9 files changed, 701 insertions, 376 deletions

diff --git a/embassy-executor/src/raw/mod.rs b/embassy-executor/src/raw/mod.rs
index d9ea5c005..913da2e25 100644
--- a/embassy-executor/src/raw/mod.rs
+++ b/embassy-executor/src/raw/mod.rs
@@ -11,13 +11,14 @@
 #[cfg_attr(not(target_has_atomic = "ptr"), path = "run_queue_critical_section.rs")]
 mod run_queue;
 
-#[cfg_attr(all(cortex_m, target_has_atomic = "8"), path = "state_atomics_arm.rs")]
+#[cfg_attr(all(cortex_m, target_has_atomic = "32"), path = "state_atomics_arm.rs")]
 #[cfg_attr(all(not(cortex_m), target_has_atomic = "8"), path = "state_atomics.rs")]
 #[cfg_attr(not(target_has_atomic = "8"), path = "state_critical_section.rs")]
 mod state;
 
-#[cfg(feature = "integrated-timers")]
-mod timer_queue;
+pub mod timer_queue;
+#[cfg(feature = "trace")]
+pub mod trace;
 pub(crate) mod util;
 #[cfg_attr(feature = "turbowakers", path = "waker_turbo.rs")]
 mod waker;
@@ -27,12 +28,13 @@ use core::marker::PhantomData;
 use core::mem;
 use core::pin::Pin;
 use core::ptr::NonNull;
+#[cfg(not(feature = "arch-avr"))]
+use core::sync::atomic::AtomicPtr;
+use core::sync::atomic::Ordering;
 use core::task::{Context, Poll};
 
-#[cfg(feature = "integrated-timers")]
-use embassy_time_driver::AlarmHandle;
-#[cfg(feature = "rtos-trace")]
-use rtos_trace::trace;
+#[cfg(feature = "arch-avr")]
+use portable_atomic::AtomicPtr;
 
 use self::run_queue::{RunQueue, RunQueueItem};
 use self::state::State;
@@ -41,20 +43,62 @@ pub use self::waker::task_from_waker;
 use super::SpawnToken;
 
 /// Raw task header for use in task pointers.
+///
+/// A task can be in one of the following states:
+///
+/// - Not spawned: the task is ready to spawn.
+/// - `SPAWNED`: the task is currently spawned and may be running.
+/// - `RUN_ENQUEUED`: the task is enqueued to be polled. Note that the task may be `!SPAWNED`.
+///    In this case, the `RUN_ENQUEUED` state will be cleared when the task is next polled, without
+///    polling the task's future.
+///
+/// A task's complete life cycle is as follows:
+///
+/// ```text
+/// ┌────────────┐   ┌────────────────────────┐
+/// │Not spawned │◄─5┤Not spawned|Run enqueued│
+/// │            ├6─►│                        │
+/// └─────┬──────┘   └──────▲─────────────────┘
+///       1                 │
+///       │    ┌────────────┘
+///       │    4
+/// ┌─────▼────┴─────────┐
+/// │Spawned|Run enqueued│
+/// │                    │
+/// └─────┬▲─────────────┘
+///       2│
+///       │3
+/// ┌─────▼┴─────┐
+/// │  Spawned   │
+/// │            │
+/// └────────────┘
+/// ```
+///
+/// Transitions:
+/// - 1: Task is spawned - `AvailableTask::claim -> Executor::spawn`
+/// - 2: During poll - `RunQueue::dequeue_all -> State::run_dequeue`
+/// - 3: Task wakes itself, waker wakes task, or task exits - `Waker::wake -> wake_task -> State::run_enqueue`
+/// - 4: A run-queued task exits - `TaskStorage::poll -> Poll::Ready`
+/// - 5: Task is dequeued. The task's future is not polled, because exiting the task replaces its `poll_fn`.
+/// - 6: A task is waken when it is not spawned - `wake_task -> State::run_enqueue`
 pub(crate) struct TaskHeader {
     pub(crate) state: State,
     pub(crate) run_queue_item: RunQueueItem,
-    pub(crate) executor: SyncUnsafeCell<Option<&'static SyncExecutor>>,
+    pub(crate) executor: AtomicPtr<SyncExecutor>,
     poll_fn: SyncUnsafeCell<Option<unsafe fn(TaskRef)>>,
 
-    #[cfg(feature = "integrated-timers")]
-    pub(crate) expires_at: SyncUnsafeCell<u64>,
-    #[cfg(feature = "integrated-timers")]
+    /// Integrated timer queue storage. This field should not be accessed outside of the timer queue.
     pub(crate) timer_queue_item: timer_queue::TimerQueueItem,
+    #[cfg(feature = "trace")]
+    pub(crate) name: Option<&'static str>,
+    #[cfg(feature = "trace")]
+    pub(crate) id: u32,
+    #[cfg(feature = "trace")]
+    all_tasks_next: AtomicPtr<TaskHeader>,
 }
 
 /// This is essentially a `&'static TaskStorage<F>` where the type of the future has been erased.
-#[derive(Clone, Copy)]
+#[derive(Clone, Copy, PartialEq)]
 pub struct TaskRef {
     ptr: NonNull<TaskHeader>,
 }
@@ -76,10 +120,31 @@ impl TaskRef {
         }
     }
 
+    /// # Safety
+    ///
+    /// The result of this function must only be compared
+    /// for equality, or stored, but not used.
+    pub const unsafe fn dangling() -> Self {
+        Self {
+            ptr: NonNull::dangling(),
+        }
+    }
+
     pub(crate) fn header(self) -> &'static TaskHeader {
         unsafe { self.ptr.as_ref() }
     }
 
+    /// Returns a reference to the executor that the task is currently running on.
+    pub unsafe fn executor(self) -> Option<&'static Executor> {
+        let executor = self.header().executor.load(Ordering::Relaxed);
+        executor.as_ref().map(|e| Executor::wrap(e))
+    }
+
+    /// Returns a reference to the timer queue item.
+    pub fn timer_queue_item(&self) -> &'static timer_queue::TimerQueueItem {
+        &self.header().timer_queue_item
+    }
+
     /// The returned pointer is valid for the entire TaskStorage.
     pub(crate) fn as_ptr(self) -> *const TaskHeader {
         self.ptr.as_ptr()
@@ -107,6 +172,10 @@ pub struct TaskStorage<F: Future + 'static> {
     future: UninitCell<F>, // Valid if STATE_SPAWNED
 }
 
+unsafe fn poll_exited(_p: TaskRef) {
+    // Nothing to do, the task is already !SPAWNED and dequeued.
+}
+
 impl<F: Future + 'static> TaskStorage<F> {
     const NEW: Self = Self::new();
 
