Rust pin-init API for pinned initialization of structs
Subsystems (4)
v4 cover Rust pin-init API for pinned initialization of structs
From: Benno Lossin <[email protected]>
This is the fourth version of the pin-init API. See [1] for v3.
The tree at [2] contains these patches applied on top of 6.3-rc1. The Rust-doc documentation of the pin-init API can be found at [3].
These patches are a long way coming, since I held a presentation on safe pinned initialization at Kangrejos [4]. And my discovery of this problem was almost a year ago [5].
- #1 enables the pin_macro feature which is already stable in Rust version 1.68.
- #2 adds a utility macro quote! for proc-macros. This macro converts the typed characters directly into Rust tokens that are the output of proc-macros. It is used by the pin-init API proc-macros.
- #3 changes the Err types of the constructor functions of Arc and UniqueArc to AllocError.
- #4 adds the assume_init function to UniqueArc<MaybeUninit<T>> that unsafely assumes the pointee to be initialized and returns a UniqueArc<T>. UniqueArc::write is modified to use this new function. Later patches use it as well.
- #5 adds Opaque::raw_get to access the value inside of an Opaque from a raw pointer only.
- #6-11 introduce the pin-init API. The commit message of #4 details the problem it solves and lays out the overall architecture. The implementation details are fairly complex; however, this is required to provide a safe API for users -- reducing the amount of unsafe code is a key goal of the Rust support in the kernel. An example of the before/after difference from the point of view of users is provided in the commit message. It is a goal to at some point have pin-init as a language feature of Rust. A first step in this direction is the Field Projection RFC [6].
- #12 adds important traits and macros from pin-init to the prelude.
- #13 adds functions for easier initialization of Opaque<T> via FFI and raw pointer initializer functions. This is necessary when writing Rust wrappers and will be used by Wedson's sync module patches.
- #14 improves the UniqueArc::try_new_uninit function by using the pin-init API. The old version first allocated uninitialized memory on the stack and then moved it into the location in the heap. The new version directly allocates this on the heap.
- #15 adds functions for initializing already allocated UniqueArcs, this will be used by the Android Binder driver.
--
Changelog:
v3 -> v4:
- Improve documentation.
- Fixing doc-tests imports, comments and full paths in macro examples.
- Implement Zeroable for many more types.
- Fix unsoundness: Zeroable allowed to create fat pointers with a null VTABLE pointer.
- Split fallible components from stack_pin_init! into stack_try_pin_init!.
- Move Invariant type alias and InitClosure into __internal.
- Change the error type of the constructor functions of Arc and UniqueArc to AllocError.
- Add try variants to InPlaceInit to allow custom error types.
- Make StackInit::init a safe function.
- Rename OnlyCallFromDrop::create to OnlyCallFromDrop::new.
- Enable the pin_macro feature for use inside of stack_pin_init!.
- Make quote! and quote_spanned! use absolute paths.
v2 -> v3:
- Split the big original commit into six smaller commits.
- Use PinnedDrop in the RawFoo code example.
- Move the init::common::ffi_init functions to Opaque::ffi_init.
- Move all #[doc(hidden)], internal types into the new __internal module.
- Specify safety guarantees and requirements of the initializer macros.
- Add a detailed example of what the expanded code of the various macros looks like.
- Require generics in the initializer macros to use turbofish syntax (::<>).
- Refactor the internals of the initializer macros, this way they have better type inference and generic arguments can be omitted more often.
- Replace init::from_value with a blanket impl of the initializer traits for ordinary objects.
- Add initializing functions for already allocated UniqueArcs.
- Add Opaque::manual_init functions akin to ffi_init, but they take an extern "Rust" fn instead.
- Documentation and inline comment improvements.
v1 -> v2:
- Split the common module and UniqueArc::assume_init into their own commits.
- Change the generics syntax of pin_init! to reflect normal struct generic syntax.
- Replace PinnedDrop::__ensure_no_unsafe_op_in_drop with an only unsafely creatable token.
- Hide StackInit<T> in the docs, because it is internal API.
- Improve macro internals of pin_init! according to Gary's review.
- Add check for PhantomPinned fields without a #[pin] attribute in #[pin_data], as those fields will not have the intended effect.
- Add docs to quote.rs.
[1] https://lore.kernel.org/rust-for-linux/[email protected]/
[2] https://github.com/y86-dev/linux.git patch/pinned-init-v4
[3] https://rust-for-linux.github.io/docs/pinned-init/kernel/init
[4] https://kangrejos.com
[5] https://github.com/Rust-for-Linux/linux/issues/772
[6] https://github.com/rust-lang/rfcs/pull/3318
Benno Lossin (14):
rust: enable the pin_macro feature
rust: sync: change error type of constructor functions
rust: sync: add assume_init to UniqueArc
rust: types: add Opaque::raw_get
rust: add pin-init API core
rust: init: add initialization macros
rust: init/sync: add InPlaceInit trait to pin-initialize smart
pointers
rust: init: add PinnedDrop trait and macros
rust: init: add stack_pin_init! macro
rust: init: add Zeroable trait and init::zeroed function
rust: prelude: add pin-init API items to prelude
rust: types: add common init-helper functions for Opaque
rust: sync: reduce stack usage of UniqueArc::try_new_uninit
rust: sync: add functions for initializing UniqueArc<MaybeUninit<T>>
Gary Guo (1):
rust: macros: add quote! macro
rust/kernel/init.rs | 1429 ++++++++++++++++++++++++++++++++ rust/kernel/init/__internal.rs | 234 ++++++ rust/kernel/init/macros.rs | 971 ++++++++++++++++++++++ rust/kernel/lib.rs | 7 + rust/kernel/prelude.rs | 6 +- rust/kernel/sync/arc.rs | 80 +- rust/kernel/types.rs | 55 ++ rust/macros/lib.rs | 80 ++ rust/macros/pin_data.rs | 79 ++ rust/macros/pinned_drop.rs | 49 ++ rust/macros/quote.rs | 143 ++++ scripts/Makefile.build | 2 +- 12 files changed, 3127 insertions(+), 8 deletions(-) create mode 100644 rust/kernel/init.rs create mode 100644 rust/kernel/init/__internal.rs create mode 100644 rust/kernel/init/macros.rs create mode 100644 rust/macros/pin_data.rs create mode 100644 rust/macros/pinned_drop.rs create mode 100644 rust/macros/quote.rs
base-commit: fe15c26ee26efa11741a7b632e9f23b01aca4cc6 --
2.39.2
PATCH v4 1/15 rust: enable the `pin_macro` feature
From: Benno Lossin <[email protected]>
This feature enables the use of the pin! macro for the stack_pin_init! macro. This feature is already stabilized in Rust version 1.68.
--- rust/kernel/lib.rs | 1 + 1 file changed, 1 insertion(+)
rust/kernel/lib.rs +1 -0
@@ -17,6 +17,7 @@#![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] #![feature(generic_associated_types)] +#![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)]
--
2.39.2
Re: [PATCH v4 01/15] rust: enable the `pin_macro` feature
Quoted 9 lines · On Fri, Mar 31, 2023 at 09:51:35PM +0000, [email protected] wrote:
From: Benno Lossin <[email protected]>
This feature enables the use of the
pin!macro for thestack_pin_init!macro. This feature is already stabilized in Rust version 1.68.Signed-off-by Benno Lossin <[email protected]>Cc Gary Guo <[email protected]>Cc Andreas Hindborg <[email protected]>Cc Alice Ryhl <[email protected]>
Regards,
Boqun
Quoted 20 lines
--- rust/kernel/lib.rs | 1 + 1 file changed, 1 insertion(+)rust/kernel/lib.rs +1
-1@@ -17,6 +17,7 @@#![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] #![feature(generic_associated_types)] +#![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)] --2.39.2
Re: [PATCH v4 01/15] rust: enable the `pin_macro` feature
Quoted 6 lines · Boqun Feng
Quoted 5 lines · On Fri, Mar 31, 2023 at 09:51:35PM +0000, [email protected] wrote:
From: Benno Lossin <[email protected]>
This feature enables the use of the
pin!macro for thestack_pin_init!macro. This feature is already stabilized in Rust version 1.68.
Quoted 7 lines
Quoted 4 lines
Signed-off-by Benno Lossin <[email protected]>Cc Gary Guo <[email protected]>Cc Andreas Hindborg <[email protected]>Cc Alice Ryhl <[email protected]>Acked-by Boqun Feng <[email protected]>
Hmm.. actually I think you will need to add this in the scripts/Makefile.build, because the user (stack_pin_init!()) is a macro, feel free to add my acked-by after that.
Regards,
Boqun
Quoted 23 lines
Regards,
BoqunQuoted 20 lines
--- rust/kernel/lib.rs | 1 + 1 file changed, 1 insertion(+)rust/kernel/lib.rs +1
-1@@ -17,6 +17,7 @@#![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] #![feature(generic_associated_types)] +#![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)] --2.39.2
PATCH v4 2/15 rust: macros: add `quote!` macro
In reply to [email protected]From: Gary Guo <[email protected]>
Add the quote! macro for creating TokenStreams directly via the given Rust tokens. It also supports repetitions using iterators.
It will be used by the pin-init API proc-macros to generate code.
--- rust/macros/lib.rs | 2 + rust/macros/quote.rs | 145 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 147 insertions(+) create mode 100644 rust/macros/quote.rs
rust/macros/lib.rs +2 -0
@@ -2,6 +2,8 @@//! Crate for all kernel procedural macros. +#[macro_use] +mod quote;mod concat_idents;mod helpers;mod module;
rust/macros/quote.rs new file +145 -0
145 lines not on this page · raw message
--
2.39.2
Re: [PATCH v4 02/15] rust: macros: add `quote!` macro
Quoted 5 lines · [email protected] writes:
From: Gary Guo <[email protected]>
Add the
quote!macro for creatingTokenStreams directly via the given Rust tokens. It also supports repetitions using iterators.
Quoted 7 lines
It will be used by the pin-init API proc-macros to generate code.
