tokio/runtime/task/
list.rs

1//! This module has containers for storing the tasks spawned on a scheduler. The
2//! `OwnedTasks` container is thread-safe but can only store tasks that
3//! implement Send. The `LocalOwnedTasks` container is not thread safe, but can
4//! store non-Send tasks.
5//!
6//! The collections can be closed to prevent adding new tasks during shutdown of
7//! the scheduler with the collection.
8
9use crate::future::Future;
10use crate::loom::cell::UnsafeCell;
11use crate::runtime::task::{JoinHandle, LocalNotified, Notified, Schedule, Task};
12use crate::util::linked_list::{Link, LinkedList};
13use crate::util::sharded_list;
14
15use crate::loom::sync::atomic::{AtomicBool, Ordering};
16use std::marker::PhantomData;
17use std::num::NonZeroU64;
18
19// The id from the module below is used to verify whether a given task is stored
20// in this OwnedTasks, or some other task. The counter starts at one so we can
21// use `None` for tasks not owned by any list.
22//
23// The safety checks in this file can technically be violated if the counter is
24// overflown, but the checks are not supposed to ever fail unless there is a
25// bug in Tokio, so we accept that certain bugs would not be caught if the two
26// mixed up runtimes happen to have the same id.
27
28cfg_has_atomic_u64! {
29    use std::sync::atomic::AtomicU64;
30
31    static NEXT_OWNED_TASKS_ID: AtomicU64 = AtomicU64::new(1);
32
33    fn get_next_id() -> NonZeroU64 {
34        loop {
35            let id = NEXT_OWNED_TASKS_ID.fetch_add(1, Ordering::Relaxed);
36            if let Some(id) = NonZeroU64::new(id) {
37                return id;
38            }
39        }
40    }
41}
42
43cfg_not_has_atomic_u64! {
44    use std::sync::atomic::AtomicU32;
45
46    static NEXT_OWNED_TASKS_ID: AtomicU32 = AtomicU32::new(1);
47
48    fn get_next_id() -> NonZeroU64 {
49        loop {
50            let id = NEXT_OWNED_TASKS_ID.fetch_add(1, Ordering::Relaxed);
51            if let Some(id) = NonZeroU64::new(u64::from(id)) {
52                return id;
53            }
54        }
55    }
56}
57
58pub(crate) struct OwnedTasks<S: 'static> {
59    list: List<S>,
60    pub(crate) id: NonZeroU64,
61    closed: AtomicBool,
62}
63
64type List<S> = sharded_list::ShardedList<Task<S>, <Task<S> as Link>::Target>;
65
66pub(crate) struct LocalOwnedTasks<S: 'static> {
67    inner: UnsafeCell<OwnedTasksInner<S>>,
68    pub(crate) id: NonZeroU64,
69    _not_send_or_sync: PhantomData<*const ()>,
70}
71
72struct OwnedTasksInner<S: 'static> {
73    list: LinkedList<Task<S>, <Task<S> as Link>::Target>,
74    closed: bool,
75}
76
77impl<S: 'static> OwnedTasks<S> {
78    pub(crate) fn new(num_cores: usize) -> Self {
79        let shard_size = Self::gen_shared_list_size(num_cores);
80        Self {
81            list: List::new(shard_size),
82            closed: AtomicBool::new(false),
83            id: get_next_id(),
84        }
85    }
86
87    /// Binds the provided task to this `OwnedTasks` instance. This fails if the
88    /// `OwnedTasks` has been closed.
89    pub(crate) fn bind<T>(
90        &self,
91        task: T,
92        scheduler: S,
93        id: super::Id,
94    ) -> (JoinHandle<T::Output>, Option<Notified<S>>)
95    where
96        S: Schedule,
97        T: Future + Send + 'static,
98        T::Output: Send + 'static,
99    {
100        let (task, notified, join) = super::new_task(task, scheduler, id);
101        let notified = unsafe { self.bind_inner(task, notified) };
102        (join, notified)
103    }
104
105    /// Bind a task that isn't safe to transfer across thread boundaries.
