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; | |
6e1819d6 | 40 | static unsigned long *buffer; |
7088a5c0 | 41 | |
25761b6e | 42 | #ifdef CONFIG_HIGHMEM |
72a97e08 RW |
43 | unsigned int count_highmem_pages(void) |
44 | { | |
45 | struct zone *zone; | |
46 | unsigned long zone_pfn; | |
47 | unsigned int n = 0; | |
48 | ||
49 | for_each_zone (zone) | |
50 | if (is_highmem(zone)) { | |
51 | mark_free_pages(zone); | |
52 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; zone_pfn++) { | |
53 | struct page *page; | |
54 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
55 | if (!pfn_valid(pfn)) | |
56 | continue; | |
57 | page = pfn_to_page(pfn); | |
58 | if (PageReserved(page)) | |
59 | continue; | |
60 | if (PageNosaveFree(page)) | |
61 | continue; | |
62 | n++; | |
63 | } | |
64 | } | |
65 | return n; | |
66 | } | |
67 | ||
25761b6e RW |
68 | struct highmem_page { |
69 | char *data; | |
70 | struct page *page; | |
71 | struct highmem_page *next; | |
72 | }; | |
73 | ||
74 | static struct highmem_page *highmem_copy; | |
75 | ||
76 | static int save_highmem_zone(struct zone *zone) | |
77 | { | |
78 | unsigned long zone_pfn; | |
79 | mark_free_pages(zone); | |
80 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
81 | struct page *page; | |
82 | struct highmem_page *save; | |
83 | void *kaddr; | |
84 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
85 | ||
ce6ed29f | 86 | if (!(pfn%10000)) |
25761b6e RW |
87 | printk("."); |
88 | if (!pfn_valid(pfn)) | |
89 | continue; | |
90 | page = pfn_to_page(pfn); | |
91 | /* | |
92 | * This condition results from rvmalloc() sans vmalloc_32() | |
93 | * and architectural memory reservations. This should be | |
94 | * corrected eventually when the cases giving rise to this | |
95 | * are better understood. | |
96 | */ | |
c8adb494 | 97 | if (PageReserved(page)) |
25761b6e | 98 | continue; |
25761b6e RW |
99 | BUG_ON(PageNosave(page)); |
100 | if (PageNosaveFree(page)) | |
101 | continue; | |
102 | save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC); | |
103 | if (!save) | |
104 | return -ENOMEM; | |
105 | save->next = highmem_copy; | |
106 | save->page = page; | |
107 | save->data = (void *) get_zeroed_page(GFP_ATOMIC); | |
108 | if (!save->data) { | |
109 | kfree(save); | |
110 | return -ENOMEM; | |
111 | } | |
112 | kaddr = kmap_atomic(page, KM_USER0); | |
113 | memcpy(save->data, kaddr, PAGE_SIZE); | |
114 | kunmap_atomic(kaddr, KM_USER0); | |
115 | highmem_copy = save; | |
116 | } | |
117 | return 0; | |
118 | } | |
25761b6e | 119 | |
0fbeb5a4 | 120 | int save_highmem(void) |
25761b6e | 121 | { |
25761b6e RW |
122 | struct zone *zone; |
123 | int res = 0; | |
124 | ||
ce6ed29f | 125 | pr_debug("swsusp: Saving Highmem"); |
e4e4d665 | 126 | drain_local_pages(); |
25761b6e RW |
127 | for_each_zone (zone) { |
128 | if (is_highmem(zone)) | |
129 | res = save_highmem_zone(zone); | |
130 | if (res) | |
131 | return res; | |
132 | } | |
ce6ed29f | 133 | printk("\n"); |
25761b6e RW |
134 | return 0; |
135 | } | |
136 | ||
137 | int restore_highmem(void) | |
138 | { | |
25761b6e RW |
139 | printk("swsusp: Restoring Highmem\n"); |
140 | while (highmem_copy) { | |
141 | struct highmem_page *save = highmem_copy; | |
142 | void *kaddr; | |
143 | highmem_copy = save->next; | |
144 | ||
145 | kaddr = kmap_atomic(save->page, KM_USER0); | |
146 | memcpy(kaddr, save->data, PAGE_SIZE); | |
147 | kunmap_atomic(kaddr, KM_USER0); | |
