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

/*
 * linux/kernel/power/swsusp.c
 *
 * This file provides code to write suspend image to swap and read it back.
 *
 * Copyright (C) 1998-2001 Gabor Kuti <seasons@fornax.hu>
 * Copyright (C) 1998,2001-2005 Pavel Machek <pavel@suse.cz>
 *
 * This file is released under the GPLv2.
 *
 * I'd like to thank the following people for their work:
 *
 * Pavel Machek <pavel@ucw.cz>:
 * Modifications, defectiveness pointing, being with me at the very beginning,
 * suspend to swap space, stop all tasks. Port to 2.4.18-ac and 2.5.17.
 *
 * Steve Doddi <dirk@loth.demon.co.uk>:
 * Support the possibility of hardware state restoring.
 *
 * Raph <grey.havens@earthling.net>:
 * Support for preserving states of network devices and virtual console
 * (including X and svgatextmode)
 *
 * Kurt Garloff <garloff@suse.de>:
 * Straightened the critical function in order to prevent compilers from
 * playing tricks with local variables.
 *
 * Andreas Mohr <a.mohr@mailto.de>
 *
 * Alex Badea <vampire@go.ro>:
 * Fixed runaway init
 *
 * Rafael J. Wysocki <rjw@sisk.pl>
 * Reworked the freeing of memory and the handling of swap
 *
 * More state savers are welcome. Especially for the scsi layer...
 *
 * For TODOs,FIXMEs also look in Documentation/power/swsusp.txt
 */

#include <linux/mm.h>
#include <linux/suspend.h>
#include <linux/spinlock.h>
#include <linux/kernel.h>
#include <linux/major.h>
#include <linux/swap.h>
#include <linux/pm.h>
#include <linux/swapops.h>
#include <linux/bootmem.h>
#include <linux/syscalls.h>
#include <linux/highmem.h>
#include <linux/time.h>

#include "power.h"

/*
 * Preferred image size in bytes (tunable via /sys/power/image_size).
 * When it is set to N, swsusp will do its best to ensure the image
 * size will not exceed N bytes, but if that is impossible, it will
 * try to create the smallest image possible.
 */
unsigned long image_size = 500 * 1024 * 1024;

int in_suspend __nosavedata = 0;

#ifdef CONFIG_HIGHMEM
unsigned int count_highmem_pages(void);
int restore_highmem(void);
#else
static inline int restore_highmem(void) { return 0; }
static inline unsigned int count_highmem_pages(void) { return 0; }
#endif

/**
 *	The following functions are used for tracing the allocated
 *	swap pages, so that they can be freed in case of an error.
 *
 *	The functions operate on a linked bitmap structure defined
 *	in power.h
 */

void free_bitmap(struct bitmap_page *bitmap)
{
	struct bitmap_page *bp;

	while (bitmap) {
		bp = bitmap->next;
		free_page((unsigned long)bitmap);
		bitmap = bp;
	}
}

struct bitmap_page *alloc_bitmap(unsigned int nr_bits)
{
	struct bitmap_page *bitmap, *bp;
	unsigned int n;

	if (!nr_bits)
		return NULL;

	bitmap = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL);
	bp = bitmap;
	for (n = BITMAP_PAGE_BITS; n < nr_bits; n += BITMAP_PAGE_BITS) {
		bp->next = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL);
		bp = bp->next;
		if (!bp) {
			free_bitmap(bitmap);
			return NULL;
		}
	}
	return bitmap;
}

static int bitmap_set(struct bitmap_page *bitmap, unsigned long bit)
{
	unsigned int n;

	n = BITMAP_PAGE_BITS;
	while (bitmap && n <= bit) {
		n += BITMAP_PAGE_BITS;
		bitmap = bitmap->next;
	}
	if (!bitmap)
		return -EINVAL;
	n -= BITMAP_PAGE_BITS;
	bit -= n;
	n = 0;
	while (bit >= BITS_PER_CHUNK) {
		bit -= BITS_PER_CHUNK;
		n++;
	}
	bitmap->chunks[n] |= (1UL << bit);
	return 0;
}

