diff options
Diffstat (limited to 'fs/btrfs/transaction.c')
| -rw-r--r-- | fs/btrfs/transaction.c | 2259 |
1 files changed, 1590 insertions, 669 deletions
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 4112d53d4f4..5f379affdf2 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -17,67 +17,219 @@ */ #include <linux/fs.h> +#include <linux/slab.h> #include <linux/sched.h> #include <linux/writeback.h> #include <linux/pagemap.h> #include <linux/blkdev.h> +#include <linux/uuid.h> #include "ctree.h" #include "disk-io.h" #include "transaction.h" #include "locking.h" -#include "ref-cache.h" #include "tree-log.h" +#include "inode-map.h" +#include "volumes.h" +#include "dev-replace.h" +#include "qgroup.h" #define BTRFS_ROOT_TRANS_TAG 0 -static noinline void put_transaction(struct btrfs_transaction *transaction) +static unsigned int btrfs_blocked_trans_types[TRANS_STATE_MAX] = { + [TRANS_STATE_RUNNING] = 0U, + [TRANS_STATE_BLOCKED] = (__TRANS_USERSPACE | + __TRANS_START), + [TRANS_STATE_COMMIT_START] = (__TRANS_USERSPACE | + __TRANS_START | + __TRANS_ATTACH), + [TRANS_STATE_COMMIT_DOING] = (__TRANS_USERSPACE | + __TRANS_START | + __TRANS_ATTACH | + __TRANS_JOIN), + [TRANS_STATE_UNBLOCKED] = (__TRANS_USERSPACE | + __TRANS_START | + __TRANS_ATTACH | + __TRANS_JOIN | + __TRANS_JOIN_NOLOCK), + [TRANS_STATE_COMPLETED] = (__TRANS_USERSPACE | + __TRANS_START | + __TRANS_ATTACH | + __TRANS_JOIN | + __TRANS_JOIN_NOLOCK), +}; + +void btrfs_put_transaction(struct btrfs_transaction *transaction) { - WARN_ON(transaction->use_count == 0); - transaction->use_count--; - if (transaction->use_count == 0) { - list_del_init(&transaction->list); - memset(transaction, 0, sizeof(*transaction)); + WARN_ON(atomic_read(&transaction->use_count) == 0); + if (atomic_dec_and_test(&transaction->use_count)) { + BUG_ON(!list_empty(&transaction->list)); + WARN_ON(!RB_EMPTY_ROOT(&transaction->delayed_refs.href_root)); + while (!list_empty(&transaction->pending_chunks)) { + struct extent_map *em; + + em = list_first_entry(&transaction->pending_chunks, + struct extent_map, list); + list_del_init(&em->list); + free_extent_map(em); + } kmem_cache_free(btrfs_transaction_cachep, transaction); } } +static noinline void switch_commit_roots(struct btrfs_transaction *trans, + struct btrfs_fs_info *fs_info) +{ + struct btrfs_root *root, *tmp; + + down_write(&fs_info->commit_root_sem); + list_for_each_entry_safe(root, tmp, &trans->switch_commits, + dirty_list) { + list_del_init(&root->dirty_list); + free_extent_buffer(root->commit_root); + root->commit_root = btrfs_root_node(root); + if (is_fstree(root->objectid)) + btrfs_unpin_free_ino(root); + } + up_write(&fs_info->commit_root_sem); +} + +static inline void extwriter_counter_inc(struct btrfs_transaction *trans, + unsigned int type) +{ + if (type & TRANS_EXTWRITERS) + atomic_inc(&trans->num_extwriters); +} + +static inline void extwriter_counter_dec(struct btrfs_transaction *trans, + unsigned int type) +{ + if (type & TRANS_EXTWRITERS) + atomic_dec(&trans->num_extwriters); +} + +static inline void extwriter_counter_init(struct btrfs_transaction *trans, + unsigned int type) +{ + atomic_set(&trans->num_extwriters, ((type & TRANS_EXTWRITERS) ? 1 : 0)); +} + +static inline int extwriter_counter_read(struct btrfs_transaction *trans) +{ + return atomic_read(&trans->num_extwriters); +} + /* * either allocate a new transaction or hop into the existing one */ -static noinline int join_transaction(struct btrfs_root *root) +static noinline int join_transaction(struct btrfs_root *root, unsigned int type) { struct btrfs_transaction *cur_trans; - cur_trans = root->fs_info->running_transaction; - if (!cur_trans) { - cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, - GFP_NOFS); - BUG_ON(!cur_trans); - root->fs_info->generation++; - root->fs_info->last_alloc = 0; - root->fs_info->last_data_alloc = 0; - cur_trans->num_writers = 1; - cur_trans->num_joined = 0; - cur_trans->transid = root->fs_info->generation; - init_waitqueue_head(&cur_trans->writer_wait); - init_waitqueue_head(&cur_trans->commit_wait); - cur_trans->in_commit = 0; - cur_trans->blocked = 0; - cur_trans->use_count = 1; - cur_trans->commit_done = 0; - cur_trans->start_time = get_seconds(); - INIT_LIST_HEAD(&cur_trans->pending_snapshots); - list_add_tail(&cur_trans->list, &root->fs_info->trans_list); - extent_io_tree_init(&cur_trans->dirty_pages, - root->fs_info->btree_inode->i_mapping, - GFP_NOFS); - spin_lock(&root->fs_info->new_trans_lock); - root->fs_info->running_transaction = cur_trans; - spin_unlock(&root->fs_info->new_trans_lock); - } else { - cur_trans->num_writers++; - cur_trans->num_joined++; + struct btrfs_fs_info *fs_info = root->fs_info; + + spin_lock(&fs_info->trans_lock); +loop: + /* The file system has been taken offline. No new transactions. */ + if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { + spin_unlock(&fs_info->trans_lock); + return -EROFS; + } + + cur_trans = fs_info->running_transaction; + if (cur_trans) { + if (cur_trans->aborted) { + spin_unlock(&fs_info->trans_lock); + return cur_trans->aborted; + } + if (btrfs_blocked_trans_types[cur_trans->state] & type) { + spin_unlock(&fs_info->trans_lock); + return -EBUSY; + } + atomic_inc(&cur_trans->use_count); + atomic_inc(&cur_trans->num_writers); + extwriter_counter_inc(cur_trans, type); + spin_unlock(&fs_info->trans_lock); + return 0; + } + spin_unlock(&fs_info->trans_lock); + + /* + * If we are ATTACH, we just want to catch the current transaction, + * and commit it. If there is no transaction, just return ENOENT. + */ + if (type == TRANS_ATTACH) + return -ENOENT; + + /* + * JOIN_NOLOCK only happens during the transaction commit, so + * it is impossible that ->running_transaction is NULL + */ + BUG_ON(type == TRANS_JOIN_NOLOCK); + + cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS); + if (!cur_trans) + return -ENOMEM; + + spin_lock(&fs_info->trans_lock); + if (fs_info->running_transaction) { + /* + * someone started a transaction after we unlocked. Make sure + * to redo the checks above + */ + kmem_cache_free(btrfs_transaction_cachep, cur_trans); + goto loop; + } else if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { + spin_unlock(&fs_info->trans_lock); + kmem_cache_free(btrfs_transaction_cachep, cur_trans); + return -EROFS; } + atomic_set(&cur_trans->num_writers, 1); + extwriter_counter_init(cur_trans, type); + init_waitqueue_head(&cur_trans->writer_wait); + init_waitqueue_head(&cur_trans->commit_wait); + cur_trans->state = TRANS_STATE_RUNNING; + /* + * One for this trans handle, one so it will live on until we + * commit the transaction. + */ + atomic_set(&cur_trans->use_count, 2); + cur_trans->start_time = get_seconds(); + + cur_trans->delayed_refs.href_root = RB_ROOT; + atomic_set(&cur_trans->delayed_refs.num_entries, 0); + cur_trans->delayed_refs.num_heads_ready = 0; + cur_trans->delayed_refs.num_heads = 0; + cur_trans->delayed_refs.flushing = 0; + cur_trans->delayed_refs.run_delayed_start = 0; + + /* + * although the tree mod log is per file system and not per transaction, + * the log must never go across transaction boundaries. + */ + smp_mb(); + if (!list_empty(&fs_info->tree_mod_seq_list)) + WARN(1, KERN_ERR "BTRFS: tree_mod_seq_list not empty when " + "creating a fresh transaction\n"); + if (!RB_EMPTY_ROOT(&fs_info->tree_mod_log)) + WARN(1, KERN_ERR "BTRFS: tree_mod_log rb tree not empty when " + "creating a fresh transaction\n"); + atomic64_set(&fs_info->tree_mod_seq, 0); + + spin_lock_init(&cur_trans->delayed_refs.lock); + + INIT_LIST_HEAD(&cur_trans->pending_snapshots); + INIT_LIST_HEAD(&cur_trans->ordered_operations); + INIT_LIST_HEAD(&cur_trans->pending_chunks); + INIT_LIST_HEAD(&cur_trans->switch_commits); + list_add_tail(&cur_trans->list, &fs_info->trans_list); + extent_io_tree_init(&cur_trans->dirty_pages, + fs_info->btree_inode->i_mapping); + fs_info->generation++; + cur_trans->transid = fs_info->generation; + fs_info->running_transaction = cur_trans; + cur_trans->aborted = 0; + spin_unlock(&fs_info->trans_lock); + return 0; } @@ -87,48 +239,93 @@ static noinline int join_transaction(struct btrfs_root *root) * to make sure the old root from before we joined the transaction is deleted * when the transaction commits */ -noinline int btrfs_record_root_in_trans(struct btrfs_root *root) +static int record_root_in_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root) { - struct btrfs_dirty_root *dirty; - u64 running_trans_id = root->fs_info->running_transaction->transid; - if (root->ref_cows && root->last_trans < running_trans_id) { + if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) && + root->last_trans < trans->transid) { WARN_ON(root == root->fs_info->extent_root); - if (root->root_item.refs != 0) { - radix_tree_tag_set(&root->fs_info->fs_roots_radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_TRANS_TAG); - - dirty = kmalloc(sizeof(*dirty), GFP_NOFS); - BUG_ON(!dirty); - dirty->root = kmalloc(sizeof(*dirty->root), GFP_NOFS); - BUG_ON(!dirty->root); - dirty->latest_root = root; - INIT_LIST_HEAD(&dirty->list); - - root->commit_root = btrfs_root_node(root); - - memcpy(dirty->root, root, sizeof(*root)); - spin_lock_init(&dirty->root->node_lock); - spin_lock_init(&dirty->root->list_lock); - mutex_init(&dirty->root->objectid_mutex); - mutex_init(&dirty->root->log_mutex); - INIT_LIST_HEAD(&dirty->root->dead_list); - dirty->root->node = root->commit_root; - dirty->root->commit_root = NULL; - - spin_lock(&root->list_lock); - list_add(&dirty->root->dead_list, &root->dead_list); - spin_unlock(&root->list_lock); - - root->dirty_root = dirty; - } else { - WARN_ON(1); + WARN_ON(root->commit_root != root->node); + + /* + * see below for IN_TRANS_SETUP usage rules + * we have the reloc mutex held now, so there + * is only one writer in this function + */ + set_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state); + + /* make sure readers find IN_TRANS_SETUP before + * they find our root->last_trans update + */ + smp_wmb(); + + spin_lock(&root->fs_info->fs_roots_radix_lock); + if (root->last_trans == trans->transid) { + spin_unlock(&root->fs_info->fs_roots_radix_lock); + return 0; } - root->last_trans = running_trans_id; + radix_tree_tag_set(&root->fs_info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_TRANS_TAG); + spin_unlock(&root->fs_info->fs_roots_radix_lock); + root->last_trans = trans->transid; + + /* this is pretty tricky. We don't want to + * take the relocation lock in btrfs_record_root_in_trans + * unless we're really doing the first setup for this root in + * this transaction. + * + * Normally we'd use root->last_trans as a flag to decide + * if we want to take the expensive mutex. + * + * But, we have to set root->last_trans before we + * init the relocation root, otherwise, we trip over warnings + * in ctree.c. The solution used here is to flag ourselves + * with root IN_TRANS_SETUP. When this is 1, we're still + * fixing up the reloc trees and everyone must wait. + * + * When this is zero, they can trust root->last_trans and fly + * through btrfs_record_root_in_trans without having to take the + * lock. smp_wmb() makes sure that all the writes above are + * done before we pop in the zero below + */ + btrfs_init_reloc_root(trans, root); + smp_mb__before_atomic(); + clear_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state); } return 0; } + +int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) + return 0; + + /* + * see record_root_in_trans for comments about IN_TRANS_SETUP usage + * and barriers + */ + smp_rmb(); + if (root->last_trans == trans->transid && + !test_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state)) + return 0; + + mutex_lock(&root->fs_info->reloc_mutex); + record_root_in_trans(trans, root); + mutex_unlock(&root->fs_info->reloc_mutex); + + return 0; +} + +static inline int is_transaction_blocked(struct btrfs_transaction *trans) +{ + return (trans->state >= TRANS_STATE_BLOCKED && + trans->state < TRANS_STATE_UNBLOCKED && + !trans->aborted); +} + /* wait for commit against the current transaction to become unblocked * when this is done, it is safe to start a new transaction, but the current * transaction might not be fully on disk. @@ -137,167 +334,464 @@ static void wait_current_trans(struct btrfs_root *root) { struct btrfs_transaction *cur_trans; + spin_lock(&root->fs_info->trans_lock); cur_trans = root->fs_info->running_transaction; - if (cur_trans && cur_trans->blocked) { - DEFINE_WAIT(wait); - cur_trans->use_count++; - while (1) { - prepare_to_wait(&root->fs_info->transaction_wait, &wait, - TASK_UNINTERRUPTIBLE); - if (cur_trans->blocked) { - mutex_unlock(&root->fs_info->trans_mutex); - schedule(); - mutex_lock(&root->fs_info->trans_mutex); - finish_wait(&root->fs_info->transaction_wait, - &wait); - } else { - finish_wait(&root->fs_info->transaction_wait, - &wait); - break; - } - } - put_transaction(cur_trans); + if (cur_trans && is_transaction_blocked(cur_trans)) { + atomic_inc(&cur_trans->use_count); + spin_unlock(&root->fs_info->trans_lock); + + wait_event(root->fs_info->transaction_wait, + cur_trans->state >= TRANS_STATE_UNBLOCKED || + cur_trans->aborted); + btrfs_put_transaction(cur_trans); + } else { + spin_unlock(&root->fs_info->trans_lock); } } -static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, - int num_blocks, int wait) +static int may_wait_transaction(struct btrfs_root *root, int type) +{ + if (root->fs_info->log_root_recovering) + return 0; + + if (type == TRANS_USERSPACE) + return 1; + + if (type == TRANS_START && + !atomic_read(&root->fs_info->open_ioctl_trans)) + return 1; + + return 0; +} + +static inline bool need_reserve_reloc_root(struct btrfs_root *root) { - struct btrfs_trans_handle *h = - kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); + if (!root->fs_info->reloc_ctl || + !test_bit(BTRFS_ROOT_REF_COWS, &root->state) || + root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID || + root->reloc_root) + return false; + + return true; +} + +static struct btrfs_trans_handle * +start_transaction(struct btrfs_root *root, u64 num_items, unsigned int type, + enum btrfs_reserve_flush_enum flush) +{ + struct btrfs_trans_handle *h; + struct btrfs_transaction *cur_trans; + u64 num_bytes = 0; + u64 qgroup_reserved = 0; + bool reloc_reserved = false; int ret; - mutex_lock(&root->fs_info->trans_mutex); - if (!root->fs_info->log_root_recovering && - ((wait == 1 && !root->fs_info->open_ioctl_trans) || wait == 2)) + /* Send isn't supposed to start transactions. */ + ASSERT(current->journal_info != (void *)BTRFS_SEND_TRANS_STUB); + + if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) + return ERR_PTR(-EROFS); + + if (current->journal_info) { + WARN_ON(type & TRANS_EXTWRITERS); + h = current->journal_info; + h->use_count++; + WARN_ON(h->use_count > 2); + h->orig_rsv = h->block_rsv; + h->block_rsv = NULL; + goto got_it; + } + + /* + * Do the reservation before we join the transaction so we can do all + * the appropriate flushing if need be. + */ + if (num_items > 0 && root != root->fs_info->chunk_root) { + if (root->fs_info->quota_enabled && + is_fstree(root->root_key.objectid)) { + qgroup_reserved = num_items * root->leafsize; + ret = btrfs_qgroup_reserve(root, qgroup_reserved); + if (ret) + return ERR_PTR(ret); + } + + num_bytes = btrfs_calc_trans_metadata_size(root, num_items); + /* + * Do the reservation for the relocation root creation + */ + if (unlikely(need_reserve_reloc_root(root))) { + num_bytes += root->nodesize; + reloc_reserved = true; + } + + ret = btrfs_block_rsv_add(root, + &root->fs_info->trans_block_rsv, + num_bytes, flush); + if (ret) + goto reserve_fail; + } +again: + h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); + if (!h) { + ret = -ENOMEM; + goto alloc_fail; + } + + /* + * If we are JOIN_NOLOCK we're already committing a transaction and + * waiting on this guy, so we don't need to do the sb_start_intwrite + * because we're already holding a ref. We need this because we could + * have raced in and did an fsync() on a file which can kick a commit + * and then we deadlock with somebody doing a freeze. + * + * If we are ATTACH, it means we just want to catch the current + * transaction and commit it, so we needn't do sb_start_intwrite(). + */ + if (type & __TRANS_FREEZABLE) + sb_start_intwrite(root->fs_info->sb); + + if (may_wait_transaction(root, type)) wait_current_trans(root); - ret = join_transaction(root); - BUG_ON(ret); - btrfs_record_root_in_trans(root); - h->transid = root->fs_info->running_transaction->transid; - h->transaction = root->fs_info->running_transaction; - h->blocks_reserved = num_blocks; + do { + ret = join_transaction(root, type); + if (ret == -EBUSY) { + wait_current_trans(root); + if (unlikely(type == TRANS_ATTACH)) + ret = -ENOENT; + } + } while (ret == -EBUSY); + + if (ret < 0) { + /* We must get the transaction if we are JOIN_NOLOCK. */ + BUG_ON(type == TRANS_JOIN_NOLOCK); + goto join_fail; + } + + cur_trans = root->fs_info->running_transaction; + + h->transid = cur_trans->transid; + h->transaction = cur_trans; h->blocks_used = 0; - h->block_group = 0; - h->alloc_exclude_nr = 0; - h->alloc_exclude_start = 0; - root->fs_info->running_transaction->use_count++; - mutex_unlock(&root->fs_info->trans_mutex); + h->bytes_reserved = 0; + h->root = root; + h->delayed_ref_updates = 0; + h->use_count = 1; + h->adding_csums = 0; + h->block_rsv = NULL; + h->orig_rsv = NULL; + h->aborted = 0; + h->qgroup_reserved = 0; + h->delayed_ref_elem.seq = 0; + h->type = type; + h->allocating_chunk = false; + h->reloc_reserved = false; + h->sync = false; + INIT_LIST_HEAD(&h->qgroup_ref_list); + INIT_LIST_HEAD(&h->new_bgs); + + smp_mb(); + if (cur_trans->state >= TRANS_STATE_BLOCKED && + may_wait_transaction(root, type)) { + current->journal_info = h; + btrfs_commit_transaction(h, root); + goto again; + } + + if (num_bytes) { + trace_btrfs_space_reservation(root->fs_info, "transaction", + h->transid, num_bytes, 1); + h->block_rsv = &root->fs_info->trans_block_rsv; + h->bytes_reserved = num_bytes; + h->reloc_reserved = reloc_reserved; + } + h->qgroup_reserved = qgroup_reserved; + +got_it: + btrfs_record_root_in_trans(h, root); + + if (!current->journal_info && type != TRANS_USERSPACE) + current->journal_info = h; return h; + +join_fail: + if (type & __TRANS_FREEZABLE) + sb_end_intwrite(root->fs_info->sb); + kmem_cache_free(btrfs_trans_handle_cachep, h); +alloc_fail: + if (num_bytes) + btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, + num_bytes); +reserve_fail: + if (qgroup_reserved) + btrfs_qgroup_free(root, qgroup_reserved); + return ERR_PTR(ret); } struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, - int num_blocks) + int num_items) { - return start_transaction(root, num_blocks, 1); + return start_transaction(root, num_items, TRANS_START, + BTRFS_RESERVE_FLUSH_ALL); } -struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root, - int num_blocks) + +struct btrfs_trans_handle *btrfs_start_transaction_lflush( + struct btrfs_root *root, int num_items) { - return start_transaction(root, num_blocks, 0); + return start_transaction(root, num_items, TRANS_START, + BTRFS_RESERVE_FLUSH_LIMIT); } -struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r, - int num_blocks) +struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root) { - return start_transaction(r, num_blocks, 2); + return start_transaction(root, 0, TRANS_JOIN, 0); } -/* wait for a transaction commit to be fully complete */ -static noinline int wait_for_commit(struct btrfs_root *root, - struct btrfs_transaction *commit) +struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root) { - DEFINE_WAIT(wait); - mutex_lock(&root->fs_info->trans_mutex); - while (!commit->commit_done) { - prepare_to_wait(&commit->commit_wait, &wait, - TASK_UNINTERRUPTIBLE); - if (commit->commit_done) - break; - mutex_unlock(&root->fs_info->trans_mutex); - schedule(); - mutex_lock(&root->fs_info->trans_mutex); - } - mutex_unlock(&root->fs_info->trans_mutex); - finish_wait(&commit->commit_wait, &wait); - return 0; + return start_transaction(root, 0, TRANS_JOIN_NOLOCK, 0); +} + +struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root) +{ + return start_transaction(root, 0, TRANS_USERSPACE, 0); } /* - * rate limit against the drop_snapshot code. This helps to slow down new - * operations if the drop_snapshot code isn't able to keep up. + * btrfs_attach_transaction() - catch the running transaction + * + * It is used when we want to commit the current the transaction, but + * don't want to start a new one. + * + * Note: If this function return -ENOENT, it just means there is no + * running transaction. But it is possible that the inactive transaction + * is still in the memory, not fully on disk. If you hope there is no + * inactive transaction in the fs when -ENOENT is returned, you should + * invoke + * btrfs_attach_transaction_barrier() */ -static void throttle_on_drops(struct btrfs_root *root) +struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root) { - struct btrfs_fs_info *info = root->fs_info; - int harder_count = 0; - -harder: - if (atomic_read(&info->throttles)) { - DEFINE_WAIT(wait); - int thr; - thr = atomic_read(&info->throttle_gen); - - do { - prepare_to_wait(&info->transaction_throttle, - &wait, TASK_UNINTERRUPTIBLE); - if (!atomic_read(&info->throttles)) { - finish_wait(&info->transaction_throttle, &wait); - break; - } - schedule(); - finish_wait(&info->transaction_throttle, &wait); - } while (thr == atomic_read(&info->throttle_gen)); - harder_count++; + return start_transaction(root, 0, TRANS_ATTACH, 0); +} + +/* + * btrfs_attach_transaction_barrier() - catch the running transaction + * + * It is similar to the above function, the differentia is this one + * will wait for all the inactive transactions until they fully + * complete. + */ +struct btrfs_trans_handle * +btrfs_attach_transaction_barrier(struct btrfs_root *root) +{ + struct btrfs_trans_handle *trans; + + trans = start_transaction(root, 0, TRANS_ATTACH, 0); + if (IS_ERR(trans) && PTR_ERR(trans) == -ENOENT) + btrfs_wait_for_commit(root, 0); + + return trans; +} - if (root->fs_info->total_ref_cache_size > 1 * 1024 * 1024 && - harder_count < 2) - goto harder; +/* wait for a transaction commit to be fully complete */ +static noinline void wait_for_commit(struct btrfs_root *root, + struct btrfs_transaction *commit) +{ + wait_event(commit->commit_wait, commit->state == TRANS_STATE_COMPLETED); +} - if (root->fs_info->total_ref_cache_size > 5 * 1024 * 1024 && - harder_count < 10) - goto harder; +int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid) +{ + struct btrfs_transaction *cur_trans = NULL, *t; + int ret = 0; - if (root->fs_info->total_ref_cache_size > 10 * 1024 * 1024 && - harder_count < 20) - goto harder; + if (transid) { + if (transid <= root->fs_info->last_trans_committed) + goto out; + + ret = -EINVAL; + /* find specified transaction */ + spin_lock(&root->fs_info->trans_lock); + list_for_each_entry(t, &root->fs_info->trans_list, list) { + if (t->transid == transid) { + cur_trans = t; + atomic_inc(&cur_trans->use_count); + ret = 0; + break; + } + if (t->transid > transid) { + ret = 0; + break; + } + } + spin_unlock(&root->fs_info->trans_lock); + /* The specified transaction doesn't exist */ + if (!cur_trans) + goto out; + } else { + /* find newest transaction that is committing | committed */ + spin_lock(&root->fs_info->trans_lock); + list_for_each_entry_reverse(t, &root->fs_info->trans_list, + list) { + if (t->state >= TRANS_STATE_COMMIT_START) { + if (t->state == TRANS_STATE_COMPLETED) + break; + cur_trans = t; + atomic_inc(&cur_trans->use_count); + break; + } + } + spin_unlock(&root->fs_info->trans_lock); + if (!cur_trans) + goto out; /* nothing committing|committed */ } + + wait_for_commit(root, cur_trans); + btrfs_put_transaction(cur_trans); +out: + return ret; } void btrfs_throttle(struct btrfs_root *root) { - mutex_lock(&root->fs_info->trans_mutex); - if (!root->fs_info->open_ioctl_trans) + if (!atomic_read(&root->fs_info->open_ioctl_trans)) wait_current_trans(root); - mutex_unlock(&root->fs_info->trans_mutex); +} - throttle_on_drops(root); +static int should_end_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + if (root->fs_info->global_block_rsv.space_info->full && + btrfs_check_space_for_delayed_refs(trans, root)) + return 1; + + return !!btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5); +} + +int btrfs_should_end_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + struct btrfs_transaction *cur_trans = trans->transaction; + int updates; + int err; + + smp_mb(); + if (cur_trans->state >= TRANS_STATE_BLOCKED || + cur_trans->delayed_refs.flushing) + return 1; + + updates = trans->delayed_ref_updates; + trans->delayed_ref_updates = 0; + if (updates) { + err = btrfs_run_delayed_refs(trans, root, updates); + if (err) /* Error code will also eval true */ + return err; + } + + return should_end_transaction(trans, root); } static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root, int throttle) { - struct btrfs_transaction *cur_trans; + struct btrfs_transaction *cur_trans = trans->transaction; struct btrfs_fs_info *info = root->fs_info; + unsigned long cur = trans->delayed_ref_updates; + int lock = (trans->type != TRANS_JOIN_NOLOCK); + int err = 0; + int must_run_delayed_refs = 0; - mutex_lock(&info->trans_mutex); - cur_trans = info->running_transaction; - WARN_ON(cur_trans != trans->transaction); - WARN_ON(cur_trans->num_writers < 1); - cur_trans->num_writers--; + if (trans->use_count > 1) { + trans->use_count--; + trans->block_rsv = trans->orig_rsv; + return 0; + } + + btrfs_trans_release_metadata(trans, root); + trans->block_rsv = NULL; + + if (!list_empty(&trans->new_bgs)) + btrfs_create_pending_block_groups(trans, root); + + trans->delayed_ref_updates = 0; + if (!trans->sync) { + must_run_delayed_refs = + btrfs_should_throttle_delayed_refs(trans, root); + cur = max_t(unsigned long, cur, 32); + + /* + * don't make the caller wait if they are from a NOLOCK + * or ATTACH transaction, it will deadlock with commit + */ + if (must_run_delayed_refs == 1 && + (trans->type & (__TRANS_JOIN_NOLOCK | __TRANS_ATTACH))) + must_run_delayed_refs = 2; + } + + if (trans->qgroup_reserved) { + /* + * the same root has to be passed here between start_transaction + * and end_transaction. Subvolume quota depends on this. + */ + btrfs_qgroup_free(trans->root, trans->qgroup_reserved); + trans->qgroup_reserved = 0; + } + + btrfs_trans_release_metadata(trans, root); + trans->block_rsv = NULL; + + if (!list_empty(&trans->new_bgs)) + btrfs_create_pending_block_groups(trans, root); + + if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) && + should_end_transaction(trans, root) && + ACCESS_ONCE(cur_trans->state) == TRANS_STATE_RUNNING) { + spin_lock(&info->trans_lock); + if (cur_trans->state == TRANS_STATE_RUNNING) + cur_trans->state = TRANS_STATE_BLOCKED; + spin_unlock(&info->trans_lock); + } + + if (lock && ACCESS_ONCE(cur_trans->state) == TRANS_STATE_BLOCKED) { + if (throttle) + return btrfs_commit_transaction(trans, root); + else + wake_up_process(info->transaction_kthread); + } + + if (trans->type & __TRANS_FREEZABLE) + sb_end_intwrite(root->fs_info->sb); + WARN_ON(cur_trans != info->running_transaction); + WARN_ON(atomic_read(&cur_trans->num_writers) < 1); + atomic_dec(&cur_trans->num_writers); + extwriter_counter_dec(cur_trans, trans->type); + + smp_mb(); if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); - put_transaction(cur_trans); - mutex_unlock(&info->trans_mutex); - memset(trans, 0, sizeof(*trans)); - kmem_cache_free(btrfs_trans_handle_cachep, trans); + btrfs_put_transaction(cur_trans); + + if (current->journal_info == trans) + current->journal_info = NULL; if (throttle) - throttle_on_drops(root); + btrfs_run_delayed_iputs(root); - return 0; + if (trans->aborted || + test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) { + wake_up_process(info->transaction_kthread); + err = -EIO; + } + assert_qgroups_uptodate(trans); + + kmem_cache_free(btrfs_trans_handle_cachep, trans); + if (must_run_delayed_refs) { + btrfs_async_run_delayed_refs(root, cur, + must_run_delayed_refs == 1); + } + return err; } int btrfs_end_transaction(struct btrfs_trans_handle *trans, @@ -315,86 +809,89 @@ int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, /* * when btree blocks are allocated, they have some corresponding bits set for * them in one of two extent_io trees. This is used to make sure all of - * those extents are on disk for transaction or log commit + * those extents are sent to disk but does not wait on them */ -int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, - struct extent_io_tree *dirty_pages) +int btrfs_write_marked_extents(struct btrfs_root *root, + struct extent_io_tree *dirty_pages, int mark) { - int ret; int err = 0; int werr = 0; - struct page *page; - struct inode *btree_inode = root->fs_info->btree_inode; + struct address_space *mapping = root->fs_info->btree_inode->i_mapping; + struct extent_state *cached_state = NULL; u64 start = 0; u64 end; - unsigned long index; - while (1) { - ret = find_first_extent_bit(dirty_pages, start, &start, &end, - EXTENT_DIRTY); - if (ret) - break; - while (start <= end) { - cond_resched(); - - index = start >> PAGE_CACHE_SHIFT; - start = (u64)(index + 1) << PAGE_CACHE_SHIFT; - page = find_get_page(btree_inode->i_mapping, index); - if (!page) - continue; - - btree_lock_page_hook(page); - if (!page->mapping) { - unlock_page(page); - page_cache_release(page); - continue; - } - - if (PageWriteback(page)) { - if (PageDirty(page)) - wait_on_page_writeback(page); - else { - unlock_page(page); - page_cache_release(page); - continue; - } - } - err = write_one_page(page, 0); - if (err) - werr = err; - page_cache_release(page); - } + while (!find_first_extent_bit(dirty_pages, start, &start, &end, + mark, &cached_state)) { + convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT, + mark, &cached_state, GFP_NOFS); + cached_state = NULL; + err = filemap_fdatawrite_range(mapping, start, end); + if (err) + werr = err; + cond_resched(); + start = end + 1; } - while (1) { - ret = find_first_extent_bit(dirty_pages, 0, &start, &end, - EXTENT_DIRTY); - if (ret) - break; + if (err) + werr = err; + return werr; +} - clear_extent_dirty(dirty_pages, start, end, GFP_NOFS); - while (start <= end) { - index = start >> PAGE_CACHE_SHIFT; - start = (u64)(index + 1) << PAGE_CACHE_SHIFT; - page = find_get_page(btree_inode->i_mapping, index); - if (!page) - continue; - if (PageDirty(page)) { - btree_lock_page_hook(page); - wait_on_page_writeback(page); - err = write_one_page(page, 0); - if (err) - werr = err; - } - wait_on_page_writeback(page); - page_cache_release(page); - cond_resched(); - } +/* + * when btree blocks are allocated, they have some corresponding bits set for + * them in one of two extent_io trees. This is used to make sure all of + * those extents are on disk for transaction or log commit. We wait + * on all the pages and clear them from the dirty pages state tree + */ +int btrfs_wait_marked_extents(struct btrfs_root *root, + struct extent_io_tree *dirty_pages, int mark) +{ + int err = 0; + int werr = 0; + struct address_space *mapping = root->fs_info->btree_inode->i_mapping; + struct extent_state *cached_state = NULL; + u64 start = 0; + u64 end; + + while (!find_first_extent_bit(dirty_pages, start, &start, &end, + EXTENT_NEED_WAIT, &cached_state)) { + clear_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT, + 0, 0, &cached_state, GFP_NOFS); + err = filemap_fdatawait_range(mapping, start, end); + if (err) + werr = err; + cond_resched(); + start = end + 1; } if (err) werr = err; return werr; } +/* + * when btree blocks are allocated, they have some corresponding bits set for + * them in one of two extent_io trees. This is used to make sure all of + * those extents are on disk for transaction or log commit + */ +static int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, + struct extent_io_tree *dirty_pages, int mark) +{ + int ret; + int ret2; + struct blk_plug plug; + + blk_start_plug(&plug); + ret = btrfs_write_marked_extents(root, dirty_pages, mark); + blk_finish_plug(&plug); + ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark); + + if (ret) + return ret; + if (ret2) + return ret2; + return 0; +} + int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) { @@ -404,7 +901,8 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, return filemap_write_and_wait(btree_inode->i_mapping); } return btrfs_write_and_wait_marked_extents(root, - &trans->transaction->dirty_pages); + &trans->transaction->dirty_pages, + EXTENT_DIRTY); } /* @@ -422,60 +920,98 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans, { int ret; u64 old_root_bytenr; + u64 old_root_used; struct btrfs_root *tree_root = root->fs_info->tree_root; - btrfs_extent_post_op(trans, root); + old_root_used = btrfs_root_used(&root->root_item); btrfs_write_dirty_block_groups(trans, root); - btrfs_extent_post_op(trans, root); while (1) { old_root_bytenr = btrfs_root_bytenr(&root->root_item); - if (old_root_bytenr == root->node->start) + if (old_root_bytenr == root->node->start && + old_root_used == btrfs_root_used(&root->root_item)) break; - btrfs_set_root_bytenr(&root->root_item, - root->node->start); - btrfs_set_root_level(&root->root_item, - btrfs_header_level(root->node)); - btrfs_set_root_generation(&root->root_item, trans->transid); - - btrfs_extent_post_op(trans, root); + btrfs_set_root_node(&root->root_item, root->node); ret = btrfs_update_root(trans, tree_root, &root->root_key, &root->root_item); - BUG_ON(ret); - btrfs_write_dirty_block_groups(trans, root); - btrfs_extent_post_op(trans, root); + if (ret) + return ret; + + old_root_used = btrfs_root_used(&root->root_item); + ret = btrfs_write_dirty_block_groups(trans, root); + if (ret) + return ret; } + return 0; } /* * update all the cowonly tree roots on disk + * + * The error handling in this function may not be obvious. Any of the + * failures will cause the file system to go offline. We still need + * to clean up the delayed refs. */ -int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, - struct btrfs_root *root) +static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans, + struct btrfs_root *root) { struct btrfs_fs_info *fs_info = root->fs_info; struct list_head *next; struct extent_buffer *eb; + int ret; - btrfs_extent_post_op(trans, fs_info->tree_root); + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) + return ret; eb = btrfs_lock_root_node(fs_info->tree_root); - btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb, 0); + ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, + 0, &eb); btrfs_tree_unlock(eb); free_extent_buffer(eb); - btrfs_extent_post_op(trans, fs_info->tree_root); + if (ret) + return ret; + + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) + return ret; + + ret = btrfs_run_dev_stats(trans, root->fs_info); + if (ret) + return ret; + ret = btrfs_run_dev_replace(trans, root->fs_info); + if (ret) + return ret; + ret = btrfs_run_qgroups(trans, root->fs_info); + if (ret) + return ret; + + /* run_qgroups might have added some more refs */ + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) + return ret; while (!list_empty(&fs_info->dirty_cowonly_roots)) { next = fs_info->dirty_cowonly_roots.next; list_del_init(next); root = list_entry(next, struct btrfs_root, dirty_list); - update_cowonly_root(trans, root); + if (root != fs_info->extent_root) + list_add_tail(&root->dirty_list, + &trans->transaction->switch_commits); + ret = update_cowonly_root(trans, root); + if (ret) + return ret; } + + list_add_tail(&fs_info->extent_root->dirty_list, + &trans->transaction->switch_commits); + btrfs_after_dev_replace_commit(fs_info); + return 0; } @@ -484,256 +1020,115 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, * a dirty root struct and adds it into the list of dead roots that need to * be deleted */ -int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest) +void btrfs_add_dead_root(struct btrfs_root *root) { - struct btrfs_dirty_root *dirty; - - dirty = kmalloc(sizeof(*dirty), GFP_NOFS); - if (!dirty) - return -ENOMEM; - dirty->root = root; - dirty->latest_root = latest; - - mutex_lock(&root->fs_info->trans_mutex); - list_add(&dirty->list, &latest->fs_info->dead_roots); - mutex_unlock(&root->fs_info->trans_mutex); - return 0; + spin_lock(&root->fs_info->trans_lock); + if (list_empty(&root->root_list)) + list_add_tail(&root->root_list, &root->fs_info->dead_roots); + spin_unlock(&root->fs_info->trans_lock); } /* - * at transaction commit time we need to schedule the old roots for - * deletion via btrfs_drop_snapshot. This runs through all the - * reference counted roots that were modified in the current - * transaction and puts them into the drop list + * update all the cowonly tree roots on disk */ -static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, - struct radix_tree_root *radix, - struct list_head *list) +static noinline int commit_fs_roots(struct btrfs_trans_handle *trans, + struct btrfs_root *root) { - struct btrfs_dirty_root *dirty; struct btrfs_root *gang[8]; - struct btrfs_root *root; + struct btrfs_fs_info *fs_info = root->fs_info; int i; int ret; int err = 0; - u32 refs; + spin_lock(&fs_info->fs_roots_radix_lock); while (1) { - ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, + ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, + (void **)gang, 0, ARRAY_SIZE(gang), BTRFS_ROOT_TRANS_TAG); if (ret == 0) break; for (i = 0; i < ret; i++) { root = gang[i]; - radix_tree_tag_clear(radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_TRANS_TAG); - - BUG_ON(!root->ref_tree); - dirty = root->dirty_root; + radix_tree_tag_clear(&fs_info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_TRANS_TAG); + spin_unlock(&fs_info->fs_roots_radix_lock); btrfs_free_log(trans, root); - btrfs_free_reloc_root(trans, root); - - if (root->commit_root == root->node) { - WARN_ON(root->node->start != - btrfs_root_bytenr(&root->root_item)); - - free_extent_buffer(root->commit_root); - root->commit_root = NULL; - root->dirty_root = NULL; + btrfs_update_reloc_root(trans, root); + btrfs_orphan_commit_root(trans, root); - spin_lock(&root->list_lock); - list_del_init(&dirty->root->dead_list); - spin_unlock(&root->list_lock); + btrfs_save_ino_cache(root, trans); - kfree(dirty->root); - kfree(dirty); + /* see comments in should_cow_block() */ + clear_bit(BTRFS_ROOT_FORCE_COW, &root->state); + smp_mb__after_atomic(); - /* make sure to update the root on disk - * so we get any updates to the block used - * counts - */ - err = btrfs_update_root(trans, - root->fs_info->tree_root, - &root->root_key, - &root->root_item); - continue; + if (root->commit_root != root->node) { + list_add_tail(&root->dirty_list, + &trans->transaction->switch_commits); + btrfs_set_root_node(&root->root_item, + root->node); } - memset(&root->root_item.drop_progress, 0, - sizeof(struct btrfs_disk_key)); - root->root_item.drop_level = 0; - root->commit_root = NULL; - root->dirty_root = NULL; - root->root_key.offset = root->fs_info->generation; - btrfs_set_root_bytenr(&root->root_item, - root->node->start); - btrfs_set_root_level(&root->root_item, - btrfs_header_level(root->node)); - btrfs_set_root_generation(&root->root_item, - root->root_key.offset); - - err = btrfs_insert_root(trans, root->fs_info->tree_root, + err = btrfs_update_root(trans, fs_info->tree_root, &root->root_key, &root->root_item); + spin_lock(&fs_info->fs_roots_radix_lock); if (err) break; - - refs = btrfs_root_refs(&dirty->root->root_item); - btrfs_set_root_refs(&dirty->root->root_item, refs - 1); - err = btrfs_update_root(trans, root->fs_info->tree_root, - &dirty->root->root_key, - &dirty->root->root_item); - - BUG_ON(err); - if (refs == 1) { - list_add(&dirty->list, list); - } else { - WARN_ON(1); - free_extent_buffer(dirty->root->node); - kfree(dirty->root); - kfree(dirty); - } } } + spin_unlock(&fs_info->fs_roots_radix_lock); return err; } /* - * defrag a given btree. If cacheonly == 1, this won't read from the disk, - * otherwise every leaf in the btree is read and defragged. + * defrag a given btree. + * Every leaf in the btree is read and defragged. */ -int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) +int btrfs_defrag_root(struct btrfs_root *root) { struct btrfs_fs_info *info = root->fs_info; - int ret; struct btrfs_trans_handle *trans; - unsigned long nr; + int ret; - smp_mb(); - if (root->defrag_running) + if (test_and_set_bit(BTRFS_ROOT_DEFRAG_RUNNING, &root->state)) return 0; - trans = btrfs_start_transaction(root, 1); + while (1) { - root->defrag_running = 1; - ret = btrfs_defrag_leaves(trans, root, cacheonly); - nr = trans->blocks_used; + trans = btrfs_start_transaction(root, 0); + if (IS_ERR(trans)) + return PTR_ERR(trans); + + ret = btrfs_defrag_leaves(trans, root); + btrfs_end_transaction(trans, root); - btrfs_btree_balance_dirty(info->tree_root, nr); + btrfs_btree_balance_dirty(info->tree_root); cond_resched(); - trans = btrfs_start_transaction(root, 1); - if (root->fs_info->closing || ret != -EAGAIN) + if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN) break; - } - root->defrag_running = 0; - smp_mb(); - btrfs_end_transaction(trans, root); - return 0; -} - -/* - * Given a list of roots that need to be deleted, call btrfs_drop_snapshot on - * all of them - */ -static noinline int drop_dirty_roots(struct btrfs_root *tree_root, - struct list_head *list) -{ - struct btrfs_dirty_root *dirty; - struct btrfs_trans_handle *trans; - unsigned long nr; - u64 num_bytes; - u64 bytes_used; - u64 max_useless; - int ret = 0; - int err; - - while (!list_empty(list)) { - struct btrfs_root *root; - - dirty = list_entry(list->prev, struct btrfs_dirty_root, list); - list_del_init(&dirty->list); - - num_bytes = btrfs_root_used(&dirty->root->root_item); - root = dirty->latest_root; - atomic_inc(&root->fs_info->throttles); - - while (1) { - trans = btrfs_start_transaction(tree_root, 1); - mutex_lock(&root->fs_info->drop_mutex); - ret = btrfs_drop_snapshot(trans, dirty->root); - if (ret != -EAGAIN) - break; - mutex_unlock(&root->fs_info->drop_mutex); - err = btrfs_update_root(trans, - tree_root, - &dirty->root->root_key, - &dirty->root->root_item); - if (err) - ret = err; - nr = trans->blocks_used; - ret = btrfs_end_transaction(trans, tree_root); - BUG_ON(ret); - - btrfs_btree_balance_dirty(tree_root, nr); - cond_resched(); - } - BUG_ON(ret); - atomic_dec(&root->fs_info->throttles); - wake_up(&root->fs_info->transaction_throttle); - - num_bytes -= btrfs_root_used(&dirty->root->root_item); - bytes_used = btrfs_root_used(&root->root_item); - if (num_bytes) { - mutex_lock(&root->fs_info->trans_mutex); - btrfs_record_root_in_trans(root); - mutex_unlock(&root->fs_info->trans_mutex); - btrfs_set_root_used(&root->root_item, - bytes_used - num_bytes); - } - - ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); - if (ret) { - BUG(); + if (btrfs_defrag_cancelled(root->fs_info)) { + pr_debug("BTRFS: defrag_root cancelled\n"); + ret = -EAGAIN; break; } - mutex_unlock(&root->fs_info->drop_mutex); - - spin_lock(&root->list_lock); - list_del_init(&dirty->root->dead_list); - if (!list_empty(&root->dead_list)) { - struct btrfs_root *oldest; - oldest = list_entry(root->dead_list.prev, - struct btrfs_root, dead_list); - max_useless = oldest->root_key.offset - 1; - } else { - max_useless = root->root_key.offset - 1; - } - spin_unlock(&root->list_lock); - - nr = trans->blocks_used; - ret = btrfs_end_transaction(trans, tree_root); - BUG_ON(ret); - - ret = btrfs_remove_leaf_refs(root, max_useless, 0); - BUG_ON(ret); - - free_extent_buffer(dirty->root->node); - kfree(dirty->root); - kfree(dirty); - - btrfs_btree_balance_dirty(tree_root, nr); - cond_resched(); } + clear_bit(BTRFS_ROOT_DEFRAG_RUNNING, &root->state); return ret; } /* * new snapshots need to be created at a very specific time in the - * transaction commit. This does the actual creation + * transaction commit. This does the actual creation. + * + * Note: + * If the error which may affect the commitment of the current transaction + * happens, we should return the error number. If the error which just affect + * the creation of the pending snapshots, just return 0. */ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, @@ -743,106 +1138,254 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root_item *new_root_item; struct btrfs_root *tree_root = fs_info->tree_root; struct btrfs_root *root = pending->root; + struct btrfs_root *parent_root; + struct btrfs_block_rsv *rsv; + struct inode *parent_inode; + struct btrfs_path *path; + struct btrfs_dir_item *dir_item; + struct dentry *dentry; struct extent_buffer *tmp; struct extent_buffer *old; - int ret; + struct timespec cur_time = CURRENT_TIME; + int ret = 0; + u64 to_reserve = 0; + u64 index = 0; u64 objectid; + u64 root_flags; + uuid_le new_uuid; + + path = btrfs_alloc_path(); + if (!path) { + pending->error = -ENOMEM; + return 0; + } new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); if (!new_root_item) { - ret = -ENOMEM; + pending->error = -ENOMEM; + goto root_item_alloc_fail; + } + + pending->error = btrfs_find_free_objectid(tree_root, &objectid); + if (pending->error) + goto no_free_objectid; + + btrfs_reloc_pre_snapshot(trans, pending, &to_reserve); + + if (to_reserve > 0) { + pending->error = btrfs_block_rsv_add(root, + &pending->block_rsv, + to_reserve, + BTRFS_RESERVE_NO_FLUSH); + if (pending->error) + goto no_free_objectid; + } + + key.objectid = objectid; + key.offset = (u64)-1; + key.type = BTRFS_ROOT_ITEM_KEY; + + rsv = trans->block_rsv; + trans->block_rsv = &pending->block_rsv; + trans->bytes_reserved = trans->block_rsv->reserved; + + dentry = pending->dentry; + parent_inode = pending->dir; + parent_root = BTRFS_I(parent_inode)->root; + record_root_in_trans(trans, parent_root); + + /* + * insert the directory item + */ + ret = btrfs_set_inode_index(parent_inode, &index); + BUG_ON(ret); /* -ENOMEM */ + + /* check if there is a file/dir which has the same name. */ + dir_item = btrfs_lookup_dir_item(NULL, parent_root, path, + btrfs_ino(parent_inode), + dentry->d_name.name, + dentry->d_name.len, 0); + if (dir_item != NULL && !IS_ERR(dir_item)) { + pending->error = -EEXIST; + goto dir_item_existed; + } else if (IS_ERR(dir_item)) { + ret = PTR_ERR(dir_item); + btrfs_abort_transaction(trans, root, ret); goto fail; } - ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid); - if (ret) + btrfs_release_path(path); + + /* + * pull in the delayed directory update + * and the delayed inode item + * otherwise we corrupt the FS during + * snapshot + */ + ret = btrfs_run_delayed_items(trans, root); + if (ret) { /* Transaction aborted */ + btrfs_abort_transaction(trans, root, ret); goto fail; + } - btrfs_record_root_in_trans(root); + record_root_in_trans(trans, root); btrfs_set_root_last_snapshot(&root->root_item, trans->transid); memcpy(new_root_item, &root->root_item, sizeof(*new_root_item)); - - key.objectid = objectid; - key.offset = trans->transid; - btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + btrfs_check_and_init_root_item(new_root_item); + + root_flags = btrfs_root_flags(new_root_item); + if (pending->readonly) + root_flags |= BTRFS_ROOT_SUBVOL_RDONLY; + else + root_flags &= ~BTRFS_ROOT_SUBVOL_RDONLY; + btrfs_set_root_flags(new_root_item, root_flags); + + btrfs_set_root_generation_v2(new_root_item, + trans->transid); + uuid_le_gen(&new_uuid); + memcpy(new_root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); + memcpy(new_root_item->parent_uuid, root->root_item.uuid, + BTRFS_UUID_SIZE); + if (!