Merge pull request #1774 from bugadani/executor

Allow custom executors without indirection

8 files changed, 124 insertions, 153 deletions

diff --git a/embassy-executor/CHANGELOG.md b/embassy-executor/CHANGELOG.md
index 86a477589..e2e7bce3a 100644
--- a/embassy-executor/CHANGELOG.md
+++ b/embassy-executor/CHANGELOG.md
@@ -5,6 +5,10 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## Unreleased
+
+- Replaced Pender. Implementations now must define an extern function called `__pender`.
+
 ## 0.2.1 - 2023-08-10
 
 - Avoid calling `pend()` when waking expired timers
diff --git a/embassy-executor/Cargo.toml b/embassy-executor/Cargo.toml
index ce5e2741f..d190c95a3 100644
--- a/embassy-executor/Cargo.toml
+++ b/embassy-executor/Cargo.toml
@@ -14,7 +14,7 @@ categories = [
 [package.metadata.embassy_docs]
 src_base = "https://github.com/embassy-rs/embassy/blob/embassy-executor-v$VERSION/embassy-executor/src/"
 src_base_git = "https://github.com/embassy-rs/embassy/blob/$COMMIT/embassy-executor/src/"
-features = ["nightly", "defmt", "pender-callback"]
+features = ["nightly", "defmt"]
 flavors = [
     { name = "std",             target = "x86_64-unknown-linux-gnu",     features = ["arch-std", "executor-thread"] },
     { name = "wasm",            target = "wasm32-unknown-unknown",       features = ["arch-wasm", "executor-thread"] },
@@ -25,7 +25,7 @@ flavors = [
 [package.metadata.docs.rs]
 default-target = "thumbv7em-none-eabi"
 targets = ["thumbv7em-none-eabi"]
-features = ["nightly", "defmt", "pender-callback", "arch-cortex-m", "executor-thread", "executor-interrupt"]
+features = ["nightly", "defmt", "arch-cortex-m", "executor-thread", "executor-interrupt"]
 
 [features]
 
@@ -37,9 +37,6 @@ arch-xtensa = ["_arch"]
 arch-riscv32 = ["_arch"]
 arch-wasm = ["_arch", "dep:wasm-bindgen", "dep:js-sys"]
 
-# Enable creating a `Pender` from an arbitrary function pointer callback.
-pender-callback = []
-
 # Enable the thread-mode executor (using WFE/SEV in Cortex-M, WFI in other embedded archs)
 executor-thread = []
 # Enable the interrupt-mode executor (available in Cortex-M only)
diff --git a/embassy-executor/src/arch/cortex_m.rs b/embassy-executor/src/arch/cortex_m.rs
index 94c8134d6..0806a22ab 100644
--- a/embassy-executor/src/arch/cortex_m.rs
+++ b/embassy-executor/src/arch/cortex_m.rs
@@ -1,3 +1,49 @@
+const THREAD_PENDER: usize = usize::MAX;
+
+#[export_name = "__pender"]
+#[cfg(any(feature = "executor-thread", feature = "executor-interrupt"))]
+fn __pender(context: *mut ()) {
+    unsafe {
+        // Safety: `context` is either `usize::MAX` created by `Executor::run`, or a valid interrupt
+        // request number given to `InterruptExecutor::start`.
+
+        let context = context as usize;
+
+        #[cfg(feature = "executor-thread")]
+        // Try to make Rust optimize the branching away if we only use thread mode.
+        if !cfg!(feature = "executor-interrupt") || context == THREAD_PENDER {
+            core::arch::asm!("sev");
+            return;
+        }
+
+        #[cfg(feature = "executor-interrupt")]
+        {
+            use cortex_m::interrupt::InterruptNumber;
+            use cortex_m::peripheral::NVIC;
+
+            #[derive(Clone, Copy)]
+            struct Irq(u16);
+            unsafe impl InterruptNumber for Irq {
+                fn number(self) -> u16 {
+                    self.0
+                }
+            }
+
+            let irq = Irq(context as u16);
+
+            // STIR is faster, but is only available in v7 and higher.
+            #[cfg(not(armv6m))]
+            {
+                let mut nvic: NVIC = core::mem::transmute(());
+                nvic.request(irq);
+            }
+
+            #[cfg(armv6m)]
+            NVIC::pend(irq);
+        }
+    }
+}
+
 #[cfg(feature = "executor-thread")]
 pub use thread::*;
 #[cfg(feature = "executor-thread")]
@@ -8,18 +54,9 @@ mod thread {
     #[cfg(feature = "nightly")]
     pub use embassy_macros::main_cortex_m as main;
 
