Commit | Line | Data |
---|---|---|
25761b6e | 1 | /* |
96bc7aec | 2 | * linux/kernel/power/snapshot.c |
25761b6e | 3 | * |
96bc7aec | 4 | * This file provide system snapshot/restore functionality. |
25761b6e RW |
5 | * |
6 | * Copyright (C) 1998-2005 Pavel Machek <pavel@suse.cz> | |
7 | * | |
8 | * This file is released under the GPLv2, and is based on swsusp.c. | |
9 | * | |
10 | */ | |
11 | ||
12 | ||
f577eb30 | 13 | #include <linux/version.h> |
25761b6e RW |
14 | #include <linux/module.h> |
15 | #include <linux/mm.h> | |
16 | #include <linux/suspend.h> | |
17 | #include <linux/smp_lock.h> | |
25761b6e | 18 | #include <linux/delay.h> |
25761b6e | 19 | #include <linux/bitops.h> |
25761b6e | 20 | #include <linux/spinlock.h> |
25761b6e | 21 | #include <linux/kernel.h> |
25761b6e RW |
22 | #include <linux/pm.h> |
23 | #include <linux/device.h> | |
25761b6e RW |
24 | #include <linux/bootmem.h> |
25 | #include <linux/syscalls.h> | |
26 | #include <linux/console.h> | |
27 | #include <linux/highmem.h> | |
25761b6e RW |
28 | |
29 | #include <asm/uaccess.h> | |
30 | #include <asm/mmu_context.h> | |
31 | #include <asm/pgtable.h> | |
32 | #include <asm/tlbflush.h> | |
33 | #include <asm/io.h> | |
34 | ||
25761b6e RW |
35 | #include "power.h" |
36 | ||
7088a5c0 | 37 | struct pbe *pagedir_nosave; |
f577eb30 RW |
38 | static unsigned int nr_copy_pages; |
39 | static unsigned int nr_meta_pages; | |
7088a5c0 | 40 | |
25761b6e | 41 | #ifdef CONFIG_HIGHMEM |
72a97e08 RW |
42 | unsigned int count_highmem_pages(void) |
43 | { | |
44 | struct zone *zone; | |
45 | unsigned long zone_pfn; | |
46 | unsigned int n = 0; | |
47 | ||
48 | for_each_zone (zone) | |
49 | if (is_highmem(zone)) { | |
50 | mark_free_pages(zone); | |
51 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; zone_pfn++) { | |
52 | struct page *page; | |
53 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
54 | if (!pfn_valid(pfn)) | |
55 | continue; | |
56 | page = pfn_to_page(pfn); | |
57 | if (PageReserved(page)) | |
58 | continue; | |
59 | if (PageNosaveFree(page)) | |
60 | continue; | |
61 | n++; | |
62 | } | |
63 | } | |
64 | return n; | |
65 | } | |
66 | ||
25761b6e RW |
67 | struct highmem_page { |
68 | char *data; | |
69 | struct page *page; | |
70 | struct highmem_page *next; | |
71 | }; | |
72 | ||
73 | static struct highmem_page *highmem_copy; | |
74 | ||
75 | static int save_highmem_zone(struct zone *zone) | |
76 | { | |
77 | unsigned long zone_pfn; | |
78 | mark_free_pages(zone); | |
79 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
80 | struct page *page; | |
81 | struct highmem_page *save; | |
82 | void *kaddr; | |
83 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
84 | ||
85 | if (!(pfn%1000)) | |
86 | printk("."); | |
87 | if (!pfn_valid(pfn)) | |
88 | continue; | |
89 | page = pfn_to_page(pfn); | |
90 | /* | |
91 | * This condition results from rvmalloc() sans vmalloc_32() | |
92 | * and architectural memory reservations. This should be | |
93 | * corrected eventually when the cases giving rise to this | |
94 | * are better understood. | |
95 | */ | |
c8adb494 | 96 | if (PageReserved(page)) |
25761b6e | 97 | continue; |
25761b6e RW |
98 | BUG_ON(PageNosave(page)); |
99 | if (PageNosaveFree(page)) | |
100 | continue; | |
101 | save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC); | |
102 | if (!save) | |
103 | return -ENOMEM; | |
104 | save->next = highmem_copy; | |
105 | save->page = page; | |
106 | save->data = (void *) get_zeroed_page(GFP_ATOMIC); | |
107 | if (!save->data) { | |
108 | kfree(save); | |
109 | return -ENOMEM; | |
110 | } | |
