jbd2: avoid pointless scanning of checkpoint lists

Yuanhan has reported that when he is running fsync(2) heavy workload
creating new files over ramdisk, significant amount of time is spent in
__jbd2_journal_clean_checkpoint_list() trying to clean old transactions
(but they cannot be cleaned up because flusher hasn't yet checkpointed
those buffers). The workload can be generated by:
  fs_mark -d /fs/ram0/1 -D 2 -N 2560 -n 1000000 -L 1 -S 1 -s 4096

Reduce the amount of scanning by stopping to scan the transaction list
once we find a transaction that cannot be checkpointed. Note that this
way of cleaning is still enough to keep freeing space in the journal
after fully checkpointed transactions.

Reported-and-tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

1 files changed, 18 insertions, 14 deletions

diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c
index 1fbf59938cc0..3ab4c5ee12ce 100644
--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -420,7 +420,6 @@ int jbd2_cleanup_journal_tail(journal_t *journal)
  * Find all the written-back checkpoint buffers in the given list and
  * release them.
  *
- * Called with the journal locked.
  * Called with j_list_lock held.
  * Returns number of buffers reaped (for debug)
  */
@@ -440,12 +439,12 @@ static int journal_clean_one_cp_list(struct journal_head *jh, int *released)
 		jh = next_jh;
 		next_jh = jh->b_cpnext;
 		ret = __try_to_free_cp_buf(jh);
-		if (ret) {
-			freed++;
-			if (ret == 2) {
-				*released = 1;
-				return freed;
-			}
+		if (!ret)
+			return freed;
+		freed++;
+		if (ret == 2) {
+			*released = 1;
+			return freed;
 		}
 		/*
 		 * This function only frees up some memory
@@ -465,7 +464,6 @@ static int journal_clean_one_cp_list(struct journal_head *jh, int *released)
  *
  * Find all the written-back checkpoint buffers in the journal and release them.
  *
- * Called with the journal locked.
  * Called with j_list_lock held.
  * Returns number of buffers reaped (for debug)
  */
@@ -473,7 +471,8 @@ static int journal_clean_one_cp_list(struct journal_head *jh, int *released)
 int __jbd2_journal_clean_checkpoint_list(journal_t *journal)
 {
 	transaction_t *transaction, *last_transaction, *next_transaction;
-	int ret = 0;
+	int ret;
+	int freed = 0;
 	int released;
 
 	transaction = journal->j_checkpoint_transactions;
@@ -485,17 +484,21 @@ int __jbd2_journal_clean_checkpoint_list(journal_t *journal)
 	do {
 		transaction = next_transaction;
 		next_transaction = transaction->t_cpnext;
-		ret += journal_clean_one_cp_list(transaction->
+		ret = journal_clean_one_cp_list(transaction->
 				t_checkpoint_list, &released);
 		/*
 		 * This function only frees up some memory if possible so we
 		 * dont have an obligation to finish processing. Bail out if
 		 * preemption requested:
 		 */
-		if (need_resched())
+		if (need_resched()) {
+			freed += ret;
 			goto out;
-		if (released)
+		}
+		if (released) {
+			freed += ret;
 			continue;
+		}
 		/*
 		 * It is essential that we are as careful as in the case of
 		 * t_checkpoint_list with removing the buffer from the list as
@@ -503,11 +506,12 @@ int __jbd2_journal_clean_checkpoint_list(journal_t *journal)
 		 */
 		ret += journal_clean_one_cp_list(transaction->
 				t_checkpoint_io_list, &released);
-		if (need_resched())
+		freed += ret;
+		if (need_resched() || !ret)
 			goto out;
 	} while (transaction != last_transaction);
 out:
-	return ret;
+	return freed;
 }
 
 /*