@@ -116,14 +185,17 @@ impl<F: Future + 'static> TaskStorage<F> {
             raw: TaskHeader {
                 state: State::new(),
                 run_queue_item: RunQueueItem::new(),
-                executor: SyncUnsafeCell::new(None),
+                executor: AtomicPtr::new(core::ptr::null_mut()),
                 // Note: this is lazily initialized so that a static `TaskStorage` will go in `.bss`
                 poll_fn: SyncUnsafeCell::new(None),
 
-                #[cfg(feature = "integrated-timers")]
-                expires_at: SyncUnsafeCell::new(0),
-                #[cfg(feature = "integrated-timers")]
                 timer_queue_item: timer_queue::TimerQueueItem::new(),
+                #[cfg(feature = "trace")]
+                name: None,
+                #[cfg(feature = "trace")]
+                id: 0,
+                #[cfg(feature = "trace")]
+                all_tasks_next: AtomicPtr::new(core::ptr::null_mut()),
             },
             future: UninitCell::uninit(),
         }
@@ -151,18 +223,30 @@ impl<F: Future + 'static> TaskStorage<F> {
     }
 
     unsafe fn poll(p: TaskRef) {
-        let this = &*(p.as_ptr() as *const TaskStorage<F>);
+        let this = &*p.as_ptr().cast::<TaskStorage<F>>();
 
         let future = Pin::new_unchecked(this.future.as_mut());
         let waker = waker::from_task(p);
         let mut cx = Context::from_waker(&waker);
         match future.poll(&mut cx) {
             Poll::Ready(_) => {
+                #[cfg(feature = "trace")]
+                let exec_ptr: *const SyncExecutor = this.raw.executor.load(Ordering::Relaxed);
+
+                // As the future has finished and this function will not be called
+                // again, we can safely drop the future here.
                 this.future.drop_in_place();
+
+                // We replace the poll_fn with a despawn function, so that the task is cleaned up
+                // when the executor polls it next.
+                this.raw.poll_fn.set(Some(poll_exited));
+
+                // Make sure we despawn last, so that other threads can only spawn the task
+                // after we're done with it.
                 this.raw.state.despawn();
 
-                #[cfg(feature = "integrated-timers")]
-                this.raw.expires_at.set(u64::MAX);
+                #[cfg(feature = "trace")]
+                trace::task_end(exec_ptr, &p);
             }
             Poll::Pending => {}
         }
@@ -316,26 +400,13 @@ impl Pender {
 pub(crate) struct SyncExecutor {
     run_queue: RunQueue,
     pender: Pender,
-
-    #[cfg(feature = "integrated-timers")]
-    pub(crate) timer_queue: timer_queue::TimerQueue,
-    #[cfg(feature = "integrated-timers")]
-    alarm: AlarmHandle,
 }
 
 impl SyncExecutor {
     pub(crate) fn new(pender: Pender) -> Self {
-        #[cfg(feature = "integrated-timers")]
-        let alarm = unsafe { unwrap!(embassy_time_driver::allocate_alarm()) };
-
         Self {
             run_queue: RunQueue::new(),
             pender,
-
-            #[cfg(feature = "integrated-timers")]
-            timer_queue: timer_queue::TimerQueue::new(),
-            #[cfg(feature = "integrated-timers")]
-            alarm,
         }
     }
 
@@ -346,90 +417,50 @@ impl SyncExecutor {
     /// - `task` must be set up to run in this executor.
     /// - `task` must NOT be already enqueued (in this executor or another one).
     #[inline(always)]
-    unsafe fn enqueue(&self, task: TaskRef) {
-        #[cfg(feature = "rtos-trace")]
-        trace::task_ready_begin(task.as_ptr() as u32);
+    unsafe fn enqueue(&self, task: TaskRef, l: state::Token) {
+        #[cfg(feature = "trace")]
+        trace::task_ready_begin(self, &task);
 
-        if self.run_queue.enqueue(task) {
+        if self.run_queue.enqueue(task, l) {
             self.pender.pend();
         }
     }
 
-    #[cfg(feature = "integrated-timers")]
-    fn alarm_callback(ctx: *mut ()) {
-        let this: &Self = unsafe { &*(ctx as *const Self) };
-        this.pender.pend();
-    }
-
     pub(super) unsafe fn spawn(&'static self, task: TaskRef) {
-        task.header().executor.set(Some(self));
+        task.header()
+            .executor
+            .store((self as *const Self).cast_mut(), Ordering::Relaxed);
 
-        #[cfg(feature = "rtos-trace")]
-        trace::task_new(task.as_ptr() as u32);
+        #[cfg(feature = "trace")]
+        trace::task_new(self, &task);
 
-        self.enqueue(task);
+        state::locked(|l| {
+            self.enqueue(task, l);
+        })
     }
 
     /// # Safety
     ///
     /// Same as [`Executor::poll`], plus you must only call this on the thread this executor was created.
     pub(crate) unsafe fn poll(&'static self) {
-        #[cfg(feature = "integrated-timers")]
-        embassy_time_driver::set_alarm_callback(self.alarm, Self::alarm_callback, self as *const _ as *mut ());
-
-        #[allow(clippy::never_loop)]
-        loop {
-            #[cfg(feature = "integrated-timers")]
-            self.timer_queue
-                .dequeue_expired(embassy_time_driver::now(), wake_task_no_pend);
-
-            self.run_queue.dequeue_all(|p| {
-                let task = p.header();
-
-                #[cfg(feature = "integrated-timers")]
-                task.expires_at.set(u64::MAX);
-
-                if !task.state.run_dequeue() {
-                    // If task is not running, ignore it. This can happen in the following scenario:
-                    //   - Task gets dequeued, poll starts
-                    //   - While task is being polled, it gets woken. It gets placed in the queue.
-                    //   - Task poll finishes, returning done=true
-                    //   - RUNNING bit is cleared, but the task is already in the queue.
-                    return;
-                }
-
-                #[cfg(feature = "rtos-trace")]
-                trace::task_exec_begin(p.as_ptr() as u32);
-
-                // Run the task
-                task.poll_fn.get().unwrap_unchecked()(p);
-
-                #[cfg(feature = "rtos-trace")]
-                trace::task_exec_end();
-
-                // Enqueue or update into timer_queue
-                #[cfg(feature = "integrated-timers")]
-                self.timer_queue.update(p);
-            });
-
-            #[cfg(feature = "integrated-timers")]
-            {
-                // If this is already in the past, set_alarm might return false
-                // In that case do another poll loop iteration.
-                let next_expiration = self.timer_queue.next_expiration();
-                if embassy_time_driver::set_alarm(self.alarm, next_expiration) {
-                    break;
-                }
-            }
+        #[cfg(feature = "trace")]
+        trace::poll_start(self);
 
-            #[cfg(not(feature = "integrated-timers"))]
-            {
-                break;
-            }
-        }
+        self.run_queue.dequeue_all(|p| {
+            let task = p.header();
+
+            #[cfg(feature = "trace")]
+            trace::task_exec_begin(self, &p);
 