111 | kaddr = kmap_atomic(page, KM_USER0); | |
112 | memcpy(save->data, kaddr, PAGE_SIZE); | |
113 | kunmap_atomic(kaddr, KM_USER0); | |
114 | highmem_copy = save; | |
115 | } | |
116 | return 0; | |
117 | } | |
25761b6e | 118 | |
0fbeb5a4 | 119 | int save_highmem(void) |
25761b6e | 120 | { |
25761b6e RW |
121 | struct zone *zone; |
122 | int res = 0; | |
123 | ||
124 | pr_debug("swsusp: Saving Highmem\n"); | |
125 | for_each_zone (zone) { | |
126 | if (is_highmem(zone)) | |
127 | res = save_highmem_zone(zone); | |
128 | if (res) | |
129 | return res; | |
130 | } | |
25761b6e RW |
131 | return 0; |
132 | } | |
133 | ||
134 | int restore_highmem(void) | |
135 | { | |
25761b6e RW |
136 | printk("swsusp: Restoring Highmem\n"); |
137 | while (highmem_copy) { | |
138 | struct highmem_page *save = highmem_copy; | |
139 | void *kaddr; | |
140 | highmem_copy = save->next; | |
141 | ||
142 | kaddr = kmap_atomic(save->page, KM_USER0); | |
143 | memcpy(kaddr, save->data, PAGE_SIZE); | |
144 | kunmap_atomic(kaddr, KM_USER0); | |
145 | free_page((long) save->data); | |
146 | kfree(save); | |
147 | } | |
25761b6e RW |
148 | return 0; |
149 | } | |
0fbeb5a4 | 150 | #endif |
25761b6e RW |
151 | |
152 | static int pfn_is_nosave(unsigned long pfn) | |
153 | { | |
154 | unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT; | |
155 | unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT; | |
156 | return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn); | |
157 | } | |
158 | ||
159 | /** | |
160 | * saveable - Determine whether a page should be cloned or not. | |
161 | * @pfn: The page | |
162 | * | |
163 | * We save a page if it's Reserved, and not in the range of pages | |
164 | * statically defined as 'unsaveable', or if it isn't reserved, and | |
165 | * isn't part of a free chunk of pages. | |
166 | */ | |
167 | ||
de491861 | 168 | static int saveable(struct zone *zone, unsigned long *zone_pfn) |
25761b6e RW |
169 | { |
170 | unsigned long pfn = *zone_pfn + zone->zone_start_pfn; | |
de491861 | 171 | struct page *page; |
25761b6e RW |
172 | |
173 | if (!pfn_valid(pfn)) | |
174 | return 0; | |
175 | ||
176 | page = pfn_to_page(pfn); | |
177 | BUG_ON(PageReserved(page) && PageNosave(page)); | |
178 | if (PageNosave(page)) | |
179 | return 0; | |
72a97e08 | 180 | if (PageReserved(page) && pfn_is_nosave(pfn)) |
25761b6e | 181 | return 0; |
25761b6e RW |
182 | if (PageNosaveFree(page)) |
183 | return 0; | |
184 | ||
185 | return 1; | |
186 | } | |
187 | ||
72a97e08 | 188 | unsigned int count_data_pages(void) |
25761b6e RW |
189 | { |
190 | struct zone *zone; | |
191 | unsigned long zone_pfn; | |
dc19d507 | 192 | unsigned int n = 0; |
25761b6e | 193 | |
25761b6e RW |
194 | for_each_zone (zone) { |
195 | if (is_highmem(zone)) | |
196 | continue; | |
197 | mark_free_pages(zone); | |
198 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
a0f49651 | 199 | n += saveable(zone, &zone_pfn); |
25761b6e | 200 | } |
a0f49651 | 201 | return n; |
25761b6e RW |
202 | } |
203 | ||
a0f49651 | 204 | static void copy_data_pages(struct pbe *pblist) |
25761b6e RW |
205 | { |
206 | struct zone *zone; | |
207 | unsigned long zone_pfn; | |
a0f49651 | 208 | struct pbe *pbe, *p; |
25761b6e | 209 | |
a0f49651 | 210 | pbe = pblist; |
25761b6e RW |
211 | for_each_zone (zone) { |
212 | if (is_highmem(zone)) | |
213 | continue; | |
214 | mark_free_pages(zone); | |
215 | /* This is necessary for swsusp_free() */ | |
a0f49651 | 216 | for_each_pb_page (p, pblist) |
25761b6e | 217 | SetPageNosaveFree(virt_to_page(p)); |
a0f49651 | 218 | for_each_pbe (p, pblist) |
