path: root/drivers/md/dm-snap.c

/*
 * dm-snapshot.c
 *
 * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
 *
 * This file is released under the GPL.
 */

#include <linux/blkdev.h>
#include <linux/config.h>
#include <linux/ctype.h>
#include <linux/device-mapper.h>
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/kdev_t.h>
#include <linux/list.h>
#include <linux/mempool.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>

#include "dm-snap.h"
#include "dm-bio-list.h"
#include "kcopyd.h"

/*
 * The percentage increment we will wake up users at
 */
#define WAKE_UP_PERCENT 5

/*
 * kcopyd priority of snapshot operations
 */
#define SNAPSHOT_COPY_PRIORITY 2

/*
 * Each snapshot reserves this many pages for io
 */
#define SNAPSHOT_PAGES 256

struct pending_exception {
	struct exception e;

	/*
	 * Origin buffers waiting for this to complete are held
	 * in a bio list
	 */
	struct bio_list origin_bios;
	struct bio_list snapshot_bios;

	/*
	 * Short-term queue of pending exceptions prior to submission.
	 */
	struct list_head list;

	/*
	 * The primary pending_exception is the one that holds
	 * the sibling_count and the list of origin_bios for a
	 * group of pending_exceptions.  It is always last to get freed.
	 * These fields get set up when writing to the origin.
	 */
	struct pending_exception *primary_pe;

	/*
	 * Number of pending_exceptions processing this chunk.
	 * When this drops to zero we must complete the origin bios.
	 * If incrementing or decrementing this, hold pe->snap->lock for
	 * the sibling concerned and not pe->primary_pe->snap->lock unless
	 * they are the same.
	 */
	atomic_t sibling_count;

	/* Pointer back to snapshot context */
	struct dm_snapshot *snap;

	/*
	 * 1 indicates the exception has already been sent to
	 * kcopyd.
	 */
	int started;
};

/*
 * Hash table mapping origin volumes to lists of snapshots and
 * a lock to protect it
 */
static kmem_cache_t *exception_cache;
static kmem_cache_t *pending_cache;
static mempool_t *pending_pool;

/*
 * One of these per registered origin, held in the snapshot_origins hash
 */
struct origin {
	/* The origin device */
	struct block_device *bdev;

	struct list_head hash_list;

	/* List of snapshots for this origin */
	struct list_head snapshots;
};

/*
 * Size of the hash table for origin volumes. If we make this
 * the size of the minors list then it should be nearly perfect
 */
#define ORIGIN_HASH_SIZE 256
#define ORIGIN_MASK      0xFF
static struct list_head *_origins;
static struct rw_semaphore _origins_lock;

static int init_origin_hash(void)
{
	int i;

	_origins = kmalloc(ORIGIN_HASH_SIZE * sizeof(struct list_head),
			   GFP_KERNEL);
	if (!_origins) {
		DMERR("Device mapper: Snapshot: unable to allocate memory");
		return -ENOMEM;
	}

	for (i = 0; i < ORIGIN_HASH_SIZE; i++)
		INIT_LIST_HEAD(_origins + i);
	init_rwsem(&_origins_lock);

	return 0;
}

static void exit_origin_hash(void)
{
	kfree(_origins);
}

static inline unsigned int origin_hash(struct block_device *bdev)
{
	return bdev->bd_dev & ORIGIN_MASK;
}

static struct origin *__lookup_origin(struct block_device *origin)
{
	struct list_head *ol;
	struct origin *o;

	ol = &_origins[origin_hash(origin)];
	list_for_each_entry (o, ol, hash_list)
		if (bdev_equal(o->bdev, origin))
			return o;

	return NULL;
}

static void __insert_origin(struct origin *o)
{
	struct list_head *sl = &_origins[origin_hash(o->bdev)];
	list_add_tail(&o->hash_list, sl);
}

/*
 * Make a note of the snapshot and its origin so we can look it
 * up when the origin has a write on it.
 */
static int register_snapshot(struct dm_snapshot *snap)
{
	struct origin *o;
	struct block_device *bdev = snap->origin->bdev;

	down_write(&_origins_lock);
	o = __lookup_origin(bdev);

	if (!o) {
		/* New origin */
		o = kmalloc(sizeof(*o), GFP_KERNEL);
		if (!o) {
			up_write(&_origins_lock);
			return -ENOMEM;
		}

		/* Initialise the struct */
		INIT_LIST_HEAD(&o->snapshots);
		o->bdev = bdev;

		__insert_origin(o);
	}

	list_add_tail(&snap->list, &o->snapshots);

	up_write(&_origins_lock);
	return 0;
}

static void unregister_snapshot(struct dm_snapshot *s)
{
	struct origin *o;

	down_write(&_origins_lock);
	o = __lookup_origin(s->origin->bdev);

	list_del(&s->list);
	if (list_empty(&o->snapshots)) {
		list_del(&o->hash_list);
		kfree(o);
	}

	up_write(&_origins_lock);
}

/*
 * Implementation of the exception hash tables.
 */
static int init_exception_table(struct exception_table *et, uint32_t size)
{
	unsigned int i;

	et->hash_mask = size - 1;
	et->table = dm_vcalloc(size, sizeof(struct list_head));
	if (!et->table)
		return -ENOMEM;

	for