[PATCH 1/4] phy: miphy365x: Add Device Tree bindings for the MiPHY365x

[Lee Jones]

> On Tuesday 17 June 2014 04:53 PM, Lee Jones wrote:
> >>> The MiPHY365x is a Generic PHY which can serve various SATA or PCIe
> >>> devices. It has 2 ports which it can use for either; both SATA, both
> >>> PCIe or one of each in any configuration.
> >>
> >> I've asked others who wrote multi-phy PHY providers to model each individual
> >> PHY as sub-node of the PHY provider. So It's only fair I ask you the same to
> >> do. So in this case the dt node should look something like:
> >>
> >> 	miphy365x_phy: miphy365x at fe382000 {
> >> 		compatible = "st,miphy365x-phy";
> >> 		#phy-cells = <2>;
> >> 		st,syscfg = <&syscfg_rear>;
> >> 		channel at 0 {
> >> 			reg =	<0xfe382000 0x100>, <0xfe394000 0x100>;
> >> 			reg-names = "sata", "pcie";
> >> 		}
> >>
> >> 		channel at 1{
> >> 			reg =	<0xfe38a000 0x100>, <0xfe804000 0x100>;
> >> 			reg-names = "sata", "pcie";
> >> 		}
> >>
> >> 	};
> > 
> > I'm interested to know why you've taken this approach, as it makes the
> > code much more complex.  The DT framework goes to the trouble of
> 
> It looks to be much closer representation of the hardware in dt. It also
> enables to have more control of each individual PHYs. For example, we can have
> something like status="disabled" for channels which is disabled.

If you wanted to disable the channels in this way, you would either
have to A) add your own code to parse that property or B) have each
channel set itself up as platform device and would probe (or not if
status = "disabled") individually.  If you choose the later option,
the platform resource would also be populated.

> > converting all addresses to to resources so drivers can easily pull
> > them out using platform_get_resource() and friends.  Pushing the reg
> 
> right. Can't we use of_address_to_resource here?

We could, but that would be an extra layer.  We'd be pulling the
address, putting it into a resource, then pulling it from the resource
for use.  If we're going to be pulling addresses out manually, we're
probably better off using of_get_address().  But again, we're just
carrying out functionality which is already provided by the
framework.

> > properties down into a child node means that we have to now iterate
> > over the sub-nodes and pull them out manually.  This will lead to a
> 
> You anyway iterate while creating PHYs based on some constant. Now you have to
> iterate over the sub-nodes.
> > pretty messy implementation IMHO.

This much is true.

> > Can you point me in the direction of previous implementations where you
> > have stipulated the same set of constraints please?
> 
> ah.. there isn't any. The author of the other multi-phy driver [1] also feels
> this will just add to the complexity of the driver.

=:)

> Maybe we should ask the opinion of others?

We could.  I'll CC Arnd as he likes this PHY stuff. :)

> [1] -> http://www.spinics.net/lists/linux-sh/msg32087.html

-- 
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog