Commit | Line | Data |
---|---|---|
55716d26 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
25761b6e | 2 | /* |
96bc7aec | 3 | * linux/kernel/power/snapshot.c |
25761b6e | 4 | * |
8357376d | 5 | * This file provides system snapshot/restore functionality for swsusp. |
25761b6e | 6 | * |
a2531293 | 7 | * Copyright (C) 1998-2005 Pavel Machek <pavel@ucw.cz> |
8357376d | 8 | * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl> |
25761b6e RW |
9 | */ |
10 | ||
7a7b99bf | 11 | #define pr_fmt(fmt) "PM: hibernation: " fmt |
64ec72a1 | 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> |
57c8a661 | 24 | #include <linux/memblock.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> |
61f6d09a | 33 | #include <linux/set_memory.h> |
25761b6e | 34 | |
7c0f6ba6 | 35 | #include <linux/uaccess.h> |
25761b6e | 36 | #include <asm/mmu_context.h> |
25761b6e RW |
37 | #include <asm/tlbflush.h> |
38 | #include <asm/io.h> | |
39 | ||
25761b6e RW |
40 | #include "power.h" |
41 | ||
49368a47 | 42 | #if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_ARCH_HAS_SET_MEMORY) |
4c0b6c10 RW |
43 | static bool hibernate_restore_protection; |
44 | static bool hibernate_restore_protection_active; | |
45 | ||
46 | void enable_restore_image_protection(void) | |
47 | { | |
48 | hibernate_restore_protection = true; | |
49 | } | |
50 | ||
51 | static inline void hibernate_restore_protection_begin(void) | |
52 | { | |
53 | hibernate_restore_protection_active = hibernate_restore_protection; | |
54 | } | |
55 | ||
56 | static inline void hibernate_restore_protection_end(void) | |
57 | { | |
58 | hibernate_restore_protection_active = false; | |
59 | } | |
60 | ||
61 | static inline void hibernate_restore_protect_page(void *page_address) | |
62 | { | |
63 | if (hibernate_restore_protection_active) | |
64 | set_memory_ro((unsigned long)page_address, 1); | |
65 | } | |
66 | ||
67 | static inline void hibernate_restore_unprotect_page(void *page_address) | |
68 | { | |
69 | if (hibernate_restore_protection_active) | |
70 | set_memory_rw((unsigned long)page_address, 1); | |
71 | } | |
72 | #else | |
73 | static inline void hibernate_restore_protection_begin(void) {} | |
74 | static inline void hibernate_restore_protection_end(void) {} | |
75 | static inline void hibernate_restore_protect_page(void *page_address) {} | |
76 | static inline void hibernate_restore_unprotect_page(void *page_address) {} | |
49368a47 | 77 | #endif /* CONFIG_STRICT_KERNEL_RWX && CONFIG_ARCH_HAS_SET_MEMORY */ |
4c0b6c10 | 78 | |
2abf962a MR |
79 | |
80 | /* | |
81 | * The calls to set_direct_map_*() should not fail because remapping a page | |
82 | * here means that we only update protection bits in an existing PTE. | |
83 | * It is still worth to have a warning here if something changes and this | |
84 | * will no longer be the case. | |
85 | */ | |
86 | static inline void hibernate_map_page(struct page *page) | |
87 | { | |
88 | if (IS_ENABLED(CONFIG_ARCH_HAS_SET_DIRECT_MAP)) { | |
89 | int ret = set_direct_map_default_noflush(page); | |
90 | ||
91 | if (ret) | |
92 | pr_warn_once("Failed to remap page\n"); | |
93 | } else { | |
94 | debug_pagealloc_map_pages(page, 1); | |
95 | } | |
96 | } | |
97 | ||
98 | static inline void hibernate_unmap_page(struct page *page) | |
99 | { | |
100 | if (IS_ENABLED(CONFIG_ARCH_HAS_SET_DIRECT_MAP)) { | |
101 | unsigned long addr = (unsigned long)page_address(page); | |
102 | int ret = set_direct_map_invalid_noflush(page); | |
103 | ||
104 | if (ret) | |
105 | pr_warn_once("Failed to remap page\n"); | |
106 | ||
107 | flush_tlb_kernel_range(addr, addr + PAGE_SIZE); | |
108 | } else { | |
109 | debug_pagealloc_unmap_pages(page, 1); | |
110 | } | |
111 | } | |
112 | ||
74dfd666 RW |
113 | static int swsusp_page_is_free(struct page *); |
114 | static void swsusp_set_page_forbidden(struct page *); | |
115 | static void swsusp_unset_page_forbidden(struct page *); | |
116 | ||
ddeb6487 RW |
117 | /* |
118 | * Number of bytes to reserve for memory allocations made by device drivers | |
119 | * from their ->freeze() and ->freeze_noirq() callbacks so that they don't | |
120 | * cause image creation to fail (tunable via /sys/power/reserved_size). | |
121 | */ | |
122 | unsigned long reserved_size; | |
123 | ||
124 | void __init hibernate_reserved_size_init(void) | |
125 | { | |
126 | reserved_size = SPARE_PAGES * PAGE_SIZE; | |
127 | } | |
128 | ||
fe419535 RW |
129 | /* |
130 | * Preferred image size in bytes (tunable via /sys/power/image_size). | |
1c1be3a9 RW |
131 | * When it is set to N, swsusp will do its best to ensure the image |
132 | * size will not exceed N bytes, but if that is impossible, it will | |
133 | * try to create the smallest image possible. | |
fe419535 | 134 | */ |
ac5c24ec RW |
135 | unsigned long image_size; |
136 | ||
137 | void __init hibernate_image_size_init(void) | |
138 | { | |
ca79b0c2 | 139 | image_size = ((totalram_pages() * 2) / 5) * PAGE_SIZE; |
ac5c24ec | 140 | } |
fe419535 | 141 | |
ef96f639 RW |
142 | /* |
143 | * List of PBEs needed for restoring the pages that were allocated before | |
8357376d RW |
144 | * the suspend and included in the suspend image, but have also been |
145 | * allocated by the "resume" kernel, so their contents cannot be written | |
146 | * directly to their "original" page frames. | |
147 | */ | |
75534b50 RW |
148 | struct pbe *restore_pblist; |
149 | ||
9c744481 RW |
150 | /* struct linked_page is used to build chains of pages */ |
151 | ||
152 | #define LINKED_PAGE_DATA_SIZE (PAGE_SIZE - sizeof(void *)) | |
153 | ||
154 | struct linked_page { | |
155 | struct linked_page *next; | |
156 | char data[LINKED_PAGE_DATA_SIZE]; | |
157 | } __packed; | |
158 | ||
159 | /* | |
160 | * List of "safe" pages (ie. pages that were not used by the image kernel | |
161 | * before hibernation) that may be used as temporary storage for image kernel | |
162 | * memory contents. | |
163 | */ | |
164 | static struct linked_page *safe_pages_list; | |
165 | ||
8357376d | 166 | /* Pointer to an auxiliary buffer (1 page) */ |
940864dd | 167 | static void *buffer; |
7088a5c0 | 168 | |
0bcd888d RW |
169 | #define PG_ANY 0 |
170 | #define PG_SAFE 1 | |
171 | #define PG_UNSAFE_CLEAR 1 | |
172 | #define PG_UNSAFE_KEEP 0 | |
173 | ||
940864dd | 174 | static unsigned int allocated_unsafe_pages; |
f6143aa6 | 175 | |
ef96f639 RW |
176 | /** |
177 | * get_image_page - Allocate a page for a hibernation image. | |
178 | * @gfp_mask: GFP mask for the allocation. | |
179 | * @safe_needed: Get pages that were not used before hibernation (restore only) | |
180 | * | |
181 | * During image restoration, for storing the PBE list and the image data, we can | |
182 | * only use memory pages that do not conflict with the pages used before | |
183 | * hibernation. The "unsafe" pages have PageNosaveFree set and we count them | |
184 | * using allocated_unsafe_pages. | |
185 | * | |
186 | * Each allocated image page is marked as PageNosave and PageNosaveFree so that | |
187 | * swsusp_free() can release it. | |
188 | */ | |
8357376d | 189 | static void *get_image_page(gfp_t gfp_mask, int safe_needed) |
f6143aa6 RW |
190 | { |
191 | void *res; | |
192 | ||
193 | res = (void *)get_zeroed_page(gfp_mask); | |
194 | if (safe_needed) | |
7be98234 | 195 | while (res && swsusp_page_is_free(virt_to_page(res))) { |
f6143aa6 | 196 | /* The page is unsafe, mark it for swsusp_free() */ |
7be98234 | 197 | swsusp_set_page_forbidden(virt_to_page(res)); |
940864dd | 198 | allocated_unsafe_pages++; |
f6143aa6 RW |
199 | res = (void *)get_zeroed_page(gfp_mask); |
200 | } | |
201 | if (res) { | |
7be98234 RW |
202 | swsusp_set_page_forbidden(virt_to_page(res)); |
203 | swsusp_set_page_free(virt_to_page(res)); | |
f6143aa6 RW |
204 | } |
205 | return res; | |
206 | } | |
207 | ||
9c744481 RW |
208 | static void *__get_safe_page(gfp_t gfp_mask) |
209 | { | |
210 | if (safe_pages_list) { | |
211 | void *ret = safe_pages_list; | |
212 | ||
213 | safe_pages_list = safe_pages_list->next; | |
214 | memset(ret, 0, PAGE_SIZE); | |
215 | return ret; | |
216 | } | |
217 | return get_image_page(gfp_mask, PG_SAFE); | |
218 | } | |
219 | ||
f6143aa6 RW |
220 | unsigned long get_safe_page(gfp_t gfp_mask) |
221 | { | |
9c744481 | 222 | return (unsigned long)__get_safe_page(gfp_mask); |
8357376d RW |
223 | } |
224 | ||
5b6d15de RW |
225 | static struct page *alloc_image_page(gfp_t gfp_mask) |
226 | { | |
8357376d RW |
227 | struct page *page; |
228 | ||
229 | page = alloc_page(gfp_mask); | |
230 | if (page) { | |
7be98234 RW |
231 | swsusp_set_page_forbidden(page); |
232 | swsusp_set_page_free(page); | |
8357376d RW |
233 | } |
234 | return page; | |
f6143aa6 RW |
235 | } |
236 | ||
307c5971 RW |
237 | static void recycle_safe_page(void *page_address) |
238 | { | |
239 | struct linked_page *lp = page_address; | |
240 | ||
241 | lp->next = safe_pages_list; | |
242 | safe_pages_list = lp; | |
243 | } | |
244 | ||
f6143aa6 | 245 | /** |
ef96f639 RW |
246 | * free_image_page - Free a page allocated for hibernation image. |
247 | * @addr: Address of the page to free. | |
248 | * @clear_nosave_free: If set, clear the PageNosaveFree bit for the page. | |
249 | * | |
250 | * The page to free should have been allocated by get_image_page() (page flags | |
251 | * set by it are affected). | |
f6143aa6 | 252 | */ |
f6143aa6 RW |
253 | static inline void free_image_page(void *addr, int clear_nosave_free) |
254 | { | |
8357376d RW |
255 | struct page *page; |
256 | ||
257 | BUG_ON(!virt_addr_valid(addr)); | |
258 | ||
259 | page = virt_to_page(addr); | |
260 | ||
7be98234 | 261 | swsusp_unset_page_forbidden(page); |
f6143aa6 | 262 | if (clear_nosave_free) |
7be98234 | 263 | swsusp_unset_page_free(page); |
8357376d RW |
264 | |
265 | __free_page(page); | |
f6143aa6 RW |
266 | } |
267 | ||
efd5a852 RW |
268 | static inline void free_list_of_pages(struct linked_page *list, |
269 | int clear_page_nosave) | |
b788db79 RW |
270 | { |
271 | while (list) { | |
272 | struct linked_page *lp = list->next; | |
273 | ||
274 | free_image_page(list, clear_page_nosave); | |
275 | list = lp; | |
276 | } | |
277 | } | |
278 | ||
ef96f639 RW |
279 | /* |
280 | * struct chain_allocator is used for allocating small objects out of | |
281 | * a linked list of pages called 'the chain'. | |
282 | * | |
283 | * The chain grows each time when there is no room for a new object in | |
284 | * the current page. The allocated objects cannot be freed individually. | |
285 | * It is only possible to free them all at once, by freeing the entire | |
286 | * chain. | |
287 | * | |
288 | * NOTE: The chain allocator may be inefficient if the allocated objects | |
289 | * are not much smaller than PAGE_SIZE. | |
290 | */ | |
b788db79 RW |
291 | struct chain_allocator { |
292 | struct linked_page *chain; /* the chain */ | |
293 | unsigned int used_space; /* total size of objects allocated out | |
ef96f639 | 294 | of the current page */ |
b788db79 RW |
295 | gfp_t gfp_mask; /* mask for allocating pages */ |
296 | int safe_needed; /* if set, only "safe" pages are allocated */ | |
297 | }; | |
298 | ||
efd5a852 RW |
299 | static void chain_init(struct chain_allocator *ca, gfp_t gfp_mask, |
300 | int safe_needed) | |
b788db79 RW |
301 | { |
302 | ca->chain = NULL; | |
303 | ca->used_space = LINKED_PAGE_DATA_SIZE; | |
304 | ca->gfp_mask = gfp_mask; | |
305 | ca->safe_needed = safe_needed; | |
306 | } | |
307 | ||
308 | static void *chain_alloc(struct chain_allocator *ca, unsigned int size) | |
309 | { | |
310 | void *ret; | |
311 | ||
312 | if (LINKED_PAGE_DATA_SIZE - ca->used_space < size) { | |
313 | struct linked_page *lp; | |
314 | ||
9c744481 RW |
315 | lp = ca->safe_needed ? __get_safe_page(ca->gfp_mask) : |
316 | get_image_page(ca->gfp_mask, PG_ANY); | |
b788db79 RW |
317 | if (!lp) |
318 | return NULL; | |
319 | ||
320 | lp->next = ca->chain; | |
321 | ca->chain = lp; | |
322 | ca->used_space = 0; | |
323 | } | |
324 | ret = ca->chain->data + ca->used_space; | |
325 | ca->used_space += size; | |
326 | return ret; | |
327 | } | |
328 | ||
b788db79 | 329 | /** |
ef96f639 | 330 | * Data types related to memory bitmaps. |
b788db79 | 331 | * |
e4b2897a | 332 | * Memory bitmap is a structure consisting of many linked lists of |
ef96f639 | 333 | * objects. The main list's elements are of type struct zone_bitmap |
6be2408a | 334 | * and each of them corresponds to one zone. For each zone bitmap |
ef96f639 RW |
335 | * object there is a list of objects of type struct bm_block that |
336 | * represent each blocks of bitmap in which information is stored. | |
b788db79 | 337 | * |
ef96f639 RW |
338 | * struct memory_bitmap contains a pointer to the main list of zone |
339 | * bitmap objects, a struct bm_position used for browsing the bitmap, | |
340 | * and a pointer to the list of pages used for allocating all of the | |
341 | * zone bitmap objects and bitmap block objects. | |
b788db79 | 342 | * |
ef96f639 RW |
343 | * NOTE: It has to be possible to lay out the bitmap in memory |
344 | * using only allocations of order 0. Additionally, the bitmap is | |
345 | * designed to work with arbitrary number of zones (this is over the | |
346 | * top for now, but let's avoid making unnecessary assumptions ;-). | |
b788db79 | 347 | * |
ef96f639 RW |
348 | * struct zone_bitmap contains a pointer to a list of bitmap block |
349 | * objects and a pointer to the bitmap block object that has been | |
350 | * most recently used for setting bits. Additionally, it contains the | |
351 | * PFNs that correspond to the start and end of the represented zone. | |
b788db79 | 352 | * |
ef96f639 RW |
