[net-next-2.6.git] / drivers / md / dm-log.c

/*
 * Copyright (C) 2003 Sistina Software
 *
 * This file is released under the LGPL.
 */

#include <linux/init.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/vmalloc.h>

#include "dm-log.h"
#include "dm-io.h"

#define DM_MSG_PREFIX "mirror log"

static LIST_HEAD(_log_types);
static DEFINE_SPINLOCK(_lock);

int dm_register_dirty_log_type(struct dirty_log_type *type)
{
	spin_lock(&_lock);
	type->use_count = 0;
	list_add(&type->list, &_log_types);
	spin_unlock(&_lock);

	return 0;
}

int dm_unregister_dirty_log_type(struct dirty_log_type *type)
{
	spin_lock(&_lock);

	if (type->use_count)
		DMWARN("Attempt to unregister a log type that is still in use");
	else
		list_del(&type->list);

	spin_unlock(&_lock);

	return 0;
}

static struct dirty_log_type *get_type(const char *type_name)
{
	struct dirty_log_type *type;

	spin_lock(&_lock);
	list_for_each_entry (type, &_log_types, list)
		if (!strcmp(type_name, type->name)) {
			if (!type->use_count && !try_module_get(type->module)){
				spin_unlock(&_lock);
				return NULL;
			}
			type->use_count++;
			spin_unlock(&_lock);
			return type;
		}

	spin_unlock(&_lock);
	return NULL;
}

static void put_type(struct dirty_log_type *type)
{
	spin_lock(&_lock);
	if (!--type->use_count)
		module_put(type->module);
	spin_unlock(&_lock);
}

struct dirty_log *dm_create_dirty_log(const char *type_name, struct dm_target *ti,
				      unsigned int argc, char **argv)
{
	struct dirty_log_type *type;
	struct dirty_log *log;

	log = kmalloc(sizeof(*log), GFP_KERNEL);
	if (!log)
		return NULL;

	type = get_type(type_name);
	if (!type) {
		kfree(log);
		return NULL;
	}

	log->type = type;
	if (type->ctr(log, ti, argc, argv)) {
		kfree(log);
		put_type(type);
		return NULL;
	}

	return log;
}

void dm_destroy_dirty_log(struct dirty_log *log)
{
	log->type->dtr(log);
	put_type(log->type);
	kfree(log);
}

/*-----------------------------------------------------------------
 * Persistent and core logs share a lot of their implementation.
 * FIXME: need a reload method to be called from a resume
 *---------------------------------------------------------------*/
/*
 * Magic for persistent mirrors: "MiRr"
 */
#define MIRROR_MAGIC 0x4D695272

/*
 * The on-disk version of the metadata.
 */
#define MIRROR_DISK_VERSION 2
#define LOG_OFFSET 2

struct log_header {
	uint32_t magic;

	/*
	 * Simple, incrementing version. no backward
	 * compatibility.
	 */
	uint32_t version;
	sector_t nr_regions;
};

struct log_c {
	struct dm_target *ti;
	int touched;
	uint32_t region_size;
	unsigned int region_count;
	region_t sync_count;

	unsigned bitset_uint32_count;
	uint32_t *clean_bits;
	uint32_t *sync_bits;
	uint32_t *recovering_bits;	/* FIXME: this seems excessive */

	int sync_search;

	/* Resync flag */
	enum sync {
		DEFAULTSYNC,	/* Synchronize if necessary */
		NOSYNC,		/* Devices known to be already in sync */
		FORCESYNC,	/* Force a sync to happen */
	} sync;

	struct dm_io_request io_req;

	/*
	 * Disk log fields
	 */
	int log_dev_failed;
	struct dm_dev *log_dev;
	struct log_header header;

	struct io_region header_location;
	struct log_header *disk_header;
};

/*
 * The touched member needs to be updated every time we access
 * one of the bitsets.
 */
static  inline int log_test_bit(uint32_t *bs, unsigned bit)
{
	return ext2_test_bit(bit, (unsigned long *) bs) ? 1 : 0;
}

static inline void log_set_bit(struct log_c *l,
			       uint32_t *bs, unsigned bit)
{
	ext2_set_bit(bit, (unsigned long *) bs);
	l->touched = 1;
}

static inline void log_clear_bit(struct log_c *l,
				 uint32_t *bs, unsigned bit)
{
	ext2_clear_bit(bit, (unsigned long *) bs);
	l->touched = 1;
}

/*----------------------------------------------------------------
 * Header IO
 *--------------------------------------------------------------*/
static void header_to_disk(struct log_header *core, struct log_header *disk)
{
	disk->magic = cpu_to_le32(core->magic);
	disk->version = cpu_to_le32(core->version);
	disk->nr_regions = cpu_to_le64(core->nr_regions);
}

static void header_from_disk(struct log_header *core, struct log_header *disk)
{
	core->magic = le32_to_cpu(disk->magic);
	core->version = le32_to_cpu(disk->version);
	core->nr_regions = le64_to_cpu(disk->nr_regions);
}

static int rw_header(struct log_c *lc, int rw)
{
	lc->io_req.bi_rw = rw;
	lc->io_req.mem.ptr.vma = lc->disk_header;
	lc->io_req.notify.fn = NULL;

	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
}

static int read_header(struct log_c *log)
{
	int r;

	r = rw_header(log, READ);
	if (r)
		return r;

	header_from_disk(&log->header, log->disk_header);

	/* New log required? */
	if (log->sync != DEFAULTSYNC || log->header.magic != MIRROR_MAGIC) {
		log->header.magic = MIRROR_MAGIC;
		log->header.version = MIRROR_DISK_VERSION;
		log->header.nr_regions = 0;
	}

#ifdef __LITTLE_ENDIAN
	if (log->header.version == 1)
		log->header.version = 2;
#endif

	if (log->header.version != MIRROR_DISK_VERSION) {
		DMWARN("incompatible disk log version");
		return -EINVAL;
	}

	return 0;
}

static inline int write_header(struct log_c *log)
{
	header_to_disk(&log->header, log->disk_header);
	return rw_header(log, WRITE);
}

