diff options
-rw-r--r-- | include/linux/swap.h | 5 | ||||
-rw-r--r-- | kernel/power/disk.c | 12 | ||||
-rw-r--r-- | kernel/power/power.h | 26 | ||||
-rw-r--r-- | kernel/power/snapshot.c | 326 | ||||
-rw-r--r-- | kernel/power/swsusp.c | 723 | ||||
-rw-r--r-- | mm/swapfile.c | 55 |
6 files changed, 656 insertions, 491 deletions
diff --git a/include/linux/swap.h b/include/linux/swap.h index 12415dd94451..54eac8a39a4c 100644 --- a/include/linux/swap.h +++ b/include/linux/swap.h | |||
@@ -234,14 +234,15 @@ extern struct page * read_swap_cache_async(swp_entry_t, struct vm_area_struct *v | |||
234 | /* linux/mm/swapfile.c */ | 234 | /* linux/mm/swapfile.c */ |
235 | extern long total_swap_pages; | 235 | extern long total_swap_pages; |
236 | extern unsigned int nr_swapfiles; | 236 | extern unsigned int nr_swapfiles; |
237 | extern struct swap_info_struct swap_info[]; | ||
238 | extern void si_swapinfo(struct sysinfo *); | 237 | extern void si_swapinfo(struct sysinfo *); |
239 | extern swp_entry_t get_swap_page(void); | 238 | extern swp_entry_t get_swap_page(void); |
240 | extern swp_entry_t get_swap_page_of_type(int type); | 239 | extern swp_entry_t get_swap_page_of_type(int); |
241 | extern int swap_duplicate(swp_entry_t); | 240 | extern int swap_duplicate(swp_entry_t); |
242 | extern int valid_swaphandles(swp_entry_t, unsigned long *); | 241 | extern int valid_swaphandles(swp_entry_t, unsigned long *); |
243 | extern void swap_free(swp_entry_t); | 242 | extern void swap_free(swp_entry_t); |
244 | extern void free_swap_and_cache(swp_entry_t); | 243 | extern void free_swap_and_cache(swp_entry_t); |
244 | extern int swap_type_of(dev_t); | ||
245 | extern unsigned int count_swap_pages(int, int); | ||
245 | extern sector_t map_swap_page(struct swap_info_struct *, pgoff_t); | 246 | extern sector_t map_swap_page(struct swap_info_struct *, pgoff_t); |
246 | extern struct swap_info_struct *get_swap_info_struct(unsigned); | 247 | extern struct swap_info_struct *get_swap_info_struct(unsigned); |
247 | extern int can_share_swap_page(struct page *); | 248 | extern int can_share_swap_page(struct page *); |
diff --git a/kernel/power/disk.c b/kernel/power/disk.c index 0b43847dc980..4eb464b71347 100644 --- a/kernel/power/disk.c +++ b/kernel/power/disk.c | |||
@@ -26,9 +26,9 @@ extern suspend_disk_method_t pm_disk_mode; | |||
26 | 26 | ||
27 | extern int swsusp_shrink_memory(void); | 27 | extern int swsusp_shrink_memory(void); |
28 | extern int swsusp_suspend(void); | 28 | extern int swsusp_suspend(void); |
29 | extern int swsusp_write(struct pbe *pblist, unsigned int nr_pages); | 29 | extern int swsusp_write(void); |
30 | extern int swsusp_check(void); | 30 | extern int swsusp_check(void); |
31 | extern int swsusp_read(struct pbe **pblist_ptr); | 31 | extern int swsusp_read(void); |
32 | extern void swsusp_close(void); | 32 | extern void swsusp_close(void); |
33 | extern int swsusp_resume(void); | 33 | extern int swsusp_resume(void); |
34 | 34 | ||
@@ -70,10 +70,6 @@ static void power_down(suspend_disk_method_t mode) | |||
70 | while(1); | 70 | while(1); |
71 | } | 71 | } |
72 | 72 | ||
73 | |||
74 | static int in_suspend __nosavedata = 0; | ||
75 | |||
76 | |||
77 | static inline void platform_finish(void) | 73 | static inline void platform_finish(void) |
78 | { | 74 | { |
79 | if (pm_disk_mode == PM_DISK_PLATFORM) { | 75 | if (pm_disk_mode == PM_DISK_PLATFORM) { |
@@ -145,7 +141,7 @@ int pm_suspend_disk(void) | |||
145 | if (in_suspend) { | 141 | if (in_suspend) { |
146 | device_resume(); | 142 | device_resume(); |
147 | pr_debug("PM: writing image.\n"); | 143 | pr_debug("PM: writing image.\n"); |
148 | error = swsusp_write(pagedir_nosave, nr_copy_pages); | 144 | error = swsusp_write(); |
149 | if (!error) | 145 | if (!error) |
150 | power_down(pm_disk_mode); | 146 | power_down(pm_disk_mode); |
151 | else { | 147 | else { |
@@ -216,7 +212,7 @@ static int software_resume(void) | |||
216 | 212 | ||
217 | pr_debug("PM: Reading swsusp image.\n"); | 213 | pr_debug("PM: Reading swsusp image.\n"); |
218 | 214 | ||
219 | if ((error = swsusp_read(&pagedir_nosave))) { | 215 | if ((error = swsusp_read())) { |
220 | swsusp_free(); | 216 | swsusp_free(); |
221 | goto Thaw; | 217 | goto Thaw; |
222 | } | 218 | } |
diff --git a/kernel/power/power.h b/kernel/power/power.h index 388dba680841..ea7132ed029b 100644 --- a/kernel/power/power.h +++ b/kernel/power/power.h | |||
@@ -37,21 +37,31 @@ extern struct subsystem power_subsys; | |||
37 | /* References to section boundaries */ | 37 | /* References to section boundaries */ |
38 | extern const void __nosave_begin, __nosave_end; | 38 | extern const void __nosave_begin, __nosave_end; |
39 | 39 | ||
40 | extern unsigned int nr_copy_pages; | ||
41 | extern struct pbe *pagedir_nosave; | 40 | extern struct pbe *pagedir_nosave; |
42 | 41 | ||
43 | /* Preferred image size in bytes (default 500 MB) */ | 42 | /* Preferred image size in bytes (default 500 MB) */ |
44 | extern unsigned long image_size; | 43 | extern unsigned long image_size; |
45 | 44 | ||
45 | extern int in_suspend; | ||
46 | |||
46 | extern asmlinkage int swsusp_arch_suspend(void); | 47 | extern asmlinkage int swsusp_arch_suspend(void); |
47 | extern asmlinkage int swsusp_arch_resume(void); | 48 | extern asmlinkage int swsusp_arch_resume(void); |
48 | 49 | ||
49 | extern unsigned int count_data_pages(void); | 50 | extern unsigned int count_data_pages(void); |
50 | extern void free_pagedir(struct pbe *pblist); | ||
51 | extern void release_eaten_pages(void); | ||
52 | extern struct pbe *alloc_pagedir(unsigned nr_pages, gfp_t gfp_mask, int safe_needed); | ||
53 | extern void swsusp_free(void); | 51 | extern void swsusp_free(void); |
54 | extern int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed); | 52 | |
55 | extern unsigned int snapshot_nr_pages(void); | 53 | struct snapshot_handle { |
56 | extern struct pbe *snapshot_pblist(void); | 54 | loff_t offset; |
57 | extern void snapshot_pblist_set(struct pbe *pblist); | 55 | unsigned int page; |
56 | unsigned int page_offset; | ||
57 | unsigned int prev; | ||
58 | struct pbe *pbe; | ||
59 | void *buffer; | ||
60 | unsigned int buf_offset; | ||
61 | }; | ||
62 | |||
63 | #define data_of(handle) ((handle).buffer + (handle).buf_offset) | ||
64 | |||
65 | extern int snapshot_read_next(struct snapshot_handle *handle, size_t count); | ||
66 | extern int snapshot_write_next(struct snapshot_handle *handle, size_t count); | ||
67 | int snapshot_image_loaded(struct snapshot_handle *handle); | ||
diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c index 8d5a5986d621..cc349437fb72 100644 --- a/kernel/power/snapshot.c +++ b/kernel/power/snapshot.c | |||
@@ -10,6 +10,7 @@ | |||
10 | */ | 10 | */ |
11 | 11 | ||
12 | 12 | ||
13 | #include <linux/version.h> | ||
13 | #include <linux/module.h> | 14 | #include <linux/module.h> |
14 | #include <linux/mm.h> | 15 | #include <linux/mm.h> |
15 | #include <linux/suspend.h> | 16 | #include <linux/suspend.h> |
@@ -34,7 +35,8 @@ | |||
34 | #include "power.h" | 35 | #include "power.h" |
35 | 36 | ||
36 | struct pbe *pagedir_nosave; | 37 | struct pbe *pagedir_nosave; |
37 | unsigned int nr_copy_pages; | 38 | static unsigned int nr_copy_pages; |
39 | static unsigned int nr_meta_pages; | ||
38 | 40 | ||
39 | #ifdef CONFIG_HIGHMEM | 41 | #ifdef CONFIG_HIGHMEM |
40 | unsigned int count_highmem_pages(void) | 42 | unsigned int count_highmem_pages(void) |
@@ -235,7 +237,7 @@ static void copy_data_pages(struct pbe *pblist) | |||
235 | * free_pagedir - free pages allocated with alloc_pagedir() | 237 | * free_pagedir - free pages allocated with alloc_pagedir() |
236 | */ | 238 | */ |
237 | 239 | ||
