iroh_quinn_proto/connection/
mod.rs

1use std::{
2    cmp,
3    collections::{BTreeMap, VecDeque, btree_map},
4    convert::TryFrom,
5    fmt, io, mem,
6    net::{IpAddr, SocketAddr},
7    num::NonZeroU32,
8    ops::Not,
9    sync::Arc,
10};
11
12use bytes::{BufMut, Bytes, BytesMut};
13use frame::StreamMetaVec;
14
15use rand::{Rng, SeedableRng, rngs::StdRng};
16use rustc_hash::{FxHashMap, FxHashSet};
17use thiserror::Error;
18use tracing::{debug, error, trace, trace_span, warn};
19
20use crate::{
21    Dir, Duration, EndpointConfig, Frame, INITIAL_MTU, Instant, MAX_CID_SIZE, MAX_STREAM_COUNT,
22    MIN_INITIAL_SIZE, Side, StreamId, TIMER_GRANULARITY, TokenStore, Transmit, TransportError,
23    TransportErrorCode, VarInt,
24    cid_generator::ConnectionIdGenerator,
25    cid_queue::CidQueue,
26    coding::BufMutExt,
27    config::{ServerConfig, TransportConfig},
28    congestion::Controller,
29    connection::{
30        qlog::{QlogRecvPacket, QlogSentPacket, QlogSink},
31        spaces::LostPacket,
32        timer::{ConnTimer, PathTimer},
33    },
34    crypto::{self, KeyPair, Keys, PacketKey},
35    frame::{self, Close, Datagram, FrameStruct, NewToken, ObservedAddr},
36    iroh_hp,
37    packet::{
38        FixedLengthConnectionIdParser, Header, InitialHeader, InitialPacket, LongType, Packet,
39        PacketNumber, PartialDecode, SpaceId,
40    },
41    range_set::ArrayRangeSet,
42    shared::{
43        ConnectionEvent, ConnectionEventInner, ConnectionId, DatagramConnectionEvent, EcnCodepoint,
44        EndpointEvent, EndpointEventInner,
45    },
46    token::{ResetToken, Token, TokenPayload},
47    transport_parameters::TransportParameters,
48};
49
50mod ack_frequency;
51use ack_frequency::AckFrequencyState;
52
53mod assembler;
54pub use assembler::Chunk;
55
56mod cid_state;
57use cid_state::CidState;
58
59mod datagrams;
60use datagrams::DatagramState;
61pub use datagrams::{Datagrams, SendDatagramError};
62
63mod mtud;
64mod pacing;
65
66mod packet_builder;
67use packet_builder::{PacketBuilder, PadDatagram};
68
69mod packet_crypto;
70use packet_crypto::{PrevCrypto, ZeroRttCrypto};
71
72mod paths;
73pub use paths::{ClosedPath, PathEvent, PathId, PathStatus, RttEstimator, SetPathStatusError};
74use paths::{PathData, PathState};
75
76pub(crate) mod qlog;
77
78mod send_buffer;
79
80mod spaces;
81#[cfg(fuzzing)]
82pub use spaces::Retransmits;
83#[cfg(not(fuzzing))]
84use spaces::Retransmits;
85use spaces::{PacketSpace, SendableFrames, SentPacket, ThinRetransmits};
86
87mod stats;
88pub use stats::{ConnectionStats, FrameStats, PathStats, UdpStats};
89
90mod streams;
91#[cfg(fuzzing)]
92pub use streams::StreamsState;
93#[cfg(not(fuzzing))]
94use streams::StreamsState;
95pub use streams::{
96    Chunks, ClosedStream, FinishError, ReadError, ReadableError, RecvStream, SendStream,
97    ShouldTransmit, StreamEvent, Streams, WriteError, Written,
98};
99
100mod timer;
101use timer::{Timer, TimerTable};
102
103mod transmit_buf;
104use transmit_buf::TransmitBuf;
105
106mod state;
107
108#[cfg(not(fuzzing))]
109use state::State;
110#[cfg(fuzzing)]
111pub use state::State;
112use state::StateType;
113
114/// Protocol state and logic for a single QUIC connection
115///
116/// Objects of this type receive [`ConnectionEvent`]s and emit [`EndpointEvent`]s and application
117/// [`Event`]s to make progress. To handle timeouts, a `Connection` returns timer updates and
118/// expects timeouts through various methods. A number of simple getter methods are exposed
119/// to allow callers to inspect some of the connection state.
120///
121/// `Connection` has roughly 4 types of methods:
122///
123/// - A. Simple getters, taking `&self`
124/// - B. Handlers for incoming events from the network or system, named `handle_*`.
125/// - C. State machine mutators, for incoming commands from the application. For convenience we
126///   refer to this as "performing I/O" below, however as per the design of this library none of the
127///   functions actually perform system-level I/O. For example, [`read`](RecvStream::read) and
128///   [`write`](SendStream::write), but also things like [`reset`](SendStream::reset).
129/// - D. Polling functions for outgoing events or actions for the caller to
130///   take, named `poll_*`.
131///
132/// The simplest way to use this API correctly is to call (B) and (C) whenever
133/// appropriate, then after each of those calls, as soon as feasible call all
134/// polling methods (D) and deal with their outputs appropriately, e.g. by
135/// passing it to the application or by making a system-level I/O call. You
136/// should call the polling functions in this order:
137///
138/// 1. [`poll_transmit`](Self::poll_transmit)
139/// 2. [`poll_timeout`](Self::poll_timeout)
140/// 3. [`poll_endpoint_events`](Self::poll_endpoint_events)
141/// 4. [`poll`](Self::poll)
142///
143/// Currently the only actual dependency is from (2) to (1), however additional
144/// dependencies may be added in future, so the above order is recommended.
145///
146/// (A) may be called whenever desired.
147///
148/// Care should be made to ensure that the input events represent monotonically
149/// increasing time. Specifically, calling [`handle_timeout`](Self::handle_timeout)
150/// with events of the same [`Instant`] may be interleaved in any order with a
151/// call to [`handle_event`](Self::handle_event) at that same instant; however
152/// events or timeouts with different instants must not be interleaved.
153pub struct Connection {
154    endpoint_config: Arc<EndpointConfig>,
155    config: Arc<TransportConfig>,
156    rng: StdRng,
157    crypto: Box<dyn crypto::Session>,
158    /// The CID we initially chose, for use during the handshake
159    handshake_cid: ConnectionId,
160    /// The CID the peer initially chose, for use during the handshake
161    rem_handshake_cid: ConnectionId,
162    /// The "real" local IP address which was was used to receive the initial packet.
163    /// This is only populated for the server case, and if known
164    local_ip: Option<IpAddr>,
165    /// The [`PathData`] for each path
166    ///
167    /// This needs to be ordered because [`Connection::poll_transmit`] needs to
168    /// deterministically select the next PathId to send on.
169    // TODO(flub): well does it really? But deterministic is nice for now.
170    paths: BTreeMap<PathId, PathState>,
171    /// Incremented every time we see a new path
172    ///
173    /// Stored separately from `path.generation` to account for aborted migrations
174    path_counter: u64,
175    /// Whether MTU detection is supported in this environment
176    allow_mtud: bool,
177    state: State,
178    side: ConnectionSide,
179    /// Whether or not 0-RTT was enabled during the handshake. Does not imply acceptance.
180    zero_rtt_enabled: bool,
181    /// Set if 0-RTT is supported, then cleared when no longer needed.
182    zero_rtt_crypto: Option<ZeroRttCrypto>,
183    key_phase: bool,
184    /// How many packets are in the current key phase. Used only for `Data` space.
185    key_phase_size: u64,
186    /// Transport parameters set by the peer
187    peer_params: TransportParameters,
188    /// Source ConnectionId of the first packet received from the peer
189    orig_rem_cid: ConnectionId,
190    /// Destination ConnectionId sent by the client on the first Initial
191    initial_dst_cid: ConnectionId,
192    /// The value that the server included in the Source Connection ID field of a Retry packet, if
193    /// one was received
194    retry_src_cid: Option<ConnectionId>,
195    /// Events returned by [`Connection::poll`]
196    events: VecDeque<Event>,
197    endpoint_events: VecDeque<EndpointEventInner>,
198    /// Whether the spin bit is in use for this connection
199    spin_enabled: bool,
200    /// Outgoing spin bit state
201    spin: bool,
202    /// Packet number spaces: initial, handshake, 1-RTT
203    spaces: [PacketSpace; 3],
204    /// Highest usable [`SpaceId`]
205    highest_space: SpaceId,
206    /// 1-RTT keys used prior to a key update
207    prev_crypto: Option<PrevCrypto>,
208    /// 1-RTT keys to be used for the next key update
209    ///
210    /// These are generated in advance to prevent timing attacks and/or DoS by third-party attackers
211    /// spoofing key updates.
212    next_crypto: Option<KeyPair<Box<dyn PacketKey>>>,
213    accepted_0rtt: bool,
214    /// Whether the idle timer should be reset the next time an ack-eliciting packet is transmitted.
215    permit_idle_reset: bool,
216    /// Negotiated idle timeout
217    idle_timeout: Option<Duration>,
218    timers: TimerTable,
219    /// Number of packets received which could not be authenticated
220    authentication_failures: u64,
221
222    //
223    // Queued non-retransmittable 1-RTT data
224    //
225    /// If the CONNECTION_CLOSE frame needs to be sent
226    close: bool,
227
228    //
229    // ACK frequency
230    //
231    ack_frequency: AckFrequencyState,
232
233    //
234    // Congestion Control
235    //
236    /// Whether the most recently received packet had an ECN codepoint set
237    receiving_ecn: bool,
238    /// Number of packets authenticated
239    total_authed_packets: u64,
240    /// Whether the last `poll_transmit` call yielded no data because there was
241    /// no outgoing application data.
242    app_limited: bool,
243
244    //
245    // ObservedAddr
246    //
247    /// Sequence number for the next observed address frame sent to the peer.
248    next_observed_addr_seq_no: VarInt,
249
250    streams: StreamsState,
251    /// Surplus remote CIDs for future use on new paths
252    ///
253    /// These are given out before multiple paths exist, also for paths that will never
254    /// exist.  So if multipath is supported the number of paths here will be higher than
255    /// the actual number of paths in use.
256    rem_cids: FxHashMap<PathId, CidQueue>,
257    /// Attributes of CIDs generated by local endpoint
258    ///
259    /// Any path that is allowed to be opened is present in this map, as well as the already
260    /// opened paths. However since CIDs are issued async by the endpoint driver via
261    /// connection events it can not be used to know if CIDs have been issued for a path or
262    /// not. See [`Connection::max_path_id_with_cids`] for this.
263    local_cid_state: FxHashMap<PathId, CidState>,
264    /// State of the unreliable datagram extension
265    datagrams: DatagramState,
266    /// Connection level statistics
267    stats: ConnectionStats,
268    /// Path level statistics
269    path_stats: FxHashMap<PathId, PathStats>,
270    /// QUIC version used for the connection.
271    version: u32,
272
273    //
274    // Multipath
275    //
276    /// Maximum number of concurrent paths
277    ///
278    /// Initially set from the [`TransportConfig::max_concurrent_multipath_paths`]. Even
279    /// when multipath is disabled this will be set to 1, it is not used in that case
280    /// though.
281    max_concurrent_paths: NonZeroU32,
282    /// Local maximum [`PathId`] to be used
283    ///
284    /// This is initially set to [`TransportConfig::get_initial_max_path_id`] when multipath
285    /// is negotiated, or to [`PathId::ZERO`] otherwise. This is essentially the value of
286    /// the highest MAX_PATH_ID frame sent.
287    ///
288    /// Any path with an ID equal or below this [`PathId`] is either:
289    ///
290    /// - Abandoned, if it is also in [`Connection::abandoned_paths`].
291    /// - Open, in this case it is present in [`Connection::paths`]
292    /// - Not yet opened, if it is in neither of these two places.
293    ///
294    /// Note that for not-yet-open there may or may not be any CIDs issued. See
295    /// [`Connection::max_path_id_with_cids`].
296    local_max_path_id: PathId,
297    /// Remote's maximum [`PathId`] to be used
298    ///
299    /// This is initially set to the peer's [`TransportParameters::initial_max_path_id`] when
300    /// multipath is negotiated, or to [`PathId::ZERO`] otherwise. A peer may increase this limit
301    /// by sending [`Frame::MaxPathId`] frames.
302    remote_max_path_id: PathId,
303    /// The greatest [`PathId`] we have issued CIDs for
304    ///
305    /// CIDs are only issued for `min(local_max_path_id, remote_max_path_id)`. It is not
306    /// possible to use [`Connection::local_cid_state`] to know if CIDs have been issued
307    /// since they are issued asynchronously by the endpoint driver.
308    max_path_id_with_cids: PathId,
309    /// The paths already abandoned
310    ///
311    /// They may still have some state left in [`Connection::paths`] or
312    /// [`Connection::local_cid_state`] since some of this has to be kept around for some
313    /// time after a path is abandoned.
314    // TODO(flub): Make this a more efficient data structure.  Like ranges of abandoned
315    //    paths.  Or a set together with a minimum.  Or something.
316    abandoned_paths: FxHashSet<PathId>,
317
318    iroh_hp: iroh_hp::State,
319    qlog: QlogSink,
320}
321
322impl Connection {
323    pub(crate) fn new(
324        endpoint_config: Arc<EndpointConfig>,
325        config: Arc<TransportConfig>,
326        init_cid: ConnectionId,
327        loc_cid: ConnectionId,
328        rem_cid: ConnectionId,
329        remote: SocketAddr,
330        local_ip: Option<IpAddr>,
331        crypto: Box<dyn crypto::Session>,
332        cid_gen: &dyn ConnectionIdGenerator,
333        now: Instant,
334        version: u32,
335        allow_mtud: bool,
336        rng_seed: [u8; 32],
337        side_args: SideArgs,
338        qlog: QlogSink,
339    ) -> Self {
340        let pref_addr_cid = side_args.pref_addr_cid();
341        let path_validated = side_args.path_validated();
342        let connection_side = ConnectionSide::from(side_args);
343        let side = connection_side.side();
344        let mut rng = StdRng::from_seed(rng_seed);
345        let initial_space = {
346            let mut space = PacketSpace::new(now, SpaceId::Initial, &mut rng);
347            space.crypto = Some(crypto.initial_keys(init_cid, side));
348            space
349        };
350        let handshake_space = PacketSpace::new(now, SpaceId::Handshake, &mut rng);
351        #[cfg(test)]
352        let data_space = match config.deterministic_packet_numbers {
353            true => PacketSpace::new_deterministic(now, SpaceId::Data),
354            false => PacketSpace::new(now, SpaceId::Data, &mut rng),
355        };
356        #[cfg(not(test))]
357        let data_space = PacketSpace::new(now, SpaceId::Data, &mut rng);
358        let state = State::handshake(state::Handshake {
359            rem_cid_set: side.is_server(),
360            expected_token: Bytes::new(),
361            client_hello: None,
362            allow_server_migration: side.is_client(),
363        });
364        let local_cid_state = FxHashMap::from_iter([(
365            PathId::ZERO,
366            CidState::new(
367                cid_gen.cid_len(),
368                cid_gen.cid_lifetime(),
369                now,
370                if pref_addr_cid.is_some() { 2 } else { 1 },
371            ),
372        )]);
373
374        let mut path = PathData::new(remote, allow_mtud, None, 0, now, &config);
375        // TODO(@divma): consider if we want to delay this until the path is validated
376        path.open = true;
377        let mut this = Self {
378            endpoint_config,
379            crypto,
380            handshake_cid: loc_cid,
381            rem_handshake_cid: rem_cid,
382            local_cid_state,
383            paths: BTreeMap::from_iter([(
384                PathId::ZERO,
385                PathState {
386                    data: path,
387                    prev: None,
388                },
389            )]),
390            path_counter: 0,
391            allow_mtud,
392            local_ip,
393            state,
394            side: connection_side,
395            zero_rtt_enabled: false,
396            zero_rtt_crypto: None,
397            key_phase: false,
398            // A small initial key phase size ensures peers that don't handle key updates correctly
399            // fail sooner rather than later. It's okay for both peers to do this, as the first one
400            // to perform an update will reset the other's key phase size in `update_keys`, and a
401            // simultaneous key update by both is just like a regular key update with a really fast
402            // response. Inspired by quic-go's similar behavior of performing the first key update
403            // at the 100th short-header packet.
404            key_phase_size: rng.random_range(10..1000),
405            peer_params: TransportParameters::default(),
406            orig_rem_cid: rem_cid,
407            initial_dst_cid: init_cid,
408            retry_src_cid: None,
409            events: VecDeque::new(),
410            endpoint_events: VecDeque::new(),
411            spin_enabled: config.allow_spin && rng.random_ratio(7, 8),
412            spin: false,
413            spaces: [initial_space, handshake_space, data_space],
414            highest_space: SpaceId::Initial,
415            prev_crypto: None,
416            next_crypto: None,
417            accepted_0rtt: false,
418            permit_idle_reset: true,
419            idle_timeout: match config.max_idle_timeout {
420                None | Some(VarInt(0)) => None,
421                Some(dur) => Some(Duration::from_millis(dur.0)),
422            },
423            timers: TimerTable::default(),
424            authentication_failures: 0,
425            close: false,
426
427            ack_frequency: AckFrequencyState::new(get_max_ack_delay(
428                &TransportParameters::default(),
429            )),
430
431            app_limited: false,
432            receiving_ecn: false,
433            total_authed_packets: 0,
434
435            next_observed_addr_seq_no: 0u32.into(),
436
437            streams: StreamsState::new(
438                side,
439                config.max_concurrent_uni_streams,
440                config.max_concurrent_bidi_streams,
441                config.send_window,
442                config.receive_window,
443                config.stream_receive_window,
444            ),
445            datagrams: DatagramState::default(),
446            config,
447            rem_cids: FxHashMap::from_iter([(PathId::ZERO, CidQueue::new(rem_cid))]),
448            rng,
449            stats: ConnectionStats::default(),
450            path_stats: Default::default(),
451            version,
452
453            // peer params are not yet known, so multipath is not enabled
454            max_concurrent_paths: NonZeroU32::MIN,
455            local_max_path_id: PathId::ZERO,
456            remote_max_path_id: PathId::ZERO,
457            max_path_id_with_cids: PathId::ZERO,
458            abandoned_paths: Default::default(),
459
460            // iroh's nat traversal
461            iroh_hp: Default::default(),
462            qlog,
463        };
464        if path_validated {
465            this.on_path_validated(PathId::ZERO);
466        }
467        if side.is_client() {
468            // Kick off the connection
469            this.write_crypto();
470            this.init_0rtt(now);
471        }
472        this.qlog.emit_tuple_assigned(PathId::ZERO, remote, now);
473        this
474    }
475
476    /// Returns the next time at which `handle_timeout` should be called
477    ///
478    /// The value returned may change after:
479    /// - the application performed some I/O on the connection
480    /// - a call was made to `handle_event`
481    /// - a call to `poll_transmit` returned `Some`
482    /// - a call was made to `handle_timeout`
483    #[must_use]
484    pub fn poll_timeout(&mut self) -> Option<Instant> {
485        self.timers.peek()
486    }
487
488    /// Returns application-facing events
489    ///
490    /// Connections should be polled for events after:
491    /// - a call was made to `handle_event`
492    /// - a call was made to `handle_timeout`
493    #[must_use]
494    pub fn poll(&mut self) -> Option<Event> {
495        if let Some(x) = self.events.pop_front() {
496            return Some(x);
497        }
498
499        if let Some(event) = self.streams.poll() {
500            return Some(Event::Stream(event));
501        }
502
503        if let Some(reason) = self.state.take_error() {
504            return Some(Event::ConnectionLost { reason });
505        }
506
507        None
508    }
509
510    /// Return endpoint-facing events
511    #[must_use]
512    pub fn poll_endpoint_events(&mut self) -> Option<EndpointEvent> {
513        self.endpoint_events.pop_front().map(EndpointEvent)
514    }
515
516    /// Provide control over streams
517    #[must_use]
518    pub fn streams(&mut self) -> Streams<'_> {
519        Streams {
520            state: &mut self.streams,
521            conn_state: &self.state,
522        }
523    }
524
525    /// Provide control over streams
526    #[must_use]
527    pub fn recv_stream(&mut self, id: StreamId) -> RecvStream<'_> {
528        assert!(id.dir() == Dir::Bi || id.initiator() != self.side.side());
529        RecvStream {
530            id,
531            state: &mut self.streams,
532            pending: &mut self.spaces[SpaceId::Data].pending,
533        }
534    }
535
536    /// Provide control over streams
537    #[must_use]
538    pub fn send_stream(&mut self, id: StreamId) -> SendStream<'_> {
539        assert!(id.dir() == Dir::Bi || id.initiator() == self.side.side());
540        SendStream {
541            id,
542            state: &mut self.streams,
543            pending: &mut self.spaces[SpaceId::Data].pending,
544            conn_state: &self.state,
545        }
546    }
547
548    /// Opens a new path only if no path to the remote address exists so far
549    ///
550    /// See [`open_path`]. Returns `(path_id, true)` if the path already existed. `(path_id,
551    /// false)` if was opened.
552    ///
553    /// [`open_path`]: Connection::open_path
554    pub fn open_path_ensure(
555        &mut self,
556        remote: SocketAddr,
557        initial_status: PathStatus,
558        now: Instant,
559    ) -> Result<(PathId, bool), PathError> {
560        match self
561            .paths
562            .iter()
563            .find(|(_id, path)| path.data.remote == remote)
564        {
565            Some((path_id, _state)) => Ok((*path_id, true)),
566            None => self
567                .open_path(remote, initial_status, now)
568                .map(|id| (id, false)),
569        }
570    }
571
572    /// Opens a new path
573    ///
574    /// Further errors might occur and they will be emitted in [`PathEvent::LocallyClosed`] events.
575    /// When the path is opened it will be reported as an [`PathEvent::Opened`].
576    pub fn open_path(
577        &mut self,
578        remote: SocketAddr,
579        initial_status: PathStatus,
580        now: Instant,
581    ) -> Result<PathId, PathError> {
582        if !self.is_multipath_negotiated() {
583            return Err(PathError::MultipathNotNegotiated);
584        }
585        if self.side().is_server() {
586            return Err(PathError::ServerSideNotAllowed);
587        }
588
589        let max_abandoned = self.abandoned_paths.iter().max().copied();
590        let max_used = self.paths.keys().last().copied();
591        let path_id = max_abandoned
592            .max(max_used)
593            .unwrap_or(PathId::ZERO)
594            .saturating_add(1u8);
595
596        if Some(path_id) > self.max_path_id() {
597            return Err(PathError::MaxPathIdReached);
598        }
599        if path_id > self.remote_max_path_id {
600            self.spaces[SpaceId::Data].pending.paths_blocked = true;
601            return Err(PathError::MaxPathIdReached);
602        }
603        if self.rem_cids.get(&path_id).map(CidQueue::active).is_none() {
604            self.spaces[SpaceId::Data]
605                .pending
606                .path_cids_blocked
607                .push(path_id);
608            return Err(PathError::RemoteCidsExhausted);
609        }
610
611        let path = self.ensure_path(path_id, remote, now, None);
612        path.status.local_update(initial_status);
613
614        Ok(path_id)
615    }
616
617    /// Closes a path by sending a PATH_ABANDON frame
618    ///
619    /// This will not allow closing the last path. It does allow closing paths which have
620    /// not yet been opened, as e.g. is the case when receiving a PATH_ABANDON from the peer
621    /// for a path that was never opened locally.
622    pub fn close_path(
623        &mut self,
624        now: Instant,
625        path_id: PathId,
626        error_code: VarInt,
627    ) -> Result<(), ClosePathError> {
628        if self.abandoned_paths.contains(&path_id)
629            || Some(path_id) > self.max_path_id()
630            || !self.paths.contains_key(&path_id)
631        {
632            return Err(ClosePathError::ClosedPath);
633        }
634        if self
635            .paths
636            .iter()
637            // Would there be any remaining, non-abandoned, validated paths
638            .any(|(id, path)| {
639                *id != path_id && !self.abandoned_paths.contains(id) && path.data.validated
640            })
641            .not()
642        {
643            return Err(ClosePathError::LastOpenPath);
644        }
645
646        // Send PATH_ABANDON
647        self.spaces[SpaceId::Data]
648            .pending
649            .path_abandon
650            .insert(path_id, error_code.into());
651
652        // Remove pending NEW CIDs for this path
653        let pending_space = &mut self.spaces[SpaceId::Data].pending;
654        pending_space.new_cids.retain(|cid| cid.path_id != path_id);
655        pending_space.path_cids_blocked.retain(|&id| id != path_id);
656        pending_space.path_status.retain(|&id| id != path_id);
657
658        // Cleanup retransmits across ALL paths (CIDs for path_id may have been transmitted on other paths)
659        for space in self.spaces[SpaceId::Data].iter_paths_mut() {
660            for sent_packet in space.sent_packets.values_mut() {
661                if let Some(retransmits) = sent_packet.retransmits.get_mut() {
662                    retransmits.new_cids.retain(|cid| cid.path_id != path_id);
663                    retransmits.path_cids_blocked.retain(|&id| id != path_id);
664                    retransmits.path_status.retain(|&id| id != path_id);
665                }
666            }
667        }
668
669        // Consider remotely issued CIDs as retired.
670        // Technically we don't have to do this just yet.  We only need to do this *after*
671        // the ABANDON_PATH frame is sent, allowing us to still send it on the
672        // to-be-abandoned path.  However it is recommended to send it on another path, and
673        // we do not allow abandoning the last path anyway.
674        self.rem_cids.remove(&path_id);
675        self.endpoint_events
676            .push_back(EndpointEventInner::RetireResetToken(path_id));
677
678        let end = self.calculate_end_timer(now, 3, SpaceId::Data);
679        let extended_end = self.calculate_end_timer(now, 6, SpaceId::Data);
680        let path = self.paths.get_mut(&path_id).expect("checked above");
681
682        // We record the time after which receiving data on this path generates a transport error.
683        path.data.last_allowed_receive = Some(end);
684        self.abandoned_paths.insert(path_id);
685
686        self.set_max_path_id(now, self.local_max_path_id.saturating_add(1u8));
687
688        // The peer MUST respond with a corresponding PATH_ABANDON frame.
689        // If we receive packets on the abandoned path after 3 * PTO we trigger a transport error.
690        // In any case, we completely discard the path after 6 * PTO whether we receive a
691        // PATH_ABANDON or not.
692        self.timers.set(
693            Timer::PerPath(path_id, PathTimer::DiscardPath),
694            extended_end,
695            self.qlog.with_time(now),
696        );
697        Ok(())
698    }
699
700    /// Gets the [`PathData`] for a known [`PathId`].
701    ///
702    /// Will panic if the path_id does not reference any known path.
703    #[track_caller]
704    fn path_data(&self, path_id: PathId) -> &PathData {
705        if let Some(data) = self.paths.get(&path_id) {
706            &data.data
707        } else {
708            panic!(
709                "unknown path: {path_id}, currently known paths: {:?}",
710                self.paths.keys().collect::<Vec<_>>()
711            );
712        }
713    }
714
715    /// Gets a reference to the [`PathData`] for a [`PathId`]
716    fn path(&self, path_id: PathId) -> Option<&PathData> {
717        self.paths.get(&path_id).map(|path_state| &path_state.data)
718    }
719
720    /// Gets a mutable reference to the [`PathData`] for a [`PathId`]
721    fn path_mut(&mut self, path_id: PathId) -> Option<&mut PathData> {
722        self.paths
723            .get_mut(&path_id)
724            .map(|path_state| &mut path_state.data)
725    }
726
727    /// Returns all known paths.
728    ///
729    /// There is no guarantee any of these paths are open or usable.
730    pub fn paths(&self) -> Vec<PathId> {
731        self.paths.keys().copied().collect()
732    }
733
734    /// Gets the local [`PathStatus`] for a known [`PathId`]
735    pub fn path_status(&self, path_id: PathId) -> Result<PathStatus, ClosedPath> {
736        self.path(path_id)
737            .map(PathData::local_status)
738            .ok_or(ClosedPath { _private: () })
739    }
740
741    /// Returns the path's remote socket address
742    pub fn path_remote_address(&self, path_id: PathId) -> Result<SocketAddr, ClosedPath> {
743        self.path(path_id)
744            .map(|path| path.remote)
745            .ok_or(ClosedPath { _private: () })
746    }
747
748    /// Sets the [`PathStatus`] for a known [`PathId`]
749    ///
750    /// Returns the previous path status on success.
751    pub fn set_path_status(
752        &mut self,
753        path_id: PathId,
754        status: PathStatus,
755    ) -> Result<PathStatus, SetPathStatusError> {
756        if !self.is_multipath_negotiated() {
757            return Err(SetPathStatusError::MultipathNotNegotiated);
758        }
759        let path = self
760            .path_mut(path_id)
761            .ok_or(SetPathStatusError::ClosedPath)?;
762        let prev = match path.status.local_update(status) {
763            Some(prev) => {
764                self.spaces[SpaceId::Data]
765                    .pending
766                    .path_status
767                    .insert(path_id);
768                prev
769            }
770            None => path.local_status(),
771        };
772        Ok(prev)
773    }
774
775    /// Returns the remote path status
776    // TODO(flub): Probably should also be some kind of path event?  Not even sure if I like
777    //    this as an API, but for now it allows me to write a test easily.
778    // TODO(flub): Technically this should be a Result<Option<PathSTatus>>?
779    pub fn remote_path_status(&self, path_id: PathId) -> Option<PathStatus> {
780        self.path(path_id).and_then(|path| path.remote_status())
781    }
782
783    /// Sets the max_idle_timeout for a specific path
784    ///
785    /// See [`TransportConfig::default_path_max_idle_timeout`] for details.
786    ///
787    /// Returns the previous value of the setting.
788    pub fn set_path_max_idle_timeout(
789        &mut self,
790        path_id: PathId,
791        timeout: Option<Duration>,
792    ) -> Result<Option<Duration>, ClosedPath> {
793        let path = self
794            .paths
795            .get_mut(&path_id)
796            .ok_or(ClosedPath { _private: () })?;
797        Ok(std::mem::replace(&mut path.data.idle_timeout, timeout))
798    }
799
800    /// Sets the keep_alive_interval for a specific path
801    ///
802    /// See [`TransportConfig::default_path_keep_alive_interval`] for details.
803    ///
804    /// Returns the previous value of the setting.
805    pub fn set_path_keep_alive_interval(
806        &mut self,
807        path_id: PathId,
808        interval: Option<Duration>,
809    ) -> Result<Option<Duration>, ClosedPath> {
810        let path = self
811            .paths
812            .get_mut(&path_id)
813            .ok_or(ClosedPath { _private: () })?;
814        Ok(std::mem::replace(&mut path.data.keep_alive, interval))
815    }
816
817    /// Gets the [`PathData`] for a known [`PathId`].
818    ///
819    /// Will panic if the path_id does not reference any known path.
820    #[track_caller]
821    fn path_data_mut(&mut self, path_id: PathId) -> &mut PathData {
822        &mut self.paths.get_mut(&path_id).expect("known path").data
823    }
824
825    fn ensure_path(
826        &mut self,
827        path_id: PathId,
828        remote: SocketAddr,
829        now: Instant,
830        pn: Option<u64>,
831    ) -> &mut PathData {
832        let vacant_entry = match self.paths.entry(path_id) {
833            btree_map::Entry::Vacant(vacant_entry) => vacant_entry,
834            btree_map::Entry::Occupied(occupied_entry) => {
835                return &mut occupied_entry.into_mut().data;
836            }
837        };
838
839        // TODO(matheus23): Add back short-circuiting path.validated = true, if we know that the
840        // path's four-tuple was already validated.
