mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Gary Guo" <gary@garyguo.net>
To: "Danilo Krummrich" <dakr@kernel.org>, "Gary Guo" <gary@garyguo.net>
Cc: <gregkh@linuxfoundation.org>, <rafael@kernel.org>,
	<acourbot@nvidia.com>, <aliceryhl@google.com>,
	<david.m.ertman@intel.com>, <ira.weiny@intel.com>,
	<leon@kernel.org>, <viresh.kumar@linaro.org>,
	<m.wilczynski@samsung.com>, <ukleinek@kernel.org>,
	<bhelgaas@google.com>, <kwilczynski@kernel.org>,
	<abdiel.janulgue@gmail.com>, <robin.murphy@arm.com>,
	<markus.probst@posteo.de>, <ojeda@kernel.org>, <boqun@kernel.org>,
	<bjorn3_gh@protonmail.com>, <lossin@kernel.org>,
	<a.hindborg@kernel.org>, <tmgross@umich.edu>,
	<igor.korotin@linux.dev>, <daniel.almeida@collabora.com>,
	<pcolberg@redhat.com>, <driver-core@lists.linux.dev>,
	<linux-kernel@vger.kernel.org>, <nova-gpu@lists.linux.dev>,
	<dri-devel@lists.freedesktop.org>, <linux-pm@vger.kernel.org>,
	<linux-pwm@vger.kernel.org>, <linux-pci@vger.kernel.org>,
	<rust-for-linux@vger.kernel.org>
Subject: Re: [PATCH v3 17/27] rust: auxiliary: generalize Registration over ForLt
Date: Wed, 20 May 2026 10:34:05 +0100	[thread overview]
Message-ID: <DINEG6ZL6D2Y.239EMAC9FOLWU@garyguo.net> (raw)
In-Reply-To: <DIN2Y5PNWGVO.EOA9SQPOFYX2@kernel.org>

On Wed May 20, 2026 at 1:33 AM BST, Danilo Krummrich wrote:
> On Tue May 19, 2026 at 6:45 PM CEST, Gary Guo wrote:
>> On Sun May 17, 2026 at 1:01 AM BST, Danilo Krummrich wrote:
>>> +    pub fn new<'bound, E>(
>>> +        parent: &'bound device::Device<device::Bound>,
>>>          name: &CStr,
>>>          id: u32,
>>>          modname: &CStr,
>>> -        data: impl PinInit<T, E>,
>>> +        data: impl PinInit<F::Of<'bound>, E>,
>>>      ) -> Result<Devres<Self>>
>>
>> I think this is unsound for the reason that I gave in another email
>> https://lore.kernel.org/rust-for-linux/DIMSJVKTYX6D.AEN6OPPC2898@garyguo.net/.
>
> Indeed, this has to be fixed. I looked into the options brought up in the linked
> reply, i.e. the "new type" approach and the "closure" approach. And after
> playing around with both of them, I'm not really satisfied with either of those.
>
> The "new type" one is simple and works, but has the disadvantage that, well, we
> need a new type and update all callsites where we would ever want to create
> registrations (mainly probe though).
>
> The "closure" one on the other hand creates a little bit of an odd API and by
> its nature does not allow to move pre-existing resources into the closure, which
> is a major limitation (maybe there is some way, but if so I didn't find it).
>
> I instead went with something else: Currently we return a
> Devres<auxiliary::Registration<_>>. However, since we're moving to lifetimes
> anyway, we can just return an auxiliary::Registration<'bound, _> instead, which
> makes the return type invariant over 'bound, which makes the problem go away
> naturally.
>
> Please find the diff below for reference.
>
> diff --git a/rust/kernel/auxiliary.rs b/rust/kernel/auxiliary.rs
> index 5acece8e369a..c784426b8092 100644
> --- a/rust/kernel/auxiliary.rs
> +++ b/rust/kernel/auxiliary.rs
> @@ -12,7 +12,7 @@
>          RawDeviceId,
>          RawDeviceIdIndex, //
>      },
> -    devres::Devres,
> +
>      driver,
>      error::{
>          from_result,
> @@ -408,12 +408,12 @@ struct RegistrationData<T> {
>  /// `self.adev` always holds a valid pointer to an initialized and registered
>  /// [`struct auxiliary_device`] whose `registration_data_rust` field points to a
>  /// valid `Pin<KBox<RegistrationData<F::Of<'static>>>>`.
> -pub struct Registration<F: ForLt + 'static> {
> +pub struct Registration<'bound, F: ForLt + 'static> {
>      adev: NonNull<bindings::auxiliary_device>,
> -    _data: PhantomData<F>,
> +    _phantom: PhantomData<(fn(&'bound ()) -> &'bound (), F)>,

