/*
 * 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;
        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(struct pool *pool,
                       struct dm_bio_prison_cell *cell)
{
        dm_cell_error(pool->prison, cell);
        dm_bio_prison_free_cell(pool->prison, cell);
}

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

/*
 * 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 quiesced:1;
        bool prepared:1;
        bool pass_discard:1;
        bool definitely_not_shared:1;

        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 __maybe_add_mapping(struct dm_thin_new_mapping *m)
{
        struct pool *pool = m->tc->pool;

        if (m->quiesced && m->prepared) {
                list_add_tail(&m->list, &pool->prepared_mappings);
                wake_worker(pool);
        }
}

static void copy_complete(int read_err, unsigned long write_err, void *context)
{
        unsigned long flags;
        struct dm_thin_new_mapping *m = context;
        struct pool *pool = m->tc->pool;

        m->err = read_err || write_err ? -EIO : 0;

        spin_lock_irqsave(&pool->lock, flags);
        m->prepared = true;
        __maybe_add_mapping(m);
        spin_unlock_irqrestore(&pool->lock, flags);
}

static void overwrite_endio(struct bio *bio, int err)
{
        unsigned long flags;
        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;
        struct pool *pool = m->tc->pool;

        m->err = err;

        spin_lock_irqsave(&pool->lock, flags);
        m->prepared = true;
        __maybe_add_mapping(m);
        spin_unlock_irqrestore(&pool->lock, flags);
}

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

/*
 * 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 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)
{
        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;

        if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
                m->quiesced = true;

        /*
         * 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 = pool->sectors_per_block;

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

                r = dm_kcopyd_copy(pool->copier, &from, 1, &to,
                                   0, copy_complete, m);
                if (r < 0) {
                        mempool_free(m, pool->mapping_pool);
                        DMERR_LIMIT("dm_kcopyd_copy() failed");
                        cell_error(pool, cell);
                }
        }
}

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);
}

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)
{
        schedule_copy(tc, virt_block, tc->origin_dev,
                      virt_block, data_dest, cell, bio);
}

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);

        m->quiesced = true;
        m->prepared = false;
        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 {
                int r;
                struct dm_io_region to;

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

                r = dm_kcopyd_zero(pool->copier, 1, &to, 0, copy_complete, m);
                if (r < 0) {
                        mempool_free(m, pool->mapping_pool);
                        DMERR_LIMIT("dm_kcopyd_zero() failed");
                        cell_error(pool, cell);
                }
        }
}

/*
 * 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 bool 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 true;

        case PM_OUT_OF_DATA_SPACE:
                return pool->pf.error_if_no_space;

        case PM_READ_ONLY:
        case PM_FAIL:
                return true;
        default:
                /* Shouldn't get here */
                DMERR_LIMIT("bio unserviceable, yet pool has an unknown mode");
                return true;
        }
}

static void handle_unserviceable_bio(struct pool *pool, struct bio *bio)
{
        if (should_error_unserviceable_bio(pool))
                bio_io_error(bio);
        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;

        if (should_error_unserviceable_bio(pool)) {
                cell_error(pool, cell);
                return;
        }

        bio_list_init(&bios);
        cell_release(pool, cell, &bios);

        if (should_error_unserviceable_bio(pool))
                while ((bio = bio_list_pop(&bios)))
                        bio_io_error(bio);
        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);

                        remap_to_origin_and_issue(tc, bio);
                } 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 noflush_work {
        struct work_struct worker;
        struct thin_c *tc;

        atomic_t complete;
        wait_queue_head_t wait;
};

static void complete_noflush_work(struct noflush_work *w)
{
        atomic_set(&w->complete, 1);
        wake_up(&w->wait);
}

static void do_noflush_start(struct work_struct *ws)
{
        struct noflush_work *w = container_of(ws, struct noflush_work, worker);
        w->tc->requeue_mode = true;
        requeue_io(w->tc);
        complete_noflush_work(w);
}

static void do_noflush_stop(struct work_struct *ws)
{
        struct noflush_work *w = container_of(ws, struct noflush_work, worker);
        w->tc->requeue_mode = false;
        complete_noflush_work(w);
}

