Re: [PATCH 1/2] compiler.h: Introduce ptr_eq() to preserve address dependency

[Mathieu Desnoyers]

On 2024-10-03 19:07, David Laight wrote:
> ...
> > What _does_ work however are the following two approaches:
> >
> > 1) Perform the equality check on the original variables, creating
> > new versions (with OPTIMIZER_HIDE_VAR) of both variables for the
> > rest of their use, therefore making sure the pointer dereference
> > are not derived from versions of the variables which were compared
> > with another pointer. (as suggested by Boqun)
>
> If that is
> 	a1 = a; OPTIMISER_HIDE_VAR(a1);
> 	b1 = b; OPTIMISER_HIDE_BAR(b1);
> 	if (a != b}
> 		return;
> 	// code using a1 and b1
> then can't the compiler first flip it to:
> 	if (a != b)
> 		return;
> 	a1 = a; OPTIMISER_HIDE_VAR(a1);
> 	b1 = b; OPTIMISER_HIDE_VAR(b1);
> and then replace the last line with:
> 	b1 = a; OPTIMISER_HIDE_VAR(b1);
> which isn't intended at all.

Good point, so I suspect Boqun's ADDRESS_EQ() suggestion did not work:

https://lore.kernel.org/lkml/ZvX12_1mK8983cXm@boqun-archlinux/
> 		
> 	
> OTOH if you do:
> 	a1 = a; OPTIMISER_HIDE_VAR(a1);
> 	b1 = b; OPTIMISER_HIDE_VAR(b1);
> 	if (a1 != b1)
> 		return;
> 	// code using a and b
> (which I think is)

This is in line with Linus' suggestion, which is the approach I
retained.

> > 2) Perform the equality check on the versions resulting of hiding
> > both variables, making sure those versions of the variables are
> > not dereferenced afterwards. (as suggested by Linus)
>
> then the compiler can't possibly reverse the asm blocks.

Indeed.

Thanks,

Mathieu

> 	David
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)

--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com