106    ///
107    /// # Safety
108    /// Only use this in `LocalRuntime` where the task cannot move
109    pub(crate) unsafe fn bind_local<T>(
110        &self,
111        task: T,
112        scheduler: S,
113        id: super::Id,
114    ) -> (JoinHandle<T::Output>, Option<Notified<S>>)
115    where
116        S: Schedule,
117        T: Future + 'static,
118        T::Output: 'static,
119    {
120        let (task, notified, join) = super::new_task(task, scheduler, id);
121        let notified = unsafe { self.bind_inner(task, notified) };
122        (join, notified)
123    }
124
125    /// The part of `bind` that's the same for every type of future.
126    unsafe fn bind_inner(&self, task: Task<S>, notified: Notified<S>) -> Option<Notified<S>>
127    where
128        S: Schedule,
129    {
130        unsafe {
131            // safety: We just created the task, so we have exclusive access
132            // to the field.
133            task.header().set_owner_id(self.id);
134        }
135
136        let shard = self.list.lock_shard(&task);
137        // Check the closed flag in the lock for ensuring all that tasks
138        // will shut down after the OwnedTasks has been closed.
139        if self.closed.load(Ordering::Acquire) {
140            drop(shard);
141            task.shutdown();
142            return None;
143        }
144        shard.push(task);
145        Some(notified)
146    }
147
148    /// Asserts that the given task is owned by this `OwnedTasks` and convert it to
149    /// a `LocalNotified`, giving the thread permission to poll this task.
150    #[inline]
151    pub(crate) fn assert_owner(&self, task: Notified<S>) -> LocalNotified<S> {
152        debug_assert_eq!(task.header().get_owner_id(), Some(self.id));
153        // safety: All tasks bound to this OwnedTasks are Send, so it is safe
154        // to poll it on this thread no matter what thread we are on.
155        LocalNotified {
156            task: task.0,
157            _not_send: PhantomData,
158        }
159    }
160
161    /// Shuts down all tasks in the collection. This call also closes the
162    /// collection, preventing new items from being added.
163    ///
164    /// The parameter start determines which shard this method will start at.
165    /// Using different values for each worker thread reduces contention.
166    pub(crate) fn close_and_shutdown_all(&self, start: usize)
167    where
168        S: Schedule,
169    {
170        self.closed.store(true, Ordering::Release);
171        for i in start..self.get_shard_size() + start {
172            loop {
173                let task = self.list.pop_back(i);
174                match task {
175                    Some(task) => {
176                        task.shutdown();
177                    }
178                    None => break,
179                }
180            }
181        }
182    }
183
184    #[inline]
185    pub(crate) fn get_shard_size(&self) -> usize {
186        self.list.shard_size()
187    }
188
189    pub(crate) fn num_alive_tasks(&self) -> usize {
190        self.list.len()
191    }
192
193    cfg_64bit_metrics! {
194        pub(crate) fn spawned_tasks_count(&self) -> u64 {
195            self.list.added()
196        }
197    }
198
199    pub(crate) fn remove(&self, task: &Task<S>) -> Option<Task<S>> {
200        // If the task's owner ID is `None` then it is not part of any list and
201        // doesn't need removing.
202        let task_id = task.header().get_owner_id()?;
203
204        assert_eq!(task_id, self.id);
205
206        // safety: We just checked that the provided task is not in some other
207        // linked list.
208        unsafe { self.list.remove(task.header_ptr()) }
209    }
210
211    pub(crate) fn is_empty(&self) -> bool {
212        self.list.is_empty()
213    }
214
215    /// Generates the size of the sharded list based on the number of worker threads.
216    ///
217    /// The sharded lock design can effectively alleviate
218    /// lock contention performance problems caused by high concurrency.
219    ///
220    /// However, as the number of shards increases, the memory continuity between
221    /// nodes in the intrusive linked list will diminish. Furthermore,
222    /// the construction time of the sharded list will also increase with a higher number of shards.
223    ///
224    /// Due to the above reasons, we set a maximum value for the shared list size,
225    /// denoted as `MAX_SHARED_LIST_SIZE`.
226    fn gen_shared_list_size(num_cores: usize) -> usize {
227        const MAX_SHARED_LIST_SIZE: usize = 1 << 16;
228        usize::min(MAX_SHARED_LIST_SIZE, num_cores.next_power_of_two() * 4)
229    }
230}
231
232cfg_taskdump! {
233    impl<S: 'static> OwnedTasks<S> {
234        /// Locks the tasks, and calls `f` on an iterator over them.