238 | void free_pagedir(struct pbe *pblist) | 240 | static void free_pagedir(struct pbe *pblist) |
239 | { | 241 | { |
240 | struct pbe *pbe; | 242 | struct pbe *pbe; |
241 | 243 | ||
@@ -301,7 +303,7 @@ struct eaten_page { | |||
301 | 303 | ||
302 | static struct eaten_page *eaten_pages = NULL; | 304 | static struct eaten_page *eaten_pages = NULL; |
303 | 305 | ||
304 | void release_eaten_pages(void) | 306 | static void release_eaten_pages(void) |
305 | { | 307 | { |
306 | struct eaten_page *p, *q; | 308 | struct eaten_page *p, *q; |
307 | 309 | ||
@@ -376,7 +378,6 @@ struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, int safe_needed | |||
376 | if (!nr_pages) | 378 | if (!nr_pages) |
377 | return NULL; | 379 | return NULL; |
378 | 380 | ||
379 | pr_debug("alloc_pagedir(): nr_pages = %d\n", nr_pages); | ||
380 | pblist = alloc_image_page(gfp_mask, safe_needed); | 381 | pblist = alloc_image_page(gfp_mask, safe_needed); |
381 | /* FIXME: rewrite this ugly loop */ | 382 | /* FIXME: rewrite this ugly loop */ |
382 | for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages; | 383 | for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages; |
@@ -414,6 +415,9 @@ void swsusp_free(void) | |||
414 | } | 415 | } |
415 | } | 416 | } |
416 | } | 417 | } |
418 | nr_copy_pages = 0; | ||
419 | nr_meta_pages = 0; | ||
420 | pagedir_nosave = NULL; | ||
417 | } | 421 | } |
418 | 422 | ||
419 | 423 | ||
@@ -437,7 +441,7 @@ static int enough_free_mem(unsigned int nr_pages) | |||
437 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE); | 441 | (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE); |
438 | } | 442 | } |
439 | 443 | ||
440 | int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed) | 444 | static int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed) |
441 | { | 445 | { |
442 | struct pbe *p; | 446 | struct pbe *p; |
443 | 447 | ||
@@ -504,7 +508,319 @@ asmlinkage int swsusp_save(void) | |||
504 | */ | 508 | */ |
505 | 509 | ||
506 | nr_copy_pages = nr_pages; | 510 | nr_copy_pages = nr_pages; |
511 | nr_meta_pages = (nr_pages * sizeof(long) + PAGE_SIZE - 1) >> PAGE_SHIFT; | ||
507 | 512 | ||
508 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); | 513 | printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages); |
509 | return 0; | 514 | return 0; |
510 | } | 515 | } |
516 | |||
517 | static void init_header(struct swsusp_info *info) | ||
518 | { | ||
519 | memset(info, 0, sizeof(struct swsusp_info)); | ||
520 | info->version_code = LINUX_VERSION_CODE; | ||
521 | info->num_physpages = num_physpages; | ||
522 | memcpy(&info->uts, &system_utsname, sizeof(system_utsname)); | ||
523 | info->cpus = num_online_cpus(); | ||
524 | info->image_pages = nr_copy_pages; | ||
525 | info->pages = nr_copy_pages + nr_meta_pages + 1; | ||
526 | } | ||
527 | |||
528 | /** | ||
529 | * pack_orig_addresses - the .orig_address fields of the PBEs from the | ||
530 | * list starting at @pbe are stored in the array @buf[] (1 page) | ||
531 | */ | ||
532 | |||
533 | static inline struct pbe *pack_orig_addresses(unsigned long *buf, struct pbe *pbe) | ||
534 | { | ||
535 | int j; | ||
536 | |||
537 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | ||
538 | buf[j] = pbe->orig_address; | ||
539 | pbe = pbe->next; | ||
540 | } | ||
541 | if (!pbe) | ||
542 | for (; j < PAGE_SIZE / sizeof(long); j++) | ||
543 | buf[j] = 0; | ||
544 | return pbe; | ||
545 | } | ||
546 | |||
547 | /** | ||
548 | * snapshot_read_next - used for reading the system memory snapshot. | ||
549 | * | ||
550 | * On the first call to it @handle should point to a zeroed | ||
551 | * snapshot_handle structure. The structure gets updated and a pointer | ||
552 | * to it should be passed to this function every next time. | ||
553 | * | ||
554 | * The @count parameter should contain the number of bytes the caller | ||
555 | * wants to read from the snapshot. It must not be zero. | ||
556 | * | ||
557 | * On success the function returns a positive number. Then, the caller | ||
558 | * is allowed to read up to the returned number of bytes from the memory | ||
559 | * location computed by the data_of() macro. The number returned | ||
560 | * may be smaller than @count, but this only happens if the read would | ||
561 | * cross a page boundary otherwise. | ||
562 | * | ||
563 | * The function returns 0 to indicate the end of data stream condition, | ||
564 | * and a negative number is returned on error. In such cases the | ||
565 | * structure pointed to by @handle is not updated and should not be used | ||
566 | * any more. | ||
567 | */ | ||
568 | |||
569 | int snapshot_read_next(struct snapshot_handle *handle, size_t count) | ||
570 | { | ||
571 | static unsigned long *buffer; | ||
572 | |||
573 | if (handle->page > nr_meta_pages + nr_copy_pages) | ||
574 | return 0; | ||
575 | if (!buffer) { | ||
576 | /* This makes the buffer be freed by swsusp_free() */ | ||
577 | buffer = alloc_image_page(GFP_ATOMIC, 0); | ||
578 | if (!buffer) | ||
579 | return -ENOMEM; | ||
580 | } | ||
581 | if (!handle->offset) { | ||
582 | init_header((struct swsusp_info *)buffer); | ||
583 | handle->buffer = buffer; | ||
584 | handle->pbe = pagedir_nosave; | ||
585 | } | ||
586 | if (handle->prev < handle->page) { | ||
587 | if (handle->page <= nr_meta_pages) { | ||
588 | handle->pbe = pack_orig_addresses(buffer, handle->pbe); | ||
589 | if (!handle->pbe) | ||
590 | handle->pbe = pagedir_nosave; | ||
591 | } else { | ||
592 | handle->buffer = (void *)handle->pbe->address; | ||
593 | handle->pbe = handle->pbe->next; | ||
594 | } | ||
595 | handle->prev = handle->page; | ||
596 | } | ||
597 | handle->buf_offset = handle->page_offset; | ||
598 | if (handle->page_offset + count >= PAGE_SIZE) { | ||
599 | count = PAGE_SIZE - handle->page_offset; | ||
600 | handle->page_offset = 0; | ||
601 | handle->page++; | ||
602 | } else { | ||
603 | handle->page_offset += count; | ||
604 | } | ||
605 | handle->offset += count; | ||
606 | return count; | ||
607 | } | ||
608 | |||
609 | /** | ||
610 | * mark_unsafe_pages - mark the pages that cannot be used for storing | ||
611 | * the image during resume, because they conflict with the pages that | ||
612 | * had been used before suspend | ||
613 | */ | ||
614 | |||
615 | static int mark_unsafe_pages(struct pbe *pblist) | ||
616 | { | ||
617 | struct zone *zone; | ||
618 | unsigned long zone_pfn; | ||
619 | struct pbe *p; | ||
620 | |||
621 | if (!pblist) /* a sanity check */ | ||
622 | return -EINVAL; | ||
623 | |||
624 | /* Clear page flags */ | ||
625 | for_each_zone (zone) { | ||
626 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | ||
627 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) | ||
628 | ClearPageNosaveFree(pfn_to_page(zone_pfn + | ||
629 | zone->zone_start_pfn)); | ||
630 | } | ||
631 | |||
632 | /* Mark orig addresses */ | ||
633 | for_each_pbe (p, pblist) { | ||
634 | if (virt_addr_valid(p->orig_address)) | ||
635 | SetPageNosaveFree(virt_to_page(p->orig_address)); | ||
636 | else | ||
637 | return -EFAULT; | ||
638 | } | ||
639 | |||
640 | return 0; | ||
641 | } | ||
642 | |||
643 | static void copy_page_backup_list(struct pbe *dst, struct pbe *src) | ||
644 | { | ||
645 | /* We assume both lists contain the same number of elements */ | ||
646 | while (src) { | ||
647 | dst->orig_address = src->orig_address; | ||
648 | dst = dst->next; | ||
649 | src = src->next; | ||
650 | } | ||
651 | } | ||
652 | |||
653 | static int check_header(struct swsusp_info *info) | ||
654 | { | ||
655 | char *reason = NULL; | ||
656 | |||
657 | if (info->version_code != LINUX_VERSION_CODE) | ||
658 | reason = "kernel version"; | ||
659 | if (info->num_physpages != num_physpages) | ||
660 | reason = "memory size"; | ||
661 | if (strcmp(info->uts.sysname,system_utsname.sysname)) | ||
662 | reason = "system type"; | ||
663 | if (strcmp(info->uts.release,system_utsname.release)) | ||
664 | reason = "kernel release"; | ||
665 | if (strcmp(info->uts.version,system_utsname.version)) | ||
666 | reason = "version"; | ||
667 | if (strcmp(info->uts.machine,system_utsname.machine)) | ||
668 | reason = "machine"; | ||
669 | if (reason) { | ||
670 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | ||
671 | return -EPERM; | ||
672 | } | ||
673 | return 0; | ||
674 | } | ||
675 | |||
676 | /** | ||
677 | * load header - check the image header and copy data from it | ||
678 | */ | ||
679 | |||
680 | static int load_header(struct snapshot_handle *handle, | ||
681 | struct swsusp_info *info) | ||
682 | { | ||
683 | int error; | ||
684 | struct pbe *pblist; | ||
685 | |||
686 | error = check_header(info); | ||
687 | if (!error) { | ||
688 | pblist = alloc_pagedir(info->image_pages, GFP_ATOMIC, 0); | ||
689 | if (!pblist) | ||
690 | return -ENOMEM; | ||
691 | pagedir_nosave = pblist; | ||
692 | handle->pbe = pblist; | ||
693 | nr_copy_pages = info->image_pages; | ||
694 | nr_meta_pages = info->pages - info->image_pages - 1; | ||
695 | } | ||
696 | return error; | ||
697 | } | ||
698 | |||
699 | /** | ||
700 | * unpack_orig_addresses - copy the elements of @buf[] (1 page) to | ||
701 | * the PBEs in the list starting at @pbe | ||
702 | */ | ||
703 | |||
704 | static inline struct pbe *unpack_orig_addresses(unsigned long *buf, | ||
705 | struct pbe *pbe) | ||
706 | { | ||
707 | int j; | ||
708 | |||
709 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | ||
710 | pbe->orig_address = buf[j]; | ||
711 | pbe = pbe->next; | ||
712 | } | ||
713 | return pbe; | ||
714 | } | ||
715 | |||
716 | /** | ||
717 | * create_image - use metadata contained in the PBE list | ||
718 | * pointed to by pagedir_nosave to mark the pages that will | ||
719 | * be overwritten in the process of restoring the system | ||
720 | * memory state from the image and allocate memory for | ||
721 | * the image avoiding these pages | ||
722 | */ | ||
723 | |||
724 | static int create_image(struct snapshot_handle *handle) | ||
725 | { | ||
726 | int error = 0; | ||
727 | struct pbe *p, *pblist; | ||
728 | |||
729 | p = pagedir_nosave; | ||
730 | error = mark_unsafe_pages(p); | ||
731 | if (!error) { | ||
732 | pblist = alloc_pagedir(nr_copy_pages, GFP_ATOMIC, 1); | ||
733 | if (pblist) | ||
734 | copy_page_backup_list(pblist, p); | ||
735 | free_pagedir(p); | ||
736 | if (!pblist) | ||
737 | error = -ENOMEM; | ||
738 | } | ||
739 | if (!error) | ||
740 | error = alloc_data_pages(pblist, GFP_ATOMIC, 1); | ||
741 | if (!error) { | ||
742 | release_eaten_pages(); | ||
743 | pagedir_nosave = pblist; | ||
744 | } else { | ||
745 | pagedir_nosave = NULL; | ||
746 | handle->pbe = NULL; | ||
747 | nr_copy_pages = 0; | ||
748 | nr_meta_pages = 0; | ||
749 | } | ||
750 | return error; | ||
751 | } | ||
752 | |||
753 | /** | ||
754 | * snapshot_write_next - used for writing the system memory snapshot. | ||
755 | * | ||
756 | * On the first call to it @handle should point to a zeroed | ||
757 | * snapshot_handle structure. The structure gets updated and a pointer | ||
758 | * to it should be passed to this function every next time. | ||
759 | * | ||
760 | * The @count parameter should contain the number of bytes the caller | ||
761 | * wants to write to the image. It must not be zero. | ||
762 | * | ||
763 | * On success the function returns a positive number. Then, the caller | ||
764 | * is allowed to write up to the returned number of bytes to the memory | ||
765 | * location computed by the data_of() macro. The number returned | ||
766 | * may be smaller than @count, but this only happens if the write would | ||
767 | * cross a page boundary otherwise. | ||
768 | * | ||
769 | * The function returns 0 to indicate the "end of file" condition, | ||
770 | * and a negative number is returned on error. In such cases the | ||
771 | * structure pointed to by @handle is not updated and should not be used | ||
772 | * any more. | ||
773 | */ | ||
774 | |||
775 | int snapshot_write_next(struct snapshot_handle *handle, size_t count) | ||
776 | { | ||
777 | static unsigned long *buffer; | ||
778 | int error = 0; | ||
779 | |||
780 | if (handle->prev && handle->page > nr_meta_pages + nr_copy_pages) | ||
781 | return 0; | ||
782 | if (!buffer) { | ||
783 | /* This makes the buffer be freed by swsusp_free() */ | ||
784 | buffer = alloc_image_page(GFP_ATOMIC, 0); | ||
785 | if (!buffer) | ||
786 | return -ENOMEM; | ||
787 | } | ||
788 | if (!handle->offset) | ||
789 | handle->buffer = buffer; | ||
790 | if (handle->prev < handle->page) { | ||
791 | if (!handle->prev) { | ||
792 | error = load_header(handle, (struct swsusp_info *)buffer); | ||
793 | if (error) | ||
794 | return error; | ||
795 | } else if (handle->prev <= nr_meta_pages) { | ||
796 | handle->pbe = unpack_orig_addresses(buffer, handle->pbe); | ||
797 | if (!handle->pbe) { | ||
798 | error = create_image(handle); | ||
799 | if (error) | ||
800 | return error; | ||
801 | handle->pbe = pagedir_nosave; | ||
802 | handle->buffer = (void *)handle->pbe->address; | ||
803 | } | ||
804 | } else { | ||
805 | handle->pbe = handle->pbe->next; | ||
806 | handle->buffer = (void *)handle->pbe->address; | ||
807 | } | ||
808 | handle->prev = handle->page; | ||
809 | } | ||
810 | handle->buf_offset = handle->page_offset; | ||
811 | if (handle->page_offset + count >= PAGE_SIZE) { | ||
812 | count = PAGE_SIZE - handle->page_offset; | ||
813 | handle->page_offset = 0; | ||
814 | handle->page++; | ||
815 | } else { | ||
816 | handle->page_offset += count; | ||
817 | } | ||
818 | handle->offset += count; | ||
819 | return count; | ||
820 | } | ||
821 | |||
822 | int snapshot_image_loaded(struct snapshot_handle *handle) | ||
823 | { | ||
824 | return !(!handle->pbe || handle->pbe->next || !nr_copy_pages || | ||
825 | handle->page <= nr_meta_pages + nr_copy_pages); | ||
826 | } | ||
diff --git a/kernel/power/swsusp.c b/kernel/power/swsusp.c index 4e90905f0e87..457084f50010 100644 --- a/kernel/power/swsusp.c +++ b/kernel/power/swsusp.c | |||
@@ -77,6 +77,8 @@ | |||
77 | */ | 77 | */ |
78 | unsigned long image_size = 500 * 1024 * 1024; | 78 | unsigned long image_size = 500 * 1024 * 1024; |
79 | 79 | ||
80 | int in_suspend __nosavedata = 0; | ||
81 | |||
80 | #ifdef CONFIG_HIGHMEM | 82 | #ifdef CONFIG_HIGHMEM |
81 | unsigned int count_highmem_pages(void); | 83 | unsigned int count_highmem_pages(void); |
82 | int save_highmem(void); | 84 | int save_highmem(void); |
@@ -98,8 +100,6 @@ static struct swsusp_header { | |||
98 | char sig[10]; | 100 | char sig[10]; |
99 | } __attribute__((packed, aligned(PAGE_SIZE))) swsusp_header; | 101 | } __attribute__((packed, aligned(PAGE_SIZE))) swsusp_header; |
100 | 102 | ||
101 | static struct swsusp_info swsusp_info; | ||
102 | |||
103 | /* | 103 | /* |
104 | * Saving part... | 104 | * Saving part... |
105 | */ | 105 | */ |
@@ -129,255 +129,261 @@ static int mark_swapfiles(swp_entry_t start) | |||
129 | return error; | 129 | return error; |
130 | } | 130 | } |
131 | 131 | ||
132 | /* | 132 | /** |
133 | * Check whether the swap device is the specified resume | 133 | * swsusp_swap_check - check if the resume device is a swap device |
134 | * device, irrespective of whether they are specified by | 134 | * and get its index (if so) |
135 | * identical names. | ||
136 | * | ||
137 | * (Thus, device inode aliasing is allowed. You can say /dev/hda4 | ||
138 | * instead of /dev/ide/host0/bus0/target0/lun0/part4 [if using devfs] | ||
139 | * and they'll be considered the same device. This is *necessary* for | ||
140 | * devfs, since the resume code can only recognize the form /dev/hda4, | ||
141 | * but the suspend code would see the long name.) | ||
142 | */ | 135 | */ |
143 | static inline int is_resume_device(const struct swap_info_struct *swap_info) | ||
144 | { | ||
145 | struct file *file = swap_info->swap_file; | ||
146 | struct inode *inode = file->f_dentry->d_inode; | ||
147 | |||
148 | return S_ISBLK(inode->i_mode) && | ||
149 | swsusp_resume_device == MKDEV(imajor(inode), iminor(inode)); | ||
150 | } | ||
151 | 136 | ||
152 | static int swsusp_swap_check(void) /* This is called before saving image */ | 137 | static int swsusp_swap_check(void) /* This is called before saving image */ |
153 | { | 138 | { |
154 | int i; | 139 | int res = swap_type_of(swsusp_resume_device); |
155 | 140 | ||
156 | if (!swsusp_resume_device) | 141 | if (res >= 0) { |
157 | return -ENODEV; | 142 | root_swap = res; |
158 | spin_lock(&swap_lock); | 143 | return 0; |
159 | for (i = 0; i < MAX_SWAPFILES; i++) { | ||
160 | if (!(swap_info[i].flags & SWP_WRITEOK)) | ||
161 | continue; | ||
162 | if (is_resume_device(swap_info + i)) { | ||
163 | spin_unlock(&swap_lock); | ||
164 | root_swap = i; | ||
165 | return 0; | ||
166 | } | ||
167 | } | 144 | } |
168 | spin_unlock(&swap_lock); | 145 | return res; |
169 | return -ENODEV; | ||
170 | } | 146 | } |
171 | 147 | ||
172 | /** | 148 | /** |
173 | * write_page - Write one page to a fresh swap location. | 149 | * The bitmap is used for tracing allocated swap pages |
174 | * @addr: Address we're writing. | ||
175 | * @loc: Place to store the entry we used. | ||
176 | * | 150 | * |
177 | * Allocate a new swap entry and 'sync' it. Note we discard -EIO | 151 | * The entire bitmap consists of a number of bitmap_page |
178 | * errors. That is an artifact left over from swsusp. It did not | 152 | * structures linked with the help of the .next member. |
179 | * check the return of rw_swap_page_sync() at all, since most pages | 153 | * Thus each page can be allocated individually, so we only |
180 | * written back to swap would return -EIO. | 154 | * need to make 0-order memory allocations to create |
181 | * This is a partial improvement, since we will at least return other | 155 | * the bitmap. |
182 | * errors, though we need to eventually fix the damn code. | ||
183 | */ | 156 | */ |
184 | static int write_page(unsigned long addr, swp_entry_t *loc) | ||
185 | { | ||
186 | swp_entry_t entry; | ||
187 | int error = -ENOSPC; | ||
188 | 157 | ||
189 | entry = get_swap_page_of_type(root_swap); | 158 | #define BITMAP_PAGE_SIZE (PAGE_SIZE - sizeof(void *)) |
190 | if (swp_offset(entry)) { | 159 | #define BITMAP_PAGE_CHUNKS (BITMAP_PAGE_SIZE / sizeof(long)) |
191 | error = rw_swap_page_sync(WRITE, entry, virt_to_page(addr)); | 160 | #define BITS_PER_CHUNK (sizeof(long) * 8) |
192 | if (!error || error == -EIO) | 161 | #define BITMAP_PAGE_BITS (BITMAP_PAGE_CHUNKS * BITS_PER_CHUNK) |
193 | *loc = entry; | 162 | |
194 | } | 163 | struct bitmap_page { |
195 | return error; | 164 | unsigned long chunks[BITMAP_PAGE_CHUNKS]; |
196 | } | 165 | struct bitmap_page *next; |
166 | }; | ||
197 | 167 | ||
198 | /** | 168 | /** |
199 | * Swap map-handling functions | 169 | * The following functions are used for tracing the allocated |
170 | * swap pages, so that they can be freed in case of an error. | ||
200 | * | 171 | * |
201 | * The swap map is a data structure used for keeping track of each page | 172 | * The functions operate on a linked bitmap structure defined |
202 | * written to the swap. It consists of many swap_map_page structures | 173 | * above |
203 | * that contain each an array of MAP_PAGE_SIZE swap entries. | ||
204 | * These structures are linked together with the help of either the | ||
205 | * .next (in memory) or the .next_swap (in swap) member. | ||
206 | * | ||
207 | * The swap map is created during suspend. At that time we need to keep | ||
208 | * it in memory, because we have to free all of the allocated swap | ||
209 | * entries if an error occurs. The memory needed is preallocated | ||
210 | * so that we know in advance if there's enough of it. | ||
211 | * | ||
212 | * The first swap_map_page structure is filled with the swap entries that | ||
213 | * correspond to the first MAP_PAGE_SIZE data pages written to swap and | ||
214 | * so on. After the all of the data pages have been written, the order | ||
215 | * of the swap_map_page structures in the map is reversed so that they | ||
216 | * can be read from swap in the original order. This causes the data | ||
217 | * pages to be loaded in exactly the same order in which they have been | ||
218 | * saved. | ||
219 | * | ||
220 | * During resume we only need to use one swap_map_page structure | ||
221 | * at a time, which means that we only need to use two memory pages for | ||
222 | * reading the image - one for reading the swap_map_page structures | ||
223 | * and the second for reading the data pages from swap. | ||
224 | */ | 174 | */ |
225 | 175 | ||
226 | #define MAP_PAGE_SIZE ((PAGE_SIZE - sizeof(swp_entry_t) - sizeof(void *)) \ | 176 | static void free_bitmap(struct bitmap_page *bitmap) |
227 | / sizeof(swp_entry_t)) | ||
228 | |||
229 | struct swap_map_page { | ||
230 | swp_entry_t entries[MAP_PAGE_SIZE]; | ||
231 | swp_entry_t next_swap; | ||
232 | struct swap_map_page *next; | ||
233 | }; | ||
234 | |||
235 | static inline void free_swap_map(struct swap_map_page *swap_map) | ||
236 | { | 177 | { |
237 | struct swap_map_page *swp; | 178 | struct bitmap_page *bp; |
238 | 179 | ||
239 | while (swap_map) { | 180 | while (bitmap) { |
240 | swp = swap_map->next; | 181 | bp = bitmap->next; |
241 | free_page((unsigned long)swap_map); | 182 | free_page((unsigned long)bitmap); |
242 | swap_map = swp; | 183 | bitmap = bp; |
243 | } | 184 | } |
244 | } | 185 | } |
245 | 186 | ||
246 | static struct swap_map_page *alloc_swap_map(unsigned int nr_pages) | 187 | static struct bitmap_page *alloc_bitmap(unsigned int nr_bits) |
247 | { | 188 | { |
248 | struct swap_map_page *swap_map, *swp; | 189 | struct bitmap_page *bitmap, *bp; |
249 | unsigned n = 0; | 190 | unsigned int n; |
250 | 191 | ||
251 | if (!nr_pages) | 192 | if (!nr_bits) |
252 | return NULL; | 193 | return NULL; |
253 | 194 | ||
254 | pr_debug("alloc_swap_map(): nr_pages = %d\n", nr_pages); | 195 | bitmap = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL); |
255 | swap_map = (struct swap_map_page *)get_zeroed_page(GFP_ATOMIC); | 196 | bp = bitmap; |
256 | swp = swap_map; | 197 | for (n = BITMAP_PAGE_BITS; n < nr_bits; n += BITMAP_PAGE_BITS) { |
257 | for (n = MAP_PAGE_SIZE; n < nr_pages; n += MAP_PAGE_SIZE) { | 198 | bp->next = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL); |
258 | swp->next = (struct swap_map_page *)get_zeroed_page(GFP_ATOMIC); | 199 | bp = bp->next; |
259 | swp = swp->next; | 200 | if (!bp) { |
260 | if (!swp) { | 201 | free_bitmap(bitmap); |
261 | free_swap_map(swap_map); | ||
262 | return NULL; | 202 | return NULL; |
263 | } | 203 | } |
264 | } | 204 | } |
265 | return swap_map; | 205 | return bitmap; |
266 | } | 206 | } |
267 | 207 | ||
268 | /** | 208 | static int bitmap_set(struct bitmap_page *bitmap, unsigned long bit) |
269 | * reverse_swap_map - reverse the order of pages in the swap map | ||
270 | * @swap_map | ||
271 | */ | ||
272 | |||
273 | static inline struct swap_map_page *reverse_swap_map(struct swap_map_page *swap_map) | ||
274 | { | 209 | { |
275 | struct swap_map_page *prev, *next; | 210 | unsigned int n; |
