347 | * of the rtree together for easy linear iteration over | |
348 | * bits and easy freeing | |
349 | */ | |
350 | struct rtree_node { | |
351 | struct list_head list; | |
352 | unsigned long *data; | |
353 | }; | |
354 | ||
355 | /* | |
356 | * struct mem_zone_bm_rtree represents a bitmap used for one | |
357 | * populated memory zone. | |
358 | */ | |
359 | struct mem_zone_bm_rtree { | |
360 | struct list_head list; /* Link Zones together */ | |
361 | struct list_head nodes; /* Radix Tree inner nodes */ | |
362 | struct list_head leaves; /* Radix Tree leaves */ | |
363 | unsigned long start_pfn; /* Zone start page frame */ | |
364 | unsigned long end_pfn; /* Zone end page frame + 1 */ | |
365 | struct rtree_node *rtree; /* Radix Tree Root */ | |
366 | int levels; /* Number of Radix Tree Levels */ | |
367 | unsigned int blocks; /* Number of Bitmap Blocks */ | |
368 | }; | |
369 | ||
b788db79 RW |
370 | /* strcut bm_position is used for browsing memory bitmaps */ |
371 | ||
372 | struct bm_position { | |
3a20cb17 JR |
373 | struct mem_zone_bm_rtree *zone; |
374 | struct rtree_node *node; | |
375 | unsigned long node_pfn; | |
376 | int node_bit; | |
b788db79 RW |
377 | }; |
378 | ||
379 | struct memory_bitmap { | |
f469f02d | 380 | struct list_head zones; |
b788db79 | 381 | struct linked_page *p_list; /* list of pages used to store zone |
ef96f639 RW |
382 | bitmap objects and bitmap block |
383 | objects */ | |
b788db79 RW |
384 | struct bm_position cur; /* most recently used bit position */ |
385 | }; | |
386 | ||
387 | /* Functions that operate on memory bitmaps */ | |
388 | ||
f469f02d JR |
389 | #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long)) |
390 | #if BITS_PER_LONG == 32 | |
391 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2) | |
392 | #else | |
393 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3) | |
394 | #endif | |
395 | #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1) | |
396 | ||
ef96f639 RW |
397 | /** |
398 | * alloc_rtree_node - Allocate a new node and add it to the radix tree. | |
f469f02d | 399 | * |
ef96f639 RW |
400 | * This function is used to allocate inner nodes as well as the |
401 | * leave nodes of the radix tree. It also adds the node to the | |
402 | * corresponding linked list passed in by the *list parameter. | |
f469f02d JR |
403 | */ |
404 | static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed, | |
405 | struct chain_allocator *ca, | |
406 | struct list_head *list) | |
407 | { | |
408 | struct rtree_node *node; | |
409 | ||
410 | node = chain_alloc(ca, sizeof(struct rtree_node)); | |
411 | if (!node) | |
412 | return NULL; | |
413 | ||
414 | node->data = get_image_page(gfp_mask, safe_needed); | |
415 | if (!node->data) | |
416 | return NULL; | |
417 | ||
418 | list_add_tail(&node->list, list); | |
419 | ||
420 | return node; | |
421 | } | |
422 | ||
ef96f639 RW |
423 | /** |
424 | * add_rtree_block - Add a new leave node to the radix tree. | |
f469f02d | 425 | * |
ef96f639 RW |
426 | * The leave nodes need to be allocated in order to keep the leaves |
427 | * linked list in order. This is guaranteed by the zone->blocks | |
428 | * counter. | |
f469f02d JR |
429 | */ |
430 | static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask, | |
431 | int safe_needed, struct chain_allocator *ca) | |
432 | { | |
433 | struct rtree_node *node, *block, **dst; | |
434 | unsigned int levels_needed, block_nr; | |
435 | int i; | |
436 | ||
437 | block_nr = zone->blocks; | |
438 | levels_needed = 0; | |
439 | ||
440 | /* How many levels do we need for this block nr? */ | |
441 | while (block_nr) { | |
442 | levels_needed += 1; | |
443 | block_nr >>= BM_RTREE_LEVEL_SHIFT; | |
444 | } | |
445 | ||
446 | /* Make sure the rtree has enough levels */ | |
447 | for (i = zone->levels; i < levels_needed; i++) { | |
448 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
449 | &zone->nodes); | |
450 | if (!node) | |
451 | return -ENOMEM; | |
452 | ||
453 | node->data[0] = (unsigned long)zone->rtree; | |
454 | zone->rtree = node; | |
455 | zone->levels += 1; | |
456 | } | |
457 | ||
458 | /* Allocate new block */ | |
459 | block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves); | |
460 | if (!block) | |
461 | return -ENOMEM; | |
462 | ||
463 | /* Now walk the rtree to insert the block */ | |
464 | node = zone->rtree; | |
465 | dst = &zone->rtree; | |
466 | block_nr = zone->blocks; | |
467 | for (i = zone->levels; i > 0; i--) { | |
468 | int index; | |
469 | ||
470 | if (!node) { | |
471 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
472 | &zone->nodes); | |
473 | if (!node) | |
474 | return -ENOMEM; | |
475 | *dst = node; | |
476 | } | |
477 | ||
478 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
479 | index &= BM_RTREE_LEVEL_MASK; | |
480 | dst = (struct rtree_node **)&((*dst)->data[index]); | |
481 | node = *dst; | |
482 | } | |
483 | ||
484 | zone->blocks += 1; | |
485 | *dst = block; | |
486 | ||
487 | return 0; | |
488 | } | |
489 | ||
490 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
491 | int clear_nosave_free); | |
492 | ||
ef96f639 RW |
493 | /** |
494 | * create_zone_bm_rtree - Create a radix tree for one zone. | |
f469f02d | 495 | * |
ef96f639 RW |
496 | * Allocated the mem_zone_bm_rtree structure and initializes it. |
497 | * This function also allocated and builds the radix tree for the | |
498 | * zone. | |
f469f02d | 499 | */ |
efd5a852 RW |
500 | static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask, |
501 | int safe_needed, | |
502 | struct chain_allocator *ca, | |
503 | unsigned long start, | |
504 | unsigned long end) | |
f469f02d JR |
505 | { |
506 | struct mem_zone_bm_rtree *zone; | |
507 | unsigned int i, nr_blocks; | |
508 | unsigned long pages; | |
509 | ||
510 | pages = end - start; | |
511 | zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree)); | |
512 | if (!zone) | |
513 | return NULL; | |
514 | ||
515 | INIT_LIST_HEAD(&zone->nodes); | |
516 | INIT_LIST_HEAD(&zone->leaves); | |
517 | zone->start_pfn = start; | |
518 | zone->end_pfn = end; | |
519 | nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); | |
520 | ||
521 | for (i = 0; i < nr_blocks; i++) { | |
522 | if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) { | |
523 | free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR); | |
524 | return NULL; | |
525 | } | |
526 | } | |
527 | ||
528 | return zone; | |
529 | } | |
530 | ||
ef96f639 RW |
531 | /** |
532 | * free_zone_bm_rtree - Free the memory of the radix tree. | |
f469f02d | 533 | * |
ef96f639 RW |
534 | * Free all node pages of the radix tree. The mem_zone_bm_rtree |
535 | * structure itself is not freed here nor are the rtree_node | |
536 | * structs. | |
f469f02d JR |
537 | */ |
538 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
539 | int clear_nosave_free) | |
540 | { | |
541 | struct rtree_node *node; | |
542 | ||
543 | list_for_each_entry(node, &zone->nodes, list) | |
544 | free_image_page(node->data, clear_nosave_free); | |
545 | ||
546 | list_for_each_entry(node, &zone->leaves, list) | |
547 | free_image_page(node->data, clear_nosave_free); | |
548 | } | |
549 | ||
b788db79 RW |
550 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
551 | { | |
3a20cb17 JR |
552 | bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree, |
553 | list); | |
554 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, | |
555 | struct rtree_node, list); | |
556 | bm->cur.node_pfn = 0; | |
557 | bm->cur.node_bit = 0; | |
b788db79 RW |
558 | } |
559 | ||
560 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
561 | ||
846705de RW |
562 | struct mem_extent { |
563 | struct list_head hook; | |
564 | unsigned long start; | |
565 | unsigned long end; | |
566 | }; | |
567 | ||
b788db79 | 568 | /** |
ef96f639 RW |
569 | * free_mem_extents - Free a list of memory extents. |
570 | * @list: List of extents to free. | |
b788db79 | 571 | */ |
846705de RW |
572 | static void free_mem_extents(struct list_head *list) |
573 | { | |
574 | struct mem_extent *ext, *aux; | |
b788db79 | 575 | |
846705de RW |
576 | list_for_each_entry_safe(ext, aux, list, hook) { |
577 | list_del(&ext->hook); | |
578 | kfree(ext); | |
579 | } | |
580 | } | |
581 | ||
582 | /** | |
ef96f639 RW |
583 | * create_mem_extents - Create a list of memory extents. |
584 | * @list: List to put the extents into. | |
585 | * @gfp_mask: Mask to use for memory allocations. | |
586 | * | |
587 | * The extents represent contiguous ranges of PFNs. | |
846705de RW |
588 | */ |
589 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 590 | { |
846705de | 591 | struct zone *zone; |
b788db79 | 592 | |
846705de | 593 | INIT_LIST_HEAD(list); |
b788db79 | 594 | |
ee99c71c | 595 | for_each_populated_zone(zone) { |
846705de RW |
596 | unsigned long zone_start, zone_end; |
597 | struct mem_extent *ext, *cur, *aux; | |
598 | ||
846705de | 599 | zone_start = zone->zone_start_pfn; |
c33bc315 | 600 | zone_end = zone_end_pfn(zone); |
846705de RW |
601 | |
602 | list_for_each_entry(ext, list, hook) | |
603 | if (zone_start <= ext->end) | |
604 | break; | |
b788db79 | 605 | |
846705de RW |
606 | if (&ext->hook == list || zone_end < ext->start) { |
607 | /* New extent is necessary */ | |
608 | struct mem_extent *new_ext; | |
609 | ||
610 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
611 | if (!new_ext) { | |
612 | free_mem_extents(list); | |
613 | return -ENOMEM; | |
614 | } | |
615 | new_ext->start = zone_start; | |
616 | new_ext->end = zone_end; | |
617 | list_add_tail(&new_ext->hook, &ext->hook); | |
618 | continue; | |
619 | } | |
620 | ||
621 | /* Merge this zone's range of PFNs with the existing one */ | |
622 | if (zone_start < ext->start) | |
623 | ext->start = zone_start; | |
624 | if (zone_end > ext->end) | |
625 | ext->end = zone_end; | |
626 | ||
627 | /* More merging may be possible */ | |
628 | cur = ext; | |
629 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
630 | if (zone_end < cur->start) | |
631 | break; | |
632 | if (zone_end < cur->end) | |
633 | ext->end = cur->end; | |
634 | list_del(&cur->hook); | |
635 | kfree(cur); | |
636 | } | |
b788db79 | 637 | } |
846705de RW |
638 | |
639 | return 0; | |
b788db79 RW |
640 | } |
641 | ||
642 | /** | |
ef96f639 RW |
643 | * memory_bm_create - Allocate memory for a memory bitmap. |
644 | */ | |
efd5a852 RW |
645 | static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, |
646 | int safe_needed) | |
b788db79 RW |
647 | { |
648 | struct chain_allocator ca; | |
846705de RW |
649 | struct list_head mem_extents; |
650 | struct mem_extent *ext; | |
651 | int error; | |
b788db79 RW |
652 | |
653 | chain_init(&ca, gfp_mask, safe_needed); | |
f469f02d | 654 | INIT_LIST_HEAD(&bm->zones); |
b788db79 | 655 | |
846705de RW |
656 | error = create_mem_extents(&mem_extents, gfp_mask); |
657 | if (error) | |
658 | return error; | |
b788db79 | 659 | |
846705de | 660 | list_for_each_entry(ext, &mem_extents, hook) { |
f469f02d | 661 | struct mem_zone_bm_rtree *zone; |
f469f02d JR |
662 | |
663 | zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca, | |
664 | ext->start, ext->end); | |
9047eb62 JR |
665 | if (!zone) { |
666 | error = -ENOMEM; | |
f469f02d | 667 | goto Error; |
9047eb62 | 668 | } |
f469f02d | 669 | list_add_tail(&zone->list, &bm->zones); |
b788db79 | 670 | } |
846705de | 671 | |
b788db79 RW |
672 | bm->p_list = ca.chain; |
673 | memory_bm_position_reset(bm); | |