sector_t alloc_swapdev_block(int swap, struct bitmap_page *bitmap)
{
	unsigned long offset;

	offset = swp_offset(get_swap_page_of_type(swap));
	if (offset) {
		if (bitmap_set(bitmap, offset))
			swap_free(swp_entry(swap, offset));
		else
			return swapdev_block(swap, offset);
	}
	return 0;
}

void free_all_swap_pages(int swap, struct bitmap_page *bitmap)
{
	unsigned int bit, n;
	unsigned long test;

	bit = 0;
	while (bitmap) {
		for (n = 0; n < BITMAP_PAGE_CHUNKS; n++)
			for (test = 1UL; test; test <<= 1) {
				if (bitmap->chunks[n] & test)
					swap_free(swp_entry(swap, bit));
				bit++;
			}
		bitmap = bitmap->next;
	}
}

/**
 *	swsusp_show_speed - print the time elapsed between two events represented by
 *	@start and @stop
 *
 *	@nr_pages -	number of pages processed between @start and @stop
 *	@msg -		introductory message to print
 */

void swsusp_show_speed(struct timeval *start, struct timeval *stop,
			unsigned nr_pages, char *msg)
{
	s64 elapsed_centisecs64;
	int centisecs;
	int k;
	int kps;

	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
	centisecs = elapsed_centisecs64;
	if (centisecs == 0)
		centisecs = 1;	/* avoid div-by-zero */
	k = nr_pages * (PAGE_SIZE / 1024);
	kps = (k * 100) / centisecs;
	printk("%s %d kbytes in %d.%02d seconds (%d.%02d MB/s)\n", msg, k,
			centisecs / 100, centisecs % 100,
			kps / 1000, (kps % 1000) / 10);
}

/**
 *	swsusp_shrink_memory -  Try to free as much memory as needed
 *
 *	... but do not OOM-kill anyone
 *
 *	Notice: all userland should be stopped before it is called, or
 *	livelock is possible.
 */

#define SHRINK_BITE	10000
static inline unsigned long __shrink_memory(long tmp)
{
	if (tmp > SHRINK_BITE)
		tmp = SHRINK_BITE;
	return shrink_all_memory(tmp);
}

int swsusp_shrink_memory(void)
{
	long tmp;
	struct zone *zone;
	unsigned long pages = 0;
	unsigned int i = 0;
	char *p = "-\\|/";
	struct timeval start, stop;

	printk("Shrinking memory...  ");
	do_gettimeofday(&start);
	do {
		long size, highmem_size;

		highmem_size = count_highmem_pages();
		size = count_data_pages() + PAGES_FOR_IO;
		tmp = size;
		size += highmem_size;
		for_each_zone (zone)
			if (populated_zone(zone)) {
				if (is_highmem(zone)) {
					highmem_size -=
					zone_page_state(zone, NR_FREE_PAGES);
				} else {
					tmp -= zone_page_state(zone, NR_FREE_PAGES);
					tmp += zone->lowmem_reserve[ZONE_NORMAL];
					tmp += snapshot_additional_pages(zone);
				}
			}

		if (highmem_size < 0)
			highmem_size = 0;

		tmp += highmem_size;
		if (tmp > 0) {
			tmp = __shrink_memory(tmp);
			if (!tmp)
				return -ENOMEM;
			pages += tmp;
		} else if (size > image_size / PAGE_SIZE) {
			tmp = __shrink_memory(size - (image_size / PAGE_SIZE));
			pages += tmp;
		}
		printk("\b%c", p[i++%4]);
	} while (tmp > 0);
	do_gettimeofday(&stop);
	printk("\bdone (%lu pages freed)\n", pages);
	swsusp_show_speed(&start, &stop, pages, "Freed");

	return 0;
}

int swsusp_suspend(void)
{
	int error;

	if ((error = arch_prepare_suspend()))
		return error;