353 | * struct bm_block contains a pointer to the memory page in which |
354 | * information is stored (in the form of a block of bitmap) | |
355 | * It also contains the pfns that correspond to the start and end of | |
356 | * the represented memory area. | |
f469f02d | 357 | * |
ef96f639 RW |
358 | * The memory bitmap is organized as a radix tree to guarantee fast random |
359 | * access to the bits. There is one radix tree for each zone (as returned | |
360 | * from create_mem_extents). | |
f469f02d | 361 | * |
ef96f639 RW |
362 | * One radix tree is represented by one struct mem_zone_bm_rtree. There are |
363 | * two linked lists for the nodes of the tree, one for the inner nodes and | |
364 | * one for the leave nodes. The linked leave nodes are used for fast linear | |
365 | * access of the memory bitmap. | |
f469f02d | 366 | * |
ef96f639 | 367 | * The struct rtree_node represents one node of the radix tree. |
b788db79 RW |
368 | */ |
369 | ||
370 | #define BM_END_OF_MAP (~0UL) | |
371 | ||
8de03073 | 372 | #define BM_BITS_PER_BLOCK (PAGE_SIZE * BITS_PER_BYTE) |
f469f02d JR |
373 | #define BM_BLOCK_SHIFT (PAGE_SHIFT + 3) |
374 | #define BM_BLOCK_MASK ((1UL << BM_BLOCK_SHIFT) - 1) | |
b788db79 | 375 | |
f469f02d JR |
376 | /* |
377 | * struct rtree_node is a wrapper struct to link the nodes | |
378 | * of the rtree together for easy linear iteration over | |
379 | * bits and easy freeing | |
380 | */ | |
381 | struct rtree_node { | |
382 | struct list_head list; | |
383 | unsigned long *data; | |
384 | }; | |
385 | ||
386 | /* | |
387 | * struct mem_zone_bm_rtree represents a bitmap used for one | |
388 | * populated memory zone. | |
389 | */ | |
390 | struct mem_zone_bm_rtree { | |
391 | struct list_head list; /* Link Zones together */ | |
392 | struct list_head nodes; /* Radix Tree inner nodes */ | |
393 | struct list_head leaves; /* Radix Tree leaves */ | |
394 | unsigned long start_pfn; /* Zone start page frame */ | |
395 | unsigned long end_pfn; /* Zone end page frame + 1 */ | |
396 | struct rtree_node *rtree; /* Radix Tree Root */ | |
397 | int levels; /* Number of Radix Tree Levels */ | |
398 | unsigned int blocks; /* Number of Bitmap Blocks */ | |
399 | }; | |
400 | ||
b788db79 RW |
401 | /* strcut bm_position is used for browsing memory bitmaps */ |
402 | ||
403 | struct bm_position { | |
3a20cb17 JR |
404 | struct mem_zone_bm_rtree *zone; |
405 | struct rtree_node *node; | |
406 | unsigned long node_pfn; | |
407 | int node_bit; | |
b788db79 RW |
408 | }; |
409 | ||
410 | struct memory_bitmap { | |
f469f02d | 411 | struct list_head zones; |
b788db79 | 412 | struct linked_page *p_list; /* list of pages used to store zone |
ef96f639 RW |
413 | bitmap objects and bitmap block |
414 | objects */ | |
b788db79 RW |
415 | struct bm_position cur; /* most recently used bit position */ |
416 | }; | |
417 | ||
418 | /* Functions that operate on memory bitmaps */ | |
419 | ||
f469f02d JR |
420 | #define BM_ENTRIES_PER_LEVEL (PAGE_SIZE / sizeof(unsigned long)) |
421 | #if BITS_PER_LONG == 32 | |
422 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 2) | |
423 | #else | |
424 | #define BM_RTREE_LEVEL_SHIFT (PAGE_SHIFT - 3) | |
425 | #endif | |
426 | #define BM_RTREE_LEVEL_MASK ((1UL << BM_RTREE_LEVEL_SHIFT) - 1) | |
427 | ||
ef96f639 RW |
428 | /** |
429 | * alloc_rtree_node - Allocate a new node and add it to the radix tree. | |
f469f02d | 430 | * |
ef96f639 RW |
431 | * This function is used to allocate inner nodes as well as the |
432 | * leave nodes of the radix tree. It also adds the node to the | |
433 | * corresponding linked list passed in by the *list parameter. | |
f469f02d JR |
434 | */ |
435 | static struct rtree_node *alloc_rtree_node(gfp_t gfp_mask, int safe_needed, | |
436 | struct chain_allocator *ca, | |
437 | struct list_head *list) | |
438 | { | |
439 | struct rtree_node *node; | |
440 | ||
441 | node = chain_alloc(ca, sizeof(struct rtree_node)); | |
442 | if (!node) | |
443 | return NULL; | |
444 | ||
445 | node->data = get_image_page(gfp_mask, safe_needed); | |
446 | if (!node->data) | |
447 | return NULL; | |
448 | ||
449 | list_add_tail(&node->list, list); | |
450 | ||
451 | return node; | |
452 | } | |
453 | ||
ef96f639 RW |
454 | /** |
455 | * add_rtree_block - Add a new leave node to the radix tree. | |
f469f02d | 456 | * |
ef96f639 RW |
457 | * The leave nodes need to be allocated in order to keep the leaves |
458 | * linked list in order. This is guaranteed by the zone->blocks | |
459 | * counter. | |
f469f02d JR |
460 | */ |
461 | static int add_rtree_block(struct mem_zone_bm_rtree *zone, gfp_t gfp_mask, | |
462 | int safe_needed, struct chain_allocator *ca) | |
463 | { | |
464 | struct rtree_node *node, *block, **dst; | |
465 | unsigned int levels_needed, block_nr; | |
466 | int i; | |
467 | ||
468 | block_nr = zone->blocks; | |
469 | levels_needed = 0; | |
470 | ||
471 | /* How many levels do we need for this block nr? */ | |
472 | while (block_nr) { | |
473 | levels_needed += 1; | |
474 | block_nr >>= BM_RTREE_LEVEL_SHIFT; | |
475 | } | |
476 | ||
477 | /* Make sure the rtree has enough levels */ | |
478 | for (i = zone->levels; i < levels_needed; i++) { | |
479 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
480 | &zone->nodes); | |
481 | if (!node) | |
482 | return -ENOMEM; | |
483 | ||
484 | node->data[0] = (unsigned long)zone->rtree; | |
485 | zone->rtree = node; | |
486 | zone->levels += 1; | |
487 | } | |
488 | ||
489 | /* Allocate new block */ | |
490 | block = alloc_rtree_node(gfp_mask, safe_needed, ca, &zone->leaves); | |
491 | if (!block) | |
492 | return -ENOMEM; | |
493 | ||
494 | /* Now walk the rtree to insert the block */ | |
495 | node = zone->rtree; | |
496 | dst = &zone->rtree; | |
497 | block_nr = zone->blocks; | |
498 | for (i = zone->levels; i > 0; i--) { | |
499 | int index; | |
500 | ||
501 | if (!node) { | |
502 | node = alloc_rtree_node(gfp_mask, safe_needed, ca, | |
503 | &zone->nodes); | |
504 | if (!node) | |
505 | return -ENOMEM; | |
506 | *dst = node; | |
507 | } | |
508 | ||
509 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
510 | index &= BM_RTREE_LEVEL_MASK; | |
511 | dst = (struct rtree_node **)&((*dst)->data[index]); | |
512 | node = *dst; | |
513 | } | |
514 | ||
515 | zone->blocks += 1; | |
516 | *dst = block; | |
517 | ||
518 | return 0; | |
519 | } | |
520 | ||
521 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
522 | int clear_nosave_free); | |
523 | ||
ef96f639 RW |
524 | /** |
525 | * create_zone_bm_rtree - Create a radix tree for one zone. | |
f469f02d | 526 | * |
ef96f639 RW |
527 | * Allocated the mem_zone_bm_rtree structure and initializes it. |
528 | * This function also allocated and builds the radix tree for the | |
529 | * zone. | |
f469f02d | 530 | */ |
efd5a852 RW |
531 | static struct mem_zone_bm_rtree *create_zone_bm_rtree(gfp_t gfp_mask, |
532 | int safe_needed, | |
533 | struct chain_allocator *ca, | |
534 | unsigned long start, | |
535 | unsigned long end) | |
f469f02d JR |
536 | { |
537 | struct mem_zone_bm_rtree *zone; | |
538 | unsigned int i, nr_blocks; | |
539 | unsigned long pages; | |
540 | ||
541 | pages = end - start; | |
542 | zone = chain_alloc(ca, sizeof(struct mem_zone_bm_rtree)); | |
543 | if (!zone) | |
544 | return NULL; | |
545 | ||
546 | INIT_LIST_HEAD(&zone->nodes); | |
547 | INIT_LIST_HEAD(&zone->leaves); | |
548 | zone->start_pfn = start; | |
549 | zone->end_pfn = end; | |
550 | nr_blocks = DIV_ROUND_UP(pages, BM_BITS_PER_BLOCK); | |
551 | ||
552 | for (i = 0; i < nr_blocks; i++) { | |
553 | if (add_rtree_block(zone, gfp_mask, safe_needed, ca)) { | |
554 | free_zone_bm_rtree(zone, PG_UNSAFE_CLEAR); | |
555 | return NULL; | |
556 | } | |
557 | } | |
558 | ||
559 | return zone; | |
560 | } | |
561 | ||
ef96f639 RW |
562 | /** |
563 | * free_zone_bm_rtree - Free the memory of the radix tree. | |
f469f02d | 564 | * |
ef96f639 RW |
565 | * Free all node pages of the radix tree. The mem_zone_bm_rtree |
566 | * structure itself is not freed here nor are the rtree_node | |
567 | * structs. | |
f469f02d JR |
568 | */ |
569 | static void free_zone_bm_rtree(struct mem_zone_bm_rtree *zone, | |
570 | int clear_nosave_free) | |
571 | { | |
572 | struct rtree_node *node; | |
573 | ||
574 | list_for_each_entry(node, &zone->nodes, list) | |
575 | free_image_page(node->data, clear_nosave_free); | |
576 | ||
577 | list_for_each_entry(node, &zone->leaves, list) | |
578 | free_image_page(node->data, clear_nosave_free); | |
579 | } | |
580 | ||
b788db79 RW |
581 | static void memory_bm_position_reset(struct memory_bitmap *bm) |
582 | { | |
3a20cb17 JR |
583 | bm->cur.zone = list_entry(bm->zones.next, struct mem_zone_bm_rtree, |
584 | list); | |
585 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, | |
586 | struct rtree_node, list); | |
587 | bm->cur.node_pfn = 0; | |
588 | bm->cur.node_bit = 0; | |
b788db79 RW |
589 | } |
590 | ||
591 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); | |
592 | ||
846705de RW |
593 | struct mem_extent { |
594 | struct list_head hook; | |
595 | unsigned long start; | |
596 | unsigned long end; | |
597 | }; | |
598 | ||
b788db79 | 599 | /** |
ef96f639 RW |
600 | * free_mem_extents - Free a list of memory extents. |
601 | * @list: List of extents to free. | |
b788db79 | 602 | */ |
846705de RW |
603 | static void free_mem_extents(struct list_head *list) |
604 | { | |
605 | struct mem_extent *ext, *aux; | |
b788db79 | 606 | |
846705de RW |
607 | list_for_each_entry_safe(ext, aux, list, hook) { |
608 | list_del(&ext->hook); | |
609 | kfree(ext); | |
610 | } | |
611 | } | |
612 | ||
613 | /** | |
ef96f639 RW |
614 | * create_mem_extents - Create a list of memory extents. |
615 | * @list: List to put the extents into. | |
616 | * @gfp_mask: Mask to use for memory allocations. | |
617 | * | |
618 | * The extents represent contiguous ranges of PFNs. | |
846705de RW |
619 | */ |
620 | static int create_mem_extents(struct list_head *list, gfp_t gfp_mask) | |
b788db79 | 621 | { |
846705de | 622 | struct zone *zone; |
b788db79 | 623 | |
846705de | 624 | INIT_LIST_HEAD(list); |
b788db79 | 625 | |
ee99c71c | 626 | for_each_populated_zone(zone) { |
846705de RW |
627 | unsigned long zone_start, zone_end; |
628 | struct mem_extent *ext, *cur, *aux; | |
629 | ||
846705de | 630 | zone_start = zone->zone_start_pfn; |
c33bc315 | 631 | zone_end = zone_end_pfn(zone); |
846705de RW |
632 | |
633 | list_for_each_entry(ext, list, hook) | |
634 | if (zone_start <= ext->end) | |
635 | break; | |
b788db79 | 636 | |
846705de RW |
637 | if (&ext->hook == list || zone_end < ext->start) { |
638 | /* New extent is necessary */ | |
639 | struct mem_extent *new_ext; | |
640 | ||
641 | new_ext = kzalloc(sizeof(struct mem_extent), gfp_mask); | |
642 | if (!new_ext) { | |
643 | free_mem_extents(list); | |
644 | return -ENOMEM; | |
645 | } | |
646 | new_ext->start = zone_start; | |
647 | new_ext->end = zone_end; | |
648 | list_add_tail(&new_ext->hook, &ext->hook); | |
649 | continue; | |
650 | } | |
651 | ||
652 | /* Merge this zone's range of PFNs with the existing one */ | |
653 | if (zone_start < ext->start) | |
654 | ext->start = zone_start; | |
655 | if (zone_end > ext->end) | |
656 | ext->end = zone_end; | |
657 | ||
658 | /* More merging may be possible */ | |
659 | cur = ext; | |
660 | list_for_each_entry_safe_continue(cur, aux, list, hook) { | |
661 | if (zone_end < cur->start) | |
662 | break; | |
663 | if (zone_end < cur->end) | |
664 | ext->end = cur->end; | |
665 | list_del(&cur->hook); | |
666 | kfree(cur); | |
667 | } | |
b788db79 | 668 | } |
846705de RW |
669 | |
670 | return 0; | |
b788db79 RW |
671 | } |
672 | ||
673 | /** | |
ef96f639 RW |
674 | * memory_bm_create - Allocate memory for a memory bitmap. |
675 | */ | |
efd5a852 RW |
676 | static int memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, |
677 | int safe_needed) | |
b788db79 RW |
678 | { |
679 | struct chain_allocator ca; | |
846705de RW |
680 | struct list_head mem_extents; |
681 | struct mem_extent *ext; | |
682 | int error; | |
b788db79 RW |
683 | |
684 | chain_init(&ca, gfp_mask, safe_needed); | |
f469f02d | 685 | INIT_LIST_HEAD(&bm->zones); |
b788db79 | 686 | |
846705de RW |
687 | error = create_mem_extents(&mem_extents, gfp_mask); |
688 | if (error) | |
689 | return error; | |
b788db79 | 690 | |
846705de | 691 | list_for_each_entry(ext, &mem_extents, hook) { |
f469f02d | 692 | struct mem_zone_bm_rtree *zone; |
f469f02d JR |
693 | |
694 | zone = create_zone_bm_rtree(gfp_mask, safe_needed, &ca, | |
695 | ext->start, ext->end); | |
9047eb62 JR |
696 | if (!zone) { |
697 | error = -ENOMEM; | |
f469f02d | 698 | goto Error; |
9047eb62 | 699 | } |
f469f02d | 700 | list_add_tail(&zone->list, &bm->zones); |
b788db79 | 701 | } |
846705de | 702 | |
b788db79 RW |
703 | bm->p_list = ca.chain; |
704 | memory_bm_position_reset(bm); | |
846705de RW |
705 | Exit: |
706 | free_mem_extents(&mem_extents); | |
707 | return error; | |
b788db79 | 708 | |
846705de | 709 | Error: |
b788db79 RW |
710 | bm->p_list = ca.chain; |
711 | memory_bm_free(bm, PG_UNSAFE_CLEAR); | |
846705de | 712 | goto Exit; |
b788db79 RW |
713 | } |
714 | ||
715 | /** | |
ef96f639 RW |
716 | * memory_bm_free - Free memory occupied by the memory bitmap. |
717 | * @bm: Memory bitmap. | |
718 | */ | |
b788db79 RW |
719 | static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free) |
720 | { | |
f469f02d | 721 | struct mem_zone_bm_rtree *zone; |
b788db79 | 722 | |
f469f02d JR |
723 | list_for_each_entry(zone, &bm->zones, list) |
724 | free_zone_bm_rtree(zone, clear_nosave_free); | |
725 | ||
b788db79 | 726 | free_list_of_pages(bm->p_list, clear_nosave_free); |
