#include <linux/device-mapper.h>
#include <linux/dm-io.h>
#include <linux/dm-kcopyd.h>
+#include <linux/log2.h>
#include <linux/list.h>
#include <linux/rculist.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/slab.h>
+#include <linux/sort.h>
#include <linux/rbtree.h>
#define DM_MSG_PREFIX "thin"
*/
#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
/*----------------------------------------------------------------*/
+#define THROTTLE_THRESHOLD (1 * HZ)
+
+struct throttle {
+ struct rw_semaphore lock;
+ unsigned long threshold;
+ bool throttle_applied;
+};
+
+static void throttle_init(struct throttle *t)
+{
+ init_rwsem(&t->lock);
+ t->throttle_applied = false;
+}
+
+static void throttle_work_start(struct throttle *t)
+{
+ t->threshold = jiffies + THROTTLE_THRESHOLD;
+}
+
+static void throttle_work_update(struct throttle *t)
+{
+ if (!t->throttle_applied && jiffies > t->threshold) {
+ down_write(&t->lock);
+ t->throttle_applied = true;
+ }
+}
+
+static void throttle_work_complete(struct throttle *t)
+{
+ if (t->throttle_applied) {
+ t->throttle_applied = false;
+ up_write(&t->lock);
+ }
+}
+
+static void throttle_lock(struct throttle *t)
+{
+ down_read(&t->lock);
+}
+
+static void throttle_unlock(struct throttle *t)
+{
+ up_read(&t->lock);
+}
+
+/*----------------------------------------------------------------*/
+
/*
* A pool device ties together a metadata device and a data device. It
* also provides the interface for creating and destroying internal
struct thin_c;
typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
+typedef void (*process_cell_fn)(struct thin_c *tc, struct dm_bio_prison_cell *cell);
typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);
+#define CELL_SORT_ARRAY_SIZE 8192
+
struct pool {
struct list_head list;
struct dm_target *ti; /* Only set if a pool target is bound */
struct dm_kcopyd_client *copier;
struct workqueue_struct *wq;
+ struct throttle throttle;
struct work_struct worker;
struct delayed_work waker;
struct delayed_work no_space_timeout;
process_bio_fn process_bio;
process_bio_fn process_discard;
+ process_cell_fn process_cell;
+ process_cell_fn process_discard_cell;
+
process_mapping_fn process_prepared_mapping;
process_mapping_fn process_prepared_discard;
+
+ struct dm_bio_prison_cell *cell_sort_array[CELL_SORT_ARRAY_SIZE];
};
static enum pool_mode get_pool_mode(struct pool *pool);
struct list_head list;
struct dm_dev *pool_dev;
struct dm_dev *origin_dev;
+ sector_t origin_size;
dm_thin_id dev_id;
struct pool *pool;
struct dm_thin_device *td;
bool requeue_mode:1;
spinlock_t lock;
+ struct list_head deferred_cells;
struct bio_list deferred_bio_list;
struct bio_list retry_on_resume_list;
struct rb_root sort_bio_list; /* sorted list of deferred bios */
dm_bio_prison_free_cell(pool->prison, cell);
}
+static void cell_visit_release(struct pool *pool,
+ void (*fn)(void *, struct dm_bio_prison_cell *),
+ void *context,
+ struct dm_bio_prison_cell *cell)
+{
+ dm_cell_visit_release(pool->prison, fn, context, cell);
+ 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_bio_prison_free_cell(pool->prison, cell);
}
-static void cell_defer_no_holder_no_free(struct thin_c *tc,
- struct dm_bio_prison_cell *cell)
-{
- struct pool *pool = tc->pool;
- unsigned long flags;
-
- spin_lock_irqsave(&tc->lock, flags);
- dm_cell_release_no_holder(pool->prison, cell, &tc->deferred_bio_list);
- spin_unlock_irqrestore(&tc->lock, flags);
-
- wake_worker(pool);
-}
-
static void cell_error_with_code(struct pool *pool,
struct dm_bio_prison_cell *cell, int error_code)
{
cell_error_with_code(pool, cell, -EIO);
}
+static void cell_success(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+ cell_error_with_code(pool, cell, 0);
+}
+
+static void cell_requeue(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+ cell_error_with_code(pool, cell, DM_ENDIO_REQUEUE);
+}
+
/*----------------------------------------------------------------*/
/*
bio_endio(bio, DM_ENDIO_REQUEUE);
}
+static void requeue_deferred_cells(struct thin_c *tc)
+{
+ struct pool *pool = tc->pool;
+ unsigned long flags;
+ struct list_head cells;
+ struct dm_bio_prison_cell *cell, *tmp;
+
+ INIT_LIST_HEAD(&cells);
+
+ spin_lock_irqsave(&tc->lock, flags);
+ list_splice_init(&tc->deferred_cells, &cells);
+ spin_unlock_irqrestore(&tc->lock, flags);
+
+ list_for_each_entry_safe(cell, tmp, &cells, user_list)
+ cell_requeue(pool, cell);
+}
+
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);
+ requeue_deferred_cells(tc);
}
static void error_thin_retry_list(struct thin_c *tc)
struct dm_thin_new_mapping {
struct list_head list;
- bool quiesced:1;
- bool prepared:1;
bool pass_discard:1;
bool definitely_not_shared:1;
+ /*
+ * Track quiescing, copying and zeroing preparation actions. When this
+ * counter hits zero the block is prepared and can be inserted into the
+ * btree.
