tokio/task/coop/mod.rs
1#![cfg_attr(not(feature = "full"), allow(dead_code))]
2#![cfg_attr(not(feature = "rt"), allow(unreachable_pub))]
3
4//! Utilities for improved cooperative scheduling.
5//!
6//! ### Cooperative scheduling
7//!
8//! A single call to [`poll`] on a top-level task may potentially do a lot of
9//! work before it returns `Poll::Pending`. If a task runs for a long period of
10//! time without yielding back to the executor, it can starve other tasks
11//! waiting on that executor to execute them, or drive underlying resources.
12//! Since Rust does not have a runtime, it is difficult to forcibly preempt a
13//! long-running task. Instead, this module provides an opt-in mechanism for
14//! futures to collaborate with the executor to avoid starvation.
15//!
16//! Consider a future like this one:
17//!
18//! ```
19//! # use tokio_stream::{Stream, StreamExt};
20//! async fn drop_all<I: Stream + Unpin>(mut input: I) {
21//! while let Some(_) = input.next().await {}
22//! }
23//! ```
24//!
25//! It may look harmless, but consider what happens under heavy load if the
26//! input stream is _always_ ready. If we spawn `drop_all`, the task will never
27//! yield, and will starve other tasks and resources on the same executor.
28//!
29//! To account for this, Tokio has explicit yield points in a number of library
30//! functions, which force tasks to return to the executor periodically.
31//!
32//!
33//! #### unconstrained
34//!
35//! If necessary, [`task::unconstrained`] lets you opt a future out of Tokio's cooperative
36//! scheduling. When a future is wrapped with `unconstrained`, it will never be forced to yield to
37//! Tokio. For example:
38//!
39//! ```
40//! # #[tokio::main]
41//! # async fn main() {
42//! use tokio::{task, sync::mpsc};
43//!
44//! let fut = async {
45//! let (tx, mut rx) = mpsc::unbounded_channel();
46//!
47//! for i in 0..1000 {
48//! let _ = tx.send(());
49//! // This will always be ready. If coop was in effect, this code would be forced to yield
50//! // periodically. However, if left unconstrained, then this code will never yield.
51//! rx.recv().await;
52//! }
53//! };
54//!
55//! task::coop::unconstrained(fut).await;
56//! # }
57//! ```
58//! [`poll`]: method@std::future::Future::poll
59//! [`task::unconstrained`]: crate::task::unconstrained()
60
61cfg_rt! {
62 mod consume_budget;
63 pub use consume_budget::consume_budget;
64
65 mod unconstrained;
66 pub use unconstrained::{unconstrained, Unconstrained};
67}
68
69// ```ignore
70// # use tokio_stream::{Stream, StreamExt};
71// async fn drop_all<I: Stream + Unpin>(mut input: I) {
72// while let Some(_) = input.next().await {
73// tokio::coop::proceed().await;
74// }
75// }
76// ```
77//
78// The `proceed` future will coordinate with the executor to make sure that
79// every so often control is yielded back to the executor so it can run other
80// tasks.
81//
82// # Placing yield points
83//
84// Voluntary yield points should be placed _after_ at least some work has been
85// done. If they are not, a future sufficiently deep in the task hierarchy may
86// end up _never_ getting to run because of the number of yield points that
87// inevitably appear before it is reached. In general, you will want yield
88// points to only appear in "leaf" futures -- those that do not themselves poll
89// other futures. By doing this, you avoid double-counting each iteration of
90// the outer future against the cooperating budget.
91
92use crate::runtime::context;
93
94/// Opaque type tracking the amount of "work" a task may still do before
95/// yielding back to the scheduler.
96#[derive(Debug, Copy, Clone)]
97pub(crate) struct Budget(Option<u8>);
98
99pub(crate) struct BudgetDecrement {
100 success: bool,
101 hit_zero: bool,
102}
103
104impl Budget {
105 /// Budget assigned to a task on each poll.
106 ///
107 /// The value itself is chosen somewhat arbitrarily. It needs to be high
108 /// enough to amortize wakeup and scheduling costs, but low enough that we
109 /// do not starve other tasks for too long. The value also needs to be high
110 /// enough that particularly deep tasks are able to do at least some useful
111 /// work at all.