The type itself may be invariant over the lifetime, but it would still accept a
shorter lifetime (which again, is a case where I think having 'bound becomes a
factor for confusion).

You can still pass in a shortened lifetime to `Registration::new` and then leak
a registration (yes, Leakpocalypse strikes again), so the invariance here isn't
actually serving purpose.

    fn bad_fn(dev: &device::Device<device::Bound>, buf: &[u8]) {
        forget(Box::pin_init(Registration::new(dev, c"foo", 0, c"bar", buf)))
    }

Best,
Gary

>  }
>  
> -impl<F: ForLt> Registration<F>
> +impl<'bound, F: ForLt> Registration<'bound, F>
>  where
>      for<'a> F::Of<'a>: Send + Sync,
>  {
> @@ -421,13 +421,13 @@ impl<F: ForLt> Registration<F>
>      ///
>      /// The `data` is owned by the registration and can be accessed through the auxiliary device
>      /// via [`Device::registration_data()`].
> -    pub fn new<'bound, E>(
> +    pub fn new<E>(
>          parent: &'bound device::Device<device::Bound>,
>          name: &CStr,
>          id: u32,
>          modname: &CStr,
>          data: impl PinInit<F::Of<'bound>, E>,
> -    ) -> Result<Devres<Self>>
> +    ) -> Result<Self>
>      where
>          Error: From<E>,
>      {
> @@ -439,8 +439,10 @@ pub fn new<'bound, E>(
>              GFP_KERNEL,
>          )?;
>  
> -        // SAFETY: Lifetimes are erased and do not affect layout, so RegistrationData<F::Of<'bound>>
> -        // and RegistrationData<F::Of<'static>> have identical representation.
> +        // SAFETY: `'bound` is invariant (via `Registration`'s `PhantomData`), guaranteeing it
> +        // represents the full binding scope. Lifetimes do not affect layout, so
> +        // RegistrationData<F::Of<'bound>> and RegistrationData<F::Of<'static>> have identical
> +        // representation.
>          let data: Pin<KBox<RegistrationData<F::Of<'static>>>> =
>              unsafe { core::mem::transmute(data) };
>  
> @@ -487,18 +489,16 @@ pub fn new<'bound, E>(
>  
>          // INVARIANT: The device will remain registered until `auxiliary_device_delete()` is
>          // called, which happens in `Self::drop()`.
> -        let reg = Self {
> +        Ok(Self {
>              // SAFETY: `adev` is guaranteed to be non-null, since the `KBox` was allocated
>              // successfully.
>              adev: unsafe { NonNull::new_unchecked(adev) },
> -            _data: PhantomData,
> -        };
> -
> -        Devres::new::<core::convert::Infallible>(parent, reg)
> +            _phantom: PhantomData,
> +        })
>      }
>  }
>  
> -impl<F: ForLt> Drop for Registration<F> {
> +impl<F: ForLt> Drop for Registration<'_, F> {
>      fn drop(&mut self) {
>          // SAFETY: By the type invariant of `Self`, `self.adev.as_ptr()` is a valid registered
>          // `struct auxiliary_device`.
> @@ -520,7 +520,7 @@ fn drop(&mut self) {
>  }
>  