25761b6e RW |
219 | SetPageNosaveFree(virt_to_page(p->address)); |
220 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
221 | if (saveable(zone, &zone_pfn)) { | |
de491861 | 222 | struct page *page; |
25761b6e RW |
223 | page = pfn_to_page(zone_pfn + zone->zone_start_pfn); |
224 | BUG_ON(!pbe); | |
225 | pbe->orig_address = (unsigned long)page_address(page); | |
226 | /* copy_page is not usable for copying task structs. */ | |
227 | memcpy((void *)pbe->address, (void *)pbe->orig_address, PAGE_SIZE); | |
228 | pbe = pbe->next; | |
229 | } | |
230 | } | |
231 | } | |
232 | BUG_ON(pbe); | |
233 | } | |
234 | ||
235 | ||
236 | /** | |
237 | * free_pagedir - free pages allocated with alloc_pagedir() | |
238 | */ | |
239 | ||
f577eb30 | 240 | static void free_pagedir(struct pbe *pblist) |
25761b6e RW |
241 | { |
242 | struct pbe *pbe; | |
243 | ||
244 | while (pblist) { | |
245 | pbe = (pblist + PB_PAGE_SKIP)->next; | |
246 | ClearPageNosave(virt_to_page(pblist)); | |
247 | ClearPageNosaveFree(virt_to_page(pblist)); | |
248 | free_page((unsigned long)pblist); | |
249 | pblist = pbe; | |
250 | } | |
251 | } | |
252 | ||
253 | /** | |
254 | * fill_pb_page - Create a list of PBEs on a given memory page | |
255 | */ | |
256 | ||
257 | static inline void fill_pb_page(struct pbe *pbpage) | |
258 | { | |
259 | struct pbe *p; | |
260 | ||
261 | p = pbpage; | |
262 | pbpage += PB_PAGE_SKIP; | |
263 | do | |
264 | p->next = p + 1; | |
265 | while (++p < pbpage); | |
266 | } | |
267 | ||
268 | /** | |
269 | * create_pbe_list - Create a list of PBEs on top of a given chain | |
270 | * of memory pages allocated with alloc_pagedir() | |
271 | */ | |
272 | ||
7088a5c0 | 273 | static inline void create_pbe_list(struct pbe *pblist, unsigned int nr_pages) |
25761b6e RW |
274 | { |
275 | struct pbe *pbpage, *p; | |
dc19d507 | 276 | unsigned int num = PBES_PER_PAGE; |
25761b6e RW |
277 | |
278 | for_each_pb_page (pbpage, pblist) { | |
279 | if (num >= nr_pages) | |
280 | break; | |
281 | ||
282 | fill_pb_page(pbpage); | |
283 | num += PBES_PER_PAGE; | |
284 | } | |
285 | if (pbpage) { | |
286 | for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++) | |
287 | p->next = p + 1; | |
288 | p->next = NULL; | |
289 | } | |
25761b6e RW |
290 | } |
291 | ||
72a97e08 RW |
292 | /** |
293 | * On resume it is necessary to trace and eventually free the unsafe | |
294 | * pages that have been allocated, because they are needed for I/O | |
295 | * (on x86-64 we likely will "eat" these pages once again while | |
296 | * creating the temporary page translation tables) | |
297 | */ | |
298 | ||
299 | struct eaten_page { | |
300 | struct eaten_page *next; | |
301 | char padding[PAGE_SIZE - sizeof(void *)]; | |
302 | }; | |
303 | ||
304 | static struct eaten_page *eaten_pages = NULL; | |
305 | ||
f577eb30 | 306 | static void release_eaten_pages(void) |
72a97e08 RW |
307 | { |
308 | struct eaten_page *p, *q; | |
309 | ||
310 | p = eaten_pages; | |
311 | while (p) { | |
312 | q = p->next; | |
313 | /* We don't want swsusp_free() to free this page again */ | |
314 | ClearPageNosave(virt_to_page(p)); | |
315 | free_page((unsigned long)p); | |
316 | p = q; | |
317 | } | |
318 | eaten_pages = NULL; | |
319 | } | |
320 | ||
054bd4c1 RW |
321 | /** |
322 | * @safe_needed - on resume, for storing the PBE list and the image, | |
323 | * we can only use memory pages that do not conflict with the pages | |
324 | * which had been used before suspend. | |
325 | * | |
326 | * The unsafe pages are marked with the PG_nosave_free flag | |
327 | * | |
328 | * Allocated but unusable (ie eaten) memory pages should be marked | |
