tor_dirmgr/
lib.rs

1#![cfg_attr(docsrs, feature(doc_auto_cfg, doc_cfg))]
2#![doc = include_str!("../README.md")]
3// @@ begin lint list maintained by maint/add_warning @@
4#![allow(renamed_and_removed_lints)] // @@REMOVE_WHEN(ci_arti_stable)
5#![allow(unknown_lints)] // @@REMOVE_WHEN(ci_arti_nightly)
6#![warn(missing_docs)]
7#![warn(noop_method_call)]
8#![warn(unreachable_pub)]
9#![warn(clippy::all)]
10#![deny(clippy::await_holding_lock)]
11#![deny(clippy::cargo_common_metadata)]
12#![deny(clippy::cast_lossless)]
13#![deny(clippy::checked_conversions)]
14#![warn(clippy::cognitive_complexity)]
15#![deny(clippy::debug_assert_with_mut_call)]
16#![deny(clippy::exhaustive_enums)]
17#![deny(clippy::exhaustive_structs)]
18#![deny(clippy::expl_impl_clone_on_copy)]
19#![deny(clippy::fallible_impl_from)]
20#![deny(clippy::implicit_clone)]
21#![deny(clippy::large_stack_arrays)]
22#![warn(clippy::manual_ok_or)]
23#![deny(clippy::missing_docs_in_private_items)]
24#![warn(clippy::needless_borrow)]
25#![warn(clippy::needless_pass_by_value)]
26#![warn(clippy::option_option)]
27#![deny(clippy::print_stderr)]
28#![deny(clippy::print_stdout)]
29#![warn(clippy::rc_buffer)]
30#![deny(clippy::ref_option_ref)]
31#![warn(clippy::semicolon_if_nothing_returned)]
32#![warn(clippy::trait_duplication_in_bounds)]
33#![deny(clippy::unchecked_duration_subtraction)]
34#![deny(clippy::unnecessary_wraps)]
35#![warn(clippy::unseparated_literal_suffix)]
36#![deny(clippy::unwrap_used)]
37#![deny(clippy::mod_module_files)]
38#![allow(clippy::let_unit_value)] // This can reasonably be done for explicitness
39#![allow(clippy::uninlined_format_args)]
40#![allow(clippy::significant_drop_in_scrutinee)] // arti/-/merge_requests/588/#note_2812945
41#![allow(clippy::result_large_err)] // temporary workaround for arti#587
42#![allow(clippy::needless_raw_string_hashes)] // complained-about code is fine, often best
43#![allow(clippy::needless_lifetimes)] // See arti#1765
44#![allow(mismatched_lifetime_syntaxes)] // temporary workaround for arti#2060
45//! <!-- @@ end lint list maintained by maint/add_warning @@ -->
46
47// This clippy lint produces a false positive on `use strum`, below.
48// Attempting to apply the lint to just the use statement fails to suppress
49// this lint and instead produces another lint about a useless clippy attribute.
50#![allow(clippy::single_component_path_imports)]
51
52pub mod authority;
53mod bootstrap;
54pub mod config;
55mod docid;
56mod docmeta;
57mod err;
58mod event;
59mod retry;
60mod shared_ref;
61mod state;
62mod storage;
63
64#[cfg(feature = "bridge-client")]
65pub mod bridgedesc;
66#[cfg(feature = "dirfilter")]
67pub mod filter;
68
69use crate::docid::{CacheUsage, ClientRequest, DocQuery};
70use crate::err::BootstrapAction;
71#[cfg(not(feature = "experimental-api"))]
72use crate::shared_ref::SharedMutArc;
73#[cfg(feature = "experimental-api")]
74pub use crate::shared_ref::SharedMutArc;
75use crate::storage::{DynStore, Store};
76use bootstrap::AttemptId;
77use event::DirProgress;
78use postage::watch;
79pub use retry::{DownloadSchedule, DownloadScheduleBuilder};
80use scopeguard::ScopeGuard;
81use tor_circmgr::CircMgr;
82use tor_dirclient::SourceInfo;
83use tor_error::{info_report, into_internal, warn_report};
84use tor_netdir::params::NetParameters;
85use tor_netdir::{DirEvent, MdReceiver, NetDir, NetDirProvider};
86
87use async_trait::async_trait;
88use futures::{stream::BoxStream, task::SpawnExt};
89use oneshot_fused_workaround as oneshot;
90use tor_netdoc::doc::netstatus::ProtoStatuses;
91use tor_rtcompat::scheduler::{TaskHandle, TaskSchedule};
92use tor_rtcompat::Runtime;
93use tracing::{debug, info, trace, warn};
94
95use std::marker::PhantomData;
96use std::sync::atomic::{AtomicBool, Ordering};
97use std::sync::{Arc, Mutex};
98use std::time::Duration;
99use std::{collections::HashMap, sync::Weak};
100use std::{fmt::Debug, time::SystemTime};
101
102use crate::state::{DirState, NetDirChange};
103pub use authority::{Authority, AuthorityBuilder};
104pub use config::{
105    DirMgrConfig, DirTolerance, DirToleranceBuilder, DownloadScheduleConfig,
106    DownloadScheduleConfigBuilder, NetworkConfig, NetworkConfigBuilder,
107};
108pub use docid::DocId;
109pub use err::Error;
110pub use event::{DirBlockage, DirBootstrapEvents, DirBootstrapStatus};
111pub use storage::DocumentText;
112pub use tor_guardmgr::fallback::{FallbackDir, FallbackDirBuilder};
113pub use tor_netdir::Timeliness;
114
115/// Re-export of `strum` crate for use by an internal macro
116use strum;
117
118/// A Result as returned by this crate.
119pub type Result<T> = std::result::Result<T, Error>;
120
121/// Storage manager used by [`DirMgr`] and
122/// [`BridgeDescMgr`](bridgedesc::BridgeDescMgr)
123///
124/// Internally, this wraps up a sqlite database.
125///
126/// This is a handle, which is cheap to clone; clones share state.
127#[derive(Clone)]
128pub struct DirMgrStore<R: Runtime> {
129    /// The actual store
130    pub(crate) store: Arc<Mutex<crate::DynStore>>,
131
132    /// Be parameterized by Runtime even though we don't use it right now
133    pub(crate) runtime: PhantomData<R>,
134}
135
136impl<R: Runtime> DirMgrStore<R> {
137    /// Open the storage, according to the specified configuration
138    pub fn new(config: &DirMgrConfig, runtime: R, offline: bool) -> Result<Self> {
139        let store = Arc::new(Mutex::new(config.open_store(offline)?));
140        drop(runtime);
141        let runtime = PhantomData;
142        Ok(DirMgrStore { store, runtime })
143    }
144}
145
146/// Trait for DirMgr implementations
147#[async_trait]
148pub trait DirProvider: NetDirProvider {
149    /// Try to change our configuration to `new_config`.
150    ///
151    /// Actual behavior will depend on the value of `how`.
152    fn reconfigure(
153        &self,
154        new_config: &DirMgrConfig,
155        how: tor_config::Reconfigure,
156    ) -> std::result::Result<(), tor_config::ReconfigureError>;
157
158    /// Bootstrap a `DirProvider` that hasn't been bootstrapped yet.
159    async fn bootstrap(&self) -> Result<()>;
160
161    /// Return a stream of [`DirBootstrapStatus`] events to tell us about changes
162    /// in the latest directory's bootstrap status.
163    ///
164    /// Note that this stream can be lossy: the caller will not necessarily
165    /// observe every event on the stream
166    fn bootstrap_events(&self) -> BoxStream<'static, DirBootstrapStatus>;
167
168    /// Return a [`TaskHandle`] that can be used to manage the download process.