	local_irq_disable();
	/* At this point, device_suspend() has been called, but *not*
	 * device_power_down(). We *must* device_power_down() now.
	 * Otherwise, drivers for some devices (e.g. interrupt controllers)
	 * become desynchronized with the actual state of the hardware
	 * at resume time, and evil weirdness ensues.
	 */
	if ((error = device_power_down(PMSG_FREEZE))) {
		printk(KERN_ERR "Some devices failed to power down, aborting suspend\n");
		goto Enable_irqs;
	}

	save_processor_state();
	if ((error = swsusp_arch_suspend()))
		printk(KERN_ERR "Error %d suspending\n", error);
	/* Restore control flow magically appears here */
	restore_processor_state();
	/* NOTE:  device_power_up() is just a resume() for devices
	 * that suspended with irqs off ... no overall powerup.
	 */
	device_power_up();
 Enable_irqs:
	local_irq_enable();
	return error;
}

int swsusp_resume(void)
{
	int error;

	local_irq_disable();
	/* NOTE:  device_power_down() is just a suspend() with irqs off;
	 * it has no special "power things down" semantics
	 */
	if (device_power_down(PMSG_PRETHAW))
		printk(KERN_ERR "Some devices failed to power down, very bad\n");
	/* We'll ignore saved state, but this gets preempt count (etc) right */
	save_processor_state();
	error = restore_highmem();
	if (!error) {
		error = swsusp_arch_resume();
		/* The code below is only ever reached in case of a failure.
		 * Otherwise execution continues at place where
		 * swsusp_arch_suspend() was called
        	 */
		BUG_ON(!error);
		/* This call to restore_highmem() undos the previous one */
		restore_highmem();
	}
	/* The only reason why swsusp_arch_resume() can fail is memory being
	 * very tight, so we have to free it as soon as we can to avoid
	 * subsequent failures
	 */
	swsusp_free();
	restore_processor_state();
	touch_softlockup_watchdog();
	device_power_up();
	local_irq_enable();
	return error;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/kernel/power/swsusp.c
	3	*
96bc7aec	4	* This file provides code to write suspend image to swap and read it back.
1da177e4 LT	5	*
1da177e4 LT	6	* Copyright (C) 1998-2001 Gabor Kuti <seasons@fornax.hu>
25761b6e	7	* Copyright (C) 1998,2001-2005 Pavel Machek <pavel@suse.cz>
1da177e4 LT	8	*
	9	* This file is released under the GPLv2.
	10	*
	11	* I'd like to thank the following people for their work:
2e4d5822	12	*
1da177e4 LT	13	* Pavel Machek <pavel@ucw.cz>:
	14	* Modifications, defectiveness pointing, being with me at the very beginning,
	15	* suspend to swap space, stop all tasks. Port to 2.4.18-ac and 2.5.17.
	16	*
2e4d5822	17	* Steve Doddi <dirk@loth.demon.co.uk>:
1da177e4 LT	18	* Support the possibility of hardware state restoring.
	19	*
	20	* Raph <grey.havens@earthling.net>:
	21	* Support for preserving states of network devices and virtual console
	22	* (including X and svgatextmode)
	23	*
	24	* Kurt Garloff <garloff@suse.de>:
	25	* Straightened the critical function in order to prevent compilers from
	26	* playing tricks with local variables.
	27	*
	28	* Andreas Mohr <a.mohr@mailto.de>
	29	*
	30	* Alex Badea <vampire@go.ro>:
	31	* Fixed runaway init
	32	*
7088a5c0	33	* Rafael J. Wysocki <rjw@sisk.pl>
61159a31	34	* Reworked the freeing of memory and the handling of swap
