Hi !

Any objection killing that ? Archs that use that in their _syscall macro
can either fix the macro or have their own "int errno" defined in
arch/something.

We just found a nasty bug in the ppc64 code that was filling that value
on in-kernel syscalls (asm shit was writing a 64 bits value to the
32 bits errno).

IMHO, Anything relying on the value of that global errno is broken,
so what about just removing it completely ?

Ben.