/*

   * Copyright (C) 2011-2012 Red Hat UK.

   *
   * This file is released under the GPL.
   */
  
  #include "dm-thin-metadata.h"

  #include "dm-bio-prison.h"

  #include "dm.h"

  
  #include <linux/device-mapper.h>
  #include <linux/dm-io.h>
  #include <linux/dm-kcopyd.h>
  #include <linux/list.h>

  #include <linux/rculist.h>

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

  #include <linux/rbtree.h>

  
  #define	DM_MSG_PREFIX	"thin"
  
  /*
   * Tunable constants
   */

  #define ENDIO_HOOK_POOL_SIZE 1024

  #define MAPPING_POOL_SIZE 1024
  #define PRISON_CELLS 1024

  #define COMMIT_PERIOD HZ

  #define NO_SPACE_TIMEOUT_SECS 60
  
  static unsigned no_space_timeout_secs = NO_SPACE_TIMEOUT_SECS;

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

  /*
   * The block size of the device holding pool data must be
   * between 64KB and 1GB.
   */
  #define DATA_DEV_BLOCK_SIZE_MIN_SECTORS (64 * 1024 >> SECTOR_SHIFT)
  #define DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT)
  
  /*

   * Device id is restricted to 24 bits.
   */
  #define MAX_DEV_ID ((1 << 24) - 1)
  
  /*
   * How do we handle breaking sharing of data blocks?
   * =================================================
   *
   * We use a standard copy-on-write btree to store the mappings for the
   * devices (note I'm talking about copy-on-write of the metadata here, not
   * the data).  When you take an internal snapshot you clone the root node
   * of the origin btree.  After this there is no concept of an origin or a
   * snapshot.  They are just two device trees that happen to point to the
   * same data blocks.
   *
   * When we get a write in we decide if it's to a shared data block using
   * some timestamp magic.  If it is, we have to break sharing.
   *
   * Let's say we write to a shared block in what was the origin.  The
   * steps are:
   *
   * i) plug io further to this physical block. (see bio_prison code).
   *
   * ii) quiesce any read io to that shared data block.  Obviously

   * including all devices that share this block.  (see dm_deferred_set code)

   *
   * iii) copy the data block to a newly allocate block.  This step can be
   * missed out if the io covers the block. (schedule_copy).
   *
   * iv) insert the new mapping into the origin's btree

   * (process_prepared_mapping).  This act of inserting breaks some

   * sharing of btree nodes between the two devices.  Breaking sharing only
   * effects the btree of that specific device.  Btrees for the other
   * devices that share the block never change.  The btree for the origin
   * device as it was after the last commit is untouched, ie. we're using
   * persistent data structures in the functional programming sense.
   *
   * v) unplug io to this physical block, including the io that triggered
   * the breaking of sharing.
   *
   * Steps (ii) and (iii) occur in parallel.
   *
   * The metadata _doesn't_ need to be committed before the io continues.  We
   * get away with this because the io is always written to a _new_ block.
   * If there's a crash, then:
   *
   * - The origin mapping will point to the old origin block (the shared
   * one).  This will contain the data as it was before the io that triggered
   * the breaking of sharing came in.
   *
   * - The snap mapping still points to the old block.  As it would after
   * the commit.
   *
   * The downside of this scheme is the timestamp magic isn't perfect, and
   * will continue to think that data block in the snapshot device is shared
   * even after the write to the origin has broken sharing.  I suspect data
   * blocks will typically be shared by many different devices, so we're
   * breaking sharing n + 1 times, rather than n, where n is the number of
   * devices that reference this data block.  At the moment I think the
   * benefits far, far outweigh the disadvantages.
   */
  
  /*----------------------------------------------------------------*/
  
  /*

   * Key building.
   */
  static void build_data_key(struct dm_thin_device *td,

  			   dm_block_t b, struct dm_cell_key *key)

  {
  	key->virtual = 0;
  	key->dev = dm_thin_dev_id(td);
  	key->block = b;
  }
  
  static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,

  			      struct dm_cell_key *key)

  {
  	key->virtual = 1;
  	key->dev = dm_thin_dev_id(td);
  	key->block = b;
  }
  
  /*----------------------------------------------------------------*/
  
  /*
   * A pool device ties together a metadata device and a data device.  It
   * also provides the interface for creating and destroying internal
   * devices.
   */

  struct dm_thin_new_mapping;

  /*

   * The pool runs in 4 modes.  Ordered in degraded order for comparisons.

   */
  enum pool_mode {
  	PM_WRITE,		/* metadata may be changed */

  	PM_OUT_OF_DATA_SPACE,	/* metadata may be changed, though data may not be allocated */

  	PM_READ_ONLY,		/* metadata may not be changed */
  	PM_FAIL,		/* all I/O fails */
  };

  struct pool_features {

  	enum pool_mode mode;

  	bool zero_new_blocks:1;
  	bool discard_enabled:1;
  	bool discard_passdown:1;

  	bool error_if_no_space:1;

  };

  struct thin_c;
  typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
  typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);

  struct pool {
  	struct list_head list;
  	struct dm_target *ti;	/* Only set if a pool target is bound */
  
  	struct mapped_device *pool_md;
  	struct block_device *md_dev;
  	struct dm_pool_metadata *pmd;

  	dm_block_t low_water_blocks;

  	uint32_t sectors_per_block;

  	int sectors_per_block_shift;

  	struct pool_features pf;

  	bool low_water_triggered:1;	/* A dm event has been sent */

  	struct dm_bio_prison *prison;

  	struct dm_kcopyd_client *copier;
  
  	struct workqueue_struct *wq;
  	struct work_struct worker;

  	struct delayed_work waker;

  	struct delayed_work no_space_timeout;

  	unsigned long last_commit_jiffies;

  	unsigned ref_count;

  
  	spinlock_t lock;

  	struct bio_list deferred_flush_bios;
  	struct list_head prepared_mappings;

  	struct list_head prepared_discards;

  	struct list_head active_thins;

  	struct dm_deferred_set *shared_read_ds;
  	struct dm_deferred_set *all_io_ds;

  	struct dm_thin_new_mapping *next_mapping;

  	mempool_t *mapping_pool;

  
  	process_bio_fn process_bio;
  	process_bio_fn process_discard;
  
  	process_mapping_fn process_prepared_mapping;
  	process_mapping_fn process_prepared_discard;

  };

  static enum pool_mode get_pool_mode(struct pool *pool);

  static void metadata_operation_failed(struct pool *pool, const char *op, int r);

  /*
   * Target context for a pool.
   */
  struct pool_c {
  	struct dm_target *ti;
  	struct pool *pool;
  	struct dm_dev *data_dev;
  	struct dm_dev *metadata_dev;
  	struct dm_target_callbacks callbacks;
  
  	dm_block_t low_water_blocks;

  	struct pool_features requested_pf; /* Features requested during table load */
  	struct pool_features adjusted_pf;  /* Features used after adjusting for constituent devices */

  };
  
  /*
   * Target context for a thin.
   */
  struct thin_c {

  	struct list_head list;

  	struct dm_dev *pool_dev;

  	struct dm_dev *origin_dev;

  	sector_t origin_size;

  	dm_thin_id dev_id;
  
  	struct pool *pool;
  	struct dm_thin_device *td;

  	bool requeue_mode:1;

  	spinlock_t lock;
  	struct bio_list deferred_bio_list;
  	struct bio_list retry_on_resume_list;

  	struct rb_root sort_bio_list; /* sorted list of deferred bios */

  
  	/*
  	 * Ensures the thin is not destroyed until the worker has finished
  	 * iterating the active_thins list.
  	 */
  	atomic_t refcount;
  	struct completion can_destroy;

  };
  
