sched: Fix proc_sched_set_task()
[deliverable/linux.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/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 * The following functions are used for tracing the allocated
43 * swap pages, so that they can be freed in case of an error.
44 */
45
46 struct swsusp_extent {
47 struct rb_node node;
48 unsigned long start;
49 unsigned long end;
50 };
51
52 static struct rb_root swsusp_extents = RB_ROOT;
53
54 static int swsusp_extents_insert(unsigned long swap_offset)
55 {
56 struct rb_node **new = &(swsusp_extents.rb_node);
57 struct rb_node *parent = NULL;
58 struct swsusp_extent *ext;
59
60 /* Figure out where to put the new node */
61 while (*new) {
62 ext = container_of(*new, struct swsusp_extent, node);
63 parent = *new;
64 if (swap_offset < ext->start) {
65 /* Try to merge */
66 if (swap_offset == ext->start - 1) {
67 ext->start--;
68 return 0;
69 }
70 new = &((*new)->rb_left);
71 } else if (swap_offset > ext->end) {
72 /* Try to merge */
73 if (swap_offset == ext->end + 1) {
74 ext->end++;
75 return 0;
76 }
77 new = &((*new)->rb_right);
78 } else {
79 /* It already is in the tree */
80 return -EINVAL;
81 }
82 }
83 /* Add the new node and rebalance the tree. */
84 ext = kzalloc(sizeof(struct swsusp_extent), GFP_KERNEL);
85 if (!ext)
86 return -ENOMEM;
87
88 ext->start = swap_offset;
89 ext->end = swap_offset;
90 rb_link_node(&ext->node, parent, new);
91 rb_insert_color(&ext->node, &swsusp_extents);
92 return 0;
93 }
94
95 /**
96 * alloc_swapdev_block - allocate a swap page and register that it has
97 * been allocated, so that it can be freed in case of an error.
98 */
99
100 sector_t alloc_swapdev_block(int swap)
101 {
102 unsigned long offset;
103
104 offset = swp_offset(get_swap_page_of_type(swap));
105 if (offset) {
106 if (swsusp_extents_insert(offset))
107 swap_free(swp_entry(swap, offset));
108 else
109 return swapdev_block(swap, offset);
110 }
111 return 0;
112 }
113
114 /**
115 * free_all_swap_pages - free swap pages allocated for saving image data.
116 * It also frees the extents used to register which swap entres had been
117 * allocated.
118 */
119
120 void free_all_swap_pages(int swap)
121 {
122 struct rb_node *node;
123
124 while ((node = swsusp_extents.rb_node)) {
125 struct swsusp_extent *ext;
126 unsigned long offset;
127
128 ext = container_of(node, struct swsusp_extent, node);
129 rb_erase(node, &swsusp_extents);
130 for (offset = ext->start; offset <= ext->end; offset++)
131 swap_free(swp_entry(swap, offset));
132
133 kfree(ext);
134 }
135 }
136
137 int swsusp_swap_in_use(void)
138 {
139 return (swsusp_extents.rb_node != NULL);
140 }
141
142 /*
143 * General things
144 */
145
146 static unsigned short root_swap = 0xffff;
147 static struct block_device *resume_bdev;
148
149 /**
150 * submit - submit BIO request.
151 * @rw: READ or WRITE.
152 * @off physical offset of page.
153 * @page: page we're reading or writing.
154 * @bio_chain: list of pending biod (for async reading)
155 *
156 * Straight from the textbook - allocate and initialize the bio.
157 * If we're reading, make sure the page is marked as dirty.
158 * Then submit it and, if @bio_chain == NULL, wait.
