ext4: Wait for proper transaction commit on fsync

(cherry picked from commit b436b9bef84de6893e86346d8fbf7104bc520645)

We cannot rely on buffer dirty bits during fsync because pdflush can come
before fsync is called and clear dirty bits without forcing a transaction
commit. What we do is that we track which transaction has last changed
the inode and which transaction last changed allocation and force it to
disk on fsync.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

7 files changed, 81 insertions, 31 deletions

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 05ce38b981cb..bd2a9dd1464c 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -703,6 +703,13 @@ struct ext4_inode_info {
 	struct list_head i_aio_dio_complete_list;
 	/* current io_end structure for async DIO write*/
 	ext4_io_end_t *cur_aio_dio;
+
+	/*
+	 * Transactions that contain inode's metadata needed to complete
+	 * fsync and fdatasync, respectively.
+	 */
+	tid_t i_sync_tid;
+	tid_t i_datasync_tid;
 };
 
 /*
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index 404ab6c824d3..1892a7763426 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -258,6 +258,19 @@ static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
 	return 0;
 }
 
+static inline void ext4_update_inode_fsync_trans(handle_t *handle,
+						 struct inode *inode,
+						 int datasync)
+{
+	struct ext4_inode_info *ei = EXT4_I(inode);
+
+	if (ext4_handle_valid(handle)) {
+		ei->i_sync_tid = handle->h_transaction->t_tid;
+		if (datasync)
+			ei->i_datasync_tid = handle->h_transaction->t_tid;
+	}
+}
+
 /* super.c */
 int ext4_force_commit(struct super_block *sb);
 
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 499d6bb9be94..43c04ff653c6 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3064,6 +3064,8 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
 	if (flags == EXT4_GET_BLOCKS_DIO_CONVERT_EXT) {
 		ret = ext4_convert_unwritten_extents_dio(handle, inode,
 							path);
+		if (ret >= 0)
+			ext4_update_inode_fsync_trans(handle, inode, 1);
 		goto out2;
 	}
 	/* buffered IO case */
@@ -3091,6 +3093,8 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
 	ret = ext4_ext_convert_to_initialized(handle, inode,
 						path, iblock,
 						max_blocks);
+	if (ret >= 0)
+		ext4_update_inode_fsync_trans(handle, inode, 1);
 out:
 	if (ret <= 0) {
 		err = ret;
@@ -3329,10 +3333,16 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
 	allocated = ext4_ext_get_actual_len(&newex);
 	set_buffer_new(bh_result);
 
-	/* Cache only when it is _not_ an uninitialized extent */
-	if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0)
+	/*
+	 * Cache the extent and update transaction to commit on fdatasync only
+	 * when it is _not_ an uninitialized extent.
+	 */
+	if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
 		ext4_ext_put_in_cache(inode, iblock, allocated, newblock,
 						EXT4_EXT_CACHE_EXTENT);
+		ext4_update_inode_fsync_trans(handle, inode, 1);
+	} else
+		ext4_update_inode_fsync_trans(handle, inode, 0);
 out:
 	if (allocated > max_blocks)
 		allocated = max_blocks;
diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
index a3c25076aef1..d6049e40b161 100644
--- a/fs/ext4/fsync.c
+++ b/fs/ext4/fsync.c
@@ -51,25 +51,30 @@
 int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync)
 {
 	struct inode *inode = dentry->d_inode;
+	struct ext4_inode_info *ei = EXT4_I(inode);
 	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
-	int err, ret = 0;
+	int ret;
+	tid_t commit_tid;
 
 	J_ASSERT(ext4_journal_current_handle() == NULL);
 
 	trace_ext4_sync_file(file, dentry, datasync);
 
+	if (inode->i_sb->s_flags & MS_RDONLY)
+		return 0;
+
 	ret = flush_aio_dio_completed_IO(inode);
 	if (ret < 0)
 		return ret;
+
+	if (!journal)
+		return simple_fsync(file, dentry, datasync);
+
 	/*
-	 * data=writeback:
+	 * data=writeback,ordered:
 	 *  The caller's filemap_fdatawrite()/wait will sync the data.
-	 *  sync_inode() will sync the metadata
-	 *
-	 * data=ordered:
-	 *  The caller's filemap_fdatawrite() will write the data and
-	 *  sync_inode() will write the inode if it is dirty.  Then the caller's
-	 *  filemap_fdatawait() will wait on the pages.
+	 *  Metadata is in the journal, we wait for proper transaction to
+	 *  commit here.
 	 *
 	 * data=journal:
 	 *  filemap_fdatawrite won't do anything (the buffers are clean).
@@ -82,27 +87,10 @@ int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync)
 	if (ext4_should_journal_data(inode))
 		return ext4_force_commit(inode->i_sb);
 
-	if (!journal)
-		ret = sync_mapping_buffers(inode->i_mapping);
-
-	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
-		goto out;
-
-	/*
-	 * The VFS has written the file data.  If the inode is unaltered
-	 * then we need not start a commit.
-	 */
-	if (inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC)) {
-		struct writeback_control wbc = {
-			.sync_mode = WB_SYNC_ALL,
-			.nr_to_write = 0, /* sys_fsync did this */
-		};
-		err = sync_inode(inode, &wbc);
-		if (ret == 0)
-			ret = err;
-	}
-out:
-	if (journal && (journal->j_flags & JBD2_BARRIER))
+	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
+	if (jbd2_log_start_commit(journal, commit_tid))
+		jbd2_log_wait_commit(journal, commit_tid);
+	else if (journal->j_flags & JBD2_BARRIER)
 		blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
 	return ret;
 }
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 64086fb96e47..1dae9a4dbb03 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1025,6 +1025,8 @@ static int ext4_ind_get_blocks(handle_t *handle, struct inode *inode,
 		goto cleanup;
 
 	set_buffer_new(bh_result);
+
+	ext4_update_inode_fsync_trans(handle, inode, 1);
 got_it:
 	map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
 	if (count > blocks_to_boundary)
@@ -4794,6 +4796,7 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
 	struct ext4_inode *raw_inode;
 	struct ext4_inode_info *ei;
 	struct inode *inode;
+	journal_t *journal = EXT4_SB(sb)->s_journal;
 	long ret;
 	int block;
 
@@ -4858,6 +4861,31 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
 		ei->i_data[block] = raw_inode->i_block[block];
 	INIT_LIST_HEAD(&ei->i_orphan);
 
+	/*
+	 * Set transaction id's of transactions that have to be committed
+	 * to finish f[data]sync. We set them to currently running transaction
+	 * as we cannot be sure that the inode or some of its metadata isn't
+	 * part of the transaction - the inode could have been reclaimed and
+	 * now it is reread from disk.
+	 */
+	if (journal) {
+		transaction_t *transaction;
+		tid_t tid;
+
+		spin_lock(&journal->j_state_lock);
+		if (journal->j_running_transaction)
+			transaction = journal->j_running_transaction;
+		else
+			transaction = journal->j_committing_transaction;
+		if (transaction)
+			tid = transaction->t_tid;
+		else
+			tid = journal->j_commit_sequence;
+		spin_unlock(&journal->j_state_lock);
+		ei->i_sync_tid = tid;
+		ei->i_datasync_tid = tid;
+	}
+
 	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
 		ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
 		if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
@@ -5112,6 +5140,7 @@ static int ext4_do_update_inode(handle_t *handle,
 		err = rc;
 	ei->i_state &= ~EXT4_STATE_NEW;
 
+	ext4_update_inode_fsync_trans(handle, inode, 0);
 out_brelse:
 	brelse(bh);
 	ext4_std_error(inode->i_sb, err);
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 803fb252c111..9ae521749b90 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -706,6 +706,8 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
 	spin_lock_init(&(ei->i_block_reservation_lock));
 	INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
 	ei->cur_aio_dio = NULL;
+	ei->i_sync_tid = 0;
+	ei->i_datasync_tid = 0;
 
 	return &ei->vfs_inode;
 }
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index c974077915fa..82c295d2a7c5 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -78,6 +78,7 @@ EXPORT_SYMBOL(jbd2_journal_errno);
 EXPORT_SYMBOL(jbd2_journal_ack_err);
 EXPORT_SYMBOL(jbd2_journal_clear_err);
 EXPORT_SYMBOL(jbd2_log_wait_commit);
+EXPORT_SYMBOL(jbd2_log_start_commit);
 EXPORT_SYMBOL(jbd2_journal_start_commit);
 EXPORT_SYMBOL(jbd2_journal_force_commit_nested);
 EXPORT_SYMBOL(jbd2_journal_wipe);