On Thu, Mar 24, 2022 at 11:04:49PM +0900, Suwan Kim
wrote:> This patch supports polling I/O via virtio-blk driver. Polling
> feature is enabled by module parameter "num_poll_queues" and it
> sets dedicated polling queues for virtio-blk. This patch improves
> the polling I/O throughput and latency.
>
> The virtio-blk driver doesn't not have a poll function and a poll
> queue and it has been operating in interrupt driven method even if
> the polling function is called in the upper layer.
>
> virtio-blk polling is implemented upon 'batched completion' of
block
> layer. virtblk_poll() queues completed request to
io_comp_batch->req_list
> and later, virtblk_complete_batch() calls unmap function and ends
> the requests in batch.
>
> virtio-blk reads the number of poll queues from module parameter
> "num_poll_queues". If VM sets queue parameter as below,
> ("num-queues=N" [QEMU property], "num_poll_queues=M"
[module parameter])
> It allocates N virtqueues to virtio_blk->vqs[N] and it uses [0..(N-M-1)]
> as default queues and [(N-M)..(N-1)] as poll queues. Unlike the default
> queues, the poll queues have no callback function.
>
> Regarding HW-SW queue mapping, the default queue mapping uses the
> existing method that condsiders MSI irq vector. But the poll queue
> doesn't have an irq, so it uses the regular blk-mq cpu mapping.
>
> For verifying the improvement, I did Fio polling I/O performance test
> with io_uring engine with the options below.
> (io_uring, hipri, randread, direct=1, bs=512, iodepth=64 numjobs=N)
> I set 4 vcpu and 4 virtio-blk queues - 2 default queues and 2 poll
> queues for VM.
>
> As a result, IOPS and average latency improved about 10%.
>
> Test result:
>
> - Fio io_uring poll without virtio-blk poll support
> -- numjobs=1 : IOPS = 339K, avg latency = 188.33us
> -- numjobs=2 : IOPS = 367K, avg latency = 347.33us
> -- numjobs=4 : IOPS = 383K, avg latency = 682.06us
>
> - Fio io_uring poll with virtio-blk poll support
> -- numjobs=1 : IOPS = 380K, avg latency = 167.87us
> -- numjobs=2 : IOPS = 409K, avg latency = 312.6us
> -- numjobs=4 : IOPS = 413K, avg latency = 619.72us
>
> Reported-by: kernel test robot <lkp at intel.com>
> Signed-off-by: Suwan Kim <suwan.kim027 at gmail.com>
> ---
> drivers/block/virtio_blk.c | 101 +++++++++++++++++++++++++++++++++++--
> 1 file changed, 97 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 8c415be86732..3d16f8b753e7 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -37,6 +37,10 @@ MODULE_PARM_DESC(num_request_queues,
> "0 for no limit. "
> "Values > nr_cpu_ids truncated to nr_cpu_ids.");
>
> +static unsigned int num_poll_queues;
> +module_param(num_poll_queues, uint, 0644);
> +MODULE_PARM_DESC(num_poll_queues, "The number of dedicated virtqueues
for polling I/O");
> +
> static int major;
> static DEFINE_IDA(vd_index_ida);
>
Is there some way to make it work reasonably without need to set
module parameters? I don't see any other devices with a num_poll_queues
parameter - how do they handle this?
