rust/src/libstd/thread.rs

656 lines
20 KiB
Rust
Raw Normal View History

// Copyright 2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! Native threads
//!
//! ## The threading model
//!
//! An executing Rust program consists of a collection of native OS threads,
//! each with their own stack and local state.
//!
//! Threads generally have their memory *isolated* from each other by virtue of
//! Rust's owned types (which of course may only be owned by a single thread at
//! a time). Communication between threads can be done through
//! [channels](../../std/comm/index.html), Rust's message-passing types, along
//! with [other forms of thread synchronization](../../std/sync/index.html) and
//! shared-memory data structures. In particular, types that are guaranteed to
//! be threadsafe are easily shared between threads using the
//! atomically-reference-counted container,
//! [`Arc`](../../std/sync/struct.Arc.html).
//!
//! Fatal logic errors in Rust cause *thread panic*, during which
//! a thread will unwind the stack, running destructors and freeing
//! owned resources. Thread panic is unrecoverable from within
//! the panicking thread (i.e. there is no 'try/catch' in Rust), but
//! panic may optionally be detected from a different thread. If
//! the main thread panics the application will exit with a non-zero
//! exit code.
//!
//! When the main thread of a Rust program terminates, the entire program shuts
//! down, even if other threads are still running. However, this module provides
//! convenient facilities for automatically waiting for the termination of a
//! child thread (i.e., join), described below.
//!
//! ## The `Thread` type
//!
//! Already-running threads are represented via the `Thread` type, which you can
//! get in one of two ways:
//!
//! * By spawning a new thread, e.g. using the `Thread::spawn` constructor;
//! * By requesting the current thread, using the `Thread::current` function.
//!
//! Threads can be named, and provide some built-in support for low-level
//! synchronization described below.
//!
//! The `Thread::current()` function is available even for threads not spawned
//! by the APIs of this module.
//!
//! ## Spawning a thread
//!
//! There are a few different ways to spawn a new thread, depending on how it
//! should relate to the parent thread.
//!
//! ### Simple detached threads
//!
//! The simplest case just spawns a completely independent (detached) thread,
//! returning a new `Thread` handle to it:
//!
//! ```rust
//! use std::thread::Thread;
//!
//! Thread::spawn(proc() {
//! println!("Hello, World!");
//! })
//! ```
//!
//! The spawned thread may outlive its parent.
//!
//! ### Joining
//!
//! Alternatively, the `with_join` constructor spawns a new thread and returns a
//! `JoinGuard` which can be used to wait until the child thread completes,
//! returning its result (or `Err` if the child thread panicked):
//!
//! ```rust
//! use std::thread::Thread;
//!
//! let guard = Thread::with_join(proc() { panic!() };
//! assert!(guard.join().is_err());
//! ```
//!
//! The guard works in RAII style, meaning that the child thread is
//! automatically joined when the guard is dropped. A handle to the thread
//! itself is available via the `thread` method on the guard.
//!
//! ### Configured threads
//!
//! Finally, a new thread can be configured independently of how it is
//! spawned. Configuration is available via the `Cfg` builder, which currently
//! allows you to set the name, stack size, and writers for `println!` and
//! `panic!` for the child thread:
//!
//! ```rust
//! use std::thread;
//!
//! thread::cfg().name("child1").spawn(proc() { println!("Hello, world!") });
//! ```
//!
//! ## Blocking support: park and unpark
//!
//! Every thread is equipped with some basic low-level blocking support, via the
//! `park` and `unpark` functions.
//!
//! Conceptually, each `Thread` handle has an associated token, which is
//! initially not present:
//!
//! * The `Thread::park()` function blocks the current thread unless or until
//! the token is available for its thread handle, at which point It atomically
//! consumes the token. It may also return *spuriously*, without consuming the
//! token.
//!
//! * The `unpark()` method on a `Thread` atomically makes the token available
//! if it wasn't already.
//!
//! In other words, each `Thread` acts a bit like a semaphore with initial count
//! 0, except that the semaphore is *saturating* (the count cannot go above 1),
//! and can return spuriously.
//!
//! The API is typically used by acquiring a handle to the current thread,
//! placing that handle in a shared data structure so that other threads can
//! find it, and then `park`ing. When some desired condition is met, another
//! thread calls `unpark` on the handle.
//!
//! The motivation for this design is twofold:
//!
//! * It avoids the need to allocate mutexes and condvars when building new
//! synchronization primitives; the threads already provide basic blocking/signaling.
//!