169    fn download_task_handle(&self) -> Option<TaskHandle> {
170        None
171    }
172}
173
174// NOTE(eta): We can't implement this for Arc<DirMgr<R>> due to trait coherence rules, so instead
175//            there's a blanket impl for Arc<T> in tor-netdir.
176impl<R: Runtime> NetDirProvider for DirMgr<R> {
177    fn netdir(&self, timeliness: Timeliness) -> tor_netdir::Result<Arc<NetDir>> {
178        use tor_netdir::Error as NetDirError;
179        let netdir = self.netdir.get().ok_or(NetDirError::NoInfo)?;
180        let lifetime = match timeliness {
181            Timeliness::Strict => netdir.lifetime().clone(),
182            Timeliness::Timely => self
183                .config
184                .get()
185                .tolerance
186                .extend_lifetime(netdir.lifetime()),
187            Timeliness::Unchecked => return Ok(netdir),
188        };
189        let now = SystemTime::now();
190        if lifetime.valid_after() > now {
191            Err(NetDirError::DirNotYetValid)
192        } else if lifetime.valid_until() < now {
193            Err(NetDirError::DirExpired)
194        } else {
195            Ok(netdir)
196        }
197    }
198
199    fn events(&self) -> BoxStream<'static, DirEvent> {
200        Box::pin(self.events.subscribe())
201    }
202
203    fn params(&self) -> Arc<dyn AsRef<tor_netdir::params::NetParameters>> {
204        if let Some(netdir) = self.netdir.get() {
205            // We have a directory, so we'd like to give it out for its
206            // parameters.
207            //
208            // We do this even if the directory is expired, since parameters
209            // don't really expire on any plausible timescale.
210            netdir
211        } else {
212            // We have no directory, so we'll give out the default parameters as
213            // modified by the provided override_net_params configuration.
214            //
215            self.default_parameters
216                .lock()
217                .expect("Poisoned lock")
218                .clone()
219        }
220        // TODO(nickm): If we felt extremely clever, we could add a third case
221        // where, if we have a pending directory with a validated consensus, we
222        // give out that consensus's network parameters even if we _don't_ yet
223        // have a full directory.  That's significant refactoring, though, for
224        // an unclear amount of benefit.
225    }
226
227    fn protocol_statuses(&self) -> Option<(SystemTime, Arc<ProtoStatuses>)> {
228        self.protocols.lock().expect("Poisoned lock").clone()
229    }
230}
231
232#[async_trait]
233impl<R: Runtime> DirProvider for Arc<DirMgr<R>> {
234    fn reconfigure(
235        &self,
236        new_config: &DirMgrConfig,
237        how: tor_config::Reconfigure,
238    ) -> std::result::Result<(), tor_config::ReconfigureError> {
239        DirMgr::reconfigure(self, new_config, how)
240    }
241
242    async fn bootstrap(&self) -> Result<()> {
243        DirMgr::bootstrap(self).await
244    }
245
246    fn bootstrap_events(&self) -> BoxStream<'static, DirBootstrapStatus> {
247        Box::pin(DirMgr::bootstrap_events(self))
248    }
249
250    fn download_task_handle(&self) -> Option<TaskHandle> {
251        Some(self.task_handle.clone())
252    }
253}
254
255/// A directory manager to download, fetch, and cache a Tor directory.
256///
257/// A DirMgr can operate in three modes:
258///   * In **offline** mode, it only reads from the cache, and can
259///     only read once.
260///   * In **read-only** mode, it reads from the cache, but checks
261///     whether it can acquire an associated lock file.  If it can, then
262///     it enters read-write mode.  If not, it checks the cache
263///     periodically for new information.
264///   * In **read-write** mode, it knows that no other process will be
265///     writing to the cache, and it takes responsibility for fetching
266///     data from the network and updating the directory with new
267///     directory information.
268pub struct DirMgr<R: Runtime> {
269    /// Configuration information: where to find directories, how to
270    /// validate them, and so on.
271    config: tor_config::MutCfg<DirMgrConfig>,
272    /// Handle to our sqlite cache.
273    // TODO(nickm): I'd like to use an rwlock, but that's not feasible, since
274    // rusqlite::Connection isn't Sync.
275    // TODO is needed?
276    store: Arc<Mutex<DynStore>>,
277    /// Our latest sufficiently bootstrapped directory, if we have one.
278    ///
279    /// We use the RwLock so that we can give this out to a bunch of other
280    /// users, and replace it once a new directory is bootstrapped.
281    // TODO(eta): Eurgh! This is so many Arcs! (especially considering this
282    //            gets wrapped in an Arc)
283    netdir: Arc<SharedMutArc<NetDir>>,
284
285    /// Our latest set of recommended protocols.
286    protocols: Mutex<Option<(SystemTime, Arc<ProtoStatuses>)>>,
287
288    /// A set of network parameters to hand out when we have no directory.
289    default_parameters: Mutex<Arc<NetParameters>>,
290
291    /// A publisher handle that we notify whenever the consensus changes.
292    events: event::FlagPublisher<DirEvent>,
293
294    /// A publisher handle that we notify whenever our bootstrapping status
295    /// changes.
296    send_status: Mutex<watch::Sender<event::DirBootstrapStatus>>,
297
298    /// A receiver handle that gets notified whenever our bootstrapping status
299    /// changes.
300    ///
301    /// We don't need to keep this drained, since `postage::watch` already knows
302    /// to discard unread events.
303    receive_status: DirBootstrapEvents,
304
305    /// A circuit manager, if this DirMgr supports downloading.
306    circmgr: Option<Arc<CircMgr<R>>>,
307
308    /// Our asynchronous runtime.
309    runtime: R,
310
311    /// Whether or not we're operating in offline mode.
312    offline: bool,
313
314    /// If we're not in offline mode, stores whether or not the `DirMgr` has attempted
315    /// to bootstrap yet or not.
316    ///
317    /// This exists in order to prevent starting two concurrent bootstrap tasks.
318    ///
319    /// (In offline mode, this does nothing.)
320    bootstrap_started: AtomicBool,
321
322    /// A filter that gets applied to directory objects before we use them.
323    #[cfg(feature = "dirfilter")]
324    filter: crate::filter::FilterConfig,
325
326    /// A task schedule that can be used if we're bootstrapping.  If this is
327    /// None, then there's currently a scheduled task in progress.
328    task_schedule: Mutex<Option<TaskSchedule<R>>>,
329
330    /// A task handle that we return to anybody who needs to manage our download process.
331    task_handle: TaskHandle,
332}
333
334/// The possible origins of a document.
335///
336/// Used (for example) to report where we got a document from if it fails to
337/// parse.
338#[derive(Debug, Clone)]
339#[non_exhaustive]
340pub enum DocSource {
341    /// We loaded the document from our cache.
342    LocalCache,
343    /// We fetched the document from a server.
344    DirServer {
345        /// Information about the server we fetched the document from.
346        source: Option<SourceInfo>,
347    },
348}
349
350impl std::fmt::Display for DocSource {
351    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
352        match self {
353            DocSource::LocalCache => write!(f, "local cache"),
354            DocSource::DirServer { source: None } => write!(f, "directory server"),
355            DocSource::DirServer { source: Some(info) } => write!(f, "directory server {}", info),
356        }
357    }
358}
359
360impl<R: Runtime> DirMgr<R> {
361    /// Try to load the directory from disk, without launching any
362    /// kind of update process.
363    ///
364    /// This function runs in **offline** mode: it will give an error
365    /// if the result is not up-to-date, or not fully downloaded.