-    use crate::raw::{Pender, PenderInner};
+    use crate::arch::THREAD_PENDER;
     use crate::{raw, Spawner};
 
-    #[derive(Copy, Clone)]
-    pub(crate) struct ThreadPender;
-
-    impl ThreadPender {
-        pub(crate) fn pend(self) {
-            unsafe { core::arch::asm!("sev") }
-        }
-    }
-
     /// Thread mode executor, using WFE/SEV.
     ///
     /// This is the simplest and most common kind of executor. It runs on
@@ -39,7 +76,7 @@ mod thread {
         /// Create a new Executor.
         pub fn new() -> Self {
             Self {
-                inner: raw::Executor::new(Pender(PenderInner::Thread(ThreadPender))),
+                inner: raw::Executor::new(THREAD_PENDER as *mut ()),
                 not_send: PhantomData,
             }
         }
@@ -86,30 +123,7 @@ mod interrupt {
     use cortex_m::interrupt::InterruptNumber;
     use cortex_m::peripheral::NVIC;
 
-    use crate::raw::{self, Pender, PenderInner};
-
-    #[derive(Clone, Copy)]
-    pub(crate) struct InterruptPender(u16);
-
-    impl InterruptPender {
-        pub(crate) fn pend(self) {
-            // STIR is faster, but is only available in v7 and higher.
-            #[cfg(not(armv6m))]
-            {
-                let mut nvic: cortex_m::peripheral::NVIC = unsafe { core::mem::transmute(()) };
-                nvic.request(self);
-            }
-
-            #[cfg(armv6m)]
-            cortex_m::peripheral::NVIC::pend(self);
-        }
-    }
-
-    unsafe impl cortex_m::interrupt::InterruptNumber for InterruptPender {
-        fn number(self) -> u16 {
-            self.0
-        }
-    }
+    use crate::raw;
 
     /// Interrupt mode executor.
     ///
@@ -194,9 +208,7 @@ mod interrupt {
             unsafe {
                 (&mut *self.executor.get())
                     .as_mut_ptr()
-                    .write(raw::Executor::new(Pender(PenderInner::Interrupt(InterruptPender(
-                        irq.number(),
-                    )))))
+                    .write(raw::Executor::new(irq.number() as *mut ()))
             }
 
             let executor = unsafe { (&*self.executor.get()).assume_init_ref() };
diff --git a/embassy-executor/src/arch/riscv32.rs b/embassy-executor/src/arch/riscv32.rs
index ff7ec1575..40c6877e2 100644
--- a/embassy-executor/src/arch/riscv32.rs
+++ b/embassy-executor/src/arch/riscv32.rs
@@ -11,22 +11,16 @@ mod thread {
     #[cfg(feature = "nightly")]
     pub use embassy_macros::main_riscv as main;
 
-    use crate::raw::{Pender, PenderInner};
     use crate::{raw, Spawner};
 
-    #[derive(Copy, Clone)]
-    pub(crate) struct ThreadPender;
-
-    impl ThreadPender {
-        #[allow(unused)]
-        pub(crate) fn pend(self) {
-            SIGNAL_WORK_THREAD_MODE.store(true, core::sync::atomic::Ordering::SeqCst);
-        }
-    }
-
     /// global atomic used to keep track of whether there is work to do since sev() is not available on RISCV
     static SIGNAL_WORK_THREAD_MODE: AtomicBool = AtomicBool::new(false);
 