112 ///
113 /// Note that as more yield points are added in the ecosystem, this value
114 /// will probably also have to be raised.
115 const fn initial() -> Budget {
116 Budget(Some(128))
117 }
118
119 /// Returns an unconstrained budget. Operations will not be limited.
120 pub(crate) const fn unconstrained() -> Budget {
121 Budget(None)
122 }
123
124 fn has_remaining(self) -> bool {
125 self.0.map_or(true, |budget| budget > 0)
126 }
127}
128
129/// Runs the given closure with a cooperative task budget. When the function
130/// returns, the budget is reset to the value prior to calling the function.
131#[inline(always)]
132pub(crate) fn budget<R>(f: impl FnOnce() -> R) -> R {
133 with_budget(Budget::initial(), f)
134}
135
136/// Runs the given closure with an unconstrained task budget. When the function returns, the budget
137/// is reset to the value prior to calling the function.
138#[inline(always)]
139pub(crate) fn with_unconstrained<R>(f: impl FnOnce() -> R) -> R {
140 with_budget(Budget::unconstrained(), f)
141}
142
143#[inline(always)]
144fn with_budget<R>(budget: Budget, f: impl FnOnce() -> R) -> R {
145 struct ResetGuard {
146 prev: Budget,
147 }
148
149 impl Drop for ResetGuard {
150 fn drop(&mut self) {
151 let _ = context::budget(|cell| {
152 cell.set(self.prev);
153 });
154 }
155 }
156
157 #[allow(unused_variables)]
158 let maybe_guard = context::budget(|cell| {
159 let prev = cell.get();
160 cell.set(budget);
161
162 ResetGuard { prev }
163 });
164
165 // The function is called regardless even if the budget is not successfully
166 // set due to the thread-local being destroyed.
167 f()
168}
169
170/// Returns `true` if there is still budget left on the task.
171///
172/// # Examples
173///
174/// This example defines a `Timeout` future that requires a given `future` to complete before the
175/// specified duration elapses. If it does, its result is returned; otherwise, an error is returned
176/// and the future is canceled.
177///
178/// Note that the future could exhaust the budget before we evaluate the timeout. Using `has_budget_remaining`,
179/// we can detect this scenario and ensure the timeout is always checked.
180///
181/// ```
182/// # use std::future::Future;
183/// # use std::pin::{pin, Pin};
184/// # use std::task::{ready, Context, Poll};
185/// # use tokio::task::coop;
186/// # use tokio::time::Sleep;
187/// pub struct Timeout<T> {
188/// future: T,
189/// delay: Pin<Box<Sleep>>,
190/// }
191///
192/// impl<T> Future for Timeout<T>
193/// where
194/// T: Future + Unpin,
195/// {
196/// type Output = Result<T::Output, ()>;
197///
198/// fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
199/// let this = Pin::into_inner(self);
200/// let future = Pin::new(&mut this.future);
201/// let delay = Pin::new(&mut this.delay);
202///
203/// // check if the future is ready
204/// let had_budget_before = coop::has_budget_remaining();
205/// if let Poll::Ready(v) = future.poll(cx) {
206/// return Poll::Ready(Ok(v));
207/// }
208/// let has_budget_now = coop::has_budget_remaining();
209///
210/// // evaluate the timeout
211/// if let (true, false) = (had_budget_before, has_budget_now) {
212/// // it is the underlying future that exhausted the budget
213/// ready!(pin!(coop::unconstrained(delay)).poll(cx));
214/// } else {
215/// ready!(delay.poll(cx));
216/// }
217/// return Poll::Ready(Err(()));
218/// }
219/// }
220///```
221#[inline(always)]
222#[cfg_attr(docsrs, doc(cfg(feature = "rt")))]
223pub fn has_budget_remaining() -> bool {
224 // If the current budget cannot be accessed due to the thread-local being
225 // shutdown, then we assume there is budget remaining.
226 context::budget(|cell| cell.get().has_remaining()).unwrap_or(true)
227}
228
229cfg_rt_multi_thread! {
230 /// Sets the current task's budget.