841        debug!(%path_id, ?remote, "path added");
842        let peer_max_udp_payload_size =
843            u16::try_from(self.peer_params.max_udp_payload_size.into_inner()).unwrap_or(u16::MAX);
844        self.path_counter = self.path_counter.wrapping_add(1);
845        let mut data = PathData::new(
846            remote,
847            self.allow_mtud,
848            Some(peer_max_udp_payload_size),
849            self.path_counter,
850            now,
851            &self.config,
852        );
853
854        let pto = self.ack_frequency.max_ack_delay_for_pto() + data.rtt.pto_base();
855        self.timers.set(
856            Timer::PerPath(path_id, PathTimer::PathOpen),
857            now + 3 * pto,
858            self.qlog.with_time(now),
859        );
860
861        // for the path to be opened we need to send a packet on the path. Sending a challenge
862        // guarantees this
863        data.send_new_challenge = true;
864
865        let path = vacant_entry.insert(PathState { data, prev: None });
866
867        let mut pn_space = spaces::PacketNumberSpace::new(now, SpaceId::Data, &mut self.rng);
868        if let Some(pn) = pn {
869            pn_space.dedup.insert(pn);
870        }
871        self.spaces[SpaceId::Data]
872            .number_spaces
873            .insert(path_id, pn_space);
874        self.qlog.emit_tuple_assigned(path_id, remote, now);
875        &mut path.data
876    }
877
878    /// Returns packets to transmit
879    ///
880    /// Connections should be polled for transmit after:
881    /// - the application performed some I/O on the connection
882    /// - a call was made to `handle_event`
883    /// - a call was made to `handle_timeout`
884    ///
885    /// `max_datagrams` specifies how many datagrams can be returned inside a
886    /// single Transmit using GSO. This must be at least 1.
887    #[must_use]
888    pub fn poll_transmit(
889        &mut self,
890        now: Instant,
891        max_datagrams: usize,
892        buf: &mut Vec<u8>,
893    ) -> Option<Transmit> {
894        if let Some(probing) = self
895            .iroh_hp
896            .server_side_mut()
897            .ok()
898            .and_then(iroh_hp::ServerState::next_probe)
899        {
900            let destination = probing.remote();
901            trace!(%destination, "RAND_DATA packet");
902            let token: u64 = self.rng.random();
903            buf.put_u64(token);
904            probing.finish(token);
905            return Some(Transmit {
906                destination,
907                ecn: None,
908                size: 8,
909                segment_size: None,
910                src_ip: None,
911            });
912        }
913
914        assert!(max_datagrams != 0);
915        let max_datagrams = match self.config.enable_segmentation_offload {
916            false => 1,
917            true => max_datagrams,
918        };
919
920        // Each call to poll_transmit can only send datagrams to one destination, because
921        // all datagrams in a GSO batch are for the same destination.  Therefore only
922        // datagrams for one Path ID are produced for each poll_transmit call.
923
924        // TODO(flub): this is wishful thinking and not actually implemented, but perhaps it
925        //   should be:
926
927        // First, if we have to send a close, select a path for that.
928        // Next, all paths that have a PATH_CHALLENGE or PATH_RESPONSE pending.
929
930        // For all, open, validated and AVAILABLE paths:
931        // - Is the path congestion blocked or pacing blocked?
932        // - call maybe_queue_ to ensure a tail-loss probe would be sent?
933        // - do we need to send a close message?
934        // - call can_send
935        // Once there's nothing more to send on the AVAILABLE paths, do the same for BACKUP paths
936
937        // Check whether we need to send a close message
938        let close = match self.state.as_type() {
939            StateType::Drained => {
940                self.app_limited = true;
941                return None;
942            }
943            StateType::Draining | StateType::Closed => {
944                // self.close is only reset once the associated packet had been
945                // encoded successfully
946                if !self.close {
947                    self.app_limited = true;
948                    return None;
949                }
950                true
951            }
952            _ => false,
953        };
954
955        // Check whether we need to send an ACK_FREQUENCY frame
956        if let Some(config) = &self.config.ack_frequency_config {
957            let rtt = self
958                .paths
959                .values()
960                .map(|p| p.data.rtt.get())
961                .min()
962                .expect("one path exists");
963            self.spaces[SpaceId::Data].pending.ack_frequency = self
964                .ack_frequency
965                .should_send_ack_frequency(rtt, config, &self.peer_params)
966                && self.highest_space == SpaceId::Data
967                && self.peer_supports_ack_frequency();
968        }
969
970        // Whether this packet can be coalesced with another one in the same datagram.
971        let mut coalesce = true;
972
973        // Whether the last packet in the datagram must be padded so the datagram takes up
974        // to at least MIN_INITIAL_SIZE, or to the maximum segment size if this is smaller.
975        let mut pad_datagram = PadDatagram::No;
976
977        // Whether congestion control stopped the next packet from being sent. Further
978        // packets could still be built, as e.g. tail-loss probes are not congestion
979        // limited.
980        let mut congestion_blocked = false;
981
982        // The packet number of the last built packet.
983        let mut last_packet_number = None;
984
985        let mut path_id = *self.paths.first_key_value().expect("one path must exist").0;
986
987        // If there is any open, validated and available path we only want to send frames to
988        // any backup path that must be sent on that backup path exclusively.
989        let have_available_path = self.paths.iter().any(|(id, path)| {
990            path.data.validated
991                && path.data.local_status() == PathStatus::Available
992                && self.rem_cids.contains_key(id)
993        });
994
995        // Setup for the first path_id
996        let mut transmit = TransmitBuf::new(
997            buf,
998            max_datagrams,
999            self.path_data(path_id).current_mtu().into(),
1000        );
1001        if let Some(challenge) = self.send_prev_path_challenge(now, &mut transmit, path_id) {
1002            return Some(challenge);
1003        }
1004        let mut space_id = match path_id {
1005            PathId::ZERO => SpaceId::Initial,
1006            _ => SpaceId::Data,
1007        };
1008
1009        loop {
1010            // check if there is at least one active CID to use for sending
1011            let Some(remote_cid) = self.rem_cids.get(&path_id).map(CidQueue::active) else {
1012                let err = PathError::RemoteCidsExhausted;
1013                if !self.abandoned_paths.contains(&path_id) {
1014                    debug!(?err, %path_id, "no active CID for path");
1015                    self.events.push_back(Event::Path(PathEvent::LocallyClosed {
1016                        id: path_id,
1017                        error: err,
1018                    }));
1019                    // Locally we should have refused to open this path, the remote should
1020                    // have given us CIDs for this path before opening it.  So we can always
1021                    // abandon this here.
1022                    self.close_path(
1023                        now,
1024                        path_id,
1025                        TransportErrorCode::NO_CID_AVAILABLE_FOR_PATH.into(),
1026                    )
1027                    .ok();
1028                    self.spaces[SpaceId::Data]
1029                        .pending
1030                        .path_cids_blocked
1031                        .push(path_id);
1032                } else {
1033                    trace!(%path_id, "remote CIDs retired for abandoned path");
1034                }
1035
1036                match self.paths.keys().find(|&&next| next > path_id) {
1037                    Some(next_path_id) => {
1038                        // See if this next path can send anything.
1039                        path_id = *next_path_id;
1040                        space_id = SpaceId::Data;
1041
1042                        // update per path state
1043                        transmit.set_segment_size(self.path_data(path_id).current_mtu().into());
1044                        if let Some(challenge) =
1045                            self.send_prev_path_challenge(now, &mut transmit, path_id)
1046                        {
1047                            return Some(challenge);
1048                        }
1049
1050                        continue;
1051                    }
1052                    None => {
1053                        // Nothing more to send.
1054                        trace!(
1055                            ?space_id,
1056                            %path_id,
1057                            "no CIDs to send on path, no more paths"
1058                        );
1059                        break;
1060                    }
1061                }
1062            };
1063
1064            // Determine if anything can be sent in this packet number space (SpaceId +
1065            // PathId).
1066            let max_packet_size = if transmit.datagram_remaining_mut() > 0 {
1067                // We are trying to coalesce another packet into this datagram.
1068                transmit.datagram_remaining_mut()
1069            } else {
1070                // A new datagram needs to be started.
1071                transmit.segment_size()
1072            };
1073            let can_send = self.space_can_send(space_id, path_id, max_packet_size, close);
1074            let path_should_send = {
1075                let path_exclusive_only = space_id == SpaceId::Data
1076                    && have_available_path
1077                    && self.path_data(path_id).local_status() == PathStatus::Backup;
1078                let path_should_send = if path_exclusive_only {
1079                    can_send.path_exclusive
1080                } else {
1081                    !can_send.is_empty()
1082                };
1083                let needs_loss_probe = self.spaces[space_id].for_path(path_id).loss_probes > 0;
1084                path_should_send || needs_loss_probe || can_send.close
1085            };
1086
1087            if !path_should_send && space_id < SpaceId::Data {
1088                if self.spaces[space_id].crypto.is_some() {
1089                    trace!(?space_id, %path_id, "nothing to send in space");
1090                }
1091                space_id = space_id.next();
1092                continue;
1093            }
1094
1095            let send_blocked = if path_should_send && transmit.datagram_remaining_mut() == 0 {
1096                // Only check congestion control if a new datagram is needed.
1097                self.path_congestion_check(space_id, path_id, &transmit, &can_send, now)
1098            } else {
1099                PathBlocked::No
1100            };
1101            if send_blocked != PathBlocked::No {
1102                trace!(?space_id, %path_id, ?send_blocked, "congestion blocked");
1103                congestion_blocked = true;
1104            }
1105            if send_blocked != PathBlocked::No && space_id < SpaceId::Data {
1106                // Higher spaces might still have tail-loss probes to send, which are not
1107                // congestion blocked.
1108                space_id = space_id.next();
1109                continue;
1110            }
1111            if !path_should_send || send_blocked != PathBlocked::No {
1112                // Nothing more to send on this path, check the next path if possible.
1113
1114                // If there are any datagrams in the transmit, packets for another path can
1115                // not be built.
1116                if transmit.num_datagrams() > 0 {
1117                    break;
1118                }
1119
1120                match self.paths.keys().find(|&&next| next > path_id) {
1121                    Some(next_path_id) => {
1122                        // See if this next path can send anything.
1123                        trace!(
1124                            ?space_id,
1125                            %path_id,
1126                            %next_path_id,
1127                            "nothing to send on path"
1128                        );
1129                        path_id = *next_path_id;
1130                        space_id = SpaceId::Data;
1131
1132                        // update per path state
1133                        transmit.set_segment_size(self.path_data(path_id).current_mtu().into());
1134                        if let Some(challenge) =
1135                            self.send_prev_path_challenge(now, &mut transmit, path_id)
1136                        {
1137                            return Some(challenge);
1138                        }
1139
1140                        continue;
1141                    }
1142                    None => {
1143                        // Nothing more to send.
1144                        trace!(
1145                            ?space_id,
1146                            %path_id,
1147                            next_path_id=?None::<PathId>,
1148                            "nothing to send on path"
1149                        );
1150                        break;
1151                    }
1152                }
1153            }
1154
1155            // If the datagram is full, we need to start a new one.
1156            if transmit.datagram_remaining_mut() == 0 {
1157                if transmit.num_datagrams() >= transmit.max_datagrams() {
1158                    // No more datagrams allowed
1159                    break;
1160                }
1161
1162                match self.spaces[space_id].for_path(path_id).loss_probes {
1163                    0 => transmit.start_new_datagram(),
1164                    _ => {
1165                        // We need something to send for a tail-loss probe.
1166                        let request_immediate_ack =
1167                            space_id == SpaceId::Data && self.peer_supports_ack_frequency();
1168                        self.spaces[space_id].maybe_queue_probe(
1169                            path_id,
1170                            request_immediate_ack,
1171                            &self.streams,
1172                        );
1173
1174                        self.spaces[space_id].for_path(path_id).loss_probes -= 1;
1175
1176                        // Clamp the datagram to at most the minimum MTU to ensure that loss
1177                        // probes can get through and enable recovery even if the path MTU
1178                        // has shrank unexpectedly.
1179                        transmit.start_new_datagram_with_size(std::cmp::min(
1180                            usize::from(INITIAL_MTU),
1181                            transmit.segment_size(),
1182                        ));
1183                    }
1184                }
1185                trace!(count = transmit.num_datagrams(), "new datagram started");
1186                coalesce = true;
1187                pad_datagram = PadDatagram::No;
1188            }
1189
1190            // If coalescing another packet into the existing datagram, there should
1191            // still be enough space for a whole packet.
1192            if transmit.datagram_start_offset() < transmit.len() {
1193                debug_assert!(transmit.datagram_remaining_mut() >= MIN_PACKET_SPACE);
1194            }
1195
1196            //
1197            // From here on, we've determined that a packet will definitely be sent.
1198            //
1199
1200            if self.spaces[SpaceId::Initial].crypto.is_some()
1201                && space_id == SpaceId::Handshake
1202                && self.side.is_client()
1203            {
1204                // A client stops both sending and processing Initial packets when it
1205                // sends its first Handshake packet.
1206                self.discard_space(now, SpaceId::Initial);
1207            }
1208            if let Some(ref mut prev) = self.prev_crypto {
1209                prev.update_unacked = false;
1210            }
1211
1212            let mut qlog = QlogSentPacket::default();
1213            let mut builder = PacketBuilder::new(
1214                now,
1215                space_id,
1216                path_id,
1217                remote_cid,
1218                &mut transmit,
1219                can_send.other,
1220                self,
1221                &mut qlog,
1222            )?;
1223            last_packet_number = Some(builder.exact_number);
1224            coalesce = coalesce && !builder.short_header;
1225
1226            if space_id == SpaceId::Initial && (self.side.is_client() || can_send.other) {
1227                // https://www.rfc-editor.org/rfc/rfc9000.html#section-14.1
1228                pad_datagram |= PadDatagram::ToMinMtu;
1229            }
1230            if space_id == SpaceId::Data && self.config.pad_to_mtu {
1231                pad_datagram |= PadDatagram::ToSegmentSize;
1232            }
1233
1234            if can_send.close {
1235                trace!("sending CONNECTION_CLOSE");
1236                // Encode ACKs before the ConnectionClose message, to give the receiver
1237                // a better approximate on what data has been processed. This is
1238                // especially important with ack delay, since the peer might not
1239                // have gotten any other ACK for the data earlier on.
1240                let mut sent_frames = SentFrames::default();
1241                let is_multipath_negotiated = self.is_multipath_negotiated();
1242                for path_id in self.spaces[space_id]
1243                    .number_spaces
1244                    .iter()
1245                    .filter(|(_, pns)| !pns.pending_acks.ranges().is_empty())
1246                    .map(|(&path_id, _)| path_id)
1247                    .collect::<Vec<_>>()
1248                {
1249                    Self::populate_acks(
1250                        now,
1251                        self.receiving_ecn,
1252                        &mut sent_frames,
1253                        path_id,
1254                        space_id,
1255                        &mut self.spaces[space_id],
1256                        is_multipath_negotiated,
1257                        &mut builder.frame_space_mut(),
1258                        &mut self.stats,
1259                        &mut qlog,
1260                    );
1261                }
1262
1263                // Since there only 64 ACK frames there will always be enough space
1264                // to encode the ConnectionClose frame too. However we still have the
1265                // check here to prevent crashes if something changes.
1266                debug_assert!(
1267                    builder.frame_space_remaining() > frame::ConnectionClose::SIZE_BOUND,
1268                    "ACKs should leave space for ConnectionClose"
1269                );
1270                if frame::ConnectionClose::SIZE_BOUND < builder.frame_space_remaining() {
1271                    let max_frame_size = builder.frame_space_remaining();
1272                    match self.state.as_type() {
1273                        StateType::Closed => {
1274                            let reason: Close =
1275                                self.state.as_closed().expect("checked").clone().into();
1276                            if space_id == SpaceId::Data || reason.is_transport_layer() {
1277                                reason.encode(&mut builder.frame_space_mut(), max_frame_size);
1278                                qlog.frame(&Frame::Close(reason));
1279                            } else {
1280                                let frame = frame::ConnectionClose {
1281                                    error_code: TransportErrorCode::APPLICATION_ERROR,
1282                                    frame_type: None,
1283                                    reason: Bytes::new(),
1284                                };
1285                                frame.encode(&mut builder.frame_space_mut(), max_frame_size);
1286                                qlog.frame(&Frame::Close(frame::Close::Connection(frame)));
1287                            }
1288                        }
1289                        StateType::Draining => {
1290                            let frame = frame::ConnectionClose {
1291                                error_code: TransportErrorCode::NO_ERROR,
1292                                frame_type: None,
1293                                reason: Bytes::new(),
1294                            };
1295                            frame.encode(&mut builder.frame_space_mut(), max_frame_size);
1296                            qlog.frame(&Frame::Close(frame::Close::Connection(frame)));
1297                        }
1298                        _ => unreachable!(
1299                            "tried to make a close packet when the connection wasn't closed"
1300                        ),
1301                    };
1302                }
1303                builder.finish_and_track(now, self, path_id, sent_frames, pad_datagram, qlog);
1304                if space_id == self.highest_space {
1305                    // Don't send another close packet. Even with multipath we only send
1306                    // CONNECTION_CLOSE on a single path since we expect our paths to work.
1307                    self.close = false;
1308                    // `CONNECTION_CLOSE` is the final packet
1309                    break;
1310                } else {
1311                    // Send a close frame in every possible space for robustness, per
1312                    // RFC9000 "Immediate Close during the Handshake". Don't bother trying
1313                    // to send anything else.
1314                    space_id = space_id.next();
1315                    continue;
1316                }
1317            }
1318
1319            // Send an off-path PATH_RESPONSE. Prioritized over on-path data to ensure that
1320            // path validation can occur while the link is saturated.
1321            if space_id == SpaceId::Data && builder.buf.num_datagrams() == 1 {
1322                let path = self.path_data_mut(path_id);
1323                if let Some((token, remote)) = path.path_responses.pop_off_path(path.remote) {
1324                    // TODO(flub): We need to use the right CID!  We shouldn't use the same
1325                    //    CID as the current active one for the path.  Though see also
1326                    //    https://github.com/quinn-rs/quinn/issues/2184
1327                    let response = frame::PathResponse(token);
1328                    trace!(%response, "(off-path)");
1329                    builder.frame_space_mut().write(response);
1330                    qlog.frame(&Frame::PathResponse(response));
1331                    self.stats.frame_tx.path_response += 1;
1332                    builder.finish_and_track(
1333                        now,
1334                        self,
1335                        path_id,
1336                        SentFrames {
1337                            non_retransmits: true,
1338                            ..SentFrames::default()
1339                        },
1340                        PadDatagram::ToMinMtu,
1341                        qlog,
1342                    );
1343                    self.stats.udp_tx.on_sent(1, transmit.len());
1344                    return Some(Transmit {
1345                        destination: remote,
1346                        size: transmit.len(),
1347                        ecn: None,
1348                        segment_size: None,
1349                        src_ip: self.local_ip,
1350                    });
1351                }
1352            }
1353
1354            let sent_frames = {
1355                let path_exclusive_only = have_available_path
1356                    && self.path_data(path_id).local_status() == PathStatus::Backup;
1357                let pn = builder.exact_number;
1358                self.populate_packet(
1359                    now,
1360                    space_id,
1361                    path_id,
1362                    path_exclusive_only,
1363                    &mut builder.frame_space_mut(),
1364                    pn,
1365                    &mut qlog,
1366                )
1367            };
1368
1369            // ACK-only packets should only be sent when explicitly allowed. If we write them due to
1370            // any other reason, there is a bug which leads to one component announcing write
1371            // readiness while not writing any data. This degrades performance. The condition is
1372            // only checked if the full MTU is available and when potentially large fixed-size
1373            // frames aren't queued, so that lack of space in the datagram isn't the reason for just
1374            // writing ACKs.
1375            debug_assert!(
1376                !(sent_frames.is_ack_only(&self.streams)
1377                    && !can_send.acks
1378                    && can_send.other
1379                    && builder.buf.segment_size()
1380                        == self.path_data(path_id).current_mtu() as usize
1381                    && self.datagrams.outgoing.is_empty()),
1382                "SendableFrames was {can_send:?}, but only ACKs have been written"
1383            );
1384            if sent_frames.requires_padding {
1385                pad_datagram |= PadDatagram::ToMinMtu;
1386            }
1387
1388            for (path_id, _pn) in sent_frames.largest_acked.iter() {
1389                self.spaces[space_id]
1390                    .for_path(*path_id)
1391                    .pending_acks
1392                    .acks_sent();
1393                self.timers.stop(
1394                    Timer::PerPath(*path_id, PathTimer::MaxAckDelay),
1395                    self.qlog.with_time(now),
1396                );
1397            }
1398
1399            // Now we need to finish the packet.  Before we do so we need to know if we will
1400            // be coalescing the next packet into this one, or will be ending the datagram
1401            // as well.  Because if this is the last packet in the datagram more padding
1402            // might be needed because of the packet type, or to fill the GSO segment size.
1403
1404            // Are we allowed to coalesce AND is there enough space for another *packet* in
1405            // this datagram AND is there another packet to send in this or the next space?
1406            if coalesce
1407                && builder
1408                    .buf
1409                    .datagram_remaining_mut()
1410                    .saturating_sub(builder.predict_packet_end())
1411                    > MIN_PACKET_SPACE
1412                && self
1413                    .next_send_space(space_id, path_id, builder.buf, close)
1414                    .is_some()
1415            {
1416                // We can append/coalesce the next packet into the current
1417                // datagram. Finish the current packet without adding extra padding.
1418                builder.finish_and_track(now, self, path_id, sent_frames, PadDatagram::No, qlog);
1419            } else {
1420                // We need a new datagram for the next packet.  Finish the current
1421                // packet with padding.
1422                if builder.buf.num_datagrams() > 1 && matches!(pad_datagram, PadDatagram::No) {
1423                    // If too many padding bytes would be required to continue the
1424                    // GSO batch after this packet, end the GSO batch here. Ensures
1425                    // that fixed-size frames with heterogeneous sizes
1426                    // (e.g. application datagrams) won't inadvertently waste large
1427                    // amounts of bandwidth. The exact threshold is a bit arbitrary
1428                    // and might benefit from further tuning, though there's no
1429                    // universally optimal value.
1430                    const MAX_PADDING: usize = 32;
1431                    if builder.buf.datagram_remaining_mut()
1432                        > builder.predict_packet_end() + MAX_PADDING
1433                    {
1434                        trace!(
1435                            "GSO truncated by demand for {} padding bytes",
1436                            builder.buf.datagram_remaining_mut() - builder.predict_packet_end()
1437                        );
1438                        builder.finish_and_track(
1439                            now,
1440                            self,
1441                            path_id,
1442                            sent_frames,
1443                            PadDatagram::No,
1444                            qlog,
1445                        );
1446                        break;
1447                    }
1448
1449                    // Pad the current datagram to GSO segment size so it can be
1450                    // included in the GSO batch.
1451                    builder.finish_and_track(
1452                        now,
1453                        self,
1454                        path_id,
1455                        sent_frames,
1456                        PadDatagram::ToSegmentSize,
1457                        qlog,
1458                    );
1459                } else {
1460                    builder.finish_and_track(now, self, path_id, sent_frames, pad_datagram, qlog);
1461                }
1462                if transmit.num_datagrams() == 1 {
1463                    transmit.clip_datagram_size();
1464                }
1465            }
1466        }
1467
1468        if let Some(last_packet_number) = last_packet_number {
1469            // Note that when sending in multiple packet spaces the last packet number will
1470            // be the one from the highest packet space.
1471            self.path_data_mut(path_id).congestion.on_sent(
1472                now,
1473                transmit.len() as u64,
1474                last_packet_number,
1475            );
1476        }
1477
1478        self.qlog.emit_recovery_metrics(
1479            path_id,
1480            &mut self.paths.get_mut(&path_id).unwrap().data,
1481            now,
1482        );
1483
1484        self.app_limited = transmit.is_empty() && !congestion_blocked;
1485
1486        // Send MTU probe if necessary
1487        if transmit.is_empty() && self.state.is_established() {
1488            // MTU probing happens only in Data space.
1489            let space_id = SpaceId::Data;
1490            path_id = *self.paths.first_key_value().expect("one path must exist").0;
1491            let probe_data = loop {
1492                // We MTU probe all paths for which all of the following is true:
1493                // - We have an active destination CID for the path.
1494                // - The remote address *and* path are validated.
1495                // - The path is not abandoned.
1496                // - The MTU Discovery subsystem wants to probe the path.
1497                let active_cid = self.rem_cids.get(&path_id).map(CidQueue::active);
1498                let eligible = self.path_data(path_id).validated
1499                    && !self.path_data(path_id).is_validating_path()
1500                    && !self.abandoned_paths.contains(&path_id);
1501                let probe_size = eligible
1502                    .then(|| {
1503                        let next_pn = self.spaces[space_id].for_path(path_id).peek_tx_number();
1504                        self.path_data_mut(path_id).mtud.poll_transmit(now, next_pn)
1505                    })
1506                    .flatten();
1507                match (active_cid, probe_size) {
1508                    (Some(active_cid), Some(probe_size)) => {
1509                        // Let's send an MTUD probe!
1510                        break Some((active_cid, probe_size));
1511                    }
1512                    _ => {
1513                        // Find the next path to check if it needs an MTUD probe.
1514                        match self.paths.keys().find(|&&next| next > path_id) {
1515                            Some(next) => {
1516                                path_id = *next;
1517                                continue;
1518                            }
1519                            None => break None,
1520                        }
1521                    }
1522                }
1523            };
1524            if let Some((active_cid, probe_size)) = probe_data {
1525                // We are definitely sending a DPLPMTUD probe.
1526                debug_assert_eq!(transmit.num_datagrams(), 0);
1527                transmit.start_new_datagram_with_size(probe_size as usize);
1528
1529                let mut qlog = QlogSentPacket::default();
1530                let mut builder = PacketBuilder::new(
1531                    now,
1532                    space_id,
1533                    path_id,
1534                    active_cid,
1535                    &mut transmit,
1536                    true,
1537                    self,
1538                    &mut qlog,
1539                )?;
1540
1541                // We implement MTU probes as ping packets padded up to the probe size
1542                trace!(?probe_size, "writing MTUD probe");
1543                trace!("PING");
1544                builder.frame_space_mut().write(frame::FrameType::PING);
1545                qlog.frame(&Frame::Ping);
1546                self.stats.frame_tx.ping += 1;
1547
1548                // If supported by the peer, we want no delays to the probe's ACK
1549                if self.peer_supports_ack_frequency() {
1550                    trace!("IMMEDIATE_ACK");
1551                    builder
1552                        .frame_space_mut()
1553                        .write(frame::FrameType::IMMEDIATE_ACK);
1554                    self.stats.frame_tx.immediate_ack += 1;
1555                    qlog.frame(&Frame::ImmediateAck);
1556                }
1557
1558                let sent_frames = SentFrames {
1559                    non_retransmits: true,
1560                    ..Default::default()
1561                };
1562                builder.finish_and_track(
1563                    now,
1564                    self,
1565                    path_id,
1566                    sent_frames,
1567                    PadDatagram::ToSize(probe_size),
1568                    qlog,
1569                );
1570
1571                self.path_stats
1572                    .entry(path_id)
1573                    .or_default()
1574                    .sent_plpmtud_probes += 1;
1575            }
1576        }
1577
1578        if transmit.is_empty() {
1579            return None;
1580        }
1581
1582        let destination = self.path_data(path_id).remote;
1583        trace!(
1584            segment_size = transmit.segment_size(),
1585            last_datagram_len = transmit.len() % transmit.segment_size(),
1586            ?destination,
1587            "sending {} bytes in {} datagrams",
1588            transmit.len(),
1589            transmit.num_datagrams()
1590        );
1591        self.path_data_mut(path_id)
1592            .inc_total_sent(transmit.len() as u64);
1593
1594        self.stats
1595            .udp_tx
1596            .on_sent(transmit.num_datagrams() as u64, transmit.len());
1597
1598        Some(Transmit {
1599            destination,
1600            size: transmit.len(),
1601            ecn: if self.path_data(path_id).sending_ecn {
1602                Some(EcnCodepoint::Ect0)
1603            } else {
1604                None
1605            },
1606            segment_size: match transmit.num_datagrams() {
1607                1 => None,
1608                _ => Some(transmit.segment_size()),
1609            },
1610            src_ip: self.local_ip,
1611        })
1612    }
1613
1614    /// Returns the [`SpaceId`] of the next packet space which has data to send
1615    ///
1616    /// This takes into account the space available to frames in the next datagram.
1617    // TODO(flub): This duplication is not nice.
1618    fn next_send_space(
1619        &mut self,
1620        current_space_id: SpaceId,
1621        path_id: PathId,
1622        buf: &TransmitBuf<'_>,
1623        close: bool,
1624    ) -> Option<SpaceId> {
1625        // Number of bytes available for frames if this is a 1-RTT packet. We're guaranteed
1626        // to be able to send an individual frame at least this large in the next 1-RTT
1627        // packet. This could be generalized to support every space, but it's only needed to
1628        // handle large fixed-size frames, which only exist in 1-RTT (application
1629        // datagrams). We don't account for coalesced packets potentially occupying space
1630        // because frames can always spill into the next datagram.
1631        let mut space_id = current_space_id;
1632        loop {
1633            let can_send = self.space_can_send(space_id, path_id, buf.segment_size(), close);
1634            if !can_send.is_empty() || (close && self.spaces[space_id].crypto.is_some()) {
1635                return Some(space_id);
1636            }
1637            space_id = match space_id {
1638                SpaceId::Initial => SpaceId::Handshake,
1639                SpaceId::Handshake => SpaceId::Data,
1640                SpaceId::Data => break,
1641            }
1642        }
1643        None
1644    }
1645
1646    /// Checks if creating a new datagram would be blocked by congestion control
1647    fn path_congestion_check(
1648        &mut self,
1649        space_id: SpaceId,
1650        path_id: PathId,
1651        transmit: &TransmitBuf<'_>,
1652        can_send: &SendableFrames,
1653        now: Instant,
1654    ) -> PathBlocked {
1655        // Anti-amplification is only based on `total_sent`, which gets updated after
1656        // the transmit is sent. Therefore we pass the amount of bytes for datagrams
1657        // that are already created, as well as 1 byte for starting another datagram. If
1658        // there is any anti-amplification budget left, we always allow a full MTU to be
1659        // sent (see https://github.com/quinn-rs/quinn/issues/1082).
1660        if self.side().is_server()
1661            && self
1662                .path_data(path_id)
1663                .anti_amplification_blocked(transmit.len() as u64 + 1)
1664        {
1665            trace!(?space_id, %path_id, "blocked by anti-amplification");
1666            return PathBlocked::AntiAmplification;
1667        }
1668
1669        // Congestion control check.
1670        // Tail loss probes must not be blocked by congestion, or a deadlock could arise.
1671        let bytes_to_send = transmit.segment_size() as u64;
1672        let need_loss_probe = self.spaces[space_id].for_path(path_id).loss_probes > 0;
1673
1674        if can_send.other && !need_loss_probe && !can_send.close {
1675            let path = self.path_data(path_id);
1676            if path.in_flight.bytes + bytes_to_send >= path.congestion.window() {
1677                trace!(?space_id, %path_id, "blocked by congestion control");
1678                return PathBlocked::Congestion;
1679            }
1680        }
1681
1682        // Pacing check.
1683        if let Some(delay) = self.path_data_mut(path_id).pacing_delay(bytes_to_send, now) {
1684            self.timers.set(
1685                Timer::PerPath(path_id, PathTimer::Pacing),
1686                delay,
1687                self.qlog.with_time(now),
1688            );
1689            // Loss probes and CONNECTION_CLOSE should be subject to pacing, even though
1690            // they are not congestion controlled.
1691            trace!(?space_id, %path_id, "blocked by pacing");
1692            return PathBlocked::Pacing;
1693        }
1694
1695        PathBlocked::No
1696    }
1697
1698    /// Send PATH_CHALLENGE for a previous path if necessary
1699    ///
1700    /// QUIC-TRANSPORT section 9.3.3
1701    /// <https://www.rfc-editor.org/rfc/rfc9000.html#name-off-path-packet-forwarding>
1702    fn send_prev_path_challenge(
1703        &mut self,
1704        now: Instant,
1705        buf: &mut TransmitBuf<'_>,
1706        path_id: PathId,
1707    ) -> Option<Transmit> {
1708        let (prev_cid, prev_path) = self.paths.get_mut(&path_id)?.prev.as_mut()?;
1709        // TODO (matheus23): We could use !prev_path.is_validating() here instead to
1710        // (possibly) also re-send challenges when they get lost.
