1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
// Copyright 2023 The Matrix.org Foundation C.I.C.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

//! Collection of small helpers that implement store-based locks.
//!
//! This is a per-process lock that may be used only for very specific use
//! cases, where multiple processes might concurrently write to the same
//! database at the same time; this would invalidate store caches, so
//! that should be done mindfully. Such a lock can be acquired multiple times by
//! the same process, and it remains active as long as there's at least one user
//! in a given process.
//!
//! The lock is implemented using time-based leases to values inserted in a
//! store. The store maintains the lock identifier (key), who's the
//! current holder (value), and an expiration timestamp on the side; see also
//! `CryptoStore::try_take_leased_lock` for more details.
//!
//! The lock is initially acquired for a certain period of time (namely, the
//! duration of a lease, aka `LEASE_DURATION_MS`), and then a "heartbeat" task
//! renews the lease to extend its duration, every so often (namely, every
//! `EXTEND_LEASE_EVERY_MS`). Since the tokio scheduler might be busy, the
//! extension request should happen way more frequently than the duration of a
//! lease, in case a deadline is missed. The current values have been chosen to
//! reflect that, with a ratio of 1:10 as of 2023-06-23.
//!
//! Releasing the lock happens naturally, by not renewing a lease. It happens
//! automatically after the duration of the last lease, at most.

use std::{
    error::Error,
    sync::{
        atomic::{self, AtomicU32},
        Arc,
    },
    time::Duration,
};

use tokio::{sync::Mutex, time::sleep};
use tracing::{debug, error, info, instrument, trace};

use crate::{
    executor::{spawn, JoinHandle},
    SendOutsideWasm,
};

/// Backing store for a cross-process lock.
#[cfg_attr(target_arch = "wasm32", async_trait::async_trait(?Send))]
#[cfg_attr(not(target_arch = "wasm32"), async_trait::async_trait)]
pub trait BackingStore {
    type Error: Error + Send + Sync;

    /// Try to take a lock using the given store.
    async fn try_lock(
        &self,
        lease_duration_ms: u32,
        key: &str,
        holder: &str,
    ) -> Result<bool, Self::Error>;
}

/// Small state machine to handle wait times.
#[derive(Clone, Debug)]
enum WaitingTime {
    /// Some time to wait, in milliseconds.
    Some(u32),
    /// Stop waiting when seeing this value.
    Stop,
}

/// A guard on the store lock.
///
/// The lock will be automatically released a short period of time after all the
/// guards have dropped.
#[derive(Debug)]
pub struct CrossProcessStoreLockGuard {
    num_holders: Arc<AtomicU32>,
}

impl Drop for CrossProcessStoreLockGuard {
    fn drop(&mut self) {
        self.num_holders.fetch_sub(1, atomic::Ordering::SeqCst);
    }
}

/// A store-based lock for a `Store`.
///
/// See the doc-comment of this module for more information.
#[derive(Clone, Debug)]
pub struct CrossProcessStoreLock<S: BackingStore + Clone + SendOutsideWasm + 'static> {
    /// The store we're using to lock.
    store: S,

    /// Number of holders of the lock in this process.
    ///
    /// If greater than 0, this means we've already acquired this lock, in this
    /// process, and the store lock mustn't be touched.
    ///
    /// When the number of holders is decreased to 0, then the lock must be
    /// released in the store.
    num_holders: Arc<AtomicU32>,

    /// A mutex to control an attempt to take the lock, to avoid making it
    /// reentrant.
    locking_attempt: Arc<Mutex<()>>,

    /// Current renew task spawned by `try_lock_once`.
    renew_task: Arc<Mutex<Option<JoinHandle<()>>>>,

    /// The key used in the key/value mapping for the lock entry.
    lock_key: String,

    /// A specific value to identify the lock's holder.
    lock_holder: String,

    /// Backoff time, in milliseconds.
    backoff: Arc<Mutex<WaitingTime>>,
}

/// Amount of time a lease of the lock should last, in milliseconds.
pub const LEASE_DURATION_MS: u32 = 500;

/// Period of time between two attempts to extend the lease. We'll
/// re-request a lease for an entire duration of `LEASE_DURATION_MS`
/// milliseconds, every `EXTEND_LEASE_EVERY_MS`, so this has to
/// be an amount safely low compared to `LEASE_DURATION_MS`, to make sure
/// that we can miss a deadline without compromising the lock.
pub const EXTEND_LEASE_EVERY_MS: u64 = 50;

