places/api/
places_api.rs

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
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

use crate::bookmark_sync::BookmarksSyncEngine;
use crate::db::db::{PlacesDb, SharedPlacesDb};
use crate::error::*;
use crate::history_sync::HistorySyncEngine;
use crate::storage::{
    self, bookmarks::bookmark_sync, delete_meta, get_meta, history::history_sync, put_meta,
};
use crate::util::normalize_path;
use error_support::handle_error;
use interrupt_support::register_interrupt;
use lazy_static::lazy_static;
use parking_lot::Mutex;
use rusqlite::OpenFlags;
use std::cell::Cell;
use std::collections::HashMap;
use std::path::{Path, PathBuf};
use std::sync::{
    atomic::{AtomicUsize, Ordering},
    Arc, Weak,
};
use sync15::client::{sync_multiple, MemoryCachedState, Sync15StorageClientInit, SyncResult};
use sync15::engine::{EngineSyncAssociation, SyncEngine, SyncEngineId};
use sync15::{telemetry, KeyBundle};

// Not clear if this should be here, but this is the "global sync state"
// which is persisted to disk and reused for all engines.
// Note that this is only ever round-tripped, and never changed by, or impacted
// by a store or collection, so it's safe to storage globally rather than
// per collection.
pub const GLOBAL_STATE_META_KEY: &str = "global_sync_state_v2";

// Our "sync manager" will use whatever is stashed here.
lazy_static::lazy_static! {
    // Mutex: just taken long enough to update the contents - needed to wrap
    //        the Weak as it isn't `Sync`
    // [Arc/Weak]: Stores the places api used to create the connection for
    //             BookmarksSyncEngine/HistorySyncEngine
    static ref PLACES_API_FOR_SYNC_MANAGER: Mutex<Weak<PlacesApi>> = Mutex::new(Weak::new());
}

// Called by the sync manager to get a sync engine via the PlacesApi previously
// registered with the sync manager.
pub fn get_registered_sync_engine(engine_id: &SyncEngineId) -> Option<Box<dyn SyncEngine>> {
    match PLACES_API_FOR_SYNC_MANAGER.lock().upgrade() {
        None => {
            log::warn!("places: get_registered_sync_engine: no PlacesApi registered");
            None
        }
        Some(places_api) => match create_sync_engine(&places_api, engine_id) {
            Ok(engine) => Some(engine),
            Err(e) => {
                // Report this to Sentry, except if it's an open database error.  That indicates
                // that there is a registered sync engine, but the connection is busy so we can't
                // open it.  This is a known issue that we don't need more reports for (see
                // https://github.com/mozilla/application-services/issues/5237 for discussion).
                if !matches!(e, Error::OpenDatabaseError(_)) {
                    error_support::report_error!(
                        "places-no-registered-sync-engine",
                        "places: get_registered_sync_engine: {}",
                        e
                    );
                }
                None
            }
        },
    }
}

fn create_sync_engine(
    places_api: &PlacesApi,
    engine_id: &SyncEngineId,
) -> Result<Box<dyn SyncEngine>> {
    let conn = places_api.get_sync_connection()?;
    match engine_id {
        SyncEngineId::Bookmarks => Ok(Box::new(BookmarksSyncEngine::new(conn)?)),
        SyncEngineId::History => Ok(Box::new(HistorySyncEngine::new(conn)?)),
        _ => unreachable!("can't provide unknown engine: {}", engine_id),
    }
}

#[repr(u8)]
#[derive(Debug, Copy, Clone, PartialEq, Eq)]
pub enum ConnectionType {
    ReadOnly = 1,
    ReadWrite = 2,
    Sync = 3,
}

impl ConnectionType {
    pub fn from_primitive(p: u8) -> Option<Self> {
        match p {
            1 => Some(ConnectionType::ReadOnly),
            2 => Some(ConnectionType::ReadWrite),
            3 => Some(ConnectionType::Sync),
            _ => None,
        }
    }
}

impl ConnectionType {
    pub fn rusqlite_flags(self) -> OpenFlags {
        let common_flags = OpenFlags::SQLITE_OPEN_NO_MUTEX | OpenFlags::SQLITE_OPEN_URI;
        match self {
            ConnectionType::ReadOnly => common_flags | OpenFlags::SQLITE_OPEN_READ_ONLY,
            ConnectionType::ReadWrite => {
                common_flags | OpenFlags::SQLITE_OPEN_CREATE | OpenFlags::SQLITE_OPEN_READ_WRITE
            }
            ConnectionType::Sync => common_flags | OpenFlags::SQLITE_OPEN_READ_WRITE,
        }
    }
}