//! * It can be implemented highly efficiently on many platforms.
use core::prelude::*;
use any::Any;
use borrow::IntoCow;
use boxed::Box;
use mem;
use sync::{Mutex, Condvar, Arc};
use string::String;
use rt::{mod, unwind};
use io::{Writer, stdio};
use sys::thread as imp;
use sys_common::{stack, thread_info};
/// Thread configuation. Provides detailed control over the properties
/// and behavior of new threads.
pub struct Cfg {
// A name for the thread-to-be, for identification in panic messages
name: Option<String>,
// The size of the stack for the spawned thread
stack_size: Option<uint>,
// Thread-local stdout
stdout: Option<Box<Writer + Send>>,
// Thread-local stderr
stderr: Option<Box<Writer + Send>>,
}
impl Cfg {
/// Generate the base configuration for spawning a thread, from which
/// configuration methods can be chained.
pub fn new() -> Cfg {
Cfg {
name: None,
stack_size: None,
stdout: None,
stderr: None,
}
}
/// Name the thread-to-be. Currently the name is used for identification
/// only in panic messages.
pub fn name(mut self, name: String) -> Cfg {
self.name = Some(name);
self
}
/// Deprecated: use `name` instead
#[deprecated = "use name instead"]
pub fn named<T: IntoCow<'static, String, str>>(self, name: T) -> Cfg {
self.name(name.into_cow().into_owned())
}
/// Set the size of the stack for the new thread.
pub fn stack_size(mut self, size: uint) -> Cfg {
self.stack_size = Some(size);
self
}
/// Redirect thread-local stdout.
#[experimental = "Will likely go away after proc removal"]
pub fn stdout(mut self, stdout: Box<Writer + Send>) -> Cfg {
self.stdout = Some(stdout);
self
}
/// Redirect thread-local stderr.
#[experimental = "Will likely go away after proc removal"]
pub fn stderr(mut self, stderr: Box<Writer + Send>) -> Cfg {
self.stderr = Some(stderr);
self
}
fn core_spawn<T: Send>(self, f: proc():Send -> T, after: proc(Result<T>):Send)
-> (imp::rust_thread, Thread)
{
let Cfg { name, stack_size, stdout, stderr } = self;
let stack_size = stack_size.unwrap_or(rt::min_stack());
let my_thread = Thread::new(name);
let their_thread = my_thread.clone();
// Spawning a new OS thread guarantees that __morestack will never get
// triggered, but we must manually set up the actual stack bounds once
// this function starts executing. This raises the lower limit by a bit
// because by the time that this function is executing we've already
// consumed at least a little bit of stack (we don't know the exact byte
// address at which our stack started).
let main = proc() {
let something_around_the_top_of_the_stack = 1;
let addr = &something_around_the_top_of_the_stack as *const int;
let my_stack_top = addr as uint;
let my_stack_bottom = my_stack_top - stack_size + 1024;
unsafe {
stack::record_os_managed_stack_bounds(my_stack_bottom, my_stack_top);
}
thread_info::set(
(my_stack_bottom, my_stack_top),
thread::current_guard_page(),
their_thread
);
// There are two primary reasons that general try/catch is
// unsafe. The first is that we do not support nested try/catch. The
// fact that this is happening in a newly-spawned thread
// suffices. The second is that unwinding while unwinding is not
// defined. We take care of that by having an 'unwinding' flag in
// the thread itself. For these reasons, this unsafety should be ok.
unsafe {
let mut output = None;
let mut f_opt = Some( // option dance
if stdout.is_some() || stderr.is_some() {
proc() {
let _ = stdout.map(stdio::set_stdout);
let _ = stderr.map(stdio::set_stderr);
f()
}
} else {
f
});
let try_result = unwind::try(|| output = Some((f_opt.take().unwrap())()));
match (output, try_result) {
(Some(data), Ok(_)) => after(Ok(data)),
(None, Err(cause)) => after(Err(cause)),
_ => unreachable!()
}
}
};
(unsafe { imp::create(stack, box main) }, my_thread)
}
/// Spawn a detached thread, and return a handle to it.
