[PATCH 2/4] of: DT quirks infrastructure

[Pantelis Antoniou]

Hi Peter,

> On Feb 20, 2015, at 18:34 , Peter Hurley <peter at hurleysoftware.com> wrote:
> 
> On 02/20/2015 10:38 AM, Pantelis Antoniou wrote:
>> Hi Peter,
>> 
>>> On Feb 20, 2015, at 17:24 , Peter Hurley <peter at hurleysoftware.com> wrote:
>>> 
>>> On 02/20/2015 10:02 AM, Pantelis Antoniou wrote:
>>>> Hi Peter,
>>>> 
>>>>> On Feb 20, 2015, at 17:00 , Peter Hurley <peter at hurleysoftware.com> wrote:
>>>>> 
>>>>> On 02/20/2015 09:35 AM, Ludovic Desroches wrote:
> 
> [...]
> 
>>>>>> As you said, we can imagine many reasons to have a failure during the
>>>>>> production, having several DTB files will increase the risk.
>>>>> 
>>>>> It's interesting that you don't see the added complexity of open-coding
>>>>> the i2c driver or mixing DTS fragments for different designs as increased risk
>>>>> (for us all).
>>>>> 
>>>>> 
>>>> 
>>>> You don’t have to use it.
>>> 
>>>> diff --git a/arch/arm/mach-omap2/Makefile b/arch/arm/mach-omap2/Makefile
>>>> index 5d27dfd..02129e7 100644
>>>> --- a/arch/arm/mach-omap2/Makefile
>>>> +++ b/arch/arm/mach-omap2/Makefile
>>>> @@ -259,6 +259,11 @@ obj-$(CONFIG_MACH_CRANEBOARD)		+= board-am3517crane.o
>>>> 
>>>> obj-$(CONFIG_MACH_SBC3530)		+= board-omap3stalker.o
>>>> 
>>>> +# DT quirks
>>>> +ifneq ($(CONFIG_OF_DYNAMIC),)
>>>> +obj-$(CONFIG_SOC_AM33XX)		+= am33xx-dt-quirks.o
>>>> +endif
>>> 
>>> Won't this automatically be included on my Black that supports DT overlays?
>>> 
>> 
>> Yes it will. It is a grand total of 498 lines of code, and the total size of
>> the code segment is about 2.2K.
>> 
>> You do realize that you’re probably booting a multi-platform kernel on the 
>> black right? Including things for all 2xxx/3xxx and 44xx platforms?
>> For instance:
>> 
>>> ~/ti/kernels/linux-github.git $ wc -l arch/arm/mach-omap2/*44xx*.c
>>>   443 arch/arm/mach-omap2/clockdomains44xx_data.c
>>>   526 arch/arm/mach-omap2/cminst44xx.c
>>>   251 arch/arm/mach-omap2/cpuidle44xx.c
>>>   250 arch/arm/mach-omap2/dpll44xx.c
>>>  4864 arch/arm/mach-omap2/omap_hwmod_44xx_data.c
>>>   295 arch/arm/mach-omap2/pm44xx.c
>>>   358 arch/arm/mach-omap2/powerdomains44xx_data.c
>>>    62 arch/arm/mach-omap2/prcm_mpu44xx.c
>>>   770 arch/arm/mach-omap2/prm44xx.c
>>>   210 arch/arm/mach-omap2/prminst44xx.c
>>>   117 arch/arm/mach-omap2/vc44xx_data.c
>>>   130 arch/arm/mach-omap2/voltagedomains44xx_data.c
>>>   104 arch/arm/mach-omap2/vp44xx_data.c
>>>  8380 total
>> 
>> I bet those things are far larger than 2.2K. And I also bet that in the
>> tradeoff analysis that the board maintainer did things came down to 
>> increasing complexity so that he can consolidate the kernels for all the
>> other platforms he has to support besides the black.
> 
> Not that it really matters, but I'm not using any of that.
> 
> 
>>>> Some people really do though. As for increased risk
>>>> I expect to see arguments instead of a statement.
>>> 
>>> No one is wasting your time with random arguments. Please keep your tone civil.
>>> 
>> 
>> A statement like 'increasing risk for all of us' is very open ended. What is
>> the risk? How much of it exists?
> 
> My point was simply that this trades reduced complexity in one area
> with increased complexity in another area.
> 
> For you, that trade-off is worth it, but for others, not so much.
> 

Of course and that’s the point. No-one is going to remove the vanilla
DTSs for the beaglebone black. If you don’t want to use this capability
there won’t be any impact.

The EEPROM probe is only triggered by the presence of the quirk node,
so if you don’t boot with a DTB that contains it there is no change
in the boot process.

> FWIW, I agree that some mechanism is required to support the other
> use cases. I just don't think ease of manufacturing, when the
> submit configuration is the BeagleBone, is where I would hang my hat.
> 

The beaglebone is just an open source friendly platform that I can
demonstrate the capability.

> 
>> If I offended you I’m really sorry though, I meant no such thing.
> 
> In re-reading it, I realize I shouldn't have taken offense. Thanks anyway
> for the apology.
> 
> Regards,
> Peter Hurley

Regards

— Pantelis