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

        INIT_WORK_ONSTACK(&w.worker, fn);
        w.tc = tc;
        atomic_set(&w.complete, 0);
        init_waitqueue_head(&w.wait);

        queue_work(tc->pool->wq, &w.worker);

        wait_event(w.wait, atomic_read(&w.complete));
}

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

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);
        struct request_queue *q;

        if (get_pool_mode(pt->pool) == PM_OUT_OF_DATA_SPACE)
                return 1;

        q = bdev_get_queue(pt->data_dev->bdev);
        return bdi_congested(&q->backing_dev_info, bdi_bits);
}

static void requeue_bios(struct pool *pool)
{
        unsigned long flags;
        struct thin_c *tc;

        rcu_read_lock();
        list_for_each_entry_rcu(tc, &pool->active_thins, list) {
                spin_lock_irqsave(&tc->lock, flags);
                bio_list_merge(&tc->deferred_bio_list, &tc->retry_on_resume_list);
                bio_list_init(&tc->retry_on_resume_list);
                spin_unlock_irqrestore(&tc->lock, flags);
        }
        rcu_read_unlock();
}

/*----------------------------------------------------------------
 * Binding of control targets to a pool object
 *--------------------------------------------------------------*/
static bool data_dev_supports_discard(struct pool_c *pt)
{
        struct request_queue *q = bdev_get_queue(pt->data_dev->bdev);

        return q && blk_queue_discard(q);
}

static bool is_factor(sector_t block_size, uint32_t n)
{
        return !sector_div(block_size, n);
}

/*
 * If discard_passdown was enabled verify that the data device
 * supports discards.  Disable discard_passdown if not.
 */
static void disable_passdown_if_not_supported(struct pool_c *pt)
{
        struct pool *pool = pt->pool;
        struct block_device *data_bdev = pt->data_dev->bdev;
        struct queue_limits *data_limits = &bdev_get_queue(data_bdev)->limits;
        sector_t block_size = pool->sectors_per_block << SECTOR_SHIFT;
        const char *reason = NULL;
        char buf[BDEVNAME_SIZE];

        if (!pt->adjusted_pf.discard_passdown)
                return;

        if (!data_dev_supports_discard(pt))
                reason = "discard unsupported";

        else if (data_limits->max_discard_sectors < pool->sectors_per_block)
                reason = "max discard sectors smaller than a block";

        else if (data_limits->discard_granularity > block_size)
                reason = "discard granularity larger than a block";

        else if (!is_factor(block_size, data_limits->discard_granularity))
                reason = "discard granularity not a factor of block size";

        if (reason) {
                DMWARN("Data device (%s) %s: Disabling discard passdown.", bdevname(data_bdev, buf), reason);
                pt->adjusted_pf.discard_passdown = false;
        }
}

static int bind_control_target(struct pool *pool, struct dm_target *ti)
{
        struct pool_c *pt = ti->private;

        /*
         * We want to make sure that a pool in PM_FAIL mode is never upgraded.
         */
        enum pool_mode old_mode = get_pool_mode(pool);
        enum pool_mode new_mode = pt->adjusted_pf.mode;

        /*
         * Don't change the pool's mode until set_pool_mode() below.
         * Otherwise the pool's process_* function pointers may
         * not match the desired pool mode.
         */
        pt->adjusted_pf.mode = old_mode;

        pool->ti = ti;
        pool->pf = pt->adjusted_pf;
        pool->low_water_blocks = pt->low_water_blocks;

        set_pool_mode(pool, new_mode);

        return 0;
}

static void unbind_control_target(struct pool *pool, struct dm_target *ti)
{
        if (pool->ti == ti)
                pool->ti = NULL;
}

/*----------------------------------------------------------------
 * Pool creation
 *--------------------------------------------------------------*/
/* Initialize pool features. */
static void pool_features_init(struct pool_features *pf)
{
        pf->mode = PM_WRITE;
        pf->zero_new_blocks = true;
        pf->discard_enabled = true;
        pf->discard_passdown = true;
        pf->error_if_no_space = false;
}

static void __pool_destroy(struct pool *pool)
{
        __pool_table_remove(pool);

        if (dm_pool_metadata_close(pool->pmd) < 0)
                DMWARN("%s: dm_pool_metadata_close() failed.", __func__);

        dm_bio_prison_destroy(pool->prison);
        dm_kcopyd_client_destroy(pool->copier);

        if (pool->wq)
                destroy_workqueue(pool->wq);

        if (pool->next_mapping)
                mempool_free(pool->next_mapping, pool->mapping_pool);
        mempool_destroy(pool->mapping_pool);
        dm_deferred_set_destroy(pool->shared_read_ds);
        dm_deferred_set_destroy(pool->all_io_ds);
        kfree(pool);
}

static struct kmem_cache *_new_mapping_cache;