+ */
+ atomic_t prepare_actions;
+
int err;
struct thin_c *tc;
dm_block_t virt_block;
bio_end_io_t *saved_bi_end_io;
};
-static void __maybe_add_mapping(struct dm_thin_new_mapping *m)
+static void __complete_mapping_preparation(struct dm_thin_new_mapping *m)
{
struct pool *pool = m->tc->pool;
- if (m->quiesced && m->prepared) {
+ if (atomic_dec_and_test(&m->prepare_actions)) {
list_add_tail(&m->list, &pool->prepared_mappings);
wake_worker(pool);
}
}
-static void copy_complete(int read_err, unsigned long write_err, void *context)
+static void complete_mapping_preparation(struct dm_thin_new_mapping *m)
{
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);
+ __complete_mapping_preparation(m);
spin_unlock_irqrestore(&pool->lock, flags);
}
+static void copy_complete(int read_err, unsigned long write_err, void *context)
+{
+ struct dm_thin_new_mapping *m = context;
+
+ m->err = read_err || write_err ? -EIO : 0;
+ complete_mapping_preparation(m);
+}
+
static void overwrite_endio(struct bio *bio, int err)
{
- 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);
+ complete_mapping_preparation(m);
}
/*----------------------------------------------------------------*/
*/
/*
- * This sends the bios in the cell back to the deferred_bios list.
+ * This sends the bios in the cell, except the original holder, back
+ * to the deferred_bios list.
*/
-static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *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(pool, cell, &tc->deferred_bio_list);
+ cell_release_no_holder(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)
+static void thin_defer_bio(struct thin_c *tc, struct bio *bio);
+
+struct remap_info {
+ struct thin_c *tc;
+ struct bio_list defer_bios;
+ struct bio_list issue_bios;
+};
+
+static void __inc_remap_and_issue_cell(void *context,
+ struct dm_bio_prison_cell *cell)
{
- struct pool *pool = tc->pool;
- unsigned long flags;
+ struct remap_info *info = context;
+ struct bio *bio;
- spin_lock_irqsave(&tc->lock, flags);
- cell_release_no_holder(pool, cell, &tc->deferred_bio_list);
- spin_unlock_irqrestore(&tc->lock, flags);
+ while ((bio = bio_list_pop(&cell->bios))) {
+ if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA))
+ bio_list_add(&info->defer_bios, bio);
+ else {
+ inc_all_io_entry(info->tc->pool, bio);
- wake_worker(pool);
+ /*
+ * We can't issue the bios with the bio prison lock
+ * held, so we add them to a list to issue on
+ * return from this function.
+ */
+ bio_list_add(&info->issue_bios, bio);
+ }
+ }
+}
+
+static void inc_remap_and_issue_cell(struct thin_c *tc,
+ struct dm_bio_prison_cell *cell,
+ dm_block_t block)
+{
+ struct bio *bio;
+ struct remap_info info;
+
+ info.tc = tc;
+ bio_list_init(&info.defer_bios);
+ bio_list_init(&info.issue_bios);
+
+ /*
+ * We have to be careful to inc any bios we're about to issue
+ * before the cell is released, and avoid a race with new bios
+ * being added to the cell.