7088a5c0	35	*
1da177e4 LT	36	* More state savers are welcome. Especially for the scsi layer...
	37	*
	38	* For TODOs,FIXMEs also look in Documentation/power/swsusp.txt
	39	*/
	40
1da177e4 LT	41	#include <linux/mm.h>
1da177e4 LT	42	#include <linux/suspend.h>
1da177e4	43	#include <linux/spinlock.h>
1da177e4 LT	44	#include <linux/kernel.h>
	45	#include <linux/major.h>
	46	#include <linux/swap.h>
	47	#include <linux/pm.h>
1da177e4 LT	48	#include <linux/swapops.h>
	49	#include <linux/bootmem.h>
	50	#include <linux/syscalls.h>
1da177e4	51	#include <linux/highmem.h>
0d3a9abe	52	#include <linux/time.h>
1da177e4 LT	53
	54	#include "power.h"
	55
ca0aec0f	56	/*
853609b6	57	* Preferred image size in bytes (tunable via /sys/power/image_size).
ca0aec0f	58	* When it is set to N, swsusp will do its best to ensure the image
853609b6	59	* size will not exceed N bytes, but if that is impossible, it will
ca0aec0f RW	60	* try to create the smallest image possible.
ca0aec0f RW	61	*/
853609b6	62	unsigned long image_size = 500 * 1024 * 1024;
ca0aec0f	63
f577eb30 RW	64	int in_suspend __nosavedata = 0;
f577eb30 RW	65
3448097f LT	66	#ifdef CONFIG_HIGHMEM
3448097f LT	67	unsigned int count_highmem_pages(void);
3448097f LT	68	int restore_highmem(void);
3448097f LT	69	#else
3448097f LT	70	static inline int restore_highmem(void) { return 0; }
	71	static inline unsigned int count_highmem_pages(void) { return 0; }
	72	#endif
	73
1da177e4	74	/**
f577eb30 RW	75	* The following functions are used for tracing the allocated
f577eb30 RW	76	* swap pages, so that they can be freed in case of an error.
7088a5c0	77	*
f577eb30	78	* The functions operate on a linked bitmap structure defined
61159a31	79	* in power.h
1da177e4	80	*/
7088a5c0	81
61159a31	82	void free_bitmap(struct bitmap_page *bitmap)
1da177e4	83	{
f577eb30	84	struct bitmap_page *bp;
1da177e4	85
f577eb30 RW	86	while (bitmap) {
	87	bp = bitmap->next;
	88	free_page((unsigned long)bitmap);
	89	bitmap = bp;
7088a5c0 RW	90	}
	91	}
	92
61159a31	93	struct bitmap_page *alloc_bitmap(unsigned int nr_bits)
7088a5c0	94	{
f577eb30 RW	95	struct bitmap_page bitmap, bp;
f577eb30 RW	96	unsigned int n;
7088a5c0	97
f577eb30	98	if (!nr_bits)
7088a5c0 RW	99	return NULL;
7088a5c0 RW	100
f577eb30 RW	101	bitmap = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL);
	102	bp = bitmap;
	103	for (n = BITMAP_PAGE_BITS; n < nr_bits; n += BITMAP_PAGE_BITS) {
	104	bp->next = (struct bitmap_page *)get_zeroed_page(GFP_KERNEL);
	105	bp = bp->next;
	106	if (!bp) {
	107	free_bitmap(bitmap);
7088a5c0 RW	108	return NULL;
7088a5c0 RW	109	}
1da177e4	110	}
f577eb30	111	return bitmap;
1da177e4 LT	112	}
1da177e4 LT	113
f577eb30	114	static int bitmap_set(struct bitmap_page *bitmap, unsigned long bit)
1da177e4	115	{
f577eb30 RW	116	unsigned int n;
	117
	118	n = BITMAP_PAGE_BITS;
	119	while (bitmap && n <= bit) {
	120	n += BITMAP_PAGE_BITS;
	121	bitmap = bitmap->next;
	122	}
	123	if (!bitmap)
	124	return -EINVAL;
	125	n -= BITMAP_PAGE_BITS;
	126	bit -= n;
	127	n = 0;
	128	while (bit >= BITS_PER_CHUNK) {
	129	bit -= BITS_PER_CHUNK;
	130	n++;
7088a5c0	131	}
f577eb30 RW	132	bitmap->chunks[n] \|= (1UL << bit);
f577eb30 RW	133	return 0;
7088a5c0	134	}
1da177e4	135
