]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/power/swsusp.c | |
3 | * | |
4 | * This file is to realize architecture-independent | |
5 | * machine suspend feature using pretty near only high-level routines | |
6 | * | |
7 | * Copyright (C) 1998-2001 Gabor Kuti <seasons@fornax.hu> | |
8 | * Copyright (C) 1998,2001-2004 Pavel Machek <pavel@suse.cz> | |
9 | * | |
10 | * This file is released under the GPLv2. | |
11 | * | |
12 | * I'd like to thank the following people for their work: | |
2e4d5822 | 13 | * |
1da177e4 LT |
14 | * Pavel Machek <pavel@ucw.cz>: |
15 | * Modifications, defectiveness pointing, being with me at the very beginning, | |
16 | * suspend to swap space, stop all tasks. Port to 2.4.18-ac and 2.5.17. | |
17 | * | |
2e4d5822 | 18 | * Steve Doddi <dirk@loth.demon.co.uk>: |
1da177e4 LT |
19 | * Support the possibility of hardware state restoring. |
20 | * | |
21 | * Raph <grey.havens@earthling.net>: | |
22 | * Support for preserving states of network devices and virtual console | |
23 | * (including X and svgatextmode) | |
24 | * | |
25 | * Kurt Garloff <garloff@suse.de>: | |
26 | * Straightened the critical function in order to prevent compilers from | |
27 | * playing tricks with local variables. | |
28 | * | |
29 | * Andreas Mohr <a.mohr@mailto.de> | |
30 | * | |
31 | * Alex Badea <vampire@go.ro>: | |
32 | * Fixed runaway init | |
33 | * | |
c2ff18f4 AS |
34 | * Andreas Steinmetz <ast@domdv.de>: |
35 | * Added encrypted suspend option | |
36 | * | |
1da177e4 LT |
37 | * More state savers are welcome. Especially for the scsi layer... |
38 | * | |
39 | * For TODOs,FIXMEs also look in Documentation/power/swsusp.txt | |
40 | */ | |
41 | ||
42 | #include <linux/module.h> | |
43 | #include <linux/mm.h> | |
44 | #include <linux/suspend.h> | |
45 | #include <linux/smp_lock.h> | |
46 | #include <linux/file.h> | |
47 | #include <linux/utsname.h> | |
48 | #include <linux/version.h> | |
49 | #include <linux/delay.h> | |
50 | #include <linux/reboot.h> | |
51 | #include <linux/bitops.h> | |
52 | #include <linux/vt_kern.h> | |
53 | #include <linux/kbd_kern.h> | |
54 | #include <linux/keyboard.h> | |
55 | #include <linux/spinlock.h> | |
56 | #include <linux/genhd.h> | |
57 | #include <linux/kernel.h> | |
58 | #include <linux/major.h> | |
59 | #include <linux/swap.h> | |
60 | #include <linux/pm.h> | |
61 | #include <linux/device.h> | |
62 | #include <linux/buffer_head.h> | |
63 | #include <linux/swapops.h> | |
64 | #include <linux/bootmem.h> | |
65 | #include <linux/syscalls.h> | |
66 | #include <linux/console.h> | |
67 | #include <linux/highmem.h> | |
68 | #include <linux/bio.h> | |
d53d9f16 | 69 | #include <linux/mount.h> |
1da177e4 LT |
70 | |
71 | #include <asm/uaccess.h> | |
72 | #include <asm/mmu_context.h> | |
73 | #include <asm/pgtable.h> | |
74 | #include <asm/tlbflush.h> | |
75 | #include <asm/io.h> | |
76 | ||
c2ff18f4 AS |
77 | #include <linux/random.h> |
78 | #include <linux/crypto.h> | |
79 | #include <asm/scatterlist.h> | |
80 | ||
1da177e4 LT |
81 | #include "power.h" |
82 | ||
c2ff18f4 AS |
83 | #define CIPHER "aes" |
84 | #define MAXKEY 32 | |
85 | #define MAXIV 32 | |
86 | ||
1da177e4 LT |
87 | /* References to section boundaries */ |
88 | extern const void __nosave_begin, __nosave_end; | |
89 | ||
90 | /* Variables to be preserved over suspend */ | |
91 | static int nr_copy_pages_check; | |
92 | ||
93 | extern char resume_file[]; | |
94 | ||
95 | /* Local variables that should not be affected by save */ | |
52c1da39 | 96 | static unsigned int nr_copy_pages __nosavedata = 0; |
1da177e4 LT |
97 | |
98 | /* Suspend pagedir is allocated before final copy, therefore it | |
2e4d5822 | 99 | must be freed after resume |
1da177e4 LT |
100 | |
101 | Warning: this is evil. There are actually two pagedirs at time of | |
102 | resume. One is "pagedir_save", which is empty frame allocated at | |
2e4d5822 | 103 | time of suspend, that must be freed. Second is "pagedir_nosave", |
1da177e4 LT |
104 | allocated at time of resume, that travels through memory not to |
105 | collide with anything. | |
106 | ||
107 | Warning: this is even more evil than it seems. Pagedirs this file | |
108 | talks about are completely different from page directories used by | |
109 | MMU hardware. | |
110 | */ | |
111 | suspend_pagedir_t *pagedir_nosave __nosavedata = NULL; | |
112 | static suspend_pagedir_t *pagedir_save; | |
113 | ||
114 | #define SWSUSP_SIG "S1SUSPEND" | |
115 | ||
116 | static struct swsusp_header { | |
c2ff18f4 AS |
117 | char reserved[PAGE_SIZE - 20 - MAXKEY - MAXIV - sizeof(swp_entry_t)]; |
118 | u8 key_iv[MAXKEY+MAXIV]; | |
1da177e4 LT |
119 | swp_entry_t swsusp_info; |
120 | char orig_sig[10]; | |
121 | char sig[10]; | |
122 | } __attribute__((packed, aligned(PAGE_SIZE))) swsusp_header; | |
123 | ||
124 | static struct swsusp_info swsusp_info; | |
125 | ||
126 | /* | |
127 | * XXX: We try to keep some more pages free so that I/O operations succeed | |
128 | * without paging. Might this be more? | |
129 | */ | |
130 | #define PAGES_FOR_IO 512 | |
131 | ||
132 | /* | |
133 | * Saving part... | |
134 | */ | |
135 | ||
136 | /* We memorize in swapfile_used what swap devices are used for suspension */ | |
137 | #define SWAPFILE_UNUSED 0 | |
138 | #define SWAPFILE_SUSPEND 1 /* This is the suspending device */ | |
139 | #define SWAPFILE_IGNORED 2 /* Those are other swap devices ignored for suspension */ | |
140 | ||
141 | static unsigned short swapfile_used[MAX_SWAPFILES]; | |
142 | static unsigned short root_swap; | |
143 | ||
c2ff18f4 AS |
144 | static int write_page(unsigned long addr, swp_entry_t * loc); |
145 | static int bio_read_page(pgoff_t page_off, void * page); | |
146 | ||
147 | static u8 key_iv[MAXKEY+MAXIV]; | |
148 | ||
149 | #ifdef CONFIG_SWSUSP_ENCRYPT | |
150 | ||
151 | static int crypto_init(int mode, void **mem) | |
152 | { | |
153 | int error = 0; | |
154 | int len; | |
155 | char *modemsg; | |
156 | struct crypto_tfm *tfm; | |
157 | ||
158 | modemsg = mode ? "suspend not possible" : "resume not possible"; | |
159 | ||
160 | tfm = crypto_alloc_tfm(CIPHER, CRYPTO_TFM_MODE_CBC); | |
161 | if(!tfm) { | |
162 | printk(KERN_ERR "swsusp: no tfm, %s\n", modemsg); | |
163 | error = -EINVAL; | |
164 | goto out; | |
165 | } | |
166 | ||
167 | if(MAXKEY < crypto_tfm_alg_min_keysize(tfm)) { | |
168 | printk(KERN_ERR "swsusp: key buffer too small, %s\n", modemsg); | |
169 | error = -ENOKEY; | |
170 | goto fail; | |
171 | } | |
172 | ||
173 | if (mode) | |
174 | get_random_bytes(key_iv, MAXKEY+MAXIV); | |