366    ///
367    /// In general, you shouldn't use this function in a long-running
368    /// program; it's only suitable for command-line or batch tools.
369    // TODO: I wish this function didn't have to be async or take a runtime.
370    pub async fn load_once(runtime: R, config: DirMgrConfig) -> Result<Arc<NetDir>> {
371        let store = DirMgrStore::new(&config, runtime.clone(), true)?;
372        let dirmgr = Arc::new(Self::from_config(config, runtime, store, None, true)?);
373
374        // TODO: add some way to return a directory that isn't up-to-date
375        let attempt = AttemptId::next();
376        trace!(%attempt, "Trying to load a full directory from cache");
377        let outcome = dirmgr.load_directory(attempt).await;
378        trace!(%attempt, "Load result: {outcome:?}");
379        let _success = outcome?;
380
381        dirmgr
382            .netdir(Timeliness::Timely)
383            .map_err(|_| Error::DirectoryNotPresent)
384    }
385
386    /// Return a current netdir, either loading it or bootstrapping it
387    /// as needed.
388    ///
389    /// Like load_once, but will try to bootstrap (or wait for another
390    /// process to bootstrap) if we don't have an up-to-date
391    /// bootstrapped directory.
392    ///
393    /// In general, you shouldn't use this function in a long-running
394    /// program; it's only suitable for command-line or batch tools.
395    pub async fn load_or_bootstrap_once(
396        config: DirMgrConfig,
397        runtime: R,
398        store: DirMgrStore<R>,
399        circmgr: Arc<CircMgr<R>>,
400    ) -> Result<Arc<NetDir>> {
401        let dirmgr = DirMgr::bootstrap_from_config(config, runtime, store, circmgr).await?;
402        dirmgr
403            .timely_netdir()
404            .map_err(|_| Error::DirectoryNotPresent)
405    }
406
407    /// Create a new `DirMgr` in online mode, but don't bootstrap it yet.
408    ///
409    /// The `DirMgr` can be bootstrapped later with `bootstrap`.
410    pub fn create_unbootstrapped(
411        config: DirMgrConfig,
412        runtime: R,
413        store: DirMgrStore<R>,
414        circmgr: Arc<CircMgr<R>>,
415    ) -> Result<Arc<Self>> {
416        Ok(Arc::new(DirMgr::from_config(
417            config,
418            runtime,
419            store,
420            Some(circmgr),
421            false,
422        )?))
423    }
424
425    /// Bootstrap a `DirMgr` created in online mode that hasn't been bootstrapped yet.
426    ///
427    /// This function will not return until the directory is bootstrapped enough to build circuits.
428    /// It will also launch a background task that fetches any missing information, and that
429    /// replaces the directory when a new one is available.
430    ///
431    /// This function is intended to be used together with `create_unbootstrapped`. There is no
432    /// need to call this function otherwise.
433    ///
434    /// If bootstrapping has already successfully taken place, returns early with success.
435    ///
436    /// # Errors
437    ///
438    /// Returns an error if bootstrapping fails. If the error is [`Error::CantAdvanceState`],
439    /// it may be possible to successfully bootstrap later on by calling this function again.
440    ///
441    /// # Panics
442    ///
443    /// Panics if the `DirMgr` passed to this function was not created in online mode, such as
444    /// via `load_once`.
445    #[allow(clippy::cognitive_complexity)] // TODO: Refactor
446    pub async fn bootstrap(self: &Arc<Self>) -> Result<()> {
447        if self.offline {
448            return Err(Error::OfflineMode);
449        }
450
451        // The semantics of this are "attempt to replace a 'false' value with 'true'.
452        // If the value in bootstrap_started was not 'false' when the attempt was made, returns
453        // `Err`; this means another bootstrap attempt is in progress or has completed, so we
454        // return early.
455
456        // NOTE(eta): could potentially weaken the `Ordering` here in future
457        if self
458            .bootstrap_started
459            .compare_exchange(false, true, Ordering::SeqCst, Ordering::SeqCst)
460            .is_err()
461        {
462            debug!("Attempted to bootstrap twice; ignoring.");
463            return Ok(());
464        }
465
466        // Use a RAII guard to reset `bootstrap_started` to `false` if we return early without
467        // completing bootstrap.
468        let reset_bootstrap_started = scopeguard::guard(&self.bootstrap_started, |v| {
469            v.store(false, Ordering::SeqCst);
470        });
471
472        let schedule = {
473            let sched = self.task_schedule.lock().expect("poisoned lock").take();
474            match sched {
475                Some(sched) => sched,
476                None => {
477                    debug!("Attempted to bootstrap twice; ignoring.");
478                    return Ok(());
479                }
480            }
481        };
482
483        // Try to load from the cache.
484        let attempt_id = AttemptId::next();
485        trace!(attempt=%attempt_id, "Starting to bootstrap directory");
486        let have_directory = self.load_directory(attempt_id).await?;
487
488        let (mut sender, receiver) = if have_directory {
489            info!("Loaded a good directory from cache.");
490            (None, None)
491        } else {
492            info!("Didn't get usable directory from cache.");
493            let (sender, receiver) = oneshot::channel();
494            (Some(sender), Some(receiver))
495        };
496
497        // Whether we loaded or not, we now start downloading.
498        let dirmgr_weak = Arc::downgrade(self);
499        self.runtime
500            .spawn(async move {
501                // Use an RAII guard to make sure that when this task exits, the
502                // TaskSchedule object is put back.
503                //
504                // TODO(nick): Putting the schedule back isn't actually useful
505                // if the task exits _after_ we've bootstrapped for the first
506                // time, because of how bootstrap_started works.
507                let mut schedule = scopeguard::guard(schedule, |schedule| {
508                    if let Some(dm) = Weak::upgrade(&dirmgr_weak) {
509                        *dm.task_schedule.lock().expect("poisoned lock") = Some(schedule);
510                    }
511                });
512
513                // Don't warn when these are Error::ManagerDropped: that
514                // means that the DirMgr has been shut down.
515                if let Err(e) =
516                    Self::reload_until_owner(&dirmgr_weak, &mut schedule, attempt_id, &mut sender)
517                        .await
518                {
519                    match e {
520                        Error::ManagerDropped => {}
521                        _ => warn_report!(e, "Unrecovered error while waiting for bootstrap",),
522                    }
523                } else if let Err(e) =
524                    Self::download_forever(dirmgr_weak.clone(), &mut schedule, attempt_id, sender)
525                        .await
526                {
527                    match e {
528                        Error::ManagerDropped => {}
529                        _ => warn_report!(e, "Unrecovered error while downloading"),
530                    }
531                }
532            })
533            .map_err(|e| Error::from_spawn("directory updater task", e))?;
534
535        if let Some(receiver) = receiver {
536            match receiver.await {
537                Ok(()) => {
538                    info!("We have enough information to build circuits.");
539                    // Disarm the RAII guard, since we succeeded.  Now bootstrap_started will remain true.
540                    let _ = ScopeGuard::into_inner(reset_bootstrap_started);
541                }
542                Err(_) => {
543                    warn!("Bootstrapping task exited before finishing.");
544                    return Err(Error::CantAdvanceState);
545                }
546            }
547        }
548        Ok(())
549    }
550
551    /// Returns `true` if a bootstrap attempt is in progress, or successfully completed.
552    pub fn bootstrap_started(&self) -> bool {
553        self.bootstrap_started.load(Ordering::SeqCst)
554    }
555
556    /// Return a new directory manager from a given configuration,
557    /// bootstrapping from the network as necessary.
