Skip to contentSkip to conversation

virtio-blk: set req->state to MQ_RQ_COMPLETE and support completion batching for the IRQ path

PATCH New Suwan Kim 1 participant 3 messages Dec 21, 2022

Subsystems (3)
lore
Share & tools
lore

Suwan Kim

author
#0 lore raw

v3 cover virtio-blk: set req->state to MQ_RQ_COMPLETE and support completion batching for the IRQ path

V3 changes
- patch#2
- Adjust tab indents for line continuation (Christoph's comment)

V2 changes
- patch#1
- use blk_mq_complete_request_remote() instead of blk_mq_set_request_complete()

- patch#2
- Modify patch based on the patch#1 v2 changes.
- Use iob.complete(&iob) instead of virtblk_complete_batch(&iob) at virtblk_done() (Stefan's comment)
- It shows almost same performance improvement as v1 patch.

Suwan Kim (2):
virtio-blk: set req->state to MQ_RQ_COMPLETE after polling I/O is
finished
virtio-blk: support completion batching for the IRQ path

 drivers/block/virtio_blk.c | 81 +++++++++++++++++++++-----------------
 1 file changed, 45 insertions(+), 36 deletions(-)
Signature · 1 line
2.26.3
author
#1 lore raw

PATCH v3 1/2 virtio-blk: set req->state to MQ_RQ_COMPLETE after polling I/O is finished

Driver should set req->state to MQ_RQ_COMPLETE after it finishes to process req. But virtio-blk doesn't set MQ_RQ_COMPLETE after virtblk_poll() handles req and req->state still remains MQ_RQ_IN_FLIGHT. Fortunately so far there is no issue about it because blk_mq_end_request_batch() sets req->state to MQ_RQ_IDLE.

In this patch, virblk_poll() calls blk_mq_complete_request_remote() to set req->state to MQ_RQ_COMPLETE before it adds req to a batch completion list. So it properly sets req->state after polling I/O is finished.

Fixes 4e0400525691 ("virtio-blk: support polling I/O")
---
 drivers/block/virtio_blk.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/block/virtio_blk.c +3 -2
index 19da5defd734..75ee51aba964 100644
@@ -858,9 +858,10 @@ static int virtblk_poll(struct blk_mq_hw_ctx *hctx, struct io_comp_batch *iob)
 		struct request *req = blk_mq_rq_from_pdu(vbr);
 
 		found++;
-		if (!blk_mq_add_to_batch(req, iob, vbr->status,
+		if (!blk_mq_complete_request_remote(req) &&
+		    !blk_mq_add_to_batch(req, iob, vbr->status,
 						virtblk_complete_batch))
-			blk_mq_complete_request(req);
+			virtblk_request_done(req);
 	}
 
 	if (found)
Signature · 1 line
2.26.3
author
#2 lore raw

PATCH v3 2/2 virtio-blk: support completion batching for the IRQ path

In reply to Suwan Kim

This patch adds completion batching to the IRQ path. It reuses batch completion code of virtblk_poll(). It collects requests to io_comp_batch and processes them all at once. It can boost up the performance by 2%.

To validate the performance improvement and stabilty, I did fio test with 4 vCPU VM and 12 vCPU VM respectively. Both VMs have 8GB ram and the same number of HW queues as vCPU.
The fio cammad is as follows and I ran the fio 5 times and got IOPS average. (io_uring, randread, direct=1, bs=512, iodepth=64 numjobs=2,4)

Test result shows about 2% improvement.

4 vcpu VM | numjobs=2 | numjobs=4
-----------------------------------------------------------
fio without patch | 367.2K IOPS | 397.6K IOPS
-----------------------------------------------------------
fio with patch | 372.8K IOPS | 407.7K IOPS

12 vcpu VM | numjobs=2 | numjobs=4
-----------------------------------------------------------
fio without patch | 363.6K IOPS | 374.8K IOPS
-----------------------------------------------------------
fio with patch | 373.8K IOPS | 385.3K IOPS

---
 drivers/block/virtio_blk.c | 82 +++++++++++++++++++++-----------------
 1 file changed, 45 insertions(+), 37 deletions(-)
drivers/block/virtio_blk.c +45 -37
index 75ee51aba964..91f1e008d6ab 100644
@@ -272,33 +272,63 @@ static inline void virtblk_request_done(struct request *req)
 	blk_mq_end_request(req, virtblk_result(vbr));
 }
 