static struct pool *pool_create(struct mapped_device *pool_md,
                                struct block_device *metadata_dev,
                                unsigned long block_size,
                                int read_only, char **error)
{
        int r;
        void *err_p;
        struct pool *pool;
        struct dm_pool_metadata *pmd;
        bool format_device = read_only ? false : true;

        pmd = dm_pool_metadata_open(metadata_dev, block_size, format_device);
        if (IS_ERR(pmd)) {
                *error = "Error creating metadata object";
                return (struct pool *)pmd;
        }

        pool = kmalloc(sizeof(*pool), GFP_KERNEL);
        if (!pool) {
                *error = "Error allocating memory for pool";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_pool;
        }

        pool->pmd = pmd;
        pool->sectors_per_block = block_size;
        if (block_size & (block_size - 1))
                pool->sectors_per_block_shift = -1;
        else
                pool->sectors_per_block_shift = __ffs(block_size);
        pool->low_water_blocks = 0;
        pool_features_init(&pool->pf);
        pool->prison = dm_bio_prison_create(PRISON_CELLS);
        if (!pool->prison) {
                *error = "Error creating pool's bio prison";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_prison;
        }

        pool->copier = dm_kcopyd_client_create(&dm_kcopyd_throttle);
        if (IS_ERR(pool->copier)) {
                r = PTR_ERR(pool->copier);
                *error = "Error creating pool's kcopyd client";
                err_p = ERR_PTR(r);
                goto bad_kcopyd_client;
        }

        /*
         * Create singlethreaded workqueue that will service all devices
         * that use this metadata.
         */
        pool->wq = alloc_ordered_workqueue("dm-" DM_MSG_PREFIX, WQ_MEM_RECLAIM);
        if (!pool->wq) {
                *error = "Error creating pool's workqueue";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_wq;
        }

        INIT_WORK(&pool->worker, do_worker);
        INIT_DELAYED_WORK(&pool->waker, do_waker);
        INIT_DELAYED_WORK(&pool->no_space_timeout, do_no_space_timeout);
        spin_lock_init(&pool->lock);
        bio_list_init(&pool->deferred_flush_bios);
        INIT_LIST_HEAD(&pool->prepared_mappings);
        INIT_LIST_HEAD(&pool->prepared_discards);
        INIT_LIST_HEAD(&pool->active_thins);
        pool->low_water_triggered = false;

        pool->shared_read_ds = dm_deferred_set_create();
        if (!pool->shared_read_ds) {
                *error = "Error creating pool's shared read deferred set";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_shared_read_ds;
        }

        pool->all_io_ds = dm_deferred_set_create();
        if (!pool->all_io_ds) {
                *error = "Error creating pool's all io deferred set";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_all_io_ds;
        }

        pool->next_mapping = NULL;
        pool->mapping_pool = mempool_create_slab_pool(MAPPING_POOL_SIZE,
                                                      _new_mapping_cache);
        if (!pool->mapping_pool) {
                *error = "Error creating pool's mapping mempool";
                err_p = ERR_PTR(-ENOMEM);
                goto bad_mapping_pool;
        }

        pool->ref_count = 1;
        pool->last_commit_jiffies = jiffies;
        pool->pool_md = pool_md;
        pool->md_dev = metadata_dev;
        __pool_table_insert(pool);

        return pool;

bad_mapping_pool:
        dm_deferred_set_destroy(pool->all_io_ds);
bad_all_io_ds:
        dm_deferred_set_destroy(pool->shared_read_ds);
bad_shared_read_ds:
        destroy_workqueue(pool->wq);
bad_wq:
        dm_kcopyd_client_destroy(pool->copier);
bad_kcopyd_client:
        dm_bio_prison_destroy(pool->prison);
bad_prison:
        kfree(pool);
bad_pool:
        if (dm_pool_metadata_close(pmd))
                DMWARN("%s: dm_pool_metadata_close() failed.", __func__);

        return err_p;
}

static void __pool_inc(struct pool *pool)
{
        BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
        pool->ref_count++;
}

static void __pool_dec(struct pool *pool)
{
        BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
        BUG_ON(!pool->ref_count);
        if (!--pool->ref_count)
                __pool_destroy(pool);
}

static struct pool *__pool_find(struct mapped_device *pool_md,
                                struct block_device *metadata_dev,
                                unsigned long block_size, int read_only,
                                char **error, int *created)
{
        struct pool *pool = __pool_table_lookup_metadata_dev(metadata_dev);

        if (pool) {
                if (pool->pool_md != pool_md) {
                        *error = "metadata device already in use by a pool";
                        return ERR_PTR(-EBUSY);
                }
                __pool_inc(pool);

