Re: [PATCH net-next 1/4] net: Wrap ndo_do_ioctl() to prepare for DSA stacked ops

[Vladimir Oltean <olteanv@gmail.com>]

On Sat, Jul 18, 2020 at 01:36:25PM -0700, Florian Fainelli wrote:
> 
> 
> On 7/18/2020 1:30 PM, Vladimir Oltean wrote:
> > Hi Florian,
> > 
> > On Fri, Jul 17, 2020 at 08:05:30PM -0700, Florian Fainelli wrote:
> >> In preparation for adding another layer of call into a DSA stacked ops
> >> singleton, wrap the ndo_do_ioctl() call into dev_do_ioctl().
> >>
> >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> >> ---
> > 
> > I missed most of the context, but this looks interesting. Am I correct
> > in saying that this could save us from having to manually pass
> > phy_mii_ioctl() and that it could be generically handled here?
> 
> The motivation for this work started with the realization while
> untangling the ethtool/netlink and PHY library that tests like those:
> dev->netdev_ops == &foo_ops would be defeated by the way DSA overloads
> the DSA net_device operations. A better solution needed to be found.
> 
> You are correct that we could just put a call to phy_mii_ioctl() here as
> well and avoid having drivers have to use phy_do_ioctl_running or roll
> their own.
> -- 
> Florian

I see.
The background for my question is that we came to realize that it would
be way cleaner if an Ethernet PHY with 1588 timestamping could just
overload .ndo_do_ioctl() and the ethtool .get_ts_info() of the host
network interface, very similarly to what DSA does.
So it's very good that you started this work, it looks like it provides
a very good foundation. Thanks!

-Vladimir