sched.c: Call debug_show_all_locks() when dumping all tasks
[safe/jmp/linux-2.6] / kernel / power / swap.c
1 /*
2  * linux/kernel/power/swap.c
3  *
4  * This file provides functions for reading the suspend image from
5  * and writing it to a swap partition.
6  *
7  * Copyright (C) 1998,2001-2005 Pavel Machek <pavel@suse.cz>
8  * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
9  *
10  * This file is released under the GPLv2.
11  *
12  */
13
14 #include <linux/module.h>
15 #include <linux/file.h>
16 #include <linux/delay.h>
17 #include <linux/bitops.h>
18 #include <linux/genhd.h>
19 #include <linux/device.h>
20 #include <linux/buffer_head.h>
21 #include <linux/bio.h>
22 #include <linux/blkdev.h>
23 #include <linux/swap.h>
24 #include <linux/swapops.h>
25 #include <linux/pm.h>
26
27 #include "power.h"
28
29 #define SWSUSP_SIG      "S1SUSPEND"
30
31 struct swsusp_header {
32         char reserved[PAGE_SIZE - 20 - sizeof(sector_t) - sizeof(int)];
33         sector_t image;
34         unsigned int flags;     /* Flags to pass to the "boot" kernel */
35         char    orig_sig[10];
36         char    sig[10];
37 } __attribute__((packed));
38
39 static struct swsusp_header *swsusp_header;
40
41 /*
42  * General things
43  */
44
45 static unsigned short root_swap = 0xffff;
46 static struct block_device *resume_bdev;
47
48 /**
49  *      submit - submit BIO request.
50  *      @rw:    READ or WRITE.
51  *      @off    physical offset of page.
52  *      @page:  page we're reading or writing.
53  *      @bio_chain: list of pending biod (for async reading)
54  *
55  *      Straight from the textbook - allocate and initialize the bio.
56  *      If we're reading, make sure the page is marked as dirty.
57  *      Then submit it and, if @bio_chain == NULL, wait.
58  */
59 static int submit(int rw, pgoff_t page_off, struct page *page,
60                         struct bio **bio_chain)
61 {
62         const int bio_rw = rw | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
63         struct bio *bio;
64
65         bio = bio_alloc(__GFP_WAIT | __GFP_HIGH, 1);
66         bio->bi_sector = page_off * (PAGE_SIZE >> 9);
67         bio->bi_bdev = resume_bdev;
68         bio->bi_end_io = end_swap_bio_read;
69
70         if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
71                 printk(KERN_ERR "PM: Adding page to bio failed at %ld\n",
72                         page_off);
73                 bio_put(bio);
74                 return -EFAULT;
75         }
76
77         lock_page(page);
78         bio_get(bio);
79
80         if (bio_chain == NULL) {
81                 submit_bio(bio_rw, bio);
82                 wait_on_page_locked(page);
83                 if (rw == READ)
84                         bio_set_pages_dirty(bio);
85                 bio_put(bio);
86         } else {
87                 if (rw == READ)
88                         get_page(page); /* These pages are freed later */
89                 bio->bi_private = *bio_chain;
90                 *bio_chain = bio;
91                 submit_bio(bio_rw, bio);
92         }
93         return 0;
94 }
95
96 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
97 {
98         return submit(READ, page_off, virt_to_page(addr), bio_chain);
99 }
100
101 static int bio_write_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
102 {
103         return submit(WRITE, page_off, virt_to_page(addr), bio_chain);
104 }
105
106 static int wait_on_bio_chain(struct bio **bio_chain)
107 {
108         struct bio *bio;
109         struct bio *next_bio;
110         int ret = 0;
111
112         if (bio_chain == NULL)
113                 return 0;
114
115         bio = *bio_chain;
116         if (bio == NULL)
117                 return 0;
118         while (bio) {