-        #[cfg(feature = "rtos-trace")]
-        trace::system_idle();
+            // Run the task
+            task.poll_fn.get().unwrap_unchecked()(p);
+
+            #[cfg(feature = "trace")]
+            trace::task_exec_end(self, &p);
+        });
+
+        #[cfg(feature = "trace")]
+        trace::executor_idle(self)
     }
 }
 
@@ -516,6 +547,8 @@ impl Executor {
     ///
     /// # Safety
     ///
+    /// You must call `initialize` before calling this method.
+    ///
     /// You must NOT call `poll` reentrantly on the same executor.
     ///
     /// In particular, note that `poll` may call the pender synchronously. Therefore, you
@@ -533,6 +566,11 @@ impl Executor {
     pub fn spawner(&'static self) -> super::Spawner {
         super::Spawner::new(self)
     }
+
+    /// Get a unique ID for this Executor.
+    pub fn id(&'static self) -> usize {
+        &self.inner as *const SyncExecutor as usize
+    }
 }
 
 /// Wake a task by `TaskRef`.
@@ -540,13 +578,13 @@ impl Executor {
 /// You can obtain a `TaskRef` from a `Waker` using [`task_from_waker`].
 pub fn wake_task(task: TaskRef) {
     let header = task.header();
-    if header.state.run_enqueue() {
+    header.state.run_enqueue(|l| {
         // We have just marked the task as scheduled, so enqueue it.
         unsafe {
-            let executor = header.executor.get().unwrap_unchecked();
-            executor.enqueue(task);
+            let executor = header.executor.load(Ordering::Relaxed).as_ref().unwrap_unchecked();
+            executor.enqueue(task, l);
         }
-    }
+    });
 }
 
 /// Wake a task by `TaskRef` without calling pend.
@@ -554,57 +592,11 @@ pub fn wake_task(task: TaskRef) {
 /// You can obtain a `TaskRef` from a `Waker` using [`task_from_waker`].
 pub fn wake_task_no_pend(task: TaskRef) {
     let header = task.header();
-    if header.state.run_enqueue() {
+    header.state.run_enqueue(|l| {
         // We have just marked the task as scheduled, so enqueue it.
         unsafe {
-            let executor = header.executor.get().unwrap_unchecked();
-            executor.run_queue.enqueue(task);
+            let executor = header.executor.load(Ordering::Relaxed).as_ref().unwrap_unchecked();
+            executor.run_queue.enqueue(task, l);
         }
-    }
+    });
 }
-
-#[cfg(feature = "integrated-timers")]
-struct TimerQueue;
-
-#[cfg(feature = "integrated-timers")]
-impl embassy_time_queue_driver::TimerQueue for TimerQueue {
-    fn schedule_wake(&'static self, at: u64, waker: &core::task::Waker) {
-        let task = waker::task_from_waker(waker);
-        let task = task.header();
-        unsafe {
-            let expires_at = task.expires_at.get();
-            task.expires_at.set(expires_at.min(at));
-        }
-    }
-}
-
-#[cfg(feature = "integrated-timers")]
-embassy_time_queue_driver::timer_queue_impl!(static TIMER_QUEUE: TimerQueue = TimerQueue);
-
-#[cfg(all(feature = "rtos-trace", feature = "integrated-timers"))]
-const fn gcd(a: u64, b: u64) -> u64 {
-    if b == 0 {
-        a
-    } else {
-        gcd(b, a % b)
-    }
-}
-
-#[cfg(feature = "rtos-trace")]
-impl rtos_trace::RtosTraceOSCallbacks for Executor {
-    fn task_list() {
-        // We don't know what tasks exist, so we can't send them.
-    }
-    #[cfg(feature = "integrated-timers")]
-    fn time() -> u64 {
-        const GCD_1M: u64 = gcd(embassy_time_driver::TICK_HZ, 1_000_000);
-        embassy_time_driver::now() * (1_000_000 / GCD_1M) / (embassy_time_driver::TICK_HZ / GCD_1M)
-    }
-    #[cfg(not(feature = "integrated-timers"))]
-    fn time() -> u64 {
-        0
-    }
-}
-
-#[cfg(feature = "rtos-trace")]
-rtos_trace::global_os_callbacks! {Executor}
diff --git a/embassy-executor/src/raw/run_queue_atomics.rs b/embassy-executor/src/raw/run_queue_atomics.rs
index 90907cfda..ce511d79a 100644
--- a/embassy-executor/src/raw/run_queue_atomics.rs
+++ b/embassy-executor/src/raw/run_queue_atomics.rs
@@ -45,7 +45,7 @@ impl RunQueue {
     ///
     /// `item` must NOT be already enqueued in any queue.
     #[inline(always)]
-    pub(crate) unsafe fn enqueue(&self, task: TaskRef) -> bool {
+    pub(crate) unsafe fn enqueue(&self, task: TaskRef, _: super::state::Token) -> bool {
         let mut was_empty = false;
 
         self.head
@@ -81,6 +81,7 @@ impl RunQueue {
             // safety: there are no concurrent accesses to `next`
             next = unsafe { task.header().run_queue_item.next.get() };
 
+            task.header().state.run_dequeue();
             on_task(task);
         }
     }
diff --git a/embassy-executor/src/raw/run_queue_critical_section.rs b/embassy-executor/src/raw/run_queue_critical_section.rs
index ba59c8f29..86c4085ed 100644
--- a/embassy-executor/src/raw/run_queue_critical_section.rs
+++ b/embassy-executor/src/raw/run_queue_critical_section.rs
@@ -44,13 +44,11 @@ impl RunQueue {
     ///
     /// `item` must NOT be already enqueued in any queue.
     #[inline(always)]
-    pub(crate) unsafe fn enqueue(&self, task: TaskRef) -> bool {
-        critical_section::with(|cs| {
-            let prev = self.head.borrow(cs).replace(Some(task));
-            task.header().run_queue_item.next.borrow(cs).set(prev);
+    pub(crate) unsafe fn enqueue(&self, task: TaskRef, cs: CriticalSection<'_>) -> bool {
+        let prev = self.head.borrow(cs).replace(Some(task));
+        task.header().run_queue_item.next.borrow(cs).set(prev);
 
-            prev.is_none()
-        })
+        prev.is_none()
     }
 
     /// Empty the queue, then call `on_task` for each task that was in the queue.
@@ -65,9 +63,10 @@ impl RunQueue {
             // If the task re-enqueues itself, the `next` pointer will get overwritten.
             // Therefore, first read the next pointer, and only then process the task.
 