846705de | 727 | |
f469f02d | 728 | INIT_LIST_HEAD(&bm->zones); |
b788db79 RW |
729 | } |
730 | ||
731 | /** | |
ef96f639 | 732 | * memory_bm_find_bit - Find the bit for a given PFN in a memory bitmap. |
07a33823 | 733 | * |
ef96f639 RW |
734 | * Find the bit in memory bitmap @bm that corresponds to the given PFN. |
735 | * The cur.zone, cur.block and cur.node_pfn members of @bm are updated. | |
736 | * | |
737 | * Walk the radix tree to find the page containing the bit that represents @pfn | |
738 | * and return the position of the bit in @addr and @bit_nr. | |
07a33823 | 739 | */ |
9047eb62 JR |
740 | static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn, |
741 | void **addr, unsigned int *bit_nr) | |
07a33823 JR |
742 | { |
743 | struct mem_zone_bm_rtree *curr, *zone; | |
744 | struct rtree_node *node; | |
745 | int i, block_nr; | |
746 | ||
3a20cb17 JR |
747 | zone = bm->cur.zone; |
748 | ||
749 | if (pfn >= zone->start_pfn && pfn < zone->end_pfn) | |
750 | goto zone_found; | |
751 | ||
07a33823 JR |
752 | zone = NULL; |
753 | ||
754 | /* Find the right zone */ | |
755 | list_for_each_entry(curr, &bm->zones, list) { | |
756 | if (pfn >= curr->start_pfn && pfn < curr->end_pfn) { | |
757 | zone = curr; | |
758 | break; | |
759 | } | |
760 | } | |
761 | ||
762 | if (!zone) | |
763 | return -EFAULT; | |
764 | ||
3a20cb17 | 765 | zone_found: |
07a33823 | 766 | /* |
ef96f639 RW |
767 | * We have found the zone. Now walk the radix tree to find the leaf node |
768 | * for our PFN. | |
07a33823 | 769 | */ |
da6043fe AW |
770 | |
771 | /* | |
7b7b8a2c | 772 | * If the zone we wish to scan is the current zone and the |
da6043fe AW |
773 | * pfn falls into the current node then we do not need to walk |
774 | * the tree. | |
775 | */ | |
3a20cb17 | 776 | node = bm->cur.node; |
da6043fe AW |
777 | if (zone == bm->cur.zone && |
778 | ((pfn - zone->start_pfn) & ~BM_BLOCK_MASK) == bm->cur.node_pfn) | |
3a20cb17 JR |
779 | goto node_found; |
780 | ||
07a33823 JR |
781 | node = zone->rtree; |
782 | block_nr = (pfn - zone->start_pfn) >> BM_BLOCK_SHIFT; | |
783 | ||
784 | for (i = zone->levels; i > 0; i--) { | |
785 | int index; | |
786 | ||
787 | index = block_nr >> ((i - 1) * BM_RTREE_LEVEL_SHIFT); | |
788 | index &= BM_RTREE_LEVEL_MASK; | |
789 | BUG_ON(node->data[index] == 0); | |
790 | node = (struct rtree_node *)node->data[index]; | |
791 | } | |
792 | ||
3a20cb17 JR |
793 | node_found: |
794 | /* Update last position */ | |
795 | bm->cur.zone = zone; | |
796 | bm->cur.node = node; | |
797 | bm->cur.node_pfn = (pfn - zone->start_pfn) & ~BM_BLOCK_MASK; | |
798 | ||
07a33823 JR |
799 | /* Set return values */ |
800 | *addr = node->data; | |
801 | *bit_nr = (pfn - zone->start_pfn) & BM_BLOCK_MASK; | |
802 | ||
803 | return 0; | |
804 | } | |
805 | ||
74dfd666 RW |
806 | static void memory_bm_set_bit(struct memory_bitmap *bm, unsigned long pfn) |
807 | { | |
808 | void *addr; | |
809 | unsigned int bit; | |
a82f7119 | 810 | int error; |
74dfd666 | 811 | |
a82f7119 RW |
812 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
813 | BUG_ON(error); | |
74dfd666 RW |
814 | set_bit(bit, addr); |
815 | } | |
816 | ||
a82f7119 RW |
817 | static int mem_bm_set_bit_check(struct memory_bitmap *bm, unsigned long pfn) |
818 | { | |
819 | void *addr; | |
820 | unsigned int bit; | |
821 | int error; | |
822 | ||
823 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); | |
07a33823 JR |
824 | if (!error) |
825 | set_bit(bit, addr); | |
826 | ||
a82f7119 RW |
827 | return error; |
828 | } | |
829 | ||
74dfd666 RW |
830 | static void memory_bm_clear_bit(struct memory_bitmap *bm, unsigned long pfn) |
831 | { | |
832 | void *addr; | |
833 | unsigned int bit; | |
a82f7119 | 834 | int error; |
74dfd666 | 835 | |
a82f7119 RW |
836 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
837 | BUG_ON(error); | |
74dfd666 RW |
838 | clear_bit(bit, addr); |
839 | } | |
840 | ||
fdd64ed5 JR |
841 | static void memory_bm_clear_current(struct memory_bitmap *bm) |
842 | { | |
843 | int bit; | |
844 | ||
845 | bit = max(bm->cur.node_bit - 1, 0); | |
846 | clear_bit(bit, bm->cur.node->data); | |
847 | } | |
848 | ||
74dfd666 RW |
849 | static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn) |
850 | { | |
851 | void *addr; | |
852 | unsigned int bit; | |
9047eb62 | 853 | int error; |
74dfd666 | 854 | |
a82f7119 RW |
855 | error = memory_bm_find_bit(bm, pfn, &addr, &bit); |
856 | BUG_ON(error); | |
9047eb62 | 857 | return test_bit(bit, addr); |
b788db79 RW |
858 | } |
859 | ||
69643279 RW |
860 | static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) |
861 | { | |
862 | void *addr; | |
863 | unsigned int bit; | |
07a33823 | 864 | |
9047eb62 | 865 | return !memory_bm_find_bit(bm, pfn, &addr, &bit); |
b788db79 RW |
866 | } |
867 | ||
3a20cb17 | 868 | /* |
ef96f639 | 869 | * rtree_next_node - Jump to the next leaf node. |
3a20cb17 | 870 | * |
ef96f639 RW |
871 | * Set the position to the beginning of the next node in the |
872 | * memory bitmap. This is either the next node in the current | |
873 | * zone's radix tree or the first node in the radix tree of the | |
874 | * next zone. | |
3a20cb17 | 875 | * |
ef96f639 | 876 | * Return true if there is a next node, false otherwise. |
3a20cb17 JR |
877 | */ |
878 | static bool rtree_next_node(struct memory_bitmap *bm) | |
879 | { | |
924d8696 JM |
880 | if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) { |
881 | bm->cur.node = list_entry(bm->cur.node->list.next, | |
882 | struct rtree_node, list); | |
3a20cb17 JR |
883 | bm->cur.node_pfn += BM_BITS_PER_BLOCK; |
884 | bm->cur.node_bit = 0; | |
0f7d83e8 | 885 | touch_softlockup_watchdog(); |
3a20cb17 JR |
886 | return true; |
887 | } | |
888 | ||
889 | /* No more nodes, goto next zone */ | |
924d8696 JM |
890 | if (!list_is_last(&bm->cur.zone->list, &bm->zones)) { |
891 | bm->cur.zone = list_entry(bm->cur.zone->list.next, | |
3a20cb17 | 892 | struct mem_zone_bm_rtree, list); |
3a20cb17 JR |
893 | bm->cur.node = list_entry(bm->cur.zone->leaves.next, |
894 | struct rtree_node, list); | |
895 | bm->cur.node_pfn = 0; | |
896 | bm->cur.node_bit = 0; | |
897 | return true; | |
898 | } | |
899 | ||
900 | /* No more zones */ | |
901 | return false; | |
902 | } | |
903 | ||
9047eb62 | 904 | /** |
ef96f639 RW |
905 | * memory_bm_rtree_next_pfn - Find the next set bit in a memory bitmap. |
906 | * @bm: Memory bitmap. | |
3a20cb17 | 907 | * |
ef96f639 RW |
908 | * Starting from the last returned position this function searches for the next |
909 | * set bit in @bm and returns the PFN represented by it. If no more bits are | |
910 | * set, BM_END_OF_MAP is returned. | |
9047eb62 | 911 | * |
ef96f639 RW |
912 | * It is required to run memory_bm_position_reset() before the first call to |
913 | * this function for the given memory bitmap. | |
3a20cb17 | 914 | */ |
9047eb62 | 915 | static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm) |
3a20cb17 JR |
916 | { |
917 | unsigned long bits, pfn, pages; | |
918 | int bit; | |
919 | ||
920 | do { | |
921 | pages = bm->cur.zone->end_pfn - bm->cur.zone->start_pfn; | |
922 | bits = min(pages - bm->cur.node_pfn, BM_BITS_PER_BLOCK); | |
923 | bit = find_next_bit(bm->cur.node->data, bits, | |
924 | bm->cur.node_bit); | |
925 | if (bit < bits) { | |
926 | pfn = bm->cur.zone->start_pfn + bm->cur.node_pfn + bit; | |
927 | bm->cur.node_bit = bit + 1; | |
928 | return pfn; | |
929 | } | |
930 | } while (rtree_next_node(bm)); | |
931 | ||
932 | return BM_END_OF_MAP; | |
933 | } | |
934 | ||
ef96f639 RW |
935 | /* |
936 | * This structure represents a range of page frames the contents of which | |
937 | * should not be saved during hibernation. | |
74dfd666 | 938 | */ |
74dfd666 RW |
939 | struct nosave_region { |
940 | struct list_head list; | |
941 | unsigned long start_pfn; | |
942 | unsigned long end_pfn; | |
943 | }; | |
944 | ||
945 | static LIST_HEAD(nosave_regions); | |
946 | ||
307c5971 RW |
947 | static void recycle_zone_bm_rtree(struct mem_zone_bm_rtree *zone) |
948 | { | |
949 | struct rtree_node *node; | |
950 | ||
951 | list_for_each_entry(node, &zone->nodes, list) | |
952 | recycle_safe_page(node->data); | |
953 | ||
954 | list_for_each_entry(node, &zone->leaves, list) | |
955 | recycle_safe_page(node->data); | |
956 | } | |
957 | ||
958 | static void memory_bm_recycle(struct memory_bitmap *bm) | |
959 | { | |
960 | struct mem_zone_bm_rtree *zone; | |
961 | struct linked_page *p_list; | |
962 | ||
963 | list_for_each_entry(zone, &bm->zones, list) | |
964 | recycle_zone_bm_rtree(zone); | |
965 | ||
966 | p_list = bm->p_list; | |
967 | while (p_list) { | |
968 | struct linked_page *lp = p_list; | |
969 | ||
970 | p_list = lp->next; | |
971 | recycle_safe_page(lp); | |
972 | } | |
973 | } | |
974 | ||
74dfd666 | 975 | /** |
ef96f639 RW |
976 | * register_nosave_region - Register a region of unsaveable memory. |
977 | * | |
978 | * Register a range of page frames the contents of which should not be saved | |
979 | * during hibernation (to be used in the early initialization code). | |
74dfd666 | 980 | */ |
33569ef3 | 981 | void __init register_nosave_region(unsigned long start_pfn, unsigned long end_pfn) |
74dfd666 RW |
982 | { |
983 | struct nosave_region *region; | |
984 | ||
985 | if (start_pfn >= end_pfn) | |
986 | return; | |
987 | ||
988 | if (!list_empty(&nosave_regions)) { | |
989 | /* Try to extend the previous region (they should be sorted) */ | |
990 | region = list_entry(nosave_regions.prev, | |
991 | struct nosave_region, list); | |
992 | if (region->end_pfn == start_pfn) { | |
993 | region->end_pfn = end_pfn; | |
994 | goto Report; | |
995 | } | |
996 | } | |
33569ef3 AS |
997 | /* This allocation cannot fail */ |
998 | region = memblock_alloc(sizeof(struct nosave_region), | |
999 | SMP_CACHE_BYTES); | |
1000 | if (!region) | |
1001 | panic("%s: Failed to allocate %zu bytes\n", __func__, | |
1002 | sizeof(struct nosave_region)); | |
74dfd666 RW |
1003 | region->start_pfn = start_pfn; |
1004 | region->end_pfn = end_pfn; | |
1005 | list_add_tail(®ion->list, &nosave_regions); | |
1006 | Report: | |
64ec72a1 | 1007 | pr_info("Registered nosave memory: [mem %#010llx-%#010llx]\n", |
cd38ca85 BH |
1008 | (unsigned long long) start_pfn << PAGE_SHIFT, |
1009 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
74dfd666 RW |
1010 | } |
1011 | ||
1012 | /* | |
1013 | * Set bits in this map correspond to the page frames the contents of which | |
1014 | * should not be saved during the suspend. | |
1015 | */ | |
1016 | static struct memory_bitmap *forbidden_pages_map; | |
1017 | ||
1018 | /* Set bits in this map correspond to free page frames. */ | |
1019 | static struct memory_bitmap *free_pages_map; | |
1020 | ||
1021 | /* | |
1022 | * Each page frame allocated for creating the image is marked by setting the | |
1023 | * corresponding bits in forbidden_pages_map and free_pages_map simultaneously | |
1024 | */ | |
1025 | ||
1026 | void swsusp_set_page_free(struct page *page) | |
1027 | { | |
1028 | if (free_pages_map) | |
1029 | memory_bm_set_bit(free_pages_map, page_to_pfn(page)); | |
1030 | } | |
1031 | ||
1032 | static int swsusp_page_is_free(struct page *page) | |
1033 | { | |
1034 | return free_pages_map ? | |
1035 | memory_bm_test_bit(free_pages_map, page_to_pfn(page)) : 0; | |
1036 | } | |
1037 | ||
1038 | void swsusp_unset_page_free(struct page *page) | |
1039 | { | |
1040 | if (free_pages_map) | |
1041 | memory_bm_clear_bit(free_pages_map, page_to_pfn(page)); | |
1042 | } | |
1043 | ||
1044 | static void swsusp_set_page_forbidden(struct page *page) | |
1045 | { | |
1046 | if (forbidden_pages_map) | |
1047 | memory_bm_set_bit(forbidden_pages_map, page_to_pfn(page)); | |
1048 | } | |
1049 | ||
1050 | int swsusp_page_is_forbidden(struct page *page) | |
1051 | { | |
1052 | return forbidden_pages_map ? | |
1053 | memory_bm_test_bit(forbidden_pages_map, page_to_pfn(page)) : 0; | |
1054 | } | |
1055 | ||
1056 | static void swsusp_unset_page_forbidden(struct page *page) | |
1057 | { | |
1058 | if (forbidden_pages_map) | |
1059 | memory_bm_clear_bit(forbidden_pages_map, page_to_pfn(page)); | |
1060 | } | |
1061 | ||
1062 | /** | |
ef96f639 RW |
1063 | * mark_nosave_pages - Mark pages that should not be saved. |
1064 | * @bm: Memory bitmap. | |
1065 | * | |
1066 | * Set the bits in @bm that correspond to the page frames the contents of which | |
1067 | * should not be saved. | |
74dfd666 | 1068 | */ |
74dfd666 RW |
1069 | static void mark_nosave_pages(struct memory_bitmap *bm) |
1070 | { | |
1071 | struct nosave_region *region; | |
1072 | ||
1073 | if (list_empty(&nosave_regions)) | |
1074 | return; | |
1075 | ||
1076 | list_for_each_entry(region, &nosave_regions, list) { | |
1077 | unsigned long pfn; | |
1078 | ||
64ec72a1 | 1079 | pr_debug("Marking nosave pages: [mem %#010llx-%#010llx]\n", |
69f1d475 BH |
1080 | (unsigned long long) region->start_pfn << PAGE_SHIFT, |
1081 | ((unsigned long long) region->end_pfn << PAGE_SHIFT) | |
1082 | - 1); | |
74dfd666 RW |
1083 | |
1084 | for (pfn = region->start_pfn; pfn < region->end_pfn; pfn++) | |
a82f7119 RW |
1085 | if (pfn_valid(pfn)) { |
1086 | /* | |
1087 | * It is safe to ignore the result of | |
1088 | * mem_bm_set_bit_check() here, since we won't | |
1089 | * touch the PFNs for which the error is | |
1090 | * returned anyway. | |
1091 | */ | |
1092 | mem_bm_set_bit_check(bm, pfn); | |
1093 | } | |
74dfd666 RW |
1094 | } |
1095 | } | |
1096 | ||
1097 | /** | |
ef96f639 RW |
1098 | * create_basic_memory_bitmaps - Create bitmaps to hold basic page information. |
1099 | * | |
