On Fri, 16 Aug 2024 08:20:44 -0400 Mina Almasry wrote:

quoted

I'd keep the current check with a WARN_ON_ONCE(), tho.
Given the absence of tests driver developers can use.
Especially those who _aren't_ supporting the feature.

Yes what I have locally is the driver setting
netdev_rx_queue->unreadable_netmem_supported when header split is
turned on, and additionally a WARN_ON_ONCE around the check in core. I
was about to send that when I read your email. I'm hoping we don't
have to go through the scope creep of adding configuration via the
queue API, which I think is a very significant undertaking.

I don't like adding more and more transient stuff to netdev_rx_queue.
It's one thing if we create a temporary solution in the core, which
we can easily redo later. It's another altogether when we expect drivers
to keep some bit up to date across all the reconfiguration paths they
have. Just to then got an replace that with another API.

If the post-check works let's go with that for now.