1711        if !prev_path.send_new_challenge {
1712            return None;
1713        };
1714        prev_path.send_new_challenge = false;
1715        let destination = prev_path.remote;
1716        let token = self.rng.random();
1717        let info = paths::SentChallengeInfo {
1718            sent_instant: now,
1719            remote: destination,
1720        };
1721        prev_path.challenges_sent.insert(token, info);
1722        debug_assert_eq!(
1723            self.highest_space,
1724            SpaceId::Data,
1725            "PATH_CHALLENGE queued without 1-RTT keys"
1726        );
1727        buf.start_new_datagram_with_size(MIN_INITIAL_SIZE as usize);
1728
1729        // Use the previous CID to avoid linking the new path with the previous path. We
1730        // don't bother accounting for possible retirement of that prev_cid because this is
1731        // sent once, immediately after migration, when the CID is known to be valid. Even
1732        // if a post-migration packet caused the CID to be retired, it's fair to pretend
1733        // this is sent first.
1734        debug_assert_eq!(buf.datagram_start_offset(), 0);
1735        let mut qlog = QlogSentPacket::default();
1736        let mut builder = PacketBuilder::new(
1737            now,
1738            SpaceId::Data,
1739            path_id,
1740            *prev_cid,
1741            buf,
1742            false,
1743            self,
1744            &mut qlog,
1745        )?;
1746        let challenge = frame::PathChallenge(token);
1747        trace!(%challenge, "validating previous path");
1748        qlog.frame(&Frame::PathChallenge(challenge));
1749        builder.frame_space_mut().write(challenge);
1750        self.stats.frame_tx.path_challenge += 1;
1751
1752        // An endpoint MUST expand datagrams that contain a PATH_CHALLENGE frame
1753        // to at least the smallest allowed maximum datagram size of 1200 bytes,
1754        // unless the anti-amplification limit for the path does not permit
1755        // sending a datagram of this size
1756        builder.pad_to(MIN_INITIAL_SIZE);
1757
1758        builder.finish(self, now, qlog);
1759        self.stats.udp_tx.on_sent(1, buf.len());
1760
1761        Some(Transmit {
1762            destination,
1763            size: buf.len(),
1764            ecn: None,
1765            segment_size: None,
1766            src_ip: self.local_ip,
1767        })
1768    }
1769
1770    /// Indicate what types of frames are ready to send for the given space
1771    ///
1772    /// *packet_size* is the number of bytes available to build the next packet.  *close*
1773    /// *indicates whether a CONNECTION_CLOSE frame needs to be sent.
1774    fn space_can_send(
1775        &mut self,
1776        space_id: SpaceId,
1777        path_id: PathId,
1778        packet_size: usize,
1779        close: bool,
1780    ) -> SendableFrames {
1781        let pn = self.spaces[SpaceId::Data]
1782            .for_path(path_id)
1783            .peek_tx_number();
1784        let frame_space_1rtt = packet_size.saturating_sub(self.predict_1rtt_overhead(pn, path_id));
1785        if self.spaces[space_id].crypto.is_none()
1786            && (space_id != SpaceId::Data
1787                || self.zero_rtt_crypto.is_none()
1788                || self.side.is_server())
1789        {
1790            // No keys available for this space
1791            return SendableFrames::empty();
1792        }
1793        let mut can_send = self.spaces[space_id].can_send(path_id, &self.streams);
1794        if space_id == SpaceId::Data {
1795            can_send |= self.can_send_1rtt(path_id, frame_space_1rtt);
1796        }
1797
1798        can_send.close = close && self.spaces[space_id].crypto.is_some();
1799
1800        can_send
1801    }
1802
1803    /// Process `ConnectionEvent`s generated by the associated `Endpoint`
1804    ///
1805    /// Will execute protocol logic upon receipt of a connection event, in turn preparing signals
1806    /// (including application `Event`s, `EndpointEvent`s and outgoing datagrams) that should be
1807    /// extracted through the relevant methods.
1808    pub fn handle_event(&mut self, event: ConnectionEvent) {
1809        use ConnectionEventInner::*;
1810        match event.0 {
1811            Datagram(DatagramConnectionEvent {
1812                now,
1813                remote,
1814                path_id,
1815                ecn,
1816                first_decode,
1817                remaining,
1818            }) => {
1819                let span = trace_span!("pkt", %path_id);
1820                let _guard = span.enter();
1821                // If this packet could initiate a migration and we're a client or a server that
1822                // forbids migration, drop the datagram. This could be relaxed to heuristically
1823                // permit NAT-rebinding-like migration.
1824                if let Some(known_remote) = self.path(path_id).map(|path| path.remote) {
1825                    if remote != known_remote && !self.side.remote_may_migrate(&self.state) {
1826                        trace!(
1827                            %path_id,
1828                            ?remote,
1829                            path_remote = ?self.path(path_id).map(|p| p.remote),
1830                            "discarding packet from unrecognized peer"
1831                        );
1832                        return;
1833                    }
1834                }
1835
1836                let was_anti_amplification_blocked = self
1837                    .path(path_id)
1838                    .map(|path| path.anti_amplification_blocked(1))
1839                    .unwrap_or(true); // if we don't know about this path it's eagerly considered as unvalidated
1840                // TODO(@divma): revisit this
1841
1842                self.stats.udp_rx.datagrams += 1;
1843                self.stats.udp_rx.bytes += first_decode.len() as u64;
1844                let data_len = first_decode.len();
1845
1846                self.handle_decode(now, remote, path_id, ecn, first_decode);
1847                // The current `path` might have changed inside `handle_decode` since the packet
1848                // could have triggered a migration. The packet might also belong to an unknown
1849                // path and have been rejected. Make sure the data received is accounted for the
1850                // most recent path by accessing `path` after `handle_decode`.
1851                if let Some(path) = self.path_mut(path_id) {
1852                    path.inc_total_recvd(data_len as u64);
1853                }
1854
1855                if let Some(data) = remaining {
1856                    self.stats.udp_rx.bytes += data.len() as u64;
1857                    self.handle_coalesced(now, remote, path_id, ecn, data);
1858                }
1859
1860                if let Some(path) = self.paths.get_mut(&path_id) {
1861                    self.qlog
1862                        .emit_recovery_metrics(path_id, &mut path.data, now);
1863                }
1864
1865                if was_anti_amplification_blocked {
1866                    // A prior attempt to set the loss detection timer may have failed due to
1867                    // anti-amplification, so ensure it's set now. Prevents a handshake deadlock if
1868                    // the server's first flight is lost.
1869                    self.set_loss_detection_timer(now, path_id);
1870                }
1871            }
1872            NewIdentifiers(ids, now, cid_len, cid_lifetime) => {
1873                let path_id = ids.first().map(|issued| issued.path_id).unwrap_or_default();
1874                debug_assert!(ids.iter().all(|issued| issued.path_id == path_id));
1875                let cid_state = self
1876                    .local_cid_state
1877                    .entry(path_id)
1878                    .or_insert_with(|| CidState::new(cid_len, cid_lifetime, now, 0));
1879                cid_state.new_cids(&ids, now);
1880
1881                ids.into_iter().rev().for_each(|frame| {
1882                    self.spaces[SpaceId::Data].pending.new_cids.push(frame);
1883                });
1884                // Always update Timer::PushNewCid
1885                self.reset_cid_retirement(now);
1886            }
1887        }
1888    }
1889
1890    /// Process timer expirations
1891    ///
1892    /// Executes protocol logic, potentially preparing signals (including application `Event`s,
1893    /// `EndpointEvent`s and outgoing datagrams) that should be extracted through the relevant
1894    /// methods.
1895    ///
1896    /// It is most efficient to call this immediately after the system clock reaches the latest
1897    /// `Instant` that was output by `poll_timeout`; however spurious extra calls will simply
1898    /// no-op and therefore are safe.
1899    pub fn handle_timeout(&mut self, now: Instant) {
1900        while let Some((timer, _time)) = self.timers.expire_before(now, &self.qlog) {
1901            // TODO(@divma): remove `at` when the unicorn is born
1902            trace!(?timer, at=?now, "timeout");
1903            match timer {
1904                Timer::Conn(timer) => match timer {
1905                    ConnTimer::Close => {
1906                        self.state.move_to_drained(None);
1907                        self.endpoint_events.push_back(EndpointEventInner::Drained);
1908                    }
1909                    ConnTimer::Idle => {
1910                        self.kill(ConnectionError::TimedOut);
1911                    }
1912                    ConnTimer::KeepAlive => {
1913                        trace!("sending keep-alive");
1914                        self.ping();
1915                    }
1916                    ConnTimer::KeyDiscard => {
1917                        self.zero_rtt_crypto = None;
1918                        self.prev_crypto = None;
1919                    }
1920                    ConnTimer::PushNewCid => {
1921                        while let Some((path_id, when)) = self.next_cid_retirement() {
1922                            if when > now {
1923                                break;
1924                            }
1925                            match self.local_cid_state.get_mut(&path_id) {
1926                                None => error!(%path_id, "No local CID state for path"),
1927                                Some(cid_state) => {
1928                                    // Update `retire_prior_to` field in NEW_CONNECTION_ID frame
1929                                    let num_new_cid = cid_state.on_cid_timeout().into();
1930                                    if !self.state.is_closed() {
1931                                        trace!(
1932                                            "push a new CID to peer RETIRE_PRIOR_TO field {}",
1933                                            cid_state.retire_prior_to()
1934                                        );
1935                                        self.endpoint_events.push_back(
1936                                            EndpointEventInner::NeedIdentifiers(
1937                                                path_id,
1938                                                now,
1939                                                num_new_cid,
1940                                            ),
1941                                        );
1942                                    }
1943                                }
1944                            }
1945                        }
1946                    }
1947                },
1948                // TODO: add path_id as span somehow
1949                Timer::PerPath(path_id, timer) => {
1950                    let span = trace_span!("per-path timer fired", %path_id, ?timer);
1951                    let _guard = span.enter();
1952                    match timer {
1953                        PathTimer::PathIdle => {
1954                            self.close_path(now, path_id, TransportErrorCode::NO_ERROR.into())
1955                                .ok();
1956                        }
1957
1958                        PathTimer::PathKeepAlive => {
1959                            trace!("sending keep-alive on path");
1960                            self.ping_path(path_id).ok();
1961                        }
1962                        PathTimer::LossDetection => {
1963                            self.on_loss_detection_timeout(now, path_id);
1964                            self.qlog.emit_recovery_metrics(
1965                                path_id,
1966                                &mut self.paths.get_mut(&path_id).unwrap().data,
1967                                now,
1968                            );
1969                        }
1970                        PathTimer::PathValidation => {
1971                            let Some(path) = self.paths.get_mut(&path_id) else {
1972                                continue;
1973                            };
1974                            self.timers.stop(
1975                                Timer::PerPath(path_id, PathTimer::PathChallengeLost),
1976                                self.qlog.with_time(now),
1977                            );
1978                            debug!("path validation failed");
1979                            if let Some((_, prev)) = path.prev.take() {
1980                                path.data = prev;
1981                            }
1982                            path.data.challenges_sent.clear();
1983                            path.data.send_new_challenge = false;
1984                        }
1985                        PathTimer::PathChallengeLost => {
1986                            let Some(path) = self.paths.get_mut(&path_id) else {
1987                                continue;
1988                            };
1989                            trace!("path challenge deemed lost");
1990                            path.data.send_new_challenge = true;
1991                        }
1992                        PathTimer::PathOpen => {
1993                            let Some(path) = self.path_mut(path_id) else {
1994                                continue;
1995                            };
1996                            path.challenges_sent.clear();
1997                            path.send_new_challenge = false;
1998                            debug!("new path validation failed");
1999                            if let Err(err) = self.close_path(
2000                                now,
2001                                path_id,
2002                                TransportErrorCode::PATH_UNSTABLE_OR_POOR.into(),
2003                            ) {
2004                                warn!(?err, "failed closing path");
2005                            }
2006
2007                            self.events.push_back(Event::Path(PathEvent::LocallyClosed {
2008                                id: path_id,
2009                                error: PathError::ValidationFailed,
2010                            }));
2011                        }
2012                        PathTimer::Pacing => trace!("pacing timer expired"),
2013                        PathTimer::MaxAckDelay => {
2014                            trace!("max ack delay reached");
2015                            // This timer is only armed in the Data space
2016                            self.spaces[SpaceId::Data]
2017                                .for_path(path_id)
2018                                .pending_acks
2019                                .on_max_ack_delay_timeout()
2020                        }
2021                        PathTimer::DiscardPath => {
2022                            // The path was abandoned and 3*PTO has expired since.  Clean up all
2023                            // remaining state and install stateless reset token.
2024                            self.timers.stop_per_path(path_id, self.qlog.with_time(now));
2025                            if let Some(loc_cid_state) = self.local_cid_state.remove(&path_id) {
2026                                let (min_seq, max_seq) = loc_cid_state.active_seq();
2027                                for seq in min_seq..=max_seq {
2028                                    self.endpoint_events.push_back(
2029                                        EndpointEventInner::RetireConnectionId(
2030                                            now, path_id, seq, false,
2031                                        ),
2032                                    );
2033                                }
2034                            }
2035                            self.discard_path(path_id, now);
2036                        }
2037                    }
2038                }
2039            }
2040        }
2041    }
2042
2043    /// Close a connection immediately
2044    ///
2045    /// This does not ensure delivery of outstanding data. It is the application's responsibility to
2046    /// call this only when all important communications have been completed, e.g. by calling
2047    /// [`SendStream::finish`] on outstanding streams and waiting for the corresponding
2048    /// [`StreamEvent::Finished`] event.
2049    ///
2050    /// If [`Streams::send_streams`] returns 0, all outstanding stream data has been
2051    /// delivered. There may still be data from the peer that has not been received.
2052    ///
2053    /// [`StreamEvent::Finished`]: crate::StreamEvent::Finished
2054    pub fn close(&mut self, now: Instant, error_code: VarInt, reason: Bytes) {
2055        self.close_inner(
2056            now,
2057            Close::Application(frame::ApplicationClose { error_code, reason }),
2058        )
2059    }
2060
2061    fn close_inner(&mut self, now: Instant, reason: Close) {
2062        let was_closed = self.state.is_closed();
2063        if !was_closed {
2064            self.close_common();
2065            self.set_close_timer(now);
2066            self.close = true;
2067            self.state.move_to_closed_local(reason);
2068        }
2069    }
2070
2071    /// Control datagrams
2072    pub fn datagrams(&mut self) -> Datagrams<'_> {
2073        Datagrams { conn: self }
2074    }
2075
2076    /// Returns connection statistics
2077    pub fn stats(&mut self) -> ConnectionStats {
2078        self.stats.clone()
2079    }
2080
2081    /// Returns path statistics
2082    pub fn path_stats(&mut self, path_id: PathId) -> Option<PathStats> {
2083        let path = self.paths.get(&path_id)?;
2084        let stats = self.path_stats.entry(path_id).or_default();
2085        stats.rtt = path.data.rtt.get();
2086        stats.cwnd = path.data.congestion.window();
2087        stats.current_mtu = path.data.mtud.current_mtu();
2088        Some(*stats)
2089    }
2090
2091    /// Ping the remote endpoint
2092    ///
2093    /// Causes an ACK-eliciting packet to be transmitted on the connection.
2094    pub fn ping(&mut self) {
2095        // TODO(flub): This is very brute-force: it pings *all* the paths.  Instead it would
2096        //    be nice if we could only send a single packet for this.
2097        for path_data in self.spaces[self.highest_space].number_spaces.values_mut() {
2098            path_data.ping_pending = true;
2099        }
2100    }
2101
2102    /// Ping the remote endpoint over a specific path
2103    ///
2104    /// Causes an ACK-eliciting packet to be transmitted on the path.
2105    pub fn ping_path(&mut self, path: PathId) -> Result<(), ClosedPath> {
2106        let path_data = self.spaces[self.highest_space]
2107            .number_spaces
2108            .get_mut(&path)
2109            .ok_or(ClosedPath { _private: () })?;
2110        path_data.ping_pending = true;
2111        Ok(())
2112    }
2113
2114    /// Update traffic keys spontaneously
2115    ///
2116    /// This can be useful for testing key updates, as they otherwise only happen infrequently.
2117    pub fn force_key_update(&mut self) {
2118        if !self.state.is_established() {
2119            debug!("ignoring forced key update in illegal state");
2120            return;
2121        }
2122        if self.prev_crypto.is_some() {
2123            // We already just updated, or are currently updating, the keys. Concurrent key updates
2124            // are illegal.
2125            debug!("ignoring redundant forced key update");
2126            return;
2127        }
2128        self.update_keys(None, false);
2129    }
2130
2131    // Compatibility wrapper for quinn < 0.11.7. Remove for 0.12.
2132    #[doc(hidden)]
2133    #[deprecated]
2134    pub fn initiate_key_update(&mut self) {
2135        self.force_key_update();
2136    }
2137
2138    /// Get a session reference
2139    pub fn crypto_session(&self) -> &dyn crypto::Session {
2140        &*self.crypto
2141    }
2142
2143    /// Whether the connection is in the process of being established
2144    ///
2145    /// If this returns `false`, the connection may be either established or closed, signaled by the
2146    /// emission of a `Connected` or `ConnectionLost` message respectively.
2147    pub fn is_handshaking(&self) -> bool {
2148        self.state.is_handshake()
2149    }
2150
2151    /// Whether the connection is closed
2152    ///
2153    /// Closed connections cannot transport any further data. A connection becomes closed when
2154    /// either peer application intentionally closes it, or when either transport layer detects an
2155    /// error such as a time-out or certificate validation failure.
2156    ///
2157    /// A `ConnectionLost` event is emitted with details when the connection becomes closed.
2158    pub fn is_closed(&self) -> bool {
2159        self.state.is_closed()
2160    }
2161
2162    /// Whether there is no longer any need to keep the connection around
2163    ///
2164    /// Closed connections become drained after a brief timeout to absorb any remaining in-flight
2165    /// packets from the peer. All drained connections have been closed.
2166    pub fn is_drained(&self) -> bool {
2167        self.state.is_drained()
2168    }
2169
2170    /// For clients, if the peer accepted the 0-RTT data packets
2171    ///
2172    /// The value is meaningless until after the handshake completes.
2173    pub fn accepted_0rtt(&self) -> bool {
2174        self.accepted_0rtt
2175    }
2176
2177    /// Whether 0-RTT is/was possible during the handshake
2178    pub fn has_0rtt(&self) -> bool {
2179        self.zero_rtt_enabled
2180    }
2181
2182    /// Whether there are any pending retransmits
2183    pub fn has_pending_retransmits(&self) -> bool {
2184        !self.spaces[SpaceId::Data].pending.is_empty(&self.streams)
2185    }
2186
2187    /// Look up whether we're the client or server of this Connection
2188    pub fn side(&self) -> Side {
2189        self.side.side()
2190    }
2191
2192    /// Get the address observed by the remote over the given path
2193    pub fn path_observed_address(&self, path_id: PathId) -> Result<Option<SocketAddr>, ClosedPath> {
2194        self.path(path_id)
2195            .map(|path_data| {
2196                path_data
2197                    .last_observed_addr_report
2198                    .as_ref()
2199                    .map(|observed| observed.socket_addr())
2200            })
2201            .ok_or(ClosedPath { _private: () })
2202    }
2203
2204    /// The local IP address which was used when the peer established
2205    /// the connection
2206    ///
2207    /// This can be different from the address the endpoint is bound to, in case
2208    /// the endpoint is bound to a wildcard address like `0.0.0.0` or `::`.
2209    ///
2210    /// This will return `None` for clients, or when no `local_ip` was passed to
2211    /// [`Endpoint::handle()`](crate::Endpoint::handle) for the datagrams establishing this
2212    /// connection.
2213    pub fn local_ip(&self) -> Option<IpAddr> {
2214        self.local_ip
2215    }
2216
2217    /// Current best estimate of this connection's latency (round-trip-time)
2218    pub fn rtt(&self, path_id: PathId) -> Option<Duration> {
2219        self.path(path_id).map(|d| d.rtt.get())
2220    }
2221
2222    /// Current state of this connection's congestion controller, for debugging purposes
2223    pub fn congestion_state(&self, path_id: PathId) -> Option<&dyn Controller> {
2224        self.path(path_id).map(|d| d.congestion.as_ref())
2225    }
2226
2227    /// Modify the number of remotely initiated streams that may be concurrently open
2228    ///
2229    /// No streams may be opened by the peer unless fewer than `count` are already open. Large
2230    /// `count`s increase both minimum and worst-case memory consumption.
2231    pub fn set_max_concurrent_streams(&mut self, dir: Dir, count: VarInt) {
2232        self.streams.set_max_concurrent(dir, count);
2233        // If the limit was reduced, then a flow control update previously deemed insignificant may
2234        // now be significant.
2235        let pending = &mut self.spaces[SpaceId::Data].pending;
2236        self.streams.queue_max_stream_id(pending);
2237    }
2238
2239    /// Modify the number of open paths allowed when multipath is enabled
2240    ///
2241    /// When reducing the number of concurrent paths this will only affect delaying sending
2242    /// new MAX_PATH_ID frames until fewer than this number of paths are possible.  To
2243    /// actively reduce paths they must be closed using [`Connection::close_path`], which
2244    /// can also be used to close not-yet-opened paths.
2245    ///
2246    /// If multipath is not negotiated (see the [`TransportConfig`]) this can not enable
2247    /// multipath and will fail.
2248    pub fn set_max_concurrent_paths(
2249        &mut self,
2250        now: Instant,
2251        count: NonZeroU32,
2252    ) -> Result<(), MultipathNotNegotiated> {
2253        if !self.is_multipath_negotiated() {
2254            return Err(MultipathNotNegotiated { _private: () });
2255        }
2256        self.max_concurrent_paths = count;
2257
2258        let in_use_count = self
2259            .local_max_path_id
2260            .next()
2261            .saturating_sub(self.abandoned_paths.len() as u32)
2262            .as_u32();
2263        let extra_needed = count.get().saturating_sub(in_use_count);
2264        let new_max_path_id = self.local_max_path_id.saturating_add(extra_needed);
2265
2266        self.set_max_path_id(now, new_max_path_id);
2267
2268        Ok(())
2269    }
2270
2271    /// If needed, issues a new MAX_PATH_ID frame and new CIDs for any newly allowed paths
2272    fn set_max_path_id(&mut self, now: Instant, max_path_id: PathId) {
2273        if max_path_id <= self.local_max_path_id {
2274            return;
2275        }
2276
2277        self.local_max_path_id = max_path_id;
2278        self.spaces[SpaceId::Data].pending.max_path_id = true;
2279
2280        self.issue_first_path_cids(now);
2281    }
2282
2283    /// Current number of remotely initiated streams that may be concurrently open
2284    ///
2285    /// If the target for this limit is reduced using [`set_max_concurrent_streams`](Self::set_max_concurrent_streams),
2286    /// it will not change immediately, even if fewer streams are open. Instead, it will
2287    /// decrement by one for each time a remotely initiated stream of matching directionality is closed.
2288    pub fn max_concurrent_streams(&self, dir: Dir) -> u64 {
2289        self.streams.max_concurrent(dir)
2290    }
2291
2292    /// See [`TransportConfig::send_window()`]
2293    pub fn set_send_window(&mut self, send_window: u64) {
2294        self.streams.set_send_window(send_window);
2295    }
2296
2297    /// See [`TransportConfig::receive_window()`]
2298    pub fn set_receive_window(&mut self, receive_window: VarInt) {
2299        if self.streams.set_receive_window(receive_window) {
2300            self.spaces[SpaceId::Data].pending.max_data = true;
2301        }
2302    }
2303
2304    /// Whether the Multipath for QUIC extension is enabled.
2305    ///
2306    /// Multipath is only enabled after the handshake is completed and if it was enabled by both
2307    /// peers.
2308    pub fn is_multipath_negotiated(&self) -> bool {
2309        !self.is_handshaking()
2310            && self.config.max_concurrent_multipath_paths.is_some()
2311            && self.peer_params.initial_max_path_id.is_some()
2312    }
2313
2314    fn on_ack_received(
2315        &mut self,
2316        now: Instant,
2317        space: SpaceId,
2318        ack: frame::Ack,
2319    ) -> Result<(), TransportError> {
2320        // All ACKs are referencing path 0
2321        let path = PathId::ZERO;
2322        self.inner_on_ack_received(now, space, path, ack)
2323    }
2324
2325    fn on_path_ack_received(
2326        &mut self,
2327        now: Instant,
2328        space: SpaceId,
2329        path_ack: frame::PathAck,
2330    ) -> Result<(), TransportError> {
2331        let (ack, path) = path_ack.into_ack();
2332        self.inner_on_ack_received(now, space, path, ack)
2333    }
2334
2335    /// Handles an ACK frame acknowledging packets sent on *path*.
2336    fn inner_on_ack_received(
2337        &mut self,
2338        now: Instant,
2339        space: SpaceId,
2340        path: PathId,
2341        ack: frame::Ack,
2342    ) -> Result<(), TransportError> {
2343        if self.abandoned_paths.contains(&path) {
2344            // See also https://www.ietf.org/archive/id/draft-ietf-quic-multipath-17.html#section-3.4.3-3
2345            // > PATH_ACK frames received with an abandoned path ID are silently ignored, as specified in Section 4.
2346            trace!("silently ignoring PATH_ACK on abandoned path");
2347            return Ok(());
2348        }
2349        if ack.largest >= self.spaces[space].for_path(path).next_packet_number {
2350            return Err(TransportError::PROTOCOL_VIOLATION("unsent packet acked"));
2351        }
2352        let new_largest = {
2353            let space = &mut self.spaces[space].for_path(path);
2354            if space.largest_acked_packet.is_none_or(|pn| ack.largest > pn) {
2355                space.largest_acked_packet = Some(ack.largest);
2356                if let Some(info) = space.sent_packets.get(ack.largest) {
2357                    // This should always succeed, but a misbehaving peer might ACK a packet we
2358                    // haven't sent. At worst, that will result in us spuriously reducing the
2359                    // congestion window.
2360                    space.largest_acked_packet_sent = info.time_sent;
2361                }
2362                true
2363            } else {
2364                false
2365            }
2366        };
2367
2368        if self.detect_spurious_loss(&ack, space, path) {
2369            self.path_data_mut(path)
2370                .congestion
2371                .on_spurious_congestion_event();
2372        }
2373
2374        // Avoid DoS from unreasonably huge ack ranges by filtering out just the new acks.
2375        let mut newly_acked = ArrayRangeSet::new();
2376        for range in ack.iter() {
2377            self.spaces[space].for_path(path).check_ack(range.clone())?;
2378            for (pn, _) in self.spaces[space]
2379                .for_path(path)
2380                .sent_packets
2381                .iter_range(range)
2382            {
2383                newly_acked.insert_one(pn);
2384            }
2385        }
2386
2387        if newly_acked.is_empty() {
2388            return Ok(());
2389        }
2390
2391        let mut ack_eliciting_acked = false;
2392        for packet in newly_acked.elts() {
2393            if let Some(info) = self.spaces[space].for_path(path).take(packet) {
2394                for (acked_path_id, acked_pn) in info.largest_acked.iter() {
2395                    // Assume ACKs for all packets below the largest acknowledged in
2396                    // `packet` have been received. This can cause the peer to spuriously
2397                    // retransmit if some of our earlier ACKs were lost, but allows for
2398                    // simpler state tracking. See discussion at
2399                    // https://www.rfc-editor.org/rfc/rfc9000.html#name-limiting-ranges-by-tracking
2400                    if let Some(pns) = self.spaces[space].path_space_mut(*acked_path_id) {
2401                        pns.pending_acks.subtract_below(*acked_pn);
2402                    }
2403                }
2404                ack_eliciting_acked |= info.ack_eliciting;
2405
2406                // Notify MTU discovery that a packet was acked, because it might be an MTU probe
2407                let path_data = self.path_data_mut(path);
2408                let mtu_updated = path_data.mtud.on_acked(space, packet, info.size);
2409                if mtu_updated {
2410                    path_data
2411                        .congestion
2412                        .on_mtu_update(path_data.mtud.current_mtu());
2413                }
2414
2415                // Notify ack frequency that a packet was acked, because it might contain an ACK_FREQUENCY frame
2416                self.ack_frequency.on_acked(path, packet);
2417
2418                self.on_packet_acked(now, path, info);
2419            }
2420        }
2421
2422        let largest_ackd = self.spaces[space].for_path(path).largest_acked_packet;
2423        let app_limited = self.app_limited;
2424        let path_data = self.path_data_mut(path);
2425        let in_flight = path_data.in_flight.bytes;
2426
2427        path_data
2428            .congestion
2429            .on_end_acks(now, in_flight, app_limited, largest_ackd);
2430
2431        if new_largest && ack_eliciting_acked {
2432            let ack_delay = if space != SpaceId::Data {
2433                Duration::from_micros(0)
2434            } else {
2435                cmp::min(
2436                    self.ack_frequency.peer_max_ack_delay,
2437                    Duration::from_micros(ack.delay << self.peer_params.ack_delay_exponent.0),
2438                )
2439            };
2440            let rtt = now.saturating_duration_since(
2441                self.spaces[space].for_path(path).largest_acked_packet_sent,
2442            );
2443
2444            let next_pn = self.spaces[space].for_path(path).next_packet_number;
2445            let path_data = self.path_data_mut(path);
2446            // TODO(@divma): should be a method of path, should be contained in a single place
2447            path_data.rtt.update(ack_delay, rtt);
2448            if path_data.first_packet_after_rtt_sample.is_none() {
2449                path_data.first_packet_after_rtt_sample = Some((space, next_pn));
2450            }
2451        }
2452
2453        // Must be called before crypto/pto_count are clobbered
2454        self.detect_lost_packets(now, space, path, true);
2455
2456        if self.peer_completed_address_validation(path) {
2457            self.path_data_mut(path).pto_count = 0;
2458        }
2459
2460        // Explicit congestion notification
2461        // TODO(@divma): this code is a good example of logic that should be contained in a single
2462        // place but it's split between the path data and the packet number space data, we should
2463        // find a way to make this work without two lookups
2464        if self.path_data(path).sending_ecn {
2465            if let Some(ecn) = ack.ecn {
2466                // We only examine ECN counters from ACKs that we are certain we received in transmit
2467                // order, allowing us to compute an increase in ECN counts to compare against the number
2468                // of newly acked packets that remains well-defined in the presence of arbitrary packet
2469                // reordering.
2470                if new_largest {
2471                    let sent = self.spaces[space].for_path(path).largest_acked_packet_sent;
2472                    self.process_ecn(now, space, path, newly_acked.len() as u64, ecn, sent);
2473                }
2474            } else {
2475                // We always start out sending ECN, so any ack that doesn't acknowledge it disables it.
2476                debug!("ECN not acknowledged by peer");
2477                self.path_data_mut(path).sending_ecn = false;
2478            }
2479        }
2480
2481        self.set_loss_detection_timer(now, path);
2482        Ok(())
2483    }
2484
2485    fn detect_spurious_loss(&mut self, ack: &frame::Ack, space: SpaceId, path: PathId) -> bool {
2486        let lost_packets = &mut self.spaces[space].for_path(path).lost_packets;
2487
2488        if lost_packets.is_empty() {
2489            return false;
2490        }
2491
2492        for range in ack.iter() {
2493            let spurious_losses: Vec<u64> = lost_packets
2494                .iter_range(range.clone())
2495                .map(|(pn, _info)| pn)
2496                .collect();
2497
2498            for pn in spurious_losses {
2499                lost_packets.remove(pn);
2500            }
2501        }
2502
2503        // If this ACK frame acknowledged all deemed lost packets,
2504        // then we have raised a spurious congestion event in the past.
2505        // We cannot conclude when there are remaining packets,
2506        // but future ACK frames might indicate a spurious loss detection.
2507        lost_packets.is_empty()
2508    }
2509
2510    /// Drain lost packets that we reasonably think will never arrive
2511    ///
2512    /// The current criterion is copied from `msquic`:
2513    /// discard packets that were sent earlier than 2 probe timeouts ago.
2514    fn drain_lost_packets(&mut self, now: Instant, space: SpaceId, path: PathId) {
2515        let two_pto = 2 * self.path_data(path).rtt.pto_base();
2516
2517        let lost_packets = &mut self.spaces[space].for_path(path).lost_packets;
2518        lost_packets.retain(|_pn, info| now.saturating_duration_since(info.time_sent) <= two_pto);
2519    }
2520
2521    /// Process a new ECN block from an in-order ACK
2522    fn process_ecn(
2523        &mut self,
2524        now: Instant,
2525        space: SpaceId,
2526        path: PathId,
2527        newly_acked: u64,
2528        ecn: frame::EcnCounts,
2529        largest_sent_time: Instant,
2530    ) {
2531        match self.spaces[space]
2532            .for_path(path)
2533            .detect_ecn(newly_acked, ecn)
2534        {
2535            Err(e) => {
2536                debug!("halting ECN due to verification failure: {}", e);
2537
2538                self.path_data_mut(path).sending_ecn = false;
2539                // Wipe out the existing value because it might be garbage and could interfere with
2540                // future attempts to use ECN on new paths.
