/*

   * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
   *
   * This file is released under the GPL.
   */
  
  #include <linux/blkdev.h>

  #include <linux/device-mapper.h>

  #include <linux/delay.h>

  #include <linux/fs.h>
  #include <linux/init.h>
  #include <linux/kdev_t.h>
  #include <linux/list.h>

  #include <linux/list_bl.h>

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

  #include <linux/log2.h>

  #include <linux/dm-kcopyd.h>

  #include "dm.h"

  #include "dm-exception-store.h"

  #define DM_MSG_PREFIX "snapshots"

  static const char dm_snapshot_merge_target_name[] = "snapshot-merge";
  
  #define dm_target_is_snapshot_merge(ti) \
  	((ti)->type->name == dm_snapshot_merge_target_name)

  /*

   * The size of the mempool used to track chunks in use.
   */
  #define MIN_IOS 256

  #define DM_TRACKED_CHUNK_HASH_SIZE	16
  #define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
  					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))

  struct dm_exception_table {

  	uint32_t hash_mask;
  	unsigned hash_shift;

  	struct hlist_bl_head *table;

  };
  
  struct dm_snapshot {

  	struct rw_semaphore lock;

  
  	struct dm_dev *origin;

  	struct dm_dev *cow;
  
  	struct dm_target *ti;

  
  	/* List of snapshots per Origin */
  	struct list_head list;

  	/*
  	 * You can't use a snapshot if this is 0 (e.g. if full).
  	 * A snapshot-merge target never clears this.
  	 */

  	int valid;

  	/*
  	 * The snapshot overflowed because of a write to the snapshot device.
  	 * We don't have to invalidate the snapshot in this case, but we need
  	 * to prevent further writes.
  	 */
  	int snapshot_overflowed;

  	/* Origin writes don't trigger exceptions until this is set */
  	int active;

  	atomic_t pending_exceptions_count;

  	spinlock_t pe_allocation_lock;
  
  	/* Protected by "pe_allocation_lock" */

  	sector_t exception_start_sequence;
  
  	/* Protected by kcopyd single-threaded callback */
  	sector_t exception_complete_sequence;
  
  	/*
  	 * A list of pending exceptions that completed out of order.
  	 * Protected by kcopyd single-threaded callback.
  	 */

  	struct rb_root out_of_order_tree;

  	mempool_t pending_pool;

  	struct dm_exception_table pending;
  	struct dm_exception_table complete;

  
  	/*
  	 * pe_lock protects all pending_exception operations and access
  	 * as well as the snapshot_bios list.
  	 */
  	spinlock_t pe_lock;

  	/* Chunks with outstanding reads */
  	spinlock_t tracked_chunk_lock;

  	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];

  	/* The on disk metadata handler */
  	struct dm_exception_store *store;

  	unsigned in_progress;
  	struct wait_queue_head in_progress_wait;

  	struct dm_kcopyd_client *kcopyd_client;

  	/* Wait for events based on state_bits */
  	unsigned long state_bits;
  
  	/* Range of chunks currently being merged. */
  	chunk_t first_merging_chunk;
  	int num_merging_chunks;

  	/*
  	 * The merge operation failed if this flag is set.
  	 * Failure modes are handled as follows:
  	 * - I/O error reading the header
  	 *   	=> don't load the target; abort.
  	 * - Header does not have "valid" flag set
  	 *   	=> use the origin; forget about the snapshot.
  	 * - I/O error when reading exceptions
  	 *   	=> don't load the target; abort.
  	 *         (We can't use the intermediate origin state.)
  	 * - I/O error while merging
  	 *	=> stop merging; set merge_failed; process I/O normally.
  	 */

  	bool merge_failed:1;
  
  	bool discard_zeroes_cow:1;
  	bool discard_passdown_origin:1;

  	/*
  	 * Incoming bios that overlap with chunks being merged must wait
  	 * for them to be committed.
  	 */
  	struct bio_list bios_queued_during_merge;

  
  	/*
  	 * Flush data after merge.
  	 */
  	struct bio flush_bio;

  };

  /*
   * state_bits:
   *   RUNNING_MERGE  - Merge operation is in progress.
   *   SHUTDOWN_MERGE - Set to signal that merge needs to be stopped;
   *                    cleared afterwards.
   */
  #define RUNNING_MERGE          0
  #define SHUTDOWN_MERGE         1

  /*
   * Maximum number of chunks being copied on write.
   *
   * The value was decided experimentally as a trade-off between memory
   * consumption, stalling the kernel's workqueues and maintaining a high enough
   * throughput.
   */
  #define DEFAULT_COW_THRESHOLD 2048

  static unsigned cow_threshold = DEFAULT_COW_THRESHOLD;
  module_param_named(snapshot_cow_threshold, cow_threshold, uint, 0644);

  MODULE_PARM_DESC(snapshot_cow_threshold, "Maximum number of chunks being copied on write");

  DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(snapshot_copy_throttle,
  		"A percentage of time allocated for copy on write");

  struct dm_dev *dm_snap_origin(struct dm_snapshot *s)
  {
  	return s->origin;
  }
  EXPORT_SYMBOL(dm_snap_origin);

  struct dm_dev *dm_snap_cow(struct dm_snapshot *s)
  {
  	return s->cow;
  }
  EXPORT_SYMBOL(dm_snap_cow);

  static sector_t chunk_to_sector(struct dm_exception_store *store,
  				chunk_t chunk)
  {
  	return chunk << store->chunk_shift;
  }
  
  static int bdev_equal(struct block_device *lhs, struct block_device *rhs)
  {
  	/*
  	 * There is only ever one instance of a particular block
  	 * device so we can compare pointers safely.
  	 */
  	return lhs == rhs;
  }

  struct dm_snap_pending_exception {

  	struct dm_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;

  	/* Pointer back to snapshot context */
  	struct dm_snapshot *snap;
  
  	/*
  	 * 1 indicates the exception has already been sent to
  	 * kcopyd.
  	 */
  	int started;

  	/* There was copying error. */
  	int copy_error;
  
  	/* A sequence number, it is used for in-order completion. */
  	sector_t exception_sequence;

  	struct rb_node out_of_order_node;

  	/*
  	 * For writing a complete chunk, bypassing the copy.
  	 */
  	struct bio *full_bio;
  	bio_end_io_t *full_bio_end_io;

  };
  
  /*
   * Hash table mapping origin volumes to lists of snapshots and
   * a lock to protect it
   */

  static struct kmem_cache *exception_cache;
  static struct kmem_cache *pending_cache;

  struct dm_snap_tracked_chunk {
  	struct hlist_node node;
  	chunk_t chunk;
  };

  static void init_tracked_chunk(struct bio *bio)
  {
  	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
  	INIT_HLIST_NODE(&c->node);
  }
  
  static bool is_bio_tracked(struct bio *bio)
  {
  	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
  	return !hlist_unhashed(&c->node);
  }
  
  static void track_chunk(struct dm_snapshot *s, struct bio *bio, chunk_t chunk)

  {

  	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));

  
  	c->chunk = chunk;

  	spin_lock_irq(&s->tracked_chunk_lock);

  	hlist_add_head(&c->node,
  		       &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)]);

  	spin_unlock_irq(&s->tracked_chunk_lock);

  }

  static void stop_tracking_chunk(struct dm_snapshot *s, struct bio *bio)

