[net-next-2.6.git] / kernel / power / snapshot.c

/*
 * linux/kernel/power/snapshot.c
 *
 * This file provide system snapshot/restore functionality.
 *
 * Copyright (C) 1998-2005 Pavel Machek <pavel@suse.cz>
 *
 * This file is released under the GPLv2, and is based on swsusp.c.
 *
 */


#include <linux/module.h>
#include <linux/mm.h>
#include <linux/suspend.h>
#include <linux/smp_lock.h>
#include <linux/delay.h>
#include <linux/bitops.h>
#include <linux/spinlock.h>
#include <linux/kernel.h>
#include <linux/pm.h>
#include <linux/device.h>
#include <linux/bootmem.h>
#include <linux/syscalls.h>
#include <linux/console.h>
#include <linux/highmem.h>

#include <asm/uaccess.h>
#include <asm/mmu_context.h>
#include <asm/pgtable.h>
#include <asm/tlbflush.h>
#include <asm/io.h>

#include "power.h"

struct pbe *pagedir_nosave;
unsigned int nr_copy_pages;

#ifdef CONFIG_HIGHMEM
unsigned int count_highmem_pages(void)
{
	struct zone *zone;
	unsigned long zone_pfn;
	unsigned int n = 0;

	for_each_zone (zone)
		if (is_highmem(zone)) {
			mark_free_pages(zone);
			for (zone_pfn = 0; zone_pfn < zone->spanned_pages; zone_pfn++) {
				struct page *page;
				unsigned long pfn = zone_pfn + zone->zone_start_pfn;
				if (!pfn_valid(pfn))
					continue;
				page = pfn_to_page(pfn);
				if (PageReserved(page))
					continue;
				if (PageNosaveFree(page))
					continue;
				n++;
			}
		}
	return n;
}

struct highmem_page {
	char *data;
	struct page *page;
	struct highmem_page *next;
};

static struct highmem_page *highmem_copy;

static int save_highmem_zone(struct zone *zone)
{
	unsigned long zone_pfn;
	mark_free_pages(zone);
	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
		struct page *page;
		struct highmem_page *save;
		void *kaddr;
		unsigned long pfn = zone_pfn + zone->zone_start_pfn;

		if (!(pfn%1000))
			printk(".");
		if (!pfn_valid(pfn))
			continue;
		page = pfn_to_page(pfn);
		/*
		 * This condition results from rvmalloc() sans vmalloc_32()
		 * and architectural memory reservations. This should be
		 * corrected eventually when the cases giving rise to this
		 * are better understood.
		 */
		if (PageReserved(page))
			continue;
		BUG_ON(PageNosave(page));
		if (PageNosaveFree(page))
			continue;
		save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC);
		if (!save)
			return -ENOMEM;
		save->next = highmem_copy;
		save->page = page;
		save->data = (void *) get_zeroed_page(GFP_ATOMIC);
		if (!save->data) {
			kfree(save);
			return -ENOMEM;
		}
		kaddr = kmap_atomic(page, KM_USER0);
		memcpy(save->data, kaddr, PAGE_SIZE);
		kunmap_atomic(kaddr, KM_USER0);
		highmem_copy = save;
	}
	return 0;
}

int save_highmem(void)
{
	struct zone *zone;
	int res = 0;

	pr_debug("swsusp: Saving Highmem\n");
	for_each_zone (zone) {
		if (is_highmem(zone))
			res = save_highmem_zone(zone);
		if (res)
			return res;
	}
	return 0;
}

int restore_highmem(void)
{
	printk("swsusp: Restoring Highmem\n");
	while (highmem_copy) {
		struct highmem_page *save = highmem_copy;
		void *kaddr;
		highmem_copy = save->next;

		kaddr = kmap_atomic(save->page, KM_USER0);
		memcpy(kaddr, save->data, PAGE_SIZE);
		kunmap_atomic(kaddr, KM_USER0);
		free_page((long) save->data);
		kfree(save);
	}
	return 0;
}
#endif

static int pfn_is_nosave(unsigned long pfn)
{
	unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT;
	unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT;
	return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn);
}

