Re: [PATCH 06/10] rust: xarray: simplify `Guard::load`

["Gary Guo" <gary@garyguo.net>]

On Thu Jan 8, 2026 at 11:07 AM GMT, Tamir Duberstein wrote:
> On Thu, Jan 8, 2026 at 4:38 AM Andreas Hindborg <a.hindborg@kernel.org> wrote:
>>
>> Tamir Duberstein <tamird@gmail.com> writes:
>>
>> > On Wed, Jan 7, 2026 at 2:37 PM Andreas Hindborg <a.hindborg@kernel.org> wrote:
>> >>
>> >> Tamir Duberstein <tamird@gmail.com> writes:
>> >>
>> >> > On Wed, Dec 3, 2025 at 5:27 PM Andreas Hindborg <a.hindborg@kernel.org> wrote:
>> >> >>
>> >> >> Simplify the implementation by removing the closure-based API from
>> >> >> `Guard::load` in favor of returning `Option<NonNull<c_void>>` directly.
>> >> >
>> >> > This is not sound. The returned pointer can now outlive the guard and
>> >> > mutation through that pointer is trivial.
>> >>
>> >> I don't think this is unsound. If we returned a reference instead, it
>> >> would be, but we are returning a raw pointer. Dereferencing the pointer
>> >> is unsafe and requires proper safety comments.
>> >
>> > You may be right, strictly speaking, but it is most definitely a
>> > footgun. This is a special pointer that requires more careful handling
>> > than other raw pointers.
>>
>> I would disagree. Dereferencing any raw pointer requires the same
>> checks, and knowing this one is valid and satisfies lifetime
>> requirements is no different than others.
>>
>> It is also a private method that is only used in this particular impl
>> block.
>>
>> At any rate, I'm fine with dropping the change. The reason I did it was
>> because I had to double take when I read the previous code. I think the
>> original code is has some unnecessary complexity with the closure.
>
> Simplification is always welcome. The current shape of this code was
> guided by the desire to avoid the footgun we're discussing here.

Using a closure isn't any better than just returning a raw pointer. If you want
a pointer you can just have a identity closure and there's nothing preventing
you from doing that.

Even with `T::borrow` inside the closure this isn't providing any security as
the code would still compile if you change `'_` in the return signature into
`'static`.

The only defence here is really the `unsafe` and safety comment.

I would prefer to not have closure in this case.

Best,
Gary