>  // SAFETY: A `Registration` of a `struct auxiliary_device` can be released from any thread.
> -unsafe impl<F: ForLt> Send for Registration<F> where for<'a> F::Of<'a>: Send {}
> +unsafe impl<F: ForLt> Send for Registration<'_, F> where for<'a> F::Of<'a>: Send {}
>  
>  // SAFETY: `Registration` does not expose any methods or fields that need synchronization.
> -unsafe impl<F: ForLt> Sync for Registration<F> where for<'a> F::Of<'a>: Send {}
> +unsafe impl<F: ForLt> Sync for Registration<'_, F> where for<'a> F::Of<'a>: Send {}
> diff --git a/samples/rust/rust_driver_auxiliary.rs b/samples/rust/rust_driver_auxiliary.rs
> index 84d18bbfafc5..efb4d97b416b 100644
> --- a/samples/rust/rust_driver_auxiliary.rs
> +++ b/samples/rust/rust_driver_auxiliary.rs
> @@ -10,7 +10,6 @@
>          Bound,
>          Core, //
>      },
> -    devres::Devres,
>      driver,
>      pci,
>      prelude::*,
> @@ -58,29 +57,29 @@ struct Data<'bound> {
>  }
>  
>  #[allow(clippy::type_complexity)]
> -struct ParentDriver {
> -    _reg0: Devres<auxiliary::Registration<ForLt!(Data<'_>)>>,
> -    _reg1: Devres<auxiliary::Registration<ForLt!(Data<'_>)>>,
> +struct ParentDriver<'bound> {
> +    _reg0: auxiliary::Registration<'bound, ForLt!(Data<'_>)>,
> +    _reg1: auxiliary::Registration<'bound, ForLt!(Data<'_>)>,
>  }
>  
>  kernel::pci_device_table!(
>      PCI_TABLE,
>      MODULE_PCI_TABLE,
> -    <ParentDriver as pci::Driver>::IdInfo,
> +    <ParentDriver<'_> as pci::Driver>::IdInfo,
>      [(pci::DeviceId::from_id(pci::Vendor::REDHAT, 0x5), ())]
>  );
>  
> -impl pci::Driver for ParentDriver {
> +impl pci::Driver for ParentDriver<'_> {
>      type IdInfo = ();
> -    type Data<'bound> = Self;
> +    type Data<'bound> = ParentDriver<'bound>;
>  
>      const ID_TABLE: pci::IdTable<Self::IdInfo> = &PCI_TABLE;
>  
>      fn probe<'bound>(
>          pdev: &'bound pci::Device<Core>,
>          _info: &'bound Self::IdInfo,
> -    ) -> impl PinInit<Self, Error> + 'bound {
> -        Ok(Self {
> +    ) -> impl PinInit<Self::Data<'bound>, Error> + 'bound {
> +        Ok(ParentDriver {
>              _reg0: auxiliary::Registration::new(
>                  pdev.as_ref(),
>                  AUXILIARY_NAME,
> @@ -105,7 +104,7 @@ fn probe<'bound>(
>      }
>  }
>  
> -impl ParentDriver {
> +impl ParentDriver<'_> {
>      fn connect(adev: &auxiliary::Device<Bound>) -> Result {
>          let data = adev.registration_data::<ForLt!(Data<'_>)>()?;
>          let pdev = data.parent;
> @@ -131,7 +130,7 @@ fn connect(adev: &auxiliary::Device<Bound>) -> Result {
>  #[pin_data]
>  struct SampleModule {
>      #[pin]
> -    _pci_driver: driver::Registration<pci::Adapter<ParentDriver>>,
> +    _pci_driver: driver::Registration<pci::Adapter<ParentDriver<'static>>>,
>      #[pin]
>      _aux_driver: driver::Registration<auxiliary::Adapter<AuxiliaryDriver>>,
>  }


  reply	other threads:[~2026-05-20  9:34 UTC|newest]

