Re: [PATCH v2] virtio-net: support the virtqueue coalescing moderation

[Alvaro Karsz <alvaro.karsz@solid-run.com>]

> Please add short description something like,
>
> When the driver prefers to use per virtqueue notifications coalescing, and if queue group (transmit or receive) level notification coalescing is enabled, driver SHOULD first disable device level notification coalescing.
> Or it should be,
>

I disagree here.
IMO "queue group level notification coalescing" is not something to
enable or disable, but a shortcut to set all TX/RX queues at once.
Why should the spec force a driver to "disable device level
notification coalescing" (I assume you mean send a
VIRTIO_NET_CTRL_NOTF_COAL_[T/R]X_SET command with zeros)?
What if the driver sends a VIRTIO_NET_CTRL_NOTF_COAL_[T/R]X_SET
command, and then a single queue traffic increases? why should it zero
the parameters to all other queues?
I think that this should be discussed in the driver implementation
stage, not in the spec.

> Virtqueue level notifications coalescing, and device level notifications can be enabled together.
> When both of them are enabled, per virtqueue notifications coalescing take priority over queue group level.

How do you enable  Virtqueue level notifications coalescing? Why are
they different entities?
I don't think that we should have priorities, but the last command
should be the one that dictates the coalescing parameters.

For example, let's look at vq0 (RX):
Device receives VIRTIO_NET_CTRL_NOTF_COAL_RX_SET, vq0 should change
the parameters accordingly (all RX vqs should do the same).
Then device receives VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET with vqn = 0,
vq0 changes the parameters accordingly (all RX vqs are still using the
"old" parameters)
Then device receives VIRTIO_NET_CTRL_NOTF_COAL_RX_SET, vq0 changes the
parameters accordingly (all RX vqs should do the same).