/*----------------------------------------------------------------
 * core log constructor/destructor
 *
 * argv contains region_size followed optionally by [no]sync
 *--------------------------------------------------------------*/
#define BYTE_SHIFT 3
static int create_log_context(struct dirty_log *log, struct dm_target *ti,
			      unsigned int argc, char **argv,
			      struct dm_dev *dev)
{
	enum sync sync = DEFAULTSYNC;

	struct log_c *lc;
	uint32_t region_size;
	unsigned int region_count;
	size_t bitset_size, buf_size;
	int r;

	if (argc < 1 || argc > 2) {
		DMWARN("wrong number of arguments to mirror log");
		return -EINVAL;
	}

	if (argc > 1) {
		if (!strcmp(argv[1], "sync"))
			sync = FORCESYNC;
		else if (!strcmp(argv[1], "nosync"))
			sync = NOSYNC;
		else {
			DMWARN("unrecognised sync argument to mirror log: %s",
			       argv[1]);
			return -EINVAL;
		}
	}

	if (sscanf(argv[0], "%u", &region_size) != 1) {
		DMWARN("invalid region size string");
		return -EINVAL;
	}

	region_count = dm_sector_div_up(ti->len, region_size);

	lc = kmalloc(sizeof(*lc), GFP_KERNEL);
	if (!lc) {
		DMWARN("couldn't allocate core log");
		return -ENOMEM;
	}

	lc->ti = ti;
	lc->touched = 0;
	lc->region_size = region_size;
	lc->region_count = region_count;
	lc->sync = sync;

	/*
	 * Work out how many "unsigned long"s we need to hold the bitset.
	 */
	bitset_size = dm_round_up(region_count,
				  sizeof(*lc->clean_bits) << BYTE_SHIFT);
	bitset_size >>= BYTE_SHIFT;

	lc->bitset_uint32_count = bitset_size / sizeof(*lc->clean_bits);

	/*
	 * Disk log?
	 */
	if (!dev) {
		lc->clean_bits = vmalloc(bitset_size);
		if (!lc->clean_bits) {
			DMWARN("couldn't allocate clean bitset");
			kfree(lc);
			return -ENOMEM;
		}
		lc->disk_header = NULL;
	} else {
		lc->log_dev = dev;
		lc->log_dev_failed = 0;
		lc->header_location.bdev = lc->log_dev->bdev;
		lc->header_location.sector = 0;

		/*
		 * Buffer holds both header and bitset.
		 */
		buf_size = dm_round_up((LOG_OFFSET << SECTOR_SHIFT) +
				       bitset_size, ti->limits.hardsect_size);
		lc->header_location.count = buf_size >> SECTOR_SHIFT;
		lc->io_req.mem.type = DM_IO_VMA;
		lc->io_req.client = dm_io_client_create(dm_div_up(buf_size,
								   PAGE_SIZE));
		if (IS_ERR(lc->io_req.client)) {
			r = PTR_ERR(lc->io_req.client);
			DMWARN("couldn't allocate disk io client");
			kfree(lc);
			return -ENOMEM;
		}

		lc->disk_header = vmalloc(buf_size);
		if (!lc->disk_header) {
			DMWARN("couldn't allocate disk log buffer");
			kfree(lc);
			return -ENOMEM;
		}

		lc->clean_bits = (void *)lc->disk_header +
				 (LOG_OFFSET << SECTOR_SHIFT);
	}

	memset(lc->clean_bits, -1, bitset_size);

	lc->sync_bits = vmalloc(bitset_size);
	if (!lc->sync_bits) {
		DMWARN("couldn't allocate sync bitset");
		if (!dev)
			vfree(lc->clean_bits);
		vfree(lc->disk_header);
		kfree(lc);
		return -ENOMEM;
	}
	memset(lc->sync_bits, (sync == NOSYNC) ? -1 : 0, bitset_size);
	lc->sync_count = (sync == NOSYNC) ? region_count : 0;

	lc->recovering_bits = vmalloc(bitset_size);
	if (!lc->recovering_bits) {
		DMWARN("couldn't allocate sync bitset");
		vfree(lc->sync_bits);
		if (!dev)
			vfree(lc->clean_bits);
		vfree(lc->disk_header);
		kfree(lc);
		return -ENOMEM;
	}
	memset(lc->recovering_bits, 0, bitset_size);
	lc->sync_search = 0;
	log->context = lc;

	return 0;
}

static int core_ctr(struct dirty_log *log, struct dm_target *ti,
		    unsigned int argc, char **argv)
{
	return create_log_context(log, ti, argc, argv, NULL);
}

static void destroy_log_context(struct log_c *lc)
{
	vfree(lc->sync_bits);
	vfree(lc->recovering_bits);
	kfree(lc);
}

static void core_dtr(struct dirty_log *log)
{
	struct log_c *lc = (struct log_c *) log->context;

	vfree(lc->clean_bits);
	destroy_log_context(lc);
}

/*----------------------------------------------------------------
 * disk log constructor/destructor
 *
 * argv contains log_device region_size followed optionally by [no]sync
 *--------------------------------------------------------------*/
static int disk_ctr(struct dirty_log *log, struct dm_target *ti,
		    unsigned int argc, char **argv)
{
	int r;
	struct dm_dev *dev;

	if (argc < 2 || argc > 3) {
		DMWARN("wrong number of arguments to disk mirror log");
		return -EINVAL;
	}

	r = dm_get_device(ti, argv[0], 0, 0 /* FIXME */,
			  FMODE_READ | FMODE_WRITE, &dev);
	if (r)
		return r;

	r = create_log_context(log, ti, argc - 1, argv + 1, dev);
	if (r) {
		dm_put_device(ti, dev);
		return r;
	}

	return 0;
}

static void disk_dtr(struct dirty_log *log)
{
	struct log_c *lc = (struct log_c *) log->context;

	dm_put_device(lc->ti, lc->log_dev);
	vfree(lc->disk_header);
	dm_io_client_destroy(lc->io_req.client);
	destroy_log_context(lc);
}

static int count_bits32(uint32_t *addr, unsigned size)
{
	int count = 0, i;

	for (i = 0; i < size; i++) {
		count += hweight32(*(addr+i));
	}
	return count;
}

static void fail_log_device(struct log_c *lc)
{
	if (lc->log_dev_failed)
		return;

	lc->log_dev_failed = 1;
	dm_table_event(lc->ti->table);
}

static int disk_resume(struct dirty_log *log)
{
	int r;
	unsigned i;
	struct log_c *lc = (struct log_c *) log->context;
	size_t size = lc->bitset_uint32_count * sizeof(uint32_t);

