Re: [RFC][PATCH 11/12] slub: Replace cmpxchg_double()

[Peter Zijlstra <peterz@infradead.org>]

On Mon, Jan 09, 2023 at 04:02:33PM -0600, Linus Torvalds wrote:
> On Mon, Jan 9, 2023 at 10:29 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > I ran into a ton of casting trouble when compiling kernel/fork.c which
> > uses this_cpu_cmpxchg() on a pointer type and the compiler hates casting
> > pointers to an integer that is not the exact same size.
> 
> Ahh. Yeah - not because that code needs or wants the 128-bit case, but
> because the macro expands to all sizes in a switch statement, so the
> compiler sees all the cases even if only one is then statically
> picked.
> 
> So the silly casts are for all the cases that never matter.
> 
> Annoying.

Yes, very.

This seems to compile (and boot). Let me go update the others and push
it out for the robots to have a go.

#define percpu_cmpxchg128_op(size, qual, _var, _oval, _nval)		\
({									\
	union {								\
		typeof(_var) full;					\
		struct {						\
			u64 low, high;					\
		};							\
	} old__, new__;							\
									\
	old__.full = _oval;						\
	new__.full = _nval;						\
									\
	asm qual ("cmpxchg16b " __percpu_arg([var])			\
		  : [var] "+m" (_var),					\
		    "+a" (old__.low),					\
		    "+d" (old__.high)					\
		  : "b" (new__.low),					\
		    "c" (new__.high)					\
		  : "memory");						\
									\
	old__.full;							\
})