231 pub(crate) fn set(budget: Budget) {
232 let _ = context::budget(|cell| cell.set(budget));
233 }
234}
235
236cfg_rt! {
237 /// Forcibly removes the budgeting constraints early.
238 ///
239 /// Returns the remaining budget
240 pub(crate) fn stop() -> Budget {
241 context::budget(|cell| {
242 let prev = cell.get();
243 cell.set(Budget::unconstrained());
244 prev
245 }).unwrap_or(Budget::unconstrained())
246 }
247}
248
249cfg_coop! {
250 use pin_project_lite::pin_project;
251 use std::cell::Cell;
252 use std::future::Future;
253 use std::pin::Pin;
254 use std::task::{ready, Context, Poll};
255
256 #[must_use]
257 pub(crate) struct RestoreOnPending(Cell<Budget>);
258
259 impl RestoreOnPending {
260 pub(crate) fn made_progress(&self) {
261 self.0.set(Budget::unconstrained());
262 }
263 }
264
265 impl Drop for RestoreOnPending {
266 fn drop(&mut self) {
267 // Don't reset if budget was unconstrained or if we made progress.
268 // They are both represented as the remembered budget being unconstrained.
269 let budget = self.0.get();
270 if !budget.is_unconstrained() {
271 let _ = context::budget(|cell| {
272 cell.set(budget);
273 });
274 }
275 }
276 }
277
278 /// Returns `Poll::Pending` if the current task has exceeded its budget and should yield.
279 ///
280 /// When you call this method, the current budget is decremented. However, to ensure that
281 /// progress is made every time a task is polled, the budget is automatically restored to its
282 /// former value if the returned `RestoreOnPending` is dropped. It is the caller's
283 /// responsibility to call `RestoreOnPending::made_progress` if it made progress, to ensure
284 /// that the budget empties appropriately.
285 ///
286 /// Note that `RestoreOnPending` restores the budget **as it was before `poll_proceed`**.
287 /// Therefore, if the budget is _further_ adjusted between when `poll_proceed` returns and
288 /// `RestRestoreOnPending` is dropped, those adjustments are erased unless the caller indicates
289 /// that progress was made.
290 #[inline]
291 pub(crate) fn poll_proceed(cx: &mut Context<'_>) -> Poll<RestoreOnPending> {
292 context::budget(|cell| {
293 let mut budget = cell.get();
294
295 let decrement = budget.decrement();
296
297 if decrement.success {
298 let restore = RestoreOnPending(Cell::new(cell.get()));
299 cell.set(budget);
300
301 // avoid double counting
302 if decrement.hit_zero {
303 inc_budget_forced_yield_count();
304 }
305
306 Poll::Ready(restore)
307 } else {
308 register_waker(cx);
309 Poll::Pending
310 }
311 }).unwrap_or(Poll::Ready(RestoreOnPending(Cell::new(Budget::unconstrained()))))
312 }
313
314 /// Returns `Poll::Ready` if the current task has budget to consume, and `Poll::Pending` otherwise.
315 ///
316 /// Note that in contrast to `poll_proceed`, this method does not consume any budget and is used when
317 /// polling for budget availability.
318 #[inline]
319 pub(crate) fn poll_budget_available(cx: &mut Context<'_>) -> Poll<()> {
320 if has_budget_remaining() {
321 Poll::Ready(())
322 } else {
323 register_waker(cx);
324
325 Poll::Pending
326 }
327 }
328
329 cfg_rt! {
330 cfg_unstable_metrics! {
331 #[inline(always)]
332 fn inc_budget_forced_yield_count() {
333 let _ = context::with_current(|handle| {
334 handle.scheduler_metrics().inc_budget_forced_yield_count();
335 });
336 }
337 }
338
339 cfg_not_unstable_metrics! {
340 #[inline(always)]
341 fn inc_budget_forced_yield_count() {}
342 }
343
344 fn register_waker(cx: &mut Context<'_>) {
345 context::defer(cx.waker());
346 }
347 }
348
349 cfg_not_rt! {
350 #[inline(always)]
351 fn inc_budget_forced_yield_count() {}
352
353 fn register_waker(cx: &mut Context<'_>) {
354 cx.waker().wake_by_ref()
355 }
356 }
357
358 impl Budget {
359 /// Decrements the budget. Returns `true` if successful. Decrementing fails
360 /// when there is not enough remaining budget.