1100 | * Create bitmaps needed for marking page frames that should not be saved and | |
1101 | * free page frames. The forbidden_pages_map and free_pages_map pointers are | |
1102 | * only modified if everything goes well, because we don't want the bits to be | |
1103 | * touched before both bitmaps are set up. | |
74dfd666 | 1104 | */ |
74dfd666 RW |
1105 | int create_basic_memory_bitmaps(void) |
1106 | { | |
1107 | struct memory_bitmap *bm1, *bm2; | |
1108 | int error = 0; | |
1109 | ||
aab17289 RW |
1110 | if (forbidden_pages_map && free_pages_map) |
1111 | return 0; | |
1112 | else | |
1113 | BUG_ON(forbidden_pages_map || free_pages_map); | |
74dfd666 | 1114 | |
0709db60 | 1115 | bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1116 | if (!bm1) |
1117 | return -ENOMEM; | |
1118 | ||
0709db60 | 1119 | error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1120 | if (error) |
1121 | goto Free_first_object; | |
1122 | ||
0709db60 | 1123 | bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL); |
74dfd666 RW |
1124 | if (!bm2) |
1125 | goto Free_first_bitmap; | |
1126 | ||
0709db60 | 1127 | error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY); |
74dfd666 RW |
1128 | if (error) |
1129 | goto Free_second_object; | |
1130 | ||
1131 | forbidden_pages_map = bm1; | |
1132 | free_pages_map = bm2; | |
1133 | mark_nosave_pages(forbidden_pages_map); | |
1134 | ||
64ec72a1 | 1135 | pr_debug("Basic memory bitmaps created\n"); |
74dfd666 RW |
1136 | |
1137 | return 0; | |
1138 | ||
1139 | Free_second_object: | |
1140 | kfree(bm2); | |
1141 | Free_first_bitmap: | |
480f0de6 | 1142 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); |
74dfd666 RW |
1143 | Free_first_object: |
1144 | kfree(bm1); | |
1145 | return -ENOMEM; | |
1146 | } | |
1147 | ||
1148 | /** | |
ef96f639 RW |
1149 | * free_basic_memory_bitmaps - Free memory bitmaps holding basic information. |
1150 | * | |
1151 | * Free memory bitmaps allocated by create_basic_memory_bitmaps(). The | |
1152 | * auxiliary pointers are necessary so that the bitmaps themselves are not | |
1153 | * referred to while they are being freed. | |
74dfd666 | 1154 | */ |
74dfd666 RW |
1155 | void free_basic_memory_bitmaps(void) |
1156 | { | |
1157 | struct memory_bitmap *bm1, *bm2; | |
1158 | ||
6a0c7cd3 RW |
1159 | if (WARN_ON(!(forbidden_pages_map && free_pages_map))) |
1160 | return; | |
74dfd666 RW |
1161 | |
1162 | bm1 = forbidden_pages_map; | |
1163 | bm2 = free_pages_map; | |
1164 | forbidden_pages_map = NULL; | |
1165 | free_pages_map = NULL; | |
1166 | memory_bm_free(bm1, PG_UNSAFE_CLEAR); | |
1167 | kfree(bm1); | |
1168 | memory_bm_free(bm2, PG_UNSAFE_CLEAR); | |
1169 | kfree(bm2); | |
1170 | ||
64ec72a1 | 1171 | pr_debug("Basic memory bitmaps freed\n"); |
74dfd666 RW |
1172 | } |
1173 | ||
03b6c9a3 VB |
1174 | static void clear_or_poison_free_page(struct page *page) |
1175 | { | |
1176 | if (page_poisoning_enabled_static()) | |
1177 | __kernel_poison_pages(page, 1); | |
1178 | else if (want_init_on_free()) | |
1179 | clear_highpage(page); | |
1180 | } | |
1181 | ||
1182 | void clear_or_poison_free_pages(void) | |
1ad1410f | 1183 | { |
1ad1410f AA |
1184 | struct memory_bitmap *bm = free_pages_map; |
1185 | unsigned long pfn; | |
1186 | ||
1187 | if (WARN_ON(!(free_pages_map))) | |
1188 | return; | |
1189 | ||
03b6c9a3 | 1190 | if (page_poisoning_enabled() || want_init_on_free()) { |
18451f9f | 1191 | memory_bm_position_reset(bm); |
1ad1410f | 1192 | pfn = memory_bm_next_pfn(bm); |
18451f9f AP |
1193 | while (pfn != BM_END_OF_MAP) { |
1194 | if (pfn_valid(pfn)) | |
03b6c9a3 | 1195 | clear_or_poison_free_page(pfn_to_page(pfn)); |
18451f9f AP |
1196 | |
1197 | pfn = memory_bm_next_pfn(bm); | |
1198 | } | |
1199 | memory_bm_position_reset(bm); | |
1200 | pr_info("free pages cleared after restore\n"); | |
1ad1410f | 1201 | } |
1ad1410f AA |
1202 | } |
1203 | ||
b788db79 | 1204 | /** |
ef96f639 RW |
1205 | * snapshot_additional_pages - Estimate the number of extra pages needed. |
1206 | * @zone: Memory zone to carry out the computation for. | |
1207 | * | |
1208 | * Estimate the number of additional pages needed for setting up a hibernation | |
1209 | * image data structures for @zone (usually, the returned value is greater than | |
1210 | * the exact number). | |
b788db79 | 1211 | */ |
b788db79 RW |
1212 | unsigned int snapshot_additional_pages(struct zone *zone) |
1213 | { | |
f469f02d | 1214 | unsigned int rtree, nodes; |
b788db79 | 1215 | |
f469f02d JR |
1216 | rtree = nodes = DIV_ROUND_UP(zone->spanned_pages, BM_BITS_PER_BLOCK); |
1217 | rtree += DIV_ROUND_UP(rtree * sizeof(struct rtree_node), | |
1218 | LINKED_PAGE_DATA_SIZE); | |
1219 | while (nodes > 1) { | |
1220 | nodes = DIV_ROUND_UP(nodes, BM_ENTRIES_PER_LEVEL); | |
1221 | rtree += nodes; | |
1222 | } | |
1223 | ||
9047eb62 | 1224 | return 2 * rtree; |
b788db79 RW |
1225 | } |
1226 | ||
8357376d RW |
1227 | #ifdef CONFIG_HIGHMEM |
1228 | /** | |
ef96f639 RW |
1229 | * count_free_highmem_pages - Compute the total number of free highmem pages. |
1230 | * | |
1231 | * The returned number is system-wide. | |
8357376d | 1232 | */ |
8357376d RW |
1233 | static unsigned int count_free_highmem_pages(void) |
1234 | { | |
1235 | struct zone *zone; | |
1236 | unsigned int cnt = 0; | |
1237 | ||
ee99c71c KM |
1238 | for_each_populated_zone(zone) |
1239 | if (is_highmem(zone)) | |
d23ad423 | 1240 | cnt += zone_page_state(zone, NR_FREE_PAGES); |
8357376d RW |
1241 | |
1242 | return cnt; | |
1243 | } | |
1244 | ||
1245 | /** | |
ef96f639 RW |
1246 | * saveable_highmem_page - Check if a highmem page is saveable. |
1247 | * | |
1248 | * Determine whether a highmem page should be included in a hibernation image. | |
8357376d | 1249 | * |
ef96f639 RW |
1250 | * We should save the page if it isn't Nosave or NosaveFree, or Reserved, |
1251 | * and it isn't part of a free chunk of pages. | |
8357376d | 1252 | */ |
846705de | 1253 | static struct page *saveable_highmem_page(struct zone *zone, unsigned long pfn) |
8357376d RW |
1254 | { |
1255 | struct page *page; | |
1256 | ||
1257 | if (!pfn_valid(pfn)) | |
1258 | return NULL; | |
1259 | ||
5b56db37 DH |
1260 | page = pfn_to_online_page(pfn); |
1261 | if (!page || page_zone(page) != zone) | |
846705de | 1262 | return NULL; |
8357376d RW |
1263 | |
1264 | BUG_ON(!PageHighMem(page)); | |
1265 | ||
abd02ac6 DH |
1266 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
1267 | return NULL; | |
1268 | ||
1269 | if (PageReserved(page) || PageOffline(page)) | |
8357376d RW |
1270 | return NULL; |
1271 | ||
c6968e73 SG |
1272 | if (page_is_guard(page)) |
1273 | return NULL; | |
1274 | ||
8357376d RW |
1275 | return page; |
1276 | } | |
1277 | ||
1278 | /** | |
ef96f639 | 1279 | * count_highmem_pages - Compute the total number of saveable highmem pages. |
8357376d | 1280 | */ |
fe419535 | 1281 | static unsigned int count_highmem_pages(void) |
8357376d RW |
1282 | { |
1283 | struct zone *zone; | |
1284 | unsigned int n = 0; | |
1285 | ||
98e73dc5 | 1286 | for_each_populated_zone(zone) { |
8357376d RW |
1287 | unsigned long pfn, max_zone_pfn; |
1288 | ||
1289 | if (!is_highmem(zone)) | |
1290 | continue; | |
1291 | ||
1292 | mark_free_pages(zone); | |
c33bc315 | 1293 | max_zone_pfn = zone_end_pfn(zone); |
8357376d | 1294 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1295 | if (saveable_highmem_page(zone, pfn)) |
8357376d RW |
1296 | n++; |
1297 | } | |
1298 | return n; | |
1299 | } | |
1300 | #else | |
846705de RW |
1301 | static inline void *saveable_highmem_page(struct zone *z, unsigned long p) |
1302 | { | |
1303 | return NULL; | |
1304 | } | |
8357376d RW |
1305 | #endif /* CONFIG_HIGHMEM */ |
1306 | ||
25761b6e | 1307 | /** |
ef96f639 RW |
1308 | * saveable_page - Check if the given page is saveable. |
1309 | * | |
1310 | * Determine whether a non-highmem page should be included in a hibernation | |
1311 | * image. | |
25761b6e | 1312 | * |
ef96f639 RW |
1313 | * We should save the page if it isn't Nosave, and is not in the range |
1314 | * of pages statically defined as 'unsaveable', and it isn't part of | |
1315 | * a free chunk of pages. | |
25761b6e | 1316 | */ |
846705de | 1317 | static struct page *saveable_page(struct zone *zone, unsigned long pfn) |
25761b6e | 1318 | { |
de491861 | 1319 | struct page *page; |
25761b6e RW |
1320 | |
1321 | if (!pfn_valid(pfn)) | |
ae83c5ee | 1322 | return NULL; |
25761b6e | 1323 | |
5b56db37 DH |
1324 | page = pfn_to_online_page(pfn); |
1325 | if (!page || page_zone(page) != zone) | |
846705de | 1326 | return NULL; |
ae83c5ee | 1327 | |
8357376d RW |
1328 | BUG_ON(PageHighMem(page)); |
1329 | ||
7be98234 | 1330 | if (swsusp_page_is_forbidden(page) || swsusp_page_is_free(page)) |
ae83c5ee | 1331 | return NULL; |
8357376d | 1332 | |
abd02ac6 DH |
1333 | if (PageOffline(page)) |
1334 | return NULL; | |
1335 | ||
8a235efa RW |
1336 | if (PageReserved(page) |
1337 | && (!kernel_page_present(page) || pfn_is_nosave(pfn))) | |
ae83c5ee | 1338 | return NULL; |
25761b6e | 1339 | |
c6968e73 SG |
1340 | if (page_is_guard(page)) |
1341 | return NULL; | |
1342 | ||
ae83c5ee | 1343 | return page; |
25761b6e RW |
1344 | } |
1345 | ||
8357376d | 1346 | /** |
ef96f639 | 1347 | * count_data_pages - Compute the total number of saveable non-highmem pages. |
8357376d | 1348 | */ |
fe419535 | 1349 | static unsigned int count_data_pages(void) |
25761b6e RW |
1350 | { |
1351 | struct zone *zone; | |
ae83c5ee | 1352 | unsigned long pfn, max_zone_pfn; |
dc19d507 | 1353 | unsigned int n = 0; |
25761b6e | 1354 | |
98e73dc5 | 1355 | for_each_populated_zone(zone) { |
25761b6e RW |
1356 | if (is_highmem(zone)) |
1357 | continue; | |
8357376d | 1358 | |
25761b6e | 1359 | mark_free_pages(zone); |
c33bc315 | 1360 | max_zone_pfn = zone_end_pfn(zone); |
ae83c5ee | 1361 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
846705de | 1362 | if (saveable_page(zone, pfn)) |
8357376d | 1363 | n++; |
25761b6e | 1364 | } |
a0f49651 | 1365 | return n; |
25761b6e RW |
1366 | } |
1367 | ||
ef96f639 RW |
1368 | /* |
1369 | * This is needed, because copy_page and memcpy are not usable for copying | |
8357376d RW |
1370 | * task structs. |
1371 | */ | |
1372 | static inline void do_copy_page(long *dst, long *src) | |
f623f0db RW |
1373 | { |
1374 | int n; | |
1375 | ||
f623f0db RW |
1376 | for (n = PAGE_SIZE / sizeof(long); n; n--) |
1377 | *dst++ = *src++; | |
1378 | } | |
1379 | ||
8a235efa | 1380 | /** |
ef96f639 RW |
1381 | * safe_copy_page - Copy a page in a safe way. |
1382 | * | |
1383 | * Check if the page we are going to copy is marked as present in the kernel | |
d6332692 RE |
1384 | * page tables. This always is the case if CONFIG_DEBUG_PAGEALLOC or |
1385 | * CONFIG_ARCH_HAS_SET_DIRECT_MAP is not set. In that case kernel_page_present() | |
1386 | * always returns 'true'. | |
8a235efa RW |
1387 | */ |
1388 | static void safe_copy_page(void *dst, struct page *s_page) | |
1389 | { | |
1390 | if (kernel_page_present(s_page)) { | |
1391 | do_copy_page(dst, page_address(s_page)); | |
1392 | } else { | |
2abf962a | 1393 | hibernate_map_page(s_page); |
8a235efa | 1394 | do_copy_page(dst, page_address(s_page)); |
2abf962a | 1395 | hibernate_unmap_page(s_page); |
8a235efa RW |
1396 | } |
1397 | } | |
1398 | ||
8357376d | 1399 | #ifdef CONFIG_HIGHMEM |
efd5a852 | 1400 | static inline struct page *page_is_saveable(struct zone *zone, unsigned long pfn) |
8357376d RW |
1401 | { |
1402 | return is_highmem(zone) ? | |
846705de | 1403 | saveable_highmem_page(zone, pfn) : saveable_page(zone, pfn); |
8357376d RW |
1404 | } |
1405 | ||
8a235efa | 1406 | static void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d RW |
1407 | { |
1408 | struct page *s_page, *d_page; | |
1409 | void *src, *dst; | |
1410 | ||
1411 | s_page = pfn_to_page(src_pfn); | |
1412 | d_page = pfn_to_page(dst_pfn); | |
1413 | if (PageHighMem(s_page)) { | |
0de9a1e2 CW |
1414 | src = kmap_atomic(s_page); |
1415 | dst = kmap_atomic(d_page); | |
8357376d | 1416 | do_copy_page(dst, src); |
0de9a1e2 CW |
1417 | kunmap_atomic(dst); |
1418 | kunmap_atomic(src); | |
8357376d | 1419 | } else { |
8357376d | 1420 | if (PageHighMem(d_page)) { |
ef96f639 RW |
1421 | /* |
1422 | * The page pointed to by src may contain some kernel | |
8357376d RW |
1423 | * data modified by kmap_atomic() |
1424 | */ | |
8a235efa | 1425 | safe_copy_page(buffer, s_page); |
0de9a1e2 | 1426 | dst = kmap_atomic(d_page); |
3ecb01df | 1427 | copy_page(dst, buffer); |
0de9a1e2 | 1428 | kunmap_atomic(dst); |
8357376d | 1429 | } else { |
8a235efa | 1430 | safe_copy_page(page_address(d_page), s_page); |
8357376d RW |
1431 | } |
1432 | } | |
1433 | } | |
1434 | #else | |
846705de | 1435 | #define page_is_saveable(zone, pfn) saveable_page(zone, pfn) |
8357376d | 1436 | |
8a235efa | 1437 | static inline void copy_data_page(unsigned long dst_pfn, unsigned long src_pfn) |
8357376d | 1438 | { |
8a235efa RW |
1439 | safe_copy_page(page_address(pfn_to_page(dst_pfn)), |
1440 | pfn_to_page(src_pfn)); | |
8357376d RW |
1441 | } |
1442 | #endif /* CONFIG_HIGHMEM */ | |
1443 | ||
efd5a852 RW |
1444 | static void copy_data_pages(struct memory_bitmap *copy_bm, |
1445 | struct memory_bitmap *orig_bm) | |
25761b6e RW |
1446 | { |
1447 | struct zone *zone; | |
b788db79 | 1448 | unsigned long pfn; |
25761b6e | 1449 | |
98e73dc5 | 1450 | for_each_populated_zone(zone) { |
b788db79 RW |
1451 | unsigned long max_zone_pfn; |
1452 | ||
25761b6e | 1453 | mark_free_pages(zone); |
c33bc315 | 1454 | max_zone_pfn = zone_end_pfn(zone); |
b788db79 | 1455 | for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) |
8357376d | 1456 | if (page_is_saveable(zone, pfn)) |