119                 struct page *page;
120
121                 next_bio = bio->bi_private;
122                 page = bio->bi_io_vec[0].bv_page;
123                 wait_on_page_locked(page);
124                 if (!PageUptodate(page) || PageError(page))
125                         ret = -EIO;
126                 put_page(page);
127                 bio_put(bio);
128                 bio = next_bio;
129         }
130         *bio_chain = NULL;
131         return ret;
132 }
133
134 /*
135  * Saving part
136  */
137
138 static int mark_swapfiles(sector_t start, unsigned int flags)
139 {
140         int error;
141
142         bio_read_page(swsusp_resume_block, swsusp_header, NULL);
143         if (!memcmp("SWAP-SPACE",swsusp_header->sig, 10) ||
144             !memcmp("SWAPSPACE2",swsusp_header->sig, 10)) {
145                 memcpy(swsusp_header->orig_sig,swsusp_header->sig, 10);
146                 memcpy(swsusp_header->sig,SWSUSP_SIG, 10);
147                 swsusp_header->image = start;
148                 swsusp_header->flags = flags;
149                 error = bio_write_page(swsusp_resume_block,
150                                         swsusp_header, NULL);
151         } else {
152                 printk(KERN_ERR "PM: Swap header not found!\n");
153                 error = -ENODEV;
154         }
155         return error;
156 }
157
158 /**
159  *      swsusp_swap_check - check if the resume device is a swap device
160  *      and get its index (if so)
161  */
162
163 static int swsusp_swap_check(void) /* This is called before saving image */
164 {
165         int res;
166
167         res = swap_type_of(swsusp_resume_device, swsusp_resume_block,
168                         &resume_bdev);
169         if (res < 0)
170                 return res;
171
172         root_swap = res;
173         res = blkdev_get(resume_bdev, FMODE_WRITE);
174         if (res)
175                 return res;
176
177         res = set_blocksize(resume_bdev, PAGE_SIZE);
178         if (res < 0)
179                 blkdev_put(resume_bdev, FMODE_WRITE);
180
181         return res;
182 }
183
184 /**
185  *      write_page - Write one page to given swap location.
186  *      @buf:           Address we're writing.
187  *      @offset:        Offset of the swap page we're writing to.
188  *      @bio_chain:     Link the next write BIO here
189  */
190
191 static int write_page(void *buf, sector_t offset, struct bio **bio_chain)
192 {
193         void *src;
194
195         if (!offset)
196                 return -ENOSPC;
197
198         if (bio_chain) {
199                 src = (void *)__get_free_page(__GFP_WAIT | __GFP_HIGH);
200                 if (src) {
201                         memcpy(src, buf, PAGE_SIZE);
202                 } else {
203                         WARN_ON_ONCE(1);
204                         bio_chain = NULL;       /* Go synchronous */
205                         src = buf;
206                 }
207         } else {
208                 src = buf;
209         }
210         return bio_write_page(offset, src, bio_chain);
211 }
212
213 /*
214  *      The swap map is a data structure used for keeping track of each page
215  *      written to a swap partition.  It consists of many swap_map_page
216  *      structures that contain each an array of MAP_PAGE_SIZE swap entries.
217  *      These structures are stored on the swap and linked together with the
218  *      help of the .next_swap member.
219  *
220  *      The swap map is created during suspend.  The swap map pages are
221  *      allocated and populated one at a time, so we only need one memory
222  *      page to set up the entire structure.
223  *
224  *      During resume we also only need to use one swap_map_page structure
225  *      at a time.