        } else {
                pool = __pool_table_lookup(pool_md);
                if (pool) {
                        if (pool->md_dev != metadata_dev) {
                                *error = "different pool cannot replace a pool";
                                return ERR_PTR(-EINVAL);
                        }
                        __pool_inc(pool);

                } else {
                        pool = pool_create(pool_md, metadata_dev, block_size, read_only, error);
                        *created = 1;
                }
        }

        return pool;
}

/*----------------------------------------------------------------
 * Pool target methods
 *--------------------------------------------------------------*/
static void pool_dtr(struct dm_target *ti)
{
        struct pool_c *pt = ti->private;

        mutex_lock(&dm_thin_pool_table.mutex);

        unbind_control_target(pt->pool, ti);
        __pool_dec(pt->pool);
        dm_put_device(ti, pt->metadata_dev);
        dm_put_device(ti, pt->data_dev);
        kfree(pt);

        mutex_unlock(&dm_thin_pool_table.mutex);
}

static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
                               struct dm_target *ti)
{
        int r;
        unsigned argc;
        const char *arg_name;

        static struct dm_arg _args[] = {
                {0, 4, "Invalid number of pool 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, "skip_block_zeroing"))
                        pf->zero_new_blocks = false;

                else if (!strcasecmp(arg_name, "ignore_discard"))
                        pf->discard_enabled = false;

                else if (!strcasecmp(arg_name, "no_discard_passdown"))
                        pf->discard_passdown = false;

                else if (!strcasecmp(arg_name, "read_only"))
                        pf->mode = PM_READ_ONLY;

                else if (!strcasecmp(arg_name, "error_if_no_space"))
                        pf->error_if_no_space = true;

                else {
                        ti->error = "Unrecognised pool feature requested";
                        r = -EINVAL;
                        break;
                }
        }

        return r;
}

static void metadata_low_callback(void *context)
{
        struct pool *pool = context;

        DMWARN("%s: reached low water mark for metadata device: sending event.",
               dm_device_name(pool->pool_md));

        dm_table_event(pool->ti->table);
}

static sector_t get_dev_size(struct block_device *bdev)
{
        return i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
}

static void warn_if_metadata_device_too_big(struct block_device *bdev)
{
        sector_t metadata_dev_size = get_dev_size(bdev);
        char buffer[BDEVNAME_SIZE];

        if (metadata_dev_size > THIN_METADATA_MAX_SECTORS_WARNING)
                DMWARN("Metadata device %s is larger than %u sectors: excess space will not be used.",
                       bdevname(bdev, buffer), THIN_METADATA_MAX_SECTORS);
}

static sector_t get_metadata_dev_size(struct block_device *bdev)
{
        sector_t metadata_dev_size = get_dev_size(bdev);

        if (metadata_dev_size > THIN_METADATA_MAX_SECTORS)
                metadata_dev_size = THIN_METADATA_MAX_SECTORS;

        return metadata_dev_size;
}

static dm_block_t get_metadata_dev_size_in_blocks(struct block_device *bdev)
{
        sector_t metadata_dev_size = get_metadata_dev_size(bdev);

        sector_div(metadata_dev_size, THIN_METADATA_BLOCK_SIZE);

        return metadata_dev_size;
}

/*
 * When a metadata threshold is crossed a dm event is triggered, and
 * userland should respond by growing the metadata device.  We could let
 * userland set the threshold, like we do with the data threshold, but I'm
 * not sure they know enough to do this well.
 */
static dm_block_t calc_metadata_threshold(struct pool_c *pt)
{
        /*
         * 4M is ample for all ops with the possible exception of thin
         * device deletion which is harmless if it fails (just retry the
         * delete after you've grown the device).
         */
        dm_block_t quarter = get_metadata_dev_size_in_blocks(pt->metadata_dev->bdev) / 4;
        return min((dm_block_t)1024ULL /* 4M */, quarter);
}

/*
 * thin-pool <metadata dev> <data dev>
 *           <data block size (sectors)>
 *           <low water mark (blocks)>
 *           [<#feature args> [<arg>]*]
 *
 * Optional feature arguments are:
 *           skip_block_zeroing: skips the zeroing of newly-provisioned blocks.
 *           ignore_discard: disable discard
 *           no_discard_passdown: don't pass discards down to the data device
 *           read_only: Don't allow any changes to be made to the pool metadata.
 *           error_if_no_space: error IOs, instead of queueing, if no space.
 */
static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
{
        int r, pool_created = 0;
        struct pool_c *pt;
        struct pool *pool;
        struct pool_features pf;
        struct dm_arg_set as;
        struct dm_dev *data_dev;
        unsigned long block_size;
        dm_block_t low_water_blocks;
        struct dm_dev *metadata_dev;
        fmode_t metadata_mode;