-            // safety: we know if the task is enqueued, no one else will touch the `next` pointer.
-            let cs = unsafe { CriticalSection::new() };
-            next = task.header().run_queue_item.next.borrow(cs).get();
+            critical_section::with(|cs| {
+                next = task.header().run_queue_item.next.borrow(cs).get();
+                task.header().state.run_dequeue(cs);
+            });
 
             on_task(task);
         }
diff --git a/embassy-executor/src/raw/state_atomics.rs b/embassy-executor/src/raw/state_atomics.rs
index e1279ac0b..e813548ae 100644
--- a/embassy-executor/src/raw/state_atomics.rs
+++ b/embassy-executor/src/raw/state_atomics.rs
@@ -1,21 +1,28 @@
-use core::sync::atomic::{AtomicU32, Ordering};
+use core::sync::atomic::{AtomicU8, Ordering};
+
+#[derive(Clone, Copy)]
+pub(crate) struct Token(());
+
+/// Creates a token and passes it to the closure.
+///
+/// This is a no-op replacement for `CriticalSection::with` because we don't need any locking.
+pub(crate) fn locked<R>(f: impl FnOnce(Token) -> R) -> R {
+    f(Token(()))
+}
 
 /// Task is spawned (has a future)
-pub(crate) const STATE_SPAWNED: u32 = 1 << 0;
+pub(crate) const STATE_SPAWNED: u8 = 1 << 0;
 /// Task is in the executor run queue
-pub(crate) const STATE_RUN_QUEUED: u32 = 1 << 1;
-/// Task is in the executor timer queue
-#[cfg(feature = "integrated-timers")]
-pub(crate) const STATE_TIMER_QUEUED: u32 = 1 << 2;
+pub(crate) const STATE_RUN_QUEUED: u8 = 1 << 1;
 
 pub(crate) struct State {
-    state: AtomicU32,
+    state: AtomicU8,
 }
 
 impl State {
     pub const fn new() -> State {
         Self {
-            state: AtomicU32::new(0),
+            state: AtomicU8::new(0),
         }
     }
 
@@ -33,41 +40,19 @@ impl State {
         self.state.fetch_and(!STATE_SPAWNED, Ordering::AcqRel);
     }
 
-    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Return true on success.
+    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Run the given
+    /// function if the task was successfully marked.
     #[inline(always)]
-    pub fn run_enqueue(&self) -> bool {
-        self.state
-            .fetch_update(Ordering::SeqCst, Ordering::SeqCst, |state| {
-                // If already scheduled, or if not started,
-                if (state & STATE_RUN_QUEUED != 0) || (state & STATE_SPAWNED == 0) {
-                    None
-                } else {
-                    // Mark it as scheduled
-                    Some(state | STATE_RUN_QUEUED)
-                }
-            })
-            .is_ok()
+    pub fn run_enqueue(&self, f: impl FnOnce(Token)) {
+        let prev = self.state.fetch_or(STATE_RUN_QUEUED, Ordering::AcqRel);
+        if prev & STATE_RUN_QUEUED == 0 {
+            locked(f);
+        }
     }
 
     /// Unmark the task as run-queued. Return whether the task is spawned.
     #[inline(always)]
-    pub fn run_dequeue(&self) -> bool {
-        let state = self.state.fetch_and(!STATE_RUN_QUEUED, Ordering::AcqRel);
-        state & STATE_SPAWNED != 0
-    }
-
-    /// Mark the task as timer-queued. Return whether it was newly queued (i.e. not queued before)
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_enqueue(&self) -> bool {
-        let old_state = self.state.fetch_or(STATE_TIMER_QUEUED, Ordering::AcqRel);
-        old_state & STATE_TIMER_QUEUED == 0
-    }
-
-    /// Unmark the task as timer-queued.
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_dequeue(&self) {
-        self.state.fetch_and(!STATE_TIMER_QUEUED, Ordering::AcqRel);
+    pub fn run_dequeue(&self) {
+        self.state.fetch_and(!STATE_RUN_QUEUED, Ordering::AcqRel);
     }
 }
diff --git a/embassy-executor/src/raw/state_atomics_arm.rs b/embassy-executor/src/raw/state_atomics_arm.rs
index e4dfe5093..b743dcc2c 100644
--- a/embassy-executor/src/raw/state_atomics_arm.rs
+++ b/embassy-executor/src/raw/state_atomics_arm.rs
@@ -1,6 +1,15 @@
-use core::arch::asm;
 use core::sync::atomic::{compiler_fence, AtomicBool, AtomicU32, Ordering};
 
+#[derive(Clone, Copy)]
+pub(crate) struct Token(());
+
+/// Creates a token and passes it to the closure.
+///
+/// This is a no-op replacement for `CriticalSection::with` because we don't need any locking.
+pub(crate) fn locked<R>(f: impl FnOnce(Token) -> R) -> R {
+    f(Token(()))
+}
+
 // Must be kept in sync with the layout of `State`!
 pub(crate) const STATE_SPAWNED: u32 = 1 << 0;
 pub(crate) const STATE_RUN_QUEUED: u32 = 1 << 8;
@@ -11,9 +20,8 @@ pub(crate) struct State {
     spawned: AtomicBool,
     /// Task is in the executor run queue
     run_queued: AtomicBool,
-    /// Task is in the executor timer queue
-    timer_queued: AtomicBool,
     pad: AtomicBool,
+    pad2: AtomicBool,
 }
 
 impl State {
@@ -21,8 +29,8 @@ impl State {
         Self {
             spawned: AtomicBool::new(false),
             run_queued: AtomicBool::new(false),
-            timer_queued: AtomicBool::new(false),
             pad: AtomicBool::new(false),
+            pad2: AtomicBool::new(false),
         }
     }
 
@@ -54,50 +62,22 @@ impl State {
         self.spawned.store(false, Ordering::Relaxed);
     }
 
-    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Return true on success.
+    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Run the given
+    /// function if the task was successfully marked.
     #[inline(always)]
-    pub fn run_enqueue(&self) -> bool {
-        unsafe {
-            loop {
-                let state: u32;
-                asm!("ldrex {}, [{}]", out(reg) state, in(reg) self, options(nostack));
+    pub fn run_enqueue(&self, f: impl FnOnce(Token)) {
+        let old = self.run_queued.swap(true, Ordering::AcqRel);
 
-                if (state & STATE_RUN_QUEUED != 0) || (state & STATE_SPAWNED == 0) {
-                    asm!("clrex", options(nomem, nostack));
-                    return false;
-                }
-
-                let outcome: usize;
-                let new_state = state | STATE_RUN_QUEUED;
-                asm!("strex {}, {}, [{}]", out(reg) outcome, in(reg) new_state, in(reg) self, options(nostack));
-                if outcome == 0 {
-                    return true;
-                }
-            }
+        if !old {
+            locked(f);
         }
     }
 
     /// Unmark the task as run-queued. Return whether the task is spawned.
     #[inline(always)]
-    pub fn run_dequeue(&self) -> bool {
+    pub fn run_dequeue(&self) {
         compiler_fence(Ordering::Release);
 