226  */
227
228 #define MAP_PAGE_ENTRIES        (PAGE_SIZE / sizeof(sector_t) - 1)
229
230 struct swap_map_page {
231         sector_t entries[MAP_PAGE_ENTRIES];
232         sector_t next_swap;
233 };
234
235 /**
236  *      The swap_map_handle structure is used for handling swap in
237  *      a file-alike way
238  */
239
240 struct swap_map_handle {
241         struct swap_map_page *cur;
242         sector_t cur_swap;
243         unsigned int k;
244 };
245
246 static void release_swap_writer(struct swap_map_handle *handle)
247 {
248         if (handle->cur)
249                 free_page((unsigned long)handle->cur);
250         handle->cur = NULL;
251 }
252
253 static int get_swap_writer(struct swap_map_handle *handle)
254 {
255         handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
256         if (!handle->cur)
257                 return -ENOMEM;
258         handle->cur_swap = alloc_swapdev_block(root_swap);
259         if (!handle->cur_swap) {
260                 release_swap_writer(handle);
261                 return -ENOSPC;
262         }
263         handle->k = 0;
264         return 0;
265 }
266
267 static int swap_write_page(struct swap_map_handle *handle, void *buf,
268                                 struct bio **bio_chain)
269 {
270         int error = 0;
271         sector_t offset;
272
273         if (!handle->cur)
274                 return -EINVAL;
275         offset = alloc_swapdev_block(root_swap);
276         error = write_page(buf, offset, bio_chain);
277         if (error)
278                 return error;
279         handle->cur->entries[handle->k++] = offset;
280         if (handle->k >= MAP_PAGE_ENTRIES) {
281                 error = wait_on_bio_chain(bio_chain);
282                 if (error)
283                         goto out;
284                 offset = alloc_swapdev_block(root_swap);
285                 if (!offset)
286                         return -ENOSPC;
287                 handle->cur->next_swap = offset;
288                 error = write_page(handle->cur, handle->cur_swap, NULL);
289                 if (error)
290                         goto out;
291                 memset(handle->cur, 0, PAGE_SIZE);
292                 handle->cur_swap = offset;
293                 handle->k = 0;
294         }
295  out:
296         return error;
297 }
298
299 static int flush_swap_writer(struct swap_map_handle *handle)
300 {
301         if (handle->cur && handle->cur_swap)
302                 return write_page(handle->cur, handle->cur_swap, NULL);
303         else
304                 return -EINVAL;
305 }
306
307 /**
308  *      save_image - save the suspend image data
309  */
310
311 static int save_image(struct swap_map_handle *handle,
312                       struct snapshot_handle *snapshot,
313                       unsigned int nr_to_write)
314 {
315         unsigned int m;
316         int ret;
317         int error = 0;
318         int nr_pages;
319         int err2;
320         struct bio *bio;
321         struct timeval start;
322         struct timeval stop;
323
324         printk(KERN_INFO "PM: Saving image data pages (%u pages) ...     ",
325                 nr_to_write);
326         m = nr_to_write / 100;
327         if (!m)
328                 m = 1;
329         nr_pages = 0;
330         bio = NULL;
331         do_gettimeofday(&start);
332         do {
333                 ret = snapshot_read_next(snapshot, PAGE_SIZE);
334                 if (ret > 0) {
335                         error = swap_write_page(handle, data_of(*snapshot),
336                                                 &bio);
337                         if (error)
338                                 break;
339                         if (!(nr_pages % m))
340                                 printk("\b\b\b\b%3d%%", nr_pages / m);
341                         nr_pages++;
342                 }
343         } while (ret > 0);
344         err2 = wait_on_bio_chain(&bio);
345         do_gettimeofday(&stop);
346         if (!error)
347                 error = err2;
348         if (!error)
349                 printk("\b\b\b\bdone\n");
350         swsusp_show_speed(&start, &stop, nr_to_write, "Wrote");
351         return error;
352 }
353
354 /**
355  *      enough_swap - Make sure we have enough swap to save the image.
356  *
357  *      Returns TRUE or FALSE after checking the total amount of swap
358  *      space avaiable from the resume partition.
359  */
360
361 static int enough_swap(unsigned int nr_pages)
362 {
363         unsigned int free_swap = count_swap_pages(root_swap, 1);
364
365         pr_debug("PM: Free swap pages: %u\n", free_swap);
366         return free_swap > nr_pages + PAGES_FOR_IO;
367 }
368
369 /**
370  *      swsusp_write - Write entire image and metadata.
371  *      @flags: flags to pass to the "boot" kernel in the image header
372  *
373  *      It is important _NOT_ to umount filesystems at this point. We want
374  *      them synced (in case something goes wrong) but we DO not want to mark
375  *      filesystem clean: it is not. (And it does not matter, if we resume
376  *      correctly, we'll mark system clean, anyway.)
377  */
378
379 int swsusp_write(unsigned int flags)
380 {
381         struct swap_map_handle handle;
382         struct snapshot_handle snapshot;
383         struct swsusp_info *header;
384         int error;
385
386         error = swsusp_swap_check();
387         if (error) {
388                 printk(KERN_ERR "PM: Cannot find swap device, try "
389                                 "swapon -a.\n");
390                 return error;
391         }
392         memset(&snapshot, 0, sizeof(struct snapshot_handle));
393         error = snapshot_read_next(&snapshot, PAGE_SIZE);
394         if (error < PAGE_SIZE) {
395                 if (error >= 0)
396                         error = -EFAULT;
397
398                 goto out;
399         }
400         header = (struct swsusp_info *)data_of(snapshot);
401         if (!enough_swap(header->pages)) {
402                 printk(KERN_ERR "PM: Not enough free swap\n");
403                 error = -ENOSPC;
404                 goto out;
405         }
406         error = get_swap_writer(&handle);
407         if (!error) {
408                 sector_t start = handle.cur_swap;
409
410                 error = swap_write_page(&handle, header, NULL);
411                 if (!error)
412                         error = save_image(&handle, &snapshot,
413                                         header->pages - 1);
414
415                 if (!error) {
416                         flush_swap_writer(&handle);
417                         printk(KERN_INFO "PM: S");
418                         error = mark_swapfiles(start, flags);
419                         printk("|\n");
420                 }
421         }
422         if (error)
423                 free_all_swap_pages(root_swap);
424
425         release_swap_writer(&handle);
426  out:
427         swsusp_close(FMODE_WRITE);
428         return error;
429 }
430
431 /**
432  *      The following functions allow us to read data using a swap map
433  *      in a file-alike way
434  */
435
436 static void release_swap_reader(struct swap_map_handle *handle)
437 {
438         if (handle->cur)
439                 free_page((unsigned long)handle->cur);
440         handle->cur = NULL;
441 }
442
443 static int get_swap_reader(struct swap_map_handle *handle, sector_t start)
444 {
445         int error;
446
447         if (!start)
448                 return -EINVAL;
449
450         handle->cur = (struct swap_map_page *)get_zeroed_page(__GFP_WAIT | __GFP_HIGH);
451         if (!handle->cur)
452                 return -ENOMEM;
453
454         error = bio_read_page(start, handle->cur, NULL);
455         if (error) {
456                 release_swap_reader(handle);
457                 return error;
458         }
459         handle->k = 0;
460         return 0;
461 }
462
463 static int swap_read_page(struct swap_map_handle *handle, void *buf,
464                                 struct bio **bio_chain)
465 {
466         sector_t offset;
467         int error;
468
469         if (!handle->cur)
470                 return -EINVAL;
471         offset = handle->cur->entries[handle->k];
472         if (!offset)
473                 return -EFAULT;
474         error = bio_read_page(offset, buf, bio_chain);
475         if (error)
476                 return error;
477         if (++handle->k >= MAP_PAGE_ENTRIES) {
478                 error = wait_on_bio_chain(bio_chain);
479                 handle->k = 0;
480                 offset = handle->cur->next_swap;
481                 if (!offset)
482                         release_swap_reader(handle);
483                 else if (!error)
484                         error = bio_read_page(offset, handle->cur, NULL);
485         }
486         return error;
487 }
488
489 /**
490  *      load_image - load the image using the swap map handle
491  *      @handle and the snapshot handle @snapshot
492  *      (assume there are @nr_pages pages to load)
493  */
494
495 static int load_image(struct swap_map_handle *handle,
496                       struct snapshot_handle *snapshot,
497                       unsigned int nr_to_read)
498 {
499         unsigned int m;
500         int error = 0;
501         struct timeval start;
502         struct timeval stop;
503         struct bio *bio;
504         int err2;
505         unsigned nr_pages;
506
507         printk(KERN_INFO "PM: Loading image data pages (%u pages) ...     ",
508                 nr_to_read);
509         m = nr_to_read / 100;
510         if (!m)
511                 m = 1;
512         nr_pages = 0;
513         bio = NULL;
514         do_gettimeofday(&start);
515         for ( ; ; ) {
516                 error = snapshot_write_next(snapshot, PAGE_SIZE);
517                 if (error <= 0)
518                         break;
519                 error = swap_read_page(handle, data_of(*snapshot), &bio);
520                 if (error)
521                         break;
522                 if (snapshot->sync_read)
523                         error = wait_on_bio_chain(&bio);
524                 if (error)
525                         break;
526                 if (!(nr_pages % m))
527                         printk("\b\b\b\b%3d%%", nr_pages / m);
528                 nr_pages++;
529         }
530         err2 = wait_on_bio_chain(&bio);
531         do_gettimeofday(&stop);
532         if (!error)
533                 error = err2;
534         if (!error) {
535                 printk("\b\b\b\bdone\n");
536                 snapshot_write_finalize(snapshot);
537                 if (!snapshot_image_loaded(snapshot))
538                         error = -ENODATA;
539         }
540         swsusp_show_speed(&start, &stop, nr_to_read, "Read");
541         return error;
542 }
543
544 /**
545  *      swsusp_read - read the hibernation image.
