On Tue, Jun 04, 2019 at 12:36:24PM +0200, Arnd Bergmann wrote:
> On Mon, Jun 3, 2019 at 4:44 PM Christian Brauner <christian@xxxxxxxxxx> wrote:
>
> > +
> > +#ifdef __ARCH_WANT_SYS_CLONE
> > +asmlinkage long sys_clone3(struct clone_args __user *uargs, size_t size);
> > +#endif
>
> I would leave it outside of __ARCH_WANT_SYS_CLONE, as far
> as I can tell the only reason for that #ifdef is so architectures that
> have their own sys_clone implementation can opt out of the generic
> one, but we don't want that for new syscalls.
>
> In fact, I'd prefer to drop the symbol entirely and have a different
> symbol with the opposite meaning such as
> __ARCH_NONSTANDARD_SYS_CLONE that only gets
> selected by sparc, ia64 and m68k. That should be a separate
> patch though, and I'm not asking you to do it, unless you
> want to clean up a little more.
I am totally up for this but I would prefer if we land clone3() in the
5.3 merge window and then for 5.3 rc{2,3} do the cleanups that David and
you suggested.
This leaves this patchset lean and easy to review.
Christian