  {

  	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));

  	unsigned long flags;
  
  	spin_lock_irqsave(&s->tracked_chunk_lock, flags);
  	hlist_del(&c->node);
  	spin_unlock_irqrestore(&s->tracked_chunk_lock, flags);

  }

  static int __chunk_is_tracked(struct dm_snapshot *s, chunk_t chunk)
  {
  	struct dm_snap_tracked_chunk *c;

  	int found = 0;
  
  	spin_lock_irq(&s->tracked_chunk_lock);

  	hlist_for_each_entry(c,

  	    &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)], node) {
  		if (c->chunk == chunk) {
  			found = 1;
  			break;
  		}
  	}
  
  	spin_unlock_irq(&s->tracked_chunk_lock);
  
  	return found;
  }

  /*

   * This conflicting I/O is extremely improbable in the caller,
   * so msleep(1) is sufficient and there is no need for a wait queue.
   */
  static void __check_for_conflicting_io(struct dm_snapshot *s, chunk_t chunk)
  {
  	while (__chunk_is_tracked(s, chunk))
  		msleep(1);
  }
  
  /*

   * 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;
  };
  
  /*

   * This structure is allocated for each origin target
   */
  struct dm_origin {
  	struct dm_dev *dev;
  	struct dm_target *ti;
  	unsigned split_boundary;
  	struct list_head hash_list;
  };
  
  /*

   * 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 list_head *_dm_origins;

  static struct rw_semaphore _origins_lock;

  static DECLARE_WAIT_QUEUE_HEAD(_pending_exceptions_done);
  static DEFINE_SPINLOCK(_pending_exceptions_done_spinlock);
  static uint64_t _pending_exceptions_done_count;

  static int init_origin_hash(void)
  {
  	int i;

  	_origins = kmalloc_array(ORIGIN_HASH_SIZE, sizeof(struct list_head),
  				 GFP_KERNEL);

  	if (!_origins) {

  		DMERR("unable to allocate memory for _origins");

  		return -ENOMEM;
  	}

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

  	_dm_origins = kmalloc_array(ORIGIN_HASH_SIZE,
  				    sizeof(struct list_head),
  				    GFP_KERNEL);

  	if (!_dm_origins) {
  		DMERR("unable to allocate memory for _dm_origins");
  		kfree(_origins);
  		return -ENOMEM;
  	}
  	for (i = 0; i < ORIGIN_HASH_SIZE; i++)
  		INIT_LIST_HEAD(_dm_origins + i);

  	init_rwsem(&_origins_lock);
  
  	return 0;
  }
  
  static void exit_origin_hash(void)
  {
  	kfree(_origins);

  	kfree(_dm_origins);

  }

  static unsigned 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);
  }

  static struct dm_origin *__lookup_dm_origin(struct block_device *origin)
  {
  	struct list_head *ol;
  	struct dm_origin *o;
  
  	ol = &_dm_origins[origin_hash(origin)];
  	list_for_each_entry (o, ol, hash_list)
  		if (bdev_equal(o->dev->bdev, origin))
  			return o;
  
  	return NULL;
  }
  
  static void __insert_dm_origin(struct dm_origin *o)
  {
  	struct list_head *sl = &_dm_origins[origin_hash(o->dev->bdev)];
  	list_add_tail(&o->hash_list, sl);
  }
  
  static void __remove_dm_origin(struct dm_origin *o)
  {
  	list_del(&o->hash_list);
  }

  /*

   * _origins_lock must be held when calling this function.
   * Returns number of snapshots registered using the supplied cow device, plus:
   * snap_src - a snapshot suitable for use as a source of exception handover
   * snap_dest - a snapshot capable of receiving exception handover.

   * snap_merge - an existing snapshot-merge target linked to the same origin.
   *   There can be at most one snapshot-merge target. The parameter is optional.

   *

   * Possible return values and states of snap_src and snap_dest.

   *   0: NULL, NULL  - first new snapshot
   *   1: snap_src, NULL - normal snapshot
   *   2: snap_src, snap_dest  - waiting for handover
   *   2: snap_src, NULL - handed over, waiting for old to be deleted
   *   1: NULL, snap_dest - source got destroyed without handover
   */
  static int __find_snapshots_sharing_cow(struct dm_snapshot *snap,
  					struct dm_snapshot **snap_src,

  					struct dm_snapshot **snap_dest,
  					struct dm_snapshot **snap_merge)

  {
  	struct dm_snapshot *s;
  	struct origin *o;
  	int count = 0;
  	int active;
  
  	o = __lookup_origin(snap->origin->bdev);
  	if (!o)
  		goto out;
  
  	list_for_each_entry(s, &o->snapshots, list) {

  		if (dm_target_is_snapshot_merge(s->ti) && snap_merge)
  			*snap_merge = s;

  		if (!bdev_equal(s->cow->bdev, snap->cow->bdev))
  			continue;

  		down_read(&s->lock);

  		active = s->active;

  		up_read(&s->lock);

  
  		if (active) {
  			if (snap_src)
  				*snap_src = s;
  		} else if (snap_dest)
  			*snap_dest = s;
  
  		count++;
  	}
  
  out:
  	return count;
  }
  
  /*
   * On success, returns 1 if this snapshot is a handover destination,
   * otherwise returns 0.
   */
  static int __validate_exception_handover(struct dm_snapshot *snap)
  {
  	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;

  	struct dm_snapshot *snap_merge = NULL;

  
  	/* Does snapshot need exceptions handed over to it? */

  	if ((__find_snapshots_sharing_cow(snap, &snap_src, &snap_dest,
  					  &snap_merge) == 2) ||

  	    snap_dest) {
  		snap->ti->error = "Snapshot cow pairing for exception "
  				  "table handover failed";
  		return -EINVAL;
  	}
  
  	/*
  	 * If no snap_src was found, snap cannot become a handover
  	 * destination.
  	 */
  	if (!snap_src)
  		return 0;

  	/*
  	 * Non-snapshot-merge handover?
  	 */
  	if (!dm_target_is_snapshot_merge(snap->ti))
  		return 1;
  
  	/*
  	 * Do not allow more than one merging snapshot.
  	 */
  	if (snap_merge) {
  		snap->ti->error = "A snapshot is already merging.";
  		return -EINVAL;
  	}

  	if (!snap_src->store->type->prepare_merge ||
  	    !snap_src->store->type->commit_merge) {
  		snap->ti->error = "Snapshot exception store does not "
  				  "support snapshot-merge.";
  		return -EINVAL;
  	}

  	return 1;
  }
  
  static void __insert_snapshot(struct origin *o, struct dm_snapshot *s)
  {
  	struct dm_snapshot *l;
  
  	/* Sort the list according to chunk size, largest-first smallest-last */
  	list_for_each_entry(l, &o->snapshots, list)
  		if (l->store->chunk_size < s->store->chunk_size)
  			break;
  	list_add_tail(&s->list, &l->list);
  }
  
  /*

   * Make a note of the snapshot and its origin so we can look it
   * up when the origin has a write on it.

   *
   * Also validate snapshot exception store handovers.
   * On success, returns 1 if this registration is a handover destination,
   * otherwise returns 0.