Signed-off-by Gary Guo <[email protected]>Signed-off-by Benno Lossin <[email protected]>Cc Alice Ryhl <[email protected]>Cc Andreas Hindborg <[email protected]>---
Quoted 169 lines
rust/macros/lib.rs | 2 + rust/macros/quote.rs | 145 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 147 insertions(+) create mode 100644 rust/macros/quote.rsrust/macros/lib.rs +2
-0@@ -2,6 +2,8 @@//! Crate for all kernel procedural macros. +#[macro_use] +mod quote;mod concat_idents;mod helpers;mod module;rust/macros/quote.rs new file +145
-0@@ -0,0 +1,145 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +use proc_macro::{TokenStream ,TokenTree }; + +pub (crate )trait ToTokens { +fn to_tokens (&self, tokens: &mut TokenStream ); +} + +impl <T :ToTokens >ToTokens for Option <T > { +fn to_tokens (&self, tokens: &mut TokenStream ) { +if let Some (v) = self { + v.to_tokens (tokens); + } + } +} + +impl ToTokens for proc_macro::Group { +fn to_tokens (&self, tokens: &mut TokenStream ) { + tokens.extend ([TokenTree ::from (self.clone ())]); + } +} + +impl ToTokens for TokenTree { +fn to_tokens (&self, tokens: &mut TokenStream ) { + tokens.extend ([self.clone ()]); + } +} + +impl ToTokens for TokenStream { +fn to_tokens (&self, tokens: &mut TokenStream ) { + tokens.extend (self.clone ()); + } +} + +/// Converts tokens into [`proc_macro::TokenStream`] and performs variable interpolations with +/// the given span. +/// +/// This is a similar to the +/// [`quote_spanned!`](https://docs.rs/quote/latest/quote/macro.quote_spanned.html) macro from the +/// `quote` crate but provides only just enough functionality needed by the current `macros` crate. +#[allow(unused_macros)] +macro_rules! quote_spanned { + ($span:expr => $($tt:tt)*) => { +#[allow(clippy::vec_init_then_push)] + { +let mut tokens = ::std::vec::Vec ::new (); +let span = $span; +quote_spanned! (@proc tokensspan $($tt)*); + ::proc_macro::TokenStream ::from_iter (tokens) + }}; + (@proc $v:ident $span:ident) => {}; + (@proc $v:ident $span:ident #$id:ident $($tt:tt)*) => { +let mut ts = ::proc_macro::TokenStream ::new (); + $crate::quote::ToTokens ::to_tokens (&$id, &mut ts); + $v.extend (ts); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident #(#$id:ident)* $($tt:tt)*) => { +for tokenin $id { +let mut ts = ::proc_macro::TokenStream ::new (); + $crate::quote::ToTokens ::to_tokens (&token, &mut ts); + $v.extend (ts); + } +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident ( $($inner:tt)* ) $($tt:tt)*) => { +let mut tokens = ::std::vec::Vec ::new (); +quote_spanned! (@proc tokens $span $($inner)*); + $v.push (::proc_macro::TokenTree ::Group (::proc_macro::Group ::new ( + ::proc_macro::Delimiter ::Parenthesis , + ::proc_macro::TokenStream ::from_iter (tokens) + ))); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident [ $($inner:tt)* ] $($tt:tt)*) => { +let mut tokens = ::std::vec::Vec ::new (); +quote_spanned! (@proc tokens $span $($inner)*); + $v.push (::proc_macro::TokenTree ::Group (::proc_macro::Group ::new ( + ::proc_macro::Delimiter ::Bracket , + ::proc_macro::TokenStream ::from_iter (tokens) + ))); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident { $($inner:tt)* } $($tt:tt)*) => { +let mut tokens = ::std::vec::Vec ::new (); +quote_spanned! (@proc tokens $span $($inner)*); + $v.push (::proc_macro::TokenTree ::Group (::proc_macro::Group ::new ( + ::proc_macro::Delimiter ::Brace , + ::proc_macro::TokenStream ::from_iter (tokens) + ))); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident :: $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new (':' , ::proc_macro::Spacing ::Joint ) + )); + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new (':' , ::proc_macro::Spacing ::Alone ) + )); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident : $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new (':' , ::proc_macro::Spacing ::Alone ) + )); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident , $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new (',' , ::proc_macro::Spacing ::Alone ) + )); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident @ $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new ('@' , ::proc_macro::Spacing ::Alone ) + )); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident ! $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Punct ( + ::proc_macro::Punct ::new ('!' , ::proc_macro::Spacing ::Alone ) + )); +quote_spanned! (@proc $v $span $($tt)*); + }; + (@proc $v:ident $span:ident $id:ident $($tt:tt)*) => { + $v.push (::proc_macro::TokenTree ::Ident (::proc_macro::Ident ::new (stringify! ($id), $span))); +quote_spanned! (@proc $v $span $($tt)*); + }; +} + +/// Converts tokens into [`proc_macro::TokenStream`] and performs variable interpolations with +/// mixed site span ([`Span::mixed_site()`]). +/// +/// This is a similar to the [`quote!`](https://docs.rs/quote/latest/quote/macro.quote.html) macro +/// from the `quote` crate but provides only just enough functionality needed by the current +/// `macros` crate. +/// +/// [`Span::mixed_site()`]: https://doc.rust-lang.org/proc_macro/struct.Span.html#method.mixed_site +#[allow(unused_macros)] +macro_rules! quote { + ($($tt:tt)*) => { +quote_spanned! (::proc_macro::Span ::mixed_site () => $($tt)*) + } +}
PATCH v4 3/15 rust: sync: change error type of constructor functions
In reply to [email protected]From: Benno Lossin <[email protected]>
Change the error type of the constructors of Arc and UniqueArc to be AllocError instead of Error. This makes the API more clear as to what can go wrong when calling try_new or its variants.
--- rust/kernel/sync/arc.rs | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
rust/kernel/sync/arc.rs +4 -4
@@ -17,11 +17,11 @@use crate ::{ bindings, - error::Result , types::{ForeignOwnable ,Opaque }, };use alloc::boxed::Box ;use core::{ + alloc::AllocError , marker::{PhantomData ,Unsize }, mem::{ManuallyDrop ,MaybeUninit }, ops::{Deref ,DerefMut },
@@ -149,7 +149,7 @@ unsafe impl<T: ?Sized + Sync + Send> Sync for Arc<T> {}impl <T >Arc <T > {/// Constructs a new reference counted instance of `T`. -pub fn try_new (contents:T ) ->Result <Self > { +pub fn try_new (contents:T ) ->Result <Self ,AllocError > {// INVARIANT: The refcount is initialised to a non-zero value. let value = ArcInner {// SAFETY: There are no safety requirements for this FFI call.
@@ -469,7 +469,7 @@ pub struct UniqueArc<T: ?Sized> {impl <T >UniqueArc <T > {/// Tries to allocate a new [`UniqueArc`] instance. -pub fn try_new (value:T ) ->Result <Self > { +pub fn try_new (value:T ) ->Result <Self ,AllocError > {Ok (Self {// INVARIANT: The newly-created object has a ref-count of 1. inner:Arc ::try_new (value)?,
@@ -477,7 +477,7 @@ impl<T> UniqueArc<T> { }/// Tries to allocate a new [`UniqueArc`] instance whose contents are not initialised yet. -pub fn try_new_uninit () ->Result <UniqueArc <MaybeUninit <T >>> { +pub fn try_new_uninit () ->Result <UniqueArc <MaybeUninit <T >>,AllocError > {Ok (UniqueArc ::<MaybeUninit <T >> {// INVARIANT: The newly-created object has a ref-count of 1. inner:Arc ::try_new (MaybeUninit ::uninit ())?,
--
2.39.2
Re: [PATCH v4 03/15] rust: sync: change error type of constructor functions
Quoted 11 lines · [email protected] writes:
From: Benno Lossin <[email protected]>
Change the error type of the constructors of
ArcandUniqueArcto beAllocErrorinstead ofError. This makes the API more clear as to what can go wrong when callingtry_newor its variants.Signed-off-by Benno Lossin <[email protected]>Cc Gary Guo <[email protected]>Cc Andreas Hindborg <[email protected]>Cc Alice Ryhl <[email protected]>---
Quoted 47 lines
rust/kernel/sync/arc.rs | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)rust/kernel/sync/arc.rs +4
-4@@ -17,11 +17,11 @@use crate ::{ bindings, - error::Result , types::{ForeignOwnable ,Opaque }, };use alloc::boxed::Box ;use core::{ + alloc::AllocError , marker::{PhantomData ,Unsize }, mem::{ManuallyDrop ,MaybeUninit }, ops::{Deref ,DerefMut },@@ -149,7 +149,7 @@ unsafe impl<T: ?Sized + Sync + Send> Sync for Arc<T> {}impl <T >Arc <T > {/// Constructs a new reference counted instance of `T`. -pub fn try_new (contents:T ) ->Result <Self > { +pub fn try_new (contents:T ) ->Result <Self ,AllocError > {// INVARIANT: The refcount is initialised to a non-zero value. let value = ArcInner {// SAFETY: There are no safety requirements for this FFI call. @@ -469,7 +469,7 @@ pub struct UniqueArc<T: ?Sized> {impl <T >UniqueArc <T > {/// Tries to allocate a new [`UniqueArc`] instance. -pub fn try_new (value:T ) ->Result <Self > { +pub fn try_new (value:T ) ->Result <Self ,AllocError > {Ok (Self {// INVARIANT: The newly-created object has a ref-count of 1. inner:Arc ::try_new (value)?,@@ -477,7 +477,7 @@ impl<T> UniqueArc<T> { }/// Tries to allocate a new [`UniqueArc`] instance whose contents are not initialised yet. -pub fn try_new_uninit () ->Result <UniqueArc <MaybeUninit <T >>> { +pub fn try_new_uninit () ->Result <UniqueArc <MaybeUninit <T >>,AllocError > {Ok (UniqueArc ::<MaybeUninit <T >> {// INVARIANT: The newly-created object has a ref-count of 1. inner:Arc ::try_new (MaybeUninit ::uninit ())?,
PATCH v4 4/15 rust: sync: add `assume_init` to `UniqueArc`
In reply to [email protected]From: Benno Lossin <[email protected]>
Adds the assume_init function to UniqueArc<MaybeUninit<T>> that unsafely assumes the value to be initialized and yields a value of type UniqueArc<T>. This function is used when manually initializing the pointee of an UniqueArc.
--- rust/kernel/sync/arc.rs | 11 +++++++++++ 1 file changed, 11 insertions(+)
rust/kernel/sync/arc.rs +11 -0
@@ -489,6 +489,17 @@ impl<T> UniqueArc<MaybeUninit<T>> {/// Converts a `UniqueArc<MaybeUninit<T>>` into a `UniqueArc<T>` by writing a value into it. pub fn write (mut self, value:T ) ->UniqueArc <T > { self.deref_mut ().write (value); +// SAFETY: We just wrote the value to be initialized. +unsafe { self.assume_init () } + } + +/// Unsafely assume that `self` is initialized. +/// +/// # Safety +/// +/// The caller guarantees that the value behind this pointer has been initialized. It is +/// *immediate* UB to call this when the value is not initialized. +pub unsafe fn assume_init (self) ->UniqueArc <T > {let inner =ManuallyDrop ::new (self).inner.ptr;UniqueArc {// SAFETY: The new `Arc` is taking over `ptr` from `self.inner` (which won't be
--
2.39.2
PATCH v4 5/15 rust: types: add `Opaque::raw_get`
In reply to [email protected]From: Benno Lossin <[email protected]>
This function mirrors UnsafeCell::raw_get. It avoids creating a reference and allows solely using raw pointers.
The pin-init API will be using this, since uninitialized memory requires raw pointers.
--- rust/kernel/types.rs | 8 ++++++++ 1 file changed, 8 insertions(+)
rust/kernel/types.rs +8 -0
@@ -238,6 +238,14 @@ impl<T> Opaque<T> {pub fn get (&self) -> *mut T {UnsafeCell ::raw_get (self.0 .as_ptr ()) } + +/// Gets the value behind `this`. +/// +/// This function is useful to get access to the value without creating intermediate +/// references. +pub const fn raw_get (this: *const Self ) -> *mut T { +UnsafeCell ::raw_get (this.cast ::<UnsafeCell <T >>()) + } }/// A sum type that always holds either a value of type `L` or `R`.
--
2.39.2
PATCH v4 6/15 rust: add pin-init API core
In reply to [email protected]From: Benno Lossin <[email protected]>
This API is used to facilitate safe pinned initialization of structs. It replaces cumbersome unsafe manual initialization with elegant safe macro invocations.
Due to the size of this change it has been split into six commits:
1. This commit introducing the basic public interface: traits and functions to represent and create initializers.
2. Adds the #[pin_data], pin_init!, try_pin_init!, init! and try_init! macros along with their internal types.
3. Adds the InPlaceInit trait that allows using an initializer to create an object inside of a Box<T> and other smart pointers.
4. Adds the PinnedDrop trait and adds macro support for it in the #[pin_data] macro.
5. Adds the stack_pin_init! macro allowing to pin-initialize a struct on the stack.
6. Adds the Zeroable trait and init::zeroed function to initialize types that have 0x00 in all bytes as a valid bit pattern.
--
In this section the problem that the new pin-init API solves is outlined. This message describes the entirety of the API, not just the parts introduced in this commit. For a more granular explanation and additional information on pinning and this issue, view [1].
Pinning is Rust's way of enforcing the address stability of a value. When a value gets pinned it will be impossible for safe code to move it to another location. This is done by wrapping pointers to said object with Pin<P>. This wrapper prevents safe code from creating mutable references to the object, preventing mutable access, which is needed to move the value. Pin<P> provides unsafe functions to circumvent this and allow modifications regardless. It is then the programmer's responsibility to uphold the pinning guarantee.
Many kernel data structures require a stable address, because there are foreign pointers to them which would get invalidated by moving the structure. Since these data structures are usually embedded in structs to use them, this pinning property propagates to the container struct. Resulting in most structs in both Rust and C code needing to be pinned.
So if we want to have a mutex field in a Rust struct, this struct also needs to be pinned, because a mutex contains a list_head. Additionally initializing a list_head requires already having the final memory location available, because it is initialized by pointing it to itself. But this presents another challenge in Rust: values have to be initialized at all times. There is the MaybeUninit<T> wrapper type, which allows handling uninitialized memory, but this requires using the unsafe raw pointers and a casting the type to the initialized variant.
This problem gets exacerbated when considering encapsulation and the normal safety requirements of Rust code. The fields of the Rust Mutex<T> should not be accessible to normal driver code. After all if anyone can modify the fields, there is no way to ensure the invariants of the Mutex<T> are upheld. But if the fields are inaccessible, then initialization of a Mutex<T> needs to be somehow achieved via a function or a macro. Because the Mutex<T> must be pinned in memory, the function cannot return it by value. It also cannot allocate a Box to put the Mutex<T> into, because that is an unnecessary allocation and indirection which would hurt performance.