/// Initial backoff, in milliseconds. This is the time we wait the first
/// time, if taking the lock initially failed.
const INITIAL_BACKOFF_MS: u32 = 10;

/// Maximal backoff, in milliseconds. This is the maximum amount of time
/// we'll wait for the lock, *between two attempts*.
pub const MAX_BACKOFF_MS: u32 = 1000;

impl<S: BackingStore + Clone + SendOutsideWasm + 'static> CrossProcessStoreLock<S> {
    /// Create a new store-based lock implemented as a value in the store.
    ///
    /// # Parameters
    ///
    /// - `lock_key`: key in the key-value store to store the lock's state.
    /// - `lock_holder`: identify the lock's holder with this given value.
    pub fn new(store: S, lock_key: String, lock_holder: String) -> Self {
        Self {
            store,
            lock_key,
            lock_holder,
            backoff: Arc::new(Mutex::new(WaitingTime::Some(INITIAL_BACKOFF_MS))),
            num_holders: Arc::new(0.into()),
            locking_attempt: Arc::new(Mutex::new(())),
            renew_task: Default::default(),
        }
    }

    /// Try to lock once, returns whether the lock was obtained or not.
    #[instrument(skip(self), fields(?self.lock_key, ?self.lock_holder))]
    pub async fn try_lock_once(
        &self,
    ) -> Result<Option<CrossProcessStoreLockGuard>, LockStoreError> {
        // Hold onto the locking attempt mutex for the entire lifetime of this
        // function, to avoid multiple reentrant calls.
        let mut _attempt = self.locking_attempt.lock().await;

        // If another thread obtained the lock, make sure to only superficially increase
        // the number of holders, and carry on.
        if self.num_holders.load(atomic::Ordering::SeqCst) > 0 {
            // Note: between the above load and the fetch_add below, another thread may
            // decrement `num_holders`. That's fine because that means the lock
            // was taken by at least one thread, and after this call it will be
            // taken by at least one thread.
            trace!("We already had the lock, incrementing holder count");
            self.num_holders.fetch_add(1, atomic::Ordering::SeqCst);
            let guard = CrossProcessStoreLockGuard { num_holders: self.num_holders.clone() };
            return Ok(Some(guard));
        }

        let acquired = self
            .store
            .try_lock(LEASE_DURATION_MS, &self.lock_key, &self.lock_holder)
            .await
            .map_err(|err| LockStoreError::BackingStoreError(Box::new(err)))?;

        if !acquired {
            trace!("Couldn't acquire the lock immediately.");
            return Ok(None);
        }

        trace!("Acquired the lock, spawning the lease extension task.");

        // This is the first time we've acquired the lock. We're going to spawn the task
        // that will renew the lease.

        // Clone data to be owned by the task.
        let this = (*self).clone();

        let mut renew_task = self.renew_task.lock().await;

        // Cancel the previous task, if any. That's safe to do, because:
        // - either the task was done,
        // - or it was still running, but taking a lock in the db has to be an atomic
        //   operation running in a transaction.

        if let Some(_prev) = renew_task.take() {
            #[cfg(not(target_arch = "wasm32"))]
            if !_prev.is_finished() {
                info!("aborting the previous renew task");
                _prev.abort();
            }
        }

        // Restart a new one.
        *renew_task = Some(spawn(async move {
            loop {
                {
                    // First, check if there are still users of this lock.
                    //
                    // This is not racy, because:
                    // - the `locking_attempt` mutex makes sure we don't have unexpected
                    // interactions with the non-atomic sequence above in `try_lock_once`
                    // (check > 0, then add 1).
                    // - other entities holding onto the `num_holders` atomic will only
                    // decrease it over time.

                    let _guard = this.locking_attempt.lock().await;

                    // If there are no more users, we can quit.
                    if this.num_holders.load(atomic::Ordering::SeqCst) == 0 {
                        info!("exiting the lease extension loop");

                        // Cancel the lease with another 0ms lease.
                        // If we don't get the lock, that's (weird but) fine.
                        let fut = this.store.try_lock(0, &this.lock_key, &this.lock_holder);
                        let _ = fut.await;