175 | ||
176 | len = crypto_tfm_alg_max_keysize(tfm); | |
177 | if (len > MAXKEY) | |
178 | len = MAXKEY; | |
179 | ||
180 | if (crypto_cipher_setkey(tfm, key_iv, len)) { | |
181 | printk(KERN_ERR "swsusp: key setup failure, %s\n", modemsg); | |
182 | error = -EKEYREJECTED; | |
183 | goto fail; | |
184 | } | |
185 | ||
186 | len = crypto_tfm_alg_ivsize(tfm); | |
187 | ||
188 | if (MAXIV < len) { | |
189 | printk(KERN_ERR "swsusp: iv buffer too small, %s\n", modemsg); | |
190 | error = -EOVERFLOW; | |
191 | goto fail; | |
192 | } | |
193 | ||
194 | crypto_cipher_set_iv(tfm, key_iv+MAXKEY, len); | |
195 | ||
196 | *mem=(void *)tfm; | |
197 | ||
198 | goto out; | |
199 | ||
200 | fail: crypto_free_tfm(tfm); | |
201 | out: return error; | |
202 | } | |
203 | ||
204 | static __inline__ void crypto_exit(void *mem) | |
205 | { | |
206 | crypto_free_tfm((struct crypto_tfm *)mem); | |
207 | } | |
208 | ||
209 | static __inline__ int crypto_write(struct pbe *p, void *mem) | |
210 | { | |
211 | int error = 0; | |
212 | struct scatterlist src, dst; | |
213 | ||
214 | src.page = virt_to_page(p->address); | |
215 | src.offset = 0; | |
216 | src.length = PAGE_SIZE; | |
217 | dst.page = virt_to_page((void *)&swsusp_header); | |
218 | dst.offset = 0; | |
219 | dst.length = PAGE_SIZE; | |
220 | ||
221 | error = crypto_cipher_encrypt((struct crypto_tfm *)mem, &dst, &src, | |
222 | PAGE_SIZE); | |
223 | ||
224 | if (!error) | |
225 | error = write_page((unsigned long)&swsusp_header, | |
226 | &(p->swap_address)); | |
227 | return error; | |
228 | } | |
229 | ||
230 | static __inline__ int crypto_read(struct pbe *p, void *mem) | |
231 | { | |
232 | int error = 0; | |
233 | struct scatterlist src, dst; | |
234 | ||
235 | error = bio_read_page(swp_offset(p->swap_address), (void *)p->address); | |
236 | if (!error) { | |
237 | src.offset = 0; | |
238 | src.length = PAGE_SIZE; | |
239 | dst.offset = 0; | |
240 | dst.length = PAGE_SIZE; | |
241 | src.page = dst.page = virt_to_page((void *)p->address); | |
242 | ||
243 | error = crypto_cipher_decrypt((struct crypto_tfm *)mem, &dst, | |
244 | &src, PAGE_SIZE); | |
245 | } | |
246 | return error; | |
247 | } | |
248 | #else | |
249 | static __inline__ int crypto_init(int mode, void *mem) | |
250 | { | |
251 | return 0; | |
252 | } | |
253 | ||
254 | static __inline__ void crypto_exit(void *mem) | |
255 | { | |
256 | } | |
257 | ||
258 | static __inline__ int crypto_write(struct pbe *p, void *mem) | |
259 | { | |
260 | return write_page(p->address, &(p->swap_address)); | |
261 | } | |
262 | ||
263 | static __inline__ int crypto_read(struct pbe *p, void *mem) | |
264 | { | |
265 | return bio_read_page(swp_offset(p->swap_address), (void *)p->address); | |
266 | } | |
267 | #endif | |
268 | ||
1da177e4 LT |
269 | static int mark_swapfiles(swp_entry_t prev) |
270 | { | |
271 | int error; | |
272 | ||
2e4d5822 | 273 | rw_swap_page_sync(READ, |
1da177e4 LT |
274 | swp_entry(root_swap, 0), |
275 | virt_to_page((unsigned long)&swsusp_header)); | |
276 | if (!memcmp("SWAP-SPACE",swsusp_header.sig, 10) || | |
277 | !memcmp("SWAPSPACE2",swsusp_header.sig, 10)) { | |
278 | memcpy(swsusp_header.orig_sig,swsusp_header.sig, 10); | |
279 | memcpy(swsusp_header.sig,SWSUSP_SIG, 10); | |
c2ff18f4 | 280 | memcpy(swsusp_header.key_iv, key_iv, MAXKEY+MAXIV); |
1da177e4 | 281 | swsusp_header.swsusp_info = prev; |
2e4d5822 | 282 | error = rw_swap_page_sync(WRITE, |
1da177e4 LT |
283 | swp_entry(root_swap, 0), |
284 | virt_to_page((unsigned long) | |
285 | &swsusp_header)); | |
286 | } else { | |
287 | pr_debug("swsusp: Partition is not swap space.\n"); | |
288 | error = -ENODEV; | |
289 | } | |
290 | return error; | |
291 | } | |
292 | ||
293 | /* | |
294 | * Check whether the swap device is the specified resume | |
295 | * device, irrespective of whether they are specified by | |
296 | * identical names. | |
297 | * | |
298 | * (Thus, device inode aliasing is allowed. You can say /dev/hda4 | |
299 | * instead of /dev/ide/host0/bus0/target0/lun0/part4 [if using devfs] | |
300 | * and they'll be considered the same device. This is *necessary* for | |
301 | * devfs, since the resume code can only recognize the form /dev/hda4, | |
302 | * but the suspend code would see the long name.) | |
303 | */ | |
304 | static int is_resume_device(const struct swap_info_struct *swap_info) | |
305 | { | |
306 | struct file *file = swap_info->swap_file; | |
307 | struct inode *inode = file->f_dentry->d_inode; | |
308 | ||
309 | return S_ISBLK(inode->i_mode) && | |
310 | swsusp_resume_device == MKDEV(imajor(inode), iminor(inode)); | |
311 | } | |
312 | ||
313 | static int swsusp_swap_check(void) /* This is called before saving image */ | |
314 | { | |
315 | int i, len; | |
2e4d5822 | 316 | |
1da177e4 LT |
317 | len=strlen(resume_file); |
318 | root_swap = 0xFFFF; | |
2e4d5822 | 319 | |
dae06ac4 | 320 | spin_lock(&swap_lock); |
2e4d5822 | 321 | for (i=0; i<MAX_SWAPFILES; i++) { |
dae06ac4 | 322 | if (!(swap_info[i].flags & SWP_WRITEOK)) { |
1da177e4 LT |
323 | swapfile_used[i]=SWAPFILE_UNUSED; |
324 | } else { | |
2e4d5822 | 325 | if (!len) { |
1da177e4 | 326 | printk(KERN_WARNING "resume= option should be used to set suspend device" ); |
2e4d5822 | 327 | if (root_swap == 0xFFFF) { |
1da177e4 LT |
328 | swapfile_used[i] = SWAPFILE_SUSPEND; |
329 | root_swap = i; | |
330 | } else | |
2e4d5822 | 331 | swapfile_used[i] = SWAPFILE_IGNORED; |
1da177e4 LT |
332 | } else { |
333 | /* we ignore all swap devices that are not the resume_file */ | |
334 | if (is_resume_device(&swap_info[i])) { | |
335 | swapfile_used[i] = SWAPFILE_SUSPEND; | |
336 | root_swap = i; | |
337 | } else { | |
338 | swapfile_used[i] = SWAPFILE_IGNORED; | |
339 | } | |
340 | } | |
341 | } | |
342 | } | |
dae06ac4 | 343 | spin_unlock(&swap_lock); |
1da177e4 LT |
344 | return (root_swap != 0xffff) ? 0 : -ENODEV; |
345 | } | |
346 | ||
347 | /** | |
348 | * This is called after saving image so modification | |
349 | * will be lost after resume... and that's what we want. | |
350 | * we make the device unusable. A new call to | |
2e4d5822 | 351 | * lock_swapdevices can unlock the devices. |
1da177e4 LT |
352 | */ |
353 | static void lock_swapdevices(void) | |
354 | { | |
355 | int i; | |
356 | ||
dae06ac4 | 357 | spin_lock(&swap_lock); |
2e4d5822 PM |
358 | for (i = 0; i< MAX_SWAPFILES; i++) |
359 | if (swapfile_used[i] == SWAPFILE_IGNORED) { | |
dae06ac4 | 360 | swap_info[i].flags ^= SWP_WRITEOK; |
1da177e4 | 361 | } |
dae06ac4 | 362 | spin_unlock(&swap_lock); |