  /*----------------------------------------------------------------*/

  /*
   * wake_worker() is used when new work is queued and when pool_resume is
   * ready to continue deferred IO processing.
   */
  static void wake_worker(struct pool *pool)
  {
  	queue_work(pool->wq, &pool->worker);
  }
  
  /*----------------------------------------------------------------*/

  static int bio_detain(struct pool *pool, struct dm_cell_key *key, struct bio *bio,
  		      struct dm_bio_prison_cell **cell_result)
  {
  	int r;
  	struct dm_bio_prison_cell *cell_prealloc;
  
  	/*
  	 * Allocate a cell from the prison's mempool.
  	 * This might block but it can't fail.
  	 */
  	cell_prealloc = dm_bio_prison_alloc_cell(pool->prison, GFP_NOIO);
  
  	r = dm_bio_detain(pool->prison, key, bio, cell_prealloc, cell_result);
  	if (r)
  		/*
  		 * We reused an old cell; we can get rid of
  		 * the new one.
  		 */
  		dm_bio_prison_free_cell(pool->prison, cell_prealloc);
  
  	return r;
  }
  
  static void cell_release(struct pool *pool,
  			 struct dm_bio_prison_cell *cell,
  			 struct bio_list *bios)
  {
  	dm_cell_release(pool->prison, cell, bios);
  	dm_bio_prison_free_cell(pool->prison, cell);
  }
  
  static void cell_release_no_holder(struct pool *pool,
  				   struct dm_bio_prison_cell *cell,
  				   struct bio_list *bios)
  {
  	dm_cell_release_no_holder(pool->prison, cell, bios);
  	dm_bio_prison_free_cell(pool->prison, cell);
  }

  static void cell_defer_no_holder_no_free(struct thin_c *tc,
  					 struct dm_bio_prison_cell *cell)
  {
  	struct pool *pool = tc->pool;
  	unsigned long flags;

  	spin_lock_irqsave(&tc->lock, flags);
  	dm_cell_release_no_holder(pool->prison, cell, &tc->deferred_bio_list);
  	spin_unlock_irqrestore(&tc->lock, flags);

  
  	wake_worker(pool);
  }

  static void cell_error_with_code(struct pool *pool,
  				 struct dm_bio_prison_cell *cell, int error_code)

  {

  	dm_cell_error(pool->prison, cell, error_code);

  	dm_bio_prison_free_cell(pool->prison, cell);
  }

  static void cell_error(struct pool *pool, struct dm_bio_prison_cell *cell)
  {
  	cell_error_with_code(pool, cell, -EIO);
  }

  /*----------------------------------------------------------------*/

  /*
   * A global list of pools that uses a struct mapped_device as a key.
   */
  static struct dm_thin_pool_table {
  	struct mutex mutex;
  	struct list_head pools;
  } dm_thin_pool_table;
  
  static void pool_table_init(void)
  {
  	mutex_init(&dm_thin_pool_table.mutex);
  	INIT_LIST_HEAD(&dm_thin_pool_table.pools);
  }
  
  static void __pool_table_insert(struct pool *pool)
  {
  	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
  	list_add(&pool->list, &dm_thin_pool_table.pools);
  }
  
  static void __pool_table_remove(struct pool *pool)
  {
  	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
  	list_del(&pool->list);
  }
  
  static struct pool *__pool_table_lookup(struct mapped_device *md)
  {
  	struct pool *pool = NULL, *tmp;
  
  	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
  
  	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
  		if (tmp->pool_md == md) {
  			pool = tmp;
  			break;
  		}
  	}
  
  	return pool;
  }
  
  static struct pool *__pool_table_lookup_metadata_dev(struct block_device *md_dev)
  {
  	struct pool *pool = NULL, *tmp;
  
  	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
  
  	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
  		if (tmp->md_dev == md_dev) {
  			pool = tmp;
  			break;
  		}
  	}
  
  	return pool;
  }
  
  /*----------------------------------------------------------------*/

  struct dm_thin_endio_hook {

  	struct thin_c *tc;

  	struct dm_deferred_entry *shared_read_entry;
  	struct dm_deferred_entry *all_io_entry;

  	struct dm_thin_new_mapping *overwrite_mapping;

  	struct rb_node rb_node;

  };

  static void requeue_bio_list(struct thin_c *tc, struct bio_list *master)

  {
  	struct bio *bio;
  	struct bio_list bios;

  	unsigned long flags;

  
  	bio_list_init(&bios);

  	spin_lock_irqsave(&tc->lock, flags);

  	bio_list_merge(&bios, master);
  	bio_list_init(master);

  	spin_unlock_irqrestore(&tc->lock, flags);

  	while ((bio = bio_list_pop(&bios)))
  		bio_endio(bio, DM_ENDIO_REQUEUE);

  }
  
  static void requeue_io(struct thin_c *tc)
  {

  	requeue_bio_list(tc, &tc->deferred_bio_list);
  	requeue_bio_list(tc, &tc->retry_on_resume_list);

  }

  static void error_thin_retry_list(struct thin_c *tc)

  {
  	struct bio *bio;
  	unsigned long flags;
  	struct bio_list bios;
  
  	bio_list_init(&bios);

  	spin_lock_irqsave(&tc->lock, flags);
  	bio_list_merge(&bios, &tc->retry_on_resume_list);
  	bio_list_init(&tc->retry_on_resume_list);
  	spin_unlock_irqrestore(&tc->lock, flags);

  
  	while ((bio = bio_list_pop(&bios)))
  		bio_io_error(bio);
  }

  static void error_retry_list(struct pool *pool)
  {
  	struct thin_c *tc;
  
  	rcu_read_lock();
  	list_for_each_entry_rcu(tc, &pool->active_thins, list)
  		error_thin_retry_list(tc);
  	rcu_read_unlock();
  }

  /*
   * This section of code contains the logic for processing a thin device's IO.
   * Much of the code depends on pool object resources (lists, workqueues, etc)
   * but most is exclusively called from the thin target rather than the thin-pool
   * target.
   */

  static bool block_size_is_power_of_two(struct pool *pool)
  {
  	return pool->sectors_per_block_shift >= 0;
  }

  static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
  {

  	struct pool *pool = tc->pool;

  	sector_t block_nr = bio->bi_iter.bi_sector;

  	if (block_size_is_power_of_two(pool))
  		block_nr >>= pool->sectors_per_block_shift;

  	else

  		(void) sector_div(block_nr, pool->sectors_per_block);

  
  	return block_nr;

  }
  
  static void remap(struct thin_c *tc, struct bio *bio, dm_block_t block)
  {
  	struct pool *pool = tc->pool;

  	sector_t bi_sector = bio->bi_iter.bi_sector;

  
  	bio->bi_bdev = tc->pool_dev->bdev;

  	if (block_size_is_power_of_two(pool))

  		bio->bi_iter.bi_sector =
  			(block << pool->sectors_per_block_shift) |
  			(bi_sector & (pool->sectors_per_block - 1));

  	else

  		bio->bi_iter.bi_sector = (block * pool->sectors_per_block) +

  				 sector_div(bi_sector, pool->sectors_per_block);

  }

  static void remap_to_origin(struct thin_c *tc, struct bio *bio)
  {
  	bio->bi_bdev = tc->origin_dev->bdev;
  }

  static int bio_triggers_commit(struct thin_c *tc, struct bio *bio)
  {
  	return (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) &&
  		dm_thin_changed_this_transaction(tc->td);
  }

  static void inc_all_io_entry(struct pool *pool, struct bio *bio)
  {
  	struct dm_thin_endio_hook *h;
  
  	if (bio->bi_rw & REQ_DISCARD)
  		return;

  	h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  	h->all_io_entry = dm_deferred_entry_inc(pool->all_io_ds);
  }

  static void issue(struct thin_c *tc, struct bio *bio)

  {
  	struct pool *pool = tc->pool;
  	unsigned long flags;

  	if (!bio_triggers_commit(tc, bio)) {
  		generic_make_request(bio);
  		return;
  	}

  	/*

  	 * Complete bio with an error if earlier I/O caused changes to
  	 * the metadata that can't be committed e.g, due to I/O errors
  	 * on the metadata device.