148 | free_page((long) save->data); | |
149 | kfree(save); | |
150 | } | |
25761b6e RW |
151 | return 0; |
152 | } | |
0fbeb5a4 | 153 | #endif |
25761b6e RW |
154 | |
155 | static int pfn_is_nosave(unsigned long pfn) | |
156 | { | |
157 | unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT; | |
158 | unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT; | |
159 | return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn); | |
160 | } | |
161 | ||
162 | /** | |
163 | * saveable - Determine whether a page should be cloned or not. | |
164 | * @pfn: The page | |
165 | * | |
166 | * We save a page if it's Reserved, and not in the range of pages | |
167 | * statically defined as 'unsaveable', or if it isn't reserved, and | |
168 | * isn't part of a free chunk of pages. | |
169 | */ | |
170 | ||
de491861 | 171 | static int saveable(struct zone *zone, unsigned long *zone_pfn) |
25761b6e RW |
172 | { |
173 | unsigned long pfn = *zone_pfn + zone->zone_start_pfn; | |
de491861 | 174 | struct page *page; |
25761b6e RW |
175 | |
176 | if (!pfn_valid(pfn)) | |
177 | return 0; | |
178 | ||
179 | page = pfn_to_page(pfn); | |
180 | BUG_ON(PageReserved(page) && PageNosave(page)); | |
181 | if (PageNosave(page)) | |
182 | return 0; | |
72a97e08 | 183 | if (PageReserved(page) && pfn_is_nosave(pfn)) |
25761b6e | 184 | return 0; |
25761b6e RW |
185 | if (PageNosaveFree(page)) |
186 | return 0; | |
187 | ||
188 | return 1; | |
189 | } | |
190 | ||
72a97e08 | 191 | unsigned int count_data_pages(void) |
25761b6e RW |
192 | { |
193 | struct zone *zone; | |
194 | unsigned long zone_pfn; | |
dc19d507 | 195 | unsigned int n = 0; |
25761b6e | 196 | |
25761b6e RW |
197 | for_each_zone (zone) { |
198 | if (is_highmem(zone)) | |
199 | continue; | |
200 | mark_free_pages(zone); | |
201 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
a0f49651 | 202 | n += saveable(zone, &zone_pfn); |
25761b6e | 203 | } |
a0f49651 | 204 | return n; |
25761b6e RW |
205 | } |
206 | ||
a0f49651 | 207 | static void copy_data_pages(struct pbe *pblist) |
25761b6e RW |
208 | { |
209 | struct zone *zone; | |
210 | unsigned long zone_pfn; | |
a0f49651 | 211 | struct pbe *pbe, *p; |
25761b6e | 212 | |
a0f49651 | 213 | pbe = pblist; |
25761b6e RW |
214 | for_each_zone (zone) { |
215 | if (is_highmem(zone)) | |
216 | continue; | |
217 | mark_free_pages(zone); | |
218 | /* This is necessary for swsusp_free() */ | |
a0f49651 | 219 | for_each_pb_page (p, pblist) |
25761b6e | 220 | SetPageNosaveFree(virt_to_page(p)); |
a0f49651 | 221 | for_each_pbe (p, pblist) |
25761b6e RW |
222 | SetPageNosaveFree(virt_to_page(p->address)); |
223 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
224 | if (saveable(zone, &zone_pfn)) { | |
de491861 | 225 | struct page *page; |
25761b6e RW |
226 | page = pfn_to_page(zone_pfn + zone->zone_start_pfn); |
227 | BUG_ON(!pbe); | |
228 | pbe->orig_address = (unsigned long)page_address(page); | |
229 | /* copy_page is not usable for copying task structs. */ | |
230 | memcpy((void *)pbe->address, (void *)pbe->orig_address, PAGE_SIZE); | |
231 | pbe = pbe->next; | |
232 | } | |
233 | } | |
234 | } | |
235 | BUG_ON(pbe); | |
236 | } | |
237 | ||
238 | ||
239 | /** | |
240 | * free_pagedir - free pages allocated with alloc_pagedir() | |
241 | */ | |
242 | ||
f577eb30 | 243 | static void free_pagedir(struct pbe *pblist) |
25761b6e RW |
244 | { |
245 | struct pbe *pbe; | |
246 | ||
247 | while (pblist) { | |
248 | pbe = (pblist + PB_PAGE_SKIP)->next; | |