// We only allow a single PlacesApi per filename.
lazy_static! {
    static ref APIS: Mutex<HashMap<PathBuf, Weak<PlacesApi>>> = Mutex::new(HashMap::new());
}

static ID_COUNTER: AtomicUsize = AtomicUsize::new(0);

pub struct SyncState {
    pub mem_cached_state: Cell<MemoryCachedState>,
    pub disk_cached_state: Cell<Option<String>>,
}

/// For uniffi we need to expose our `Arc` returning constructor as a global function :(
/// https://github.com/mozilla/uniffi-rs/pull/1063 would fix this, but got some pushback
/// meaning we are forced into this unfortunate workaround.
#[handle_error(crate::Error)]
pub fn places_api_new(db_name: impl AsRef<Path>) -> ApiResult<Arc<PlacesApi>> {
    PlacesApi::new(db_name)
}

/// The entry-point to the places API. This object gives access to database
/// connections and other helpers. It enforces that only 1 write connection
/// can exist to the database at once.
pub struct PlacesApi {
    db_name: PathBuf,
    write_connection: Mutex<Option<PlacesDb>>,
    sync_state: Mutex<Option<SyncState>>,
    coop_tx_lock: Arc<Mutex<()>>,
    // Used for get_sync_connection()
    // - The inner mutex synchronizes sync operation (for example one of the [SyncEngine] methods).
    //   This avoids issues like #867
    // - The weak facilitates connection sharing.  When `get_sync_connection()` returns an Arc, we
    //   keep a weak reference to it.  If the Arc is still alive when `get_sync_connection()` is
    //   called again, we reuse it.
    // - The outer mutex synchronizes the `get_sync_connection()` operation.  If multiple threads
    //   ran that at the same time there would be issues.
    sync_connection: Mutex<Weak<SharedPlacesDb>>,
    id: usize,
}

impl PlacesApi {
    /// Create a new, or fetch an already open, PlacesApi backed by a file on disk.
    pub fn new(db_name: impl AsRef<Path>) -> Result<Arc<Self>> {
        let db_name = normalize_path(db_name)?;
        Self::new_or_existing(db_name)
    }

    /// Create a new, or fetch an already open, memory-based PlacesApi. You must
    /// provide a name, but you are still able to have a single writer and many
    ///  reader connections to the same memory DB open.
    pub fn new_memory(db_name: &str) -> Result<Arc<Self>> {
        let name = PathBuf::from(format!("file:{}?mode=memory&cache=shared", db_name));
        Self::new_or_existing(name)
    }
    fn new_or_existing_into(
        target: &mut HashMap<PathBuf, Weak<PlacesApi>>,
        db_name: PathBuf,
    ) -> Result<Arc<Self>> {
        let id = ID_COUNTER.fetch_add(1, Ordering::SeqCst);
        match target.get(&db_name).and_then(Weak::upgrade) {
            Some(existing) => Ok(existing),
            None => {
                // We always create a new read-write connection for an initial open so
                // we can create the schema and/or do version upgrades.
                let coop_tx_lock = Arc::new(Mutex::new(()));
                let connection = PlacesDb::open(
                    &db_name,
                    ConnectionType::ReadWrite,
                    id,
                    coop_tx_lock.clone(),
                )?;
                let new = PlacesApi {
                    db_name: db_name.clone(),
                    write_connection: Mutex::new(Some(connection)),
                    sync_state: Mutex::new(None),
                    sync_connection: Mutex::new(Weak::new()),
                    id,
                    coop_tx_lock,
                };
                let arc = Arc::new(new);
                target.insert(db_name, Arc::downgrade(&arc));
                Ok(arc)
            }
        }
    }

    fn new_or_existing(db_name: PathBuf) -> Result<Arc<Self>> {
        let mut guard = APIS.lock();
        Self::new_or_existing_into(&mut guard, db_name)
    }