  	 */

  	if (dm_thin_aborted_changes(tc->td)) {
  		bio_io_error(bio);
  		return;
  	}
  
  	/*
  	 * Batch together any bios that trigger commits and then issue a
  	 * single commit for them in process_deferred_bios().
  	 */
  	spin_lock_irqsave(&pool->lock, flags);
  	bio_list_add(&pool->deferred_flush_bios, bio);
  	spin_unlock_irqrestore(&pool->lock, flags);

  }

  static void remap_to_origin_and_issue(struct thin_c *tc, struct bio *bio)
  {
  	remap_to_origin(tc, bio);
  	issue(tc, bio);
  }
  
  static void remap_and_issue(struct thin_c *tc, struct bio *bio,
  			    dm_block_t block)
  {
  	remap(tc, bio, block);
  	issue(tc, bio);
  }

  /*----------------------------------------------------------------*/
  
  /*
   * Bio endio functions.
   */

  struct dm_thin_new_mapping {

  	struct list_head list;

  	bool pass_discard:1;
  	bool definitely_not_shared:1;

  	/*
  	 * Track quiescing, copying and zeroing preparation actions.  When this
  	 * counter hits zero the block is prepared and can be inserted into the
  	 * btree.
  	 */
  	atomic_t prepare_actions;

  	int err;

  	struct thin_c *tc;
  	dm_block_t virt_block;
  	dm_block_t data_block;

  	struct dm_bio_prison_cell *cell, *cell2;

  
  	/*
  	 * If the bio covers the whole area of a block then we can avoid
  	 * zeroing or copying.  Instead this bio is hooked.  The bio will
  	 * still be in the cell, so care has to be taken to avoid issuing
  	 * the bio twice.
  	 */
  	struct bio *bio;
  	bio_end_io_t *saved_bi_end_io;
  };

  static void __complete_mapping_preparation(struct dm_thin_new_mapping *m)

  {
  	struct pool *pool = m->tc->pool;

  	if (atomic_dec_and_test(&m->prepare_actions)) {

  		list_add_tail(&m->list, &pool->prepared_mappings);

  		wake_worker(pool);
  	}
  }

  static void complete_mapping_preparation(struct dm_thin_new_mapping *m)

  {
  	unsigned long flags;

  	struct pool *pool = m->tc->pool;

  	spin_lock_irqsave(&pool->lock, flags);

  	__complete_mapping_preparation(m);

  	spin_unlock_irqrestore(&pool->lock, flags);
  }

  static void copy_complete(int read_err, unsigned long write_err, void *context)
  {
  	struct dm_thin_new_mapping *m = context;
  
  	m->err = read_err || write_err ? -EIO : 0;
  	complete_mapping_preparation(m);
  }

  static void overwrite_endio(struct bio *bio, int err)
  {

  	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  	struct dm_thin_new_mapping *m = h->overwrite_mapping;

  
  	m->err = err;

  	complete_mapping_preparation(m);

  }

  /*----------------------------------------------------------------*/
  
  /*
   * Workqueue.
   */
  
  /*
   * Prepared mapping jobs.
   */
  
  /*
   * This sends the bios in the cell back to the deferred_bios list.
   */

  static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *cell)

  {
  	struct pool *pool = tc->pool;
  	unsigned long flags;

  	spin_lock_irqsave(&tc->lock, flags);
  	cell_release(pool, cell, &tc->deferred_bio_list);
  	spin_unlock_irqrestore(&tc->lock, flags);

  
  	wake_worker(pool);
  }
  
  /*

   * Same as cell_defer above, except it omits the original holder of the cell.

   */

  static void cell_defer_no_holder(struct thin_c *tc, struct dm_bio_prison_cell *cell)

  {

  	struct pool *pool = tc->pool;
  	unsigned long flags;

  	spin_lock_irqsave(&tc->lock, flags);
  	cell_release_no_holder(pool, cell, &tc->deferred_bio_list);
  	spin_unlock_irqrestore(&tc->lock, flags);

  
  	wake_worker(pool);
  }

  static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
  {

  	if (m->bio) {

  		m->bio->bi_end_io = m->saved_bi_end_io;

  		atomic_inc(&m->bio->bi_remaining);
  	}

  	cell_error(m->tc->pool, m->cell);

  	list_del(&m->list);
  	mempool_free(m, m->tc->pool->mapping_pool);
  }

  static void process_prepared_mapping(struct dm_thin_new_mapping *m)

  {
  	struct thin_c *tc = m->tc;

  	struct pool *pool = tc->pool;

  	struct bio *bio;
  	int r;
  
  	bio = m->bio;

  	if (bio) {

  		bio->bi_end_io = m->saved_bi_end_io;

  		atomic_inc(&bio->bi_remaining);
  	}

  
  	if (m->err) {

  		cell_error(pool, m->cell);

  		goto out;

  	}
  
  	/*
  	 * Commit the prepared block into the mapping btree.
  	 * Any I/O for this block arriving after this point will get
  	 * remapped to it directly.
  	 */
  	r = dm_thin_insert_block(tc->td, m->virt_block, m->data_block);
  	if (r) {

  		metadata_operation_failed(pool, "dm_thin_insert_block", r);

  		cell_error(pool, m->cell);

  		goto out;

  	}
  
  	/*
  	 * Release any bios held while the block was being provisioned.
  	 * If we are processing a write bio that completely covers the block,
  	 * we already processed it so can ignore it now when processing
  	 * the bios in the cell.
  	 */
  	if (bio) {

  		cell_defer_no_holder(tc, m->cell);

  		bio_endio(bio, 0);
  	} else

  		cell_defer(tc, m->cell);

  out:

  	list_del(&m->list);

  	mempool_free(m, pool->mapping_pool);

  }

  static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)

  {

  	struct thin_c *tc = m->tc;

  	bio_io_error(m->bio);

  	cell_defer_no_holder(tc, m->cell);
  	cell_defer_no_holder(tc, m->cell2);

  	mempool_free(m, tc->pool->mapping_pool);
  }
  
  static void process_prepared_discard_passdown(struct dm_thin_new_mapping *m)
  {
  	struct thin_c *tc = m->tc;

  	inc_all_io_entry(tc->pool, m->bio);

  	cell_defer_no_holder(tc, m->cell);
  	cell_defer_no_holder(tc, m->cell2);

  	if (m->pass_discard)

  		if (m->definitely_not_shared)
  			remap_and_issue(tc, m->bio, m->data_block);
  		else {
  			bool used = false;
  			if (dm_pool_block_is_used(tc->pool->pmd, m->data_block, &used) || used)
  				bio_endio(m->bio, 0);
  			else
  				remap_and_issue(tc, m->bio, m->data_block);
  		}

  	else
  		bio_endio(m->bio, 0);

  	mempool_free(m, tc->pool->mapping_pool);
  }

  static void process_prepared_discard(struct dm_thin_new_mapping *m)
  {
  	int r;
  	struct thin_c *tc = m->tc;
  
  	r = dm_thin_remove_block(tc->td, m->virt_block);
  	if (r)

  		DMERR_LIMIT("dm_thin_remove_block() failed");

  
  	process_prepared_discard_passdown(m);
  }

  static void process_prepared(struct pool *pool, struct list_head *head,

  			     process_mapping_fn *fn)

  {
  	unsigned long flags;
  	struct list_head maps;

  	struct dm_thin_new_mapping *m, *tmp;

  
  	INIT_LIST_HEAD(&maps);
  	spin_lock_irqsave(&pool->lock, flags);

  	list_splice_init(head, &maps);

  	spin_unlock_irqrestore(&pool->lock, flags);
  
  	list_for_each_entry_safe(m, tmp, &maps, list)