+    #[export_name = "__pender"]
+    fn __pender(_context: *mut ()) {
+        SIGNAL_WORK_THREAD_MODE.store(true, Ordering::SeqCst);
+    }
+
     /// RISCV32 Executor
     pub struct Executor {
         inner: raw::Executor,
@@ -37,7 +31,7 @@ mod thread {
         /// Create a new Executor.
         pub fn new() -> Self {
             Self {
-                inner: raw::Executor::new(Pender(PenderInner::Thread(ThreadPender))),
+                inner: raw::Executor::new(core::ptr::null_mut()),
                 not_send: PhantomData,
             }
         }
diff --git a/embassy-executor/src/arch/std.rs b/embassy-executor/src/arch/std.rs
index 4e4a178f0..5b2f7e2e4 100644
--- a/embassy-executor/src/arch/std.rs
+++ b/embassy-executor/src/arch/std.rs
@@ -11,17 +11,12 @@ mod thread {
     #[cfg(feature = "nightly")]
     pub use embassy_macros::main_std as main;
 
-    use crate::raw::{Pender, PenderInner};
     use crate::{raw, Spawner};
 
-    #[derive(Copy, Clone)]
-    pub(crate) struct ThreadPender(&'static Signaler);
-
-    impl ThreadPender {
-        #[allow(unused)]
-        pub(crate) fn pend(self) {
-            self.0.signal()
-        }
+    #[export_name = "__pender"]
+    fn __pender(context: *mut ()) {
+        let signaler: &'static Signaler = unsafe { std::mem::transmute(context) };
+        signaler.signal()
     }
 
     /// Single-threaded std-based executor.
@@ -34,9 +29,9 @@ mod thread {
     impl Executor {
         /// Create a new Executor.
         pub fn new() -> Self {
-            let signaler = &*Box::leak(Box::new(Signaler::new()));
+            let signaler = Box::leak(Box::new(Signaler::new()));
             Self {
-                inner: raw::Executor::new(Pender(PenderInner::Thread(ThreadPender(signaler)))),
+                inner: raw::Executor::new(signaler as *mut Signaler as *mut ()),
                 not_send: PhantomData,
                 signaler,
             }
diff --git a/embassy-executor/src/arch/wasm.rs b/embassy-executor/src/arch/wasm.rs
index 08ab16b99..934fd69e5 100644
--- a/embassy-executor/src/arch/wasm.rs
+++ b/embassy-executor/src/arch/wasm.rs
@@ -14,14 +14,12 @@ mod thread {
     use wasm_bindgen::prelude::*;
 
     use crate::raw::util::UninitCell;
-    use crate::raw::{Pender, PenderInner};
     use crate::{raw, Spawner};
 
-    /// WASM executor, wasm_bindgen to schedule tasks on the JS event loop.
-    pub struct Executor {
-        inner: raw::Executor,
-        ctx: &'static WasmContext,
-        not_send: PhantomData<*mut ()>,
+    #[export_name = "__pender"]
+    fn __pender(context: *mut ()) {
+        let signaler: &'static WasmContext = unsafe { std::mem::transmute(context) };
+        let _ = signaler.promise.then(unsafe { signaler.closure.as_mut() });
     }
 
     pub(crate) struct WasmContext {
@@ -29,16 +27,6 @@ mod thread {
         closure: UninitCell<Closure<dyn FnMut(JsValue)>>,
     }
 
-    #[derive(Copy, Clone)]
-    pub(crate) struct ThreadPender(&'static WasmContext);
-
-    impl ThreadPender {
-        #[allow(unused)]
-        pub(crate) fn pend(self) {
-            let _ = self.0.promise.then(unsafe { self.0.closure.as_mut() });
-        }
-    }
-
     impl WasmContext {
         pub fn new() -> Self {
             Self {
@@ -48,14 +36,21 @@ mod thread {
         }
     }
 
+    /// WASM executor, wasm_bindgen to schedule tasks on the JS event loop.
+    pub struct Executor {
+        inner: raw::Executor,
+        ctx: &'static WasmContext,
+        not_send: PhantomData<*mut ()>,
+    }
+
     impl Executor {
         /// Create a new Executor.
         pub fn new() -> Self {
-            let ctx = &*Box::leak(Box::new(WasmContext::new()));
+            let ctx = Box::leak(Box::new(WasmContext::new()));
             Self {
-                inner: raw::Executor::new(Pender(PenderInner::Thread(ThreadPender(ctx)))),
-                not_send: PhantomData,
+                inner: raw::Executor::new(ctx as *mut WasmContext as *mut ()),
                 ctx,
+                not_send: PhantomData,
             }
         }
 
diff --git a/embassy-executor/src/arch/xtensa.rs b/embassy-executor/src/arch/xtensa.rs
index 017b2c52b..601d85002 100644
--- a/embassy-executor/src/arch/xtensa.rs
+++ b/embassy-executor/src/arch/xtensa.rs
@@ -8,22 +8,16 @@ mod thread {
     use core::marker::PhantomData;
     use core::sync::atomic::{AtomicBool, Ordering};
 