    /// Open a connection to the database.
    pub fn open_connection(&self, conn_type: ConnectionType) -> Result<PlacesDb> {
        match conn_type {
            ConnectionType::ReadOnly => {
                // make a new one - we can have as many of these as we want.
                PlacesDb::open(
                    self.db_name.clone(),
                    ConnectionType::ReadOnly,
                    self.id,
                    self.coop_tx_lock.clone(),
                )
            }
            ConnectionType::ReadWrite => {
                // We only allow one of these.
                let mut guard = self.write_connection.lock();
                match guard.take() {
                    None => Err(Error::ConnectionAlreadyOpen),
                    Some(db) => Ok(db),
                }
            }
            ConnectionType::Sync => {
                panic!("Use `get_sync_connection` to open a sync connection");
            }
        }
    }

    // Get a database connection to sync with
    //
    // This function provides a couple features to facilitate sharing the connection between
    // different sync engines:
    //   - Each connection is wrapped in a `Mutex<>` to synchronize access.
    //   - The mutex is then wrapped in an Arc<>.  If the last Arc<> returned is still alive, then
    //     get_sync_connection() will reuse it.
    pub fn get_sync_connection(&self) -> Result<Arc<SharedPlacesDb>> {
        // First step: lock the outer mutex
        let mut conn = self.sync_connection.lock();
        match conn.upgrade() {
            // If our Weak is still alive, then re-use that
            Some(db) => Ok(db),
            // If not, create a new connection
            None => {
                let db = Arc::new(SharedPlacesDb::new(PlacesDb::open(
                    self.db_name.clone(),
                    ConnectionType::Sync,
                    self.id,
                    self.coop_tx_lock.clone(),
                )?));
                register_interrupt(Arc::<SharedPlacesDb>::downgrade(&db));
                // Store a weakref for next time
                *conn = Arc::downgrade(&db);
                Ok(db)
            }
        }
    }

    /// Close a connection to the database. If the connection is the write
    /// connection, you can re-fetch it using open_connection.
    pub fn close_connection(&self, connection: PlacesDb) -> Result<()> {
        if connection.api_id() != self.id {
            return Err(Error::WrongApiForClose);
        }
        if connection.conn_type() == ConnectionType::ReadWrite {
            // We only allow one of these.
            let mut guard = self.write_connection.lock();
            assert!((*guard).is_none());
            *guard = Some(connection);
        }
        Ok(())
    }

    fn get_disk_persisted_state(&self, conn: &PlacesDb) -> Result<Option<String>> {
        get_meta::<String>(conn, GLOBAL_STATE_META_KEY)
    }

    fn set_disk_persisted_state(&self, conn: &PlacesDb, state: &Option<String>) -> Result<()> {
        match state {
            Some(ref s) => put_meta(conn, GLOBAL_STATE_META_KEY, s),
            None => delete_meta(conn, GLOBAL_STATE_META_KEY),
        }
    }

    // This allows the embedding app to say "make this instance available to
    // the sync manager". The implementation is more like "offer to sync mgr"
    // (thereby avoiding us needing to link with the sync manager) but
    // `register_with_sync_manager()` is logically what's happening so that's
    // the name it gets.
    pub fn register_with_sync_manager(self: Arc<Self>) {
        *PLACES_API_FOR_SYNC_MANAGER.lock() = Arc::downgrade(&self);
    }

    // NOTE: These should be deprecated as soon as possible - that will be once
    // all consumers have been updated to use the .sync() method below, and/or
    // we have implemented the sync manager and migrated consumers to that.
    pub fn sync_history(
        &self,
        client_init: &Sync15StorageClientInit,
        key_bundle: &KeyBundle,
    ) -> Result<telemetry::SyncTelemetryPing> {
        self.do_sync_one(
            "history",
            move |conn, mem_cached_state, disk_cached_state| {
                let engine = HistorySyncEngine::new(conn)?;
                Ok(sync_multiple(
                    &[&engine],
                    disk_cached_state,
                    mem_cached_state,
                    client_init,
                    key_bundle,
                    &interrupt_support::ShutdownInterruptee,
                    None,
                ))
            },
        )
    }

