From: Pawel Chmielewski <redacted>
Date: Thu, 14 Sep 2023 16:27:28 +0200

On Sun, Sep 03, 2023 at 04:00:57PM +0200, Andrew Lunn wrote:

quoted

quoted

Let me check if I understand correctly- is that what was sent with the
v3 [1] , with the initialization helper (ethtool_forced_speed_maps_init)
and the structure map in the ethtool code? Or do you have another helper
in mind?

Sorry for the late reply, been on vacation.

The main thing is you try to reuse the table:

static const struct phy_setting settings[] = {}

If you can build your helper on top of phy_lookup_setting() even
better. You don't need a phy_device to use those.

	Andrew

Thank for the hint Andrew! I took a look into the phy-core code,
and a little into phylink. However, I still have the same concern

Here I'd like to add that we're planning to try to use Phylink in ice
soon. It may take a while and will most likely require core code
expansion, since Phylink was originally developed for embedded HW and
DeviceTree and doesn't fully support PCI devices.
Let's see how it goes.

regarding modes that are supported/unsupported by hardware (managed
by the firmware in our case). Let's say I'm only looking for duplex
modes and iterate over speeds with advertised modes map as an argument
for phy_lookup_setting. In this case, I still need another table/map of
hardware compatible link modes to check against. Theese are actually
the maps we'd like to keep in the driver (and proposed in [1]), so
maybe the simple intersect check between them and the advertised modes
is sufficient?

[1] https://lore.kernel.org/netdev/20230823180633.2450617-4-pawel.chmielewski@intel.com/ (local)

Thanks,
Olek