f2fs: wait discard IO completion without cmd_lock held

Wait discard IO completion outside cmd_lock to avoid long latency
of holding cmd_lock in IO busy scenario.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index e304b89..f55a1da 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -997,17 +997,34 @@
 	struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
 	struct list_head *wait_list = &(dcc->wait_list);
 	struct discard_cmd *dc, *tmp;
+	bool need_wait;
+
+next:
+	need_wait = false;
 
 	mutex_lock(&dcc->cmd_lock);
 	list_for_each_entry_safe(dc, tmp, wait_list, list) {
-		if (!wait_cond || dc->state == D_DONE) {
-			if (dc->ref)
-				continue;
+		if (!wait_cond || (dc->state == D_DONE && !dc->ref)) {
 			wait_for_completion_io(&dc->wait);
 			__remove_discard_cmd(sbi, dc);
+		} else {
+			dc->ref++;
+			need_wait = true;
+			break;
 		}
 	}
 	mutex_unlock(&dcc->cmd_lock);
+
+	if (need_wait) {
+		wait_for_completion_io(&dc->wait);
+		mutex_lock(&dcc->cmd_lock);
+		f2fs_bug_on(sbi, dc->state != D_DONE);
+		dc->ref--;
+		if (!dc->ref)
+			__remove_discard_cmd(sbi, dc);
+		mutex_unlock(&dcc->cmd_lock);
+		goto next;
+	}
 }
 
 /* This should be covered by global mutex, &sit_i->sentry_lock */