jbd2: Remove a large array of bh's from the stack of the checkpoint routine

jbd2_log_do_checkpoint()n is one of the kernel's largest stack users.
Move the array of buffer head's from the stack of jbd2_log_do_checkpoint()
to the in-core journal structure.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

fs/jbd2/checkpoint.c

@@ -249,16 +249,14 @@ restart:
 	return ret;
 }
 
-#define NR_BATCH	64
-
 static void
-__flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count)
+__flush_batch(journal_t *journal, int *batch_count)
 {
 	int i;
 
-	ll_rw_block(SWRITE, *batch_count, bhs);
+	ll_rw_block(SWRITE, *batch_count, journal->j_chkpt_bhs);
 	for (i = 0; i < *batch_count; i++) {
-		struct buffer_head *bh = bhs[i];
+		struct buffer_head *bh = journal->j_chkpt_bhs[i];
 		clear_buffer_jwrite(bh);
 		BUFFER_TRACE(bh, "brelse");
 		__brelse(bh);
@@ -277,8 +275,7 @@ __flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count)
  * Called under jbd_lock_bh_state(jh2bh(jh)), and drops it
  */
 static int __process_buffer(journal_t *journal, struct journal_head *jh,
-			struct buffer_head **bhs, int *batch_count,
-			transaction_t *transaction)
+			    int *batch_count, transaction_t *transaction)
 {
 	struct buffer_head *bh = jh2bh(jh);
 	int ret = 0;
@@ -325,14 +322,14 @@ static int __process_buffer(journal_t *journal, struct journal_head *jh,
 		get_bh(bh);
 		J_ASSERT_BH(bh, !buffer_jwrite(bh));
 		set_buffer_jwrite(bh);
-		bhs[*batch_count] = bh;
+		journal->j_chkpt_bhs[*batch_count] = bh;
 		__buffer_relink_io(jh);
 		jbd_unlock_bh_state(bh);
 		transaction->t_chp_stats.cs_written++;
 		(*batch_count)++;
-		if (*batch_count == NR_BATCH) {
+		if (*batch_count == JBD2_NR_BATCH) {
 			spin_unlock(&journal->j_list_lock);
-			__flush_batch(journal, bhs, batch_count);
+			__flush_batch(journal, batch_count);
 			ret = 1;
 		}
 	}
@@ -388,7 +385,6 @@ restart:
 	if (journal->j_checkpoint_transactions == transaction &&
 			transaction->t_tid == this_tid) {
 		int batch_count = 0;
-		struct buffer_head *bhs[NR_BATCH];
 		struct journal_head *jh;
 		int retry = 0, err;
 
@@ -402,7 +398,7 @@ restart:
 				retry = 1;
 				break;
 			}
-			retry = __process_buffer(journal, jh, bhs, &batch_count,
+			retry = __process_buffer(journal, jh, &batch_count,
 						 transaction);
 			if (retry < 0 && !result)
 				result = retry;
@@ -419,7 +415,7 @@ restart:
 				spin_unlock(&journal->j_list_lock);
 				retry = 1;
 			}
-			__flush_batch(journal, bhs, &batch_count);
+			__flush_batch(journal, &batch_count);
 		}
 
 		if (retry) {

fs/jbd2/journal.c

@@ -1477,7 +1477,9 @@ int jbd2_journal_destroy(journal_t *journal)
 	spin_lock(&journal->j_list_lock);
 	while (journal->j_checkpoint_transactions != NULL) {
 		spin_unlock(&journal->j_list_lock);
+		mutex_lock(&journal->j_checkpoint_mutex);
 		jbd2_log_do_checkpoint(journal);
+		mutex_unlock(&journal->j_checkpoint_mutex);
 		spin_lock(&journal->j_list_lock);
 	}
 

include/linux/jbd2.h

@@ -687,6 +687,8 @@ jbd2_time_diff(unsigned long start, unsigned long end)
 	return end + (MAX_JIFFY_OFFSET - start);
 }
 
+#define JBD2_NR_BATCH	64
+
 /**
  * struct journal_s - The journal_s type is the concrete type associated with
  *     journal_t.
@@ -830,6 +832,14 @@ struct journal_s
 	/* Semaphore for locking against concurrent checkpoints */
 	struct mutex		j_checkpoint_mutex;
 
+	/*
+	 * List of buffer heads used by the checkpoint routine.  This
+	 * was moved from jbd2_log_do_checkpoint() to reduce stack
+	 * usage.  Access to this array is controlled by the
+	 * j_checkpoint_mutex.  [j_checkpoint_mutex]
+	 */
+	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
+	
 	/*
 	 * Journal head: identifies the first unused block in the journal.
 	 * [j_state_lock]