jbd2: Implement data=ordered mode handling via inodes

This patch adds necessary framework into JBD2 to be able to track inodes
with each transaction and write-out their dirty data during transaction
commit time.

This new ordered mode brings all sorts of advantages such as possibility 
to get rid of journal heads and buffer heads for data buffers in ordered 
mode, better ordering of writes on transaction commit, simplification of
 some JBD code, no more anonymous pages when truncate of data being 
committed happens.  Also with this new ordered mode, delayed allocation 
on ordered mode is much simpler.
Signed-off-by: NJan Kara <jack@suse.cz>

fs/jbd2/commit.c

@@ -355,6 +355,81 @@ static void journal_submit_data_buffers(journal_t *journal,
 	journal_do_submit_data(wbuf, bufs);
 }
 
+/*
+ * Submit all the data buffers of inode associated with the transaction to
+ * disk.
+ *
+ * We are in a committing transaction. Therefore no new inode can be added to
+ * our inode list. We use JI_COMMIT_RUNNING flag to protect inode we currently
+ * operate on from being released while we write out pages.
+ */
+static int journal_submit_inode_data_buffers(journal_t *journal,
+		transaction_t *commit_transaction)
+{
+	struct jbd2_inode *jinode;
+	int err, ret = 0;
+	struct address_space *mapping;
+
+	spin_lock(&journal->j_list_lock);
+	list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
+		mapping = jinode->i_vfs_inode->i_mapping;
+		jinode->i_flags |= JI_COMMIT_RUNNING;
+		spin_unlock(&journal->j_list_lock);
+		err = filemap_fdatawrite_range(mapping, 0,
+					i_size_read(jinode->i_vfs_inode));
+		if (!ret)
+			ret = err;
+		spin_lock(&journal->j_list_lock);
+		J_ASSERT(jinode->i_transaction == commit_transaction);
+		jinode->i_flags &= ~JI_COMMIT_RUNNING;
+		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
+	}
+	spin_unlock(&journal->j_list_lock);
+	return ret;
+}
+
+/*
+ * Wait for data submitted for writeout, refile inodes to proper
+ * transaction if needed.
+ *
+ */
+static int journal_finish_inode_data_buffers(journal_t *journal,
+		transaction_t *commit_transaction)
+{
+	struct jbd2_inode *jinode, *next_i;
+	int err, ret = 0;
+
+	/* For locking, see the comment in journal_submit_inode_data_buffers() */
+	spin_lock(&journal->j_list_lock);
+	list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
+		jinode->i_flags |= JI_COMMIT_RUNNING;
+		spin_unlock(&journal->j_list_lock);
+		err = filemap_fdatawait(jinode->i_vfs_inode->i_mapping);
+		if (!ret)
+			ret = err;
+		spin_lock(&journal->j_list_lock);
+		jinode->i_flags &= ~JI_COMMIT_RUNNING;
+		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
+	}
+
+	/* Now refile inode to proper lists */
+	list_for_each_entry_safe(jinode, next_i,
+				 &commit_transaction->t_inode_list, i_list) {
+		list_del(&jinode->i_list);
+		if (jinode->i_next_transaction) {
+			jinode->i_transaction = jinode->i_next_transaction;
+			jinode->i_next_transaction = NULL;
+			list_add(&jinode->i_list,
+				&jinode->i_transaction->t_inode_list);
+		} else {
+			jinode->i_transaction = NULL;
+		}
+	}
+	spin_unlock(&journal->j_list_lock);
+
+	return ret;
+}
+
 static __u32 jbd2_checksum_data(__u32 crc32_sum, struct buffer_head *bh)
 {
 	struct page *page = bh->b_page;
@@ -529,6 +604,9 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 	 */
 	err = 0;
 	journal_submit_data_buffers(journal, commit_transaction);
+	err = journal_submit_inode_data_buffers(journal, commit_transaction);
+	if (err)
+		jbd2_journal_abort(journal, err);
 
 	/*
 	 * Wait for all previously submitted IO to complete if commit
@@ -760,6 +838,17 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 			__jbd2_journal_abort_hard(journal);
 	}
 
+	/*
+	 * This is the right place to wait for data buffers both for ASYNC
+	 * and !ASYNC commit. If commit is ASYNC, we need to wait only after
+	 * the commit block went to disk (which happens above). If commit is
+	 * SYNC, we need to wait for data buffers before we start writing
+	 * commit block, which happens below in such setting.
+	 */
+	err = journal_finish_inode_data_buffers(journal, commit_transaction);
+	if (err)
+		jbd2_journal_abort(journal, err);
+
 	/* Lo and behold: we have just managed to send a transaction to
            the log.  Before we can commit it, wait for the IO so far to
            complete.  Control buffers being written are on the
@@ -880,6 +969,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 	jbd_debug(3, "JBD: commit phase 7\n");
 
 	J_ASSERT(commit_transaction->t_sync_datalist == NULL);
+	J_ASSERT(list_empty(&commit_transaction->t_inode_list));
 	J_ASSERT(commit_transaction->t_buffers == NULL);
 	J_ASSERT(commit_transaction->t_checkpoint_list == NULL);
 	J_ASSERT(commit_transaction->t_iobuf_list == NULL);

fs/jbd2/journal.c

@@ -82,6 +82,10 @@ EXPORT_SYMBOL(jbd2_journal_blocks_per_page);
 EXPORT_SYMBOL(jbd2_journal_invalidatepage);
 EXPORT_SYMBOL(jbd2_journal_try_to_free_buffers);
 EXPORT_SYMBOL(jbd2_journal_force_commit);
+EXPORT_SYMBOL(jbd2_journal_file_inode);
+EXPORT_SYMBOL(jbd2_journal_init_jbd_inode);
+EXPORT_SYMBOL(jbd2_journal_release_jbd_inode);
+EXPORT_SYMBOL(jbd2_journal_begin_ordered_truncate);
 
 static int journal_convert_superblock_v1(journal_t *, journal_superblock_t *);
 static void __journal_abort_soft (journal_t *journal, int errno);
@@ -2194,6 +2198,54 @@ void jbd2_journal_put_journal_head(struct journal_head *jh)
 	jbd_unlock_bh_journal_head(bh);
 }
 
+/*
+ * Initialize jbd inode head
+ */
+void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode)
+{
+	jinode->i_transaction = NULL;
+	jinode->i_next_transaction = NULL;
+	jinode->i_vfs_inode = inode;
+	jinode->i_flags = 0;
+	INIT_LIST_HEAD(&jinode->i_list);
+}
+
+/*
+ * Function to be called before we start removing inode from memory (i.e.,
+ * clear_inode() is a fine place to be called from). It removes inode from
+ * transaction's lists.
+ */
+void jbd2_journal_release_jbd_inode(journal_t *journal,
+				    struct jbd2_inode *jinode)
+{
+	int writeout = 0;
+
+	if (!journal)
+		return;
+restart:
+	spin_lock(&journal->j_list_lock);
+	/* Is commit writing out inode - we have to wait */
+	if (jinode->i_flags & JI_COMMIT_RUNNING) {
+		wait_queue_head_t *wq;
+		DEFINE_WAIT_BIT(wait, &jinode->i_flags, __JI_COMMIT_RUNNING);
+		wq = bit_waitqueue(&jinode->i_flags, __JI_COMMIT_RUNNING);
+		prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
+		spin_unlock(&journal->j_list_lock);
+		schedule();
+		finish_wait(wq, &wait.wait);
+		goto restart;
+	}
+
+	/* Do we need to wait for data writeback? */
+	if (journal->j_committing_transaction == jinode->i_transaction)
+		writeout = 1;
+	if (jinode->i_transaction) {
+		list_del(&jinode->i_list);
+		jinode->i_transaction = NULL;
+	}
+	spin_unlock(&journal->j_list_lock);
+}
+
 /*
  * debugfs tunables
  */

fs/jbd2/transaction.c

@@ -51,6 +51,7 @@ jbd2_get_transaction(journal_t *journal, transaction_t *transaction)
 	transaction->t_tid = journal->j_transaction_sequence++;
 	transaction->t_expires = jiffies + journal->j_commit_interval;
 	spin_lock_init(&transaction->t_handle_lock);
+	INIT_LIST_HEAD(&transaction->t_inode_list);
 