3aef83e0	136	sector_t alloc_swapdev_block(int swap, struct bitmap_page *bitmap)
7088a5c0	137	{
f577eb30 RW	138	unsigned long offset;
	139
	140	offset = swp_offset(get_swap_page_of_type(swap));
	141	if (offset) {
3aef83e0	142	if (bitmap_set(bitmap, offset))
f577eb30	143	swap_free(swp_entry(swap, offset));
3aef83e0 RW	144	else
3aef83e0 RW	145	return swapdev_block(swap, offset);
7088a5c0	146	}
3aef83e0	147	return 0;
7088a5c0	148	}
1da177e4	149
61159a31	150	void free_all_swap_pages(int swap, struct bitmap_page *bitmap)
7088a5c0	151	{
f577eb30 RW	152	unsigned int bit, n;
f577eb30 RW	153	unsigned long test;
7088a5c0	154
f577eb30 RW	155	bit = 0;
	156	while (bitmap) {
	157	for (n = 0; n < BITMAP_PAGE_CHUNKS; n++)
	158	for (test = 1UL; test; test <<= 1) {
	159	if (bitmap->chunks[n] & test)
	160	swap_free(swp_entry(swap, bit));
	161	bit++;
	162	}
	163	bitmap = bitmap->next;
1da177e4	164	}
7088a5c0 RW	165	}
7088a5c0 RW	166
0d3a9abe RW	167	/**
	168	* swsusp_show_speed - print the time elapsed between two events represented by
	169	* @start and @stop
	170	*
	171	* @nr_pages - number of pages processed between @start and @stop
	172	* @msg - introductory message to print
	173	*/
	174
	175	void swsusp_show_speed(struct timeval start, struct timeval stop,
	176	unsigned nr_pages, char *msg)
	177	{
	178	s64 elapsed_centisecs64;
	179	int centisecs;
	180	int k;
	181	int kps;
	182
	183	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
	184	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
	185	centisecs = elapsed_centisecs64;
	186	if (centisecs == 0)
	187	centisecs = 1; /* avoid div-by-zero */
	188	k = nr_pages * (PAGE_SIZE / 1024);
	189	kps = (k * 100) / centisecs;
	190	printk("%s %d kbytes in %d.%02d seconds (%d.%02d MB/s)\n", msg, k,
	191	centisecs / 100, centisecs % 100,
	192	kps / 1000, (kps % 1000) / 10);
	193	}
	194
72a97e08 RW	195	/**
	196	* swsusp_shrink_memory - Try to free as much memory as needed
	197	*
	198	* ... but do not OOM-kill anyone
	199	*
	200	* Notice: all userland should be stopped before it is called, or
	201	* livelock is possible.
	202	*/
	203
	204	#define SHRINK_BITE 10000
d6277db4 RW	205	static inline unsigned long __shrink_memory(long tmp)
	206	{
	207	if (tmp > SHRINK_BITE)
	208	tmp = SHRINK_BITE;
	209	return shrink_all_memory(tmp);
	210	}
72a97e08 RW	211
	212	int swsusp_shrink_memory(void)
	213	{
8357376d	214	long tmp;
72a97e08 RW	215	struct zone *zone;
	216	unsigned long pages = 0;
	217	unsigned int i = 0;
	218	char *p = "-\\\|/";
0d3a9abe	219	struct timeval start, stop;
72a97e08 RW	220
72a97e08 RW	221	printk("Shrinking memory... ");
0d3a9abe	222	do_gettimeofday(&start);
72a97e08	223	do {
8357376d RW	224	long size, highmem_size;
	225
	226	highmem_size = count_highmem_pages();
	227	size = count_data_pages() + PAGES_FOR_IO;
b3a93a25	228	tmp = size;
8357376d	229	size += highmem_size;
72a97e08	230	for_each_zone (zone)
8357376d RW	231	if (populated_zone(zone)) {
8357376d RW	232	if (is_highmem(zone)) {
d23ad423 CL	233	highmem_size -=
d23ad423 CL	234	zone_page_state(zone, NR_FREE_PAGES);
8357376d	235	} else {
d23ad423	236	tmp -= zone_page_state(zone, NR_FREE_PAGES);
8357376d RW	237	tmp += zone->lowmem_reserve[ZONE_NORMAL];