   */
  static int register_snapshot(struct dm_snapshot *snap)
  {

  	struct origin *o, *new_o = NULL;

  	struct block_device *bdev = snap->origin->bdev;

  	int r = 0;

  	new_o = kmalloc(sizeof(*new_o), GFP_KERNEL);
  	if (!new_o)
  		return -ENOMEM;

  	down_write(&_origins_lock);

  	r = __validate_exception_handover(snap);
  	if (r < 0) {
  		kfree(new_o);
  		goto out;
  	}
  
  	o = __lookup_origin(bdev);

  	if (o)
  		kfree(new_o);
  	else {

  		/* New origin */

  		o = new_o;

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

  	__insert_snapshot(o, snap);
  
  out:
  	up_write(&_origins_lock);
  
  	return r;
  }
  
  /*
   * Move snapshot to correct place in list according to chunk size.
   */
  static void reregister_snapshot(struct dm_snapshot *s)
  {
  	struct block_device *bdev = s->origin->bdev;
  
  	down_write(&_origins_lock);
  
  	list_del(&s->list);
  	__insert_snapshot(__lookup_origin(bdev), s);

  
  	up_write(&_origins_lock);

  }
  
  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 (o && list_empty(&o->snapshots)) {

  		list_del(&o->hash_list);
  		kfree(o);
  	}
  
  	up_write(&_origins_lock);
  }
  
  /*
   * Implementation of the exception hash tables.

   * The lowest hash_shift bits of the chunk number are ignored, allowing
   * some consecutive chunks to be grouped together.

   */

  static uint32_t exception_hash(struct dm_exception_table *et, chunk_t chunk);
  
  /* Lock to protect access to the completed and pending exception hash tables. */
  struct dm_exception_table_lock {
  	struct hlist_bl_head *complete_slot;
  	struct hlist_bl_head *pending_slot;
  };
  
  static void dm_exception_table_lock_init(struct dm_snapshot *s, chunk_t chunk,
  					 struct dm_exception_table_lock *lock)
  {
  	struct dm_exception_table *complete = &s->complete;
  	struct dm_exception_table *pending = &s->pending;
  
  	lock->complete_slot = &complete->table[exception_hash(complete, chunk)];
  	lock->pending_slot = &pending->table[exception_hash(pending, chunk)];
  }
  
  static void dm_exception_table_lock(struct dm_exception_table_lock *lock)
  {
  	hlist_bl_lock(lock->complete_slot);
  	hlist_bl_lock(lock->pending_slot);
  }
  
  static void dm_exception_table_unlock(struct dm_exception_table_lock *lock)
  {
  	hlist_bl_unlock(lock->pending_slot);
  	hlist_bl_unlock(lock->complete_slot);
  }

  static int dm_exception_table_init(struct dm_exception_table *et,
  				   uint32_t size, unsigned hash_shift)

  {
  	unsigned int i;

  	et->hash_shift = hash_shift;

  	et->hash_mask = size - 1;

  	et->table = dm_vcalloc(size, sizeof(struct hlist_bl_head));

  	if (!et->table)
  		return -ENOMEM;
  
  	for (i = 0; i < size; i++)

  		INIT_HLIST_BL_HEAD(et->table + i);

  
  	return 0;
  }

  static void dm_exception_table_exit(struct dm_exception_table *et,
  				    struct kmem_cache *mem)

  {

  	struct hlist_bl_head *slot;
  	struct dm_exception *ex;
  	struct hlist_bl_node *pos, *n;

  	int i, size;
  
  	size = et->hash_mask + 1;
  	for (i = 0; i < size; i++) {
  		slot = et->table + i;

  		hlist_bl_for_each_entry_safe(ex, pos, n, slot, hash_list)

  			kmem_cache_free(mem, ex);
  	}
  
  	vfree(et->table);
  }

  static uint32_t exception_hash(struct dm_exception_table *et, chunk_t chunk)

  {

  	return (chunk >> et->hash_shift) & et->hash_mask;

  }

  static void dm_remove_exception(struct dm_exception *e)

  {

  	hlist_bl_del(&e->hash_list);

  }
  
  /*
   * Return the exception data for a sector, or NULL if not
   * remapped.
   */

  static struct dm_exception *dm_lookup_exception(struct dm_exception_table *et,
  						chunk_t chunk)

  {

  	struct hlist_bl_head *slot;
  	struct hlist_bl_node *pos;

  	struct dm_exception *e;

  
  	slot = &et->table[exception_hash(et, chunk)];

  	hlist_bl_for_each_entry(e, pos, slot, hash_list)

  		if (chunk >= e->old_chunk &&
  		    chunk <= e->old_chunk + dm_consecutive_chunk_count(e))

  			return e;
  
  	return NULL;
  }

  static struct dm_exception *alloc_completed_exception(gfp_t gfp)

  {

  	struct dm_exception *e;

  	e = kmem_cache_alloc(exception_cache, gfp);
  	if (!e && gfp == GFP_NOIO)

  		e = kmem_cache_alloc(exception_cache, GFP_ATOMIC);
  
  	return e;
  }

  static void free_completed_exception(struct dm_exception *e)

  {
  	kmem_cache_free(exception_cache, e);
  }

  static struct dm_snap_pending_exception *alloc_pending_exception(struct dm_snapshot *s)

  {

  	struct dm_snap_pending_exception *pe = mempool_alloc(&s->pending_pool,

  							     GFP_NOIO);

  	atomic_inc(&s->pending_exceptions_count);

  	pe->snap = s;
  
  	return pe;

  }

  static void free_pending_exception(struct dm_snap_pending_exception *pe)

  {

  	struct dm_snapshot *s = pe->snap;

  	mempool_free(pe, &s->pending_pool);

  	smp_mb__before_atomic();

  	atomic_dec(&s->pending_exceptions_count);

  }

  static void dm_insert_exception(struct dm_exception_table *eh,
  				struct dm_exception *new_e)

  {

  	struct hlist_bl_head *l;
  	struct hlist_bl_node *pos;

  	struct dm_exception *e = NULL;

  
  	l = &eh->table[exception_hash(eh, new_e->old_chunk)];
  
  	/* Add immediately if this table doesn't support consecutive chunks */
  	if (!eh->hash_shift)
  		goto out;
  
  	/* List is ordered by old_chunk */

  	hlist_bl_for_each_entry(e, pos, l, hash_list) {

  		/* Insert after an existing chunk? */
  		if (new_e->old_chunk == (e->old_chunk +
  					 dm_consecutive_chunk_count(e) + 1) &&
  		    new_e->new_chunk == (dm_chunk_number(e->new_chunk) +
  					 dm_consecutive_chunk_count(e) + 1)) {
  			dm_consecutive_chunk_count_inc(e);

  			free_completed_exception(new_e);

  			return;
  		}
  
  		/* Insert before an existing chunk? */
  		if (new_e->old_chunk == (e->old_chunk - 1) &&
  		    new_e->new_chunk == (dm_chunk_number(e->new_chunk) - 1)) {
  			dm_consecutive_chunk_count_inc(e);
  			e->old_chunk--;
  			e->new_chunk--;

  			free_completed_exception(new_e);

  			return;
  		}

  		if (new_e->old_chunk < e->old_chunk)

  			break;
  	}
  
  out:

  	if (!e) {
  		/*
  		 * Either the table doesn't support consecutive chunks or slot
  		 * l is empty.
  		 */
  		hlist_bl_add_head(&new_e->hash_list, l);
  	} else if (new_e->old_chunk < e->old_chunk) {
  		/* Add before an existing exception */
  		hlist_bl_add_before(&new_e->hash_list, &e->hash_list);
  	} else {
  		/* Add to l's tail: e is the last exception in this slot */
  		hlist_bl_add_behind(&new_e->hash_list, &e->hash_list);
  	}

