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>
29 #define SWSUSP_SIG "S1SUSPEND"
31 struct swsusp_header {
32 char reserved[PAGE_SIZE - 20 - sizeof(sector_t) - sizeof(int)];
34 unsigned int flags; /* Flags to pass to the "boot" kernel */
37 } __attribute__((packed));
39 static struct swsusp_header *swsusp_header;
42 * The following functions are used for tracing the allocated
43 * swap pages, so that they can be freed in case of an error.
46 struct swsusp_extent {
52 static struct rb_root swsusp_extents = RB_ROOT;
54 static int swsusp_extents_insert(unsigned long swap_offset)
56 struct rb_node **new = &(swsusp_extents.rb_node);
57 struct rb_node *parent = NULL;
58 struct swsusp_extent *ext;
60 /* Figure out where to put the new node */
62 ext = container_of(*new, struct swsusp_extent, node);
64 if (swap_offset < ext->start) {
66 if (swap_offset == ext->start - 1) {
70 new = &((*new)->rb_left);
71 } else if (swap_offset > ext->end) {
73 if (swap_offset == ext->end + 1) {
77 new = &((*new)->rb_right);
79 /* It already is in the tree */
83 /* Add the new node and rebalance the tree. */
84 ext = kzalloc(sizeof(struct swsusp_extent), GFP_KERNEL);
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);
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.
100 sector_t alloc_swapdev_block(int swap)
102 unsigned long offset;
104 offset = swp_offset(get_swap_page_of_type(swap));
106 if (swsusp_extents_insert(offset))
107 swap_free(swp_entry(swap, offset));
109 return swapdev_block(swap, offset);
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
120 void free_all_swap_pages(int swap)
122 struct rb_node *node;
124 while ((node = swsusp_extents.rb_node)) {
125 struct swsusp_extent *ext;
126 unsigned long offset;
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));
137 int swsusp_swap_in_use(void)
139 return (swsusp_extents.rb_node != NULL);
146 static unsigned short root_swap = 0xffff;
147 static struct block_device *resume_bdev;
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)
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.
160 static int submit(int rw, pgoff_t page_off, struct page *page,
161 struct bio **bio_chain)
163 const int bio_rw = rw | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
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;
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",
181 if (bio_chain == NULL) {
182 submit_bio(bio_rw, bio);
183 wait_on_page_locked(page);
185 bio_set_pages_dirty(bio);
189 get_page(page); /* These pages are freed later */
190 bio->bi_private = *bio_chain;
192 submit_bio(bio_rw, bio);
197 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
199 return submit(READ, page_off, virt_to_page(addr), bio_chain);
202 static int bio_write_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
204 return submit(WRITE, page_off, virt_to_page(addr), bio_chain);
207 static int wait_on_bio_chain(struct bio **bio_chain)
210 struct bio *next_bio;
213 if (bio_chain == NULL)
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))
239 static int mark_swapfiles(sector_t start, unsigned int flags)
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);
253 printk(KERN_ERR "PM: Swap header not found!\n");
260 * swsusp_swap_check - check if the resume device is a swap device
261 * and get its index (if so)
264 static int swsusp_swap_check(void) /* This is called before saving image */
268 res = swap_type_of(swsusp_resume_device, swsusp_resume_block,
274 res = blkdev_get(resume_bdev, FMODE_WRITE);
278 res = set_blocksize(resume_bdev, PAGE_SIZE);
280 blkdev_put(resume_bdev, FMODE_WRITE);
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
292 static int write_page(void *buf, sector_t offset, struct bio **bio_chain)
300 src = (void *)__get_free_page(__GFP_WAIT | __GFP_HIGH);
302 memcpy(src, buf, PAGE_SIZE);
305 bio_chain = NULL; /* Go synchronous */
311 return bio_write_page(offset, src, bio_chain);
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.
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.
325 * During resume we also only need to use one swap_map_page structure
329 #define MAP_PAGE_ENTRIES (PAGE_SIZE / sizeof(sector_t) - 1)
331 struct swap_map_page {
332 sector_t entries[MAP_PAGE_ENTRIES];
337 * The swap_map_handle structure is used for handling swap in
341 struct swap_map_handle {
342 struct swap_map_page *cur;
347 static void release_swap_writer(struct swap_map_handle *handle)
350 free_page((unsigned long)handle->cur);
354 static int get_swap_writer(struct swap_map_handle *handle)
356 handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
359 handle->cur_swap = alloc_swapdev_block(root_swap);
360 if (!handle->cur_swap) {
361 release_swap_writer(handle);
368 static int swap_write_page(struct swap_map_handle *handle, void *buf,
369 struct bio **bio_chain)
376 offset = alloc_swapdev_block(root_swap);
377 error = write_page(buf, offset, bio_chain);
380 handle->cur->entries[handle->k++] = offset;
381 if (handle->k >= MAP_PAGE_ENTRIES) {
382 error = wait_on_bio_chain(bio_chain);
385 offset = alloc_swapdev_block(root_swap);
388 handle->cur->next_swap = offset;
389 error = write_page(handle->cur, handle->cur_swap, NULL);
392 memset(handle->cur, 0, PAGE_SIZE);
393 handle->cur_swap = offset;
400 static int flush_swap_writer(struct swap_map_handle *handle)
402 if (handle->cur && handle->cur_swap)
403 return write_page(handle->cur, handle->cur_swap, NULL);
409 * save_image - save the suspend image data
412 static int save_image(struct swap_map_handle *handle,
413 struct snapshot_handle *snapshot,
414 unsigned int nr_to_write)
421 struct timeval start;
424 printk(KERN_INFO "PM: Saving image data pages (%u pages) ... ",
426 m = nr_to_write / 100;
431 do_gettimeofday(&start);
433 ret = snapshot_read_next(snapshot, PAGE_SIZE);
436 ret = swap_write_page(handle, data_of(*snapshot), &bio);
440 printk(KERN_CONT "\b\b\b\b%3d%%", nr_pages / m);
443 err2 = wait_on_bio_chain(&bio);
444 do_gettimeofday(&stop);
448 printk(KERN_CONT "\b\b\b\bdone\n");
450 printk(KERN_CONT "\n");
451 swsusp_show_speed(&start, &stop, nr_to_write, "Wrote");
456 * enough_swap - Make sure we have enough swap to save the image.