276 | 211 | ||
277 | prev = NULL; | 212 | n = BITMAP_PAGE_BITS; |
278 | while (swap_map) { | 213 | while (bitmap && n <= bit) { |
279 | next = swap_map->next; | 214 | n += BITMAP_PAGE_BITS; |
280 | swap_map->next = prev; | 215 | bitmap = bitmap->next; |
281 | prev = swap_map; | ||
282 | swap_map = next; | ||
283 | } | 216 | } |
284 | return prev; | 217 | if (!bitmap) |
218 | return -EINVAL; | ||
219 | n -= BITMAP_PAGE_BITS; | ||
220 | bit -= n; | ||
221 | n = 0; | ||
222 | while (bit >= BITS_PER_CHUNK) { | ||
223 | bit -= BITS_PER_CHUNK; | ||
224 | n++; | ||
225 | } | ||
226 | bitmap->chunks[n] |= (1UL << bit); | ||
227 | return 0; | ||
285 | } | 228 | } |
286 | 229 | ||
287 | /** | 230 | static unsigned long alloc_swap_page(int swap, struct bitmap_page *bitmap) |
288 | * free_swap_map_entries - free the swap entries allocated to store | ||
289 | * the swap map @swap_map (this is only called in case of an error) | ||
290 | */ | ||
291 | static inline void free_swap_map_entries(struct swap_map_page *swap_map) | ||
292 | { | 231 | { |
293 | while (swap_map) { | 232 | unsigned long offset; |
294 | if (swap_map->next_swap.val) | 233 | |
295 | swap_free(swap_map->next_swap); | 234 | offset = swp_offset(get_swap_page_of_type(swap)); |
296 | swap_map = swap_map->next; | 235 | if (offset) { |
236 | if (bitmap_set(bitmap, offset)) { | ||
237 | swap_free(swp_entry(swap, offset)); | ||
238 | offset = 0; | ||
239 | } | ||
297 | } | 240 | } |
241 | return offset; | ||
298 | } | 242 | } |
299 | 243 | ||
300 | /** | 244 | static void free_all_swap_pages(int swap, struct bitmap_page *bitmap) |
301 | * save_swap_map - save the swap map used for tracing the data pages | ||
302 | * stored in the swap | ||
303 | */ | ||
304 | |||
305 | static int save_swap_map(struct swap_map_page *swap_map, swp_entry_t *start) | ||
306 | { | 245 | { |
307 | swp_entry_t entry = (swp_entry_t){0}; | 246 | unsigned int bit, n; |
308 | int error; | 247 | unsigned long test; |
309 | 248 | ||
310 | while (swap_map) { | 249 | bit = 0; |
311 | swap_map->next_swap = entry; | 250 | while (bitmap) { |
312 | if ((error = write_page((unsigned long)swap_map, &entry))) | 251 | for (n = 0; n < BITMAP_PAGE_CHUNKS; n++) |
313 | return error; | 252 | for (test = 1UL; test; test <<= 1) { |
314 | swap_map = swap_map->next; | 253 | if (bitmap->chunks[n] & test) |
254 | swap_free(swp_entry(swap, bit)); | ||
255 | bit++; | ||
256 | } | ||
257 | bitmap = bitmap->next; | ||
315 | } | 258 | } |
316 | *start = entry; | ||
317 | return 0; | ||
318 | } | 259 | } |
319 | 260 | ||
320 | /** | 261 | /** |
321 | * free_image_entries - free the swap entries allocated to store | 262 | * write_page - Write one page to given swap location. |
322 | * the image data pages (this is only called in case of an error) | 263 | * @buf: Address we're writing. |
264 | * @offset: Offset of the swap page we're writing to. | ||
323 | */ | 265 | */ |
324 | 266 | ||
325 | static inline void free_image_entries(struct swap_map_page *swp) | 267 | static int write_page(void *buf, unsigned long offset) |
326 | { | 268 | { |
327 | unsigned k; | 269 | swp_entry_t entry; |
270 | int error = -ENOSPC; | ||
328 | 271 | ||
329 | while (swp) { | 272 | if (offset) { |
330 | for (k = 0; k < MAP_PAGE_SIZE; k++) | 273 | entry = swp_entry(root_swap, offset); |
331 | if (swp->entries[k].val) | 274 | error = rw_swap_page_sync(WRITE, entry, virt_to_page(buf)); |
332 | swap_free(swp->entries[k]); | ||
333 | swp = swp->next; | ||
334 | } | 275 | } |
276 | return error; | ||
335 | } | 277 | } |
336 | 278 | ||
279 | /* | ||
280 | * The swap map is a data structure used for keeping track of each page | ||
281 | * written to a swap partition. It consists of many swap_map_page | ||
282 | * structures that contain each an array of MAP_PAGE_SIZE swap entries. | ||
283 | * These structures are stored on the swap and linked together with the | ||
284 | * help of the .next_swap member. | ||
285 | * | ||
286 | * The swap map is created during suspend. The swap map pages are | ||
287 | * allocated and populated one at a time, so we only need one memory | ||
288 | * page to set up the entire structure. | ||
289 | * | ||
290 | * During resume we also only need to use one swap_map_page structure | ||
291 | * at a time. | ||
292 | */ | ||
293 | |||
294 | #define MAP_PAGE_ENTRIES (PAGE_SIZE / sizeof(long) - 1) | ||
295 | |||
296 | struct swap_map_page { | ||
297 | unsigned long entries[MAP_PAGE_ENTRIES]; | ||
298 | unsigned long next_swap; | ||
299 | }; | ||
300 | |||
337 | /** | 301 | /** |
338 | * The swap_map_handle structure is used for handling the swap map in | 302 | * The swap_map_handle structure is used for handling swap in |
339 | * a file-alike way | 303 | * a file-alike way |
340 | */ | 304 | */ |
341 | 305 | ||
342 | struct swap_map_handle { | 306 | struct swap_map_handle { |
343 | struct swap_map_page *cur; | 307 | struct swap_map_page *cur; |
308 | unsigned long cur_swap; | ||
309 | struct bitmap_page *bitmap; | ||
344 | unsigned int k; | 310 | unsigned int k; |
345 | }; | 311 | }; |
346 | 312 | ||
347 | static inline void init_swap_map_handle(struct swap_map_handle *handle, | 313 | static void release_swap_writer(struct swap_map_handle *handle) |
348 | struct swap_map_page *map) | ||
349 | { | 314 | { |
350 | handle->cur = map; | 315 | if (handle->cur) |
316 | free_page((unsigned long)handle->cur); | ||
317 | handle->cur = NULL; | ||
318 | if (handle->bitmap) | ||
319 | free_bitmap(handle->bitmap); | ||
320 | handle->bitmap = NULL; | ||
321 | } | ||
322 | |||
323 | static int get_swap_writer(struct swap_map_handle *handle) | ||
324 | { | ||
325 | handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL); | ||
326 | if (!handle->cur) | ||
327 | return -ENOMEM; | ||
328 | handle->bitmap = alloc_bitmap(count_swap_pages(root_swap, 0)); | ||
329 | if (!handle->bitmap) { | ||
330 | release_swap_writer(handle); | ||
331 | return -ENOMEM; | ||
332 | } | ||
333 | handle->cur_swap = alloc_swap_page(root_swap, handle->bitmap); | ||
334 | if (!handle->cur_swap) { | ||
335 | release_swap_writer(handle); | ||
336 | return -ENOSPC; | ||
337 | } | ||
351 | handle->k = 0; | 338 | handle->k = 0; |
339 | return 0; | ||
352 | } | 340 | } |
353 | 341 | ||
354 | static inline int swap_map_write_page(struct swap_map_handle *handle, | 342 | static int swap_write_page(struct swap_map_handle *handle, void *buf) |
355 | unsigned long addr) | ||
356 | { | 343 | { |
357 | int error; | 344 | int error; |
345 | unsigned long offset; | ||
358 | 346 | ||
359 | error = write_page(addr, handle->cur->entries + handle->k); | 347 | if (!handle->cur) |
348 | return -EINVAL; | ||
349 | offset = alloc_swap_page(root_swap, handle->bitmap); | ||
350 | error = write_page(buf, offset); | ||
360 | if (error) | 351 | if (error) |
361 | return error; | 352 | return error; |
362 | if (++handle->k >= MAP_PAGE_SIZE) { | 353 | handle->cur->entries[handle->k++] = offset; |
363 | handle->cur = handle->cur->next; | 354 | if (handle->k >= MAP_PAGE_ENTRIES) { |
355 | offset = alloc_swap_page(root_swap, handle->bitmap); | ||
356 | if (!offset) | ||
357 | return -ENOSPC; | ||
358 | handle->cur->next_swap = offset; | ||
359 | error = write_page(handle->cur, handle->cur_swap); | ||
360 | if (error) | ||
361 | return error; | ||
362 | memset(handle->cur, 0, PAGE_SIZE); | ||
363 | handle->cur_swap = offset; | ||
364 | handle->k = 0; | 364 | handle->k = 0; |
365 | } | 365 | } |
366 | return 0; | 366 | return 0; |
367 | } | 367 | } |
368 | 368 | ||
369 | static int flush_swap_writer(struct swap_map_handle *handle) | ||
370 | { | ||
371 | if (handle->cur && handle->cur_swap) | ||
372 | return write_page(handle->cur, handle->cur_swap); | ||
373 | else | ||
374 | return -EINVAL; | ||
375 | } | ||
376 | |||
369 | /** | 377 | /** |
370 | * save_image_data - save the data pages pointed to by the PBEs | 378 | * save_image - save the suspend image data |
371 | * from the list @pblist using the swap map handle @handle | ||
372 | * (assume there are @nr_pages data pages to save) | ||
373 | */ | 379 | */ |
374 | 380 | ||
375 | static int save_image_data(struct pbe *pblist, | 381 | static int save_image(struct swap_map_handle *handle, |
376 | struct swap_map_handle *handle, | 382 | struct snapshot_handle *snapshot, |
377 | unsigned int nr_pages) | 383 | unsigned int nr_pages) |
378 | { | 384 | { |
379 | unsigned int m; | 385 | unsigned int m; |
380 | struct pbe *p; | 386 | int ret; |
381 | int error = 0; | 387 | int error = 0; |
382 | 388 | ||
383 | printk("Saving image data pages (%u pages) ... ", nr_pages); | 389 | printk("Saving image data pages (%u pages) ... ", nr_pages); |
@@ -385,98 +391,22 @@ static int save_image_data(struct pbe *pblist, | |||
385 | if (!m) | 391 | if (!m) |
386 | m = 1; | 392 | m = 1; |
387 | nr_pages = 0; | 393 | nr_pages = 0; |
388 | for_each_pbe (p, pblist) { | 394 | do { |
389 | error = swap_map_write_page(handle, p->address); | 395 | ret = snapshot_read_next(snapshot, PAGE_SIZE); |
390 | if (error) | 396 | if (ret > 0) { |
391 | break; | 397 | error = swap_write_page(handle, data_of(*snapshot)); |
392 | if (!(nr_pages % m)) | 398 | if (error) |
393 | printk("\b\b\b\b%3d%%", nr_pages / m); | 399 | break; |
394 | nr_pages++; | 400 | if (!(nr_pages % m)) |
395 | } | 401 | printk("\b\b\b\b%3d%%", nr_pages / m); |
402 | nr_pages++; | ||
403 | } | ||
404 | } while (ret > 0); | ||
396 | if (!error) | 405 | if (!error) |
397 | printk("\b\b\b\bdone\n"); | 406 | printk("\b\b\b\bdone\n"); |
398 | return error; | 407 | return error; |
399 | } | 408 | } |
400 | 409 | ||
401 | static void dump_info(void) | ||
402 | { | ||
403 | pr_debug(" swsusp: Version: %u\n",swsusp_info.version_code); | ||
404 | pr_debug(" swsusp: Num Pages: %ld\n",swsusp_info.num_physpages); | ||
405 | pr_debug(" swsusp: UTS Sys: %s\n",swsusp_info.uts.sysname); | ||
406 | pr_debug(" swsusp: UTS Node: %s\n",swsusp_info.uts.nodename); | ||
407 | pr_debug(" swsusp: UTS Release: %s\n",swsusp_info.uts.release); | ||
408 | pr_debug(" swsusp: UTS Version: %s\n",swsusp_info.uts.version); | ||
409 | pr_debug(" swsusp: UTS Machine: %s\n",swsusp_info.uts.machine); | ||
410 | pr_debug(" swsusp: UTS Domain: %s\n",swsusp_info.uts.domainname); | ||
411 | pr_debug(" swsusp: CPUs: %d\n",swsusp_info.cpus); | ||
412 | pr_debug(" swsusp: Image: %ld Pages\n",swsusp_info.image_pages); | ||
413 | pr_debug(" swsusp: Total: %ld Pages\n", swsusp_info.pages); | ||
414 | } | ||
415 | |||
416 | static void init_header(unsigned int nr_pages) | ||
417 | { | ||
418 | memset(&swsusp_info, 0, sizeof(swsusp_info)); | ||
419 | swsusp_info.version_code = LINUX_VERSION_CODE; | ||
420 | swsusp_info.num_physpages = num_physpages; | ||
421 | memcpy(&swsusp_info.uts, &system_utsname, sizeof(system_utsname)); | ||
422 | |||
423 | swsusp_info.cpus = num_online_cpus(); | ||
424 | swsusp_info.image_pages = nr_pages; | ||
425 | swsusp_info.pages = nr_pages + | ||
426 | ((nr_pages * sizeof(long) + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1; | ||
427 | } | ||
428 | |||
429 | /** | ||
430 | * pack_orig_addresses - the .orig_address fields of the PBEs from the | ||
431 | * list starting at @pbe are stored in the array @buf[] (1 page) | ||
432 | */ | ||
433 | |||
434 | static inline struct pbe *pack_orig_addresses(unsigned long *buf, | ||
435 | struct pbe *pbe) | ||
436 | { | ||
437 | int j; | ||
438 | |||
439 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | ||
440 | buf[j] = pbe->orig_address; | ||
441 | pbe = pbe->next; | ||
442 | } | ||
443 | if (!pbe) | ||
444 | for (; j < PAGE_SIZE / sizeof(long); j++) | ||
445 | buf[j] = 0; | ||
446 | return pbe; | ||
447 | } | ||
448 | |||
449 | /** | ||
450 | * save_image_metadata - save the .orig_address fields of the PBEs | ||
451 | * from the list @pblist using the swap map handle @handle | ||
452 | */ | ||
453 | |||
454 | static int save_image_metadata(struct pbe *pblist, | ||
455 | struct swap_map_handle *handle) | ||
456 | { | ||
457 | unsigned long *buf; | ||
458 | unsigned int n = 0; | ||
459 | struct pbe *p; | ||
460 | int error = 0; | ||
461 | |||
462 | printk("Saving image metadata ... "); | ||
463 | buf = (unsigned long *)get_zeroed_page(GFP_ATOMIC); | ||
464 | if (!buf) | ||
465 | return -ENOMEM; | ||
466 | p = pblist; | ||
467 | while (p) { | ||
468 | p = pack_orig_addresses(buf, p); | ||
469 | error = swap_map_write_page(handle, (unsigned long)buf); | ||
470 | if (error) | ||
471 | break; | ||
472 | n++; | ||
473 | } | ||
474 | free_page((unsigned long)buf); | ||
475 | if (!error) | ||
476 | printk("done (%u pages saved)\n", n); | ||
477 | return error; | ||
478 | } | ||
479 | |||
480 | /** | 410 | /** |
481 | * enough_swap - Make sure we have enough swap to save the image. | 411 | * enough_swap - Make sure we have enough swap to save the image. |
482 | * | 412 | * |
@@ -486,8 +416,7 @@ static int save_image_metadata(struct pbe *pblist, | |||
486 | 416 | ||
487 | static int enough_swap(unsigned int nr_pages) | 417 | static int enough_swap(unsigned int nr_pages) |
488 | { | 418 | { |
489 | unsigned int free_swap = swap_info[root_swap].pages - | 419 | unsigned int free_swap = count_swap_pages(root_swap, 1); |
490 | swap_info[root_swap].inuse_pages; | ||
491 | 420 | ||
492 | pr_debug("swsusp: free swap pages: %u\n", free_swap); | 421 | pr_debug("swsusp: free swap pages: %u\n", free_swap); |
493 | return free_swap > (nr_pages + PAGES_FOR_IO + | 422 | return free_swap > (nr_pages + PAGES_FOR_IO + |
@@ -503,57 +432,44 @@ static int enough_swap(unsigned int nr_pages) | |||
503 | * correctly, we'll mark system clean, anyway.) | 432 | * correctly, we'll mark system clean, anyway.) |
504 | */ | 433 | */ |
505 | 434 | ||
506 | int swsusp_write(struct pbe *pblist, unsigned int nr_pages) | 435 | int swsusp_write(void) |
507 | { | 436 | { |
508 | struct swap_map_page *swap_map; | ||
509 | struct swap_map_handle handle; | 437 | struct swap_map_handle handle; |
510 | swp_entry_t start; | 438 | struct snapshot_handle snapshot; |
439 | struct swsusp_info *header; | ||
440 | unsigned long start; | ||
511 | int error; | 441 | int error; |
512 | 442 | ||
513 | if ((error = swsusp_swap_check())) { | 443 | if ((error = swsusp_swap_check())) { |
514 | printk(KERN_ERR "swsusp: Cannot find swap device, try swapon -a.\n"); | 444 | printk(KERN_ERR "swsusp: Cannot find swap device, try swapon -a.\n"); |
515 | return error; | 445 | return error; |
516 | } | 446 | } |
517 | if (!enough_swap(nr_pages)) { | 447 | memset(&snapshot, 0, sizeof(struct snapshot_handle)); |
448 | error = snapshot_read_next(&snapshot, PAGE_SIZE); | ||
449 | if (error < PAGE_SIZE) | ||
450 | return error < 0 ? error : -EFAULT; | ||
451 | header = (struct swsusp_info *)data_of(snapshot); | ||
452 | if (!enough_swap(header->pages)) { | ||
518 | printk(KERN_ERR "swsusp: Not enough free swap\n"); | 453 | printk(KERN_ERR "swsusp: Not enough free swap\n"); |
519 | return -ENOSPC; | 454 | return -ENOSPC; |
520 | } | 455 | } |
521 | 456 | error = get_swap_writer(&handle); | |
522 | init_header(nr_pages); | 457 | if (!error) { |
523 | swap_map = alloc_swap_map(swsusp_info.pages); | 458 | start = handle.cur_swap; |
524 | if (!swap_map) | 459 | error = swap_write_page(&handle, header); |
525 | return -ENOMEM; | 460 | } |
526 | init_swap_map_handle(&handle, swap_map); | ||
527 | |||