546  *      @flags_p: flags passed by the "frozen" kernel in the image header should
547  *                be written into this memeory location
548  */
549
550 int swsusp_read(unsigned int *flags_p)
551 {
552         int error;
553         struct swap_map_handle handle;
554         struct snapshot_handle snapshot;
555         struct swsusp_info *header;
556
557         *flags_p = swsusp_header->flags;
558         if (IS_ERR(resume_bdev)) {
559                 pr_debug("PM: Image device not initialised\n");
560                 return PTR_ERR(resume_bdev);
561         }
562
563         memset(&snapshot, 0, sizeof(struct snapshot_handle));
564         error = snapshot_write_next(&snapshot, PAGE_SIZE);
565         if (error < PAGE_SIZE)
566                 return error < 0 ? error : -EFAULT;
567         header = (struct swsusp_info *)data_of(snapshot);
568         error = get_swap_reader(&handle, swsusp_header->image);
569         if (!error)
570                 error = swap_read_page(&handle, header, NULL);
571         if (!error)
572                 error = load_image(&handle, &snapshot, header->pages - 1);
573         release_swap_reader(&handle);
574
575         blkdev_put(resume_bdev, FMODE_READ);
576
577         if (!error)
578                 pr_debug("PM: Image successfully loaded\n");
579         else
580                 pr_debug("PM: Error %d resuming\n", error);
581         return error;
582 }
583
584 /**
585  *      swsusp_check - Check for swsusp signature in the resume device
586  */
587
588 int swsusp_check(void)
589 {
590         int error;
591
592         resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ);
593         if (!IS_ERR(resume_bdev)) {
594                 set_blocksize(resume_bdev, PAGE_SIZE);
595                 memset(swsusp_header, 0, PAGE_SIZE);
596                 error = bio_read_page(swsusp_resume_block,
597                                         swsusp_header, NULL);
598                 if (error)
599                         return error;
600
601                 if (!memcmp(SWSUSP_SIG, swsusp_header->sig, 10)) {
602                         memcpy(swsusp_header->sig, swsusp_header->orig_sig, 10);
603                         /* Reset swap signature now */
604                         error = bio_write_page(swsusp_resume_block,
605                                                 swsusp_header, NULL);
606                 } else {
607                         return -EINVAL;
608                 }
609                 if (error)
610                         blkdev_put(resume_bdev, FMODE_READ);
611                 else
612                         pr_debug("PM: Signature found, resuming\n");
613         } else {
614                 error = PTR_ERR(resume_bdev);
615         }
616
617         if (error)
618                 pr_debug("PM: Error %d checking image file\n", error);
619
620         return error;
621 }
622
623 /**
624  *      swsusp_close - close swap device.
625  */
626
627 void swsusp_close(fmode_t mode)
628 {
629         if (IS_ERR(resume_bdev)) {
630                 pr_debug("PM: Image device not initialised\n");
631                 return;
632         }
633
634         blkdev_put(resume_bdev, mode);
635 }
636
637 static int swsusp_header_init(void)
638 {
639         swsusp_header = (struct swsusp_header*) __get_free_page(GFP_KERNEL);
640         if (!swsusp_header)
641                 panic("Could not allocate memory for swsusp_header\n");
642         return 0;
643 }
644
645 core_initcall(swsusp_header_init);