329 | * so that swsusp_free() can release them | |
330 | */ | |
331 | ||
332 | static inline void *alloc_image_page(gfp_t gfp_mask, int safe_needed) | |
25761b6e | 333 | { |
054bd4c1 RW |
334 | void *res; |
335 | ||
336 | if (safe_needed) | |
337 | do { | |
338 | res = (void *)get_zeroed_page(gfp_mask); | |
72a97e08 | 339 | if (res && PageNosaveFree(virt_to_page(res))) { |
054bd4c1 RW |
340 | /* This is for swsusp_free() */ |
341 | SetPageNosave(virt_to_page(res)); | |
72a97e08 RW |
342 | ((struct eaten_page *)res)->next = eaten_pages; |
343 | eaten_pages = res; | |
344 | } | |
054bd4c1 RW |
345 | } while (res && PageNosaveFree(virt_to_page(res))); |
346 | else | |
347 | res = (void *)get_zeroed_page(gfp_mask); | |
25761b6e RW |
348 | if (res) { |
349 | SetPageNosave(virt_to_page(res)); | |
350 | SetPageNosaveFree(virt_to_page(res)); | |
351 | } | |
352 | return res; | |
353 | } | |
354 | ||
054bd4c1 RW |
355 | unsigned long get_safe_page(gfp_t gfp_mask) |
356 | { | |
357 | return (unsigned long)alloc_image_page(gfp_mask, 1); | |
358 | } | |
359 | ||
25761b6e RW |
360 | /** |
361 | * alloc_pagedir - Allocate the page directory. | |
362 | * | |
363 | * First, determine exactly how many pages we need and | |
364 | * allocate them. | |
365 | * | |
366 | * We arrange the pages in a chain: each page is an array of PBES_PER_PAGE | |
367 | * struct pbe elements (pbes) and the last element in the page points | |
368 | * to the next page. | |
369 | * | |
370 | * On each page we set up a list of struct_pbe elements. | |
371 | */ | |
372 | ||
054bd4c1 | 373 | struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, int safe_needed) |
25761b6e | 374 | { |
dc19d507 | 375 | unsigned int num; |
25761b6e RW |
376 | struct pbe *pblist, *pbe; |
377 | ||
378 | if (!nr_pages) | |
379 | return NULL; | |
380 | ||
054bd4c1 | 381 | pblist = alloc_image_page(gfp_mask, safe_needed); |
25761b6e RW |
382 | /* FIXME: rewrite this ugly loop */ |
383 | for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages; | |
384 | pbe = pbe->next, num += PBES_PER_PAGE) { | |
385 | pbe += PB_PAGE_SKIP; | |
054bd4c1 | 386 | pbe->next = alloc_image_page(gfp_mask, safe_needed); |
25761b6e RW |
387 | } |
388 | if (!pbe) { /* get_zeroed_page() failed */ | |
389 | free_pagedir(pblist); | |
390 | pblist = NULL; | |
7088a5c0 RW |
391 | } else |
392 | create_pbe_list(pblist, nr_pages); | |
25761b6e RW |
393 | return pblist; |
394 | } | |
395 | ||
396 | /** | |
397 | * Free pages we allocated for suspend. Suspend pages are alocated | |
398 | * before atomic copy, so we need to free them after resume. | |
399 | */ | |
400 | ||
401 | void swsusp_free(void) | |
402 | { | |
403 | struct zone *zone; | |
404 | unsigned long zone_pfn; | |
405 | ||
406 | for_each_zone(zone) { | |
407 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
408 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) { | |
dc19d507 | 409 | struct page *page; |
25761b6e RW |
410 | page = pfn_to_page(zone_pfn + zone->zone_start_pfn); |
411 | if (PageNosave(page) && PageNosaveFree(page)) { | |
412 | ClearPageNosave(page); | |
413 | ClearPageNosaveFree(page); | |
414 | free_page((long) page_address(page)); | |
415 | } | |
416 | } | |
417 | } | |
f577eb30 RW |
418 | nr_copy_pages = 0; |
419 | nr_meta_pages = 0; | |
420 | pagedir_nosave = NULL; | |
25761b6e RW |
421 | } |
422 | ||
423 | ||
424 | /** | |
425 | * enough_free_mem - Make sure we enough free memory to snapshot. | |
426 | * | |
427 | * Returns TRUE or FALSE after checking the number of available | |
428 | * free pages. | |
