[dpdk-dev] net/virtio-user: send kick to tx queue to notify backend on initialization
Checks
Commit Message
Acccording to the spec, https://fossies.org/linux/qemu/docs/specs/vhost-user.txt
client must start ring upon receiving a kick (that is, detecting that file
descriptor is reachable) on the descriptor specified by
VHOST_USER_SET_VRING_KICK.
The code sends a kick to the rx queue. It is missing sending a kick for the
tx queue. This patch is to add the missing code to comply with the spec.
Signed-off-by: Steven <sluong@cisco.com>
---
drivers/net/virtio/virtio_ethdev.c | 5 +++++
1 file changed, 5 insertions(+)
Comments
Dear Maxime and Yuanhan,
Do you have any more question on this patch? According to the text in the spec, the vring/queue is supposed to be disable until a kick is received. This is to ensure that the kickfd is working properly prior to the operation and I think it makes sense. The only exception to this rule is when the driver opts out for the interrupt support (no kickfd) in VHOST_USER_SET_VRING_KICK message.
Steven
On 8/3/17, 7:12 AM, "dev on behalf of Steven Luong (sluong)" <dev-bounces@dpdk.org on behalf of sluong@cisco.com> wrote:
Maxime,
Thank you so much for the reply.
1. It’s about conforming to the spec. Please read the text as I quoted in the email. A non-conforming implementation does not communicate with a conforming implementation such as VPP.
2. QEMU’s implementation is conforming and it is sending kick for both TX and RX queues upon initialization.
Steven
On 8/3/17, 2:37 AM, "dev on behalf of Maxime Coquelin" <dev-bounces@dpdk.org on behalf of maxime.coquelin@redhat.com> wrote:
Hi Steven,
On 08/01/2017 06:17 PM, Steven wrote:
> Acccording to the spec, https://fossies.org/linux/qemu/docs/specs/vhost-user.txt
>
> client must start ring upon receiving a kick (that is, detecting that file
> descriptor is reachable) on the descriptor specified by
> VHOST_USER_SET_VRING_KICK.
>
> The code sends a kick to the rx queue. It is missing sending a kick for the
> tx queue. This patch is to add the missing code to comply with the spec.
>
> Signed-off-by: Steven <sluong@cisco.com>
> ---
> drivers/net/virtio/virtio_ethdev.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
> index 00a3122..6362e14 100644
> --- a/drivers/net/virtio/virtio_ethdev.c
> +++ b/drivers/net/virtio/virtio_ethdev.c
> @@ -1747,6 +1747,11 @@ virtio_dev_start(struct rte_eth_dev *dev)
> virtqueue_notify(rxvq->vq);
> }
>
> + for (i = 0; i < dev->data->nb_tx_queues; i++) {
> + txvq = dev->data->tx_queues[i];
> + virtqueue_notify(txvq->vq);
> + }
> +
I'm not sure to get why we would need to send Txq notification whereas
no packet have been enqueued. That said, I don't think it hurts.
Steven, does it solve a real problem you are facing with virtio-user?
Yuanhan, what's your opinion on this?
Cheers,
Maxime
On Tue, Aug 01, 2017 at 09:17:36AM -0700, Steven wrote:
> Acccording to the spec, https://fossies.org/linux/qemu/docs/specs/vhost-user.txt
>
> client must start ring upon receiving a kick (that is, detecting that file
> descriptor is reachable) on the descriptor specified by
> VHOST_USER_SET_VRING_KICK.
>
> The code sends a kick to the rx queue. It is missing sending a kick for the
> tx queue. This patch is to add the missing code to comply with the spec.
>
> Signed-off-by: Steven <sluong@cisco.com>
Applied to dpdk-next-virtio.
Thanks.
--yliu
@@ -1747,6 +1747,11 @@ virtio_dev_start(struct rte_eth_dev *dev)
virtqueue_notify(rxvq->vq);
}
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ txvq = dev->data->tx_queues[i];
+ virtqueue_notify(txvq->vq);
+ }
+
PMD_INIT_LOG(DEBUG, "Notified backend at initialization");
for (i = 0; i < dev->data->nb_rx_queues; i++) {