  }

  /*
   * Callback used by the exception stores to load exceptions when
   * initialising.
   */
  static int dm_add_exception(void *context, chunk_t old, chunk_t new)

  {

  	struct dm_exception_table_lock lock;

  	struct dm_snapshot *s = context;

  	struct dm_exception *e;

  	e = alloc_completed_exception(GFP_KERNEL);

  	if (!e)
  		return -ENOMEM;
  
  	e->old_chunk = old;

  
  	/* Consecutive_count is implicitly initialised to zero */

  	e->new_chunk = new;

  	/*
  	 * Although there is no need to lock access to the exception tables
  	 * here, if we don't then hlist_bl_add_head(), called by
  	 * dm_insert_exception(), will complain about accessing the
  	 * corresponding list without locking it first.
  	 */
  	dm_exception_table_lock_init(s, old, &lock);
  
  	dm_exception_table_lock(&lock);

  	dm_insert_exception(&s->complete, e);

  	dm_exception_table_unlock(&lock);

  	return 0;
  }

  /*
   * Return a minimum chunk size of all snapshots that have the specified origin.
   * Return zero if the origin has no snapshots.
   */

  static uint32_t __minimum_chunk_size(struct origin *o)

  {
  	struct dm_snapshot *snap;
  	unsigned chunk_size = 0;
  
  	if (o)
  		list_for_each_entry(snap, &o->snapshots, list)
  			chunk_size = min_not_zero(chunk_size,
  						  snap->store->chunk_size);

  	return (uint32_t) chunk_size;

  }

  /*
   * Hard coded magic.
   */
  static int calc_max_buckets(void)
  {
  	/* use a fixed size of 2MB */
  	unsigned long mem = 2 * 1024 * 1024;

  	mem /= sizeof(struct hlist_bl_head);

  
  	return mem;
  }
  
  /*

   * Allocate room for a suitable hash table.
   */

  static int init_hash_tables(struct dm_snapshot *s)

  {

  	sector_t hash_size, cow_dev_size, max_buckets;

  
  	/*
  	 * Calculate based on the size of the original volume or
  	 * the COW volume...
  	 */

  	cow_dev_size = get_dev_size(s->cow->bdev);

  	max_buckets = calc_max_buckets();

  	hash_size = cow_dev_size >> s->store->chunk_shift;

  	hash_size = min(hash_size, max_buckets);

  	if (hash_size < 64)
  		hash_size = 64;

  	hash_size = rounddown_pow_of_two(hash_size);

  	if (dm_exception_table_init(&s->complete, hash_size,
  				    DM_CHUNK_CONSECUTIVE_BITS))

  		return -ENOMEM;
  
  	/*
  	 * Allocate hash table for in-flight exceptions
  	 * Make this smaller than the real hash table
  	 */
  	hash_size >>= 3;
  	if (hash_size < 64)
  		hash_size = 64;

  	if (dm_exception_table_init(&s->pending, hash_size, 0)) {
  		dm_exception_table_exit(&s->complete, exception_cache);

  		return -ENOMEM;
  	}
  
  	return 0;
  }

  static void merge_shutdown(struct dm_snapshot *s)
  {
  	clear_bit_unlock(RUNNING_MERGE, &s->state_bits);

  	smp_mb__after_atomic();

  	wake_up_bit(&s->state_bits, RUNNING_MERGE);
  }

  static struct bio *__release_queued_bios_after_merge(struct dm_snapshot *s)
  {
  	s->first_merging_chunk = 0;
  	s->num_merging_chunks = 0;
  
  	return bio_list_get(&s->bios_queued_during_merge);
  }

  /*
   * Remove one chunk from the index of completed exceptions.
   */
  static int __remove_single_exception_chunk(struct dm_snapshot *s,
  					   chunk_t old_chunk)
  {
  	struct dm_exception *e;

  	e = dm_lookup_exception(&s->complete, old_chunk);
  	if (!e) {
  		DMERR("Corruption detected: exception for block %llu is "
  		      "on disk but not in memory",
  		      (unsigned long long)old_chunk);
  		return -EINVAL;
  	}
  
  	/*
  	 * If this is the only chunk using this exception, remove exception.
  	 */
  	if (!dm_consecutive_chunk_count(e)) {
  		dm_remove_exception(e);
  		free_completed_exception(e);
  		return 0;
  	}
  
  	/*
  	 * The chunk may be either at the beginning or the end of a
  	 * group of consecutive chunks - never in the middle.  We are
  	 * removing chunks in the opposite order to that in which they
  	 * were added, so this should always be true.
  	 * Decrement the consecutive chunk counter and adjust the
  	 * starting point if necessary.
  	 */
  	if (old_chunk == e->old_chunk) {
  		e->old_chunk++;
  		e->new_chunk++;
  	} else if (old_chunk != e->old_chunk +
  		   dm_consecutive_chunk_count(e)) {
  		DMERR("Attempt to merge block %llu from the "
  		      "middle of a chunk range [%llu - %llu]",
  		      (unsigned long long)old_chunk,
  		      (unsigned long long)e->old_chunk,
  		      (unsigned long long)
  		      e->old_chunk + dm_consecutive_chunk_count(e));
  		return -EINVAL;
  	}
  
  	dm_consecutive_chunk_count_dec(e);
  
  	return 0;
  }

  static void flush_bios(struct bio *bio);
  
  static int remove_single_exception_chunk(struct dm_snapshot *s)

  {

  	struct bio *b = NULL;
  	int r;
  	chunk_t old_chunk = s->first_merging_chunk + s->num_merging_chunks - 1;

  	down_write(&s->lock);

  
  	/*
  	 * Process chunks (and associated exceptions) in reverse order
  	 * so that dm_consecutive_chunk_count_dec() accounting works.
  	 */
  	do {
  		r = __remove_single_exception_chunk(s, old_chunk);
  		if (r)
  			goto out;
  	} while (old_chunk-- > s->first_merging_chunk);
  
  	b = __release_queued_bios_after_merge(s);
  
  out:

  	up_write(&s->lock);

  	if (b)
  		flush_bios(b);

  
  	return r;
  }

  static int origin_write_extent(struct dm_snapshot *merging_snap,
  			       sector_t sector, unsigned chunk_size);

  static void merge_callback(int read_err, unsigned long write_err,
  			   void *context);

  static uint64_t read_pending_exceptions_done_count(void)
  {
  	uint64_t pending_exceptions_done;
  
  	spin_lock(&_pending_exceptions_done_spinlock);
  	pending_exceptions_done = _pending_exceptions_done_count;
  	spin_unlock(&_pending_exceptions_done_spinlock);
  
  	return pending_exceptions_done;
  }
  
  static void increment_pending_exceptions_done_count(void)
  {
  	spin_lock(&_pending_exceptions_done_spinlock);
  	_pending_exceptions_done_count++;
  	spin_unlock(&_pending_exceptions_done_spinlock);
  
  	wake_up_all(&_pending_exceptions_done);
  }

  static void snapshot_merge_next_chunks(struct dm_snapshot *s)
  {

  	int i, linear_chunks;

  	chunk_t old_chunk, new_chunk;
  	struct dm_io_region src, dest;

  	sector_t io_size;

  	uint64_t previous_count;

  
  	BUG_ON(!test_bit(RUNNING_MERGE, &s->state_bits));
  	if (unlikely(test_bit(SHUTDOWN_MERGE, &s->state_bits)))
  		goto shut;
  