249 | ClearPageNosave(virt_to_page(pblist)); | |
250 | ClearPageNosaveFree(virt_to_page(pblist)); | |
251 | free_page((unsigned long)pblist); | |
252 | pblist = pbe; | |
253 | } | |
254 | } | |
255 | ||
256 | /** | |
257 | * fill_pb_page - Create a list of PBEs on a given memory page | |
258 | */ | |
259 | ||
260 | static inline void fill_pb_page(struct pbe *pbpage) | |
261 | { | |
262 | struct pbe *p; | |
263 | ||
264 | p = pbpage; | |
265 | pbpage += PB_PAGE_SKIP; | |
266 | do | |
267 | p->next = p + 1; | |
268 | while (++p < pbpage); | |
269 | } | |
270 | ||
271 | /** | |
272 | * create_pbe_list - Create a list of PBEs on top of a given chain | |
273 | * of memory pages allocated with alloc_pagedir() | |
274 | */ | |
275 | ||
7088a5c0 | 276 | static inline void create_pbe_list(struct pbe *pblist, unsigned int nr_pages) |
25761b6e RW |
277 | { |
278 | struct pbe *pbpage, *p; | |
dc19d507 | 279 | unsigned int num = PBES_PER_PAGE; |
25761b6e RW |
280 | |
281 | for_each_pb_page (pbpage, pblist) { | |
282 | if (num >= nr_pages) | |
283 | break; | |
284 | ||
285 | fill_pb_page(pbpage); | |
286 | num += PBES_PER_PAGE; | |
287 | } | |
288 | if (pbpage) { | |
289 | for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++) | |
290 | p->next = p + 1; | |
291 | p->next = NULL; | |
292 | } | |
25761b6e RW |
293 | } |
294 | ||
72a97e08 RW |
295 | /** |
296 | * On resume it is necessary to trace and eventually free the unsafe | |
297 | * pages that have been allocated, because they are needed for I/O | |
298 | * (on x86-64 we likely will "eat" these pages once again while | |
299 | * creating the temporary page translation tables) | |
300 | */ | |
301 | ||
302 | struct eaten_page { | |
303 | struct eaten_page *next; | |
304 | char padding[PAGE_SIZE - sizeof(void *)]; | |
305 | }; | |
306 | ||
307 | static struct eaten_page *eaten_pages = NULL; | |
308 | ||
f577eb30 | 309 | static void release_eaten_pages(void) |
72a97e08 RW |
310 | { |
311 | struct eaten_page *p, *q; | |
312 | ||
313 | p = eaten_pages; | |
314 | while (p) { | |
315 | q = p->next; | |
316 | /* We don't want swsusp_free() to free this page again */ | |
317 | ClearPageNosave(virt_to_page(p)); | |
318 | free_page((unsigned long)p); | |
319 | p = q; | |
320 | } | |
321 | eaten_pages = NULL; | |
322 | } | |
323 | ||
054bd4c1 RW |
324 | /** |
325 | * @safe_needed - on resume, for storing the PBE list and the image, | |
326 | * we can only use memory pages that do not conflict with the pages | |
327 | * which had been used before suspend. | |
328 | * | |
329 | * The unsafe pages are marked with the PG_nosave_free flag | |
330 | * | |
331 | * Allocated but unusable (ie eaten) memory pages should be marked | |
332 | * so that swsusp_free() can release them | |
333 | */ | |
334 | ||
335 | static inline void *alloc_image_page(gfp_t gfp_mask, int safe_needed) | |
25761b6e | 336 | { |
054bd4c1 RW |
337 | void *res; |
338 | ||
339 | if (safe_needed) | |
340 | do { | |
341 | res = (void *)get_zeroed_page(gfp_mask); | |
72a97e08 | 342 | if (res && PageNosaveFree(virt_to_page(res))) { |
054bd4c1 RW |
343 | /* This is for swsusp_free() */ |
344 | SetPageNosave(virt_to_page(res)); | |
72a97e08 RW |
345 | ((struct eaten_page *)res)->next = eaten_pages; |
346 | eaten_pages = res; | |
347 | } | |
054bd4c1 RW |
348 | } while (res && PageNosaveFree(virt_to_page(res))); |
349 | else | |
350 | res = (void *)get_zeroed_page(gfp_mask); | |
25761b6e RW |