558    pub async fn bootstrap_from_config(
559        config: DirMgrConfig,
560        runtime: R,
561        store: DirMgrStore<R>,
562        circmgr: Arc<CircMgr<R>>,
563    ) -> Result<Arc<Self>> {
564        let dirmgr = Self::create_unbootstrapped(config, runtime, store, circmgr)?;
565
566        dirmgr.bootstrap().await?;
567
568        Ok(dirmgr)
569    }
570
571    /// Try forever to either lock the storage (and thereby become the
572    /// owner), or to reload the database.
573    ///
574    /// If we have begin to have a bootstrapped directory, send a
575    /// message using `on_complete`.
576    ///
577    /// If we eventually become the owner, return Ok().
578    #[allow(clippy::cognitive_complexity)] // TODO: Refactor?
579    async fn reload_until_owner(
580        weak: &Weak<Self>,
581        schedule: &mut TaskSchedule<R>,
582        attempt_id: AttemptId,
583        on_complete: &mut Option<oneshot::Sender<()>>,
584    ) -> Result<()> {
585        let mut logged = false;
586        let mut bootstrapped;
587        {
588            let dirmgr = upgrade_weak_ref(weak)?;
589            bootstrapped = dirmgr.netdir.get().is_some();
590        }
591
592        loop {
593            {
594                let dirmgr = upgrade_weak_ref(weak)?;
595                trace!("Trying to take ownership of the directory cache lock");
596                if dirmgr.try_upgrade_to_readwrite()? {
597                    // We now own the lock!  (Maybe we owned it before; the
598                    // upgrade_to_readwrite() function is idempotent.)  We can
599                    // do our own bootstrapping.
600                    if logged {
601                        info!("The previous owning process has given up the lock. We are now in charge of managing the directory.");
602                    }
603                    return Ok(());
604                }
605            }
606
607            if !logged {
608                logged = true;
609                if bootstrapped {
610                    info!("Another process is managing the directory. We'll use its cache.");
611                } else {
612                    info!("Another process is bootstrapping the directory. Waiting till it finishes or exits.");
613                }
614            }
615
616            // We don't own the lock.  Somebody else owns the cache.  They
617            // should be updating it.  Wait a bit, then try again.
618            let pause = if bootstrapped {
619                std::time::Duration::new(120, 0)
620            } else {
621                std::time::Duration::new(5, 0)
622            };
623            schedule.sleep(pause).await?;
624            // TODO: instead of loading the whole thing we should have a
625            // database entry that says when the last update was, or use
626            // our state functions.
627            {
628                let dirmgr = upgrade_weak_ref(weak)?;
629                trace!("Trying to load from the directory cache");
630                if dirmgr.load_directory(attempt_id).await? {
631                    // Successfully loaded a bootstrapped directory.
632                    if let Some(send_done) = on_complete.take() {
633                        let _ = send_done.send(());
634                    }
635                    if !bootstrapped {
636                        info!("The directory is now bootstrapped.");
637                    }
638                    bootstrapped = true;
639                }
640            }
641        }
642    }
643
644    /// Try to fetch our directory info and keep it updated, indefinitely.
645    ///
646    /// If we have begin to have a bootstrapped directory, send a
647    /// message using `on_complete`.
648    #[allow(clippy::cognitive_complexity)] // TODO: Refactor?
649    async fn download_forever(
650        weak: Weak<Self>,
651        schedule: &mut TaskSchedule<R>,
652        mut attempt_id: AttemptId,
653        mut on_complete: Option<oneshot::Sender<()>>,
654    ) -> Result<()> {
655        let mut state: Box<dyn DirState> = {
656            let dirmgr = upgrade_weak_ref(&weak)?;
657            Box::new(state::GetConsensusState::new(
658                dirmgr.runtime.clone(),
659                dirmgr.config.get(),
660                CacheUsage::CacheOkay,
661                Some(dirmgr.netdir.clone()),
662                #[cfg(feature = "dirfilter")]
663                dirmgr
664                    .filter
665                    .clone()
666                    .unwrap_or_else(|| Arc::new(crate::filter::NilFilter)),
667            ))
668        };
669
670        trace!("Entering download loop.");
671
672        loop {
673            let mut usable = false;
674
675            let retry_config = {
676                let dirmgr = upgrade_weak_ref(&weak)?;
677                // TODO(nickm): instead of getting this every time we loop, it
678                // might be a good idea to refresh it with each attempt, at
679                // least at the point of checking the number of attempts.
680                dirmgr.config.get().schedule.retry_bootstrap
681            };
682            let mut retry_delay = retry_config.schedule();
683
684            'retry_attempt: for try_num in retry_config.attempts() {
685                trace!(attempt=%attempt_id, ?try_num, "Trying to download a directory.");
686                let outcome = bootstrap::download(
687                    Weak::clone(&weak),
688                    &mut state,
689                    schedule,
690                    attempt_id,
691                    &mut on_complete,
692                )
693                .await;
694                trace!(attempt=%attempt_id, ?try_num, ?outcome, "Download is over.");
695
696                if let Err(err) = outcome {
697                    if state.is_ready(Readiness::Usable) {
698                        usable = true;
699                        info_report!(err, "Unable to completely download a directory. (Nevertheless, the directory is usable, so we'll pause for now)");
700                        break 'retry_attempt;
701                    }
702
703                    match err.bootstrap_action() {
704                        BootstrapAction::Nonfatal => {
705                            return Err(into_internal!(
706                                "Nonfatal error should not have propagated here"
707                            )(err)
708                            .into());
709                        }
710                        BootstrapAction::Reset => {}
711                        BootstrapAction::Fatal => return Err(err),
712                    }
713
714                    let delay = retry_delay.next_delay(&mut rand::rng());
715                    warn_report!(
716                        err,
717                        "Unable to download a usable directory. (We will restart in {})",
718                        humantime::format_duration(delay),
719                    );
720                    {
721                        let dirmgr = upgrade_weak_ref(&weak)?;
722                        dirmgr.note_reset(attempt_id);
723                    }
724                    schedule.sleep(delay).await?;
725                    state = state.reset();
726                } else {
727                    info!(attempt=%attempt_id, "Directory is complete.");
728                    usable = true;
729                    break 'retry_attempt;
730                }
731            }
732
733            if !usable {
734                // we ran out of attempts.
735                warn!(
736                    "We failed {} times to bootstrap a directory. We're going to give up.",
737                    retry_config.n_attempts()
738                );
739                return Err(Error::CantAdvanceState);
740            } else {
741                // Report success, if appropriate.
742                if let Some(send_done) = on_complete.take() {
743                    let _ = send_done.send(());
744                }
745            }
746
747            let reset_at = state.reset_time();
748            match reset_at {
749                Some(t) => {
750                    trace!("Sleeping until {}", time::OffsetDateTime::from(t));
751                    schedule.sleep_until_wallclock(t).await?;
752                }
753                None => return Ok(()),
754            }
755            attempt_id = bootstrap::AttemptId::next();
756            trace!(attempt=%attempt_id, "Beginning new attempt to bootstrap directory");
757            state = state.reset();
758        }
759    }
760
761    /// Get a reference to the circuit manager, if we have one.
762    fn circmgr(&self) -> Result<Arc<CircMgr<R>>> {
763        self.circmgr.clone().ok_or(Error::NoDownloadSupport)
764    }
765
766    /// Try to change our configuration to `new_config`.
767    ///
768    /// Actual behavior will depend on the value of `how`.
769    pub fn reconfigure(
770        &self,
771        new_config: &DirMgrConfig,
772        how: tor_config::Reconfigure,
773    ) -> std::result::Result<(), tor_config::ReconfigureError> {
774        let config = self.config.get();
775        // We don't support changing these: doing so basically would require us
776        // to abort all our in-progress downloads, since they might be based on
777        // no-longer-viable information.
