Re: [PATCH v13] virtio-net: support device stats

[Jason Wang <jasowang@redhat.com>]

On Mon, Jul 31, 2023 at 12:32âPM Parav Pandit <parav@nvidia.com> wrote:
>
>
> > From: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> > Sent: Thursday, July 27, 2023 7:18 AM
>
> > > why don't we have a config space field for supported types?
> > > more straight forward than complex rules...
> >
> > I am ok for adding a config space field.
>
> During 1.3 we completed the discussion that any new config field addition must adhere to DMA interface and not a register.
> Multiple capabilities need this.
> So lets please work on that.
>
> Counters needs to use the dma interface and it can wait one time to utilize the modern interface.
>
> I have two proposals to for the dma interface and we should discuss that in separate thread to not mix the cvq parts of the stats.
> I will start the thread shortly and put pointer to this requirement of counters + many others that we are discussing including ntuple.

It's really the charge of the transport to determine the way to access
the configuration space, not the device itself.

CCW had already used DMA to access device configuration space. PCI can
do the same.

Thanks

>
>