+ */
+ cell_visit_release(tc->pool, __inc_remap_and_issue_cell,
+ &info, cell);
+
+ while ((bio = bio_list_pop(&info.defer_bios)))
+ thin_defer_bio(tc, bio);
+
+ while ((bio = bio_list_pop(&info.issue_bios)))
+ remap_and_issue(info.tc, bio, block);
}
static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
* the bios in the cell.
*/
if (bio) {
- cell_defer_no_holder(tc, m->cell);
+ inc_remap_and_issue_cell(tc, m->cell, m->data_block);
bio_endio(bio, 0);
- } else
- cell_defer(tc, m->cell);
+ } else {
+ inc_all_io_entry(tc->pool, m->cell->holder);
+ remap_and_issue(tc, m->cell->holder, m->data_block);
+ inc_remap_and_issue_cell(tc, m->cell, m->data_block);
+ }
out:
list_del(&m->list);
return m;
}
+static void ll_zero(struct thin_c *tc, struct dm_thin_new_mapping *m,
+ sector_t begin, sector_t end)
+{
+ int r;
+ struct dm_io_region to;
+
+ to.bdev = tc->pool_dev->bdev;
+ to.sector = begin;
+ to.count = end - begin;
+
+ r = dm_kcopyd_zero(tc->pool->copier, 1, &to, 0, copy_complete, m);
+ if (r < 0) {
+ DMERR_LIMIT("dm_kcopyd_zero() failed");
+ copy_complete(1, 1, m);
+ }
+}
+
+static void remap_and_issue_overwrite(struct thin_c *tc, struct bio *bio,
+ dm_block_t data_block,
+ struct dm_thin_new_mapping *m)
+{
+ struct pool *pool = tc->pool;
+ 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);
+}
+
+/*
+ * A partial copy also needs to zero the uncopied region.
+ */
static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
struct dm_dev *origin, dm_block_t data_origin,
dm_block_t data_dest,
- struct dm_bio_prison_cell *cell, struct bio *bio)
+ struct dm_bio_prison_cell *cell, struct bio *bio,
+ sector_t len)
{
int r;
struct pool *pool = tc->pool;
m->data_block = data_dest;
m->cell = cell;
+ /*
+ * quiesce action + copy action + an extra reference held for the
+ * duration of this function (we may need to inc later for a
+ * partial zero).
+ */
+ atomic_set(&m->prepare_actions, 3);
+
if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
- m->quiesced = true;
+ complete_mapping_preparation(m); /* already quiesced */
/*
* IO to pool_dev remaps to the pool target's data_dev.
* If the whole block of data is being overwritten, we can issue the
* bio immediately. Otherwise we use kcopyd to clone the data first.
*/
- if (io_overwrites_block(pool, bio)) {
- struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
-
- h->overwrite_mapping = m;
- m->bio = bio;
- save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
- inc_all_io_entry(pool, bio);
- remap_and_issue(tc, bio, data_dest);
- } else {
+ if (io_overwrites_block(pool, bio))
+ remap_and_issue_overwrite(tc, bio, data_dest, m);
+ 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;
+ from.count = len;
to.bdev = tc->pool_dev->bdev;
to.sector = data_dest * pool->sectors_per_block;
- to.count = pool->sectors_per_block;
+ to.count = len;
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);
+ copy_complete(1, 1, m);
+
+ /*
+ * We allow the zero to be issued, to simplify the
+ * error path. Otherwise we'd need to start
+ * worrying about decrementing the prepare_actions
+ * counter.
+ */
+ }
+
+ /*
+ * Do we need to zero a tail region?
+ */
+ if (len < pool->sectors_per_block && pool->pf.zero_new_blocks) {
+ atomic_inc(&m->prepare_actions);
+ ll_zero(tc, m,
+ data_dest * pool->sectors_per_block + len,
+ (data_dest + 1) * pool->sectors_per_block);
}
}
+
+ complete_mapping_preparation(m); /* drop our ref */
}
static void schedule_internal_copy(struct thin_c *tc, dm_block_t virt_block,
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);
+ data_origin, data_dest, cell, bio,
+ tc->pool->sectors_per_block);
}
static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
struct pool *pool = tc->pool;
struct dm_thin_new_mapping *m = get_next_mapping(pool);
- m->quiesced = true;
- m->prepared = false;
+ atomic_set(&m->prepare_actions, 1); /* no need to quiesce */
m->tc = tc;
m->virt_block = virt_block;
m->data_block = data_block;
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));
+ else if (io_overwrites_block(pool, bio))
+ remap_and_issue_overwrite(tc, bio, data_block, m);
- 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;
+ else
+ ll_zero(tc, m,
+ data_block * pool->sectors_per_block,
+ (data_block + 1) * pool->sectors_per_block);
+}
- to.bdev = tc->pool_dev->bdev;
- to.sector = data_block * pool->sectors_per_block;
- to.count = pool->sectors_per_block;
+static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
+ dm_block_t data_dest,
+ struct dm_bio_prison_cell *cell, struct bio *bio)
+{
+ struct pool *pool = tc->pool;
+ sector_t virt_block_begin = virt_block * pool->sectors_per_block;
+ sector_t virt_block_end = (virt_block + 1) * pool->sectors_per_block;
- 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);
- }
- }
+ if (virt_block_end <= tc->origin_size)
+ schedule_copy(tc, virt_block, tc->origin_dev,
+ virt_block, data_dest, cell, bio,
+ pool->sectors_per_block);
+
+ else if (virt_block_begin < tc->origin_size)
+ schedule_copy(tc, virt_block, tc->origin_dev,
+ virt_block, data_dest, cell, bio,
+ tc->origin_size - virt_block_begin);
+
+ else
+ schedule_zero(tc, virt_block, data_dest, cell, bio);
}
/*
retry_on_resume(bio);
}
-static void process_discard(struct thin_c *tc, struct bio *bio)
+static void process_discard_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
{
int r;
- unsigned long flags;
+ struct bio *bio = cell->holder;
struct pool *pool = tc->pool;
- struct dm_bio_prison_cell *cell, *cell2;
- struct dm_cell_key key, key2;
+ struct dm_bio_prison_cell *cell2;
+ struct dm_cell_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))
+ if (tc->requeue_mode) {
+ cell_requeue(pool, cell);
return;
+ }
r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
switch (r) {
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);
- }
+ if (!dm_deferred_set_add_work(pool->all_io_ds, &m->list))
+ pool->process_prepared_discard(m);
+
} else {
inc_all_io_entry(pool, bio);
cell_defer_no_holder(tc, cell);
}
}
+static void process_discard_bio(struct thin_c *tc, struct bio *bio)
+{
+ struct dm_bio_prison_cell *cell;
+ struct dm_cell_key key;
+ dm_block_t block = get_bio_block(tc, bio);
+
+ build_virtual_key(tc->td, block, &key);
+ if (bio_detain(tc->pool, &key, bio, &cell))
+ return;
+
+ process_discard_cell(tc, cell);
+}
+
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,
}
}
+static void __remap_and_issue_shared_cell(void *context,
+ struct dm_bio_prison_cell *cell)
+{
+ struct remap_info *info = context;
+ struct bio *bio;
+
+ while ((bio = bio_list_pop(&cell->bios))) {
+ if ((bio_data_dir(bio) == WRITE) ||
+ (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)))
+ bio_list_add(&info->defer_bios, bio);
+ 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(info->tc->pool->shared_read_ds);
+ inc_all_io_entry(info->tc->pool, bio);
+ bio_list_add(&info->issue_bios, bio);
+ }
+ }
+}
+
+static void remap_and_issue_shared_cell(struct thin_c *tc,
+ struct dm_bio_prison_cell *cell,
+ dm_block_t block)
+{
+ struct bio *bio;
+ struct remap_info info;
+
+ info.tc = tc;
+ bio_list_init(&info.defer_bios);
+ bio_list_init(&info.issue_bios);
+
+ cell_visit_release(tc->pool, __remap_and_issue_shared_cell,
+ &info, cell);
+
+ while ((bio = bio_list_pop(&info.defer_bios)))
+ thin_defer_bio(tc, bio);
+
+ while ((bio = bio_list_pop(&info.issue_bios)))
+ remap_and_issue(tc, bio, block);
+}
+
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_thin_lookup_result *lookup_result,
+ struct dm_bio_prison_cell *virt_cell)
{
- struct dm_bio_prison_cell *cell;
+ struct dm_bio_prison_cell *data_cell;
struct pool *pool = tc->pool;
struct dm_cell_key key;
* 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))
+ if (bio_detain(pool, &key, bio, &data_cell)) {
+ cell_defer_no_holder(tc, virt_cell);
return;
+ }
- if (bio_data_dir(bio) == WRITE && bio->bi_iter.bi_size)
- break_sharing(tc, bio, block, &key, lookup_result, cell);
- else {
+ if (bio_data_dir(bio) == WRITE && bio->bi_iter.bi_size) {
+ break_sharing(tc, bio, block, &key, lookup_result, data_cell);
+ cell_defer_no_holder(tc, virt_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);
+
+ remap_and_issue_shared_cell(tc, data_cell, lookup_result->block);
+ remap_and_issue_shared_cell(tc, virt_cell, lookup_result->block);
}
}
}
}
-static void process_bio(struct thin_c *tc, struct bio *bio)
+static void process_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
{
int r;
struct pool *pool = tc->pool;
+ struct bio *bio = cell->holder;
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))
+ if (tc->requeue_mode) {
+ cell_requeue(pool, 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 {
+ if (lookup_result.shared)
+ process_shared_bio(tc, bio, block, &lookup_result, cell);
+ else {
inc_all_io_entry(pool, bio);
- cell_defer_no_holder(tc, cell);
-
remap_and_issue(tc, bio, lookup_result.block);
+ inc_remap_and_issue_cell(tc, cell, lookup_result.block);
}
break;
inc_all_io_entry(pool, bio);
cell_defer_no_holder(tc, cell);
- remap_to_origin_and_issue(tc, bio);
+ if (bio_end_sector(bio) <= tc->origin_size)
+ remap_to_origin_and_issue(tc, bio);
+
+ else if (bio->bi_iter.bi_sector < tc->origin_size) {
+ zero_fill_bio(bio);
+ bio->bi_iter.bi_size = (tc->origin_size - bio->bi_iter.bi_sector) << SECTOR_SHIFT;
+ remap_to_origin_and_issue(tc, bio);
+
+ } else {
+ zero_fill_bio(bio);
+ bio_endio(bio, 0);
+ }
} else
provision_block(tc, bio, block, cell);
break;
}
}
-static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
+static void process_bio(struct thin_c *tc, struct bio *bio)
+{
+ 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;
+
+ /*
+ * 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;
+
+ process_cell(tc, cell);
+}
+
+static void __process_bio_read_only(struct thin_c *tc, struct bio *bio,
+ struct dm_bio_prison_cell *cell)
{
int r;
int rw = bio_data_dir(bio);
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)
+ if (lookup_result.shared && (rw == WRITE) && bio->bi_iter.bi_size) {
handle_unserviceable_bio(tc->pool, bio);
- else {
+ if (cell)
+ cell_defer_no_holder(tc, cell);
+ } else {
inc_all_io_entry(tc->pool, bio);
remap_and_issue(tc, bio, lookup_result.block);
+ if (cell)
+ inc_remap_and_issue_cell(tc, cell, lookup_result.block);
}
break;
case -ENODATA:
+ if (cell)
+ cell_defer_no_holder(tc, cell);
if (rw != READ) {
handle_unserviceable_bio(tc->pool, bio);
break;
default:
DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
__func__, r);
+ if (cell)
+ cell_defer_no_holder(tc, cell);
bio_io_error(bio);
break;
}
}
+static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
+{
+ __process_bio_read_only(tc, bio, NULL);
+}
+
+static void process_cell_read_only(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+ __process_bio_read_only(tc, cell->holder, cell);
+}
+
static void process_bio_success(struct thin_c *tc, struct bio *bio)
{
bio_endio(bio, 0);
bio_io_error(bio);
}
+static void process_cell_success(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+ cell_success(tc->pool, cell);
+}
+
+static void process_cell_fail(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+ cell_error(tc->pool, cell);
+}
+
/*
* FIXME: should we also commit due to size of transaction, measured in
* metadata blocks?
struct bio *bio;
struct bio_list bios;
struct blk_plug plug;
+ unsigned count = 0;
if (tc->requeue_mode) {
requeue_bio_list(tc, &tc->deferred_bio_list);
pool->process_discard(tc, bio);
else
pool->process_bio(tc, bio);
+
+ if ((count++ & 127) == 0) {
+ throttle_work_update(&pool->throttle);
+ dm_pool_issue_prefetches(pool->pmd);
+ }
}
blk_finish_plug(&plug);
}
+static int cmp_cells(const void *lhs, const void *rhs)
+{
+ struct dm_bio_prison_cell *lhs_cell = *((struct dm_bio_prison_cell **) lhs);
+ struct dm_bio_prison_cell *rhs_cell = *((struct dm_bio_prison_cell **) rhs);
+
+ BUG_ON(!lhs_cell->holder);
+ BUG_ON(!rhs_cell->holder);
+
+ if (lhs_cell->holder->bi_iter.bi_sector < rhs_cell->holder->bi_iter.bi_sector)
+ return -1;
+
+ if (lhs_cell->holder->bi_iter.bi_sector > rhs_cell->holder->bi_iter.bi_sector)
+ return 1;
+
+ return 0;
+}
+
+static unsigned sort_cells(struct pool *pool, struct list_head *cells)
+{
+ unsigned count = 0;
+ struct dm_bio_prison_cell *cell, *tmp;
+
+ list_for_each_entry_safe(cell, tmp, cells, user_list) {
+ if (count >= CELL_SORT_ARRAY_SIZE)
+ break;
+
+ pool->cell_sort_array[count++] = cell;
+ list_del(&cell->user_list);
+ }
+
+ sort(pool->cell_sort_array, count, sizeof(cell), cmp_cells, NULL);
+
+ return count;
+}
+
+static void process_thin_deferred_cells(struct thin_c *tc)
+{
+ struct pool *pool = tc->pool;
+ unsigned long flags;
+ struct list_head cells;
+ struct dm_bio_prison_cell *cell;
+ unsigned i, j, count;
+
+ INIT_LIST_HEAD(&cells);
+
+ spin_lock_irqsave(&tc->lock, flags);
+ list_splice_init(&tc->deferred_cells, &cells);
+ spin_unlock_irqrestore(&tc->lock, flags);
+
+ if (list_empty(&cells))
+ return;
+
+ do {
+ count = sort_cells(tc->pool, &cells);
+
+ for (i = 0; i < count; i++) {
+ cell = pool->cell_sort_array[i];
+ BUG_ON(!cell->holder);
+
+ /*
+ * 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)) {
+ for (j = i; j < count; j++)
+ list_add(&pool->cell_sort_array[j]->user_list, &cells);
+
+ spin_lock_irqsave(&tc->lock, flags);
+ list_splice(&cells, &tc->deferred_cells);
+ spin_unlock_irqrestore(&tc->lock, flags);
+ return;
+ }
+
+ if (cell->holder->bi_rw & REQ_DISCARD)
+ pool->process_discard_cell(tc, cell);
+ else
+ pool->process_cell(tc, cell);
+ }
+ } while (!list_empty(&cells));
+}
+
static void thin_get(struct thin_c *tc);
static void thin_put(struct thin_c *tc);
tc = get_first_thin(pool);
while (tc) {
+ process_thin_deferred_cells(tc);
process_thin_deferred_bios(tc);
tc = get_next_thin(pool, tc);
}
{
struct pool *pool = container_of(ws, struct pool, worker);
+ throttle_work_start(&pool->throttle);
+ dm_pool_issue_prefetches(pool->pmd);
+ throttle_work_update(&pool->throttle);
process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
+ throttle_work_update(&pool->throttle);
process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
+ throttle_work_update(&pool->throttle);
process_deferred_bios(pool);
+ throttle_work_complete(&pool->throttle);
}
/*
dm_pool_metadata_read_only(pool->pmd);
pool->process_bio = process_bio_fail;
pool->process_discard = process_bio_fail;
+ pool->process_cell = process_cell_fail;
+ pool->process_discard_cell = process_cell_fail;
pool->process_prepared_mapping = process_prepared_mapping_fail;
pool->process_prepared_discard = process_prepared_discard_fail;
dm_pool_metadata_read_only(pool->pmd);
pool->process_bio = process_bio_read_only;
pool->process_discard = process_bio_success;
+ pool->process_cell = process_cell_read_only;
+ pool->process_discard_cell = process_cell_success;
pool->process_prepared_mapping = process_prepared_mapping_fail;
pool->process_prepared_discard = process_prepared_discard_passdown;
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_discard = process_discard_bio;
+ pool->process_cell = process_cell_read_only;
+ pool->process_discard_cell = process_discard_cell;
pool->process_prepared_mapping = process_prepared_mapping;
pool->process_prepared_discard = process_prepared_discard_passdown;
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_discard = process_discard_bio;
+ pool->process_cell = process_cell;
+ pool->process_discard_cell = process_discard_cell;
pool->process_prepared_mapping = process_prepared_mapping;
pool->process_prepared_discard = process_prepared_discard;
break;
wake_worker(pool);
}
+static void thin_defer_bio_with_throttle(struct thin_c *tc, struct bio *bio)
+{
+ struct pool *pool = tc->pool;
+
+ throttle_lock(&pool->throttle);
+ thin_defer_bio(tc, bio);
+ throttle_unlock(&pool->throttle);
+}
+
+static void thin_defer_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+ unsigned long flags;
+ struct pool *pool = tc->pool;
+
+ throttle_lock(&pool->throttle);
+ spin_lock_irqsave(&tc->lock, flags);
+ list_add_tail(&cell->user_list, &tc->deferred_cells);
+ spin_unlock_irqrestore(&tc->lock, flags);
+ throttle_unlock(&pool->throttle);
+
+ 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));
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_bio_prison_cell *virt_cell, *data_cell;
struct dm_cell_key key;
thin_hook_bio(tc, bio);
}
if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)) {
- thin_defer_bio(tc, bio);
+ thin_defer_bio_with_throttle(tc, bio);
return DM_MAPIO_SUBMITTED;
}
+ /*
+ * We must hold the virtual cell before doing the lookup, otherwise
+ * there's a race with discard.
+ */
+ build_virtual_key(tc->td, block, &key);
+ if (bio_detain(tc->pool, &key, bio, &virt_cell))
+ return DM_MAPIO_SUBMITTED;
+
r = dm_thin_find_block(td, block, 0, &result);
/*
* More distant ancestors are irrelevant. The
* shared flag will be set in their case.
*/
- thin_defer_bio(tc, bio);
+ thin_defer_cell(tc, virt_cell);
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);
+ if (bio_detain(tc->pool, &key, bio, &data_cell)) {
+ cell_defer_no_holder(tc, virt_cell);
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);
+ cell_defer_no_holder(tc, data_cell);
+ cell_defer_no_holder(tc, virt_cell);
remap(tc, bio, result.block);
return DM_MAPIO_REMAPPED;
* of doing so.
*/
handle_unserviceable_bio(tc->pool, bio);
+ cell_defer_no_holder(tc, virt_cell);
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);
+ thin_defer_cell(tc, virt_cell);
return DM_MAPIO_SUBMITTED;
default:
* pool is switched to fail-io mode.
*/
bio_io_error(bio);
+ cell_defer_no_holder(tc, virt_cell);
return DM_MAPIO_SUBMITTED;
}
}
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);
+ pool->prison = dm_bio_prison_create();
if (!pool->prison) {
*error = "Error creating pool's bio prison";
err_p = ERR_PTR(-ENOMEM);
goto bad_wq;
}
+ throttle_init(&pool->throttle);
INIT_WORK(&pool->worker, do_worker);
INIT_DELAYED_WORK(&pool->waker, do_waker);
INIT_DELAYED_WORK(&pool->no_space_timeout, do_no_space_timeout);
{
struct pool_c *pt = ti->private;
struct pool *pool = pt->pool;
- uint64_t io_opt_sectors = limits->io_opt >> SECTOR_SHIFT;
+ sector_t io_opt_sectors = limits->io_opt >> SECTOR_SHIFT;
+
+ /*
+ * Adjust max_sectors_kb to highest possible power-of-2
+ * factor of pool->sectors_per_block.
+ */
+ if (limits->max_hw_sectors & (limits->max_hw_sectors - 1))
+ limits->max_sectors = rounddown_pow_of_two(limits->max_hw_sectors);
+ else
+ limits->max_sectors = limits->max_hw_sectors;
+
+ if (limits->max_sectors < pool->sectors_per_block) {
+ while (!is_factor(pool->sectors_per_block, limits->max_sectors)) {
+ if ((limits->max_sectors & (limits->max_sectors - 1)) == 0)
+ limits->max_sectors--;
+ limits->max_sectors = rounddown_pow_of_two(limits->max_sectors);
+ }
+ } else if (block_size_is_power_of_two(pool)) {
+ /* max_sectors_kb is >= power-of-2 thinp blocksize */
+ while (!is_factor(limits->max_sectors, pool->sectors_per_block)) {
+ if ((limits->max_sectors & (limits->max_sectors - 1)) == 0)
+ limits->max_sectors--;
+ limits->max_sectors = rounddown_pow_of_two(limits->max_sectors);
+ }
+ }
/*
* If the system-determined stacked limits are compatible with the
* pool's blocksize (io_opt is a factor) do not override them.
*/
if (io_opt_sectors < pool->sectors_per_block ||
- do_div(io_opt_sectors, pool->sectors_per_block)) {
- blk_limits_io_min(limits, 0);
+ !is_factor(io_opt_sectors, pool->sectors_per_block)) {
+ if (is_factor(pool->sectors_per_block, limits->max_sectors))
+ blk_limits_io_min(limits, limits->max_sectors << SECTOR_SHIFT);
+ else
+ blk_limits_io_min(limits, pool->sectors_per_block << SECTOR_SHIFT);
blk_limits_io_opt(limits, pool->sectors_per_block << SECTOR_SHIFT);
}
.name = "thin-pool",
.features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
DM_TARGET_IMMUTABLE,
- .version = {1, 12, 0},
+ .version = {1, 14, 0},
.module = THIS_MODULE,
.ctr = pool_ctr,
.dtr = pool_dtr,
goto out_unlock;
}
spin_lock_init(&tc->lock);
+ INIT_LIST_HEAD(&tc->deferred_cells);
bio_list_init(&tc->deferred_bio_list);
bio_list_init(&tc->retry_on_resume_list);
tc->sort_bio_list = RB_ROOT;
spin_lock_irqsave(&pool->lock, flags);
list_for_each_entry_safe(m, tmp, &work, list) {
list_del(&m->list);
- m->quiesced = true;
- __maybe_add_mapping(m);
+ __complete_mapping_preparation(m);
}
spin_unlock_irqrestore(&pool->lock, flags);
}
noflush_work(tc, do_noflush_stop);
}
+static int thin_preresume(struct dm_target *ti)
+{
+ struct thin_c *tc = ti->private;
+
+ if (tc->origin_dev)
+ tc->origin_size = get_dev_size(tc->origin_dev->bdev);
+
+ return 0;
+}
+
/*
* <nr mapped sectors> <highest mapped sector>
*/
DMEMIT("Error");
}
+static int thin_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+ struct bio_vec *biovec, int max_size)
+{
+ struct thin_c *tc = ti->private;
+ struct request_queue *q = bdev_get_queue(tc->pool_dev->bdev);
+
+ if (!q->merge_bvec_fn)
+ return max_size;
+
+ bvm->bi_bdev = tc->pool_dev->bdev;
+ bvm->bi_sector = dm_target_offset(ti, bvm->bi_sector);
+
+ return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
static int thin_iterate_devices(struct dm_target *ti,
iterate_devices_callout_fn fn, void *data)
{
static struct target_type thin_target = {
.name = "thin",
- .version = {1, 12, 0},
+ .version = {1, 14, 0},
.module = THIS_MODULE,
.ctr = thin_ctr,
.dtr = thin_dtr,
.map = thin_map,
.end_io = thin_endio,
+ .preresume = thin_preresume,
.presuspend = thin_presuspend,
.postsuspend = thin_postsuspend,
.status = thin_status,
+ .merge = thin_merge,
.iterate_devices = thin_iterate_devices,
};