361 fn decrement(&mut self) -> BudgetDecrement {
362 if let Some(num) = &mut self.0 {
363 if *num > 0 {
364 *num -= 1;
365
366 let hit_zero = *num == 0;
367
368 BudgetDecrement { success: true, hit_zero }
369 } else {
370 BudgetDecrement { success: false, hit_zero: false }
371 }
372 } else {
373 BudgetDecrement { success: true, hit_zero: false }
374 }
375 }
376
377 fn is_unconstrained(self) -> bool {
378 self.0.is_none()
379 }
380 }
381
382 pin_project! {
383 /// Future wrapper to ensure cooperative scheduling.
384 ///
385 /// When being polled `poll_proceed` is called before the inner future is polled to check
386 /// if the inner future has exceeded its budget. If the inner future resolves, this will
387 /// automatically call `RestoreOnPending::made_progress` before resolving this future with
388 /// the result of the inner one. If polling the inner future is pending, polling this future
389 /// type will also return a `Poll::Pending`.
390 #[must_use = "futures do nothing unless polled"]
391 pub(crate) struct Coop<F: Future> {
392 #[pin]
393 pub(crate) fut: F,
394 }
395 }
396
397 impl<F: Future> Future for Coop<F> {
398 type Output = F::Output;
399
400 fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
401 let coop = ready!(poll_proceed(cx));
402 let me = self.project();
403 if let Poll::Ready(ret) = me.fut.poll(cx) {
404 coop.made_progress();
405 Poll::Ready(ret)
406 } else {
407 Poll::Pending
408 }
409 }
410 }
411
412 /// Run a future with a budget constraint for cooperative scheduling.
413 /// If the future exceeds its budget while being polled, control is yielded back to the
414 /// runtime.
415 #[inline]
416 pub(crate) fn cooperative<F: Future>(fut: F) -> Coop<F> {
417 Coop { fut }
418 }
419}
420
421#[cfg(all(test, not(loom)))]
422mod test {
423 use super::*;
424
425 #[cfg(all(target_family = "wasm", not(target_os = "wasi")))]
426 use wasm_bindgen_test::wasm_bindgen_test as test;
427
428 fn get() -> Budget {
429 context::budget(|cell| cell.get()).unwrap_or(Budget::unconstrained())
430 }
431
432 #[test]
433 fn budgeting() {
434 use std::future::poll_fn;
435 use tokio_test::*;
436
437 assert!(get().0.is_none());
438
439 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
440
441 assert!(get().0.is_none());
442 drop(coop);
443 assert!(get().0.is_none());
444
445 budget(|| {
446 assert_eq!(get().0, Budget::initial().0);
447
448 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
449 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 1);
450 drop(coop);
451 // we didn't make progress
452 assert_eq!(get().0, Budget::initial().0);
453
454 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
455 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 1);
456 coop.made_progress();
457 drop(coop);
458 // we _did_ make progress
459 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 1);
460
461 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
462 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 2);
463 coop.made_progress();
464 drop(coop);
465 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 2);
466
467 budget(|| {
468 assert_eq!(get().0, Budget::initial().0);
469
470 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
471 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 1);
472 coop.made_progress();
473 drop(coop);
474 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 1);
475 });
476
477 assert_eq!(get().0.unwrap(), Budget::initial().0.unwrap() - 2);
478 });
479
480 assert!(get().0.is_none());
481
482 budget(|| {
483 let n = get().0.unwrap();
484
485 for _ in 0..n {
486 let coop = assert_ready!(task::spawn(()).enter(|cx, _| poll_proceed(cx)));
487 coop.made_progress();
488 }
489
490 let mut task = task::spawn(poll_fn(|cx| {
491 let coop = std::task::ready!(poll_proceed(cx));
492 coop.made_progress();
493 Poll::Ready(())
494 }));
495
496 assert_pending!(task.poll());
497 });
498 }
499}