	/* read the disk header */
	r = read_header(lc);
	if (r) {
		DMWARN("%s: Failed to read header on mirror log device",
		       lc->log_dev->name);
		fail_log_device(lc);
		/*
		 * If the log device cannot be read, we must assume
		 * all regions are out-of-sync.  If we simply return
		 * here, the state will be uninitialized and could
		 * lead us to return 'in-sync' status for regions
		 * that are actually 'out-of-sync'.
		 */
		lc->header.nr_regions = 0;
	}

	/* set or clear any new bits -- device has grown */
	if (lc->sync == NOSYNC)
		for (i = lc->header.nr_regions; i < lc->region_count; i++)
			/* FIXME: amazingly inefficient */
			log_set_bit(lc, lc->clean_bits, i);
	else
		for (i = lc->header.nr_regions; i < lc->region_count; i++)
			/* FIXME: amazingly inefficient */
			log_clear_bit(lc, lc->clean_bits, i);

	/* clear any old bits -- device has shrunk */
	for (i = lc->region_count; i % (sizeof(*lc->clean_bits) << BYTE_SHIFT); i++)
		log_clear_bit(lc, lc->clean_bits, i);

	/* copy clean across to sync */
	memcpy(lc->sync_bits, lc->clean_bits, size);
	lc->sync_count = count_bits32(lc->clean_bits, lc->bitset_uint32_count);
	lc->sync_search = 0;

	/* set the correct number of regions in the header */
	lc->header.nr_regions = lc->region_count;

	/* write the new header */
	r = write_header(lc);
	if (r) {
		DMWARN("%s: Failed to write header on mirror log device",
		       lc->log_dev->name);
		fail_log_device(lc);
	}

	return r;
}

static uint32_t core_get_region_size(struct dirty_log *log)
{
	struct log_c *lc = (struct log_c *) log->context;
	return lc->region_size;
}

static int core_resume(struct dirty_log *log)
{
	struct log_c *lc = (struct log_c *) log->context;
	lc->sync_search = 0;
	return 0;
}

static int core_is_clean(struct dirty_log *log, region_t region)
{
	struct log_c *lc = (struct log_c *) log->context;
	return log_test_bit(lc->clean_bits, region);
}

static int core_in_sync(struct dirty_log *log, region_t region, int block)
{
	struct log_c *lc = (struct log_c *) log->context;
	return log_test_bit(lc->sync_bits, region);
}

static int core_flush(struct dirty_log *log)
{
	/* no op */
	return 0;
}

static int disk_flush(struct dirty_log *log)
{
	int r;
	struct log_c *lc = (struct log_c *) log->context;

	/* only write if the log has changed */
	if (!lc->touched)
		return 0;

	r = write_header(lc);
	if (r)
		fail_log_device(lc);
	else
		lc->touched = 0;

	return r;
}

static void core_mark_region(struct dirty_log *log, region_t region)
{
	struct log_c *lc = (struct log_c *) log->context;
	log_clear_bit(lc, lc->clean_bits, region);
}

static void core_clear_region(struct dirty_log *log, region_t region)
{
	struct log_c *lc = (struct log_c *) log->context;
	log_set_bit(lc, lc->clean_bits, region);
}

static int core_get_resync_work(struct dirty_log *log, region_t *region)
{
	struct log_c *lc = (struct log_c *) log->context;

	if (lc->sync_search >= lc->region_count)
		return 0;

	do {
		*region = ext2_find_next_zero_bit(
					     (unsigned long *) lc->sync_bits,
					     lc->region_count,
					     lc->sync_search);
		lc->sync_search = *region + 1;

		if (*region >= lc->region_count)
			return 0;

	} while (log_test_bit(lc->recovering_bits, *region));

	log_set_bit(lc, lc->recovering_bits, *region);
	return 1;
}

static void core_set_region_sync(struct dirty_log *log, region_t region,
				 int in_sync)
{
	struct log_c *lc = (struct log_c *) log->context;

	log_clear_bit(lc, lc->recovering_bits, region);
	if (in_sync) {
		log_set_bit(lc, lc->sync_bits, region);
                lc->sync_count++;
        } else if (log_test_bit(lc->sync_bits, region)) {
		lc->sync_count--;
		log_clear_bit(lc, lc->sync_bits, region);
	}
}

static region_t core_get_sync_count(struct dirty_log *log)
{
        struct log_c *lc = (struct log_c *) log->context;

        return lc->sync_count;
}

#define	DMEMIT_SYNC \
	if (lc->sync != DEFAULTSYNC) \
		DMEMIT("%ssync ", lc->sync == NOSYNC ? "no" : "")

static int core_status(struct dirty_log *log, status_type_t status,
		       char *result, unsigned int maxlen)
{
	int sz = 0;
	struct log_c *lc = log->context;

	switch(status) {
	case STATUSTYPE_INFO:
		DMEMIT("1 %s", log->type->name);
		break;

	case STATUSTYPE_TABLE:
		DMEMIT("%s %u %u ", log->type->name,
		       lc->sync == DEFAULTSYNC ? 1 : 2, lc->region_size);
		DMEMIT_SYNC;
	}

	return sz;
}

static int disk_status(struct dirty_log *log, status_type_t status,
		       char *result, unsigned int maxlen)
{
	int sz = 0;
	struct log_c *lc = log->context;

	switch(status) {
	case STATUSTYPE_INFO:
		DMEMIT("3 %s %s %c", log->type->name, lc->log_dev->name,
		       lc->log_dev_failed ? 'D' : 'A');
		break;

	case STATUSTYPE_TABLE:
		DMEMIT("%s %u %s %u ", log->type->name,
		       lc->sync == DEFAULTSYNC ? 2 : 3, lc->log_dev->name,
		       lc->region_size);
		DMEMIT_SYNC;
	}

	return sz;
}

static struct dirty_log_type _core_type = {
	.name = "core",
	.module = THIS_MODULE,
	.ctr = core_ctr,
	.dtr = core_dtr,
	.resume = core_resume,
	.get_region_size = core_get_region_size,
	.is_clean = core_is_clean,
	.in_sync = core_in_sync,
	.flush = core_flush,
	.mark_region = core_mark_region,
	.clear_region = core_clear_region,
	.get_resync_work = core_get_resync_work,
	.set_region_sync = core_set_region_sync,
	.get_sync_count = core_get_sync_count,
	.status = core_status,
};

static struct dirty_log_type _disk_type = {
	.name = "disk",
	.module = THIS_MODULE,
	.ctr = disk_ctr,
	.dtr = disk_dtr,
	.postsuspend = disk_flush,
	.resume = disk_resume,
	.get_region_size = core_get_region_size,
	.is_clean = core_is_clean,
	.in_sync = core_in_sync,
	.flush = disk_flush,
	.mark_region = core_mark_region,
	.clear_region = core_clear_region,
	.get_resync_work = core_get_resync_work,
	.set_region_sync = core_set_region_sync,
	.get_sync_count = core_get_sync_count,
	.status = disk_status,
};

int __init dm_dirty_log_init(void)
{
	int r;

	r = dm_register_dirty_log_type(&_core_type);
	if (r)
		DMWARN("couldn't register core log");

	r = dm_register_dirty_log_type(&_disk_type);
	if (r) {
		DMWARN("couldn't register disk type");
		dm_unregister_dirty_log_type(&_core_type);
	}

	return r;
}

void dm_dirty_log_exit(void)
{
	dm_unregister_dirty_log_type(&_disk_type);
	dm_unregister_dirty_log_type(&_core_type);
}