b788db79 | 1457 | memory_bm_set_bit(orig_bm, pfn); |
25761b6e | 1458 | } |
b788db79 RW |
1459 | memory_bm_position_reset(orig_bm); |
1460 | memory_bm_position_reset(copy_bm); | |
df7c4872 | 1461 | for(;;) { |
b788db79 | 1462 | pfn = memory_bm_next_pfn(orig_bm); |
df7c4872 FW |
1463 | if (unlikely(pfn == BM_END_OF_MAP)) |
1464 | break; | |
1465 | copy_data_page(memory_bm_next_pfn(copy_bm), pfn); | |
1466 | } | |
25761b6e RW |
1467 | } |
1468 | ||
8357376d RW |
1469 | /* Total number of image pages */ |
1470 | static unsigned int nr_copy_pages; | |
1471 | /* Number of pages needed for saving the original pfns of the image pages */ | |
1472 | static unsigned int nr_meta_pages; | |
64a473cb RW |
1473 | /* |
1474 | * Numbers of normal and highmem page frames allocated for hibernation image | |
1475 | * before suspending devices. | |
1476 | */ | |
0bae5fd3 | 1477 | static unsigned int alloc_normal, alloc_highmem; |
64a473cb RW |
1478 | /* |
1479 | * Memory bitmap used for marking saveable pages (during hibernation) or | |
1480 | * hibernation image pages (during restore) | |
1481 | */ | |
1482 | static struct memory_bitmap orig_bm; | |
1483 | /* | |
1484 | * Memory bitmap used during hibernation for marking allocated page frames that | |
1485 | * will contain copies of saveable pages. During restore it is initially used | |
1486 | * for marking hibernation image pages, but then the set bits from it are | |
1487 | * duplicated in @orig_bm and it is released. On highmem systems it is next | |
1488 | * used for marking "safe" highmem pages, but it has to be reinitialized for | |
1489 | * this purpose. | |
1490 | */ | |
1491 | static struct memory_bitmap copy_bm; | |
8357376d | 1492 | |
25761b6e | 1493 | /** |
ef96f639 | 1494 | * swsusp_free - Free pages allocated for hibernation image. |
cd560bb2 | 1495 | * |
6be2408a | 1496 | * Image pages are allocated before snapshot creation, so they need to be |
ef96f639 | 1497 | * released after resume. |
25761b6e | 1498 | */ |
25761b6e RW |
1499 | void swsusp_free(void) |
1500 | { | |
fdd64ed5 | 1501 | unsigned long fb_pfn, fr_pfn; |
6efde38f | 1502 | |
fdd64ed5 JR |
1503 | if (!forbidden_pages_map || !free_pages_map) |
1504 | goto out; | |
1505 | ||
1506 | memory_bm_position_reset(forbidden_pages_map); | |
1507 | memory_bm_position_reset(free_pages_map); | |
1508 | ||
1509 | loop: | |
1510 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1511 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1512 | ||
1513 | /* | |
1514 | * Find the next bit set in both bitmaps. This is guaranteed to | |
1515 | * terminate when fb_pfn == fr_pfn == BM_END_OF_MAP. | |
1516 | */ | |
1517 | do { | |
1518 | if (fb_pfn < fr_pfn) | |
1519 | fb_pfn = memory_bm_next_pfn(forbidden_pages_map); | |
1520 | if (fr_pfn < fb_pfn) | |
1521 | fr_pfn = memory_bm_next_pfn(free_pages_map); | |
1522 | } while (fb_pfn != fr_pfn); | |
1523 | ||
1524 | if (fr_pfn != BM_END_OF_MAP && pfn_valid(fr_pfn)) { | |
1525 | struct page *page = pfn_to_page(fr_pfn); | |
1526 | ||
1527 | memory_bm_clear_current(forbidden_pages_map); | |
1528 | memory_bm_clear_current(free_pages_map); | |
4c0b6c10 | 1529 | hibernate_restore_unprotect_page(page_address(page)); |
fdd64ed5 JR |
1530 | __free_page(page); |
1531 | goto loop; | |
25761b6e | 1532 | } |
fdd64ed5 JR |
1533 | |
1534 | out: | |
f577eb30 RW |
1535 | nr_copy_pages = 0; |
1536 | nr_meta_pages = 0; | |
75534b50 | 1537 | restore_pblist = NULL; |
6e1819d6 | 1538 | buffer = NULL; |
64a473cb RW |
1539 | alloc_normal = 0; |
1540 | alloc_highmem = 0; | |
4c0b6c10 | 1541 | hibernate_restore_protection_end(); |
25761b6e RW |
1542 | } |
1543 | ||
4bb33435 RW |
1544 | /* Helper functions used for the shrinking of memory. */ |
1545 | ||
1546 | #define GFP_IMAGE (GFP_KERNEL | __GFP_NOWARN) | |
1547 | ||
fe419535 | 1548 | /** |
ef96f639 | 1549 | * preallocate_image_pages - Allocate a number of pages for hibernation image. |
4bb33435 RW |
1550 | * @nr_pages: Number of page frames to allocate. |
1551 | * @mask: GFP flags to use for the allocation. | |
fe419535 | 1552 | * |
4bb33435 RW |
1553 | * Return value: Number of page frames actually allocated |
1554 | */ | |
1555 | static unsigned long preallocate_image_pages(unsigned long nr_pages, gfp_t mask) | |
1556 | { | |
1557 | unsigned long nr_alloc = 0; | |
1558 | ||
1559 | while (nr_pages > 0) { | |
64a473cb RW |
1560 | struct page *page; |
1561 | ||
1562 | page = alloc_image_page(mask); | |
1563 | if (!page) | |
4bb33435 | 1564 | break; |
64a473cb RW |
1565 | memory_bm_set_bit(©_bm, page_to_pfn(page)); |
1566 | if (PageHighMem(page)) | |
1567 | alloc_highmem++; | |
1568 | else | |
1569 | alloc_normal++; | |
4bb33435 RW |
1570 | nr_pages--; |
1571 | nr_alloc++; | |
1572 | } | |
1573 | ||
1574 | return nr_alloc; | |
1575 | } | |
1576 | ||
6715045d RW |
1577 | static unsigned long preallocate_image_memory(unsigned long nr_pages, |
1578 | unsigned long avail_normal) | |
4bb33435 | 1579 | { |
6715045d RW |
1580 | unsigned long alloc; |
1581 | ||
1582 | if (avail_normal <= alloc_normal) | |
1583 | return 0; | |
1584 | ||
1585 | alloc = avail_normal - alloc_normal; | |
1586 | if (nr_pages < alloc) | |
1587 | alloc = nr_pages; | |
1588 | ||
1589 | return preallocate_image_pages(alloc, GFP_IMAGE); | |
4bb33435 RW |
1590 | } |
1591 | ||
1592 | #ifdef CONFIG_HIGHMEM | |
1593 | static unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1594 | { | |
1595 | return preallocate_image_pages(nr_pages, GFP_IMAGE | __GFP_HIGHMEM); | |
1596 | } | |
1597 | ||
1598 | /** | |
ef96f639 | 1599 | * __fraction - Compute (an approximation of) x * (multiplier / base). |
fe419535 | 1600 | */ |
4bb33435 RW |
1601 | static unsigned long __fraction(u64 x, u64 multiplier, u64 base) |
1602 | { | |
809ed78a | 1603 | return div64_u64(x * multiplier, base); |
4bb33435 | 1604 | } |
fe419535 | 1605 | |
4bb33435 | 1606 | static unsigned long preallocate_highmem_fraction(unsigned long nr_pages, |
efd5a852 RW |
1607 | unsigned long highmem, |
1608 | unsigned long total) | |
fe419535 | 1609 | { |
4bb33435 RW |
1610 | unsigned long alloc = __fraction(nr_pages, highmem, total); |
1611 | ||
1612 | return preallocate_image_pages(alloc, GFP_IMAGE | __GFP_HIGHMEM); | |
fe419535 | 1613 | } |
4bb33435 RW |
1614 | #else /* CONFIG_HIGHMEM */ |
1615 | static inline unsigned long preallocate_image_highmem(unsigned long nr_pages) | |
1616 | { | |
1617 | return 0; | |
1618 | } | |
1619 | ||
1620 | static inline unsigned long preallocate_highmem_fraction(unsigned long nr_pages, | |
efd5a852 RW |
1621 | unsigned long highmem, |
1622 | unsigned long total) | |
4bb33435 RW |
1623 | { |
1624 | return 0; | |
1625 | } | |
1626 | #endif /* CONFIG_HIGHMEM */ | |
fe419535 | 1627 | |
4bb33435 | 1628 | /** |
ef96f639 | 1629 | * free_unnecessary_pages - Release preallocated pages not needed for the image. |
64a473cb | 1630 | */ |
a64fc82c | 1631 | static unsigned long free_unnecessary_pages(void) |
64a473cb | 1632 | { |
a64fc82c | 1633 | unsigned long save, to_free_normal, to_free_highmem, free; |
64a473cb | 1634 | |
6715045d RW |
1635 | save = count_data_pages(); |
1636 | if (alloc_normal >= save) { | |
1637 | to_free_normal = alloc_normal - save; | |
1638 | save = 0; | |
1639 | } else { | |
1640 | to_free_normal = 0; | |
1641 | save -= alloc_normal; | |
1642 | } | |
1643 | save += count_highmem_pages(); | |
1644 | if (alloc_highmem >= save) { | |
1645 | to_free_highmem = alloc_highmem - save; | |
64a473cb RW |
1646 | } else { |
1647 | to_free_highmem = 0; | |
4d4cf23c RW |
1648 | save -= alloc_highmem; |
1649 | if (to_free_normal > save) | |
1650 | to_free_normal -= save; | |
1651 | else | |
1652 | to_free_normal = 0; | |
64a473cb | 1653 | } |
a64fc82c | 1654 | free = to_free_normal + to_free_highmem; |
64a473cb RW |
1655 | |
1656 | memory_bm_position_reset(©_bm); | |
1657 | ||
a9c9b442 | 1658 | while (to_free_normal > 0 || to_free_highmem > 0) { |
64a473cb RW |
1659 | unsigned long pfn = memory_bm_next_pfn(©_bm); |
1660 | struct page *page = pfn_to_page(pfn); | |
1661 | ||
1662 | if (PageHighMem(page)) { | |
1663 | if (!to_free_highmem) | |
1664 | continue; | |
1665 | to_free_highmem--; | |
1666 | alloc_highmem--; | |
1667 | } else { | |
1668 | if (!to_free_normal) | |
1669 | continue; | |
1670 | to_free_normal--; | |
1671 | alloc_normal--; | |
1672 | } | |
1673 | memory_bm_clear_bit(©_bm, pfn); | |
1674 | swsusp_unset_page_forbidden(page); | |
1675 | swsusp_unset_page_free(page); | |
1676 | __free_page(page); | |
1677 | } | |
a64fc82c WK |
1678 | |
1679 | return free; | |
64a473cb RW |
1680 | } |
1681 | ||
ef4aede3 | 1682 | /** |
ef96f639 | 1683 | * minimum_image_size - Estimate the minimum acceptable size of an image. |
ef4aede3 RW |
1684 | * @saveable: Number of saveable pages in the system. |
1685 | * | |
1686 | * We want to avoid attempting to free too much memory too hard, so estimate the | |
1687 | * minimum acceptable size of a hibernation image to use as the lower limit for | |
1688 | * preallocating memory. | |
1689 | * | |
1690 | * We assume that the minimum image size should be proportional to | |
1691 | * | |
1692 | * [number of saveable pages] - [number of pages that can be freed in theory] | |
1693 | * | |
1694 | * where the second term is the sum of (1) reclaimable slab pages, (2) active | |
bdbc98ab | 1695 | * and (3) inactive anonymous pages, (4) active and (5) inactive file pages. |
ef4aede3 RW |
1696 | */ |
1697 | static unsigned long minimum_image_size(unsigned long saveable) | |
1698 | { | |
1699 | unsigned long size; | |
1700 | ||
d42f3245 | 1701 | size = global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B) |
599d0c95 MG |
1702 | + global_node_page_state(NR_ACTIVE_ANON) |
1703 | + global_node_page_state(NR_INACTIVE_ANON) | |
1704 | + global_node_page_state(NR_ACTIVE_FILE) | |
bdbc98ab | 1705 | + global_node_page_state(NR_INACTIVE_FILE); |
ef4aede3 RW |
1706 | |
1707 | return saveable <= size ? 0 : saveable - size; | |
1708 | } | |
1709 | ||
64a473cb | 1710 | /** |
ef96f639 | 1711 | * hibernate_preallocate_memory - Preallocate memory for hibernation image. |
4bb33435 RW |
1712 | * |
1713 | * To create a hibernation image it is necessary to make a copy of every page | |
1714 | * frame in use. We also need a number of page frames to be free during | |
1715 | * hibernation for allocations made while saving the image and for device | |
1716 | * drivers, in case they need to allocate memory from their hibernation | |
ddeb6487 | 1717 | * callbacks (these two numbers are given by PAGES_FOR_IO (which is a rough |
b0c609ab | 1718 | * estimate) and reserved_size divided by PAGE_SIZE (which is tunable through |
ddeb6487 RW |
1719 | * /sys/power/reserved_size, respectively). To make this happen, we compute the |
1720 | * total number of available page frames and allocate at least | |
4bb33435 | 1721 | * |
ddeb6487 RW |
1722 | * ([page frames total] + PAGES_FOR_IO + [metadata pages]) / 2 |
1723 | * + 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE) | |
4bb33435 RW |
1724 | * |
1725 | * of them, which corresponds to the maximum size of a hibernation image. | |
1726 | * | |
1727 | * If image_size is set below the number following from the above formula, | |
1728 | * the preallocation of memory is continued until the total number of saveable | |
ef4aede3 RW |
1729 | * pages in the system is below the requested image size or the minimum |
1730 | * acceptable image size returned by minimum_image_size(), whichever is greater. | |
4bb33435 | 1731 | */ |
64a473cb | 1732 | int hibernate_preallocate_memory(void) |
fe419535 | 1733 | { |
fe419535 | 1734 | struct zone *zone; |
4bb33435 | 1735 | unsigned long saveable, size, max_size, count, highmem, pages = 0; |
6715045d | 1736 | unsigned long alloc, save_highmem, pages_highmem, avail_normal; |
db597605 | 1737 | ktime_t start, stop; |
64a473cb | 1738 | int error; |
fe419535 | 1739 | |
7a7b99bf | 1740 | pr_info("Preallocating image memory\n"); |
db597605 | 1741 | start = ktime_get(); |
fe419535 | 1742 | |
64a473cb | 1743 | error = memory_bm_create(&orig_bm, GFP_IMAGE, PG_ANY); |
7a7b99bf LS |
1744 | if (error) { |
1745 | pr_err("Cannot allocate original bitmap\n"); | |
64a473cb | 1746 | goto err_out; |
7a7b99bf | 1747 | } |
64a473cb RW |
1748 | |
1749 | error = memory_bm_create(©_bm, GFP_IMAGE, PG_ANY); | |
7a7b99bf LS |
1750 | if (error) { |
1751 | pr_err("Cannot allocate copy bitmap\n"); | |
64a473cb | 1752 | goto err_out; |
7a7b99bf | 1753 | } |
64a473cb RW |
1754 | |
1755 | alloc_normal = 0; | |
1756 | alloc_highmem = 0; | |
1757 | ||
4bb33435 | 1758 | /* Count the number of saveable data pages. */ |
64a473cb | 1759 | save_highmem = count_highmem_pages(); |
4bb33435 | 1760 | saveable = count_data_pages(); |
fe419535 | 1761 | |
4bb33435 RW |
1762 | /* |
1763 | * Compute the total number of page frames we can use (count) and the | |
1764 | * number of pages needed for image metadata (size). | |
1765 | */ | |
1766 | count = saveable; | |
64a473cb RW |
1767 | saveable += save_highmem; |
1768 | highmem = save_highmem; | |
4bb33435 RW |
1769 | size = 0; |
1770 | for_each_populated_zone(zone) { | |
1771 | size += snapshot_additional_pages(zone); | |
1772 | if (is_highmem(zone)) | |
1773 | highmem += zone_page_state(zone, NR_FREE_PAGES); | |
1774 | else | |
1775 | count += zone_page_state(zone, NR_FREE_PAGES); | |
1776 | } | |
6715045d | 1777 | avail_normal = count; |
4bb33435 RW |
1778 | count += highmem; |
1779 | count -= totalreserve_pages; | |
1780 | ||
1781 | /* Compute the maximum number of saveable pages to leave in memory. */ | |
ddeb6487 RW |
1782 | max_size = (count - (size + PAGES_FOR_IO)) / 2 |