846705de RW |
674 | Exit: |
675 | free_mem_extents(&mem_extents); | |
676 | return error; | |
b788db79 | 677 | |
846705de | 678 | Error: |
b788db79 RW |
679 | bm->p_list = ca.chain; |
680 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 681 | goto Exit; |
b788db79 RW |
682 | } |
683 | ||
684 | /** | |
ef96f639 RW |
685 | * memory_bm_free - Free memory occupied by the memory bitmap. |
686 | * @bm: Memory bitmap. | |
687 | */ | |
b788db79 RW |
688 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
689 | { | |
f469f02d | 690 | struct mem_zone_bm_rtree *zone; |
b788db79 | 691 | |
f469f02d JR |
692 | list_for_each_entry(zone, &bm->zones, list) |
693 | free_zone_bm_rtree(zone, clear_nosave_free); | |
694 | ||
b788db79 | 695 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de | 696 | |
f469f02d | 697 | INIT_LIST_HEAD(&bm->zones); |
b788db79 RW |
698 | } |
699 | ||
700 | /** | |
ef96f639 | 701 | * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap. |
07a33823 | 702 | * |
ef96f639 RW |
703 | * Find the bit in memory bitmap @bm that corresponds to the given PFN. |
704 | * The cur.zone, cur.block and cur.node_pfn members of @bm are updated. | |
705 | * | |
706 | * Walk the radix tree to find the page containing the bit that represents @pfn | |
707 | * and return the position of the bit in @addr and @bit_nr. | |
07a33823 | 708 | */ |
9047eb62 JR |
709 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
710 | void **addr, unsigned int *bit_nr) | |
07a33823 JR |
711 | { |
712 | struct mem_zone_bm_rtree *curr, *zone; | |
713 | struct rtree_node *node; | |
714 | int i, block_nr; | |
715 | ||
3a20cb17 JR |
716 | zone = bm->cur.zone; |
717 | ||
718 | if (pfn >= zone->start_pfn && pfn < zone->end_pfn) | |
719 | goto zone_found; | |
720 | ||
07a33823 JR |
721 | zone = NULL; |
722 | ||
723 | /* Find the right zone */ | |
724 | list_for_each_entry(curr, &bm->zones, list) { | |
725 | if (pfn >= curr->start_pfn && pfn < curr->end_pfn) { | |
726 | zone = curr; | |
727 | break; | |
728 | } | |
729 | } | |
730 | ||
731 | if (!zone) | |
732 | return -EFAULT; | |
733 | ||
3a20cb17 | 734 | zone_found: |
07a33823 | 735 | /* |
ef96f639 RW |
736 | * We have found the zone. Now walk the radix tree to find the leaf node |
737 | * for our PFN. | |
07a33823 | 738 | */ |
3a20cb17 JR |
739 | node = bm->cur.node; |
740 | if (((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn) | |
741 | goto node_found; | |
742 | ||
07a33823 JR |
743 | node = zone->rtree; |
744 | block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT; | |
745 | ||
746 | for (i = zone->levels; i > 0; i--) { | |
747 | int index; | |
748 | ||
749 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
750 | index &= BM_RTREE_LEVEL_MASK; | |
751 | BUG_ON(node->data[index] == 0); | |
752 | node = (struct rtree_node *)node->data[index]; | |
753 | } | |
754 | ||
3a20cb17 JR |
755 | node_found: |
756 | /* Update last position */ | |
757 | bm->cur.zone = zone; | |
758 | bm->cur.node = node; | |
759 | bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK; | |
760 | ||
07a33823 JR |
761 | /* Set return values */ |
762 | *addr = node->data; | |
763 | *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK; | |
764 | ||
765 | return 0; | |
766 | } | |
767 | ||
74dfd666 RW |
768 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) |
769 | { | |
770 | void *addr; | |
771 | unsigned int bit; | |
a82f7119 | 772 | int error; |
74dfd666 | 773 | |
a82f7119 RW |
774 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
775 | BUG_ON(error); | |
74dfd666 RW |
776 | set_bit(bit, addr); |
777 | } | |
778 | ||
a82f7119 RW |
779 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
780 | { | |
781 | void *addr; | |
782 | unsigned int bit; | |
783 | int error; | |
784 | ||
785 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
07a33823 JR |
786 | if (!error) |
787 | set_bit(bit, addr); | |
788 | ||
a82f7119 RW |
789 | return error; |
790 | } | |
791 | ||
74dfd666 RW |
792 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
793 | { | |
794 | void *addr; | |
795 | unsigned int bit; | |
a82f7119 | 796 | int error; |
74dfd666 | 797 | |
a82f7119 RW |
798 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
799 | BUG_ON(error); | |
74dfd666 RW |
800 | clear_bit(bit, addr); |
801 | } | |
802 | ||
fdd64ed5 JR |
803 | static void memory_bm_clear_current(struct memory_bitmap *bm) |
804 | { | |
805 | int bit; | |
806 | ||
807 | bit = max(bm->cur.node_bit - 1, 0); | |
808 | clear_bit(bit, bm->cur.node->data); | |
809 | } | |
810 | ||
74dfd666 RW |
811 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) |
812 | { | |
813 | void *addr; | |
814 | unsigned int bit; | |
9047eb62 | 815 | int error; |
74dfd666 | 816 | |
a82f7119 RW |
817 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
818 | BUG_ON(error); | |
9047eb62 | 819 | return test_bit(bit, addr); |
b788db79 RW |
820 | } |
821 | ||
69643279 RW |
822 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
823 | { | |
824 | void *addr; | |
825 | unsigned int bit; | |
07a33823 | 826 | |
9047eb62 | 827 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); |
b788db79 RW |
828 | } |
829 | ||
3a20cb17 | 830 | /* |
ef96f639 | 831 | * rtree_next_node - Jump to the next leaf node. |
3a20cb17 | 832 | * |
ef96f639 RW |
833 | * Set the position to the beginning of the next node in the |
834 | * memory bitmap. This is either the next node in the current | |
835 | * zone's radix tree or the first node in the radix tree of the | |
836 | * next zone. | |
3a20cb17 | 837 | * |
ef96f639 | 838 | * Return true if there is a next node, false otherwise. |
3a20cb17 JR |
839 | */ |
840 | static bool rtree_next_node(struct memory_bitmap *bm) | |
841 | { | |
924d8696 JM |
842 | if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) { |
843 | bm->cur.node = list_entry(bm->cur.node->list.next, | |
844 | struct rtree_node, list); | |
3a20cb17 JR |
845 | bm->cur.node_pfn += BM_BITS_PER_BLOCK; |
846 | bm->cur.node_bit = 0; | |
0f7d83e8 | 847 | touch_softlockup_watchdog(); |
3a20cb17 JR |
848 | return true; |
849 | } | |
850 | ||
851 | /* No more nodes, goto next zone */ | |
924d8696 JM |
852 | if (!list_is_last(&bm->cur.zone->list, &bm->zones)) { |
853 | bm->cur.zone = list_entry(bm->cur.zone->list.next, | |
3a20cb17 | 854 | struct mem_zone_bm_rtree, list); |
3a20cb17 JR |
855 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, |
856 | struct rtree_node, list); | |
857 | bm->cur.node_pfn = 0; | |
858 | bm->cur.node_bit = 0; | |
859 | return true; | |
860 | } | |
861 | ||
862 | /* No more zones */ | |
863 | return false; | |
864 | } | |
865 | ||
9047eb62 | 866 | /** |
ef96f639 RW |
867 | * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap. |
868 | * @bm: Memory bitmap. | |
3a20cb17 | 869 | * |
ef96f639 RW |
870 | * Starting from the last returned position this function searches for the next |
871 | * set bit in @bm and returns the PFN represented by it. If no more bits are | |
872 | * set, BM_END_OF_MAP is returned. | |
9047eb62 | 873 | * |
ef96f639 RW |
874 | * It is required to run memory_bm_position_reset() before the first call to |
875 | * this function for the given memory bitmap. | |
3a20cb17 | 876 | */ |
9047eb62 | 877 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) |
3a20cb17 JR |
878 | { |
879 | unsigned long bits, pfn, pages; | |
880 | int bit; | |
881 | ||
882 | do { | |
883 | pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn; | |
884 | bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK); | |
885 | bit = find_next_bit(bm->cur.node->data, bits, | |
886 | bm->cur.node_bit); | |
887 | if (bit < bits) { | |
888 | pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit; | |
889 | bm->cur.node_bit = bit + 1; | |
890 | return pfn; | |
891 | } | |
892 | } while (rtree_next_node(bm)); | |
893 | ||
894 | return BM_END_OF_MAP; | |
895 | } | |
896 | ||
ef96f639 RW |
897 | /* |
898 | * This structure represents a range of page frames the contents of which | |
899 | * should not be saved during hibernation. | |
74dfd666 | 900 | */ |
74dfd666 RW |
901 | struct nosave_region { |
902 | struct list_head list; | |
903 | unsigned long start_pfn; | |
904 | unsigned long end_pfn; | |
905 | }; | |
906 | ||
907 | static LIST_HEAD(nosave_regions); | |
908 | ||
307c5971 RW |
909 | static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone) |
910 | { | |
911 | struct rtree_node *node; | |
912 | ||
913 | list_for_each_entry(node, &zone->nodes, list) | |
914 | recycle_safe_page(node->data); | |
915 | ||
916 | list_for_each_entry(node, &zone->leaves, list) | |
917 | recycle_safe_page(node->data); | |
918 | } | |
919 | ||
920 | static void memory_bm_recycle(struct memory_bitmap *bm) | |
921 | { | |
922 | struct mem_zone_bm_rtree *zone; | |
923 | struct linked_page *p_list; | |
924 | ||
925 | list_for_each_entry(zone, &bm->zones, list) | |
926 | recycle_zone_bm_rtree(zone); | |
927 | ||
928 | p_list = bm->p_list; | |
929 | while (p_list) { | |
930 | struct linked_page *lp = p_list; | |
931 | ||
932 | p_list = lp->next; | |
933 | recycle_safe_page(lp); | |
934 | } | |
935 | } | |
936 | ||
74dfd666 | 937 | /** |
ef96f639 RW |
938 | * register_nosave_region - Register a region of unsaveable memory. |
939 | * | |
940 | * Register a range of page frames the contents of which should not be saved | |
941 | * during hibernation (to be used in the early initialization code). | |
74dfd666 | 942 | */ |
efd5a852 RW |
943 | void __init __register_nosave_region(unsigned long start_pfn, |
944 | unsigned long end_pfn, int use_kmalloc) | |
74dfd666 RW |
945 | { |
946 | struct nosave_region *region; | |
947 | ||
948 | if (start_pfn >= end_pfn) | |
949 | return; | |
950 | ||
951 | if (!list_empty(&nosave_regions)) { | |
952 | /* Try to extend the previous region (they should be sorted) */ | |
953 | region = list_entry(nosave_regions.prev, | |
954 | struct nosave_region, list); | |
955 | if (region->end_pfn == start_pfn) { | |
956 | region->end_pfn = end_pfn; | |
957 | goto Report; | |
958 | } | |
959 | } | |
940d67f6 | 960 | if (use_kmalloc) { |
ef96f639 | 961 | /* During init, this shouldn't fail */ |
940d67f6 JB |
962 | region = kmalloc(sizeof(struct nosave_region), GFP_KERNEL); |
963 | BUG_ON(!region); | |
d5f32af3 | 964 | } else { |
940d67f6 | 965 | /* This allocation cannot fail */ |
c2f69cda | 966 | region = memblock_virt_alloc(sizeof(struct nosave_region), 0); |
d5f32af3 | 967 | } |
74dfd666 RW |
968 | region->start_pfn = start_pfn; |
969 | region->end_pfn = end_pfn; | |
970 | list_add_tail(®ion->list, &nosave_regions); | |
971 | Report: | |
cd38ca85 BH |
972 | printk(KERN_INFO "PM: Registered nosave memory: [mem %#010llx-%#010llx]\n", |
973 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
974 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
74dfd666 RW |
975 | } |
976 | ||
977 | /* | |
978 | * Set bits in this map correspond to the page frames the contents of which | |
979 | * should not be saved during the suspend. | |
980 | */ | |
981 | static struct memory_bitmap *forbidden_pages_map; | |
982 | ||
983 | /* Set bits in this map correspond to free page frames. */ | |
984 | static struct memory_bitmap *free_pages_map; | |
985 | ||
986 | /* | |
987 | * Each page frame allocated for creating the image is marked by setting the | |
988 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
989 | */ | |
990 | ||
991 | void swsusp_set_page_free(struct page *page) | |