528 | error = swap_map_write_page(&handle, (unsigned long)&swsusp_info); | ||
529 | if (!error) | ||
530 | error = save_image_metadata(pblist, &handle); | ||
531 | if (!error) | 461 | if (!error) |
532 | error = save_image_data(pblist, &handle, nr_pages); | 462 | error = save_image(&handle, &snapshot, header->pages - 1); |
533 | if (error) | 463 | if (!error) { |
534 | goto Free_image_entries; | 464 | flush_swap_writer(&handle); |
535 | 465 | printk("S"); | |
536 | swap_map = reverse_swap_map(swap_map); | 466 | error = mark_swapfiles(swp_entry(root_swap, start)); |
537 | error = save_swap_map(swap_map, &start); | 467 | printk("|\n"); |
538 | if (error) | 468 | } |
539 | goto Free_map_entries; | ||
540 | |||
541 | dump_info(); | ||
542 | printk( "S" ); | ||
543 | error = mark_swapfiles(start); | ||
544 | printk( "|\n" ); | ||
545 | if (error) | 469 | if (error) |
546 | goto Free_map_entries; | 470 | free_all_swap_pages(root_swap, handle.bitmap); |
547 | 471 | release_swap_writer(&handle); | |
548 | Free_swap_map: | ||
549 | free_swap_map(swap_map); | ||
550 | return error; | 472 | return error; |
551 | |||
552 | Free_map_entries: | ||
553 | free_swap_map_entries(swap_map); | ||
554 | Free_image_entries: | ||
555 | free_image_entries(swap_map); | ||
556 | goto Free_swap_map; | ||
557 | } | 473 | } |
558 | 474 | ||
559 | /** | 475 | /** |
@@ -663,45 +579,6 @@ int swsusp_resume(void) | |||
663 | return error; | 579 | return error; |
664 | } | 580 | } |
665 | 581 | ||
666 | /** | ||
667 | * mark_unsafe_pages - mark the pages that cannot be used for storing | ||
668 | * the image during resume, because they conflict with the pages that | ||
669 | * had been used before suspend | ||
670 | */ | ||
671 | |||
672 | static void mark_unsafe_pages(struct pbe *pblist) | ||
673 | { | ||
674 | struct zone *zone; | ||
675 | unsigned long zone_pfn; | ||
676 | struct pbe *p; | ||
677 | |||
678 | if (!pblist) /* a sanity check */ | ||
679 | return; | ||
680 | |||
681 | /* Clear page flags */ | ||
682 | for_each_zone (zone) { | ||
683 | for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) | ||
684 | if (pfn_valid(zone_pfn + zone->zone_start_pfn)) | ||
685 | ClearPageNosaveFree(pfn_to_page(zone_pfn + | ||
686 | zone->zone_start_pfn)); | ||
687 | } | ||
688 | |||
689 | /* Mark orig addresses */ | ||
690 | for_each_pbe (p, pblist) | ||
691 | SetPageNosaveFree(virt_to_page(p->orig_address)); | ||
692 | |||
693 | } | ||
694 | |||
695 | static void copy_page_backup_list(struct pbe *dst, struct pbe *src) | ||
696 | { | ||
697 | /* We assume both lists contain the same number of elements */ | ||
698 | while (src) { | ||
699 | dst->orig_address = src->orig_address; | ||
700 | dst = dst->next; | ||
701 | src = src->next; | ||
702 | } | ||
703 | } | ||
704 | |||
705 | /* | 582 | /* |
706 | * Using bio to read from swap. | 583 | * Using bio to read from swap. |
707 | * This code requires a bit more work than just using buffer heads | 584 | * This code requires a bit more work than just using buffer heads |
@@ -779,14 +656,14 @@ static int bio_write_page(pgoff_t page_off, void *page) | |||
779 | * in a file-alike way | 656 | * in a file-alike way |
780 | */ | 657 | */ |
781 | 658 | ||
782 | static inline void release_swap_map_reader(struct swap_map_handle *handle) | 659 | static void release_swap_reader(struct swap_map_handle *handle) |
783 | { | 660 | { |
784 | if (handle->cur) | 661 | if (handle->cur) |
785 | free_page((unsigned long)handle->cur); | 662 | free_page((unsigned long)handle->cur); |
786 | handle->cur = NULL; | 663 | handle->cur = NULL; |
787 | } | 664 | } |
788 | 665 | ||
789 | static inline int get_swap_map_reader(struct swap_map_handle *handle, | 666 | static int get_swap_reader(struct swap_map_handle *handle, |
790 | swp_entry_t start) | 667 | swp_entry_t start) |
791 | { | 668 | { |
792 | int error; | 669 | int error; |
@@ -798,149 +675,80 @@ static inline int get_swap_map_reader(struct swap_map_handle *handle, | |||
798 | return -ENOMEM; | 675 | return -ENOMEM; |
799 | error = bio_read_page(swp_offset(start), handle->cur); | 676 | error = bio_read_page(swp_offset(start), handle->cur); |
800 | if (error) { | 677 | if (error) { |
801 | release_swap_map_reader(handle); | 678 | release_swap_reader(handle); |
802 | return error; | 679 | return error; |
803 | } | 680 | } |
804 | handle->k = 0; | 681 | handle->k = 0; |
805 | return 0; | 682 | return 0; |
806 | } | 683 | } |
807 | 684 | ||
808 | static inline int swap_map_read_page(struct swap_map_handle *handle, void *buf) | 685 | static int swap_read_page(struct swap_map_handle *handle, void *buf) |
809 | { | 686 | { |
810 | unsigned long offset; | 687 | unsigned long offset; |
811 | int error; | 688 | int error; |
812 | 689 | ||
813 | if (!handle->cur) | 690 | if (!handle->cur) |
814 | return -EINVAL; | 691 | return -EINVAL; |
815 | offset = swp_offset(handle->cur->entries[handle->k]); | 692 | offset = handle->cur->entries[handle->k]; |
816 | if (!offset) | 693 | if (!offset) |
817 | return -EINVAL; | 694 | return -EFAULT; |
818 | error = bio_read_page(offset, buf); | 695 | error = bio_read_page(offset, buf); |
819 | if (error) | 696 | if (error) |
820 | return error; | 697 | return error; |
821 | if (++handle->k >= MAP_PAGE_SIZE) { | 698 | if (++handle->k >= MAP_PAGE_ENTRIES) { |
822 | handle->k = 0; | 699 | handle->k = 0; |
823 | offset = swp_offset(handle->cur->next_swap); | 700 | offset = handle->cur->next_swap; |
824 | if (!offset) | 701 | if (!offset) |
825 | release_swap_map_reader(handle); | 702 | release_swap_reader(handle); |
826 | else | 703 | else |
827 | error = bio_read_page(offset, handle->cur); | 704 | error = bio_read_page(offset, handle->cur); |
828 | } | 705 | } |
829 | return error; | 706 | return error; |
830 | } | 707 | } |
831 | 708 | ||
832 | static int check_header(void) | ||
833 | { | ||
834 | char *reason = NULL; | ||
835 | |||
836 | dump_info(); | ||
837 | if (swsusp_info.version_code != LINUX_VERSION_CODE) | ||
838 | reason = "kernel version"; | ||
839 | if (swsusp_info.num_physpages != num_physpages) | ||
840 | reason = "memory size"; | ||
841 | if (strcmp(swsusp_info.uts.sysname,system_utsname.sysname)) | ||
842 | reason = "system type"; | ||
843 | if (strcmp(swsusp_info.uts.release,system_utsname.release)) | ||
844 | reason = "kernel release"; | ||
845 | if (strcmp(swsusp_info.uts.version,system_utsname.version)) | ||
846 | reason = "version"; | ||
847 | if (strcmp(swsusp_info.uts.machine,system_utsname.machine)) | ||
848 | reason = "machine"; | ||
849 | if (reason) { | ||
850 | printk(KERN_ERR "swsusp: Resume mismatch: %s\n", reason); | ||
851 | return -EPERM; | ||
852 | } | ||
853 | return 0; | ||
854 | } | ||
855 | |||
856 | /** | 709 | /** |
857 | * load_image_data - load the image data using the swap map handle | 710 | * load_image - load the image using the swap map handle |
858 | * @handle and store them using the page backup list @pblist | 711 | * @handle and the snapshot handle @snapshot |
859 | * (assume there are @nr_pages pages to load) | 712 | * (assume there are @nr_pages pages to load) |
860 | */ | 713 | */ |
861 | 714 | ||
862 | static int load_image_data(struct pbe *pblist, | 715 | static int load_image(struct swap_map_handle *handle, |
863 | struct swap_map_handle *handle, | 716 | struct snapshot_handle *snapshot, |
864 | unsigned int nr_pages) | 717 | unsigned int nr_pages) |
865 | { | 718 | { |
866 | int error; | ||
867 | unsigned int m; | 719 | unsigned int m; |
868 | struct pbe *p; | 720 | int ret; |
721 | int error = 0; | ||
869 | 722 | ||
870 | if (!pblist) | ||
871 | return -EINVAL; | ||
