Re: [dpdk-dev] [PATCH 3/3] ethdev: add async queue-based flow rules operations

[Ajit Khaparde <ajit.khaparde@broadcom.com>]

On Tue, Oct 5, 2021 at 9:49 PM Alexander Kozyrev <akozyrev@nvidia.com> wrote:
>
> A new, faster, queue-based flow rules management mechanism is needed for
> applications offloading rules inside the datapath. This asynchronous
> and lockless mechanism frees the CPU for further packet processing and
> reduces the performance impact of the flow rules creation/destruction
> on the datapath. Note that queues are not thread-safe and queue-based
> operations can be safely invoked without any locks from a single thread.
>
> The rte_flow_q_flow_create() function enqueues a flow creation to the
> requested queue. It benefits from already configured resources and sets
> unique values on top of item and action templates. A flow rule is enqueued
> on the specified flow queue and offloaded asynchronously to the hardware.
> The function returns immediately to spare CPU for further packet
> processing. The application must invoke the rte_flow_q_dequeue() function
> to complete the flow rule operation offloading, to clear the queue, and to
> receive the operation status. The rte_flow_q_flow_destroy() function
> enqueues a flow destruction to the requested queue.
>
> Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
> Suggested-by: Ori Kam <orika@nvidia.com>
> ---
>  lib/ethdev/rte_flow.h | 288 ++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 288 insertions(+)
>
> diff --git a/lib/ethdev/rte_flow.h b/lib/ethdev/rte_flow.h
> index ba3204b17e..8cdffd8d2e 100644
> --- a/lib/ethdev/rte_flow.h
> +++ b/lib/ethdev/rte_flow.h
> @@ -4298,6 +4298,13 @@ struct rte_flow_port_attr {
>          * Version of the struct layout, should be 0.
>          */
>         uint32_t version;
> +       /**
> +        * Number of flow queues to be configured.
> +        * Flow queues are used for asyncronous flow rule creation/destruction.
> +        * The order of operations is not guaranteed inside a queue.
> +        * Flow queues are not thread-safe.
> +        */
> +       uint16_t nb_queues;
Will it matter if PMD can create a smaller set of queues? Or may be just one?
Should the application set this based on get_infos_get() or some other
mechanism?

::::