-    use crate::raw::{Pender, PenderInner};
     use crate::{raw, Spawner};
 
-    #[derive(Copy, Clone)]
-    pub(crate) struct ThreadPender;
-
-    impl ThreadPender {
-        #[allow(unused)]
-        pub(crate) fn pend(self) {
-            SIGNAL_WORK_THREAD_MODE.store(true, core::sync::atomic::Ordering::SeqCst);
-        }
-    }
-
     /// global atomic used to keep track of whether there is work to do since sev() is not available on Xtensa
     static SIGNAL_WORK_THREAD_MODE: AtomicBool = AtomicBool::new(false);
 
+    #[export_name = "__pender"]
+    fn __pender(_context: *mut ()) {
+        SIGNAL_WORK_THREAD_MODE.store(true, Ordering::SeqCst);
+    }
+
     /// Xtensa Executor
     pub struct Executor {
         inner: raw::Executor,
@@ -34,7 +28,7 @@ mod thread {
         /// Create a new Executor.
         pub fn new() -> Self {
             Self {
-                inner: raw::Executor::new(Pender(PenderInner::Thread(ThreadPender))),
+                inner: raw::Executor::new(core::ptr::null_mut()),
                 not_send: PhantomData,
             }
         }
@@ -77,8 +71,8 @@ mod thread {
                         SIGNAL_WORK_THREAD_MODE.store(false, Ordering::SeqCst);
 
                         core::arch::asm!(
-                            "wsr.ps {0}",
-                            "rsync", in(reg) token)
+                        "wsr.ps {0}",
+                        "rsync", in(reg) token)
                     } else {
                         // waiti sets the PS.INTLEVEL when slipping into sleep
                         // because critical sections in Xtensa are implemented via increasing
diff --git a/embassy-executor/src/raw/mod.rs b/embassy-executor/src/raw/mod.rs
index 25c2ab0da..7caa3302f 100644
--- a/embassy-executor/src/raw/mod.rs
+++ b/embassy-executor/src/raw/mod.rs
@@ -292,53 +292,17 @@ impl<F: Future + 'static, const N: usize> TaskPool<F, N> {
 }
 
 #[derive(Clone, Copy)]
-pub(crate) enum PenderInner {
-    #[cfg(feature = "executor-thread")]
-    Thread(crate::arch::ThreadPender),
-    #[cfg(feature = "executor-interrupt")]
-    Interrupt(crate::arch::InterruptPender),
-    #[cfg(feature = "pender-callback")]
-    Callback { func: fn(*mut ()), context: *mut () },
-}
-
-unsafe impl Send for PenderInner {}
-unsafe impl Sync for PenderInner {}
+pub(crate) struct Pender(*mut ());
 
-/// Platform/architecture-specific action executed when an executor has pending work.
-///
-/// When a task within an executor is woken, the `Pender` is called. This does a
-/// platform/architecture-specific action to signal there is pending work in the executor.
-/// When this happens, you must arrange for [`Executor::poll`] to be called.
-///
-/// You can think of it as a waker, but for the whole executor.
-pub struct Pender(pub(crate) PenderInner);
+unsafe impl Send for Pender {}
+unsafe impl Sync for Pender {}
 