  	/*
  	 * valid flag never changes during merge, so no lock required.
  	 */
  	if (!s->valid) {
  		DMERR("Snapshot is invalid: can't merge");
  		goto shut;
  	}

  	linear_chunks = s->store->type->prepare_merge(s->store, &old_chunk,
  						      &new_chunk);
  	if (linear_chunks <= 0) {

  		if (linear_chunks < 0) {

  			DMERR("Read error in exception store: "
  			      "shutting down merge");

  			down_write(&s->lock);

  			s->merge_failed = true;

  			up_write(&s->lock);

  		}

  		goto shut;
  	}

  	/* Adjust old_chunk and new_chunk to reflect start of linear region */
  	old_chunk = old_chunk + 1 - linear_chunks;
  	new_chunk = new_chunk + 1 - linear_chunks;
  
  	/*
  	 * Use one (potentially large) I/O to copy all 'linear_chunks'
  	 * from the exception store to the origin
  	 */
  	io_size = linear_chunks * s->store->chunk_size;

  	dest.bdev = s->origin->bdev;
  	dest.sector = chunk_to_sector(s->store, old_chunk);

  	dest.count = min(io_size, get_dev_size(dest.bdev) - dest.sector);

  
  	src.bdev = s->cow->bdev;
  	src.sector = chunk_to_sector(s->store, new_chunk);
  	src.count = dest.count;

  	/*
  	 * Reallocate any exceptions needed in other snapshots then
  	 * wait for the pending exceptions to complete.
  	 * Each time any pending exception (globally on the system)
  	 * completes we are woken and repeat the process to find out
  	 * if we can proceed.  While this may not seem a particularly
  	 * efficient algorithm, it is not expected to have any
  	 * significant impact on performance.
  	 */
  	previous_count = read_pending_exceptions_done_count();

  	while (origin_write_extent(s, dest.sector, io_size)) {

  		wait_event(_pending_exceptions_done,
  			   (read_pending_exceptions_done_count() !=
  			    previous_count));
  		/* Retry after the wait, until all exceptions are done. */
  		previous_count = read_pending_exceptions_done_count();
  	}

  	down_write(&s->lock);

  	s->first_merging_chunk = old_chunk;

  	s->num_merging_chunks = linear_chunks;

  	up_write(&s->lock);

  	/* Wait until writes to all 'linear_chunks' drain */
  	for (i = 0; i < linear_chunks; i++)
  		__check_for_conflicting_io(s, old_chunk + i);

  	dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, merge_callback, s);
  	return;
  
  shut:
  	merge_shutdown(s);
  }

  static void error_bios(struct bio *bio);

  static int flush_data(struct dm_snapshot *s)
  {
  	struct bio *flush_bio = &s->flush_bio;
  
  	bio_reset(flush_bio);
  	bio_set_dev(flush_bio, s->origin->bdev);
  	flush_bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
  
  	return submit_bio_wait(flush_bio);
  }

  static void merge_callback(int read_err, unsigned long write_err, void *context)
  {
  	struct dm_snapshot *s = context;

  	struct bio *b = NULL;

  
  	if (read_err || write_err) {
  		if (read_err)
  			DMERR("Read error: shutting down merge.");
  		else
  			DMERR("Write error: shutting down merge.");
  		goto shut;
  	}

  	if (flush_data(s) < 0) {
  		DMERR("Flush after merge failed: shutting down merge");
  		goto shut;
  	}

  	if (s->store->type->commit_merge(s->store,
  					 s->num_merging_chunks) < 0) {

  		DMERR("Write error in exception store: shutting down merge");
  		goto shut;
  	}

  	if (remove_single_exception_chunk(s) < 0)
  		goto shut;

  	snapshot_merge_next_chunks(s);
  
  	return;
  
  shut:

  	down_write(&s->lock);

  	s->merge_failed = true;

  	b = __release_queued_bios_after_merge(s);

  	up_write(&s->lock);

  	error_bios(b);

  	merge_shutdown(s);
  }
  
  static void start_merge(struct dm_snapshot *s)
  {
  	if (!test_and_set_bit(RUNNING_MERGE, &s->state_bits))
  		snapshot_merge_next_chunks(s);
  }

  /*
   * Stop the merging process and wait until it finishes.
   */
  static void stop_merge(struct dm_snapshot *s)
  {
  	set_bit(SHUTDOWN_MERGE, &s->state_bits);

  	wait_on_bit(&s->state_bits, RUNNING_MERGE, TASK_UNINTERRUPTIBLE);

  	clear_bit(SHUTDOWN_MERGE, &s->state_bits);
  }

  static int parse_snapshot_features(struct dm_arg_set *as, struct dm_snapshot *s,
  				   struct dm_target *ti)
  {
  	int r;
  	unsigned argc;
  	const char *arg_name;
  
  	static const struct dm_arg _args[] = {
  		{0, 2, "Invalid number of feature arguments"},
  	};
  
  	/*
  	 * No feature arguments supplied.
  	 */
  	if (!as->argc)
  		return 0;
  
  	r = dm_read_arg_group(_args, as, &argc, &ti->error);
  	if (r)
  		return -EINVAL;
  
  	while (argc && !r) {
  		arg_name = dm_shift_arg(as);
  		argc--;
  
  		if (!strcasecmp(arg_name, "discard_zeroes_cow"))
  			s->discard_zeroes_cow = true;
  
  		else if (!strcasecmp(arg_name, "discard_passdown_origin"))
  			s->discard_passdown_origin = true;
  
  		else {
  			ti->error = "Unrecognised feature requested";
  			r = -EINVAL;
  			break;
  		}
  	}
  
  	if (!s->discard_zeroes_cow && s->discard_passdown_origin) {
  		/*
  		 * TODO: really these are disjoint.. but ti->num_discard_bios
  		 * and dm_bio_get_target_bio_nr() require rigid constraints.
  		 */
  		ti->error = "discard_passdown_origin feature depends on discard_zeroes_cow";
  		r = -EINVAL;
  	}
  
  	return r;
  }

  /*

   * Construct a snapshot mapping:
   * <origin_dev> <COW-dev> <p|po|n> <chunk-size> [<# feature args> [<arg>]*]

   */
  static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
  {
  	struct dm_snapshot *s;

  	struct dm_arg_set as;

  	int i;

  	int r = -EINVAL;

  	char *origin_path, *cow_path;

  	dev_t origin_dev, cow_dev;

  	unsigned args_used, num_flush_bios = 1;

  	fmode_t origin_mode = FMODE_READ;

  	if (argc < 4) {
  		ti->error = "requires 4 or more arguments";

  		r = -EINVAL;

  		goto bad;