The current solution was to split this function into two parts:
1. A new function that returns a partially initialized Mutex<T>,
2. An init function that requires the Mutex<T> to be pinned and that fully initializes the Mutex<T>.
Both of these functions have to be marked unsafe, since a call to new needs to be accompanied with a call to init, otherwise using the Mutex<T> could result in UB. And because calling init twice also is not safe. While Mutex<T> initialization cannot fail, other structs might also have to allocate memory, which would result in conditional successful initialization requiring even more manual accommodation work.
Combine this with the problem of pin-projections -- the way of accessing fields of a pinned struct -- which also have an unsafe API, pinned initialization is riddled with unsafe resulting in very poor ergonomics. Not only that, but also having to call two functions possibly multiple lines apart makes it very easy to forget it outright or during refactoring.
Here is an example of the current way of initializing a struct with two synchronization primitives (see [2] for the full example):
struct SharedState {
state_changed: CondVar,
inner: Mutex<SharedStateInner>,
}
impl SharedState {
fn try_new() -> Result<Arc<Self>> {
let mut state = Pin::from(UniqueArc::try_new(Self {
// SAFETY: `condvar_init!` is called below.
state_changed: unsafe { CondVar::new() },
// SAFETY: `mutex_init!` is called below.
inner: unsafe {
Mutex::new(SharedStateInner { token_count: 0 })
},
})?);
// SAFETY: `state_changed` is pinned when `state` is.
let pinned = unsafe {
state.as_mut().map_unchecked_mut(|s| &mut s.state_changed)
};
kernel::condvar_init!(pinned, "SharedState::state_changed");
// SAFETY: `inner` is pinned when `state` is.
let pinned = unsafe {
state.as_mut().map_unchecked_mut(|s| &mut s.inner)
};
kernel::mutex_init!(pinned, "SharedState::inner");
Ok(state.into())
}
}The pin-init API of this patch solves this issue by providing a comprehensive solution comprised of macros and traits. Here is the example from above using the pin-init API:
#[pin_data]
struct SharedState {
#[pin]
state_changed: CondVar,
#[pin]
inner: Mutex<SharedStateInner>,
}
impl SharedState {
fn new() -> impl PinInit<Self> {
pin_init!(Self {
state_changed <- new_condvar!("SharedState::state_changed"),
inner <- new_mutex!(
SharedStateInner { token_count: 0 },
"SharedState::inner",
),
})
}
}Notably the way the macro is used here requires no unsafe and thus comes with the usual Rust promise of safe code not introducing any memory violations. Additionally it is now up to the caller of new() to decide the memory location of the SharedState. They can choose at the moment Arc<T>, Box<T> or the stack.
--
The API has the following architecture:
1. Initializer traits PinInit<T, E> and Init<T, E> that act like closures.
2. Macros to create these initializer traits safely.
3. Functions to allow manually writing initializers.
The initializers (an impl PinInit<T, E>) receive a raw pointer pointing to uninitialized memory and their job is to fully initialize a T at that location. If initialization fails, they return an error (E) by value.
This way of initializing cannot be safely exposed to the user, since it relies upon these properties outside of the control of the trait:
- the memory location (slot) needs to be valid memory,
- if initialization fails, the slot should not be read from,
- the value in the slot should be pinned, so it cannot move and the memory cannot be deallocated until the value is dropped.
This is why using an initializer is facilitated by another trait that ensures these requirements.
These initializers can be created manually by just supplying a closure that fulfills the same safety requirements as PinInit<T, E>. But this is an unsafe operation. To allow safe initializer creation, the pin_init! is provided along with three other variants: try_pin_init!, try_init! and init!. These take a modified struct initializer as a parameter and generate a closure that initializes the fields in sequence.
The macros take great care in upholding the safety requirements:
- A shadowed struct type is used as the return type of the closure instead of (). This is to prevent early returns, as these would prevent full initialization.
- To ensure every field is only initialized once, a normal struct initializer is placed in unreachable code. The type checker will emit errors if a field is missing or specified multiple times.
- When initializing a field fails, the whole initializer will fail and automatically drop fields that have been initialized earlier.
- Only the correct initializer type is allowed for unpinned fields. You cannot use a impl PinInit<T, E> to initialize a structurally not pinned field.
To ensure the last point, an additional macro #[pin_data] is needed. This macro annotates the struct itself and the user specifies structurally pinned and not pinned fields.
Because dropping a pinned struct is also not allowed to break the pinning invariants, another macro attribute #[pinned_drop] is needed. This macro is introduced in a following commit.
These two macros also have mechanisms to ensure the overall safety of the API. Additionally, they utilize a combined proc-macro, declarative macro
design: first a proc-macro enables the outer attribute syntax #[...] and does some important pre-parsing. Notably this prepares the generics such that the declarative macro can handle them using token trees. Then the actual parsing of the structure and the emission of code is handled by a declarative macro.
For pin-projections the crates pin-project [3] and pin-project-lite [4] had been considered, but were ultimately rejected:
- pin-project depends on syn [5] which is a very big dependency, around 50k lines of code.
- pin-project-lite is a more reasonable 5k lines of code, but contains a very complex declarative macro to parse generics. On top of that it would require modification that would need to be maintained independently.
--- rust/kernel/init.rs | 183 +++++++++++++++++++++++++++++++++ rust/kernel/init/__internal.rs | 45 ++++++++ rust/kernel/lib.rs | 7 ++ scripts/Makefile.build | 2 +- 4 files changed, 236 insertions(+), 1 deletion(-) create mode 100644 rust/kernel/init.rs create mode 100644 rust/kernel/init/__internal.rs
rust/kernel/init.rs new file +183 -0
@@ -0,0 +1,183 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +//! API to safely and fallibly initialize pinned `struct`s using in-place constructors. +//! +//! It also allows in-place initialization of big `struct`s that would otherwise produce a stack +//! overflow. +//! +//! Most `struct`s from the [`sync`] module need to be pinned, because they contain self-referential +//! `struct`s from C. [Pinning][pinning] is Rust's way of ensuring data does not move. +//! +//! # Overview +//! +//! To initialize a `struct` with an in-place constructor you will need two things: +//! - an in-place constructor, +//! - a memory location that can hold your `struct`. +//! +//! To get an in-place constructor there are generally two options: +//! - a custom function/macro returning an in-place constructor provided by someone else, +//! - using the unsafe function [`pin_init_from_closure()`] to manually create an initializer. +//! +//! Aside from pinned initialization, this API also supports in-place construction without pinning, +//! the macros/types/functions are generally named like the pinned variants without the `pin` +//! prefix. +//! +//! [`sync`]: kernel::sync +//! [pinning]: https://doc.rust-lang.org/std/pin/index.html +//! [structurally pinned fields]: +//! https://doc.rust-lang.org/std/pin/index.html#pinning-is-structural-for-field +//! [`Arc<T>`]: crate::sync::Arc +//! [`impl PinInit<Foo>`]: PinInit +//! [`impl PinInit<T, E>`]: PinInit +//! [`impl Init<T, E>`]: Init +//! [`Opaque`]: kernel::types::Opaque +//! [`pin_data`]: ::macros::pin_data +//! [`UniqueArc<T>`]: kernel::sync::UniqueArc +//! [`Box<T>`]: alloc::boxed::Box + +use core::{convert::Infallible , marker::PhantomData , mem::MaybeUninit }; + +#[doc(hidden)] +pub mod __internal; + +/// A pin-initializer for the type `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`pin_init_from_closure`] where possible. +/// +/// The [`PinInit::__pinned_init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`Arc::pin_init`]: crate::sync::Arc::pin_init +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ] +pub unsafe trait PinInit <T : ?Sized ,E =Infallible >:Sized { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +/// - `slot` will not move until it is dropped, i.e. it will be pinned. +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),E >; +} + +/// An initializer for `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. Because [`PinInit<T, E>`] is a super trait, you can +/// use every function that takes it as well. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`init_from_closure`] where possible. +/// +/// The [`Init::__init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// The `__pinned_init` function from the supertrait [`PinInit`] needs to execute the exact same +/// code as `__init`. +/// +/// Contrary to its supertype [`PinInit<T, E>`] the caller is allowed to +/// move the pointee after initialization. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ] +pub unsafe trait Init <T : ?Sized ,E =Infallible >:PinInit <T ,E > { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +unsafe fn __init (self, slot: *mut T ) ->Result <(),E >; +} + +/// Creates a new [`PinInit<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - may assume that the `slot` does not move if `T: !Unpin`, +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn pin_init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl PinInit <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// Creates a new [`Init<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - the `slot` may move after initialization. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl Init <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// An initializer that leaves the memory uninitialized. +/// +/// The initializer is a no-op. The `slot` memory is not changed. +#[inline] +pub fn uninit <T ,E >() ->impl Init <MaybeUninit <T >,E > { +// SAFETY: The memory is allowed to be uninitialized. +unsafe {init_from_closure (|_|Ok (())) } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >PinInit <T >for T { +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >Init <T >for T { +unsafe fn __init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +}
rust/kernel/init/__internal.rs new file +45 -0
45 lines not on this page · raw message
rust/kernel/lib.rs +7 -0
@@ -16,7 +16,9 @@#![feature(coerce_unsized)] #![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] +#![feature(explicit_generic_args_with_impl_trait)] #![feature(generic_associated_types)] +#![feature(new_uninit)] #![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)]
@@ -26,11 +28,16 @@#[cfg(not( CONFIG_RUST ))] compile_error! ("Missing kernel configuration for conditional compilation" ); +#[allow(unused_extern_crates)] +// Allow proc-macros to refer to `::kernel` inside the `kernel` crate (this crate). +extern crate selfas kernel; +#[cfg(not(test))] #[cfg(not(testlib))] mod allocator;mod build_assert;pub mod error; +pub mod init;pub mod prelude;pub mod print;mod static_assert;
scripts/Makefile.build +1 -1
@@ -277,7 +277,7 @@ $(obj)/%.lst: $(src)/%.c FORCE # Compile Rust sources (.rs) # --------------------------------------------------------------------------- -rust_allowed_features := core_ffi_c +rust_allowed_features := core_ffi_c,explicit_generic_args_with_impl_trait rust_common_cmd = \ RUST_MODFILE=$(modfile) $(RUSTC_OR_CLIPPY) $(rust_flags) \
--
2.39.2
Re: [PATCH v4 06/15] rust: add pin-init API core
Quoted 62 lines · On Fri, Mar 31, 2023 at 09:52:59PM +0000, [email protected] wrote:
From: Benno Lossin <[email protected]>
This API is used to facilitate safe pinned initialization of structs. It replaces cumbersome
unsafemanual initialization with elegant safe macro invocations.Due to the size of this change it has been split into six commits:
1. This commit introducing the basic public interface: traits and functions to represent and create initializers.
2. Adds the#[pin_data],pin_init!,try_pin_init!,init!andtry_init!macros along with their internal types.
3. Adds theInPlaceInittrait that allows using an initializer to create an object inside of aBox<T>and other smart pointers.
4. Adds thePinnedDroptrait and adds macro support for it in the#[pin_data]macro.
5. Adds thestack_pin_init!macro allowing to pin-initialize a struct on the stack.
6. Adds theZeroabletrait andinit::zeroedfunction to initialize types that have0x00in all bytes as a valid bit pattern.--
In this section the problem that the new pin-init API solves is outlined. This message describes the entirety of the API, not just the parts introduced in this commit. For a more granular explanation and additional information on pinning and this issue, view [1].
Pinning is Rust's way of enforcing the address stability of a value. When a value gets pinned it will be impossible for safe code to move it to another location. This is done by wrapping pointers to said object with
Pin<P>. This wrapper prevents safe code from creating mutable references to the object, preventing mutable access, which is needed to move the value.Pin<P>providesunsafefunctions to circumvent this and allow modifications regardless. It is then the programmer's responsibility to uphold the pinning guarantee.Many kernel data structures require a stable address, because there are foreign pointers to them which would get invalidated by moving the structure. Since these data structures are usually embedded in structs to use them, this pinning property propagates to the container struct. Resulting in most structs in both Rust and C code needing to be pinned.