-        let r = self.spawned.load(Ordering::Relaxed);
         self.run_queued.store(false, Ordering::Relaxed);
-        r
-    }
-
-    /// Mark the task as timer-queued. Return whether it was newly queued (i.e. not queued before)
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_enqueue(&self) -> bool {
-        !self.timer_queued.swap(true, Ordering::Relaxed)
-    }
-
-    /// Unmark the task as timer-queued.
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_dequeue(&self) {
-        self.timer_queued.store(false, Ordering::Relaxed);
     }
 }
diff --git a/embassy-executor/src/raw/state_critical_section.rs b/embassy-executor/src/raw/state_critical_section.rs
index c3cc1b0b7..ec08f2f58 100644
--- a/embassy-executor/src/raw/state_critical_section.rs
+++ b/embassy-executor/src/raw/state_critical_section.rs
@@ -1,17 +1,15 @@
 use core::cell::Cell;
 
-use critical_section::Mutex;
+pub(crate) use critical_section::{with as locked, CriticalSection as Token};
+use critical_section::{CriticalSection, Mutex};
 
 /// Task is spawned (has a future)
-pub(crate) const STATE_SPAWNED: u32 = 1 << 0;
+pub(crate) const STATE_SPAWNED: u8 = 1 << 0;
 /// Task is in the executor run queue
-pub(crate) const STATE_RUN_QUEUED: u32 = 1 << 1;
-/// Task is in the executor timer queue
-#[cfg(feature = "integrated-timers")]
-pub(crate) const STATE_TIMER_QUEUED: u32 = 1 << 2;
+pub(crate) const STATE_RUN_QUEUED: u8 = 1 << 1;
 
 pub(crate) struct State {
-    state: Mutex<Cell<u32>>,
+    state: Mutex<Cell<u8>>,
 }
 
 impl State {
@@ -21,14 +19,16 @@ impl State {
         }
     }
 
-    fn update<R>(&self, f: impl FnOnce(&mut u32) -> R) -> R {
-        critical_section::with(|cs| {
-            let s = self.state.borrow(cs);
-            let mut val = s.get();
-            let r = f(&mut val);
-            s.set(val);
-            r
-        })
+    fn update<R>(&self, f: impl FnOnce(&mut u8) -> R) -> R {
+        critical_section::with(|cs| self.update_with_cs(cs, f))
+    }
+
+    fn update_with_cs<R>(&self, cs: CriticalSection<'_>, f: impl FnOnce(&mut u8) -> R) -> R {
+        let s = self.state.borrow(cs);
+        let mut val = s.get();
+        let r = f(&mut val);
+        s.set(val);
+        r
     }
 
     /// If task is idle, mark it as spawned + run_queued and return true.
@@ -50,44 +50,24 @@ impl State {
         self.update(|s| *s &= !STATE_SPAWNED);
     }
 
-    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Return true on success.
+    /// Mark the task as run-queued if it's spawned and isn't already run-queued. Run the given
+    /// function if the task was successfully marked.
     #[inline(always)]
-    pub fn run_enqueue(&self) -> bool {
-        self.update(|s| {
-            if (*s & STATE_RUN_QUEUED != 0) || (*s & STATE_SPAWNED == 0) {
-                false
-            } else {
+    pub fn run_enqueue(&self, f: impl FnOnce(Token)) {
+        critical_section::with(|cs| {
+            if self.update_with_cs(cs, |s| {
+                let ok = *s & STATE_RUN_QUEUED == 0;
                 *s |= STATE_RUN_QUEUED;
-                true
+                ok
+            }) {
+                f(cs);
             }
-        })
+        });
     }
 
     /// Unmark the task as run-queued. Return whether the task is spawned.
     #[inline(always)]
-    pub fn run_dequeue(&self) -> bool {
-        self.update(|s| {
-            let ok = *s & STATE_SPAWNED != 0;
-            *s &= !STATE_RUN_QUEUED;
-            ok
-        })
-    }
-
-    /// Mark the task as timer-queued. Return whether it was newly queued (i.e. not queued before)
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_enqueue(&self) -> bool {
-        self.update(|s| {
-            let ok = *s & STATE_TIMER_QUEUED == 0;
-            *s |= STATE_TIMER_QUEUED;
-            ok
-        })
-    }
-
-    /// Unmark the task as timer-queued.
-    #[cfg(feature = "integrated-timers")]
-    #[inline(always)]
-    pub fn timer_dequeue(&self) {
-        self.update(|s| *s &= !STATE_TIMER_QUEUED);
+    pub fn run_dequeue(&self, cs: CriticalSection<'_>) {
+        self.update_with_cs(cs, |s| *s &= !STATE_RUN_QUEUED)
     }
 }
diff --git a/embassy-executor/src/raw/timer_queue.rs b/embassy-executor/src/raw/timer_queue.rs
index 94a5f340b..e52453be4 100644
--- a/embassy-executor/src/raw/timer_queue.rs
+++ b/embassy-executor/src/raw/timer_queue.rs
@@ -1,76 +1,73 @@
-use core::cmp::min;
+//! Timer queue operations.
+
+use core::cell::Cell;
 
 use super::TaskRef;
-use crate::raw::util::SyncUnsafeCell;
 
-pub(crate) struct TimerQueueItem {
-    next: SyncUnsafeCell<Option<TaskRef>>,
-}
+#[cfg(feature = "_timer-item-payload")]
+macro_rules! define_opaque {
+    ($size:tt) => {
+        /// An opaque data type.
+        #[repr(align($size))]
+        pub struct OpaqueData {
+            data: [u8; $size],
+        }
 
-impl TimerQueueItem {
-    pub const fn new() -> Self {
-        Self {
-            next: SyncUnsafeCell::new(None),
+        impl OpaqueData {
+            const fn new() -> Self {
+                Self { data: [0; $size] }
+            }
+
+            /// Access the data as a reference to a type `T`.
+            ///
+            /// Safety:
+            ///
+            /// The caller must ensure that the size of the type `T` is less than, or equal to
+            /// the size of the payload, and must ensure that the alignment of the type `T` is
+            /// less than, or equal to the alignment of the payload.
+            ///
+            /// The type must be valid when zero-initialized.
+            pub unsafe fn as_ref<T>(&self) -> &T {
+                &*(self.data.as_ptr() as *const T)
+            }
         }
-    }
+    };
 }
 
-pub(crate) struct TimerQueue {
-    head: SyncUnsafeCell<Option<TaskRef>>,
-}
+#[cfg(feature = "timer-item-payload-size-1")]
+define_opaque!(1);
+#[cfg(feature = "timer-item-payload-size-2")]
+define_opaque!(2);
+#[cfg(feature = "timer-item-payload-size-4")]
+define_opaque!(4);
+#[cfg(feature = "timer-item-payload-size-8")]
+define_opaque!(8);
 
-impl TimerQueue {
-    pub const fn new() -> Self {
-        Self {
-            head: SyncUnsafeCell::new(None),
-        }
-    }
+/// An item in the timer queue.
+pub struct TimerQueueItem {
+    /// The next item in the queue.
+    ///
+    /// If this field contains `Some`, the item is in the queue. The last item in the queue has a
+    /// value of `Some(dangling_pointer)`
+    pub next: Cell<Option<TaskRef>>,
 