> @@ -81,6 +85,7 @@ struct virtio_blk {
>
> /* num of vqs */
> int num_vqs;
> + int io_queues[HCTX_MAX_TYPES];
> struct virtio_blk_vq *vqs;
> };
>
> @@ -548,6 +553,7 @@ static int init_vq(struct virtio_blk *vblk)
> const char **names;
> struct virtqueue **vqs;
> unsigned short num_vqs;
> + unsigned int num_poll_vqs;
> struct virtio_device *vdev = vblk->vdev;
> struct irq_affinity desc = { 0, };
>
> @@ -556,6 +562,7 @@ static int init_vq(struct virtio_blk *vblk)
> &num_vqs);
> if (err)
> num_vqs = 1;
> +
> if (!err && !num_vqs) {
> dev_err(&vdev->dev, "MQ advertised but zero queues
reported\n");
> return -EINVAL;
> @@ -565,6 +572,13 @@ static int init_vq(struct virtio_blk *vblk)
> min_not_zero(num_request_queues, nr_cpu_ids),
> num_vqs);
>
> + num_poll_vqs = min_t(unsigned int, num_poll_queues, num_vqs - 1);
> +
> + memset(vblk->io_queues, 0, sizeof(int) * HCTX_MAX_TYPES);
> + vblk->io_queues[HCTX_TYPE_DEFAULT] = num_vqs - num_poll_vqs;
> + vblk->io_queues[HCTX_TYPE_READ] = 0;
> + vblk->io_queues[HCTX_TYPE_POLL] = num_poll_vqs;
> +
> vblk->vqs = kmalloc_array(num_vqs, sizeof(*vblk->vqs), GFP_KERNEL);
> if (!vblk->vqs)
> return -ENOMEM;
> @@ -578,8 +592,13 @@ static int init_vq(struct virtio_blk *vblk)
> }
>
> for (i = 0; i < num_vqs; i++) {
> - callbacks[i] = virtblk_done;
> - snprintf(vblk->vqs[i].name, VQ_NAME_LEN, "req.%d", i);
> + if (i < num_vqs - num_poll_vqs) {
> + callbacks[i] = virtblk_done;
> + snprintf(vblk->vqs[i].name, VQ_NAME_LEN, "req.%d", i);
> + } else {
> + callbacks[i] = NULL;
> + snprintf(vblk->vqs[i].name, VQ_NAME_LEN, "req_poll.%d",
i);
> + }
> names[i] = vblk->vqs[i].name;
> }
>
> @@ -728,16 +747,87 @@ static const struct attribute_group
*virtblk_attr_groups[] = {
> static int virtblk_map_queues(struct blk_mq_tag_set *set)
> {
> struct virtio_blk *vblk = set->driver_data;
> + int i, qoff;
> +
> + for (i = 0, qoff = 0; i < set->nr_maps; i++) {
> + struct blk_mq_queue_map *map = &set->map[i];
> +
> + map->nr_queues = vblk->io_queues[i];
> + map->queue_offset = qoff;
> + qoff += map->nr_queues;
> +
> + if (map->nr_queues == 0)
> + continue;
> +
> + /*
> + * Regular queues have interrupts and hence CPU affinity is
> + * defined by the core virtio code, but polling queues have
> + * no interrupts so we let the block layer assign CPU affinity.
> + */
> + if (i == HCTX_TYPE_DEFAULT)
> + blk_mq_virtio_map_queues(&set->map[i], vblk->vdev, 0);
> + else
> + blk_mq_map_queues(&set->map[i]);
> + }
> +
> + return 0;
> +}
> +
> +static void virtblk_complete_batch(struct io_comp_batch *iob)
> +{
> + struct request *req;
> + struct virtblk_req *vbr;
>
> - return blk_mq_virtio_map_queues(&set->map[HCTX_TYPE_DEFAULT],
> - vblk->vdev, 0);
> + rq_list_for_each(&iob->req_list, req) {
> + vbr = blk_mq_rq_to_pdu(req);
> + virtblk_unmap_data(req, vbr);
> + 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_vq *vq = hctx->driver_data;
> + 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_add_to_batch(req, iob, vbr->status,
> + virtblk_complete_batch))
> + blk_mq_complete_request(req);
> + }
> +
> + spin_unlock_irqrestore(&vq->lock, flags);
> +
> + return found;
> +}
> +
> +static int virtblk_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
> + unsigned int hctx_idx)
> +{
> + struct virtio_blk *vblk = data;
> + struct virtio_blk_vq *vq = &vblk->vqs[hctx_idx];
> +
> + WARN_ON(vblk->tag_set.tags[hctx_idx] != hctx->tags);
> + hctx->driver_data = vq;
> + return 0;
> }
>
> static const struct blk_mq_ops virtio_mq_ops = {
> .queue_rq = virtio_queue_rq,
> .commit_rqs = virtio_commit_rqs,
> + .init_hctx = virtblk_init_hctx,
> .complete = virtblk_request_done,
> .map_queues = virtblk_map_queues,
> + .poll = virtblk_poll,
> };
>
> static unsigned int virtblk_queue_depth;
> @@ -816,6 +906,9 @@ static int virtblk_probe(struct virtio_device *vdev)
> sizeof(struct scatterlist) * VIRTIO_BLK_INLINE_SG_CNT;
> vblk->tag_set.driver_data = vblk;
> vblk->tag_set.nr_hw_queues = vblk->num_vqs;
> + vblk->tag_set.nr_maps = 1;
> + if (vblk->io_queues[HCTX_TYPE_POLL])
> + vblk->tag_set.nr_maps = 3;
>
> err = blk_mq_alloc_tag_set(&vblk->tag_set);
> if (err)
> --
> 2.26.3