So if we want to have a
mutexfield in a Rust struct, this struct also needs to be pinned, because amutexcontains alist_head. Additionally initializing alist_headrequires already having the final memory location available, because it is initialized by pointing it to itself. But this presents another challenge in Rust: values have to be initialized at all times. There is theMaybeUninit<T>wrapper type, which allows handling uninitialized memory, but this requires using theunsaferaw pointers and a casting the type to the initialized variant.This problem gets exacerbated when considering encapsulation and the normal safety requirements of Rust code. The fields of the Rust
Mutex<T>should not be accessible to normal driver code. After all if anyone can modify the fields, there is no way to ensure the invariants of theMutex<T>are upheld. But if the fields are inaccessible, then initialization of aMutex<T>needs to be somehow achieved via a function or a macro. Because theMutex<T>must be pinned in memory, the function cannot return it by value. It also cannot allocate aBoxto put theMutex<T>into, because that is an unnecessary allocation and indirection which would hurt performance.
Quoted 1 line
The current solution was to split this function into two parts:
nit: since this is the only solution that is going upstream, perhaps here we should say, the "current solution in the the rust tree" or something to that effect so that a reader with less context would know what we're talking about.
Quoted 149 lines
1. A
newfunction that returns a partially initializedMutex<T>,
2. Aninitfunction that requires theMutex<T>to be pinned and that fully initializes theMutex<T>.Both of these functions have to be marked
unsafe, since a call tonewneeds to be accompanied with a call toinit, otherwise using theMutex<T>could result in UB. And because callinginittwice also is not safe. WhileMutex<T>initialization cannot fail, other structs might also have to allocate memory, which would result in conditional successful initialization requiring even more manual accommodation work.Combine this with the problem of pin-projections -- the way of accessing fields of a pinned struct -- which also have an
unsafeAPI, pinned initialization is riddled withunsaferesulting in very poor ergonomics. Not only that, but also having to call two functions possibly multiple lines apart makes it very easy to forget it outright or during refactoring.Here is an example of the current way of initializing a struct with two synchronization primitives (see [2] for the full example):
struct SharedState { state_changed: CondVar, inner: Mutex<SharedStateInner>, } impl SharedState { fn try_new() -> Result<Arc<Self>> { let mut state = Pin::from(UniqueArc::try_new(Self { // SAFETY: `condvar_init!` is called below. state_changed: unsafe { CondVar::new() }, // SAFETY: `mutex_init!` is called below. inner: unsafe { Mutex::new(SharedStateInner { token_count: 0 }) }, })?); // SAFETY: `state_changed` is pinned when `state` is. let pinned = unsafe { state.as_mut().map_unchecked_mut(|s| &mut s.state_changed) }; kernel::condvar_init!(pinned, "SharedState::state_changed"); // SAFETY: `inner` is pinned when `state` is. let pinned = unsafe { state.as_mut().map_unchecked_mut(|s| &mut s.inner) }; kernel::mutex_init!(pinned, "SharedState::inner"); Ok(state.into()) } }The pin-init API of this patch solves this issue by providing a comprehensive solution comprised of macros and traits. Here is the example from above using the pin-init API:
#[pin_data] struct SharedState { #[pin] state_changed: CondVar, #[pin] inner: Mutex<SharedStateInner>, } impl SharedState { fn new() -> impl PinInit<Self> { pin_init!(Self { state_changed <- new_condvar!("SharedState::state_changed"), inner <- new_mutex!( SharedStateInner { token_count: 0 }, "SharedState::inner", ), }) } }Notably the way the macro is used here requires no
unsafeand thus comes with the usual Rust promise of safe code not introducing any memory violations. Additionally it is now up to the caller ofnew()to decide the memory location of theSharedState. They can choose at the momentArc<T>,Box<T>or the stack.--
The API has the following architecture:
1. Initializer traitsPinInit<T, E>andInit<T, E>that act like closures.
2. Macros to create these initializer traits safely.
3. Functions to allow manually writing initializers.The initializers (an
impl PinInit<T, E>) receive a raw pointer pointing to uninitialized memory and their job is to fully initialize aTat that location. If initialization fails, they return an error (E) by value.This way of initializing cannot be safely exposed to the user, since it relies upon these properties outside of the control of the trait:
- the memory location (slot) needs to be valid memory,
- if initialization fails, the slot should not be read from,
- the value in the slot should be pinned, so it cannot move and the memory cannot be deallocated until the value is dropped.This is why using an initializer is facilitated by another trait that ensures these requirements.
These initializers can be created manually by just supplying a closure that fulfills the same safety requirements as
PinInit<T, E>. But this is anunsafeoperation. To allow safe initializer creation, thepin_init!is provided along with three other variants:try_pin_init!,try_init!andinit!. These take a modified struct initializer as a parameter and generate a closure that initializes the fields in sequence.
The macros take great care in upholding the safety requirements:
- A shadowed struct type is used as the return type of the closure instead of(). This is to prevent early returns, as these would prevent full initialization.
- To ensure every field is only initialized once, a normal struct initializer is placed in unreachable code. The type checker will emit errors if a field is missing or specified multiple times.
- When initializing a field fails, the whole initializer will fail and automatically drop fields that have been initialized earlier.
- Only the correct initializer type is allowed for unpinned fields. You cannot use aimpl PinInit<T, E>to initialize a structurally not pinned field.To ensure the last point, an additional macro
#[pin_data]is needed. This macro annotates the struct itself and the user specifies structurally pinned and not pinned fields.Because dropping a pinned struct is also not allowed to break the pinning invariants, another macro attribute
#[pinned_drop]is needed. This macro is introduced in a following commit.These two macros also have mechanisms to ensure the overall safety of the API. Additionally, they utilize a combined proc-macro, declarative macro
design: first a proc-macro enables the outer attribute syntax#[...]and does some important pre-parsing. Notably this prepares the generics such that the declarative macro can handle them using token trees. Then the actual parsing of the structure and the emission of code is handled by a declarative macro.For pin-projections the crates
pin-project[3] andpin-project-lite[4] had been considered, but were ultimately rejected:
-pin-projectdepends onsyn[5] which is a very big dependency, around 50k lines of code.
-pin-project-liteis a more reasonable 5k lines of code, but contains a very complex declarative macro to parse generics. On top of that it would require modification that would need to be maintained independently.
Quoted 10 lines
Link https://crates.io/crates/pin-project [3]Link https://crates.io/crates/syn [5]Co-developed-by Gary Guo <[email protected]>Signed-off-by Gary Guo <[email protected]>Signed-off-by Benno Lossin <[email protected]>Cc Andreas Hindborg <[email protected]>Cc Alice Ryhl <[email protected]>
Quoted 125 lines
--- rust/kernel/init.rs | 183 +++++++++++++++++++++++++++++++++ rust/kernel/init/__internal.rs | 45 ++++++++ rust/kernel/lib.rs | 7 ++ scripts/Makefile.build | 2 +- 4 files changed, 236 insertions(+), 1 deletion(-) create mode 100644 rust/kernel/init.rs create mode 100644 rust/kernel/init/__internal.rsrust/kernel/init.rs new file +110
-0@@ -0,0 +1,183 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +//! API to safely and fallibly initialize pinned `struct`s using in-place constructors. +//! +//! It also allows in-place initialization of big `struct`s that would otherwise produce a stack +//! overflow. +//! +//! Most `struct`s from the [`sync`] module need to be pinned, because they contain self-referential +//! `struct`s from C. [Pinning][pinning] is Rust's way of ensuring data does not move. +//! +//! # Overview +//! +//! To initialize a `struct` with an in-place constructor you will need two things: +//! - an in-place constructor, +//! - a memory location that can hold your `struct`. +//! +//! To get an in-place constructor there are generally two options: +//! - a custom function/macro returning an in-place constructor provided by someone else, +//! - using the unsafe function [`pin_init_from_closure()`] to manually create an initializer. +//! +//! Aside from pinned initialization, this API also supports in-place construction without pinning, +//! the macros/types/functions are generally named like the pinned variants without the `pin` +//! prefix. +//! +//! [`sync`]: kernel::sync +//! [pinning]: https://doc.rust-lang.org/std/pin/index.html +//! [structurally pinned fields]: +//! https://doc.rust-lang.org/std/pin/index.html#pinning-is-structural-for-field +//! [`Arc<T>`]: crate::sync::Arc +//! [`impl PinInit<Foo>`]: PinInit +//! [`impl PinInit<T, E>`]: PinInit +//! [`impl Init<T, E>`]: Init +//! [`Opaque`]: kernel::types::Opaque +//! [`pin_data`]: ::macros::pin_data +//! [`UniqueArc<T>`]: kernel::sync::UniqueArc +//! [`Box<T>`]: alloc::boxed::Box + +use core::{convert::Infallible , marker::PhantomData , mem::MaybeUninit }; + +#[doc(hidden)] +pub mod __internal; + +/// A pin-initializer for the type `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`pin_init_from_closure`] where possible. +/// +/// The [`PinInit::__pinned_init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`Arc::pin_init`]: crate::sync::Arc::pin_init +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ] +pub unsafe trait PinInit <T : ?Sized ,E =Infallible >:Sized { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +/// - `slot` will not move until it is dropped, i.e. it will be pinned. +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),E >; +} + +/// An initializer for `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. Because [`PinInit<T, E>`] is a super trait, you can +/// use every function that takes it as well. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`init_from_closure`] where possible. +/// +/// The [`Init::__init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// The `__pinned_init` function from the supertrait [`PinInit`] needs to execute the exact same +/// code as `__init`. +/// +/// Contrary to its supertype [`PinInit<T, E>`] the caller is allowed to +/// move the pointee after initialization. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ]
rust/kernel/init.rs line 120
Quoted 10 lines
rust/kernel/init.rs new file +10
-0+pub unsafe trait Init <T : ?Sized ,E =Infallible >:PinInit <T ,E > { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +unsafe fn __init (self, slot: *mut T ) ->Result <(),E >; +}
Rather than making PinInit<T, E> a supertrait here, wouldn't it make sense to just implement PinInit<T, E> for every type that implements Init<T, E>? This way we could relax the unsafe requirement that "__pinned_init needs to mimick __init" by simply making it so. Something like (plus a proper safety comment):
unsafe impl<T, U: ?Sized, E> PinInit<U, E> for T where T: Init<U, E> {
unsafe fn __pinned_init(self, slot: *mut U) -> Result<(), E> {
unsafe { T::__init(self, slot) }
}
}Quoted 104 lines
rust/kernel/init.rs new file +63
-0+ +/// Creates a new [`PinInit<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - may assume that the `slot` does not move if `T: !Unpin`, +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn pin_init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl PinInit <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// Creates a new [`Init<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - the `slot` may move after initialization. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl Init <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// An initializer that leaves the memory uninitialized. +/// +/// The initializer is a no-op. The `slot` memory is not changed. +#[inline] +pub fn uninit <T ,E >() ->impl Init <MaybeUninit <T >,E > { +// SAFETY: The memory is allowed to be uninitialized. +unsafe {init_from_closure (|_|Ok (())) } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >PinInit <T >for T { +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >Init <T >for T { +unsafe fn __init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +}rust/kernel/init/__internal.rs new file +35
-0@@ -0,0 +1,45 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +//! This module contains API-internal items for pin-init. +//! +//! These items must not be used outside of +//! - `kernel/init.rs` +//! - `macros/pin_data.rs` +//! - `macros/pinned_drop.rs` + +use super ::*; + +/// See the [nomicon] for what subtyping is. See also [this table]. +/// +/// [nomicon]: https://doc.rust-lang.org/nomicon/subtyping.html +/// [this table]: https://doc.rust-lang.org/nomicon/phantom-data.html#table-of-phantomdata-patterns +type Invariant <T > =PhantomData <fn (*mut T ) -> *mut T >; + +/// This is the module-internal type implementing `PinInit` and `Init`. It is unsafe to create this +/// type, since the closure needs to fulfill the same safety requirement as the +/// `__pinned_init`/`__init` functions. +pub (crate )struct InitClosure <F ,T : ?Sized ,E >(pub (crate )F ,pub (crate )Invariant <(E ,T )>); + +// SAFETY: While constructing the `InitClosure`, the user promised that it upholds the +// `__pinned_init` invariants. +unsafe impl <T : ?Sized ,F ,E >PinInit <T ,E >for InitClosure <F ,T ,E > +where +F :FnOnce (*mut T ) ->Result <(),E >, +{ +#[inline] +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),E > { + (self.0 )(slot) + } +} + +// SAFETY: While constructing the `InitClosure`, the user promised that it upholds the
rust/kernel/init/__internal.rs line 45
Quoted 10 lines
rust/kernel/init/__internal.rs new file +10
-0+// `__init` invariants. +unsafe impl <T : ?Sized ,F ,E >Init <T ,E >for InitClosure <F ,T ,E > +where +F :FnOnce (*mut T ) ->Result <(),E >, +{ +#[inline] +unsafe fn __init (self, slot: *mut T ) ->Result <(),E > { + (self.0 )(slot) + } +}
Since these are only used to implement init_from_closure and pin_init_from_closure, would it make sense to move them into init_from_closure, and have pin_init_from_closure just call init_from_closure (given that if a type implements Init<T, E>, it also implements PinInit<T, E>)?