(root_flags & BTRFS_ROOT_SUBVOL_RDONLY)) { + memset(new_root_item->received_uuid, 0, + sizeof(new_root_item->received_uuid)); + memset(&new_root_item->stime, 0, sizeof(new_root_item->stime)); + memset(&new_root_item->rtime, 0, sizeof(new_root_item->rtime)); + btrfs_set_root_stransid(new_root_item, 0); + btrfs_set_root_rtransid(new_root_item, 0); + } + btrfs_set_stack_timespec_sec(&new_root_item->otime, cur_time.tv_sec); + btrfs_set_stack_timespec_nsec(&new_root_item->otime, cur_time.tv_nsec); + btrfs_set_root_otransid(new_root_item, trans->transid); old = btrfs_lock_root_node(root); - btrfs_cow_block(trans, root, old, NULL, 0, &old, 0); + ret = btrfs_cow_block(trans, root, old, NULL, 0, &old); + if (ret) { + btrfs_tree_unlock(old); + free_extent_buffer(old); + btrfs_abort_transaction(trans, root, ret); + goto fail; + } + + btrfs_set_lock_blocking(old); - btrfs_copy_root(trans, root, old, &tmp, objectid); + ret = btrfs_copy_root(trans, root, old, &tmp, objectid); + /* clean up in any case */ btrfs_tree_unlock(old); free_extent_buffer(old); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } - btrfs_set_root_bytenr(new_root_item, tmp->start); - btrfs_set_root_level(new_root_item, btrfs_header_level(tmp)); - btrfs_set_root_generation(new_root_item, trans->transid); - ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, - new_root_item); - btrfs_tree_unlock(tmp); - free_extent_buffer(tmp); - if (ret) + /* + * We need to flush delayed refs in order to make sure all of our quota + * operations have been done before we call btrfs_qgroup_inherit. + */ + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) { + btrfs_abort_transaction(trans, root, ret); goto fail; + } - key.offset = (u64)-1; - memcpy(&pending->root_key, &key, sizeof(key)); -fail: - kfree(new_root_item); - return ret; -} + ret = btrfs_qgroup_inherit(trans, fs_info, + root->root_key.objectid, + objectid, pending->inherit); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } -static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info, - struct btrfs_pending_snapshot *pending) -{ - int ret; - int namelen; - u64 index = 0; - struct btrfs_trans_handle *trans; - struct inode *parent_inode; - struct inode *inode; - struct btrfs_root *parent_root; + /* see comments in should_cow_block() */ + set_bit(BTRFS_ROOT_FORCE_COW, &root->state); + smp_wmb(); - parent_inode = pending->dentry->d_parent->d_inode; - parent_root = BTRFS_I(parent_inode)->root; - trans = btrfs_join_transaction(parent_root, 1); + btrfs_set_root_node(new_root_item, tmp); + /* record when the snapshot was created in key.offset */ + key.offset = trans->transid; + ret = btrfs_insert_root(trans, tree_root, &key, new_root_item); + btrfs_tree_unlock(tmp); + free_extent_buffer(tmp); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } /* - * insert the directory item + * insert root back/forward references */ - namelen = strlen(pending->name); - ret = btrfs_set_inode_index(parent_inode, &index); - ret = btrfs_insert_dir_item(trans, parent_root, - pending->name, namelen, - parent_inode->i_ino, - &pending->root_key, BTRFS_FT_DIR, index); - - if (ret) + ret = btrfs_add_root_ref(trans, tree_root, objectid, + parent_root->root_key.objectid, + btrfs_ino(parent_inode), index, + dentry->d_name.name, dentry->d_name.len); + if (ret) { + btrfs_abort_transaction(trans, root, ret); goto fail; + } - btrfs_i_size_write(parent_inode, parent_inode->i_size + namelen * 2); - ret = btrfs_update_inode(trans, parent_root, parent_inode); - BUG_ON(ret); + key.offset = (u64)-1; + pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key); + if (IS_ERR(pending->snap)) { + ret = PTR_ERR(pending->snap); + btrfs_abort_transaction(trans, root, ret); + goto fail; + } - /* add the backref first */ - ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root, - pending->root_key.objectid, - BTRFS_ROOT_BACKREF_KEY, - parent_root->root_key.objectid, - parent_inode->i_ino, index, pending->name, - namelen); + ret = btrfs_reloc_post_snapshot(trans, pending); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } - BUG_ON(ret); + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } - /* now add the forward ref */ - ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root, - parent_root->root_key.objectid, - BTRFS_ROOT_REF_KEY, - pending->root_key.objectid, - parent_inode->i_ino, index, pending->name, - namelen); + ret = btrfs_insert_dir_item(trans, parent_root, + dentry->d_name.name, dentry->d_name.len, + parent_inode, &key, + BTRFS_FT_DIR, index); + /* We have check then name at the beginning, so it is impossible. */ + BUG_ON(ret == -EEXIST || ret == -EOVERFLOW); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } - inode = btrfs_lookup_dentry(parent_inode, pending->dentry); - d_instantiate(pending->dentry, inode); + btrfs_i_size_write(parent_inode, parent_inode->i_size + + dentry->d_name.len * 2); + parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; + ret = btrfs_update_inode_fallback(trans, parent_root, parent_inode); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } + ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root, new_uuid.b, + BTRFS_UUID_KEY_SUBVOL, objectid); + if (ret) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } + if (!btrfs_is_empty_uuid(new_root_item->received_uuid)) { + ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root, + new_root_item->received_uuid, + BTRFS_UUID_KEY_RECEIVED_SUBVOL, + objectid); + if (ret && ret != -EEXIST) { + btrfs_abort_transaction(trans, root, ret); + goto fail; + } + } fail: - btrfs_end_transaction(trans, fs_info->fs_root); + pending->error = ret; +dir_item_existed: + trans->block_rsv = rsv; + trans->bytes_reserved = 0; +no_free_objectid: + kfree(new_root_item); +root_item_alloc_fail: + btrfs_free_path(path); return ret; } @@ -852,122 +1395,427 @@ fail: static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info) { - struct btrfs_pending_snapshot *pending; + struct btrfs_pending_snapshot *pending, *next; struct list_head *head = &trans->transaction->pending_snapshots; - int ret; + int ret = 0; - list_for_each_entry(pending, head, list) { + list_for_each_entry_safe(pending, next, head, list) { + list_del(&pending->list); ret = create_pending_snapshot(trans, fs_info, pending); - BUG_ON(ret); + if (ret) + break; } + return ret; +} + +static void update_super_roots(struct btrfs_root *root) +{ + struct btrfs_root_item *root_item; + struct btrfs_super_block *super; + + super = root->fs_info->super_copy; + + root_item = &root->fs_info->chunk_root->root_item; + super->chunk_root = root_item->bytenr; + super->chunk_root_generation = root_item->generation; + super->chunk_root_level = root_item->level; + + root_item = &root->fs_info->tree_root->root_item; + super->root = root_item->bytenr; + super->generation = root_item->generation; + super->root_level = root_item->level; + if (btrfs_test_opt(root, SPACE_CACHE)) + super->cache_generation = root_item->generation; + if (root->fs_info->update_uuid_tree_gen) + super->uuid_tree_generation = root_item->generation; +} + +int btrfs_transaction_in_commit(struct btrfs_fs_info *info) +{ + struct btrfs_transaction *trans; + int ret = 0; + + spin_lock(&info->trans_lock); + trans = info->running_transaction; + if (trans) + ret = (trans->state >= TRANS_STATE_COMMIT_START); + spin_unlock(&info->trans_lock); + return ret; +} + +int btrfs_transaction_blocked(struct btrfs_fs_info *info) +{ + struct btrfs_transaction *trans; + int ret = 0; + + spin_lock(&info->trans_lock); + trans = info->running_transaction; + if (trans) + ret = is_transaction_blocked(trans); + spin_unlock(&info->trans_lock); + return ret; +} + +/* + * wait for the current transaction commit to start and block subsequent + * transaction joins + */ +static void wait_current_trans_commit_start(struct btrfs_root *root, + struct btrfs_transaction *trans) +{ + wait_event(root->fs_info->transaction_blocked_wait, + trans->state >= TRANS_STATE_COMMIT_START || + trans->aborted); +} + +/* + * wait for the current transaction to start and then become unblocked. + * caller holds ref. + */ +static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root, + struct btrfs_transaction *trans) +{ + wait_event(root->fs_info->transaction_wait, + trans->state >= TRANS_STATE_UNBLOCKED || + trans->aborted); +} + +/* + * commit transactions asynchronously. once btrfs_commit_transaction_async + * returns, any subsequent transaction will not be allowed to join. + */ +struct btrfs_async_commit { + struct btrfs_trans_handle *newtrans; + struct btrfs_root *root; + struct work_struct work; +}; + +static void do_async_commit(struct work_struct *work) +{ + struct btrfs_async_commit *ac = + container_of(work, struct btrfs_async_commit, work); + + /* + * We've got freeze protection passed with the transaction. + * Tell lockdep about it. + */ + if (ac->newtrans->type & __TRANS_FREEZABLE) + rwsem_acquire_read( + &ac->root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1], + 0, 1, _THIS_IP_); + + current->journal_info = ac->newtrans; + + btrfs_commit_transaction(ac->newtrans, ac->root); + kfree(ac); +} + +int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + int wait_for_unblock) +{ + struct btrfs_async_commit *ac; + struct btrfs_transaction *cur_trans; + + ac = kmalloc(sizeof(*ac), GFP_NOFS); + if (!ac) + return -ENOMEM; + + INIT_WORK(&ac->work, do_async_commit); + ac->root = root; + ac->newtrans = btrfs_join_transaction(root); + if (IS_ERR(ac->newtrans)) { + int err = PTR_ERR(ac->newtrans); + kfree(ac); + return err; + } + + /* take transaction reference */ + cur_trans = trans->transaction; + atomic_inc(&cur_trans->use_count); + + btrfs_end_transaction(trans, root); + + /* + * Tell lockdep we've released the freeze rwsem, since the + * async commit thread will be the one to unlock it. + */ + if (ac->newtrans->type & __TRANS_FREEZABLE) + rwsem_release( + &root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1], + 1, _THIS_IP_); + + schedule_work(&ac->work); + + /* wait for transaction to start and unblock */ + if (wait_for_unblock) + wait_current_trans_commit_start_and_unblock(root, cur_trans); + else + wait_current_trans_commit_start(root, cur_trans); + + if (current->journal_info == trans) + current->journal_info = NULL; + + btrfs_put_transaction(cur_trans); return 0; } -static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans, - struct btrfs_fs_info *fs_info) + +static void cleanup_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int err) { - struct btrfs_pending_snapshot *pending; - struct list_head *head = &trans->transaction->pending_snapshots; - int ret; + struct btrfs_transaction *cur_trans = trans->transaction; + DEFINE_WAIT(wait); - while (!list_empty(head)) { - pending = list_entry(head->next, - struct btrfs_pending_snapshot, list); - ret = finish_pending_snapshot(fs_info, pending); - BUG_ON(ret); - list_del(&pending->list); - kfree(pending->name); - kfree(pending); + WARN_ON(trans->use_count > 1); + + btrfs_abort_transaction(trans, root, err); + + spin_lock(&root->fs_info->trans_lock); + + /* + * If the transaction is removed from the list, it means this + * transaction has been committed successfully, so it is impossible + * to call the cleanup function. + */ + BUG_ON(list_empty(&cur_trans->list)); + + list_del_init(&cur_trans->list); + if (cur_trans == root->fs_info->running_transaction) { + cur_trans->state = TRANS_STATE_COMMIT_DOING; + spin_unlock(&root->fs_info->trans_lock); + wait_event(cur_trans->writer_wait, + atomic_read(&cur_trans->num_writers) == 1); + + spin_lock(&root->fs_info->trans_lock); } + spin_unlock(&root->fs_info->trans_lock); + + btrfs_cleanup_one_transaction(trans->transaction, root); + + spin_lock(&root->fs_info->trans_lock); + if (cur_trans == root->fs_info->running_transaction) + root->fs_info->running_transaction = NULL; + spin_unlock(&root->fs_info->trans_lock); + + if (trans->type & __TRANS_FREEZABLE) + sb_end_intwrite(root->fs_info->sb); + btrfs_put_transaction(cur_trans); + btrfs_put_transaction(cur_trans); + + trace_btrfs_transaction_commit(root); + + if (current->journal_info == trans) + current->journal_info = NULL; + btrfs_scrub_cancel(root->fs_info); + + kmem_cache_free(btrfs_trans_handle_cachep, trans); +} + +static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + int ret; + + ret = btrfs_run_delayed_items(trans, root); + if (ret) + return ret; + + /* + * rename don't use btrfs_join_transaction, so, once we + * set the transaction to blocked above, we aren't going + * to get any new ordered operations. We can safely run + * it here and no for sure that nothing new will be added + * to the list + */ + ret = btrfs_run_ordered_operations(trans, root, 1); + + return ret; +} + +static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info) +{ + if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT)) + return btrfs_start_delalloc_roots(fs_info, 1, -1); return 0; } +static inline void btrfs_wait_delalloc_flush(struct btrfs_fs_info *fs_info) +{ + if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT)) + btrfs_wait_ordered_roots(fs_info, -1); +} + int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - unsigned long joined = 0; - unsigned long timeout = 1; - struct btrfs_transaction *cur_trans; + struct btrfs_transaction *cur_trans = trans->transaction; struct btrfs_transaction *prev_trans = NULL; - struct btrfs_root *chunk_root = root->fs_info->chunk_root; - struct list_head dirty_fs_roots; - struct extent_io_tree *pinned_copy; - DEFINE_WAIT(wait); int ret; - INIT_LIST_HEAD(&dirty_fs_roots); - mutex_lock(&root->fs_info->trans_mutex); - if (trans->transaction->in_commit) { - cur_trans = trans->transaction; - trans->transaction->use_count++; - mutex_unlock(&root->fs_info->trans_mutex); + ret = btrfs_run_ordered_operations(trans, root, 0); + if (ret) { + btrfs_abort_transaction(trans, root, ret); btrfs_end_transaction(trans, root); + return ret; + } - ret = wait_for_commit(root, cur_trans); - BUG_ON(ret); + /* Stop the commit early if ->aborted is set */ + if (unlikely(ACCESS_ONCE(cur_trans->aborted))) { + ret = cur_trans->aborted; + btrfs_end_transaction(trans, root); + return ret; + } - mutex_lock(&root->fs_info->trans_mutex); - put_transaction(cur_trans); - mutex_unlock(&root->fs_info->trans_mutex); + /* make a pass through all the delayed refs we have so far + * any runnings procs may add more while we are here + */ + ret = btrfs_run_delayed_refs(trans, root, 0); + if (ret) { + btrfs_end_transaction(trans, root); + return ret; + } - return 0; + btrfs_trans_release_metadata(trans, root); + trans->block_rsv = NULL; + if (trans->qgroup_reserved) { + btrfs_qgroup_free(root, trans->qgroup_reserved); + trans->qgroup_reserved = 0; } - pinned_copy = kmalloc(sizeof(*pinned_copy), GFP_NOFS); - if (!pinned_copy) - return -ENOMEM; + cur_trans = trans->transaction; - extent_io_tree_init(pinned_copy, - root->fs_info->btree_inode->i_mapping, GFP_NOFS); + /* + * set the flushing flag so procs in this transaction have to + * start sending their work down. + */ + cur_trans->delayed_refs.flushing = 1; + smp_wmb(); + + if (!list_empty(&trans->new_bgs)) + btrfs_create_pending_block_groups(trans, root); + + ret = btrfs_run_delayed_refs(trans, root, 0); + if (ret) { + btrfs_end_transaction(trans, root); + return ret; + } + + spin_lock(&root->fs_info->trans_lock); + if (cur_trans->state >= TRANS_STATE_COMMIT_START) { + spin_unlock(&root->fs_info->trans_lock); + atomic_inc(&cur_trans->use_count); + ret = btrfs_end_transaction(trans, root); + + wait_for_commit(root, cur_trans); + + btrfs_put_transaction(cur_trans); + + return ret; + } + + cur_trans->state = TRANS_STATE_COMMIT_START; + wake_up(&root->fs_info->transaction_blocked_wait); - trans->transaction->in_commit = 1; - trans->transaction->blocked = 1; - cur_trans = trans->transaction; if (cur_trans->list.prev != &root->fs_info->trans_list) { prev_trans = list_entry(cur_trans->list.prev, struct btrfs_transaction, list); - if (!prev_trans->commit_done) { - prev_trans->use_count++; - mutex_unlock(&root->fs_info->trans_mutex); + if (prev_trans->state != TRANS_STATE_COMPLETED) { + atomic_inc(&prev_trans->use_count); + spin_unlock(&root->fs_info->trans_lock); wait_for_commit(root, prev_trans); - mutex_lock(&root->fs_info->trans_mutex); - put_transaction(prev_trans); + btrfs_put_transaction(prev_trans); + } else { + spin_unlock(&root->fs_info->trans_lock); } + } else { + spin_unlock(&root->fs_info->trans_lock); } - do { - int snap_pending = 0; - joined = cur_trans->num_joined; - if (!list_empty(&trans->transaction->pending_snapshots)) - snap_pending = 1; + extwriter_counter_dec(cur_trans, trans->type); - WARN_ON(cur_trans != trans->transaction); - prepare_to_wait(&cur_trans->writer_wait, &wait, - TASK_UNINTERRUPTIBLE); + ret = btrfs_start_delalloc_flush(root->fs_info); + if (ret) + goto cleanup_transaction; - if (cur_trans->num_writers > 1) - timeout = MAX_SCHEDULE_TIMEOUT; - else - timeout = 1; + ret = btrfs_flush_all_pending_stuffs(trans, root); + if (ret) + goto cleanup_transaction; - mutex_unlock(&root->fs_info->trans_mutex); + wait_event(cur_trans->writer_wait, + extwriter_counter_read(cur_trans) == 0); - if (snap_pending) { - ret = btrfs_wait_ordered_extents(root, 1); - BUG_ON(ret); - } + /* some pending stuffs might be added after the previous flush. */ + ret = btrfs_flush_all_pending_stuffs(trans, root); + if (ret) + goto cleanup_transaction; - schedule_timeout(timeout); + btrfs_wait_delalloc_flush(root->fs_info); - mutex_lock(&root->fs_info->trans_mutex); - finish_wait(&cur_trans->writer_wait, &wait); - } while (cur_trans->num_writers > 1 || - (cur_trans->num_joined != joined)); + btrfs_scrub_pause(root); + /* + * Ok now we need to make sure to block out any other joins while we + * commit the transaction. We could have started a join before setting + * COMMIT_DOING so make sure to wait for num_writers to == 1 again. + */ + spin_lock(&root->fs_info->trans_lock); + cur_trans->state = TRANS_STATE_COMMIT_DOING; + spin_unlock(&root->fs_info->trans_lock); + wait_event(cur_trans->writer_wait, + atomic_read(&cur_trans->num_writers) == 1); + + /* ->aborted might be set after the previous check, so check it */ + if (unlikely(ACCESS_ONCE(cur_trans->aborted))) { + ret = cur_trans->aborted; + goto scrub_continue; + } + /* + * the reloc mutex makes sure that we stop + * the balancing code