split embassy-util into embassy-futures, embassy-sync.

This commit is contained in:
Dario Nieuwenhuis
2022-08-22 21:46:09 +02:00
parent 61356181b2
commit 21072bee48
118 changed files with 391 additions and 142 deletions

View File

@ -0,0 +1,189 @@
//! Blocking mutex.
//!
//! This module provides a blocking mutex that can be used to synchronize data.
pub mod raw;
use core::cell::UnsafeCell;
use self::raw::RawMutex;
/// Blocking mutex (not async)
///
/// Provides a blocking mutual exclusion primitive backed by an implementation of [`raw::RawMutex`].
///
/// Which implementation you select depends on the context in which you're using the mutex, and you can choose which kind
/// of interior mutability fits your use case.
///
/// Use [`CriticalSectionMutex`] when data can be shared between threads and interrupts.
///
/// Use [`NoopMutex`] when data is only shared between tasks running on the same executor.
///
/// Use [`ThreadModeMutex`] when data is shared between tasks running on the same executor but you want a global singleton.
///
/// In all cases, the blocking mutex is intended to be short lived and not held across await points.
/// Use the async [`Mutex`](crate::mutex::Mutex) if you need a lock that is held across await points.
pub struct Mutex<R, T: ?Sized> {
// NOTE: `raw` must be FIRST, so when using ThreadModeMutex the "can't drop in non-thread-mode" gets
// to run BEFORE dropping `data`.
raw: R,
data: UnsafeCell<T>,
}
unsafe impl<R: RawMutex + Send, T: ?Sized + Send> Send for Mutex<R, T> {}
unsafe impl<R: RawMutex + Sync, T: ?Sized + Send> Sync for Mutex<R, T> {}
impl<R: RawMutex, T> Mutex<R, T> {
/// Creates a new mutex in an unlocked state ready for use.
#[inline]
pub const fn new(val: T) -> Mutex<R, T> {
Mutex {
raw: R::INIT,
data: UnsafeCell::new(val),
}
}
/// Creates a critical section and grants temporary access to the protected data.
pub fn lock<U>(&self, f: impl FnOnce(&T) -> U) -> U {
self.raw.lock(|| {
let ptr = self.data.get() as *const T;
let inner = unsafe { &*ptr };
f(inner)
})
}
}
impl<R, T> Mutex<R, T> {
/// Creates a new mutex based on a pre-existing raw mutex.
///
/// This allows creating a mutex in a constant context on stable Rust.
#[inline]
pub const fn const_new(raw_mutex: R, val: T) -> Mutex<R, T> {
Mutex {
raw: raw_mutex,
data: UnsafeCell::new(val),
}
}
/// Consumes this mutex, returning the underlying data.
#[inline]
pub fn into_inner(self) -> T {
self.data.into_inner()
}
/// Returns a mutable reference to the underlying data.
///
/// Since this call borrows the `Mutex` mutably, no actual locking needs to
/// take place---the mutable borrow statically guarantees no locks exist.
#[inline]
pub fn get_mut(&mut self) -> &mut T {
unsafe { &mut *self.data.get() }
}
}
/// A mutex that allows borrowing data across executors and interrupts.
///
/// # Safety
///
/// This mutex is safe to share between different executors and interrupts.
pub type CriticalSectionMutex<T> = Mutex<raw::CriticalSectionRawMutex, T>;
/// A mutex that allows borrowing data in the context of a single executor.
///
/// # Safety
///
/// **This Mutex is only safe within a single executor.**
pub type NoopMutex<T> = Mutex<raw::NoopRawMutex, T>;
impl<T> Mutex<raw::CriticalSectionRawMutex, T> {
/// Borrows the data for the duration of the critical section
pub fn borrow<'cs>(&'cs self, _cs: critical_section::CriticalSection<'cs>) -> &'cs T {
let ptr = self.data.get() as *const T;
unsafe { &*ptr }
}
}
impl<T> Mutex<raw::NoopRawMutex, T> {
/// Borrows the data
pub fn borrow(&self) -> &T {
let ptr = self.data.get() as *const T;
unsafe { &*ptr }
}
}
// ThreadModeMutex does NOT use the generic mutex from above because it's special:
// it's Send+Sync even if T: !Send. There's no way to do that without specialization (I think?).
//
// There's still a ThreadModeRawMutex for use with the generic Mutex (handy with Channel, for example),
// but that will require T: Send even though it shouldn't be needed.
#[cfg(any(cortex_m, feature = "std"))]
pub use thread_mode_mutex::*;
#[cfg(any(cortex_m, feature = "std"))]
mod thread_mode_mutex {
use super::*;
/// A "mutex" that only allows borrowing from thread mode.
///
/// # Safety
///
/// **This Mutex is only safe on single-core systems.**
///
/// On multi-core systems, a `ThreadModeMutex` **is not sufficient** to ensure exclusive access.
pub struct ThreadModeMutex<T: ?Sized> {
inner: UnsafeCell<T>,
}
// NOTE: ThreadModeMutex only allows borrowing from one execution context ever: thread mode.
// Therefore it cannot be used to send non-sendable stuff between execution contexts, so it can
// be Send+Sync even if T is not Send (unlike CriticalSectionMutex)
unsafe impl<T: ?Sized> Sync for ThreadModeMutex<T> {}
unsafe impl<T: ?Sized> Send for ThreadModeMutex<T> {}
impl<T> ThreadModeMutex<T> {
/// Creates a new mutex
pub const fn new(value: T) -> Self {
ThreadModeMutex {
inner: UnsafeCell::new(value),
}
}
}
impl<T: ?Sized> ThreadModeMutex<T> {
/// Lock the `ThreadModeMutex`, granting access to the data.
///
/// # Panics
///
/// This will panic if not currently running in thread mode.
pub fn lock<R>(&self, f: impl FnOnce(&T) -> R) -> R {
f(self.borrow())
}
/// Borrows the data
///
/// # Panics
///
/// This will panic if not currently running in thread mode.
pub fn borrow(&self) -> &T {
assert!(
raw::in_thread_mode(),
"ThreadModeMutex can only be borrowed from thread mode."
);
unsafe { &*self.inner.get() }
}
}
impl<T: ?Sized> Drop for ThreadModeMutex<T> {
fn drop(&mut self) {
// Only allow dropping from thread mode. Dropping calls drop on the inner `T`, so
// `drop` needs the same guarantees as `lock`. `ThreadModeMutex<T>` is Send even if
// T isn't, so without this check a user could create a ThreadModeMutex in thread mode,
// send it to interrupt context and drop it there, which would "send" a T even if T is not Send.
assert!(
raw::in_thread_mode(),
"ThreadModeMutex can only be dropped from thread mode."
);
// Drop of the inner `T` happens after this.
}
}
}

View File

@ -0,0 +1,149 @@
//! Mutex primitives.
//!
//! This module provides a trait for mutexes that can be used in different contexts.
use core::marker::PhantomData;
/// Raw mutex trait.
///
/// This mutex is "raw", which means it does not actually contain the protected data, it
/// just implements the mutex mechanism. For most uses you should use [`super::Mutex`] instead,
/// which is generic over a RawMutex and contains the protected data.
///
/// Note that, unlike other mutexes, implementations only guarantee no
/// concurrent access from other threads: concurrent access from the current
/// thread is allwed. For example, it's possible to lock the same mutex multiple times reentrantly.
///
/// Therefore, locking a `RawMutex` is only enough to guarantee safe shared (`&`) access
/// to the data, it is not enough to guarantee exclusive (`&mut`) access.
///
/// # Safety
///
/// RawMutex implementations must ensure that, while locked, no other thread can lock
/// the RawMutex concurrently.
///
/// Unsafe code is allowed to rely on this fact, so incorrect implementations will cause undefined behavior.
pub unsafe trait RawMutex {
/// Create a new `RawMutex` instance.
///
/// This is a const instead of a method to allow creating instances in const context.
const INIT: Self;
/// Lock this `RawMutex`.
fn lock<R>(&self, f: impl FnOnce() -> R) -> R;
}
/// A mutex that allows borrowing data across executors and interrupts.
///
/// # Safety
///
/// This mutex is safe to share between different executors and interrupts.
pub struct CriticalSectionRawMutex {
_phantom: PhantomData<()>,
}
unsafe impl Send for CriticalSectionRawMutex {}
unsafe impl Sync for CriticalSectionRawMutex {}
impl CriticalSectionRawMutex {
/// Create a new `CriticalSectionRawMutex`.
pub const fn new() -> Self {
Self { _phantom: PhantomData }
}
}
unsafe impl RawMutex for CriticalSectionRawMutex {
const INIT: Self = Self::new();
fn lock<R>(&self, f: impl FnOnce() -> R) -> R {
critical_section::with(|_| f())
}
}
// ================
/// A mutex that allows borrowing data in the context of a single executor.
///
/// # Safety
///
/// **This Mutex is only safe within a single executor.**
pub struct NoopRawMutex {
_phantom: PhantomData<*mut ()>,
}
unsafe impl Send for NoopRawMutex {}
impl NoopRawMutex {
/// Create a new `NoopRawMutex`.
pub const fn new() -> Self {
Self { _phantom: PhantomData }
}
}
unsafe impl RawMutex for NoopRawMutex {
const INIT: Self = Self::new();
fn lock<R>(&self, f: impl FnOnce() -> R) -> R {
f()
}
}
// ================
#[cfg(any(cortex_m, feature = "std"))]
mod thread_mode {
use super::*;
/// A "mutex" that only allows borrowing from thread mode.
///
/// # Safety
///
/// **This Mutex is only safe on single-core systems.**
///
/// On multi-core systems, a `ThreadModeRawMutex` **is not sufficient** to ensure exclusive access.
pub struct ThreadModeRawMutex {
_phantom: PhantomData<()>,
}
unsafe impl Send for ThreadModeRawMutex {}
unsafe impl Sync for ThreadModeRawMutex {}
impl ThreadModeRawMutex {
/// Create a new `ThreadModeRawMutex`.
pub const fn new() -> Self {
Self { _phantom: PhantomData }
}
}
unsafe impl RawMutex for ThreadModeRawMutex {
const INIT: Self = Self::new();
fn lock<R>(&self, f: impl FnOnce() -> R) -> R {
assert!(in_thread_mode(), "ThreadModeMutex can only be locked from thread mode.");
f()
}
}
impl Drop for ThreadModeRawMutex {
fn drop(&mut self) {
// Only allow dropping from thread mode. Dropping calls drop on the inner `T`, so
// `drop` needs the same guarantees as `lock`. `ThreadModeMutex<T>` is Send even if
// T isn't, so without this check a user could create a ThreadModeMutex in thread mode,
// send it to interrupt context and drop it there, which would "send" a T even if T is not Send.
assert!(
in_thread_mode(),
"ThreadModeMutex can only be dropped from thread mode."
);
// Drop of the inner `T` happens after this.
}
}
pub(crate) fn in_thread_mode() -> bool {
#[cfg(feature = "std")]
return Some("main") == std::thread::current().name();
#[cfg(not(feature = "std"))]
// ICSR.VECTACTIVE == 0
return unsafe { (0xE000ED04 as *const u32).read_volatile() } & 0x1FF == 0;
}
}
#[cfg(any(cortex_m, feature = "std"))]
pub use thread_mode::*;