Quoted 48 lines
rust/kernel/lib.rs +7
-0@@ -16,7 +16,9 @@#![feature(coerce_unsized)] #![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] +#![feature(explicit_generic_args_with_impl_trait)] #![feature(generic_associated_types)] +#![feature(new_uninit)] #![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)] @@ -26,11 +28,16 @@#[cfg(not( CONFIG_RUST ))] compile_error! ("Missing kernel configuration for conditional compilation" ); +#[allow(unused_extern_crates)] +// Allow proc-macros to refer to `::kernel` inside the `kernel` crate (this crate). +extern crate selfas kernel; +#[cfg(not(test))] #[cfg(not(testlib))] mod allocator;mod build_assert;pub mod error; +pub mod init;pub mod prelude;pub mod print;mod static_assert;scripts/Makefile.build +1
-2@@ -277,7 +277,7 @@ $(obj)/%.lst: $(src)/%.c FORCE # Compile Rust sources (.rs) # --------------------------------------------------------------------------- -rust_allowed_features := core_ffi_c +rust_allowed_features := core_ffi_c,explicit_generic_args_with_impl_trait rust_common_cmd = \ RUST_MODFILE=$(modfile) $(RUSTC_OR_CLIPPY) $(rust_flags) \ --2.39.2
Re: [PATCH v4 06/15] rust: add pin-init API core
Quoted 63 lines · Wedson Almeida Filho
Quoted 62 lines · On Fri, Mar 31, 2023 at 09:52:59PM +0000, [email protected] wrote:
From: Benno Lossin <[email protected]>
This API is used to facilitate safe pinned initialization of structs. It replaces cumbersome
unsafemanual initialization with elegant safe macro invocations.Due to the size of this change it has been split into six commits:
1. This commit introducing the basic public interface: traits and functions to represent and create initializers.
2. Adds the#[pin_data],pin_init!,try_pin_init!,init!andtry_init!macros along with their internal types.
3. Adds theInPlaceInittrait that allows using an initializer to create an object inside of aBox<T>and other smart pointers.
4. Adds thePinnedDroptrait and adds macro support for it inthe `#[pin_data]` macro. 5. Adds the `stack_pin_init!` macro allowing to pin-initialize a struct on the stack. 6. Adds the `Zeroable` trait and `init::zeroed` function to initialize types that have `0x00` in all bytes as a valid bit pattern.--
In this section the problem that the new pin-init API solves is outlined. This message describes the entirety of the API, not just the parts introduced in this commit. For a more granular explanation and additional information on pinning and this issue, view [1].
Pinning is Rust's way of enforcing the address stability of a value. When a value gets pinned it will be impossible for safe code to move it to another location. This is done by wrapping pointers to said object with
Pin<P>. This wrapper prevents safe code from creating mutable references to the object, preventing mutable access, which is needed to move the value.Pin<P>providesunsafefunctions to circumvent this and allow modifications regardless. It is then the programmer's responsibility to uphold the pinning guarantee.Many kernel data structures require a stable address, because there are foreign pointers to them which would get invalidated by moving the structure. Since these data structures are usually embedded in structs to use them, this pinning property propagates to the container struct. Resulting in most structs in both Rust and C code needing to be pinned.
So if we want to have a
mutexfield in a Rust struct, this struct also needs to be pinned, because amutexcontains alist_head. Additionally initializing alist_headrequires already having the final memory location available, because it is initialized by pointing it to itself. But this presents another challenge in Rust: values have to be initialized at all times. There is theMaybeUninit<T>wrapper type, which allows handling uninitialized memory, but this requires using theunsaferaw pointers and a casting the type to the initialized variant.This problem gets exacerbated when considering encapsulation and the normal safety requirements of Rust code. The fields of the Rust
Mutex<T>should not be accessible to normal driver code. After all if anyone can modify the fields, there is no way to ensure the invariants of theMutex<T>are upheld. But if the fields are inaccessible, then initialization of aMutex<T>needs to be somehow achieved via a function or a macro. Because theMutex<T>must be pinned in memory, the function cannot return it by value. It also cannot allocate aBoxto put theMutex<T>into, because that is an unnecessary allocation and indirection which would hurt performance.
Quoted 5 lines
Quoted 1 line
The current solution was to split this function into two parts:
nit: since this is the only solution that is going upstream, perhaps here we should say, the "current solution in the the
rusttree" or something to that effect so that a reader with less context would know what we're talking abo
Sure.
Quoted 303 lines
Quoted 149 lines
1. A
newfunction that returns a partially initializedMutex<T>,
2. Aninitfunction that requires theMutex<T>to be pinned and that fully initializes theMutex<T>.Both of these functions have to be marked
unsafe, since a call tonewneeds to be accompanied with a call toinit, otherwise using theMutex<T>could result in UB. And because callinginittwice also is not safe. WhileMutex<T>initialization cannot fail, other structs might also have to allocate memory, which would result in conditional successful initialization requiring even more manual accommodation work.Combine this with the problem of pin-projections -- the way of accessing fields of a pinned struct -- which also have an
unsafeAPI, pinned initialization is riddled withunsaferesulting in very poor ergonomics. Not only that, but also having to call two functions possibly multiple lines apart makes it very easy to forget it outright or during refactoring.Here is an example of the current way of initializing a struct with two synchronization primitives (see [2] for the full example):
struct SharedState { state_changed: CondVar, inner: Mutex<SharedStateInner>, } impl SharedState { fn try_new() -> Result<Arc<Self>> { let mut state = Pin::from(UniqueArc::try_new(Self { // SAFETY: `condvar_init!` is called below. state_changed: unsafe { CondVar::new() }, // SAFETY: `mutex_init!` is called below. inner: unsafe { Mutex::new(SharedStateInner { token_count: 0 }) }, })?); // SAFETY: `state_changed` is pinned when `state` is. let pinned = unsafe { state.as_mut().map_unchecked_mut(|s| &mut s.state_changed) }; kernel::condvar_init!(pinned, "SharedState::state_changed"); // SAFETY: `inner` is pinned when `state` is. let pinned = unsafe { state.as_mut().map_unchecked_mut(|s| &mut s.inner) }; kernel::mutex_init!(pinned, "SharedState::inner"); Ok(state.into()) } }The pin-init API of this patch solves this issue by providing a comprehensive solution comprised of macros and traits. Here is the example from above using the pin-init API:
#[pin_data] struct SharedState { #[pin] state_changed: CondVar, #[pin] inner: Mutex<SharedStateInner>, } impl SharedState { fn new() -> impl PinInit<Self> { pin_init!(Self { state_changed <- new_condvar!("SharedState::state_changed"), inner <- new_mutex!( SharedStateInner { token_count: 0 }, "SharedState::inner", ), }) } }Notably the way the macro is used here requires no
unsafeand thus comes with the usual Rust promise of safe code not introducing any memory violations. Additionally it is now up to the caller ofnew()to decide the memory location of theSharedState. They can choose at the momentArc<T>,Box<T>or the stack.--
The API has the following architecture:
1. Initializer traitsPinInit<T, E>andInit<T, E>that act like closures.
2. Macros to create these initializer traits safely.
3. Functions to allow manually writing initializers.The initializers (an
impl PinInit<T, E>) receive a raw pointer pointing to uninitialized memory and their job is to fully initialize aTat that location. If initialization fails, they return an error (E) by value.This way of initializing cannot be safely exposed to the user, since it relies upon these properties outside of the control of the trait:
- the memory location (slot) needs to be valid memory,
- if initialization fails, the slot should not be read from,
- the value in the slot should be pinned, so it cannot move and the memory cannot be deallocated until the value is dropped.This is why using an initializer is facilitated by another trait that ensures these requirements.
These initializers can be created manually by just supplying a closure that fulfills the same safety requirements as
PinInit<T, E>. But this is anunsafeoperation. To allow safe initializer creation, thepin_init!is provided along with three other variants:try_pin_init!,try_init!andinit!. These take a modified struct initializer as a parameter and generate a closure that initializes the fields in sequence.
The macros take great care in upholding the safety requirements:
- A shadowed struct type is used as the return type of the closure instead of(). This is to prevent early returns, as these would prevent full initialization.
- To ensure every field is only initialized once, a normal struct initializer is placed in unreachable code. The type checker will emit errors if a field is missing or specified multiple times.
- When initializing a field fails, the whole initializer will fail and automatically drop fields that have been initialized earlier.
- Only the correct initializer type is allowed for unpinned fields. You cannot use aimpl PinInit<T, E>to initialize a structurally not pinned field.To ensure the last point, an additional macro
#[pin_data]is needed. This macro annotates the struct itself and the user specifies structurally pinned and not pinned fields.Because dropping a pinned struct is also not allowed to break the pinning invariants, another macro attribute
#[pinned_drop]is needed. This macro is introduced in a following commit.These two macros also have mechanisms to ensure the overall safety of the API. Additionally, they utilize a combined proc-macro, declarative macro
design: first a proc-macro enables the outer attribute syntax#[...]and does some important pre-parsing. Notably this prepares the generics such that the declarative macro can handle them using token trees. Then the actual parsing of the structure and the emission of code is handled by a declarative macro.For pin-projections the crates
pin-project[3] andpin-project-lite[4] had been considered, but were ultimately rejected:
-pin-projectdepends onsyn[5] which is a very big dependency, around 50k lines of code.
-pin-project-liteis a more reasonable 5k lines of code, but contains a very complex declarative macro to parse generics. On top of that it would require modification that would need to be maintained independently.Quoted 10 lines
Link https://crates.io/crates/pin-project [3]Link https://crates.io/crates/syn [5]Co-developed-by Gary Guo <[email protected]>Signed-off-by Gary Guo <[email protected]>Signed-off-by Benno Lossin <[email protected]>Cc Andreas Hindborg <[email protected]>Cc Alice Ryhl <[email protected]>Reviewed-by Wedson Almeida Filho <[email protected]>Quoted 125 lines
--- rust/kernel/init.rs | 183 +++++++++++++++++++++++++++++++++ rust/kernel/init/__internal.rs | 45 ++++++++ rust/kernel/lib.rs | 7 ++ scripts/Makefile.build | 2 +- 4 files changed, 236 insertions(+), 1 deletion(-)create mode 100644
rust/kernel/init.rscreate mode 100644rust/kernel/init/__internal.rsrust/kernel/init.rs new file +110
-0@@ -0,0 +1,183 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +//! API to safely and fallibly initialize pinned `struct`s using in-place constructors. +//! +//! It also allows in-place initialization of big `struct`s that would otherwise produce a stack +//! overflow. +//! +//! Most `struct`s from the [`sync`] module need to be pinned, because they contain self-referential +//! `struct`s from C. [Pinning][pinning] is Rust's way of ensuring data does not move. +//! +//! # Overview +//! +//! To initialize a `struct` with an in-place constructor you will need two things: +//! - an in-place constructor, +//! - a memory location that can hold your `struct`. +//! +//! To get an in-place constructor there are generally two options: +//! - a custom function/macro returning an in-place constructor provided by someone else, +//! - using the unsafe function [`pin_init_from_closure()`] to manually create an initializer. +//! +//! Aside from pinned initialization, this API also supports in-place construction without pinning, +//! the macros/types/functions are generally named like the pinned variants without the `pin` +//! prefix. +//! +//! [`sync`]: kernel::sync +//! [pinning]: https://doc.rust-lang.org/std/pin/index.html +//! [structurally pinned fields]: +//! https://doc.rust-lang.org/std/pin/index.html#pinning-is-structural-for-field +//! [`Arc<T>`]: crate::sync::Arc +//! [`impl PinInit<Foo>`]: PinInit +//! [`impl PinInit<T, E>`]: PinInit +//! [`impl Init<T, E>`]: Init +//! [`Opaque`]: kernel::types::Opaque +//! [`pin_data`]: ::macros::pin_data +//! [`UniqueArc<T>`]: kernel::sync::UniqueArc +//! [`Box<T>`]: alloc::boxed::Box + +use core::{convert::Infallible , marker::PhantomData , mem::MaybeUninit }; + +#[doc(hidden)] +pub mod __internal; + +/// A pin-initializer for the type `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`pin_init_from_closure`] where possible. +/// +/// The [`PinInit::__pinned_init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`Arc::pin_init`]: crate::sync::Arc::pin_init +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ] +pub unsafe trait PinInit <T : ?Sized ,E =Infallible >:Sized { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +/// - `slot` will not move until it is dropped, i.e. it will be pinned. +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),E >; +} + +/// An initializer for `T`. +/// +/// To use this initializer, you will need a suitable memory location that can hold a `T`. This can +/// be [`Box<T>`], [`Arc<T>`], [`UniqueArc<T>`]. Because [`PinInit<T, E>`] is a super trait, you can +/// use every function that takes it as well. +/// +/// Also see the [module description](self). +/// +/// # Safety +/// +/// When implementing this type you will need to take great care. Also there are probably very few +/// cases where a manual implementation is necessary. Use [`init_from_closure`] where possible. +/// +/// The [`Init::__init`] function +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +/// +/// The `__pinned_init` function from the supertrait [`PinInit`] needs to execute the exact same +/// code as `__init`. +/// +/// Contrary to its supertype [`PinInit<T, E>`] the caller is allowed to +/// move the pointee after initialization. +/// +/// [`Arc<T>`]: crate::sync::Arc +/// [`UniqueArc<T>`]: kernel::sync::UniqueArc +/// [`Box<T>`]: alloc::boxed::Box +#[must_use = "An initializer must be used in order to create its value." ] rust/kernel/init.rs line 120
Quoted 10 lines
rust/kernel/init.rs new file +10
-0+pub unsafe trait Init <T : ?Sized ,E =Infallible >:PinInit <T ,E > { +/// Initializes `slot`. +/// +/// # Safety +/// +/// - `slot` is a valid pointer to uninitialized memory. +/// - the caller does not touch `slot` when `Err` is returned, they are only permitted to +/// deallocate. +unsafe fn __init (self, slot: *mut T ) ->Result <(),E >; +}Rather than making
PinInit<T, E>a supertrait here, wouldn't it make sense to just implementPinInit<T, E>for every type that implementsInit<T, E>? This way we could relax the unsafe requirement that "__pinned_initneeds to mimick__init" by simply making it so. Something like (plus a proper safety comment):
Quoted 5 lines
unsafe impl<T, U: ?Sized, E> PinInit<U, E> for T where T: Init<U, E> {
unsafe fn __pinned_init(self, slot: *mut U) -> Result<(), E> {
unsafe { T::__init(self, slot) }
}
}This makes sense.
