parking_lot/fair_mutex.rs
1// Copyright 2016 Amanieu d'Antras
2//
3// Licensed under the Apache License, Version 2.0, <LICENSE-APACHE or
4// http://apache.org/licenses/LICENSE-2.0> or the MIT license <LICENSE-MIT or
5// http://opensource.org/licenses/MIT>, at your option. This file may not be
6// copied, modified, or distributed except according to those terms.
7
8use crate::raw_fair_mutex::RawFairMutex;
9use lock_api;
10
11/// A mutual exclusive primitive that is always fair, useful for protecting shared data
12///
13/// This mutex will block threads waiting for the lock to become available. The
14/// mutex can be statically initialized or created by the `new`
15/// constructor. Each mutex has a type parameter which represents the data that
16/// it is protecting. The data can only be accessed through the RAII guards
17/// returned from `lock` and `try_lock`, which guarantees that the data is only
18/// ever accessed when the mutex is locked.
19///
20/// The regular mutex provided by `parking_lot` uses eventual fairness
21/// (after some time it will default to the fair algorithm), but eventual
22/// fairness does not provide the same guarantees an always fair method would.
23/// Fair mutexes are generally slower, but sometimes needed.
24///
25/// In a fair mutex the waiters form a queue, and the lock is always granted to
26/// the next requester in the queue, in first-in first-out order. This ensures
27/// that one thread cannot starve others by quickly re-acquiring the lock after
28/// releasing it.
29///
30/// A fair mutex may not be interesting if threads have different priorities (this is known as
31/// priority inversion).
32///
33/// # Differences from the standard library `Mutex`
34///
35/// - No poisoning, the lock is released normally on panic.
36/// - Only requires 1 byte of space, whereas the standard library boxes the
37/// `FairMutex` due to platform limitations.
38/// - Can be statically constructed (requires the `const_fn` nightly feature).
39/// - Does not require any drop glue when dropped.
40/// - Inline fast path for the uncontended case.
41/// - Efficient handling of micro-contention using adaptive spinning.
42/// - Allows raw locking & unlocking without a guard.
43///
44/// # Examples
45///
46/// ```
47/// use parking_lot::FairMutex;
48/// use std::sync::{Arc, mpsc::channel};
49/// use std::thread;
50///
51/// const N: usize = 10;
52///
53/// // Spawn a few threads to increment a shared variable (non-atomically), and
54/// // let the main thread know once all increments are done.
55/// //
56/// // Here we're using an Arc to share memory among threads, and the data inside
57/// // the Arc is protected with a mutex.
58/// let data = Arc::new(FairMutex::new(0));
59///
60/// let (tx, rx) = channel();
61/// for _ in 0..10 {
62/// let (data, tx) = (Arc::clone(&data), tx.clone());
63/// thread::spawn(move || {
64/// // The shared state can only be accessed once the lock is held.
65/// // Our non-atomic increment is safe because we're the only thread
66/// // which can access the shared state when the lock is held.
67/// let mut data = data.lock();
68/// *data += 1;
69/// if *data == N {
70/// tx.send(()).unwrap();
71/// }
72/// // the lock is unlocked here when `data` goes out of scope.
73/// });
74/// }
75///
76/// rx.recv().unwrap();
77/// ```
78pub type FairMutex<T> = lock_api::Mutex<RawFairMutex, T>;
79
80/// Creates a new fair mutex in an unlocked state ready for use.
81///
82/// This allows creating a fair mutex in a constant context on stable Rust.
83pub const fn const_fair_mutex<T>(val: T) -> FairMutex<T> {
84 FairMutex::const_new(<RawFairMutex as lock_api::RawMutex>::INIT, val)
85}
86
87/// An RAII implementation of a "scoped lock" of a mutex. When this structure is
88/// dropped (falls out of scope), the lock will be unlocked.
89///
90/// The data protected by the mutex can be accessed through this guard via its
91/// `Deref` and `DerefMut` implementations.
92pub type FairMutexGuard<'a, T> = lock_api::MutexGuard<'a, RawFairMutex, T>;
93
94/// An RAII mutex guard returned by `FairMutexGuard::map`, which can point to a
95/// subfield of the protected data.
96///
97/// The main difference between `MappedFairMutexGuard` and `FairMutexGuard` is that the
98/// former doesn't support temporarily unlocking and re-locking, since that
99/// could introduce soundness issues if the locked object is modified by another
100/// thread.