429 | */ | |
430 | ||
dc19d507 | 431 | static int enough_free_mem(unsigned int nr_pages) |
25761b6e | 432 | { |
e5e2fa78 RW |
433 | struct zone *zone; |
434 | unsigned int n = 0; | |
435 | ||
436 | for_each_zone (zone) | |
437 | if (!is_highmem(zone)) | |
438 | n += zone->free_pages; | |
439 | pr_debug("swsusp: available memory: %u pages\n", n); | |
440 | return n > (nr_pages + PAGES_FOR_IO + | |
a0f49651 | 441 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE); |
25761b6e RW |
442 | } |
443 | ||
f577eb30 | 444 | static int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed) |
054bd4c1 RW |
445 | { |
446 | struct pbe *p; | |
447 | ||
448 | for_each_pbe (p, pblist) { | |
449 | p->address = (unsigned long)alloc_image_page(gfp_mask, safe_needed); | |
450 | if (!p->address) | |
451 | return -ENOMEM; | |
452 | } | |
453 | return 0; | |
454 | } | |
25761b6e | 455 | |
dc19d507 | 456 | static struct pbe *swsusp_alloc(unsigned int nr_pages) |
25761b6e | 457 | { |
054bd4c1 | 458 | struct pbe *pblist; |
25761b6e | 459 | |
054bd4c1 | 460 | if (!(pblist = alloc_pagedir(nr_pages, GFP_ATOMIC | __GFP_COLD, 0))) { |
25761b6e | 461 | printk(KERN_ERR "suspend: Allocating pagedir failed.\n"); |
a0f49651 | 462 | return NULL; |
25761b6e | 463 | } |
25761b6e | 464 | |
054bd4c1 RW |
465 | if (alloc_data_pages(pblist, GFP_ATOMIC | __GFP_COLD, 0)) { |
466 | printk(KERN_ERR "suspend: Allocating image pages failed.\n"); | |
467 | swsusp_free(); | |
468 | return NULL; | |
25761b6e RW |
469 | } |
470 | ||
a0f49651 | 471 | return pblist; |
25761b6e RW |
472 | } |
473 | ||
2e32a43e | 474 | asmlinkage int swsusp_save(void) |
25761b6e | 475 | { |
dc19d507 | 476 | unsigned int nr_pages; |
25761b6e RW |
477 | |
478 | pr_debug("swsusp: critical section: \n"); | |
25761b6e RW |
479 | |
480 | drain_local_pages(); | |
a0f49651 RW |
481 | nr_pages = count_data_pages(); |
482 | printk("swsusp: Need to copy %u pages\n", nr_pages); | |
25761b6e RW |
483 | |
484 | pr_debug("swsusp: pages needed: %u + %lu + %u, free: %u\n", | |
a0f49651 RW |
485 | nr_pages, |
486 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE, | |
25761b6e RW |
487 | PAGES_FOR_IO, nr_free_pages()); |
488 | ||
a0f49651 | 489 | if (!enough_free_mem(nr_pages)) { |
25761b6e RW |
490 | printk(KERN_ERR "swsusp: Not enough free memory\n"); |
491 | return -ENOMEM; | |
492 | } | |
493 | ||
a0f49651 RW |
494 | pagedir_nosave = swsusp_alloc(nr_pages); |
495 | if (!pagedir_nosave) | |
496 | return -ENOMEM; | |
25761b6e RW |
497 | |
498 | /* During allocating of suspend pagedir, new cold pages may appear. | |
499 | * Kill them. | |
500 | */ | |
501 | drain_local_pages(); | |
a0f49651 | 502 | copy_data_pages(pagedir_nosave); |
25761b6e RW |
503 | |
504 | /* | |
505 | * End of critical section. From now on, we can write to memory, | |
506 | * but we should not touch disk. This specially means we must _not_ | |
507 | * touch swap space! Except we must write out our image of course. | |
508 | */ | |
509 | ||
a0f49651 | 510 | nr_copy_pages = nr_pages; |
f577eb30 | 511 | nr_meta_pages = (nr_pages * sizeof(long) + PAGE_SIZE - 1) >> PAGE_SHIFT; |
a0f49651 RW |
512 | |
513 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); | |
25761b6e RW |
514 | return 0; |
515 | } | |
f577eb30 RW |
516 | |
517 | static void init_header(struct swsusp_info *info) | |
518 | { | |
519 | memset(info, 0, sizeof(struct swsusp_info)); | |
520 | info->version_code = LINUX_VERSION_CODE; | |
521 | info->num_physpages = num_physpages; | |
522 | memcpy(&info->uts, &system_utsname, sizeof(system_utsname)); | |
523 | info->cpus = num_online_cpus(); | |