Quoted 112 lines
Quoted 112 lines
rust/kernel/init.rs new file +63
-0+ +/// Creates a new [`PinInit<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - may assume that the `slot` does not move if `T: !Unpin`, +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn pin_init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl PinInit <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// Creates a new [`Init<T, E>`] from the given closure. +/// +/// # Safety +/// +/// The closure: +/// - returns `Ok(())` if it initialized every field of `slot`, +/// - returns `Err(err)` if it encountered an error and then cleaned `slot`, this means: +/// - `slot` can be deallocated without UB occurring, +/// - `slot` does not need to be dropped, +/// - `slot` is not partially initialized. +/// - the `slot` may move after initialization. +/// - while constructing the `T` at `slot` it upholds the pinning invariants of `T`. +#[inline] +pub const unsafe fn init_from_closure <T : ?Sized ,E >( + f:impl FnOnce (*mut T ) ->Result <(),E >, +) ->impl Init <T ,E > { + __internal::InitClosure (f,PhantomData ) +} + +/// An initializer that leaves the memory uninitialized. +/// +/// The initializer is a no-op. The `slot` memory is not changed. +#[inline] +pub fn uninit <T ,E >() ->impl Init <MaybeUninit <T >,E > { +// SAFETY: The memory is allowed to be uninitialized. +unsafe {init_from_closure (|_|Ok (())) } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >PinInit <T >for T { +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +} + +// SAFETY: Every type can be initialized by-value. +unsafe impl <T >Init <T >for T { +unsafe fn __init (self, slot: *mut T ) ->Result <(),Infallible > { +unsafe { slot.write (self) }; +Ok (()) + } +}rust/kernel/init/__internal.rs new file +43
-0@@ -0,0 +1,45 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +//! This module contains API-internal items for pin-init. +//! +//! These items must not be used outside of +//! - `kernel/init.rs` +//! - `macros/pin_data.rs` +//! - `macros/pinned_drop.rs` + +use super ::*; + +/// See the [nomicon] for what subtyping is. See also [this table]. +/// +/// [nomicon]: https://doc.rust-lang.org/nomicon/subtyping.html +/// [this table]: https://doc.rust-lang.org/nomicon/phantom-data.html#table-of-phantomdata-patterns +type Invariant <T > =PhantomData <fn (*mut T ) -> *mut T >; + +/// This is the module-internal type implementing `PinInit` and `Init`. It is unsafe to create this +/// type, since the closure needs to fulfill the same safety requirement as the +/// `__pinned_init`/`__init` functions. +pub (crate )struct InitClosure <F ,T : ?Sized ,E >(pub (crate )F ,pub (crate )Invariant <(E ,T )>); + +// SAFETY: While constructing the `InitClosure`, the user promised that it upholds the +// `__pinned_init` invariants. +unsafe impl <T : ?Sized ,F ,E >PinInit <T ,E >for InitClosure <F ,T ,E > +where +F :FnOnce (*mut T ) ->Result <(),E >, +{ +#[inline] +unsafe fn __pinned_init (self, slot: *mut T ) ->Result <(),E > { + (self.0 )(slot) + } +} + +// SAFETY: While constructing the `InitClosure`, the user promised that it upholds the +// `__init` invariants. +unsafe impl <T : ?Sized ,F ,E >Init <T ,E >for InitClosure <F ,T ,E > +where +F :FnOnce (*mut T ) ->Result <(),E >, +{ +#[inline] +unsafe fn __init (self, slot: *mut T ) ->Result <(),E > { + (self.0 )(slot)
Quoted 8 lines
rust/kernel/init/__internal.rs line 45
Quoted 2 lines
rust/kernel/init/__internal.rs new file +2
-0+ } +}Since these are only used to implement
init_from_closureandpin_init_from_closure, would it make sense to move them intoinit_from_closure, and havepin_init_from_closurejust callinit_from_closure(given that if a type implements Init<T, E>, it also implements PinInit<T, E>)?
Calling init_from_closure from pin_init_from_closure results in a strange compile error about a cycle when computing the return type of pin_init_from_closure.
Quoted 42 lines
Quoted 42 lines
rust/kernel/lib.rs +7
-0@@ -16,7 +16,9 @@#![feature(coerce_unsized)] #![feature(core_ffi_c)] #![feature(dispatch_from_dyn)] +#![feature(explicit_generic_args_with_impl_trait)] #![feature(generic_associated_types)] +#![feature(new_uninit)] #![feature(pin_macro)] #![feature(receiver_trait)] #![feature(unsize)] @@ -26,11 +28,16 @@#[cfg(not( CONFIG_RUST ))] compile_error! ("Missing kernel configuration for conditional compilation" ); +#[allow(unused_extern_crates)] +// Allow proc-macros to refer to `::kernel` inside the `kernel` crate (this crate). +extern crate selfas kernel; +#[cfg(not(test))] #[cfg(not(testlib))] mod allocator;mod build_assert;pub mod error; +pub mod init;pub mod prelude;pub mod print;mod static_assert;scripts/Makefile.build +1
-1@@ -277,7 +277,7 @@ $(obj)/%.lst: $(src)/%.c FORCE # Compile Rust sources (.rs) # --------------------------------------------------------------------------- -rust_allowed_features := core_ffi_c +rust_allowed_features := core_ffi_c,explicit_generic_args_with_impl_trait
Quoted 6 lines
Quoted 6 lines
scripts/Makefile.build +0
-1rust_common_cmd = \ RUST_MODFILE=$(modfile) $(RUSTC_OR_CLIPPY) $(rust_flags) \ --2.39.2
--
Cheers,
Benno
PATCH v4 7/15 rust: init: add initialization macros
In reply to [email protected]From: Benno Lossin <[email protected]>
Add the following initializer macros:
- #[pin_data] to annotate structurally pinned fields of structs, needed for pin_init! and try_pin_init! to select the correct initializer of fields.
- pin_init! create a pin-initializer for a struct with the Infallible error type.
- try_pin_init! create a pin-initializer for a struct with a custom error type (kernel::error::Error is the default).
- init! create an in-place-initializer for a struct with the Infallible error type.
- try_init! create an in-place-initializer for a struct with a custom error type (kernel::error::Error is the default).
Also add their needed internal helper traits and structs.
--- rust/kernel/init.rs | 810 ++++++++++++++++++++++++++++++++- rust/kernel/init/__internal.rs | 124 +++++ rust/kernel/init/macros.rs | 707 ++++++++++++++++++++++++++++ rust/macros/lib.rs | 29 ++ rust/macros/pin_data.rs | 79 ++++ rust/macros/quote.rs | 2 - 6 files changed, 1744 insertions(+), 7 deletions(-) create mode 100644 rust/kernel/init/macros.rs create mode 100644 rust/macros/pin_data.rs
rust/kernel/init.rs +805 -5
@@ -14,7 +14,8 @@//! - an in-place constructor, //! - a memory location that can hold your `struct`. //! -//! To get an in-place constructor there are generally two options: +//! To get an in-place constructor there are generally three options: +//! - directly creating an in-place constructor using the [`pin_init!`] macro, //! - a custom function/macro returning an in-place constructor provided by someone else, //! - using the unsafe function [`pin_init_from_closure()`] to manually create an initializer. //!
@@ -22,6 +23,87 @@//! the macros/types/functions are generally named like the pinned variants without the `pin` //! prefix. //! +//! # Examples +//! +//! ## Using the [`pin_init!`] macro +//! +//! If you want to use [`PinInit`], then you will have to annotate your `struct` with +//! `#[`[`pin_data`]`]`. It is a macro that uses `#[pin]` as a marker for +//! [structurally pinned fields]. After doing this, you can then create an in-place constructor via +//! [`pin_init!`]. The syntax is almost the same as normal `struct` initializers. The difference is +//! that you need to write `<-` instead of `:` for fields that you want to initialize in-place. +//! +//! ```rust +//! # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +//! use kernel::{prelude::*, sync::Mutex, new_mutex}; +//! # use core::pin::Pin; +//! #[pin_data] +//! struct Foo { +//! #[pin] +//! a: Mutex<usize>, +//! b: u32, +//! } +//! +//! let foo = pin_init!(Foo { +//! a <- new_mutex!(42, "Foo::a"), +//! b: 24, +//! }); +//! ``` +//! +//! `foo` now is of the type [`impl PinInit<Foo>`]. We can now use any smart pointer that we like +//! (or just the stack) to actually initialize a `Foo`: +//! +//! ```rust +//! # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +//! # use kernel::{prelude::*, sync::Mutex, new_mutex}; +//! # use core::pin::Pin; +//! # #[pin_data] +//! # struct Foo { +//! # #[pin] +//! # a: Mutex<usize>, +//! # b: u32, +//! # } +//! # let foo = pin_init!(Foo { +//! # a <- new_mutex!(42, "Foo::a"), +//! # b: 24, +//! # }); +//! let foo: Result<Pin<Box<Foo>>> = Box::pin_init(foo); +//! ``` +//! +//! For more information see the [`pin_init!`] macro. +//! +//! ## Using a custom function/macro that returns an initializer +//! +//! Many types from the kernel supply a function/macro that returns an initializer, because the +//! above method only works for types where you can access the fields. +//! +//! ```rust +//! # use kernel::{new_mutex, sync::{Arc, Mutex}}; +//! let mtx: Result<Arc<Mutex<usize>>> = Arc::pin_init(new_mutex!(42, "example::mtx")); +//! ``` +//! +//! To declare an init macro/function you just return an [`impl PinInit<T, E>`]: +//! +//! ```rust +//! # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +//! # use kernel::{sync::Mutex, prelude::*, new_mutex, init::PinInit, try_pin_init}; +//! #[pin_data] +//! struct DriverData { +//! #[pin] +//! status: Mutex<i32>, +//! buffer: Box<[u8; 1_000_000]>, +//! } +//! +//! impl DriverData { +//! fn new() -> impl PinInit<Self, Error> { +//! try_pin_init!(Self { +//! status <- new_mutex!(0, "DriverData::status"), +//! buffer: Box::init(kernel::init::zeroed())?, +//! }) +//! } +//! } +//! ``` +//! //! [`sync`]: kernel::sync //! [pinning]: https://doc.rust-lang.org/std/pin/index.html //! [structurally pinned fields]:
@@ -33,12 +115,732 @@//! [`Opaque`]: kernel::types::Opaque //! [`pin_data`]: ::macros::pin_data //! [`UniqueArc<T>`]: kernel::sync::UniqueArc -//! [`Box<T>`]: alloc::boxed::Box -use core::{convert::Infallible , marker::PhantomData , mem::MaybeUninit }; +use alloc::boxed::Box ; +use core::{cell::Cell , convert::Infallible , marker::PhantomData , mem::MaybeUninit , ptr};#[doc(hidden)] pub mod __internal; +#[doc(hidden)] +pub mod macros; + +/// Construct an in-place, pinned initializer for `struct`s. +/// +/// This macro defaults the error to [`Infallible`]. If you need [`Error`], then use +/// [`try_pin_init!`]. +/// +/// The syntax is almost identical to that of a normal `struct` initializer: +/// +/// ```rust +/// # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +/// # use kernel::{init, pin_init, macros::pin_data, init::*}; +/// # use core::pin::Pin; +/// #[pin_data] +/// struct Foo { +/// a: usize, +/// b: Bar, +/// } +/// +/// #[pin_data] +/// struct Bar { +/// x: u32, +/// } +/// +/// # fn demo() -> impl PinInit<Foo> { +/// let a = 42; +/// +/// let initializer = pin_init!(Foo { +/// a, +/// b: Bar { +/// x: 64, +/// }, +/// }); +/// # initializer } +/// # Box::pin_init(demo()).unwrap(); +/// ``` +/// +/// Arbitrary Rust expressions can be used to set the value of a variable. +/// +/// The fields are initialized in the order that they appear in the initializer. So it is possible +/// to read already initialized fields using raw pointers. +/// +/// IMPORTANT: You are not allowed to create references to fields of the struct inside of the +/// initializer. +/// +/// # Init-functions +/// +/// When working with this API it is often desired to let others construct your types without +/// giving access to all fields. This is where you would normally write a plain function `new` +/// that would return a new instance of your type. With this API that is also possible. +/// However, there are a few extra things to keep in mind. +/// +/// To create an initializer function, simply declare it like this: +/// +/// ```rust +/// # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +/// # use kernel::{init, pin_init, prelude::*, init::*}; +/// # use core::pin::Pin; +/// # #[pin_data] +/// # struct Foo { +/// # a: usize, +/// # b: Bar, +/// # } +/// # #[pin_data] +/// # struct Bar { +/// # x: u32, +/// # } +/// impl Foo { +/// fn new() -> impl PinInit<Self> { +/// pin_init!(Self { +/// a: 42, +/// b: Bar { +/// x: 64, +/// }, +/// }) +/// } +/// } +/// ``` +/// +/// Users of `Foo` can now create it like this: +/// +/// ```rust +/// # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +/// # use kernel::{init, pin_init, macros::pin_data, init::*}; +/// # use core::pin::Pin; +/// # #[pin_data] +/// # struct Foo { +/// # a: usize, +/// # b: Bar, +/// # } +/// # #[pin_data] +/// # struct Bar { +/// # x: u32, +/// # } +/// # impl Foo { +/// # fn new() -> impl PinInit<Self> { +/// # pin_init!(Self { +/// # a: 42, +/// # b: Bar { +/// # x: 64, +/// # }, +/// # }) +/// # } +/// # } +/// let foo = Box::pin_init(Foo::new()); +/// ``` +/// +/// They can also easily embed it into their own `struct`s: +/// +/// ```rust +/// # #![allow(clippy::disallowed_names, clippy::new_ret_no_self)] +/// # use kernel::{init, pin_init, macros::pin_data, init::*}; +/// # use core::pin::Pin; +/// # #[pin_data] +/// # struct Foo { +/// # a: usize, +/// # b: Bar, +/// # } +/// # #[pin_data] +/// # struct Bar { +/// # x: u32, +/// # } +/// # impl Foo { +/// # fn new() -> impl PinInit<Self> { +/// # pin_init!(Self { +/// # a: 42, +/// # b: Bar { +/// # x: 64, +/// # }, +/// # }) +/// # } +/// # } +/// #[pin_data] +/// struct FooContainer { +/// #[pin] +/// foo1: Foo, +/// #[pin] +/// foo2: Foo, +/// other: u32, +/// } +/// +/// impl FooContainer { +/// fn new(other: u32) -> impl PinInit<Self> { +/// pin_init!(Self { +/// foo1 <- Foo::new(), +/// foo2 <- Foo::new(), +/// other, +/// }) +/// } +/// } +/// ``` +/// +/// Here we see that when using `pin_init!` with `PinInit`, one needs to write `<-` instead of `:`. +/// This signifies that the given field is initialized in-place. As with `struct` initializers, just +/// writing the field (in this case `other`) without `:` or `<-` means `other: other,`. +/// +/// # Syntax +/// +/// As already mentioned in the examples above, inside of `pin_init!` a `struct` initializer with +/// the following modifications is expected: +/// - Fields that you want to initialize in-place have to use `<-` instead of `:`. +/// - In front of the initializer you can write `&this in` to have access to a [`NonNull<Self>`] +/// pointer named `this` inside of the initializer. +/// +/// For instance: +/// +/// ```rust +/// # use kernel::pin_init; +/// # use macros::pin_data; +/// # use core::{ptr::addr_of_mut, marker::PhantomPinned}; +/// #[pin_data] +/// struct Buf { +/// // `ptr` points into `buf`. +/// ptr: *mut u8, +/// buf: [u8; 64], +/// #[pin] +/// pin: PhantomPinned, +/// } +/// pin_init!(&this in Buf { +/// buf: [0; 64], +/// ptr: unsafe { addr_of_mut!((*this.as_ptr()).buf).cast() }, +/// pin: PhantomPinned, +/// }); +/// ``` +/// +/// [`try_pin_init!`]: kernel::try_pin_init +/// [`NonNull<Self>`]: core::ptr::NonNull +/// [`Error`]: kernel::error::Error +// For a detailed example of how this macro works, see the module documentation of the hidden +// module `__internal` inside of `init/__internal.rs`. +#[macro_export] +macro_rules! pin_init { + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }) => { + $crate::try_pin_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)?), + @fields ($($fields)*), + @error (::core::convert::Infallible ), + ) + }; +} + +/// Construct an in-place, fallible pinned initializer for `struct`s. +/// +/// If the initialization can complete without error (or [`Infallible`]), then use [`pin_init!`]. +/// +/// You can use the `?` operator or use `return Err(err)` inside the initializer to stop +/// initialization and return the error. +/// +/// IMPORTANT: if you have `unsafe` code inside of the initializer you have to ensure that when +/// initialization fails, the memory can be safely deallocated without any further modifications. +/// +/// This macro defaults the error to [`Error`]. +/// +/// The syntax is identical to [`pin_init!`] with the following exception: you can append `? $type` +/// after the `struct` initializer to specify the error type you want to use. +/// +/// # Examples +/// +/// ```rust +/// # #![feature(new_uninit)] +/// use kernel::{init::{self, PinInit}, error::Error}; +/// #[pin_data] +/// struct BigBuf { +/// big: Box<[u8; 1024 * 1024 * 1024]>, +/// small: [u8; 1024 * 1024], +/// ptr: *mut u8, +/// } +/// +/// impl BigBuf { +/// fn new() -> impl PinInit<Self, Error> { +/// try_pin_init!(Self { +/// big: Box::init(init::zeroed())?, +/// small: [0; 1024 * 1024], +/// ptr: core::ptr::null_mut(), +/// }? Error) +/// } +/// } +/// ``` +/// +/// [`Error`]: kernel::error::Error +// For a detailed example of how this macro works, see the module documentation of the hidden +// module `__internal` inside of `init/__internal.rs`. +#[macro_export] +macro_rules! try_pin_init { + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }) => { + $crate::try_pin_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)? ), + @fields ($($fields)*), + @error ($crate ::error::Error ), + ) + }; + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }? $err:ty) => { + $crate::try_pin_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)? ), + @fields ($($fields)*), + @error ($err), + ) + }; + ( + @this($($this:ident)?), + @typ($t:ident $(::<$($generics:ty),*>)?), + @fields($($fields:tt)*), + @error($err:ty), + ) => {{ +// We do not want to allow arbitrary returns, so we declare this type as the `Ok` return +// type and shadow it later when we insert the arbitrary user code. That way there will be +// no possibility of returning without `unsafe`. +struct __InitOk ; +// Get the pin data from the supplied type. +let data =unsafe { +use $crate::init::__internal::HasPinData ; + $t$(::<$($generics),*>)?::__pin_data () + }; +// Ensure that `data` really is of type `PinData` and help with type inference: +let init = $crate::init::__internal::PinData ::make_closure ::<_ ,__InitOk , $err>( + data, +move |slot| { + { +// Shadow the structure so it cannot be used to return early. +struct __InitOk ; +// Create the `this` so it can be referenced by the user inside of the +// expressions creating the individual fields. + $(let $this =unsafe { ::core::ptr::NonNull ::new_unchecked (slot) };)? +// Initialize every field. + $crate::try_pin_init!(init_slot : + @data(data), + @slot (slot), + @munch_fields ($($fields)*,), + ); +// We use unreachable code to ensure that all fields have been mentioned exactly +// once, this struct initializer will still be type-checked and complain with a +// very natural error message if a field is forgotten/mentioned more than once. +#[allow(unreachable_code, clippy::diverging_sub_expression)] +if false { + $crate::try_pin_init!(make_initializer : + @slot(slot), + @type_name ($t), + @munch_fields ($($fields)*,), + @acc (), + ); + } +// Forget all guards, since initialization was a success. + $crate::try_pin_init!(forget_guards : + @munch_fields($($fields)*,), + ); + } +Ok (__InitOk) + } + ); +let init =move |slot| -> ::core::result::Result <(), $err> { +init (slot).map (|__InitOk| ()) + }; +let init =unsafe { $crate::init::pin_init_from_closure ::<_ , $err>(init) }; + init + }}; + (init_slot: + @data($data:ident), + @slot($slot:ident), + @munch_fields($(,)?), + ) => { +// Endpoint of munching, no fields are left. + }; + (init_slot: + @data($data:ident), + @slot($slot:ident), +// In-place initialization syntax. + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + ) => { +let $field = $val; +// Call the initializer. +// +// SAFETY: `slot` is valid, because we are inside of an initializer closure, we +// return when an error/panic occurs. +// We also use the `data` to require the correct trait (`Init` or `PinInit`) for `$field`. +unsafe { $data.$field(::core::ptr::addr_of_mut!((*$slot).$field), $field)? }; +// Create the drop guard. +// +// We only give access to `&DropGuard`, so it cannot be forgotten via safe code. +// +// SAFETY: We forget the guard later when initialization has succeeded. +let $field = &unsafe { + $crate::init::__internal::DropGuard ::new (::core::ptr::addr_of_mut!