Re: [PATCH v9 3/7] rust: irq: add support for non-threaded IRQs and handlers

[Daniel Almeida <daniel.almeida@collabora.com>]

Hi Andreas,

> On 18 Aug 2025, at 05:14, Andreas Hindborg <a.hindborg@kernel.org> wrote:
> 
> "Daniel Almeida" <daniel.almeida@collabora.com> writes:
> 
>> This patch adds support for non-threaded IRQs and handlers through
>> irq::Registration and the irq::Handler trait.
>> 
>> Registering an irq is dependent upon having a IrqRequest that was
>> previously allocated by a given device. This will be introduced in
>> subsequent patches.
>> 
>> Tested-by: Joel Fernandes <joelagnelf@nvidia.com>
>> Tested-by: Dirk Behme <dirk.behme@de.bosch.com>
>> Reviewed-by: Alice Ryhl <aliceryhl@google.com>
>> Signed-off-by: Daniel Almeida <daniel.almeida@collabora.com>
>> ---
>> rust/bindings/bindings_helper.h |   1 +
>> rust/helpers/helpers.c          |   1 +
>> rust/helpers/irq.c              |   9 ++
>> rust/kernel/irq.rs              |   5 +
>> rust/kernel/irq/request.rs      | 264 ++++++++++++++++++++++++++++++++++++++++
>> 5 files changed, 280 insertions(+)
>> 
>> diff --git a/rust/bindings/bindings_helper.h b/rust/bindings/bindings_helper.h
>> index 84d60635e8a9baef1f1a1b2752dc0fa044f8542f..69a975da829f0c35760f71a1b32b8fcb12c8a8dc 100644
>> --- a/rust/bindings/bindings_helper.h
>> +++ b/rust/bindings/bindings_helper.h
>> @@ -52,6 +52,7 @@
>> #include <linux/ethtool.h>
>> #include <linux/file.h>
>> #include <linux/firmware.h>
>> +#include <linux/interrupt.h>
>> #include <linux/fs.h>
>> #include <linux/ioport.h>
>> #include <linux/jiffies.h>
>> diff --git a/rust/helpers/helpers.c b/rust/helpers/helpers.c
>> index 7cf7fe95e41dd51717050648d6160bebebdf4b26..44b2005d50140d34a44ae37d01c2ddbae6aeaa32 100644
>> --- a/rust/helpers/helpers.c
>> +++ b/rust/helpers/helpers.c
>> @@ -22,6 +22,7 @@
>> #include "dma.c"
>> #include "drm.c"
>> #include "err.c"
>> +#include "irq.c"
>> #include "fs.c"
>> #include "io.c"
>> #include "jump_label.c"
>> diff --git a/rust/helpers/irq.c b/rust/helpers/irq.c
>> new file mode 100644
>> index 0000000000000000000000000000000000000000..1faca428e2c047a656dec3171855c1508d67e60b
>> --- /dev/null
>> +++ b/rust/helpers/irq.c
>> @@ -0,0 +1,9 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +
>> +#include <linux/interrupt.h>
>> +
>> +int rust_helper_request_irq(unsigned int irq, irq_handler_t handler,
>> +     unsigned long flags, const char *name, void *dev)
>> +{
>> + return request_irq(irq, handler, flags, name, dev);
>> +}
>> diff --git a/rust/kernel/irq.rs b/rust/kernel/irq.rs
>> index 068df2fea31de51115c30344f7ebdb4da4ad86cc..c1019bc36ad1e7ae7dd3af8a8b5c14780bf70712 100644
>> --- a/rust/kernel/irq.rs
>> +++ b/rust/kernel/irq.rs
>> @@ -13,4 +13,9 @@
>> /// Flags to be used when registering IRQ handlers.
>> mod flags;
>> 
>> +/// IRQ allocation and handling.
>> +mod request;
>> +
>> pub use flags::Flags;
>> +
>> +pub use request::{Handler, IrqRequest, IrqReturn, Registration};
>> diff --git a/rust/kernel/irq/request.rs b/rust/kernel/irq/request.rs
>> new file mode 100644
>> index 0000000000000000000000000000000000000000..57e00ebf694d8e6e870d9ed57af7ee2ecf86ec05
>> --- /dev/null
>> +++ b/rust/kernel/irq/request.rs
>> @@ -0,0 +1,264 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +// SPDX-FileCopyrightText: Copyright 2025 Collabora ltd.
>> +
>> +//! This module provides types like [`Registration`] which allow users to
>> +//! register handlers for a given IRQ line.
>> +
>> +use core::marker::PhantomPinned;
>> +
>> +use crate::alloc::Allocator;
>> +use crate::device::{Bound, Device};
> 
> nit: I would suggest either normalizing all the imports, or using one
> import per line consistently.
> 
>> +use crate::devres::Devres;
>> +use crate::error::to_result;
>> +use crate::irq::flags::Flags;
>> +use crate::prelude::*;
>> +use crate::str::CStr;
>> +use crate::sync::Arc;
>> +
>> +/// The value that can be returned from a [`Handler`] or a [`ThreadedHandler`].
> 
> error: unresolved link to `ThreadedHandler`
>  --> /home/aeh/src/linux-rust/request-irq/rust/kernel/irq/request.rs:18:62
>   |
> 18 | /// The value that can be returned from a [`Handler`] or a [`ThreadedHandler`].
>   |                                                              ^^^^^^^^^^^^^^^ no item named `ThreadedHandler` in scope
>   |