-    pub(crate) unsafe fn update(&self, p: TaskRef) {
-        let task = p.header();
-        if task.expires_at.get() != u64::MAX {
-            if task.state.timer_enqueue() {
-                task.timer_queue_item.next.set(self.head.get());
-                self.head.set(Some(p));
-            }
-        }
-    }
+    /// The time at which this item expires.
+    pub expires_at: Cell<u64>,
 
-    pub(crate) unsafe fn next_expiration(&self) -> u64 {
-        let mut res = u64::MAX;
-        self.retain(|p| {
-            let task = p.header();
-            let expires = task.expires_at.get();
-            res = min(res, expires);
-            expires != u64::MAX
-        });
-        res
-    }
+    /// Some implementation-defined, zero-initialized piece of data.
+    #[cfg(feature = "_timer-item-payload")]
+    pub payload: OpaqueData,
+}
 
-    pub(crate) unsafe fn dequeue_expired(&self, now: u64, on_task: impl Fn(TaskRef)) {
-        self.retain(|p| {
-            let task = p.header();
-            if task.expires_at.get() <= now {
-                on_task(p);
-                false
-            } else {
-                true
-            }
-        });
-    }
+unsafe impl Sync for TimerQueueItem {}
 
-    pub(crate) unsafe fn retain(&self, mut f: impl FnMut(TaskRef) -> bool) {
-        let mut prev = &self.head;
-        while let Some(p) = prev.get() {
-            let task = p.header();
-            if f(p) {
-                // Skip to next
-                prev = &task.timer_queue_item.next;
-            } else {
-                // Remove it
-                prev.set(task.timer_queue_item.next.get());
-                task.state.timer_dequeue();
-            }
+impl TimerQueueItem {
+    pub(crate) const fn new() -> Self {
+        Self {
+            next: Cell::new(None),
+            expires_at: Cell::new(0),
+            #[cfg(feature = "_timer-item-payload")]
+            payload: OpaqueData::new(),
         }
     }
 }
diff --git a/embassy-executor/src/raw/trace.rs b/embassy-executor/src/raw/trace.rs
new file mode 100644
index 000000000..6c9cfda25
--- /dev/null
+++ b/embassy-executor/src/raw/trace.rs
@@ -0,0 +1,412 @@
+//! # Tracing
+//!
+//! The `trace` feature enables a number of callbacks that can be used to track the
+//! lifecycle of tasks and/or executors.
+//!
+//! Callbacks will have one or both of the following IDs passed to them:
+//!
+//! 1. A `task_id`, a `u32` value unique to a task for the duration of the time it is valid
+//! 2. An `executor_id`, a `u32` value unique to an executor for the duration of the time it is
+//!    valid
+//!
+//! Today, both `task_id` and `executor_id` are u32s containing the least significant 32 bits of
+//! the address of the task or executor, however this is NOT a stable guarantee, and MAY change
+//! at any time.
+//!
+//! IDs are only guaranteed to be unique for the duration of time the item is valid. If a task
+//! ends, and is re-spawned, it MAY or MAY NOT have the same ID. For tasks, this valid time is defined
+//! as the time between `_embassy_trace_task_new` and `_embassy_trace_task_end` for a given task.
+//! For executors, this time is not defined, but is often "forever" for practical embedded
+//! programs.
+//!
+//! Callbacks can be used by enabling the `trace` feature, and providing implementations of the
+//! `extern "Rust"` functions below. All callbacks must be implemented.
+//!
+//! ## Task Tracing lifecycle
+//!
+//! ```text
+//! ┌ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─
+//!        │(1)                                            │
+//! │      │
+//!   ╔════▼════╗ (2) ┌─────────┐ (3) ┌─────────┐          │
+//! │ ║ SPAWNED ║────▶│ WAITING │────▶│ RUNNING │
+//!   ╚═════════╝     └─────────┘     └─────────┘          │
+//! │                 ▲         ▲     │    │    │
+//!                   │           (4)      │    │(6)       │
+//! │                 │(7)      └ ─ ─ ┘    │    │
+//!                   │                    │    │          │
+//! │             ┌──────┐             (5) │    │  ┌─────┐
+//!               │ IDLE │◀────────────────┘    └─▶│ END │ │
+//! │             └──────┘                         └─────┘
+//!   ┌──────────────────────┐                             │
+//! └ ┤ Task Trace Lifecycle │─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─
+//!   └──────────────────────┘
+//! ```
+//!
+//! 1. A task is spawned, `_embassy_trace_task_new` is called
+//! 2. A task is enqueued for the first time, `_embassy_trace_task_ready_begin` is called
+//! 3. A task is polled, `_embassy_trace_task_exec_begin` is called
+//! 4. WHILE a task is polled, the task is re-awoken, and `_embassy_trace_task_ready_begin` is
+//!      called. The task does not IMMEDIATELY move state, until polling is complete and the
+//!      RUNNING state is existed. `_embassy_trace_task_exec_end` is called when polling is
+//!      complete, marking the transition to WAITING
+//! 5. Polling is complete, `_embassy_trace_task_exec_end` is called
+//! 6. The task has completed, and `_embassy_trace_task_end` is called
+//! 7. A task is awoken, `_embassy_trace_task_ready_begin` is called
+//!
+//! ## Executor Tracing lifecycle
+//!
+//! ```text
+//! ┌ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─
+//!       │(1)                                             │
+//! │     │
+//!   ╔═══▼══╗   (2)     ┌────────────┐  (3)  ┌─────────┐  │
+//! │ ║ IDLE ║──────────▶│ SCHEDULING │──────▶│ POLLING │
+//!   ╚══════╝           └────────────┘       └─────────┘  │
+//! │     ▲              │            ▲            │
+//!       │      (5)     │            │  (4)       │       │
+//! │     └──────────────┘            └────────────┘
+//!   ┌──────────────────────────┐                         │
+//! └ ┤ Executor Trace Lifecycle │─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─
+//!   └──────────────────────────┘
+//! ```
+//!
+//! 1. The executor is started (no associated trace)
+//! 2. A task on this executor is awoken. `_embassy_trace_task_ready_begin` is called
+//!      when this occurs, and `_embassy_trace_poll_start` is called when the executor
+//!      actually begins running
+//! 3. The executor has decided a task to poll. `_embassy_trace_task_exec_begin` is called
+//! 4. The executor finishes polling the task. `_embassy_trace_task_exec_end` is called
+//! 5. The executor has finished polling tasks. `_embassy_trace_executor_idle` is called
+
+#![allow(unused)]
+
+use core::cell::UnsafeCell;
+use core::sync::atomic::{AtomicPtr, AtomicUsize, Ordering};
+
+use rtos_trace::TaskInfo;
+
+use crate::raw::{SyncExecutor, TaskHeader, TaskRef};
+use crate::spawner::{SpawnError, SpawnToken, Spawner};
+
+/// Global task tracker instance
+///
+/// This static provides access to the global task tracker which maintains
+/// a list of all tasks in the system. It's automatically updated by the
+/// task lifecycle hooks in the trace module.
+pub static TASK_TRACKER: TaskTracker = TaskTracker::new();
+
+/// A thread-safe tracker for all tasks in the system
+///
+/// This struct uses an intrusive linked list approach to track all tasks
+/// without additional memory allocations. It maintains a global list of
+/// tasks that can be traversed to find all currently existing tasks.
+pub struct TaskTracker {
+    head: AtomicPtr<TaskHeader>,
+}
+
+impl TaskTracker {
+    /// Creates a new empty task tracker
+    ///
+    /// Initializes a tracker with no tasks in its list.
+    pub const fn new() -> Self {
+        Self {
+            head: AtomicPtr::new(core::ptr::null_mut()),
+        }
+    }
+
+    /// Adds a task to the tracker
+    ///
+    /// This method inserts a task at the head of the intrusive linked list.
+    /// The operation is thread-safe and lock-free, using atomic operations
+    /// to ensure consistency even when called from different contexts.
+    ///
+    /// # Arguments
+    /// * `task` - The task reference to add to the tracker
+    pub fn add(&self, task: TaskRef) {
+        let task_ptr = task.as_ptr() as *mut TaskHeader;
+
+        loop {
+            let current_head = self.head.load(Ordering::Acquire);
+            unsafe {
+                (*task_ptr).all_tasks_next.store(current_head, Ordering::Relaxed);
+            }
+
+            if self
+                .head
+                .compare_exchange(current_head, task_ptr, Ordering::Release, Ordering::Relaxed)
+                .is_ok()
+            {
+                break;
+            }
+        }
+    }
+
+    /// Performs an operation on each task in the tracker
+    ///
+    /// This method traverses the entire list of tasks and calls the provided
+    /// function for each task. This allows inspecting or processing all tasks
+    /// in the system without modifying the tracker's structure.
+    ///
+    /// # Arguments
+    /// * `f` - A function to call for each task in the tracker
+    pub fn for_each<F>(&self, mut f: F)
+    where
+        F: FnMut(TaskRef),
+    {
+        let mut current = self.head.load(Ordering::Acquire);
+        while !current.is_null() {
+            let task = unsafe { TaskRef::from_ptr(current) };
+            f(task);
+
+            current = unsafe { (*current).all_tasks_next.load(Ordering::Acquire) };
+        }
+    }
+}
+
+/// Extension trait for `TaskRef` that provides tracing functionality.
+///
+/// This trait is only available when the `trace` feature is enabled.
+/// It extends `TaskRef` with methods for accessing and modifying task identifiers
+/// and names, which are useful for debugging, logging, and performance analysis.
+pub trait TaskRefTrace {
+    /// Get the name for a task
+    fn name(&self) -> Option<&'static str>;
+
+    /// Set the name for a task
+    fn set_name(&self, name: Option<&'static str>);
+
+    /// Get the ID for a task
+    fn id(&self) -> u32;
+
+    /// Set the ID for a task
+    fn set_id(&self, id: u32);
+}
+
+impl TaskRefTrace for TaskRef {
+    fn name(&self) -> Option<&'static str> {
+        self.header().name
+    }
+
+    fn set_name(&self, name: Option<&'static str>) {
+        unsafe {
+            let header_ptr = self.ptr.as_ptr() as *mut TaskHeader;
+            (*header_ptr).name = name;
+        }
+    }
+
+    fn id(&self) -> u32 {
+        self.header().id
+    }
+
+    fn set_id(&self, id: u32) {
+        unsafe {
+            let header_ptr = self.ptr.as_ptr() as *mut TaskHeader;
+            (*header_ptr).id = id;
+        }
+    }
+}
+
+#[cfg(not(feature = "rtos-trace"))]
+extern "Rust" {
+    /// This callback is called when the executor begins polling. This will always
+    /// be paired with a later call to `_embassy_trace_executor_idle`.
+    ///
+    /// This marks the EXECUTOR state transition from IDLE -> SCHEDULING.
+    fn _embassy_trace_poll_start(executor_id: u32);
+
+    /// This callback is called AFTER a task is initialized/allocated, and BEFORE
+    /// it is enqueued to run for the first time. If the task ends (and does not
+    /// loop "forever"), there will be a matching call to `_embassy_trace_task_end`.
+    ///
+    /// Tasks start life in the SPAWNED state.
+    fn _embassy_trace_task_new(executor_id: u32, task_id: u32);
+
+    /// This callback is called AFTER a task is destructed/freed. This will always
+    /// have a prior matching call to `_embassy_trace_task_new`.
+    fn _embassy_trace_task_end(executor_id: u32, task_id: u32);
+
+    /// This callback is called AFTER a task has been dequeued from the runqueue,
+    /// and BEFORE the task is polled. There will always be a matching call to
+    /// `_embassy_trace_task_exec_end`.
+    ///
+    /// This marks the TASK state transition from WAITING -> RUNNING
+    /// This marks the EXECUTOR state transition from SCHEDULING -> POLLING
+    fn _embassy_trace_task_exec_begin(executor_id: u32, task_id: u32);
+
+    /// This callback is called AFTER a task has completed polling. There will
+    /// always be a matching call to `_embassy_trace_task_exec_begin`.
+    ///
+    /// This marks the TASK state transition from either:
+    /// * RUNNING -> IDLE - if there were no `_embassy_trace_task_ready_begin` events
+    ///     for this task since the last `_embassy_trace_task_exec_begin` for THIS task
+    /// * RUNNING -> WAITING - if there WAS a `_embassy_trace_task_ready_begin` event
+    ///     for this task since the last `_embassy_trace_task_exec_begin` for THIS task
+    ///
+    /// This marks the EXECUTOR state transition from POLLING -> SCHEDULING
+    fn _embassy_trace_task_exec_end(excutor_id: u32, task_id: u32);
+
+    /// This callback is called AFTER the waker for a task is awoken, and BEFORE it
+    /// is added to the run queue.
+    ///
+    /// If the given task is currently RUNNING, this marks no state change, BUT the
+    /// RUNNING task will then move to the WAITING stage when polling is complete.
+    ///
+    /// If the given task is currently IDLE, this marks the TASK state transition
+    /// from IDLE -> WAITING.
+    ///
+    /// NOTE: This may be called from an interrupt, outside the context of the current
+    /// task or executor.
+    fn _embassy_trace_task_ready_begin(executor_id: u32, task_id: u32);
+
+    /// This callback is called AFTER all dequeued tasks in a single call to poll
+    /// have been processed. This will always be paired with a call to
+    /// `_embassy_trace_executor_idle`.
+    ///
+    /// This marks the EXECUTOR state transition from SCHEDULING -> IDLE
+    fn _embassy_trace_executor_idle(executor_id: u32);
+}
+
+#[inline]
+pub(crate) fn poll_start(executor: &SyncExecutor) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_poll_start(executor as *const _ as u32)
+    }
+}
+
+#[inline]
+pub(crate) fn task_new(executor: &SyncExecutor, task: &TaskRef) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_task_new(executor as *const _ as u32, task.as_ptr() as u32)
+    }
+
+    #[cfg(feature = "rtos-trace")]
+    rtos_trace::trace::task_new(task.as_ptr() as u32);
+
+    #[cfg(feature = "rtos-trace")]
+    TASK_TRACKER.add(*task);
+}
+
+#[inline]
+pub(crate) fn task_end(executor: *const SyncExecutor, task: &TaskRef) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_task_end(executor as u32, task.as_ptr() as u32)
+    }
+}
+
+#[inline]
+pub(crate) fn task_ready_begin(executor: &SyncExecutor, task: &TaskRef) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_task_ready_begin(executor as *const _ as u32, task.as_ptr() as u32)
+    }
+    #[cfg(feature = "rtos-trace")]
+    rtos_trace::trace::task_ready_begin(task.as_ptr() as u32);
+}
+
+#[inline]
+pub(crate) fn task_exec_begin(executor: &SyncExecutor, task: &TaskRef) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_task_exec_begin(executor as *const _ as u32, task.as_ptr() as u32)
+    }
+    #[cfg(feature = "rtos-trace")]
+    rtos_trace::trace::task_exec_begin(task.as_ptr() as u32);
+}
+
+#[inline]
+pub(crate) fn task_exec_end(executor: &SyncExecutor, task: &TaskRef) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_task_exec_end(executor as *const _ as u32, task.as_ptr() as u32)
+    }
+    #[cfg(feature = "rtos-trace")]
+    rtos_trace::trace::task_exec_end();
+}
+
+#[inline]
+pub(crate) fn executor_idle(executor: &SyncExecutor) {
+    #[cfg(not(feature = "rtos-trace"))]
+    unsafe {
+        _embassy_trace_executor_idle(executor as *const _ as u32)
+    }
+    #[cfg(feature = "rtos-trace")]
+    rtos_trace::trace::system_idle();
+}
+
+/// Returns an iterator over all active tasks in the system
+///
+/// This function provides a convenient way to iterate over all tasks
+/// that are currently tracked in the system. The returned iterator
+/// yields each task in the global task tracker.
+///
+/// # Returns
+/// An iterator that yields `TaskRef` items for each task
+fn get_all_active_tasks() -> impl Iterator<Item = TaskRef> + 'static {
+    struct TaskIterator<'a> {
+        tracker: &'a TaskTracker,
+        current: *mut TaskHeader,
+    }
+
+    impl<'a> Iterator for TaskIterator<'a> {
+        type Item = TaskRef;
+
+        fn next(&mut self) -> Option<Self::Item> {
+            if self.current.is_null() {
+                return None;
+            }
+
+            let task = unsafe { TaskRef::from_ptr(self.current) };
+            self.current = unsafe { (*self.current).all_tasks_next.load(Ordering::Acquire) };
+
+            Some(task)
+        }
+    }
+
+    TaskIterator {
+        tracker: &TASK_TRACKER,
+        current: TASK_TRACKER.head.load(Ordering::Acquire),
+    }
+}
+
+/// Perform an action on each active task
+fn with_all_active_tasks<F>(f: F)
+where
+    F: FnMut(TaskRef),
+{
+    TASK_TRACKER.for_each(f);
+}
+
+#[cfg(feature = "rtos-trace")]
+impl rtos_trace::RtosTraceOSCallbacks for crate::raw::SyncExecutor {
+    fn task_list() {
+        with_all_active_tasks(|task| {
+            let name = task.name().unwrap_or("unnamed task\0");
+            let info = rtos_trace::TaskInfo {
+                name,
+                priority: 0,
+                stack_base: 0,
+                stack_size: 0,
+            };
+            rtos_trace::trace::task_send_info(task.id(), info);
+        });
+    }
+    fn time() -> u64 {
+        const fn gcd(a: u64, b: u64) -> u64 {
+            if b == 0 {
+                a
+            } else {
+                gcd(b, a % b)
+            }
+        }
+
+        const GCD_1M: u64 = gcd(embassy_time_driver::TICK_HZ, 1_000_000);
+        embassy_time_driver::now() * (1_000_000 / GCD_1M) / (embassy_time_driver::TICK_HZ / GCD_1M)
+    }
+}
+
+#[cfg(feature = "rtos-trace")]
+rtos_trace::global_os_callbacks! {SyncExecutor}
diff --git a/embassy-executor/src/raw/waker.rs b/embassy-executor/src/raw/waker.rs
index 8d3910a25..d0d7b003d 100644
--- a/embassy-executor/src/raw/waker.rs
+++ b/embassy-executor/src/raw/waker.rs
@@ -26,38 +26,17 @@ pub(crate) unsafe fn from_task(p: TaskRef) -> Waker {
 /// (1 word) instead of full Wakers (2 words). This saves a bit of RAM and helps
 /// avoid dynamic dispatch.
 ///
-/// You can use the returned task pointer to wake the task with [`wake_task`](super::wake_task).
+/// You can use the returned task pointer to wake the task with [`wake_task`].
 ///
 /// # Panics
 ///
 /// Panics if the waker is not created by the Embassy executor.
 pub fn task_from_waker(waker: &Waker) -> TaskRef {
-    let (vtable, data) = {
-        #[cfg(not(feature = "nightly"))]
-        {
-            struct WakerHack {
-                data: *const (),
-                vtable: &'static RawWakerVTable,
-            }
-
-            // safety: OK because WakerHack has the same layout as Waker.
-            // This is not really guaranteed because the structs are `repr(Rust)`, it is
-            // indeed the case in the current implementation.
-            // TODO use waker_getters when stable. https://github.com/rust-lang/rust/issues/96992
-            let hack: &WakerHack = unsafe { core::mem::transmute(waker) };
-            (hack.vtable, hack.data)
-        }
-
-        #[cfg(feature = "nightly")]
-        {
-            let raw_waker = waker.as_raw();
-            (raw_waker.vtable(), raw_waker.data())
-        }
-    };
-
-    if vtable != &VTABLE {
+    // make sure to compare vtable addresses. Doing `==` on the references
+    // will compare the contents, which is slower.
+    if waker.vtable() as *const _ != &VTABLE as *const _ {
         panic!("Found waker not created by the Embassy executor. `embassy_time::Timer` only works with the Embassy executor.")
     }
     // safety: our wakers are always created with `TaskRef::as_ptr`
-    unsafe { TaskRef::from_ptr(data as *const TaskHeader) }
+    unsafe { TaskRef::from_ptr(waker.data() as *const TaskHeader) }
 }