On Wed, Oct 8, 2025 at 7:45 AM Luca Vizzarro <luca.vizzarro@arm.com> wrote:
Hi Patrick,

I am assuming the move of send_packets is because we can't send
arbitrary packets with a perf generator. I believe this should be
explained as it's not clear why the move and it feels counterintuitive
at first.

Sort of. We can send arbitrary packets with a perf generator, but there is a particular use case for send_packet (send a small number of arbitrary packets) and this is just "better" done by a capturing packet generator like Scapy since it is concerned with reliability. The performance traffic generators have a different focus (send in high volume), so in my view it makes sense to encourage people to use a capturing traffic generator when they want to send a small amount of arbitrary packets. If you think this idea is okay, yes I can update the commit message. Or, I can reverse the decision altogether if you don't think my justification is sound.
 

On Wed, Oct 01, 2025 at 07:16:57PM +0000, Patrick Robb wrote:
> diff --git a/dts/framework/testbed_model/traffic_generator/performance_traffic_generator.py b/dts/framework/testbed_model/traffic_generator/performance_traffic_generator.py
> new file mode 100644
> index 0000000000..6b23faa1a5

> +    def setup(self, topology: Topology) -> None:
> +        """Overrides :meth:`.traffic_generator.TrafficGenerator.setup`."""
> +        for port in self._tg_node.ports:
> +            self._tg_node.main_session.configure_port_mtu(2000, port)

should do:
  port.configure_mtu(2000)

Thanks for starting a review. Do you have the hardware available to test the patch as well? I am wondering whether my docs additions are sufficient (I think they are, but a second opinion is always good). If you can't, I'll ask Andrew to test it.