EXPORT_SYMBOL(dm_register_dirty_log_type);
EXPORT_SYMBOL(dm_unregister_dirty_log_type);
EXPORT_SYMBOL(dm_create_dirty_log);
EXPORT_SYMBOL(dm_destroy_dirty_log);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Copyright (C) 2003 Sistina Software
	3	*
	4	* This file is released under the LGPL.
	5	*/
	6
	7	#include <linux/init.h>
	8	#include <linux/slab.h>
	9	#include <linux/module.h>
	10	#include <linux/vmalloc.h>
	11
	12	#include "dm-log.h"
	13	#include "dm-io.h"
	14
72d94861 AK	15	#define DM_MSG_PREFIX "mirror log"
72d94861 AK	16
1da177e4 LT	17	static LIST_HEAD(_log_types);
	18	static DEFINE_SPINLOCK(_lock);
	19
	20	int dm_register_dirty_log_type(struct dirty_log_type *type)
	21	{
	22	spin_lock(&_lock);
	23	type->use_count = 0;
	24	list_add(&type->list, &_log_types);
	25	spin_unlock(&_lock);
	26
	27	return 0;
	28	}
	29
	30	int dm_unregister_dirty_log_type(struct dirty_log_type *type)
	31	{
	32	spin_lock(&_lock);
	33
	34	if (type->use_count)
	35	DMWARN("Attempt to unregister a log type that is still in use");
	36	else
	37	list_del(&type->list);
	38
	39	spin_unlock(&_lock);
	40
	41	return 0;
	42	}
	43
	44	static struct dirty_log_type get_type(const char type_name)
	45	{
	46	struct dirty_log_type *type;
	47
	48	spin_lock(&_lock);
	49	list_for_each_entry (type, &_log_types, list)
	50	if (!strcmp(type_name, type->name)) {
	51	if (!type->use_count && !try_module_get(type->module)){
	52	spin_unlock(&_lock);
	53	return NULL;
	54	}
	55	type->use_count++;
	56	spin_unlock(&_lock);
	57	return type;
	58	}
	59
	60	spin_unlock(&_lock);
	61	return NULL;
	62	}
	63
	64	static void put_type(struct dirty_log_type *type)
	65	{
	66	spin_lock(&_lock);
	67	if (!--type->use_count)
	68	module_put(type->module);
	69	spin_unlock(&_lock);
	70	}
	71
	72	struct dirty_log dm_create_dirty_log(const char type_name, struct dm_target *ti,
	73	unsigned int argc, char **argv)
	74	{
	75	struct dirty_log_type *type;
	76	struct dirty_log *log;
	77
	78	log = kmalloc(sizeof(*log), GFP_KERNEL);
	79	if (!log)
	80	return NULL;
81
82	type = get_type(type_name);
83	if (!type) {
84	kfree(log);
85	return NULL;
86	}
87
88	log->type = type;
89	if (type->ctr(log, ti, argc, argv)) {
90	kfree(log);
91	put_type(type);
92	return NULL;
93	}
94
95	return log;
96	}
97
98	void dm_destroy_dirty_log(struct dirty_log *log)
99	{
100	log->type->dtr(log);
101	put_type(log->type);
102	kfree(log);
103	}
104
105	/*-----------------------------------------------------------------
106	* Persistent and core logs share a lot of their implementation.
107	* FIXME: need a reload method to be called from a resume
108	---------------------------------------------------------------/
109	/*
110	* Magic for persistent mirrors: "MiRr"
111	*/
112	#define MIRROR_MAGIC 0x4D695272
113
114	/*
115	* The on-disk version of the metadata.
116	*/
a4fc4717	117	#define MIRROR_DISK_VERSION 2
1da177e4 LT	118	#define LOG_OFFSET 2
	119
	120	struct log_header {
	121	uint32_t magic;
	122
	123	/*
	124	* Simple, incrementing version. no backward
	125	* compatibility.
	126	*/
	127	uint32_t version;
	128	sector_t nr_regions;
	129	};
	130
	131	struct log_c {
	132	struct dm_target *ti;
	133	int touched;
	134	uint32_t region_size;
	135	unsigned int region_count;
	136	region_t sync_count;
	137
	138	unsigned bitset_uint32_count;
	139	uint32_t *clean_bits;
	140	uint32_t *sync_bits;
	141	uint32_t recovering_bits; / FIXME: this seems excessive */
	142
	143	int sync_search;
	144
	145	/* Resync flag */
	146	enum sync {
	147	DEFAULTSYNC, /* Synchronize if necessary */
	148	NOSYNC, /* Devices known to be already in sync */
	149	FORCESYNC, /* Force a sync to happen */
	150	} sync;
	151
5d234d1e HM	152	struct dm_io_request io_req;
5d234d1e HM	153
1da177e4 LT	154	/*
	155	* Disk log fields
	156	*/
01d03a66	157	int log_dev_failed;
1da177e4 LT	158	struct dm_dev *log_dev;
	159	struct log_header header;
	160
	161	struct io_region header_location;
	162	struct log_header *disk_header;
1da177e4 LT	163	};
	164
	165	/*
	166	* The touched member needs to be updated every time we access
	167	* one of the bitsets.
	168	*/
	169	static inline int log_test_bit(uint32_t *bs, unsigned bit)
	170	{
a4fc4717	171	return ext2_test_bit(bit, (unsigned long *) bs) ? 1 : 0;
1da177e4 LT	172	}
	173
	174	static inline void log_set_bit(struct log_c *l,
	175	uint32_t *bs, unsigned bit)
	176	{
a4fc4717	177	ext2_set_bit(bit, (unsigned long *) bs);
1da177e4 LT	178	l->touched = 1;
	179	}
	180
	181	static inline void log_clear_bit(struct log_c *l,
	182	uint32_t *bs, unsigned bit)
	183	{
a4fc4717	184	ext2_clear_bit(bit, (unsigned long *) bs);
1da177e4 LT	185	l->touched = 1;
	186	}
	187
	188	/*----------------------------------------------------------------
	189	* Header IO
	190	--------------------------------------------------------------/
	191	static void header_to_disk(struct log_header core, struct log_header disk)
	192	{
	193	disk->magic = cpu_to_le32(core->magic);
	194	disk->version = cpu_to_le32(core->version);
	195	disk->nr_regions = cpu_to_le64(core->nr_regions);
	196	}
	197
	198	static void header_from_disk(struct log_header core, struct log_header disk)
	199	{
	200	core->magic = le32_to_cpu(disk->magic);
	201	core->version = le32_to_cpu(disk->version);