2541                self.spaces[space].for_path(path).ecn_feedback = frame::EcnCounts::ZERO;
2542            }
2543            Ok(false) => {}
2544            Ok(true) => {
2545                self.path_stats.entry(path).or_default().congestion_events += 1;
2546                self.path_data_mut(path).congestion.on_congestion_event(
2547                    now,
2548                    largest_sent_time,
2549                    false,
2550                    true,
2551                    0,
2552                );
2553            }
2554        }
2555    }
2556
2557    // Not timing-aware, so it's safe to call this for inferred acks, such as arise from
2558    // high-latency handshakes
2559    fn on_packet_acked(&mut self, now: Instant, path_id: PathId, info: SentPacket) {
2560        self.paths
2561            .get_mut(&path_id)
2562            .expect("known path")
2563            .remove_in_flight(&info);
2564        let app_limited = self.app_limited;
2565        let path = self.path_data_mut(path_id);
2566        if info.ack_eliciting && !path.is_validating_path() {
2567            // Only pass ACKs to the congestion controller if we are not validating the current
2568            // path, so as to ignore any ACKs from older paths still coming in.
2569            let rtt = path.rtt;
2570            path.congestion
2571                .on_ack(now, info.time_sent, info.size.into(), app_limited, &rtt);
2572        }
2573
2574        // Update state for confirmed delivery of frames
2575        if let Some(retransmits) = info.retransmits.get() {
2576            for (id, _) in retransmits.reset_stream.iter() {
2577                self.streams.reset_acked(*id);
2578            }
2579        }
2580
2581        for frame in info.stream_frames {
2582            self.streams.received_ack_of(frame);
2583        }
2584    }
2585
2586    fn set_key_discard_timer(&mut self, now: Instant, space: SpaceId) {
2587        let start = if self.zero_rtt_crypto.is_some() {
2588            now
2589        } else {
2590            self.prev_crypto
2591                .as_ref()
2592                .expect("no previous keys")
2593                .end_packet
2594                .as_ref()
2595                .expect("update not acknowledged yet")
2596                .1
2597        };
2598
2599        // QUIC-MULTIPATH § 2.5 Key Phase Update Process: use largest PTO off all paths.
2600        let end = self.calculate_end_timer(start, 3, space);
2601        self.timers.set(
2602            Timer::Conn(ConnTimer::KeyDiscard),
2603            end,
2604            self.qlog.with_time(now),
2605        );
2606    }
2607
2608    /// Handle a [`PathTimer::LossDetection`] timeout.
2609    ///
2610    /// This timer expires for two reasons:
2611    /// - An ACK-eliciting packet we sent should be considered lost.
2612    /// - The PTO may have expired and a tail-loss probe needs to be scheduled.
2613    ///
2614    /// The former needs us to schedule re-transmission of the lost data.
2615    ///
2616    /// The latter means we have not received an ACK for an ack-eliciting packet we sent
2617    /// within the PTO time-window. We need to schedule a tail-loss probe, an ack-eliciting
2618    /// packet, to try and elicit new acknowledgements. These new acknowledgements will
2619    /// indicate whether the previously sent packets were lost or not.
2620    fn on_loss_detection_timeout(&mut self, now: Instant, path_id: PathId) {
2621        if let Some((_, pn_space)) = self.loss_time_and_space(path_id) {
2622            // Time threshold loss Detection
2623            self.detect_lost_packets(now, pn_space, path_id, false);
2624            self.set_loss_detection_timer(now, path_id);
2625            return;
2626        }
2627
2628        let (_, space) = match self.pto_time_and_space(now, path_id) {
2629            Some(x) => x,
2630            None => {
2631                error!(%path_id, "PTO expired while unset");
2632                return;
2633            }
2634        };
2635        trace!(
2636            in_flight = self.path_data(path_id).in_flight.bytes,
2637            count = self.path_data(path_id).pto_count,
2638            ?space,
2639            %path_id,
2640            "PTO fired"
2641        );
2642
2643        let count = match self.path_data(path_id).in_flight.ack_eliciting {
2644            // A PTO when we're not expecting any ACKs must be due to handshake anti-amplification
2645            // deadlock preventions
2646            0 => {
2647                debug_assert!(!self.peer_completed_address_validation(path_id));
2648                1
2649            }
2650            // Conventional loss probe
2651            _ => 2,
2652        };
2653        let pns = self.spaces[space].for_path(path_id);
2654        pns.loss_probes = pns.loss_probes.saturating_add(count);
2655        let path_data = self.path_data_mut(path_id);
2656        path_data.pto_count = path_data.pto_count.saturating_add(1);
2657        self.set_loss_detection_timer(now, path_id);
2658    }
2659
2660    /// Detect any lost packets
2661    ///
2662    /// There are two cases in which we detects lost packets:
2663    ///
2664    /// - We received an ACK packet.
2665    /// - The [`PathTimer::LossDetection`] timer expired. So there is an un-acknowledged packet
2666    ///   that was followed by an acknowledged packet. The loss timer for this
2667    ///   un-acknowledged packet expired and we need to detect that packet as lost.
2668    ///
2669    /// Packets are lost if they are both (See RFC9002 §6.1):
2670    ///
2671    /// - Unacknowledged, in flight and sent prior to an acknowledged packet.
2672    /// - Old enough by either:
2673    ///   - Having a packet number [`TransportConfig::packet_threshold`] lower then the last
2674    ///     acknowledged packet.
2675    ///   - Being sent [`TransportConfig::time_threshold`] * RTT in the past.
2676    fn detect_lost_packets(
2677        &mut self,
2678        now: Instant,
2679        pn_space: SpaceId,
2680        path_id: PathId,
2681        due_to_ack: bool,
2682    ) {
2683        let mut lost_packets = Vec::<u64>::new();
2684        let mut lost_mtu_probe = None;
2685        let mut in_persistent_congestion = false;
2686        let mut size_of_lost_packets = 0u64;
2687        self.spaces[pn_space].for_path(path_id).loss_time = None;
2688
2689        // Find all the lost packets, populating all variables initialised above.
2690
2691        let path = self.path_data(path_id);
2692        let in_flight_mtu_probe = path.mtud.in_flight_mtu_probe();
2693        let loss_delay = path
2694            .rtt
2695            .conservative()
2696            .mul_f32(self.config.time_threshold)
2697            .max(TIMER_GRANULARITY);
2698        let first_packet_after_rtt_sample = path.first_packet_after_rtt_sample;
2699
2700        let largest_acked_packet = self.spaces[pn_space]
2701            .for_path(path_id)
2702            .largest_acked_packet
2703            .expect("detect_lost_packets only to be called if path received at least one ACK");
2704        let packet_threshold = self.config.packet_threshold as u64;
2705
2706        // InPersistentCongestion: Determine if all packets in the time period before the newest
2707        // lost packet, including the edges, are marked lost. PTO computation must always
2708        // include max ACK delay, i.e. operate as if in Data space (see RFC9001 §7.6.1).
2709        let congestion_period = self
2710            .pto(SpaceId::Data, path_id)
2711            .saturating_mul(self.config.persistent_congestion_threshold);
2712        let mut persistent_congestion_start: Option<Instant> = None;
2713        let mut prev_packet = None;
2714        let space = self.spaces[pn_space].for_path(path_id);
2715
2716        for (packet, info) in space.sent_packets.iter_range(0..largest_acked_packet) {
2717            if prev_packet != Some(packet.wrapping_sub(1)) {
2718                // An intervening packet was acknowledged
2719                persistent_congestion_start = None;
2720            }
2721
2722            // Packets sent before now - loss_delay are deemed lost.
2723            // However, we avoid subtraction as it can panic and there's no
2724            // saturating equivalent of this subtraction operation with a Duration.
2725            let packet_too_old = now.saturating_duration_since(info.time_sent) >= loss_delay;
2726            if packet_too_old || largest_acked_packet >= packet + packet_threshold {
2727                // The packet should be declared lost.
2728                if Some(packet) == in_flight_mtu_probe {
2729                    // Lost MTU probes are not included in `lost_packets`, because they
2730                    // should not trigger a congestion control response
2731                    lost_mtu_probe = in_flight_mtu_probe;
2732                } else {
2733                    lost_packets.push(packet);
2734                    size_of_lost_packets += info.size as u64;
2735                    if info.ack_eliciting && due_to_ack {
2736                        match persistent_congestion_start {
2737                            // Two ACK-eliciting packets lost more than
2738                            // congestion_period apart, with no ACKed packets in between
2739                            Some(start) if info.time_sent - start > congestion_period => {
2740                                in_persistent_congestion = true;
2741                            }
2742                            // Persistent congestion must start after the first RTT sample
2743                            None if first_packet_after_rtt_sample
2744                                .is_some_and(|x| x < (pn_space, packet)) =>
2745                            {
2746                                persistent_congestion_start = Some(info.time_sent);
2747                            }
2748                            _ => {}
2749                        }
2750                    }
2751                }
2752            } else {
2753                // The packet should not yet be declared lost.
2754                if space.loss_time.is_none() {
2755                    // Since we iterate in order the lowest packet number's loss time will
2756                    // always be the earliest.
2757                    space.loss_time = Some(info.time_sent + loss_delay);
2758                }
2759                persistent_congestion_start = None;
2760            }
2761
2762            prev_packet = Some(packet);
2763        }
2764
2765        self.handle_lost_packets(
2766            pn_space,
2767            path_id,
2768            now,
2769            lost_packets,
2770            lost_mtu_probe,
2771            loss_delay,
2772            in_persistent_congestion,
2773            size_of_lost_packets,
2774        );
2775    }
2776
2777    /// Drops the path state, declaring any remaining in-flight packets as lost
2778    fn discard_path(&mut self, path_id: PathId, now: Instant) {
2779        trace!(%path_id, "dropping path state");
2780        let path = self.path_data(path_id);
2781        let in_flight_mtu_probe = path.mtud.in_flight_mtu_probe();
2782
2783        let mut size_of_lost_packets = 0u64; // add to path_stats.lost_bytes;
2784        let lost_pns: Vec<_> = self.spaces[SpaceId::Data]
2785            .for_path(path_id)
2786            .sent_packets
2787            .iter()
2788            .filter(|(pn, _info)| Some(*pn) != in_flight_mtu_probe)
2789            .map(|(pn, info)| {
2790                size_of_lost_packets += info.size as u64;
2791                pn
2792            })
2793            .collect();
2794
2795        if !lost_pns.is_empty() {
2796            trace!(
2797                %path_id,
2798                count = lost_pns.len(),
2799                lost_bytes = size_of_lost_packets,
2800                "packets lost on path abandon"
2801            );
2802            self.handle_lost_packets(
2803                SpaceId::Data,
2804                path_id,
2805                now,
2806                lost_pns,
2807                in_flight_mtu_probe,
2808                Duration::ZERO,
2809                false,
2810                size_of_lost_packets,
2811            );
2812        }
2813        self.paths.remove(&path_id);
2814        self.spaces[SpaceId::Data].number_spaces.remove(&path_id);
2815
2816        let path_stats = self.path_stats.remove(&path_id).unwrap_or_default();
2817        self.events.push_back(
2818            PathEvent::Abandoned {
2819                id: path_id,
2820                path_stats,
2821            }
2822            .into(),
2823        );
2824    }
2825
2826    fn handle_lost_packets(
2827        &mut self,
2828        pn_space: SpaceId,
2829        path_id: PathId,
2830        now: Instant,
2831        lost_packets: Vec<u64>,
2832        lost_mtu_probe: Option<u64>,
2833        loss_delay: Duration,
2834        in_persistent_congestion: bool,
2835        size_of_lost_packets: u64,
2836    ) {
2837        debug_assert!(
2838            {
2839                let mut sorted = lost_packets.clone();
2840                sorted.sort();
2841                sorted == lost_packets
2842            },
2843            "lost_packets must be sorted"
2844        );
2845
2846        self.drain_lost_packets(now, pn_space, path_id);
2847
2848        // OnPacketsLost
2849        if let Some(largest_lost) = lost_packets.last().cloned() {
2850            let old_bytes_in_flight = self.path_data_mut(path_id).in_flight.bytes;
2851            let largest_lost_sent = self.spaces[pn_space]
2852                .for_path(path_id)
2853                .sent_packets
2854                .get(largest_lost)
2855                .unwrap()
2856                .time_sent;
2857            let path_stats = self.path_stats.entry(path_id).or_default();
2858            path_stats.lost_packets += lost_packets.len() as u64;
2859            path_stats.lost_bytes += size_of_lost_packets;
2860            trace!(
2861                %path_id,
2862                count = lost_packets.len(),
2863                lost_bytes = size_of_lost_packets,
2864                "packets lost",
2865            );
2866
2867            for &packet in &lost_packets {
2868                let Some(info) = self.spaces[pn_space].for_path(path_id).take(packet) else {
2869                    continue;
2870                };
2871                self.qlog
2872                    .emit_packet_lost(packet, &info, loss_delay, pn_space, now);
2873                self.paths
2874                    .get_mut(&path_id)
2875                    .unwrap()
2876                    .remove_in_flight(&info);
2877
2878                for frame in info.stream_frames {
2879                    self.streams.retransmit(frame);
2880                }
2881                self.spaces[pn_space].pending |= info.retransmits;
2882                self.path_data_mut(path_id)
2883                    .mtud
2884                    .on_non_probe_lost(packet, info.size);
2885
2886                self.spaces[pn_space].for_path(path_id).lost_packets.insert(
2887                    packet,
2888                    LostPacket {
2889                        time_sent: info.time_sent,
2890                    },
2891                );
2892            }
2893
2894            let path = self.path_data_mut(path_id);
2895            if path.mtud.black_hole_detected(now) {
2896                path.congestion.on_mtu_update(path.mtud.current_mtu());
2897                if let Some(max_datagram_size) = self.datagrams().max_size() {
2898                    self.datagrams.drop_oversized(max_datagram_size);
2899                }
2900                self.path_stats
2901                    .entry(path_id)
2902                    .or_default()
2903                    .black_holes_detected += 1;
2904            }
2905
2906            // Don't apply congestion penalty for lost ack-only packets
2907            let lost_ack_eliciting =
2908                old_bytes_in_flight != self.path_data_mut(path_id).in_flight.bytes;
2909
2910            if lost_ack_eliciting {
2911                self.path_stats
2912                    .entry(path_id)
2913                    .or_default()
2914                    .congestion_events += 1;
2915                self.path_data_mut(path_id).congestion.on_congestion_event(
2916                    now,
2917                    largest_lost_sent,
2918                    in_persistent_congestion,
2919                    false,
2920                    size_of_lost_packets,
2921                );
2922            }
2923        }
2924
2925        // Handle a lost MTU probe
2926        if let Some(packet) = lost_mtu_probe {
2927            let info = self.spaces[SpaceId::Data]
2928                .for_path(path_id)
2929                .take(packet)
2930                .unwrap(); // safe: lost_mtu_probe is omitted from lost_packets, and
2931            // therefore must not have been removed yet
2932            self.paths
2933                .get_mut(&path_id)
2934                .unwrap()
2935                .remove_in_flight(&info);
2936            self.path_data_mut(path_id).mtud.on_probe_lost();
2937            self.path_stats
2938                .entry(path_id)
2939                .or_default()
2940                .lost_plpmtud_probes += 1;
2941        }
2942    }
2943
2944    /// Returns the earliest time packets should be declared lost for all spaces on a path.
2945    ///
2946    /// If a path has an acknowledged packet with any prior un-acknowledged packets, the
2947    /// earliest un-acknowledged packet can be declared lost after a timeout has elapsed.
2948    /// The time returned is when this packet should be declared lost.
2949    fn loss_time_and_space(&self, path_id: PathId) -> Option<(Instant, SpaceId)> {
2950        SpaceId::iter()
2951            .filter_map(|id| {
2952                self.spaces[id]
2953                    .number_spaces
2954                    .get(&path_id)
2955                    .and_then(|pns| pns.loss_time)
2956                    .map(|time| (time, id))
2957            })
2958            .min_by_key(|&(time, _)| time)
2959    }
2960
2961    /// Returns the earliest next PTO should fire for all spaces on a path.
2962    fn pto_time_and_space(&mut self, now: Instant, path_id: PathId) -> Option<(Instant, SpaceId)> {
2963        let path = self.path(path_id)?;
2964        let pto_count = path.pto_count;
2965        let backoff = 2u32.pow(pto_count.min(MAX_BACKOFF_EXPONENT));
2966        let mut duration = path.rtt.pto_base() * backoff;
2967
2968        if path_id == PathId::ZERO
2969            && path.in_flight.ack_eliciting == 0
2970            && !self.peer_completed_address_validation(PathId::ZERO)
2971        {
2972            // Address Validation during Connection Establishment:
2973            // https://www.rfc-editor.org/rfc/rfc9000.html#section-8.1. To prevent a
2974            // deadlock if an Initial or Handshake packet from the server is lost and the
2975            // server can not send more due to its anti-amplification limit the client must
2976            // send another packet on PTO.
2977            let space = match self.highest_space {
2978                SpaceId::Handshake => SpaceId::Handshake,
2979                _ => SpaceId::Initial,
2980            };
2981
2982            return Some((now + duration, space));
2983        }
2984
2985        let mut result = None;
2986        for space in SpaceId::iter() {
2987            let Some(pns) = self.spaces[space].number_spaces.get(&path_id) else {
2988                continue;
2989            };
2990
2991            if !pns.has_in_flight() {
2992                continue;
2993            }
2994            if space == SpaceId::Data {
2995                // Skip ApplicationData until handshake completes.
2996                if self.is_handshaking() {
2997                    return result;
2998                }
2999                // Include max_ack_delay and backoff for ApplicationData.
3000                duration += self.ack_frequency.max_ack_delay_for_pto() * backoff;
3001            }
3002            let Some(last_ack_eliciting) = pns.time_of_last_ack_eliciting_packet else {
3003                continue;
3004            };
3005            let pto = last_ack_eliciting + duration;
3006            if result.is_none_or(|(earliest_pto, _)| pto < earliest_pto) {
3007                if path.anti_amplification_blocked(1) {
3008                    // Nothing would be able to be sent.
3009                    continue;
3010                }
3011                if path.in_flight.ack_eliciting == 0 {
3012                    // Nothing ack-eliciting, no PTO to arm/fire.
3013                    continue;
3014                }
3015                result = Some((pto, space));
3016            }
3017        }
3018        result
3019    }
3020
3021    fn peer_completed_address_validation(&self, path: PathId) -> bool {
3022        // TODO(flub): This logic needs updating for multipath
3023        if self.side.is_server() || self.state.is_closed() {
3024            return true;
3025        }
3026        // The server is guaranteed to have validated our address if any of our handshake or 1-RTT
3027        // packets are acknowledged or we've seen HANDSHAKE_DONE and discarded handshake keys.
3028        self.spaces[SpaceId::Handshake]
3029            .path_space(PathId::ZERO)
3030            .and_then(|pns| pns.largest_acked_packet)
3031            .is_some()
3032            || self.spaces[SpaceId::Data]
3033                .path_space(path)
3034                .and_then(|pns| pns.largest_acked_packet)
3035                .is_some()
3036            || (self.spaces[SpaceId::Data].crypto.is_some()
3037                && self.spaces[SpaceId::Handshake].crypto.is_none())
3038    }
3039
3040    /// Resets the the [`PathTimer::LossDetection`] timer to the next instant it may be needed
3041    ///
3042    /// The timer must fire if either:
3043    /// - An ack-eliciting packet we sent needs to be declared lost.
3044    /// - A tail-loss probe needs to be sent.
3045    ///
3046    /// See [`Connection::on_loss_detection_timeout`] for details.
3047    fn set_loss_detection_timer(&mut self, now: Instant, path_id: PathId) {
3048        if self.state.is_closed() {
3049            // No loss detection takes place on closed connections, and `close_common` already
3050            // stopped time timer. Ensure we don't restart it inadvertently, e.g. in response to a
3051            // reordered packet being handled by state-insensitive code.
3052            return;
3053        }
3054
3055        if let Some((loss_time, _)) = self.loss_time_and_space(path_id) {
3056            // Time threshold loss detection.
3057            self.timers.set(
3058                Timer::PerPath(path_id, PathTimer::LossDetection),
3059                loss_time,
3060                self.qlog.with_time(now),
3061            );
3062            return;
3063        }
3064
3065        // Determine which PN space to arm PTO for.
3066        // Calculate PTO duration
3067        if let Some((timeout, _)) = self.pto_time_and_space(now, path_id) {
3068            self.timers.set(
3069                Timer::PerPath(path_id, PathTimer::LossDetection),
3070                timeout,
3071                self.qlog.with_time(now),
3072            );
3073        } else {
3074            self.timers.stop(
3075                Timer::PerPath(path_id, PathTimer::LossDetection),
3076                self.qlog.with_time(now),
3077            );
3078        }
3079    }
3080
3081    /// Probe Timeout
3082    ///
3083    /// The PTO is logically the time in which you'd expect to receive an acknowledgement
3084    /// for a packet. So approximately RTT + max_ack_delay.
3085    fn pto(&self, space: SpaceId, path_id: PathId) -> Duration {
3086        let max_ack_delay = match space {
3087            SpaceId::Initial | SpaceId::Handshake => Duration::ZERO,
3088            SpaceId::Data => self.ack_frequency.max_ack_delay_for_pto(),
3089        };
3090        self.path_data(path_id).rtt.pto_base() + max_ack_delay
3091    }
3092
3093    fn on_packet_authenticated(
3094        &mut self,
3095        now: Instant,
3096        space_id: SpaceId,
3097        path_id: PathId,
3098        ecn: Option<EcnCodepoint>,
3099        packet: Option<u64>,
3100        spin: bool,
3101        is_1rtt: bool,
3102    ) {
3103        self.total_authed_packets += 1;
3104        if let Some(last_allowed_receive) = self
3105            .paths
3106            .get(&path_id)
3107            .and_then(|path| path.data.last_allowed_receive)
3108        {
3109            if now > last_allowed_receive {
3110                warn!("received data on path which we abandoned more than 3 * PTO ago");
3111                // The peer failed to respond with a PATH_ABANDON in time.
3112                if !self.state.is_closed() {
3113                    // TODO(flub): What should the error code be?
3114                    self.state.move_to_closed(TransportError::NO_ERROR(
3115                        "peer failed to respond with PATH_ABANDON in time",
3116                    ));
3117                    self.close_common();
3118                    self.set_close_timer(now);
3119                    self.close = true;
3120                }
3121                return;
3122            }
3123        }
3124
3125        self.reset_keep_alive(path_id, now);
3126        self.reset_idle_timeout(now, space_id, path_id);
3127        self.permit_idle_reset = true;
3128        self.receiving_ecn |= ecn.is_some();
3129        if let Some(x) = ecn {
3130            let space = &mut self.spaces[space_id];
3131            space.for_path(path_id).ecn_counters += x;
3132
3133            if x.is_ce() {
3134                space
3135                    .for_path(path_id)
3136                    .pending_acks
3137                    .set_immediate_ack_required();
3138            }
3139        }
3140
3141        let packet = match packet {
3142            Some(x) => x,
3143            None => return,
3144        };
3145        match &self.side {
3146            ConnectionSide::Client { .. } => {
3147                // If we received a handshake packet that authenticated, then we're talking to
3148                // the real server.  From now on we should no longer allow the server to migrate
3149                // its address.
3150                if space_id == SpaceId::Handshake {
3151                    if let Some(hs) = self.state.as_handshake_mut() {
3152                        hs.allow_server_migration = false;
3153                    }
3154                }
3155            }
3156            ConnectionSide::Server { .. } => {
3157                if self.spaces[SpaceId::Initial].crypto.is_some() && space_id == SpaceId::Handshake
3158                {
3159                    // A server stops sending and processing Initial packets when it receives its first Handshake packet.
3160                    self.discard_space(now, SpaceId::Initial);
3161                }
3162                if self.zero_rtt_crypto.is_some() && is_1rtt {
3163                    // Discard 0-RTT keys soon after receiving a 1-RTT packet
3164                    self.set_key_discard_timer(now, space_id)
3165                }
3166            }
3167        }
3168        let space = self.spaces[space_id].for_path(path_id);
3169        space.pending_acks.insert_one(packet, now);
3170        if packet >= space.rx_packet.unwrap_or_default() {
3171            space.rx_packet = Some(packet);
3172            // Update outgoing spin bit, inverting iff we're the client
3173            self.spin = self.side.is_client() ^ spin;
3174        }
3175    }
3176
3177    /// Resets the idle timeout timers
3178    ///
3179    /// Without multipath there is only the connection-wide idle timeout. When multipath is
3180    /// enabled there is an additional per-path idle timeout.
3181    fn reset_idle_timeout(&mut self, now: Instant, space: SpaceId, path_id: PathId) {
3182        // First reset the global idle timeout.
3183        if let Some(timeout) = self.idle_timeout {
3184            if self.state.is_closed() {
3185                self.timers
3186                    .stop(Timer::Conn(ConnTimer::Idle), self.qlog.with_time(now));
3187            } else {
3188                // TODO: include max(timeout)
3189                let end = self.calculate_end_timer(now, 3, space);
3190                self.timers
3191                    .set(Timer::Conn(ConnTimer::Idle), end, self.qlog.with_time(now));
3192            }
3193        }
3194
3195        // Now handle the per-path state
3196        if let Some(timeout) = self.path_data(path_id).idle_timeout {
3197            if self.state.is_closed() {
3198                self.timers.stop(
3199                    Timer::PerPath(path_id, PathTimer::PathIdle),
3200                    self.qlog.with_time(now),
3201                );
3202            } else {
3203                let pto = self.pto(space, path_id);
3204                let end = self.path_data(path_id).timer_offset(now, 3 * pto);
3205                let dt = cmp::max(now + timeout, end);
3206                self.timers.set(
3207                    Timer::PerPath(path_id, PathTimer::PathIdle),
3208                    dt,
3209                    self.qlog.with_time(now),
3210                );
3211            }
3212        }
3213    }
3214
3215    /// Resets both the [`ConnTimer::KeepAlive`] and [`PathTimer::PathKeepAlive`] timers
3216    fn reset_keep_alive(&mut self, path_id: PathId, now: Instant) {
3217        if !self.state.is_established() {
3218            return;
3219        }
3220
3221        if let Some(interval) = self.config.keep_alive_interval {
3222            self.timers.set(
3223                Timer::Conn(ConnTimer::KeepAlive),
3224                now + interval,
3225                self.qlog.with_time(now),
3226            );
3227        }
3228
3229        if let Some(interval) = self.path_data(path_id).keep_alive {
3230            self.timers.set(
3231                Timer::PerPath(path_id, PathTimer::PathKeepAlive),
3232                now + interval,
3233                self.qlog.with_time(now),
3234            );
3235        }
3236    }
3237
3238    /// Sets the timer for when a previously issued CID should be retired next
3239    fn reset_cid_retirement(&mut self, now: Instant) {
3240        if let Some((_path, t)) = self.next_cid_retirement() {
3241            self.timers.set(
3242                Timer::Conn(ConnTimer::PushNewCid),
3243                t,
3244                self.qlog.with_time(now),
3245            );
3246        }
3247    }
3248
3249    /// The next time when a previously issued CID should be retired
3250    fn next_cid_retirement(&self) -> Option<(PathId, Instant)> {
3251        self.local_cid_state
3252            .iter()
3253            .filter_map(|(path_id, cid_state)| cid_state.next_timeout().map(|t| (*path_id, t)))
3254            .min_by_key(|(_path_id, timeout)| *timeout)
3255    }
3256
3257    /// Handle the already-decrypted first packet from the client
3258    ///
3259    /// Decrypting the first packet in the `Endpoint` allows stateless packet handling to be more
3260    /// efficient.
3261    pub(crate) fn handle_first_packet(
3262        &mut self,
3263        now: Instant,
3264        remote: SocketAddr,
3265        ecn: Option<EcnCodepoint>,
3266        packet_number: u64,
3267        packet: InitialPacket,
3268        remaining: Option<BytesMut>,
3269    ) -> Result<(), ConnectionError> {
3270        let span = trace_span!("first recv");
3271        let _guard = span.enter();
3272        debug_assert!(self.side.is_server());
3273        let len = packet.header_data.len() + packet.payload.len();
3274        let path_id = PathId::ZERO;
3275        self.path_data_mut(path_id).total_recvd = len as u64;
3276
3277        if let Some(hs) = self.state.as_handshake_mut() {
3278            hs.expected_token = packet.header.token.clone();
3279        } else {
3280            unreachable!("first packet must be delivered in Handshake state");
3281        }
3282
3283        // The first packet is always on PathId::ZERO
3284        self.on_packet_authenticated(
3285            now,
3286            SpaceId::Initial,
3287            path_id,
3288            ecn,
3289            Some(packet_number),
3290            false,
3291            false,
3292        );
3293
3294        let packet: Packet = packet.into();
3295
3296        let mut qlog = QlogRecvPacket::new(len);
3297        qlog.header(&packet.header, Some(packet_number), path_id);
3298
3299        self.process_decrypted_packet(
3300            now,
3301            remote,
3302            path_id,
3303            Some(packet_number),
3304            packet,
3305            &mut qlog,
3306        )?;
3307        self.qlog.emit_packet_received(qlog, now);
3308        if let Some(data) = remaining {
3309            self.handle_coalesced(now, remote, path_id, ecn, data);
3310        }
3311
3312        self.qlog.emit_recovery_metrics(
3313            path_id,
3314            &mut self.paths.get_mut(&path_id).unwrap().data,
3315            now,
3316        );
3317
3318        Ok(())
3319    }
3320
3321    fn init_0rtt(&mut self, now: Instant) {
3322        let (header, packet) = match self.crypto.early_crypto() {
3323            Some(x) => x,
3324            None => return,
3325        };
3326        if self.side.is_client() {
3327            match self.crypto.transport_parameters() {
3328                Ok(params) => {
3329                    let params = params
3330                        .expect("crypto layer didn't supply transport parameters with ticket");
3331                    // Certain values must not be cached
3332                    let params = TransportParameters {
3333                        initial_src_cid: None,
3334                        original_dst_cid: None,
3335                        preferred_address: None,
3336                        retry_src_cid: None,
3337                        stateless_reset_token: None,
3338                        min_ack_delay: None,
3339                        ack_delay_exponent: TransportParameters::default().ack_delay_exponent,
3340                        max_ack_delay: TransportParameters::default().max_ack_delay,
3341                        initial_max_path_id: None,
3342                        ..params
3343                    };
3344                    self.set_peer_params(params);
3345                    self.qlog.emit_peer_transport_params_restored(self, now);
3346                }
3347                Err(e) => {
3348                    error!("session ticket has malformed transport parameters: {}", e);
3349                    return;
3350                }
3351            }
3352        }
3353        trace!("0-RTT enabled");
3354        self.zero_rtt_enabled = true;
3355        self.zero_rtt_crypto = Some(ZeroRttCrypto { header, packet });
3356    }
3357
3358    fn read_crypto(
3359        &mut self,
3360        space: SpaceId,
3361        crypto: &frame::Crypto,
3362        payload_len: usize,
3363    ) -> Result<(), TransportError> {
3364        let expected = if !self.state.is_handshake() {
3365            SpaceId::Data
3366        } else if self.highest_space == SpaceId::Initial {
3367            SpaceId::Initial
3368        } else {
3369            // On the server, self.highest_space can be Data after receiving the client's first
3370            // flight, but we expect Handshake CRYPTO until the handshake is complete.
3371            SpaceId::Handshake
3372        };
3373        // We can't decrypt Handshake packets when highest_space is Initial, CRYPTO frames in 0-RTT
3374        // packets are illegal, and we don't process 1-RTT packets until the handshake is
3375        // complete. Therefore, we will never see CRYPTO data from a later-than-expected space.
