Re: [RFC v1 10/16] RISC-V: The ABI implementation for the 32-bit

[Adhemerval Zanella <adhemerval dot zanella at linaro dot org>]

On 24/06/2019 06:51, Arnd Bergmann wrote:
> On Mon, Jun 24, 2019 at 4:49 AM Palmer Dabbelt <palmer@sifive.com> wrote:
>>
>> On Sun, 23 Jun 2019 12:41:20 PDT (-0700), Arnd Bergmann wrote:
>>> On Sat, Jun 22, 2019 at 6:47 AM Alistair Francis
>>> <alistair.francis@wdc.com> wrote:
>>>
>>>>  sysdeps/unix/sysv/linux/riscv/rv32/lockf64.c  | 70 +++++++++++++++++++
>>>
>>> Why is this file architecture specific? It seems everything else just uses
>>> #include <sysdeps/unix/sysv/linux/i386/lockf64.c>
>>> Is your code different from that version?
>>
>> As part of the code review it was indicated that we shouldn't #include C files
>> but should instead duplicate them.  At the time it was exactly the same, we
>> should scrutinize any differences as that was months ago.
> 
> I would expect that you would want it to be consistent across architectures
> as well: so either replace all the other #include versions with direct copies,
> or move the file into a more generic location that avoids the #include if you
> can't add another one any more.
> 
>       Arnd
> 

Since e442e40de5646e93bf31ace3e0c5159085a7259b glibc lock{64} is based on
fcntl{64}.  And sysdeps/unix/sysv/linux/fcntl{64}.c should already call
__NR_fcntl64 for the required cases.  So why do you need to add an arch
specific lock64 implementation here?