RE: [PATCH v14 10/11] virtio-net: Describe RSS using rss rq id

[Parav Pandit <parav@nvidia.com>]

> From: Michael S. Tsirkin <mst@redhat.com>
> Sent: Monday, April 24, 2023 11:54 AM

> > The data type of the unclassified_queue is struct rss_rq id, so we don't need
> to emphasis it again.
> 
> oh, you can drop that:
Bit 1 to 16 mechanics to be defined once in rss_rq_id definition.
I don't want to repeat it 3 times for unclassified q, indirection table and on device side.

So I didn't follow this comment about comment.

> 	\field{struct rss_rq_id} contains a virtqueue index:
> \field{vq_index_1_16}
> 	consists of bits 1 to 16 of a virtqueue index. For example, a
> 	\field{vq_index_1_16} value of 3 corresponds to virtqueue index 6,
> which maps
> 	to receiveq4.
> 
> 	Field \field{unclassified_queue} contains the virtqueue index of the
> receive
> 	virtqueue to place unclassified packets in.
> 
What is proposed is this patch is:

"Field unclassified_queue _contains_ the receive virtqueue in which to place unclassified packets."

This we can change to,

Field unclassified_queue _specifies_ the receive virtqueue in which to place unclassified packets.

> but in any case, there is no point in saying "\field{rss_rq_id} is a receive
> virtqueue id." This concept of "receive virtqueue id"
> is not really useful.
> 
Without repeating text 3 times rss_rq_id serve the purpose.

> 
> 
> > For example, we don't say, max_tx_vq is in le16 format.
> >
> > I am going to keep the current version as it is better than then extra
> verbosity.
> 
> Maybe "specifies" instead of "contains" then?
> 
> +Field \field{unclassified_queue} specifies the receive virtqueue in
> +which to place unclassified packets.
> 
Yes "specifies" is good. This helps to define the format also at one place.