2 * linux/kernel/power/swap.c
4 * This file provides functions for reading the suspend image from
5 * and writing it to a swap partition.
7 * Copyright (C) 1998,2001-2005 Pavel Machek <pavel@suse.cz>
8 * Copyright (C) 2006 Rafael J. Wysocki <rjw@sisk.pl>
10 * This file is released under the GPLv2.
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>
26 #include <linux/slab.h>
30 #define SWSUSP_SIG "S1SUSPEND"
32 struct swsusp_header {
33 char reserved[PAGE_SIZE - 20 - sizeof(sector_t) - sizeof(int)];
35 unsigned int flags; /* Flags to pass to the "boot" kernel */
38 } __attribute__((packed));
40 static struct swsusp_header *swsusp_header;
43 * The following functions are used for tracing the allocated
44 * swap pages, so that they can be freed in case of an error.
47 struct swsusp_extent {
53 static struct rb_root swsusp_extents = RB_ROOT;
55 static int swsusp_extents_insert(unsigned long swap_offset)
57 struct rb_node **new = &(swsusp_extents.rb_node);
58 struct rb_node *parent = NULL;
59 struct swsusp_extent *ext;
61 /* Figure out where to put the new node */
63 ext = container_of(*new, struct swsusp_extent, node);
65 if (swap_offset < ext->start) {
67 if (swap_offset == ext->start - 1) {
71 new = &((*new)->rb_left);
72 } else if (swap_offset > ext->end) {
74 if (swap_offset == ext->end + 1) {
78 new = &((*new)->rb_right);
80 /* It already is in the tree */
84 /* Add the new node and rebalance the tree. */
85 ext = kzalloc(sizeof(struct swsusp_extent), GFP_KERNEL);
89 ext->start = swap_offset;
90 ext->end = swap_offset;
91 rb_link_node(&ext->node, parent, new);
92 rb_insert_color(&ext->node, &swsusp_extents);
97 * alloc_swapdev_block - allocate a swap page and register that it has
98 * been allocated, so that it can be freed in case of an error.
101 sector_t alloc_swapdev_block(int swap)
103 unsigned long offset;
105 offset = swp_offset(get_swap_page_of_type(swap));
107 if (swsusp_extents_insert(offset))
108 swap_free(swp_entry(swap, offset));
110 return swapdev_block(swap, offset);
116 * free_all_swap_pages - free swap pages allocated for saving image data.
117 * It also frees the extents used to register which swap entres had been
121 void free_all_swap_pages(int swap)
123 struct rb_node *node;
125 while ((node = swsusp_extents.rb_node)) {
126 struct swsusp_extent *ext;
127 unsigned long offset;
129 ext = container_of(node, struct swsusp_extent, node);
130 rb_erase(node, &swsusp_extents);
131 for (offset = ext->start; offset <= ext->end; offset++)
132 swap_free(swp_entry(swap, offset));
138 int swsusp_swap_in_use(void)
140 return (swsusp_extents.rb_node != NULL);
147 static unsigned short root_swap = 0xffff;
148 static struct block_device *resume_bdev;
151 * submit - submit BIO request.
152 * @rw: READ or WRITE.
153 * @off physical offset of page.
154 * @page: page we're reading or writing.
155 * @bio_chain: list of pending biod (for async reading)
157 * Straight from the textbook - allocate and initialize the bio.
158 * If we're reading, make sure the page is marked as dirty.
159 * Then submit it and, if @bio_chain == NULL, wait.