This was introduced by the next commit. I wonder what is the right thing to do
here now?

> 
>> +#[repr(u32)]
>> +pub enum IrqReturn {
>> +    /// The interrupt was not from this device or was not handled.
>> +    None = bindings::irqreturn_IRQ_NONE,
>> +
>> +    /// The interrupt was handled by this device.
>> +    Handled = bindings::irqreturn_IRQ_HANDLED,
>> +}
>> +
>> +/// Callbacks for an IRQ handler.
>> +pub trait Handler: Sync {
>> +    /// The hard IRQ handler.
> 
> Could you do a vocabulary somewhere? What does it mean that the handler
> is hard?


This nomenclature is borrowed from the C part of the kernel. The hard part is
what runs immediately in interrupt context while the bottom half runs later. In
this case, the bottom half is a threaded handler, i.e.: code running in a
separate kthread.

I think this is immediately understandable most of the time because it's a term
that is often used in the kernel. Do you still feel that I should expand the
docs in this case?

> 
>> +    ///
>> +    /// This is executed in interrupt context, hence all corresponding
>> +    /// limitations do apply.
>> +    ///
>> +    /// All work that does not necessarily need to be executed from
>> +    /// interrupt context, should be deferred to a threaded handler.
>> +    /// See also [`ThreadedRegistration`].
> 
> error: unresolved link to `ThreadedRegistration`
>  --> /home/aeh/src/linux-rust/request-irq/rust/kernel/irq/request.rs:37:20
>   |
> 37 |     /// See also [`ThreadedRegistration`].
>   |                    ^^^^^^^^^^^^^^^^^^^^ no item named `ThreadedRegistration` in scope
>   |
> 

Same as the previous doc issue you highlighted.

>> +    fn handle(&self) -> IrqReturn;
>> +}
>> +
>> +impl<T: ?Sized + Handler + Send> Handler for Arc<T> {
>> +    fn handle(&self) -> IrqReturn {
>> +        T::handle(self)
>> +    }
>> +}
>> +
>> +impl<T: ?Sized + Handler, A: Allocator> Handler for Box<T, A> {
>> +    fn handle(&self) -> IrqReturn {
>> +        T::handle(self)
>> +    }
>> +}
>> +
>> +/// # Invariants
>> +///
>> +/// - `self.irq` is the same as the one passed to `request_{threaded}_irq`.
>> +/// - `cookie` was passed to `request_{threaded}_irq` as the cookie. It is guaranteed to be unique
>> +///   by the type system, since each call to `new` will return a different instance of
>> +///   `Registration`.
> 
> This seems like a mix of invariant declaration and conformance. I don't
> think the following belongs here:
> 
>  It is guaranteed to be unique
>  by the type system, since each call to `new` will return a different instance of
>  `Registration`.
> 
> You could replace it with a uniqueness requirement.

WDYM? This invariant is indeed provided by the type system and we do rely on it
to make the abstraction work.

