[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87sek82bgf.fsf@kernel.org>
Date: Tue, 10 Jun 2025 11:27:28 +0200
From: Andreas Hindborg <a.hindborg@...nel.org>
To: "Benno Lossin" <lossin@...nel.org>
Cc: "Danilo Krummrich" <dakr@...nel.org>, "Miguel Ojeda"
<ojeda@...nel.org>, "Alex Gaynor" <alex.gaynor@...il.com>, "Boqun Feng"
<boqun.feng@...il.com>, "Gary Guo" <gary@...yguo.net>, Björn Roy Baron
<bjorn3_gh@...tonmail.com>, "Alice Ryhl" <aliceryhl@...gle.com>, "Trevor
Gross" <tmgross@...ch.edu>, "Bjorn Helgaas" <bhelgaas@...gle.com>,
Krzysztof Wilczyński <kwilczynski@...nel.org>, "Greg
Kroah-Hartman"
<gregkh@...uxfoundation.org>, "Rafael J. Wysocki" <rafael@...nel.org>,
"Tamir Duberstein" <tamird@...il.com>, "Viresh Kumar"
<viresh.kumar@...aro.org>, <rust-for-linux@...r.kernel.org>,
<linux-kernel@...r.kernel.org>, <linux-pci@...r.kernel.org>,
Maíra
Canal <mcanal@...lia.com>
Subject: Re: [PATCH] rust: types: add FOREIGN_ALIGN to ForeignOwnable
Hi Benno,
"Benno Lossin" <lossin@...nel.org> writes:
> The title should probably also mention that it removes `PointedTo`.
>
> On Thu Jun 5, 2025 at 9:55 PM CEST, Andreas Hindborg wrote:
>> The current implementation of `ForeignOwnable` is leaking the type of the
>> opaque pointer to consumers of the API. This allows consumers of the opaque
>> pointer to rely on the information that can be extracted from the pointer
>> type.
>>
>> To prevent this, change the API to the version suggested by Maira
>> Canal (link below): Remove `ForeignOwnable::PointedTo` in favor of a
>> constant, which specifies the alignment of the pointers returned by
>> `into_foreign`.
>>
>> Suggested-by: Alice Ryhl <aliceryhl@...gle.com>
>> Suggested-by: Maíra Canal <mcanal@...lia.com>
>> Link: https://lore.kernel.org/r/20240309235927.168915-3-mcanal@igalia.com
>> Signed-off-by: Andreas Hindborg <a.hindborg@...nel.org>
>
> A couple nits and documentation review below, with those fixed:
>
> Reviewed-by: Benno Lossin <lossin@...nel.org>
>
>> ---
>> rust/kernel/alloc/kbox.rs | 40 ++++++++++++++++++++++------------------
>> rust/kernel/miscdevice.rs | 10 +++++-----
>> rust/kernel/pci.rs | 2 +-
>> rust/kernel/platform.rs | 2 +-
>> rust/kernel/sync/arc.rs | 23 ++++++++++++-----------
>> rust/kernel/types.rs | 46 +++++++++++++++++++++-------------------------
>> rust/kernel/xarray.rs | 8 ++++----
>> 7 files changed, 66 insertions(+), 65 deletions(-)
>>
>> diff --git a/rust/kernel/alloc/kbox.rs b/rust/kernel/alloc/kbox.rs
>> index c386ff771d50..97f45bc4d74f 100644
>> --- a/rust/kernel/alloc/kbox.rs
>> +++ b/rust/kernel/alloc/kbox.rs
>> @@ -398,70 +398,74 @@ fn try_init<E>(init: impl Init<T, E>, flags: Flags) -> Result<Self, E>
>> }
>> }
>>
>> -// SAFETY: The `into_foreign` function returns a pointer that is well-aligned.
>> +// SAFETY: The pointer returned by `into_foreign` comes from a well aligned
>> +// pointer to `T`.
>> unsafe impl<T: 'static, A> ForeignOwnable for Box<T, A>
>> where
>> A: Allocator,
>> {
>> - type PointedTo = T;
>> + const FOREIGN_ALIGN: usize = core::mem::align_of::<T>();
>> type Borrowed<'a> = &'a T;
>> type BorrowedMut<'a> = &'a mut T;
>>
>> - fn into_foreign(self) -> *mut Self::PointedTo {
>> - Box::into_raw(self)
>> + fn into_foreign(self) -> *mut crate::ffi::c_void {
>
> How about we import the prelude, then you can just write `*mut c_void`
> everywhere instead of having to write `crate::ffi` all the time.
OK.
>
>> diff --git a/rust/kernel/sync/arc.rs b/rust/kernel/sync/arc.rs
>> index c7af0aa48a0a..6603079b05af 100644
>> --- a/rust/kernel/sync/arc.rs
>> +++ b/rust/kernel/sync/arc.rs
>> @@ -140,10 +140,9 @@ pub struct Arc<T: ?Sized> {
>> _p: PhantomData<ArcInner<T>>,
>> }
>>
>> -#[doc(hidden)]
>> #[pin_data]
>> #[repr(C)]
>> -pub struct ArcInner<T: ?Sized> {
>> +struct ArcInner<T: ?Sized> {
>
> I agree with this change, but let's mention it in the commit message.
Right.
>
>> refcount: Opaque<bindings::refcount_t>,
>> data: T,
>> }
>
>> diff --git a/rust/kernel/types.rs b/rust/kernel/types.rs
>> index 22985b6f6982..025c619a2195 100644
>> --- a/rust/kernel/types.rs
>> +++ b/rust/kernel/types.rs
>> @@ -21,15 +21,11 @@
>> ///
>> /// # Safety
>> ///
>> -/// Implementers must ensure that [`into_foreign`] returns a pointer which meets the alignment
>> -/// requirements of [`PointedTo`].
>> -///
>> -/// [`into_foreign`]: Self::into_foreign
>> -/// [`PointedTo`]: Self::PointedTo
>> +/// Implementers must ensure that [`Self::into_foreign`] return pointers with alignment that is an
>
> s/return/returns/
Thanks.
>
>> +/// integer multiple of [`Self::FOREIGN_ALIGN`].
>
> I would just write "returns pointers aligned to [`Self::FOREIGN_ALIGN`]".
OK.
>
>> pub unsafe trait ForeignOwnable: Sized {
>> - /// Type used when the value is foreign-owned. In practical terms only defines the alignment of
>> - /// the pointer.
>> - type PointedTo;
>> + /// The alignment of pointers returned by `into_foreign`.
>> + const FOREIGN_ALIGN: usize;
>>
>> /// Type used to immutably borrow a value that is currently foreign-owned.
>> type Borrowed<'a>;
>> @@ -39,18 +35,17 @@ pub unsafe trait ForeignOwnable: Sized {
>>
>> /// Converts a Rust-owned object to a foreign-owned one.
>> ///
>> - /// # Guarantees
>
> Why remove this section? I think we should streamline it, (make it use
> bullet points, shorten the sentences etc). We can keep the paragraph you
> wrote below as normal docs.
Not sure exactly what you are going for here. How is this:
Converts a Rust-owned object to a foreign-owned one.
The foreign representation is a pointer to void.
# Guarantees
- Minimum alignment of returned pointer is [`Self::FOREIGN_ALIGN`].
There are no other guarantees for this pointer. For example, it might be invalid, dangling
or pointing to uninitialized memory. Using it in any way except for [`from_foreign`],
[`try_from_foreign`], [`borrow`], or [`borrow_mut`] can result in undefined behavior.
Best regards,
Andreas Hindborg
Powered by blists - more mailing lists