	202	core->nr_regions = le64_to_cpu(disk->nr_regions);
	203	}
	204
5d234d1e HM	205	static int rw_header(struct log_c *lc, int rw)
	206	{
	207	lc->io_req.bi_rw = rw;
	208	lc->io_req.mem.ptr.vma = lc->disk_header;
	209	lc->io_req.notify.fn = NULL;
	210
	211	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
	212	}
	213
1da177e4 LT	214	static int read_header(struct log_c *log)
	215	{
	216	int r;
1da177e4	217
5d234d1e	218	r = rw_header(log, READ);
1da177e4 LT	219	if (r)
	220	return r;
	221
	222	header_from_disk(&log->header, log->disk_header);
	223
	224	/* New log required? */
	225	if (log->sync != DEFAULTSYNC \|\| log->header.magic != MIRROR_MAGIC) {
	226	log->header.magic = MIRROR_MAGIC;
	227	log->header.version = MIRROR_DISK_VERSION;
	228	log->header.nr_regions = 0;
	229	}
	230
a4fc4717 PC	231	#ifdef __LITTLE_ENDIAN
	232	if (log->header.version == 1)
	233	log->header.version = 2;
	234	#endif
	235
1da177e4 LT	236	if (log->header.version != MIRROR_DISK_VERSION) {
	237	DMWARN("incompatible disk log version");
	238	return -EINVAL;
	239	}
	240
	241	return 0;
	242	}
	243
	244	static inline int write_header(struct log_c *log)
	245	{
1da177e4	246	header_to_disk(&log->header, log->disk_header);
5d234d1e	247	return rw_header(log, WRITE);
1da177e4 LT	248	}
1da177e4 LT	249
1da177e4 LT	250	/*----------------------------------------------------------------
	251	* core log constructor/destructor
	252	*
	253	* argv contains region_size followed optionally by [no]sync
	254	--------------------------------------------------------------/
	255	#define BYTE_SHIFT 3
b7cca195 AK	256	static int create_log_context(struct dirty_log log, struct dm_target ti,
	257	unsigned int argc, char **argv,
	258	struct dm_dev *dev)
1da177e4 LT	259	{
	260	enum sync sync = DEFAULTSYNC;
	261
	262	struct log_c *lc;
	263	uint32_t region_size;
	264	unsigned int region_count;
b7cca195	265	size_t bitset_size, buf_size;
5d234d1e	266	int r;
1da177e4 LT	267
	268	if (argc < 1 \|\| argc > 2) {
	269	DMWARN("wrong number of arguments to mirror log");
	270	return -EINVAL;
	271	}
	272
	273	if (argc > 1) {
	274	if (!strcmp(argv[1], "sync"))
	275	sync = FORCESYNC;
	276	else if (!strcmp(argv[1], "nosync"))
	277	sync = NOSYNC;
	278	else {
	279	DMWARN("unrecognised sync argument to mirror log: %s",
	280	argv[1]);
	281	return -EINVAL;
	282	}
	283	}
	284
	285	if (sscanf(argv[0], "%u", &region_size) != 1) {
	286	DMWARN("invalid region size string");
	287	return -EINVAL;
	288	}
	289
	290	region_count = dm_sector_div_up(ti->len, region_size);
	291
	292	lc = kmalloc(sizeof(*lc), GFP_KERNEL);
	293	if (!lc) {
	294	DMWARN("couldn't allocate core log");
	295	return -ENOMEM;
	296	}
	297
	298	lc->ti = ti;
	299	lc->touched = 0;
	300	lc->region_size = region_size;
	301	lc->region_count = region_count;
	302	lc->sync = sync;
	303
	304	/*
0e56822d	305	* Work out how many "unsigned long"s we need to hold the bitset.
1da177e4 LT	306	*/
1da177e4 LT	307	bitset_size = dm_round_up(region_count,
29121bd0	308	sizeof(*lc->clean_bits) << BYTE_SHIFT);
1da177e4 LT	309	bitset_size >>= BYTE_SHIFT;
1da177e4 LT	310
29121bd0	311	lc->bitset_uint32_count = bitset_size / sizeof(*lc->clean_bits);
b7cca195 AK	312
	313	/*
	314	* Disk log?
	315	*/
	316	if (!dev) {
	317	lc->clean_bits = vmalloc(bitset_size);
	318	if (!lc->clean_bits) {
	319	DMWARN("couldn't allocate clean bitset");
	320	kfree(lc);
	321	return -ENOMEM;
	322	}
	323	lc->disk_header = NULL;
	324	} else {
	325	lc->log_dev = dev;
01d03a66	326	lc->log_dev_failed = 0;
b7cca195 AK	327	lc->header_location.bdev = lc->log_dev->bdev;
	328	lc->header_location.sector = 0;
	329
	330	/*
	331	* Buffer holds both header and bitset.
	332	*/
	333	buf_size = dm_round_up((LOG_OFFSET << SECTOR_SHIFT) +
	334	bitset_size, ti->limits.hardsect_size);
	335	lc->header_location.count = buf_size >> SECTOR_SHIFT;
5d234d1e HM	336	lc->io_req.mem.type = DM_IO_VMA;
	337	lc->io_req.client = dm_io_client_create(dm_div_up(buf_size,
	338	PAGE_SIZE));
	339	if (IS_ERR(lc->io_req.client)) {
	340	r = PTR_ERR(lc->io_req.client);
	341	DMWARN("couldn't allocate disk io client");
	342	kfree(lc);
	343	return -ENOMEM;
	344	}
b7cca195 AK	345
	346	lc->disk_header = vmalloc(buf_size);
	347	if (!lc->disk_header) {
	348	DMWARN("couldn't allocate disk log buffer");
	349	kfree(lc);
	350	return -ENOMEM;
	351	}
	352
	353	lc->clean_bits = (void *)lc->disk_header +
	354	(LOG_OFFSET << SECTOR_SHIFT);
1da177e4	355	}
b7cca195	356
1da177e4 LT	357	memset(lc->clean_bits, -1, bitset_size);
	358
	359	lc->sync_bits = vmalloc(bitset_size);
	360	if (!lc->sync_bits) {
	361	DMWARN("couldn't allocate sync bitset");
b7cca195 AK	362	if (!dev)
	363	vfree(lc->clean_bits);
	364	vfree(lc->disk_header);
1da177e4 LT	365	kfree(lc);
	366	return -ENOMEM;
	367	}
	368	memset(lc->sync_bits, (sync == NOSYNC) ? -1 : 0, bitset_size);
	369	lc->sync_count = (sync == NOSYNC) ? region_count : 0;
	370
	371	lc->recovering_bits = vmalloc(bitset_size);
	372	if (!lc->recovering_bits) {
	373	DMWARN("couldn't allocate sync bitset");
	374	vfree(lc->sync_bits);
b7cca195 AK	375	if (!dev)
	376	vfree(lc->clean_bits);
	377	vfree(lc->disk_header);
1da177e4 LT	378	kfree(lc);
	379	return -ENOMEM;
	380	}
