Re: [virtio-comment] [PATCH] Introduction of Virtio Network device interrupt coalescing feature

[Jason Wang <jasowang@redhat.com>]

On Mon, May 16, 2022 at 4:58 PM Stefan Hajnoczi <stefanha@redhat.com> wrote:
>
> On Thu, May 12, 2022 at 11:07:39AM +0300, Alvaro Karsz wrote:
> > Control a network device interrupt coalescing parameters using the control virtqueue.
> > A new control class was added: VIRTIO_NET_CTRL_INTR_COAL.
> >
> > This class provides 2 commands:
> > - VIRTIO_NET_CTRL_INTR_COAL_USECS_SET:
> >   Ask the network device to change the rx-usecs and tx-usecs parameters.
> >   rx-usecs - Time to delay an RX interrupt after packet arrival in microseconds.
> >   tx-usecs - Time to delay a TX interrupt after a sending a packet in microseconds.
> >
> > - VIRTIO_NET_CTRL_INTR_COAL_FRAMES_SET:
> >   Ask the network device to change the rx-max-frames and tx-max-frames parameters.
> >   rx-max-frames - Number of packets to delay an RX interrupt after packet arrival.
> >   tx-max-frames - Number of packets to delay a TX interrupt after sending a packet.
> >
> >
> > Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
> > ---
> >  content.tex | 41 +++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 41 insertions(+)
>
> Please add a sentence about how this interacts with Used Buffer
> Notification Suppression. My guess is that suppression takes effect in
> addition to virtio-net-specific coalescing, so it's possible that the
> notification will be suppressed even after the timer expires or frame
> count is exceeded?

My understanding is that, as you suggest, it's actually a coalsing of
notification. So this should work after

1) Notification Suppression: if driver doesn't want notification, the
device does not care about coalescing
2) Event idx: the notification sent for event must still be coalesced

Anyhow, I agree we need to clarify these.

Thanks

>
> Also, since the VIRTIO specification usually talks about Notifications
> instead of "interrupts" (see 2.3 Notifications) I think it makes sense
> to use that terminology instead of "interrupts". Interrupts are more of
> transport-level (e.g. PCI, MMIO) concept than a VIRTIO device model
> concept.