[PATCH v2 8/8] ARM: DRA7: dts: Add the dts files for dra7 SoC and dra7-evm board

[Benoit Cousson]

Hi Rajendra,

On 30/07/2013 15:01, Rajendra Nayak wrote:
> On Tuesday 30 July 2013 06:29 PM, Nishanth Menon wrote:
>> On 07/30/2013 07:56 AM, Rajendra Nayak wrote:
>>> On Tuesday 30 July 2013 06:16 PM, Nishanth Menon wrote:
>>>> On 07/30/2013 07:41 AM, Rajendra Nayak wrote:
>>>>> On Tuesday 30 July 2013 06:00 PM, Nishanth Menon wrote:
>>>>>> On 07/30/2013 06:25 AM, Rajendra Nayak wrote:
>>>>>>> From: R Sricharan <r.sricharan at ti.com>
>>>> [...]
>>>>>>> +        mcspi4: spi at 480ba000 {
>>>>>>> +            compatible = "ti,omap4-mcspi";
>>>>>>> +            reg = <0x480ba000 0x200>;
>>>>>>> +            interrupts = <0 48 0x4>;
>>>>>>> +            #address-cells = <1>;
>>>>>>> +            #size-cells = <0>;
>>>>>>> +            ti,hwmods = "mcspi4";
>>>>>>> +            ti,spi-num-cs = <1>;
>>>>>>> +            dmas = <&sdma 70>, <&sdma 71>;
>>>>>>> +            dma-names = "tx0", "rx0";
>>>>>>> +        };
>>>>>>> +    };
>>>>>>> +};
>>>>>>>
>>>>>> ref: [1], we discussed that we should now be able to introduce all instances of h/w blocks into the dra7.dts. Further, considering [2]
>>>>> hmm, thats a long discussion on crossbar driver that [1] points to. Do you want to summarize what you mean by 'introduce all instances of h/w blocks'
>>>> I recommend reading the last few emails on the thread about how we could do this with pinctrl. unfortunately, this patch is not informative enough to indicate that not all instances of the potential IP blocks are listed here.
>>>>
>>>>>> would you not want to follow "status = disabled" for all modules by default and enable required modules in board file, so that we dont have to respin this yet again?
>>>>> Well, I was just following the convention of whats already followed on existing OMAPs. See [3] for some views on these.
>>>> DRA7 case, I would not think it makes sense due to the number of product variants being done, not all will use the same set. Further, rationale for DRA7 and my suggestion for Grant's option (1) is mainly because the product variants will require more dtsis rather than board files using the product variants use just the necessary modules from a common dtsi. Makes support of variants like OMAP57xx etc trivial and constrainted to board file usage, rather than spinning off new dtsis.
>>> Makes sense with the different product variants for DRA7, AM335x already does it this way, but the rest of OMAP3/4/5 are doing it the other way.
>>> I think its just too confusing to follow different conventions for different SoCs. We should stick to just one, either this way or that.
>>>
>> I think bucketing DRA7(with multitude of SoC variants) with OMAP family(usually with <5 variants) will be a wrong approach. we should choose the approach appropriate for the SoC. hence, OMAPx having all default enabled makes sense (as the delta is usually trivial), but on DRA7, the variants are larger :(
>>
>> just my 2 cents.
> I can respin with the changes, but before I do so, Benoit do you agree with the rationale for these and fine with the approach?
>
>

Sorry for the very late reply. I've just seen it because Mark's answer 
put in again in my gmail box.

I'm not sure to understand what Nishanth and you are arguing for :-)

If this is about the default status state, I think that this flag is 
anyway confusing because you cannot know if we are considering an IP 
that does not exist at all in a variant or an IP that is not used on a 
certain board...

This potential issue we can have with the second case is that the IP 
might be enabled by the boot loader and never disabled because 
status="disabled" prevent DT to create the device, and thus prevent the 
driver to properly idle the device.

That's why I did not like the usage of status="disabled" by default, 
because the way DT core handle that might not be appropriate for an IP 
that is there but not use.

For a variant that does not contain physically the IP, that flag can be 
relevant.
Regarding which default status is the best, I guess as soon as it is 
consistant across a chip family, both are fine to me.

Not necessarily related to your discussion, but I still think that DT is 
missing a flag to make the difference between the two cases.

Regards,
Benoit