  		(*fn)(m);

  }
  
  /*
   * Deferred bio jobs.
   */

  static int io_overlaps_block(struct pool *pool, struct bio *bio)

  {

  	return bio->bi_iter.bi_size ==
  		(pool->sectors_per_block << SECTOR_SHIFT);

  }
  
  static int io_overwrites_block(struct pool *pool, struct bio *bio)
  {
  	return (bio_data_dir(bio) == WRITE) &&
  		io_overlaps_block(pool, bio);

  }
  
  static void save_and_set_endio(struct bio *bio, bio_end_io_t **save,
  			       bio_end_io_t *fn)
  {
  	*save = bio->bi_end_io;
  	bio->bi_end_io = fn;
  }
  
  static int ensure_next_mapping(struct pool *pool)
  {
  	if (pool->next_mapping)
  		return 0;
  
  	pool->next_mapping = mempool_alloc(pool->mapping_pool, GFP_ATOMIC);
  
  	return pool->next_mapping ? 0 : -ENOMEM;
  }

  static struct dm_thin_new_mapping *get_next_mapping(struct pool *pool)

  {

  	struct dm_thin_new_mapping *m = pool->next_mapping;

  
  	BUG_ON(!pool->next_mapping);

  	memset(m, 0, sizeof(struct dm_thin_new_mapping));
  	INIT_LIST_HEAD(&m->list);
  	m->bio = NULL;

  	pool->next_mapping = NULL;

  	return m;

  }

  static void ll_zero(struct thin_c *tc, struct dm_thin_new_mapping *m,
  		    sector_t begin, sector_t end)
  {
  	int r;
  	struct dm_io_region to;
  
  	to.bdev = tc->pool_dev->bdev;
  	to.sector = begin;
  	to.count = end - begin;
  
  	r = dm_kcopyd_zero(tc->pool->copier, 1, &to, 0, copy_complete, m);
  	if (r < 0) {
  		DMERR_LIMIT("dm_kcopyd_zero() failed");
  		copy_complete(1, 1, m);
  	}
  }
  
  /*
   * A partial copy also needs to zero the uncopied region.
   */

  static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,

  			  struct dm_dev *origin, dm_block_t data_origin,
  			  dm_block_t data_dest,

  			  struct dm_bio_prison_cell *cell, struct bio *bio,
  			  sector_t len)

  {
  	int r;
  	struct pool *pool = tc->pool;

  	struct dm_thin_new_mapping *m = get_next_mapping(pool);

  	m->tc = tc;
  	m->virt_block = virt_block;
  	m->data_block = data_dest;
  	m->cell = cell;

  	/*
  	 * quiesce action + copy action + an extra reference held for the
  	 * duration of this function (we may need to inc later for a
  	 * partial zero).
  	 */
  	atomic_set(&m->prepare_actions, 3);

  	if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))

  		complete_mapping_preparation(m); /* already quiesced */

  
  	/*
  	 * IO to pool_dev remaps to the pool target's data_dev.
  	 *
  	 * If the whole block of data is being overwritten, we can issue the
  	 * bio immediately. Otherwise we use kcopyd to clone the data first.
  	 */
  	if (io_overwrites_block(pool, bio)) {

  		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  		h->overwrite_mapping = m;

  		m->bio = bio;
  		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);

  		inc_all_io_entry(pool, bio);

  		remap_and_issue(tc, bio, data_dest);
  	} else {
  		struct dm_io_region from, to;

  		from.bdev = origin->bdev;

  		from.sector = data_origin * pool->sectors_per_block;

  		from.count = len;

  
  		to.bdev = tc->pool_dev->bdev;
  		to.sector = data_dest * pool->sectors_per_block;

  		to.count = len;

  
  		r = dm_kcopyd_copy(pool->copier, &from, 1, &to,
  				   0, copy_complete, m);
  		if (r < 0) {

  			DMERR_LIMIT("dm_kcopyd_copy() failed");

  			copy_complete(1, 1, m);
  
  			/*
  			 * We allow the zero to be issued, to simplify the
  			 * error path.  Otherwise we'd need to start
  			 * worrying about decrementing the prepare_actions
  			 * counter.
  			 */
  		}
  
  		/*
  		 * Do we need to zero a tail region?
  		 */
  		if (len < pool->sectors_per_block && pool->pf.zero_new_blocks) {
  			atomic_inc(&m->prepare_actions);
  			ll_zero(tc, m,
  				data_dest * pool->sectors_per_block + len,
  				(data_dest + 1) * pool->sectors_per_block);

  		}
  	}

  
  	complete_mapping_preparation(m); /* drop our ref */

  }

  static void schedule_internal_copy(struct thin_c *tc, dm_block_t virt_block,
  				   dm_block_t data_origin, dm_block_t data_dest,

  				   struct dm_bio_prison_cell *cell, struct bio *bio)

  {
  	schedule_copy(tc, virt_block, tc->pool_dev,

  		      data_origin, data_dest, cell, bio,
  		      tc->pool->sectors_per_block);

  }

  static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,

  			  dm_block_t data_block, struct dm_bio_prison_cell *cell,

  			  struct bio *bio)
  {
  	struct pool *pool = tc->pool;

  	struct dm_thin_new_mapping *m = get_next_mapping(pool);

  	atomic_set(&m->prepare_actions, 1); /* no need to quiesce */

  	m->tc = tc;
  	m->virt_block = virt_block;
  	m->data_block = data_block;
  	m->cell = cell;

  
  	/*
  	 * If the whole block of data is being overwritten or we are not
  	 * zeroing pre-existing data, we can issue the bio immediately.
  	 * Otherwise we use kcopyd to zero the data first.
  	 */

  	if (!pool->pf.zero_new_blocks)

  		process_prepared_mapping(m);
  
  	else if (io_overwrites_block(pool, bio)) {

  		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  		h->overwrite_mapping = m;

  		m->bio = bio;
  		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);

  		inc_all_io_entry(pool, bio);

  		remap_and_issue(tc, bio, data_block);

  	} else
  		ll_zero(tc, m,
  			data_block * pool->sectors_per_block,
  			(data_block + 1) * pool->sectors_per_block);
  }

  static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
  				   dm_block_t data_dest,
  				   struct dm_bio_prison_cell *cell, struct bio *bio)
  {
  	struct pool *pool = tc->pool;
  	sector_t virt_block_begin = virt_block * pool->sectors_per_block;
  	sector_t virt_block_end = (virt_block + 1) * pool->sectors_per_block;
  
  	if (virt_block_end <= tc->origin_size)
  		schedule_copy(tc, virt_block, tc->origin_dev,
  			      virt_block, data_dest, cell, bio,
  			      pool->sectors_per_block);
  
  	else if (virt_block_begin < tc->origin_size)
  		schedule_copy(tc, virt_block, tc->origin_dev,
  			      virt_block, data_dest, cell, bio,
  			      tc->origin_size - virt_block_begin);
  
  	else
  		schedule_zero(tc, virt_block, data_dest, cell, bio);

  }

  /*
   * A non-zero return indicates read_only or fail_io mode.
   * Many callers don't care about the return value.
   */

  static int commit(struct pool *pool)