992 | { | |
993 | if (free_pages_map) | |
994 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
995 | } | |
996 | ||
997 | static int swsusp_page_is_free(struct page *page) | |
998 | { | |
999 | return free_pages_map ? | |
1000 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
1001 | } | |
1002 | ||
1003 | void swsusp_unset_page_free(struct page *page) | |
1004 | { | |
1005 | if (free_pages_map) | |
1006 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
1007 | } | |
1008 | ||
1009 | static void swsusp_set_page_forbidden(struct page *page) | |
1010 | { | |
1011 | if (forbidden_pages_map) | |
1012 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
1013 | } | |
1014 | ||
1015 | int swsusp_page_is_forbidden(struct page *page) | |
1016 | { | |
1017 | return forbidden_pages_map ? | |
1018 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
1019 | } | |
1020 | ||
1021 | static void swsusp_unset_page_forbidden(struct page *page) | |
1022 | { | |
1023 | if (forbidden_pages_map) | |
1024 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
1025 | } | |
1026 | ||
1027 | /** | |
ef96f639 RW |
1028 | * mark_nosave_pages - Mark pages that should not be saved. |
1029 | * @bm: Memory bitmap. | |
1030 | * | |
1031 | * Set the bits in @bm that correspond to the page frames the contents of which | |
1032 | * should not be saved. | |
74dfd666 | 1033 | */ |
74dfd666 RW |
1034 | static void mark_nosave_pages(struct memory_bitmap *bm) |
1035 | { | |
1036 | struct nosave_region *region; | |
1037 | ||
1038 | if (list_empty(&nosave_regions)) | |
1039 | return; | |
1040 | ||
1041 | list_for_each_entry(region, &nosave_regions, list) { | |
1042 | unsigned long pfn; | |
1043 | ||
69f1d475 BH |
1044 | pr_debug("PM: Marking nosave pages: [mem %#010llx-%#010llx]\n", |
1045 | (unsigned long long) region->start_pfn << PAGE_SHIFT, | |
1046 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
1047 | - 1); | |
74dfd666 RW |
1048 | |
1049 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
1050 | if (pfn_valid(pfn)) { |
1051 | /* | |
1052 | * It is safe to ignore the result of | |
1053 | * mem_bm_set_bit_check() here, since we won't | |
1054 | * touch the PFNs for which the error is | |
1055 | * returned anyway. | |
1056 | */ | |
1057 | mem_bm_set_bit_check(bm, pfn); | |
1058 | } | |
74dfd666 RW |
1059 | } |
1060 | } | |
1061 | ||
1062 | /** | |
ef96f639 RW |
1063 | * create_basic_memory_bitmaps - Create bitmaps to hold basic page information. |
1064 | * | |
1065 | * Create bitmaps needed for marking page frames that should not be saved and | |
1066 | * free page frames. The forbidden_pages_map and free_pages_map pointers are | |
1067 | * only modified if everything goes well, because we don't want the bits to be | |
1068 | * touched before both bitmaps are set up. | |
74dfd666 | 1069 | */ |
74dfd666 RW |
1070 | int create_basic_memory_bitmaps(void) |
1071 | { | |
1072 | struct memory_bitmap *bm1, *bm2; | |
1073 | int error = 0; | |
1074 | ||
aab17289 RW |
1075 | if (forbidden_pages_map && free_pages_map) |
1076 | return 0; | |
1077 | else | |
1078 | BUG_ON(forbidden_pages_map || free_pages_map); | |
74dfd666 | 1079 | |
0709db60 | 1080 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1081 | if (!bm1) |
1082 | return -ENOMEM; | |
1083 | ||
0709db60 | 1084 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1085 | if (error) |
1086 | goto Free_first_object; | |
1087 | ||
0709db60 | 1088 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1089 | if (!bm2) |
1090 | goto Free_first_bitmap; | |
1091 | ||
0709db60 | 1092 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1093 | if (error) |
1094 | goto Free_second_object; | |
1095 | ||
1096 | forbidden_pages_map = bm1; | |
1097 | free_pages_map = bm2; | |
1098 | mark_nosave_pages(forbidden_pages_map); | |
1099 | ||
23976728 | 1100 | pr_debug("PM: Basic memory bitmaps created\n"); |
74dfd666 RW |
1101 | |
1102 | return 0; | |
1103 | ||
1104 | Free_second_object: | |
1105 | kfree(bm2); | |
1106 | Free_first_bitmap: | |
1107 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1108 | Free_first_object: | |
1109 | kfree(bm1); | |
1110 | return -ENOMEM; | |
1111 | } | |
1112 | ||
1113 | /** | |
ef96f639 RW |
1114 | * free_basic_memory_bitmaps - Free memory bitmaps holding basic information. |
1115 | * | |
1116 | * Free memory bitmaps allocated by create_basic_memory_bitmaps(). The | |
1117 | * auxiliary pointers are necessary so that the bitmaps themselves are not | |
1118 | * referred to while they are being freed. | |
74dfd666 | 1119 | */ |
74dfd666 RW |
1120 | void free_basic_memory_bitmaps(void) |
1121 | { | |
1122 | struct memory_bitmap *bm1, *bm2; | |
1123 | ||
6a0c7cd3 RW |
1124 | if (WARN_ON(!(forbidden_pages_map && free_pages_map))) |
1125 | return; | |
74dfd666 RW |
1126 | |
1127 | bm1 = forbidden_pages_map; | |
1128 | bm2 = free_pages_map; | |
1129 | forbidden_pages_map = NULL; | |
1130 | free_pages_map = NULL; | |
1131 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1132 | kfree(bm1); | |
1133 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
1134 | kfree(bm2); | |
1135 | ||
23976728 | 1136 | pr_debug("PM: Basic memory bitmaps freed\n"); |
74dfd666 RW |
1137 | } |
1138 | ||
1ad1410f AA |
1139 | void clear_free_pages(void) |
1140 | { | |
1141 | #ifdef CONFIG_PAGE_POISONING_ZERO | |
1142 | struct memory_bitmap *bm = free_pages_map; | |
1143 | unsigned long pfn; | |
1144 | ||
1145 | if (WARN_ON(!(free_pages_map))) | |
1146 | return; | |
1147 | ||
1148 | memory_bm_position_reset(bm); | |
1149 | pfn = memory_bm_next_pfn(bm); | |
1150 | while (pfn != BM_END_OF_MAP) { | |
1151 | if (pfn_valid(pfn)) | |
1152 | clear_highpage(pfn_to_page(pfn)); | |
1153 | ||
1154 | pfn = memory_bm_next_pfn(bm); | |
1155 | } | |
1156 | memory_bm_position_reset(bm); | |
1157 | pr_info("PM: free pages cleared after restore\n"); | |
1158 | #endif /* PAGE_POISONING_ZERO */ | |
1159 | } | |
1160 | ||
b788db79 | 1161 | /** |
ef96f639 RW |
1162 | * snapshot_additional_pages - Estimate the number of extra pages needed. |
1163 | * @zone: Memory zone to carry out the computation for. | |
1164 | * | |
1165 | * Estimate the number of additional pages needed for setting up a hibernation | |
1166 | * image data structures for @zone (usually, the returned value is greater than | |
1167 | * the exact number). | |
b788db79 | 1168 | */ |
b788db79 RW |
1169 | unsigned int snapshot_additional_pages(struct zone *zone) |
1170 | { | |
f469f02d | 1171 | unsigned int rtree, nodes; |
b788db79 | 1172 | |
f469f02d JR |
1173 | rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); |
1174 | rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node), | |
1175 | LINKED_PAGE_DATA_SIZE); | |
1176 | while (nodes > 1) { | |
1177 | nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL); | |
1178 | rtree += nodes; | |
1179 | } | |
1180 | ||
9047eb62 | 1181 | return 2 * rtree; |
b788db79 RW |
1182 | } |
1183 | ||
8357376d RW |
1184 | #ifdef CONFIG_HIGHMEM |
1185 | /** | |
ef96f639 RW |
1186 | * count_free_highmem_pages - Compute the total number of free highmem pages. |
1187 | * | |
1188 | * The returned number is system-wide. | |
8357376d | 1189 | */ |
8357376d RW |
1190 | static unsigned int count_free_highmem_pages(void) |
1191 | { | |
1192 | struct zone *zone; | |
1193 | unsigned int cnt = 0; | |
1194 | ||
ee99c71c KM |
1195 | for_each_populated_zone(zone) |
1196 | if (is_highmem(zone)) | |
d23ad423 | 1197 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
1198 | |
1199 | return cnt; | |
1200 | } | |
1201 | ||
1202 | /** | |
ef96f639 RW |
1203 | * saveable_highmem_page - Check if a highmem page is saveable. |
1204 | * | |
1205 | * Determine whether a highmem page should be included in a hibernation image. | |
8357376d | 1206 | * |
ef96f639 RW |
1207 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, |
1208 | * and it isn't part of a free chunk of pages. | |
8357376d | 1209 | */ |
846705de | 1210 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
1211 | { |
1212 | struct page *page; | |
1213 | ||
1214 | if (!pfn_valid(pfn)) | |
1215 | return NULL; | |
1216 | ||
1217 | page = pfn_to_page(pfn); | |
846705de RW |
1218 | if (page_zone(page) != zone) |
1219 | return NULL; | |
8357376d RW |
1220 | |
1221 | BUG_ON(!PageHighMem(page)); | |
1222 | ||
7be98234 RW |
1223 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page) || |
1224 | PageReserved(page)) | |
8357376d RW |
1225 | return NULL; |
1226 | ||
c6968e73 SG |
1227 | if (page_is_guard(page)) |
1228 | return NULL; | |
1229 | ||
8357376d RW |
1230 | return page; |
1231 | } | |
1232 | ||
1233 | /** | |
ef96f639 | 1234 | * count_highmem_pages - Compute the total number of saveable highmem pages. |
8357376d | 1235 | */ |
fe419535 | 1236 | static unsigned int count_highmem_pages(void) |
8357376d RW |
1237 | { |
1238 | struct zone *zone; | |
1239 | unsigned int n = 0; | |
1240 | ||
98e73dc5 | 1241 | for_each_populated_zone(zone) { |
8357376d RW |
1242 | unsigned long pfn, max_zone_pfn; |
1243 | ||
1244 | if (!is_highmem(zone)) | |
1245 | continue; | |
1246 | ||
1247 | mark_free_pages(zone); | |
c33bc315 | 1248 | max_zone_pfn = zone_end_pfn(zone); |
8357376d | 1249 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1250 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
1251 | n++; |
1252 | } | |
1253 | return n; | |
1254 | } | |
1255 | #else | |
846705de RW |
1256 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
1257 | { | |
1258 | return NULL; | |
1259 | } | |
8357376d RW |
1260 | #endif /* CONFIG_HIGHMEM */ |
1261 | ||
25761b6e | 1262 | /** |
ef96f639 RW |
1263 | * saveable_page - Check if the given page is saveable. |
1264 | * | |
1265 | * Determine whether a non-highmem page should be included in a hibernation | |
1266 | * image. | |
25761b6e | 1267 | * |
ef96f639 RW |
1268 | * We should save the page if it isn't Nosave, and is not in the range |
1269 | * of pages statically defined as 'unsaveable', and it isn't part of | |
1270 | * a free chunk of pages. | |
25761b6e | 1271 | */ |
846705de | 1272 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 1273 | { |
de491861 | 1274 | struct page *page; |
25761b6e RW |
1275 | |
1276 | if (!pfn_valid(pfn)) | |
ae83c5ee | 1277 | return NULL; |
25761b6e RW |
1278 | |
1279 | page = pfn_to_page(pfn); | |
846705de RW |
1280 | if (page_zone(page) != zone) |
1281 | return NULL; | |
ae83c5ee | 1282 | |
8357376d RW |
1283 | BUG_ON(PageHighMem(page)); |
1284 | ||
7be98234 | 1285 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 1286 | return NULL; |
8357376d | 1287 | |
8a235efa RW |
1288 | if (PageReserved(page) |
1289 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 1290 | return NULL; |
25761b6e | 1291 | |
c6968e73 SG |
1292 | if (page_is_guard(page)) |
1293 | return NULL; | |
1294 | ||
ae83c5ee | 1295 | return page; |
25761b6e RW |
1296 | } |
1297 | ||
8357376d | 1298 | /** |
ef96f639 | 1299 | * count_data_pages - Compute the total number of saveable non-highmem pages. |
8357376d | 1300 | */ |
fe419535 | 1301 | static unsigned int count_data_pages(void) |
25761b6e RW |
1302 | { |
1303 | struct zone *zone; | |
ae83c5ee | 1304 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 1305 | unsigned int n = 0; |
25761b6e | 1306 | |
98e73dc5 | 1307 | for_each_populated_zone(zone) { |
25761b6e RW |
1308 | if (is_highmem(zone)) |
1309 | continue; | |
8357376d | 1310 | |
25761b6e | 1311 | mark_free_pages(zone); |