524 | info->image_pages = nr_copy_pages; | |
525 | info->pages = nr_copy_pages + nr_meta_pages + 1; | |
526 | } | |
527 | ||
528 | /** | |
529 | * pack_orig_addresses - the .orig_address fields of the PBEs from the | |
530 | * list starting at @pbe are stored in the array @buf[] (1 page) | |
531 | */ | |
532 | ||
533 | static inline struct pbe *pack_orig_addresses(unsigned long *buf, struct pbe *pbe) | |
534 | { | |
535 | int j; | |
536 | ||
537 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | |
538 | buf[j] = pbe->orig_address; | |
539 | pbe = pbe->next; | |
540 | } | |
541 | if (!pbe) | |
542 | for (; j < PAGE_SIZE / sizeof(long); j++) | |
543 | buf[j] = 0; | |
544 | return pbe; | |
545 | } | |
546 | ||
547 | /** | |
548 | * snapshot_read_next - used for reading the system memory snapshot. | |
549 | * | |
550 | * On the first call to it @handle should point to a zeroed | |
551 | * snapshot_handle structure. The structure gets updated and a pointer | |
552 | * to it should be passed to this function every next time. | |
553 | * | |
554 | * The @count parameter should contain the number of bytes the caller | |
555 | * wants to read from the snapshot. It must not be zero. | |
556 | * | |
557 | * On success the function returns a positive number. Then, the caller | |
558 | * is allowed to read up to the returned number of bytes from the memory | |
559 | * location computed by the data_of() macro. The number returned | |
560 | * may be smaller than @count, but this only happens if the read would | |
561 | * cross a page boundary otherwise. | |
562 | * | |
563 | * The function returns 0 to indicate the end of data stream condition, | |
564 | * and a negative number is returned on error. In such cases the | |
565 | * structure pointed to by @handle is not updated and should not be used | |
566 | * any more. | |
567 | */ | |
568 | ||
569 | int snapshot_read_next(struct snapshot_handle *handle, size_t count) | |
570 | { | |
571 | static unsigned long *buffer; | |
572 | ||
573 | if (handle->page > nr_meta_pages + nr_copy_pages) | |
574 | return 0; | |
575 | if (!buffer) { | |
576 | /* This makes the buffer be freed by swsusp_free() */ | |
577 | buffer = alloc_image_page(GFP_ATOMIC, 0); | |
578 | if (!buffer) | |
579 | return -ENOMEM; | |
580 | } | |
581 | if (!handle->offset) { | |
582 | init_header((struct swsusp_info *)buffer); | |
583 | handle->buffer = buffer; | |
584 | handle->pbe = pagedir_nosave; | |
585 | } | |
586 | if (handle->prev < handle->page) { | |
587 | if (handle->page <= nr_meta_pages) { | |
588 | handle->pbe = pack_orig_addresses(buffer, handle->pbe); | |
589 | if (!handle->pbe) | |
590 | handle->pbe = pagedir_nosave; | |
591 | } else { | |
592 | handle->buffer = (void *)handle->pbe->address; | |
593 | handle->pbe = handle->pbe->next; | |
594 | } | |
595 | handle->prev = handle->page; | |
596 | } | |
597 | handle->buf_offset = handle->page_offset; | |
598 | if (handle->page_offset + count >= PAGE_SIZE) { | |
599 | count = PAGE_SIZE - handle->page_offset; | |
600 | handle->page_offset = 0; | |
601 | handle->page++; | |
602 | } else { | |
603 | handle->page_offset += count; | |
604 | } | |
605 | handle->offset += count; | |
606 | return count; | |
607 | } | |
608 | ||
609 | /** | |
610 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
611 | * the image during resume, because they conflict with the pages that | |
612 | * had been used before suspend | |
613 | */ | |
614 | ||
615 | static int mark_unsafe_pages(struct pbe *pblist) | |
616 | { | |
617 | struct zone *zone; | |
618 | unsigned long zone_pfn; | |
