On Tue, 5 Mar 2024 06:59:39 -0800 Jakub Kicinski <kuba@xxxxxxxxxx> wrote: > On Tue, 5 Mar 2024 11:10:21 +0100 Köry Maincent wrote: > > > No. In the case of mvpp2 + marvell PHY, the two PTP implementations are > > > entirely separate. > > > > Yes the PTP clock can be independent from the netdev. > > We need to know which software layer register the PHC to be able to call its > > callbacks. > > > > My commit log is a bit small here. I will enhance it in the next version. > > Still, wouldn't it be simpler to store all accessible PTP instances > in the netdev? You are talking about something like the phy topology but for the ptp? Then when asking information on a PHC (tsinfo or hwtstamp config) from ethtool we would have to look at the PHC topology of the netdev. This could work. Not sure it is much simpler, do you see other advantages that it could have? Regards, -- Köry Maincent, Bootlin Embedded Linux and kernel engineering https://bootlin.com