351 | if (res) { |
352 | SetPageNosave(virt_to_page(res)); | |
353 | SetPageNosaveFree(virt_to_page(res)); | |
354 | } | |
355 | return res; | |
356 | } | |
357 | ||
054bd4c1 RW |
358 | unsigned long get_safe_page(gfp_t gfp_mask) |
359 | { | |
360 | return (unsigned long)alloc_image_page(gfp_mask, 1); | |
361 | } | |
362 | ||
25761b6e RW |
363 | /** |
364 | * alloc_pagedir - Allocate the page directory. | |
365 | * | |
366 | * First, determine exactly how many pages we need and | |
367 | * allocate them. | |
368 | * | |
369 | * We arrange the pages in a chain: each page is an array of PBES_PER_PAGE | |
370 | * struct pbe elements (pbes) and the last element in the page points | |
371 | * to the next page. | |
372 | * | |
373 | * On each page we set up a list of struct_pbe elements. | |
374 | */ | |
375 | ||
054bd4c1 | 376 | struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, int safe_needed) |
25761b6e | 377 | { |
dc19d507 | 378 | unsigned int num; |
25761b6e RW |
379 | struct pbe *pblist, *pbe; |
380 | ||
381 | if (!nr_pages) | |
382 | return NULL; | |
383 | ||
054bd4c1 | 384 | pblist = alloc_image_page(gfp_mask, safe_needed); |
25761b6e RW |
385 | /* FIXME: rewrite this ugly loop */ |
386 | for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages; | |
387 | pbe = pbe->next, num += PBES_PER_PAGE) { | |
388 | pbe += PB_PAGE_SKIP; | |
054bd4c1 | 389 | pbe->next = alloc_image_page(gfp_mask, safe_needed); |
25761b6e RW |
390 | } |
391 | if (!pbe) { /* get_zeroed_page() failed */ | |
392 | free_pagedir(pblist); | |
393 | pblist = NULL; | |
7088a5c0 | 394 | } else |
6e1819d6 | 395 | create_pbe_list(pblist, nr_pages); |
25761b6e RW |
396 | return pblist; |
397 | } | |
398 | ||
399 | /** | |
400 | * Free pages we allocated for suspend. Suspend pages are alocated | |
401 | * before atomic copy, so we need to free them after resume. | |
402 | */ | |
403 | ||
404 | void swsusp_free(void) | |
405 | { | |
406 | struct zone *zone; | |
407 | unsigned long zone_pfn; | |
408 | ||
409 | for_each_zone(zone) { | |
410 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
411 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) { | |
dc19d507 | 412 | struct page *page; |
25761b6e RW |
413 | page = pfn_to_page(zone_pfn + zone->zone_start_pfn); |
414 | if (PageNosave(page) && PageNosaveFree(page)) { | |
415 | ClearPageNosave(page); | |
416 | ClearPageNosaveFree(page); | |
417 | free_page((long) page_address(page)); | |
418 | } | |
419 | } | |
420 | } | |
f577eb30 RW |
421 | nr_copy_pages = 0; |
422 | nr_meta_pages = 0; | |
423 | pagedir_nosave = NULL; | |
6e1819d6 | 424 | buffer = NULL; |
25761b6e RW |
425 | } |
426 | ||
427 | ||
428 | /** | |
429 | * enough_free_mem - Make sure we enough free memory to snapshot. | |
430 | * | |
431 | * Returns TRUE or FALSE after checking the number of available | |
432 | * free pages. | |
433 | */ | |
434 | ||
dc19d507 | 435 | static int enough_free_mem(unsigned int nr_pages) |
25761b6e | 436 | { |
e5e2fa78 RW |
437 | struct zone *zone; |
438 | unsigned int n = 0; | |
439 | ||
440 | for_each_zone (zone) | |
441 | if (!is_highmem(zone)) | |
442 | n += zone->free_pages; | |
443 | pr_debug("swsusp: available memory: %u pages\n", n); | |
444 | return n > (nr_pages + PAGES_FOR_IO + | |
a0f49651 | 445 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE); |
25761b6e RW |
446 | } |
447 | ||
f577eb30 | 448 | static int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed) |
