qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v2 5/5] vhost: send VHOST_USER_SET_VRING_ENABLE


From: Michael S. Tsirkin
Subject: Re: [Qemu-devel] [PATCH v2 5/5] vhost: send VHOST_USER_SET_VRING_ENABLE at start/stop
Date: Wed, 21 Oct 2015 13:39:11 +0300

On Wed, Oct 21, 2015 at 05:07:18PM +0800, Yuanhan Liu wrote:
> Send VHOST_USER_SET_VRING_ENABLE at start/stop when multiple queue
> is negotiated, to inform the backend that we are ready or not.

OK but that's only if MQ is set. If now, we need to do
RESET_OWNER followed by SET_OWNER.

> And exclude VHOST_USER_GET_QUEUE_NUM as one time request, as we need
> to get max_queues for each vhost_dev.

Pls split this part out.

> Suggested-by: Michael S. Tsirkin <address@hidden>
> Signed-off-by: Yuanhan Liu <address@hidden>
> ---
>  hw/virtio/vhost-user.c |  1 -
>  hw/virtio/vhost.c      | 18 ++++++++++++++++++
>  2 files changed, 18 insertions(+), 1 deletion(-)
> 
> diff --git a/hw/virtio/vhost-user.c b/hw/virtio/vhost-user.c
> index 12a9104..6532a73 100644
> --- a/hw/virtio/vhost-user.c
> +++ b/hw/virtio/vhost-user.c
> @@ -194,7 +194,6 @@ static bool vhost_user_one_time_request(VhostUserRequest 
> request)
>      case VHOST_USER_SET_OWNER:
>      case VHOST_USER_RESET_OWNER:
>      case VHOST_USER_SET_MEM_TABLE:
> -    case VHOST_USER_GET_QUEUE_NUM:
>          return true;
>      default:
>          return false;
> diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> index c0ed5b2..54a4633 100644
> --- a/hw/virtio/vhost.c
> +++ b/hw/virtio/vhost.c
> @@ -1146,6 +1146,19 @@ int vhost_dev_start(struct vhost_dev *hdev, 
> VirtIODevice *vdev)
>          }
>      }
>  
> +    /*
> +     * Send VHOST_USER_SET_VRING_ENABLE message when multiple queue
> +     * is negotiated to inform the back end that we are ready.
> +     *
> +     * Only set enable to 1 for first queue pair, as we enable one
> +     * queue pair by default.
> +     */
> +    if (hdev->max_queues > 1 &&

this makes no sense in the generic code.
This is a work around for a protocol bug - belongs in
vhost user internally.
And that's not the way to test this. MQ could be negotiated
even if there is a single pair of queues.

> +        hdev->vhost_ops->vhost_backend_set_vring_enable) {
> +        hdev->vhost_ops->vhost_backend_set_vring_enable(hdev,
> +                                                        hdev->vq_index == 0);
> +    }
> +
>      return 0;
>  fail_log:
>      vhost_log_put(hdev, false);
> @@ -1180,5 +1193,10 @@ void vhost_dev_stop(struct vhost_dev *hdev, 
> VirtIODevice *vdev)
>      hdev->started = false;
>      hdev->log = NULL;
>      hdev->log_size = 0;
> +
> +    if (hdev->max_queues > 1 &&
> +        hdev->vhost_ops->vhost_backend_set_vring_enable) {
> +        hdev->vhost_ops->vhost_backend_set_vring_enable(hdev, 0);
> +    }
>  }
>  
> -- 
> 1.9.0
> 



reply via email to

[Prev in Thread] Current Thread [Next in Thread]