  {
  	int r;

  	if (get_pool_mode(pool) >= PM_READ_ONLY)

  		return -EINVAL;

  	r = dm_pool_commit_metadata(pool->pmd);

  	if (r)
  		metadata_operation_failed(pool, "dm_pool_commit_metadata", r);

  
  	return r;
  }

  static void check_low_water_mark(struct pool *pool, dm_block_t free_blocks)
  {
  	unsigned long flags;
  
  	if (free_blocks <= pool->low_water_blocks && !pool->low_water_triggered) {
  		DMWARN("%s: reached low water mark for data device: sending event.",
  		       dm_device_name(pool->pool_md));
  		spin_lock_irqsave(&pool->lock, flags);
  		pool->low_water_triggered = true;
  		spin_unlock_irqrestore(&pool->lock, flags);
  		dm_table_event(pool->ti->table);
  	}
  }

  static void set_pool_mode(struct pool *pool, enum pool_mode new_mode);

  static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
  {
  	int r;
  	dm_block_t free_blocks;

  	struct pool *pool = tc->pool;

  	if (WARN_ON(get_pool_mode(pool) != PM_WRITE))

  		return -EINVAL;

  	r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);

  	if (r) {
  		metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);

  		return r;

  	}

  	check_low_water_mark(pool, free_blocks);

  
  	if (!free_blocks) {

  		/*
  		 * Try to commit to see if that will free up some
  		 * more space.
  		 */

  		r = commit(pool);
  		if (r)
  			return r;

  		r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);

  		if (r) {
  			metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);

  			return r;

  		}

  		if (!free_blocks) {

  			set_pool_mode(pool, PM_OUT_OF_DATA_SPACE);

  			return -ENOSPC;

  		}
  	}
  
  	r = dm_pool_alloc_data_block(pool->pmd, result);

  	if (r) {

  		metadata_operation_failed(pool, "dm_pool_alloc_data_block", r);

  		return r;

  	}

  
  	return 0;
  }
  
  /*
   * If we have run out of space, queue bios until the device is
   * resumed, presumably after having been reloaded with more space.
   */
  static void retry_on_resume(struct bio *bio)
  {

  	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  	struct thin_c *tc = h->tc;

  	unsigned long flags;

  	spin_lock_irqsave(&tc->lock, flags);
  	bio_list_add(&tc->retry_on_resume_list, bio);
  	spin_unlock_irqrestore(&tc->lock, flags);

  }

  static int should_error_unserviceable_bio(struct pool *pool)

  {

  	enum pool_mode m = get_pool_mode(pool);
  
  	switch (m) {
  	case PM_WRITE:
  		/* Shouldn't get here */
  		DMERR_LIMIT("bio unserviceable, yet pool is in PM_WRITE mode");

  		return -EIO;

  
  	case PM_OUT_OF_DATA_SPACE:

  		return pool->pf.error_if_no_space ? -ENOSPC : 0;

  
  	case PM_READ_ONLY:
  	case PM_FAIL:

  		return -EIO;

  	default:
  		/* Shouldn't get here */
  		DMERR_LIMIT("bio unserviceable, yet pool has an unknown mode");

  		return -EIO;

  	}
  }

  static void handle_unserviceable_bio(struct pool *pool, struct bio *bio)
  {

  	int error = should_error_unserviceable_bio(pool);
  
  	if (error)
  		bio_endio(bio, error);

  	else
  		retry_on_resume(bio);

  }

  static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *cell)

  {
  	struct bio *bio;
  	struct bio_list bios;

  	int error;

  	error = should_error_unserviceable_bio(pool);
  	if (error) {
  		cell_error_with_code(pool, cell, error);

  		return;
  	}

  	bio_list_init(&bios);

  	cell_release(pool, cell, &bios);

  	error = should_error_unserviceable_bio(pool);
  	if (error)

  		while ((bio = bio_list_pop(&bios)))

  			bio_endio(bio, error);

  	else
  		while ((bio = bio_list_pop(&bios)))
  			retry_on_resume(bio);

  }

  static void process_discard(struct thin_c *tc, struct bio *bio)
  {
  	int r;

  	unsigned long flags;

  	struct pool *pool = tc->pool;

  	struct dm_bio_prison_cell *cell, *cell2;

  	struct dm_cell_key key, key2;

  	dm_block_t block = get_bio_block(tc, bio);
  	struct dm_thin_lookup_result lookup_result;

  	struct dm_thin_new_mapping *m;

  
  	build_virtual_key(tc->td, block, &key);

  	if (bio_detain(tc->pool, &key, bio, &cell))

  		return;
  
  	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
  	switch (r) {
  	case 0:
  		/*
  		 * Check nobody is fiddling with this pool block.  This can
  		 * happen if someone's in the process of breaking sharing
  		 * on this block.
  		 */
  		build_data_key(tc->td, lookup_result.block, &key2);

  		if (bio_detain(tc->pool, &key2, bio, &cell2)) {

  			cell_defer_no_holder(tc, cell);

  			break;
  		}
  
  		if (io_overlaps_block(pool, bio)) {
  			/*
  			 * IO may still be going to the destination block.  We must
  			 * quiesce before we can do the removal.
  			 */
  			m = get_next_mapping(pool);
  			m->tc = tc;

  			m->pass_discard = pool->pf.discard_passdown;
  			m->definitely_not_shared = !lookup_result.shared;

  			m->virt_block = block;
  			m->data_block = lookup_result.block;
  			m->cell = cell;
  			m->cell2 = cell2;

  			m->bio = bio;

  			if (!dm_deferred_set_add_work(pool->all_io_ds, &m->list)) {

  				spin_lock_irqsave(&pool->lock, flags);

  				list_add_tail(&m->list, &pool->prepared_discards);

  				spin_unlock_irqrestore(&pool->lock, flags);

  				wake_worker(pool);
  			}
  		} else {

  			inc_all_io_entry(pool, bio);

  			cell_defer_no_holder(tc, cell);
  			cell_defer_no_holder(tc, cell2);

  			/*

  			 * The DM core makes sure that the discard doesn't span
  			 * a block boundary.  So we submit the discard of a
  			 * partial block appropriately.

  			 */

  			if ((!lookup_result.shared) && pool->pf.discard_passdown)
  				remap_and_issue(tc, bio, lookup_result.block);
  			else
  				bio_endio(bio, 0);

  		}
  		break;
  
  	case -ENODATA:
  		/*
  		 * It isn't provisioned, just forget it.
  		 */

  		cell_defer_no_holder(tc, cell);

  		bio_endio(bio, 0);
  		break;
  
  	default:

  		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
  			    __func__, r);

  		cell_defer_no_holder(tc, cell);

  		bio_io_error(bio);
  		break;
  	}
  }

  static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,

  			  struct dm_cell_key *key,

  			  struct dm_thin_lookup_result *lookup_result,

  			  struct dm_bio_prison_cell *cell)

  {
  	int r;
  	dm_block_t data_block;

  	struct pool *pool = tc->pool;

  
  	r = alloc_data_block(tc, &data_block);
  	switch (r) {
  	case 0:

  		schedule_internal_copy(tc, block, lookup_result->block,
  				       data_block, cell, bio);

  		break;
  
  	case -ENOSPC:

  		retry_bios_on_resume(pool, cell);

  		break;
  
  	default:

  		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
  			    __func__, r);

  		cell_error(pool, cell);

  		break;
  	}
  }
  
  static void process_shared_bio(struct thin_c *tc, struct bio *bio,
  			       dm_block_t block,
  			       struct dm_thin_lookup_result *lookup_result)
  {

  	struct dm_bio_prison_cell *cell;

  	struct pool *pool = tc->pool;

  	struct dm_cell_key key;

  
  	/*
  	 * If cell is already occupied, then sharing is already in the process
  	 * of being broken so we have nothing further to do here.
  	 */
  	build_data_key(tc->td, lookup_result->block, &key);

  	if (bio_detain(pool, &key, bio, &cell))

  		return;

  	if (bio_data_dir(bio) == WRITE && bio->bi_iter.bi_size)

  		break_sharing(tc, bio, block, &key, lookup_result, cell);
  	else {

  		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  		h->shared_read_entry = dm_deferred_entry_inc(pool->shared_read_ds);

  		inc_all_io_entry(pool, bio);

  		cell_defer_no_holder(tc, cell);

  		remap_and_issue(tc, bio, lookup_result->block);
  	}
  }
  
  static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block,

  			    struct dm_bio_prison_cell *cell)