        /*
         * FIXME Remove validation from scope of lock.
         */
        mutex_lock(&dm_thin_pool_table.mutex);

        if (argc < 4) {
                ti->error = "Invalid argument count";
                r = -EINVAL;
                goto out_unlock;
        }

        as.argc = argc;
        as.argv = argv;

        /*
         * Set default pool features.
         */
        pool_features_init(&pf);

        dm_consume_args(&as, 4);
        r = parse_pool_features(&as, &pf, ti);
        if (r)
                goto out_unlock;

        metadata_mode = FMODE_READ | ((pf.mode == PM_READ_ONLY) ? 0 : FMODE_WRITE);
        r = dm_get_device(ti, argv[0], metadata_mode, &metadata_dev);
        if (r) {
                ti->error = "Error opening metadata block device";
                goto out_unlock;
        }
        warn_if_metadata_device_too_big(metadata_dev->bdev);

        r = dm_get_device(ti, argv[1], FMODE_READ | FMODE_WRITE, &data_dev);
        if (r) {
                ti->error = "Error getting data device";
                goto out_metadata;
        }

        if (kstrtoul(argv[2], 10, &block_size) || !block_size ||
            block_size < DATA_DEV_BLOCK_SIZE_MIN_SECTORS ||
            block_size > DATA_DEV_BLOCK_SIZE_MAX_SECTORS ||
            block_size & (DATA_DEV_BLOCK_SIZE_MIN_SECTORS - 1)) {
                ti->error = "Invalid block size";
                r = -EINVAL;
                goto out;
        }

        if (kstrtoull(argv[3], 10, (unsigned long long *)&low_water_blocks)) {
                ti->error = "Invalid low water mark";
                r = -EINVAL;
                goto out;
        }

        pt = kzalloc(sizeof(*pt), GFP_KERNEL);
        if (!pt) {
                r = -ENOMEM;
                goto out;
        }

        pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev,
                           block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created);
        if (IS_ERR(pool)) {
                r = PTR_ERR(pool);
                goto out_free_pt;
        }

        /*
         * 'pool_created' reflects whether this is the first table load.
         * Top level discard support is not allowed to be changed after
         * initial load.  This would require a pool reload to trigger thin
         * device changes.
         */
        if (!pool_created && pf.discard_enabled != pool->pf.discard_enabled) {
                ti->error = "Discard support cannot be disabled once enabled";
                r = -EINVAL;
                goto out_flags_changed;
        }

        pt->pool = pool;
        pt->ti = ti;
        pt->metadata_dev = metadata_dev;
        pt->data_dev = data_dev;
        pt->low_water_blocks = low_water_blocks;
        pt->adjusted_pf = pt->requested_pf = pf;
        ti->num_flush_bios = 1;

        /*
         * Only need to enable discards if the pool should pass
         * them down to the data device.  The thin device's discard
         * processing will cause mappings to be removed from the btree.
         */
        ti->discard_zeroes_data_unsupported = true;
        if (pf.discard_enabled && pf.discard_passdown) {
                ti->num_discard_bios = 1;

                /*
                 * Setting 'discards_supported' circumvents the normal
                 * stacking of discard limits (this keeps the pool and
                 * thin devices' discard limits consistent).
                 */
                ti->discards_supported = true;
        }
        ti->private = pt;

        r = dm_pool_register_metadata_threshold(pt->pool->pmd,
                                                calc_metadata_threshold(pt),
                                                metadata_low_callback,
                                                pool);
        if (r)
                goto out_free_pt;

        pt->callbacks.congested_fn = pool_is_congested;
        dm_table_add_target_callbacks(ti->table, &pt->callbacks);

        mutex_unlock(&dm_thin_pool_table.mutex);

        return 0;

out_flags_changed:
        __pool_dec(pool);
out_free_pt:
        kfree(pt);
out:
        dm_put_device(ti, data_dev);
out_metadata:
        dm_put_device(ti, metadata_dev);
out_unlock:
        mutex_unlock(&dm_thin_pool_table.mutex);

        return r;
}

static int pool_map(struct dm_target *ti, struct bio *bio)
{
        int r;
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;
        unsigned long flags;

        /*
         * As this is a singleton target, ti->begin is always zero.
         */
        spin_lock_irqsave(&pool->lock, flags);
        bio->bi_bdev = pt->data_dev->bdev;
        r = DM_MAPIO_REMAPPED;
        spin_unlock_irqrestore(&pool->lock, flags);

        return r;
}

static int maybe_resize_data_dev(struct dm_target *ti, bool *need_commit)
{
        int r;
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;
        sector_t data_size = ti->len;
        dm_block_t sb_data_size;

        *need_commit = false;

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

        r = dm_pool_get_data_dev_size(pool->pmd, &sb_data_size);
        if (r) {
                DMERR("%s: failed to retrieve data device size",
                      dm_device_name(pool->pool_md));
                return r;
        }

        if (data_size < sb_data_size) {
                DMERR("%s: pool target (%llu blocks) too small: expected %llu",
                      dm_device_name(pool->pool_md),
                      (unsigned long long)data_size, sb_data_size);
                return -EINVAL;

        } else if (data_size > sb_data_size) {
                if (dm_pool_metadata_needs_check(pool->pmd)) {
                        DMERR("%s: unable to grow the data device until repaired.",
                              dm_device_name(pool->pool_md));
                        return 0;
                }

                if (sb_data_size)
                        DMINFO("%s: growing the data device from %llu to %llu blocks",
                               dm_device_name(pool->pool_md),
                               sb_data_size, (unsigned long long)data_size);
                r = dm_pool_resize_data_dev(pool->pmd, data_size);
                if (r) {
                        metadata_operation_failed(pool, "dm_pool_resize_data_dev", r);
                        return r;
                }

                *need_commit = true;
        }

        return 0;
}

static int maybe_resize_metadata_dev(struct dm_target *ti, bool *need_commit)
{
        int r;
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;
        dm_block_t metadata_dev_size, sb_metadata_dev_size;

        *need_commit = false;

        metadata_dev_size = get_metadata_dev_size_in_blocks(pool->md_dev);

        r = dm_pool_get_metadata_dev_size(pool->pmd, &sb_metadata_dev_size);
        if (r) {
                DMERR("%s: failed to retrieve metadata device size",
                      dm_device_name(pool->pool_md));
                return r;
        }

        if (metadata_dev_size < sb_metadata_dev_size) {
                DMERR("%s: metadata device (%llu blocks) too small: expected %llu",
                      dm_device_name(pool->pool_md),
                      metadata_dev_size, sb_metadata_dev_size);
                return -EINVAL;

        } else if (metadata_dev_size > sb_metadata_dev_size) {
                if (dm_pool_metadata_needs_check(pool->pmd)) {
                        DMERR("%s: unable to grow the metadata device until repaired.",
                              dm_device_name(pool->pool_md));
                        return 0;
                }

                warn_if_metadata_device_too_big(pool->md_dev);
                DMINFO("%s: growing the metadata device from %llu to %llu blocks",
                       dm_device_name(pool->pool_md),
                       sb_metadata_dev_size, metadata_dev_size);
                r = dm_pool_resize_metadata_dev(pool->pmd, metadata_dev_size);
                if (r) {
                        metadata_operation_failed(pool, "dm_pool_resize_metadata_dev", r);
                        return r;
                }

                *need_commit = true;
        }

        return 0;
}

/*
 * Retrieves the number of blocks of the data device from
 * the superblock and compares it to the actual device size,
 * thus resizing the data device in case it has grown.
 *
 * This both copes with opening preallocated data devices in the ctr
 * being followed by a resume
 * -and-
 * calling the resume method individually after userspace has
 * grown the data device in reaction to a table event.
 */
static int pool_preresume(struct dm_target *ti)
{
        int r;
        bool need_commit1, need_commit2;
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;