1da177e4 LT |
363 | } |
364 | ||
365 | /** | |
8686bcd0 | 366 | * write_page - Write one page to a fresh swap location. |
1da177e4 LT |
367 | * @addr: Address we're writing. |
368 | * @loc: Place to store the entry we used. | |
369 | * | |
370 | * Allocate a new swap entry and 'sync' it. Note we discard -EIO | |
2e4d5822 | 371 | * errors. That is an artifact left over from swsusp. It did not |
1da177e4 LT |
372 | * check the return of rw_swap_page_sync() at all, since most pages |
373 | * written back to swap would return -EIO. | |
374 | * This is a partial improvement, since we will at least return other | |
375 | * errors, though we need to eventually fix the damn code. | |
376 | */ | |
377 | static int write_page(unsigned long addr, swp_entry_t * loc) | |
378 | { | |
379 | swp_entry_t entry; | |
380 | int error = 0; | |
381 | ||
382 | entry = get_swap_page(); | |
2e4d5822 | 383 | if (swp_offset(entry) && |
1da177e4 LT |
384 | swapfile_used[swp_type(entry)] == SWAPFILE_SUSPEND) { |
385 | error = rw_swap_page_sync(WRITE, entry, | |
386 | virt_to_page(addr)); | |
387 | if (error == -EIO) | |
388 | error = 0; | |
389 | if (!error) | |
390 | *loc = entry; | |
391 | } else | |
392 | error = -ENOSPC; | |
393 | return error; | |
394 | } | |
395 | ||
396 | /** | |
397 | * data_free - Free the swap entries used by the saved image. | |
398 | * | |
2e4d5822 | 399 | * Walk the list of used swap entries and free each one. |
1da177e4 LT |
400 | * This is only used for cleanup when suspend fails. |
401 | */ | |
402 | static void data_free(void) | |
403 | { | |
404 | swp_entry_t entry; | |
254b5477 | 405 | struct pbe * p; |
1da177e4 | 406 | |
254b5477 RW |
407 | for_each_pbe(p, pagedir_nosave) { |
408 | entry = p->swap_address; | |
1da177e4 LT |
409 | if (entry.val) |
410 | swap_free(entry); | |
411 | else | |
412 | break; | |
1da177e4 LT |
413 | } |
414 | } | |
415 | ||
416 | /** | |
417 | * data_write - Write saved image to swap. | |
418 | * | |
419 | * Walk the list of pages in the image and sync each one to swap. | |
420 | */ | |
421 | static int data_write(void) | |
422 | { | |
423 | int error = 0, i = 0; | |
424 | unsigned int mod = nr_copy_pages / 100; | |
425 | struct pbe *p; | |
c2ff18f4 AS |
426 | void *tfm; |
427 | ||
428 | if ((error = crypto_init(1, &tfm))) | |
429 | return error; | |
1da177e4 LT |
430 | |
431 | if (!mod) | |
432 | mod = 1; | |
433 | ||
434 | printk( "Writing data to swap (%d pages)... ", nr_copy_pages ); | |
2e4d5822 | 435 | for_each_pbe (p, pagedir_nosave) { |
1da177e4 LT |
436 | if (!(i%mod)) |
437 | printk( "\b\b\b\b%3d%%", i / mod ); | |
c2ff18f4 AS |
438 | if ((error = crypto_write(p, tfm))) { |
439 | crypto_exit(tfm); | |
1da177e4 | 440 | return error; |
c2ff18f4 | 441 | } |
1da177e4 LT |
442 | i++; |
443 | } | |
444 | printk("\b\b\b\bdone\n"); | |
c2ff18f4 | 445 | crypto_exit(tfm); |
1da177e4 LT |
446 | return error; |
447 | } | |
448 | ||
449 | static void dump_info(void) | |
450 | { | |
451 | pr_debug(" swsusp: Version: %u\n",swsusp_info.version_code); | |
452 | pr_debug(" swsusp: Num Pages: %ld\n",swsusp_info.num_physpages); | |
453 | pr_debug(" swsusp: UTS Sys: %s\n",swsusp_info.uts.sysname); | |
454 | pr_debug(" swsusp: UTS Node: %s\n",swsusp_info.uts.nodename); | |
455 | pr_debug(" swsusp: UTS Release: %s\n",swsusp_info.uts.release); | |
456 | pr_debug(" swsusp: UTS Version: %s\n",swsusp_info.uts.version); | |
457 | pr_debug(" swsusp: UTS Machine: %s\n",swsusp_info.uts.machine); | |
458 | pr_debug(" swsusp: UTS Domain: %s\n",swsusp_info.uts.domainname); | |
459 | pr_debug(" swsusp: CPUs: %d\n",swsusp_info.cpus); | |
460 | pr_debug(" swsusp: Image: %ld Pages\n",swsusp_info.image_pages); | |
461 | pr_debug(" swsusp: Pagedir: %ld Pages\n",swsusp_info.pagedir_pages); | |
462 | } | |
463 | ||
464 | static void init_header(void) | |
465 | { | |
466 | memset(&swsusp_info, 0, sizeof(swsusp_info)); | |
467 | swsusp_info.version_code = LINUX_VERSION_CODE; | |
468 | swsusp_info.num_physpages = num_physpages; | |
469 | memcpy(&swsusp_info.uts, &system_utsname, sizeof(system_utsname)); | |
470 | ||
471 | swsusp_info.suspend_pagedir = pagedir_nosave; | |
472 | swsusp_info.cpus = num_online_cpus(); | |
473 | swsusp_info.image_pages = nr_copy_pages; | |
474 | } | |
475 | ||
476 | static int close_swap(void) | |
477 | { | |
478 | swp_entry_t entry; | |
479 | int error; | |
480 | ||
481 | dump_info(); | |
482 | error = write_page((unsigned long)&swsusp_info, &entry); | |
2e4d5822 | 483 | if (!error) { |
1da177e4 LT |
484 | printk( "S" ); |
485 | error = mark_swapfiles(entry); | |
486 | printk( "|\n" ); | |
487 | } | |
488 | return error; | |
489 | } | |
490 | ||
491 | /** | |
492 | * free_pagedir_entries - Free pages used by the page directory. | |
493 | * | |
494 | * This is used during suspend for error recovery. | |
495 | */ | |
496 | ||
497 | static void free_pagedir_entries(void) | |
498 | { | |
499 | int i; | |
500 | ||
501 | for (i = 0; i < swsusp_info.pagedir_pages; i++) | |
502 | swap_free(swsusp_info.pagedir[i]); | |
503 | } | |
504 | ||
505 | ||
506 | /** | |
507 | * write_pagedir - Write the array of pages holding the page directory. | |
508 | * @last: Last swap entry we write (needed for header). | |
509 | */ | |
510 | ||
511 | static int write_pagedir(void) | |
512 | { | |
513 | int error = 0; | |
514 | unsigned n = 0; | |
515 | struct pbe * pbe; | |
516 | ||
517 | printk( "Writing pagedir..."); | |
2e4d5822 | 518 | for_each_pb_page (pbe, pagedir_nosave) { |
1da177e4 LT |
519 | if ((error = write_page((unsigned long)pbe, &swsusp_info.pagedir[n++]))) |
520 | return error; | |
521 | } | |
522 | ||
523 | swsusp_info.pagedir_pages = n; | |
524 | printk("done (%u pages)\n", n); | |
525 | return error; | |
526 | } | |
527 | ||
528 | /** | |
529 | * write_suspend_image - Write entire image and metadata. | |
530 | * | |
531 | */ | |
1da177e4 LT |
532 | static int write_suspend_image(void) |
533 | { | |
534 | int error; | |
535 | ||
536 | init_header(); | |
537 | if ((error = data_write())) | |
538 | goto FreeData; | |
539 | ||
540 | if ((error = write_pagedir())) | |
541 | goto FreePagedir; | |
542 | ||
543 | if ((error = close_swap())) | |
544 | goto FreePagedir; | |
545 | Done: | |
c2ff18f4 | 546 | memset(key_iv, 0, MAXKEY+MAXIV); |
1da177e4 LT |
547 | return error; |
548 | FreePagedir: | |
549 | free_pagedir_entries(); | |
550 | FreeData: | |
551 | data_free(); | |
552 | goto Done; | |
553 | } | |
554 | ||
555 | ||
556 | #ifdef CONFIG_HIGHMEM | |
557 | struct highmem_page { | |
