Struct tor_guardmgr::GuardMgr

pub struct GuardMgr<R: Runtime> { /* private fields */ }

A “guard manager” that selects and remembers a persistent set of guard nodes.

Implementations

impl<R: Runtime> GuardMgr<R>

pub fn new<S>(
    runtime: R,
    state_mgr: S,
    fallbacks: FallbackList
) -> Result<Self, GuardMgrError> where
    S: StateMgr + Send + Sync + 'static, 

Create a new “empty” guard manager and launch its background tasks.

It won’t be able to hand out any guards until GuardMgr::update_network has been called.

pub fn install_netdir_provider(
    &self,
    provider: &Arc<dyn NetDirProvider>
) -> Result<(), GuardMgrError>

Install a NetDirProvider for use by this guard manager.

It will be used to keep the guards up-to-date with changes from the network directory, and to find new guards when no NetDir is provided to select_guard().

TODO: we should eventually return some kind of a task handle from this task, even though it is not strictly speaking periodic.

pub fn store_persistent_state(&self) -> Result<(), GuardMgrError>

Flush our current guard state to the state manager, if there is any unsaved state.

pub fn reload_persistent_state(&self) -> Result<(), GuardMgrError>

Reload state from the state manager.

We only call this method if we don’t have the lock on the state files. If we have the lock, we only want to save.

pub fn upgrade_to_owned_persistent_state(&self) -> Result<(), GuardMgrError>

Switch from having an unowned persistent state to having an owned one.

Requires that we hold the lock on the state files.

pub fn netdir_is_sufficient(&self, netdir: &NetDir) -> bool

Return true if netdir has enough information to safely become our new netdir.

pub fn mark_all_guards_retriable(&self)

Mark every guard as potentially retriable, regardless of how recently we failed to connect to it.

pub fn update_network(&self, netdir: &NetDir)

Update the state of this GuardMgr based on a new or modified NetDir object.

This method can add new guards, or notice that existing guards have become unusable. It needs a NetDir so it can identify potential candidate guards.

Call this method whenever the NetDir changes, unless you have used install_netdir_provider.

pub fn replace_fallback_list(&self, list: FallbackList)

Replace the fallback list held by this GuardMgr with new_list.

pub fn set_filter(&self, filter: GuardFilter, netdir: Option<&NetDir>)

Replace the current GuardFilter used by this GuardMgr.

pub fn select_guard(
    &self,
    usage: GuardUsage,
    netdir: Option<&NetDir>
) -> Result<(FirstHop, GuardMonitor, GuardUsable), PickGuardError>

Select a guard for a given GuardUsage.

On success, we return a FirstHop object to identify which guard we have picked, a GuardMonitor object that the caller can use to report whether its attempt to use the guard succeeded or failed, and a GuardUsable future that the caller can use to decide whether a circuit built through the guard is actually safe to use.

That last point is important: It’s okay to build a circuit through the guard returned by this function, but you can’t actually use it for traffic unless the GuardUsable future yields “true”.

Limitations

This function will never return a guard that isn’t listed in the most recent NetDir.

That’s usually what you’d want, but when we’re trying to bootstrap we might want to use all guards as possible directory caches. That’s not implemented yet. (See ticket #220).

This function only looks at netdir when all of the known guards are down; to force an update, use GuardMgr::update_network.

pub fn note_external_failure(
    &self,
    ed_identity: &Ed25519Identity,
    rsa_identity: &RsaIdentity,
    external_failure: ExternalActivity
)

Record that after we built a circuit with a guard, something described in external_failure went wrong with it.

pub fn note_external_success(
    &self,
    ed_identity: &Ed25519Identity,
    rsa_identity: &RsaIdentity,
    external_activity: ExternalActivity
)

Record that after we built a circuit with a guard, some activity described in external_activity was successful with it.

pub fn skew_events(&self) -> ClockSkewEvents

Return a stream of events about our estimated clock skew; these events are None when we don’t have enough information to make an estimate, and Some(SkewEstimate) otherwise.

Note that this stream can be lossy: if the estimate changes more than one before you read from the stream, you might only get the most recent update.

Trait Implementations

impl<R: Clone + Runtime> Clone for GuardMgr<R>

fn clone(&self) -> GuardMgr<R>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.