872 | printk("Loading image data pages (%u pages) ... ", nr_pages); | 723 | printk("Loading image data pages (%u pages) ... ", nr_pages); |
873 | m = nr_pages / 100; | 724 | m = nr_pages / 100; |
874 | if (!m) | 725 | if (!m) |
875 | m = 1; | 726 | m = 1; |
876 | nr_pages = 0; | 727 | nr_pages = 0; |
877 | p = pblist; | 728 | do { |
878 | while (p) { | 729 | ret = snapshot_write_next(snapshot, PAGE_SIZE); |
879 | error = swap_map_read_page(handle, (void *)p->address); | 730 | if (ret > 0) { |
880 | if (error) | 731 | error = swap_read_page(handle, data_of(*snapshot)); |
881 | break; | 732 | if (error) |
882 | p = p->next; | 733 | break; |
883 | if (!(nr_pages % m)) | 734 | if (!(nr_pages % m)) |
884 | printk("\b\b\b\b%3d%%", nr_pages / m); | 735 | printk("\b\b\b\b%3d%%", nr_pages / m); |
885 | nr_pages++; | 736 | nr_pages++; |
886 | } | 737 | } |
738 | } while (ret > 0); | ||
887 | if (!error) | 739 | if (!error) |
888 | printk("\b\b\b\bdone\n"); | 740 | printk("\b\b\b\bdone\n"); |
741 | if (!snapshot_image_loaded(snapshot)) | ||
742 | error = -ENODATA; | ||
889 | return error; | 743 | return error; |
890 | } | 744 | } |
891 | 745 | ||
892 | /** | 746 | int swsusp_read(void) |
893 | * unpack_orig_addresses - copy the elements of @buf[] (1 page) to | ||
894 | * the PBEs in the list starting at @pbe | ||
895 | */ | ||
896 | |||
897 | static inline struct pbe *unpack_orig_addresses(unsigned long *buf, | ||
898 | struct pbe *pbe) | ||
899 | { | ||
900 | int j; | ||
901 | |||
902 | for (j = 0; j < PAGE_SIZE / sizeof(long) && pbe; j++) { | ||
903 | pbe->orig_address = buf[j]; | ||
904 | pbe = pbe->next; | ||
905 | } | ||
906 | return pbe; | ||
907 | } | ||
908 | |||
909 | /** | ||
910 | * load_image_metadata - load the image metadata using the swap map | ||
911 | * handle @handle and put them into the PBEs in the list @pblist | ||
912 | */ | ||
913 | |||
914 | static int load_image_metadata(struct pbe *pblist, struct swap_map_handle *handle) | ||
915 | { | ||
916 | struct pbe *p; | ||
917 | unsigned long *buf; | ||
918 | unsigned int n = 0; | ||
919 | int error = 0; | ||
920 | |||
921 | printk("Loading image metadata ... "); | ||
922 | buf = (unsigned long *)get_zeroed_page(GFP_ATOMIC); | ||
923 | if (!buf) | ||
924 | return -ENOMEM; | ||
925 | p = pblist; | ||
926 | while (p) { | ||
927 | error = swap_map_read_page(handle, buf); | ||
928 | if (error) | ||
929 | break; | ||
930 | p = unpack_orig_addresses(buf, p); | ||
931 | n++; | ||
932 | } | ||
933 | free_page((unsigned long)buf); | ||
934 | if (!error) | ||
935 | printk("done (%u pages loaded)\n", n); | ||
936 | return error; | ||
937 | } | ||
938 | |||
939 | int swsusp_read(struct pbe **pblist_ptr) | ||
940 | { | 747 | { |
941 | int error; | 748 | int error; |
942 | struct pbe *p, *pblist; | ||
943 | struct swap_map_handle handle; | 749 | struct swap_map_handle handle; |
750 | struct snapshot_handle snapshot; | ||
751 | struct swsusp_info *header; | ||
944 | unsigned int nr_pages; | 752 | unsigned int nr_pages; |
945 | 753 | ||
946 | if (IS_ERR(resume_bdev)) { | 754 | if (IS_ERR(resume_bdev)) { |
@@ -948,38 +756,19 @@ int swsusp_read(struct pbe **pblist_ptr) | |||
948 | return PTR_ERR(resume_bdev); | 756 | return PTR_ERR(resume_bdev); |
949 | } | 757 | } |
950 | 758 | ||
951 | error = get_swap_map_reader(&handle, swsusp_header.image); | 759 | memset(&snapshot, 0, sizeof(struct snapshot_handle)); |
760 | error = snapshot_write_next(&snapshot, PAGE_SIZE); | ||
761 | if (error < PAGE_SIZE) | ||
762 | return error < 0 ? error : -EFAULT; | ||
763 | header = (struct swsusp_info *)data_of(snapshot); | ||
764 | error = get_swap_reader(&handle, swsusp_header.image); | ||
952 | if (!error) | 765 | if (!error) |
953 | error = swap_map_read_page(&handle, &swsusp_info); | 766 | error = swap_read_page(&handle, header); |
954 | if (!error) | ||
955 | error = check_header(); | ||
956 | if (error) | ||
957 | return error; | ||
958 | nr_pages = swsusp_info.image_pages; | ||
959 | p = alloc_pagedir(nr_pages, GFP_ATOMIC, 0); | ||
960 | if (!p) | ||
961 | return -ENOMEM; | ||
962 | error = load_image_metadata(p, &handle); | ||
963 | if (!error) { | 767 | if (!error) { |
964 | mark_unsafe_pages(p); | 768 | nr_pages = header->image_pages; |
965 | pblist = alloc_pagedir(nr_pages, GFP_ATOMIC, 1); | 769 | error = load_image(&handle, &snapshot, nr_pages); |
966 | if (pblist) | ||
967 | copy_page_backup_list(pblist, p); | ||
968 | free_pagedir(p); | ||
969 | if (!pblist) | ||
970 | error = -ENOMEM; | ||
971 | |||
972 | /* Allocate memory for the image and read the data from swap */ | ||
973 | if (!error) | ||
974 | error = alloc_data_pages(pblist, GFP_ATOMIC, 1); | ||
975 | if (!error) { | ||
976 | release_eaten_pages(); | ||
977 | error = load_image_data(pblist, &handle, nr_pages); | ||
978 | } | ||
979 | if (!error) | ||
980 | *pblist_ptr = pblist; | ||
981 | } | 770 | } |
982 | release_swap_map_reader(&handle); | 771 | release_swap_reader(&handle); |
983 | 772 | ||
984 | blkdev_put(resume_bdev); | 773 | blkdev_put(resume_bdev); |
985 | 774 | ||
diff --git a/mm/swapfile.c b/mm/swapfile.c index 365ed6ff182d..4d11f9d84666 100644 --- a/mm/swapfile.c +++ b/mm/swapfile.c | |||
@@ -45,7 +45,7 @@ static const char Unused_offset[] = "Unused swap offset entry "; | |||
45 | 45 | ||
46 | struct swap_list_t swap_list = {-1, -1}; | 46 | struct swap_list_t swap_list = {-1, -1}; |
47 | 47 | ||
48 | struct swap_info_struct swap_info[MAX_SWAPFILES]; | 48 | static struct swap_info_struct swap_info[MAX_SWAPFILES]; |
49 | 49 | ||
50 | static DEFINE_MUTEX(swapon_mutex); | 50 | static DEFINE_MUTEX(swapon_mutex); |
51 | 51 | ||
@@ -417,6 +417,59 @@ void free_swap_and_cache(swp_entry_t entry) | |||
417 | } | 417 | } |
418 | } | 418 | } |
419 | 419 | ||
420 | #ifdef CONFIG_SOFTWARE_SUSPEND | ||
421 | /* | ||
422 | * Find the swap type that corresponds to given device (if any) | ||
423 | * | ||
424 | * This is needed for software suspend and is done in such a way that inode | ||
425 | * aliasing is allowed. | ||
426 | */ | ||
427 | int swap_type_of(dev_t device) | ||
428 | { | ||
429 | int i; | ||
430 | |||
431 | if (!device) | ||
432 | return -EINVAL; | ||
433 | spin_lock(&swap_lock); | ||
434 | for (i = 0; i < nr_swapfiles; i++) { | ||
435 | struct inode *inode; | ||
436 | |||
437 | if (!(swap_info[i].flags & SWP_WRITEOK)) | ||
438 | continue; | ||
439 | inode = swap_info->swap_file->f_dentry->d_inode; | ||
440 | if (S_ISBLK(inode->i_mode) && | ||
441 | device == MKDEV(imajor(inode), iminor(inode))) { | ||
442 | spin_unlock(&swap_lock); | ||
443 | return i; | ||
444 | } | ||
445 | } | ||
446 | spin_unlock(&swap_lock); | ||
447 | return -ENODEV; | ||
448 | } | ||
449 | |||
450 | /* | ||
451 | * Return either the total number of swap pages of given type, or the number | ||
452 | * of free pages of that type (depending on @free) | ||
453 | * | ||
454 | * This is needed for software suspend | ||
455 | */ | ||
456 | unsigned int count_swap_pages(int type, int free) | ||
457 | { | ||
458 | unsigned int n = 0; | ||
459 | |||
460 | if (type < nr_swapfiles) { | ||
461 | spin_lock(&swap_lock); | ||
462 | if (swap_info[type].flags & SWP_WRITEOK) { | ||
463 | n = swap_info[type].pages; | ||
464 | if (free) | ||
465 | n -= swap_info[type].inuse_pages; | ||
466 | } | ||
467 | spin_unlock(&swap_lock); | ||
468 | } | ||
469 | return n; | ||
470 | } | ||
471 | #endif | ||
472 | |||
420 | /* | 473 | /* |
421 | * No need to decide whether this PTE shares the swap entry with others, | 474 | * No need to decide whether this PTE shares the swap entry with others, |
422 | * just let do_wp_page work it out if a write is requested later - to | 475 | * just let do_wp_page work it out if a write is requested later - to |