jbd2: protect all log tail updates with j_checkpoint_mutex

There are some log tail updates that are not protected by j_checkpoint_mutex.
Some of these are harmless because they happen during startup or shutdown but
updates in jbd2_journal_commit_transaction() and jbd2_journal_flush() can
really race with other log tail updates (e.g. someone doing
jbd2_journal_flush() with someone running jbd2_cleanup_journal_tail()). So
protect all log tail updates with j_checkpoint_mutex.

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

2 files changed, 18 insertions, 3 deletions

diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index 19371a8..6705717 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -340,7 +340,9 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 	/* Do we need to erase the effects of a prior jbd2_journal_flush? */
 	if (journal->j_flags & JBD2_FLUSHED) {
 		jbd_debug(3, "super block updated\n");
+		mutex_lock(&journal->j_checkpoint_mutex);
 		jbd2_journal_update_sb_log_tail(journal);
+		mutex_unlock(&journal->j_checkpoint_mutex);
 	} else {
 		jbd_debug(3, "superblock not updated\n");
 	}
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index 6e75fbd..fc5f2ac 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -1123,8 +1123,11 @@ static int journal_reset(journal_t *journal)
 			journal->j_errno);
 		journal->j_flags |= JBD2_FLUSHED;
 	} else {
+		/* Lock here to make assertions happy... */
+		mutex_lock(&journal->j_checkpoint_mutex);
 		/* Add the dynamic fields and write it to disk. */
 		jbd2_journal_update_sb_log_tail(journal);
+		mutex_unlock(&journal->j_checkpoint_mutex);
 	}
 	return jbd2_journal_start_thread(journal);
 }
@@ -1173,6 +1176,7 @@ void jbd2_journal_update_sb_log_tail(journal_t *journal)
 {
 	journal_superblock_t *sb = journal->j_superblock;
 
+	BUG_ON(!mutex_is_locked(&journal->j_checkpoint_mutex));
 	read_lock(&journal->j_state_lock);
 	jbd_debug(1, "JBD2: updating superblock (start %ld, seq %d)\n",
 		  journal->j_tail, journal->j_tail_sequence);
@@ -1201,6 +1205,7 @@ static void jbd2_mark_journal_empty(journal_t *journal)
 {
 	journal_superblock_t *sb = journal->j_superblock;
 
+	BUG_ON(!mutex_is_locked(&journal->j_checkpoint_mutex));
 	read_lock(&journal->j_state_lock);
 	jbd_debug(1, "JBD2: Marking journal as empty (seq %d)\n",
 		  journal->j_tail_sequence);
@@ -1434,9 +1439,11 @@ int jbd2_journal_destroy(journal_t *journal)
 	spin_unlock(&journal->j_list_lock);
 
 	if (journal->j_sb_buffer) {
-		if (!is_journal_aborted(journal))
+		if (!is_journal_aborted(journal)) {
+			mutex_lock(&journal->j_checkpoint_mutex);
 			jbd2_mark_journal_empty(journal);
-		else
+			mutex_unlock(&journal->j_checkpoint_mutex);
+		} else
 			err = -EIO;
 		brelse(journal->j_sb_buffer);
 	}
@@ -1630,6 +1637,7 @@ int jbd2_journal_flush(journal_t *journal)
 	if (is_journal_aborted(journal))
 		return -EIO;
 
+	mutex_lock(&journal->j_checkpoint_mutex);
 	jbd2_cleanup_journal_tail(journal);
 
 	/* Finally, mark the journal as really needing no recovery.
@@ -1638,6 +1646,7 @@ int jbd2_journal_flush(journal_t *journal)
 	 * commits of data to the journal will restore the current
 	 * s_start value. */
 	jbd2_mark_journal_empty(journal);
+	mutex_unlock(&journal->j_checkpoint_mutex);
 	write_lock(&journal->j_state_lock);
 	J_ASSERT(!journal->j_running_transaction);
 	J_ASSERT(!journal->j_committing_transaction);
@@ -1678,8 +1687,12 @@ int jbd2_journal_wipe(journal_t *journal, int write)
 		write ? "Clearing" : "Ignoring");
 
 	err = jbd2_journal_skip_recovery(journal);
-	if (write)
+	if (write) {
+		/* Lock to make assertions happy... */
+		mutex_lock(&journal->j_checkpoint_mutex);
 		jbd2_mark_journal_empty(journal);
+		mutex_unlock(&journal->j_checkpoint_mutex);
+	}
 
  no_recovery:
 	return err;