///
/// The new child thread may outlive its parent.
pub fn spawn(self, f: proc():Send) -> Thread {
let (native, thread) = self.core_spawn(f, proc(_) {});
unsafe { imp::detach(native) };
thread
}
/// Spawn a joinable thread, and return an RAII guard for it.
pub fn with_join<T: Send>(self, f: proc():Send -> T) -> JoinGuard<T> {
// We need the address of the packet to fill in to be stable so when
// `main` fills it in it's still valid, so allocate an extra box to do
// so.
let my_packet = box Err(box 0); // sentinel value
let their_packet: *mut Result<T> = unsafe {
*mem::transmute::<&Box<Result<T>>, *const *mut Result<T>>(&my_packet)
};
let (native, thread) = self.core_spawn(f, proc(result) {
*their_packet = result;
});
JoinGuard {
native: native,
joined: false,
packet: my_packet,
thread: thread,
}
}
}
/// A convenience function for creating configurations.
pub fn cfg() -> Cfg { Cfg::new() }
struct Inner {
name: Option<String>,
lock: Mutex<bool>, // true when there is a buffered unpark
cvar: Condvar,
}
#[deriving(Clone)]
/// A handle to a thread.
pub struct Thread {
inner: Arc<Inner>,
}
impl Thread {
fn new(name: Option<String>) -> Thread {
Thread {
inner: Arc::new(Inner {
name: name,
lock: Mutex::new(false),
cvar: Condvar::new(),
})
}
}
/// Spawn a detached thread, and return a handle to it.
///
/// The new child thread may outlive its parent.
pub fn spawn(f: proc():Send) -> Thread {
Cfg::new().spawn(f)
}
/// Spawn a joinable thread, and return an RAII guard for it.
pub fn with_join<T: Send>(f: proc():Send -> T) -> JoinGuard<T> {
Cfg::new().with_join(f)
}
/// Gets a handle to the thread that invokes it.
pub fn current() -> Thread {
ThreadInfo::current_thread()
}
/// Cooperatively give up a timeslice to the OS scheduler.
pub fn yield_now() {
unsafe { imp::yield_now() }
}
/// Determines whether the current thread is panicking.
pub fn panicking() -> bool {
ThreadInfo::panicking()
}
// http://cr.openjdk.java.net/~stefank/6989984.1/raw_files/new/src/os/linux/vm/os_linux.cpp
/// Block unless or until the current thread's token is made available (may wake spuriously).
///
/// See the module doc for more detail.
pub fn park() {
let thread = Thread::current();
let guard = thread.inner.lock.lock();
while !*guard {
thread.inner.cvar.wait(guard);
}
*guard = false;
}
/// Atomically makes the handle's token available if it is not already.
///
/// See the module doc for more detail.
pub fn unpark(&self) {
let guard = self.inner.lock();
if !*guard {
*guard = true;
self.inner.cvar.notify_one();
}
}
/// Get the thread's name.
pub fn name(&self) -> Option<&str> {
self.inner.name.as_ref()
}
}
// a hack to get around privacy restrictions
impl thread_info::NewThread for Thread {
fn new(name: Option<String>) -> Thread { Thread::new(name) }
}
/// Indicates the manner in which a thread exited.
///
/// A thread that completes without panicking is considered to exit successfully.
pub type Result<T> = result::Result<T, Box<Any + Send>>;
#[must_use]
/// An RAII guard that will block until thread termination when dropped.
pub struct JoinGuard<T> {
native: imp::rust_thread,
thread: Thread,
joined: bool,
packet: Box<Result<T>>,
}
impl<T: Send> JoinGuard<T> {
/// Extract a handle to the thread this guard will join on.
pub fn thread(&self) -> Thread {
self.thread.clone()
}
/// Wait for the associated thread to finish, returning the result of the thread's
/// calculation.
pub fn join(mut self) -> Result<T> {
assert!(!self.joined);
unsafe { imp::join(self.native) };
self.joined = true;
let box res = self.packet.take().unwrap();
res
}
}
#[unsafe_destructor]
impl<T: Send> Drop for JoinGuard<T> {
fn drop(&mut self) {
// This is required for correctness. If this is not done then the thread
// would fill in a return box which no longer exists.
if !self.joined {
unsafe { imp::join(self.native) };
}
}
}
// TODO: fix tests
#[cfg(test)]
mod test {
use any::{Any, AnyRefExt};
use boxed::BoxAny;
use prelude::*;
use result::Result::{Ok, Err};
use result;
use std::io::{ChanReader, ChanWriter};
use string::String;
use super::{Thread, cfg};
// !!! These tests are dangerous. If something is buggy, they will hang, !!!
// !!! instead of exiting cleanly. This might wedge the buildbots. !!!
