Re: [PATCH v8 net-next 5/5] net: ena: Add PHC documentation

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 7 Apr 2025 16:02:33 +0200 Andrew Lunn wrote:
> > Thanks for suggesting the devlink params option for enable/disable, we will
> > explore the option and provide a revised patchset.
> > 
> > Given the pushback on custom sysfs utilization, what can be the alternative for exposing 
> > the PHC statistics? If `ethtool -S` is not an option, is there another framework that
> > allows outputting statistics?
> > We've explored devlink health reporter dump, would that be acceptable?  
> 
> We seem to be going backwards and forwards between this is connected
> to a netdev and it is not connected to a netdev. You have to destroy
> and recreate the netdev in order to make us if it, which might just be
> FUBAR design, but that is what you have. So maybe ethtool -S is an
> option?

The device has to be reinitialized, I don't think full re-initialization
makes the feature connected to other functionality.

> Or take a step back. Are your statistics specific to your hardware, or
> generic about any PTP clock? Could you expand the PTP infrastructure
> to return generic statistics? The problem being, PTP is currently
> IOCTL based, so not very expandable, unlike netlink used for ethtool.

Historic parts are IOCTL-based, but extended functionality for PTP core
and drivers is currently implemented using sysfs.