778        // NOTE: keep this in sync with the behaviour of `DirMgrConfig::update_from_config`
779        if new_config.cache_dir != config.cache_dir {
780            how.cannot_change("storage.cache_dir")?;
781        }
782        if new_config.cache_trust != config.cache_trust {
783            how.cannot_change("storage.permissions")?;
784        }
785        if new_config.authorities() != config.authorities() {
786            how.cannot_change("network.authorities")?;
787        }
788
789        if how == tor_config::Reconfigure::CheckAllOrNothing {
790            return Ok(());
791        }
792
793        let params_changed = new_config.override_net_params != config.override_net_params;
794
795        self.config
796            .map_and_replace(|cfg| cfg.update_from_config(new_config));
797
798        if params_changed {
799            let _ignore_err = self.netdir.mutate(|netdir| {
800                netdir.replace_overridden_parameters(&new_config.override_net_params);
801                Ok(())
802            });
803            {
804                let mut params = self.default_parameters.lock().expect("lock failed");
805                *params = Arc::new(NetParameters::from_map(&new_config.override_net_params));
806            }
807
808            // (It's okay to ignore the error, since it just means that there
809            // was no current netdir.)
810            self.events.publish(DirEvent::NewConsensus);
811        }
812
813        Ok(())
814    }
815
816    /// Return a stream of [`DirBootstrapStatus`] events to tell us about changes
817    /// in the latest directory's bootstrap status.
818    ///
819    /// Note that this stream can be lossy: the caller will not necessarily
820    /// observe every event on the stream
821    pub fn bootstrap_events(&self) -> event::DirBootstrapEvents {
822        self.receive_status.clone()
823    }
824
825    /// Replace the latest status with `progress` and broadcast to anybody
826    /// watching via a [`DirBootstrapEvents`] stream.
827    fn update_progress(&self, attempt_id: AttemptId, progress: DirProgress) {
828        // TODO(nickm): can I kill off this lock by having something else own the sender?
829        let mut sender = self.send_status.lock().expect("poisoned lock");
830        let mut status = sender.borrow_mut();
831
832        status.update_progress(attempt_id, progress);
833    }
834
835    /// Update our status tracker to note that some number of errors has
836    /// occurred.
837    fn note_errors(&self, attempt_id: AttemptId, n_errors: usize) {
838        if n_errors == 0 {
839            return;
840        }
841        let mut sender = self.send_status.lock().expect("poisoned lock");
842        let mut status = sender.borrow_mut();
843
844        status.note_errors(attempt_id, n_errors);
845    }
846
847    /// Update our status tracker to note that we've needed to reset our download attempt.
848    fn note_reset(&self, attempt_id: AttemptId) {
849        let mut sender = self.send_status.lock().expect("poisoned lock");
850        let mut status = sender.borrow_mut();
851
852        status.note_reset(attempt_id);
853    }
854
855    /// Try to make this a directory manager with read-write access to its
856    /// storage.
857    ///
858    /// Return true if we got the lock, or if we already had it.
859    ///
860    /// Return false if another process has the lock
861    fn try_upgrade_to_readwrite(&self) -> Result<bool> {
862        self.store
863            .lock()
864            .expect("Directory storage lock poisoned")
865            .upgrade_to_readwrite()
866    }
867
868    /// Return a reference to the store, if it is currently read-write.
869    #[cfg(test)]
870    fn store_if_rw(&self) -> Option<&Mutex<DynStore>> {
871        let rw = !self
872            .store
873            .lock()
874            .expect("Directory storage lock poisoned")
875            .is_readonly();
876        // A race-condition is possible here, but I believe it's harmless.
877        if rw {
878            Some(&self.store)
879        } else {
880            None
881        }
882    }
883
884    /// Construct a DirMgr from a DirMgrConfig.
885    ///
886    /// If `offline` is set, opens the SQLite store read-only and sets the offline flag in the
887    /// returned manager.
888    #[allow(clippy::unnecessary_wraps)] // API compat and future-proofing
889    fn from_config(
890        config: DirMgrConfig,
891        runtime: R,
892        store: DirMgrStore<R>,
893        circmgr: Option<Arc<CircMgr<R>>>,
894        offline: bool,
895    ) -> Result<Self> {
896        let netdir = Arc::new(SharedMutArc::new());
897        let events = event::FlagPublisher::new();
898        let default_parameters = NetParameters::from_map(&config.override_net_params);
899        let default_parameters = Mutex::new(Arc::new(default_parameters));
900
901        let (send_status, receive_status) = postage::watch::channel();
902        let send_status = Mutex::new(send_status);
903        let receive_status = DirBootstrapEvents {
904            inner: receive_status,
905        };
906        #[cfg(feature = "dirfilter")]
907        let filter = config.extensions.filter.clone();
908
909        // We create these early so the client code can access task_handle before bootstrap() returns.
910        let (task_schedule, task_handle) = TaskSchedule::new(runtime.clone());
911        let task_schedule = Mutex::new(Some(task_schedule));
912
913        // We load the cached protocol recommendations unconditionally: the caller needs them even
914        // if it does not try to load the reset of the cache.
915        let protocols = {
916            let store = store.store.lock().expect("lock poisoned");
917            store
918                .cached_protocol_recommendations()?
919                .map(|(t, p)| (t, Arc::new(p)))
920        };
921
922        Ok(DirMgr {
923            config: config.into(),
924            store: store.store,
925            netdir,
926            protocols: Mutex::new(protocols),
927            default_parameters,
928            events,
929            send_status,
930            receive_status,
931            circmgr,
932            runtime,
933            offline,
934            bootstrap_started: AtomicBool::new(false),
935            #[cfg(feature = "dirfilter")]
936            filter,
937            task_schedule,
938            task_handle,
939        })
940    }
941
942    /// Load the latest non-pending non-expired directory from the
943    /// cache, if it is newer than the one we have.
944    ///
945    /// Return false if there is no such consensus.
946    async fn load_directory(self: &Arc<Self>, attempt_id: AttemptId) -> Result<bool> {
947        let state = state::GetConsensusState::new(
948            self.runtime.clone(),
949            self.config.get(),
950            CacheUsage::CacheOnly,
951            None,
952            #[cfg(feature = "dirfilter")]
953            self.filter
954                .clone()
955                .unwrap_or_else(|| Arc::new(crate::filter::NilFilter)),
956        );
957        let _ = bootstrap::load(Arc::clone(self), Box::new(state), attempt_id).await?;
958
959        Ok(self.netdir.get().is_some())
960    }
961
962    /// Return a new asynchronous stream that will receive notification
963    /// whenever the consensus has changed.
964    ///
965    /// Multiple events may be batched up into a single item: each time
966    /// this stream yields an event, all you can assume is that the event has
967    /// occurred at least once.
968    pub fn events(&self) -> impl futures::Stream<Item = DirEvent> {
969        self.events.subscribe()
970    }
971
972    /// Try to load the text of a single document described by `doc` from
973    /// storage.
974    pub fn text(&self, doc: &DocId) -> Result<Option<DocumentText>> {
975        use itertools::Itertools;
976        let mut result = HashMap::new();
977        let query: DocQuery = (*doc).into();
978        let store = self.store.lock().expect("store lock poisoned");
979        query.load_from_store_into(&mut result, &**store)?;
980        let item = result.into_iter().at_most_one().map_err(|_| {
981            Error::CacheCorruption("Found more than one entry in storage for given docid")
982        })?;
983        if let Some((docid, doctext)) = item {
984            if &docid != doc {
985                return Err(Error::CacheCorruption(
986                    "Item from storage had incorrect docid.",
987                ));
988            }
989            Ok(Some(doctext))
990        } else {
991            Ok(None)
992        }
993    }
994
995    /// Load the text for a collection of documents.
