jbd2: Fix assertion failure in fs/jbd2/checkpoint.c

Before we start committing a transaction, we call
__journal_clean_checkpoint_list() to cleanup transaction's written-back
buffers.

If this call happens to remove all of them (and there were already some
buffers), __journal_remove_checkpoint() will decide to free the transaction
because it isn't (yet) a committing transaction and soon we fail some
assertion - the transaction really isn't ready to be freed :).

We change the check in __journal_remove_checkpoint() to free only a
transaction in T_FINISHED state.  The locking there is subtle though (as
everywhere in JBD ;().  We use j_list_lock to protect the check and a
subsequent call to __journal_drop_transaction() and do the same in the end
of journal_commit_transaction() which is the only place where a transaction
can get to T_FINISHED state.

Probably I'm too paranoid here and such locking is not really necessary -
checkpoint lists are processed only from log_do_checkpoint() where a
transaction must be already committed to be processed or from
__journal_clean_checkpoint_list() where kjournald itself calls it and thus
transaction cannot change state either.  Better be safe if something
changes in future...

Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

3 files changed, 12 insertions, 10 deletions

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index 3fccde7ba008..7e958c86242f 100644
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -602,15 +602,15 @@ int __jbd2_journal_remove_checkpoint(struct journal_head *jh)
 
 	/*
 	 * There is one special case to worry about: if we have just pulled the
-	 * buffer off a committing transaction's forget list, then even if the
-	 * checkpoint list is empty, the transaction obviously cannot be
-	 * dropped!
+	 * buffer off a running or committing transaction's checkpoing list,
+	 * then even if the checkpoint list is empty, the transaction obviously
+	 * cannot be dropped!
 	 *
-	 * The locking here around j_committing_transaction is a bit sleazy.
+	 * The locking here around t_state is a bit sleazy.
 	 * See the comment at the end of jbd2_journal_commit_transaction().
 	 */
-	if (transaction == journal->j_committing_transaction) {
-		JBUFFER_TRACE(jh, "belongs to committing transaction");
+	if (transaction->t_state != T_FINISHED) {
+		JBUFFER_TRACE(jh, "belongs to running/committing transaction");
 		goto out;
 	}
 
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index 6986f334c643..39b5cee3dd8a 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -867,10 +867,10 @@ restart_loop:
 	}
 	spin_unlock(&journal->j_list_lock);
 	/*
-	 * This is a bit sleazy.  We borrow j_list_lock to protect
-	 * journal->j_committing_transaction in __jbd2_journal_remove_checkpoint.
-	 * Really, __jbd2_journal_remove_checkpoint should be using j_state_lock but
-	 * it's a bit hassle to hold that across __jbd2_journal_remove_checkpoint
+	 * This is a bit sleazy.  We use j_list_lock to protect transition
+	 * of a transaction into T_FINISHED state and calling
+	 * __jbd2_journal_drop_transaction(). Otherwise we could race with
+	 * other checkpointing code processing the transaction...
 	 */
 	spin_lock(&journal->j_state_lock);
 	spin_lock(&journal->j_list_lock);
diff --git a/include/linux/jbd2.h b/include/linux/jbd2.h
index d5f7cff4cb28..d861ffd49821 100644
--- a/include/linux/jbd2.h
+++ b/include/linux/jbd2.h
@@ -442,6 +442,8 @@ struct transaction_s
 	/*
 	 * Transaction's current state
 	 * [no locking - only kjournald2 alters this]
+	 * [j_list_lock] guards transition of a transaction into T_FINISHED
+	 * state and subsequent call of __jbd2_journal_drop_transaction()
 	 * FIXME: needs barriers
 	 * KLUDGE: [use j_state_lock]
 	 */