Re: [PATCH net-next 06/16] net: pcs: xpcs: Avoid creating dummy XPCS MDIO device

[Serge Semin]

Hi Andrew, Russell

Sorry for the delay with response. I had to refresh my understanding
of the series since it was created sometime ago and I already managed
to forget some of its aspects (particularly regarding the MDIO-bus
PHY-mask semantics).

On Tue, Dec 05, 2023 at 02:50:58PM +0000, Russell King (Oracle) wrote:
> On Tue, Dec 05, 2023 at 02:52:24PM +0100, Andrew Lunn wrote:
> > On Tue, Dec 05, 2023 at 02:31:41PM +0300, Serge Semin wrote:
> > > On Tue, Dec 05, 2023 at 10:49:47AM +0000, Russell King (Oracle) wrote:
> > > > On Tue, Dec 05, 2023 at 01:35:27PM +0300, Serge Semin wrote:
> > > > > If the DW XPCS MDIO devices are either left unmasked for being auto-probed
> > > > > or explicitly registered in the MDIO subsystem by means of the
> > > > > mdiobus_register_board_info() method there is no point in creating the
> > > > > dummy MDIO device instance in order to get the DW XPCS handler since the
> > > > > MDIO core subsystem will create the device during the MDIO bus
> > > > > registration procedure.
> > > > 
> > > 
> > > > Please reword this overly long sentence.
> > > 
> > > Ok.
> > > 
> > > > 
> > > > If they're left unmasked, what prevents them being created as PHY
> > > > devices?
> > > 
> > > Not sure I fully get what you meant. If they are left unmasked the
> > > MDIO-device descriptor will be created by the MDIO subsystem anyway.
> > > What the point in creating another one?
> > 

> > Saying what Russell said, in a different way:
> > 
> > /*
> >  * Return true if the child node is for a phy. It must either:
> >  * o Compatible string of "ethernet-phy-idX.X"
> >  * o Compatible string of "ethernet-phy-ieee802.3-c45"
> >  * o Compatible string of "ethernet-phy-ieee802.3-c22"
> >  * o In the white list above (and issue a warning)
> >  * o No compatibility string
> >  *
> >  * A device which is not a phy is expected to have a compatible string
> >  * indicating what sort of device it is.
> >  */
> > bool of_mdiobus_child_is_phy(struct device_node *child)
> > 
> > So when walking the bus, if a node is found which fits these criteria,
> > its assumed to be a PHY. 
> > 
> > Anything on the MDIO bus which is not a PHY needs to use a compatible.
> 
> Right. I'd actually forgotten about the firmware-based walking, and
> was thinking more of the non-firmware bus scanning as the commit
> message was talking about being _unmasked_ and the only mask we have
> is bus->phy_mask.

Back then when I was working on the series and up until last week I
had thought that having a device unmasked in mii_bus->phy_mask was a
correct way to do for _any_ device including our DW XPCS (which BTW
looks like a normal C45 PHY and if synthesized with a PMA attached
could be passed to be handled by the PHY subsystem). Can't remember
why exactly I came to that thought, but likely it was due to finding
out examples of having mii_bus->phy_mask uninitialized in some of the
PCS use-cases, like in drivers/net/dsa/ocelot/felix_vsc9959.c (but in
case of DW XPCS the mask is always set indeed). Anyway obviously I was
wrong and PHY-device is supposed to be created only if a device is
actual PHY and handled by the PHY subsystem drivers. So the correct
ways to create PHY MDIO-devices are:

1. Call mdiobus_register() with PHY-addresses unmasked
2. Call of_mdiobus_register() for a DT-node with sub-nodes for which
of_mdiobus_child_is_phy() returns true.

and the correct ways to create non-PHY MDIO-devices are:

1. Call mdiobus_register() with non-PHY-addresses masked and have
those non-PHY device registered by mdiobus_register_board_info()
beforehand.
2. Call of_mdiobus_register() with DT sub-nodes having specific
compatible string (based on the of_mdiobus_child_is_phy() semantics).

Only in case of having a non-PHY device registered it's allowed to
use it in in non-PHY MDIO driver, like PCS, etc. Right?

Please correct me if I am wrong or miss something.

> 
> It seems to me that this is yet another case of a really confusing
> commit message making review harder than it needs to be.

>From the perspective described above the patch log is indeed partly
wrong. Sorry about that. I shouldn't have mentioned the mask at all
but instead just listed two use-cases of creating the non-PHY
MDIO-devices. I'll fix that in v2.

-Serge(y)

> 
> -- 
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
>