558 | char *data; | |
559 | struct page *page; | |
560 | struct highmem_page *next; | |
561 | }; | |
562 | ||
563 | static struct highmem_page *highmem_copy; | |
564 | ||
565 | static int save_highmem_zone(struct zone *zone) | |
566 | { | |
567 | unsigned long zone_pfn; | |
568 | mark_free_pages(zone); | |
569 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
570 | struct page *page; | |
571 | struct highmem_page *save; | |
572 | void *kaddr; | |
573 | unsigned long pfn = zone_pfn + zone->zone_start_pfn; | |
574 | ||
575 | if (!(pfn%1000)) | |
576 | printk("."); | |
577 | if (!pfn_valid(pfn)) | |
578 | continue; | |
579 | page = pfn_to_page(pfn); | |
580 | /* | |
b5810039 NP |
581 | * PageReserved results from rvmalloc() sans vmalloc_32() |
582 | * and architectural memory reservations. | |
583 | * | |
584 | * rvmalloc should not cause this, because all implementations | |
585 | * appear to always be using vmalloc_32 on architectures with | |
586 | * highmem. This is a good thing, because we would like to save | |
587 | * rvmalloc pages. | |
588 | * | |
589 | * It appears to be triggered by pages which do not point to | |
590 | * valid memory (see arch/i386/mm/init.c:one_highpage_init(), | |
591 | * which sets PageReserved if the page does not point to valid | |
592 | * RAM. | |
593 | * | |
594 | * XXX: must remove usage of PageReserved! | |
1da177e4 | 595 | */ |
b5810039 | 596 | if (PageReserved(page)) |
1da177e4 | 597 | continue; |
1da177e4 LT |
598 | BUG_ON(PageNosave(page)); |
599 | if (PageNosaveFree(page)) | |
600 | continue; | |
601 | save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC); | |
602 | if (!save) | |
603 | return -ENOMEM; | |
604 | save->next = highmem_copy; | |
605 | save->page = page; | |
606 | save->data = (void *) get_zeroed_page(GFP_ATOMIC); | |
607 | if (!save->data) { | |
608 | kfree(save); | |
609 | return -ENOMEM; | |
610 | } | |
611 | kaddr = kmap_atomic(page, KM_USER0); | |
612 | memcpy(save->data, kaddr, PAGE_SIZE); | |
613 | kunmap_atomic(kaddr, KM_USER0); | |
614 | highmem_copy = save; | |
615 | } | |
616 | return 0; | |
617 | } | |
618 | #endif /* CONFIG_HIGHMEM */ | |
619 | ||
620 | ||
621 | static int save_highmem(void) | |
622 | { | |
623 | #ifdef CONFIG_HIGHMEM | |
624 | struct zone *zone; | |
625 | int res = 0; | |
626 | ||
627 | pr_debug("swsusp: Saving Highmem\n"); | |
2e4d5822 | 628 | for_each_zone (zone) { |
1da177e4 LT |
629 | if (is_highmem(zone)) |
630 | res = save_highmem_zone(zone); | |
631 | if (res) | |
632 | return res; | |
633 | } | |
634 | #endif | |
635 | return 0; | |
636 | } | |
637 | ||
638 | static int restore_highmem(void) | |
639 | { | |
640 | #ifdef CONFIG_HIGHMEM | |
641 | printk("swsusp: Restoring Highmem\n"); | |
642 | while (highmem_copy) { | |
643 | struct highmem_page *save = highmem_copy; | |
644 | void *kaddr; | |
645 | highmem_copy = save->next; | |
646 | ||
647 | kaddr = kmap_atomic(save->page, KM_USER0); | |
648 | memcpy(kaddr, save->data, PAGE_SIZE); | |
649 | kunmap_atomic(kaddr, KM_USER0); | |
650 | free_page((long) save->data); | |
651 | kfree(save); | |
652 | } | |
653 | #endif | |
654 | return 0; | |
655 | } | |
656 | ||
657 | ||
658 | static int pfn_is_nosave(unsigned long pfn) | |
659 | { | |
660 | unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT; | |
661 | unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT; | |
662 | return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn); | |
663 | } | |
664 | ||
665 | /** | |
666 | * saveable - Determine whether a page should be cloned or not. | |
667 | * @pfn: The page | |
668 | * | |
669 | * We save a page if it's Reserved, and not in the range of pages | |
670 | * statically defined as 'unsaveable', or if it isn't reserved, and | |
671 | * isn't part of a free chunk of pages. | |
672 | */ | |
673 | ||
674 | static int saveable(struct zone * zone, unsigned long * zone_pfn) | |
675 | { | |
676 | unsigned long pfn = *zone_pfn + zone->zone_start_pfn; | |
677 | struct page * page; | |
678 | ||
679 | if (!pfn_valid(pfn)) | |
680 | return 0; | |
681 | ||
682 | page = pfn_to_page(pfn); | |
1da177e4 LT |
683 | if (PageNosave(page)) |
684 | return 0; | |
b5810039 | 685 | if (pfn_is_nosave(pfn)) { |
1da177e4 LT |
686 | pr_debug("[nosave pfn 0x%lx]", pfn); |
687 | return 0; | |
688 | } | |
689 | if (PageNosaveFree(page)) | |
690 | return 0; | |
691 | ||
692 | return 1; | |
693 | } | |
694 | ||
695 | static void count_data_pages(void) | |
696 | { | |
697 | struct zone *zone; | |
698 | unsigned long zone_pfn; | |
699 | ||
700 | nr_copy_pages = 0; | |
701 | ||
2e4d5822 | 702 | for_each_zone (zone) { |
1da177e4 LT |
703 | if (is_highmem(zone)) |
704 | continue; | |
705 | mark_free_pages(zone); | |
706 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | |
707 | nr_copy_pages += saveable(zone, &zone_pfn); | |
708 | } | |
709 | } | |
710 | ||
711 | ||
712 | static void copy_data_pages(void) | |
713 | { | |
714 | struct zone *zone; | |
715 | unsigned long zone_pfn; | |
716 | struct pbe * pbe = pagedir_nosave; | |
2e4d5822 | 717 | |
1da177e4 | 718 | pr_debug("copy_data_pages(): pages to copy: %d\n", nr_copy_pages); |
2e4d5822 | 719 | for_each_zone (zone) { |
1da177e4 LT |
720 | if (is_highmem(zone)) |
721 | continue; | |
722 | mark_free_pages(zone); | |
723 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) { | |
724 | if (saveable(zone, &zone_pfn)) { | |
725 | struct page * page; | |
726 | page = pfn_to_page(zone_pfn + zone->zone_start_pfn); | |
727 | BUG_ON(!pbe); | |
728 | pbe->orig_address = (long) page_address(page); | |
729 | /* copy_page is not usable for copying task structs. */ | |
730 | memcpy((void *)pbe->address, (void *)pbe->orig_address, PAGE_SIZE); | |
731 | pbe = pbe->next; | |
732 | } | |
733 | } | |
734 | } | |
735 | BUG_ON(pbe); | |
736 | } | |
737 | ||
738 | ||
739 | /** | |
740 | * calc_nr - Determine the number of pages needed for a pbe list. | |
741 | */ | |
742 | ||
743 | static int calc_nr(int nr_copy) | |
744 | { | |
56057e1a | 745 | return nr_copy + (nr_copy+PBES_PER_PAGE-2)/(PBES_PER_PAGE-1); |
1da177e4 LT |
746 | } |
747 | ||
748 | /** | |
749 | * free_pagedir - free pages allocated with alloc_pagedir() | |
750 | */ | |
751 | ||
752 | static inline void free_pagedir(struct pbe *pblist) | |
753 | { | |
754 | struct pbe *pbe; | |
755 | ||
756 | while (pblist) { | |
757 | pbe = (pblist + PB_PAGE_SKIP)->next; | |
758 | free_page((unsigned long)pblist); | |
759 | pblist = pbe; | |
760 | } | |
761 | } | |
762 | ||
763 | /** | |