c33bc315 | 1312 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee | 1313 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1314 | if (saveable_page(zone, pfn)) |
8357376d | 1315 | n++; |
25761b6e | 1316 | } |
a0f49651 | 1317 | return n; |
25761b6e RW |
1318 | } |
1319 | ||
ef96f639 RW |
1320 | /* |
1321 | * This is needed, because copy_page and memcpy are not usable for copying | |
8357376d RW |
1322 | * task structs. |
1323 | */ | |
1324 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
1325 | { |
1326 | int n; | |
1327 | ||
f623f0db RW |
1328 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
1329 | *dst++ = *src++; | |
1330 | } | |
1331 | ||
8a235efa | 1332 | /** |
ef96f639 RW |
1333 | * safe_copy_page - Copy a page in a safe way. |
1334 | * | |
1335 | * Check if the page we are going to copy is marked as present in the kernel | |
1336 | * page tables (this always is the case if CONFIG_DEBUG_PAGEALLOC is not set | |
1337 | * and in that case kernel_page_present() always returns 'true'). | |
8a235efa RW |
1338 | */ |
1339 | static void safe_copy_page(void *dst, struct page *s_page) | |
1340 | { | |
1341 | if (kernel_page_present(s_page)) { | |
1342 | do_copy_page(dst, page_address(s_page)); | |
1343 | } else { | |
1344 | kernel_map_pages(s_page, 1, 1); | |
1345 | do_copy_page(dst, page_address(s_page)); | |
1346 | kernel_map_pages(s_page, 1, 0); | |
1347 | } | |
1348 | } | |
1349 | ||
8357376d | 1350 | #ifdef CONFIG_HIGHMEM |
efd5a852 | 1351 | static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn) |
8357376d RW |
1352 | { |
1353 | return is_highmem(zone) ? | |
846705de | 1354 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
1355 | } |
1356 | ||
8a235efa | 1357 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
1358 | { |
1359 | struct page *s_page, *d_page; | |
1360 | void *src, *dst; | |
1361 | ||
1362 | s_page = pfn_to_page(src_pfn); | |
1363 | d_page = pfn_to_page(dst_pfn); | |
1364 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1365 | src = kmap_atomic(s_page); |
1366 | dst = kmap_atomic(d_page); | |
8357376d | 1367 | do_copy_page(dst, src); |
0de9a1e2 CW |
1368 | kunmap_atomic(dst); |
1369 | kunmap_atomic(src); | |
8357376d | 1370 | } else { |
8357376d | 1371 | if (PageHighMem(d_page)) { |
ef96f639 RW |
1372 | /* |
1373 | * The page pointed to by src may contain some kernel | |
8357376d RW |
1374 | * data modified by kmap_atomic() |
1375 | */ | |
8a235efa | 1376 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1377 | dst = kmap_atomic(d_page); |
3ecb01df | 1378 | copy_page(dst, buffer); |
0de9a1e2 | 1379 | kunmap_atomic(dst); |
8357376d | 1380 | } else { |
8a235efa | 1381 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1382 | } |
1383 | } | |
1384 | } | |
1385 | #else | |
846705de | 1386 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1387 | |
8a235efa | 1388 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1389 | { |
8a235efa RW |
1390 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1391 | pfn_to_page(src_pfn)); | |
8357376d RW |
1392 | } |
1393 | #endif /* CONFIG_HIGHMEM */ | |
1394 | ||
efd5a852 RW |
1395 | static void copy_data_pages(struct memory_bitmap *copy_bm, |
1396 | struct memory_bitmap *orig_bm) | |
25761b6e RW |
1397 | { |
1398 | struct zone *zone; | |
b788db79 | 1399 | unsigned long pfn; |
25761b6e | 1400 | |
98e73dc5 | 1401 | for_each_populated_zone(zone) { |
b788db79 RW |
1402 | unsigned long max_zone_pfn; |
1403 | ||
25761b6e | 1404 | mark_free_pages(zone); |
c33bc315 | 1405 | max_zone_pfn = zone_end_pfn(zone); |
b788db79 | 1406 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1407 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1408 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1409 | } |
b788db79 RW |
1410 | memory_bm_position_reset(orig_bm); |
1411 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1412 | for(;;) { |
b788db79 | 1413 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1414 | if (unlikely(pfn == BM_END_OF_MAP)) |
1415 | break; | |
1416 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1417 | } | |
25761b6e RW |
1418 | } |
1419 | ||
8357376d RW |
1420 | /* Total number of image pages */ |
1421 | static unsigned int nr_copy_pages; | |
1422 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1423 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1424 | /* |
1425 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1426 | * before suspending devices. | |
1427 | */ | |
1428 | unsigned int alloc_normal, alloc_highmem; | |
1429 | /* | |
1430 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1431 | * hibernation image pages (during restore) | |
1432 | */ | |
1433 | static struct memory_bitmap orig_bm; | |
1434 | /* | |
1435 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1436 | * will contain copies of saveable pages. During restore it is initially used | |
1437 | * for marking hibernation image pages, but then the set bits from it are | |
1438 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1439 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1440 | * this purpose. | |
1441 | */ | |
1442 | static struct memory_bitmap copy_bm; | |
8357376d | 1443 | |
25761b6e | 1444 | /** |
ef96f639 | 1445 | * swsusp_free - Free pages allocated for hibernation image. |
cd560bb2 | 1446 | * |
ef96f639 RW |
1447 | * Image pages are alocated before snapshot creation, so they need to be |
1448 | * released after resume. | |
25761b6e | 1449 | */ |
25761b6e RW |
1450 | void swsusp_free(void) |
1451 | { | |
fdd64ed5 | 1452 | unsigned long fb_pfn, fr_pfn; |
6efde38f | 1453 | |
fdd64ed5 JR |
1454 | if (!forbidden_pages_map || !free_pages_map) |
1455 | goto out; | |
1456 | ||
1457 | memory_bm_position_reset(forbidden_pages_map); | |
1458 | memory_bm_position_reset(free_pages_map); | |
1459 | ||
1460 | loop: | |
1461 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1462 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1463 | ||
1464 | /* | |
1465 | * Find the next bit set in both bitmaps. This is guaranteed to | |
1466 | * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP. | |
1467 | */ | |
1468 | do { | |
1469 | if (fb_pfn < fr_pfn) | |
1470 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1471 | if (fr_pfn < fb_pfn) | |
1472 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1473 | } while (fb_pfn != fr_pfn); | |
1474 | ||
1475 | if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) { | |
1476 | struct page *page = pfn_to_page(fr_pfn); | |
1477 | ||
1478 | memory_bm_clear_current(forbidden_pages_map); | |
1479 | memory_bm_clear_current(free_pages_map); | |
4c0b6c10 | 1480 | hibernate_restore_unprotect_page(page_address(page)); |
fdd64ed5 JR |
1481 | __free_page(page); |
1482 | goto loop; | |
25761b6e | 1483 | } |
fdd64ed5 JR |
1484 | |
1485 | out: | |
f577eb30 RW |
1486 | nr_copy_pages = 0; |
1487 | nr_meta_pages = 0; | |
75534b50 | 1488 | restore_pblist = NULL; |
6e1819d6 | 1489 | buffer = NULL; |
64a473cb RW |
1490 | alloc_normal = 0; |
1491 | alloc_highmem = 0; | |
4c0b6c10 | 1492 | hibernate_restore_protection_end(); |
25761b6e RW |
1493 | } |
1494 | ||
4bb33435 RW |
1495 | /* Helper functions used for the shrinking of memory. */ |
1496 | ||
1497 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1498 | ||
fe419535 | 1499 | /** |
ef96f639 | 1500 | * preallocate_image_pages - Allocate a number of pages for hibernation image. |
4bb33435 RW |
1501 | * @nr_pages: Number of page frames to allocate. |
1502 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1503 | * |
4bb33435 RW |
1504 | * Return value: Number of page frames actually allocated |
1505 | */ | |
1506 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1507 | { | |
1508 | unsigned long nr_alloc = 0; | |
1509 | ||
1510 | while (nr_pages > 0) { | |
64a473cb RW |
1511 | struct page *page; |
1512 | ||
1513 | page = alloc_image_page(mask); | |
1514 | if (!page) | |
4bb33435 | 1515 | break; |
64a473cb RW |
1516 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1517 | if (PageHighMem(page)) | |
1518 | alloc_highmem++; | |
1519 | else | |
1520 | alloc_normal++; | |
4bb33435 RW |
1521 | nr_pages--; |
1522 | nr_alloc++; | |
1523 | } | |
1524 | ||
1525 | return nr_alloc; | |
1526 | } | |
1527 | ||
6715045d RW |
1528 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1529 | unsigned long avail_normal) | |
4bb33435 | 1530 | { |
6715045d RW |
1531 | unsigned long alloc; |
1532 | ||
1533 | if (avail_normal <= alloc_normal) | |
1534 | return 0; | |
1535 | ||
1536 | alloc = avail_normal - alloc_normal; | |
1537 | if (nr_pages < alloc) | |
1538 | alloc = nr_pages; | |
1539 | ||
1540 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1541 | } |
1542 | ||
1543 | #ifdef CONFIG_HIGHMEM | |
1544 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1545 | { | |
1546 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1547 | } | |
1548 | ||
1549 | /** | |
ef96f639 | 1550 | * __fraction - Compute (an approximation of) x * (multiplier / base). |
fe419535 | 1551 | */ |
4bb33435 RW |
1552 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1553 | { | |
1554 | x *= multiplier; | |
1555 | do_div(x, base); | |
1556 | return (unsigned long)x; | |
1557 | } | |
fe419535 | 1558 | |
4bb33435 | 1559 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
efd5a852 RW |
1560 | unsigned long highmem, |
1561 | unsigned long total) | |
fe419535 | 1562 | { |
4bb33435 RW |
1563 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1564 | ||
1565 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1566 | } |
4bb33435 RW |
1567 | #else /* CONFIG_HIGHMEM */ |
1568 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1569 | { | |
1570 | return 0; | |
1571 | } | |
1572 | ||
1573 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
efd5a852 RW |
1574 | unsigned long highmem, |
1575 | unsigned long total) | |
4bb33435 RW |
1576 | { |
1577 | return 0; | |
1578 | } | |
1579 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1580 | |
4bb33435 | 1581 | /** |
ef96f639 | 1582 | * free_unnecessary_pages - Release preallocated pages not needed for the image. |
64a473cb | 1583 | */ |
a64fc82c | 1584 | static unsigned long free_unnecessary_pages(void) |
64a473cb | 1585 | { |
a64fc82c | 1586 | unsigned long save, to_free_normal, to_free_highmem, free; |
64a473cb | 1587 | |
6715045d RW |
1588 | save = count_data_pages(); |
1589 | if (alloc_normal >= save) { | |
1590 | to_free_normal = alloc_normal - save; | |
1591 | save = 0; | |
1592 | } else { | |
1593 | to_free_normal = 0; | |
1594 | save -= alloc_normal; | |
1595 | } | |
1596 | save += count_highmem_pages(); | |
1597 | if (alloc_highmem >= save) { | |
1598 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1599 | } else { |
1600 | to_free_highmem = 0; | |
4d4cf23c RW |
1601 | save -= alloc_highmem; |
1602 | if (to_free_normal > save) | |
1603 | to_free_normal -= save; | |
1604 | else | |
1605 | to_free_normal = 0; | |
64a473cb | 1606 | } |
a64fc82c | 1607 | free = to_free_normal + to_free_highmem; |
64a473cb RW |
1608 | |
1609 | memory_bm_position_reset(©_bm); | |
1610 | ||
a9c9b442 | 1611 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1612 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1613 | struct page *page = pfn_to_page(pfn); | |
1614 | ||
1615 | if (PageHighMem(page)) { | |
1616 | if (!to_free_highmem) | |
1617 | continue; | |
1618 | to_free_highmem--; | |
1619 | alloc_highmem--; | |
1620 | } else { | |