                        // Exit the loop.
                        break;
                    }
                }

                sleep(Duration::from_millis(EXTEND_LEASE_EVERY_MS)).await;

                let fut = this.store.try_lock(LEASE_DURATION_MS, &this.lock_key, &this.lock_holder);
                if let Err(err) = fut.await {
                    error!("error when extending lock lease: {err:#}");
                    // Exit the loop.
                    break;
                }
            }
        }));

        self.num_holders.fetch_add(1, atomic::Ordering::SeqCst);

        let guard = CrossProcessStoreLockGuard { num_holders: self.num_holders.clone() };
        Ok(Some(guard))
    }

    /// Attempt to take the lock, with exponential backoff if the lock has
    /// already been taken before.
    ///
    /// The `max_backoff` parameter is the maximum time (in milliseconds) that
    /// should be waited for, between two attempts. When that time is
    /// reached a second time, the lock will stop attempting to get the lock
    /// and will return a timeout error upon locking. If not provided,
    /// will wait for [`MAX_BACKOFF_MS`].
    #[instrument(skip(self), fields(?self.lock_key, ?self.lock_holder))]
    pub async fn spin_lock(
        &self,
        max_backoff: Option<u32>,
    ) -> Result<CrossProcessStoreLockGuard, LockStoreError> {
        let max_backoff = max_backoff.unwrap_or(MAX_BACKOFF_MS);

        // Note: reads/writes to the backoff are racy across threads in theory, but the
        // lock in `try_lock_once` should sequentialize it all.

        loop {
            if let Some(guard) = self.try_lock_once().await? {
                // Reset backoff before returning, for the next attempt to lock.
                *self.backoff.lock().await = WaitingTime::Some(INITIAL_BACKOFF_MS);
                return Ok(guard);
            }

            // Exponential backoff! Multiply by 2 the time we've waited before, cap it to
            // max_backoff.
            let mut backoff = self.backoff.lock().await;

            let wait = match &mut *backoff {
                WaitingTime::Some(ref mut val) => {
                    let wait = *val;
                    *val = val.saturating_mul(2);
                    if *val >= max_backoff {
                        *backoff = WaitingTime::Stop;
                    }
                    wait
                }
                WaitingTime::Stop => {
                    // We've reached the maximum backoff, abandon.
                    return Err(LockStoreError::LockTimeout);
                }
            };

            debug!("Waiting {wait} before re-attempting to take the lock");
            sleep(Duration::from_millis(wait.into())).await;
        }
    }

    /// Returns the value in the database that represents the holder's
    /// identifier.
    pub fn lock_holder(&self) -> &str {
        &self.lock_holder
    }
}

/// Error related to the locking API of the store.
#[derive(Debug, thiserror::Error)]
pub enum LockStoreError {
    /// Spent too long waiting for a database lock.
    #[error("a lock timed out")]
    LockTimeout,

    #[error(transparent)]
    BackingStoreError(#[from] Box<dyn Error + Send + Sync>),
}

#[cfg(test)]
#[cfg(not(target_arch = "wasm32"))] // These tests require tokio::time, which is not implemented on wasm.
mod tests {
    use std::{
        collections::HashMap,
        sync::{atomic, Arc, Mutex},
        time::Instant,
    };

    use assert_matches::assert_matches;
    use matrix_sdk_test::async_test;
    use tokio::{
        spawn,
        time::{sleep, Duration},
    };

    use super::{
        BackingStore, CrossProcessStoreLock, CrossProcessStoreLockGuard, LockStoreError,
        EXTEND_LEASE_EVERY_MS,
    };

    #[derive(Clone, Default)]
    struct TestStore {
        leases: Arc<Mutex<HashMap<String, (String, Instant)>>>,
    }

    impl TestStore {
        fn try_take_leased_lock(&self, lease_duration_ms: u32, key: &str, holder: &str) -> bool {
            let now = Instant::now();
            let expiration = now + Duration::from_millis(lease_duration_ms.into());
            let mut leases = self.leases.lock().unwrap();
            if let Some(prev) = leases.get_mut(key) {
                if prev.0 == holder {
                    // We had the lease before, extend it.
                    prev.1 = expiration;
                    true
                } else {
                    // We didn't have it.
                    if prev.1 < now {
                        // Steal it!
                        prev.0 = holder.to_owned();
                        prev.1 = expiration;
                        true
                    } else {
                        // We tried our best.
                        false
                    }
                }
            } else {
                leases.insert(
                    key.to_owned(),
                    (
                        holder.to_owned(),
                        Instant::now() + Duration::from_millis(lease_duration_ms.into()),
                    ),
                );
                true
            }
        }
    }