159 */
160 static int submit(int rw, pgoff_t page_off, struct page *page,
161 struct bio **bio_chain)
162 {
163 const int bio_rw = rw | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
164 struct bio *bio;
165
166 bio = bio_alloc(__GFP_WAIT | __GFP_HIGH, 1);
167 bio->bi_sector = page_off * (PAGE_SIZE >> 9);
168 bio->bi_bdev = resume_bdev;
169 bio->bi_end_io = end_swap_bio_read;
170
171 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
172 printk(KERN_ERR "PM: Adding page to bio failed at %ld\n",
173 page_off);
174 bio_put(bio);
175 return -EFAULT;
176 }
177
178 lock_page(page);
179 bio_get(bio);
180
181 if (bio_chain == NULL) {
182 submit_bio(bio_rw, bio);
183 wait_on_page_locked(page);
184 if (rw == READ)
185 bio_set_pages_dirty(bio);
186 bio_put(bio);
187 } else {
188 if (rw == READ)
189 get_page(page); /* These pages are freed later */
190 bio->bi_private = *bio_chain;
191 *bio_chain = bio;
192 submit_bio(bio_rw, bio);
193 }
194 return 0;
195 }
196
197 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
198 {
199 return submit(READ, page_off, virt_to_page(addr), bio_chain);
200 }
201
202 static int bio_write_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
203 {
204 return submit(WRITE, page_off, virt_to_page(addr), bio_chain);
205 }
206
207 static int wait_on_bio_chain(struct bio **bio_chain)
208 {
209 struct bio *bio;
210 struct bio *next_bio;
211 int ret = 0;
212
213 if (bio_chain == NULL)
214 return 0;
215
216 bio = *bio_chain;
217 if (bio == NULL)
218 return 0;
219 while (bio) {
220 struct page *page;
221
222 next_bio = bio->bi_private;
223 page = bio->bi_io_vec[0].bv_page;
224 wait_on_page_locked(page);
225 if (!PageUptodate(page) || PageError(page))
226 ret = -EIO;
227 put_page(page);
228 bio_put(bio);
229 bio = next_bio;
230 }
231 *bio_chain = NULL;
232 return ret;
233 }
234
235 /*
236 * Saving part
237 */
238
239 static int mark_swapfiles(sector_t start, unsigned int flags)
240 {
241 int error;
242
243 bio_read_page(swsusp_resume_block, swsusp_header, NULL);
244 if (!memcmp("SWAP-SPACE",swsusp_header->sig, 10) ||
245 !memcmp("SWAPSPACE2",swsusp_header->sig, 10)) {
246 memcpy(swsusp_header->orig_sig,swsusp_header->sig, 10);
247 memcpy(swsusp_header->sig,SWSUSP_SIG, 10);
248 swsusp_header->image = start;
249 swsusp_header->flags = flags;
250 error = bio_write_page(swsusp_resume_block,
251 swsusp_header, NULL);
252 } else {
253 printk(KERN_ERR "PM: Swap header not found!\n");
254 error = -ENODEV;
255 }
256 return error;
257 }
258
259 /**
260 * swsusp_swap_check - check if the resume device is a swap device
261 * and get its index (if so)
262 */
263
264 static int swsusp_swap_check(void) /* This is called before saving image */
265 {
266 int res;
267
268 res = swap_type_of(swsusp_resume_device, swsusp_resume_block,
269 &resume_bdev);
270 if (res < 0)
271 return res;
272
273 root_swap = res;
274 res = blkdev_get(resume_bdev, FMODE_WRITE);
275 if (res)
276 return res;
277
278 res = set_blocksize(resume_bdev, PAGE_SIZE);
279 if (res < 0)
280 blkdev_put(resume_bdev, FMODE_WRITE);
281
282 return res;
283 }
284
285 /**
286 * write_page - Write one page to given swap location.
287 * @buf: Address we're writing.
288 * @offset: Offset of the swap page we're writing to.
289 * @bio_chain: Link the next write BIO here
290 */
291
292 static int write_page(void *buf, sector_t offset, struct bio **bio_chain)
293 {
294 void *src;
295
296 if (!offset)
297 return -ENOSPC;
298
299 if (bio_chain) {
300 src = (void *)__get_free_page(__GFP_WAIT | __GFP_HIGH);
301 if (src) {
302 memcpy(src, buf, PAGE_SIZE);
303 } else {
304 WARN_ON_ONCE(1);
305 bio_chain = NULL; /* Go synchronous */
306 src = buf;
307 }
308 } else {
309 src = buf;
310 }
311 return bio_write_page(offset, src, bio_chain);
312 }
313
314 /*
315 * The swap map is a data structure used for keeping track of each page
316 * written to a swap partition. It consists of many swap_map_page
317 * structures that contain each an array of MAP_PAGE_SIZE swap entries.
318 * These structures are stored on the swap and linked together with the
319 * help of the .next_swap member.
320 *
321 * The swap map is created during suspend. The swap map pages are
322 * allocated and populated one at a time, so we only need one memory
323 * page to set up the entire structure.