1621 | if (!to_free_normal) | |
1622 | continue; | |
1623 | to_free_normal--; | |
1624 | alloc_normal--; | |
1625 | } | |
1626 | memory_bm_clear_bit(©_bm, pfn); | |
1627 | swsusp_unset_page_forbidden(page); | |
1628 | swsusp_unset_page_free(page); | |
1629 | __free_page(page); | |
1630 | } | |
a64fc82c WK |
1631 | |
1632 | return free; | |
64a473cb RW |
1633 | } |
1634 | ||
ef4aede3 | 1635 | /** |
ef96f639 | 1636 | * minimum_image_size - Estimate the minimum acceptable size of an image. |
ef4aede3 RW |
1637 | * @saveable: Number of saveable pages in the system. |
1638 | * | |
1639 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1640 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1641 | * preallocating memory. | |
1642 | * | |
1643 | * We assume that the minimum image size should be proportional to | |
1644 | * | |
1645 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1646 | * | |
1647 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
4d434820 | 1648 | * and (3) inactive anonymous pages, (4) active and (5) inactive file pages, |
ef4aede3 RW |
1649 | * minus mapped file pages. |
1650 | */ | |
1651 | static unsigned long minimum_image_size(unsigned long saveable) | |
1652 | { | |
1653 | unsigned long size; | |
1654 | ||
1655 | size = global_page_state(NR_SLAB_RECLAIMABLE) | |
599d0c95 MG |
1656 | + global_node_page_state(NR_ACTIVE_ANON) |
1657 | + global_node_page_state(NR_INACTIVE_ANON) | |
1658 | + global_node_page_state(NR_ACTIVE_FILE) | |
1659 | + global_node_page_state(NR_INACTIVE_FILE) | |
1660 | - global_node_page_state(NR_FILE_MAPPED); | |
ef4aede3 RW |
1661 | |
1662 | return saveable <= size ? 0 : saveable - size; | |
1663 | } | |
1664 | ||
64a473cb | 1665 | /** |
ef96f639 | 1666 | * hibernate_preallocate_memory - Preallocate memory for hibernation image. |
4bb33435 RW |
1667 | * |
1668 | * To create a hibernation image it is necessary to make a copy of every page | |
1669 | * frame in use. We also need a number of page frames to be free during | |
1670 | * hibernation for allocations made while saving the image and for device | |
1671 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 RW |
1672 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
1673 | * estimate) and reserverd_size divided by PAGE_SIZE (which is tunable through | |
1674 | * /sys/power/reserved_size, respectively). To make this happen, we compute the | |
1675 | * total number of available page frames and allocate at least | |
4bb33435 | 1676 | * |
ddeb6487 RW |
1677 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1678 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1679 | * |
1680 | * of them, which corresponds to the maximum size of a hibernation image. | |
1681 | * | |
1682 | * If image_size is set below the number following from the above formula, | |
1683 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1684 | * pages in the system is below the requested image size or the minimum |
1685 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1686 | */ |
64a473cb | 1687 | int hibernate_preallocate_memory(void) |
fe419535 | 1688 | { |
fe419535 | 1689 | struct zone *zone; |
4bb33435 | 1690 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1691 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
db597605 | 1692 | ktime_t start, stop; |
64a473cb | 1693 | int error; |
fe419535 | 1694 | |
64a473cb | 1695 | printk(KERN_INFO "PM: Preallocating image memory... "); |
db597605 | 1696 | start = ktime_get(); |
fe419535 | 1697 | |
64a473cb RW |
1698 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
1699 | if (error) | |
1700 | goto err_out; | |
1701 | ||
1702 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
1703 | if (error) | |
1704 | goto err_out; | |
1705 | ||
1706 | alloc_normal = 0; | |
1707 | alloc_highmem = 0; | |
1708 | ||
4bb33435 | 1709 | /* Count the number of saveable data pages. */ |
64a473cb | 1710 | save_highmem = count_highmem_pages(); |
4bb33435 | 1711 | saveable = count_data_pages(); |
fe419535 | 1712 | |
4bb33435 RW |
1713 | /* |
1714 | * Compute the total number of page frames we can use (count) and the | |
1715 | * number of pages needed for image metadata (size). | |
1716 | */ | |
1717 | count = saveable; | |
64a473cb RW |
1718 | saveable += save_highmem; |
1719 | highmem = save_highmem; | |
4bb33435 RW |
1720 | size = 0; |
1721 | for_each_populated_zone(zone) { | |
1722 | size += snapshot_additional_pages(zone); | |
1723 | if (is_highmem(zone)) | |
1724 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1725 | else | |
1726 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1727 | } | |
6715045d | 1728 | avail_normal = count; |
4bb33435 RW |
1729 | count += highmem; |
1730 | count -= totalreserve_pages; | |
1731 | ||
85055dd8 MS |
1732 | /* Add number of pages required for page keys (s390 only). */ |
1733 | size += page_key_additional_pages(saveable); | |
1734 | ||
4bb33435 | 1735 | /* Compute the maximum number of saveable pages to leave in memory. */ |
ddeb6487 RW |
1736 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1737 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1738 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1739 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1740 | if (size > max_size) | |
1741 | size = max_size; | |
1742 | /* | |
266f1a25 RW |
1743 | * If the desired number of image pages is at least as large as the |
1744 | * current number of saveable pages in memory, allocate page frames for | |
1745 | * the image and we're done. | |
4bb33435 | 1746 | */ |
64a473cb RW |
1747 | if (size >= saveable) { |
1748 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1749 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1750 | goto out; |
64a473cb | 1751 | } |
4bb33435 | 1752 | |
ef4aede3 RW |
1753 | /* Estimate the minimum size of the image. */ |
1754 | pages = minimum_image_size(saveable); | |
6715045d RW |
1755 | /* |
1756 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1757 | * accommodate an image of the minimum size (unless it's already too | |
1758 | * small, in which case don't preallocate pages from it at all). | |
1759 | */ | |
1760 | if (avail_normal > pages) | |
1761 | avail_normal -= pages; | |
1762 | else | |
1763 | avail_normal = 0; | |
ef4aede3 RW |
1764 | if (size < pages) |
1765 | size = min_t(unsigned long, pages, max_size); | |
1766 | ||
4bb33435 RW |
1767 | /* |
1768 | * Let the memory management subsystem know that we're going to need a | |
1769 | * large number of page frames to allocate and make it free some memory. | |
1770 | * NOTE: If this is not done, performance will be hurt badly in some | |
1771 | * test cases. | |
1772 | */ | |
1773 | shrink_all_memory(saveable - size); | |
1774 | ||
1775 | /* | |
1776 | * The number of saveable pages in memory was too high, so apply some | |
1777 | * pressure to decrease it. First, make room for the largest possible | |
1778 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1779 | * of the image as much as indicated by 'size' using allocations from |
1780 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1781 | */ |
1782 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
fd432b9f AL |
1783 | alloc = count - max_size; |
1784 | if (alloc > pages_highmem) | |
1785 | alloc -= pages_highmem; | |
1786 | else | |
1787 | alloc = 0; | |
6715045d RW |
1788 | pages = preallocate_image_memory(alloc, avail_normal); |
1789 | if (pages < alloc) { | |
1790 | /* We have exhausted non-highmem pages, try highmem. */ | |
1791 | alloc -= pages; | |
1792 | pages += pages_highmem; | |
1793 | pages_highmem = preallocate_image_highmem(alloc); | |
1794 | if (pages_highmem < alloc) | |
1795 | goto err_out; | |
1796 | pages += pages_highmem; | |
1797 | /* | |
1798 | * size is the desired number of saveable pages to leave in | |
1799 | * memory, so try to preallocate (all memory - size) pages. | |
1800 | */ | |
1801 | alloc = (count - pages) - size; | |
1802 | pages += preallocate_image_highmem(alloc); | |
1803 | } else { | |
1804 | /* | |
1805 | * There are approximately max_size saveable pages at this point | |
1806 | * and we want to reduce this number down to size. | |
1807 | */ | |
1808 | alloc = max_size - size; | |
1809 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1810 | pages_highmem += size; | |
1811 | alloc -= size; | |
1812 | size = preallocate_image_memory(alloc, avail_normal); | |
1813 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1814 | pages += pages_highmem + size; | |
1815 | } | |
4bb33435 | 1816 | |
64a473cb RW |
1817 | /* |
1818 | * We only need as many page frames for the image as there are saveable | |
1819 | * pages in memory, but we have allocated more. Release the excessive | |
1820 | * ones now. | |
1821 | */ | |
a64fc82c | 1822 | pages -= free_unnecessary_pages(); |
4bb33435 RW |
1823 | |
1824 | out: | |
db597605 | 1825 | stop = ktime_get(); |
64a473cb | 1826 | printk(KERN_CONT "done (allocated %lu pages)\n", pages); |
db597605 | 1827 | swsusp_show_speed(start, stop, pages, "Allocated"); |
fe419535 RW |
1828 | |
1829 | return 0; | |
64a473cb RW |
1830 | |
1831 | err_out: | |
1832 | printk(KERN_CONT "\n"); | |
1833 | swsusp_free(); | |
1834 | return -ENOMEM; | |
fe419535 RW |
1835 | } |
1836 | ||
8357376d RW |
1837 | #ifdef CONFIG_HIGHMEM |
1838 | /** | |
ef96f639 RW |
1839 | * count_pages_for_highmem - Count non-highmem pages needed for copying highmem. |
1840 | * | |
1841 | * Compute the number of non-highmem pages that will be necessary for creating | |
1842 | * copies of highmem pages. | |
1843 | */ | |
8357376d RW |
1844 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) |
1845 | { | |
64a473cb | 1846 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1847 | |
1848 | if (free_highmem >= nr_highmem) | |
1849 | nr_highmem = 0; | |
1850 | else | |
1851 | nr_highmem -= free_highmem; | |
1852 | ||
1853 | return nr_highmem; | |
1854 | } | |
1855 | #else | |
efd5a852 | 1856 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; } |
8357376d | 1857 | #endif /* CONFIG_HIGHMEM */ |
25761b6e RW |
1858 | |
1859 | /** | |
ef96f639 | 1860 | * enough_free_mem - Check if there is enough free memory for the image. |
25761b6e | 1861 | */ |
8357376d | 1862 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1863 | { |
e5e2fa78 | 1864 | struct zone *zone; |
64a473cb | 1865 | unsigned int free = alloc_normal; |
e5e2fa78 | 1866 | |
98e73dc5 | 1867 | for_each_populated_zone(zone) |
8357376d | 1868 | if (!is_highmem(zone)) |
d23ad423 | 1869 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1870 | |
8357376d | 1871 | nr_pages += count_pages_for_highmem(nr_highmem); |
64a473cb RW |
1872 | pr_debug("PM: Normal pages needed: %u + %u, available pages: %u\n", |
1873 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1874 | |
64a473cb | 1875 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1876 | } |
1877 | ||
8357376d RW |
1878 | #ifdef CONFIG_HIGHMEM |
1879 | /** | |
ef96f639 RW |
1880 | * get_highmem_buffer - Allocate a buffer for highmem pages. |
1881 | * | |
1882 | * If there are some highmem pages in the hibernation image, we may need a | |
1883 | * buffer to copy them and/or load their data. | |
8357376d | 1884 | */ |
8357376d RW |
1885 | static inline int get_highmem_buffer(int safe_needed) |
1886 | { | |
1887 | buffer = get_image_page(GFP_ATOMIC | __GFP_COLD, safe_needed); | |
1888 | return buffer ? 0 : -ENOMEM; | |
1889 | } | |
1890 | ||
1891 | /** | |
