Pull ia64-clocksource into release branch
[pandora-kernel.git] / 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/utsname.h>
17 #include <linux/version.h>
18 #include <linux/delay.h>
19 #include <linux/bitops.h>
20 #include <linux/genhd.h>
21 #include <linux/device.h>
22 #include <linux/buffer_head.h>
23 #include <linux/bio.h>
24 #include <linux/blkdev.h>
25 #include <linux/swap.h>
26 #include <linux/swapops.h>
27 #include <linux/pm.h>
28
29 #include "power.h"
30
31 extern char resume_file[];
32
33 #define SWSUSP_SIG      "S1SUSPEND"
34
35 struct swsusp_header {
36         char reserved[PAGE_SIZE - 20 - sizeof(sector_t) - sizeof(int)];
37         sector_t image;
38         unsigned int flags;     /* Flags to pass to the "boot" kernel */
39         char    orig_sig[10];
40         char    sig[10];
41 } __attribute__((packed));
42
43 static struct swsusp_header *swsusp_header;
44
45 /*
46  * General things
47  */
48
49 static unsigned short root_swap = 0xffff;
50 static struct block_device *resume_bdev;
51
52 /**
53  *      submit - submit BIO request.
54  *      @rw:    READ or WRITE.
55  *      @off    physical offset of page.
56  *      @page:  page we're reading or writing.
57  *      @bio_chain: list of pending biod (for async reading)
58  *
59  *      Straight from the textbook - allocate and initialize the bio.
60  *      If we're reading, make sure the page is marked as dirty.
61  *      Then submit it and, if @bio_chain == NULL, wait.
62  */
63 static int submit(int rw, pgoff_t page_off, struct page *page,
64                         struct bio **bio_chain)
65 {
66         struct bio *bio;
67
68         bio = bio_alloc(__GFP_WAIT | __GFP_HIGH, 1);
69         if (!bio)
70                 return -ENOMEM;
71         bio->bi_sector = page_off * (PAGE_SIZE >> 9);
72         bio->bi_bdev = resume_bdev;
73         bio->bi_end_io = end_swap_bio_read;
74
75         if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
76                 printk("swsusp: ERROR: adding page to bio at %ld\n", page_off);
77                 bio_put(bio);
78                 return -EFAULT;
79         }
80
81         lock_page(page);
82         bio_get(bio);
83
84         if (bio_chain == NULL) {
85                 submit_bio(rw | (1 << BIO_RW_SYNC), bio);
86                 wait_on_page_locked(page);
87                 if (rw == READ)
88                         bio_set_pages_dirty(bio);
89                 bio_put(bio);
90         } else {
91                 if (rw == READ)
92                         get_page(page); /* These pages are freed later */
93                 bio->bi_private = *bio_chain;
94                 *bio_chain = bio;
95                 submit_bio(rw | (1 << BIO_RW_SYNC), bio);
96         }
97         return 0;
98 }
99
100 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
101 {
102         return submit(READ, page_off, virt_to_page(addr), bio_chain);
103 }
104
105 static int bio_write_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
106 {
107         return submit(WRITE, page_off, virt_to_page(addr), bio_chain);
108 }
109
110 static int wait_on_bio_chain(struct bio **bio_chain)
111 {
112         struct bio *bio;
113         struct bio *next_bio;
114         int ret = 0;
115
116         if (bio_chain == NULL)
117                 return 0;
118
119         bio = *bio_chain;
120         if (bio == NULL)
121                 return 0;
122         while (bio) {
123                 struct page *page;
124
125                 next_bio = bio->bi_private;
126                 page = bio->bi_io_vec[0].bv_page;
127                 wait_on_page_locked(page);