619 | struct pbe *p; | |
620 | ||
621 | if (!pblist) /* a sanity check */ | |
622 | return -EINVAL; | |
623 | ||
624 | /* Clear page flags */ | |
625 | for_each_zone (zone) { | |
626 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
627 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) | |
628 | ClearPageNosaveFree(pfn_to_page(zone_pfn + | |
629 | zone->zone_start_pfn)); | |
630 | } | |
631 | ||
632 | /* Mark orig addresses */ | |
633 | for_each_pbe (p, pblist) { | |
634 | if (virt_addr_valid(p->orig_address)) | |
635 | SetPageNosaveFree(virt_to_page(p->orig_address)); | |
636 | else | |
637 | return -EFAULT; | |
638 | } | |
639 | ||
640 | return 0; | |
641 | } | |
642 | ||
643 | static void copy_page_backup_list(struct pbe *dst, struct pbe *src) | |
644 | { | |
645 | /* We assume both lists contain the same number of elements */ | |
646 | while (src) { | |
647 | dst->orig_address = src->orig_address; | |
648 | dst = dst->next; | |
649 | src = src->next; | |
650 | } | |
651 | } | |
652 | ||
653 | static int check_header(struct swsusp_info *info) | |
654 | { | |
655 | char *reason = NULL; | |
656 | ||
657 | if (info->version_code != LINUX_VERSION_CODE) | |
658 | reason = "kernel version"; | |
659 | if (info->num_physpages != num_physpages) | |
660 | reason = "memory size"; | |
661 | if (strcmp(info->uts.sysname,system_utsname.sysname)) | |
662 | reason = "system type"; | |
663 | if (strcmp(info->uts.release,system_utsname.release)) | |
664 | reason = "kernel release"; | |
665 | if (strcmp(info->uts.version,system_utsname.version)) | |
666 | reason = "version"; | |
667 | if (strcmp(info->uts.machine,system_utsname.machine)) | |
668 | reason = "machine"; | |
669 | if (reason) { | |
670 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | |
671 | return -EPERM; | |
672 | } | |
673 | return 0; | |
674 | } | |
675 | ||
676 | /** | |
677 | * load header - check the image header and copy data from it | |
678 | */ | |
679 | ||
680 | static int load_header(struct snapshot_handle *handle, | |
681 | struct swsusp_info *info) | |
682 | { | |
683 | int error; | |
684 | struct pbe *pblist; | |
685 | ||
686 | error = check_header(info); | |
687 | if (!error) { | |
688 | pblist = alloc_pagedir(info->image_pages, GFP_ATOMIC, 0); | |
689 | if (!pblist) | |
690 | return -ENOMEM; | |
691 | pagedir_nosave = pblist; | |
692 | handle->pbe = pblist; | |
693 | nr_copy_pages = info->image_pages; | |
694 | nr_meta_pages = info->pages - info->image_pages - 1; | |
695 | } | |
696 | return error; | |
697 | } | |
698 | ||
699 | /** | |
700 | * unpack_orig_addresses - copy the elements of @buf[] (1 page) to | |
701 | * the PBEs in the list starting at @pbe | |
702 | */ | |
703 | ||
704 | static inline struct pbe *unpack_orig_addresses(unsigned long *buf, | |
705 | struct pbe *pbe) | |
706 | { | |
707 | int j; | |
708 | ||
709 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | |
710 | pbe->orig_address = buf[j]; | |
711 | pbe = pbe->next; | |
712 | } | |
713 | return pbe; | |
714 | } | |
715 | ||
716 | /** | |
717 | * create_image - use metadata contained in the PBE list | |
718 | * pointed to by pagedir_nosave to mark the pages that will | |
719 | * be overwritten in the process of restoring the system | |
720 | * memory state from the image and allocate memory for | |
721 | * the image avoiding these pages | |
722 | */ | |
723 | ||
724 | static int create_image(struct snapshot_handle *handle) | |
725 | { | |
726 | int error = 0; | |
727 | struct pbe *p, *pblist; | |
728 | ||
729 | p = pagedir_nosave; | |
730 | error = mark_unsafe_pages(p); | |