764 | * fill_pb_page - Create a list of PBEs on a given memory page | |
765 | */ | |
766 | ||
767 | static inline void fill_pb_page(struct pbe *pbpage) | |
768 | { | |
769 | struct pbe *p; | |
770 | ||
771 | p = pbpage; | |
772 | pbpage += PB_PAGE_SKIP; | |
773 | do | |
774 | p->next = p + 1; | |
775 | while (++p < pbpage); | |
776 | } | |
777 | ||
778 | /** | |
779 | * create_pbe_list - Create a list of PBEs on top of a given chain | |
780 | * of memory pages allocated with alloc_pagedir() | |
781 | */ | |
782 | ||
783 | static void create_pbe_list(struct pbe *pblist, unsigned nr_pages) | |
784 | { | |
785 | struct pbe *pbpage, *p; | |
786 | unsigned num = PBES_PER_PAGE; | |
787 | ||
788 | for_each_pb_page (pbpage, pblist) { | |
789 | if (num >= nr_pages) | |
790 | break; | |
791 | ||
792 | fill_pb_page(pbpage); | |
793 | num += PBES_PER_PAGE; | |
794 | } | |
795 | if (pbpage) { | |
796 | for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++) | |
797 | p->next = p + 1; | |
798 | p->next = NULL; | |
799 | } | |
800 | pr_debug("create_pbe_list(): initialized %d PBEs\n", num); | |
801 | } | |
802 | ||
803 | /** | |
804 | * alloc_pagedir - Allocate the page directory. | |
805 | * | |
806 | * First, determine exactly how many pages we need and | |
807 | * allocate them. | |
808 | * | |
809 | * We arrange the pages in a chain: each page is an array of PBES_PER_PAGE | |
810 | * struct pbe elements (pbes) and the last element in the page points | |
811 | * to the next page. | |
812 | * | |
813 | * On each page we set up a list of struct_pbe elements. | |
814 | */ | |
815 | ||
816 | static struct pbe * alloc_pagedir(unsigned nr_pages) | |
817 | { | |
818 | unsigned num; | |
819 | struct pbe *pblist, *pbe; | |
820 | ||
821 | if (!nr_pages) | |
822 | return NULL; | |
823 | ||
824 | pr_debug("alloc_pagedir(): nr_pages = %d\n", nr_pages); | |
825 | pblist = (struct pbe *)get_zeroed_page(GFP_ATOMIC | __GFP_COLD); | |
826 | for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages; | |
827 | pbe = pbe->next, num += PBES_PER_PAGE) { | |
828 | pbe += PB_PAGE_SKIP; | |
829 | pbe->next = (struct pbe *)get_zeroed_page(GFP_ATOMIC | __GFP_COLD); | |
830 | } | |
831 | if (!pbe) { /* get_zeroed_page() failed */ | |
832 | free_pagedir(pblist); | |
833 | pblist = NULL; | |
834 | } | |
835 | return pblist; | |
836 | } | |
837 | ||
838 | /** | |
839 | * free_image_pages - Free pages allocated for snapshot | |
840 | */ | |
841 | ||
842 | static void free_image_pages(void) | |
843 | { | |
844 | struct pbe * p; | |
845 | ||
2e4d5822 | 846 | for_each_pbe (p, pagedir_save) { |
1da177e4 LT |
847 | if (p->address) { |
848 | ClearPageNosave(virt_to_page(p->address)); | |
849 | free_page(p->address); | |
850 | p->address = 0; | |
851 | } | |
852 | } | |
853 | } | |
854 | ||
855 | /** | |
856 | * alloc_image_pages - Allocate pages for the snapshot. | |
857 | */ | |
858 | ||
859 | static int alloc_image_pages(void) | |
860 | { | |
861 | struct pbe * p; | |
862 | ||
2e4d5822 | 863 | for_each_pbe (p, pagedir_save) { |
1da177e4 LT |
864 | p->address = get_zeroed_page(GFP_ATOMIC | __GFP_COLD); |
865 | if (!p->address) | |
866 | return -ENOMEM; | |
867 | SetPageNosave(virt_to_page(p->address)); | |
868 | } | |
869 | return 0; | |
870 | } | |
871 | ||
8686bcd0 PM |
872 | /* Free pages we allocated for suspend. Suspend pages are alocated |
873 | * before atomic copy, so we need to free them after resume. | |
874 | */ | |
1da177e4 LT |
875 | void swsusp_free(void) |
876 | { | |
877 | BUG_ON(PageNosave(virt_to_page(pagedir_save))); | |
878 | BUG_ON(PageNosaveFree(virt_to_page(pagedir_save))); | |
879 | free_image_pages(); | |
880 | free_pagedir(pagedir_save); | |
881 | } | |
882 | ||
883 | ||
884 | /** | |
885 | * enough_free_mem - Make sure we enough free memory to snapshot. | |
886 | * | |
2e4d5822 | 887 | * Returns TRUE or FALSE after checking the number of available |
1da177e4 LT |
888 | * free pages. |
889 | */ | |
890 | ||
891 | static int enough_free_mem(void) | |
892 | { | |
893 | if (nr_free_pages() < (nr_copy_pages + PAGES_FOR_IO)) { | |
894 | pr_debug("swsusp: Not enough free pages: Have %d\n", | |
895 | nr_free_pages()); | |
896 | return 0; | |
897 | } | |
898 | return 1; | |
899 | } | |
900 | ||
901 | ||
902 | /** | |
903 | * enough_swap - Make sure we have enough swap to save the image. | |
904 | * | |
2e4d5822 | 905 | * Returns TRUE or FALSE after checking the total amount of swap |
1da177e4 LT |
906 | * space avaiable. |
907 | * | |
908 | * FIXME: si_swapinfo(&i) returns all swap devices information. | |
2e4d5822 | 909 | * We should only consider resume_device. |
1da177e4 LT |
910 | */ |
911 | ||
912 | static int enough_swap(void) | |
913 | { | |
914 | struct sysinfo i; | |
915 | ||
916 | si_swapinfo(&i); | |
917 | if (i.freeswap < (nr_copy_pages + PAGES_FOR_IO)) { | |
918 | pr_debug("swsusp: Not enough swap. Need %ld\n",i.freeswap); | |
919 | return 0; | |
920 | } | |
921 | return 1; | |
922 | } | |
923 | ||
924 | static int swsusp_alloc(void) | |
925 | { | |
926 | int error; | |
927 | ||
c61978b3 PM |
928 | pagedir_nosave = NULL; |
929 | nr_copy_pages = calc_nr(nr_copy_pages); | |
57487f43 | 930 | nr_copy_pages_check = nr_copy_pages; |
c61978b3 | 931 | |
1da177e4 LT |
932 | pr_debug("suspend: (pages needed: %d + %d free: %d)\n", |
933 | nr_copy_pages, PAGES_FOR_IO, nr_free_pages()); | |
934 | ||
1da177e4 LT |
935 | if (!enough_free_mem()) |
936 | return -ENOMEM; | |
937 | ||
938 | if (!enough_swap()) | |
939 | return -ENOSPC; | |
940 | ||
0f7347c2 RW |
941 | if (MAX_PBES < nr_copy_pages / PBES_PER_PAGE + |
942 | !!(nr_copy_pages % PBES_PER_PAGE)) | |
943 | return -ENOSPC; | |
944 | ||
1da177e4 LT |
945 | if (!(pagedir_save = alloc_pagedir(nr_copy_pages))) { |
946 | printk(KERN_ERR "suspend: Allocating pagedir failed.\n"); | |
947 | return -ENOMEM; | |
948 | } | |
949 | create_pbe_list(pagedir_save, nr_copy_pages); | |
950 | pagedir_nosave = pagedir_save; | |
951 | if ((error = alloc_image_pages())) { | |
952 | printk(KERN_ERR "suspend: Allocating image pages failed.\n"); | |
953 | swsusp_free(); | |
954 | return error; | |
955 | } | |
956 | ||
1da177e4 LT |
957 | return 0; |
958 | } | |
959 | ||
960 | static int suspend_prepare_image(void) | |
961 | { | |
962 | int error; | |
963 | ||
964 | pr_debug("swsusp: critical section: \n"); | |
965 | if (save_highmem()) { | |
966 | printk(KERN_CRIT "Suspend machine: Not enough free pages for highmem\n"); | |
967 | restore_highmem(); | |
968 | return -ENOMEM; | |
969 | } | |
970 | ||