  {
  	int r;
  	dm_block_t data_block;

  	struct pool *pool = tc->pool;

  
  	/*
  	 * Remap empty bios (flushes) immediately, without provisioning.
  	 */

  	if (!bio->bi_iter.bi_size) {

  		inc_all_io_entry(pool, bio);

  		cell_defer_no_holder(tc, cell);

  		remap_and_issue(tc, bio, 0);
  		return;
  	}
  
  	/*
  	 * Fill read bios with zeroes and complete them immediately.
  	 */
  	if (bio_data_dir(bio) == READ) {
  		zero_fill_bio(bio);

  		cell_defer_no_holder(tc, cell);

  		bio_endio(bio, 0);
  		return;
  	}
  
  	r = alloc_data_block(tc, &data_block);
  	switch (r) {
  	case 0:

  		if (tc->origin_dev)
  			schedule_external_copy(tc, block, data_block, cell, bio);
  		else
  			schedule_zero(tc, block, data_block, cell, bio);

  		break;
  
  	case -ENOSPC:

  		retry_bios_on_resume(pool, cell);

  		break;
  
  	default:

  		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
  			    __func__, r);

  		cell_error(pool, cell);

  		break;
  	}
  }
  
  static void process_bio(struct thin_c *tc, struct bio *bio)
  {
  	int r;

  	struct pool *pool = tc->pool;

  	dm_block_t block = get_bio_block(tc, bio);

  	struct dm_bio_prison_cell *cell;

  	struct dm_cell_key key;

  	struct dm_thin_lookup_result lookup_result;
  
  	/*
  	 * If cell is already occupied, then the block is already
  	 * being provisioned so we have nothing further to do here.
  	 */
  	build_virtual_key(tc->td, block, &key);

  	if (bio_detain(pool, &key, bio, &cell))

  		return;
  
  	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
  	switch (r) {
  	case 0:

  		if (lookup_result.shared) {

  			process_shared_bio(tc, bio, block, &lookup_result);

  			cell_defer_no_holder(tc, cell); /* FIXME: pass this cell into process_shared? */

  		} else {

  			inc_all_io_entry(pool, bio);

  			cell_defer_no_holder(tc, cell);

  			remap_and_issue(tc, bio, lookup_result.block);

  		}

  		break;
  
  	case -ENODATA:

  		if (bio_data_dir(bio) == READ && tc->origin_dev) {

  			inc_all_io_entry(pool, bio);

  			cell_defer_no_holder(tc, cell);

  			if (bio_end_sector(bio) <= tc->origin_size)
  				remap_to_origin_and_issue(tc, bio);
  
  			else if (bio->bi_iter.bi_sector < tc->origin_size) {
  				zero_fill_bio(bio);
  				bio->bi_iter.bi_size = (tc->origin_size - bio->bi_iter.bi_sector) << SECTOR_SHIFT;
  				remap_to_origin_and_issue(tc, bio);
  
  			} else {
  				zero_fill_bio(bio);
  				bio_endio(bio, 0);
  			}

  		} else
  			provision_block(tc, bio, block, cell);

  		break;
  
  	default:

  		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
  			    __func__, r);

  		cell_defer_no_holder(tc, cell);

  		bio_io_error(bio);
  		break;
  	}
  }

  static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
  {
  	int r;
  	int rw = bio_data_dir(bio);
  	dm_block_t block = get_bio_block(tc, bio);
  	struct dm_thin_lookup_result lookup_result;
  
  	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
  	switch (r) {
  	case 0:

  		if (lookup_result.shared && (rw == WRITE) && bio->bi_iter.bi_size)

  			handle_unserviceable_bio(tc->pool, bio);

  		else {
  			inc_all_io_entry(tc->pool, bio);

  			remap_and_issue(tc, bio, lookup_result.block);

  		}

  		break;
  
  	case -ENODATA:
  		if (rw != READ) {

  			handle_unserviceable_bio(tc->pool, bio);

  			break;
  		}
  
  		if (tc->origin_dev) {

  			inc_all_io_entry(tc->pool, bio);

  			remap_to_origin_and_issue(tc, bio);
  			break;
  		}
  
  		zero_fill_bio(bio);
  		bio_endio(bio, 0);
  		break;
  
  	default:

  		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
  			    __func__, r);

  		bio_io_error(bio);
  		break;
  	}
  }

  static void process_bio_success(struct thin_c *tc, struct bio *bio)
  {
  	bio_endio(bio, 0);
  }

  static void process_bio_fail(struct thin_c *tc, struct bio *bio)
  {
  	bio_io_error(bio);
  }

  /*
   * FIXME: should we also commit due to size of transaction, measured in
   * metadata blocks?
   */

  static int need_commit_due_to_time(struct pool *pool)
  {
  	return jiffies < pool->last_commit_jiffies ||
  	       jiffies > pool->last_commit_jiffies + COMMIT_PERIOD;
  }

  #define thin_pbd(node) rb_entry((node), struct dm_thin_endio_hook, rb_node)
  #define thin_bio(pbd) dm_bio_from_per_bio_data((pbd), sizeof(struct dm_thin_endio_hook))
  
  static void __thin_bio_rb_add(struct thin_c *tc, struct bio *bio)
  {
  	struct rb_node **rbp, *parent;
  	struct dm_thin_endio_hook *pbd;
  	sector_t bi_sector = bio->bi_iter.bi_sector;
  
  	rbp = &tc->sort_bio_list.rb_node;
  	parent = NULL;
  	while (*rbp) {
  		parent = *rbp;
  		pbd = thin_pbd(parent);
  
  		if (bi_sector < thin_bio(pbd)->bi_iter.bi_sector)
  			rbp = &(*rbp)->rb_left;
  		else
  			rbp = &(*rbp)->rb_right;
  	}
  
  	pbd = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
  	rb_link_node(&pbd->rb_node, parent, rbp);
  	rb_insert_color(&pbd->rb_node, &tc->sort_bio_list);
  }
  
  static void __extract_sorted_bios(struct thin_c *tc)
  {
  	struct rb_node *node;
  	struct dm_thin_endio_hook *pbd;
  	struct bio *bio;
  
  	for (node = rb_first(&tc->sort_bio_list); node; node = rb_next(node)) {
  		pbd = thin_pbd(node);
  		bio = thin_bio(pbd);
  
  		bio_list_add(&tc->deferred_bio_list, bio);
  		rb_erase(&pbd->rb_node, &tc->sort_bio_list);
  	}
  
  	WARN_ON(!RB_EMPTY_ROOT(&tc->sort_bio_list));
  }
  
  static void __sort_thin_deferred_bios(struct thin_c *tc)
  {
  	struct bio *bio;
  	struct bio_list bios;
  
  	bio_list_init(&bios);
  	bio_list_merge(&bios, &tc->deferred_bio_list);
  	bio_list_init(&tc->deferred_bio_list);
  
  	/* Sort deferred_bio_list using rb-tree */
  	while ((bio = bio_list_pop(&bios)))
  		__thin_bio_rb_add(tc, bio);
  
  	/*
  	 * Transfer the sorted bios in sort_bio_list back to
  	 * deferred_bio_list to allow lockless submission of
  	 * all bios.
  	 */
  	__extract_sorted_bios(tc);
  }

  static void process_thin_deferred_bios(struct thin_c *tc)

  {

  	struct pool *pool = tc->pool;

  	unsigned long flags;
  	struct bio *bio;
  	struct bio_list bios;

  	struct blk_plug plug;

  	if (tc->requeue_mode) {
  		requeue_bio_list(tc, &tc->deferred_bio_list);
  		return;
  	}

  	bio_list_init(&bios);

  	spin_lock_irqsave(&tc->lock, flags);

  
  	if (bio_list_empty(&tc->deferred_bio_list)) {
  		spin_unlock_irqrestore(&tc->lock, flags);
  		return;
  	}
  