458 * Returns TRUE or FALSE after checking the total amount of swap
459 * space avaiable from the resume partition.
462 static int enough_swap(unsigned int nr_pages)
464 unsigned int free_swap = count_swap_pages(root_swap, 1);
466 pr_debug("PM: Free swap pages: %u\n", free_swap);
467 return free_swap > nr_pages + PAGES_FOR_IO;
471 * swsusp_write - Write entire image and metadata.
472 * @flags: flags to pass to the "boot" kernel in the image header
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.)
480 int swsusp_write(unsigned int flags)
482 struct swap_map_handle handle;
483 struct snapshot_handle snapshot;
484 struct swsusp_info *header;
487 error = swsusp_swap_check();
489 printk(KERN_ERR "PM: Cannot find swap device, try "
493 memset(&snapshot, 0, sizeof(struct snapshot_handle));
494 error = snapshot_read_next(&snapshot, PAGE_SIZE);
495 if (error < PAGE_SIZE) {
501 header = (struct swsusp_info *)data_of(snapshot);
502 if (!enough_swap(header->pages)) {
503 printk(KERN_ERR "PM: Not enough free swap\n");
507 error = get_swap_writer(&handle);
509 sector_t start = handle.cur_swap;
511 error = swap_write_page(&handle, header, NULL);
513 error = save_image(&handle, &snapshot,
517 flush_swap_writer(&handle);
518 printk(KERN_INFO "PM: S");
519 error = mark_swapfiles(start, flags);
524 free_all_swap_pages(root_swap);
526 release_swap_writer(&handle);
528 swsusp_close(FMODE_WRITE);
533 * The following functions allow us to read data using a swap map
534 * in a file-alike way
537 static void release_swap_reader(struct swap_map_handle *handle)
540 free_page((unsigned long)handle->cur);
544 static int get_swap_reader(struct swap_map_handle *handle, sector_t start)
551 handle->cur = (struct swap_map_page *)get_zeroed_page(__GFP_WAIT | __GFP_HIGH);
555 error = bio_read_page(start, handle->cur, NULL);
557 release_swap_reader(handle);
564 static int swap_read_page(struct swap_map_handle *handle, void *buf,
565 struct bio **bio_chain)
572 offset = handle->cur->entries[handle->k];
575 error = bio_read_page(offset, buf, bio_chain);
578 if (++handle->k >= MAP_PAGE_ENTRIES) {
579 error = wait_on_bio_chain(bio_chain);
581 offset = handle->cur->next_swap;
583 release_swap_reader(handle);
585 error = bio_read_page(offset, handle->cur, NULL);
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)
596 static int load_image(struct swap_map_handle *handle,
597 struct snapshot_handle *snapshot,
598 unsigned int nr_to_read)
602 struct timeval start;
608 printk(KERN_INFO "PM: Loading image data pages (%u pages) ... ",
610 m = nr_to_read / 100;
615 do_gettimeofday(&start);
617 error = snapshot_write_next(snapshot, PAGE_SIZE);
620 error = swap_read_page(handle, data_of(*snapshot), &bio);
623 if (snapshot->sync_read)
624 error = wait_on_bio_chain(&bio);
628 printk("\b\b\b\b%3d%%", nr_pages / m);
631 err2 = wait_on_bio_chain(&bio);
632 do_gettimeofday(&stop);
636 printk("\b\b\b\bdone\n");
637 snapshot_write_finalize(snapshot);
638 if (!snapshot_image_loaded(snapshot))
642 swsusp_show_speed(&start, &stop, nr_to_read, "Read");
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
652 int swsusp_read(unsigned int *flags_p)
655 struct swap_map_handle handle;
656 struct snapshot_handle snapshot;
657 struct swsusp_info *header;
659 *flags_p = swsusp_header->flags;
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);
668 error = swap_read_page(&handle, header, NULL);
670 error = load_image(&handle, &snapshot, header->pages - 1);
671 release_swap_reader(&handle);
674 pr_debug("PM: Image successfully loaded\n");
676 pr_debug("PM: Error %d resuming\n", error);
681 * swsusp_check - Check for swsusp signature in the resume device
684 int swsusp_check(void)
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);
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);
708 blkdev_put(resume_bdev, FMODE_READ);
710 pr_debug("PM: Signature found, resuming\n");
712 error = PTR_ERR(resume_bdev);
716 pr_debug("PM: Error %d checking image file\n", error);
722 * swsusp_close - close swap device.
725 void swsusp_close(fmode_t mode)
727 if (IS_ERR(resume_bdev)) {
728 pr_debug("PM: Image device not initialised\n");
732 blkdev_put(resume_bdev, mode);
735 static int swsusp_header_init(void)
737 swsusp_header = (struct swsusp_header*) __get_free_page(GFP_KERNEL);
739 panic("Could not allocate memory for swsusp_header\n");
743 core_initcall(swsusp_header_init);