101pub type MappedFairMutexGuard<'a, T> = lock_api::MappedMutexGuard<'a, RawFairMutex, T>;
102
103#[cfg(test)]
104mod tests {
105 use crate::FairMutex;
106 use std::sync::atomic::{AtomicUsize, Ordering};
107 use std::sync::mpsc::channel;
108 use std::sync::Arc;
109 use std::thread;
110
111 #[cfg(feature = "serde")]
112 use bincode::{deserialize, serialize};
113
114 #[derive(Eq, PartialEq, Debug)]
115 struct NonCopy(i32);
116
117 #[test]
118 fn smoke() {
119 let m = FairMutex::new(());
120 drop(m.lock());
121 drop(m.lock());
122 }
123
124 #[test]
125 fn lots_and_lots() {
126 const J: u32 = 1000;
127 const K: u32 = 3;
128
129 let m = Arc::new(FairMutex::new(0));
130
131 fn inc(m: &FairMutex<u32>) {
132 for _ in 0..J {
133 *m.lock() += 1;
134 }
135 }
136
137 let (tx, rx) = channel();
138 for _ in 0..K {
139 let tx2 = tx.clone();
140 let m2 = m.clone();
141 thread::spawn(move || {
142 inc(&m2);
143 tx2.send(()).unwrap();
144 });
145 let tx2 = tx.clone();
146 let m2 = m.clone();
147 thread::spawn(move || {
148 inc(&m2);
149 tx2.send(()).unwrap();
150 });
151 }
152
153 drop(tx);
154 for _ in 0..2 * K {
155 rx.recv().unwrap();
156 }
157 assert_eq!(*m.lock(), J * K * 2);
158 }
159
160 #[test]
161 fn try_lock() {
162 let m = FairMutex::new(());
163 *m.try_lock().unwrap() = ();
164 }
165
166 #[test]
167 fn test_into_inner() {
168 let m = FairMutex::new(NonCopy(10));
169 assert_eq!(m.into_inner(), NonCopy(10));
170 }
171
172 #[test]
173 fn test_into_inner_drop() {
174 struct Foo(Arc<AtomicUsize>);
175 impl Drop for Foo {
176 fn drop(&mut self) {
177 self.0.fetch_add(1, Ordering::SeqCst);
178 }
179 }
180 let num_drops = Arc::new(AtomicUsize::new(0));
181 let m = FairMutex::new(Foo(num_drops.clone()));
182 assert_eq!(num_drops.load(Ordering::SeqCst), 0);
183 {
184 let _inner = m.into_inner();
185 assert_eq!(num_drops.load(Ordering::SeqCst), 0);
186 }
187 assert_eq!(num_drops.load(Ordering::SeqCst), 1);
188 }
189
190 #[test]
191 fn test_get_mut() {
192 let mut m = FairMutex::new(NonCopy(10));
193 *m.get_mut() = NonCopy(20);
194 assert_eq!(m.into_inner(), NonCopy(20));
195 }
196
197 #[test]
198 fn test_mutex_arc_nested() {
199 // Tests nested mutexes and access
200 // to underlying data.
201 let arc = Arc::new(FairMutex::new(1));
202 let arc2 = Arc::new(FairMutex::new(arc));
203 let (tx, rx) = channel();
204 let _t = thread::spawn(move || {
205 let lock = arc2.lock();
206 let lock2 = lock.lock();
207 assert_eq!(*lock2, 1);
208 tx.send(()).unwrap();
209 });
210 rx.recv().unwrap();
211 }
212
213 #[test]
214 fn test_mutex_arc_access_in_unwind() {
215 let arc = Arc::new(FairMutex::new(1));
216 let arc2 = arc.clone();
217 let _ = thread::spawn(move || {
218 struct Unwinder {
219 i: Arc<FairMutex<i32>>,
220 }
221 impl Drop for Unwinder {
222 fn drop(&mut self) {
223 *self.i.lock() += 1;
224 }
225 }
226 let _u = Unwinder { i: arc2 };
227 panic!();
228 })
229 .join();
230 let lock = arc.lock();
231 assert_eq!(*lock, 2);
232 }
233
234 #[test]
235 fn test_mutex_unsized() {
236 let mutex: &FairMutex<[i32]> = &FairMutex::new([1, 2, 3]);
237 {
238 let b = &mut *mutex.lock();
239 b[0] = 4;
240 b[2] = 5;
241 }
242 let comp: &[i32] = &[4, 2, 5];
243 assert_eq!(&*mutex.lock(), comp);
244 }
245
246 #[test]
247 fn test_mutexguard_sync() {
248 fn sync<T: Sync>(_: T) {}
249
250 let mutex = FairMutex::new(());
251 sync(mutex.lock());
252 }
253
254 #[test]
255 fn test_mutex_debug() {
256 let mutex = FairMutex::new(vec![0u8, 10]);
257
258 assert_eq!(format!("{:?}", mutex), "Mutex { data: [0, 10] }");
259 let _lock = mutex.lock();
260 assert_eq!(format!("{:?}", mutex), "Mutex { data: <locked> }");
261 }
262
263 #[cfg(feature = "serde")]
264 #[test]
265 fn test_serde() {
266 let contents: Vec<u8> = vec![0, 1, 2];
267 let mutex = FairMutex::new(contents.clone());
268
269 let serialized = serialize(&mutex).unwrap();
270 let deserialized: FairMutex<Vec<u8>> = deserialize(&serialized).unwrap();
271
272 assert_eq!(*(mutex.lock()), *(deserialized.lock()));
273 assert_eq!(contents, *(deserialized.lock()));
274 }
275}