    pub fn sync_bookmarks(
        &self,
        client_init: &Sync15StorageClientInit,
        key_bundle: &KeyBundle,
    ) -> Result<telemetry::SyncTelemetryPing> {
        self.do_sync_one(
            "bookmarks",
            move |conn, mem_cached_state, disk_cached_state| {
                let engine = BookmarksSyncEngine::new(conn)?;
                Ok(sync_multiple(
                    &[&engine],
                    disk_cached_state,
                    mem_cached_state,
                    client_init,
                    key_bundle,
                    &interrupt_support::ShutdownInterruptee,
                    None,
                ))
            },
        )
    }

    pub fn do_sync_one<F>(
        &self,
        name: &'static str,
        syncer: F,
    ) -> Result<telemetry::SyncTelemetryPing>
    where
        F: FnOnce(
            Arc<SharedPlacesDb>,
            &mut MemoryCachedState,
            &mut Option<String>,
        ) -> Result<SyncResult>,
    {
        let mut guard = self.sync_state.lock();
        let conn = self.get_sync_connection()?;
        if guard.is_none() {
            *guard = Some(SyncState {
                mem_cached_state: Cell::default(),
                disk_cached_state: Cell::new(self.get_disk_persisted_state(&conn.lock())?),
            });
        }

        let sync_state = guard.as_ref().unwrap();

        let mut mem_cached_state = sync_state.mem_cached_state.take();
        let mut disk_cached_state = sync_state.disk_cached_state.take();
        let mut result = syncer(conn.clone(), &mut mem_cached_state, &mut disk_cached_state)?;
        // even on failure we set the persisted state - sync itself takes care
        // to ensure this has been None'd out if necessary.
        self.set_disk_persisted_state(&conn.lock(), &disk_cached_state)?;
        sync_state.mem_cached_state.replace(mem_cached_state);
        sync_state.disk_cached_state.replace(disk_cached_state);

        // for b/w compat reasons, we do some dances with the result.
        if let Err(e) = result.result {
            return Err(e.into());
        }
        match result.engine_results.remove(name) {
            None | Some(Ok(())) => Ok(result.telemetry),
            Some(Err(e)) => Err(e.into()),
        }
    }

    // This is the new sync API until the sync manager lands. It's currently
    // not wired up via the FFI - it's possible we'll do declined engines too
    // before we do.
    // Note we've made a policy decision about the return value - even though
    // it is Result<SyncResult>, we will only return an Err() if there's a
    // fatal error that prevents us starting a sync, such as failure to open
    // the DB. Any errors that happen *after* sync must not escape - ie, once
    // we have a SyncResult, we must return it.
    pub fn sync(
        &self,
        client_init: &Sync15StorageClientInit,
        key_bundle: &KeyBundle,
    ) -> Result<SyncResult> {
        let mut guard = self.sync_state.lock();
        let conn = self.get_sync_connection()?;
        if guard.is_none() {
            *guard = Some(SyncState {
                mem_cached_state: Cell::default(),
                disk_cached_state: Cell::new(self.get_disk_persisted_state(&conn.lock())?),
            });
        }

        let sync_state = guard.as_ref().unwrap();

        let bm_engine = BookmarksSyncEngine::new(conn.clone())?;
        let history_engine = HistorySyncEngine::new(conn.clone())?;
        let mut mem_cached_state = sync_state.mem_cached_state.take();
        let mut disk_cached_state = sync_state.disk_cached_state.take();

        // NOTE: After here we must never return Err()!
        let result = sync_multiple(
            &[&history_engine, &bm_engine],
            &mut disk_cached_state,
            &mut mem_cached_state,
            client_init,
            key_bundle,
            &interrupt_support::ShutdownInterruptee,
            None,
        );
        // even on failure we set the persisted state - sync itself takes care
        // to ensure this has been None'd out if necessary.
        if let Err(e) = self.set_disk_persisted_state(&conn.lock(), &disk_cached_state) {
            error_support::report_error!(
                "places-sync-persist-failure",
                "Failed to persist the sync state: {:?}",
                e
            );
        }
        sync_state.mem_cached_state.replace(mem_cached_state);
        sync_state.disk_cached_state.replace(disk_cached_state);

        Ok(result)
    }

    pub fn wipe_bookmarks(&self) -> Result<()> {
        // Take the lock to prevent syncing while we're doing this.
        let _guard = self.sync_state.lock();
        let conn = self.get_sync_connection()?;

        storage::bookmarks::delete_everything(&conn.lock())?;
        Ok(())
    }

    pub fn reset_bookmarks(&self) -> Result<()> {
        // Take the lock to prevent syncing while we're doing this.
        let _guard = self.sync_state.lock();
        let conn = self.get_sync_connection()?;

        bookmark_sync::reset(&conn.lock(), &EngineSyncAssociation::Disconnected)?;
        Ok(())
    }

    #[handle_error(crate::Error)]
    pub fn reset_history(&self) -> ApiResult<()> {
        // Take the lock to prevent syncing while we're doing this.
        let _guard = self.sync_state.lock();
        let conn = self.get_sync_connection()?;

        history_sync::reset(&conn.lock(), &EngineSyncAssociation::Disconnected)?;
        Ok(())
    }
}

#[cfg(test)]
pub mod test {
    use super::*;
    use std::sync::atomic::{AtomicUsize, Ordering};

    // A helper for our tests to get their own memory Api.
    static ATOMIC_COUNTER: AtomicUsize = AtomicUsize::new(0);

    pub fn new_mem_api() -> Arc<PlacesApi> {
        // A bit hacky, but because this is a test-only function that almost all tests use,
        // it's a convenient place to initialize logging for tests.
        let _ = env_logger::try_init();

        let counter = ATOMIC_COUNTER.fetch_add(1, Ordering::Relaxed);
        PlacesApi::new_memory(&format!("test-api-{}", counter)).expect("should get an API")
    }

    pub fn new_mem_connection() -> PlacesDb {
        new_mem_api()
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get a connection")
    }

    pub struct MemConnections {
        pub read: PlacesDb,
        pub write: PlacesDb,
        pub api: Arc<PlacesApi>,
    }

    pub fn new_mem_connections() -> MemConnections {
        let api = new_mem_api();
        let read = api
            .open_connection(ConnectionType::ReadOnly)
            .expect("should get a read connection");
        let write = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get a write connection");
        MemConnections { read, write, api }
    }
}

#[cfg(test)]
mod tests {
    use super::test::*;
    use super::*;
    use sql_support::ConnExt;

    #[test]
    fn test_multi_writers_fails() {
        let api = new_mem_api();
        let writer1 = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer");
        api.open_connection(ConnectionType::ReadWrite)
            .expect_err("should fail to get second writer");
        // But we should be able to re-get it after closing it.
        api.close_connection(writer1)
            .expect("should be able to close");
        api.open_connection(ConnectionType::ReadWrite)
            .expect("should get a writer after closing the other");
    }

    #[test]
    fn test_shared_memory() {
        let api = new_mem_api();
        let writer = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer");
        writer
            .execute_batch(
                "CREATE TABLE test_table (test_value INTEGER);
                              INSERT INTO test_table VALUES (999)",
            )
            .expect("should insert");
        let reader = api
            .open_connection(ConnectionType::ReadOnly)
            .expect("should get reader");
        let val = reader
            .query_one::<i64>("SELECT test_value FROM test_table")
            .expect("should get value");
        assert_eq!(val, 999);
    }

    #[test]
    fn test_reader_before_writer() {
        let api = new_mem_api();
        let reader = api
            .open_connection(ConnectionType::ReadOnly)
            .expect("should get reader");
        let writer = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer");
        writer
            .execute_batch(
                "CREATE TABLE test_table (test_value INTEGER);
                              INSERT INTO test_table VALUES (999)",
            )
            .expect("should insert");
        let val = reader
            .query_one::<i64>("SELECT test_value FROM test_table")
            .expect("should get value");
        assert_eq!(val, 999);
    }

    #[test]
    fn test_wrong_writer_close() {
        let api = new_mem_api();
        // Grab this so `api` doesn't think it still has a writer.
        let _writer = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer");

        let fake_api = new_mem_api();
        let fake_writer = fake_api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer 2");

        assert!(matches!(
            api.close_connection(fake_writer).unwrap_err(),
            Error::WrongApiForClose
        ));
    }

    #[test]
    fn test_valid_writer_close() {
        let api = new_mem_api();
        let writer = api
            .open_connection(ConnectionType::ReadWrite)
            .expect("should get writer");

        api.close_connection(writer)
            .expect("Should allow closing own connection");

        // Make sure we can open it again.
        assert!(api.open_connection(ConnectionType::ReadWrite).is_ok());
    }
}