054bd4c1 RW |
449 | { |
450 | struct pbe *p; | |
451 | ||
452 | for_each_pbe (p, pblist) { | |
453 | p->address = (unsigned long)alloc_image_page(gfp_mask, safe_needed); | |
454 | if (!p->address) | |
455 | return -ENOMEM; | |
456 | } | |
457 | return 0; | |
458 | } | |
25761b6e | 459 | |
dc19d507 | 460 | static struct pbe *swsusp_alloc(unsigned int nr_pages) |
25761b6e | 461 | { |
054bd4c1 | 462 | struct pbe *pblist; |
25761b6e | 463 | |
054bd4c1 | 464 | if (!(pblist = alloc_pagedir(nr_pages, GFP_ATOMIC | __GFP_COLD, 0))) { |
25761b6e | 465 | printk(KERN_ERR "suspend: Allocating pagedir failed.\n"); |
a0f49651 | 466 | return NULL; |
25761b6e | 467 | } |
25761b6e | 468 | |
054bd4c1 RW |
469 | if (alloc_data_pages(pblist, GFP_ATOMIC | __GFP_COLD, 0)) { |
470 | printk(KERN_ERR "suspend: Allocating image pages failed.\n"); | |
471 | swsusp_free(); | |
472 | return NULL; | |
25761b6e RW |
473 | } |
474 | ||
a0f49651 | 475 | return pblist; |
25761b6e RW |
476 | } |
477 | ||
2e32a43e | 478 | asmlinkage int swsusp_save(void) |
25761b6e | 479 | { |
dc19d507 | 480 | unsigned int nr_pages; |
25761b6e RW |
481 | |
482 | pr_debug("swsusp: critical section: \n"); | |
25761b6e RW |
483 | |
484 | drain_local_pages(); | |
a0f49651 RW |
485 | nr_pages = count_data_pages(); |
486 | printk("swsusp: Need to copy %u pages\n", nr_pages); | |
25761b6e RW |
487 | |
488 | pr_debug("swsusp: pages needed: %u + %lu + %u, free: %u\n", | |
a0f49651 RW |
489 | nr_pages, |
490 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE, | |
25761b6e RW |
491 | PAGES_FOR_IO, nr_free_pages()); |
492 | ||
a0f49651 | 493 | if (!enough_free_mem(nr_pages)) { |
25761b6e RW |
494 | printk(KERN_ERR "swsusp: Not enough free memory\n"); |
495 | return -ENOMEM; | |
496 | } | |
497 | ||
a0f49651 RW |
498 | pagedir_nosave = swsusp_alloc(nr_pages); |
499 | if (!pagedir_nosave) | |
500 | return -ENOMEM; | |
25761b6e RW |
501 | |
502 | /* During allocating of suspend pagedir, new cold pages may appear. | |
503 | * Kill them. | |
504 | */ | |
505 | drain_local_pages(); | |
a0f49651 | 506 | copy_data_pages(pagedir_nosave); |
25761b6e RW |
507 | |
508 | /* | |
509 | * End of critical section. From now on, we can write to memory, | |
510 | * but we should not touch disk. This specially means we must _not_ | |
511 | * touch swap space! Except we must write out our image of course. | |
512 | */ | |
513 | ||
a0f49651 | 514 | nr_copy_pages = nr_pages; |
f577eb30 | 515 | nr_meta_pages = (nr_pages * sizeof(long) + PAGE_SIZE - 1) >> PAGE_SHIFT; |
a0f49651 RW |
516 | |
517 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); | |
25761b6e RW |
518 | return 0; |
519 | } | |
f577eb30 RW |
520 | |
521 | static void init_header(struct swsusp_info *info) | |
522 | { | |
523 | memset(info, 0, sizeof(struct swsusp_info)); | |
524 | info->version_code = LINUX_VERSION_CODE; | |
525 | info->num_physpages = num_physpages; | |
526 | memcpy(&info->uts, &system_utsname, sizeof(system_utsname)); | |
527 | info->cpus = num_online_cpus(); | |
528 | info->image_pages = nr_copy_pages; | |
529 | info->pages = nr_copy_pages + nr_meta_pages + 1; | |
6e1819d6 RW |
530 | info->size = info->pages; |
531 | info->size <<= PAGE_SHIFT; | |
f577eb30 RW |
532 | } |
533 | ||
534 | /** | |
535 | * pack_orig_addresses - the .orig_address fields of the PBEs from the | |
536 | * list starting at @pbe are stored in the array @buf[] (1 page) | |
537 | */ | |
538 | ||