ef96f639 RW |
1892 | * alloc_highmem_image_pages - Allocate some highmem pages for the image. |
1893 | * | |
1894 | * Try to allocate as many pages as needed, but if the number of free highmem | |
1895 | * pages is less than that, allocate them all. | |
8357376d | 1896 | */ |
efd5a852 RW |
1897 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1898 | unsigned int nr_highmem) | |
8357376d RW |
1899 | { |
1900 | unsigned int to_alloc = count_free_highmem_pages(); | |
1901 | ||
1902 | if (to_alloc > nr_highmem) | |
1903 | to_alloc = nr_highmem; | |
1904 | ||
1905 | nr_highmem -= to_alloc; | |
1906 | while (to_alloc-- > 0) { | |
1907 | struct page *page; | |
1908 | ||
d0164adc | 1909 | page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM); |
8357376d RW |
1910 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1911 | } | |
1912 | return nr_highmem; | |
1913 | } | |
1914 | #else | |
1915 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1916 | ||
efd5a852 RW |
1917 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1918 | unsigned int n) { return 0; } | |
8357376d RW |
1919 | #endif /* CONFIG_HIGHMEM */ |
1920 | ||
1921 | /** | |
ef96f639 | 1922 | * swsusp_alloc - Allocate memory for hibernation image. |
8357376d | 1923 | * |
ef96f639 RW |
1924 | * We first try to allocate as many highmem pages as there are |
1925 | * saveable highmem pages in the system. If that fails, we allocate | |
1926 | * non-highmem pages for the copies of the remaining highmem ones. | |
8357376d | 1927 | * |
ef96f639 RW |
1928 | * In this approach it is likely that the copies of highmem pages will |
1929 | * also be located in the high memory, because of the way in which | |
1930 | * copy_data_pages() works. | |
8357376d | 1931 | */ |
efd5a852 RW |
1932 | static int swsusp_alloc(struct memory_bitmap *orig_bm, |
1933 | struct memory_bitmap *copy_bm, | |
1934 | unsigned int nr_pages, unsigned int nr_highmem) | |
054bd4c1 | 1935 | { |
8357376d | 1936 | if (nr_highmem > 0) { |
2e725a06 | 1937 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1938 | goto err_out; |
1939 | if (nr_highmem > alloc_highmem) { | |
1940 | nr_highmem -= alloc_highmem; | |
1941 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1942 | } | |
8357376d | 1943 | } |
64a473cb RW |
1944 | if (nr_pages > alloc_normal) { |
1945 | nr_pages -= alloc_normal; | |
1946 | while (nr_pages-- > 0) { | |
1947 | struct page *page; | |
1948 | ||
1949 | page = alloc_image_page(GFP_ATOMIC | __GFP_COLD); | |
1950 | if (!page) | |
1951 | goto err_out; | |
1952 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
1953 | } | |
25761b6e | 1954 | } |
64a473cb | 1955 | |
b788db79 | 1956 | return 0; |
25761b6e | 1957 | |
64a473cb | 1958 | err_out: |
b788db79 | 1959 | swsusp_free(); |
2e725a06 | 1960 | return -ENOMEM; |
25761b6e RW |
1961 | } |
1962 | ||
722a9f92 | 1963 | asmlinkage __visible int swsusp_save(void) |
25761b6e | 1964 | { |
8357376d | 1965 | unsigned int nr_pages, nr_highmem; |
25761b6e | 1966 | |
07c3bb57 | 1967 | printk(KERN_INFO "PM: Creating hibernation image:\n"); |
25761b6e | 1968 | |
9f8f2172 | 1969 | drain_local_pages(NULL); |
a0f49651 | 1970 | nr_pages = count_data_pages(); |
8357376d | 1971 | nr_highmem = count_highmem_pages(); |
23976728 | 1972 | printk(KERN_INFO "PM: Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 1973 | |
8357376d | 1974 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
23976728 | 1975 | printk(KERN_ERR "PM: Not enough free memory\n"); |
25761b6e RW |
1976 | return -ENOMEM; |
1977 | } | |
1978 | ||
8357376d | 1979 | if (swsusp_alloc(&orig_bm, ©_bm, nr_pages, nr_highmem)) { |
23976728 | 1980 | printk(KERN_ERR "PM: Memory allocation failed\n"); |
a0f49651 | 1981 | return -ENOMEM; |
8357376d | 1982 | } |
25761b6e | 1983 | |
ef96f639 RW |
1984 | /* |
1985 | * During allocating of suspend pagedir, new cold pages may appear. | |
25761b6e RW |
1986 | * Kill them. |
1987 | */ | |
9f8f2172 | 1988 | drain_local_pages(NULL); |
b788db79 | 1989 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
1990 | |
1991 | /* | |
1992 | * End of critical section. From now on, we can write to memory, | |
1993 | * but we should not touch disk. This specially means we must _not_ | |
1994 | * touch swap space! Except we must write out our image of course. | |
1995 | */ | |
1996 | ||
8357376d | 1997 | nr_pages += nr_highmem; |
a0f49651 | 1998 | nr_copy_pages = nr_pages; |
8357376d | 1999 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 2000 | |
23976728 RW |
2001 | printk(KERN_INFO "PM: Hibernation image created (%d pages copied)\n", |
2002 | nr_pages); | |
8357376d | 2003 | |
25761b6e RW |
2004 | return 0; |
2005 | } | |
f577eb30 | 2006 | |
d307c4a8 RW |
2007 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
2008 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 2009 | { |
d307c4a8 | 2010 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 2011 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
2012 | return 0; |
2013 | } | |
2014 | ||
2015 | static char *check_image_kernel(struct swsusp_info *info) | |
2016 | { | |
2017 | if (info->version_code != LINUX_VERSION_CODE) | |
2018 | return "kernel version"; | |
2019 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
2020 | return "system type"; | |
2021 | if (strcmp(info->uts.release,init_utsname()->release)) | |
2022 | return "kernel release"; | |
2023 | if (strcmp(info->uts.version,init_utsname()->version)) | |
2024 | return "version"; | |
2025 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
2026 | return "machine"; | |
2027 | return NULL; | |
2028 | } | |
2029 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
2030 | ||
af508b34 RW |
2031 | unsigned long snapshot_get_image_size(void) |
2032 | { | |
2033 | return nr_copy_pages + nr_meta_pages + 1; | |
2034 | } | |
2035 | ||
d307c4a8 RW |
2036 | static int init_header(struct swsusp_info *info) |
2037 | { | |
2038 | memset(info, 0, sizeof(struct swsusp_info)); | |
0ed5fd13 | 2039 | info->num_physpages = get_num_physpages(); |
f577eb30 | 2040 | info->image_pages = nr_copy_pages; |
af508b34 | 2041 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
2042 | info->size = info->pages; |
2043 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 2044 | return init_header_complete(info); |
f577eb30 RW |
2045 | } |
2046 | ||
2047 | /** | |
ef96f639 RW |
2048 | * pack_pfns - Prepare PFNs for saving. |
2049 | * @bm: Memory bitmap. | |
2050 | * @buf: Memory buffer to store the PFNs in. | |
2051 | * | |
2052 | * PFNs corresponding to set bits in @bm are stored in the area of memory | |
2053 | * pointed to by @buf (1 page at a time). | |
f577eb30 | 2054 | */ |
efd5a852 | 2055 | static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2056 | { |
2057 | int j; | |
2058 | ||
b788db79 | 2059 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
2060 | buf[j] = memory_bm_next_pfn(bm); |
2061 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 2062 | break; |
85055dd8 MS |
2063 | /* Save page key for data page (s390 only). */ |
2064 | page_key_read(buf + j); | |
f577eb30 | 2065 | } |
f577eb30 RW |
2066 | } |
2067 | ||
2068 | /** | |
ef96f639 RW |
2069 | * snapshot_read_next - Get the address to read the next image page from. |
2070 | * @handle: Snapshot handle to be used for the reading. | |
f577eb30 | 2071 | * |
ef96f639 RW |
2072 | * On the first call, @handle should point to a zeroed snapshot_handle |
2073 | * structure. The structure gets populated then and a pointer to it should be | |
2074 | * passed to this function every next time. | |
f577eb30 | 2075 | * |
ef96f639 RW |
2076 | * On success, the function returns a positive number. Then, the caller |
2077 | * is allowed to read up to the returned number of bytes from the memory | |
2078 | * location computed by the data_of() macro. | |
f577eb30 | 2079 | * |
ef96f639 RW |
2080 | * The function returns 0 to indicate the end of the data stream condition, |
2081 | * and negative numbers are returned on errors. If that happens, the structure | |
2082 | * pointed to by @handle is not updated and should not be used any more. | |
f577eb30 | 2083 | */ |
d3c1b24c | 2084 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 2085 | { |
fb13a28b | 2086 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2087 | return 0; |
b788db79 | 2088 | |
f577eb30 RW |
2089 | if (!buffer) { |
2090 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 2091 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
2092 | if (!buffer) |
2093 | return -ENOMEM; | |
2094 | } | |
d3c1b24c | 2095 | if (!handle->cur) { |
d307c4a8 RW |
2096 | int error; |
2097 | ||
2098 | error = init_header((struct swsusp_info *)buffer); | |
2099 | if (error) | |
2100 | return error; | |
f577eb30 | 2101 | handle->buffer = buffer; |
b788db79 RW |
2102 | memory_bm_position_reset(&orig_bm); |
2103 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 2104 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 2105 | clear_page(buffer); |
d3c1b24c JS |
2106 | pack_pfns(buffer, &orig_bm); |
2107 | } else { | |
2108 | struct page *page; | |
b788db79 | 2109 | |
d3c1b24c JS |
2110 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
2111 | if (PageHighMem(page)) { | |
ef96f639 RW |
2112 | /* |
2113 | * Highmem pages are copied to the buffer, | |
d3c1b24c JS |
2114 | * because we can't return with a kmapped |
2115 | * highmem page (we may not be called again). | |
2116 | */ | |
2117 | void *kaddr; | |
8357376d | 2118 | |
0de9a1e2 | 2119 | kaddr = kmap_atomic(page); |
3ecb01df | 2120 | copy_page(buffer, kaddr); |
0de9a1e2 | 2121 | kunmap_atomic(kaddr); |
d3c1b24c JS |
2122 | handle->buffer = buffer; |
2123 | } else { | |
2124 | handle->buffer = page_address(page); | |
f577eb30 | 2125 | } |
f577eb30 | 2126 | } |
d3c1b24c JS |
2127 | handle->cur++; |
2128 | return PAGE_SIZE; | |
f577eb30 RW |
2129 | } |
2130 | ||
6dbecfd3 RW |
2131 | static void duplicate_memory_bitmap(struct memory_bitmap *dst, |
2132 | struct memory_bitmap *src) | |
2133 | { | |
2134 | unsigned long pfn; | |
2135 | ||
2136 | memory_bm_position_reset(src); | |
2137 | pfn = memory_bm_next_pfn(src); | |
2138 | while (pfn != BM_END_OF_MAP) { | |
2139 | memory_bm_set_bit(dst, pfn); | |
2140 | pfn = memory_bm_next_pfn(src); | |
2141 | } | |
2142 | } | |
2143 | ||
f577eb30 | 2144 | /** |
ef96f639 RW |
2145 | * mark_unsafe_pages - Mark pages that were used before hibernation. |
2146 | * | |
2147 | * Mark the pages that cannot be used for storing the image during restoration, | |
2148 | * because they conflict with the pages that had been used before hibernation. | |
f577eb30 | 2149 | */ |
6dbecfd3 | 2150 | static void mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 | 2151 | { |
6dbecfd3 | 2152 | unsigned long pfn; |
f577eb30 | 2153 | |
6dbecfd3 RW |
2154 | /* Clear the "free"/"unsafe" bit for all PFNs */ |
2155 | memory_bm_position_reset(free_pages_map); | |
2156 | pfn = memory_bm_next_pfn(free_pages_map); | |
2157 | while (pfn != BM_END_OF_MAP) { | |
2158 | memory_bm_clear_current(free_pages_map); | |
2159 | pfn = memory_bm_next_pfn(free_pages_map); | |
f577eb30 RW |
2160 | } |
2161 | ||
6dbecfd3 RW |
2162 | /* Mark pages that correspond to the "original" PFNs as "unsafe" */ |
2163 | duplicate_memory_bitmap(free_pages_map, bm); | |
f577eb30 | 2164 | |
940864dd | 2165 | allocated_unsafe_pages = 0; |
f577eb30 RW |
2166 | } |
2167 | ||
d307c4a8 | 2168 | static int check_header(struct swsusp_info *info) |