324 *
325 * During resume we also only need to use one swap_map_page structure
326 * at a time.
327 */
328
329 #define MAP_PAGE_ENTRIES (PAGE_SIZE / sizeof(sector_t) - 1)
330
331 struct swap_map_page {
332 sector_t entries[MAP_PAGE_ENTRIES];
333 sector_t next_swap;
334 };
335
336 /**
337 * The swap_map_handle structure is used for handling swap in
338 * a file-alike way
339 */
340
341 struct swap_map_handle {
342 struct swap_map_page *cur;
343 sector_t cur_swap;
344 unsigned int k;
345 };
346
347 static void release_swap_writer(struct swap_map_handle *handle)
348 {
349 if (handle->cur)
350 free_page((unsigned long)handle->cur);
351 handle->cur = NULL;
352 }
353
354 static int get_swap_writer(struct swap_map_handle *handle)
355 {
356 handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
357 if (!handle->cur)
358 return -ENOMEM;
359 handle->cur_swap = alloc_swapdev_block(root_swap);
360 if (!handle->cur_swap) {
361 release_swap_writer(handle);
362 return -ENOSPC;
363 }
364 handle->k = 0;
365 return 0;
366 }
367
368 static int swap_write_page(struct swap_map_handle *handle, void *buf,
369 struct bio **bio_chain)
370 {
371 int error = 0;
372 sector_t offset;
373
374 if (!handle->cur)
375 return -EINVAL;
376 offset = alloc_swapdev_block(root_swap);
377 error = write_page(buf, offset, bio_chain);
378 if (error)
379 return error;
380 handle->cur->entries[handle->k++] = offset;
381 if (handle->k >= MAP_PAGE_ENTRIES) {
382 error = wait_on_bio_chain(bio_chain);
383 if (error)
384 goto out;
385 offset = alloc_swapdev_block(root_swap);
386 if (!offset)
387 return -ENOSPC;
388 handle->cur->next_swap = offset;
389 error = write_page(handle->cur, handle->cur_swap, NULL);
390 if (error)
391 goto out;
392 memset(handle->cur, 0, PAGE_SIZE);
393 handle->cur_swap = offset;
394 handle->k = 0;
395 }
396 out:
397 return error;
398 }
399
400 static int flush_swap_writer(struct swap_map_handle *handle)
401 {
402 if (handle->cur && handle->cur_swap)
403 return write_page(handle->cur, handle->cur_swap, NULL);
404 else
405 return -EINVAL;
406 }
407
408 /**
409 * save_image - save the suspend image data
410 */
411
412 static int save_image(struct swap_map_handle *handle,
413 struct snapshot_handle *snapshot,
414 unsigned int nr_to_write)
415 {
416 unsigned int m;
417 int ret;
418 int nr_pages;
419 int err2;
420 struct bio *bio;
421 struct timeval start;
422 struct timeval stop;
423
424 printk(KERN_INFO "PM: Saving image data pages (%u pages) ... ",
425 nr_to_write);
426 m = nr_to_write / 100;
427 if (!m)
428 m = 1;
429 nr_pages = 0;
430 bio = NULL;
431 do_gettimeofday(&start);
432 while (1) {
433 ret = snapshot_read_next(snapshot, PAGE_SIZE);
434 if (ret <= 0)
435 break;
436 ret = swap_write_page(handle, data_of(*snapshot), &bio);
437 if (ret)
438 break;
439 if (!(nr_pages % m))
440 printk(KERN_CONT "\b\b\b\b%3d%%", nr_pages / m);
441 nr_pages++;
442 }
443 err2 = wait_on_bio_chain(&bio);
444 do_gettimeofday(&stop);
445 if (!ret)
446 ret = err2;
447 if (!ret)
448 printk(KERN_CONT "\b\b\b\bdone\n");
449 else
450 printk(KERN_CONT "\n");
451 swsusp_show_speed(&start, &stop, nr_to_write, "Wrote");
452 return ret;
453 }
454
455 /**
456 * enough_swap - Make sure we have enough swap to save the image.
457 *
458 * Returns TRUE or FALSE after checking the total amount of swap
459 * space avaiable from the resume partition.
