18 #include "transaction.h"
39  * Transaction states and transitions
41  * No running transaction (fs tree blocks are not modified)
46  * Transaction N [[TRANS_STATE_RUNNING]]
48  * | New trans handles can be attached to transaction N by calling all
53  * |  transaction N
55  * Transaction N [[TRANS_STATE_COMMIT_PREP]]
58  * | the race and the rest will wait for the winner to commit the transaction.
60  * | The winner will wait for previous running transaction to completely finish
63  * Transaction N [[TRANS_STATE_COMMIT_START]]
68  * | - Wait for current transaction to be committed by others.
72  * | to this running transaction.
74  * | transaction N+1.
77  * |  Caller is chosen to commit transaction N, and all other trans handle
80  * Transaction N [[TRANS_STATE_COMMIT_DOING]]
82  * | The heavy lifting transaction work is started.
89  * | transaction to finish and attach to transaction N+1.
94  * Transaction N [[TRANS_STATE_UNBLOCKED]]
95  * |						    Transaction N+1
100  * | At this stage, new transaction is allowed to   |
109  * Transaction N [[TRANS_STATE_COMPLETED]]	    V
110  *   All tree blocks and super blocks are written.  Transaction N+1
111  *   This transaction is finished and all its	    [[TRANS_STATE_COMMIT_START]]
139 void btrfs_put_transaction(struct btrfs_transaction *transaction)  in btrfs_put_transaction()  argument
141 	WARN_ON(refcount_read(&transaction->use_count) == 0);  in btrfs_put_transaction()
142 	if (refcount_dec_and_test(&transaction->use_count)) {  in btrfs_put_transaction()
143 		BUG_ON(!list_empty(&transaction->list));  in btrfs_put_transaction()
145 				&transaction->delayed_refs.href_root.rb_root));  in btrfs_put_transaction()
146 		WARN_ON(!xa_empty(&transaction->delayed_refs.dirty_extents));  in btrfs_put_transaction()
147 		if (transaction->delayed_refs.pending_csums)  in btrfs_put_transaction()
148 			btrfs_err(transaction->fs_info,  in btrfs_put_transaction()
150 				  transaction->delayed_refs.pending_csums);  in btrfs_put_transaction()
153 		 * because the transaction was aborted and a commit did not  in btrfs_put_transaction()
158 		while (!list_empty(&transaction->deleted_bgs)) {  in btrfs_put_transaction()
161 			cache = list_first_entry(&transaction->deleted_bgs,  in btrfs_put_transaction()
168 		WARN_ON(!list_empty(&transaction->dev_update_list));  in btrfs_put_transaction()
169 		kfree(transaction);  in btrfs_put_transaction()
175 	struct btrfs_transaction *cur_trans = trans->transaction;  in switch_commit_roots()
180 	 * At this point no one can be using this transaction to modify any tree  in switch_commit_roots()
181 	 * and no one can start another transaction to modify any tree either.  in switch_commit_roots()
260  * either allocate a new transaction or hop into the existing one
297 	 * current transaction, and commit it. If there is no transaction, just  in join_transaction()
304 	 * JOIN_NOLOCK only happens during the transaction commit, so  in join_transaction()
319 		 * someone started a transaction after we unlocked.  Make sure  in join_transaction()
344 	 * commit the transaction.  in join_transaction()
357 	 * although the tree mod log is per file system and not per transaction,  in join_transaction()
358 	 * the log must never go across transaction boundaries.  in join_transaction()
362 		WARN(1, KERN_ERR "BTRFS: tree_mod_seq_list not empty when creating a fresh transaction\n");  in join_transaction()
364 		WARN(1, KERN_ERR "BTRFS: tree_mod_log rb tree not empty when creating a fresh transaction\n");  in join_transaction()
395  * is properly recorded in a given transaction.  This is required to make sure
396  * the old root from before we joined the transaction is deleted when the
397  * transaction commits.
436 		 * this transaction.  in record_root_in_trans()
464 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_add_dropped_root()
466 	/* Add ourselves to the transaction dropped list */  in btrfs_add_dropped_root()
511 /* wait for commit against the current transaction to become unblocked
512  * when this is done, it is safe to start a new transaction, but the current
513  * transaction might not be fully on disk.
570 	 * do 1 enospc flushing cycle per transaction start.  in btrfs_reserve_trans_metadata()
619 	 * Do the reservation before we join the transaction so we can do all  in start_transaction()
626 		 * transaction that could free this reserved space prematurely  in start_transaction()
683 	 * If we are JOIN_NOLOCK we're already committing a transaction and  in start_transaction()
690 	 * transaction and commit it, so we needn't do sb_start_intwrite().   in start_transaction()
714 	h->transaction = cur_trans;  in start_transaction()
731 		trace_btrfs_space_reservation(fs_info, "transaction",  in start_transaction()
767 	 * transaction.  in start_transaction()
775 		 * The transaction handle is fully initialized and linked with  in start_transaction()
783 	 * Now that we have found a transaction to be a part of, convert the  in start_transaction()
784 	 * qgroup reservation from prealloc to pertrans. A different transaction  in start_transaction()
835  * Similar to regular join but it never starts a transaction when none is
838  * happen if the transaction commit already started but it's not yet in the
848  * Catch the running transaction.
850  * It is used when we want to commit the current the transaction, but
854  * running transaction. But it is possible that the inactive transaction
856  * inactive transaction in the fs when -ENOENT is returned, you should
867  * Catch the running transaction.
891 /* Wait for a transaction commit to reach at least the given state. */
917 		 * A transaction isn't really completed until all of the  in wait_for_commit()
920 		 * transaction. Wait for those.  in wait_for_commit()
946 		/* find specified transaction */  in btrfs_wait_for_commit()
963 		 * The specified transaction doesn't exist, or we  in btrfs_wait_for_commit()
972 		/* find newest transaction that is committing | committed */  in btrfs_wait_for_commit()
1003 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_should_end_transaction()
1032 	trace_btrfs_space_reservation(fs_info, "transaction",  in btrfs_trans_release_metadata()
1053 	struct btrfs_transaction *cur_trans = trans->transaction;  in __btrfs_end_transaction()
1144 		 * the transaction (through extent_io_tree_release()).  in btrfs_write_marked_extents()
1167  * those extents are on disk for transaction or log commit.  We wait
1183 		 * When committing the transaction, we'll remove any entries  in __btrfs_wait_marked_extents()
1186 		 * concurrently - we do it only at transaction commit time when  in __btrfs_wait_marked_extents()
1245  * This function ensures such extents are persisted on disk for transaction or
1248  * @trans: transaction whose dirty pages we'd like to write
1254 	struct extent_io_tree *dirty_pages = &trans->transaction->dirty_pages;  in btrfs_write_and_wait_transaction()
1263 	extent_io_tree_release(&trans->transaction->dirty_pages);  in btrfs_write_and_wait_transaction()
1323 	struct list_head *dirty_bgs = &trans->transaction->dirty_bgs;  in commit_cowonly_roots()
1324 	struct list_head *io_bgs = &trans->transaction->io_bgs;  in commit_cowonly_roots()
1330 	 * At this point no one can be using this transaction to modify any tree  in commit_cowonly_roots()
1331 	 * and no one can start another transaction to modify any tree either.  in commit_cowonly_roots()
1333 	ASSERT(trans->transaction->state == TRANS_STATE_COMMIT_DOING);  in commit_cowonly_roots()
1367 			      &trans->transaction->switch_commits);  in commit_cowonly_roots()
1464 	 * At this point no one can be using this transaction to modify any tree  in commit_fs_roots()
1465 	 * and no one can start another transaction to modify any tree either.  in commit_fs_roots()
1467 	ASSERT(trans->transaction->state == TRANS_STATE_COMMIT_DOING);  in commit_fs_roots()
1506 					&trans->transaction->switch_commits);  in commit_fs_roots()
1527  * roots inside one transaction and write all btree into disk, to make
1587 	 * Now we do a simplified commit transaction, which will:  in qgroup_account_snapshot()
1605 			"Error while writing out transaction for qgroup");  in qgroup_account_snapshot()
1621  * transaction commit.  This does the actual creation.