539 | static inline struct pbe *pack_orig_addresses(unsigned long *buf, struct pbe *pbe) | |
540 | { | |
541 | int j; | |
542 | ||
543 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | |
544 | buf[j] = pbe->orig_address; | |
545 | pbe = pbe->next; | |
546 | } | |
547 | if (!pbe) | |
548 | for (; j < PAGE_SIZE / sizeof(long); j++) | |
549 | buf[j] = 0; | |
550 | return pbe; | |
551 | } | |
552 | ||
553 | /** | |
554 | * snapshot_read_next - used for reading the system memory snapshot. | |
555 | * | |
556 | * On the first call to it @handle should point to a zeroed | |
557 | * snapshot_handle structure. The structure gets updated and a pointer | |
558 | * to it should be passed to this function every next time. | |
559 | * | |
560 | * The @count parameter should contain the number of bytes the caller | |
561 | * wants to read from the snapshot. It must not be zero. | |
562 | * | |
563 | * On success the function returns a positive number. Then, the caller | |
564 | * is allowed to read up to the returned number of bytes from the memory | |
565 | * location computed by the data_of() macro. The number returned | |
566 | * may be smaller than @count, but this only happens if the read would | |
567 | * cross a page boundary otherwise. | |
568 | * | |
569 | * The function returns 0 to indicate the end of data stream condition, | |
570 | * and a negative number is returned on error. In such cases the | |
571 | * structure pointed to by @handle is not updated and should not be used | |
572 | * any more. | |
573 | */ | |
574 | ||
575 | int snapshot_read_next(struct snapshot_handle *handle, size_t count) | |
576 | { | |
f577eb30 RW |
577 | if (handle->page > nr_meta_pages + nr_copy_pages) |
578 | return 0; | |
579 | if (!buffer) { | |
580 | /* This makes the buffer be freed by swsusp_free() */ | |
581 | buffer = alloc_image_page(GFP_ATOMIC, 0); | |
582 | if (!buffer) | |
583 | return -ENOMEM; | |
584 | } | |
585 | if (!handle->offset) { | |
586 | init_header((struct swsusp_info *)buffer); | |
587 | handle->buffer = buffer; | |
588 | handle->pbe = pagedir_nosave; | |
589 | } | |
590 | if (handle->prev < handle->page) { | |
591 | if (handle->page <= nr_meta_pages) { | |
592 | handle->pbe = pack_orig_addresses(buffer, handle->pbe); | |
593 | if (!handle->pbe) | |
594 | handle->pbe = pagedir_nosave; | |
595 | } else { | |
596 | handle->buffer = (void *)handle->pbe->address; | |
597 | handle->pbe = handle->pbe->next; | |
598 | } | |
599 | handle->prev = handle->page; | |
600 | } | |
601 | handle->buf_offset = handle->page_offset; | |
602 | if (handle->page_offset + count >= PAGE_SIZE) { | |
603 | count = PAGE_SIZE - handle->page_offset; | |
604 | handle->page_offset = 0; | |
605 | handle->page++; | |
606 | } else { | |
607 | handle->page_offset += count; | |
608 | } | |
609 | handle->offset += count; | |
610 | return count; | |
611 | } | |
612 | ||
613 | /** | |
614 | * mark_unsafe_pages - mark the pages that cannot be used for storing | |
615 | * the image during resume, because they conflict with the pages that | |
616 | * had been used before suspend | |
617 | */ | |
618 | ||
619 | static int mark_unsafe_pages(struct pbe *pblist) | |
620 | { | |
621 | struct zone *zone; | |
622 | unsigned long zone_pfn; | |
623 | struct pbe *p; | |
624 | ||
625 | if (!pblist) /* a sanity check */ | |
626 | return -EINVAL; | |
627 | ||
628 | /* Clear page flags */ | |
629 | for_each_zone (zone) { | |
630 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