1783 | - 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE); | |
266f1a25 | 1784 | /* Compute the desired number of image pages specified by image_size. */ |
4bb33435 RW |
1785 | size = DIV_ROUND_UP(image_size, PAGE_SIZE); |
1786 | if (size > max_size) | |
1787 | size = max_size; | |
1788 | /* | |
266f1a25 RW |
1789 | * If the desired number of image pages is at least as large as the |
1790 | * current number of saveable pages in memory, allocate page frames for | |
1791 | * the image and we're done. | |
4bb33435 | 1792 | */ |
64a473cb RW |
1793 | if (size >= saveable) { |
1794 | pages = preallocate_image_highmem(save_highmem); | |
6715045d | 1795 | pages += preallocate_image_memory(saveable - pages, avail_normal); |
4bb33435 | 1796 | goto out; |
64a473cb | 1797 | } |
4bb33435 | 1798 | |
ef4aede3 RW |
1799 | /* Estimate the minimum size of the image. */ |
1800 | pages = minimum_image_size(saveable); | |
6715045d RW |
1801 | /* |
1802 | * To avoid excessive pressure on the normal zone, leave room in it to | |
1803 | * accommodate an image of the minimum size (unless it's already too | |
1804 | * small, in which case don't preallocate pages from it at all). | |
1805 | */ | |
1806 | if (avail_normal > pages) | |
1807 | avail_normal -= pages; | |
1808 | else | |
1809 | avail_normal = 0; | |
ef4aede3 RW |
1810 | if (size < pages) |
1811 | size = min_t(unsigned long, pages, max_size); | |
1812 | ||
4bb33435 RW |
1813 | /* |
1814 | * Let the memory management subsystem know that we're going to need a | |
1815 | * large number of page frames to allocate and make it free some memory. | |
1816 | * NOTE: If this is not done, performance will be hurt badly in some | |
1817 | * test cases. | |
1818 | */ | |
1819 | shrink_all_memory(saveable - size); | |
1820 | ||
1821 | /* | |
1822 | * The number of saveable pages in memory was too high, so apply some | |
1823 | * pressure to decrease it. First, make room for the largest possible | |
1824 | * image and fail if that doesn't work. Next, try to decrease the size | |
ef4aede3 RW |
1825 | * of the image as much as indicated by 'size' using allocations from |
1826 | * highmem and non-highmem zones separately. | |
4bb33435 RW |
1827 | */ |
1828 | pages_highmem = preallocate_image_highmem(highmem / 2); | |
fd432b9f AL |
1829 | alloc = count - max_size; |
1830 | if (alloc > pages_highmem) | |
1831 | alloc -= pages_highmem; | |
1832 | else | |
1833 | alloc = 0; | |
6715045d RW |
1834 | pages = preallocate_image_memory(alloc, avail_normal); |
1835 | if (pages < alloc) { | |
1836 | /* We have exhausted non-highmem pages, try highmem. */ | |
1837 | alloc -= pages; | |
1838 | pages += pages_highmem; | |
1839 | pages_highmem = preallocate_image_highmem(alloc); | |
7a7b99bf LS |
1840 | if (pages_highmem < alloc) { |
1841 | pr_err("Image allocation is %lu pages short\n", | |
1842 | alloc - pages_highmem); | |
6715045d | 1843 | goto err_out; |
7a7b99bf | 1844 | } |
6715045d RW |
1845 | pages += pages_highmem; |
1846 | /* | |
1847 | * size is the desired number of saveable pages to leave in | |
1848 | * memory, so try to preallocate (all memory - size) pages. | |
1849 | */ | |
1850 | alloc = (count - pages) - size; | |
1851 | pages += preallocate_image_highmem(alloc); | |
1852 | } else { | |
1853 | /* | |
1854 | * There are approximately max_size saveable pages at this point | |
1855 | * and we want to reduce this number down to size. | |
1856 | */ | |
1857 | alloc = max_size - size; | |
1858 | size = preallocate_highmem_fraction(alloc, highmem, count); | |
1859 | pages_highmem += size; | |
1860 | alloc -= size; | |
1861 | size = preallocate_image_memory(alloc, avail_normal); | |
1862 | pages_highmem += preallocate_image_highmem(alloc - size); | |
1863 | pages += pages_highmem + size; | |
1864 | } | |
4bb33435 | 1865 | |
64a473cb RW |
1866 | /* |
1867 | * We only need as many page frames for the image as there are saveable | |
1868 | * pages in memory, but we have allocated more. Release the excessive | |
1869 | * ones now. | |
1870 | */ | |
a64fc82c | 1871 | pages -= free_unnecessary_pages(); |
4bb33435 RW |
1872 | |
1873 | out: | |
db597605 | 1874 | stop = ktime_get(); |
5c0e9de0 | 1875 | pr_info("Allocated %lu pages for snapshot\n", pages); |
db597605 | 1876 | swsusp_show_speed(start, stop, pages, "Allocated"); |
fe419535 RW |
1877 | |
1878 | return 0; | |
64a473cb RW |
1879 | |
1880 | err_out: | |
64a473cb RW |
1881 | swsusp_free(); |
1882 | return -ENOMEM; | |
fe419535 RW |
1883 | } |
1884 | ||
8357376d RW |
1885 | #ifdef CONFIG_HIGHMEM |
1886 | /** | |
ef96f639 RW |
1887 | * count_pages_for_highmem - Count non-highmem pages needed for copying highmem. |
1888 | * | |
1889 | * Compute the number of non-highmem pages that will be necessary for creating | |
1890 | * copies of highmem pages. | |
1891 | */ | |
8357376d RW |
1892 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) |
1893 | { | |
64a473cb | 1894 | unsigned int free_highmem = count_free_highmem_pages() + alloc_highmem; |
8357376d RW |
1895 | |
1896 | if (free_highmem >= nr_highmem) | |
1897 | nr_highmem = 0; | |
1898 | else | |
1899 | nr_highmem -= free_highmem; | |
1900 | ||
1901 | return nr_highmem; | |
1902 | } | |
1903 | #else | |
efd5a852 | 1904 | static unsigned int count_pages_for_highmem(unsigned int nr_highmem) { return 0; } |
8357376d | 1905 | #endif /* CONFIG_HIGHMEM */ |
25761b6e RW |
1906 | |
1907 | /** | |
ef96f639 | 1908 | * enough_free_mem - Check if there is enough free memory for the image. |
25761b6e | 1909 | */ |
8357376d | 1910 | static int enough_free_mem(unsigned int nr_pages, unsigned int nr_highmem) |
25761b6e | 1911 | { |
e5e2fa78 | 1912 | struct zone *zone; |
64a473cb | 1913 | unsigned int free = alloc_normal; |
e5e2fa78 | 1914 | |
98e73dc5 | 1915 | for_each_populated_zone(zone) |
8357376d | 1916 | if (!is_highmem(zone)) |
d23ad423 | 1917 | free += zone_page_state(zone, NR_FREE_PAGES); |
940864dd | 1918 | |
8357376d | 1919 | nr_pages += count_pages_for_highmem(nr_highmem); |
64ec72a1 JP |
1920 | pr_debug("Normal pages needed: %u + %u, available pages: %u\n", |
1921 | nr_pages, PAGES_FOR_IO, free); | |
940864dd | 1922 | |
64a473cb | 1923 | return free > nr_pages + PAGES_FOR_IO; |
25761b6e RW |
1924 | } |
1925 | ||
8357376d RW |
1926 | #ifdef CONFIG_HIGHMEM |
1927 | /** | |
ef96f639 RW |
1928 | * get_highmem_buffer - Allocate a buffer for highmem pages. |
1929 | * | |
1930 | * If there are some highmem pages in the hibernation image, we may need a | |
1931 | * buffer to copy them and/or load their data. | |
8357376d | 1932 | */ |
8357376d RW |
1933 | static inline int get_highmem_buffer(int safe_needed) |
1934 | { | |
453f85d4 | 1935 | buffer = get_image_page(GFP_ATOMIC, safe_needed); |
8357376d RW |
1936 | return buffer ? 0 : -ENOMEM; |
1937 | } | |
1938 | ||
1939 | /** | |
ef96f639 RW |
1940 | * alloc_highmem_image_pages - Allocate some highmem pages for the image. |
1941 | * | |
1942 | * Try to allocate as many pages as needed, but if the number of free highmem | |
1943 | * pages is less than that, allocate them all. | |
8357376d | 1944 | */ |
efd5a852 RW |
1945 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1946 | unsigned int nr_highmem) | |
8357376d RW |
1947 | { |
1948 | unsigned int to_alloc = count_free_highmem_pages(); | |
1949 | ||
1950 | if (to_alloc > nr_highmem) | |
1951 | to_alloc = nr_highmem; | |
1952 | ||
1953 | nr_highmem -= to_alloc; | |
1954 | while (to_alloc-- > 0) { | |
1955 | struct page *page; | |
1956 | ||
d0164adc | 1957 | page = alloc_image_page(__GFP_HIGHMEM|__GFP_KSWAPD_RECLAIM); |
8357376d RW |
1958 | memory_bm_set_bit(bm, page_to_pfn(page)); |
1959 | } | |
1960 | return nr_highmem; | |
1961 | } | |
1962 | #else | |
1963 | static inline int get_highmem_buffer(int safe_needed) { return 0; } | |
1964 | ||
efd5a852 RW |
1965 | static inline unsigned int alloc_highmem_pages(struct memory_bitmap *bm, |
1966 | unsigned int n) { return 0; } | |
8357376d RW |
1967 | #endif /* CONFIG_HIGHMEM */ |
1968 | ||
1969 | /** | |
ef96f639 | 1970 | * swsusp_alloc - Allocate memory for hibernation image. |
8357376d | 1971 | * |
ef96f639 RW |
1972 | * We first try to allocate as many highmem pages as there are |
1973 | * saveable highmem pages in the system. If that fails, we allocate | |
1974 | * non-highmem pages for the copies of the remaining highmem ones. | |
8357376d | 1975 | * |
ef96f639 RW |
1976 | * In this approach it is likely that the copies of highmem pages will |
1977 | * also be located in the high memory, because of the way in which | |
1978 | * copy_data_pages() works. | |
8357376d | 1979 | */ |
eba74c29 | 1980 | static int swsusp_alloc(struct memory_bitmap *copy_bm, |
efd5a852 | 1981 | unsigned int nr_pages, unsigned int nr_highmem) |
054bd4c1 | 1982 | { |
8357376d | 1983 | if (nr_highmem > 0) { |
2e725a06 | 1984 | if (get_highmem_buffer(PG_ANY)) |
64a473cb RW |
1985 | goto err_out; |
1986 | if (nr_highmem > alloc_highmem) { | |
1987 | nr_highmem -= alloc_highmem; | |
1988 | nr_pages += alloc_highmem_pages(copy_bm, nr_highmem); | |
1989 | } | |
8357376d | 1990 | } |
64a473cb RW |
1991 | if (nr_pages > alloc_normal) { |
1992 | nr_pages -= alloc_normal; | |
1993 | while (nr_pages-- > 0) { | |
1994 | struct page *page; | |
1995 | ||
453f85d4 | 1996 | page = alloc_image_page(GFP_ATOMIC); |
64a473cb RW |
1997 | if (!page) |
1998 | goto err_out; | |
1999 | memory_bm_set_bit(copy_bm, page_to_pfn(page)); | |
2000 | } | |
25761b6e | 2001 | } |
64a473cb | 2002 | |
b788db79 | 2003 | return 0; |
25761b6e | 2004 | |
64a473cb | 2005 | err_out: |
b788db79 | 2006 | swsusp_free(); |
2e725a06 | 2007 | return -ENOMEM; |
25761b6e RW |
2008 | } |
2009 | ||
722a9f92 | 2010 | asmlinkage __visible int swsusp_save(void) |
25761b6e | 2011 | { |
8357376d | 2012 | unsigned int nr_pages, nr_highmem; |
25761b6e | 2013 | |
7a7b99bf | 2014 | pr_info("Creating image:\n"); |
25761b6e | 2015 | |
9f8f2172 | 2016 | drain_local_pages(NULL); |
a0f49651 | 2017 | nr_pages = count_data_pages(); |
8357376d | 2018 | nr_highmem = count_highmem_pages(); |
64ec72a1 | 2019 | pr_info("Need to copy %u pages\n", nr_pages + nr_highmem); |
25761b6e | 2020 | |
8357376d | 2021 | if (!enough_free_mem(nr_pages, nr_highmem)) { |
64ec72a1 | 2022 | pr_err("Not enough free memory\n"); |
25761b6e RW |
2023 | return -ENOMEM; |
2024 | } | |
2025 | ||
eba74c29 | 2026 | if (swsusp_alloc(©_bm, nr_pages, nr_highmem)) { |
64ec72a1 | 2027 | pr_err("Memory allocation failed\n"); |
a0f49651 | 2028 | return -ENOMEM; |
8357376d | 2029 | } |
25761b6e | 2030 | |
ef96f639 RW |
2031 | /* |
2032 | * During allocating of suspend pagedir, new cold pages may appear. | |
25761b6e RW |
2033 | * Kill them. |
2034 | */ | |
9f8f2172 | 2035 | drain_local_pages(NULL); |
b788db79 | 2036 | copy_data_pages(©_bm, &orig_bm); |
25761b6e RW |
2037 | |
2038 | /* | |
2039 | * End of critical section. From now on, we can write to memory, | |
2040 | * but we should not touch disk. This specially means we must _not_ | |
2041 | * touch swap space! Except we must write out our image of course. | |
2042 | */ | |
2043 | ||
8357376d | 2044 | nr_pages += nr_highmem; |
a0f49651 | 2045 | nr_copy_pages = nr_pages; |
8357376d | 2046 | nr_meta_pages = DIV_ROUND_UP(nr_pages * sizeof(long), PAGE_SIZE); |
a0f49651 | 2047 | |
7a7b99bf | 2048 | pr_info("Image created (%d pages copied)\n", nr_pages); |
8357376d | 2049 | |
25761b6e RW |
2050 | return 0; |
2051 | } | |
f577eb30 | 2052 | |
d307c4a8 RW |
2053 | #ifndef CONFIG_ARCH_HIBERNATION_HEADER |
2054 | static int init_header_complete(struct swsusp_info *info) | |
f577eb30 | 2055 | { |
d307c4a8 | 2056 | memcpy(&info->uts, init_utsname(), sizeof(struct new_utsname)); |
f577eb30 | 2057 | info->version_code = LINUX_VERSION_CODE; |
d307c4a8 RW |
2058 | return 0; |
2059 | } | |
2060 | ||
02d7f400 | 2061 | static const char *check_image_kernel(struct swsusp_info *info) |
d307c4a8 RW |
2062 | { |
2063 | if (info->version_code != LINUX_VERSION_CODE) | |
2064 | return "kernel version"; | |
2065 | if (strcmp(info->uts.sysname,init_utsname()->sysname)) | |
2066 | return "system type"; | |
2067 | if (strcmp(info->uts.release,init_utsname()->release)) | |
2068 | return "kernel release"; | |
2069 | if (strcmp(info->uts.version,init_utsname()->version)) | |
2070 | return "version"; | |
2071 | if (strcmp(info->uts.machine,init_utsname()->machine)) | |
2072 | return "machine"; | |
2073 | return NULL; | |
2074 | } | |
2075 | #endif /* CONFIG_ARCH_HIBERNATION_HEADER */ | |
2076 | ||
af508b34 RW |
2077 | unsigned long snapshot_get_image_size(void) |
2078 | { | |
2079 | return nr_copy_pages + nr_meta_pages + 1; | |
2080 | } | |
2081 | ||
d307c4a8 RW |
2082 | static int init_header(struct swsusp_info *info) |
2083 | { | |
2084 | memset(info, 0, sizeof(struct swsusp_info)); | |
0ed5fd13 | 2085 | info->num_physpages = get_num_physpages(); |
f577eb30 | 2086 | info->image_pages = nr_copy_pages; |
af508b34 | 2087 | info->pages = snapshot_get_image_size(); |
6e1819d6 RW |
2088 | info->size = info->pages; |
2089 | info->size <<= PAGE_SHIFT; | |
d307c4a8 | 2090 | return init_header_complete(info); |
f577eb30 RW |
2091 | } |
2092 | ||
2093 | /** | |
ef96f639 RW |
2094 | * pack_pfns - Prepare PFNs for saving. |
2095 | * @bm: Memory bitmap. | |
2096 | * @buf: Memory buffer to store the PFNs in. | |
2097 | * | |
2098 | * PFNs corresponding to set bits in @bm are stored in the area of memory | |
2099 | * pointed to by @buf (1 page at a time). | |
f577eb30 | 2100 | */ |
efd5a852 | 2101 | static inline void pack_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2102 | { |
2103 | int j; | |
2104 | ||
b788db79 | 2105 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