Thread overview: 80+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-17  0:00 [PATCH v3 00/27] rust: device: Higher-Ranked Lifetime Types for device drivers Danilo Krummrich
2026-05-17  0:00 ` [PATCH v3 01/27] rust: alloc: remove `'static` bound on `ForeignOwnable` Danilo Krummrich
2026-05-18 14:42   ` Alexandre Courbot
2026-05-17  0:00 ` [PATCH v3 02/27] rust: driver: move 'static bounds to constructor Danilo Krummrich
2026-05-18 14:42   ` Alexandre Courbot
2026-05-17  0:00 ` [PATCH v3 03/27] rust: driver: decouple driver private data from driver type Danilo Krummrich
2026-05-17 14:32   ` Danilo Krummrich
2026-05-19 12:47     ` Gary Guo
2026-05-18 14:43   ` Alexandre Courbot
2026-05-17  0:00 ` [PATCH v3 04/27] rust: driver core: drop drvdata before devres release Danilo Krummrich
2026-05-18 14:45   ` Alexandre Courbot
2026-05-19 12:47   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 05/27] rust: pci: implement Sync for Device<Bound> Danilo Krummrich
2026-05-18 14:46   ` Alexandre Courbot
2026-05-19 13:01   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 06/27] rust: platform: " Danilo Krummrich
2026-05-18 14:46   ` Alexandre Courbot
2026-05-19 13:01   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 07/27] rust: auxiliary: " Danilo Krummrich
2026-05-18 14:47   ` Alexandre Courbot
2026-05-19 13:02   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 08/27] rust: usb: " Danilo Krummrich
2026-05-18 14:47   ` Alexandre Courbot
2026-05-19 13:02   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 09/27] rust: device: " Danilo Krummrich
2026-05-18 14:48   ` Alexandre Courbot
2026-05-19 13:02   ` Gary Guo
2026-05-17  0:00 ` [PATCH v3 10/27] rust: pci: make Driver trait lifetime-parameterized Danilo Krummrich
2026-05-18 14:53   ` Alexandre Courbot
2026-05-18 15:36   ` Gary Guo
2026-05-18 16:10     ` Danilo Krummrich
2026-05-19  4:52   ` Eliot Courtney
2026-05-19 10:39     ` Danilo Krummrich
2026-05-19 11:48       ` Gary Guo
2026-05-19 12:36         ` Danilo Krummrich
2026-05-20  6:14           ` Eliot Courtney
2026-05-17  0:00 ` [PATCH v3 11/27] rust: platform: " Danilo Krummrich
2026-05-18 14:55   ` Alexandre Courbot
2026-05-17  0:01 ` [PATCH v3 12/27] rust: auxiliary: " Danilo Krummrich
2026-05-18 15:39   ` Alexandre Courbot
2026-05-17  0:01 ` [PATCH v3 13/27] rust: usb: " Danilo Krummrich
2026-05-18 15:40   ` Alexandre Courbot
2026-05-17  0:01 ` [PATCH v3 14/27] rust: i2c: " Danilo Krummrich
2026-05-18 15:41   ` Alexandre Courbot
2026-05-17  0:01 ` [PATCH v3 15/27] rust: driver: update module documentation for GAT-based Data type Danilo Krummrich
2026-05-18 15:46   ` Alexandre Courbot
2026-05-17  0:01 ` [PATCH v3 16/27] rust: types: add `ForLt` trait for higher-ranked lifetime support Danilo Krummrich
2026-05-19  6:02   ` Eliot Courtney
2026-05-19 11:23     ` Gary Guo
2026-05-19 11:07   ` Alexandre Courbot
2026-05-19 11:39     ` Gary Guo
2026-05-19 13:03       ` Danilo Krummrich
2026-05-19 13:34         ` Miguel Ojeda
2026-05-17  0:01 ` [PATCH v3 17/27] rust: auxiliary: generalize Registration over ForLt Danilo Krummrich
2026-05-19  7:56   ` Eliot Courtney
2026-05-19 10:39     ` Danilo Krummrich
2026-05-19 11:20       ` Gary Guo
2026-05-19 16:45   ` Gary Guo
2026-05-20  0:33     ` Danilo Krummrich
2026-05-20  9:34       ` Gary Guo [this message]
2026-05-17  0:01 ` [PATCH v3 18/27] samples: rust: rust_driver_auxiliary: showcase lifetime-bound registration data Danilo Krummrich
2026-05-19  6:52   ` Eliot Courtney
2026-05-19 15:48   ` Gary Guo
2026-05-17  0:01 ` [PATCH v3 19/27] rust: pci: make Bar lifetime-parameterized Danilo Krummrich
2026-05-19  6:36   ` Eliot Courtney
2026-05-19 16:24   ` Gary Guo
2026-05-19 17:27     ` Danilo Krummrich
2026-05-17  0:01 ` [PATCH v3 20/27] rust: io: make IoMem and ExclusiveIoMem lifetime-parameterized Danilo Krummrich
2026-05-19  6:39   ` Eliot Courtney
2026-05-17  0:01 ` [PATCH v3 21/27] samples: rust: rust_driver_pci: use HRT lifetime for Bar Danilo Krummrich
2026-05-19  6:41   ` Eliot Courtney
2026-05-17  0:01 ` [PATCH v3 22/27] rust: driver-core: rename 'a lifetime to 'bound Danilo Krummrich
2026-05-19  6:42   ` Eliot Courtney
2026-05-19 16:56   ` Gary Guo
2026-05-19 17:23     ` Danilo Krummrich
2026-05-17  0:01 ` [PATCH REF v3 23/27] gpu: nova-core: " Danilo Krummrich
2026-05-17  0:01 ` [PATCH REF v3 24/27] gpu: nova-core: use lifetime for Bar Danilo Krummrich
2026-05-17  0:01 ` [PATCH REF v3 25/27] gpu: nova-core: unregister sysmem flush page from Drop Danilo Krummrich
2026-05-17  0:01 ` [PATCH REF v3 26/27] gpu: nova-core: replace ARef<Device> with &'bound Device in SysmemFlush Danilo Krummrich
2026-05-17  0:01 ` [PATCH REF v3 27/27] gpu: drm: tyr: use lifetime for IoMem Danilo Krummrich