/**
 *	saveable - Determine whether a page should be cloned or not.
 *	@pfn:	The page
 *
 *	We save a page if it's Reserved, and not in the range of pages
 *	statically defined as 'unsaveable', or if it isn't reserved, and
 *	isn't part of a free chunk of pages.
 */

static int saveable(struct zone *zone, unsigned long *zone_pfn)
{
	unsigned long pfn = *zone_pfn + zone->zone_start_pfn;
	struct page *page;

	if (!pfn_valid(pfn))
		return 0;

	page = pfn_to_page(pfn);
	BUG_ON(PageReserved(page) && PageNosave(page));
	if (PageNosave(page))
		return 0;
	if (PageReserved(page) && pfn_is_nosave(pfn))
		return 0;
	if (PageNosaveFree(page))
		return 0;

	return 1;
}

unsigned int count_data_pages(void)
{
	struct zone *zone;
	unsigned long zone_pfn;
	unsigned int n = 0;

	for_each_zone (zone) {
		if (is_highmem(zone))
			continue;
		mark_free_pages(zone);
		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
			n += saveable(zone, &zone_pfn);
	}
	return n;
}

static void copy_data_pages(struct pbe *pblist)
{
	struct zone *zone;
	unsigned long zone_pfn;
	struct pbe *pbe, *p;

	pbe = pblist;
	for_each_zone (zone) {
		if (is_highmem(zone))
			continue;
		mark_free_pages(zone);
		/* This is necessary for swsusp_free() */
		for_each_pb_page (p, pblist)
			SetPageNosaveFree(virt_to_page(p));
		for_each_pbe (p, pblist)
			SetPageNosaveFree(virt_to_page(p->address));
		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
			if (saveable(zone, &zone_pfn)) {
				struct page *page;
				page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
				BUG_ON(!pbe);
				pbe->orig_address = (unsigned long)page_address(page);
				/* copy_page is not usable for copying task structs. */
				memcpy((void *)pbe->address, (void *)pbe->orig_address, PAGE_SIZE);
				pbe = pbe->next;
			}
		}
	}
	BUG_ON(pbe);
}


/**
 *	free_pagedir - free pages allocated with alloc_pagedir()
 */

void free_pagedir(struct pbe *pblist)
{
	struct pbe *pbe;

	while (pblist) {
		pbe = (pblist + PB_PAGE_SKIP)->next;
		ClearPageNosave(virt_to_page(pblist));
		ClearPageNosaveFree(virt_to_page(pblist));
		free_page((unsigned long)pblist);
		pblist = pbe;
	}
}

/**
 *	fill_pb_page - Create a list of PBEs on a given memory page
 */

static inline void fill_pb_page(struct pbe *pbpage)
{
	struct pbe *p;

	p = pbpage;
	pbpage += PB_PAGE_SKIP;
	do
		p->next = p + 1;
	while (++p < pbpage);
}

/**
 *	create_pbe_list - Create a list of PBEs on top of a given chain
 *	of memory pages allocated with alloc_pagedir()
 */

static inline void create_pbe_list(struct pbe *pblist, unsigned int nr_pages)
{
	struct pbe *pbpage, *p;
	unsigned int num = PBES_PER_PAGE;

	for_each_pb_page (pbpage, pblist) {
		if (num >= nr_pages)
			break;

		fill_pb_page(pbpage);
		num += PBES_PER_PAGE;
	}
	if (pbpage) {
		for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++)
			p->next = p + 1;
		p->next = NULL;
	}
}

/**
 *	On resume it is necessary to trace and eventually free the unsafe
 *	pages that have been allocated, because they are needed for I/O
 *	(on x86-64 we likely will "eat" these pages once again while
 *	creating the temporary page translation tables)
 */

struct eaten_page {
	struct eaten_page *next;
	char padding[PAGE_SIZE - sizeof(void *)];
};

static struct eaten_page *eaten_pages = NULL;

void release_eaten_pages(void)
{
	struct eaten_page *p, *q;

	p = eaten_pages;
	while (p) {
		q = p->next;
		/* We don't want swsusp_free() to free this page again */
		ClearPageNosave(virt_to_page(p));
		free_page((unsigned long)p);
		p = q;
	}
	eaten_pages = NULL;
}