631 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) | |
632 | ClearPageNosaveFree(pfn_to_page(zone_pfn + | |
633 | zone->zone_start_pfn)); | |
634 | } | |
635 | ||
636 | /* Mark orig addresses */ | |
637 | for_each_pbe (p, pblist) { | |
638 | if (virt_addr_valid(p->orig_address)) | |
639 | SetPageNosaveFree(virt_to_page(p->orig_address)); | |
640 | else | |
641 | return -EFAULT; | |
642 | } | |
643 | ||
644 | return 0; | |
645 | } | |
646 | ||
647 | static void copy_page_backup_list(struct pbe *dst, struct pbe *src) | |
648 | { | |
649 | /* We assume both lists contain the same number of elements */ | |
650 | while (src) { | |
651 | dst->orig_address = src->orig_address; | |
652 | dst = dst->next; | |
653 | src = src->next; | |
654 | } | |
655 | } | |
656 | ||
657 | static int check_header(struct swsusp_info *info) | |
658 | { | |
659 | char *reason = NULL; | |
660 | ||
661 | if (info->version_code != LINUX_VERSION_CODE) | |
662 | reason = "kernel version"; | |
663 | if (info->num_physpages != num_physpages) | |
664 | reason = "memory size"; | |
665 | if (strcmp(info->uts.sysname,system_utsname.sysname)) | |
666 | reason = "system type"; | |
667 | if (strcmp(info->uts.release,system_utsname.release)) | |
668 | reason = "kernel release"; | |
669 | if (strcmp(info->uts.version,system_utsname.version)) | |
670 | reason = "version"; | |
671 | if (strcmp(info->uts.machine,system_utsname.machine)) | |
672 | reason = "machine"; | |
673 | if (reason) { | |
674 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | |
675 | return -EPERM; | |
676 | } | |
677 | return 0; | |
678 | } | |
679 | ||
680 | /** | |
681 | * load header - check the image header and copy data from it | |
682 | */ | |
683 | ||
684 | static int load_header(struct snapshot_handle *handle, | |
685 | struct swsusp_info *info) | |
686 | { | |
687 | int error; | |
688 | struct pbe *pblist; | |
689 | ||
690 | error = check_header(info); | |
691 | if (!error) { | |
692 | pblist = alloc_pagedir(info->image_pages, GFP_ATOMIC, 0); | |
693 | if (!pblist) | |
694 | return -ENOMEM; | |
695 | pagedir_nosave = pblist; | |
696 | handle->pbe = pblist; | |
697 | nr_copy_pages = info->image_pages; | |
698 | nr_meta_pages = info->pages - info->image_pages - 1; | |
699 | } | |
700 | return error; | |
701 | } | |
702 | ||
703 | /** | |
704 | * unpack_orig_addresses - copy the elements of @buf[] (1 page) to | |
705 | * the PBEs in the list starting at @pbe | |
706 | */ | |
707 | ||
708 | static inline struct pbe *unpack_orig_addresses(unsigned long *buf, | |
709 | struct pbe *pbe) | |
710 | { | |
711 | int j; | |
712 | ||
713 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | |
714 | pbe->orig_address = buf[j]; | |
715 | pbe = pbe->next; | |
716 | } | |
717 | return pbe; | |
718 | } | |
719 | ||
720 | /** | |
721 | * create_image - use metadata contained in the PBE list | |
722 | * pointed to by pagedir_nosave to mark the pages that will | |
723 | * be overwritten in the process of restoring the system | |
724 | * memory state from the image and allocate memory for | |
725 | * the image avoiding these pages | |
726 | */ | |
727 | ||
728 | static int create_image(struct snapshot_handle *handle) | |
729 | { | |
730 | int error = 0; | |
731 | struct pbe *p, *pblist; | |
732 | ||
733 | p = pagedir_nosave; | |
734 | error = mark_unsafe_pages(p); | |
735 | if (!error) { | |
736 | pblist = alloc_pagedir(nr_copy_pages, GFP_ATOMIC, 1); | |
737 | if (pblist) | |