	381	memset(lc->recovering_bits, 0, bitset_size);
	382	lc->sync_search = 0;
	383	log->context = lc;
b7cca195	384
1da177e4 LT	385	return 0;
	386	}
	387
b7cca195 AK	388	static int core_ctr(struct dirty_log log, struct dm_target ti,
	389	unsigned int argc, char **argv)
	390	{
	391	return create_log_context(log, ti, argc, argv, NULL);
	392	}
	393
	394	static void destroy_log_context(struct log_c *lc)
1da177e4	395	{
1da177e4 LT	396	vfree(lc->sync_bits);
	397	vfree(lc->recovering_bits);
	398	kfree(lc);
	399	}
	400
b7cca195 AK	401	static void core_dtr(struct dirty_log *log)
	402	{
	403	struct log_c lc = (struct log_c ) log->context;
	404
	405	vfree(lc->clean_bits);
	406	destroy_log_context(lc);
	407	}
	408
1da177e4 LT	409	/*----------------------------------------------------------------
	410	* disk log constructor/destructor
	411	*
	412	* argv contains log_device region_size followed optionally by [no]sync
	413	--------------------------------------------------------------/
	414	static int disk_ctr(struct dirty_log log, struct dm_target ti,
	415	unsigned int argc, char **argv)
	416	{
	417	int r;
1da177e4 LT	418	struct dm_dev *dev;
	419
	420	if (argc < 2 \|\| argc > 3) {
	421	DMWARN("wrong number of arguments to disk mirror log");
	422	return -EINVAL;
	423	}
	424
	425	r = dm_get_device(ti, argv[0], 0, 0 /* FIXME */,
	426	FMODE_READ \| FMODE_WRITE, &dev);
	427	if (r)
	428	return r;
	429
b7cca195	430	r = create_log_context(log, ti, argc - 1, argv + 1, dev);
1da177e4 LT	431	if (r) {
	432	dm_put_device(ti, dev);
	433	return r;
	434	}
	435
1da177e4	436	return 0;
1da177e4 LT	437	}
	438
	439	static void disk_dtr(struct dirty_log *log)
	440	{
	441	struct log_c lc = (struct log_c ) log->context;
b7cca195	442
1da177e4 LT	443	dm_put_device(lc->ti, lc->log_dev);
1da177e4 LT	444	vfree(lc->disk_header);
5d234d1e	445	dm_io_client_destroy(lc->io_req.client);
b7cca195	446	destroy_log_context(lc);
1da177e4 LT	447	}
	448
	449	static int count_bits32(uint32_t *addr, unsigned size)
	450	{
	451	int count = 0, i;
	452
	453	for (i = 0; i < size; i++) {
	454	count += hweight32(*(addr+i));
	455	}
	456	return count;
	457	}
	458
01d03a66 JB	459	static void fail_log_device(struct log_c *lc)
	460	{
	461	if (lc->log_dev_failed)
	462	return;
	463
	464	lc->log_dev_failed = 1;
	465	dm_table_event(lc->ti->table);
	466	}
	467
1da177e4 LT	468	static int disk_resume(struct dirty_log *log)
	469	{
	470	int r;
	471	unsigned i;
	472	struct log_c lc = (struct log_c ) log->context;
	473	size_t size = lc->bitset_uint32_count * sizeof(uint32_t);
	474
	475	/* read the disk header */
	476	r = read_header(lc);
01d03a66 JB	477	if (r) {
	478	DMWARN("%s: Failed to read header on mirror log device",
	479	lc->log_dev->name);
	480	fail_log_device(lc);
ba8b45ce JB	481	/*
	482	* If the log device cannot be read, we must assume
	483	* all regions are out-of-sync. If we simply return
	484	* here, the state will be uninitialized and could
	485	* lead us to return 'in-sync' status for regions
	486	* that are actually 'out-of-sync'.
	487	*/
	488	lc->header.nr_regions = 0;
01d03a66	489	}
1da177e4	490
8a835f11	491	/* set or clear any new bits -- device has grown */
1da177e4 LT	492	if (lc->sync == NOSYNC)
	493	for (i = lc->header.nr_regions; i < lc->region_count; i++)
	494	/* FIXME: amazingly inefficient */
	495	log_set_bit(lc, lc->clean_bits, i);
	496	else
	497	for (i = lc->header.nr_regions; i < lc->region_count; i++)
	498	/* FIXME: amazingly inefficient */
	499	log_clear_bit(lc, lc->clean_bits, i);
	500
8a835f11 AK	501	/* clear any old bits -- device has shrunk */
	502	for (i = lc->region_count; i % (sizeof(*lc->clean_bits) << BYTE_SHIFT); i++)
	503	log_clear_bit(lc, lc->clean_bits, i);
	504
1da177e4 LT	505	/* copy clean across to sync */
	506	memcpy(lc->sync_bits, lc->clean_bits, size);
	507	lc->sync_count = count_bits32(lc->clean_bits, lc->bitset_uint32_count);
88b20a1a	508	lc->sync_search = 0;
1da177e4	509
1da177e4 LT	510	/* set the correct number of regions in the header */
	511	lc->header.nr_regions = lc->region_count;
	512
	513	/* write the new header */
01d03a66 JB	514	r = write_header(lc);
	515	if (r) {
	516	DMWARN("%s: Failed to write header on mirror log device",
	517	lc->log_dev->name);
	518	fail_log_device(lc);
	519	}
	520
	521	return r;
1da177e4 LT	522	}
	523
	524	static uint32_t core_get_region_size(struct dirty_log *log)
	525	{
	526	struct log_c lc = (struct log_c ) log->context;
	527	return lc->region_size;
	528	}
	529
88b20a1a JB	530	static int core_resume(struct dirty_log *log)
	531	{
	532	struct log_c lc = (struct log_c ) log->context;
	533	lc->sync_search = 0;
	534	return 0;
	535	}
	536
1da177e4 LT	537	static int core_is_clean(struct dirty_log *log, region_t region)
	538	{
	539	struct log_c lc = (struct log_c ) log->context;
	540	return log_test_bit(lc->clean_bits, region);
	541	}
	542
	543	static int core_in_sync(struct dirty_log *log, region_t region, int block)
	544	{
	545	struct log_c lc = (struct log_c ) log->context;
	546	return log_test_bit(lc->sync_bits, region);
	547	}
	548
	549	static int core_flush(struct dirty_log *log)
	550	{
	551	/* no op */
	552	return 0;
	553	}
	554
	555	static int disk_flush(struct dirty_log *log)
	556	{
	557	int r;
	558	struct log_c lc = (struct log_c ) log->context;
	559
	560	/* only write if the log has changed */
	561	if (!lc->touched)
	562	return 0;
	563
702ca6f0	564	r = write_header(lc);
01d03a66 JB	565	if (r)
	566	fail_log_device(lc);
	567	else
1da177e4 LT	568	lc->touched = 0;
	569
	570	return r;
	571	}
	572
	573	static void core_mark_region(struct dirty_log *log, region_t region)
	574	{
	575	struct log_c lc = (struct log_c ) log->context;
	576	log_clear_bit(lc, lc->clean_bits, region);
	577	}
	578
	579	static void core_clear_region(struct dirty_log *log, region_t region)
	580	{
	581	struct log_c lc = (struct log_c ) log->context;
	582	log_set_bit(lc, lc->clean_bits, region);
	583	}
	584
	585	static int core_get_resync_work(struct dirty_log log, region_t region)
	586	{
	587	struct log_c lc = (struct log_c ) log->context;
	588
	589	if (lc->sync_search >= lc->region_count)
	590	return 0;
	591
	592	do {
1113a7e9 SB	593	*region = ext2_find_next_zero_bit(
1113a7e9 SB	594	(unsigned long *) lc->sync_bits,
1da177e4 LT	595	lc->region_count,
	596	lc->sync_search);
	597	lc->sync_search = *region + 1;
	598
ac81b2ee	599	if (*region >= lc->region_count)
1da177e4 LT	600	return 0;
	601
	602	} while (log_test_bit(lc->recovering_bits, *region));
	603
	604	log_set_bit(lc, lc->recovering_bits, *region);
	605	return 1;
	606	}
	607
f3ee6b2f JB	608	static void core_set_region_sync(struct dirty_log *log, region_t region,
f3ee6b2f JB	609	int in_sync)
1da177e4 LT	610	{
	611	struct log_c lc = (struct log_c ) log->context;
	612
	613	log_clear_bit(lc, lc->recovering_bits, region);
f3ee6b2f	614	if (in_sync) {
1da177e4 LT	615	log_set_bit(lc, lc->sync_bits, region);
1da177e4 LT	616	lc->sync_count++;
f3ee6b2f JB	617	} else if (log_test_bit(lc->sync_bits, region)) {
	618	lc->sync_count--;
	619	log_clear_bit(lc, lc->sync_bits, region);
	620	}
1da177e4 LT	621	}
	622
	623	static region_t core_get_sync_count(struct dirty_log *log)
	624	{
	625	struct log_c lc = (struct log_c ) log->context;
	626
	627	return lc->sync_count;
	628	}
	629
	630	#define DMEMIT_SYNC \
	631	if (lc->sync != DEFAULTSYNC) \
	632	DMEMIT("%ssync ", lc->sync == NOSYNC ? "no" : "")
	633
	634	static int core_status(struct dirty_log *log, status_type_t status,
	635	char *result, unsigned int maxlen)
	636	{
	637	int sz = 0;
	638	struct log_c *lc = log->context;
	639
	640	switch(status) {
	641	case STATUSTYPE_INFO:
315dcc22	642	DMEMIT("1 %s", log->type->name);
1da177e4 LT	643	break;
	644
	645	case STATUSTYPE_TABLE:
	646	DMEMIT("%s %u %u ", log->type->name,
	647	lc->sync == DEFAULTSYNC ? 1 : 2, lc->region_size);
	648	DMEMIT_SYNC;
	649	}
	650
	651	return sz;
	652	}
	653
	654	static int disk_status(struct dirty_log *log, status_type_t status,
	655	char *result, unsigned int maxlen)
	656	{
	657	int sz = 0;
1da177e4 LT	658	struct log_c *lc = log->context;
	659
	660	switch(status) {
	661	case STATUSTYPE_INFO:
315dcc22 JB	662	DMEMIT("3 %s %s %c", log->type->name, lc->log_dev->name,
315dcc22 JB	663	lc->log_dev_failed ? 'D' : 'A');
1da177e4 LT	664	break;
	665
	666	case STATUSTYPE_TABLE:
1da177e4	667	DMEMIT("%s %u %s %u ", log->type->name,
315dcc22	668	lc->sync == DEFAULTSYNC ? 2 : 3, lc->log_dev->name,
1da177e4 LT	669	lc->region_size);
	670	DMEMIT_SYNC;
	671	}
	672
	673	return sz;
	674	}
	675
	676	static struct dirty_log_type _core_type = {
	677	.name = "core",
	678	.module = THIS_MODULE,
	679	.ctr = core_ctr,
	680	.dtr = core_dtr,
88b20a1a	681	.resume = core_resume,
1da177e4 LT	682	.get_region_size = core_get_region_size,
	683	.is_clean = core_is_clean,
	684	.in_sync = core_in_sync,
	685	.flush = core_flush,
	686	.mark_region = core_mark_region,
	687	.clear_region = core_clear_region,
	688	.get_resync_work = core_get_resync_work,
f3ee6b2f	689	.set_region_sync = core_set_region_sync,
1da177e4 LT	690	.get_sync_count = core_get_sync_count,
	691	.status = core_status,
	692	};
	693
	694	static struct dirty_log_type _disk_type = {
	695	.name = "disk",
	696	.module = THIS_MODULE,
	697	.ctr = disk_ctr,
	698	.dtr = disk_dtr,
6b3df0d7	699	.postsuspend = disk_flush,
1da177e4 LT	700	.resume = disk_resume,
	701	.get_region_size = core_get_region_size,
	702	.is_clean = core_is_clean,
	703	.in_sync = core_in_sync,
	704	.flush = disk_flush,
	705	.mark_region = core_mark_region,
	706	.clear_region = core_clear_region,
	707	.get_resync_work = core_get_resync_work,
f3ee6b2f	708	.set_region_sync = core_set_region_sync,
1da177e4 LT	709	.get_sync_count = core_get_sync_count,
	710	.status = disk_status,
	711	};
	712
	713	int __init dm_dirty_log_init(void)
	714	{
	715	int r;
	716
	717	r = dm_register_dirty_log_type(&_core_type);
	718	if (r)
	719	DMWARN("couldn't register core log");
	720
	721	r = dm_register_dirty_log_type(&_disk_type);
	722	if (r) {
	723	DMWARN("couldn't register disk type");
	724	dm_unregister_dirty_log_type(&_core_type);
	725	}
	726
	727	return r;
	728	}
	729
	730	void dm_dirty_log_exit(void)
	731	{
	732	dm_unregister_dirty_log_type(&_disk_type);
	733	dm_unregister_dirty_log_type(&_core_type);
	734	}
	735
	736	EXPORT_SYMBOL(dm_register_dirty_log_type);
	737	EXPORT_SYMBOL(dm_unregister_dirty_log_type);
	738	EXPORT_SYMBOL(dm_create_dirty_log);
	739	EXPORT_SYMBOL(dm_destroy_dirty_log);