On 08/18/2017 10:22 AM, Michael S. Tsirkin wrote:
+static void send_balloon_page_sg(struct virtio_balloon *vb, + struct virtqueue *vq, + void *addr, + uint32_t size) +{ + unsigned int len; + int ret; + + do { + ret = add_one_sg(vq, addr, size); + virtqueue_kick(vq); + wait_event(vb->acked, virtqueue_get_buf(vq, &len)); + /* + * It is uncommon to see the vq is full, because the sg is sent + * one by one and the device is able to handle it in time. But + * if that happens, we go back to retry after an entry gets + * released. + */ Why send one by one though? Why not batch some s/gs and wait for all of them to be completed? If memory if fragmented, waiting every time is worse than what we have now (VIRTIO_BALLOON_ARRAY_PFNS_MAX at a time).
OK, I'll do batching in some fashion. Best, Wei