/**
 *	@safe_needed - on resume, for storing the PBE list and the image,
 *	we can only use memory pages that do not conflict with the pages
 *	which had been used before suspend.
 *
 *	The unsafe pages are marked with the PG_nosave_free flag
 *
 *	Allocated but unusable (ie eaten) memory pages should be marked
 *	so that swsusp_free() can release them
 */

static inline void *alloc_image_page(gfp_t gfp_mask, int safe_needed)
{
	void *res;

	if (safe_needed)
		do {
			res = (void *)get_zeroed_page(gfp_mask);
			if (res && PageNosaveFree(virt_to_page(res))) {
				/* This is for swsusp_free() */
				SetPageNosave(virt_to_page(res));
				((struct eaten_page *)res)->next = eaten_pages;
				eaten_pages = res;
			}
		} while (res && PageNosaveFree(virt_to_page(res)));
	else
		res = (void *)get_zeroed_page(gfp_mask);
	if (res) {
		SetPageNosave(virt_to_page(res));
		SetPageNosaveFree(virt_to_page(res));
	}
	return res;
}

unsigned long get_safe_page(gfp_t gfp_mask)
{
	return (unsigned long)alloc_image_page(gfp_mask, 1);
}

/**
 *	alloc_pagedir - Allocate the page directory.
 *
 *	First, determine exactly how many pages we need and
 *	allocate them.
 *
 *	We arrange the pages in a chain: each page is an array of PBES_PER_PAGE
 *	struct pbe elements (pbes) and the last element in the page points
 *	to the next page.
 *
 *	On each page we set up a list of struct_pbe elements.
 */

struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, int safe_needed)
{
	unsigned int num;
	struct pbe *pblist, *pbe;

	if (!nr_pages)
		return NULL;

	pr_debug("alloc_pagedir(): nr_pages = %d\n", nr_pages);
	pblist = alloc_image_page(gfp_mask, safe_needed);
	/* FIXME: rewrite this ugly loop */
	for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages;
        		pbe = pbe->next, num += PBES_PER_PAGE) {
		pbe += PB_PAGE_SKIP;
		pbe->next = alloc_image_page(gfp_mask, safe_needed);
	}
	if (!pbe) { /* get_zeroed_page() failed */
		free_pagedir(pblist);
		pblist = NULL;
        } else
        	create_pbe_list(pblist, nr_pages);
	return pblist;
}

/**
 * Free pages we allocated for suspend. Suspend pages are alocated
 * before atomic copy, so we need to free them after resume.
 */

void swsusp_free(void)
{
	struct zone *zone;
	unsigned long zone_pfn;

	for_each_zone(zone) {
		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
			if (pfn_valid(zone_pfn + zone->zone_start_pfn)) {
				struct page *page;
				page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
				if (PageNosave(page) && PageNosaveFree(page)) {
					ClearPageNosave(page);
					ClearPageNosaveFree(page);
					free_page((long) page_address(page));
				}
			}
	}
}


/**
 *	enough_free_mem - Make sure we enough free memory to snapshot.
 *
 *	Returns TRUE or FALSE after checking the number of available
 *	free pages.
 */

static int enough_free_mem(unsigned int nr_pages)
{
	struct zone *zone;
	unsigned int n = 0;

	for_each_zone (zone)
		if (!is_highmem(zone))
			n += zone->free_pages;
	pr_debug("swsusp: available memory: %u pages\n", n);
	return n > (nr_pages + PAGES_FOR_IO +
		(nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE);
}

int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed)
{
	struct pbe *p;

	for_each_pbe (p, pblist) {
		p->address = (unsigned long)alloc_image_page(gfp_mask, safe_needed);
		if (!p->address)
			return -ENOMEM;
	}
	return 0;
}

static struct pbe *swsusp_alloc(unsigned int nr_pages)
{
	struct pbe *pblist;

	if (!(pblist = alloc_pagedir(nr_pages, GFP_ATOMIC | __GFP_COLD, 0))) {
		printk(KERN_ERR "suspend: Allocating pagedir failed.\n");
		return NULL;
	}