 impl Pender {
-    /// Create a `Pender` that will call an arbitrary function pointer.
-    ///
-    /// # Arguments
-    ///
-    /// - `func`: The function pointer to call.
-    /// - `context`: Opaque context pointer, that will be passed to the function pointer.
-    #[cfg(feature = "pender-callback")]
-    pub fn new_from_callback(func: fn(*mut ()), context: *mut ()) -> Self {
-        Self(PenderInner::Callback {
-            func,
-            context: context.into(),
-        })
-    }
-}
-
-impl Pender {
-    pub(crate) fn pend(&self) {
-        match self.0 {
-            #[cfg(feature = "executor-thread")]
-            PenderInner::Thread(x) => x.pend(),
-            #[cfg(feature = "executor-interrupt")]
-            PenderInner::Interrupt(x) => x.pend(),
-            #[cfg(feature = "pender-callback")]
-            PenderInner::Callback { func, context } => func(context),
+    pub(crate) fn pend(self) {
+        extern "Rust" {
+            fn __pender(context: *mut ());
         }
+        unsafe { __pender(self.0) };
     }
 }
 
@@ -472,15 +436,31 @@ impl SyncExecutor {
 ///
 /// - To get the executor to do work, call `poll()`. This will poll all queued tasks (all tasks
 ///   that "want to run").
-/// - You must supply a [`Pender`]. The executor will call it to notify you it has work
-///   to do. You must arrange for `poll()` to be called as soon as possible.
+/// - You must supply a pender function, as shown below. The executor will call it to notify you
+///   it has work to do. You must arrange for `poll()` to be called as soon as possible.
+/// - Enabling `arch-xx` features will define a pender function for you. This means that you
+///   are limited to using the executors provided to you by the architecture/platform
+///   implementation. If you need a different executor, you must not enable `arch-xx` features.
 ///
-/// The [`Pender`] can be called from *any* context: any thread, any interrupt priority
+/// The pender can be called from *any* context: any thread, any interrupt priority
 /// level, etc. It may be called synchronously from any `Executor` method call as well.
 /// You must deal with this correctly.
 ///
 /// In particular, you must NOT call `poll` directly from the pender callback, as this violates
 /// the requirement for `poll` to not be called reentrantly.
+///
+/// The pender function must be exported with the name `__pender` and have the following signature:
+///
+/// ```rust
+/// #[export_name = "__pender"]
+/// fn pender(context: *mut ()) {
+///    // schedule `poll()` to be called
+/// }
+/// ```
+///
+/// The `context` argument is a piece of arbitrary data the executor will pass to the pender.
+/// You can set the `context` when calling [`Executor::new()`]. You can use it to, for example,
+/// differentiate between executors, or to pass a pointer to a callback that should be called.
 #[repr(transparent)]
 pub struct Executor {
     pub(crate) inner: SyncExecutor,
@@ -495,12 +475,12 @@ impl Executor {
 
     /// Create a new executor.
     ///
-    /// When the executor has work to do, it will call the [`Pender`].
+    /// When the executor has work to do, it will call the pender function and pass `context` to it.
     ///
-    /// See [`Executor`] docs for details on `Pender`.
-    pub fn new(pender: Pender) -> Self {
+    /// See [`Executor`] docs for details on the pender.
+    pub fn new(context: *mut ()) -> Self {
         Self {
-            inner: SyncExecutor::new(pender),
+            inner: SyncExecutor::new(Pender(context)),
             _not_sync: PhantomData,
         }
     }
@@ -523,16 +503,16 @@ impl Executor {
     /// This loops over all tasks that are queued to be polled (i.e. they're
     /// freshly spawned or they've been woken). Other tasks are not polled.
     ///
-    /// You must call `poll` after receiving a call to the [`Pender`]. It is OK
-    /// to call `poll` even when not requested by the `Pender`, but it wastes
+    /// You must call `poll` after receiving a call to the pender. It is OK
+    /// to call `poll` even when not requested by the pender, but it wastes
     /// energy.
     ///
     /// # Safety
     ///
     /// You must NOT call `poll` reentrantly on the same executor.
     ///
-    /// In particular, note that `poll` may call the `Pender` synchronously. Therefore, you
-    /// must NOT directly call `poll()` from the `Pender` callback. Instead, the callback has to
+    /// In particular, note that `poll` may call the pender synchronously. Therefore, you
+    /// must NOT directly call `poll()` from the pender callback. Instead, the callback has to
     /// somehow schedule for `poll()` to be called later, at a time you know for sure there's
     /// no `poll()` already running.
     pub unsafe fn poll(&'static self) {