Normally, we have to issue a cache flush before we can update journal tail in
journal superblock, effectively wiping out old transactions from the journal.
So use the fact that during transaction commit we issue cache flush anyway and
opportunistically push journal tail as far as we can. Since update of journal
superblock is still costly (we have to use WRITE_FUA), we update log tail only
if we can free significant amount of space.

Signed-off-by: Jan Kara <jack@suse.cz>
---
fs/jbd2/commit.c     |   32 ++++++++++++++++++++++++++++++++
fs/jbd2/journal.c    |   13 +++++++++++++
include/linux/jbd2.h |    1 +
3 files changed, 46 insertions(+), 0 deletions(-)

diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index f37b783..245201c 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c

@@ -331,6 +331,10 @@ void jbd2_journal_commit_transaction(journal_t *journal)

	struct buffer_head *cbh = NULL; /* For transactional checksums */
	__u32 crc32_sum = ~0;
	struct blk_plug plug;
+	/* Tail of the journal */
+	unsigned long first_block;
+	tid_t first_tid;
+	int update_tail;

	/*
	 * First job: lock down the current transaction and wait for

@@ -682,10 +686,30 @@ start_journal_io:

		err = 0;
	}

+	/*
+	 * Get current oldest transaction in the log before we issue flush
+	 * to the filesystem device. After the flush we can be sure that
+	 * blocks of all older transactions are checkpointed to persistent
+	 * storage and we will be safe to update journal start in the
+	 * superblock with the numbers we get here.
+	 */
+	update_tail =
+		jbd2_journal_get_log_tail(journal, &first_tid, &first_block);
+
	write_lock(&journal->j_state_lock);
+	if (update_tail) {
+		long freed = first_block - journal->j_tail;
+
+		if (first_block < journal->j_tail)
+			freed += journal->j_last - journal->j_first;
+		/* Update tail only if we free significant amount of space */
+		if (freed < journal->j_maxlen / 4)
+			update_tail = 0;
+	}