  	}

  	if (dm_target_is_snapshot_merge(ti)) {

  		num_flush_bios = 2;

  		origin_mode = FMODE_WRITE;
  	}

  	s = kzalloc(sizeof(*s), GFP_KERNEL);

  	if (!s) {

  		ti->error = "Cannot allocate private snapshot structure";

  		r = -ENOMEM;
  		goto bad;
  	}

  	as.argc = argc;
  	as.argv = argv;
  	dm_consume_args(&as, 4);
  	r = parse_snapshot_features(&as, s, ti);
  	if (r)
  		goto bad_features;

  	origin_path = argv[0];
  	argv++;
  	argc--;
  
  	r = dm_get_device(ti, origin_path, origin_mode, &s->origin);
  	if (r) {
  		ti->error = "Cannot get origin device";
  		goto bad_origin;
  	}

  	origin_dev = s->origin->bdev->bd_dev;

  	cow_path = argv[0];
  	argv++;
  	argc--;

  	cow_dev = dm_get_dev_t(cow_path);
  	if (cow_dev && cow_dev == origin_dev) {
  		ti->error = "COW device cannot be the same as origin device";
  		r = -EINVAL;
  		goto bad_cow;
  	}

  	r = dm_get_device(ti, cow_path, dm_table_get_mode(ti->table), &s->cow);

  	if (r) {
  		ti->error = "Cannot get COW device";
  		goto bad_cow;
  	}
  
  	r = dm_exception_store_create(ti, argc, argv, s, &args_used, &s->store);

  	if (r) {
  		ti->error = "Couldn't create exception store";

  		r = -EINVAL;

  		goto bad_store;

  	}

  	argv += args_used;
  	argc -= args_used;

  	s->ti = ti;

  	s->valid = 1;

  	s->snapshot_overflowed = 0;

  	s->active = 0;

  	atomic_set(&s->pending_exceptions_count, 0);

  	spin_lock_init(&s->pe_allocation_lock);

  	s->exception_start_sequence = 0;
  	s->exception_complete_sequence = 0;

  	s->out_of_order_tree = RB_ROOT;

  	init_rwsem(&s->lock);

  	INIT_LIST_HEAD(&s->list);

  	spin_lock_init(&s->pe_lock);

  	s->state_bits = 0;

  	s->merge_failed = false;

  	s->first_merging_chunk = 0;
  	s->num_merging_chunks = 0;
  	bio_list_init(&s->bios_queued_during_merge);

  	bio_init(&s->flush_bio, NULL, 0);

  
  	/* Allocate hash table for COW data */

  	if (init_hash_tables(s)) {

  		ti->error = "Unable to allocate hash table space";
  		r = -ENOMEM;

  		goto bad_hash_tables;

  	}

  	init_waitqueue_head(&s->in_progress_wait);

  	s->kcopyd_client = dm_kcopyd_client_create(&dm_kcopyd_throttle);

  	if (IS_ERR(s->kcopyd_client)) {
  		r = PTR_ERR(s->kcopyd_client);

  		ti->error = "Could not create kcopyd client";

  		goto bad_kcopyd;

  	}

  	r = mempool_init_slab_pool(&s->pending_pool, MIN_IOS, pending_cache);
  	if (r) {

  		ti->error = "Could not allocate mempool for pending exceptions";

  		goto bad_pending_pool;

  	}

  	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
  		INIT_HLIST_HEAD(&s->tracked_chunk_hash[i]);
  
  	spin_lock_init(&s->tracked_chunk_lock);

  	ti->private = s;

  	ti->num_flush_bios = num_flush_bios;

  	if (s->discard_zeroes_cow)
  		ti->num_discard_bios = (s->discard_passdown_origin ? 2 : 1);

  	ti->per_io_data_size = sizeof(struct dm_snap_tracked_chunk);

  
  	/* Add snapshot to the list of snapshots for this origin */
  	/* Exceptions aren't triggered till snapshot_resume() is called */
  	r = register_snapshot(s);
  	if (r == -ENOMEM) {
  		ti->error = "Snapshot origin struct allocation failed";
  		goto bad_load_and_register;
  	} else if (r < 0) {
  		/* invalid handover, register_snapshot has set ti->error */
  		goto bad_load_and_register;
  	}
  
  	/*
  	 * Metadata must only be loaded into one table at once, so skip this
  	 * if metadata will be handed over during resume.
  	 * Chunk size will be set during the handover - set it to zero to
  	 * ensure it's ignored.
  	 */
  	if (r > 0) {
  		s->store->chunk_size = 0;
  		return 0;
  	}

  	r = s->store->type->read_metadata(s->store, dm_add_exception,
  					  (void *)s);

  	if (r < 0) {

  		ti->error = "Failed to read snapshot metadata";

  		goto bad_read_metadata;

  	} else if (r > 0) {
  		s->valid = 0;
  		DMWARN("Snapshot is marked invalid.");

  	}

  	if (!s->store->chunk_size) {
  		ti->error = "Chunk size not set";

  		goto bad_read_metadata;

  	}

  
  	r = dm_set_target_max_io_len(ti, s->store->chunk_size);
  	if (r)
  		goto bad_read_metadata;

  
  	return 0;

  bad_read_metadata:
  	unregister_snapshot(s);

  bad_load_and_register:

  	mempool_exit(&s->pending_pool);

  bad_pending_pool:

  	dm_kcopyd_client_destroy(s->kcopyd_client);

  bad_kcopyd:

  	dm_exception_table_exit(&s->pending, pending_cache);
  	dm_exception_table_exit(&s->complete, exception_cache);

  bad_hash_tables:

  	dm_exception_store_destroy(s->store);

  bad_store:
  	dm_put_device(ti, s->cow);

  bad_cow:

  	dm_put_device(ti, s->origin);

  bad_origin:

  bad_features:

  	kfree(s);

  bad:

  	return r;
  }

  static void __free_exceptions(struct dm_snapshot *s)
  {

  	dm_kcopyd_client_destroy(s->kcopyd_client);

  	s->kcopyd_client = NULL;

  	dm_exception_table_exit(&s->pending, pending_cache);
  	dm_exception_table_exit(&s->complete, exception_cache);

  }

  static void __handover_exceptions(struct dm_snapshot *snap_src,
  				  struct dm_snapshot *snap_dest)
  {
  	union {
  		struct dm_exception_table table_swap;
  		struct dm_exception_store *store_swap;
  	} u;
  
  	/*
  	 * Swap all snapshot context information between the two instances.
  	 */
  	u.table_swap = snap_dest->complete;
  	snap_dest->complete = snap_src->complete;
  	snap_src->complete = u.table_swap;
  
  	u.store_swap = snap_dest->store;
  	snap_dest->store = snap_src->store;

  	snap_dest->store->userspace_supports_overflow = u.store_swap->userspace_supports_overflow;

  	snap_src->store = u.store_swap;
  
  	snap_dest->store->snap = snap_dest;
  	snap_src->store->snap = snap_src;

  	snap_dest->ti->max_io_len = snap_dest->store->chunk_size;

  	snap_dest->valid = snap_src->valid;

  	snap_dest->snapshot_overflowed = snap_src->snapshot_overflowed;

  
  	/*
  	 * Set source invalid to ensure it receives no further I/O.
  	 */
  	snap_src->valid = 0;
  }

  static void snapshot_dtr(struct dm_target *ti)
  {

  #ifdef CONFIG_DM_DEBUG
  	int i;
  #endif

  	struct dm_snapshot *s = ti->private;

  	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;

  	down_read(&_origins_lock);
  	/* Check whether exception handover must be cancelled */

  	(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL);

  	if (snap_src && snap_dest && (s == snap_src)) {

  		down_write(&snap_dest->lock);

  		snap_dest->valid = 0;

  		up_write(&snap_dest->lock);

  		DMERR("Cancelling snapshot handover.");
  	}
  	up_read(&_origins_lock);

  	if (dm_target_is_snapshot_merge(ti))
  		stop_merge(s);