> 
>> +#[pin_data(PinnedDrop)]
>> +struct RegistrationInner {
>> +    irq: u32,
>> +    cookie: *mut c_void,
>> +}
>> +
>> +impl RegistrationInner {
>> +    fn synchronize(&self) {
>> +        // SAFETY: safe as per the invariants of `RegistrationInner`
>> +        unsafe { bindings::synchronize_irq(self.irq) };
>> +    }
>> +}
>> +
>> +#[pinned_drop]
>> +impl PinnedDrop for RegistrationInner {
>> +    fn drop(self: Pin<&mut Self>) {
>> +        // SAFETY:
>> +        //
>> +        // Safe as per the invariants of `RegistrationInner` and:
>> +        //
>> +        // - The containing struct is `!Unpin` and was initialized using
>> +        // pin-init, so it occupied the same memory location for the entirety of
>> +        // its lifetime.
>> +        //
>> +        // Notice that this will block until all handlers finish executing,
>> +        // i.e.: at no point will &self be invalid while the handler is running.
>> +        unsafe { bindings::free_irq(self.irq, self.cookie) };
>> +    }
>> +}
>> +
>> +// SAFETY: We only use `inner` on drop, which called at most once with no
>> +// concurrent access.
>> +unsafe impl Sync for RegistrationInner {}
>> +
>> +// SAFETY: It is safe to send `RegistrationInner` across threads.
> 
> Why?

It's a u32 and an opaque pointer. The pointer itself (which is what demands a
manual send/sync impl) is only used on drop() and there are no restrictions
that prevent freeing an irq from another thread.

The cookie points to the handler, i.e.:

+                    cookie: unsafe { &raw mut (*this.as_ptr()).handler }cast(),

Perhaps we can add a bound on Send for Handler if that helps?

> 
>> +unsafe impl Send for RegistrationInner {}
>> +
>> +/// A request for an IRQ line for a given device.
>> +///
>> +/// # Invariants
>> +///
>> +/// - `ìrq` is the number of an interrupt source of `dev`.
>> +/// - `irq` has not been registered yet.
>> +pub struct IrqRequest<'a> {
>> +    dev: &'a Device<Bound>,
>> +    irq: u32,
>> +}
>> +
>> +impl<'a> IrqRequest<'a> {
>> +    /// Creates a new IRQ request for the given device and IRQ number.
>> +    ///
>> +    /// # Safety
>> +    ///
>> +    /// - `irq` should be a valid IRQ number for `dev`.
>> +    pub(crate) unsafe fn new(dev: &'a Device<Bound>, irq: u32) -> Self {
> 
> This needs `#[expect(dead_code)]`.

Danilo did that already before applying.

> 
>> +        // INVARIANT: `irq` is a valid IRQ number for `dev`.
> 
> I would suggest rephrasing:
> 
>  By function safety requirement, irq` is a valid IRQ number for `dev`.

Ack, I can send a patch.

> 
>> +        IrqRequest { dev, irq }
>> +    }
>> +
>> +    /// Returns the IRQ number of an [`IrqRequest`].
>> +    pub fn irq(&self) -> u32 {
>> +        self.irq
>> +    }
>> +}
>> +
>> +/// A registration of an IRQ handler for a given IRQ line.
>> +///
>> +/// # Examples
>> +///
>> +/// The following is an example of using `Registration`. It uses a
>> +/// [`Completion`] to coordinate between the IRQ
>> +/// handler and process context. [`Completion`] uses interior mutability, so the
>> +/// handler can signal with [`Completion::complete_all()`] and the process
>> +/// context can wait with [`Completion::wait_for_completion()`] even though
>> +/// there is no way to get a mutable reference to the any of the fields in
>> +/// `Data`.
>> +///
>> +/// [`Completion`]: kernel::sync::Completion
>> +/// [`Completion::complete_all()`]: kernel::sync::Completion::complete_all
>> +/// [`Completion::wait_for_completion()`]: kernel::sync::Completion::wait_for_completion
>> +///
>> +/// ```
>> +/// use kernel::c_str;
>> +/// use kernel::device::Bound;
>> +/// use kernel::irq::{self, Flags, IrqRequest, IrqReturn, Registration};
>> +/// use kernel::prelude::*;
>> +/// use kernel::sync::{Arc, Completion};
>> +///
>> +/// // Data shared between process and IRQ context.
>> +/// #[pin_data]
>> +/// struct Data {
>> +///     #[pin]
>> +///     completion: Completion,
>> +/// }
>> +///
>> +/// impl irq::Handler for Data {
>> +///     // Executed in IRQ context.
>> +///     fn handle(&self) -> IrqReturn {
>> +///         self.completion.complete_all();
>> +///         IrqReturn::Handled
>> +///     }
>> +/// }
>> +///
>> +/// // Registers an IRQ handler for the given IrqRequest.
>> +/// //
>> +/// // This runs in process context and assumes `request` was previously acquired from a device.
>> +/// fn register_irq(
>> +///     handler: impl PinInit<Data, Error>,
>> +///     request: IrqRequest<'_>,
>> +/// ) -> Result<Arc<Registration<Data>>> {
>> +///     let registration = Registration::new(request, Flags::SHARED, c_str!("my_device"), handler);
>> +///
>> +///     let registration = Arc::pin_init(registration, GFP_KERNEL)?;
>> +///
>> +///     registration.handler().completion.wait_for_completion();
>> +///
>> +///     Ok(registration)
>> +/// }
>> +/// # Ok::<(), Error>(())
>> +/// ```
>> +///
>> +/// # Invariants
>> +///
>> +/// * We own an irq handler using `&self.handler` as its private data.
>> +#[pin_data]
>> +pub struct Registration<T: Handler + 'static> {
>> +    #[pin]
>> +    inner: Devres<RegistrationInner>,
> 
> Soundness of this API requires `inner` to be dropped before `handler`.
> Maybe we should have a comment specifying that the order of these fields
> is important?

