Struct tor_hsservice::ipt_set::PublishIptSet

pub(crate) struct PublishIptSet {
    pub(crate) ipts: Option<IptSet>,
    pub(crate) last_descriptor_expiry_including_slop: HashMap<IptLocalId, Instant>,
    storage: StorageHandle<StateRecord>,
}

Information shared between the IPT manager and the IPT publisher

The principal information is ipts, which is calculated by the IPT Manager. See IptManager::compute_iptsetstatus_publish for more detailed information about how this is calculated.

Fields

ipts: Option<IptSet>

Set of introduction points to be advertised in a descriptor (if we are to publish)

If Some, the publisher will try to maintain a published descriptor, of lifetime lifetime, listing ipts.

If None, the publisher will not try to publish. (Already-published descriptors will not be deleted.)

These instructions ultimately come from IptManager::compute_iptsetstatus_publish.

last_descriptor_expiry_including_slop: HashMap<IptLocalId, Instant>

Record of publication attempts

Time until which the manager ought we to try to maintain each ipt, even after we stop publishing it.

This is a ceiling on:

• The last time we finished publishing the descriptor (we can estimate this by taking the time we started to publish plus our timeout on the publication attempt).

• Plus the lifetime that was used for publication.

• Plus the length of time between a client obtaining the descriptor and its introduction request reaching us through the intro point (IPT_PUBLISH_EXPIRY_SLOP)

This field is updated by the publisher, using note_publication_attempt, and read by the manager.

A separate copy of the information is stored by the manager, in ipt_mgr::Ipt::last_descriptor_expiry_including_slop.

There may be entries in this table that don’t correspond to introduction points in ipts. The publisher mustn’t create such entries (since that would imply publishing IPTs contrary to the manager’s instructions) but it can occur, for example, on restart.

It is the manager’s job to remove expired entries.

storage: StorageHandle<StateRecord>

The on-disk state storage handle.

Implementations

impl PublishIptSet

pub(crate) fn note_publication_attempt( &mut self, runtime: &impl SleepProvider, worst_case_end: Instant ) -> Result<(), IptStoreError>

Update all the last_descriptor_expiry_including_slop for a publication attempt

Called by the publisher when it starts a publication attempt which will advertise this set of introduction points.

When calling this, the publisher promises that the publication attempt will either complete, or be abandoned, before worst_case_end.

impl PublishIptSet

fn save(&mut self, runtime: &impl SleepProvider) -> Result<(), IptStoreError>

Save the publication times to the persistent state

fn load( storage: StorageHandle<StateRecord>, runtime: &impl SleepProvider ) -> Result<PublishIptSet, StartupError>

Load the publication times from the persistent state

Trait Implementations

impl Debug for PublishIptSet

fn fmt(&self, formatter: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.