3376        debug_assert!(space <= expected, "received out-of-order CRYPTO data");
3377
3378        let end = crypto.offset + crypto.data.len() as u64;
3379        if space < expected && end > self.spaces[space].crypto_stream.bytes_read() {
3380            warn!(
3381                "received new {:?} CRYPTO data when expecting {:?}",
3382                space, expected
3383            );
3384            return Err(TransportError::PROTOCOL_VIOLATION(
3385                "new data at unexpected encryption level",
3386            ));
3387        }
3388
3389        let space = &mut self.spaces[space];
3390        let max = end.saturating_sub(space.crypto_stream.bytes_read());
3391        if max > self.config.crypto_buffer_size as u64 {
3392            return Err(TransportError::CRYPTO_BUFFER_EXCEEDED(""));
3393        }
3394
3395        space
3396            .crypto_stream
3397            .insert(crypto.offset, crypto.data.clone(), payload_len);
3398        while let Some(chunk) = space.crypto_stream.read(usize::MAX, true) {
3399            trace!("consumed {} CRYPTO bytes", chunk.bytes.len());
3400            if self.crypto.read_handshake(&chunk.bytes)? {
3401                self.events.push_back(Event::HandshakeDataReady);
3402            }
3403        }
3404
3405        Ok(())
3406    }
3407
3408    fn write_crypto(&mut self) {
3409        loop {
3410            let space = self.highest_space;
3411            let mut outgoing = Vec::new();
3412            if let Some(crypto) = self.crypto.write_handshake(&mut outgoing) {
3413                match space {
3414                    SpaceId::Initial => {
3415                        self.upgrade_crypto(SpaceId::Handshake, crypto);
3416                    }
3417                    SpaceId::Handshake => {
3418                        self.upgrade_crypto(SpaceId::Data, crypto);
3419                    }
3420                    _ => unreachable!("got updated secrets during 1-RTT"),
3421                }
3422            }
3423            if outgoing.is_empty() {
3424                if space == self.highest_space {
3425                    break;
3426                } else {
3427                    // Keys updated, check for more data to send
3428                    continue;
3429                }
3430            }
3431            let offset = self.spaces[space].crypto_offset;
3432            let outgoing = Bytes::from(outgoing);
3433            if let Some(hs) = self.state.as_handshake_mut() {
3434                if space == SpaceId::Initial && offset == 0 && self.side.is_client() {
3435                    hs.client_hello = Some(outgoing.clone());
3436                }
3437            }
3438            self.spaces[space].crypto_offset += outgoing.len() as u64;
3439            trace!("wrote {} {:?} CRYPTO bytes", outgoing.len(), space);
3440            self.spaces[space].pending.crypto.push_back(frame::Crypto {
3441                offset,
3442                data: outgoing,
3443            });
3444        }
3445    }
3446
3447    /// Switch to stronger cryptography during handshake
3448    fn upgrade_crypto(&mut self, space: SpaceId, crypto: Keys) {
3449        debug_assert!(
3450            self.spaces[space].crypto.is_none(),
3451            "already reached packet space {space:?}"
3452        );
3453        trace!("{:?} keys ready", space);
3454        if space == SpaceId::Data {
3455            // Precompute the first key update
3456            self.next_crypto = Some(
3457                self.crypto
3458                    .next_1rtt_keys()
3459                    .expect("handshake should be complete"),
3460            );
3461        }
3462
3463        self.spaces[space].crypto = Some(crypto);
3464        debug_assert!(space as usize > self.highest_space as usize);
3465        self.highest_space = space;
3466        if space == SpaceId::Data && self.side.is_client() {
3467            // Discard 0-RTT keys because 1-RTT keys are available.
3468            self.zero_rtt_crypto = None;
3469        }
3470    }
3471
3472    fn discard_space(&mut self, now: Instant, space_id: SpaceId) {
3473        debug_assert!(space_id != SpaceId::Data);
3474        trace!("discarding {:?} keys", space_id);
3475        if space_id == SpaceId::Initial {
3476            // No longer needed
3477            if let ConnectionSide::Client { token, .. } = &mut self.side {
3478                *token = Bytes::new();
3479            }
3480        }
3481        let space = &mut self.spaces[space_id];
3482        space.crypto = None;
3483        let pns = space.for_path(PathId::ZERO);
3484        pns.time_of_last_ack_eliciting_packet = None;
3485        pns.loss_time = None;
3486        pns.loss_probes = 0;
3487        let sent_packets = mem::take(&mut pns.sent_packets);
3488        let path = self.paths.get_mut(&PathId::ZERO).unwrap();
3489        for (_, packet) in sent_packets.into_iter() {
3490            path.data.remove_in_flight(&packet);
3491        }
3492
3493        self.set_loss_detection_timer(now, PathId::ZERO)
3494    }
3495
3496    fn handle_coalesced(
3497        &mut self,
3498        now: Instant,
3499        remote: SocketAddr,
3500        path_id: PathId,
3501        ecn: Option<EcnCodepoint>,
3502        data: BytesMut,
3503    ) {
3504        self.path_data_mut(path_id)
3505            .inc_total_recvd(data.len() as u64);
3506        let mut remaining = Some(data);
3507        let cid_len = self
3508            .local_cid_state
3509            .values()
3510            .map(|cid_state| cid_state.cid_len())
3511            .next()
3512            .expect("one cid_state must exist");
3513        while let Some(data) = remaining {
3514            match PartialDecode::new(
3515                data,
3516                &FixedLengthConnectionIdParser::new(cid_len),
3517                &[self.version],
3518                self.endpoint_config.grease_quic_bit,
3519            ) {
3520                Ok((partial_decode, rest)) => {
3521                    remaining = rest;
3522                    self.handle_decode(now, remote, path_id, ecn, partial_decode);
3523                }
3524                Err(e) => {
3525                    trace!("malformed header: {}", e);
3526                    return;
3527                }
3528            }
3529        }
3530    }
3531
3532    fn handle_decode(
3533        &mut self,
3534        now: Instant,
3535        remote: SocketAddr,
3536        path_id: PathId,
3537        ecn: Option<EcnCodepoint>,
3538        partial_decode: PartialDecode,
3539    ) {
3540        let qlog = QlogRecvPacket::new(partial_decode.len());
3541        if let Some(decoded) = packet_crypto::unprotect_header(
3542            partial_decode,
3543            &self.spaces,
3544            self.zero_rtt_crypto.as_ref(),
3545            self.peer_params.stateless_reset_token,
3546        ) {
3547            self.handle_packet(
3548                now,
3549                remote,
3550                path_id,
3551                ecn,
3552                decoded.packet,
3553                decoded.stateless_reset,
3554                qlog,
3555            );
3556        }
3557    }
3558
3559    fn handle_packet(
3560        &mut self,
3561        now: Instant,
3562        remote: SocketAddr,
3563        path_id: PathId,
3564        ecn: Option<EcnCodepoint>,
3565        packet: Option<Packet>,
3566        stateless_reset: bool,
3567        mut qlog: QlogRecvPacket,
3568    ) {
3569        self.stats.udp_rx.ios += 1;
3570        if let Some(ref packet) = packet {
3571            trace!(
3572                "got {:?} packet ({} bytes) from {} using id {}",
3573                packet.header.space(),
3574                packet.payload.len() + packet.header_data.len(),
3575                remote,
3576                packet.header.dst_cid(),
3577            );
3578        }
3579
3580        if self.is_handshaking() {
3581            if path_id != PathId::ZERO {
3582                debug!(%remote, %path_id, "discarding multipath packet during handshake");
3583                return;
3584            }
3585            if remote != self.path_data_mut(path_id).remote {
3586                if let Some(hs) = self.state.as_handshake() {
3587                    if hs.allow_server_migration {
3588                        trace!(?remote, prev = ?self.path_data(path_id).remote, "server migrated to new remote");
3589                        self.path_data_mut(path_id).remote = remote;
3590                        self.qlog.emit_tuple_assigned(path_id, remote, now);
3591                    } else {
3592                        debug!("discarding packet with unexpected remote during handshake");
3593                        return;
3594                    }
3595                } else {
3596                    debug!("discarding packet with unexpected remote during handshake");
3597                    return;
3598                }
3599            }
3600        }
3601
3602        let was_closed = self.state.is_closed();
3603        let was_drained = self.state.is_drained();
3604
3605        let decrypted = match packet {
3606            None => Err(None),
3607            Some(mut packet) => self
3608                .decrypt_packet(now, path_id, &mut packet)
3609                .map(move |number| (packet, number)),
3610        };
3611        let result = match decrypted {
3612            _ if stateless_reset => {
3613                debug!("got stateless reset");
3614                Err(ConnectionError::Reset)
3615            }
3616            Err(Some(e)) => {
3617                warn!("illegal packet: {}", e);
3618                Err(e.into())
3619            }
3620            Err(None) => {
3621                debug!("failed to authenticate packet");
3622                self.authentication_failures += 1;
3623                let integrity_limit = self.spaces[self.highest_space]
3624                    .crypto
3625                    .as_ref()
3626                    .unwrap()
3627                    .packet
3628                    .local
3629                    .integrity_limit();
3630                if self.authentication_failures > integrity_limit {
3631                    Err(TransportError::AEAD_LIMIT_REACHED("integrity limit violated").into())
3632                } else {
3633                    return;
3634                }
3635            }
3636            Ok((packet, number)) => {
3637                qlog.header(&packet.header, number, path_id);
3638                let span = match number {
3639                    Some(pn) => trace_span!("recv", space = ?packet.header.space(), pn),
3640                    None => trace_span!("recv", space = ?packet.header.space()),
3641                };
3642                let _guard = span.enter();
3643
3644                let dedup = self.spaces[packet.header.space()]
3645                    .path_space_mut(path_id)
3646                    .map(|pns| &mut pns.dedup);
3647                if number.zip(dedup).is_some_and(|(n, d)| d.insert(n)) {
3648                    debug!("discarding possible duplicate packet");
3649                    self.qlog.emit_packet_received(qlog, now);
3650                    return;
3651                } else if self.state.is_handshake() && packet.header.is_short() {
3652                    // TODO: SHOULD buffer these to improve reordering tolerance.
3653                    trace!("dropping short packet during handshake");
3654                    self.qlog.emit_packet_received(qlog, now);
3655                    return;
3656                } else {
3657                    if let Header::Initial(InitialHeader { ref token, .. }) = packet.header {
3658                        if let Some(hs) = self.state.as_handshake() {
3659                            if self.side.is_server() && token != &hs.expected_token {
3660                                // Clients must send the same retry token in every Initial. Initial
3661                                // packets can be spoofed, so we discard rather than killing the
3662                                // connection.
3663                                warn!("discarding Initial with invalid retry token");
3664                                self.qlog.emit_packet_received(qlog, now);
3665                                return;
3666                            }
3667                        }
3668                    }
3669
3670                    if !self.state.is_closed() {
3671                        let spin = match packet.header {
3672                            Header::Short { spin, .. } => spin,
3673                            _ => false,
3674                        };
3675
3676                        if self.side().is_server() && !self.abandoned_paths.contains(&path_id) {
3677                            // Only the client is allowed to open paths
3678                            self.ensure_path(path_id, remote, now, number);
3679                        }
3680                        if self.paths.contains_key(&path_id) {
3681                            self.on_packet_authenticated(
3682                                now,
3683                                packet.header.space(),
3684                                path_id,
3685                                ecn,
3686                                number,
3687                                spin,
3688                                packet.header.is_1rtt(),
3689                            );
3690                        }
3691                    }
3692
3693                    let res = self
3694                        .process_decrypted_packet(now, remote, path_id, number, packet, &mut qlog);
3695
3696                    self.qlog.emit_packet_received(qlog, now);
3697                    res
3698                }
3699            }
3700        };
3701
3702        // State transitions for error cases
3703        if let Err(conn_err) = result {
3704            match conn_err {
3705                ConnectionError::ApplicationClosed(reason) => self.state.move_to_closed(reason),
3706                ConnectionError::ConnectionClosed(reason) => self.state.move_to_closed(reason),
3707                ConnectionError::Reset
3708                | ConnectionError::TransportError(TransportError {
3709                    code: TransportErrorCode::AEAD_LIMIT_REACHED,
3710                    ..
3711                }) => {
3712                    self.state.move_to_drained(Some(conn_err));
3713                }
3714                ConnectionError::TimedOut => {
3715                    unreachable!("timeouts aren't generated by packet processing");
3716                }
3717                ConnectionError::TransportError(err) => {
3718                    debug!("closing connection due to transport error: {}", err);
3719                    self.state.move_to_closed(err);
3720                }
3721                ConnectionError::VersionMismatch => {
3722                    self.state.move_to_draining(Some(conn_err));
3723                }
3724                ConnectionError::LocallyClosed => {
3725                    unreachable!("LocallyClosed isn't generated by packet processing");
3726                }
3727                ConnectionError::CidsExhausted => {
3728                    unreachable!("CidsExhausted isn't generated by packet processing");
3729                }
3730            };
3731        }
3732
3733        if !was_closed && self.state.is_closed() {
3734            self.close_common();
3735            if !self.state.is_drained() {
3736                self.set_close_timer(now);
3737            }
3738        }
3739        if !was_drained && self.state.is_drained() {
3740            self.endpoint_events.push_back(EndpointEventInner::Drained);
3741            // Close timer may have been started previously, e.g. if we sent a close and got a
3742            // stateless reset in response
3743            self.timers
3744                .stop(Timer::Conn(ConnTimer::Close), self.qlog.with_time(now));
3745        }
3746
3747        // Transmit CONNECTION_CLOSE if necessary
3748        if matches!(self.state.as_type(), StateType::Closed) {
3749            // If there is no PathData for this PathId the packet was for a brand new
3750            // path. It was a valid packet however, so the remote is valid and we want to
3751            // send CONNECTION_CLOSE.
3752            let path_remote = self
3753                .paths
3754                .get(&path_id)
3755                .map(|p| p.data.remote)
3756                .unwrap_or(remote);
3757            self.close = remote == path_remote;
3758        }
3759    }
3760
3761    fn process_decrypted_packet(
3762        &mut self,
3763        now: Instant,
3764        remote: SocketAddr,
3765        path_id: PathId,
3766        number: Option<u64>,
3767        packet: Packet,
3768        qlog: &mut QlogRecvPacket,
3769    ) -> Result<(), ConnectionError> {
3770        if !self.paths.contains_key(&path_id) {
3771            // There is a chance this is a server side, first (for this path) packet, which would
3772            // be a protocol violation. It's more likely, however, that this is a packet of a
3773            // pruned path
3774            trace!(%path_id, ?number, "discarding packet for unknown path");
3775            return Ok(());
3776        }
3777        let state = match self.state.as_type() {
3778            StateType::Established => {
3779                match packet.header.space() {
3780                    SpaceId::Data => {
3781                        self.process_payload(now, remote, path_id, number.unwrap(), packet, qlog)?
3782                    }
3783                    _ if packet.header.has_frames() => {
3784                        self.process_early_payload(now, path_id, packet, qlog)?
3785                    }
3786                    _ => {
3787                        trace!("discarding unexpected pre-handshake packet");
3788                    }
3789                }
3790                return Ok(());
3791            }
3792            StateType::Closed => {
3793                for result in frame::Iter::new(packet.payload.freeze())? {
3794                    let frame = match result {
3795                        Ok(frame) => frame,
3796                        Err(err) => {
3797                            debug!("frame decoding error: {err:?}");
3798                            continue;
3799                        }
3800                    };
3801                    qlog.frame(&frame);
3802
3803                    if let Frame::Padding = frame {
3804                        continue;
3805                    };
3806
3807                    self.stats.frame_rx.record(&frame);
3808
3809                    if let Frame::Close(_error) = frame {
3810                        trace!("draining");
3811                        self.state.move_to_draining(None);
3812                        break;
3813                    }
3814                }
3815                return Ok(());
3816            }
3817            StateType::Draining | StateType::Drained => return Ok(()),
3818            StateType::Handshake => self.state.as_handshake_mut().expect("checked"),
3819        };
3820
3821        match packet.header {
3822            Header::Retry {
3823                src_cid: rem_cid, ..
3824            } => {
3825                debug_assert_eq!(path_id, PathId::ZERO);
3826                if self.side.is_server() {
3827                    return Err(TransportError::PROTOCOL_VIOLATION("client sent Retry").into());
3828                }
3829
3830                let is_valid_retry = self
3831                    .rem_cids
3832                    .get(&path_id)
3833                    .map(|cids| cids.active())
3834                    .map(|orig_dst_cid| {
3835                        self.crypto.is_valid_retry(
3836                            orig_dst_cid,
3837                            &packet.header_data,
3838                            &packet.payload,
3839                        )
3840                    })
3841                    .unwrap_or_default();
3842                if self.total_authed_packets > 1
3843                            || packet.payload.len() <= 16 // token + 16 byte tag
3844                            || !is_valid_retry
3845                {
3846                    trace!("discarding invalid Retry");
3847                    // - After the client has received and processed an Initial or Retry
3848                    //   packet from the server, it MUST discard any subsequent Retry
3849                    //   packets that it receives.
3850                    // - A client MUST discard a Retry packet with a zero-length Retry Token
3851                    //   field.
3852                    // - Clients MUST discard Retry packets that have a Retry Integrity Tag
3853                    //   that cannot be validated
3854                    return Ok(());
3855                }
3856
3857                trace!("retrying with CID {}", rem_cid);
3858                let client_hello = state.client_hello.take().unwrap();
3859                self.retry_src_cid = Some(rem_cid);
3860                self.rem_cids
3861                    .get_mut(&path_id)
3862                    .expect("PathId::ZERO not yet abandoned, is_valid_retry would have been false")
3863                    .update_initial_cid(rem_cid);
3864                self.rem_handshake_cid = rem_cid;
3865
3866                let space = &mut self.spaces[SpaceId::Initial];
3867                if let Some(info) = space.for_path(PathId::ZERO).take(0) {
3868                    self.on_packet_acked(now, PathId::ZERO, info);
3869                };
3870
3871                self.discard_space(now, SpaceId::Initial); // Make sure we clean up after
3872                // any retransmitted Initials
3873                self.spaces[SpaceId::Initial] = {
3874                    let mut space = PacketSpace::new(now, SpaceId::Initial, &mut self.rng);
3875                    space.crypto = Some(self.crypto.initial_keys(rem_cid, self.side.side()));
3876                    space.crypto_offset = client_hello.len() as u64;
3877                    space.for_path(path_id).next_packet_number = self.spaces[SpaceId::Initial]
3878                        .for_path(path_id)
3879                        .next_packet_number;
3880                    space.pending.crypto.push_back(frame::Crypto {
3881                        offset: 0,
3882                        data: client_hello,
3883                    });
3884                    space
3885                };
3886
3887                // Retransmit all 0-RTT data
3888                let zero_rtt = mem::take(
3889                    &mut self.spaces[SpaceId::Data]
3890                        .for_path(PathId::ZERO)
3891                        .sent_packets,
3892                );
3893                for (_, info) in zero_rtt.into_iter() {
3894                    self.paths
3895                        .get_mut(&PathId::ZERO)
3896                        .unwrap()
3897                        .remove_in_flight(&info);
3898                    self.spaces[SpaceId::Data].pending |= info.retransmits;
3899                }
3900                self.streams.retransmit_all_for_0rtt();
3901
3902                let token_len = packet.payload.len() - 16;
3903                let ConnectionSide::Client { ref mut token, .. } = self.side else {
3904                    unreachable!("we already short-circuited if we're server");
3905                };
3906                *token = packet.payload.freeze().split_to(token_len);
3907
3908                self.state = State::handshake(state::Handshake {
3909                    expected_token: Bytes::new(),
3910                    rem_cid_set: false,
3911                    client_hello: None,
3912                    allow_server_migration: true,
3913                });
3914                Ok(())
3915            }
3916            Header::Long {
3917                ty: LongType::Handshake,
3918                src_cid: rem_cid,
3919                dst_cid: loc_cid,
3920                ..
3921            } => {
3922                debug_assert_eq!(path_id, PathId::ZERO);
3923                if rem_cid != self.rem_handshake_cid {
3924                    debug!(
3925                        "discarding packet with mismatched remote CID: {} != {}",
3926                        self.rem_handshake_cid, rem_cid
3927                    );
3928                    return Ok(());
3929                }
3930                self.on_path_validated(path_id);
3931
3932                self.process_early_payload(now, path_id, packet, qlog)?;
3933                if self.state.is_closed() {
3934                    return Ok(());
3935                }
3936
3937                if self.crypto.is_handshaking() {
3938                    trace!("handshake ongoing");
3939                    return Ok(());
3940                }
3941
3942                if self.side.is_client() {
3943                    // Client-only because server params were set from the client's Initial
3944                    let params = self.crypto.transport_parameters()?.ok_or_else(|| {
3945                        TransportError::new(
3946                            TransportErrorCode::crypto(0x6d),
3947                            "transport parameters missing".to_owned(),
3948                        )
3949                    })?;
3950
3951                    if self.has_0rtt() {
3952                        if !self.crypto.early_data_accepted().unwrap() {
3953                            debug_assert!(self.side.is_client());
3954                            debug!("0-RTT rejected");
3955                            self.accepted_0rtt = false;
3956                            self.streams.zero_rtt_rejected();
3957
3958                            // Discard already-queued frames
3959                            self.spaces[SpaceId::Data].pending = Retransmits::default();
3960
3961                            // Discard 0-RTT packets
3962                            let sent_packets = mem::take(
3963                                &mut self.spaces[SpaceId::Data].for_path(path_id).sent_packets,
3964                            );
3965                            for (_, packet) in sent_packets.into_iter() {
3966                                self.paths
3967                                    .get_mut(&path_id)
3968                                    .unwrap()
3969                                    .remove_in_flight(&packet);
3970                            }
3971                        } else {
3972                            self.accepted_0rtt = true;
3973                            params.validate_resumption_from(&self.peer_params)?;
3974                        }
3975                    }
3976                    if let Some(token) = params.stateless_reset_token {
3977                        let remote = self.path_data(path_id).remote;
3978                        self.endpoint_events
3979                            .push_back(EndpointEventInner::ResetToken(path_id, remote, token));
3980                    }
3981                    self.handle_peer_params(params, loc_cid, rem_cid, now)?;
3982                    self.issue_first_cids(now);
3983                } else {
3984                    // Server-only
3985                    self.spaces[SpaceId::Data].pending.handshake_done = true;
3986                    self.discard_space(now, SpaceId::Handshake);
3987                    self.events.push_back(Event::HandshakeConfirmed);
3988                    trace!("handshake confirmed");
3989                }
3990
3991                self.events.push_back(Event::Connected);
3992                self.state.move_to_established();
3993                trace!("established");
3994
3995                // Multipath can only be enabled after the state has reached Established.
3996                // So this can not happen any earlier.
3997                self.issue_first_path_cids(now);
3998                Ok(())
3999            }
4000            Header::Initial(InitialHeader {
4001                src_cid: rem_cid,
4002                dst_cid: loc_cid,
4003                ..
4004            }) => {
4005                debug_assert_eq!(path_id, PathId::ZERO);
4006                if !state.rem_cid_set {
4007                    trace!("switching remote CID to {}", rem_cid);
4008                    let mut state = state.clone();
4009                    self.rem_cids
4010                        .get_mut(&path_id)
4011                        .expect("PathId::ZERO not yet abandoned")
4012                        .update_initial_cid(rem_cid);
4013                    self.rem_handshake_cid = rem_cid;
4014                    self.orig_rem_cid = rem_cid;
4015                    state.rem_cid_set = true;
4016                    self.state.move_to_handshake(state);
4017                } else if rem_cid != self.rem_handshake_cid {
4018                    debug!(
4019                        "discarding packet with mismatched remote CID: {} != {}",
4020                        self.rem_handshake_cid, rem_cid
4021                    );
4022                    return Ok(());
4023                }
4024
4025                let starting_space = self.highest_space;
4026                self.process_early_payload(now, path_id, packet, qlog)?;
4027
4028                if self.side.is_server()
4029                    && starting_space == SpaceId::Initial
4030                    && self.highest_space != SpaceId::Initial
4031                {
4032                    let params = self.crypto.transport_parameters()?.ok_or_else(|| {
4033                        TransportError::new(
4034                            TransportErrorCode::crypto(0x6d),
4035                            "transport parameters missing".to_owned(),
4036                        )
4037                    })?;
4038                    self.handle_peer_params(params, loc_cid, rem_cid, now)?;
4039                    self.issue_first_cids(now);
4040                    self.init_0rtt(now);
4041                }
4042                Ok(())
4043            }
4044            Header::Long {
4045                ty: LongType::ZeroRtt,
4046                ..
4047            } => {
4048                self.process_payload(now, remote, path_id, number.unwrap(), packet, qlog)?;
4049                Ok(())
4050            }
4051            Header::VersionNegotiate { .. } => {
4052                if self.total_authed_packets > 1 {
4053                    return Ok(());
4054                }
4055                let supported = packet
4056                    .payload
4057                    .chunks(4)
4058                    .any(|x| match <[u8; 4]>::try_from(x) {
4059                        Ok(version) => self.version == u32::from_be_bytes(version),
4060                        Err(_) => false,
4061                    });
4062                if supported {
4063                    return Ok(());
4064                }
4065                debug!("remote doesn't support our version");
4066                Err(ConnectionError::VersionMismatch)
4067            }
4068            Header::Short { .. } => unreachable!(
4069                "short packets received during handshake are discarded in handle_packet"
4070            ),
4071        }
4072    }
4073
4074    /// Process an Initial or Handshake packet payload
4075    fn process_early_payload(
4076        &mut self,
4077        now: Instant,
4078        path_id: PathId,
4079        packet: Packet,
4080        #[allow(unused)] qlog: &mut QlogRecvPacket,
4081    ) -> Result<(), TransportError> {
4082        debug_assert_ne!(packet.header.space(), SpaceId::Data);
4083        debug_assert_eq!(path_id, PathId::ZERO);
4084        let payload_len = packet.payload.len();
4085        let mut ack_eliciting = false;
4086        for result in frame::Iter::new(packet.payload.freeze())? {
4087            let frame = result?;
4088            qlog.frame(&frame);
4089            let span = match frame {
4090                Frame::Padding => continue,
4091                _ => Some(trace_span!("frame", ty = %frame.ty(), path = tracing::field::Empty)),
4092            };
4093
4094            self.stats.frame_rx.record(&frame);
4095
4096            let _guard = span.as_ref().map(|x| x.enter());
4097            ack_eliciting |= frame.is_ack_eliciting();
4098
4099            // Process frames
4100            if frame.is_1rtt() && packet.header.space() != SpaceId::Data {
4101                return Err(TransportError::PROTOCOL_VIOLATION(
4102                    "illegal frame type in handshake",
4103                ));
4104            }
4105
4106            match frame {
4107                Frame::Padding | Frame::Ping => {}
4108                Frame::Crypto(frame) => {
4109                    self.read_crypto(packet.header.space(), &frame, payload_len)?;
4110                }
4111                Frame::Ack(ack) => {
4112                    self.on_ack_received(now, packet.header.space(), ack)?;
4113                }
4114                Frame::PathAck(ack) => {
4115                    span.as_ref()
4116                        .map(|span| span.record("path", tracing::field::debug(&ack.path_id)));
4117                    self.on_path_ack_received(now, packet.header.space(), ack)?;
4118                }
4119                Frame::Close(reason) => {
4120                    self.state.move_to_draining(Some(reason.into()));
4121                    return Ok(());
4122                }
4123                _ => {
4124                    let mut err =
4125                        TransportError::PROTOCOL_VIOLATION("illegal frame type in handshake");
4126                    err.frame = Some(frame.ty());
4127                    return Err(err);
4128                }
4129            }
4130        }
4131
4132        if ack_eliciting {
4133            // In the initial and handshake spaces, ACKs must be sent immediately
4134            self.spaces[packet.header.space()]
4135                .for_path(path_id)
4136                .pending_acks
4137                .set_immediate_ack_required();
4138        }
4139
4140        self.write_crypto();
4141        Ok(())
4142    }
4143
4144    /// Processes the packet payload, always in the data space.
4145    fn process_payload(
4146        &mut self,
4147        now: Instant,
4148        remote: SocketAddr,
4149        path_id: PathId,
4150        number: u64,
4151        packet: Packet,
4152        #[allow(unused)] qlog: &mut QlogRecvPacket,
4153    ) -> Result<(), TransportError> {
4154        let payload = packet.payload.freeze();
4155        let mut is_probing_packet = true;
4156        let mut close = None;
4157        let payload_len = payload.len();
4158        let mut ack_eliciting = false;
4159        // if this packet triggers a path migration and includes a observed address frame, it's
4160        // stored here
4161        let mut migration_observed_addr = None;
4162        for result in frame::Iter::new(payload)? {
4163            let frame = result?;
4164            qlog.frame(&frame);
4165            let span = match frame {
4166                Frame::Padding => continue,
4167                _ => trace_span!("frame", ty = %frame.ty(), path = tracing::field::Empty),
4168            };
4169
4170            self.stats.frame_rx.record(&frame);
4171            // Crypto, Stream and Datagram frames are special cased in order no pollute
4172            // the log with payload data
4173            match &frame {
4174                Frame::Crypto(f) => {
4175                    trace!(offset = f.offset, len = f.data.len(), "got crypto frame");
4176                }
4177                Frame::Stream(f) => {
4178                    trace!(id = %f.id, offset = f.offset, len = f.data.len(), fin = f.fin, "got stream frame");
4179                }
4180                Frame::Datagram(f) => {
4181                    trace!(len = f.data.len(), "got datagram frame");
4182                }
4183                f => {
4184                    trace!("got frame {:?}", f);
4185                }
4186            }
4187
4188            let _guard = span.enter();
4189            if packet.header.is_0rtt() {
4190                match frame {
4191                    Frame::Crypto(_) | Frame::Close(Close::Application(_)) => {
4192                        return Err(TransportError::PROTOCOL_VIOLATION(
4193                            "illegal frame type in 0-RTT",
4194                        ));
4195                    }
4196                    _ => {
4197                        if frame.is_1rtt() {
4198                            return Err(TransportError::PROTOCOL_VIOLATION(
4199                                "illegal frame type in 0-RTT",
4200                            ));
4201                        }
4202                    }
4203                }
4204            }
4205            ack_eliciting |= frame.is_ack_eliciting();
4206
4207            // Check whether this could be a probing packet
4208            match frame {
4209                Frame::Padding
4210                | Frame::PathChallenge(_)
4211                | Frame::PathResponse(_)
4212                | Frame::NewConnectionId(_)
4213                | Frame::ObservedAddr(_) => {}
4214                _ => {
4215                    is_probing_packet = false;
4216                }
4217            }
4218
4219            match frame {
4220                Frame::Crypto(frame) => {
4221                    self.read_crypto(SpaceId::Data, &frame, payload_len)?;
4222                }
4223                Frame::Stream(frame) => {
4224                    if self.streams.received(frame, payload_len)?.should_transmit() {
4225                        self.spaces[SpaceId::Data].pending.max_data = true;
4226                    }
4227                }
4228                Frame::Ack(ack) => {
4229                    self.on_ack_received(now, SpaceId::Data, ack)?;
4230                }
4231                Frame::PathAck(ack) => {
4232                    span.record("path", tracing::field::debug(&ack.path_id));
4233                    self.on_path_ack_received(now, SpaceId::Data, ack)?;
4234                }
4235                Frame::Padding | Frame::Ping => {}
4236                Frame::Close(reason) => {
4237                    close = Some(reason);
4238                }
4239                Frame::PathChallenge(challenge) => {
4240                    let path = &mut self
4241                        .path_mut(path_id)
4242                        .expect("payload is processed only after the path becomes known");
4243                    path.path_responses.push(number, challenge.0, remote);
4244                    if remote == path.remote {
4245                        // PATH_CHALLENGE on active path, possible off-path packet forwarding
4246                        // attack. Send a non-probing packet to recover the active path.
4247                        // TODO(flub): No longer true! We now path_challege also to validate
4248                        //    the path if the path is new, without an RFC9000-style
4249                        //    migration involved. This means we add in an extra
4250                        //    IMMEDIATE_ACK on some challenges. It isn't really wrong to do
4251                        //    so, but it still is something untidy. We should instead
4252                        //    suppress this when we know the remote is still validating the
4253                        //    path.
4254                        match self.peer_supports_ack_frequency() {
4255                            true => self.immediate_ack(path_id),
4256                            false => {
4257                                self.ping_path(path_id).ok();
4258                            }
4259                        }
4260                    }
4261                }
4262                Frame::PathResponse(response) => {
4263                    let path = self
4264                        .paths
4265                        .get_mut(&path_id)
4266                        .expect("payload is processed only after the path becomes known");
4267
4268                    use PathTimer::*;
4269                    use paths::OnPathResponseReceived::*;
4270                    match path.data.on_path_response_received(now, response.0, remote) {
4271                        OnPath { was_open } => {
4272                            let qlog = self.qlog.with_time(now);
4273
4274                            self.timers
4275                                .stop(Timer::PerPath(path_id, PathValidation), qlog.clone());
4276                            self.timers
4277                                .stop(Timer::PerPath(path_id, PathOpen), qlog.clone());
4278
4279                            let next_challenge = path
4280                                .data
4281                                .earliest_expiring_challenge()
4282                                .map(|time| time + self.ack_frequency.max_ack_delay_for_pto());
4283                            self.timers.set_or_stop(
4284                                Timer::PerPath(path_id, PathChallengeLost),
4285                                next_challenge,
4286                                qlog,
4287                            );
4288
4289                            if !was_open {
4290                                self.events
4291                                    .push_back(Event::Path(PathEvent::Opened { id: path_id }));
4292                                if let Some(observed) = path.data.last_observed_addr_report.as_ref()
4293                                {
4294                                    self.events.push_back(Event::Path(PathEvent::ObservedAddr {
4295                                        id: path_id,
4296                                        addr: observed.socket_addr(),
4297                                    }));
4298                                }
4299                            }
4300                            if let Some((_, ref mut prev)) = path.prev {
4301                                prev.challenges_sent.clear();
4302                                prev.send_new_challenge = false;
4303                            }
4304                        }
4305                        OffPath => {
4306                            debug!("Response to off-path PathChallenge!");
4307                            let next_challenge = path
4308                                .data
4309                                .earliest_expiring_challenge()
4310                                .map(|time| time + self.ack_frequency.max_ack_delay_for_pto());
4311                            self.timers.set_or_stop(
4312                                Timer::PerPath(path_id, PathChallengeLost),
4313                                next_challenge,
4314                                self.qlog.with_time(now),
4315                            );
4316                        }
4317                        Invalid { expected } => {
4318                            debug!(%response, from=%remote, %expected, "ignoring invalid PATH_RESPONSE")
4319                        }
4320                        Unknown => debug!(%response, "ignoring invalid PATH_RESPONSE"),
4321                    }
4322                }
4323                Frame::MaxData(bytes) => {
4324                    self.streams.received_max_data(bytes);
4325                }
4326                Frame::MaxStreamData { id, offset } => {
4327                    self.streams.received_max_stream_data(id, offset)?;
4328                }
4329                Frame::MaxStreams { dir, count } => {
4330                    self.streams.received_max_streams(dir, count)?;
4331                }
4332                Frame::ResetStream(frame) => {
4333                    if self.streams.received_reset(frame)?.should_transmit() {
4334                        self.spaces[SpaceId::Data].pending.max_data = true;
4335                    }
4336                }
4337                Frame::DataBlocked { offset } => {
4338                    debug!(offset, "peer claims to be blocked at connection level");
4339                }
4340                Frame::StreamDataBlocked { id, offset } => {
4341                    if id.initiator() == self.side.side() && id.dir() == Dir::Uni {
4342                        debug!("got STREAM_DATA_BLOCKED on send-only {}", id);
4343                        return Err(TransportError::STREAM_STATE_ERROR(
4344                            "STREAM_DATA_BLOCKED on send-only stream",
4345                        ));
4346                    }
4347                    debug!(
4348                        stream = %id,
4349                        offset, "peer claims to be blocked at stream level"
4350                    );
4351                }
4352                Frame::StreamsBlocked { dir, limit } => {
4353                    if limit > MAX_STREAM_COUNT {
4354                        return Err(TransportError::FRAME_ENCODING_ERROR(
4355                            "unrepresentable stream limit",
4356                        ));
4357                    }
4358                    debug!(
4359                        "peer claims to be blocked opening more than {} {} streams",
4360                        limit, dir
4361                    );
4362                }
4363                Frame::StopSending(frame::StopSending { id, error_code }) => {
4364                    if id.initiator() != self.side.side() {
4365                        if id.dir() == Dir::Uni {
4366                            debug!("got STOP_SENDING on recv-only {}", id);
4367                            return Err(TransportError::STREAM_STATE_ERROR(
4368                                "STOP_SENDING on recv-only stream",
4369                            ));
4370                        }
4371                    } else if self.streams.is_local_unopened(id) {
4372                        return Err(TransportError::STREAM_STATE_ERROR(
4373                            "STOP_SENDING on unopened stream",
4374                        ));
4375                    }
4376                    self.streams.received_stop_sending(id, error_code);
4377                }
4378                Frame::RetireConnectionId(frame::RetireConnectionId { path_id, sequence }) => {
4379                    if let Some(ref path_id) = path_id {
4380                        span.record("path", tracing::field::debug(&path_id));
4381                    }
4382                    let path_id = path_id.unwrap_or_default();
4383                    match self.local_cid_state.get_mut(&path_id) {
4384                        None => error!(?path_id, "RETIRE_CONNECTION_ID for unknown path"),
4385                        Some(cid_state) => {
4386                            let allow_more_cids = cid_state
4387                                .on_cid_retirement(sequence, self.peer_params.issue_cids_limit())?;
4388
4389                            // If the path has closed, we do not issue more CIDs for this path
4390                            // For details see  https://www.ietf.org/archive/id/draft-ietf-quic-multipath-17.html#section-3.2.2
4391                            // > an endpoint SHOULD provide new connection IDs for that path, if still open, using PATH_NEW_CONNECTION_ID frames.
4392                            let has_path = !self.abandoned_paths.contains(&path_id);
4393                            let allow_more_cids = allow_more_cids && has_path;
4394
4395                            self.endpoint_events
4396                                .push_back(EndpointEventInner::RetireConnectionId(
4397                                    now,
4398                                    path_id,
4399                                    sequence,
4400                                    allow_more_cids,
4401                                ));
4402                        }
4403                    }
4404                }
4405                Frame::NewConnectionId(frame) => {
4406                    let path_id = if let Some(path_id) = frame.path_id {
4407                        if !self.is_multipath_negotiated() {
4408                            return Err(TransportError::PROTOCOL_VIOLATION(
4409                                "received PATH_NEW_CONNECTION_ID frame when multipath was not negotiated",
4410                            ));
4411                        }
4412                        if path_id > self.local_max_path_id {
4413                            return Err(TransportError::PROTOCOL_VIOLATION(
4414                                "PATH_NEW_CONNECTION_ID contains path_id exceeding current max",
4415                            ));
4416                        }
4417                        path_id
4418                    } else {
4419                        PathId::ZERO
4420                    };
4421
4422                    if self.abandoned_paths.contains(&path_id) {
4423                        trace!("ignoring issued CID for abandoned path");
4424                        continue;
4425                    }
4426                    if let Some(ref path_id) = frame.path_id {
4427                        span.record("path", tracing::field::debug(&path_id));
4428                    }
4429                    let rem_cids = self
4430                        .rem_cids
4431                        .entry(path_id)
4432                        .or_insert_with(|| CidQueue::new(frame.id));
4433                    if rem_cids.active().is_empty() {
4434                        // TODO(@divma): is the entry removed later? (rem_cids.entry)
4435                        return Err(TransportError::PROTOCOL_VIOLATION(
4436                            "NEW_CONNECTION_ID when CIDs aren't in use",
4437                        ));
4438                    }
4439                    if frame.retire_prior_to > frame.sequence {
4440                        return Err(TransportError::PROTOCOL_VIOLATION(
4441                            "NEW_CONNECTION_ID retiring unissued CIDs",
4442                        ));
4443                    }
4444
4445                    use crate::cid_queue::InsertError;
4446                    match rem_cids.insert(frame) {
4447                        Ok(None) => {}
4448                        Ok(Some((retired, reset_token))) => {
4449                            let pending_retired =
4450                                &mut self.spaces[SpaceId::Data].pending.retire_cids;
4451                            /// Ensure `pending_retired` cannot grow without bound. Limit is
4452                            /// somewhat arbitrary but very permissive.
4453                            const MAX_PENDING_RETIRED_CIDS: u64 = CidQueue::LEN as u64 * 10;
4454                            // We don't bother counting in-flight frames because those are bounded
4455                            // by congestion control.
4456                            if (pending_retired.len() as u64)
4457                                .saturating_add(retired.end.saturating_sub(retired.start))
4458                                > MAX_PENDING_RETIRED_CIDS
4459                            {
4460                                return Err(TransportError::CONNECTION_ID_LIMIT_ERROR(
4461                                    "queued too many retired CIDs",
4462                                ));
4463                            }
4464                            pending_retired.extend(retired.map(|seq| (path_id, seq)));
4465                            self.set_reset_token(path_id, remote, reset_token);
4466                        }
4467                        Err(InsertError::ExceedsLimit) => {
4468                            return Err(TransportError::CONNECTION_ID_LIMIT_ERROR(""));
4469                        }
4470                        Err(InsertError::Retired) => {
4471                            trace!("discarding already-retired");
4472                            // RETIRE_CONNECTION_ID might not have been previously sent if e.g. a
4473                            // range of connection IDs larger than the active connection ID limit
4474                            // was retired all at once via retire_prior_to.
4475                            self.spaces[SpaceId::Data]
4476                                .pending
4477                                .retire_cids
4478                                .push((path_id, frame.sequence));
4479                            continue;
4480                        }
4481                    };
4482
4483                    if self.side.is_server()
4484                        && path_id == PathId::ZERO
4485                        && self
4486                            .rem_cids
4487                            .get(&PathId::ZERO)
4488                            .map(|cids| cids.active_seq() == 0)
4489                            .unwrap_or_default()
4490                    {
4491                        // We're a server still using the initial remote CID for the client, so
4492                        // let's switch immediately to enable clientside stateless resets.
4493                        self.update_rem_cid(PathId::ZERO);
4494                    }
4495                }
4496                Frame::NewToken(NewToken { token }) => {
4497                    let ConnectionSide::Client {
4498                        token_store,
4499                        server_name,
4500                        ..
4501                    } = &self.side
4502                    else {
4503                        return Err(TransportError::PROTOCOL_VIOLATION("client sent NEW_TOKEN"));
4504                    };
4505                    if token.is_empty() {
4506                        return Err(TransportError::FRAME_ENCODING_ERROR("empty token"));
4507                    }
4508                    trace!("got new token");
4509                    token_store.insert(server_name, token);
4510                }
4511                Frame::Datagram(datagram) => {
4512                    if self
4513                        .datagrams
4514                        .received(datagram, &self.config.datagram_receive_buffer_size)?
4515                    {
4516                        self.events.push_back(Event::DatagramReceived);
4517                    }
4518                }
4519                Frame::AckFrequency(ack_frequency) => {
4520                    // This frame can only be sent in the Data space
4521
4522                    if !self.ack_frequency.ack_frequency_received(&ack_frequency)? {
4523                        // The AckFrequency frame is stale (we have already received a more
4524                        // recent one)
4525                        continue;
4526                    }
4527
4528                    // Update the params for all of our paths
4529                    for (path_id, space) in self.spaces[SpaceId::Data].number_spaces.iter_mut() {
4530                        space.pending_acks.set_ack_frequency_params(&ack_frequency);
4531
4532                        // Our `max_ack_delay` has been updated, so we may need to adjust
4533                        // its associated timeout
4534                        if let Some(timeout) = space
4535                            .pending_acks
4536                            .max_ack_delay_timeout(self.ack_frequency.max_ack_delay)
4537                        {
4538                            self.timers.set(
4539                                Timer::PerPath(*path_id, PathTimer::MaxAckDelay),
4540                                timeout,
4541                                self.qlog.with_time(now),
4542                            );
4543                        }
4544                    }
4545                }
4546                Frame::ImmediateAck => {
4547                    // This frame can only be sent in the Data space
4548                    for pns in self.spaces[SpaceId::Data].iter_paths_mut() {
4549                        pns.pending_acks.set_immediate_ack_required();
4550                    }
4551                }
4552                Frame::HandshakeDone => {
4553                    if self.side.is_server() {
4554                        return Err(TransportError::PROTOCOL_VIOLATION(
4555                            "client sent HANDSHAKE_DONE",
4556                        ));
4557                    }
4558                    if self.spaces[SpaceId::Handshake].crypto.is_some() {
4559                        self.discard_space(now, SpaceId::Handshake);
4560                    }
4561                    self.events.push_back(Event::HandshakeConfirmed);
4562                    trace!("handshake confirmed");
4563                }
4564                Frame::ObservedAddr(observed) => {
4565                    // check if params allows the peer to send report and this node to receive it
4566                    trace!(seq_no = %observed.seq_no, ip = %observed.ip, port = observed.port);
4567                    if !self
4568                        .peer_params
4569                        .address_discovery_role
4570                        .should_report(&self.config.address_discovery_role)
4571                    {
4572                        return Err(TransportError::PROTOCOL_VIOLATION(
4573                            "received OBSERVED_ADDRESS frame when not negotiated",
4574                        ));
4575                    }
4576                    // must only be sent in data space
4577                    if packet.header.space() != SpaceId::Data {
4578                        return Err(TransportError::PROTOCOL_VIOLATION(
4579                            "OBSERVED_ADDRESS frame outside data space",
4580                        ));
4581                    }
4582
4583                    let path = self.path_data_mut(path_id);
4584                    if remote == path.remote {
4585                        if let Some(updated) = path.update_observed_addr_report(observed) {
4586                            if path.open {
4587                                self.events.push_back(Event::Path(PathEvent::ObservedAddr {
4588                                    id: path_id,
4589                                    addr: updated,
4590                                }));
4591                            }
4592                            // otherwise the event is reported when the path is deemed open
4593                        }
4594                    } else {
4595                        // include in migration
4596                        migration_observed_addr = Some(observed)
4597                    }
4598                }
4599                Frame::PathAbandon(frame::PathAbandon {
4600                    path_id,
4601                    error_code,
4602                }) => {
4603                    span.record("path", tracing::field::debug(&path_id));
4604                    // TODO(flub): don't really know which error code to use here.
4605                    let already_abandoned = match self.close_path(now, path_id, error_code.into()) {
4606                        Ok(()) => {
4607                            trace!("peer abandoned path");
4608                            false
4609                        }
4610                        Err(ClosePathError::LastOpenPath) => {
4611                            trace!("peer abandoned last path, closing connection");
4612                            // TODO(flub): which error code?
4613                            return Err(TransportError::NO_ERROR("last path abandoned by peer"));
4614                        }
4615                        Err(ClosePathError::ClosedPath) => {
4616                            trace!("peer abandoned already closed path");
4617                            true
4618                        }
4619                    };
4620                    // If we receive a retransmit of PATH_ABANDON then we may already have
4621                    // abandoned this path locally.  In that case the DiscardPath timer
4622                    // may already have fired and we no longer have any state for this path.
4623                    // Only set this timer if we still have path state.
4624                    if self.path(path_id).is_some() && !already_abandoned {
4625                        // TODO(flub): Checking is_some() here followed by a number of calls
4626                        //    that would panic if it was None is really ugly.  If only we
4627                        //    could do something like PathData::pto().  One day we'll have
4628                        //    unified SpaceId and PathId and this will be possible.
4629                        let delay = self.pto(SpaceId::Data, path_id) * 3;
4630                        self.timers.set(
4631                            Timer::PerPath(path_id, PathTimer::DiscardPath),
4632                            now + delay,
4633                            self.qlog.with_time(now),
4634                        );
4635                    }
4636                }
4637                Frame::PathStatusAvailable(info) => {
4638                    span.record("path", tracing::field::debug(&info.path_id));
4639                    if self.is_multipath_negotiated() {
4640                        self.on_path_status(
4641                            info.path_id,
4642                            PathStatus::Available,
4643                            info.status_seq_no,
4644                        );
4645                    } else {
4646                        return Err(TransportError::PROTOCOL_VIOLATION(
4647                            "received PATH_STATUS_AVAILABLE frame when multipath was not negotiated",
4648                        ));
4649                    }
4650                }
4651                Frame::PathStatusBackup(info) => {
4652                    span.record("path", tracing::field::debug(&info.path_id));
4653                    if self.is_multipath_negotiated() {
4654                        self.on_path_status(info.path_id, PathStatus::Backup, info.status_seq_no);
4655                    } else {
4656                        return Err(TransportError::PROTOCOL_VIOLATION(
4657                            "received PATH_STATUS_BACKUP frame when multipath was not negotiated",
4658                        ));
4659                    }
4660                }
4661                Frame::MaxPathId(frame::MaxPathId(path_id)) => {
4662                    span.record("path", tracing::field::debug(&path_id));
4663                    if !self.is_multipath_negotiated() {
4664                        return Err(TransportError::PROTOCOL_VIOLATION(
4665                            "received MAX_PATH_ID frame when multipath was not negotiated",
4666                        ));
4667                    }
4668                    // frames that do not increase the path id are ignored
4669                    if path_id > self.remote_max_path_id {
4670                        self.remote_max_path_id = path_id;
4671                        self.issue_first_path_cids(now);
4672                    }
4673                }
4674                Frame::PathsBlocked(frame::PathsBlocked(max_path_id)) => {
4675                    // Receipt of a value of Maximum Path Identifier or Path Identifier that is higher than the local maximum value MUST
4676                    // be treated as a connection error of type PROTOCOL_VIOLATION.
4677                    // Ref <https://www.ietf.org/archive/id/draft-ietf-quic-multipath-14.html#name-paths_blocked-and-path_cids>
4678                    if self.is_multipath_negotiated() {
4679                        if self.local_max_path_id > max_path_id {
4680                            return Err(TransportError::PROTOCOL_VIOLATION(
4681                                "PATHS_BLOCKED maximum path identifier was larger than local maximum",
4682                            ));
4683                        }
4684                        debug!("received PATHS_BLOCKED({:?})", max_path_id);
4685                        // TODO(@divma): ensure max concurrent paths
4686                    } else {
4687                        return Err(TransportError::PROTOCOL_VIOLATION(
4688                            "received PATHS_BLOCKED frame when not multipath was not negotiated",
4689                        ));
4690                    }
4691                }
4692                Frame::PathCidsBlocked(frame::PathCidsBlocked { path_id, next_seq }) => {
4693                    // Nothing to do.  This is recorded in the frame stats, but otherwise we
4694                    // always issue all CIDs we're allowed to issue, so either this is an
4695                    // impatient peer or a bug on our side.
4696
4697                    // Receipt of a value of Maximum Path Identifier or Path Identifier that is higher than the local maximum value MUST
4698                    // be treated as a connection error of type PROTOCOL_VIOLATION.
4699                    // Ref <https://www.ietf.org/archive/id/draft-ietf-quic-multipath-14.html#name-paths_blocked-and-path_cids>
4700                    if self.is_multipath_negotiated() {
4701                        if path_id > self.local_max_path_id {
4702                            return Err(TransportError::PROTOCOL_VIOLATION(
4703                                "PATH_CIDS_BLOCKED path identifier was larger than local maximum",
4704                            ));
4705                        }
4706                        if next_seq.0
4707                            > self
4708                                .local_cid_state
4709                                .get(&path_id)
4710                                .map(|cid_state| cid_state.active_seq().1 + 1)
4711                                .unwrap_or_default()
4712                        {
4713                            return Err(TransportError::PROTOCOL_VIOLATION(
4714                                "PATH_CIDS_BLOCKED next sequence number larger than in local state",
4715                            ));
4716                        }
4717                        debug!(%path_id, %next_seq, "received PATH_CIDS_BLOCKED");
4718                    } else {
4719                        return Err(TransportError::PROTOCOL_VIOLATION(
4720                            "received PATH_CIDS_BLOCKED frame when not multipath was not negotiated",
4721                        ));
4722                    }
4723                }
4724                Frame::AddAddress(addr) => {
4725                    let client_state = match self.iroh_hp.client_side_mut() {
4726                        Ok(state) => state,
4727                        Err(err) => {
4728                            return Err(TransportError::PROTOCOL_VIOLATION(format!(
4729                                "Nat traversal(ADD_ADDRESS): {err}"
4730                            )));
4731                        }
4732                    };
4733
4734                    if !client_state.check_remote_address(&addr) {
4735                        // if the address is not valid we flag it, but update anyway
4736                        warn!(?addr, "server sent illegal ADD_ADDRESS frame");
4737                    }
4738
4739                    match client_state.add_remote_address(addr) {
4740                        Ok(maybe_added) => {
4741                            if let Some(added) = maybe_added {
4742                                self.events.push_back(Event::NatTraversal(
4743                                    iroh_hp::Event::AddressAdded(added),
4744                                ));
4745                            }
4746                        }
4747                        Err(e) => {
4748                            warn!(%e, "failed to add remote address")
4749                        }
4750                    }
4751                }
4752                Frame::RemoveAddress(addr) => {
4753                    let client_state = match self.iroh_hp.client_side_mut() {
4754                        Ok(state) => state,
4755                        Err(err) => {
4756                            return Err(TransportError::PROTOCOL_VIOLATION(format!(
4757                                "Nat traversal(REMOVE_ADDRESS): {err}"
4758                            )));
4759                        }
4760                    };
4761                    if let Some(removed_addr) = client_state.remove_remote_address(addr) {
4762                        self.events
4763                            .push_back(Event::NatTraversal(iroh_hp::Event::AddressRemoved(
4764                                removed_addr,
4765                            )));
4766                    }
4767                }
4768                Frame::ReachOut(reach_out) => {
4769                    let server_state = match self.iroh_hp.server_side_mut() {
4770                        Ok(state) => state,
4771                        Err(err) => {
4772                            return Err(TransportError::PROTOCOL_VIOLATION(format!(
4773                                "Nat traversal(REACH_OUT): {err}"
4774                            )));
4775                        }
4776                    };
4777
4778                    if let Err(err) = server_state.handle_reach_out(reach_out) {
4779                        return Err(TransportError::PROTOCOL_VIOLATION(format!(
4780                            "Nat traversal(REACH_OUT): {err}"
4781                        )));
4782                    }
4783                }
4784            }
4785        }
4786
4787        let space = self.spaces[SpaceId::Data].for_path(path_id);
4788        if space
4789            .pending_acks
4790            .packet_received(now, number, ack_eliciting, &space.dedup)
4791        {
4792            if self.abandoned_paths.contains(&path_id) {
4793                // § 3.4.3 QUIC-MULTIPATH: promptly send ACKs for packets received from
4794                // abandoned paths.
4795                space.pending_acks.set_immediate_ack_required();
4796            } else {
4797                self.timers.set(
4798                    Timer::PerPath(path_id, PathTimer::MaxAckDelay),
4799                    now + self.ack_frequency.max_ack_delay,
4800                    self.qlog.with_time(now),
4801                );
4802            }
4803        }
4804
4805        // Issue stream ID credit due to ACKs of outgoing finish/resets and incoming finish/resets
4806        // on stopped streams. Incoming finishes/resets on open streams are not handled here as they
4807        // are only freed, and hence only issue credit, once the application has been notified
4808        // during a read on the stream.
4809        let pending = &mut self.spaces[SpaceId::Data].pending;
4810        self.streams.queue_max_stream_id(pending);
4811
4812        if let Some(reason) = close {
4813            self.state.move_to_draining(Some(reason.into()));
4814            self.close = true;
4815        }
4816
4817        if Some(number) == self.spaces[SpaceId::Data].for_path(path_id).rx_packet
4818            && !is_probing_packet
4819            && remote != self.path_data(path_id).remote
4820        {
4821            let ConnectionSide::Server { ref server_config } = self.side else {
4822                panic!("packets from unknown remote should be dropped by clients");
4823            };
4824            debug_assert!(
4825                server_config.migration,
4826                "migration-initiating packets should have been dropped immediately"
4827            );
4828            self.migrate(path_id, now, remote, migration_observed_addr);
4829            // Break linkability, if possible
4830            self.update_rem_cid(path_id);
4831            self.spin = false;
4832        }
4833
4834        Ok(())
4835    }
4836
4837    fn migrate(
4838        &mut self,
4839        path_id: PathId,
4840        now: Instant,
4841        remote: SocketAddr,
4842        observed_addr: Option<ObservedAddr>,
4843    ) {
4844        trace!(%remote, %path_id, "migration initiated");
4845        self.path_counter = self.path_counter.wrapping_add(1);
4846        // TODO(@divma): conditions for path migration in multipath are very specific, check them
4847        // again to prevent path migrations that should actually create a new path
4848
4849        // Reset rtt/congestion state for new path unless it looks like a NAT rebinding.
4850        // Note that the congestion window will not grow until validation terminates. Helps mitigate
4851        // amplification attacks performed by spoofing source addresses.
4852        let prev_pto = self.pto(SpaceId::Data, path_id);
4853        let known_path = self.paths.get_mut(&path_id).expect("known path");
4854        let path = &mut known_path.data;
4855        let mut new_path = if remote.is_ipv4() && remote.ip() == path.remote.ip() {
4856            PathData::from_previous(remote, path, self.path_counter, now)
4857        } else {
4858            let peer_max_udp_payload_size =
4859                u16::try_from(self.peer_params.max_udp_payload_size.into_inner())
4860                    .unwrap_or(u16::MAX);
4861            PathData::new(
4862                remote,
4863                self.allow_mtud,
4864                Some(peer_max_udp_payload_size),
4865                self.path_counter,
4866                now,
4867                &self.config,
4868            )
4869        };
4870        new_path.last_observed_addr_report = path.last_observed_addr_report.clone();
4871        if let Some(report) = observed_addr {
4872            if let Some(updated) = new_path.update_observed_addr_report(report) {
4873                tracing::info!("adding observed addr event from migration");
4874                self.events.push_back(Event::Path(PathEvent::ObservedAddr {
4875                    id: path_id,
4876                    addr: updated,
4877                }));
4878            }
4879        }
4880        new_path.send_new_challenge = true;
4881
4882        let mut prev = mem::replace(path, new_path);
4883        // Don't clobber the original path if the previous one hasn't been validated yet
4884        if !prev.is_validating_path() {
4885            prev.send_new_challenge = true;
4886            // We haven't updated the remote CID yet, this captures the remote CID we were using on
4887            // the previous path.
4888
4889            known_path.prev = Some((self.rem_cids.get(&path_id).unwrap().active(), prev));
4890        }
4891
4892        // We need to re-assign the correct remote to this path in qlog
4893        self.qlog.emit_tuple_assigned(path_id, remote, now);
4894
4895        self.timers.set(
4896            Timer::PerPath(path_id, PathTimer::PathValidation),
4897            now + 3 * cmp::max(self.pto(SpaceId::Data, path_id), prev_pto),
4898            self.qlog.with_time(now),
4899        );
4900    }
4901
4902    /// Handle a change in the local address, i.e. an active migration
4903    pub fn local_address_changed(&mut self) {
4904        // TODO(flub): if multipath is enabled this needs to create a new path entirely.
4905        self.update_rem_cid(PathId::ZERO);
4906        self.ping();
4907    }
4908
4909    /// Switch to a previously unused remote connection ID, if possible
4910    fn update_rem_cid(&mut self, path_id: PathId) {
4911        let Some((reset_token, retired)) =
4912            self.rem_cids.get_mut(&path_id).and_then(|cids| cids.next())
4913        else {
4914            return;
4915        };
4916
4917        // Retire the current remote CID and any CIDs we had to skip.
4918        self.spaces[SpaceId::Data]
4919            .pending
4920            .retire_cids
4921            .extend(retired.map(|seq| (path_id, seq)));
4922        let remote = self.path_data(path_id).remote;
4923        self.set_reset_token(path_id, remote, reset_token);
4924    }
4925
4926    /// Sends this reset token to the endpoint
4927    ///
4928    /// The endpoint needs to know the reset tokens issued by the peer, so that if the peer
4929    /// sends a reset token it knows to route it to this connection. See RFC 9000 section
4930    /// 10.3. Stateless Reset.
4931    ///
4932    /// Reset tokens are different for each path, the endpoint identifies paths by peer
4933    /// socket address however, not by path ID.
4934    fn set_reset_token(&mut self, path_id: PathId, remote: SocketAddr, reset_token: ResetToken) {
4935        self.endpoint_events
4936            .push_back(EndpointEventInner::ResetToken(path_id, remote, reset_token));
4937
4938        // During the handshake the server sends a reset token in the transport
4939        // parameters. When we are the client and we receive the reset token during the
4940        // handshake we want this to affect our peer transport parameters.
4941        // TODO(flub): Pretty sure this is pointless, the entire params is overwritten
4942        //    shortly after this was called.  And then the params don't have this anymore.
4943        if path_id == PathId::ZERO {
4944            self.peer_params.stateless_reset_token = Some(reset_token);
4945        }
4946    }
4947
4948    /// Issue an initial set of connection IDs to the peer upon connection
4949    fn issue_first_cids(&mut self, now: Instant) {
4950        if self
4951            .local_cid_state
4952            .get(&PathId::ZERO)
4953            .expect("PathId::ZERO exists when the connection is created")
4954            .cid_len()
4955            == 0
4956        {
4957            return;
4958        }
4959
4960        // Subtract 1 to account for the CID we supplied while handshaking
4961        let mut n = self.peer_params.issue_cids_limit() - 1;
4962        if let ConnectionSide::Server { server_config } = &self.side {
4963            if server_config.has_preferred_address() {
4964                // We also sent a CID in the transport parameters
4965                n -= 1;
4966            }
4967        }
4968        self.endpoint_events
4969            .push_back(EndpointEventInner::NeedIdentifiers(PathId::ZERO, now, n));
4970    }
4971
4972    /// Issues an initial set of CIDs for paths that have not yet had any CIDs issued
4973    ///
4974    /// Later CIDs are issued when CIDs expire or are retired by the peer.
4975    fn issue_first_path_cids(&mut self, now: Instant) {
4976        if let Some(max_path_id) = self.max_path_id() {
4977            let mut path_id = self.max_path_id_with_cids.next();
4978            while path_id <= max_path_id {
4979                self.endpoint_events
4980                    .push_back(EndpointEventInner::NeedIdentifiers(
4981                        path_id,
4982                        now,
4983                        self.peer_params.issue_cids_limit(),
4984                    ));
4985                path_id = path_id.next();
4986            }
4987            self.max_path_id_with_cids = max_path_id;
4988        }
4989    }
4990
4991    /// Populates a packet with frames
4992    ///
4993    /// This tries to fit as many frames as possible into the packet.
4994    ///
4995    /// *path_exclusive_only* means to only build frames which can only be sent on this
4996    /// *path.  This is used in multipath for backup paths while there is still an active
4997    /// *path.
4998    fn populate_packet(
4999        &mut self,
5000        now: Instant,
5001        space_id: SpaceId,
5002        path_id: PathId,
5003        path_exclusive_only: bool,
5004        buf: &mut impl BufMut,
5005        pn: u64,
5006        #[allow(unused)] qlog: &mut QlogSentPacket,
5007    ) -> SentFrames {
5008        let mut sent = SentFrames::default();
5009        let is_multipath_negotiated = self.is_multipath_negotiated();
5010        let space = &mut self.spaces[space_id];
5011        let path = &mut self.paths.get_mut(&path_id).expect("known path").data;
5012        let is_0rtt = space_id == SpaceId::Data && space.crypto.is_none();
5013        space
5014            .for_path(path_id)
5015            .pending_acks
5016            .maybe_ack_non_eliciting();
5017
5018        // HANDSHAKE_DONE
5019        if !is_0rtt && mem::replace(&mut space.pending.handshake_done, false) {
5020            trace!("HANDSHAKE_DONE");
5021            buf.write(frame::FrameType::HANDSHAKE_DONE);
5022            qlog.frame(&Frame::HandshakeDone);
5023            sent.retransmits.get_or_create().handshake_done = true;
5024            // This is just a u8 counter and the frame is typically just sent once
5025            self.stats.frame_tx.handshake_done =
5026                self.stats.frame_tx.handshake_done.saturating_add(1);
5027        }
5028
5029        // REACH_OUT
5030        // TODO(@divma): path explusive considerations
5031        if let Some((round, addresses)) = space.pending.reach_out.as_mut() {
5032            while let Some(local_addr) = addresses.pop() {
5033                let reach_out = frame::ReachOut::new(*round, local_addr);
5034                if buf.remaining_mut() > reach_out.size() {
5035                    trace!(%round, ?local_addr, "REACH_OUT");
5036                    reach_out.write(buf);
5037                    let sent_reachouts = sent
5038                        .retransmits
5039                        .get_or_create()
5040                        .reach_out
5041                        .get_or_insert_with(|| (*round, Default::default()));
5042                    sent_reachouts.1.push(local_addr);
5043                    self.stats.frame_tx.reach_out = self.stats.frame_tx.reach_out.saturating_add(1);
5044                    qlog.frame(&Frame::ReachOut(reach_out));
5045                } else {
5046                    addresses.push(local_addr);
5047                    break;
5048                }
5049            }
5050            if addresses.is_empty() {
5051                space.pending.reach_out = None;
5052            }
5053        }
5054
5055        // OBSERVED_ADDR
5056        if !path_exclusive_only
5057            && space_id == SpaceId::Data
5058            && self
5059                .config
5060                .address_discovery_role
5061                .should_report(&self.peer_params.address_discovery_role)
5062            && (!path.observed_addr_sent || space.pending.observed_addr)
5063        {
5064            let frame = frame::ObservedAddr::new(path.remote, self.next_observed_addr_seq_no);
5065            if buf.remaining_mut() > frame.size() {
5066                trace!(seq = %frame.seq_no, ip = %frame.ip, port = frame.port, "OBSERVED_ADDRESS");
5067                frame.write(buf);
5068
5069                self.next_observed_addr_seq_no = self.next_observed_addr_seq_no.saturating_add(1u8);
5070                path.observed_addr_sent = true;
5071
5072                self.stats.frame_tx.observed_addr += 1;
5073                sent.retransmits.get_or_create().observed_addr = true;
5074                space.pending.observed_addr = false;
5075                qlog.frame(&Frame::ObservedAddr(frame));
5076            }
5077        }
5078
5079        // PING
5080        if mem::replace(&mut space.for_path(path_id).ping_pending, false) {
5081            trace!("PING");
5082            buf.write(frame::FrameType::PING);
5083            sent.non_retransmits = true;
5084            self.stats.frame_tx.ping += 1;
5085            qlog.frame(&Frame::Ping);
5086        }
5087
5088        // IMMEDIATE_ACK
5089        if mem::replace(&mut space.for_path(path_id).immediate_ack_pending, false) {
5090            debug_assert_eq!(
5091                space_id,
5092                SpaceId::Data,
5093                "immediate acks must be sent in the data space"
5094            );
5095            trace!("IMMEDIATE_ACK");
5096            buf.write(frame::FrameType::IMMEDIATE_ACK);
5097            sent.non_retransmits = true;
5098            self.stats.frame_tx.immediate_ack += 1;
5099            qlog.frame(&Frame::ImmediateAck);
5100        }
5101
5102        // ACK
5103        // TODO(flub): Should this send acks for this path anyway?
5104
5105        if !path_exclusive_only {
5106            for path_id in space
5107                .number_spaces
5108                .iter_mut()
5109                .filter(|(_, pns)| pns.pending_acks.can_send())
5110                .map(|(&path_id, _)| path_id)
5111                .collect::<Vec<_>>()
5112            {
5113                Self::populate_acks(
5114                    now,
5115                    self.receiving_ecn,
5116                    &mut sent,
5117                    path_id,
5118                    space_id,
5119                    space,
5120                    is_multipath_negotiated,
5121                    buf,
5122                    &mut self.stats,
5123                    qlog,
5124                );
5125            }
5126        }
5127
5128        // ACK_FREQUENCY
5129        if !path_exclusive_only && mem::replace(&mut space.pending.ack_frequency, false) {
5130            let sequence_number = self.ack_frequency.next_sequence_number();
5131
5132            // Safe to unwrap because this is always provided when ACK frequency is enabled
5133            let config = self.config.ack_frequency_config.as_ref().unwrap();
5134
5135            // Ensure the delay is within bounds to avoid a PROTOCOL_VIOLATION error
5136            let max_ack_delay = self.ack_frequency.candidate_max_ack_delay(
5137                path.rtt.get(),
5138                config,
5139                &self.peer_params,
5140            );
5141
5142            trace!(?max_ack_delay, "ACK_FREQUENCY");
5143
5144            let frame = frame::AckFrequency {
5145                sequence: sequence_number,
5146                ack_eliciting_threshold: config.ack_eliciting_threshold,
5147                request_max_ack_delay: max_ack_delay.as_micros().try_into().unwrap_or(VarInt::MAX),
5148                reordering_threshold: config.reordering_threshold,
5149            };
5150            frame.encode(buf);
5151            qlog.frame(&Frame::AckFrequency(frame));
5152
5153            sent.retransmits.get_or_create().ack_frequency = true;
5154
5155            self.ack_frequency
5156                .ack_frequency_sent(path_id, pn, max_ack_delay);
5157            self.stats.frame_tx.ack_frequency += 1;
5158        }
5159
5160        // PATH_CHALLENGE
5161        if buf.remaining_mut() > frame::PathChallenge::SIZE_BOUND
5162            && space_id == SpaceId::Data
5163            && path.send_new_challenge
5164            && !self.state.is_closed()
5165        // we don't want to send new challenges if we are already closing
5166        {
5167            path.send_new_challenge = false;
5168
5169            // Generate a new challenge every time we send a new PATH_CHALLENGE
5170            let token = self.rng.random();
5171            let info = paths::SentChallengeInfo {
5172                sent_instant: now,
5173                remote: path.remote,
5174            };
5175            path.challenges_sent.insert(token, info);
5176            sent.non_retransmits = true;
5177            sent.requires_padding = true;
5178            let challenge = frame::PathChallenge(token);
5179            trace!(%challenge, "sending new challenge");
5180            buf.write(challenge);
5181            qlog.frame(&Frame::PathChallenge(challenge));
5182            self.stats.frame_tx.path_challenge += 1;
5183            let pto = self.ack_frequency.max_ack_delay_for_pto() + path.rtt.pto_base();
5184            self.timers.set(
5185                Timer::PerPath(path_id, PathTimer::PathChallengeLost),
5186                now + pto,
5187                self.qlog.with_time(now),
5188            );
5189
5190            if is_multipath_negotiated && !path.validated && path.send_new_challenge {
5191                // queue informing the path status along with the challenge
5192                space.pending.path_status.insert(path_id);
5193            }
5194
5195            // Always include an OBSERVED_ADDR frame with a PATH_CHALLENGE, regardless
5196            // of whether one has already been sent on this path.
5197            if space_id == SpaceId::Data
5198                && self
5199                    .config
5200                    .address_discovery_role
5201                    .should_report(&self.peer_params.address_discovery_role)
5202            {
5203                let frame = frame::ObservedAddr::new(path.remote, self.next_observed_addr_seq_no);
5204                if buf.remaining_mut() > frame.size() {
5205                    frame.write(buf);
5206                    qlog.frame(&Frame::ObservedAddr(frame));
5207
5208                    self.next_observed_addr_seq_no =
5209                        self.next_observed_addr_seq_no.saturating_add(1u8);
5210                    path.observed_addr_sent = true;
5211
5212                    self.stats.frame_tx.observed_addr += 1;
5213                    sent.retransmits.get_or_create().observed_addr = true;
5214                    space.pending.observed_addr = false;
5215                }
5216            }
5217        }
5218
5219        // PATH_RESPONSE
5220        if buf.remaining_mut() > frame::PathResponse::SIZE_BOUND && space_id == SpaceId::Data {
5221            if let Some(token) = path.path_responses.pop_on_path(path.remote) {
5222                sent.non_retransmits = true;
5223                sent.requires_padding = true;
5224                let response = frame::PathResponse(token);
5225                trace!(%response, "sending response");
5226                buf.write(response);
5227                qlog.frame(&Frame::PathResponse(response));
5228                self.stats.frame_tx.path_response += 1;
5229
5230                // NOTE: this is technically not required but might be useful to ride the
5231                // request/response nature of path challenges to refresh an observation
5232                // Since PATH_RESPONSE is a probing frame, this is allowed by the spec.
5233                if space_id == SpaceId::Data
5234                    && self
5235                        .config
5236                        .address_discovery_role
5237                        .should_report(&self.peer_params.address_discovery_role)
5238                {
5239                    let frame =
5240                        frame::ObservedAddr::new(path.remote, self.next_observed_addr_seq_no);
5241                    if buf.remaining_mut() > frame.size() {
5242                        frame.write(buf);
5243                        qlog.frame(&Frame::ObservedAddr(frame));
5244
5245                        self.next_observed_addr_seq_no =
5246                            self.next_observed_addr_seq_no.saturating_add(1u8);
5247                        path.observed_addr_sent = true;
5248
5249                        self.stats.frame_tx.observed_addr += 1;
5250                        sent.retransmits.get_or_create().observed_addr = true;
5251                        space.pending.observed_addr = false;
5252                    }
5253                }
5254            }
5255        }
5256
5257        // CRYPTO
5258        while !path_exclusive_only && buf.remaining_mut() > frame::Crypto::SIZE_BOUND && !is_0rtt {
5259            let mut frame = match space.pending.crypto.pop_front() {
5260                Some(x) => x,
5261                None => break,
5262            };
5263
5264            // Calculate the maximum amount of crypto data we can store in the buffer.
5265            // Since the offset is known, we can reserve the exact size required to encode it.
5266            // For length we reserve 2bytes which allows to encode up to 2^14,
5267            // which is more than what fits into normally sized QUIC frames.
5268            let max_crypto_data_size = buf.remaining_mut()
5269                - 1 // Frame Type
5270                - VarInt::size(unsafe { VarInt::from_u64_unchecked(frame.offset) })
5271                - 2; // Maximum encoded length for frame size, given we send less than 2^14 bytes
5272
5273            let len = frame
5274                .data
5275                .len()
5276                .min(2usize.pow(14) - 1)
5277                .min(max_crypto_data_size);
5278
5279            let data = frame.data.split_to(len);
5280            let truncated = frame::Crypto {
5281                offset: frame.offset,
5282                data,
5283            };
5284            trace!(
5285                "CRYPTO: off {} len {}",
5286                truncated.offset,
5287                truncated.data.len()
5288            );
5289            truncated.encode(buf);
5290            self.stats.frame_tx.crypto += 1;
5291
5292            // The clone is cheap but we still cfg it out if qlog is disabled.
5293            #[cfg(feature = "qlog")]
5294            qlog.frame(&Frame::Crypto(truncated.clone()));
5295            sent.retransmits.get_or_create().crypto.push_back(truncated);
5296            if !frame.data.is_empty() {
5297                frame.offset += len as u64;
5298                space.pending.crypto.push_front(frame);
5299            }
5300        }
5301
5302        // TODO(flub): maybe this is much higher priority?
5303        // PATH_ABANDON
5304        while !path_exclusive_only
5305            && space_id == SpaceId::Data
5306            && frame::PathAbandon::SIZE_BOUND <= buf.remaining_mut()
5307        {
5308            let Some((path_id, error_code)) = space.pending.path_abandon.pop_first() else {
5309                break;
5310            };
5311            let frame = frame::PathAbandon {
5312                path_id,
5313                error_code,
5314            };
5315            frame.encode(buf);
5316            qlog.frame(&Frame::PathAbandon(frame));
5317            self.stats.frame_tx.path_abandon += 1;
5318            trace!(%path_id, "PATH_ABANDON");
5319            sent.retransmits
5320                .get_or_create()
5321                .path_abandon
5322                .entry(path_id)
5323                .or_insert(error_code);
5324        }
5325
5326        // PATH_STATUS_AVAILABLE & PATH_STATUS_BACKUP
5327        while !path_exclusive_only
5328            && space_id == SpaceId::Data
5329            && frame::PathStatusAvailable::SIZE_BOUND <= buf.remaining_mut()
5330        {
5331            let Some(path_id) = space.pending.path_status.pop_first() else {
5332                break;
5333            };
5334            let Some(path) = self.paths.get(&path_id).map(|path_state| &path_state.data) else {
5335                trace!(%path_id, "discarding queued path status for unknown path");
5336                continue;
5337            };
5338
5339            let seq = path.status.seq();
5340            sent.retransmits.get_or_create().path_status.insert(path_id);
5341            match path.local_status() {
5342                PathStatus::Available => {
5343                    let frame = frame::PathStatusAvailable {
5344                        path_id,
5345                        status_seq_no: seq,
5346                    };
5347                    frame.encode(buf);
5348                    qlog.frame(&Frame::PathStatusAvailable(frame));
5349                    self.stats.frame_tx.path_status_available += 1;
5350                    trace!(%path_id, %seq, "PATH_STATUS_AVAILABLE")
5351                }
5352                PathStatus::Backup => {
5353                    let frame = frame::PathStatusBackup {
5354                        path_id,
5355                        status_seq_no: seq,
5356                    };
5357                    frame.encode(buf);
5358                    qlog.frame(&Frame::PathStatusBackup(frame));
5359                    self.stats.frame_tx.path_status_backup += 1;
5360                    trace!(%path_id, %seq, "PATH_STATUS_BACKUP")
5361                }
5362            }
5363        }
5364
5365        // MAX_PATH_ID
5366        if space_id == SpaceId::Data
5367            && space.pending.max_path_id
5368            && frame::MaxPathId::SIZE_BOUND <= buf.remaining_mut()
5369        {
5370            let frame = frame::MaxPathId(self.local_max_path_id);
5371            frame.encode(buf);
5372            qlog.frame(&Frame::MaxPathId(frame));
5373            space.pending.max_path_id = false;
5374            sent.retransmits.get_or_create().max_path_id = true;
5375            trace!(val = %self.local_max_path_id, "MAX_PATH_ID");
5376            self.stats.frame_tx.max_path_id += 1;
5377        }
5378
5379        // PATHS_BLOCKED
5380        if space_id == SpaceId::Data
5381            && space.pending.paths_blocked
5382            && frame::PathsBlocked::SIZE_BOUND <= buf.remaining_mut()
5383        {
5384            let frame = frame::PathsBlocked(self.remote_max_path_id);
5385            frame.encode(buf);
5386            qlog.frame(&Frame::PathsBlocked(frame));
5387            space.pending.paths_blocked = false;
5388            sent.retransmits.get_or_create().paths_blocked = true;
5389            trace!(max_path_id = ?self.remote_max_path_id, "PATHS_BLOCKED");
5390            self.stats.frame_tx.paths_blocked += 1;
5391        }
5392
5393        // PATH_CIDS_BLOCKED
5394        while space_id == SpaceId::Data && frame::PathCidsBlocked::SIZE_BOUND <= buf.remaining_mut()
5395        {
5396            let Some(path_id) = space.pending.path_cids_blocked.pop() else {
5397                break;
5398            };
5399            let next_seq = match self.rem_cids.get(&path_id) {
5400                Some(cid_queue) => cid_queue.active_seq() + 1,
5401                None => 0,
5402            };
5403            let frame = frame::PathCidsBlocked {
5404                path_id,
5405                next_seq: VarInt(next_seq),
5406            };
5407            frame.encode(buf);
5408            qlog.frame(&Frame::PathCidsBlocked(frame));
5409            sent.retransmits
5410                .get_or_create()
5411                .path_cids_blocked
5412                .push(path_id);
5413            trace!(%path_id, next_seq, "PATH_CIDS_BLOCKED");
5414            self.stats.frame_tx.path_cids_blocked += 1;
5415        }
5416
5417        // RESET_STREAM, STOP_SENDING, MAX_DATA, MAX_STREAM_DATA, MAX_STREAMS
5418        if space_id == SpaceId::Data {
5419            self.streams.write_control_frames(
5420                buf,
5421                &mut space.pending,
5422                &mut sent.retransmits,
5423                &mut self.stats.frame_tx,
5424                qlog,
5425            );
5426        }
5427
5428        // NEW_CONNECTION_ID
5429        let cid_len = self
5430            .local_cid_state
5431            .values()
5432            .map(|cid_state| cid_state.cid_len())
5433            .max()
5434            .expect("some local CID state must exist");
5435        let new_cid_size_bound =
5436            frame::NewConnectionId::size_bound(is_multipath_negotiated, cid_len);
5437        while !path_exclusive_only && buf.remaining_mut() > new_cid_size_bound {
5438            let issued = match space.pending.new_cids.pop() {
5439                Some(x) => x,
5440                None => break,
5441            };
5442            let retire_prior_to = self
5443                .local_cid_state
5444                .get(&issued.path_id)
5445                .map(|cid_state| cid_state.retire_prior_to())
5446                .unwrap_or_else(|| panic!("missing local CID state for path={}", issued.path_id));
5447
5448            let cid_path_id = match is_multipath_negotiated {
5449                true => {
5450                    trace!(
5451                        path_id = ?issued.path_id,
5452                        sequence = issued.sequence,
5453                        id = %issued.id,
5454                        "PATH_NEW_CONNECTION_ID",
5455                    );
5456                    self.stats.frame_tx.path_new_connection_id += 1;
5457                    Some(issued.path_id)
5458                }
5459                false => {
5460                    trace!(
5461                        sequence = issued.sequence,
5462                        id = %issued.id,
5463                        "NEW_CONNECTION_ID"
5464                    );
5465                    debug_assert_eq!(issued.path_id, PathId::ZERO);
5466                    self.stats.frame_tx.new_connection_id += 1;
5467                    None
5468                }
5469            };
5470            let frame = frame::NewConnectionId {
5471                path_id: cid_path_id,
5472                sequence: issued.sequence,
5473                retire_prior_to,
5474                id: issued.id,
5475                reset_token: issued.reset_token,
5476            };
5477            frame.encode(buf);
5478            sent.retransmits.get_or_create().new_cids.push(issued);
5479            qlog.frame(&Frame::NewConnectionId(frame));
5480        }
5481
5482        // RETIRE_CONNECTION_ID
5483        let retire_cid_bound = frame::RetireConnectionId::size_bound(is_multipath_negotiated);
5484        while !path_exclusive_only && buf.remaining_mut() > retire_cid_bound {
5485            let (path_id, sequence) = match space.pending.retire_cids.pop() {
5486                Some((PathId::ZERO, seq)) if !is_multipath_negotiated => {
5487                    trace!(sequence = seq, "RETIRE_CONNECTION_ID");
5488                    self.stats.frame_tx.retire_connection_id += 1;
5489                    (None, seq)
5490                }
5491                Some((path_id, seq)) => {
5492                    trace!(%path_id, sequence = seq, "PATH_RETIRE_CONNECTION_ID");
5493                    self.stats.frame_tx.path_retire_connection_id += 1;
5494                    (Some(path_id), seq)
5495                }
5496                None => break,
5497            };
5498            let frame = frame::RetireConnectionId { path_id, sequence };
5499            frame.encode(buf);
5500            qlog.frame(&Frame::RetireConnectionId(frame));
5501            sent.retransmits
5502                .get_or_create()
5503                .retire_cids
5504                .push((path_id.unwrap_or_default(), sequence));
5505        }
5506
5507        // DATAGRAM
5508        let mut sent_datagrams = false;
5509        while !path_exclusive_only
5510            && buf.remaining_mut() > Datagram::SIZE_BOUND
5511            && space_id == SpaceId::Data
5512        {
5513            let prev_remaining = buf.remaining_mut();
5514            match self.datagrams.write(buf) {
5515                true => {
5516                    sent_datagrams = true;
5517                    sent.non_retransmits = true;
5518                    self.stats.frame_tx.datagram += 1;
5519                    qlog.frame_datagram((prev_remaining - buf.remaining_mut()) as u64);
5520                }
5521                false => break,
5522            }
5523        }
5524        if self.datagrams.send_blocked && sent_datagrams {
5525            self.events.push_back(Event::DatagramsUnblocked);
5526            self.datagrams.send_blocked = false;
5527        }
5528
5529        let path = &mut self.paths.get_mut(&path_id).expect("known path").data;
5530
5531        // NEW_TOKEN
5532        while let Some(remote_addr) = space.pending.new_tokens.pop() {
5533            if path_exclusive_only {
5534                break;
5535            }
5536            debug_assert_eq!(space_id, SpaceId::Data);
5537            let ConnectionSide::Server { server_config } = &self.side else {
5538                panic!("NEW_TOKEN frames should not be enqueued by clients");
5539            };
5540
5541            if remote_addr != path.remote {
5542                // NEW_TOKEN frames contain tokens bound to a client's IP address, and are only
5543                // useful if used from the same IP address.  Thus, we abandon enqueued NEW_TOKEN
5544                // frames upon an path change. Instead, when the new path becomes validated,
5545                // NEW_TOKEN frames may be enqueued for the new path instead.
5546                continue;
5547            }
5548
5549            let token = Token::new(
5550                TokenPayload::Validation {
5551                    ip: remote_addr.ip(),
5552                    issued: server_config.time_source.now(),
5553                },
5554                &mut self.rng,
5555            );
5556            let new_token = NewToken {
5557                token: token.encode(&*server_config.token_key).into(),
5558            };
5559
5560            if buf.remaining_mut() < new_token.size() {
5561                space.pending.new_tokens.push(remote_addr);
5562                break;
5563            }
5564
5565            trace!("NEW_TOKEN");
5566            new_token.encode(buf);
5567            qlog.frame(&Frame::NewToken(new_token));
5568            sent.retransmits
5569                .get_or_create()
5570                .new_tokens
5571                .push(remote_addr);
5572            self.stats.frame_tx.new_token += 1;
5573        }
5574
5575        // STREAM
5576        if !path_exclusive_only && space_id == SpaceId::Data {
5577            sent.stream_frames =
5578                self.streams
5579                    .write_stream_frames(buf, self.config.send_fairness, qlog);
5580            self.stats.frame_tx.stream += sent.stream_frames.len() as u64;
5581        }
5582
5583        // ADD_ADDRESS
5584        // TODO(@divma): check if we need to do path exclusive filters
5585        while space_id == SpaceId::Data && frame::AddAddress::SIZE_BOUND <= buf.remaining_mut() {
5586            if let Some(added_address) = space.pending.add_address.pop_last() {
5587                trace!(
5588                    seq = %added_address.seq_no,
5589                    ip = ?added_address.ip,
5590                    port = added_address.port,
5591                    "ADD_ADDRESS",
5592                );
5593                added_address.write(buf);
5594                sent.retransmits
5595                    .get_or_create()
5596                    .add_address
5597                    .insert(added_address);
5598                self.stats.frame_tx.add_address = self.stats.frame_tx.add_address.saturating_add(1);
5599                qlog.frame(&Frame::AddAddress(added_address));
5600            } else {
5601                break;
5602            }
5603        }
5604
5605        // REMOVE_ADDRESS
5606        while space_id == SpaceId::Data && frame::RemoveAddress::SIZE_BOUND <= buf.remaining_mut() {
5607            if let Some(removed_address) = space.pending.remove_address.pop_last() {
5608                trace!(seq = %removed_address.seq_no, "REMOVE_ADDRESS");
5609                removed_address.write(buf);
5610                sent.retransmits
5611                    .get_or_create()
5612                    .remove_address
5613                    .insert(removed_address);
5614                self.stats.frame_tx.remove_address =
5615                    self.stats.frame_tx.remove_address.saturating_add(1);
5616                qlog.frame(&Frame::RemoveAddress(removed_address));
5617            } else {
5618                break;
5619            }
5620        }
5621
5622        sent
5623    }
5624
5625    /// Write pending ACKs into a buffer
5626    fn populate_acks(
5627        now: Instant,
5628        receiving_ecn: bool,
5629        sent: &mut SentFrames,
5630        path_id: PathId,
5631        space_id: SpaceId,
5632        space: &mut PacketSpace,
5633        is_multipath_negotiated: bool,
5634        buf: &mut impl BufMut,
5635        stats: &mut ConnectionStats,
5636        #[allow(unused)] qlog: &mut QlogSentPacket,
5637    ) {
5638        // 0-RTT packets must never carry acks (which would have to be of handshake packets)
5639        debug_assert!(space.crypto.is_some(), "tried to send ACK in 0-RTT");
5640
5641        debug_assert!(
5642            is_multipath_negotiated || path_id == PathId::ZERO,
5643            "Only PathId::ZERO allowed without multipath (have {path_id:?})"
5644        );
5645        if is_multipath_negotiated {
5646            debug_assert!(
5647                space_id == SpaceId::Data || path_id == PathId::ZERO,
5648                "path acks must be sent in 1RTT space (have {space_id:?})"
5649            );
5650        }
5651
5652        let pns = space.for_path(path_id);
5653        let ranges = pns.pending_acks.ranges();
5654        debug_assert!(!ranges.is_empty(), "can not send empty ACK range");
5655        let ecn = if receiving_ecn {
5656            Some(&pns.ecn_counters)
5657        } else {
5658            None
5659        };
5660        if let Some(max) = ranges.max() {
5661            sent.largest_acked.insert(path_id, max);
5662        }
5663
5664        let delay_micros = pns.pending_acks.ack_delay(now).as_micros() as u64;
5665        // TODO: This should come from `TransportConfig` if that gets configurable.
5666        let ack_delay_exp = TransportParameters::default().ack_delay_exponent;
5667        let delay = delay_micros >> ack_delay_exp.into_inner();
5668
5669        if is_multipath_negotiated && space_id == SpaceId::Data {
5670            if !ranges.is_empty() {
5671                trace!("PATH_ACK {path_id:?} {ranges:?}, Delay = {delay_micros}us");
5672                frame::PathAck::encode(path_id, delay as _, ranges, ecn, buf);
5673                qlog.frame_path_ack(path_id, delay as _, ranges, ecn);
5674                stats.frame_tx.path_acks += 1;
5675            }
5676        } else {
5677            trace!("ACK {ranges:?}, Delay = {delay_micros}us");
5678            frame::Ack::encode(delay as _, ranges, ecn, buf);
5679            stats.frame_tx.acks += 1;
5680            qlog.frame_ack(delay, ranges, ecn);
5681        }
5682    }
5683
5684    fn close_common(&mut self) {
5685        trace!("connection closed");
5686        self.timers.reset();
5687    }
5688
5689    fn calculate_end_timer(&self, now: Instant, factor: u32, space: SpaceId) -> Instant {
5690        match space {
5691            SpaceId::Initial | SpaceId::Handshake => {
5692                let duration = factor * self.pto(space, PathId::ZERO);
5693                self.path_data(PathId::ZERO).timer_offset(now, duration)
5694            }
5695            SpaceId::Data => self
5696                .paths
5697                .iter()
5698                .filter_map(|(path_id, state)| {
5699                    if state.data.total_sent == 0 && state.data.total_recvd == 0 {
5700                        // If we are closing and haven't sent anything yet, do not include
5701                        None
5702                    } else {
5703                        let duration = factor * self.pto(self.highest_space, *path_id);
5704                        Some(self.path_data(*path_id).timer_offset(now, duration))
5705                    }
5706                })
5707                .max()
5708                .unwrap_or(now), // TODO:: reevaluate
5709        }
5710    }
5711
5712    fn set_close_timer(&mut self, now: Instant) {
5713        // QUIC-MULTIPATH § 2.6 Connection Closure: draining for 3*PTO with PTO the max of
5714        // the PTO for all paths.
5715
5716        let end = self.calculate_end_timer(now, 3, self.highest_space);
5717        self.timers
5718            .set(Timer::Conn(ConnTimer::Close), end, self.qlog.with_time(now));
5719    }
5720
5721    /// Handle transport parameters received from the peer
5722    ///
5723    /// *rem_cid* and *loc_cid* are the source and destination CIDs respectively of the
5724    /// *packet into which the transport parameters arrived.
5725    fn handle_peer_params(
5726        &mut self,
5727        params: TransportParameters,
5728        loc_cid: ConnectionId,
5729        rem_cid: ConnectionId,
5730        now: Instant,
5731    ) -> Result<(), TransportError> {
5732        if Some(self.orig_rem_cid) != params.initial_src_cid
5733            || (self.side.is_client()
5734                && (Some(self.initial_dst_cid) != params.original_dst_cid
5735                    || self.retry_src_cid != params.retry_src_cid))
5736        {
5737            return Err(TransportError::TRANSPORT_PARAMETER_ERROR(
5738                "CID authentication failure",
5739            ));
5740        }
5741        if params.initial_max_path_id.is_some() && (loc_cid.is_empty() || rem_cid.is_empty()) {
5742            return Err(TransportError::PROTOCOL_VIOLATION(
5743                "multipath must not use zero-length CIDs",
5744            ));
5745        }
5746
5747        self.set_peer_params(params);
5748        self.qlog.emit_peer_transport_params_received(self, now);
5749
5750        Ok(())
5751    }
5752
5753    fn set_peer_params(&mut self, params: TransportParameters) {
5754        self.streams.set_params(&params);
5755        self.idle_timeout =
5756            negotiate_max_idle_timeout(self.config.max_idle_timeout, Some(params.max_idle_timeout));
5757        trace!("negotiated max idle timeout {:?}", self.idle_timeout);
5758
5759        if let Some(ref info) = params.preferred_address {
5760            // During the handshake PathId::ZERO exists.
5761            self.rem_cids.get_mut(&PathId::ZERO).expect("not yet abandoned").insert(frame::NewConnectionId {
5762                path_id: None,
5763                sequence: 1,
5764                id: info.connection_id,
5765                reset_token: info.stateless_reset_token,
5766                retire_prior_to: 0,
5767            })
5768            .expect(
5769                "preferred address CID is the first received, and hence is guaranteed to be legal",
5770            );
5771            let remote = self.path_data(PathId::ZERO).remote;
5772            self.set_reset_token(PathId::ZERO, remote, info.stateless_reset_token);
5773        }
5774        self.ack_frequency.peer_max_ack_delay = get_max_ack_delay(&params);
5775
5776        let mut multipath_enabled = None;
5777        if let (Some(local_max_path_id), Some(remote_max_path_id)) = (
5778            self.config.get_initial_max_path_id(),
5779            params.initial_max_path_id,
5780        ) {
5781            // multipath is enabled, register the local and remote maximums
5782            self.local_max_path_id = local_max_path_id;
5783            self.remote_max_path_id = remote_max_path_id;
5784            let initial_max_path_id = local_max_path_id.min(remote_max_path_id);
5785            debug!(%initial_max_path_id, "multipath negotiated");
5786            multipath_enabled = Some(initial_max_path_id);
5787        }
5788
5789        if let Some((max_locally_allowed_remote_addresses, max_remotely_allowed_remote_addresses)) =
5790            self.config
5791                .max_remote_nat_traversal_addresses
5792                .zip(params.max_remote_nat_traversal_addresses)
5793        {
5794            if let Some(max_initial_paths) =
5795                multipath_enabled.map(|path_id| path_id.saturating_add(1u8))
5796            {
5797                let max_local_addresses = max_remotely_allowed_remote_addresses.get();
5798                let max_remote_addresses = max_locally_allowed_remote_addresses.get();
5799                self.iroh_hp =
5800                    iroh_hp::State::new(max_remote_addresses, max_local_addresses, self.side());
5801                debug!(
5802                    %max_remote_addresses, %max_local_addresses,
5803                    "iroh hole punching negotiated"
5804                );
5805
5806                match self.side() {
5807                    Side::Client => {
5808                        if max_initial_paths.as_u32() < max_remote_addresses as u32 + 1 {
5809                            // in this case the client might try to open `max_remote_addresses` new
5810                            // paths, but the current multipath configuration will not allow it
5811                            warn!(%max_initial_paths, %max_remote_addresses, "local client configuration might cause nat traversal issues")
5812                        } else if max_local_addresses as u64
5813                            > params.active_connection_id_limit.into_inner()
5814                        {
5815                            // the server allows us to send at most `params.active_connection_id_limit`
5816                            // but they might need at least `max_local_addresses` to effectively send
5817                            // `PATH_CHALLENGE` frames to each advertised local address
5818                            warn!(%max_local_addresses, remote_cid_limit=%params.active_connection_id_limit.into_inner(), "remote server configuration might cause nat traversal issues")
5819                        }
5820                    }
5821                    Side::Server => {
5822                        if (max_initial_paths.as_u32() as u64) < crate::LOC_CID_COUNT {
5823                            warn!(%max_initial_paths, local_cid_limit=%crate::LOC_CID_COUNT, "local server configuration might cause nat traversal issues")
5824                        }
5825                    }
5826                }
5827            } else {
5828                debug!("iroh nat traversal enabled for both endpoints, but multipath is missing")
5829            }
5830        }
5831
5832        self.peer_params = params;
5833        let peer_max_udp_payload_size =
5834            u16::try_from(self.peer_params.max_udp_payload_size.into_inner()).unwrap_or(u16::MAX);
5835        self.path_data_mut(PathId::ZERO)
5836            .mtud
5837            .on_peer_max_udp_payload_size_received(peer_max_udp_payload_size);
5838    }
5839
5840    /// Decrypts a packet, returning the packet number on success
5841    fn decrypt_packet(
5842        &mut self,
5843        now: Instant,
5844        path_id: PathId,
5845        packet: &mut Packet,
5846    ) -> Result<Option<u64>, Option<TransportError>> {
5847        let result = packet_crypto::decrypt_packet_body(
5848            packet,
5849            path_id,
5850            &self.spaces,
5851            self.zero_rtt_crypto.as_ref(),
5852            self.key_phase,
5853            self.prev_crypto.as_ref(),
5854            self.next_crypto.as_ref(),
5855        )?;
5856
5857        let result = match result {
5858            Some(r) => r,
5859            None => return Ok(None),
5860        };
5861
5862        if result.outgoing_key_update_acked {
5863            if let Some(prev) = self.prev_crypto.as_mut() {
5864                prev.end_packet = Some((result.number, now));
5865                self.set_key_discard_timer(now, packet.header.space());
5866            }
5867        }
5868
5869        if result.incoming_key_update {
5870            trace!("key update authenticated");
5871            self.update_keys(Some((result.number, now)), true);
5872            self.set_key_discard_timer(now, packet.header.space());
5873        }
5874
5875        Ok(Some(result.number))
5876    }
5877
5878    fn update_keys(&mut self, end_packet: Option<(u64, Instant)>, remote: bool) {
5879        trace!("executing key update");
5880        // Generate keys for the key phase after the one we're switching to, store them in
5881        // `next_crypto`, make the contents of `next_crypto` current, and move the current keys into
5882        // `prev_crypto`.
5883        let new = self
5884            .crypto
5885            .next_1rtt_keys()
5886            .expect("only called for `Data` packets");
5887        self.key_phase_size = new
5888            .local
5889            .confidentiality_limit()
5890            .saturating_sub(KEY_UPDATE_MARGIN);
5891        let old = mem::replace(
5892            &mut self.spaces[SpaceId::Data]
5893                .crypto
5894                .as_mut()
5895                .unwrap() // safe because update_keys() can only be triggered by short packets
5896                .packet,
5897            mem::replace(self.next_crypto.as_mut().unwrap(), new),
5898        );
5899        self.spaces[SpaceId::Data]
5900            .iter_paths_mut()
5901            .for_each(|s| s.sent_with_keys = 0);
5902        self.prev_crypto = Some(PrevCrypto {
5903            crypto: old,
5904            end_packet,
5905            update_unacked: remote,
5906        });
5907        self.key_phase = !self.key_phase;
5908    }
5909
5910    fn peer_supports_ack_frequency(&self) -> bool {
5911        self.peer_params.min_ack_delay.is_some()
5912    }
5913
5914    /// Send an IMMEDIATE_ACK frame to the remote endpoint
5915    ///
5916    /// According to the spec, this will result in an error if the remote endpoint does not support
5917    /// the Acknowledgement Frequency extension
5918    pub(crate) fn immediate_ack(&mut self, path_id: PathId) {
5919        debug_assert_eq!(
5920            self.highest_space,
5921            SpaceId::Data,
5922            "immediate ack must be written in the data space"
5923        );
5924        self.spaces[self.highest_space]
5925            .for_path(path_id)
5926            .immediate_ack_pending = true;
5927    }
5928
5929    /// Decodes a packet, returning its decrypted payload, so it can be inspected in tests
5930    #[cfg(test)]
5931    pub(crate) fn decode_packet(&self, event: &ConnectionEvent) -> Option<Vec<u8>> {
5932        let (path_id, first_decode, remaining) = match &event.0 {
5933            ConnectionEventInner::Datagram(DatagramConnectionEvent {
5934                path_id,
5935                first_decode,
5936                remaining,
5937                ..
5938            }) => (path_id, first_decode, remaining),
5939            _ => return None,
5940        };
5941
5942        if remaining.is_some() {
5943            panic!("Packets should never be coalesced in tests");
5944        }
5945
5946        let decrypted_header = packet_crypto::unprotect_header(
5947            first_decode.clone(),
5948            &self.spaces,
5949            self.zero_rtt_crypto.as_ref(),
5950            self.peer_params.stateless_reset_token,
5951        )?;
5952
5953        let mut packet = decrypted_header.packet?;
5954        packet_crypto::decrypt_packet_body(
5955            &mut packet,
5956            *path_id,
5957            &self.spaces,
5958            self.zero_rtt_crypto.as_ref(),
5959            self.key_phase,
5960            self.prev_crypto.as_ref(),
5961            self.next_crypto.as_ref(),
5962        )
5963        .ok()?;
5964
5965        Some(packet.payload.to_vec())
5966    }
5967
5968    /// The number of bytes of packets containing retransmittable frames that have not been
5969    /// acknowledged or declared lost.
5970    #[cfg(test)]
5971    pub(crate) fn bytes_in_flight(&self) -> u64 {
5972        // TODO(@divma): consider including for multipath?
5973        self.path_data(PathId::ZERO).in_flight.bytes
5974    }
5975
5976    /// Number of bytes worth of non-ack-only packets that may be sent
5977    #[cfg(test)]
5978    pub(crate) fn congestion_window(&self) -> u64 {
5979        let path = self.path_data(PathId::ZERO);
5980        path.congestion
5981            .window()
5982            .saturating_sub(path.in_flight.bytes)
5983    }
5984
5985    /// Whether no timers but keepalive, idle, rtt, pushnewcid, and key discard are running
5986    #[cfg(test)]
5987    pub(crate) fn is_idle(&self) -> bool {
5988        let current_timers = self.timers.values();
5989        current_timers
5990            .into_iter()
5991            .filter(|(timer, _)| {
5992                !matches!(
5993                    timer,
5994                    Timer::Conn(ConnTimer::KeepAlive)
5995                        | Timer::PerPath(_, PathTimer::PathKeepAlive)
5996                        | Timer::Conn(ConnTimer::PushNewCid)
5997                        | Timer::Conn(ConnTimer::KeyDiscard)
5998                )
5999            })
6000            .min_by_key(|(_, time)| *time)
6001            .is_none_or(|(timer, _)| timer == Timer::Conn(ConnTimer::Idle))
6002    }
6003
6004    /// Whether explicit congestion notification is in use on outgoing packets.
6005    #[cfg(test)]
6006    pub(crate) fn using_ecn(&self) -> bool {
6007        self.path_data(PathId::ZERO).sending_ecn
6008    }
6009
6010    /// The number of received bytes in the current path
6011    #[cfg(test)]
6012    pub(crate) fn total_recvd(&self) -> u64 {
6013        self.path_data(PathId::ZERO).total_recvd
6014    }
6015
6016    #[cfg(test)]
6017    pub(crate) fn active_local_cid_seq(&self) -> (u64, u64) {
6018        self.local_cid_state
6019            .get(&PathId::ZERO)
6020            .unwrap()
6021            .active_seq()
6022    }
6023
6024    #[cfg(test)]
6025    #[track_caller]
6026    pub(crate) fn active_local_path_cid_seq(&self, path_id: u32) -> (u64, u64) {
6027        self.local_cid_state
6028            .get(&PathId(path_id))
6029            .unwrap()
6030            .active_seq()
6031    }
6032
6033    /// Instruct the peer to replace previously issued CIDs by sending a NEW_CONNECTION_ID frame
6034    /// with updated `retire_prior_to` field set to `v`
6035    #[cfg(test)]
6036    pub(crate) fn rotate_local_cid(&mut self, v: u64, now: Instant) {
6037        let n = self
6038            .local_cid_state
6039            .get_mut(&PathId::ZERO)
6040            .unwrap()
6041            .assign_retire_seq(v);
6042        self.endpoint_events
6043            .push_back(EndpointEventInner::NeedIdentifiers(PathId::ZERO, now, n));
6044    }
6045
6046    /// Check the current active remote CID sequence for `PathId::ZERO`
6047    #[cfg(test)]
6048    pub(crate) fn active_rem_cid_seq(&self) -> u64 {
6049        self.rem_cids.get(&PathId::ZERO).unwrap().active_seq()
6050    }
6051
6052    /// Returns the detected maximum udp payload size for the current path
6053    #[cfg(test)]
6054    pub(crate) fn path_mtu(&self, path_id: PathId) -> u16 {
6055        self.path_data(path_id).current_mtu()
6056    }
6057
6058    /// Triggers path validation on all paths
6059    #[cfg(test)]
6060    pub(crate) fn trigger_path_validation(&mut self) {
6061        for path in self.paths.values_mut() {
6062            path.data.send_new_challenge = true;
6063        }
6064    }
6065
6066    /// Whether we have 1-RTT data to send
6067    ///
6068    /// This checks for frames that can only be sent in the data space (1-RTT):
6069    /// - Pending PATH_CHALLENGE frames on the active and previous path if just migrated.
6070    /// - Pending PATH_RESPONSE frames.
6071    /// - Pending data to send in STREAM frames.
6072    /// - Pending DATAGRAM frames to send.
6073    ///
6074    /// See also [`PacketSpace::can_send`] which keeps track of all other frame types that
6075    /// may need to be sent.
6076    fn can_send_1rtt(&self, path_id: PathId, max_size: usize) -> SendableFrames {
6077        let path_exclusive = self.paths.get(&path_id).is_some_and(|path| {
6078            path.data.send_new_challenge
6079                || path
6080                    .prev
6081                    .as_ref()
6082                    .is_some_and(|(_, path)| path.send_new_challenge)
6083                || !path.data.path_responses.is_empty()
6084        });
6085        let other = self.streams.can_send_stream_data()
6086            || self
6087                .datagrams
6088                .outgoing
6089                .front()
6090                .is_some_and(|x| x.size(true) <= max_size);
6091        SendableFrames {
6092            acks: false,
6093            other,
6094            close: false,
6095            path_exclusive,
6096        }
6097    }
6098
6099    /// Terminate the connection instantly, without sending a close packet
6100    fn kill(&mut self, reason: ConnectionError) {
6101        self.close_common();
6102        self.state.move_to_drained(Some(reason));
6103        self.endpoint_events.push_back(EndpointEventInner::Drained);
6104    }
6105
6106    /// Storage size required for the largest packet that can be transmitted on all currently
6107    /// available paths
6108    ///
6109    /// Buffers passed to [`Connection::poll_transmit`] should be at least this large.
6110    ///
6111    /// When multipath is enabled, this value is the minimum MTU across all available paths.
6112    pub fn current_mtu(&self) -> u16 {
6113        self.paths
6114            .iter()
6115            .filter(|&(path_id, _path_state)| !self.abandoned_paths.contains(path_id))
6116            .map(|(_path_id, path_state)| path_state.data.current_mtu())
6117            .min()
6118            .expect("There is always at least one available path")
6119    }
6120
6121    /// Size of non-frame data for a 1-RTT packet
6122    ///
6123    /// Quantifies space consumed by the QUIC header and AEAD tag. All other bytes in a packet are
6124    /// frames. Changes if the length of the remote connection ID changes, which is expected to be
6125    /// rare. If `pn` is specified, may additionally change unpredictably due to variations in
6126    /// latency and packet loss.
6127    fn predict_1rtt_overhead(&mut self, pn: u64, path: PathId) -> usize {
6128        let pn_len = PacketNumber::new(
6129            pn,
6130            self.spaces[SpaceId::Data]
6131                .for_path(path)
6132                .largest_acked_packet
6133                .unwrap_or(0),
6134        )
6135        .len();
6136
6137        // 1 byte for flags
6138        1 + self
6139            .rem_cids
6140            .get(&path)
6141            .map(|cids| cids.active().len())
6142            .unwrap_or(20)      // Max CID len in QUIC v1
6143            + pn_len
6144            + self.tag_len_1rtt()
6145    }
6146
6147    fn predict_1rtt_overhead_no_pn(&self) -> usize {
6148        let pn_len = 4;
6149
6150        let cid_len = self
6151            .rem_cids
6152            .values()
6153            .map(|cids| cids.active().len())
6154            .max()
6155            .unwrap_or(20); // Max CID len in QUIC v1
6156
6157        // 1 byte for flags
6158        1 + cid_len + pn_len + self.tag_len_1rtt()
6159    }
6160
6161    fn tag_len_1rtt(&self) -> usize {
6162        let key = match self.spaces[SpaceId::Data].crypto.as_ref() {
6163            Some(crypto) => Some(&*crypto.packet.local),
6164            None => self.zero_rtt_crypto.as_ref().map(|x| &*x.packet),
6165        };
6166        // If neither Data nor 0-RTT keys are available, make a reasonable tag length guess. As of
6167        // this writing, all QUIC cipher suites use 16-byte tags. We could return `None` instead,
6168        // but that would needlessly prevent sending datagrams during 0-RTT.
6169        key.map_or(16, |x| x.tag_len())
6170    }
6171
6172    /// Mark the path as validated, and enqueue NEW_TOKEN frames to be sent as appropriate
6173    fn on_path_validated(&mut self, path_id: PathId) {
6174        self.path_data_mut(path_id).validated = true;
6175        let ConnectionSide::Server { server_config } = &self.side else {
6176            return;
6177        };
6178        let remote_addr = self.path_data(path_id).remote;
6179        let new_tokens = &mut self.spaces[SpaceId::Data as usize].pending.new_tokens;
6180        new_tokens.clear();
6181        for _ in 0..server_config.validation_token.sent {
6182            new_tokens.push(remote_addr);
6183        }
6184    }
6185
6186    /// Handle new path status information: PATH_STATUS_AVAILABLE, PATH_STATUS_BACKUP
6187    fn on_path_status(&mut self, path_id: PathId, status: PathStatus, status_seq_no: VarInt) {
6188        if let Some(path) = self.paths.get_mut(&path_id) {
6189            path.data.status.remote_update(status, status_seq_no);
6190        } else {
6191            debug!("PATH_STATUS_AVAILABLE received unknown path {:?}", path_id);
6192        }
6193        self.events.push_back(
6194            PathEvent::RemoteStatus {
6195                id: path_id,
6196                status,
6197            }
6198            .into(),
6199        );
6200    }
6201
6202    /// Returns the maximum [`PathId`] to be used for sending in this connection.
6203    ///
6204    /// This is calculated as minimum between the local and remote's maximums when multipath is
6205    /// enabled, or `None` when disabled.
6206    ///
6207    /// For data that's received, we should use [`Self::local_max_path_id`] instead.
6208    /// The reasoning is that the remote might already have updated to its own newer
6209    /// [`Self::max_path_id`] after sending out a `MAX_PATH_ID` frame, but it got re-ordered.
6210    fn max_path_id(&self) -> Option<PathId> {
6211        if self.is_multipath_negotiated() {
6212            Some(self.remote_max_path_id.min(self.local_max_path_id))
6213        } else {
6214            None
6215        }
6216    }
6217
6218    /// Add addresses the local endpoint considers are reachable for nat traversal
6219    pub fn add_nat_traversal_address(&mut self, address: SocketAddr) -> Result<(), iroh_hp::Error> {
6220        if let Some(added) = self.iroh_hp.add_local_address(address)? {
6221            self.spaces[SpaceId::Data].pending.add_address.insert(added);
6222        };
6223        Ok(())
6224    }
6225
6226    /// Removes an address the endpoing no longer considers reachable for nat traversal
6227    ///
6228    /// Addresses not present in the set will be silently ignored.
6229    pub fn remove_nat_traversal_address(
6230        &mut self,
6231        address: SocketAddr,
6232    ) -> Result<(), iroh_hp::Error> {
6233        if let Some(removed) = self.iroh_hp.remove_local_address(address)? {
6234            self.spaces[SpaceId::Data]
6235                .pending
6236                .remove_address
6237                .insert(removed);
6238        }
6239        Ok(())
6240    }
6241
6242    /// Get the current local nat traversal addresses
6243    pub fn get_local_nat_traversal_addresses(&self) -> Result<Vec<SocketAddr>, iroh_hp::Error> {
6244        self.iroh_hp.get_local_nat_traversal_addresses()
6245    }
6246
6247    /// Get the currently advertised nat traversal addresses by the server
6248    pub fn get_remote_nat_traversal_addresses(&self) -> Result<Vec<SocketAddr>, iroh_hp::Error> {
6249        Ok(self
6250            .iroh_hp
6251            .client_side()?
6252            .get_remote_nat_traversal_addresses())
6253    }
6254
6255    /// Initiates a new nat traversal round
6256    ///
6257    /// A nat traversal round involves advertising the client's local addresses in `REACH_OUT`
6258    /// frames, and initiating probing of the known remote addresses. When a new round is
6259    /// initiated, the previous one is cancelled, and paths that have not been opened are closed.
6260    ///
6261    /// Returns the server addresses that are now being probed.
6262    pub fn initiate_nat_traversal_round(
6263        &mut self,
6264        now: Instant,
6265    ) -> Result<Vec<SocketAddr>, iroh_hp::Error> {
6266        if self.state.is_closed() {
6267            return Err(iroh_hp::Error::Closed);
6268        }
6269
6270        let client_state = self.iroh_hp.client_side_mut()?;
6271        let iroh_hp::NatTraversalRound {
6272            new_round,
6273            reach_out_at,
6274            addresses_to_probe,
6275            prev_round_path_ids,
6276        } = client_state.initiate_nat_traversal_round()?;
6277
6278        self.spaces[SpaceId::Data].pending.reach_out = Some((new_round, reach_out_at));
6279
6280        for path_id in prev_round_path_ids {
6281            // TODO(@divma): this sounds reasonable but we need if this actually works for the
6282            // purposes of the protocol
6283            let validated = self
6284                .path(path_id)
6285                .map(|path| path.validated)
6286                .unwrap_or(false);
6287
6288            if !validated {
6289                let _ = self.close_path(
6290                    now,
6291                    path_id,
6292                    TransportErrorCode::APPLICATION_ABANDON_PATH.into(),
6293                );
6294            }
6295        }
6296
6297        let mut err = None;
6298
6299        let mut path_ids = Vec::with_capacity(addresses_to_probe.len());
6300        let mut probed_addresses = Vec::with_capacity(addresses_to_probe.len());
6301        let ipv6 = self.paths.values().any(|p| p.data.remote.is_ipv6());
6302
6303        for (ip, port) in addresses_to_probe {
6304            // If this endpoint is an IPv6 endpoint we use IPv6 addresses for all remotes.
6305            let remote = match ip {
6306                IpAddr::V4(addr) if ipv6 => SocketAddr::new(addr.to_ipv6_mapped().into(), port),
6307                IpAddr::V4(addr) => SocketAddr::new(addr.into(), port),
6308                IpAddr::V6(_) if ipv6 => SocketAddr::new(ip, port),
6309                IpAddr::V6(_) => {
6310                    trace!("not using IPv6 nat candidate for IPv4 socket");
6311                    continue;
6312                }
6313            };
6314            match self.open_path_ensure(remote, PathStatus::Backup, now) {
6315                Ok((path_id, path_was_known)) if !path_was_known => {
6316                    path_ids.push(path_id);
6317                    probed_addresses.push(remote);
6318                }
6319                Ok((path_id, _)) => {
6320                    trace!(%path_id, %remote,"nat traversal: path existed for remote")
6321                }
6322                Err(e) => {
6323                    debug!(%remote, %e,"nat traversal: failed to probe remote");
6324                    err.get_or_insert(e);
6325                }
6326            }
6327        }
6328
6329        if let Some(err) = err {
6330            // We failed to probe any addresses, bail out
6331            if probed_addresses.is_empty() {
6332                return Err(iroh_hp::Error::Multipath(err));
6333            }
6334        }
6335
6336        self.iroh_hp
6337            .client_side_mut()
6338            .expect("connection side validated")
6339            .set_round_path_ids(path_ids);
6340
6341        Ok(probed_addresses)
6342    }
6343}
6344
6345impl fmt::Debug for Connection {
6346    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
6347        f.debug_struct("Connection")
6348            .field("handshake_cid", &self.handshake_cid)
6349            .finish()
6350    }
6351}
6352
6353#[derive(Debug, Copy, Clone, PartialEq, Eq)]
6354enum PathBlocked {
6355    No,
6356    AntiAmplification,
6357    Congestion,
6358    Pacing,
6359}
6360
6361/// Fields of `Connection` specific to it being client-side or server-side
6362enum ConnectionSide {
6363    Client {
6364        /// Sent in every outgoing Initial packet. Always empty after Initial keys are discarded
6365        token: Bytes,
6366        token_store: Arc<dyn TokenStore>,
6367        server_name: String,
6368    },
6369    Server {
6370        server_config: Arc<ServerConfig>,
6371    },
6372}
6373
6374impl ConnectionSide {
6375    fn remote_may_migrate(&self, state: &State) -> bool {
6376        match self {
6377            Self::Server { server_config } => server_config.migration,
6378            Self::Client { .. } => {
6379                if let Some(hs) = state.as_handshake() {
6380                    hs.allow_server_migration
6381                } else {
6382                    false
6383                }
6384            }
6385        }
6386    }
6387
6388    fn is_client(&self) -> bool {
6389        self.side().is_client()
6390    }
6391
6392    fn is_server(&self) -> bool {
6393        self.side().is_server()
6394    }
6395
6396    fn side(&self) -> Side {
6397        match *self {
6398            Self::Client { .. } => Side::Client,
6399            Self::Server { .. } => Side::Server,
6400        }
6401    }
6402}
6403
6404impl From<SideArgs> for ConnectionSide {
6405    fn from(side: SideArgs) -> Self {
6406        match side {
6407            SideArgs::Client {
6408                token_store,
6409                server_name,
6410            } => Self::Client {
6411                token: token_store.take(&server_name).unwrap_or_default(),
6412                token_store,
6413                server_name,
6414            },
6415            SideArgs::Server {
6416                server_config,
6417                pref_addr_cid: _,
6418                path_validated: _,
6419            } => Self::Server { server_config },
6420        }
6421    }
6422}
6423
6424/// Parameters to `Connection::new` specific to it being client-side or server-side
6425pub(crate) enum SideArgs {
6426    Client {
6427        token_store: Arc<dyn TokenStore>,
6428        server_name: String,
6429    },
6430    Server {
6431        server_config: Arc<ServerConfig>,
6432        pref_addr_cid: Option<ConnectionId>,
6433        path_validated: bool,
6434    },
6435}
6436
6437impl SideArgs {
6438    pub(crate) fn pref_addr_cid(&self) -> Option<ConnectionId> {
6439        match *self {
6440            Self::Client { .. } => None,
6441            Self::Server { pref_addr_cid, .. } => pref_addr_cid,
6442        }
6443    }
6444
6445    pub(crate) fn path_validated(&self) -> bool {
6446        match *self {
6447            Self::Client { .. } => true,
6448            Self::Server { path_validated, .. } => path_validated,
6449        }
6450    }
6451
6452    pub(crate) fn side(&self) -> Side {
6453        match *self {
6454            Self::Client { .. } => Side::Client,
6455            Self::Server { .. } => Side::Server,
6456        }
6457    }
6458}
6459
6460/// Reasons why a connection might be lost
6461#[derive(Debug, Error, Clone, PartialEq, Eq)]
6462pub enum ConnectionError {
6463    /// The peer doesn't implement any supported version
6464    #[error("peer doesn't implement any supported version")]
6465    VersionMismatch,
6466    /// The peer violated the QUIC specification as understood by this implementation
6467    #[error(transparent)]
6468    TransportError(#[from] TransportError),
6469    /// The peer's QUIC stack aborted the connection automatically
6470    #[error("aborted by peer: {0}")]
6471    ConnectionClosed(frame::ConnectionClose),
6472    /// The peer closed the connection
6473    #[error("closed by peer: {0}")]
6474    ApplicationClosed(frame::ApplicationClose),
6475    /// The peer is unable to continue processing this connection, usually due to having restarted
6476    #[error("reset by peer")]
6477    Reset,
6478    /// Communication with the peer has lapsed for longer than the negotiated idle timeout
6479    ///
6480    /// If neither side is sending keep-alives, a connection will time out after a long enough idle
6481    /// period even if the peer is still reachable. See also [`TransportConfig::max_idle_timeout()`]
6482    /// and [`TransportConfig::keep_alive_interval()`].
6483    #[error("timed out")]
6484    TimedOut,
6485    /// The local application closed the connection
6486    #[error("closed")]
6487    LocallyClosed,
6488    /// The connection could not be created because not enough of the CID space is available
6489    ///
6490    /// Try using longer connection IDs.
6491    #[error("CIDs exhausted")]
6492    CidsExhausted,
6493}
6494
6495impl From<Close> for ConnectionError {
6496    fn from(x: Close) -> Self {
6497        match x {
6498            Close::Connection(reason) => Self::ConnectionClosed(reason),
6499            Close::Application(reason) => Self::ApplicationClosed(reason),
6500        }
6501    }
6502}
6503
6504// For compatibility with API consumers
6505impl From<ConnectionError> for io::Error {
6506    fn from(x: ConnectionError) -> Self {
6507        use ConnectionError::*;
6508        let kind = match x {
6509            TimedOut => io::ErrorKind::TimedOut,
6510            Reset => io::ErrorKind::ConnectionReset,
6511            ApplicationClosed(_) | ConnectionClosed(_) => io::ErrorKind::ConnectionAborted,
6512            TransportError(_) | VersionMismatch | LocallyClosed | CidsExhausted => {
6513                io::ErrorKind::Other
6514            }
6515        };
6516        Self::new(kind, x)
6517    }
6518}
6519
6520/// Errors that might trigger a path being closed
6521// TODO(@divma): maybe needs to be reworked based on what we want to do with the public API
6522#[derive(Debug, Error, PartialEq, Eq, Clone, Copy)]
6523pub enum PathError {
6524    /// The extension was not negotiated with the peer
6525    #[error("multipath extension not negotiated")]
6526    MultipathNotNegotiated,
6527    /// Paths can only be opened client-side
6528    #[error("the server side may not open a path")]
6529    ServerSideNotAllowed,
6530    /// Current limits do not allow us to open more paths
6531    #[error("maximum number of concurrent paths reached")]
6532    MaxPathIdReached,
6533    /// No remote CIDs available to open a new path
6534    #[error("remoted CIDs exhausted")]
6535    RemoteCidsExhausted,
6536    /// Path could not be validated and will be abandoned
6537    #[error("path validation failed")]
6538    ValidationFailed,
6539    /// The remote address for the path is not supported by the endpoint
6540    #[error("invalid remote address")]
6541    InvalidRemoteAddress(SocketAddr),
6542}
6543
6544/// Errors triggered when abandoning a path
6545#[derive(Debug, Error, Clone, Eq, PartialEq)]
6546pub enum ClosePathError {
6547    /// The path is already closed or was never opened
6548    #[error("closed path")]
6549    ClosedPath,
6550    /// This is the last path, which can not be abandoned
6551    #[error("last open path")]
6552    LastOpenPath,
6553}
6554
6555#[derive(Debug, Error, Clone, Copy)]
6556#[error("Multipath extension not negotiated")]
6557pub struct MultipathNotNegotiated {
6558    _private: (),
6559}
6560
6561/// Events of interest to the application
6562#[derive(Debug)]
6563pub enum Event {
6564    /// The connection's handshake data is ready
6565    HandshakeDataReady,
6566    /// The connection was successfully established
6567    Connected,
6568    /// The TLS handshake was confirmed
6569    HandshakeConfirmed,
6570    /// The connection was lost
6571    ///
6572    /// Emitted if the peer closes the connection or an error is encountered.
6573    ConnectionLost {
6574        /// Reason that the connection was closed
6575        reason: ConnectionError,
6576    },
6577    /// Stream events
6578    Stream(StreamEvent),
6579    /// One or more application datagrams have been received
6580    DatagramReceived,
6581    /// One or more application datagrams have been sent after blocking
6582    DatagramsUnblocked,
6583    /// (Multi)Path events
6584    Path(PathEvent),
6585    /// Iroh's nat traversal events
6586    NatTraversal(iroh_hp::Event),
6587}
6588
6589impl From<PathEvent> for Event {
6590    fn from(source: PathEvent) -> Self {
6591        Self::Path(source)
6592    }
6593}
6594
6595fn get_max_ack_delay(params: &TransportParameters) -> Duration {
6596    Duration::from_micros(params.max_ack_delay.0 * 1000)
6597}
6598
6599// Prevents overflow and improves behavior in extreme circumstances
6600const MAX_BACKOFF_EXPONENT: u32 = 16;
6601
6602/// Minimal remaining size to allow packet coalescing, excluding cryptographic tag
6603///
6604/// This must be at least as large as the header for a well-formed empty packet to be coalesced,
6605/// plus some space for frames. We only care about handshake headers because short header packets
6606/// necessarily have smaller headers, and initial packets are only ever the first packet in a
6607/// datagram (because we coalesce in ascending packet space order and the only reason to split a
6608/// packet is when packet space changes).
6609const MIN_PACKET_SPACE: usize = MAX_HANDSHAKE_OR_0RTT_HEADER_SIZE + 32;
6610
6611/// Largest amount of space that could be occupied by a Handshake or 0-RTT packet's header
6612///
6613/// Excludes packet-type-specific fields such as packet number or Initial token
6614// https://www.rfc-editor.org/rfc/rfc9000.html#name-0-rtt: flags + version + dcid len + dcid +
6615// scid len + scid + length + pn
6616const MAX_HANDSHAKE_OR_0RTT_HEADER_SIZE: usize =
6617    1 + 4 + 1 + MAX_CID_SIZE + 1 + MAX_CID_SIZE + VarInt::from_u32(u16::MAX as u32).size() + 4;
6618
6619/// Perform key updates this many packets before the AEAD confidentiality limit.
6620///
6621/// Chosen arbitrarily, intended to be large enough to prevent spurious connection loss.
6622const KEY_UPDATE_MARGIN: u64 = 10_000;
6623
6624#[derive(Default)]
6625struct SentFrames {
6626    retransmits: ThinRetransmits,
6627    /// The packet number of the largest acknowledged packet for each path
6628    largest_acked: FxHashMap<PathId, u64>,
6629    stream_frames: StreamMetaVec,
6630    /// Whether the packet contains non-retransmittable frames (like datagrams)
6631    non_retransmits: bool,
6632    /// If the datagram containing these frames should be padded to the min MTU
6633    requires_padding: bool,
6634}
6635
6636impl SentFrames {
6637    /// Returns whether the packet contains only ACKs
6638    fn is_ack_only(&self, streams: &StreamsState) -> bool {
6639        !self.largest_acked.is_empty()
6640            && !self.non_retransmits
6641            && self.stream_frames.is_empty()
6642            && self.retransmits.is_empty(streams)
6643    }
6644}
6645
6646/// Compute the negotiated idle timeout based on local and remote max_idle_timeout transport parameters.
6647///
6648/// According to the definition of max_idle_timeout, a value of `0` means the timeout is disabled; see <https://www.rfc-editor.org/rfc/rfc9000#section-18.2-4.4.1.>
6649///
6650/// According to the negotiation procedure, either the minimum of the timeouts or one specified is used as the negotiated value; see <https://www.rfc-editor.org/rfc/rfc9000#section-10.1-2.>
6651///
6652/// Returns the negotiated idle timeout as a `Duration`, or `None` when both endpoints have opted out of idle timeout.
6653fn negotiate_max_idle_timeout(x: Option<VarInt>, y: Option<VarInt>) -> Option<Duration> {
6654    match (x, y) {
6655        (Some(VarInt(0)) | None, Some(VarInt(0)) | None) => None,
6656        (Some(VarInt(0)) | None, Some(y)) => Some(Duration::from_millis(y.0)),
6657        (Some(x), Some(VarInt(0)) | None) => Some(Duration::from_millis(x.0)),
6658        (Some(x), Some(y)) => Some(Duration::from_millis(cmp::min(x, y).0)),
6659    }
6660}
6661
6662#[cfg(test)]
6663mod tests {
6664    use super::*;
6665
6666    #[test]
6667    fn negotiate_max_idle_timeout_commutative() {
6668        let test_params = [
6669            (None, None, None),
6670            (None, Some(VarInt(0)), None),
6671            (None, Some(VarInt(2)), Some(Duration::from_millis(2))),
6672            (Some(VarInt(0)), Some(VarInt(0)), None),
6673            (
6674                Some(VarInt(2)),
6675                Some(VarInt(0)),
6676                Some(Duration::from_millis(2)),
6677            ),
6678            (
6679                Some(VarInt(1)),
6680                Some(VarInt(4)),
6681                Some(Duration::from_millis(1)),
6682            ),
6683        ];
6684
6685        for (left, right, result) in test_params {
6686            assert_eq!(negotiate_max_idle_timeout(left, right), result);
6687            assert_eq!(negotiate_max_idle_timeout(right, left), result);
6688        }
6689    }
6690}