#[test]
fn test_unnamed_thread() {
Thread::with_join(proc() {
assert!(Thread::current().name().is_none());
}).join().map_err(|_| ()).unwrap();
}
#[test]
fn test_named_thread() {
cfg().name("ada lovelace".to_string()).with_join(proc() {
assert!(Thread::current().name().unwrap() == "ada lovelace".to_string());
}).join().map_err(|_| ()).unwrap();
}
#[test]
fn test_run_basic() {
let (tx, rx) = channel();
Thread::spawn(proc() {
tx.send(());
});
rx.recv();
}
#[test]
fn test_join_success() {
match Thread::with_join::<String>(proc() {
"Success!".to_string()
}).join().as_ref().map(|s| s.as_slice()) {
result::Result::Ok("Success!") => (),
_ => panic!()
}
}
#[test]
fn test_join_panic() {
match Thread::with_join(proc() {
panic!()
}).join() {
result::Result::Err(_) => (),
result::Result::Ok(()) => panic!()
}
}
#[test]
fn test_spawn_sched() {
use clone::Clone;
let (tx, rx) = channel();
fn f(i: int, tx: Sender<()>) {
let tx = tx.clone();
Thread::spawn(proc() {
if i == 0 {
tx.send(());
} else {
f(i - 1, tx);
}
});
}
f(10, tx);
rx.recv();
}
#[test]
fn test_spawn_sched_childs_on_default_sched() {
let (tx, rx) = channel();
Thread::spawn(proc() {
Thread::spawn(proc() {
tx.send(());
});
});
rx.recv();
}
fn avoid_copying_the_body(spawnfn: |v: proc():Send|) {
let (tx, rx) = channel::<uint>();
let x = box 1;
let x_in_parent = (&*x) as *const int as uint;
spawnfn(proc() {
let x_in_child = (&*x) as *const int as uint;
tx.send(x_in_child);
});
let x_in_child = rx.recv();
assert_eq!(x_in_parent, x_in_child);
}
#[test]
fn test_avoid_copying_the_body_spawn() {
avoid_copying_the_body(|v| { Thread::spawn(v); });
}
#[test]
fn test_avoid_copying_the_body_thread_spawn() {
avoid_copying_the_body(|f| {
let builder = cfg();
builder.spawn(proc() {
f();
});
})
}
#[test]
fn test_avoid_copying_the_body_join() {
avoid_copying_the_body(|f| {
let _ = Thread::with_join(proc() {
f()
}).join();
})
}
#[test]
fn test_child_doesnt_ref_parent() {
// If the child refcounts the parent task, this will stack overflow when
// climbing the task tree to dereference each ancestor. (See #1789)
// (well, it would if the constant were 8000+ - I lowered it to be more
// valgrind-friendly. try this at home, instead..!)
static GENERATIONS: uint = 16;
fn child_no(x: uint) -> proc(): Send {
return proc() {
if x < GENERATIONS {
Thread::spawn(child_no(x+1));
}
}
}
Thread::spawn(child_no(0));
}
#[test]
fn test_simple_newsched_spawn() {
Thread::spawn(proc()());
}
#[test]
fn test_try_panic_message_static_str() {
match Thread::with_join(proc() {
panic!("static string");
}).join() {
Err(e) => {
type T = &'static str;
assert!(e.is::<T>());
assert_eq!(*e.downcast::<T>().unwrap(), "static string");
}
Ok(()) => panic!()
}
}
#[test]
fn test_try_panic_message_owned_str() {
match Thread::with_join(proc() {
panic!("owned string".to_string());
}).join() {
Err(e) => {
type T = String;
assert!(e.is::<T>());
assert_eq!(*e.downcast::<T>().unwrap(), "owned string".to_string());
}
Ok(()) => panic!()
}
}
#[test]
fn test_try_panic_message_any() {
match Thread::with_join(proc() {
panic!(box 413u16 as Box<Any + Send>);
}).join() {
Err(e) => {
type T = Box<Any + Send>;
assert!(e.is::<T>());
let any = e.downcast::<T>().unwrap();
assert!(any.is::<u16>());
assert_eq!(*any.downcast::<u16>().unwrap(), 413u16);
}
Ok(()) => panic!()
}
}
#[test]
fn test_try_panic_message_unit_struct() {
struct Juju;
match Thread::with_join(proc() {
panic!(Juju)
}).join() {
Err(ref e) if e.is::<Juju>() => {}
Err(_) | Ok(()) => panic!()
}
}
#[test]
fn test_stdout() {
let (tx, rx) = channel();
let mut reader = ChanReader::new(rx);
let stdout = ChanWriter::new(tx);
let r = cfg().stdout(box stdout as Box<Writer + Send>).with_join(proc() {
print!("Hello, world!");
}).join();
assert!(r.is_ok());
let output = reader.read_to_string().unwrap();
assert_eq!(output, "Hello, world!".to_string());
}
// NOTE: the corresponding test for stderr is in run-pass/task-stderr, due
// to the test harness apparently interfering with stderr configuration.
}