971 | drain_local_pages(); | |
972 | count_data_pages(); | |
973 | printk("swsusp: Need to copy %u pages\n", nr_copy_pages); | |
974 | ||
975 | error = swsusp_alloc(); | |
976 | if (error) | |
977 | return error; | |
2e4d5822 PM |
978 | |
979 | /* During allocating of suspend pagedir, new cold pages may appear. | |
1da177e4 LT |
980 | * Kill them. |
981 | */ | |
982 | drain_local_pages(); | |
983 | copy_data_pages(); | |
984 | ||
985 | /* | |
986 | * End of critical section. From now on, we can write to memory, | |
987 | * but we should not touch disk. This specially means we must _not_ | |
988 | * touch swap space! Except we must write out our image of course. | |
989 | */ | |
990 | ||
991 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_copy_pages ); | |
992 | return 0; | |
993 | } | |
994 | ||
995 | ||
996 | /* It is important _NOT_ to umount filesystems at this point. We want | |
997 | * them synced (in case something goes wrong) but we DO not want to mark | |
998 | * filesystem clean: it is not. (And it does not matter, if we resume | |
999 | * correctly, we'll mark system clean, anyway.) | |
1000 | */ | |
1001 | int swsusp_write(void) | |
1002 | { | |
1003 | int error; | |
1004 | device_resume(); | |
1005 | lock_swapdevices(); | |
1006 | error = write_suspend_image(); | |
1007 | /* This will unlock ignored swap devices since writing is finished */ | |
1008 | lock_swapdevices(); | |
1009 | return error; | |
1010 | ||
1011 | } | |
1012 | ||
1013 | ||
1014 | extern asmlinkage int swsusp_arch_suspend(void); | |
1015 | extern asmlinkage int swsusp_arch_resume(void); | |
1016 | ||
1017 | ||
1018 | asmlinkage int swsusp_save(void) | |
1019 | { | |
1da177e4 LT |
1020 | return suspend_prepare_image(); |
1021 | } | |
1022 | ||
1023 | int swsusp_suspend(void) | |
1024 | { | |
1025 | int error; | |
1026 | if ((error = arch_prepare_suspend())) | |
1027 | return error; | |
1028 | local_irq_disable(); | |
1029 | /* At this point, device_suspend() has been called, but *not* | |
1030 | * device_power_down(). We *must* device_power_down() now. | |
1031 | * Otherwise, drivers for some devices (e.g. interrupt controllers) | |
1032 | * become desynchronized with the actual state of the hardware | |
1033 | * at resume time, and evil weirdness ensues. | |
1034 | */ | |
1035 | if ((error = device_power_down(PMSG_FREEZE))) { | |
99dc7d63 | 1036 | printk(KERN_ERR "Some devices failed to power down, aborting suspend\n"); |
1da177e4 | 1037 | local_irq_enable(); |
1da177e4 LT |
1038 | return error; |
1039 | } | |
47b724f3 PM |
1040 | |
1041 | if ((error = swsusp_swap_check())) { | |
99dc7d63 PM |
1042 | printk(KERN_ERR "swsusp: cannot find swap device, try swapon -a.\n"); |
1043 | device_power_up(); | |
47b724f3 PM |
1044 | local_irq_enable(); |
1045 | return error; | |
1046 | } | |
1047 | ||
1da177e4 LT |
1048 | save_processor_state(); |
1049 | if ((error = swsusp_arch_suspend())) | |
99dc7d63 | 1050 | printk(KERN_ERR "Error %d suspending\n", error); |
1da177e4 LT |
1051 | /* Restore control flow magically appears here */ |
1052 | restore_processor_state(); | |
1053 | BUG_ON (nr_copy_pages_check != nr_copy_pages); | |
1054 | restore_highmem(); | |
1055 | device_power_up(); | |
1056 | local_irq_enable(); | |
1057 | return error; | |
1058 | } | |
1059 | ||
1060 | int swsusp_resume(void) | |
1061 | { | |
1062 | int error; | |
1063 | local_irq_disable(); | |
1064 | if (device_power_down(PMSG_FREEZE)) | |
1065 | printk(KERN_ERR "Some devices failed to power down, very bad\n"); | |
1066 | /* We'll ignore saved state, but this gets preempt count (etc) right */ | |
1067 | save_processor_state(); | |
1068 | error = swsusp_arch_resume(); | |
1069 | /* Code below is only ever reached in case of failure. Otherwise | |
1070 | * execution continues at place where swsusp_arch_suspend was called | |
1071 | */ | |
1072 | BUG_ON(!error); | |
1073 | restore_processor_state(); | |
1074 | restore_highmem(); | |
8446f1d3 | 1075 | touch_softlockup_watchdog(); |
1da177e4 LT |
1076 | device_power_up(); |
1077 | local_irq_enable(); | |
1078 | return error; | |
1079 | } | |
1080 | ||
1da177e4 LT |
1081 | /** |
1082 | * On resume, for storing the PBE list and the image, | |
1083 | * we can only use memory pages that do not conflict with the pages | |
1084 | * which had been used before suspend. | |
1085 | * | |
1086 | * We don't know which pages are usable until we allocate them. | |
1087 | * | |
1088 | * Allocated but unusable (ie eaten) memory pages are linked together | |
1089 | * to create a list, so that we can free them easily | |
1090 | * | |
1091 | * We could have used a type other than (void *) | |
1092 | * for this purpose, but ... | |
1093 | */ | |
1094 | static void **eaten_memory = NULL; | |
1095 | ||
1096 | static inline void eat_page(void *page) | |
1097 | { | |
1098 | void **c; | |
1099 | ||
1100 | c = eaten_memory; | |
1101 | eaten_memory = page; | |
1102 | *eaten_memory = c; | |
1103 | } | |
1104 | ||
9796fdd8 | 1105 | unsigned long get_usable_page(gfp_t gfp_mask) |
1da177e4 LT |
1106 | { |
1107 | unsigned long m; | |
1108 | ||
1109 | m = get_zeroed_page(gfp_mask); | |
8f9bdf15 | 1110 | while (!PageNosaveFree(virt_to_page(m))) { |
1da177e4 LT |
1111 | eat_page((void *)m); |
1112 | m = get_zeroed_page(gfp_mask); | |
1113 | if (!m) | |
1114 | break; | |
1115 | } | |
1116 | return m; | |
1117 | } | |
1118 | ||
3dd08325 | 1119 | void free_eaten_memory(void) |
1da177e4 LT |
1120 | { |
1121 | unsigned long m; | |
1122 | void **c; | |
1123 | int i = 0; | |
1124 | ||
1125 | c = eaten_memory; | |
1126 | while (c) { | |
1127 | m = (unsigned long)c; | |
1128 | c = *c; | |
1129 | free_page(m); | |
1130 | i++; | |
1131 | } | |
1132 | eaten_memory = NULL; | |
1133 | pr_debug("swsusp: %d unused pages freed\n", i); | |
1134 | } | |
1135 | ||
1136 | /** | |
1137 | * check_pagedir - We ensure here that pages that the PBEs point to | |
1138 | * won't collide with pages where we're going to restore from the loaded | |
1139 | * pages later | |
1140 | */ | |
1141 | ||
1142 | static int check_pagedir(struct pbe *pblist) | |
1143 | { | |
1144 | struct pbe *p; | |
1145 | ||
1146 | /* This is necessary, so that we can free allocated pages | |
1147 | * in case of failure | |
1148 | */ | |
1149 | for_each_pbe (p, pblist) | |
1150 | p->address = 0UL; | |
1151 | ||
1152 | for_each_pbe (p, pblist) { | |
1153 | p->address = get_usable_page(GFP_ATOMIC); | |
1154 | if (!p->address) | |
1155 | return -ENOMEM; | |
1156 | } | |
1157 | return 0; | |
1158 | } | |
1159 | ||
1160 | /** | |
1161 | * swsusp_pagedir_relocate - It is possible, that some memory pages | |
1162 | * occupied by the list of PBEs collide with pages where we're going to | |
1163 | * restore from the loaded pages later. We relocate them here. | |
1164 | */ | |
1165 | ||
1166 | static struct pbe * swsusp_pagedir_relocate(struct pbe *pblist) | |
1167 | { | |
1168 | struct zone *zone; | |
1169 | unsigned long zone_pfn; | |
1170 | struct pbe *pbpage, *tail, *p; | |
1171 | void *m; | |
1172 | int rel = 0, error = 0; | |
1173 | ||
1174 | if (!pblist) /* a sanity check */ | |
1175 | return NULL; | |
1176 | ||
1177 | pr_debug("swsusp: Relocating pagedir (%lu pages to check)\n", | |
1178 | swsusp_info.pagedir_pages); | |
1179 | ||
1180 | /* Set page flags */ | |
1181 | ||
2e4d5822 | 1182 | for_each_zone (zone) { |
1da177e4 LT |
1183 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) |
1184 | SetPageNosaveFree(pfn_to_page(zone_pfn + | |
1185 | zone->zone_start_pfn)); | |
1186 | } | |
1187 | ||
1188 | /* Clear orig addresses */ | |
1189 | ||
1190 | for_each_pbe (p, pblist) | |
1191 | ClearPageNosaveFree(virt_to_page(p->orig_address)); | |
1192 | ||
1193 | tail = pblist + PB_PAGE_SKIP; | |
1194 | ||
1195 | /* Relocate colliding pages */ | |
1196 | ||
1197 | for_each_pb_page (pbpage, pblist) { | |
8f9bdf15 | 1198 | if (!PageNosaveFree(virt_to_page((unsigned long)pbpage))) { |
1da177e4 LT |
1199 | m = (void *)get_usable_page(GFP_ATOMIC | __GFP_COLD); |
1200 | if (!m) { | |
1201 | error = -ENOMEM; | |
1202 | break; | |
1203 | } | |
1204 | memcpy(m, (void *)pbpage, PAGE_SIZE); | |
1205 | if (pbpage == pblist) | |
1206 | pblist = (struct pbe *)m; | |
1207 | else | |
1208 | tail->next = (struct pbe *)m; | |
1209 | ||
1210 | eat_page((void *)pbpage); | |
1211 | pbpage = (struct pbe *)m; | |
1212 | ||
1213 | /* We have to link the PBEs again */ | |
1214 | ||
1215 | for (p = pbpage; p < pbpage + PB_PAGE_SKIP; p++) | |
1216 | if (p->next) /* needed to save the end */ | |
1217 | p->next = p + 1; | |
1218 | ||
1219 | rel++; | |
1220 | } | |
1221 | tail = pbpage + PB_PAGE_SKIP; | |
1222 | } | |
1223 | ||
1224 | if (error) { | |
1225 | printk("\nswsusp: Out of memory\n\n"); | |
1226 | free_pagedir(pblist); | |
1227 | free_eaten_memory(); | |
1228 | pblist = NULL; | |
8686bcd0 PM |
1229 | /* Is this even worth handling? It should never ever happen, and we |
1230 | have just lost user's state, anyway... */ | |
1231 | } else | |
1da177e4 LT |
1232 | printk("swsusp: Relocated %d pages\n", rel); |
1233 | ||
1234 | return pblist; | |
1235 | } | |
1236 | ||
4dc3b16b | 1237 | /* |
1da177e4 LT |
1238 | * Using bio to read from swap. |
1239 | * This code requires a bit more work than just using buffer heads | |
1240 | * but, it is the recommended way for 2.5/2.6. | |
1241 | * The following are to signal the beginning and end of I/O. Bios | |
1242 | * finish asynchronously, while we want them to happen synchronously. | |
1243 | * A simple atomic_t, and a wait loop take care of this problem. | |
1244 | */ | |
1245 | ||
1246 | static atomic_t io_done = ATOMIC_INIT(0); | |
1247 | ||
1248 | static int end_io(struct bio * bio, unsigned int num, int err) | |
1249 | { | |
1250 | if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) | |
1251 | panic("I/O error reading memory image"); | |
1252 | atomic_set(&io_done, 0); | |
1253 | return 0; | |
1254 | } | |
1255 | ||
1256 | static struct block_device * resume_bdev; | |
1257 | ||
1258 | /** | |
1259 | * submit - submit BIO request. | |
1260 | * @rw: READ or WRITE. | |
1261 | * @off physical offset of page. | |
1262 | * @page: page we're reading or writing. | |
1263 | * | |
1264 | * Straight from the textbook - allocate and initialize the bio. | |
1265 | * If we're writing, make sure the page is marked as dirty. | |
1266 | * Then submit it and wait. | |
1267 | */ | |
1268 | ||
1269 | static int submit(int rw, pgoff_t page_off, void * page) | |
1270 | { | |
1271 | int error = 0; | |
1272 | struct bio * bio; | |
1273 | ||
1274 | bio = bio_alloc(GFP_ATOMIC, 1); | |
1275 | if (!bio) | |
1276 | return -ENOMEM; | |
1277 | bio->bi_sector = page_off * (PAGE_SIZE >> 9); | |
1278 | bio_get(bio); | |
1279 | bio->bi_bdev = resume_bdev; | |
1280 | bio->bi_end_io = end_io; | |
1281 | ||
1282 | if (bio_add_page(bio, virt_to_page(page), PAGE_SIZE, 0) < PAGE_SIZE) { | |
1283 | printk("swsusp: ERROR: adding page to bio at %ld\n",page_off); | |
1284 | error = -EFAULT; | |
1285 | goto Done; | |
1286 | } | |
1287 | ||
1288 | if (rw == WRITE) | |
1289 | bio_set_pages_dirty(bio); | |
1290 | ||
1291 | atomic_set(&io_done, 1); | |
1292 | submit_bio(rw | (1 << BIO_RW_SYNC), bio); | |
1293 | while (atomic_read(&io_done)) | |
1294 | yield(); | |
1295 | ||
1296 | Done: | |
1297 | bio_put(bio); | |
1298 | return error; | |
1299 | } | |
1300 | ||
1301 | static int bio_read_page(pgoff_t page_off, void * page) | |
1302 | { | |
1303 | return submit(READ, page_off, page); | |
1304 | } | |
1305 | ||
1306 | static int bio_write_page(pgoff_t page_off, void * page) | |
1307 | { | |
1308 | return submit(WRITE, page_off, page); | |
1309 | } | |
1310 | ||
1311 | /* | |
1312 | * Sanity check if this image makes sense with this kernel/swap context | |
1313 | * I really don't think that it's foolproof but more than nothing.. | |
1314 | */ | |
1315 | ||
1316 | static const char * sanity_check(void) | |
1317 | { | |
1318 | dump_info(); | |
47b724f3 | 1319 | if (swsusp_info.version_code != LINUX_VERSION_CODE) |
1da177e4 | 1320 | return "kernel version"; |
47b724f3 | 1321 | if (swsusp_info.num_physpages != num_physpages) |
1da177e4 LT |
1322 | return "memory size"; |
1323 | if (strcmp(swsusp_info.uts.sysname,system_utsname.sysname)) | |
1324 | return "system type"; | |
1325 | if (strcmp(swsusp_info.uts.release,system_utsname.release)) | |
1326 | return "kernel release"; | |
1327 | if (strcmp(swsusp_info.uts.version,system_utsname.version)) | |
1328 | return "version"; | |
1329 | if (strcmp(swsusp_info.uts.machine,system_utsname.machine)) | |
1330 | return "machine"; | |
5a72e04d | 1331 | #if 0 |
99dc7d63 PM |
1332 | /* We can't use number of online CPUs when we use hotplug to remove them ;-))) */ |
1333 | if (swsusp_info.cpus != num_possible_cpus()) | |
1da177e4 | 1334 | return "number of cpus"; |
5a72e04d | 1335 | #endif |
1da177e4 LT |
1336 | return NULL; |
1337 | } | |
1338 | ||
1339 | ||
1340 | static int check_header(void) | |
1341 | { | |
1342 | const char * reason = NULL; | |
1343 | int error; | |
1344 | ||