  	/* Prevent further origin writes from using this snapshot. */
  	/* After this returns there can be no new kcopyd jobs. */

  	unregister_snapshot(s);

  	while (atomic_read(&s->pending_exceptions_count))

  		msleep(1);

  	/*

  	 * Ensure instructions in mempool_exit aren't reordered

  	 * before atomic_read.
  	 */
  	smp_mb();

  #ifdef CONFIG_DM_DEBUG
  	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
  		BUG_ON(!hlist_empty(&s->tracked_chunk_hash[i]));
  #endif

  	__free_exceptions(s);

  	mempool_exit(&s->pending_pool);

  	dm_exception_store_destroy(s->store);

  	bio_uninit(&s->flush_bio);

  	dm_put_device(ti, s->cow);

  	dm_put_device(ti, s->origin);

  	WARN_ON(s->in_progress);

  	kfree(s);
  }

  static void account_start_copy(struct dm_snapshot *s)
  {

  	spin_lock(&s->in_progress_wait.lock);
  	s->in_progress++;
  	spin_unlock(&s->in_progress_wait.lock);

  }
  
  static void account_end_copy(struct dm_snapshot *s)
  {

  	spin_lock(&s->in_progress_wait.lock);
  	BUG_ON(!s->in_progress);
  	s->in_progress--;
  	if (likely(s->in_progress <= cow_threshold) &&
  	    unlikely(waitqueue_active(&s->in_progress_wait)))
  		wake_up_locked(&s->in_progress_wait);
  	spin_unlock(&s->in_progress_wait.lock);
  }
  
  static bool wait_for_in_progress(struct dm_snapshot *s, bool unlock_origins)
  {
  	if (unlikely(s->in_progress > cow_threshold)) {
  		spin_lock(&s->in_progress_wait.lock);
  		if (likely(s->in_progress > cow_threshold)) {
  			/*
  			 * NOTE: this throttle doesn't account for whether
  			 * the caller is servicing an IO that will trigger a COW
  			 * so excess throttling may result for chunks not required
  			 * to be COW'd.  But if cow_threshold was reached, extra
  			 * throttling is unlikely to negatively impact performance.
  			 */
  			DECLARE_WAITQUEUE(wait, current);
  			__add_wait_queue(&s->in_progress_wait, &wait);
  			__set_current_state(TASK_UNINTERRUPTIBLE);
  			spin_unlock(&s->in_progress_wait.lock);
  			if (unlock_origins)
  				up_read(&_origins_lock);
  			io_schedule();
  			remove_wait_queue(&s->in_progress_wait, &wait);
  			return false;
  		}
  		spin_unlock(&s->in_progress_wait.lock);
  	}
  	return true;

  }

  /*
   * Flush a list of buffers.
   */
  static void flush_bios(struct bio *bio)
  {
  	struct bio *n;
  
  	while (bio) {
  		n = bio->bi_next;
  		bio->bi_next = NULL;

  		submit_bio_noacct(bio);

  		bio = n;
  	}
  }

  static int do_origin(struct dm_dev *origin, struct bio *bio, bool limit);

  
  /*
   * Flush a list of buffers.
   */
  static void retry_origin_bios(struct dm_snapshot *s, struct bio *bio)
  {
  	struct bio *n;
  	int r;
  
  	while (bio) {
  		n = bio->bi_next;
  		bio->bi_next = NULL;

  		r = do_origin(s->origin, bio, false);

  		if (r == DM_MAPIO_REMAPPED)

  			submit_bio_noacct(bio);

  		bio = n;
  	}
  }

  /*
   * Error a list of buffers.
   */
  static void error_bios(struct bio *bio)
  {
  	struct bio *n;
  
  	while (bio) {
  		n = bio->bi_next;
  		bio->bi_next = NULL;

  		bio_io_error(bio);

  		bio = n;
  	}
  }

  static void __invalidate_snapshot(struct dm_snapshot *s, int err)

  {
  	if (!s->valid)
  		return;
  
  	if (err == -EIO)
  		DMERR("Invalidating snapshot: Error reading/writing.");
  	else if (err == -ENOMEM)
  		DMERR("Invalidating snapshot: Unable to allocate exception.");

  	if (s->store->type->drop_snapshot)
  		s->store->type->drop_snapshot(s->store);

  
  	s->valid = 0;

  	dm_table_event(s->ti->table);

  }

  static void invalidate_snapshot(struct dm_snapshot *s, int err)
  {
  	down_write(&s->lock);
  	__invalidate_snapshot(s, err);
  	up_write(&s->lock);
  }

  static void pending_complete(void *context, int success)

  {

  	struct dm_snap_pending_exception *pe = context;

  	struct dm_exception *e;

  	struct dm_snapshot *s = pe->snap;

  	struct bio *origin_bios = NULL;
  	struct bio *snapshot_bios = NULL;

  	struct bio *full_bio = NULL;

  	struct dm_exception_table_lock lock;

  	int error = 0;

  	dm_exception_table_lock_init(s, pe->e.old_chunk, &lock);

  	if (!success) {
  		/* Read/write error - snapshot is unusable */

  		invalidate_snapshot(s, -EIO);

  		error = 1;

  
  		dm_exception_table_lock(&lock);

  		goto out;
  	}

  	e = alloc_completed_exception(GFP_NOIO);

  	if (!e) {

  		invalidate_snapshot(s, -ENOMEM);

  		error = 1;

  
  		dm_exception_table_lock(&lock);

  		goto out;
  	}
  	*e = pe->e;

  	down_read(&s->lock);

  	dm_exception_table_lock(&lock);

  	if (!s->valid) {

  		up_read(&s->lock);

  		free_completed_exception(e);

  		error = 1;

  		goto out;

  	}

  	/*

  	 * Add a proper exception. After inserting the completed exception all
  	 * subsequent snapshot reads to this chunk will be redirected to the
  	 * COW device.  This ensures that we do not starve. Moreover, as long
  	 * as the pending exception exists, neither origin writes nor snapshot
  	 * merging can overwrite the chunk in origin.

  	 */

  	dm_insert_exception(&s->complete, e);

  	up_read(&s->lock);

  	/* Wait for conflicting reads to drain */
  	if (__chunk_is_tracked(s, pe->e.old_chunk)) {

  		dm_exception_table_unlock(&lock);

  		__check_for_conflicting_io(s, pe->e.old_chunk);

  		dm_exception_table_lock(&lock);

  	}

  out:

  	/* Remove the in-flight exception from the list */

  	dm_remove_exception(&pe->e);

  
  	dm_exception_table_unlock(&lock);

  	snapshot_bios = bio_list_get(&pe->snapshot_bios);

  	origin_bios = bio_list_get(&pe->origin_bios);

  	full_bio = pe->full_bio;

  	if (full_bio)

  		full_bio->bi_end_io = pe->full_bio_end_io;

  	increment_pending_exceptions_done_count();

  	/* Submit any pending write bios */

  	if (error) {
  		if (full_bio)
  			bio_io_error(full_bio);

  		error_bios(snapshot_bios);

  	} else {
  		if (full_bio)

  			bio_endio(full_bio);

  		flush_bios(snapshot_bios);

  	}

  	retry_origin_bios(s, origin_bios);

  
  	free_pending_exception(pe);

  }

  static void complete_exception(struct dm_snap_pending_exception *pe)
  {
  	struct dm_snapshot *s = pe->snap;

  	/* Update the metadata if we are persistent */
  	s->store->type->commit_exception(s->store, &pe->e, !pe->copy_error,
  					 pending_complete, pe);