996    ///
997    /// If many of the documents have the same type, this can be more
998    /// efficient than calling [`text`](Self::text).
999    pub fn texts<T>(&self, docs: T) -> Result<HashMap<DocId, DocumentText>>
1000    where
1001        T: IntoIterator<Item = DocId>,
1002    {
1003        let partitioned = docid::partition_by_type(docs);
1004        let mut result = HashMap::new();
1005        let store = self.store.lock().expect("store lock poisoned");
1006        for (_, query) in partitioned.into_iter() {
1007            query.load_from_store_into(&mut result, &**store)?;
1008        }
1009        Ok(result)
1010    }
1011
1012    /// Given a request we sent and the response we got from a
1013    /// directory server, see whether we should expand that response
1014    /// into "something larger".
1015    ///
1016    /// Currently, this handles expanding consensus diffs, and nothing
1017    /// else.  We do it at this stage of our downloading operation
1018    /// because it requires access to the store.
1019    fn expand_response_text(&self, req: &ClientRequest, text: String) -> Result<String> {
1020        if let ClientRequest::Consensus(req) = req {
1021            if tor_consdiff::looks_like_diff(&text) {
1022                if let Some(old_d) = req.old_consensus_digests().next() {
1023                    let db_val = {
1024                        let s = self.store.lock().expect("Directory storage lock poisoned");
1025                        s.consensus_by_sha3_digest_of_signed_part(old_d)?
1026                    };
1027                    if let Some((old_consensus, meta)) = db_val {
1028                        info!("Applying a consensus diff");
1029                        let new_consensus = tor_consdiff::apply_diff(
1030                            old_consensus.as_str()?,
1031                            &text,
1032                            Some(*meta.sha3_256_of_signed()),
1033                        )?;
1034                        new_consensus.check_digest()?;
1035                        return Ok(new_consensus.to_string());
1036                    }
1037                }
1038                return Err(Error::Unwanted(
1039                    "Received a consensus diff we did not ask for",
1040                ));
1041            }
1042        }
1043        Ok(text)
1044    }
1045
1046    /// If `state` has netdir changes to apply, apply them to our netdir.
1047    #[allow(clippy::cognitive_complexity)]
1048    fn apply_netdir_changes(
1049        self: &Arc<Self>,
1050        state: &mut Box<dyn DirState>,
1051        store: &mut dyn Store,
1052    ) -> Result<()> {
1053        if let Some(change) = state.get_netdir_change() {
1054            match change {
1055                NetDirChange::AttemptReplace {
1056                    netdir,
1057                    consensus_meta,
1058                } => {
1059                    // Check the new netdir is sufficient, if we have a circmgr.
1060                    // (Unwraps are fine because the `Option` is `Some` until we take it.)
1061                    if let Some(ref cm) = self.circmgr {
1062                        if !cm
1063                            .netdir_is_sufficient(netdir.as_ref().expect("AttemptReplace had None"))
1064                        {
1065                            debug!("Got a new NetDir, but it doesn't have enough guards yet.");
1066                            return Ok(());
1067                        }
1068                    }
1069                    let is_stale = {
1070                        // Done inside a block to not hold a long-lived copy of the NetDir.
1071                        self.netdir
1072                            .get()
1073                            .map(|x| {
1074                                x.lifetime().valid_after()
1075                                    > netdir
1076                                        .as_ref()
1077                                        .expect("AttemptReplace had None")
1078                                        .lifetime()
1079                                        .valid_after()
1080                            })
1081                            .unwrap_or(false)
1082                    };
1083                    if is_stale {
1084                        warn!("Got a new NetDir, but it's older than the one we currently have!");
1085                        return Err(Error::NetDirOlder);
1086                    }
1087                    let cfg = self.config.get();
1088                    let mut netdir = netdir.take().expect("AttemptReplace had None");
1089                    netdir.replace_overridden_parameters(&cfg.override_net_params);
1090                    self.netdir.replace(netdir);
1091                    self.events.publish(DirEvent::NewConsensus);
1092                    self.events.publish(DirEvent::NewDescriptors);
1093
1094                    info!("Marked consensus usable.");
1095                    if !store.is_readonly() {
1096                        store.mark_consensus_usable(consensus_meta)?;
1097                        // Now that a consensus is usable, older consensuses may
1098                        // need to expire.
1099                        store.expire_all(&crate::storage::EXPIRATION_DEFAULTS)?;
1100                    }
1101                    Ok(())
1102                }
1103                NetDirChange::AddMicrodescs(mds) => {
1104                    self.netdir.mutate(|netdir| {
1105                        for md in mds.drain(..) {
1106                            netdir.add_microdesc(md);
1107                        }
1108                        Ok(())
1109                    })?;
1110                    self.events.publish(DirEvent::NewDescriptors);
1111                    Ok(())
1112                }
1113                NetDirChange::SetRequiredProtocol { timestamp, protos } => {
1114                    if !store.is_readonly() {
1115                        store.update_protocol_recommendations(timestamp, protos.as_ref())?;
1116                    }
1117                    let mut pr = self.protocols.lock().expect("Poisoned lock");
1118                    *pr = Some((timestamp, protos));
1119                    self.events.publish(DirEvent::NewProtocolRecommendation);
1120                    Ok(())
1121                }
1122            }
1123        } else {
1124            Ok(())
1125        }
1126    }
1127}
1128
1129/// A degree of readiness for a given directory state object.
1130#[derive(Debug, Copy, Clone)]
1131enum Readiness {
1132    /// There is no more information to download.
1133    Complete,
1134    /// There is more information to download, but we don't need to
1135    Usable,
1136}
1137
1138/// Try to upgrade a weak reference to a DirMgr, and give an error on
1139/// failure.
1140fn upgrade_weak_ref<T>(weak: &Weak<T>) -> Result<Arc<T>> {
1141    Weak::upgrade(weak).ok_or(Error::ManagerDropped)
1142}
1143
1144/// Given a time `now`, and an amount of tolerated clock skew `tolerance`,
1145/// return the age of the oldest consensus that we should request at that time.
1146pub(crate) fn default_consensus_cutoff(
1147    now: SystemTime,
1148    tolerance: &DirTolerance,
1149) -> Result<SystemTime> {
1150    /// We _always_ allow at least this much age in our consensuses, to account
1151    /// for the fact that consensuses have some lifetime.
1152    const MIN_AGE_TO_ALLOW: Duration = Duration::from_secs(3 * 3600);
1153    let allow_skew = std::cmp::max(MIN_AGE_TO_ALLOW, tolerance.post_valid_tolerance);
1154    let cutoff = time::OffsetDateTime::from(now - allow_skew);
1155    // We now round cutoff to the next hour, so that we aren't leaking our exact
1156    // time to the directory cache.
1157    //
1158    // With the time crate, it's easier to calculate the "next hour" by rounding
1159    // _down_ then adding an hour; rounding up would sometimes require changing
1160    // the date too.
1161    let (h, _m, _s) = cutoff.to_hms();
1162    let cutoff = cutoff.replace_time(
1163        time::Time::from_hms(h, 0, 0)
1164            .map_err(tor_error::into_internal!("Failed clock calculation"))?,
1165    );
1166    let cutoff = cutoff + Duration::from_secs(3600);
1167
1168    Ok(cutoff.into())
1169}
1170
1171/// Return a list of the protocols [supported](tor_protover::doc_supported) by this crate
1172/// when running as a client.