        /*
         * Take control of the pool object.
         */
        r = bind_control_target(pool, ti);
        if (r)
                return r;

        r = maybe_resize_data_dev(ti, &need_commit1);
        if (r)
                return r;

        r = maybe_resize_metadata_dev(ti, &need_commit2);
        if (r)
                return r;

        if (need_commit1 || need_commit2)
                (void) commit(pool);

        return 0;
}

static void pool_resume(struct dm_target *ti)
{
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;
        unsigned long flags;

        spin_lock_irqsave(&pool->lock, flags);
        pool->low_water_triggered = false;
        spin_unlock_irqrestore(&pool->lock, flags);
        requeue_bios(pool);

        do_waker(&pool->waker.work);
}

static void pool_postsuspend(struct dm_target *ti)
{
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;

        cancel_delayed_work(&pool->waker);
        cancel_delayed_work(&pool->no_space_timeout);
        flush_workqueue(pool->wq);
        (void) commit(pool);
}

static int check_arg_count(unsigned argc, unsigned args_required)
{
        if (argc != args_required) {
                DMWARN("Message received with %u arguments instead of %u.",
                       argc, args_required);
                return -EINVAL;
        }

        return 0;
}

static int read_dev_id(char *arg, dm_thin_id *dev_id, int warning)
{
        if (!kstrtoull(arg, 10, (unsigned long long *)dev_id) &&
            *dev_id <= MAX_DEV_ID)
                return 0;

        if (warning)
                DMWARN("Message received with invalid device id: %s", arg);

        return -EINVAL;
}

static int process_create_thin_mesg(unsigned argc, char **argv, struct pool *pool)
{
        dm_thin_id dev_id;
        int r;

        r = check_arg_count(argc, 2);
        if (r)
                return r;

        r = read_dev_id(argv[1], &dev_id, 1);
        if (r)
                return r;

        r = dm_pool_create_thin(pool->pmd, dev_id);
        if (r) {
                DMWARN("Creation of new thinly-provisioned device with id %s failed.",
                       argv[1]);
                return r;
        }

        return 0;
}

static int process_create_snap_mesg(unsigned argc, char **argv, struct pool *pool)
{
        dm_thin_id dev_id;
        dm_thin_id origin_dev_id;
        int r;

        r = check_arg_count(argc, 3);
        if (r)
                return r;

        r = read_dev_id(argv[1], &dev_id, 1);
        if (r)
                return r;

        r = read_dev_id(argv[2], &origin_dev_id, 1);
        if (r)
                return r;

        r = dm_pool_create_snap(pool->pmd, dev_id, origin_dev_id);
        if (r) {
                DMWARN("Creation of new snapshot %s of device %s failed.",
                       argv[1], argv[2]);
                return r;
        }

        return 0;
}

static int process_delete_mesg(unsigned argc, char **argv, struct pool *pool)
{
        dm_thin_id dev_id;
        int r;

        r = check_arg_count(argc, 2);
        if (r)
                return r;

        r = read_dev_id(argv[1], &dev_id, 1);
        if (r)
                return r;

        r = dm_pool_delete_thin_device(pool->pmd, dev_id);
        if (r)
                DMWARN("Deletion of thin device %s failed.", argv[1]);

        return r;
}

static int process_set_transaction_id_mesg(unsigned argc, char **argv, struct pool *pool)
{
        dm_thin_id old_id, new_id;
        int r;

        r = check_arg_count(argc, 3);
        if (r)
                return r;

        if (kstrtoull(argv[1], 10, (unsigned long long *)&old_id)) {
                DMWARN("set_transaction_id message: Unrecognised id %s.", argv[1]);
                return -EINVAL;
        }

        if (kstrtoull(argv[2], 10, (unsigned long long *)&new_id)) {
                DMWARN("set_transaction_id message: Unrecognised new id %s.", argv[2]);
                return -EINVAL;
        }

        r = dm_pool_set_metadata_transaction_id(pool->pmd, old_id, new_id);
        if (r) {
                DMWARN("Failed to change transaction id from %s to %s.",
                       argv[1], argv[2]);
                return r;
        }

        return 0;
}

static int process_reserve_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
{
        int r;

        r = check_arg_count(argc, 1);
        if (r)
                return r;

        (void) commit(pool);

        r = dm_pool_reserve_metadata_snap(pool->pmd);
        if (r)
                DMWARN("reserve_metadata_snap message failed.");

        return r;
}

static int process_release_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
{
        int r;

        r = check_arg_count(argc, 1);
        if (r)
                return r;

        r = dm_pool_release_metadata_snap(pool->pmd);
        if (r)
                DMWARN("release_metadata_snap message failed.");

        return r;
}

/*
 * Messages supported:
 *   create_thin        <dev_id>
 *   create_snap        <dev_id> <origin_id>
 *   delete             <dev_id>
 *   trim               <dev_id> <new_size_in_sectors>
 *   set_transaction_id <current_trans_id> <new_trans_id>
 *   reserve_metadata_snap
 *   release_metadata_snap
 */
static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
{
        int r = -EINVAL;
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;