f577eb30 | 2169 | { |
d307c4a8 | 2170 | char *reason; |
f577eb30 | 2171 | |
d307c4a8 | 2172 | reason = check_image_kernel(info); |
0ed5fd13 | 2173 | if (!reason && info->num_physpages != get_num_physpages()) |
f577eb30 | 2174 | reason = "memory size"; |
f577eb30 | 2175 | if (reason) { |
23976728 | 2176 | printk(KERN_ERR "PM: Image mismatch: %s\n", reason); |
f577eb30 RW |
2177 | return -EPERM; |
2178 | } | |
2179 | return 0; | |
2180 | } | |
2181 | ||
2182 | /** | |
ef96f639 | 2183 | * load header - Check the image header and copy the data from it. |
f577eb30 | 2184 | */ |
efd5a852 | 2185 | static int load_header(struct swsusp_info *info) |
f577eb30 RW |
2186 | { |
2187 | int error; | |
f577eb30 | 2188 | |
940864dd | 2189 | restore_pblist = NULL; |
f577eb30 RW |
2190 | error = check_header(info); |
2191 | if (!error) { | |
f577eb30 RW |
2192 | nr_copy_pages = info->image_pages; |
2193 | nr_meta_pages = info->pages - info->image_pages - 1; | |
2194 | } | |
2195 | return error; | |
2196 | } | |
2197 | ||
2198 | /** | |
ef96f639 RW |
2199 | * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap. |
2200 | * @bm: Memory bitmap. | |
2201 | * @buf: Area of memory containing the PFNs. | |
2202 | * | |
2203 | * For each element of the array pointed to by @buf (1 page at a time), set the | |
2204 | * corresponding bit in @bm. | |
f577eb30 | 2205 | */ |
69643279 | 2206 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2207 | { |
2208 | int j; | |
2209 | ||
940864dd RW |
2210 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
2211 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
2212 | break; | |
2213 | ||
85055dd8 MS |
2214 | /* Extract and buffer page key for data page (s390 only). */ |
2215 | page_key_memorize(buf + j); | |
2216 | ||
6dbecfd3 | 2217 | if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j])) |
69643279 RW |
2218 | memory_bm_set_bit(bm, buf[j]); |
2219 | else | |
2220 | return -EFAULT; | |
f577eb30 | 2221 | } |
69643279 RW |
2222 | |
2223 | return 0; | |
f577eb30 RW |
2224 | } |
2225 | ||
8357376d | 2226 | #ifdef CONFIG_HIGHMEM |
ef96f639 RW |
2227 | /* |
2228 | * struct highmem_pbe is used for creating the list of highmem pages that | |
8357376d RW |
2229 | * should be restored atomically during the resume from disk, because the page |
2230 | * frames they have occupied before the suspend are in use. | |
2231 | */ | |
2232 | struct highmem_pbe { | |
2233 | struct page *copy_page; /* data is here now */ | |
2234 | struct page *orig_page; /* data was here before the suspend */ | |
2235 | struct highmem_pbe *next; | |
2236 | }; | |
2237 | ||
ef96f639 RW |
2238 | /* |
2239 | * List of highmem PBEs needed for restoring the highmem pages that were | |
8357376d RW |
2240 | * allocated before the suspend and included in the suspend image, but have |
2241 | * also been allocated by the "resume" kernel, so their contents cannot be | |
2242 | * written directly to their "original" page frames. | |
2243 | */ | |
2244 | static struct highmem_pbe *highmem_pblist; | |
2245 | ||
2246 | /** | |
ef96f639 RW |
2247 | * count_highmem_image_pages - Compute the number of highmem pages in the image. |
2248 | * @bm: Memory bitmap. | |
2249 | * | |
2250 | * The bits in @bm that correspond to image pages are assumed to be set. | |
8357376d | 2251 | */ |
8357376d RW |
2252 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) |
2253 | { | |
2254 | unsigned long pfn; | |
2255 | unsigned int cnt = 0; | |
2256 | ||
2257 | memory_bm_position_reset(bm); | |
2258 | pfn = memory_bm_next_pfn(bm); | |
2259 | while (pfn != BM_END_OF_MAP) { | |
2260 | if (PageHighMem(pfn_to_page(pfn))) | |
2261 | cnt++; | |
2262 | ||
2263 | pfn = memory_bm_next_pfn(bm); | |
2264 | } | |
2265 | return cnt; | |
2266 | } | |
2267 | ||
8357376d RW |
2268 | static unsigned int safe_highmem_pages; |
2269 | ||
2270 | static struct memory_bitmap *safe_highmem_bm; | |
2271 | ||
ef96f639 RW |
2272 | /** |
2273 | * prepare_highmem_image - Allocate memory for loading highmem data from image. | |
2274 | * @bm: Pointer to an uninitialized memory bitmap structure. | |
2275 | * @nr_highmem_p: Pointer to the number of highmem image pages. | |
2276 | * | |
2277 | * Try to allocate as many highmem pages as there are highmem image pages | |
2278 | * (@nr_highmem_p points to the variable containing the number of highmem image | |
2279 | * pages). The pages that are "safe" (ie. will not be overwritten when the | |
2280 | * hibernation image is restored entirely) have the corresponding bits set in | |
2281 | * @bm (it must be unitialized). | |
2282 | * | |
2283 | * NOTE: This function should not be called if there are no highmem image pages. | |
2284 | */ | |
efd5a852 RW |
2285 | static int prepare_highmem_image(struct memory_bitmap *bm, |
2286 | unsigned int *nr_highmem_p) | |
8357376d RW |
2287 | { |
2288 | unsigned int to_alloc; | |
2289 | ||
2290 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
2291 | return -ENOMEM; | |
2292 | ||
2293 | if (get_highmem_buffer(PG_SAFE)) | |
2294 | return -ENOMEM; | |
2295 | ||
2296 | to_alloc = count_free_highmem_pages(); | |
2297 | if (to_alloc > *nr_highmem_p) | |
2298 | to_alloc = *nr_highmem_p; | |
2299 | else | |
2300 | *nr_highmem_p = to_alloc; | |
2301 | ||
2302 | safe_highmem_pages = 0; | |
2303 | while (to_alloc-- > 0) { | |
2304 | struct page *page; | |
2305 | ||
2306 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 2307 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
2308 | /* The page is "safe", set its bit the bitmap */ |
2309 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
2310 | safe_highmem_pages++; | |
2311 | } | |
2312 | /* Mark the page as allocated */ | |
7be98234 RW |
2313 | swsusp_set_page_forbidden(page); |
2314 | swsusp_set_page_free(page); | |
8357376d RW |
2315 | } |
2316 | memory_bm_position_reset(bm); | |
2317 | safe_highmem_bm = bm; | |
2318 | return 0; | |
2319 | } | |
2320 | ||
ef96f639 RW |
2321 | static struct page *last_highmem_page; |
2322 | ||
8357376d | 2323 | /** |
ef96f639 RW |
2324 | * get_highmem_page_buffer - Prepare a buffer to store a highmem image page. |
2325 | * | |
2326 | * For a given highmem image page get a buffer that suspend_write_next() should | |
2327 | * return to its caller to write to. | |
8357376d | 2328 | * |
ef96f639 RW |
2329 | * If the page is to be saved to its "original" page frame or a copy of |
2330 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
2331 | * the copy of the page is to be made in normal memory, so the address of | |
2332 | * the copy is returned. | |
8357376d | 2333 | * |
ef96f639 RW |
2334 | * If @buffer is returned, the caller of suspend_write_next() will write |
2335 | * the page's contents to @buffer, so they will have to be copied to the | |
2336 | * right location on the next call to suspend_write_next() and it is done | |
2337 | * with the help of copy_last_highmem_page(). For this purpose, if | |
2338 | * @buffer is returned, @last_highmem_page is set to the page to which | |
2339 | * the data will have to be copied from @buffer. | |
8357376d | 2340 | */ |
efd5a852 RW |
2341 | static void *get_highmem_page_buffer(struct page *page, |
2342 | struct chain_allocator *ca) | |
8357376d RW |
2343 | { |
2344 | struct highmem_pbe *pbe; | |
2345 | void *kaddr; | |
2346 | ||
7be98234 | 2347 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
ef96f639 RW |
2348 | /* |
2349 | * We have allocated the "original" page frame and we can | |
8357376d RW |
2350 | * use it directly to store the loaded page. |
2351 | */ | |
2352 | last_highmem_page = page; | |
2353 | return buffer; | |
2354 | } | |
ef96f639 RW |
2355 | /* |
2356 | * The "original" page frame has not been allocated and we have to | |
8357376d RW |
2357 | * use a "safe" page frame to store the loaded page. |
2358 | */ | |
2359 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
2360 | if (!pbe) { | |
2361 | swsusp_free(); | |
69643279 | 2362 | return ERR_PTR(-ENOMEM); |
8357376d RW |
2363 | } |
2364 | pbe->orig_page = page; | |
2365 | if (safe_highmem_pages > 0) { | |
2366 | struct page *tmp; | |
2367 | ||
2368 | /* Copy of the page will be stored in high memory */ | |
2369 | kaddr = buffer; | |
2370 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
2371 | safe_highmem_pages--; | |
2372 | last_highmem_page = tmp; | |
2373 | pbe->copy_page = tmp; | |
2374 | } else { | |
2375 | /* Copy of the page will be stored in normal memory */ | |
2376 | kaddr = safe_pages_list; | |
2377 | safe_pages_list = safe_pages_list->next; | |
2378 | pbe->copy_page = virt_to_page(kaddr); | |
2379 | } | |
2380 | pbe->next = highmem_pblist; | |
2381 | highmem_pblist = pbe; | |
2382 | return kaddr; | |
2383 | } | |
2384 | ||
2385 | /** | |
ef96f639 RW |
2386 | * copy_last_highmem_page - Copy most the most recent highmem image page. |
2387 | * | |
2388 | * Copy the contents of a highmem image from @buffer, where the caller of | |
2389 | * snapshot_write_next() has stored them, to the right location represented by | |
2390 | * @last_highmem_page . | |
8357376d | 2391 | */ |
8357376d RW |
2392 | static void copy_last_highmem_page(void) |
2393 | { | |
2394 | if (last_highmem_page) { | |
2395 | void *dst; | |
2396 | ||
0de9a1e2 | 2397 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2398 | copy_page(dst, buffer); |
0de9a1e2 | 2399 | kunmap_atomic(dst); |
8357376d RW |
2400 | last_highmem_page = NULL; |
2401 | } | |
2402 | } | |
2403 | ||
2404 | static inline int last_highmem_page_copied(void) | |
2405 | { | |
2406 | return !last_highmem_page; | |
2407 | } | |
2408 | ||
2409 | static inline void free_highmem_data(void) | |
2410 | { | |
2411 | if (safe_highmem_bm) | |
2412 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2413 | ||
2414 | if (buffer) | |
2415 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2416 | } | |
2417 | #else | |
efd5a852 | 2418 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } |
8357376d | 2419 | |
efd5a852 RW |
2420 | static inline int prepare_highmem_image(struct memory_bitmap *bm, |
2421 | unsigned int *nr_highmem_p) { return 0; } | |
8357376d | 2422 | |
efd5a852 RW |
2423 | static inline void *get_highmem_page_buffer(struct page *page, |
2424 | struct chain_allocator *ca) | |
8357376d | 2425 | { |
69643279 | 2426 | return ERR_PTR(-EINVAL); |
8357376d RW |
2427 | } |
2428 | ||
2429 | static inline void copy_last_highmem_page(void) {} | |
2430 | static inline int last_highmem_page_copied(void) { return 1; } | |
2431 | static inline void free_highmem_data(void) {} | |
2432 | #endif /* CONFIG_HIGHMEM */ | |
2433 | ||
ef96f639 RW |
2434 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2435 | ||
f577eb30 | 2436 | /** |
ef96f639 RW |
2437 | * prepare_image - Make room for loading hibernation image. |
2438 | * @new_bm: Unitialized memory bitmap structure. | |
2439 | * @bm: Memory bitmap with unsafe pages marked. | |
2440 | * | |
2441 | * Use @bm to mark the pages that will be overwritten in the process of | |
2442 | * restoring the system memory state from the suspend image ("unsafe" pages) | |
2443 | * and allocate memory for the image. | |
968808b8 | 2444 | * |
ef96f639 RW |
2445 | * The idea is to allocate a new memory bitmap first and then allocate |
2446 | * as many pages as needed for image data, but without specifying what those | |
2447 | * pages will be used for just yet. Instead, we mark them all as allocated and | |