1173pub fn supported_client_protocols() -> tor_protover::Protocols {
1174    use tor_protover::named::*;
1175    // WARNING: REMOVING ELEMENTS FROM THIS LIST CAN BE DANGEROUS!
1176    // SEE [`tor_protover::doc_changing`]
1177    [
1178        //
1179        DIRCACHE_CONSDIFF,
1180    ]
1181    .into_iter()
1182    .collect()
1183}
1184
1185#[cfg(test)]
1186mod test {
1187    // @@ begin test lint list maintained by maint/add_warning @@
1188    #![allow(clippy::bool_assert_comparison)]
1189    #![allow(clippy::clone_on_copy)]
1190    #![allow(clippy::dbg_macro)]
1191    #![allow(clippy::mixed_attributes_style)]
1192    #![allow(clippy::print_stderr)]
1193    #![allow(clippy::print_stdout)]
1194    #![allow(clippy::single_char_pattern)]
1195    #![allow(clippy::unwrap_used)]
1196    #![allow(clippy::unchecked_duration_subtraction)]
1197    #![allow(clippy::useless_vec)]
1198    #![allow(clippy::needless_pass_by_value)]
1199    //! <!-- @@ end test lint list maintained by maint/add_warning @@ -->
1200    use super::*;
1201    use crate::docmeta::{AuthCertMeta, ConsensusMeta};
1202    use std::time::Duration;
1203    use tempfile::TempDir;
1204    use tor_basic_utils::test_rng::testing_rng;
1205    use tor_netdoc::doc::netstatus::ConsensusFlavor;
1206    use tor_netdoc::doc::{authcert::AuthCertKeyIds, netstatus::Lifetime};
1207    use tor_rtcompat::SleepProvider;
1208
1209    #[test]
1210    fn protocols() {
1211        let pr = supported_client_protocols();
1212        let expected = "DirCache=2".parse().unwrap();
1213        assert_eq!(pr, expected);
1214    }
1215
1216    pub(crate) fn new_mgr<R: Runtime>(runtime: R) -> (TempDir, DirMgr<R>) {
1217        let dir = TempDir::new().unwrap();
1218        let config = DirMgrConfig {
1219            cache_dir: dir.path().into(),
1220            ..Default::default()
1221        };
1222        let store = DirMgrStore::new(&config, runtime.clone(), false).unwrap();
1223        let dirmgr = DirMgr::from_config(config, runtime, store, None, false).unwrap();
1224
1225        (dir, dirmgr)
1226    }
1227
1228    #[test]
1229    fn failing_accessors() {
1230        tor_rtcompat::test_with_one_runtime!(|rt| async {
1231            let (_tempdir, mgr) = new_mgr(rt);
1232
1233            assert!(mgr.circmgr().is_err());
1234            assert!(mgr.netdir(Timeliness::Unchecked).is_err());
1235        });
1236    }
1237
1238    #[test]
1239    fn load_and_store_internals() {
1240        tor_rtcompat::test_with_one_runtime!(|rt| async {
1241            let now = rt.wallclock();
1242            let tomorrow = now + Duration::from_secs(86400);
1243            let later = tomorrow + Duration::from_secs(86400);
1244
1245            let (_tempdir, mgr) = new_mgr(rt);
1246
1247            // Seed the storage with a bunch of junk.
1248            let d1 = [5_u8; 32];
1249            let d2 = [7; 32];
1250            let d3 = [42; 32];
1251            let d4 = [99; 20];
1252            let d5 = [12; 20];
1253            let certid1 = AuthCertKeyIds {
1254                id_fingerprint: d4.into(),
1255                sk_fingerprint: d5.into(),
1256            };
1257            let certid2 = AuthCertKeyIds {
1258                id_fingerprint: d5.into(),
1259                sk_fingerprint: d4.into(),
1260            };
1261
1262            {
1263                let mut store = mgr.store.lock().unwrap();
1264
1265                store
1266                    .store_microdescs(
1267                        &[
1268                            ("Fake micro 1", &d1),
1269                            ("Fake micro 2", &d2),
1270                            ("Fake micro 3", &d3),
1271                        ],
1272                        now,
1273                    )
1274                    .unwrap();
1275
1276                #[cfg(feature = "routerdesc")]
1277                store
1278                    .store_routerdescs(&[("Fake rd1", now, &d4), ("Fake rd2", now, &d5)])
1279                    .unwrap();
1280
1281                store
1282                    .store_authcerts(&[
1283                        (
1284                            AuthCertMeta::new(certid1, now, tomorrow),
1285                            "Fake certificate one",
1286                        ),
1287                        (
1288                            AuthCertMeta::new(certid2, now, tomorrow),
1289                            "Fake certificate two",
1290                        ),
1291                    ])
1292                    .unwrap();
1293
1294                let cmeta = ConsensusMeta::new(
1295                    Lifetime::new(now, tomorrow, later).unwrap(),
1296                    [102; 32],
1297                    [103; 32],
1298                );
1299                store
1300                    .store_consensus(&cmeta, ConsensusFlavor::Microdesc, false, "Fake consensus!")
1301                    .unwrap();
1302            }
1303
1304            // Try to get it with text().
1305            let t1 = mgr.text(&DocId::Microdesc(d1)).unwrap().unwrap();
1306            assert_eq!(t1.as_str(), Ok("Fake micro 1"));
1307
1308            let t2 = mgr
1309                .text(&DocId::LatestConsensus {
1310                    flavor: ConsensusFlavor::Microdesc,
1311                    cache_usage: CacheUsage::CacheOkay,
1312                })
1313                .unwrap()
1314                .unwrap();
1315            assert_eq!(t2.as_str(), Ok("Fake consensus!"));
1316
1317            let t3 = mgr.text(&DocId::Microdesc([255; 32])).unwrap();
1318            assert!(t3.is_none());
1319
1320            // Now try texts()
1321            let d_bogus = DocId::Microdesc([255; 32]);
1322            let res = mgr
1323                .texts(vec![
1324                    DocId::Microdesc(d2),
1325                    DocId::Microdesc(d3),
1326                    d_bogus,
1327                    DocId::AuthCert(certid2),
1328                    #[cfg(feature = "routerdesc")]
1329                    DocId::RouterDesc(d5),
1330                ])
1331                .unwrap();
1332            assert_eq!(
1333                res.get(&DocId::Microdesc(d2)).unwrap().as_str(),
1334                Ok("Fake micro 2")
1335            );
1336            assert_eq!(
1337                res.get(&DocId::Microdesc(d3)).unwrap().as_str(),
1338                Ok("Fake micro 3")
1339            );
1340            assert!(!res.contains_key(&d_bogus));
1341            assert_eq!(
1342                res.get(&DocId::AuthCert(certid2)).unwrap().as_str(),
1343                Ok("Fake certificate two")
1344            );
1345            #[cfg(feature = "routerdesc")]
1346            assert_eq!(
1347                res.get(&DocId::RouterDesc(d5)).unwrap().as_str(),
1348                Ok("Fake rd2")
1349            );
1350        });
1351    }
1352
1353    #[test]
1354    fn make_consensus_request() {
1355        tor_rtcompat::test_with_one_runtime!(|rt| async {
1356            let now = rt.wallclock();
1357            let tomorrow = now + Duration::from_secs(86400);
1358            let later = tomorrow + Duration::from_secs(86400);
1359
1360            let (_tempdir, mgr) = new_mgr(rt);
1361            let config = DirMgrConfig::default();
1362
1363            // Try with an empty store.