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DINEG6ZL6D2Y.239EMAC9FOLWU@garyguo.net \
    --to=gary@garyguo.net \
    --cc=a.hindborg@kernel.org \
    --cc=abdiel.janulgue@gmail.com \
    --cc=acourbot@nvidia.com \
    --cc=aliceryhl@google.com \
    --cc=bhelgaas@google.com \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun@kernel.org \
    --cc=dakr@kernel.org \
    --cc=daniel.almeida@collabora.com \
    --cc=david.m.ertman@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=driver-core@lists.linux.dev \
    --cc=gregkh@linuxfoundation.org \
    --cc=igor.korotin@linux.dev \
    --cc=ira.weiny@intel.com \
    --cc=kwilczynski@kernel.org \
    --cc=leon@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=linux-pwm@vger.kernel.org \
    --cc=lossin@kernel.org \
    --cc=m.wilczynski@samsung.com \
    --cc=markus.probst@posteo.de \
    --cc=nova-gpu@lists.linux.dev \
    --cc=ojeda@kernel.org \
    --cc=pcolberg@redhat.com \
    --cc=rafael@kernel.org \
    --cc=robin.murphy@arm.com \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=tmgross@umich.edu \
    --cc=ukleinek@kernel.org \
    --cc=viresh.kumar@linaro.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox