[PATCH 1/1] ARM: Exynos: Add generic compatible string

Tomasz Figa tomasz.figa at gmail.com
Wed Mar 5 07:12:52 EST 2014


On 05.03.2014 09:25, Sachin Kamat wrote:
> On 25 February 2014 17:12, Arnd Bergmann <arnd at arndb.de> wrote:
>> On Tuesday 25 February 2014, Olof Johansson wrote:
>>> I disagree. I don't know what Samsung has in mind, but the revision of
>>> the CPU doesn't have all that much to do with the rest of the SoC.
>>> It's quite likely that some vendors (maybe not Samsung, but the same
>>> concept applies) will ship 64-bit SoCs that are very similar to their
>>> preceding 32-bit ones, same IP, similar busses, etc. I'm pretty sure
>>> at least some vendors will do very close to that.
>>
>> Right.
>>
>>> So, if EXYNOS4 and EXYNOS5 can share a compatible value when they use
>>> different CPUs, then there's no reason that whatever future 64-bit
>>> ones can also share it.
>>
>> How about putting both 'samsung,exynos' and 'samsung,exynos4' in DT then
>> and having the platform code match exynos4 and exynos5 but not exynos?
>>
>> That way, I think we are consistent and future-proof. Any code that needs
>> to know if it's running on some exynos version can just check for the
>> 'samsung,exynos' compatible value and that will work on both arm32 and
>> arm64. Also, if we ever decide we want to run a 32-bit kernel on a 64-bit
>> exynos, we can just add 'samsung,exynos6' (or whatever number that will
>> be) to the list.
>>
>> My usual disclaimer for this: You should never ever consider actually
>> running a 32-bit kernel on a 64-bit CPU, but at the same time there
>> shouldn't be any reason why it won't work either, given that we require
>> arm64 based systems to have all SoC specific code in drivers and we
>> can use the same drivers on arm32.
>
> Kukjin, Tomasz,
>
> What is your opinion about Arnd's suggestion?
>

I would still prefer introducing a generic string for 32-bit Exynos 
SoCs, but I don't think it really matters a lot. I guess we can stick to 
just exynos4 and exynos5 compatible strings then, as long as we can 
merge the "board"-files and common.c together, since the code is pretty 
much SoC-independent now.

Best regards,
Tomasz



More information about the linux-arm-kernel mailing list