On 3/28/24 7:44 PM, Zhiguo Niu wrote:
diff --git a/block/mq-deadline.c b/block/mq-deadline.c
index 02a916b..89c516e 100644
--- a/block/mq-deadline.c
+++ b/block/mq-deadline.c
@@ -646,10 +646,12 @@ static void dd_depth_updated(struct blk_mq_hw_ctx *hctx)
struct request_queue *q = hctx->queue;
struct deadline_data *dd = q->elevator->elevator_data;
struct blk_mq_tags *tags = hctx->sched_tags;
+ unsigned int shift = tags->bitmap_tags.sb.shift;
+ unsigned int dd_min_depth = max(1U, 3 * (1U << shift) / 4);
dd->async_depth = max(1UL, 3 * q->nr_requests / 4);
- sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, dd->async_depth);
+ sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, dd_min_depth);
}
The above patch sets min_shallow_depth to the same value as commit
d47f9717e5cf ("block/mq-deadline: use correct way to throttling write
requests"). That commit got reverted because it was causing performance
problems. So the above patch reintroduces the performance problem that
has been fixed by commit 256aab46e316 ("Revert "block/mq-deadline: use
correct way to throttling write requests"").
Thank you for attempting to reintroduce a problem that just got fixed
without even mentioning that this is an attempt to reintroduce a
performance problem.
Bart.
