[RFC PATCH 3/5] powerpc/syscalls: Split SPU-ness out of ABI

[Arnd Bergmann]

On Mon, Jan 14, 2019 at 12:36 PM Michael Ellerman <mpe at ellerman.id.au> wrote:
>
> Using the ABI field to encode whether a syscall is usable by SPU
> programs or not is a bit of kludge.
>
> The ABI of the syscall doesn't change depending on the SPU-ness, but
> in order to make the syscall generation work we have to pretend that
> it does.
>
> It also means we have more duplicated syscall lines than we need to,
> and the SPU logic is not well contained, instead all of the syscall
> generation targets need to know if they are spu or nospu.
>
> So instead add another column to the table which is a boolean SPU
> field. If the field has the value "spu" then the syscall is available
> to SPU programs, any other value means the syscall is not available to
> SPu programs.

This change unfortunately makes the format of the file incompatible
with the other architectures, which prevents us from adding common
scripts.

There are obviously the existing scripts that we want to generalize
across architectures, but also further ones, like a suggestion by
David Miller to check that each native 32-bit syscall comes with
a matching compat syscall.

       Arnd