Ack.

> 
>> +
>> +    #[pin]
>> +    handler: T,
>> +
>> +    /// Pinned because we need address stability so that we can pass a pointer
>> +    /// to the callback.
>> +    #[pin]
>> +    _pin: PhantomPinned,
>> +}
>> +
>> +impl<T: Handler + 'static> Registration<T> {
>> +    /// Registers the IRQ handler with the system for the given IRQ number.
> 
> Could this be "... for the IRQ number represented by `request`."? Because we
> don't pass in an actual number here.

Ack.

> 
>> +    pub fn new<'a>(
>> +        request: IrqRequest<'a>,
>> +        flags: Flags,
>> +        name: &'static CStr,
>> +        handler: impl PinInit<T, Error> + 'a,
>> +    ) -> impl PinInit<Self, Error> + 'a {
>> +        try_pin_init!(&this in Self {
>> +            handler <- handler,
>> +            inner <- Devres::new(
>> +                request.dev,
>> +                try_pin_init!(RegistrationInner {
>> +                    // SAFETY: `this` is a valid pointer to the `Registration` instance
>> +                    cookie: unsafe { &raw mut (*this.as_ptr()).handler }cast(),
>> +                    irq: {
>> +                        // SAFETY:
>> +                        // - The callbacks are valid for use with request_irq.
>> +                        // - If this succeeds, the slot is guaranteed to be valid until the
>> +                        //   destructor of Self runs, which will deregister the callbacks
>> +                        //   before the memory location becomes invalid.
>> +                        to_result(unsafe {
>> +                            bindings::request_irq(
>> +                                request.irq,
>> +                                Some(handle_irq_callback::<T>),
>> +                                flags.into_inner(),
>> +                                name.as_char_ptr(),
>> +                                (&raw mut (*this.as_ptr()).handler).cast(),
>> +                            )
>> +                        })?;
>> +                        request.irq
>> +                    }
>> +                })
>> +            ),
>> +            _pin: PhantomPinned,
>> +        })
>> +    }
>> +
>> +    /// Returns a reference to the handler that was registered with the system.
>> +    pub fn handler(&self) -> &T {
>> +        &self.handler
>> +    }
>> +
>> +    /// Wait for pending IRQ handlers on other CPUs.
>> +    ///
>> +    /// This will attempt to access the inner [`Devres`] container.
>> +    pub fn try_synchronize(&self) -> Result {
>> +        let inner = self.inner.try_access().ok_or(ENODEV)?;
>> +        inner.synchronize();
>> +        Ok(())
>> +    }
>> +
>> +    /// Wait for pending IRQ handlers on other CPUs.
>> +    pub fn synchronize(&self, dev: &Device<Bound>) -> Result {
>> +        let inner = self.inner.access(dev)?;
>> +        inner.synchronize();
>> +        Ok(())
>> +    }
>> +}
>> +
>> +/// # Safety
>> +///
>> +/// This function should be only used as the callback in `request_irq`.
> 
> I think this safety requirement is inadequate. We must require `ptr` to
> be valid for use as a reference to `T`. When we install the pointer to
> this function, we should certify why safety requirements are fulfilled
> when C calls through the pointer.

Ack.

> 
>> +unsafe extern "C" fn handle_irq_callback<T: Handler>(_irq: i32, ptr: *mut c_void) -> c_uint {
>> +    // SAFETY: `ptr` is a pointer to T set in `Registration::new`
>> +    let handler = unsafe { &*(ptr as *const T) };
>> +    T::handle(handler) as c_uint
>> +}
> 
> 
> Best regards,
> Andreas Hindborg