128                 if (!PageUptodate(page) || PageError(page))
129                         ret = -EIO;
130                 put_page(page);
131                 bio_put(bio);
132                 bio = next_bio;
133         }
134         *bio_chain = NULL;
135         return ret;
136 }
137
138 /*
139  * Saving part
140  */
141
142 static int mark_swapfiles(sector_t start, unsigned int flags)
143 {
144         int error;
145
146         bio_read_page(swsusp_resume_block, swsusp_header, NULL);
147         if (!memcmp("SWAP-SPACE",swsusp_header->sig, 10) ||
148             !memcmp("SWAPSPACE2",swsusp_header->sig, 10)) {
149                 memcpy(swsusp_header->orig_sig,swsusp_header->sig, 10);
150                 memcpy(swsusp_header->sig,SWSUSP_SIG, 10);
151                 swsusp_header->image = start;
152                 swsusp_header->flags = flags;
153                 error = bio_write_page(swsusp_resume_block,
154                                         swsusp_header, NULL);
155         } else {
156                 printk(KERN_ERR "swsusp: Swap header not found!\n");
157                 error = -ENODEV;
158         }
159         return error;
160 }
161
162 /**
163  *      swsusp_swap_check - check if the resume device is a swap device
164  *      and get its index (if so)
165  */
166
167 static int swsusp_swap_check(void) /* This is called before saving image */
168 {
169         int res;
170
171         res = swap_type_of(swsusp_resume_device, swsusp_resume_block,
172                         &resume_bdev);
173         if (res < 0)
174                 return res;
175
176         root_swap = res;
177         res = blkdev_get(resume_bdev, FMODE_WRITE, O_RDWR);
178         if (res)
179                 return res;
180
181         res = set_blocksize(resume_bdev, PAGE_SIZE);
182         if (res < 0)
183                 blkdev_put(resume_bdev);
184
185         return res;
186 }
187
188 /**
189  *      write_page - Write one page to given swap location.
190  *      @buf:           Address we're writing.
191  *      @offset:        Offset of the swap page we're writing to.
192  *      @bio_chain:     Link the next write BIO here
193  */
194
195 static int write_page(void *buf, sector_t offset, struct bio **bio_chain)
196 {
197         void *src;
198
199         if (!offset)
200                 return -ENOSPC;
201
202         if (bio_chain) {
203                 src = (void *)__get_free_page(__GFP_WAIT | __GFP_HIGH);
204                 if (src) {
205                         memcpy(src, buf, PAGE_SIZE);
206                 } else {
207                         WARN_ON_ONCE(1);
208                         bio_chain = NULL;       /* Go synchronous */
209                         src = buf;
210                 }
211         } else {
212                 src = buf;
213         }
214         return bio_write_page(offset, src, bio_chain);
215 }
216
217 /*
218  *      The swap map is a data structure used for keeping track of each page
219  *      written to a swap partition.  It consists of many swap_map_page
220  *      structures that contain each an array of MAP_PAGE_SIZE swap entries.
221  *      These structures are stored on the swap and linked together with the
222  *      help of the .next_swap member.
223  *
224  *      The swap map is created during suspend.  The swap map pages are
225  *      allocated and populated one at a time, so we only need one memory
226  *      page to set up the entire structure.
227  *
228  *      During resume we also only need to use one swap_map_page structure
229  *      at a time.
