1 // Copyright 2013-2014 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
11 //! Tasks implemented on top of OS threads
13 //! This module contains the implementation of the 1:1 threading module required
14 //! by rust tasks. This implements the necessary API traits laid out by std::rt
15 //! in order to spawn new tasks and deschedule the current task.
19 use std::rt::bookkeeping;
20 use std::rt::local::Local;
21 use std::rt::mutex::NativeMutex;
24 use std::rt::task::{Task, BlockedTask, TaskOpts};
25 use std::rt::thread::Thread;
31 /// Creates a new Task which is ready to execute as a 1:1 task.
32 pub fn new(stack_bounds: (uint, uint)) -> Box<Task> {
33 let mut task = box Task::new();
35 ops.stack_bounds = stack_bounds;
36 task.put_runtime(ops);
40 fn ops() -> Box<Ops> {
42 lock: unsafe { NativeMutex::new() },
44 io: io::IoFactory::new(),
45 // these *should* get overwritten
50 /// Spawns a function with the default configuration
51 pub fn spawn(f: proc():Send) {
52 spawn_opts(TaskOpts { name: None, stack_size: None, on_exit: None }, f)
55 /// Spawns a new task given the configuration options and a procedure to run
57 pub fn spawn_opts(opts: TaskOpts, f: proc():Send) {
58 let TaskOpts { name, stack_size, on_exit } = opts;
60 let mut task = box Task::new();
62 task.death.on_exit = on_exit;
64 let stack = stack_size.unwrap_or(rt::min_stack());
68 // Note that this increment must happen *before* the spawn in order to
69 // guarantee that if this task exits it will always end up waiting for the
70 // spawned task to exit.
71 bookkeeping::increment();
73 // Spawning a new OS thread guarantees that __morestack will never get
74 // triggered, but we must manually set up the actual stack bounds once this
75 // function starts executing. This raises the lower limit by a bit because
76 // by the time that this function is executing we've already consumed at
77 // least a little bit of stack (we don't know the exact byte address at
78 // which our stack started).
79 Thread::spawn_stack(stack, proc() {
80 let something_around_the_top_of_the_stack = 1;
81 let addr = &something_around_the_top_of_the_stack as *int;
82 let my_stack = addr as uint;
84 stack::record_stack_bounds(my_stack - stack + 1024, my_stack);
87 ops.stack_bounds = (my_stack - stack + 1024, my_stack);
91 task.put_runtime(ops);
92 let t = task.run(|| { f.take_unwrap()() });
94 bookkeeping::decrement();
98 // This structure is the glue between channels and the 1:1 scheduling mode. This
99 // structure is allocated once per task.
101 lock: NativeMutex, // native synchronization
102 awoken: bool, // used to prevent spurious wakeups
103 io: io::IoFactory, // local I/O factory
105 // This field holds the known bounds of the stack in (lo, hi) form. Not all
106 // native tasks necessarily know their precise bounds, hence this is
108 stack_bounds: (uint, uint),
111 impl rt::Runtime for Ops {
112 fn yield_now(~self, mut cur_task: Box<Task>) {
113 // put the task back in TLS and then invoke the OS thread yield
114 cur_task.put_runtime(self);
115 Local::put(cur_task);
119 fn maybe_yield(~self, mut cur_task: Box<Task>) {
120 // just put the task back in TLS, on OS threads we never need to
121 // opportunistically yield b/c the OS will do that for us (preemption)
122 cur_task.put_runtime(self);
123 Local::put(cur_task);
126 fn wrap(~self) -> Box<Any> {
130 fn stack_bounds(&self) -> (uint, uint) { self.stack_bounds }
132 fn can_block(&self) -> bool { true }
134 // This function gets a little interesting. There are a few safety and
135 // ownership violations going on here, but this is all done in the name of
136 // shared state. Additionally, all of the violations are protected with a
137 // mutex, so in theory there are no races.
139 // The first thing we need to do is to get a pointer to the task's internal
140 // mutex. This address will not be changing (because the task is allocated
141 // on the heap). We must have this handle separately because the task will
142 // have its ownership transferred to the given closure. We're guaranteed,
143 // however, that this memory will remain valid because *this* is the current
144 // task's execution thread.
146 // The next weird part is where ownership of the task actually goes. We
147 // relinquish it to the `f` blocking function, but upon returning this
148 // function needs to replace the task back in TLS. There is no communication
149 // from the wakeup thread back to this thread about the task pointer, and
150 // there's really no need to. In order to get around this, we cast the task
151 // to a `uint` which is then used at the end of this function to cast back
152 // to a `Box<Task>` object. Naturally, this looks like it violates
153 // ownership semantics in that there may be two `Box<Task>` objects.