  	__sort_thin_deferred_bios(tc);

  	bio_list_merge(&bios, &tc->deferred_bio_list);
  	bio_list_init(&tc->deferred_bio_list);

  	spin_unlock_irqrestore(&tc->lock, flags);

  	blk_start_plug(&plug);

  	while ((bio = bio_list_pop(&bios))) {

  		/*
  		 * If we've got no free new_mapping structs, and processing
  		 * this bio might require one, we pause until there are some
  		 * prepared mappings to process.
  		 */
  		if (ensure_next_mapping(pool)) {

  			spin_lock_irqsave(&tc->lock, flags);
  			bio_list_add(&tc->deferred_bio_list, bio);
  			bio_list_merge(&tc->deferred_bio_list, &bios);
  			spin_unlock_irqrestore(&tc->lock, flags);

  			break;
  		}

  
  		if (bio->bi_rw & REQ_DISCARD)

  			pool->process_discard(tc, bio);

  		else

  			pool->process_bio(tc, bio);

  	}

  	blk_finish_plug(&plug);

  }

  static void thin_get(struct thin_c *tc);
  static void thin_put(struct thin_c *tc);
  
  /*
   * We can't hold rcu_read_lock() around code that can block.  So we
   * find a thin with the rcu lock held; bump a refcount; then drop
   * the lock.
   */
  static struct thin_c *get_first_thin(struct pool *pool)
  {
  	struct thin_c *tc = NULL;
  
  	rcu_read_lock();
  	if (!list_empty(&pool->active_thins)) {
  		tc = list_entry_rcu(pool->active_thins.next, struct thin_c, list);
  		thin_get(tc);
  	}
  	rcu_read_unlock();
  
  	return tc;
  }
  
  static struct thin_c *get_next_thin(struct pool *pool, struct thin_c *tc)
  {
  	struct thin_c *old_tc = tc;
  
  	rcu_read_lock();
  	list_for_each_entry_continue_rcu(tc, &pool->active_thins, list) {
  		thin_get(tc);
  		thin_put(old_tc);
  		rcu_read_unlock();
  		return tc;
  	}
  	thin_put(old_tc);
  	rcu_read_unlock();
  
  	return NULL;
  }

  static void process_deferred_bios(struct pool *pool)
  {
  	unsigned long flags;
  	struct bio *bio;
  	struct bio_list bios;
  	struct thin_c *tc;

  	tc = get_first_thin(pool);
  	while (tc) {

  		process_thin_deferred_bios(tc);

  		tc = get_next_thin(pool, tc);
  	}

  
  	/*
  	 * If there are any deferred flush bios, we must commit
  	 * the metadata before issuing them.
  	 */
  	bio_list_init(&bios);
  	spin_lock_irqsave(&pool->lock, flags);
  	bio_list_merge(&bios, &pool->deferred_flush_bios);
  	bio_list_init(&pool->deferred_flush_bios);
  	spin_unlock_irqrestore(&pool->lock, flags);

  	if (bio_list_empty(&bios) &&
  	    !(dm_pool_changed_this_transaction(pool->pmd) && need_commit_due_to_time(pool)))

  		return;

  	if (commit(pool)) {

  		while ((bio = bio_list_pop(&bios)))
  			bio_io_error(bio);
  		return;
  	}

  	pool->last_commit_jiffies = jiffies;

  
  	while ((bio = bio_list_pop(&bios)))
  		generic_make_request(bio);
  }
  
  static void do_worker(struct work_struct *ws)
  {
  	struct pool *pool = container_of(ws, struct pool, worker);

  	process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
  	process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);

  	process_deferred_bios(pool);
  }

  /*
   * We want to commit periodically so that not too much
   * unwritten data builds up.
   */
  static void do_waker(struct work_struct *ws)
  {
  	struct pool *pool = container_of(to_delayed_work(ws), struct pool, waker);
  	wake_worker(pool);
  	queue_delayed_work(pool->wq, &pool->waker, COMMIT_PERIOD);
  }

  /*
   * We're holding onto IO to allow userland time to react.  After the
   * timeout either the pool will have been resized (and thus back in
   * PM_WRITE mode), or we degrade to PM_READ_ONLY and start erroring IO.
   */
  static void do_no_space_timeout(struct work_struct *ws)
  {
  	struct pool *pool = container_of(to_delayed_work(ws), struct pool,
  					 no_space_timeout);
  
  	if (get_pool_mode(pool) == PM_OUT_OF_DATA_SPACE && !pool->pf.error_if_no_space)
  		set_pool_mode(pool, PM_READ_ONLY);
  }

  /*----------------------------------------------------------------*/

  struct pool_work {

  	struct work_struct worker;

  	struct completion complete;
  };
  
  static struct pool_work *to_pool_work(struct work_struct *ws)
  {
  	return container_of(ws, struct pool_work, worker);
  }
  
  static void pool_work_complete(struct pool_work *pw)
  {
  	complete(&pw->complete);
  }

  static void pool_work_wait(struct pool_work *pw, struct pool *pool,
  			   void (*fn)(struct work_struct *))
  {
  	INIT_WORK_ONSTACK(&pw->worker, fn);
  	init_completion(&pw->complete);
  	queue_work(pool->wq, &pw->worker);
  	wait_for_completion(&pw->complete);
  }
  
  /*----------------------------------------------------------------*/
  
  struct noflush_work {
  	struct pool_work pw;
  	struct thin_c *tc;

  };

  static struct noflush_work *to_noflush(struct work_struct *ws)

  {

  	return container_of(to_pool_work(ws), struct noflush_work, pw);

  }
  
  static void do_noflush_start(struct work_struct *ws)
  {

  	struct noflush_work *w = to_noflush(ws);

  	w->tc->requeue_mode = true;
  	requeue_io(w->tc);

  	pool_work_complete(&w->pw);

  }
  
  static void do_noflush_stop(struct work_struct *ws)
  {

  	struct noflush_work *w = to_noflush(ws);

  	w->tc->requeue_mode = false;

  	pool_work_complete(&w->pw);

  }
  
  static void noflush_work(struct thin_c *tc, void (*fn)(struct work_struct *))
  {
  	struct noflush_work w;

  	w.tc = tc;

  	pool_work_wait(&w.pw, tc->pool, fn);

  }
  
  /*----------------------------------------------------------------*/

  static enum pool_mode get_pool_mode(struct pool *pool)
  {
  	return pool->pf.mode;
  }

  static void notify_of_pool_mode_change(struct pool *pool, const char *new_mode)
  {
  	dm_table_event(pool->ti->table);
  	DMINFO("%s: switching pool to %s mode",
  	       dm_device_name(pool->pool_md), new_mode);
  }

  static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)

  {

  	struct pool_c *pt = pool->ti->private;

  	bool needs_check = dm_pool_metadata_needs_check(pool->pmd);
  	enum pool_mode old_mode = get_pool_mode(pool);

  	unsigned long no_space_timeout = ACCESS_ONCE(no_space_timeout_secs) * HZ;