738 | copy_page_backup_list(pblist, p); | |
739 | free_pagedir(p); | |
740 | if (!pblist) | |
741 | error = -ENOMEM; | |
742 | } | |
743 | if (!error) | |
744 | error = alloc_data_pages(pblist, GFP_ATOMIC, 1); | |
745 | if (!error) { | |
746 | release_eaten_pages(); | |
747 | pagedir_nosave = pblist; | |
748 | } else { | |
749 | pagedir_nosave = NULL; | |
750 | handle->pbe = NULL; | |
751 | nr_copy_pages = 0; | |
752 | nr_meta_pages = 0; | |
753 | } | |
754 | return error; | |
755 | } | |
756 | ||
757 | /** | |
758 | * snapshot_write_next - used for writing the system memory snapshot. | |
759 | * | |
760 | * On the first call to it @handle should point to a zeroed | |
761 | * snapshot_handle structure. The structure gets updated and a pointer | |
762 | * to it should be passed to this function every next time. | |
763 | * | |
764 | * The @count parameter should contain the number of bytes the caller | |
765 | * wants to write to the image. It must not be zero. | |
766 | * | |
767 | * On success the function returns a positive number. Then, the caller | |
768 | * is allowed to write up to the returned number of bytes to the memory | |
769 | * location computed by the data_of() macro. The number returned | |
770 | * may be smaller than @count, but this only happens if the write would | |
771 | * cross a page boundary otherwise. | |
772 | * | |
773 | * The function returns 0 to indicate the "end of file" condition, | |
774 | * and a negative number is returned on error. In such cases the | |
775 | * structure pointed to by @handle is not updated and should not be used | |
776 | * any more. | |
777 | */ | |
778 | ||
779 | int snapshot_write_next(struct snapshot_handle *handle, size_t count) | |
780 | { | |
f577eb30 RW |
781 | int error = 0; |
782 | ||
783 | if (handle->prev && handle->page > nr_meta_pages + nr_copy_pages) | |
784 | return 0; | |
785 | if (!buffer) { | |
786 | /* This makes the buffer be freed by swsusp_free() */ | |
787 | buffer = alloc_image_page(GFP_ATOMIC, 0); | |
788 | if (!buffer) | |
789 | return -ENOMEM; | |
790 | } | |
791 | if (!handle->offset) | |
792 | handle->buffer = buffer; | |
793 | if (handle->prev < handle->page) { | |
794 | if (!handle->prev) { | |
795 | error = load_header(handle, (struct swsusp_info *)buffer); | |
796 | if (error) | |
797 | return error; | |
798 | } else if (handle->prev <= nr_meta_pages) { | |
799 | handle->pbe = unpack_orig_addresses(buffer, handle->pbe); | |
800 | if (!handle->pbe) { | |
801 | error = create_image(handle); | |
802 | if (error) | |
803 | return error; | |
804 | handle->pbe = pagedir_nosave; | |
805 | handle->buffer = (void *)handle->pbe->address; | |
806 | } | |
807 | } else { | |
808 | handle->pbe = handle->pbe->next; | |
809 | handle->buffer = (void *)handle->pbe->address; | |
810 | } | |
811 | handle->prev = handle->page; | |
812 | } | |
813 | handle->buf_offset = handle->page_offset; | |
814 | if (handle->page_offset + count >= PAGE_SIZE) { | |
815 | count = PAGE_SIZE - handle->page_offset; | |
816 | handle->page_offset = 0; | |
817 | handle->page++; | |
818 | } else { | |
819 | handle->page_offset += count; | |
820 | } | |
821 | handle->offset += count; | |
822 | return count; | |
823 | } | |
824 | ||
825 | int snapshot_image_loaded(struct snapshot_handle *handle) | |
826 | { | |
827 | return !(!handle->pbe || handle->pbe->next || !nr_copy_pages || | |
828 | handle->page <= nr_meta_pages + nr_copy_pages); | |
829 | } |