 	/* Set up the commit timer for the new transaction. */
 	journal->j_commit_timer.expires = round_jiffies(transaction->t_expires);
@@ -2195,3 +2196,88 @@ void jbd2_journal_refile_buffer(journal_t *journal, struct journal_head *jh)
 	spin_unlock(&journal->j_list_lock);
 	__brelse(bh);
 }
+
+/*
+ * File inode in the inode list of the handle's transaction
+ */
+int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *jinode)
+{
+	transaction_t *transaction = handle->h_transaction;
+	journal_t *journal = transaction->t_journal;
+
+	if (is_handle_aborted(handle))
+		return -EIO;
+
+	jbd_debug(4, "Adding inode %lu, tid:%d\n", jinode->i_vfs_inode->i_ino,
+			transaction->t_tid);
+
+	/*
+	 * First check whether inode isn't already on the transaction's
+	 * lists without taking the lock. Note that this check is safe
+	 * without the lock as we cannot race with somebody removing inode
+	 * from the transaction. The reason is that we remove inode from the
+	 * transaction only in journal_release_jbd_inode() and when we commit
+	 * the transaction. We are guarded from the first case by holding
+	 * a reference to the inode. We are safe against the second case
+	 * because if jinode->i_transaction == transaction, commit code
+	 * cannot touch the transaction because we hold reference to it,
+	 * and if jinode->i_next_transaction == transaction, commit code
+	 * will only file the inode where we want it.
+	 */
+	if (jinode->i_transaction == transaction ||
+	    jinode->i_next_transaction == transaction)
+		return 0;
+
+	spin_lock(&journal->j_list_lock);
+
+	if (jinode->i_transaction == transaction ||
+	    jinode->i_next_transaction == transaction)
+		goto done;
+
+	/* On some different transaction's list - should be
+	 * the committing one */
+	if (jinode->i_transaction) {
+		J_ASSERT(jinode->i_next_transaction == NULL);
+		J_ASSERT(jinode->i_transaction ==
+					journal->j_committing_transaction);
+		jinode->i_next_transaction = transaction;
+		goto done;
+	}
+	/* Not on any transaction list... */
+	J_ASSERT(!jinode->i_next_transaction);
+	jinode->i_transaction = transaction;
+	list_add(&jinode->i_list, &transaction->t_inode_list);
+done:
+	spin_unlock(&journal->j_list_lock);
+
+	return 0;
+}
+
+/*
+ * This function must be called when inode is journaled in ordered mode
+ * before truncation happens. It starts writeout of truncated part in
+ * case it is in the committing transaction so that we stand to ordered
+ * mode consistency guarantees.
+ */
+int jbd2_journal_begin_ordered_truncate(struct jbd2_inode *inode,
+					loff_t new_size)
+{
+	journal_t *journal;
+	transaction_t *commit_trans;
+	int ret = 0;
+
+	if (!inode->i_transaction && !inode->i_next_transaction)
+		goto out;
+	journal = inode->i_transaction->t_journal;
+	spin_lock(&journal->j_state_lock);
+	commit_trans = journal->j_committing_transaction;
+	spin_unlock(&journal->j_state_lock);
+	if (inode->i_transaction == commit_trans) {
+		ret = filemap_fdatawrite_range(inode->i_vfs_inode->i_mapping,
+			new_size, LLONG_MAX);
+		if (ret)
+			jbd2_journal_abort(journal, ret);
+	}
+out:
+	return ret;
+}

include/linux/jbd2.h

@@ -381,6 +381,38 @@ static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
 }
 
+/* Flags in jbd_inode->i_flags */
+#define __JI_COMMIT_RUNNING 0
+/* Commit of the inode data in progress. We use this flag to protect us from
+ * concurrent deletion of inode. We cannot use reference to inode for this
+ * since we cannot afford doing last iput() on behalf of kjournald
+ */
+#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
+
+/**
+ * struct jbd_inode is the structure linking inodes in ordered mode
+ *   present in a transaction so that we can sync them during commit.
+ */
+struct jbd2_inode {
+	/* Which transaction does this inode belong to? Either the running
+	 * transaction or the committing one. [j_list_lock] */
+	transaction_t *i_transaction;
+
+	/* Pointer to the running transaction modifying inode's data in case
+	 * there is already a committing transaction touching it. [j_list_lock] */
+	transaction_t *i_next_transaction;
+
+	/* List of inodes in the i_transaction [j_list_lock] */
+	struct list_head i_list;
+
+	/* VFS inode this inode belongs to [constant during the lifetime
+	 * of the structure] */
+	struct inode *i_vfs_inode;
+
+	/* Flags of inode [j_list_lock] */
+	unsigned int i_flags;
+};
+
 struct jbd2_revoke_table_s;
 
 /**
@@ -566,6 +598,12 @@ struct transaction_s
 	 */
 	struct journal_head	*t_log_list;
 
+	/*
+	 * List of inodes whose data we've modified in data=ordered mode.
+	 * [j_list_lock]
+	 */
+	struct list_head	t_inode_list;
+
 	/*
 	 * Protects info related to handles
 	 */
@@ -1046,6 +1084,10 @@ extern void	   jbd2_journal_ack_err    (journal_t *);
 extern int	   jbd2_journal_clear_err  (journal_t *);
 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
 extern int	   jbd2_journal_force_commit(journal_t *);
+extern int	   jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
+extern int	   jbd2_journal_begin_ordered_truncate(struct jbd2_inode *inode, loff_t new_size);
+extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
+extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
 
 /*
  * journal_head management