  
  	/*
  	 * Never allow the pool to transition to PM_WRITE mode if user
  	 * intervention is required to verify metadata and data consistency.
  	 */
  	if (new_mode == PM_WRITE && needs_check) {
  		DMERR("%s: unable to switch pool to write mode until repaired.",
  		      dm_device_name(pool->pool_md));
  		if (old_mode != new_mode)
  			new_mode = old_mode;
  		else
  			new_mode = PM_READ_ONLY;
  	}
  	/*
  	 * If we were in PM_FAIL mode, rollback of metadata failed.  We're
  	 * not going to recover without a thin_repair.	So we never let the
  	 * pool move out of the old mode.
  	 */
  	if (old_mode == PM_FAIL)
  		new_mode = old_mode;

  	switch (new_mode) {

  	case PM_FAIL:

  		if (old_mode != new_mode)

  			notify_of_pool_mode_change(pool, "failure");

  		dm_pool_metadata_read_only(pool->pmd);

  		pool->process_bio = process_bio_fail;
  		pool->process_discard = process_bio_fail;
  		pool->process_prepared_mapping = process_prepared_mapping_fail;
  		pool->process_prepared_discard = process_prepared_discard_fail;

  
  		error_retry_list(pool);

  		break;
  
  	case PM_READ_ONLY:

  		if (old_mode != new_mode)

  			notify_of_pool_mode_change(pool, "read-only");
  		dm_pool_metadata_read_only(pool->pmd);
  		pool->process_bio = process_bio_read_only;
  		pool->process_discard = process_bio_success;
  		pool->process_prepared_mapping = process_prepared_mapping_fail;
  		pool->process_prepared_discard = process_prepared_discard_passdown;
  
  		error_retry_list(pool);
  		break;
  
  	case PM_OUT_OF_DATA_SPACE:
  		/*
  		 * Ideally we'd never hit this state; the low water mark
  		 * would trigger userland to extend the pool before we
  		 * completely run out of data space.  However, many small
  		 * IOs to unprovisioned space can consume data space at an
  		 * alarming rate.  Adjust your low water mark if you're
  		 * frequently seeing this mode.
  		 */
  		if (old_mode != new_mode)
  			notify_of_pool_mode_change(pool, "out-of-data-space");
  		pool->process_bio = process_bio_read_only;
  		pool->process_discard = process_discard;
  		pool->process_prepared_mapping = process_prepared_mapping;
  		pool->process_prepared_discard = process_prepared_discard_passdown;

  		if (!pool->pf.error_if_no_space && no_space_timeout)
  			queue_delayed_work(pool->wq, &pool->no_space_timeout, no_space_timeout);

  		break;
  
  	case PM_WRITE:

  		if (old_mode != new_mode)

  			notify_of_pool_mode_change(pool, "write");

  		dm_pool_metadata_read_write(pool->pmd);

  		pool->process_bio = process_bio;
  		pool->process_discard = process_discard;
  		pool->process_prepared_mapping = process_prepared_mapping;
  		pool->process_prepared_discard = process_prepared_discard;
  		break;
  	}

  
  	pool->pf.mode = new_mode;

  	/*
  	 * The pool mode may have changed, sync it so bind_control_target()
  	 * doesn't cause an unexpected mode transition on resume.
  	 */
  	pt->adjusted_pf.mode = new_mode;

  }

  static void abort_transaction(struct pool *pool)

  {

  	const char *dev_name = dm_device_name(pool->pool_md);
  
  	DMERR_LIMIT("%s: aborting current metadata transaction", dev_name);
  	if (dm_pool_abort_metadata(pool->pmd)) {
  		DMERR("%s: failed to abort metadata transaction", dev_name);
  		set_pool_mode(pool, PM_FAIL);
  	}
  
  	if (dm_pool_metadata_set_needs_check(pool->pmd)) {
  		DMERR("%s: failed to set 'needs_check' flag in metadata", dev_name);
  		set_pool_mode(pool, PM_FAIL);
  	}
  }

  static void metadata_operation_failed(struct pool *pool, const char *op, int r)
  {

  	DMERR_LIMIT("%s: metadata operation '%s' failed: error = %d",
  		    dm_device_name(pool->pool_md), op, r);

  	abort_transaction(pool);

  	set_pool_mode(pool, PM_READ_ONLY);
  }

  /*----------------------------------------------------------------*/

  /*
   * Mapping functions.
   */
  
  /*
   * Called only while mapping a thin bio to hand it over to the workqueue.
   */
  static void thin_defer_bio(struct thin_c *tc, struct bio *bio)
  {
  	unsigned long flags;
  	struct pool *pool = tc->pool;

  	spin_lock_irqsave(&tc->lock, flags);
  	bio_list_add(&tc->deferred_bio_list, bio);
  	spin_unlock_irqrestore(&tc->lock, flags);

  
  	wake_worker(pool);
  }

  static void thin_hook_bio(struct thin_c *tc, struct bio *bio)

  {

  	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));

  
  	h->tc = tc;
  	h->shared_read_entry = NULL;

  	h->all_io_entry = NULL;

  	h->overwrite_mapping = NULL;

  }

  /*
   * Non-blocking function called from the thin target's map function.
   */

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

  {
  	int r;
  	struct thin_c *tc = ti->private;
  	dm_block_t block = get_bio_block(tc, bio);
  	struct dm_thin_device *td = tc->td;
  	struct dm_thin_lookup_result result;

  	struct dm_bio_prison_cell cell1, cell2;
  	struct dm_bio_prison_cell *cell_result;

  	struct dm_cell_key key;

  	thin_hook_bio(tc, bio);

  	if (tc->requeue_mode) {
  		bio_endio(bio, DM_ENDIO_REQUEUE);
  		return DM_MAPIO_SUBMITTED;
  	}

  	if (get_pool_mode(tc->pool) == PM_FAIL) {
  		bio_io_error(bio);
  		return DM_MAPIO_SUBMITTED;
  	}

  	if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)) {

  		thin_defer_bio(tc, bio);
  		return DM_MAPIO_SUBMITTED;
  	}
  
  	r = dm_thin_find_block(td, block, 0, &result);
  
  	/*
  	 * Note that we defer readahead too.
  	 */
  	switch (r) {
  	case 0:
  		if (unlikely(result.shared)) {
  			/*
  			 * We have a race condition here between the
  			 * result.shared value returned by the lookup and
  			 * snapshot creation, which may cause new
  			 * sharing.
  			 *
  			 * To avoid this always quiesce the origin before
  			 * taking the snap.  You want to do this anyway to
  			 * ensure a consistent application view
  			 * (i.e. lockfs).
  			 *
  			 * More distant ancestors are irrelevant. The
  			 * shared flag will be set in their case.
  			 */
  			thin_defer_bio(tc, bio);

  			return DM_MAPIO_SUBMITTED;

  		}

  
  		build_virtual_key(tc->td, block, &key);

  		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1, &cell_result))

  			return DM_MAPIO_SUBMITTED;
  
  		build_data_key(tc->td, result.block, &key);

  		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell2, &cell_result)) {
  			cell_defer_no_holder_no_free(tc, &cell1);

  			return DM_MAPIO_SUBMITTED;
  		}
  
  		inc_all_io_entry(tc->pool, bio);

  		cell_defer_no_holder_no_free(tc, &cell2);
  		cell_defer_no_holder_no_free(tc, &cell1);

  
  		remap(tc, bio, result.block);
  		return DM_MAPIO_REMAPPED;

  
  	case -ENODATA:

  		if (get_pool_mode(tc->pool) == PM_READ_ONLY) {
  			/*
  			 * This block isn't provisioned, and we have no way

  			 * of doing so.

  			 */

  			handle_unserviceable_bio(tc->pool, bio);

  			return DM_MAPIO_SUBMITTED;

  		}
  		/* fall through */
  
  	case -EWOULDBLOCK:

  		/*
  		 * In future, the failed dm_thin_find_block above could
  		 * provide the hint to load the metadata into cache.
  		 */

  		thin_defer_bio(tc, bio);

  		return DM_MAPIO_SUBMITTED;

  
  	default:
  		/*
  		 * Must always call bio_io_error on failure.
  		 * dm_thin_find_block can fail with -EINVAL if the
  		 * pool is switched to fail-io mode.
  		 */
  		bio_io_error(bio);

  		return DM_MAPIO_SUBMITTED;

  	}

  }
  
  static int pool_is_congested(struct dm_target_callbacks *cb, int bdi_bits)
  {

  	struct pool_c *pt = container_of(cb, struct pool_c, callbacks);