((*$slot).$field)) + }; + + $crate::try_pin_init!(init_slot : + @data($data), + @slot ($slot), + @munch_fields ($($rest)*), + ); + }; + (init_slot: + @data($data:ident), + @slot($slot:ident), +// Direct value init, this is safe for every field. + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + ) => { + $(let $field = $val;)? +// Initialize the field. +// +// SAFETY: The memory at `slot` is uninitialized. +unsafe { ::core::ptr::write (::core::ptr::addr_of_mut!((*$slot).$field), $field) }; +// Create the drop guard: +// +// We only give access to `&DropGuard`, so it cannot be accidentally forgotten. +// +// SAFETY: We forget the guard later when initialization has succeeded. +let $field = &unsafe { + $crate::init::__internal::DropGuard ::new (::core::ptr::addr_of_mut!((*$slot).$field)) + }; + + $crate::try_pin_init!(init_slot : + @data($data), + @slot ($slot), + @munch_fields ($($rest)*), + ); + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields($(,)?), + @acc($($acc:tt)*), + ) => { +// Endpoint, nothing more to munch, create the initializer. +// Since we are in the `if false` branch, this will never get executed. We abuse `slot` to +// get the correct type inference here: +unsafe { + ::core::ptr::write ($slot, $t { + $($acc)* + }); + } + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + @acc($($acc:tt)*), + ) => { + $crate::try_pin_init!(make_initializer : + @slot($slot), + @type_name ($t), + @munch_fields ($($rest)*), + @acc ($($acc)* $field: ::core::panic!(),), + ); + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + @acc($($acc:tt)*), + ) => { + $crate::try_pin_init!(make_initializer : + @slot($slot), + @type_name ($t), + @munch_fields ($($rest)*), + @acc ($($acc)* $field: ::core::panic!(),), + ); + }; + (forget_guards: + @munch_fields($(,)?), + ) => { +// Munching finished. + }; + (forget_guards: + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + ) => { +unsafe { $crate::init::__internal::DropGuard ::forget ($field) }; + + $crate::try_pin_init!(forget_guards : + @munch_fields($($rest)*), + ); + }; + (forget_guards: + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + ) => { +unsafe { $crate::init::__internal::DropGuard ::forget ($field) }; + + $crate::try_pin_init!(forget_guards : + @munch_fields($($rest)*), + ); + }; +} + +/// Construct an in-place initializer for `struct`s. +/// +/// This macro defaults the error to [`Infallible`]. If you need [`Error`], then use +/// [`try_init!`]. +/// +/// The syntax is identical to [`pin_init!`] and its safety caveats also apply: +/// - `unsafe` code must guarantee either full initialization or return an error and allow +/// deallocation of the memory. +/// - the fields are initialized in the order given in the initializer. +/// - no references to fields are allowed to be created inside of the initializer. +/// +/// This initializer is for initializing data in-place that might later be moved. If you want to +/// pin-initialize, use [`pin_init!`]. +/// +/// [`Error`]: kernel::error::Error +// For a detailed example of how this macro works, see the module documentation of the hidden +// module `__internal` inside of `init/__internal.rs`. +#[macro_export] +macro_rules! init { + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }) => { + $crate::try_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)?), + @fields ($($fields)*), + @error (::core::convert::Infallible ), + ) + } +} + +/// Construct an in-place fallible initializer for `struct`s. +/// +/// This macro defaults the error to [`Error`]. If you need [`Infallible`], then use +/// [`init!`]. +/// +/// The syntax is identical to [`try_pin_init!`]. If you want to specify a custom error, +/// append `? $type` after the `struct` initializer. +/// The safety caveats from [`try_pin_init!`] also apply: +/// - `unsafe` code must guarantee either full initialization or return an error and allow +/// deallocation of the memory. +/// - the fields are initialized in the order given in the initializer. +/// - no references to fields are allowed to be created inside of the initializer. +/// +/// # Examples +/// +/// ```rust +/// use kernel::{init::PinInit, error::Error}; +/// struct BigBuf { +/// big: Box<[u8; 1024 * 1024 * 1024]>, +/// small: [u8; 1024 * 1024], +/// } +/// +/// impl BigBuf { +/// fn new() -> impl Init<Self, Error> { +/// try_init!(Self { +/// big: { +/// let zero = Box::try_new_zeroed()?; +/// unsafe { zero.assume_init() } +/// }, +/// small: [0; 1024 * 1024], +/// }? Error) +/// } +/// } +/// ``` +/// +/// [`Error`]: kernel::error::Error +// For a detailed example of how this macro works, see the module documentation of the hidden +// module `__internal` inside of `init/__internal.rs`. +#[macro_export] +macro_rules! try_init { + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }) => { + $crate::try_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)?), + @fields ($($fields)*), + @error ($crate ::error::Error ), + ) + }; + ($(&$this:ident in)? $t:ident $(::<$($generics:ty),* $(,)?>)? { + $($fields:tt)* + }? $err:ty) => { + $crate::try_init!( + @this ($($this)?), + @typ ($t $(::<$($generics),*>)?), + @fields ($($fields)*), + @error ($err), + ) + }; + ( + @this($($this:ident)?), + @typ($t:ident $(::<$($generics:ty),*>)?), + @fields($($fields:tt)*), + @error($err:ty), + ) => {{ +// We do not want to allow arbitrary returns, so we declare this type as the `Ok` return +// type and shadow it later when we insert the arbitrary user code. That way there will be +// no possibility of returning without `unsafe`. +struct __InitOk ; +// Get the init data from the supplied type. +let data =unsafe { +use $crate::init::__internal::HasInitData ; + $t$(::<$($generics),*>)?::__init_data () + }; +// Ensure that `data` really is of type `InitData` and help with type inference: +let init = $crate::init::__internal::InitData ::make_closure ::<_ ,__InitOk , $err>( + data, +move |slot| { + { +// Shadow the structure so it cannot be used to return early. +struct __InitOk ; +// Create the `this` so it can be referenced by the user inside of the +// expressions creating the individual fields. + $(let $this =unsafe { ::core::ptr::NonNull ::new_unchecked (slot) };)? +// Initialize every field. + $crate::try_init!(init_slot : + @slot(slot), + @munch_fields ($($fields)*,), + ); +// We use unreachable code to ensure that all fields have been mentioned exactly +// once, this struct initializer will still be type-checked and complain with a +// very natural error message if a field is forgotten/mentioned more than once. +#[allow(unreachable_code, clippy::diverging_sub_expression)] +if false { + $crate::try_init!(make_initializer : + @slot(slot), + @type_name ($t), + @munch_fields ($($fields)*,), + @acc (), + ); + } +// Forget all guards, since initialization was a success. + $crate::try_init!(forget_guards : + @munch_fields($($fields)*,), + ); + } +Ok (__InitOk) + } + ); +let init =move |slot| -> ::core::result::Result <(), $err> { +init (slot).map (|__InitOk| ()) + }; +let init =unsafe { $crate::init::init_from_closure ::<_ , $err>(init) }; + init + }}; + (init_slot: + @slot($slot:ident), + @munch_fields( $(,)?), + ) => { +// Endpoint of munching, no fields are left. + }; + (init_slot: + @slot($slot:ident), + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + ) => { +let $field = $val; +// Call the initializer. +// +// SAFETY: `slot` is valid, because we are inside of an initializer closure, we +// return when an error/panic occurs. +unsafe { + $crate::init::Init ::__init ($field, ::core::ptr::addr_of_mut!((*$slot).$field))?; + } +// Create the drop guard. +// +// We only give access to `&DropGuard`, so it cannot be accidentally forgotten. +// +// SAFETY: We forget the guard later when initialization has succeeded. +let $field = &unsafe { + $crate::init::__internal::DropGuard ::new (::core::ptr::addr_of_mut!((*$slot).$field)) + }; + + $crate::try_init!(init_slot : + @slot($slot), + @munch_fields ($($rest)*), + ); + }; + (init_slot: + @slot($slot:ident), +// Direct value init. + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + ) => { + $(let $field = $val;)? +// Call the initializer. +// +// SAFETY: The memory at `slot` is uninitialized. +unsafe { ::core::ptr::write (::core::ptr::addr_of_mut!((*$slot).$field), $field) }; +// Create the drop guard. +// +// We only give access to `&DropGuard`, so it cannot be accidentally forgotten. +// +// SAFETY: We forget the guard later when initialization has succeeded. +let $field = &unsafe { + $crate::init::__internal::DropGuard ::new (::core::ptr::addr_of_mut!((*$slot).$field)) + }; + + $crate::try_init!(init_slot : + @slot($slot), + @munch_fields ($($rest)*), + ); + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields( $(,)?), + @acc($($acc:tt)*), + ) => { +// Endpoint, nothing more to munch, create the initializer. +// Since we are in the `if false` branch, this will never get executed. We abuse `slot` to +// get the correct type inference here: +unsafe { + ::core::ptr::write ($slot, $t { + $($acc)* + }); + } + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + @acc($($acc:tt)*), + ) => { + $crate::try_init!(make_initializer : + @slot($slot), + @type_name ($t), + @munch_fields ($($rest)*), + @acc ($($acc)*$field: ::core::panic!(),), + ); + }; + (make_initializer: + @slot($slot:ident), + @type_name($t:ident), + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + @acc($($acc:tt)*), + ) => { + $crate::try_init!(make_initializer : + @slot($slot), + @type_name ($t), + @munch_fields ($($rest)*), + @acc ($($acc)*$field: ::core::panic!(),), + ); + }; + (forget_guards: + @munch_fields($(,)?), + ) => { +// Munching finished. + }; + (forget_guards: + @munch_fields($field:ident <- $val:expr, $($rest:tt)*), + ) => { +unsafe { $crate::init::__internal::DropGuard ::forget ($field) }; + + $crate::try_init!(forget_guards : + @munch_fields($($rest)*), + ); + }; + (forget_guards: + @munch_fields($field:ident $(: $val:expr)?, $($rest:tt)*), + ) => { +unsafe { $crate::init::__internal::DropGuard ::forget ($field) }; + + $crate::try_init!(forget_guards : + @munch_fields($($rest)*), + ); + }; +}/// A pin-initializer for the type `T`. ///
@@ -63,7 +865,6 @@ pub mod __internal;/// [`Arc<T>`]: crate::sync::Arc /// [`Arc::pin_init`]: crate::sync::Arc::pin_init /// [`UniqueArc<T>`]: kernel::sync::UniqueArc -/// [`Box<T>`]: alloc::boxed::Box #[must_use = "An initializer must be used in order to create its value." ] pub unsafe trait PinInit <T : ?Sized ,E =Infallible >:Sized {/// Initializes `slot`.
@@ -106,7 +907,6 @@ pub unsafe trait PinInit<T: ?Sized, E = Infallible>: Sized {/// /// [`Arc<T>`]: crate::sync::Arc /// [`UniqueArc<T>`]: kernel::sync::UniqueArc -/// [`Box<T>`]: alloc::boxed::Box #[must_use = "An initializer must be used in order to create its value." ] pub unsafe trait Init <T : ?Sized ,E =Infallible >:PinInit <T ,E > {/// Initializes `slot`.
rust/kernel/init/__internal.rs +124 -0
127 lines not on this page · raw message
rust/kernel/init/macros.rs new file +707 -0
707 lines not on this page · raw message
rust/macros/lib.rs +29 -0
@@ -7,6 +7,7 @@ mod quote;mod concat_idents;mod helpers;mod module; +mod pin_data;mod vtable;use proc_macro::TokenStream ;
@@ -168,3 +169,31 @@ pub fn vtable(attr: TokenStream, ts: TokenStream) -> TokenStream {pub fn concat_idents (ts:TokenStream ) ->TokenStream { concat_idents::concat_idents (ts) } + +/// Used to specify the pinning information of the fields of a struct. +/// +/// This is somewhat similar in purpose as +/// [pin-project-lite](https://crates.io/crates/pin-project-lite). +/// Place this macro on a struct definition and then `#[pin]` in front of the attributes of each +/// field you want to structurally pin. +/// +/// This macro enables the use of the [`pin_init!`] macro. When pin-initializing a `struct`, +/// then `#[pin]` directs the type of intializer that is required. +/// +/// # Examples +/// +/// ```rust,ignore +/// #[pin_data] +/// struct DriverData { +/// #[pin] +/// queue: Mutex<Vec<Command>>, +/// buf: Box<[u8; 1024 * 1024]>, +/// } +/// ``` +/// +/// [`pin_init!`]: ../kernel/macro.pin_init.html +// ^ cannot use direct link, since `kernel` is not a dependency of `macros`. +#[proc_macro_attribute] +pub fn pin_data (inner:TokenStream , item:TokenStream ) ->TokenStream { + pin_data::pin_data (inner, item) +}
rust/macros/pin_data.rs new file +79 -0
79 lines not on this page · raw message
rust/macros/quote.rs +0 -2
@@ -38,7 +38,6 @@ impl ToTokens for TokenStream {/// This is a similar to the /// [`quote_spanned!`](https://docs.rs/quote/latest/quote/macro.quote_spanned.html) macro from the /// `quote` crate but provides only just enough functionality needed by the current `macros` crate. -#[allow(unused_macros)] macro_rules! quote_spanned { ($span:expr => $($tt:tt)*) => {#[allow(clippy::vec_init_then_push)]
@@ -137,7 +136,6 @@ macro_rules! quote_spanned {/// `macros` crate. /// /// [`Span::mixed_site()`]: https://doc.rust-lang.org/proc_macro/struct.Span.html#method.mixed_site -#[allow(unused_macros)] macro_rules! quote { ($($tt:tt)*) => {quote_spanned! (::proc_macro::Span ::mixed_site () => $($tt)*)
--
2.39.2