from coming in and moving + * extents around in the middle of the commit + */ + mutex_lock(&root->fs_info->reloc_mutex); + /* + * We needn't worry about the delayed items because we will + * deal with them in create_pending_snapshot(), which is the + * core function of the snapshot creation. + */ ret = create_pending_snapshots(trans, root->fs_info); - BUG_ON(ret); + if (ret) { + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } + + /* + * We insert the dir indexes of the snapshots and update the inode + * of the snapshots' parents after the snapshot creation, so there + * are some delayed items which are not dealt with. Now deal with + * them. + * + * We needn't worry that this operation will corrupt the snapshots, + * because all the tree which are snapshoted will be forced to COW + * the nodes and leaves. + */ + ret = btrfs_run_delayed_items(trans, root); + if (ret) { + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } + + ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1); + if (ret) { + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } + + /* + * make sure none of the code above managed to slip in a + * delayed item + */ + btrfs_assert_delayed_root_empty(root); WARN_ON(cur_trans != trans->transaction); @@ -985,60 +1833,91 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, * with the tree-log code. */ mutex_lock(&root->fs_info->tree_log_mutex); - /* - * keep tree reloc code from adding new reloc trees - */ - mutex_lock(&root->fs_info->tree_reloc_mutex); + ret = commit_fs_roots(trans, root); + if (ret) { + mutex_unlock(&root->fs_info->tree_log_mutex); + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } - ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, - &dirty_fs_roots); - BUG_ON(ret); + /* + * Since the transaction is done, we should set the inode map cache flag + * before any other comming transaction. + */ + if (btrfs_test_opt(root, CHANGE_INODE_CACHE)) + btrfs_set_opt(root->fs_info->mount_opt, INODE_MAP_CACHE); + else + btrfs_clear_opt(root->fs_info->mount_opt, INODE_MAP_CACHE); - /* add_dirty_roots gets rid of all the tree log roots, it is now + /* commit_fs_roots gets rid of all the tree log roots, it is now * safe to free the root of tree log roots */ btrfs_free_log_root_tree(trans, root->fs_info); - ret = btrfs_commit_tree_roots(trans, root); - BUG_ON(ret); + ret = commit_cowonly_roots(trans, root); + if (ret) { + mutex_unlock(&root->fs_info->tree_log_mutex); + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } + + /* + * The tasks which save the space cache and inode cache may also + * update ->aborted, check it. + */ + if (unlikely(ACCESS_ONCE(cur_trans->aborted))) { + ret = cur_trans->aborted; + mutex_unlock(&root->fs_info->tree_log_mutex); + mutex_unlock(&root->fs_info->reloc_mutex); + goto scrub_continue; + } + + btrfs_prepare_extent_commit(trans, root); cur_trans = root->fs_info->running_transaction; - spin_lock(&root->fs_info->new_trans_lock); - root->fs_info->running_transaction = NULL; - spin_unlock(&root->fs_info->new_trans_lock); - btrfs_set_super_generation(&root->fs_info->super_copy, - cur_trans->transid); - btrfs_set_super_root(&root->fs_info->super_copy, - root->fs_info->tree_root->node->start); - btrfs_set_super_root_level(&root->fs_info->super_copy, - btrfs_header_level(root->fs_info->tree_root->node)); - btrfs_set_super_chunk_root(&root->fs_info->super_copy, - chunk_root->node->start); - btrfs_set_super_chunk_root_level(&root->fs_info->super_copy, - btrfs_header_level(chunk_root->node)); - btrfs_set_super_chunk_root_generation(&root->fs_info->super_copy, - btrfs_header_generation(chunk_root->node)); + btrfs_set_root_node(&root->fs_info->tree_root->root_item, + root->fs_info->tree_root->node); + list_add_tail(&root->fs_info->tree_root->dirty_list, + &cur_trans->switch_commits); - if (!root->fs_info->log_root_recovering) { - btrfs_set_super_log_root(&root->fs_info->super_copy, 0); - btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0); - } + btrfs_set_root_node(&root->fs_info->chunk_root->root_item, + root->fs_info->chunk_root->node); + list_add_tail(&root->fs_info->chunk_root->dirty_list, + &cur_trans->switch_commits); + + switch_commit_roots(cur_trans, root->fs_info); + + assert_qgroups_uptodate(trans); + update_super_roots(root); - memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy, - sizeof(root->fs_info->super_copy)); + btrfs_set_super_log_root(root->fs_info->super_copy, 0); + btrfs_set_super_log_root_level(root->fs_info->super_copy, 0); + memcpy(root->fs_info->super_for_commit, root->fs_info->super_copy, + sizeof(*root->fs_info->super_copy)); - btrfs_copy_pinned(root, pinned_copy); + spin_lock(&root->fs_info->trans_lock); + cur_trans->state = TRANS_STATE_UNBLOCKED; + root->fs_info->running_transaction = NULL; + spin_unlock(&root->fs_info->trans_lock); + mutex_unlock(&root->fs_info->reloc_mutex); - trans->transaction->blocked = 0; - wake_up(&root->fs_info->transaction_throttle); wake_up(&root->fs_info->transaction_wait); - mutex_unlock(&root->fs_info->trans_mutex); ret = btrfs_write_and_wait_transaction(trans, root); - BUG_ON(ret); - write_ctree_super(trans, root, 0); + if (ret) { + btrfs_error(root->fs_info, ret, + "Error while writing out transaction"); + mutex_unlock(&root->fs_info->tree_log_mutex); + goto scrub_continue; + } + + ret = write_ctree_super(trans, root, 0); + if (ret) { + mutex_unlock(&root->fs_info->tree_log_mutex); + goto scrub_continue; + } /* * the super is written, we can safely allow the tree-loggers @@ -1046,52 +1925,94 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, */ mutex_unlock(&root->fs_info->tree_log_mutex); - btrfs_finish_extent_commit(trans, root, pinned_copy); - kfree(pinned_copy); + btrfs_finish_extent_commit(trans, root); - btrfs_drop_dead_reloc_roots(root); - mutex_unlock(&root->fs_info->tree_reloc_mutex); + root->fs_info->last_trans_committed = cur_trans->transid; + /* + * We needn't acquire the lock here because there is no other task + * which can change it. + */ + cur_trans->state = TRANS_STATE_COMPLETED; + wake_up(&cur_trans->commit_wait); - /* do the directory inserts of any pending snapshot creations */ - finish_pending_snapshots(trans, root->fs_info); + spin_lock(&root->fs_info->trans_lock); + list_del_init(&cur_trans->list); + spin_unlock(&root->fs_info->trans_lock); - mutex_lock(&root->fs_info->trans_mutex); + btrfs_put_transaction(cur_trans); + btrfs_put_transaction(cur_trans); - cur_trans->commit_done = 1; - root->fs_info->last_trans_committed = cur_trans->transid; - wake_up(&cur_trans->commit_wait); + if (trans->type & __TRANS_FREEZABLE) + sb_end_intwrite(root->fs_info->sb); - put_transaction(cur_trans); - put_transaction(cur_trans); + trace_btrfs_transaction_commit(root); - list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); - if (root->fs_info->closing) - list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); + btrfs_scrub_continue(root); - mutex_unlock(&root->fs_info->trans_mutex); + if (current->journal_info == trans) + current->journal_info = NULL; kmem_cache_free(btrfs_trans_handle_cachep, trans); - if (root->fs_info->closing) - drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots); + if (current != root->fs_info->transaction_kthread) + btrfs_run_delayed_iputs(root); + + return ret; + +scrub_continue: + btrfs_scrub_continue(root); +cleanup_transaction: + btrfs_trans_release_metadata(trans, root); + trans->block_rsv = NULL; + if (trans->qgroup_reserved) { + btrfs_qgroup_free(root, trans->qgroup_reserved); + trans->qgroup_reserved = 0; + } + btrfs_warn(root->fs_info, "Skipping commit of aborted transaction."); + if (current->journal_info == trans) + current->journal_info = NULL; + cleanup_transaction(trans, root, ret); + return ret; } /* - * interface function to delete all the snapshots we have scheduled for deletion + * return < 0 if error + * 0 if there are no more dead_roots at the time of call + * 1 there are more to be processed, call me again + * + * The return value indicates there are certainly more snapshots to delete, but + * if there comes a new one during processing, it may return 0. We don't mind, + * because btrfs_commit_super will poke cleaner thread and it will process it a + * few seconds later. */ -int btrfs_clean_old_snapshots(struct btrfs_root *root) +int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root) { - struct list_head dirty_roots; - INIT_LIST_HEAD(&dirty_roots); -again: - mutex_lock(&root->fs_info->trans_mutex); - list_splice_init(&root->fs_info->dead_roots, &dirty_roots); - mutex_unlock(&root->fs_info->trans_mutex); + int ret; + struct btrfs_fs_info *fs_info = root->fs_info; - if (!list_empty(&dirty_roots)) { - drop_dirty_roots(root, &dirty_roots); - goto again; + spin_lock(&fs_info->trans_lock); + if (list_empty(&fs_info->dead_roots)) { + spin_unlock(&fs_info->trans_lock); + return 0; } - return 0; + root = list_first_entry(&fs_info->dead_roots, + struct btrfs_root, root_list); + list_del_init(&root->root_list); + spin_unlock(&fs_info->trans_lock); + + pr_debug("BTRFS: cleaner removing %llu\n", root->objectid); + + btrfs_kill_all_delayed_nodes(root); + + if (btrfs_header_backref_rev(root->node) < + BTRFS_MIXED_BACKREF_REV) + ret = btrfs_drop_snapshot(root, NULL, 0, 0); + else + ret = btrfs_drop_snapshot(root, NULL, 1, 0); + /* + * If we encounter a transaction abort during snapshot cleaning, we + * don't want to crash here + */ + return (ret < 0) ? 0 : 1; } |