155 // The fun part is that the wakeup half of this implementation knows to
156 // "forget" the task on the other end. This means that the awakening half of
157 // things silently relinquishes ownership back to this thread, but not in a
158 // way that the compiler can understand. The task's memory is always valid
159 // for both tasks because these operations are all done inside of a mutex.
161 // You'll also find that if blocking fails (the `f` function hands the
162 // BlockedTask back to us), we will `mem::forget` the handles. The
163 // reasoning for this is the same logic as above in that the task silently
164 // transfers ownership via the `uint`, not through normal compiler
167 // On a mildly unrelated note, it should also be pointed out that OS
168 // condition variables are susceptible to spurious wakeups, which we need to
169 // be ready for. In order to accomodate for this fact, we have an extra
170 // `awoken` field which indicates whether we were actually woken up via some
171 // invocation of `reawaken`. This flag is only ever accessed inside the
172 // lock, so there's no need to make it atomic.
173 fn deschedule(mut ~self, times: uint, mut cur_task: Box<Task>,
174 f: |BlockedTask| -> Result<(), BlockedTask>) {
175 let me = &mut *self as *mut Ops;
176 cur_task.put_runtime(self);
179 let cur_task_dupe = &*cur_task as *Task;
180 let task = BlockedTask::block(cur_task);
183 let guard = (*me).lock.lock();
184 (*me).awoken = false;
187 while !(*me).awoken {
191 Err(task) => { mem::forget(task.wake()); }
194 let iter = task.make_selectable(times);
195 let guard = (*me).lock.lock();
196 (*me).awoken = false;
198 // Apply the given closure to all of the "selectable tasks",
199 // bailing on the first one that produces an error. Note that
200 // care must be taken such that when an error is occurred, we
201 // may not own the task, so we may still have to wait for the
202 // task to become available. In other words, if task.wake()
203 // returns `None`, then someone else has ownership and we must
204 // wait for their signal.
205 match iter.map(f).filter_map(|a| a.err()).next() {
217 while !(*me).awoken {
221 // re-acquire ownership of the task
222 cur_task = mem::transmute(cur_task_dupe);
225 // put the task back in TLS, and everything is as it once was.
226 Local::put(cur_task);
229 // See the comments on `deschedule` for why the task is forgotten here, and
230 // why it's valid to do so.
231 fn reawaken(mut ~self, mut to_wake: Box<Task>) {
233 let me = &mut *self as *mut Ops;
234 to_wake.put_runtime(self);
235 mem::forget(to_wake);
236 let guard = (*me).lock.lock();
242 fn spawn_sibling(~self,
243 mut cur_task: Box<Task>,
246 cur_task.put_runtime(self);
247 Local::put(cur_task);
249 task::spawn_opts(opts, f);
252 fn local_io<'a>(&'a mut self) -> Option<rtio::LocalIo<'a>> {
253 Some(rtio::LocalIo::new(&mut self.io as &mut rtio::IoFactory))
259 use std::rt::local::Local;
260 use std::rt::task::{Task, TaskOpts};
262 use super::{spawn, spawn_opts, Ops};
266 let (tx, rx) = channel();
275 let (tx, rx) = channel::<()>();
280 assert_eq!(rx.recv_opt(), Err(()));
285 let mut opts = TaskOpts::new();
286 opts.name = Some("test".into_maybe_owned());
287 opts.stack_size = Some(20 * 4096);
288 let (tx, rx) = channel();
289 opts.on_exit = Some(proc(r) tx.send(r));
290 spawn_opts(opts, proc() {});
291 assert!(rx.recv().is_ok());
295 fn smoke_opts_fail() {
296 let mut opts = TaskOpts::new();
297 let (tx, rx) = channel();
298 opts.on_exit = Some(proc(r) tx.send(r));
299 spawn_opts(opts, proc() { fail!() });
300 assert!(rx.recv().is_err());
305 let (tx, rx) = channel();
307 for _ in range(0, 10) { task::deschedule(); }
314 fn spawn_children() {
315 let (tx1, rx) = channel();
317 let (tx2, rx) = channel();
319 let (tx3, rx) = channel();
333 fn spawn_inherits() {
334 let (tx, rx) = channel();
337 let mut task: Box<Task> = Local::take();
338 match task.maybe_take_runtime::<Ops>() {
340 task.put_runtime(ops);