On Mon, 19 Feb 2024 16:44:34 +0100 Andrew Lunn <andrew@xxxxxxx> wrote: > On Mon, Feb 19, 2024 at 04:04:56PM +0100, Köry Maincent wrote: > > On Fri, 16 Feb 2024 17:36:38 -0800 > > Jakub Kicinski <kuba@xxxxxxxxxx> wrote: > > > > > but why the separate header? Is it going to be used in other parts of > > > > > uAPI than just in ethtool? > > > > > > > > We might use it in pse core if capabilities between PoE and PoDL differ > > > > but I am not sure about it. > > > > Do you prefer to move it to ethtool header and add prefix ETHTOOL_ to > > > > the enum values? > > > > > > I don't know enough to have an opinion :) Whatever you end up doing, > > > it's probably worth documenting the reason for the choice in the commit > > > message? > > > > Mmh, I am still not sure of the best choice on this. I think I will move it > > to ethtool as you suggested. > > kAPI is hard to change. So it is worth thinking about it. > > Can you think of any possible kAPI not using ethtool netlink? Its not > going to be ioctl. We generally don't export new things in /sysfs if > we have netlink, etc. > > So to me, it is only going to be used be the ethtool API, so i would > follow the usual conventions for ethtool. Oops sorry forgot to reply to you. Indeed I reached to the same conclusion on my side. Regards, -- Köry Maincent, Bootlin Embedded Linux and kernel engineering https://bootlin.com
