RE: [PATCH v10 net-next 4/8] net: ena: Control PHC enable through devlink

["Arinzon, David" <darinzon@amazon.com>]

> >> >> >+enum ena_devlink_param_id {
> >> >> >+      ENA_DEVLINK_PARAM_ID_BASE =
> >> >> DEVLINK_PARAM_GENERIC_ID_MAX,
> >> >> >+      ENA_DEVLINK_PARAM_ID_PHC_ENABLE,
> >> >>
> >> >> What exactly is driver/vendor specific about this? Sounds quite
> >> >> generic to me.
> >> >
> >> >Can you please clarify the question?
> >> >If you refer to the need of ENA_DEVLINK_PARAM_ID_PHC_ENABLE, it
> was
> >> >discussed as part of patchset v8 in
> >> >https://lore.kernel.org/netdev/20250304190504.3743-6-
> >> darinzon@amazon.co
> >> >m/ More specifically in
> >>
> >https://lore.kernel.org/netdev/55f9df6241d052a91dfde950af04c70969ea2
> >> >8
> >> b2
> >> >.camel@infradead.org/
> >> >
> >>
> >> Could you please read "Generic configuration parameters" section of
> >> Documentation/networking/devlink/devlink-params.rst? Perhaps that
> >> would help. So basically my question is, why your new param can't go in
> that list?
> >
> >Thanks for the clarification.
> >This is a topic that has been discussed in the versions of this
> >patchset, specifically in
> >https://lore.kernel.org/netdev/20250304190504.3743-6-
> darinzon@amazon.co
> >m/
> 
> Where exactly?

The discussion is from https://lore.kernel.org/netdev/20250304190504.3743-6-darinzon@amazon.com/ and split into two sub-threads
The suggestion to use devlink is from https://lore.kernel.org/netdev/20250402092344.5a12a26a@kernel.org/

> 
> 
> >Other modules in the kernel enable PHC unconditionally, due to
> >potential blast radius concerns, we've decided to not enable the feature
> unconditionally and allow customers to enable it if they choose to use the
> functionality.
> >As this is a specific behavior for the ENA driver, we've added a specific
> devlink parameter (was a sysfs entry previously and changed to devlink in v9
> due to feedback).
> 
> Why is this specific to ENA? Why any other driver can't have the same knob
> to enable/disable PHC?
> 

If there are no objections, I can add a generic command instead of a specific one.