230  */
231
232 #define MAP_PAGE_ENTRIES        (PAGE_SIZE / sizeof(sector_t) - 1)
233
234 struct swap_map_page {
235         sector_t entries[MAP_PAGE_ENTRIES];
236         sector_t next_swap;
237 };
238
239 /**
240  *      The swap_map_handle structure is used for handling swap in
241  *      a file-alike way
242  */
243
244 struct swap_map_handle {
245         struct swap_map_page *cur;
246         sector_t cur_swap;
247         unsigned int k;
248 };
249
250 static void release_swap_writer(struct swap_map_handle *handle)
251 {
252         if (handle->cur)
253                 free_page((unsigned long)handle->cur);
254         handle->cur = NULL;
255 }
256
257 static int get_swap_writer(struct swap_map_handle *handle)
258 {
259         handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
260         if (!handle->cur)
261                 return -ENOMEM;
262         handle->cur_swap = alloc_swapdev_block(root_swap);
263         if (!handle->cur_swap) {
264                 release_swap_writer(handle);
265                 return -ENOSPC;
266         }
267         handle->k = 0;
268         return 0;
269 }
270
271 static int swap_write_page(struct swap_map_handle *handle, void *buf,
272                                 struct bio **bio_chain)
273 {
274         int error = 0;
275         sector_t offset;
276
277         if (!handle->cur)
278                 return -EINVAL;
279         offset = alloc_swapdev_block(root_swap);
280         error = write_page(buf, offset, bio_chain);
281         if (error)
282                 return error;
283         handle->cur->entries[handle->k++] = offset;
284         if (handle->k >= MAP_PAGE_ENTRIES) {
285                 error = wait_on_bio_chain(bio_chain);
286                 if (error)
287                         goto out;
288                 offset = alloc_swapdev_block(root_swap);
289                 if (!offset)
290                         return -ENOSPC;
291                 handle->cur->next_swap = offset;
292                 error = write_page(handle->cur, handle->cur_swap, NULL);
293                 if (error)
294                         goto out;
295                 memset(handle->cur, 0, PAGE_SIZE);
296                 handle->cur_swap = offset;
297                 handle->k = 0;
298         }
299  out:
300         return error;
301 }
302
303 static int flush_swap_writer(struct swap_map_handle *handle)
304 {
305         if (handle->cur && handle->cur_swap)
306                 return write_page(handle->cur, handle->cur_swap, NULL);
307         else
308                 return -EINVAL;
309 }
310
311 /**
312  *      save_image - save the suspend image data
313  */
314
315 static int save_image(struct swap_map_handle *handle,
316                       struct snapshot_handle *snapshot,
317                       unsigned int nr_to_write)
318 {
319         unsigned int m;
320         int ret;
321         int error = 0;
322         int nr_pages;
323         int err2;
324         struct bio *bio;
325         struct timeval start;
326         struct timeval stop;
327
328         printk("Saving image data pages (%u pages) ...     ", nr_to_write);
329         m = nr_to_write / 100;
330         if (!m)
331                 m = 1;
332         nr_pages = 0;
333         bio = NULL;
334         do_gettimeofday(&start);
335         do {
336                 ret = snapshot_read_next(snapshot, PAGE_SIZE);
337                 if (ret > 0) {
338                         error = swap_write_page(handle, data_of(*snapshot),
339                                                 &bio);
340                         if (error)
341                                 break;
342                         if (!(nr_pages % m))
343                                 printk("\b\b\b\b%3d%%", nr_pages / m);
344                         nr_pages++;
345                 }
346         } while (ret > 0);
347         err2 = wait_on_bio_chain(&bio);
348         do_gettimeofday(&stop);
349         if (!error)
350                 error = err2;
351         if (!error)
352                 printk("\b\b\b\bdone\n");
353         swsusp_show_speed(&start, &stop, nr_to_write, "Wrote");
354         return error;
355 }
356
357 /**
358  *      enough_swap - Make sure we have enough swap to save the image.
359  *
360  *      Returns TRUE or FALSE after checking the total amount of swap
361  *      space avaiable from the resume partition.
362  */
363
364 static int enough_swap(unsigned int nr_pages)
365 {
366         unsigned int free_swap = count_swap_pages(root_swap, 1);
367
368         pr_debug("swsusp: free swap pages: %u\n", free_swap);
369         return free_swap > nr_pages + PAGES_FOR_IO;
370 }
371
372 /**
373  *      swsusp_write - Write entire image and metadata.
374  *      @flags: flags to pass to the "boot" kernel in the image header
375  *
376  *      It is important _NOT_ to umount filesystems at this point. We want
377  *      them synced (in case something goes wrong) but we DO not want to mark
378  *      filesystem clean: it is not. (And it does not matter, if we resume
379  *      correctly, we'll mark system clean, anyway.)
380  */
381
382 int swsusp_write(unsigned int flags)
383 {
384         struct swap_map_handle handle;
385         struct snapshot_handle snapshot;
386         struct swsusp_info *header;
387         int error;
388
389         error = swsusp_swap_check();
390         if (error) {
391                 printk(KERN_ERR "swsusp: Cannot find swap device, try "
392                                 "swapon -a.\n");
393                 return error;
394         }
395         memset(&snapshot, 0, sizeof(struct snapshot_handle));
396         error = snapshot_read_next(&snapshot, PAGE_SIZE);
397         if (error < PAGE_SIZE) {
398                 if (error >= 0)
399                         error = -EFAULT;
400
401                 goto out;
402         }
403         header = (struct swsusp_info *)data_of(snapshot);
404         if (!enough_swap(header->pages)) {
405                 printk(KERN_ERR "swsusp: Not enough free swap\n");
406                 error = -ENOSPC;
407                 goto out;
408         }
409         error = get_swap_writer(&handle);
410         if (!error) {
411                 sector_t start = handle.cur_swap;
412
413                 error = swap_write_page(&handle, header, NULL);
414                 if (!error)
415                         error = save_image(&handle, &snapshot,
416                                         header->pages - 1);
417
418                 if (!error) {
419                         flush_swap_writer(&handle);
420                         printk("S");
421                         error = mark_swapfiles(start, flags);
422                         printk("|\n");
423                 }
424         }
425         if (error)
426                 free_all_swap_pages(root_swap);
427
428         release_swap_writer(&handle);
429  out:
430         swsusp_close();
431         return error;
432 }
433
434 /**
435  *      The following functions allow us to read data using a swap map
436  *      in a file-alike way
437  */
438
439 static void release_swap_reader(struct swap_map_handle *handle)
440 {
441         if (handle->cur)
442                 free_page((unsigned long)handle->cur);
443         handle->cur = NULL;
444 }
445
446 static int get_swap_reader(struct swap_map_handle *handle, sector_t start)
447 {
448         int error;
449
450         if (!start)
451                 return -EINVAL;
452
453         handle->cur = (struct swap_map_page *)get_zeroed_page(__GFP_WAIT | __GFP_HIGH);
454         if (!handle->cur)
455                 return -ENOMEM;
456
457         error = bio_read_page(start, handle->cur, NULL);
458         if (error) {
459                 release_swap_reader(handle);
460                 return error;
461         }
462         handle->k = 0;
463         return 0;
464 }
465
466 static int swap_read_page(struct swap_map_handle *handle, void *buf,
467                                 struct bio **bio_chain)
468 {
469         sector_t offset;
470         int error;
471
472         if (!handle->cur)
473                 return -EINVAL;
474         offset = handle->cur->entries[handle->k];
475         if (!offset)
476                 return -EFAULT;
477         error = bio_read_page(offset, buf, bio_chain);
478         if (error)
479                 return error;
480         if (++handle->k >= MAP_PAGE_ENTRIES) {
481                 error = wait_on_bio_chain(bio_chain);
482                 handle->k = 0;
483                 offset = handle->cur->next_swap;
484                 if (!offset)
485                         release_swap_reader(handle);
486                 else if (!error)
487                         error = bio_read_page(offset, handle->cur, NULL);
488         }
489         return error;
490 }
491
492 /**
493  *      load_image - load the image using the swap map handle
494  *      @handle and the snapshot handle @snapshot
495  *      (assume there are @nr_pages pages to load)
496  */
497
498 static int load_image(struct swap_map_handle *handle,
499                       struct snapshot_handle *snapshot,
500                       unsigned int nr_to_read)
501 {
502         unsigned int m;
503         int error = 0;
504         struct timeval start;
505         struct timeval stop;
506         struct bio *bio;
507         int err2;
508         unsigned nr_pages;
509
510         printk("Loading image data pages (%u pages) ...     ", nr_to_read);
511         m = nr_to_read / 100;
512         if (!m)
513                 m = 1;
514         nr_pages = 0;
515         bio = NULL;
516         do_gettimeofday(&start);
517         for ( ; ; ) {
518                 error = snapshot_write_next(snapshot, PAGE_SIZE);
519                 if (error <= 0)
520                         break;
521                 error = swap_read_page(handle, data_of(*snapshot), &bio);
522                 if (error)
523                         break;
524                 if (snapshot->sync_read)
525                         error = wait_on_bio_chain(&bio);
526                 if (error)
527                         break;
528                 if (!(nr_pages % m))
529                         printk("\b\b\b\b%3d%%", nr_pages / m);
530                 nr_pages++;
531         }
532         err2 = wait_on_bio_chain(&bio);
533         do_gettimeofday(&stop);
534         if (!error)
535                 error = err2;
536         if (!error) {
537                 printk("\b\b\b\bdone\n");
538                 snapshot_write_finalize(snapshot);
539                 if (!snapshot_image_loaded(snapshot))
540                         error = -ENODATA;
541         }
542         swsusp_show_speed(&start, &stop, nr_to_read, "Read");
543         return error;
544 }
545
546 /**
547  *      swsusp_read - read the hibernation image.
