Patchwork [1/2] jbd2: don't wake kjournald unnecessarily

login
register
mail settings
Submitter Theodore Ts'o
Date Jan. 30, 2013, 5:39 a.m.
Message ID <1359524369-31234-1-git-send-email-tytso@mit.edu>
Download mbox | patch
Permalink /patch/216768/
State Superseded
Headers show

Comments

Theodore Ts'o - Jan. 30, 2013, 5:39 a.m.
Don't send an extra wakeup to kjournald in the case where we
already have the proper target in j_commit_request, i.e. that
transaction has already been requested for commit.

commit deeeaf13 "jbd2: fix fsync() tid wraparound bug" changed
the logic leading to a wakeup, but it caused some extra wakeups
which were found to lead to a measurable performance regression.

Reported-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: mszeredi@suse.cz
---

Eric,

I think this might be clearer than the patch which you proposed; what do
you think?

 fs/jbd2/journal.c | 4 ++++
 1 file changed, 4 insertions(+)

Patch

diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index dbf41f9..1a80e31 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -513,6 +513,10 @@  int __jbd2_log_space_left(journal_t *journal)
  */
 int __jbd2_log_start_commit(journal_t *journal, tid_t target)
 {
+	/* Return if the txn has already requested to be committed */
+	if (journal->j_commit_request == target)
+		return 0;
+
 	/*
 	 * The only transaction we can possibly wait upon is the
 	 * currently running transaction (if it exists).  Otherwise,