1624  * If the error which may affect the commitment of the current transaction
1660 	 * We're inside a transaction and must make sure that any potential  in create_pending_snapshot()
1700 	trace_btrfs_space_reservation(fs_info, "transaction",  in create_pending_snapshot()
1745 	if (ret) {	/* Transaction aborted */  in create_pending_snapshot()
1916 	struct list_head *head = &trans->transaction->pending_snapshots;  in create_pending_snapshots()
1970 	/* Kick the transaction kthread. */  in btrfs_commit_transaction_async()
1974 	/* take transaction reference */  in btrfs_commit_transaction_async()
1975 	cur_trans = trans->transaction;  in btrfs_commit_transaction_async()
1981 	 * Wait for the current transaction commit to start and block  in btrfs_commit_transaction_async()
1982 	 * subsequent transaction joins  in btrfs_commit_transaction_async()
1992  * If there is a running transaction commit it or if it's already committing,
1993  * wait for its commit to complete. Does not start and commit a new transaction
2013 	struct btrfs_transaction *cur_trans = trans->transaction;  in cleanup_transaction()
2022 	 * If the transaction is removed from the list, it means this  in cleanup_transaction()
2023 	 * transaction has been committed successfully, so it is impossible  in cleanup_transaction()
2044 	 * Now that we know no one else is still using the transaction we can  in cleanup_transaction()
2045 	 * remove the transaction from the list of transactions. This avoids  in cleanup_transaction()
2046 	 * the transaction kthread from cleaning up the transaction while some  in cleanup_transaction()
2048 	 * on things like log trees, as it forces the transaction kthread to  in cleanup_transaction()
2049 	 * wait for this transaction to be cleaned up by us.  in cleanup_transaction()
2055 	btrfs_cleanup_one_transaction(trans->transaction, fs_info);  in cleanup_transaction()
2075 	 * pauses scrub, then starts and commits a transaction before unpausing  in cleanup_transaction()
2076 	 * scrub. If the transaction commit is being done by the relocation  in cleanup_transaction()
2091  * transaction and remove them from the list
2112 	 * from already being in a transaction and our join_transaction doesn't  in btrfs_start_delalloc_flush()
2135  * Add a pending snapshot associated with the given transaction handle to the
2136  * respective handle. This must be called after the transaction commit started
2144 	struct btrfs_transaction *cur_trans = trans->transaction;  in add_pending_snapshot()
2167 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_commit_transaction()
2188 	 * We only want one transaction commit doing the flushing so we do not  in btrfs_commit_transaction()
2212 		 * will wait for the transaction to commit if it  in btrfs_commit_transaction()
2285 		 * The previous transaction was aborted and was already removed  in btrfs_commit_transaction()
2318 	 * The thread has started/joined the transaction thus it holds the  in btrfs_commit_transaction()
2338 	 * transaction. Otherwise if this transaction commits before the ordered  in btrfs_commit_transaction()
2348 	 * commit the transaction.  We could have started a join before setting  in btrfs_commit_transaction()
2357 	 * The thread has started/joined the transaction thus it holds the  in btrfs_commit_transaction()
2378 	 * do an async commit but somebody else started before the transaction  in btrfs_commit_transaction()
2428 	WARN_ON(cur_trans != trans->transaction);  in btrfs_commit_transaction()
2498 	 * Before changing the transaction state to TRANS_STATE_UNBLOCKED and  in btrfs_commit_transaction()
2501 	 * start a new transaction and commit a log tree before we commit our  in btrfs_commit_transaction()
2524 				      "Error while writing out transaction");  in btrfs_commit_transaction()
2597 	btrfs_warn(fs_info, "Skipping commit of aborted transaction.");  in btrfs_commit_transaction()
2655  * We only mark the transaction aborted and then set the file system read-only.
2661  * further cleanup. The transaction should complete as it normally would
2674 	WRITE_ONCE(trans->transaction->aborted, error);  in __btrfs_abort_transaction()
2677 	/* Wake up anybody who may be waiting on this transaction */  in __btrfs_abort_transaction()