    #[derive(Debug, thiserror::Error)]
    enum DummyError {}

    #[cfg_attr(target_arch = "wasm32", async_trait::async_trait(?Send))]
    #[cfg_attr(not(target_arch = "wasm32"), async_trait::async_trait)]
    impl BackingStore for TestStore {
        type Error = DummyError;

        /// Try to take a lock using the given store.
        async fn try_lock(
            &self,
            lease_duration_ms: u32,
            key: &str,
            holder: &str,
        ) -> Result<bool, Self::Error> {
            Ok(self.try_take_leased_lock(lease_duration_ms, key, holder))
        }
    }

    async fn release_lock(guard: Option<CrossProcessStoreLockGuard>) {
        drop(guard);
        sleep(Duration::from_millis(EXTEND_LEASE_EVERY_MS)).await;
    }

    type TestResult = Result<(), LockStoreError>;

    #[async_test]
    async fn test_simple_lock_unlock() -> TestResult {
        let store = TestStore::default();
        let lock = CrossProcessStoreLock::new(store, "key".to_owned(), "first".to_owned());

        // The lock plain works when used with a single holder.
        let acquired = lock.try_lock_once().await?;
        assert!(acquired.is_some());
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 1);

        // Releasing works.
        release_lock(acquired).await;
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 0);

        // Spin locking on the same lock always works, assuming no concurrent access.
        let acquired = lock.spin_lock(None).await.unwrap();

        // Releasing still works.
        release_lock(Some(acquired)).await;
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 0);

        Ok(())
    }

    #[async_test]
    async fn test_self_recovery() -> TestResult {
        let store = TestStore::default();
        let lock = CrossProcessStoreLock::new(store.clone(), "key".to_owned(), "first".to_owned());

        // When a lock is acquired...
        let acquired = lock.try_lock_once().await?;
        assert!(acquired.is_some());
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 1);

        // But then forgotten... (note: no need to release the guard)
        drop(lock);

        // And when rematerializing the lock with the same key/value...
        let lock = CrossProcessStoreLock::new(store.clone(), "key".to_owned(), "first".to_owned());

        // We still got it.
        let acquired = lock.try_lock_once().await?;
        assert!(acquired.is_some());
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 1);

        Ok(())
    }

    #[async_test]
    async fn test_multiple_holders_same_process() -> TestResult {
        let store = TestStore::default();
        let lock = CrossProcessStoreLock::new(store, "key".to_owned(), "first".to_owned());

        // Taking the lock twice...
        let acquired = lock.try_lock_once().await?;
        assert!(acquired.is_some());

        let acquired2 = lock.try_lock_once().await?;
        assert!(acquired2.is_some());

        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 2);

        // ...means we can release it twice.
        release_lock(acquired).await;
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 1);

        release_lock(acquired2).await;
        assert_eq!(lock.num_holders.load(atomic::Ordering::SeqCst), 0);

        Ok(())
    }

    #[async_test]
    async fn test_multiple_processes() -> TestResult {
        let store = TestStore::default();
        let lock1 = CrossProcessStoreLock::new(store.clone(), "key".to_owned(), "first".to_owned());
        let lock2 = CrossProcessStoreLock::new(store, "key".to_owned(), "second".to_owned());

        // When the first process takes the lock...
        let acquired1 = lock1.try_lock_once().await?;
        assert!(acquired1.is_some());

        // The second can't take it immediately.
        let acquired2 = lock2.try_lock_once().await?;
        assert!(acquired2.is_none());

        let lock2_clone = lock2.clone();
        let handle = spawn(async move { lock2_clone.spin_lock(Some(1000)).await });

        sleep(Duration::from_millis(100)).await;

        drop(acquired1);

        // lock2 in the background manages to get the lock at some point.
        let _acquired2 = handle
            .await
            .expect("join handle is properly awaited")
            .expect("lock was obtained after spin-locking");

        // Now if lock1 tries to get the lock with a small timeout, it will fail.
        assert_matches!(lock1.spin_lock(Some(200)).await, Err(LockStoreError::LockTimeout));

        Ok(())
    }
}