1364            let req = {
1365                let store = mgr.store.lock().unwrap();
1366                bootstrap::make_consensus_request(
1367                    now,
1368                    ConsensusFlavor::Microdesc,
1369                    &**store,
1370                    &config,
1371                )
1372                .unwrap()
1373            };
1374            let tolerance = DirTolerance::default().post_valid_tolerance;
1375            match req {
1376                ClientRequest::Consensus(r) => {
1377                    assert_eq!(r.old_consensus_digests().count(), 0);
1378                    let date = r.last_consensus_date().unwrap();
1379                    assert!(date >= now - tolerance);
1380                    assert!(date <= now - tolerance + Duration::from_secs(3600));
1381                }
1382                _ => panic!("Wrong request type"),
1383            }
1384
1385            // Add a fake consensus record.
1386            let d_prev = [42; 32];
1387            {
1388                let mut store = mgr.store.lock().unwrap();
1389
1390                let cmeta = ConsensusMeta::new(
1391                    Lifetime::new(now, tomorrow, later).unwrap(),
1392                    d_prev,
1393                    [103; 32],
1394                );
1395                store
1396                    .store_consensus(&cmeta, ConsensusFlavor::Microdesc, false, "Fake consensus!")
1397                    .unwrap();
1398            }
1399
1400            // Now try again.
1401            let req = {
1402                let store = mgr.store.lock().unwrap();
1403                bootstrap::make_consensus_request(
1404                    now,
1405                    ConsensusFlavor::Microdesc,
1406                    &**store,
1407                    &config,
1408                )
1409                .unwrap()
1410            };
1411            match req {
1412                ClientRequest::Consensus(r) => {
1413                    let ds: Vec<_> = r.old_consensus_digests().collect();
1414                    assert_eq!(ds.len(), 1);
1415                    assert_eq!(ds[0], &d_prev);
1416                    assert_eq!(r.last_consensus_date(), Some(now));
1417                }
1418                _ => panic!("Wrong request type"),
1419            }
1420        });
1421    }
1422
1423    #[test]
1424    fn make_other_requests() {
1425        tor_rtcompat::test_with_one_runtime!(|rt| async {
1426            use rand::Rng;
1427            let (_tempdir, mgr) = new_mgr(rt);
1428
1429            let certid1 = AuthCertKeyIds {
1430                id_fingerprint: [99; 20].into(),
1431                sk_fingerprint: [100; 20].into(),
1432            };
1433            let mut rng = testing_rng();
1434            #[cfg(feature = "routerdesc")]
1435            let rd_ids: Vec<DocId> = (0..1000).map(|_| DocId::RouterDesc(rng.random())).collect();
1436            let md_ids: Vec<DocId> = (0..1000).map(|_| DocId::Microdesc(rng.random())).collect();
1437            let config = DirMgrConfig::default();
1438
1439            // Try an authcert.
1440            let query = DocId::AuthCert(certid1);
1441            let store = mgr.store.lock().unwrap();
1442            let reqs =
1443                bootstrap::make_requests_for_documents(&mgr.runtime, &[query], &**store, &config)
1444                    .unwrap();
1445            assert_eq!(reqs.len(), 1);
1446            let req = &reqs[0];
1447            if let ClientRequest::AuthCert(r) = req {
1448                assert_eq!(r.keys().next(), Some(&certid1));
1449            } else {
1450                panic!();
1451            }
1452
1453            // Try a bunch of mds.
1454            let reqs =
1455                bootstrap::make_requests_for_documents(&mgr.runtime, &md_ids, &**store, &config)
1456                    .unwrap();
1457            assert_eq!(reqs.len(), 2);
1458            assert!(matches!(reqs[0], ClientRequest::Microdescs(_)));
1459
1460            // Try a bunch of rds.
1461            #[cfg(feature = "routerdesc")]
1462            {
1463                let reqs = bootstrap::make_requests_for_documents(
1464                    &mgr.runtime,
1465                    &rd_ids,
1466                    &**store,
1467                    &config,
1468                )
1469                .unwrap();
1470                assert_eq!(reqs.len(), 2);
1471                assert!(matches!(reqs[0], ClientRequest::RouterDescs(_)));
1472            }
1473        });
1474    }
1475
1476    #[test]
1477    fn expand_response() {
1478        tor_rtcompat::test_with_one_runtime!(|rt| async {
1479            let now = rt.wallclock();
1480            let day = Duration::from_secs(86400);
1481            let config = DirMgrConfig::default();
1482
1483            let (_tempdir, mgr) = new_mgr(rt);
1484
1485            // Try a simple request: nothing should happen.
1486            let q = DocId::Microdesc([99; 32]);
1487            let r = {
1488                let store = mgr.store.lock().unwrap();
1489                bootstrap::make_requests_for_documents(&mgr.runtime, &[q], &**store, &config)
1490                    .unwrap()
1491            };
1492            let expanded = mgr.expand_response_text(&r[0], "ABC".to_string());
1493            assert_eq!(&expanded.unwrap(), "ABC");
1494
1495            // Try a consensus response that doesn't look like a diff in
1496            // response to a query that doesn't ask for one.
1497            let latest_id = DocId::LatestConsensus {
1498                flavor: ConsensusFlavor::Microdesc,
1499                cache_usage: CacheUsage::CacheOkay,
1500            };
1501            let r = {
1502                let store = mgr.store.lock().unwrap();
1503                bootstrap::make_requests_for_documents(
1504                    &mgr.runtime,
1505                    &[latest_id],
1506                    &**store,
1507                    &config,
1508                )
1509                .unwrap()
1510            };
1511            let expanded = mgr.expand_response_text(&r[0], "DEF".to_string());
1512            assert_eq!(&expanded.unwrap(), "DEF");
1513
1514            // Now stick some metadata and a string into the storage so that
1515            // we can ask for a diff.
1516            {
1517                let mut store = mgr.store.lock().unwrap();
1518                let d_in = [0x99; 32]; // This one, we can fake.
1519                let cmeta = ConsensusMeta::new(
1520                    Lifetime::new(now, now + day, now + 2 * day).unwrap(),
1521                    d_in,
1522                    d_in,
1523                );
1524                store
1525                    .store_consensus(
1526                        &cmeta,
1527                        ConsensusFlavor::Microdesc,
1528                        false,
1529                        "line 1\nline2\nline 3\n",
1530                    )
1531                    .unwrap();
1532            }
1533
1534            // Try expanding something that isn't a consensus, even if we'd like
1535            // one.
1536            let r = {
1537                let store = mgr.store.lock().unwrap();
1538                bootstrap::make_requests_for_documents(
1539                    &mgr.runtime,
1540                    &[latest_id],
1541                    &**store,
1542                    &config,
1543                )
1544                .unwrap()
1545            };
1546            let expanded = mgr.expand_response_text(&r[0], "hello".to_string());
1547            assert_eq!(&expanded.unwrap(), "hello");
1548
1549            // Finally, try "expanding" a diff (by applying it and checking the digest.
1550            let diff = "network-status-diff-version 1
1551hash 9999999999999999999999999999999999999999999999999999999999999999 8382374ca766873eb0d2530643191c6eaa2c5e04afa554cbac349b5d0592d300
15522c
1553replacement line
1554.
1555".to_string();
1556            let expanded = mgr.expand_response_text(&r[0], diff);
1557
1558            assert_eq!(expanded.unwrap(), "line 1\nreplacement line\nline 3\n");
1559
1560            // If the digest is wrong, that should get rejected.
1561            let diff = "network-status-diff-version 1
1562hash 9999999999999999999999999999999999999999999999999999999999999999 9999999999999999999999999999999999999999999999999999999999999999
15632c
1564replacement line
1565.
1566".to_string();
1567            let expanded = mgr.expand_response_text(&r[0], diff);
1568            assert!(expanded.is_err());
1569        });
1570    }
1571}