jbd2: fix a race when checking checkpoint buffer busy

Before removing checkpoint buffer from the t_checkpoint_list, we have to
check both BH_Dirty and BH_Lock bits together to distinguish buffers
have not been or were being written back. But __cp_buffer_busy() checks
them separately, it first check lock state and then check dirty, the
window between these two checks could be raced by writing back
procedure, which locks buffer and clears buffer dirty before I/O
completes. So it cannot guarantee checkpointing buffers been written
back to disk if some error happens later. Finally, it may clean
checkpoint transactions and lead to inconsistent filesystem.

jbd2_journal_forget() and __journal_try_to_free_buffer() also have the
same problem (journal_unmap_buffer() escape from this issue since it's
running under the buffer lock), so fix them through introducing a new
helper to try holding the buffer lock and remove really clean buffer.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=217490
Cc: stable@vger.kernel.org
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230606135928.434610-6-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

2 files changed, 40 insertions, 15 deletions

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index 42b34cab64fb..9ec91017a7f3 100644
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -376,11 +376,15 @@ static unsigned long journal_shrink_one_cp_list(struct journal_head *jh,
 		jh = next_jh;
 		next_jh = jh->b_cpnext;
 
-		if (!destroy && __cp_buffer_busy(jh))
-			continue;
+		if (destroy) {
+			ret = __jbd2_journal_remove_checkpoint(jh);
+		} else {
+			ret = jbd2_journal_try_remove_checkpoint(jh);
+			if (ret < 0)
+				continue;
+		}
 
 		nr_freed++;
-		ret = __jbd2_journal_remove_checkpoint(jh);
 		if (ret) {
 			*released = true;
 			break;
@@ -617,6 +621,34 @@ int __jbd2_journal_remove_checkpoint(struct journal_head *jh)
 }
 
 /*
+ * Check the checkpoint buffer and try to remove it from the checkpoint
+ * list if it's clean. Returns -EBUSY if it is not clean, returns 1 if
+ * it frees the transaction, 0 otherwise.
+ *
+ * This function is called with j_list_lock held.
+ */
+int jbd2_journal_try_remove_checkpoint(struct journal_head *jh)
+{
+	struct buffer_head *bh = jh2bh(jh);
+
+	if (!trylock_buffer(bh))
+		return -EBUSY;
+	if (buffer_dirty(bh)) {
+		unlock_buffer(bh);
+		return -EBUSY;
+	}
+	unlock_buffer(bh);
+
+	/*
+	 * Buffer is clean and the IO has finished (we held the buffer
+	 * lock) so the checkpoint is done. We can safely remove the
+	 * buffer from this transaction.
+	 */
+	JBUFFER_TRACE(jh, "remove from checkpoint list");
+	return __jbd2_journal_remove_checkpoint(jh);
+}
+
+/*
  * journal_insert_checkpoint: put a committed buffer onto a checkpoint
  * list so that we know when it is safe to clean the transaction out of
  * the log.
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index 18611241f451..6ef5022949c4 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -1784,8 +1784,7 @@ int jbd2_journal_forget(handle_t *handle, struct buffer_head *bh)
 		 * Otherwise, if the buffer has been written to disk,
 		 * it is safe to remove the checkpoint and drop it.
 		 */
-		if (!buffer_dirty(bh)) {
-			__jbd2_journal_remove_checkpoint(jh);
+		if (jbd2_journal_try_remove_checkpoint(jh) >= 0) {
 			spin_unlock(&journal->j_list_lock);
 			goto drop;
 		}
@@ -2112,20 +2111,14 @@ __journal_try_to_free_buffer(journal_t *journal, struct buffer_head *bh)
 
 	jh = bh2jh(bh);
 
-	if (buffer_locked(bh) || buffer_dirty(bh))
-		goto out;
-
 	if (jh->b_next_transaction != NULL || jh->b_transaction != NULL)
-		goto out;
+		return;
 
 	spin_lock(&journal->j_list_lock);
-	if (jh->b_cp_transaction != NULL) {
-		/* written-back checkpointed metadata buffer */
-		JBUFFER_TRACE(jh, "remove from checkpoint list");
-		__jbd2_journal_remove_checkpoint(jh);
-	}
+	/* Remove written-back checkpointed metadata buffer */
+	if (jh->b_cp_transaction != NULL)
+		jbd2_journal_try_remove_checkpoint(jh);
 	spin_unlock(&journal->j_list_lock);
-out:
 	return;
 }