Re: [RFC PATCH 1/4] hazptr: Add initial implementation of hazard pointers

[Linus Torvalds <torvalds@linux-foundation.org>]

On Fri, 27 Sept 2024 at 12:12, Jonas Oberhauser
<jonas.oberhauser@huaweicloud.com> wrote:
>
> I think it depends on which one you hide.

No.

Dammit, people, read the code I posted.

> But for
>
>   z = a;
>   hide(z);
>   if (z==b) { *b; }

No.

I *intentionally* made it an inline function, and only hid the
arguments to the equality comparison.

So the "hide(z)" hides the argument to the inline function - NOT THE ORIGINAL.

> then it would still know that b == z, and could replace *b with *z
> (which really is *a).

No.

The hiding is literally *ONLY* for the comparison. It's inside the
helper function. It doesn't affect the originals at all.

Which means that the compiler CANNOT KNOW anything about the original
pointers when it compares for equality (or inequality).

Basically, the comparison is now a black box to the compiler, and the
compiler cannot use the result of the comparison to make ANY judgment
on whether the two original pointers were related or not.

               Linus