Re: [PATCH v3 1/2] rust: irq: add support for request_irq()

["Benno Lossin" <lossin@kernel.org>]

On Wed Jun 4, 2025 at 9:48 AM CEST, Alice Ryhl wrote:
> On Wed, Jun 4, 2025 at 9:37 AM Benno Lossin <lossin@kernel.org> wrote:
>> On Mon Jun 2, 2025 at 5:20 PM CEST, Alice Ryhl wrote:
>> > On Wed, May 14, 2025 at 10:04:43PM +0200, Benno Lossin wrote:
>> >> On Wed May 14, 2025 at 9:20 PM CEST, Daniel Almeida wrote:
>> >> > +                )
>> >> > +            });
>> >> > +
>> >> > +            if res.is_err() {
>> >> > +                // SAFETY: We are returning an error, so we can destroy the slot.
>> >> > +                unsafe { core::ptr::drop_in_place(addr_of_mut!((*slot).handler)) };
>> >> > +            }
>> >> > +
>> >> > +            res
>> >> > +        };
>> >> > +
>> >> > +        // SAFETY:
>> >> > +        // - if this returns Ok, then every field of `slot` is fully
>> >> > +        // initialized.
>> >> > +        // - if this returns an error, then the slot does not need to remain
>> >> > +        // valid.
>> >> > +        unsafe { pin_init_from_closure(closure) }
>> >>
>> >> Please don't use `pin_init_from_closure`, instead do this:
>> >>
>> >>     pin_init!(Self {
>> >>         irq,
>> >>         handler,
>> >>         _pin: PhantomPinned
>> >>     })
>> >>     .pin_chain(|this| {
>> >>         // SAFETY: TODO: correct FFI safety requirements
>> >>         to_result(unsafe {
>> >>             bindings::request_irq(...)
>> >>         })
>> >>     })
>> >>
>> >> The `pin_chain` function is exactly for this use-case, doing some
>> >> operation that might fail after initializing & it will drop the value
>> >> when the closure fails.
>> >
>> > No, that doesn't work. Using pin_chain will call free_irq if the call to
>> > request_irq fails, which is incorrect.
>>
>> Good catch. That's a bit annoying then... I wonder if there is a
>> primitive missing in pin-init that could help with this... Any ideas?
>
> I believe initializers for underscore fields would do it. We could
> potentially abuse the _pin field, but frankly I think that's too
> confusing to the reader.

Oh yeah that's the feature we want here, will add it to my list for the
next cycle :)

---
Cheers,
Benno