548  *      @flags_p: flags passed by the "frozen" kernel in the image header should
549  *                be written into this memeory location
550  */
551
552 int swsusp_read(unsigned int *flags_p)
553 {
554         int error;
555         struct swap_map_handle handle;
556         struct snapshot_handle snapshot;
557         struct swsusp_info *header;
558
559         *flags_p = swsusp_header->flags;
560         if (IS_ERR(resume_bdev)) {
561                 pr_debug("swsusp: block device not initialised\n");
562                 return PTR_ERR(resume_bdev);
563         }
564
565         memset(&snapshot, 0, sizeof(struct snapshot_handle));
566         error = snapshot_write_next(&snapshot, PAGE_SIZE);
567         if (error < PAGE_SIZE)
568                 return error < 0 ? error : -EFAULT;
569         header = (struct swsusp_info *)data_of(snapshot);
570         error = get_swap_reader(&handle, swsusp_header->image);
571         if (!error)
572                 error = swap_read_page(&handle, header, NULL);
573         if (!error)
574                 error = load_image(&handle, &snapshot, header->pages - 1);
575         release_swap_reader(&handle);
576
577         blkdev_put(resume_bdev);
578
579         if (!error)
580                 pr_debug("swsusp: Reading resume file was successful\n");
581         else
582                 pr_debug("swsusp: Error %d resuming\n", error);
583         return error;
584 }
585
586 /**
587  *      swsusp_check - Check for swsusp signature in the resume device
588  */
589
590 int swsusp_check(void)
591 {
592         int error;
593
594         resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ);
595         if (!IS_ERR(resume_bdev)) {
596                 set_blocksize(resume_bdev, PAGE_SIZE);
597                 memset(swsusp_header, 0, PAGE_SIZE);
598                 error = bio_read_page(swsusp_resume_block,
599                                         swsusp_header, NULL);
600                 if (error)
601                         return error;
602
603                 if (!memcmp(SWSUSP_SIG, swsusp_header->sig, 10)) {
604                         memcpy(swsusp_header->sig, swsusp_header->orig_sig, 10);
605                         /* Reset swap signature now */
606                         error = bio_write_page(swsusp_resume_block,
607                                                 swsusp_header, NULL);
608                 } else {
609                         return -EINVAL;
610                 }
611                 if (error)
612                         blkdev_put(resume_bdev);
613                 else
614                         pr_debug("swsusp: Signature found, resuming\n");
615         } else {
616                 error = PTR_ERR(resume_bdev);
617         }
618
619         if (error)
620                 pr_debug("swsusp: Error %d check for resume file\n", error);
621
622         return error;
623 }
624
625 /**
626  *      swsusp_close - close swap device.
627  */
628
629 void swsusp_close(void)
630 {
631         if (IS_ERR(resume_bdev)) {
632                 pr_debug("swsusp: block device not initialised\n");
633                 return;
634         }
635
636         blkdev_put(resume_bdev);
637 }
638
639 static int swsusp_header_init(void)
640 {
641         swsusp_header = (struct swsusp_header*) __get_free_page(GFP_KERNEL);
642         if (!swsusp_header)
643                 panic("Could not allocate memory for swsusp_header\n");
644         return 0;
645 }
646
647 core_initcall(swsusp_header_init);