diff options
Diffstat (limited to 'drivers/md/dm-thin.c')
| -rw-r--r-- | drivers/md/dm-thin.c | 60 |
1 files changed, 53 insertions, 7 deletions
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index 5a2c494cb552..fa8d5464c1fb 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -231,6 +231,7 @@ struct pool { struct dm_target *ti; /* Only set if a pool target is bound */ struct mapped_device *pool_md; + struct block_device *data_dev; struct block_device *md_dev; struct dm_pool_metadata *pmd; @@ -281,6 +282,8 @@ struct pool { struct dm_bio_prison_cell **cell_sort_array; mempool_t mapping_pool; + + struct bio flush_bio; }; static void metadata_operation_failed(struct pool *pool, const char *op, int r); @@ -2383,8 +2386,16 @@ static void process_deferred_bios(struct pool *pool) while ((bio = bio_list_pop(&bio_completions))) bio_endio(bio); - while ((bio = bio_list_pop(&bios))) - generic_make_request(bio); + while ((bio = bio_list_pop(&bios))) { + /* + * The data device was flushed as part of metadata commit, + * so complete redundant flushes immediately. + */ + if (bio->bi_opf & REQ_PREFLUSH) + bio_endio(bio); + else + generic_make_request(bio); + } } static void do_worker(struct work_struct *ws) @@ -2915,6 +2926,7 @@ static void __pool_destroy(struct pool *pool) if (pool->next_mapping) mempool_free(pool->next_mapping, &pool->mapping_pool); mempool_exit(&pool->mapping_pool); + bio_uninit(&pool->flush_bio); dm_deferred_set_destroy(pool->shared_read_ds); dm_deferred_set_destroy(pool->all_io_ds); kfree(pool); @@ -2924,6 +2936,7 @@ static struct kmem_cache *_new_mapping_cache; static struct pool *pool_create(struct mapped_device *pool_md, struct block_device *metadata_dev, + struct block_device *data_dev, unsigned long block_size, int read_only, char **error) { @@ -2994,6 +3007,7 @@ static struct pool *pool_create(struct mapped_device *pool_md, pool->low_water_triggered = false; pool->suspended = true; pool->out_of_data_space = false; + bio_init(&pool->flush_bio, NULL, 0); pool->shared_read_ds = dm_deferred_set_create(); if (!pool->shared_read_ds) { @@ -3031,6 +3045,7 @@ static struct pool *pool_create(struct mapped_device *pool_md, pool->last_commit_jiffies = jiffies; pool->pool_md = pool_md; pool->md_dev = metadata_dev; + pool->data_dev = data_dev; __pool_table_insert(pool); return pool; @@ -3072,6 +3087,7 @@ static void __pool_dec(struct pool *pool) static struct pool *__pool_find(struct mapped_device *pool_md, struct block_device *metadata_dev, + struct block_device *data_dev, unsigned long block_size, int read_only, char **error, int *created) { @@ -3082,19 +3098,23 @@ static struct pool *__pool_find(struct mapped_device *pool_md, *error = "metadata device already in use by a pool"; return ERR_PTR(-EBUSY); } + if (pool->data_dev != data_dev) { + *error = "data 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) { + if (pool->md_dev != metadata_dev || pool->data_dev != data_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); + pool = pool_create(pool_md, metadata_dev, data_dev, block_size, read_only, error); *created = 1; } } @@ -3180,6 +3200,29 @@ static void metadata_low_callback(void *context) dm_table_event(pool->ti->table); } +/* + * We need to flush the data device **before** committing the metadata. + * + * This ensures that the data blocks of any newly inserted mappings are + * properly written to non-volatile storage and won't be lost in case of a + * crash. + * + * Failure to do so can result in data corruption in the case of internal or + * external snapshots and in the case of newly provisioned blocks, when block + * zeroing is enabled. + */ +static int metadata_pre_commit_callback(void *context) +{ + struct pool *pool = context; + struct bio *flush_bio = &pool->flush_bio; + + bio_reset(flush_bio); + bio_set_dev(flush_bio, pool->data_dev); + flush_bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH; + + return submit_bio_wait(flush_bio); +} + static sector_t get_dev_size(struct block_device *bdev) { return i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; @@ -3323,7 +3366,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv) goto out; } - pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev, + pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev, data_dev->bdev, block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created); if (IS_ERR(pool)) { r = PTR_ERR(pool); @@ -3374,6 +3417,9 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv) if (r) goto out_flags_changed; + dm_pool_register_pre_commit_callback(pool->pmd, + metadata_pre_commit_callback, pool); + pt->callbacks.congested_fn = pool_is_congested; dm_table_add_target_callbacks(ti->table, &pt->callbacks); @@ -4061,7 +4107,7 @@ static struct target_type pool_target = { .name = "thin-pool", .features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE | DM_TARGET_IMMUTABLE, - .version = {1, 21, 0}, + .version = {1, 22, 0}, .module = THIS_MODULE, .ctr = pool_ctr, .dtr = pool_dtr, @@ -4438,7 +4484,7 @@ static void thin_io_hints(struct dm_target *ti, struct queue_limits *limits) static struct target_type thin_target = { .name = "thin", - .version = {1, 21, 0}, + .version = {1, 22, 0}, .module = THIS_MODULE, .ctr = thin_ctr, .dtr = thin_dtr, |