731 | if (!error) { | |
732 | pblist = alloc_pagedir(nr_copy_pages, GFP_ATOMIC, 1); | |
733 | if (pblist) | |
734 | copy_page_backup_list(pblist, p); | |
735 | free_pagedir(p); | |
736 | if (!pblist) | |
737 | error = -ENOMEM; | |
738 | } | |
739 | if (!error) | |
740 | error = alloc_data_pages(pblist, GFP_ATOMIC, 1); | |
741 | if (!error) { | |
742 | release_eaten_pages(); | |
743 | pagedir_nosave = pblist; | |
744 | } else { | |
745 | pagedir_nosave = NULL; | |
746 | handle->pbe = NULL; | |
747 | nr_copy_pages = 0; | |
748 | nr_meta_pages = 0; | |
749 | } | |
750 | return error; | |
751 | } | |
752 | ||
753 | /** | |
754 | * snapshot_write_next - used for writing the system memory snapshot. | |
755 | * | |
756 | * On the first call to it @handle should point to a zeroed | |
757 | * snapshot_handle structure. The structure gets updated and a pointer | |
758 | * to it should be passed to this function every next time. | |
759 | * | |
760 | * The @count parameter should contain the number of bytes the caller | |
761 | * wants to write to the image. It must not be zero. | |
762 | * | |
763 | * On success the function returns a positive number. Then, the caller | |
764 | * is allowed to write up to the returned number of bytes to the memory | |
765 | * location computed by the data_of() macro. The number returned | |
766 | * may be smaller than @count, but this only happens if the write would | |
767 | * cross a page boundary otherwise. | |
768 | * | |
769 | * The function returns 0 to indicate the "end of file" condition, | |
770 | * and a negative number is returned on error. In such cases the | |
771 | * structure pointed to by @handle is not updated and should not be used | |
772 | * any more. | |
773 | */ | |
774 | ||
775 | int snapshot_write_next(struct snapshot_handle *handle, size_t count) | |
776 | { | |
777 | static unsigned long *buffer; | |
778 | int error = 0; | |
779 | ||
780 | if (handle->prev && handle->page > nr_meta_pages + nr_copy_pages) | |
781 | return 0; | |
782 | if (!buffer) { | |
783 | /* This makes the buffer be freed by swsusp_free() */ | |
784 | buffer = alloc_image_page(GFP_ATOMIC, 0); | |
785 | if (!buffer) | |
786 | return -ENOMEM; | |
787 | } | |
788 | if (!handle->offset) | |
789 | handle->buffer = buffer; | |
790 | if (handle->prev < handle->page) { | |
791 | if (!handle->prev) { | |
792 | error = load_header(handle, (struct swsusp_info *)buffer); | |
793 | if (error) | |
794 | return error; | |
795 | } else if (handle->prev <= nr_meta_pages) { | |
796 | handle->pbe = unpack_orig_addresses(buffer, handle->pbe); | |
797 | if (!handle->pbe) { | |
798 | error = create_image(handle); | |
799 | if (error) | |
800 | return error; | |
801 | handle->pbe = pagedir_nosave; | |
802 | handle->buffer = (void *)handle->pbe->address; | |
803 | } | |
804 | } else { | |
805 | handle->pbe = handle->pbe->next; | |
806 | handle->buffer = (void *)handle->pbe->address; | |
807 | } | |
808 | handle->prev = handle->page; | |
809 | } | |
810 | handle->buf_offset = handle->page_offset; | |
811 | if (handle->page_offset + count >= PAGE_SIZE) { | |
812 | count = PAGE_SIZE - handle->page_offset; | |
813 | handle->page_offset = 0; | |
814 | handle->page++; | |
815 | } else { | |
816 | handle->page_offset += count; | |
817 | } | |
818 | handle->offset += count; | |
819 | return count; | |
820 | } | |
821 | ||
822 | int snapshot_image_loaded(struct snapshot_handle *handle) | |
823 | { | |
824 | return !(!handle->pbe || handle->pbe->next || !nr_copy_pages || | |
825 | handle->page <= nr_meta_pages + nr_copy_pages); | |
826 | } |