[PATCHv5 27/35] ARM: dts: am33xx: add minimal l4 bus layout with control module support

[Suman Anna]

On 03/20/2015 05:35 PM, Tony Lindgren wrote:
> * Suman Anna <s-anna at ti.com> [150320 14:44]:
>> On 03/20/2015 01:44 PM, Kristo, Tero wrote:
>>> +			scm: scm at 210000 {
>>> +				compatible = "ti,am3-scm", "simple-bus";
>>> +				reg = <0x210000 0x2000>;
>>> +				#address-cells = <1>;
>>> +				#size-cells = <1>;
>>> +				ranges = <0 0x210000 0x2000>;
>>> +
>>> +				am33xx_pinmux: pinmux at 800 {
>>> +					compatible = "pinctrl-single";
>>> +					reg = <0x800 0x238>;
>>> +					#address-cells = <1>;
>>> +					#size-cells = <0>;
>>> +					pinctrl-single,register-width = <32>;
>>> +					pinctrl-single,function-mask = <0x7f>;
>>> +				};
>>> +
>>> +				scm_conf: scm_conf at 0 {
>>> +					compatible = "syscon";
>>> +					reg = <0x0 0x7fc>;
>>
>> Hmm, you are consolidating the am33xx_control_module and cm nodes, so is
>> this supposed to be 0x800 or 0x7fc? I would think it should be 0x800.
> 
> Seems correct to me, it's offset 0, size 0x7fc. So that's the scm_conf
> syscon area before pinctrl-single at 0x44c00000 + 0x210000 + 0.
> 
> The io area for pinctrl-single starts at 0x800, so the scm_conf should
> be before it in the dts file.

Well, I understand that it is how it was before, but we won't be mapping
or covering the last register efuse_sma before the pinctrl cfg
registers. Any reason for just leaving out that register?

regards
Suman

> 
>> Also, are we ordering the child nodes of scm by node names or addresses.
>> I have to add the wkup_m3 node, and prefer ordering by addresses.
> 
> Yeah address ordering makes most sense here IMO.
> 
> Note that you should follow the TRM "Table 2-2. L4_WKUP Peripheral Memory
> Map" and set up things as separate devices as shown there. Pretty much
> each row in that table is a separate device on the interconnect. That's
> especially true if the device has registers like revision, sysc, syss
> and so on. In that case they can be clocked and idled separately.
> 
> So with these changes we follow the hardware mapping, although only
> partially have it populated now for l4_wkup:
> 
> l3 (ocp) +-> l4_per  +-> ...
>          |           |-> ...
>          |
>          +-> l4_wkup +-> prcm
>          |           |-> scm 
> 	 |           |-> ...
>          |
>          +-> ...     +-> ...
> 
> 
> Regards,
> 
> Tony
>