161 static int submit(int rw, pgoff_t page_off, struct page *page,
162 struct bio **bio_chain)
164 const int bio_rw = rw | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
167 bio = bio_alloc(__GFP_WAIT | __GFP_HIGH, 1);
168 bio->bi_sector = page_off * (PAGE_SIZE >> 9);
169 bio->bi_bdev = resume_bdev;
170 bio->bi_end_io = end_swap_bio_read;
172 if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
173 printk(KERN_ERR "PM: Adding page to bio failed at %ld\n",
182 if (bio_chain == NULL) {
183 submit_bio(bio_rw, bio);
184 wait_on_page_locked(page);
186 bio_set_pages_dirty(bio);
190 get_page(page); /* These pages are freed later */
191 bio->bi_private = *bio_chain;
193 submit_bio(bio_rw, bio);
198 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
200 return submit(READ, page_off, virt_to_page(addr), bio_chain);
203 static int bio_write_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
205 return submit(WRITE, page_off, virt_to_page(addr), bio_chain);
208 static int wait_on_bio_chain(struct bio **bio_chain)
211 struct bio *next_bio;
214 if (bio_chain == NULL)
223 next_bio = bio->bi_private;
224 page = bio->bi_io_vec[0].bv_page;
225 wait_on_page_locked(page);
226 if (!PageUptodate(page) || PageError(page))
240 static int mark_swapfiles(sector_t start, unsigned int flags)
244 bio_read_page(swsusp_resume_block, swsusp_header, NULL);
245 if (!memcmp("SWAP-SPACE",swsusp_header->sig, 10) ||
246 !memcmp("SWAPSPACE2",swsusp_header->sig, 10)) {
247 memcpy(swsusp_header->orig_sig,swsusp_header->sig, 10);
248 memcpy(swsusp_header->sig,SWSUSP_SIG, 10);
249 swsusp_header->image = start;
250 swsusp_header->flags = flags;
251 error = bio_write_page(swsusp_resume_block,
252 swsusp_header, NULL);
254 printk(KERN_ERR "PM: Swap header not found!\n");
261 * swsusp_swap_check - check if the resume device is a swap device
262 * and get its index (if so)
265 static int swsusp_swap_check(void) /* This is called before saving image */
269 res = swap_type_of(swsusp_resume_device, swsusp_resume_block,
275 res = blkdev_get(resume_bdev, FMODE_WRITE);
279 res = set_blocksize(resume_bdev, PAGE_SIZE);
281 blkdev_put(resume_bdev, FMODE_WRITE);
287 * write_page - Write one page to given swap location.
288 * @buf: Address we're writing.
289 * @offset: Offset of the swap page we're writing to.
290 * @bio_chain: Link the next write BIO here
293 static int write_page(void *buf, sector_t offset, struct bio **bio_chain)
301 src = (void *)__get_free_page(__GFP_WAIT | __GFP_HIGH);
303 memcpy(src, buf, PAGE_SIZE);
306 bio_chain = NULL; /* Go synchronous */
312 return bio_write_page(offset, src, bio_chain);
316 * The swap map is a data structure used for keeping track of each page
317 * written to a swap partition. It consists of many swap_map_page
318 * structures that contain each an array of MAP_PAGE_SIZE swap entries.
319 * These structures are stored on the swap and linked together with the
320 * help of the .next_swap member.
322 * The swap map is created during suspend. The swap map pages are
323 * allocated and populated one at a time, so we only need one memory
324 * page to set up the entire structure.
326 * During resume we also only need to use one swap_map_page structure
330 #define MAP_PAGE_ENTRIES (PAGE_SIZE / sizeof(sector_t) - 1)
332 struct swap_map_page {
333 sector_t entries[MAP_PAGE_ENTRIES];
338 * The swap_map_handle structure is used for handling swap in
342 struct swap_map_handle {
343 struct swap_map_page *cur;
348 static void release_swap_writer(struct swap_map_handle *handle)
351 free_page((unsigned long)handle->cur);
355 static int get_swap_writer(struct swap_map_handle *handle)
357 handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
360 handle->cur_swap = alloc_swapdev_block(root_swap);
361 if (!handle->cur_swap) {
362 release_swap_writer(handle);
369 static int swap_write_page(struct swap_map_handle *handle, void *buf,
370 struct bio **bio_chain)
377 offset = alloc_swapdev_block(root_swap);
378 error = write_page(buf, offset, bio_chain);
381 handle->cur->entries[handle->k++] = offset;
382 if (handle->k >= MAP_PAGE_ENTRIES) {
383 error = wait_on_bio_chain(bio_chain);
386 offset = alloc_swapdev_block(root_swap);
389 handle->cur->next_swap = offset;
390 error = write_page(handle->cur, handle->cur_swap, NULL);
393 memset(handle->cur, 0, PAGE_SIZE);
394 handle->cur_swap = offset;
401 static int flush_swap_writer(struct swap_map_handle *handle)
403 if (handle->cur && handle->cur_swap)
404 return write_page(handle->cur, handle->cur_swap, NULL);
410 * save_image - save the suspend image data
413 static int save_image(struct swap_map_handle *handle,
414 struct snapshot_handle *snapshot,
415 unsigned int nr_to_write)
422 struct timeval start;
425 printk(KERN_INFO "PM: Saving image data pages (%u pages) ... ",
427 m = nr_to_write / 100;
432 do_gettimeofday(&start);
434 ret = snapshot_read_next(snapshot, PAGE_SIZE);
437 ret = swap_write_page(handle, data_of(*snapshot), &bio);
441 printk(KERN_CONT "\b\b\b\b%3d%%", nr_pages / m);
444 err2 = wait_on_bio_chain(&bio);
445 do_gettimeofday(&stop);
449 printk(KERN_CONT "\b\b\b\bdone\n");
451 printk(KERN_CONT "\n");
452 swsusp_show_speed(&start, &stop, nr_to_write, "Wrote");
457 * enough_swap - Make sure we have enough swap to save the image.
459 * Returns TRUE or FALSE after checking the total amount of swap
460 * space avaiable from the resume partition.
463 static int enough_swap(unsigned int nr_pages)
465 unsigned int free_swap = count_swap_pages(root_swap, 1);
467 pr_debug("PM: Free swap pages: %u\n", free_swap);
468 return free_swap > nr_pages + PAGES_FOR_IO;
472 * swsusp_write - Write entire image and metadata.
473 * @flags: flags to pass to the "boot" kernel in the image header
475 * It is important _NOT_ to umount filesystems at this point. We want
476 * them synced (in case something goes wrong) but we DO not want to mark
477 * filesystem clean: it is not. (And it does not matter, if we resume
478 * correctly, we'll mark system clean, anyway.)
481 int swsusp_write(unsigned int flags)
483 struct swap_map_handle handle;
484 struct snapshot_handle snapshot;
485 struct swsusp_info *header;
488 error = swsusp_swap_check();
490 printk(KERN_ERR "PM: Cannot find swap device, try "
494 memset(&snapshot, 0, sizeof(struct snapshot_handle));
495 error = snapshot_read_next(&snapshot, PAGE_SIZE);
496 if (error < PAGE_SIZE) {
502 header = (struct swsusp_info *)data_of(snapshot);
503 if (!enough_swap(header->pages)) {
504 printk(KERN_ERR "PM: Not enough free swap\n");
508 error = get_swap_writer(&handle);
510 sector_t start = handle.cur_swap;
512 error = swap_write_page(&handle, header, NULL);
514 error = save_image(&handle, &snapshot,
518 flush_swap_writer(&handle);
519 printk(KERN_INFO "PM: S");
520 error = mark_swapfiles(start, flags);
525 free_all_swap_pages(root_swap);
527 release_swap_writer(&handle);
529 swsusp_close(FMODE_WRITE);