1345 | if ((error = bio_read_page(swp_offset(swsusp_header.swsusp_info), &swsusp_info))) | |
1346 | return error; | |
1347 | ||
1348 | /* Is this same machine? */ | |
1349 | if ((reason = sanity_check())) { | |
1350 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n",reason); | |
1351 | return -EPERM; | |
1352 | } | |
1353 | nr_copy_pages = swsusp_info.image_pages; | |
1354 | return error; | |
1355 | } | |
1356 | ||
1357 | static int check_sig(void) | |
1358 | { | |
1359 | int error; | |
1360 | ||
1361 | memset(&swsusp_header, 0, sizeof(swsusp_header)); | |
1362 | if ((error = bio_read_page(0, &swsusp_header))) | |
1363 | return error; | |
1364 | if (!memcmp(SWSUSP_SIG, swsusp_header.sig, 10)) { | |
1365 | memcpy(swsusp_header.sig, swsusp_header.orig_sig, 10); | |
c2ff18f4 AS |
1366 | memcpy(key_iv, swsusp_header.key_iv, MAXKEY+MAXIV); |
1367 | memset(swsusp_header.key_iv, 0, MAXKEY+MAXIV); | |
1da177e4 LT |
1368 | |
1369 | /* | |
1370 | * Reset swap signature now. | |
1371 | */ | |
1372 | error = bio_write_page(0, &swsusp_header); | |
1373 | } else { | |
1da177e4 LT |
1374 | return -EINVAL; |
1375 | } | |
1376 | if (!error) | |
1377 | pr_debug("swsusp: Signature found, resuming\n"); | |
1378 | return error; | |
1379 | } | |
1380 | ||
1381 | /** | |
1382 | * data_read - Read image pages from swap. | |
1383 | * | |
1384 | * You do not need to check for overlaps, check_pagedir() | |
1385 | * already did that. | |
1386 | */ | |
1387 | ||
1388 | static int data_read(struct pbe *pblist) | |
1389 | { | |
1390 | struct pbe * p; | |
1391 | int error = 0; | |
1392 | int i = 0; | |
1393 | int mod = swsusp_info.image_pages / 100; | |
c2ff18f4 AS |
1394 | void *tfm; |
1395 | ||
1396 | if ((error = crypto_init(0, &tfm))) | |
1397 | return error; | |
1da177e4 LT |
1398 | |
1399 | if (!mod) | |
1400 | mod = 1; | |
1401 | ||
1402 | printk("swsusp: Reading image data (%lu pages): ", | |
1403 | swsusp_info.image_pages); | |
1404 | ||
1405 | for_each_pbe (p, pblist) { | |
1406 | if (!(i % mod)) | |
1407 | printk("\b\b\b\b%3d%%", i / mod); | |
1408 | ||
c2ff18f4 AS |
1409 | if ((error = crypto_read(p, tfm))) { |
1410 | crypto_exit(tfm); | |
1da177e4 | 1411 | return error; |
c2ff18f4 | 1412 | } |
1da177e4 LT |
1413 | |
1414 | i++; | |
1415 | } | |
1416 | printk("\b\b\b\bdone\n"); | |
c2ff18f4 | 1417 | crypto_exit(tfm); |
1da177e4 LT |
1418 | return error; |
1419 | } | |
1420 | ||
1da177e4 LT |
1421 | /** |
1422 | * read_pagedir - Read page backup list pages from swap | |
1423 | */ | |
1424 | ||
1425 | static int read_pagedir(struct pbe *pblist) | |
1426 | { | |
1427 | struct pbe *pbpage, *p; | |
1428 | unsigned i = 0; | |
1429 | int error; | |
1430 | ||
1431 | if (!pblist) | |
1432 | return -EFAULT; | |
1433 | ||
1434 | printk("swsusp: Reading pagedir (%lu pages)\n", | |
1435 | swsusp_info.pagedir_pages); | |
1436 | ||
1437 | for_each_pb_page (pbpage, pblist) { | |
1438 | unsigned long offset = swp_offset(swsusp_info.pagedir[i++]); | |
1439 | ||
1440 | error = -EFAULT; | |
1441 | if (offset) { | |
1442 | p = (pbpage + PB_PAGE_SKIP)->next; | |
1443 | error = bio_read_page(offset, (void *)pbpage); | |
1444 | (pbpage + PB_PAGE_SKIP)->next = p; | |
1445 | } | |
1446 | if (error) | |
1447 | break; | |
1448 | } | |
1449 | ||
1450 | if (error) | |
f2d61379 RW |
1451 | free_pagedir(pblist); |
1452 | else | |
1453 | BUG_ON(i != swsusp_info.pagedir_pages); | |
1da177e4 LT |
1454 | |
1455 | return error; | |
1456 | } | |
1457 | ||
1458 | ||
1459 | static int check_suspend_image(void) | |
1460 | { | |
1461 | int error = 0; | |
1462 | ||
1463 | if ((error = check_sig())) | |
1464 | return error; | |
1465 | ||
1466 | if ((error = check_header())) | |
1467 | return error; | |
1468 | ||
1469 | return 0; | |
1470 | } | |
1471 | ||
1472 | static int read_suspend_image(void) | |
1473 | { | |
1474 | int error = 0; | |
1475 | struct pbe *p; | |
1476 | ||
1477 | if (!(p = alloc_pagedir(nr_copy_pages))) | |
1478 | return -ENOMEM; | |
1479 | ||
1480 | if ((error = read_pagedir(p))) | |
1481 | return error; | |
1482 | ||
1483 | create_pbe_list(p, nr_copy_pages); | |
1484 | ||
1485 | if (!(pagedir_nosave = swsusp_pagedir_relocate(p))) | |
1486 | return -ENOMEM; | |
1487 | ||
1488 | /* Allocate memory for the image and read the data from swap */ | |
1489 | ||
1490 | error = check_pagedir(pagedir_nosave); | |
3dd08325 | 1491 | |
1da177e4 LT |
1492 | if (!error) |
1493 | error = data_read(pagedir_nosave); | |
1494 | ||
1495 | if (error) { /* We fail cleanly */ | |
3dd08325 | 1496 | free_eaten_memory(); |
1da177e4 LT |
1497 | for_each_pbe (p, pagedir_nosave) |
1498 | if (p->address) { | |
1499 | free_page(p->address); | |
1500 | p->address = 0UL; | |
1501 | } | |
1502 | free_pagedir(pagedir_nosave); | |
1503 | } | |
1504 | return error; | |
1505 | } | |
1506 | ||
1507 | /** | |
1508 | * swsusp_check - Check for saved image in swap | |
1509 | */ | |
1510 | ||
1511 | int swsusp_check(void) | |
1512 | { | |
1513 | int error; | |
1514 | ||
1da177e4 LT |
1515 | resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ); |
1516 | if (!IS_ERR(resume_bdev)) { | |
1517 | set_blocksize(resume_bdev, PAGE_SIZE); | |
1518 | error = check_suspend_image(); | |
1519 | if (error) | |
1520 | blkdev_put(resume_bdev); | |
1521 | } else | |
1522 | error = PTR_ERR(resume_bdev); | |
1523 | ||
1524 | if (!error) | |
1525 | pr_debug("swsusp: resume file found\n"); | |
1526 | else | |
1527 | pr_debug("swsusp: Error %d check for resume file\n", error); | |
1528 | return error; | |
1529 | } | |
1530 | ||
1531 | /** | |
1532 | * swsusp_read - Read saved image from swap. | |
1533 | */ | |
1534 | ||
1535 | int swsusp_read(void) | |
1536 | { | |
1537 | int error; | |
1538 | ||
1539 | if (IS_ERR(resume_bdev)) { | |
1540 | pr_debug("swsusp: block device not initialised\n"); | |
1541 | return PTR_ERR(resume_bdev); | |
1542 | } | |
1543 | ||
1544 | error = read_suspend_image(); | |
1545 | blkdev_put(resume_bdev); | |
c2ff18f4 | 1546 | memset(key_iv, 0, MAXKEY+MAXIV); |
1da177e4 LT |
1547 | |
1548 | if (!error) | |
1549 | pr_debug("swsusp: Reading resume file was successful\n"); | |
1550 | else | |
1551 | pr_debug("swsusp: Error %d resuming\n", error); | |
1552 | return error; | |
1553 | } | |
1554 | ||
1555 | /** | |
1556 | * swsusp_close - close swap device. | |
1557 | */ | |
1558 | ||
1559 | void swsusp_close(void) | |
1560 | { | |
1561 | if (IS_ERR(resume_bdev)) { | |
1562 | pr_debug("swsusp: block device not initialised\n"); | |
1563 | return; | |
1564 | } | |
1565 | ||
1566 | blkdev_put(resume_bdev); | |
1567 | } |