940864dd RW |
2106 | buf[j] = memory_bm_next_pfn(bm); |
2107 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
b788db79 | 2108 | break; |
f577eb30 | 2109 | } |
f577eb30 RW |
2110 | } |
2111 | ||
2112 | /** | |
ef96f639 RW |
2113 | * snapshot_read_next - Get the address to read the next image page from. |
2114 | * @handle: Snapshot handle to be used for the reading. | |
f577eb30 | 2115 | * |
ef96f639 RW |
2116 | * On the first call, @handle should point to a zeroed snapshot_handle |
2117 | * structure. The structure gets populated then and a pointer to it should be | |
2118 | * passed to this function every next time. | |
f577eb30 | 2119 | * |
ef96f639 RW |
2120 | * On success, the function returns a positive number. Then, the caller |
2121 | * is allowed to read up to the returned number of bytes from the memory | |
2122 | * location computed by the data_of() macro. | |
f577eb30 | 2123 | * |
ef96f639 RW |
2124 | * The function returns 0 to indicate the end of the data stream condition, |
2125 | * and negative numbers are returned on errors. If that happens, the structure | |
2126 | * pointed to by @handle is not updated and should not be used any more. | |
f577eb30 | 2127 | */ |
d3c1b24c | 2128 | int snapshot_read_next(struct snapshot_handle *handle) |
f577eb30 | 2129 | { |
fb13a28b | 2130 | if (handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2131 | return 0; |
b788db79 | 2132 | |
f577eb30 RW |
2133 | if (!buffer) { |
2134 | /* This makes the buffer be freed by swsusp_free() */ | |
8357376d | 2135 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); |
f577eb30 RW |
2136 | if (!buffer) |
2137 | return -ENOMEM; | |
2138 | } | |
d3c1b24c | 2139 | if (!handle->cur) { |
d307c4a8 RW |
2140 | int error; |
2141 | ||
2142 | error = init_header((struct swsusp_info *)buffer); | |
2143 | if (error) | |
2144 | return error; | |
f577eb30 | 2145 | handle->buffer = buffer; |
b788db79 RW |
2146 | memory_bm_position_reset(&orig_bm); |
2147 | memory_bm_position_reset(©_bm); | |
d3c1b24c | 2148 | } else if (handle->cur <= nr_meta_pages) { |
3ecb01df | 2149 | clear_page(buffer); |
d3c1b24c JS |
2150 | pack_pfns(buffer, &orig_bm); |
2151 | } else { | |
2152 | struct page *page; | |
b788db79 | 2153 | |
d3c1b24c JS |
2154 | page = pfn_to_page(memory_bm_next_pfn(©_bm)); |
2155 | if (PageHighMem(page)) { | |
ef96f639 RW |
2156 | /* |
2157 | * Highmem pages are copied to the buffer, | |
d3c1b24c JS |
2158 | * because we can't return with a kmapped |
2159 | * highmem page (we may not be called again). | |
2160 | */ | |
2161 | void *kaddr; | |
8357376d | 2162 | |
0de9a1e2 | 2163 | kaddr = kmap_atomic(page); |
3ecb01df | 2164 | copy_page(buffer, kaddr); |
0de9a1e2 | 2165 | kunmap_atomic(kaddr); |
d3c1b24c JS |
2166 | handle->buffer = buffer; |
2167 | } else { | |
2168 | handle->buffer = page_address(page); | |
f577eb30 | 2169 | } |
f577eb30 | 2170 | } |
d3c1b24c JS |
2171 | handle->cur++; |
2172 | return PAGE_SIZE; | |
f577eb30 RW |
2173 | } |
2174 | ||
6dbecfd3 RW |
2175 | static void duplicate_memory_bitmap(struct memory_bitmap *dst, |
2176 | struct memory_bitmap *src) | |
2177 | { | |
2178 | unsigned long pfn; | |
2179 | ||
2180 | memory_bm_position_reset(src); | |
2181 | pfn = memory_bm_next_pfn(src); | |
2182 | while (pfn != BM_END_OF_MAP) { | |
2183 | memory_bm_set_bit(dst, pfn); | |
2184 | pfn = memory_bm_next_pfn(src); | |
2185 | } | |
2186 | } | |
2187 | ||
f577eb30 | 2188 | /** |
ef96f639 RW |
2189 | * mark_unsafe_pages - Mark pages that were used before hibernation. |
2190 | * | |
2191 | * Mark the pages that cannot be used for storing the image during restoration, | |
2192 | * because they conflict with the pages that had been used before hibernation. | |
f577eb30 | 2193 | */ |
6dbecfd3 | 2194 | static void mark_unsafe_pages(struct memory_bitmap *bm) |
f577eb30 | 2195 | { |
6dbecfd3 | 2196 | unsigned long pfn; |
f577eb30 | 2197 | |
6dbecfd3 RW |
2198 | /* Clear the "free"/"unsafe" bit for all PFNs */ |
2199 | memory_bm_position_reset(free_pages_map); | |
2200 | pfn = memory_bm_next_pfn(free_pages_map); | |
2201 | while (pfn != BM_END_OF_MAP) { | |
2202 | memory_bm_clear_current(free_pages_map); | |
2203 | pfn = memory_bm_next_pfn(free_pages_map); | |
f577eb30 RW |
2204 | } |
2205 | ||
6dbecfd3 RW |
2206 | /* Mark pages that correspond to the "original" PFNs as "unsafe" */ |
2207 | duplicate_memory_bitmap(free_pages_map, bm); | |
f577eb30 | 2208 | |
940864dd | 2209 | allocated_unsafe_pages = 0; |
f577eb30 RW |
2210 | } |
2211 | ||
d307c4a8 | 2212 | static int check_header(struct swsusp_info *info) |
f577eb30 | 2213 | { |
02d7f400 | 2214 | const char *reason; |
f577eb30 | 2215 | |
d307c4a8 | 2216 | reason = check_image_kernel(info); |
0ed5fd13 | 2217 | if (!reason && info->num_physpages != get_num_physpages()) |
f577eb30 | 2218 | reason = "memory size"; |
f577eb30 | 2219 | if (reason) { |
64ec72a1 | 2220 | pr_err("Image mismatch: %s\n", reason); |
f577eb30 RW |
2221 | return -EPERM; |
2222 | } | |
2223 | return 0; | |
2224 | } | |
2225 | ||
2226 | /** | |
ef96f639 | 2227 | * load header - Check the image header and copy the data from it. |
f577eb30 | 2228 | */ |
efd5a852 | 2229 | static int load_header(struct swsusp_info *info) |
f577eb30 RW |
2230 | { |
2231 | int error; | |
f577eb30 | 2232 | |
940864dd | 2233 | restore_pblist = NULL; |
f577eb30 RW |
2234 | error = check_header(info); |
2235 | if (!error) { | |
f577eb30 RW |
2236 | nr_copy_pages = info->image_pages; |
2237 | nr_meta_pages = info->pages - info->image_pages - 1; | |
2238 | } | |
2239 | return error; | |
2240 | } | |
2241 | ||
2242 | /** | |
ef96f639 RW |
2243 | * unpack_orig_pfns - Set bits corresponding to given PFNs in a memory bitmap. |
2244 | * @bm: Memory bitmap. | |
2245 | * @buf: Area of memory containing the PFNs. | |
2246 | * | |
2247 | * For each element of the array pointed to by @buf (1 page at a time), set the | |
2248 | * corresponding bit in @bm. | |
f577eb30 | 2249 | */ |
69643279 | 2250 | static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm) |
f577eb30 RW |
2251 | { |
2252 | int j; | |
2253 | ||
940864dd RW |
2254 | for (j = 0; j < PAGE_SIZE / sizeof(long); j++) { |
2255 | if (unlikely(buf[j] == BM_END_OF_MAP)) | |
2256 | break; | |
2257 | ||
6dbecfd3 | 2258 | if (pfn_valid(buf[j]) && memory_bm_pfn_present(bm, buf[j])) |
69643279 RW |
2259 | memory_bm_set_bit(bm, buf[j]); |
2260 | else | |
2261 | return -EFAULT; | |
f577eb30 | 2262 | } |
69643279 RW |
2263 | |
2264 | return 0; | |
f577eb30 RW |
2265 | } |
2266 | ||
8357376d | 2267 | #ifdef CONFIG_HIGHMEM |
ef96f639 RW |
2268 | /* |
2269 | * struct highmem_pbe is used for creating the list of highmem pages that | |
8357376d RW |
2270 | * should be restored atomically during the resume from disk, because the page |
2271 | * frames they have occupied before the suspend are in use. | |
2272 | */ | |
2273 | struct highmem_pbe { | |
2274 | struct page *copy_page; /* data is here now */ | |
2275 | struct page *orig_page; /* data was here before the suspend */ | |
2276 | struct highmem_pbe *next; | |
2277 | }; | |
2278 | ||
ef96f639 RW |
2279 | /* |
2280 | * List of highmem PBEs needed for restoring the highmem pages that were | |
8357376d RW |
2281 | * allocated before the suspend and included in the suspend image, but have |
2282 | * also been allocated by the "resume" kernel, so their contents cannot be | |
2283 | * written directly to their "original" page frames. | |
2284 | */ | |
2285 | static struct highmem_pbe *highmem_pblist; | |
2286 | ||
2287 | /** | |
ef96f639 RW |
2288 | * count_highmem_image_pages - Compute the number of highmem pages in the image. |
2289 | * @bm: Memory bitmap. | |
2290 | * | |
2291 | * The bits in @bm that correspond to image pages are assumed to be set. | |
8357376d | 2292 | */ |
8357376d RW |
2293 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) |
2294 | { | |
2295 | unsigned long pfn; | |
2296 | unsigned int cnt = 0; | |
2297 | ||
2298 | memory_bm_position_reset(bm); | |
2299 | pfn = memory_bm_next_pfn(bm); | |
2300 | while (pfn != BM_END_OF_MAP) { | |
2301 | if (PageHighMem(pfn_to_page(pfn))) | |
2302 | cnt++; | |
2303 | ||
2304 | pfn = memory_bm_next_pfn(bm); | |
2305 | } | |
2306 | return cnt; | |
2307 | } | |
2308 | ||
8357376d RW |
2309 | static unsigned int safe_highmem_pages; |
2310 | ||
2311 | static struct memory_bitmap *safe_highmem_bm; | |
2312 | ||
ef96f639 RW |
2313 | /** |
2314 | * prepare_highmem_image - Allocate memory for loading highmem data from image. | |
2315 | * @bm: Pointer to an uninitialized memory bitmap structure. | |
2316 | * @nr_highmem_p: Pointer to the number of highmem image pages. | |
2317 | * | |
2318 | * Try to allocate as many highmem pages as there are highmem image pages | |
2319 | * (@nr_highmem_p points to the variable containing the number of highmem image | |
2320 | * pages). The pages that are "safe" (ie. will not be overwritten when the | |
2321 | * hibernation image is restored entirely) have the corresponding bits set in | |
6be2408a | 2322 | * @bm (it must be uninitialized). |
ef96f639 RW |
2323 | * |
2324 | * NOTE: This function should not be called if there are no highmem image pages. | |
2325 | */ | |
efd5a852 RW |
2326 | static int prepare_highmem_image(struct memory_bitmap *bm, |
2327 | unsigned int *nr_highmem_p) | |
8357376d RW |
2328 | { |
2329 | unsigned int to_alloc; | |
2330 | ||
2331 | if (memory_bm_create(bm, GFP_ATOMIC, PG_SAFE)) | |
2332 | return -ENOMEM; | |
2333 | ||
2334 | if (get_highmem_buffer(PG_SAFE)) | |
2335 | return -ENOMEM; | |
2336 | ||
2337 | to_alloc = count_free_highmem_pages(); | |
2338 | if (to_alloc > *nr_highmem_p) | |
2339 | to_alloc = *nr_highmem_p; | |
2340 | else | |
2341 | *nr_highmem_p = to_alloc; | |
2342 | ||
2343 | safe_highmem_pages = 0; | |
2344 | while (to_alloc-- > 0) { | |
2345 | struct page *page; | |
2346 | ||
2347 | page = alloc_page(__GFP_HIGHMEM); | |
7be98234 | 2348 | if (!swsusp_page_is_free(page)) { |
8357376d RW |
2349 | /* The page is "safe", set its bit the bitmap */ |
2350 | memory_bm_set_bit(bm, page_to_pfn(page)); | |
2351 | safe_highmem_pages++; | |
2352 | } | |
2353 | /* Mark the page as allocated */ | |
7be98234 RW |
2354 | swsusp_set_page_forbidden(page); |
2355 | swsusp_set_page_free(page); | |
8357376d RW |
2356 | } |
2357 | memory_bm_position_reset(bm); | |
2358 | safe_highmem_bm = bm; | |
2359 | return 0; | |
2360 | } | |
2361 | ||
ef96f639 RW |
2362 | static struct page *last_highmem_page; |
2363 | ||
8357376d | 2364 | /** |
ef96f639 RW |
2365 | * get_highmem_page_buffer - Prepare a buffer to store a highmem image page. |
2366 | * | |
2367 | * For a given highmem image page get a buffer that suspend_write_next() should | |
2368 | * return to its caller to write to. | |
8357376d | 2369 | * |
ef96f639 RW |
2370 | * If the page is to be saved to its "original" page frame or a copy of |
2371 | * the page is to be made in the highmem, @buffer is returned. Otherwise, | |
2372 | * the copy of the page is to be made in normal memory, so the address of | |
2373 | * the copy is returned. | |
8357376d | 2374 | * |
ef96f639 RW |
2375 | * If @buffer is returned, the caller of suspend_write_next() will write |
2376 | * the page's contents to @buffer, so they will have to be copied to the | |
2377 | * right location on the next call to suspend_write_next() and it is done | |
2378 | * with the help of copy_last_highmem_page(). For this purpose, if | |
2379 | * @buffer is returned, @last_highmem_page is set to the page to which | |
2380 | * the data will have to be copied from @buffer. | |
8357376d | 2381 | */ |
efd5a852 RW |
2382 | static void *get_highmem_page_buffer(struct page *page, |
2383 | struct chain_allocator *ca) | |
8357376d RW |
2384 | { |
2385 | struct highmem_pbe *pbe; | |
2386 | void *kaddr; | |
2387 | ||
7be98234 | 2388 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) { |
ef96f639 RW |
2389 | /* |
2390 | * We have allocated the "original" page frame and we can | |
8357376d RW |
2391 | * use it directly to store the loaded page. |
2392 | */ | |
2393 | last_highmem_page = page; | |
2394 | return buffer; | |
2395 | } | |
ef96f639 RW |
2396 | /* |
2397 | * The "original" page frame has not been allocated and we have to | |
8357376d RW |
2398 | * use a "safe" page frame to store the loaded page. |
2399 | */ | |
2400 | pbe = chain_alloc(ca, sizeof(struct highmem_pbe)); | |
2401 | if (!pbe) { | |
2402 | swsusp_free(); | |
69643279 | 2403 | return ERR_PTR(-ENOMEM); |
8357376d RW |
2404 | } |
2405 | pbe->orig_page = page; | |
2406 | if (safe_highmem_pages > 0) { | |
2407 | struct page *tmp; | |
2408 | ||
2409 | /* Copy of the page will be stored in high memory */ | |
2410 | kaddr = buffer; | |
2411 | tmp = pfn_to_page(memory_bm_next_pfn(safe_highmem_bm)); | |
2412 | safe_highmem_pages--; | |
2413 | last_highmem_page = tmp; | |
2414 | pbe->copy_page = tmp; | |
2415 | } else { | |
2416 | /* Copy of the page will be stored in normal memory */ | |
2417 | kaddr = safe_pages_list; | |
2418 | safe_pages_list = safe_pages_list->next; | |
2419 | pbe->copy_page = virt_to_page(kaddr); | |
2420 | } | |
2421 | pbe->next = highmem_pblist; | |
2422 | highmem_pblist = pbe; | |
2423 | return kaddr; | |
2424 | } | |
2425 | ||
2426 | /** | |
ef96f639 RW |
2427 | * copy_last_highmem_page - Copy most the most recent highmem image page. |
2428 | * | |
2429 | * Copy the contents of a highmem image from @buffer, where the caller of | |
2430 | * snapshot_write_next() has stored them, to the right location represented by | |
2431 | * @last_highmem_page . | |
8357376d | 2432 | */ |
8357376d RW |