534 * The following functions allow us to read data using a swap map
535 * in a file-alike way
538 static void release_swap_reader(struct swap_map_handle *handle)
541 free_page((unsigned long)handle->cur);
545 static int get_swap_reader(struct swap_map_handle *handle, sector_t start)
552 handle->cur = (struct swap_map_page *)get_zeroed_page(__GFP_WAIT | __GFP_HIGH);
556 error = bio_read_page(start, handle->cur, NULL);
558 release_swap_reader(handle);
565 static int swap_read_page(struct swap_map_handle *handle, void *buf,
566 struct bio **bio_chain)
573 offset = handle->cur->entries[handle->k];
576 error = bio_read_page(offset, buf, bio_chain);
579 if (++handle->k >= MAP_PAGE_ENTRIES) {
580 error = wait_on_bio_chain(bio_chain);
582 offset = handle->cur->next_swap;
584 release_swap_reader(handle);
586 error = bio_read_page(offset, handle->cur, NULL);
592 * load_image - load the image using the swap map handle
593 * @handle and the snapshot handle @snapshot
594 * (assume there are @nr_pages pages to load)
597 static int load_image(struct swap_map_handle *handle,
598 struct snapshot_handle *snapshot,
599 unsigned int nr_to_read)
603 struct timeval start;
609 printk(KERN_INFO "PM: Loading image data pages (%u pages) ... ",
611 m = nr_to_read / 100;
616 do_gettimeofday(&start);
618 error = snapshot_write_next(snapshot, PAGE_SIZE);
621 error = swap_read_page(handle, data_of(*snapshot), &bio);
624 if (snapshot->sync_read)
625 error = wait_on_bio_chain(&bio);
629 printk("\b\b\b\b%3d%%", nr_pages / m);
632 err2 = wait_on_bio_chain(&bio);
633 do_gettimeofday(&stop);
637 printk("\b\b\b\bdone\n");
638 snapshot_write_finalize(snapshot);
639 if (!snapshot_image_loaded(snapshot))
643 swsusp_show_speed(&start, &stop, nr_to_read, "Read");
648 * swsusp_read - read the hibernation image.
649 * @flags_p: flags passed by the "frozen" kernel in the image header should
650 * be written into this memeory location
653 int swsusp_read(unsigned int *flags_p)
656 struct swap_map_handle handle;
657 struct snapshot_handle snapshot;
658 struct swsusp_info *header;
660 *flags_p = swsusp_header->flags;
662 memset(&snapshot, 0, sizeof(struct snapshot_handle));
663 error = snapshot_write_next(&snapshot, PAGE_SIZE);
664 if (error < PAGE_SIZE)
665 return error < 0 ? error : -EFAULT;
666 header = (struct swsusp_info *)data_of(snapshot);
667 error = get_swap_reader(&handle, swsusp_header->image);
669 error = swap_read_page(&handle, header, NULL);
671 error = load_image(&handle, &snapshot, header->pages - 1);
672 release_swap_reader(&handle);
675 pr_debug("PM: Image successfully loaded\n");
677 pr_debug("PM: Error %d resuming\n", error);
682 * swsusp_check - Check for swsusp signature in the resume device
685 int swsusp_check(void)
689 resume_bdev = open_by_devnum(swsusp_resume_device, FMODE_READ);
690 if (!IS_ERR(resume_bdev)) {
691 set_blocksize(resume_bdev, PAGE_SIZE);
692 memset(swsusp_header, 0, PAGE_SIZE);
693 error = bio_read_page(swsusp_resume_block,
694 swsusp_header, NULL);
698 if (!memcmp(SWSUSP_SIG, swsusp_header->sig, 10)) {
699 memcpy(swsusp_header->sig, swsusp_header->orig_sig, 10);
700 /* Reset swap signature now */
701 error = bio_write_page(swsusp_resume_block,
702 swsusp_header, NULL);
709 blkdev_put(resume_bdev, FMODE_READ);
711 pr_debug("PM: Signature found, resuming\n");
713 error = PTR_ERR(resume_bdev);
717 pr_debug("PM: Error %d checking image file\n", error);
723 * swsusp_close - close swap device.
726 void swsusp_close(fmode_t mode)
728 if (IS_ERR(resume_bdev)) {
729 pr_debug("PM: Image device not initialised\n");
733 blkdev_put(resume_bdev, mode);
736 static int swsusp_header_init(void)
738 swsusp_header = (struct swsusp_header*) __get_free_page(GFP_KERNEL);
740 panic("Could not allocate memory for swsusp_header\n");
744 core_initcall(swsusp_header_init);