	238	tmp += snapshot_additional_pages(zone);
	239	}
a938c356	240	}
8357376d RW	241
	242	if (highmem_size < 0)
	243	highmem_size = 0;
	244
	245	tmp += highmem_size;
72a97e08	246	if (tmp > 0) {
d6277db4	247	tmp = __shrink_memory(tmp);
72a97e08 RW	248	if (!tmp)
	249	return -ENOMEM;
	250	pages += tmp;
853609b6	251	} else if (size > image_size / PAGE_SIZE) {
d6277db4	252	tmp = __shrink_memory(size - (image_size / PAGE_SIZE));
b3a93a25	253	pages += tmp;
72a97e08	254	}
72a97e08 RW	255	printk("\b%c", p[i++%4]);
72a97e08 RW	256	} while (tmp > 0);
0d3a9abe	257	do_gettimeofday(&stop);
72a97e08	258	printk("\bdone (%lu pages freed)\n", pages);
0d3a9abe	259	swsusp_show_speed(&start, &stop, pages, "Freed");
72a97e08 RW	260
	261	return 0;
	262	}
	263
1da177e4 LT	264	int swsusp_suspend(void)
	265	{
	266	int error;
0fbeb5a4	267
1da177e4 LT	268	if ((error = arch_prepare_suspend()))
1da177e4 LT	269	return error;
8357376d	270
1da177e4 LT	271	local_irq_disable();
	272	/* At this point, device_suspend() has been called, but not
	273	* device_power_down(). We must device_power_down() now.
	274	* Otherwise, drivers for some devices (e.g. interrupt controllers)
	275	* become desynchronized with the actual state of the hardware
	276	* at resume time, and evil weirdness ensues.
	277	*/
	278	if ((error = device_power_down(PMSG_FREEZE))) {
99dc7d63	279	printk(KERN_ERR "Some devices failed to power down, aborting suspend\n");
0fbeb5a4	280	goto Enable_irqs;
1da177e4	281	}
47b724f3	282
1da177e4 LT	283	save_processor_state();
1da177e4 LT	284	if ((error = swsusp_arch_suspend()))
99dc7d63	285	printk(KERN_ERR "Error %d suspending\n", error);
1da177e4 LT	286	/* Restore control flow magically appears here */
1da177e4 LT	287	restore_processor_state();
f1cc0a89 DB	288	/* NOTE: device_power_up() is just a resume() for devices
	289	* that suspended with irqs off ... no overall powerup.
	290	*/
1da177e4	291	device_power_up();
59a49335	292	Enable_irqs:
1da177e4 LT	293	local_irq_enable();
	294	return error;
	295	}
	296
	297	int swsusp_resume(void)
	298	{
	299	int error;
f1cc0a89	300
1da177e4	301	local_irq_disable();
f1cc0a89 DB	302	/* NOTE: device_power_down() is just a suspend() with irqs off;
	303	* it has no special "power things down" semantics
	304	*/
	305	if (device_power_down(PMSG_PRETHAW))
1da177e4 LT	306	printk(KERN_ERR "Some devices failed to power down, very bad\n");
	307	/* We'll ignore saved state, but this gets preempt count (etc) right */
	308	save_processor_state();
8357376d RW	309	error = restore_highmem();
	310	if (!error) {
	311	error = swsusp_arch_resume();
	312	/* The code below is only ever reached in case of a failure.
	313	* Otherwise execution continues at place where
	314	* swsusp_arch_suspend() was called
	315	*/
	316	BUG_ON(!error);
	317	/* This call to restore_highmem() undos the previous one */
	318	restore_highmem();
	319	}
2c1b4a5c RW	320	/* The only reason why swsusp_arch_resume() can fail is memory being
	321	* very tight, so we have to free it as soon as we can to avoid
	322	* subsequent failures
	323	*/
	324	swsusp_free();
1da177e4	325	restore_processor_state();
8446f1d3	326	touch_softlockup_watchdog();
1da177e4 LT	327	device_power_up();
	328	local_irq_enable();
	329	return error;
	330	}