2448 | * create a lists of "safe" pages to be used later. On systems with high | |
2449 | * memory a list of "safe" highmem pages is created too. | |
f577eb30 | 2450 | */ |
efd5a852 | 2451 | static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) |
f577eb30 | 2452 | { |
8357376d | 2453 | unsigned int nr_pages, nr_highmem; |
9c744481 | 2454 | struct linked_page *lp; |
940864dd | 2455 | int error; |
f577eb30 | 2456 | |
8357376d RW |
2457 | /* If there is no highmem, the buffer will not be necessary */ |
2458 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2459 | buffer = NULL; | |
2460 | ||
2461 | nr_highmem = count_highmem_image_pages(bm); | |
6dbecfd3 | 2462 | mark_unsafe_pages(bm); |
940864dd RW |
2463 | |
2464 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2465 | if (error) | |
2466 | goto Free; | |
2467 | ||
2468 | duplicate_memory_bitmap(new_bm, bm); | |
2469 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2470 | if (nr_highmem > 0) { |
2471 | error = prepare_highmem_image(bm, &nr_highmem); | |
2472 | if (error) | |
2473 | goto Free; | |
2474 | } | |
ef96f639 RW |
2475 | /* |
2476 | * Reserve some safe pages for potential later use. | |
940864dd RW |
2477 | * |
2478 | * NOTE: This way we make sure there will be enough safe pages for the | |
2479 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2480 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
9c744481 RW |
2481 | * |
2482 | * nr_copy_pages cannot be less than allocated_unsafe_pages too. | |
940864dd | 2483 | */ |
8357376d | 2484 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2485 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2486 | while (nr_pages > 0) { | |
8357376d | 2487 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2488 | if (!lp) { |
f577eb30 | 2489 | error = -ENOMEM; |
940864dd RW |
2490 | goto Free; |
2491 | } | |
9c744481 RW |
2492 | lp->next = safe_pages_list; |
2493 | safe_pages_list = lp; | |
940864dd | 2494 | nr_pages--; |
f577eb30 | 2495 | } |
940864dd | 2496 | /* Preallocate memory for the image */ |
8357376d | 2497 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2498 | while (nr_pages > 0) { |
2499 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2500 | if (!lp) { | |
2501 | error = -ENOMEM; | |
2502 | goto Free; | |
2503 | } | |
7be98234 | 2504 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2505 | /* The page is "safe", add it to the list */ |
2506 | lp->next = safe_pages_list; | |
2507 | safe_pages_list = lp; | |
968808b8 | 2508 | } |
940864dd | 2509 | /* Mark the page as allocated */ |
7be98234 RW |
2510 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2511 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2512 | nr_pages--; |
968808b8 | 2513 | } |
940864dd RW |
2514 | return 0; |
2515 | ||
59a49335 | 2516 | Free: |
940864dd | 2517 | swsusp_free(); |
f577eb30 RW |
2518 | return error; |
2519 | } | |
2520 | ||
940864dd | 2521 | /** |
ef96f639 RW |
2522 | * get_buffer - Get the address to store the next image data page. |
2523 | * | |
2524 | * Get the address that snapshot_write_next() should return to its caller to | |
2525 | * write to. | |
940864dd | 2526 | */ |
940864dd | 2527 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) |
968808b8 | 2528 | { |
940864dd | 2529 | struct pbe *pbe; |
69643279 RW |
2530 | struct page *page; |
2531 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2532 | |
69643279 RW |
2533 | if (pfn == BM_END_OF_MAP) |
2534 | return ERR_PTR(-EFAULT); | |
2535 | ||
2536 | page = pfn_to_page(pfn); | |
8357376d RW |
2537 | if (PageHighMem(page)) |
2538 | return get_highmem_page_buffer(page, ca); | |
2539 | ||
7be98234 | 2540 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
ef96f639 RW |
2541 | /* |
2542 | * We have allocated the "original" page frame and we can | |
940864dd | 2543 | * use it directly to store the loaded page. |
968808b8 | 2544 | */ |
940864dd RW |
2545 | return page_address(page); |
2546 | ||
ef96f639 RW |
2547 | /* |
2548 | * The "original" page frame has not been allocated and we have to | |
940864dd | 2549 | * use a "safe" page frame to store the loaded page. |
968808b8 | 2550 | */ |
940864dd RW |
2551 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2552 | if (!pbe) { | |
2553 | swsusp_free(); | |
69643279 | 2554 | return ERR_PTR(-ENOMEM); |
940864dd | 2555 | } |
8357376d RW |
2556 | pbe->orig_address = page_address(page); |
2557 | pbe->address = safe_pages_list; | |
940864dd RW |
2558 | safe_pages_list = safe_pages_list->next; |
2559 | pbe->next = restore_pblist; | |
2560 | restore_pblist = pbe; | |
8357376d | 2561 | return pbe->address; |
968808b8 RW |
2562 | } |
2563 | ||
f577eb30 | 2564 | /** |
ef96f639 RW |
2565 | * snapshot_write_next - Get the address to store the next image page. |
2566 | * @handle: Snapshot handle structure to guide the writing. | |
f577eb30 | 2567 | * |
ef96f639 RW |
2568 | * On the first call, @handle should point to a zeroed snapshot_handle |
2569 | * structure. The structure gets populated then and a pointer to it should be | |
2570 | * passed to this function every next time. | |
f577eb30 | 2571 | * |
ef96f639 RW |
2572 | * On success, the function returns a positive number. Then, the caller |
2573 | * is allowed to write up to the returned number of bytes to the memory | |
2574 | * location computed by the data_of() macro. | |
f577eb30 | 2575 | * |
ef96f639 RW |
2576 | * The function returns 0 to indicate the "end of file" condition. Negative |
2577 | * numbers are returned on errors, in which cases the structure pointed to by | |
2578 | * @handle is not updated and should not be used any more. | |
f577eb30 | 2579 | */ |
d3c1b24c | 2580 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2581 | { |
940864dd | 2582 | static struct chain_allocator ca; |
f577eb30 RW |
2583 | int error = 0; |
2584 | ||
940864dd | 2585 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2586 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2587 | return 0; |
940864dd | 2588 | |
d3c1b24c JS |
2589 | handle->sync_read = 1; |
2590 | ||
2591 | if (!handle->cur) { | |
8357376d RW |
2592 | if (!buffer) |
2593 | /* This makes the buffer be freed by swsusp_free() */ | |
2594 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2595 | ||
f577eb30 RW |
2596 | if (!buffer) |
2597 | return -ENOMEM; | |
8357376d | 2598 | |
f577eb30 | 2599 | handle->buffer = buffer; |
d3c1b24c JS |
2600 | } else if (handle->cur == 1) { |
2601 | error = load_header(buffer); | |
2602 | if (error) | |
2603 | return error; | |
940864dd | 2604 | |
9c744481 RW |
2605 | safe_pages_list = NULL; |
2606 | ||
d3c1b24c JS |
2607 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2608 | if (error) | |
2609 | return error; | |
2610 | ||
85055dd8 MS |
2611 | /* Allocate buffer for page keys. */ |
2612 | error = page_key_alloc(nr_copy_pages); | |
2613 | if (error) | |
2614 | return error; | |
2615 | ||
4c0b6c10 | 2616 | hibernate_restore_protection_begin(); |
d3c1b24c JS |
2617 | } else if (handle->cur <= nr_meta_pages + 1) { |
2618 | error = unpack_orig_pfns(buffer, ©_bm); | |
2619 | if (error) | |
2620 | return error; | |
940864dd | 2621 | |
d3c1b24c JS |
2622 | if (handle->cur == nr_meta_pages + 1) { |
2623 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2624 | if (error) |
2625 | return error; | |
2626 | ||
d3c1b24c JS |
2627 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2628 | memory_bm_position_reset(&orig_bm); | |
2629 | restore_pblist = NULL; | |
940864dd | 2630 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2631 | handle->sync_read = 0; |
69643279 RW |
2632 | if (IS_ERR(handle->buffer)) |
2633 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2634 | } |
f577eb30 | 2635 | } else { |
d3c1b24c | 2636 | copy_last_highmem_page(); |
85055dd8 MS |
2637 | /* Restore page key for data page (s390 only). */ |
2638 | page_key_write(handle->buffer); | |
4c0b6c10 | 2639 | hibernate_restore_protect_page(handle->buffer); |
d3c1b24c JS |
2640 | handle->buffer = get_buffer(&orig_bm, &ca); |
2641 | if (IS_ERR(handle->buffer)) | |
2642 | return PTR_ERR(handle->buffer); | |
2643 | if (handle->buffer != buffer) | |
2644 | handle->sync_read = 0; | |
f577eb30 | 2645 | } |
d3c1b24c JS |
2646 | handle->cur++; |
2647 | return PAGE_SIZE; | |
f577eb30 RW |
2648 | } |
2649 | ||
8357376d | 2650 | /** |
ef96f639 RW |
2651 | * snapshot_write_finalize - Complete the loading of a hibernation image. |
2652 | * | |
2653 | * Must be called after the last call to snapshot_write_next() in case the last | |
2654 | * page in the image happens to be a highmem page and its contents should be | |
2655 | * stored in highmem. Additionally, it recycles bitmap memory that's not | |
2656 | * necessary any more. | |
8357376d | 2657 | */ |
8357376d RW |
2658 | void snapshot_write_finalize(struct snapshot_handle *handle) |
2659 | { | |
2660 | copy_last_highmem_page(); | |
85055dd8 MS |
2661 | /* Restore page key for data page (s390 only). */ |
2662 | page_key_write(handle->buffer); | |
2663 | page_key_free(); | |
4c0b6c10 | 2664 | hibernate_restore_protect_page(handle->buffer); |
307c5971 | 2665 | /* Do that only if we have loaded the image entirely */ |
d3c1b24c | 2666 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
307c5971 | 2667 | memory_bm_recycle(&orig_bm); |
8357376d RW |
2668 | free_highmem_data(); |
2669 | } | |
2670 | } | |
2671 | ||
f577eb30 RW |
2672 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2673 | { | |
8357376d | 2674 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2675 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2676 | } | |
2677 | ||
8357376d RW |
2678 | #ifdef CONFIG_HIGHMEM |
2679 | /* Assumes that @buf is ready and points to a "safe" page */ | |
efd5a852 RW |
2680 | static inline void swap_two_pages_data(struct page *p1, struct page *p2, |
2681 | void *buf) | |
940864dd | 2682 | { |
8357376d RW |
2683 | void *kaddr1, *kaddr2; |
2684 | ||
0de9a1e2 CW |
2685 | kaddr1 = kmap_atomic(p1); |
2686 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2687 | copy_page(buf, kaddr1); |
2688 | copy_page(kaddr1, kaddr2); | |
2689 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2690 | kunmap_atomic(kaddr2); |
2691 | kunmap_atomic(kaddr1); | |
8357376d RW |
2692 | } |
2693 | ||
2694 | /** | |
ef96f639 RW |
2695 | * restore_highmem - Put highmem image pages into their original locations. |
2696 | * | |
2697 | * For each highmem page that was in use before hibernation and is included in | |
2698 | * the image, and also has been allocated by the "restore" kernel, swap its | |
2699 | * current contents with the previous (ie. "before hibernation") ones. | |
8357376d | 2700 | * |
ef96f639 RW |
2701 | * If the restore eventually fails, we can call this function once again and |
2702 | * restore the highmem state as seen by the restore kernel. | |
8357376d | 2703 | */ |
8357376d RW |
2704 | int restore_highmem(void) |
2705 | { | |
2706 | struct highmem_pbe *pbe = highmem_pblist; | |
2707 | void *buf; | |
2708 | ||
2709 | if (!pbe) | |
2710 | return 0; | |
2711 | ||
2712 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2713 | if (!buf) | |
2714 | return -ENOMEM; | |
2715 | ||
2716 | while (pbe) { | |
2717 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2718 | pbe = pbe->next; | |
2719 | } | |
2720 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2721 | return 0; | |
f577eb30 | 2722 | } |
8357376d | 2723 | #endif /* CONFIG_HIGHMEM */ |