On Tue, Sep 29, 2020 at 05:09:48PM -0700, Jakub Kicinski wrote:
> On Tue, 29 Sep 2020 20:47:23 +0200 Andrew Lunn wrote:
> > > Do you mean report supported range via extack?
> >
> > Yes.
> >
> > 811ac400ea33 ("net: phy: dp83869: Add speed optimization feature")
> >
> > was merged recently. It has:
> >
> > + default:
> > + phydev_err(phydev,
> > + "Downshift count must be 1, 2, 4 or 8\n");
> > + return -EINVAL;
> >
> > and there are more examples in PHY drivers where it would be good to
> > tell the uses what the valid values are. I guess most won't see this
> > kernel message, but if netlink ethtool printed:
> >
> > Invalid Argument: Downshift count must be 1, 2, 4 or 8
> >
> > it would be a lot more user friendly.
>
> Ah, now I recall, we already discussed this.
>
> FWIW we could provision for the extack and just pass NULL for now?
> Would that be too ugly?
If Michal does not have any code lying around in a drawer, what might
be a good idea. For the old IOCTL we will need to pass a NULL anyway.
Andrew
