vhost: try avoiding avail index access when getting descriptor (e3b56cdd) · Commits · 戴 / test

If last avail idx is not equal to cached avail idx, we're sure there's still available buffers in the virtqueue so there's no need to re-read avail idx. So let's skip this to avoid unnecessary userspace memory access and memory barrier. Pktgen test show about 3% improvement on rx pps. Signed-off-by:

Jason Wang <jasowang@redhat.com> Signed-off-by:

Michael S. Tsirkin <mst@redhat.com>

drivers/vhost/vhost.c

+23 −16

Original line number	Diff line number	Diff line
		@@ -1930,6 +1930,8 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq,

		/* Check it isn't doing very strange things with descriptor numbers. */
		last_avail_idx = vq->last_avail_idx;

		if (vq->avail_idx == vq->last_avail_idx) {
		if (unlikely(vhost_get_user(vq, avail_idx, &vq->avail->idx))) {
		vq_err(vq, "Failed to access avail idx at %p\n",
		&vq->avail->idx);
		@@ -1943,12 +1945,17 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq,
		return -EFAULT;
		}

		/* If there's nothing new since last we looked, return invalid. */
		/* If there's nothing new since last we looked, return
		* invalid.
		*/
		if (vq->avail_idx == last_avail_idx)
		return vq->num;

		/* Only get avail ring entries after they have been exposed by guest. */
		/* Only get avail ring entries after they have been
		* exposed by guest.
		*/
		smp_rmb();
		}

		/* Grab the next descriptor number they're advertising, and increment
		* the index we've seen. */

Admin message