	if (alloc_data_pages(pblist, GFP_ATOMIC | __GFP_COLD, 0)) {
		printk(KERN_ERR "suspend: Allocating image pages failed.\n");
		swsusp_free();
		return NULL;
	}

	return pblist;
}

asmlinkage int swsusp_save(void)
{
	unsigned int nr_pages;

	pr_debug("swsusp: critical section: \n");

	drain_local_pages();
	nr_pages = count_data_pages();
	printk("swsusp: Need to copy %u pages\n", nr_pages);

	pr_debug("swsusp: pages needed: %u + %lu + %u, free: %u\n",
		 nr_pages,
		 (nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE,
		 PAGES_FOR_IO, nr_free_pages());

	if (!enough_free_mem(nr_pages)) {
		printk(KERN_ERR "swsusp: Not enough free memory\n");
		return -ENOMEM;
	}

	pagedir_nosave = swsusp_alloc(nr_pages);
	if (!pagedir_nosave)
		return -ENOMEM;

	/* During allocating of suspend pagedir, new cold pages may appear.
	 * Kill them.
	 */
	drain_local_pages();
	copy_data_pages(pagedir_nosave);

	/*
	 * End of critical section. From now on, we can write to memory,
	 * but we should not touch disk. This specially means we must _not_
	 * touch swap space! Except we must write out our image of course.
	 */

	nr_copy_pages = nr_pages;

	printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages);
	return 0;
}
Commit	Line	Data
25761b6e	1	/*
96bc7aec	2	* linux/kernel/power/snapshot.c
25761b6e	3	*
96bc7aec	4	* This file provide system snapshot/restore functionality.
25761b6e RW	5	*
	6	* Copyright (C) 1998-2005 Pavel Machek <pavel@suse.cz>
	7	*
	8	* This file is released under the GPLv2, and is based on swsusp.c.
	9	*
	10	*/
	11
	12
	13	#include <linux/module.h>
	14	#include <linux/mm.h>
	15	#include <linux/suspend.h>
	16	#include <linux/smp_lock.h>
25761b6e	17	#include <linux/delay.h>
25761b6e	18	#include <linux/bitops.h>
25761b6e	19	#include <linux/spinlock.h>
25761b6e	20	#include <linux/kernel.h>
25761b6e RW	21	#include <linux/pm.h>
25761b6e RW	22	#include <linux/device.h>
25761b6e RW	23	#include <linux/bootmem.h>
	24	#include <linux/syscalls.h>
	25	#include <linux/console.h>
	26	#include <linux/highmem.h>
25761b6e RW	27
	28	#include <asm/uaccess.h>
	29	#include <asm/mmu_context.h>
	30	#include <asm/pgtable.h>
	31	#include <asm/tlbflush.h>
	32	#include <asm/io.h>
	33
25761b6e RW	34	#include "power.h"
25761b6e RW	35
7088a5c0 RW	36	struct pbe *pagedir_nosave;
	37	unsigned int nr_copy_pages;
	38
25761b6e	39	#ifdef CONFIG_HIGHMEM
72a97e08 RW	40	unsigned int count_highmem_pages(void)
	41	{
	42	struct zone *zone;
	43	unsigned long zone_pfn;
	44	unsigned int n = 0;
	45
	46	for_each_zone (zone)
	47	if (is_highmem(zone)) {
	48	mark_free_pages(zone);
	49	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; zone_pfn++) {
	50	struct page *page;
	51	unsigned long pfn = zone_pfn + zone->zone_start_pfn;
	52	if (!pfn_valid(pfn))
	53	continue;
	54	page = pfn_to_page(pfn);
	55	if (PageReserved(page))
	56	continue;
	57	if (PageNosaveFree(page))
	58	continue;
	59	n++;
	60	}
	61	}
	62	return n;
	63	}
	64
25761b6e RW	65	struct highmem_page {
	66	char *data;
	67	struct page *page;
	68	struct highmem_page *next;
	69	};
	70
	71	static struct highmem_page *highmem_copy;
	72
	73	static int save_highmem_zone(struct zone *zone)
	74	{
	75	unsigned long zone_pfn;
	76	mark_free_pages(zone);
	77	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
	78	struct page *page;
	79	struct highmem_page *save;
	80	void *kaddr;
	81	unsigned long pfn = zone_pfn + zone->zone_start_pfn;
	82
	83	if (!(pfn%1000))
	84	printk(".");
	85	if (!pfn_valid(pfn))
	86	continue;
	87	page = pfn_to_page(pfn);
	88	/*
	89	* This condition results from rvmalloc() sans vmalloc_32()
	90	* and architectural memory reservations. This should be
	91	* corrected eventually when the cases giving rise to this
	92	* are better understood.
	93	*/
c8adb494	94	if (PageReserved(page))
25761b6e	95	continue;
25761b6e RW	96	BUG_ON(PageNosave(page));
	97	if (PageNosaveFree(page))
	98	continue;
	99	save = kmalloc(sizeof(struct highmem_page), GFP_ATOMIC);
	100	if (!save)
	101	return -ENOMEM;
	102	save->next = highmem_copy;
	103	save->page = page;
	104	save->data = (void *) get_zeroed_page(GFP_ATOMIC);
	105	if (!save->data) {
	106	kfree(save);
	107	return -ENOMEM;
	108	}
	109	kaddr = kmap_atomic(page, KM_USER0);
	110	memcpy(save->data, kaddr, PAGE_SIZE);
	111	kunmap_atomic(kaddr, KM_USER0);
	112	highmem_copy = save;
	113	}
	114	return 0;
	115	}
25761b6e	116
0fbeb5a4	117	int save_highmem(void)
25761b6e	118	{
25761b6e RW	119	struct zone *zone;
	120	int res = 0;
	121
	122	pr_debug("swsusp: Saving Highmem\n");
	123	for_each_zone (zone) {
	124	if (is_highmem(zone))
	125	res = save_highmem_zone(zone);
	126	if (res)
	127	return res;
	128	}
25761b6e RW	129	return 0;
	130	}
	131
	132	int restore_highmem(void)
	133	{
25761b6e RW	134	printk("swsusp: Restoring Highmem\n");
	135	while (highmem_copy) {
	136	struct highmem_page *save = highmem_copy;
	137	void *kaddr;
	138	highmem_copy = save->next;
	139