+static void virtblk_complete_batch(struct io_comp_batch *iob)
+{
+	struct request *req;
+
+	rq_list_for_each(&iob->req_list, req) {
+		virtblk_unmap_data(req, blk_mq_rq_to_pdu(req));
+		virtblk_cleanup_cmd(req);
+	}
+	blk_mq_end_request_batch(iob);
+}
+
+static int virtblk_handle_req(struct virtio_blk_vq *vq,
+			      struct io_comp_batch *iob)
+{
+	struct virtblk_req *vbr;
+	int req_done = 0;
+	unsigned int len;
+
+	while ((vbr = virtqueue_get_buf(vq->vq, &len)) != NULL) {
+		struct request *req = blk_mq_rq_from_pdu(vbr);
+
+		if (likely(!blk_should_fake_timeout(req->q)) &&
+		    !blk_mq_complete_request_remote(req) &&
+		    !blk_mq_add_to_batch(req, iob, vbr->status,
+					 virtblk_complete_batch))
+			virtblk_request_done(req);
+		req_done++;
+	}
+
+	return req_done;
+}
+
 static void virtblk_done(struct virtqueue *vq)
 {
 	struct virtio_blk *vblk = vq->vdev->priv;
-	bool req_done = false;
-	int qid = vq->index;
-	struct virtblk_req *vbr;
+	struct virtio_blk_vq *vblk_vq = &vblk->vqs[vq->index];
+	int req_done = 0;
 	unsigned long flags;
-	unsigned int len;
+	DEFINE_IO_COMP_BATCH(iob);
 
-	spin_lock_irqsave(&vblk->vqs[qid].lock, flags);
+	spin_lock_irqsave(&vblk_vq->lock, flags);
 	do {
 		virtqueue_disable_cb(vq);
-		while ((vbr = virtqueue_get_buf(vblk->vqs[qid].vq, &len)) != NULL) {
-			struct request *req = blk_mq_rq_from_pdu(vbr);
+		req_done += virtblk_handle_req(vblk_vq, &iob);
 
-			if (likely(!blk_should_fake_timeout(req->q)))
-				blk_mq_complete_request(req);
-			req_done = true;
-		}
 		if (unlikely(virtqueue_is_broken(vq)))
 			break;
 	} while (!virtqueue_enable_cb(vq));
 
-	/* In case queue is stopped waiting for more buffers. */
-	if (req_done)
+	if (req_done) {
+		if (!rq_list_empty(iob.req_list))
+			iob.complete(&iob);
+
+		/* In case queue is stopped waiting for more buffers. */
 		blk_mq_start_stopped_hw_queues(vblk->disk->queue, true);
-	spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
+	}
+	spin_unlock_irqrestore(&vblk_vq->lock, flags);
 }
 
 static void virtio_commit_rqs(struct blk_mq_hw_ctx *hctx)
@@ -832,37 +862,15 @@ static void virtblk_map_queues(struct blk_mq_tag_set *set)
 	}
 }
 
-static void virtblk_complete_batch(struct io_comp_batch *iob)
-{
-	struct request *req;
-
-	rq_list_for_each(&iob->req_list, req) {
-		virtblk_unmap_data(req, blk_mq_rq_to_pdu(req));
-		virtblk_cleanup_cmd(req);
-	}
-	blk_mq_end_request_batch(iob);
-}
-
 static int virtblk_poll(struct blk_mq_hw_ctx *hctx, struct io_comp_batch *iob)
 {
 	struct virtio_blk *vblk = hctx->queue->queuedata;
 	struct virtio_blk_vq *vq = get_virtio_blk_vq(hctx);
-	struct virtblk_req *vbr;
 	unsigned long flags;
-	unsigned int len;
 	int found = 0;
 
 	spin_lock_irqsave(&vq->lock, flags);
-
-	while ((vbr = virtqueue_get_buf(vq->vq, &len)) != NULL) {
-		struct request *req = blk_mq_rq_from_pdu(vbr);
-
-		found++;
-		if (!blk_mq_complete_request_remote(req) &&
-		    !blk_mq_add_to_batch(req, iob, vbr->status,
-						virtblk_complete_batch))
-			virtblk_request_done(req);
-	}
+	found = virtblk_handle_req(vq, iob);
 
 	if (found)
 		blk_mq_start_stopped_hw_queues(vblk->disk->queue, true);
Signature · 1 line
2.26.3