On Thursday 15 May 2014 02:47 PM, Sebastian Hesselbarth wrote: > On 05/15/2014 10:46 AM, Kishon Vijay Abraham I wrote: >> On Thursday 15 May 2014 12:32 PM, Sebastian Hesselbarth wrote: >>> On 05/15/2014 08:45 AM, Kishon Vijay Abraham I wrote: >>>> On Thursday 15 May 2014 12:12 AM, Sebastian Hesselbarth wrote: >>>>> On 05/14/2014 08:12 PM, Arnd Bergmann wrote: >>>>>> On Wednesday 14 May 2014 19:57:46 Sebastian Hesselbarth wrote: >>>>>>> Let's assume we have one dual-port SATA controller and one PCIe >>>>>>> controller with either x1 or x2 support. The only sane DT binding, >>>>>>> I can think of then would be: >>>>>>> >>>>>>> berlin2q.dtsi: >>>>>>> >>>>>>> genphy: lvds@ea00ff { >>>>>>> compatible = "marvell,berlin-lvds-phy"; >>>>>>> reg = <0xea00ff 0x100>; >>>>>>> #phy-cells = <2>; >>>>>>> }; > [...] >>> >>> Depends on what you call PHY. In the example above the PHY is what >>> allows you to control both lanes. >>> >>> So you want sub-nodes for each individual lane given the nomenclature >>> of the example? >>> >>> Or like it is used in the example above, a single PHY node with an index >>> in the phy-specifier to pick an individual lane. >>> >>> IMHO, having both phy-specifier index _and_ PHY sub-node per lane >>> has no benefit at all. You cannot even use the PHY sub-nodes for any >>> setup properties, as they depend on the consumer claiming the lane. >> >> IMO the dt data should completely describe the HW. So just by looking at the >> PHY node, we won't be able to tell the no of PHYs implemented in the IP if we >> have a single PHY node (In this case the lanes in the IP). >> >> However if you think it's an overkill for having sub-nodes for each lane then >> single PHY node is fine too. > > Yeah, I see your point. I just wonder how many Marvell PHYs we may hit > that require the _same_ magic setup inside but have _different_ number > of lanes. And even if, we can deal with it using a different compatible > string. > > Currently, I feel a single PHY provider node and a set of compatibles > will be most likely, i.e. no per-lane sub-nodes. OTOH, the per-lane > sub-nodes is more generic as it allows us to deal PHYs that may > suddenly skip one lane in the numbering scheme. The difference for the > driver is marginal, i.e. some SoC-specific struct with a field for the > number of lanes vs. of_count_child_nodes() and a reg = <n> property for > the per-lane sub-nodes. > > I used to agree to "DT should descibe HW", but with no datasheet > available, it quickly becomes fuzzy what it really looks like. > > Anyway, I'll have some discussion with Antoine and Alexandre to sort out > the differences and the things in common for the PHY and SoCs in > question. cool, thanks. -Kishon -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html