460 */
461
462 static int enough_swap(unsigned int nr_pages)
463 {
464 unsigned int free_swap = count_swap_pages(root_swap, 1);
465
466 pr_debug("PM: Free swap pages: %u\n", free_swap);
467 return free_swap > nr_pages + PAGES_FOR_IO;
468 }
469
470 /**
471 * swsusp_write - Write entire image and metadata.
472 * @flags: flags to pass to the "boot" kernel in the image header
473 *
474 * It is important _NOT_ to umount filesystems at this point. We want
475 * them synced (in case something goes wrong) but we DO not want to mark
476 * filesystem clean: it is not. (And it does not matter, if we resume
477 * correctly, we'll mark system clean, anyway.)
478 */
479
480 int swsusp_write(unsigned int flags)
481 {
482 struct swap_map_handle handle;
483 struct snapshot_handle snapshot;
484 struct swsusp_info *header;
485 int error;
486
487 error = swsusp_swap_check();
488 if (error) {
489 printk(KERN_ERR "PM: Cannot find swap device, try "
490 "swapon -a.\n");
491 return error;
492 }
493 memset(&snapshot, 0, sizeof(struct snapshot_handle));
494 error = snapshot_read_next(&snapshot, PAGE_SIZE);
495 if (error < PAGE_SIZE) {
496 if (error >= 0)
497 error = -EFAULT;
498
499 goto out;
500 }
501 header = (struct swsusp_info *)data_of(snapshot);
502 if (!enough_swap(header->pages)) {
503 printk(KERN_ERR "PM: Not enough free swap\n");
504 error = -ENOSPC;
505 goto out;
506 }
507 error = get_swap_writer(&handle);
508 if (!error) {
509 sector_t start = handle.cur_swap;
510
511 error = swap_write_page(&handle, header, NULL);
512 if (!error)
513 error = save_image(&handle, &snapshot,
514 header->pages - 1);
515
516 if (!error) {
517 flush_swap_writer(&handle);
518 printk(KERN_INFO "PM: S");
519 error = mark_swapfiles(start, flags);
520 printk("|\n");
521 }
522 }
523 if (error)
524 free_all_swap_pages(root_swap);
525
526 release_swap_writer(&handle);
527 out:
528 swsusp_close(FMODE_WRITE);
529 return error;
530 }
531
532 /**
533 * The following functions allow us to read data using a swap map
534 * in a file-alike way
535 */
536
537 static void release_swap_reader(struct swap_map_handle *handle)
538 {
539 if (handle->cur)
540 free_page((unsigned long)handle->cur);
541 handle->cur = NULL;
542 }
543
544 static int get_swap_reader(struct swap_map_handle *handle, sector_t start)
545 {
546 int error;
547
548 if (!start)
549 return -EINVAL;
550
551 handle->cur = (struct swap_map_page *)get_zeroed_page(__GFP_WAIT | __GFP_HIGH);
552 if (!handle->cur)
553 return -ENOMEM;
554
555 error = bio_read_page(start, handle->cur, NULL);
556 if (error) {
557 release_swap_reader(handle);
558 return error;
559 }
560 handle->k = 0;
561 return 0;
562 }
563
564 static int swap_read_page(struct swap_map_handle *handle, void *buf,
565 struct bio **bio_chain)
566 {
567 sector_t offset;
568 int error;
569
570 if (!handle->cur)
571 return -EINVAL;
572 offset = handle->cur->entries[handle->k];
573 if (!offset)
574 return -EFAULT;
575 error = bio_read_page(offset, buf, bio_chain);
576 if (error)
577 return error;
578 if (++handle->k >= MAP_PAGE_ENTRIES) {
579 error = wait_on_bio_chain(bio_chain);
580 handle->k = 0;
581 offset = handle->cur->next_swap;
582 if (!offset)
583 release_swap_reader(handle);
584 else if (!error)
585 error = bio_read_page(offset, handle->cur, NULL);
586 }
587 return error;
588 }
589
590 /**
591 * load_image - load the image using the swap map handle
592 * @handle and the snapshot handle @snapshot
593 * (assume there are @nr_pages pages to load)
594 */
595
596 static int load_image(struct swap_map_handle *handle,
597 struct snapshot_handle *snapshot,
598 unsigned int nr_to_read)
599 {
600 unsigned int m;
601 int error = 0;
602 struct timeval start;
603 struct timeval stop;
604 struct bio *bio;
605 int err2;
606 unsigned nr_pages;
607
608 printk(KERN_INFO "PM: Loading image data pages (%u pages) ... ",
609 nr_to_read);
610 m = nr_to_read / 100;
611 if (!m)
612 m = 1;
613 nr_pages = 0;
614 bio = NULL;
615 do_gettimeofday(&start);
616 for ( ; ; ) {
617 error = snapshot_write_next(snapshot, PAGE_SIZE);
618 if (error <= 0)
619 break;
620 error = swap_read_page(handle, data_of(*snapshot), &bio);
621 if (error)
622 break;
623 if (snapshot->sync_read)
624 error = wait_on_bio_chain(&bio);
625 if (error)
626 break;
627 if (!(nr_pages % m))
628 printk("\b\b\b\b%3d%%", nr_pages / m);
629 nr_pages++;
630 }
631 err2 = wait_on_bio_chain(&bio);
632 do_gettimeofday(&stop);
633 if (!error)
634 error = err2;
635 if (!error) {
636 printk("\b\b\b\bdone\n");
637 snapshot_write_finalize(snapshot);
638 if (!snapshot_image_loaded(snapshot))
639 error = -ENODATA;
640 } else
641 printk("\n");
642 swsusp_show_speed(&start, &stop, nr_to_read, "Read");
643 return error;
644 }
645
646 /**
647 * swsusp_read - read the hibernation image.
648 * @flags_p: flags passed by the "frozen" kernel in the image header should
649 * be written into this memeory location
650 */
651
652 int swsusp_read(unsigned int *flags_p)
653 {
654 int error;
655 struct swap_map_handle handle;
656 struct snapshot_handle snapshot;
657 struct swsusp_info *header;
658
659 *flags_p = swsusp_header->flags;
660
661 memset(&snapshot, 0, sizeof(struct snapshot_handle));
662 error = snapshot_write_next(&snapshot, PAGE_SIZE);
663 if (error < PAGE_SIZE)
664 return error < 0 ? error : -EFAULT;
665 header = (struct swsusp_info *)data_of(snapshot);
666 error = get_swap_reader(&handle, swsusp_header->image);
667 if (!error)
668 error = swap_read_page(&handle, header, NULL);
669 if (!error)
670 error = load_image(&handle, &snapshot, header->pages - 1);
671 release_swap_reader(&handle);
672
673 if (!error)
674 pr_debug("PM: Image successfully loaded\n");
675 else
676 pr_debug("PM: Error %d resuming\n", error);
677 return error;
678 }
679
680 /**
681 * swsusp_check - Check for swsusp signature in the resume device
682 */
683
684 int swsusp_check(void)
685 {
686 int error;
687
688 resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ);
689 if (!IS_ERR(resume_bdev)) {
690 set_blocksize(resume_bdev, PAGE_SIZE);
691 memset(swsusp_header, 0, PAGE_SIZE);
692 error = bio_read_page(swsusp_resume_block,
693 swsusp_header, NULL);
694 if (error)
695 goto put;
696
697 if (!memcmp(SWSUSP_SIG, swsusp_header->sig, 10)) {
698 memcpy(swsusp_header->sig, swsusp_header->orig_sig, 10);
699 /* Reset swap signature now */
700 error = bio_write_page(swsusp_resume_block,
701 swsusp_header, NULL);
702 } else {
703 error = -EINVAL;
704 }
705
706 put:
707 if (error)
708 blkdev_put(resume_bdev, FMODE_READ);
709 else
710 pr_debug("PM: Signature found, resuming\n");
711 } else {
712 error = PTR_ERR(resume_bdev);
713 }
714
715 if (error)
716 pr_debug("PM: Error %d checking image file\n", error);
717
718 return error;
719 }
720
721 /**
722 * swsusp_close - close swap device.
723 */
724
725 void swsusp_close(fmode_t mode)
726 {
727 if (IS_ERR(resume_bdev)) {
728 pr_debug("PM: Image device not initialised\n");
729 return;
730 }
731
732 blkdev_put(resume_bdev, mode);
733 }
734
735 static int swsusp_header_init(void)
736 {
737 swsusp_header = (struct swsusp_header*) __get_free_page(GFP_KERNEL);
738 if (!swsusp_header)
739 panic("Could not allocate memory for swsusp_header\n");
740 return 0;
741 }
742
743 core_initcall(swsusp_header_init);
This page took 0.058536 seconds and 5 git commands to generate.