	140	kaddr = kmap_atomic(save->page, KM_USER0);
	141	memcpy(kaddr, save->data, PAGE_SIZE);
	142	kunmap_atomic(kaddr, KM_USER0);
	143	free_page((long) save->data);
	144	kfree(save);
	145	}
25761b6e RW	146	return 0;
25761b6e RW	147	}
0fbeb5a4	148	#endif
25761b6e RW	149
	150	static int pfn_is_nosave(unsigned long pfn)
	151	{
	152	unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT;
	153	unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT;
	154	return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn);
	155	}
	156
	157	/**
	158	* saveable - Determine whether a page should be cloned or not.
	159	* @pfn: The page
	160	*
	161	* We save a page if it's Reserved, and not in the range of pages
	162	* statically defined as 'unsaveable', or if it isn't reserved, and
	163	* isn't part of a free chunk of pages.
	164	*/
	165
de491861	166	static int saveable(struct zone zone, unsigned long zone_pfn)
25761b6e RW	167	{
25761b6e RW	168	unsigned long pfn = *zone_pfn + zone->zone_start_pfn;
de491861	169	struct page *page;
25761b6e RW	170
	171	if (!pfn_valid(pfn))
	172	return 0;
	173
	174	page = pfn_to_page(pfn);
	175	BUG_ON(PageReserved(page) && PageNosave(page));
	176	if (PageNosave(page))
	177	return 0;
72a97e08	178	if (PageReserved(page) && pfn_is_nosave(pfn))
25761b6e	179	return 0;
25761b6e RW	180	if (PageNosaveFree(page))
	181	return 0;
	182
	183	return 1;
	184	}
	185
72a97e08	186	unsigned int count_data_pages(void)
25761b6e RW	187	{
	188	struct zone *zone;
	189	unsigned long zone_pfn;
dc19d507	190	unsigned int n = 0;
25761b6e	191
25761b6e RW	192	for_each_zone (zone) {
	193	if (is_highmem(zone))
	194	continue;
	195	mark_free_pages(zone);
	196	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
a0f49651	197	n += saveable(zone, &zone_pfn);
25761b6e	198	}
a0f49651	199	return n;
25761b6e RW	200	}
25761b6e RW	201
a0f49651	202	static void copy_data_pages(struct pbe *pblist)
25761b6e RW	203	{
	204	struct zone *zone;
	205	unsigned long zone_pfn;
a0f49651	206	struct pbe pbe, p;
25761b6e	207
a0f49651	208	pbe = pblist;
25761b6e RW	209	for_each_zone (zone) {
	210	if (is_highmem(zone))
	211	continue;
	212	mark_free_pages(zone);
	213	/* This is necessary for swsusp_free() */
a0f49651	214	for_each_pb_page (p, pblist)
25761b6e	215	SetPageNosaveFree(virt_to_page(p));
a0f49651	216	for_each_pbe (p, pblist)
25761b6e RW	217	SetPageNosaveFree(virt_to_page(p->address));
	218	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
	219	if (saveable(zone, &zone_pfn)) {
de491861	220	struct page *page;
25761b6e RW	221	page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
	222	BUG_ON(!pbe);
	223	pbe->orig_address = (unsigned long)page_address(page);
	224	/* copy_page is not usable for copying task structs. */
	225	memcpy((void )pbe->address, (void )pbe->orig_address, PAGE_SIZE);
	226	pbe = pbe->next;
	227	}
	228	}
	229	}
	230	BUG_ON(pbe);
	231	}
	232
	233
	234	/**
	235	* free_pagedir - free pages allocated with alloc_pagedir()
	236	*/
	237
ed14b527	238	void free_pagedir(struct pbe *pblist)
25761b6e RW	239	{
	240	struct pbe *pbe;
	241
	242	while (pblist) {
	243	pbe = (pblist + PB_PAGE_SKIP)->next;
	244	ClearPageNosave(virt_to_page(pblist));
	245	ClearPageNosaveFree(virt_to_page(pblist));
	246	free_page((unsigned long)pblist);
	247	pblist = pbe;
	248	}
	249	}
	250
	251	/**
	252	* fill_pb_page - Create a list of PBEs on a given memory page
	253	*/
	254
	255	static inline void fill_pb_page(struct pbe *pbpage)
	256	{
	257	struct pbe *p;
	258
	259	p = pbpage;
	260	pbpage += PB_PAGE_SKIP;
	261	do
	262	p->next = p + 1;
	263	while (++p < pbpage);
	264	}
	265
	266	/**
	267	* create_pbe_list - Create a list of PBEs on top of a given chain
	268	* of memory pages allocated with alloc_pagedir()
	269	*/
	270
7088a5c0	271	static inline void create_pbe_list(struct pbe *pblist, unsigned int nr_pages)
25761b6e RW	272	{
25761b6e RW	273	struct pbe pbpage, p;
dc19d507	274	unsigned int num = PBES_PER_PAGE;
25761b6e RW	275
	276	for_each_pb_page (pbpage, pblist) {
	277	if (num >= nr_pages)
	278	break;
	279
	280	fill_pb_page(pbpage);
	281	num += PBES_PER_PAGE;
	282	}
	283	if (pbpage) {
	284	for (num -= PBES_PER_PAGE - 1, p = pbpage; num < nr_pages; p++, num++)
	285	p->next = p + 1;
	286	p->next = NULL;
	287	}
25761b6e RW	288	}
25761b6e RW	289
72a97e08 RW	290	/**
	291	* On resume it is necessary to trace and eventually free the unsafe
	292	* pages that have been allocated, because they are needed for I/O
	293	* (on x86-64 we likely will "eat" these pages once again while
	294	* creating the temporary page translation tables)
	295	*/
	296
	297	struct eaten_page {
	298	struct eaten_page *next;
	299	char padding[PAGE_SIZE - sizeof(void *)];
	300	};
	301
	302	static struct eaten_page *eaten_pages = NULL;
	303
	304	void release_eaten_pages(void)
	305	{
	306	struct eaten_page p, q;
	307
	308	p = eaten_pages;
	309	while (p) {
	310	q = p->next;
	311	/* We don't want swsusp_free() to free this page again */
	312	ClearPageNosave(virt_to_page(p));
	313	free_page((unsigned long)p);
	314	p = q;
	315	}
	316	eaten_pages = NULL;
	317	}
	318
054bd4c1 RW	319	/**
	320	* @safe_needed - on resume, for storing the PBE list and the image,
	321	* we can only use memory pages that do not conflict with the pages
	322	* which had been used before suspend.
	323	*
	324	* The unsafe pages are marked with the PG_nosave_free flag
	325	*
	326	* Allocated but unusable (ie eaten) memory pages should be marked
	327	* so that swsusp_free() can release them
	328	*/
	329
	330	static inline void *alloc_image_page(gfp_t gfp_mask, int safe_needed)
25761b6e	331	{
054bd4c1 RW	332	void *res;
	333
	334	if (safe_needed)
	335	do {
	336	res = (void *)get_zeroed_page(gfp_mask);
72a97e08	337	if (res && PageNosaveFree(virt_to_page(res))) {
054bd4c1 RW	338	/* This is for swsusp_free() */
054bd4c1 RW	339	SetPageNosave(virt_to_page(res));
72a97e08 RW	340	((struct eaten_page *)res)->next = eaten_pages;
	341	eaten_pages = res;
	342	}
054bd4c1 RW	343	} while (res && PageNosaveFree(virt_to_page(res)));
	344	else
	345	res = (void *)get_zeroed_page(gfp_mask);
25761b6e RW	346	if (res) {
	347	SetPageNosave(virt_to_page(res));
	348	SetPageNosaveFree(virt_to_page(res));
	349	}
	350	return res;
	351	}
	352
054bd4c1 RW	353	unsigned long get_safe_page(gfp_t gfp_mask)
	354	{
	355	return (unsigned long)alloc_image_page(gfp_mask, 1);
	356	}
	357
25761b6e RW	358	/**
	359	* alloc_pagedir - Allocate the page directory.
	360	*
	361	* First, determine exactly how many pages we need and
	362	* allocate them.
	363	*
	364	* We arrange the pages in a chain: each page is an array of PBES_PER_PAGE
	365	* struct pbe elements (pbes) and the last element in the page points
	366	* to the next page.
	367	*
	368	* On each page we set up a list of struct_pbe elements.
	369	*/
	370
054bd4c1	371	struct pbe *alloc_pagedir(unsigned int nr_pages, gfp_t gfp_mask, int safe_needed)
25761b6e	372	{
dc19d507	373	unsigned int num;
25761b6e RW	374	struct pbe pblist, pbe;
	375
	376	if (!nr_pages)
	377	return NULL;
	378
	379	pr_debug("alloc_pagedir(): nr_pages = %d\n", nr_pages);
054bd4c1	380	pblist = alloc_image_page(gfp_mask, safe_needed);
25761b6e RW	381	/* FIXME: rewrite this ugly loop */
	382	for (pbe = pblist, num = PBES_PER_PAGE; pbe && num < nr_pages;
	383	pbe = pbe->next, num += PBES_PER_PAGE) {
	384	pbe += PB_PAGE_SKIP;
054bd4c1	385	pbe->next = alloc_image_page(gfp_mask, safe_needed);
25761b6e RW	386	}
	387	if (!pbe) { /* get_zeroed_page() failed */
	388	free_pagedir(pblist);
	389	pblist = NULL;
7088a5c0 RW	390	} else
