Re: CPU_BIG_ENDIAN in generic code (was: Re: [PATCH v3 3/7] arch/sparc: Define config parameter CPU_BIG_ENDIAN)

[Arnd Bergmann]

On Wed, May 24, 2017 at 4:45 PM, Babu Moger <babu.moger@xxxxxxxxxx> wrote:
> On 5/24/2017 5:18 AM, Arnd Bergmann wrote:
>> On Wed, May 24, 2017 at 11:59 AM, Geert Uytterhoeven
>> <geert@xxxxxxxxxxxxxx> wrote:
>>> On Tue, May 23, 2017 at 11:45 PM, Babu Moger <babu.moger@xxxxxxxxxx>
>>> wrote:
>>> include/asm-generic/qrwlock.h is also generic, but depends on the
>>> architecture to select ARCH_USE_QUEUED_RWLOCKS, which only very few do
>>> (x86, and now sparc).
>>>
>>> I guess the time is ripe for adding (both) symbols to all architectures?
>>
>> Good idea. I think we can do most of this by adding a few lines to
>> arch/Kconfig:
>>
>> config CPU_BIG_ENDIAN
>>          bool
>>
>> config CPU_LITTLE_ENDIAN
>>         def_bool !CPU_BIG_ENDIAN
>
> I noticed that even x86 does not define CPU_LITTLE_ENDIAN.  Strange.

There is no architecture-independent code that tests for
CONFIG_CPU_LITTLE_ENDIAN, unlike CONFIG_CPU_BIG_ENDIAN,
so that's not very suprising.

> With this code all the architecture will default to
> CONFIG_CPU_LITTLE_ENDIAN.

What I meant is that we have to 'select CPU_BIG_ENDIAN' on all architectures
that actually are big-endian:

These are all configurable:
$ git grep -l linux/byteorder/big_endian.h | xargs grep -l
linux/byteorder/little_endian.h
arch/arc/include/uapi/asm/byteorder.h
arch/arm/include/uapi/asm/byteorder.h
arch/arm64/include/uapi/asm/byteorder.h
arch/c6x/include/uapi/asm/byteorder.h
arch/m32r/include/uapi/asm/byteorder.h
arch/microblaze/include/uapi/asm/byteorder.h
arch/mips/include/uapi/asm/byteorder.h
arch/powerpc/include/uapi/asm/byteorder.h
arch/sh/include/uapi/asm/byteorder.h
arch/tile/include/uapi/asm/byteorder.h

These are always big-endian:
$ git grep -l linux/byteorder/big_endian.h | xargs grep -L
linux/byteorder/little_endian.h
arch/avr32/include/uapi/asm/byteorder.h
arch/frv/include/uapi/asm/byteorder.h
arch/m68k/include/uapi/asm/byteorder.h
arch/openrisc/include/uapi/asm/byteorder.h
arch/parisc/include/uapi/asm/byteorder.h
arch/s390/include/uapi/asm/byteorder.h
arch/sparc/include/uapi/asm/byteorder.h

And these are always little-endian:
arch/alpha/include/uapi/asm/byteorder.h
arch/blackfin/include/uapi/asm/byteorder.h
arch/cris/include/uapi/asm/byteorder.h
arch/hexagon/include/uapi/asm/byteorder.h
arch/ia64/include/uapi/asm/byteorder.h
arch/metag/include/uapi/asm/byteorder.h
arch/mn10300/include/uapi/asm/byteorder.h
arch/score/include/uapi/asm/byteorder.h
arch/unicore32/include/uapi/asm/byteorder.h
arch/x86/include/uapi/asm/byteorder.h

So if we 'select CPU_BIG_ENDIAN' from avr32, frv, m68k, openrisc, parisc,
s390 and sparc, this covers all the fixed-endian architectures, and the
other ones are those that already have either CPU_BIG_ENDIAN
as a 'bool' option, or both as a 'choice'.

> I can make it as a separate patch. But I can only test SPARC and little bit
> of x86. Is that ok?

I think that's ok.

        Arnd