  }

  /*
   * Called when the copy I/O has finished.  kcopyd actually runs
   * this code so don't block.
   */

  static void copy_callback(int read_err, unsigned long write_err, void *context)

  {

  	struct dm_snap_pending_exception *pe = context;

  	struct dm_snapshot *s = pe->snap;

  	pe->copy_error = read_err || write_err;

  	if (pe->exception_sequence == s->exception_complete_sequence) {

  		struct rb_node *next;

  		s->exception_complete_sequence++;
  		complete_exception(pe);

  		next = rb_first(&s->out_of_order_tree);
  		while (next) {
  			pe = rb_entry(next, struct dm_snap_pending_exception,
  					out_of_order_node);

  			if (pe->exception_sequence != s->exception_complete_sequence)
  				break;

  			next = rb_next(next);

  			s->exception_complete_sequence++;

  			rb_erase(&pe->out_of_order_node, &s->out_of_order_tree);

  			complete_exception(pe);

  			cond_resched();

  		}
  	} else {

  		struct rb_node *parent = NULL;
  		struct rb_node **p = &s->out_of_order_tree.rb_node;

  		struct dm_snap_pending_exception *pe2;

  		while (*p) {
  			pe2 = rb_entry(*p, struct dm_snap_pending_exception, out_of_order_node);
  			parent = *p;
  
  			BUG_ON(pe->exception_sequence == pe2->exception_sequence);
  			if (pe->exception_sequence < pe2->exception_sequence)
  				p = &((*p)->rb_left);
  			else
  				p = &((*p)->rb_right);

  		}

  
  		rb_link_node(&pe->out_of_order_node, parent, p);
  		rb_insert_color(&pe->out_of_order_node, &s->out_of_order_tree);

  	}

  	account_end_copy(s);

  }
  
  /*
   * Dispatches the copy operation to kcopyd.
   */

  static void start_copy(struct dm_snap_pending_exception *pe)

  {
  	struct dm_snapshot *s = pe->snap;

  	struct dm_io_region src, dest;

  	struct block_device *bdev = s->origin->bdev;
  	sector_t dev_size;
  
  	dev_size = get_dev_size(bdev);
  
  	src.bdev = bdev;

  	src.sector = chunk_to_sector(s->store, pe->e.old_chunk);

  	src.count = min((sector_t)s->store->chunk_size, dev_size - src.sector);

  	dest.bdev = s->cow->bdev;

  	dest.sector = chunk_to_sector(s->store, pe->e.new_chunk);

  	dest.count = src.count;
  
  	/* Hand over to kcopyd */

  	account_start_copy(s);

  	dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, copy_callback, pe);

  }

  static void full_bio_end_io(struct bio *bio)

  {
  	void *callback_data = bio->bi_private;

  	dm_kcopyd_do_callback(callback_data, 0, bio->bi_status ? 1 : 0);

  }
  
  static void start_full_bio(struct dm_snap_pending_exception *pe,
  			   struct bio *bio)
  {
  	struct dm_snapshot *s = pe->snap;
  	void *callback_data;
  
  	pe->full_bio = bio;
  	pe->full_bio_end_io = bio->bi_end_io;

  	account_start_copy(s);

  	callback_data = dm_kcopyd_prepare_callback(s->kcopyd_client,
  						   copy_callback, pe);
  
  	bio->bi_end_io = full_bio_end_io;
  	bio->bi_private = callback_data;

  	submit_bio_noacct(bio);

  }

  static struct dm_snap_pending_exception *
  __lookup_pending_exception(struct dm_snapshot *s, chunk_t chunk)
  {

  	struct dm_exception *e = dm_lookup_exception(&s->pending, chunk);

  
  	if (!e)
  		return NULL;
  
  	return container_of(e, struct dm_snap_pending_exception, e);
  }

  /*

   * Inserts a pending exception into the pending table.

   *

   * NOTE: a write lock must be held on the chunk's pending exception table slot
   * before calling this.

   */

  static struct dm_snap_pending_exception *

  __insert_pending_exception(struct dm_snapshot *s,
  			   struct dm_snap_pending_exception *pe, chunk_t chunk)

  {

  	pe->e.old_chunk = chunk;
  	bio_list_init(&pe->origin_bios);
  	bio_list_init(&pe->snapshot_bios);

  	pe->started = 0;

  	pe->full_bio = NULL;

  	spin_lock(&s->pe_allocation_lock);

  	if (s->store->type->prepare_exception(s->store, &pe->e)) {

  		spin_unlock(&s->pe_allocation_lock);

  		free_pending_exception(pe);
  		return NULL;
  	}

  	pe->exception_sequence = s->exception_start_sequence++;

  	spin_unlock(&s->pe_allocation_lock);

  	dm_insert_exception(&s->pending, &pe->e);

  	return pe;
  }

  /*
   * Looks to see if this snapshot already has a pending exception
   * for this chunk, otherwise it allocates a new one and inserts
   * it into the pending table.
   *

   * NOTE: a write lock must be held on the chunk's pending exception table slot
   * before calling this.

   */
  static struct dm_snap_pending_exception *
  __find_pending_exception(struct dm_snapshot *s,
  			 struct dm_snap_pending_exception *pe, chunk_t chunk)
  {
  	struct dm_snap_pending_exception *pe2;
  
  	pe2 = __lookup_pending_exception(s, chunk);
  	if (pe2) {
  		free_pending_exception(pe);
  		return pe2;
  	}
  
  	return __insert_pending_exception(s, pe, chunk);
  }

  static void remap_exception(struct dm_snapshot *s, struct dm_exception *e,

  			    struct bio *bio, chunk_t chunk)

  {

  	bio_set_dev(bio, s->cow->bdev);

  	bio->bi_iter.bi_sector =
  		chunk_to_sector(s->store, dm_chunk_number(e->new_chunk) +
  				(chunk - e->old_chunk)) +
  		(bio->bi_iter.bi_sector & s->store->chunk_mask);

  }

  static void zero_callback(int read_err, unsigned long write_err, void *context)
  {
  	struct bio *bio = context;
  	struct dm_snapshot *s = bio->bi_private;

  	account_end_copy(s);

  	bio->bi_status = write_err ? BLK_STS_IOERR : 0;
  	bio_endio(bio);
  }
  
  static void zero_exception(struct dm_snapshot *s, struct dm_exception *e,
  			   struct bio *bio, chunk_t chunk)
  {
  	struct dm_io_region dest;
  
  	dest.bdev = s->cow->bdev;
  	dest.sector = bio->bi_iter.bi_sector;
  	dest.count = s->store->chunk_size;

  	account_start_copy(s);

  	WARN_ON_ONCE(bio->bi_private);
  	bio->bi_private = s;
  	dm_kcopyd_zero(s->kcopyd_client, 1, &dest, 0, zero_callback, bio);
  }
  
  static bool io_overlaps_chunk(struct dm_snapshot *s, struct bio *bio)
  {
  	return bio->bi_iter.bi_size ==
  		(s->store->chunk_size << SECTOR_SHIFT);
  }

  static int snapshot_map(struct dm_target *ti, struct bio *bio)

  {

  	struct dm_exception *e;

  	struct dm_snapshot *s = ti->private;

  	int r = DM_MAPIO_REMAPPED;

  	chunk_t chunk;

  	struct dm_snap_pending_exception *pe = NULL;

  	struct dm_exception_table_lock lock;