Mon, Nov 10, 2025 at 11:58:31PM +0100, kuba@kernel.org wrote:

On Mon, 10 Nov 2025 08:05:57 -0500 Daniel Zahka wrote:

quoted

On 11/9/25 5:39 AM, Jiri Pirko wrote:

quoted

Daniel, I asked twice if this could be a non-driver param. Jakub asked
for clearer definition of this know in that context.

Not sure why you are ignoring this :/
 

My apologies. I think there was a miscommunication. I assumed Jakub's 
question was directed towards you. I have no objection to making it a 
generic param; I will do so in v4. It sounded to me like Jakub was 
wanting more information on what exactly this setting does beyond what I 
was able to provide in the commit message and mlx5 devlink 
documentation. My understanding is that this setting pertains to tx 
csums and how the device expects the driver to prepare partial csums 
when doing tx cso. I don't really know more than that. Especially not 
something like what the FW's role in implementing this is.

Right, per To: field of my email I as asking Jiri for clarifications.

Since we struggle to understand the semantics nack on making this
generic. Chances are whoever reuses the "generic" param will have a
different interpretation of its meaning, so what's the point of making
it generic.

Okay, I don't mind that much.