235        pub(crate) fn for_each<F>(&self, f: F)
236        where
237            F: FnMut(&Task<S>),
238        {
239            self.list.for_each(f);
240        }
241    }
242}
243
244impl<S: 'static> LocalOwnedTasks<S> {
245    pub(crate) fn new() -> Self {
246        Self {
247            inner: UnsafeCell::new(OwnedTasksInner {
248                list: LinkedList::new(),
249                closed: false,
250            }),
251            id: get_next_id(),
252            _not_send_or_sync: PhantomData,
253        }
254    }
255
256    pub(crate) fn bind<T>(
257        &self,
258        task: T,
259        scheduler: S,
260        id: super::Id,
261    ) -> (JoinHandle<T::Output>, Option<Notified<S>>)
262    where
263        S: Schedule,
264        T: Future + 'static,
265        T::Output: 'static,
266    {
267        let (task, notified, join) = super::new_task(task, scheduler, id);
268
269        unsafe {
270            // safety: We just created the task, so we have exclusive access
271            // to the field.
272            task.header().set_owner_id(self.id);
273        }
274
275        if self.is_closed() {
276            drop(notified);
277            task.shutdown();
278            (join, None)
279        } else {
280            self.with_inner(|inner| {
281                inner.list.push_front(task);
282            });
283            (join, Some(notified))
284        }
285    }
286
287    /// Shuts down all tasks in the collection. This call also closes the
288    /// collection, preventing new items from being added.
289    pub(crate) fn close_and_shutdown_all(&self)
290    where
291        S: Schedule,
292    {
293        self.with_inner(|inner| inner.closed = true);
294
295        while let Some(task) = self.with_inner(|inner| inner.list.pop_back()) {
296            task.shutdown();
297        }
298    }
299
300    pub(crate) fn remove(&self, task: &Task<S>) -> Option<Task<S>> {
301        // If the task's owner ID is `None` then it is not part of any list and
302        // doesn't need removing.
303        let task_id = task.header().get_owner_id()?;
304
305        assert_eq!(task_id, self.id);
306
307        self.with_inner(|inner|
308            // safety: We just checked that the provided task is not in some
309            // other linked list.
310            unsafe { inner.list.remove(task.header_ptr()) })
311    }
312
313    /// Asserts that the given task is owned by this `LocalOwnedTasks` and convert
314    /// it to a `LocalNotified`, giving the thread permission to poll this task.
315    #[inline]
316    pub(crate) fn assert_owner(&self, task: Notified<S>) -> LocalNotified<S> {
317        assert_eq!(task.header().get_owner_id(), Some(self.id));
318
319        // safety: The task was bound to this LocalOwnedTasks, and the
320        // LocalOwnedTasks is not Send or Sync, so we are on the right thread
321        // for polling this task.
322        LocalNotified {
323            task: task.0,
324            _not_send: PhantomData,
325        }
326    }
327
328    #[inline]
329    fn with_inner<F, T>(&self, f: F) -> T
330    where
331        F: FnOnce(&mut OwnedTasksInner<S>) -> T,
332    {
333        // safety: This type is not Sync, so concurrent calls of this method
334        // can't happen.  Furthermore, all uses of this method in this file make
335        // sure that they don't call `with_inner` recursively.
336        self.inner.with_mut(|ptr| unsafe { f(&mut *ptr) })
337    }
338
339    pub(crate) fn is_closed(&self) -> bool {
340        self.with_inner(|inner| inner.closed)
341    }
342
343    pub(crate) fn is_empty(&self) -> bool {
344        self.with_inner(|inner| inner.list.is_empty())
345    }
346}
347
348#[cfg(test)]
349mod tests {
350    use super::*;
351
352    // This test may run in parallel with other tests, so we only test that ids
353    // come in increasing order.
354    #[test]
355    fn test_id_not_broken() {
356        let mut last_id = get_next_id();
357
358        for _ in 0..1000 {
359            let next_id = get_next_id();
360            assert!(last_id < next_id);
361            last_id = next_id;
362        }
363    }
364}