7088a5c0 RW	391	create_pbe_list(pblist, nr_pages);
25761b6e RW	392	return pblist;
	393	}
	394
	395	/**
	396	* Free pages we allocated for suspend. Suspend pages are alocated
	397	* before atomic copy, so we need to free them after resume.
	398	*/
	399
	400	void swsusp_free(void)
	401	{
	402	struct zone *zone;
	403	unsigned long zone_pfn;
	404
	405	for_each_zone(zone) {
	406	for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
	407	if (pfn_valid(zone_pfn + zone->zone_start_pfn)) {
dc19d507	408	struct page *page;
25761b6e RW	409	page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
	410	if (PageNosave(page) && PageNosaveFree(page)) {
	411	ClearPageNosave(page);
	412	ClearPageNosaveFree(page);
	413	free_page((long) page_address(page));
	414	}
	415	}
	416	}
	417	}
	418
	419
	420	/**
	421	* enough_free_mem - Make sure we enough free memory to snapshot.
	422	*
	423	* Returns TRUE or FALSE after checking the number of available
	424	* free pages.
	425	*/
	426
dc19d507	427	static int enough_free_mem(unsigned int nr_pages)
25761b6e	428	{
e5e2fa78 RW	429	struct zone *zone;
	430	unsigned int n = 0;
	431
	432	for_each_zone (zone)
	433	if (!is_highmem(zone))
	434	n += zone->free_pages;
	435	pr_debug("swsusp: available memory: %u pages\n", n);
	436	return n > (nr_pages + PAGES_FOR_IO +
a0f49651	437	(nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE);
25761b6e RW	438	}
25761b6e RW	439
054bd4c1 RW	440	int alloc_data_pages(struct pbe *pblist, gfp_t gfp_mask, int safe_needed)
	441	{
	442	struct pbe *p;
	443
	444	for_each_pbe (p, pblist) {
	445	p->address = (unsigned long)alloc_image_page(gfp_mask, safe_needed);
	446	if (!p->address)
	447	return -ENOMEM;
	448	}
	449	return 0;
	450	}
25761b6e	451
dc19d507	452	static struct pbe *swsusp_alloc(unsigned int nr_pages)
25761b6e	453	{
054bd4c1	454	struct pbe *pblist;
25761b6e	455
054bd4c1	456	if (!(pblist = alloc_pagedir(nr_pages, GFP_ATOMIC \| __GFP_COLD, 0))) {
25761b6e	457	printk(KERN_ERR "suspend: Allocating pagedir failed.\n");
a0f49651	458	return NULL;
25761b6e	459	}
25761b6e	460
054bd4c1 RW	461	if (alloc_data_pages(pblist, GFP_ATOMIC \| __GFP_COLD, 0)) {
	462	printk(KERN_ERR "suspend: Allocating image pages failed.\n");
	463	swsusp_free();
	464	return NULL;
25761b6e RW	465	}
25761b6e RW	466
a0f49651	467	return pblist;
25761b6e RW	468	}
25761b6e RW	469
2e32a43e	470	asmlinkage int swsusp_save(void)
25761b6e	471	{
dc19d507	472	unsigned int nr_pages;
25761b6e RW	473
25761b6e RW	474	pr_debug("swsusp: critical section: \n");
25761b6e RW	475
25761b6e RW	476	drain_local_pages();
a0f49651 RW	477	nr_pages = count_data_pages();
a0f49651 RW	478	printk("swsusp: Need to copy %u pages\n", nr_pages);
25761b6e RW	479
25761b6e RW	480	pr_debug("swsusp: pages needed: %u + %lu + %u, free: %u\n",
a0f49651 RW	481	nr_pages,
a0f49651 RW	482	(nr_pages + PBES_PER_PAGE - 1) / PBES_PER_PAGE,
25761b6e RW	483	PAGES_FOR_IO, nr_free_pages());
25761b6e RW	484
a0f49651	485	if (!enough_free_mem(nr_pages)) {
25761b6e RW	486	printk(KERN_ERR "swsusp: Not enough free memory\n");
	487	return -ENOMEM;
	488	}
	489
a0f49651 RW	490	pagedir_nosave = swsusp_alloc(nr_pages);
	491	if (!pagedir_nosave)
	492	return -ENOMEM;
25761b6e RW	493
	494	/* During allocating of suspend pagedir, new cold pages may appear.
	495	* Kill them.
	496	*/
	497	drain_local_pages();
a0f49651	498	copy_data_pages(pagedir_nosave);
25761b6e RW	499
	500	/*
	501	* End of critical section. From now on, we can write to memory,
	502	* but we should not touch disk. This specially means we must _not_
	503	* touch swap space! Except we must write out our image of course.
	504	*/
	505
a0f49651 RW	506	nr_copy_pages = nr_pages;
	507
	508	printk("swsusp: critical section/: done (%d pages copied)\n", nr_pages);
25761b6e RW	509	return 0;
25761b6e RW	510	}