[PATCH net-next 3/3] net: ethtool: tsconfig: Re-configure hwtstamp upon provider change

[Russell King (Oracle)]

On Wed, Oct 15, 2025 at 02:45:26PM +0200, Kory Maincent wrote:
> On Wed, 15 Oct 2025 12:27:23 +0200
> Maxime Chevallier <maxime.chevallier at bootlin.com> wrote:
> 
> > When a hwprov timestamping source is changed, but without updating the
> > timestamping parameters, we may want to reconfigure the timestamping
> > source to enable the new provider.
> > 
> > This is especially important if the same HW unit implements 2 providers,
> > a precise and an approx one. In this case, we need to make sure we call
> > the hwtstamp_set operation for the newly selected provider.
> 
> This is a design choice.
> Do we want to preserve the hwtstamp config if only the hwtstamp source is
> changed from ethtool?

This depends what is meant by "preserve". If the hwtstamp capabilities
of the two sources being switched between are the same in terms of how
userspace configures them, then it's fine.

However, it's my understanding that the hwtstamp configuration is a
negotiation between kernel and userspace - drivers are required to
return back to userspace what they're actually doing when userspace
requests a certain configuration. If the hwtstamp capabilities are
different, it breaks this model because what the previous instance
reports back to userspace for a certain configuration could be
different to that which the new instance would report back.

This could get worse when a configuration is set on the previous
instance that isn't supported by the new instance.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!