        if (!strcasecmp(argv[0], "create_thin"))
                r = process_create_thin_mesg(argc, argv, pool);

        else if (!strcasecmp(argv[0], "create_snap"))
                r = process_create_snap_mesg(argc, argv, pool);

        else if (!strcasecmp(argv[0], "delete"))
                r = process_delete_mesg(argc, argv, pool);

        else if (!strcasecmp(argv[0], "set_transaction_id"))
                r = process_set_transaction_id_mesg(argc, argv, pool);

        else if (!strcasecmp(argv[0], "reserve_metadata_snap"))
                r = process_reserve_metadata_snap_mesg(argc, argv, pool);

        else if (!strcasecmp(argv[0], "release_metadata_snap"))
                r = process_release_metadata_snap_mesg(argc, argv, pool);

        else
                DMWARN("Unrecognised thin pool target message received: %s", argv[0]);

        if (!r)
                (void) commit(pool);

        return r;
}

static void emit_flags(struct pool_features *pf, char *result,
                       unsigned sz, unsigned maxlen)
{
        unsigned count = !pf->zero_new_blocks + !pf->discard_enabled +
                !pf->discard_passdown + (pf->mode == PM_READ_ONLY) +
                pf->error_if_no_space;
        DMEMIT("%u ", count);

        if (!pf->zero_new_blocks)
                DMEMIT("skip_block_zeroing ");

        if (!pf->discard_enabled)
                DMEMIT("ignore_discard ");

        if (!pf->discard_passdown)
                DMEMIT("no_discard_passdown ");

        if (pf->mode == PM_READ_ONLY)
                DMEMIT("read_only ");

        if (pf->error_if_no_space)
                DMEMIT("error_if_no_space ");
}

/*
 * Status line is:
 *    <transaction id> <used metadata sectors>/<total metadata sectors>
 *    <used data sectors>/<total data sectors> <held metadata root>
 */
static void pool_status(struct dm_target *ti, status_type_t type,
                        unsigned status_flags, char *result, unsigned maxlen)
{
        int r;
        unsigned sz = 0;
        uint64_t transaction_id;
        dm_block_t nr_free_blocks_data;
        dm_block_t nr_free_blocks_metadata;
        dm_block_t nr_blocks_data;
        dm_block_t nr_blocks_metadata;
        dm_block_t held_root;
        char buf[BDEVNAME_SIZE];
        char buf2[BDEVNAME_SIZE];
        struct pool_c *pt = ti->private;
        struct pool *pool = pt->pool;

        switch (type) {
        case STATUSTYPE_INFO:
                if (get_pool_mode(pool) == PM_FAIL) {
                        DMEMIT("Fail");
                        break;
                }

                /* Commit to ensure statistics aren't out-of-date */
                if (!(status_flags & DM_STATUS_NOFLUSH_FLAG) && !dm_suspended(ti))
                        (void) commit(pool);

                r = dm_pool_get_metadata_transaction_id(pool->pmd, &transaction_id);
                if (r) {
                        DMERR("%s: dm_pool_get_metadata_transaction_id returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                r = dm_pool_get_free_metadata_block_count(pool->pmd, &nr_free_blocks_metadata);
                if (r) {
                        DMERR("%s: dm_pool_get_free_metadata_block_count returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                r = dm_pool_get_metadata_dev_size(pool->pmd, &nr_blocks_metadata);
                if (r) {
                        DMERR("%s: dm_pool_get_metadata_dev_size returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                r = dm_pool_get_free_block_count(pool->pmd, &nr_free_blocks_data);
                if (r) {
                        DMERR("%s: dm_pool_get_free_block_count returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                r = dm_pool_get_data_dev_size(pool->pmd, &nr_blocks_data);
                if (r) {
                        DMERR("%s: dm_pool_get_data_dev_size returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                r = dm_pool_get_metadata_snap(pool->pmd, &held_root);
                if (r) {
                        DMERR("%s: dm_pool_get_metadata_snap returned %d",
                              dm_device_name(pool->pool_md), r);
                        goto err;
                }

                DMEMIT("%llu %llu/%llu %llu/%llu ",
                       (unsigned long long)transaction_id,
                       (unsigned long long)(nr_blocks_metadata - nr_free_blocks_metadata),
                       (unsigned long long)nr_blocks_metadata,
                       (unsigned long long)(nr_blocks_data - nr_free_blocks_data),
                       (unsigned long long)nr_blocks_data);

                if (held_root)
                        DMEMIT("%llu ", held_root);
                else
                        DMEMIT("- ");