2433 | static void copy_last_highmem_page(void) |
2434 | { | |
2435 | if (last_highmem_page) { | |
2436 | void *dst; | |
2437 | ||
0de9a1e2 | 2438 | dst = kmap_atomic(last_highmem_page); |
3ecb01df | 2439 | copy_page(dst, buffer); |
0de9a1e2 | 2440 | kunmap_atomic(dst); |
8357376d RW |
2441 | last_highmem_page = NULL; |
2442 | } | |
2443 | } | |
2444 | ||
2445 | static inline int last_highmem_page_copied(void) | |
2446 | { | |
2447 | return !last_highmem_page; | |
2448 | } | |
2449 | ||
2450 | static inline void free_highmem_data(void) | |
2451 | { | |
2452 | if (safe_highmem_bm) | |
2453 | memory_bm_free(safe_highmem_bm, PG_UNSAFE_CLEAR); | |
2454 | ||
2455 | if (buffer) | |
2456 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2457 | } | |
2458 | #else | |
efd5a852 | 2459 | static unsigned int count_highmem_image_pages(struct memory_bitmap *bm) { return 0; } |
8357376d | 2460 | |
efd5a852 RW |
2461 | static inline int prepare_highmem_image(struct memory_bitmap *bm, |
2462 | unsigned int *nr_highmem_p) { return 0; } | |
8357376d | 2463 | |
efd5a852 RW |
2464 | static inline void *get_highmem_page_buffer(struct page *page, |
2465 | struct chain_allocator *ca) | |
8357376d | 2466 | { |
69643279 | 2467 | return ERR_PTR(-EINVAL); |
8357376d RW |
2468 | } |
2469 | ||
2470 | static inline void copy_last_highmem_page(void) {} | |
2471 | static inline int last_highmem_page_copied(void) { return 1; } | |
2472 | static inline void free_highmem_data(void) {} | |
2473 | #endif /* CONFIG_HIGHMEM */ | |
2474 | ||
ef96f639 RW |
2475 | #define PBES_PER_LINKED_PAGE (LINKED_PAGE_DATA_SIZE / sizeof(struct pbe)) |
2476 | ||
f577eb30 | 2477 | /** |
ef96f639 | 2478 | * prepare_image - Make room for loading hibernation image. |
6be2408a | 2479 | * @new_bm: Uninitialized memory bitmap structure. |
ef96f639 RW |
2480 | * @bm: Memory bitmap with unsafe pages marked. |
2481 | * | |
2482 | * Use @bm to mark the pages that will be overwritten in the process of | |
2483 | * restoring the system memory state from the suspend image ("unsafe" pages) | |
2484 | * and allocate memory for the image. | |
968808b8 | 2485 | * |
ef96f639 RW |
2486 | * The idea is to allocate a new memory bitmap first and then allocate |
2487 | * as many pages as needed for image data, but without specifying what those | |
2488 | * pages will be used for just yet. Instead, we mark them all as allocated and | |
2489 | * create a lists of "safe" pages to be used later. On systems with high | |
2490 | * memory a list of "safe" highmem pages is created too. | |
f577eb30 | 2491 | */ |
efd5a852 | 2492 | static int prepare_image(struct memory_bitmap *new_bm, struct memory_bitmap *bm) |
f577eb30 | 2493 | { |
8357376d | 2494 | unsigned int nr_pages, nr_highmem; |
9c744481 | 2495 | struct linked_page *lp; |
940864dd | 2496 | int error; |
f577eb30 | 2497 | |
8357376d RW |
2498 | /* If there is no highmem, the buffer will not be necessary */ |
2499 | free_image_page(buffer, PG_UNSAFE_CLEAR); | |
2500 | buffer = NULL; | |
2501 | ||
2502 | nr_highmem = count_highmem_image_pages(bm); | |
6dbecfd3 | 2503 | mark_unsafe_pages(bm); |
940864dd RW |
2504 | |
2505 | error = memory_bm_create(new_bm, GFP_ATOMIC, PG_SAFE); | |
2506 | if (error) | |
2507 | goto Free; | |
2508 | ||
2509 | duplicate_memory_bitmap(new_bm, bm); | |
2510 | memory_bm_free(bm, PG_UNSAFE_KEEP); | |
8357376d RW |
2511 | if (nr_highmem > 0) { |
2512 | error = prepare_highmem_image(bm, &nr_highmem); | |
2513 | if (error) | |
2514 | goto Free; | |
2515 | } | |
ef96f639 RW |
2516 | /* |
2517 | * Reserve some safe pages for potential later use. | |
940864dd RW |
2518 | * |
2519 | * NOTE: This way we make sure there will be enough safe pages for the | |
2520 | * chain_alloc() in get_buffer(). It is a bit wasteful, but | |
2521 | * nr_copy_pages cannot be greater than 50% of the memory anyway. | |
9c744481 RW |
2522 | * |
2523 | * nr_copy_pages cannot be less than allocated_unsafe_pages too. | |
940864dd | 2524 | */ |
8357376d | 2525 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2526 | nr_pages = DIV_ROUND_UP(nr_pages, PBES_PER_LINKED_PAGE); |
2527 | while (nr_pages > 0) { | |
8357376d | 2528 | lp = get_image_page(GFP_ATOMIC, PG_SAFE); |
940864dd | 2529 | if (!lp) { |
f577eb30 | 2530 | error = -ENOMEM; |
940864dd RW |
2531 | goto Free; |
2532 | } | |
9c744481 RW |
2533 | lp->next = safe_pages_list; |
2534 | safe_pages_list = lp; | |
940864dd | 2535 | nr_pages--; |
f577eb30 | 2536 | } |
940864dd | 2537 | /* Preallocate memory for the image */ |
8357376d | 2538 | nr_pages = nr_copy_pages - nr_highmem - allocated_unsafe_pages; |
940864dd RW |
2539 | while (nr_pages > 0) { |
2540 | lp = (struct linked_page *)get_zeroed_page(GFP_ATOMIC); | |
2541 | if (!lp) { | |
2542 | error = -ENOMEM; | |
2543 | goto Free; | |
2544 | } | |
7be98234 | 2545 | if (!swsusp_page_is_free(virt_to_page(lp))) { |
940864dd RW |
2546 | /* The page is "safe", add it to the list */ |
2547 | lp->next = safe_pages_list; | |
2548 | safe_pages_list = lp; | |
968808b8 | 2549 | } |
940864dd | 2550 | /* Mark the page as allocated */ |
7be98234 RW |
2551 | swsusp_set_page_forbidden(virt_to_page(lp)); |
2552 | swsusp_set_page_free(virt_to_page(lp)); | |
940864dd | 2553 | nr_pages--; |
968808b8 | 2554 | } |
940864dd RW |
2555 | return 0; |
2556 | ||
59a49335 | 2557 | Free: |
940864dd | 2558 | swsusp_free(); |
f577eb30 RW |
2559 | return error; |
2560 | } | |
2561 | ||
940864dd | 2562 | /** |
ef96f639 RW |
2563 | * get_buffer - Get the address to store the next image data page. |
2564 | * | |
2565 | * Get the address that snapshot_write_next() should return to its caller to | |
2566 | * write to. | |
940864dd | 2567 | */ |
940864dd | 2568 | static void *get_buffer(struct memory_bitmap *bm, struct chain_allocator *ca) |
968808b8 | 2569 | { |
940864dd | 2570 | struct pbe *pbe; |
69643279 RW |
2571 | struct page *page; |
2572 | unsigned long pfn = memory_bm_next_pfn(bm); | |
968808b8 | 2573 | |
69643279 RW |
2574 | if (pfn == BM_END_OF_MAP) |
2575 | return ERR_PTR(-EFAULT); | |
2576 | ||
2577 | page = pfn_to_page(pfn); | |
8357376d RW |
2578 | if (PageHighMem(page)) |
2579 | return get_highmem_page_buffer(page, ca); | |
2580 | ||
7be98234 | 2581 | if (swsusp_page_is_forbidden(page) && swsusp_page_is_free(page)) |
ef96f639 RW |
2582 | /* |
2583 | * We have allocated the "original" page frame and we can | |
940864dd | 2584 | * use it directly to store the loaded page. |
968808b8 | 2585 | */ |
940864dd RW |
2586 | return page_address(page); |
2587 | ||
ef96f639 RW |
2588 | /* |
2589 | * The "original" page frame has not been allocated and we have to | |
940864dd | 2590 | * use a "safe" page frame to store the loaded page. |
968808b8 | 2591 | */ |
940864dd RW |
2592 | pbe = chain_alloc(ca, sizeof(struct pbe)); |
2593 | if (!pbe) { | |
2594 | swsusp_free(); | |
69643279 | 2595 | return ERR_PTR(-ENOMEM); |
940864dd | 2596 | } |
8357376d RW |
2597 | pbe->orig_address = page_address(page); |
2598 | pbe->address = safe_pages_list; | |
940864dd RW |
2599 | safe_pages_list = safe_pages_list->next; |
2600 | pbe->next = restore_pblist; | |
2601 | restore_pblist = pbe; | |
8357376d | 2602 | return pbe->address; |
968808b8 RW |
2603 | } |
2604 | ||
f577eb30 | 2605 | /** |
ef96f639 RW |
2606 | * snapshot_write_next - Get the address to store the next image page. |
2607 | * @handle: Snapshot handle structure to guide the writing. | |
f577eb30 | 2608 | * |
ef96f639 RW |
2609 | * On the first call, @handle should point to a zeroed snapshot_handle |
2610 | * structure. The structure gets populated then and a pointer to it should be | |
2611 | * passed to this function every next time. | |
f577eb30 | 2612 | * |
ef96f639 RW |
2613 | * On success, the function returns a positive number. Then, the caller |
2614 | * is allowed to write up to the returned number of bytes to the memory | |
2615 | * location computed by the data_of() macro. | |
f577eb30 | 2616 | * |
ef96f639 RW |
2617 | * The function returns 0 to indicate the "end of file" condition. Negative |
2618 | * numbers are returned on errors, in which cases the structure pointed to by | |
2619 | * @handle is not updated and should not be used any more. | |
f577eb30 | 2620 | */ |
d3c1b24c | 2621 | int snapshot_write_next(struct snapshot_handle *handle) |
f577eb30 | 2622 | { |
940864dd | 2623 | static struct chain_allocator ca; |
f577eb30 RW |
2624 | int error = 0; |
2625 | ||
940864dd | 2626 | /* Check if we have already loaded the entire image */ |
d3c1b24c | 2627 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) |
f577eb30 | 2628 | return 0; |
940864dd | 2629 | |
d3c1b24c JS |
2630 | handle->sync_read = 1; |
2631 | ||
2632 | if (!handle->cur) { | |
8357376d RW |
2633 | if (!buffer) |
2634 | /* This makes the buffer be freed by swsusp_free() */ | |
2635 | buffer = get_image_page(GFP_ATOMIC, PG_ANY); | |
2636 | ||
f577eb30 RW |
2637 | if (!buffer) |
2638 | return -ENOMEM; | |
8357376d | 2639 | |
f577eb30 | 2640 | handle->buffer = buffer; |
d3c1b24c JS |
2641 | } else if (handle->cur == 1) { |
2642 | error = load_header(buffer); | |
2643 | if (error) | |
2644 | return error; | |
940864dd | 2645 | |
9c744481 RW |
2646 | safe_pages_list = NULL; |
2647 | ||
d3c1b24c JS |
2648 | error = memory_bm_create(©_bm, GFP_ATOMIC, PG_ANY); |
2649 | if (error) | |
2650 | return error; | |
2651 | ||
4c0b6c10 | 2652 | hibernate_restore_protection_begin(); |
d3c1b24c JS |
2653 | } else if (handle->cur <= nr_meta_pages + 1) { |
2654 | error = unpack_orig_pfns(buffer, ©_bm); | |
2655 | if (error) | |
2656 | return error; | |
940864dd | 2657 | |
d3c1b24c JS |
2658 | if (handle->cur == nr_meta_pages + 1) { |
2659 | error = prepare_image(&orig_bm, ©_bm); | |
69643279 RW |
2660 | if (error) |
2661 | return error; | |
2662 | ||
d3c1b24c JS |
2663 | chain_init(&ca, GFP_ATOMIC, PG_SAFE); |
2664 | memory_bm_position_reset(&orig_bm); | |
2665 | restore_pblist = NULL; | |
940864dd | 2666 | handle->buffer = get_buffer(&orig_bm, &ca); |
d3c1b24c | 2667 | handle->sync_read = 0; |
69643279 RW |
2668 | if (IS_ERR(handle->buffer)) |
2669 | return PTR_ERR(handle->buffer); | |
f577eb30 | 2670 | } |
f577eb30 | 2671 | } else { |
d3c1b24c | 2672 | copy_last_highmem_page(); |
4c0b6c10 | 2673 | hibernate_restore_protect_page(handle->buffer); |
d3c1b24c JS |
2674 | handle->buffer = get_buffer(&orig_bm, &ca); |
2675 | if (IS_ERR(handle->buffer)) | |
2676 | return PTR_ERR(handle->buffer); | |
2677 | if (handle->buffer != buffer) | |
2678 | handle->sync_read = 0; | |
f577eb30 | 2679 | } |
d3c1b24c JS |
2680 | handle->cur++; |
2681 | return PAGE_SIZE; | |
f577eb30 RW |
2682 | } |
2683 | ||
8357376d | 2684 | /** |
ef96f639 RW |
2685 | * snapshot_write_finalize - Complete the loading of a hibernation image. |
2686 | * | |
2687 | * Must be called after the last call to snapshot_write_next() in case the last | |
2688 | * page in the image happens to be a highmem page and its contents should be | |
2689 | * stored in highmem. Additionally, it recycles bitmap memory that's not | |
2690 | * necessary any more. | |
8357376d | 2691 | */ |
8357376d RW |
2692 | void snapshot_write_finalize(struct snapshot_handle *handle) |
2693 | { | |
2694 | copy_last_highmem_page(); | |
4c0b6c10 | 2695 | hibernate_restore_protect_page(handle->buffer); |
307c5971 | 2696 | /* Do that only if we have loaded the image entirely */ |
d3c1b24c | 2697 | if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) { |
307c5971 | 2698 | memory_bm_recycle(&orig_bm); |
8357376d RW |
2699 | free_highmem_data(); |
2700 | } | |
2701 | } | |
2702 | ||
f577eb30 RW |
2703 | int snapshot_image_loaded(struct snapshot_handle *handle) |
2704 | { | |
8357376d | 2705 | return !(!nr_copy_pages || !last_highmem_page_copied() || |
940864dd RW |
2706 | handle->cur <= nr_meta_pages + nr_copy_pages); |
2707 | } | |
2708 | ||
8357376d RW |
2709 | #ifdef CONFIG_HIGHMEM |
2710 | /* Assumes that @buf is ready and points to a "safe" page */ | |
efd5a852 RW |
2711 | static inline void swap_two_pages_data(struct page *p1, struct page *p2, |
2712 | void *buf) | |
940864dd | 2713 | { |
8357376d RW |
2714 | void *kaddr1, *kaddr2; |
2715 | ||
0de9a1e2 CW |
2716 | kaddr1 = kmap_atomic(p1); |
2717 | kaddr2 = kmap_atomic(p2); | |
3ecb01df JB |
2718 | copy_page(buf, kaddr1); |
2719 | copy_page(kaddr1, kaddr2); | |
2720 | copy_page(kaddr2, buf); | |
0de9a1e2 CW |
2721 | kunmap_atomic(kaddr2); |
2722 | kunmap_atomic(kaddr1); | |
8357376d RW |
2723 | } |
2724 | ||
2725 | /** | |
ef96f639 RW |
2726 | * restore_highmem - Put highmem image pages into their original locations. |
2727 | * | |
2728 | * For each highmem page that was in use before hibernation and is included in | |
2729 | * the image, and also has been allocated by the "restore" kernel, swap its | |
2730 | * current contents with the previous (ie. "before hibernation") ones. | |
8357376d | 2731 | * |
ef96f639 RW |
2732 | * If the restore eventually fails, we can call this function once again and |
2733 | * restore the highmem state as seen by the restore kernel. | |
8357376d | 2734 | */ |
8357376d RW |
2735 | int restore_highmem(void) |
2736 | { | |
2737 | struct highmem_pbe *pbe = highmem_pblist; | |
2738 | void *buf; | |
2739 | ||
2740 | if (!pbe) | |
2741 | return 0; | |
2742 | ||
2743 | buf = get_image_page(GFP_ATOMIC, PG_SAFE); | |
2744 | if (!buf) | |
2745 | return -ENOMEM; | |
2746 | ||
2747 | while (pbe) { | |
2748 | swap_two_pages_data(pbe->copy_page, pbe->orig_page, buf); | |
2749 | pbe = pbe->next; | |
2750 | } | |
2751 | free_image_page(buf, PG_UNSAFE_CLEAR); | |
2752 | return 0; | |
f577eb30 | 2753 | } |
8357376d | 2754 | #endif /* CONFIG_HIGHMEM */ |