Re: [PATCH 2/6] phy: improved lookup method

[Heikki Krogerus]

> >>>>>>> Assume you have 2 phys in your system..
> >>>>>>> static struct phy_lookup usb_lookup = {
> >>>>>>> 	.phy_name	= "phy-usb.0",
> >>>>>>> 	.dev_id		= "usb.0",
> >>>>>>> 	.con_id		= "usb",
> >>>>>>> };
> >>>>>>>
> >>>>>>> static struct phy_lookup sata_lookup = {
> >>>>>>> 	.phy_name	= "sata-usb.1",
> >>>>>>> 	.dev_id		= "sata.0",
> >>>>>>> 	.con_id		= "sata",
> >>>>>>> };
> >>>>>>>
> >>>>>>> First you do modprobe phy-usb, the probe of USB PHY driver gets invoked and it
> >>>>>>> creates the PHY. The phy-core will find a free id (now it will be 0) and then
> >>>>>>> name the phy as phy-usb.0.
> >>>>>>> Then with modprobe phy-sata, the phy-core will create phy-sata.1.
> >>>>>>>
> >>>>>>> This is an ideal case where the .phy_name in phy_lookup matches.
> >>>>>>>
> >>>>>>> Consider if the order is flipped and the user does modprobe phy-sata first. The
> >>>>>>> phy_names won't match anymore (the sata phy device name would be "sata-usb.0").
> >>>>>
> >>>>> Actually, I don't think there would be this problem if we used the
> >>>>> name of the actual device which is the parent of phy devices, right?
> >>>>
> >>>> hmm.. but if the parent is a multi-phy phy provider (like pipe3 PHY driver), we
> >>>> might end up with the same problem.
> >>>
> >>> I'm not completely sure what you mean? If you are talking about
> >>> platforms with multiple instances of a single phy, I don't see how
> >>> there could ever be a scenario where we did not know the order in
> >>> which they were enumerated. Can you give an example again?
> >>
> >> If a single IP implements multiple PHYs (phy-miphy365x.c in linux-phy next),
> >> the parent for all the phy devices would be the same.

Hold on...

Let's take a step back here. Where could we actually have a scenario
where the phy device, the dev_id (consumer) and the con_id would all
be the same? There can't be such a case.

It's not like you could ever have a driver requesting multiple phys
with the same con_id. You would just get the same phy handle even if
you used dt.

        phy1 = phy_get(dev, "phy");
        ...
        phy2 = phy_get(dev, "phy");

And if the drivers requesting those phys are different, your consumers
are different.

> Isn't making the PHY to be aware of it's user much simpler?

No it's not. I'm not going into this again. We have already gone
through this in the past.


Cheers,

-- 
heikki
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/