soe-protocol 1.0.3

Sony Online Entertainment protocol implementation in Rust.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
//! The reliable data output channel: converts application data into ordered,
//! fragmented reliable data packets, and resends them until acknowledged.
//!
//! This is a port of the reference implementation's simplified
//! `ReliableDataOutputChannel2`, which trades the original's multi-packet bundling
//! for a much simpler (and less bug-prone) go-back-N style window.
//!
//! Like the input channel, this is an I/O-agnostic component: enqueued data is fragmented
//! into outgoing packets which accumulate in an internal queue. Calling
//! [`ReliableDataOutputChannel::run_tick`] moves due packets into the outgoing buffer
//! (drained via [`ReliableDataOutputChannel::take_outgoing`]). Acknowledgements are
//! fed back in via [`ReliableDataOutputChannel::notify_of_acknowledge`] /
//! [`ReliableDataOutputChannel::notify_of_acknowledge_all`]. Time is supplied by the
//! caller as [`Instant`] values.

use std::collections::VecDeque;
use std::time::{Duration, Instant};

use bytes::{BufMut, Bytes, BytesMut};

use crate::protocol::OpCode;
use crate::rc4::Rc4KeyState;

use super::true_incoming_sequence;

/// The size of a reliable data packet's sequence prefix.
const SEQUENCE_SIZE: usize = 2;
/// The size of a master fragment's total-length prefix.
const FRAGMENT_LENGTH_SIZE: usize = 4;

/// Adaptive retransmit-timeout (RTO) tuning. Uses the Jacobson/Karels SRTT/RTTVAR
/// estimator (SIGCOMM '88, "Congestion Avoidance and Control"), as later standardized
/// for TCP in RFC 6298.
/// `RTO = SRTT + max(RTO_GRANULARITY, RTT_K * RTTVAR)`, clamped to [RTO_MIN, RTO_MAX].
const RTT_K: u32 = 4;
/// Floor on the variance term so the RTO keeps headroom above a perfectly steady RTT
/// (otherwise RTTVAR -> 0 makes RTO == RTT and the timer fires right as acks arrive).
const RTO_GRANULARITY: Duration = Duration::from_millis(100);
/// Lower clamp on the computed RTO; avoids spurious resends on very low-RTT links.
const RTO_MIN: Duration = Duration::from_millis(200);
/// Upper clamp on the computed RTO (also the ceiling for exponential backoff).
const RTO_MAX: Duration = Duration::from_secs(8);

/// Statistics gathered while sending reliable data.
#[derive(Debug, Default, Clone)]
pub struct DataOutputStats {
    /// Total reliable data packets dispatched, including re-sends.
    pub total_sent: u64,
    /// Total reliable data packets that were re-sent.
    pub total_resent: u64,
    /// Total acknowledgement packets received (including ack-alls).
    pub incoming_acknowledge_count: u64,
    /// Total reliable data packets acknowledged (including via ack-all).
    pub actual_acknowledge_count: u64,
}

/// Configuration controlling the output channel's behaviour.
#[derive(Debug, Clone)]
pub struct OutputConfig {
    /// The maximum length, in bytes, of the data portion (sequence + data) of a
    /// single reliable data packet. This is the remote UDP length minus the OP code
    /// and CRC.
    pub max_data_length: usize,
    /// The maximum number of unacknowledged reliable data packets that may be in
    /// flight at once (the send window).
    pub max_queued_outgoing: usize,
    /// The INITIAL retransmit timeout, used before any round-trip time has been
    /// measured. Once acknowledgements start arriving the channel derives an adaptive
    /// RTO from the measured RTT (see `RTT_K`/`RTO_MIN`/`RTO_MAX`), so this value only
    /// governs the very first window.
    pub ack_wait: Duration,
}

impl Default for OutputConfig {
    fn default() -> Self {
        Self {
            max_data_length: 508,
            max_queued_outgoing: 196,
            ack_wait: Duration::from_millis(500),
        }
    }
}

/// A reliable data packet the channel wishes to send (without OP code or CRC
/// framing, which the session layer applies).
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct OutgoingReliable {
    /// The OP code of the packet ([`OpCode::ReliableData`] or
    /// [`OpCode::ReliableDataFragment`]).
    pub op_code: OpCode,
    /// The packet payload: a big-endian `u16` sequence, an optional big-endian `u32`
    /// total-length prefix (master fragments only), and the data chunk.
    pub payload: Bytes,
}

#[derive(Debug)]
struct StashedOutputPacket {
    is_fragment: bool,
    data: Bytes,
    sent: bool,
    /// When this packet was most recently (re)sent, for RTT measurement and per-packet
    /// retransmit timing. `None` until first dispatched.
    sent_at: Option<Instant>,
    /// Set once the packet has been retransmitted: its RTT becomes ambiguous, so it must
    /// not be used as a round-trip sample (Karn's algorithm).
    resent: bool,
}

/// Converts application data into ordered, fragmented reliable data packets.
#[derive(Debug)]
pub struct ReliableDataOutputChannel {
    config: OutputConfig,
    cipher: Option<Rc4KeyState>,

    dispatch_queue: VecDeque<(i64, StashedOutputPacket)>,

    /// The total number of sequences that have been output.
    total_sequence: i64,
    /// The maximum sequence number that the client is known to have received.
    max_client_sequence: i64,
    /// The index into `dispatch_queue` of the next packet to dispatch.
    current_dispatch_index: usize,

    /// Smoothed round-trip time estimate (`None` until the first RTT sample).
    srtt: Option<Duration>,
    /// Round-trip time variation estimate (the RTTVAR term of the Jacobson/Karels estimator).
    rttvar: Duration,
    /// Current retransmit timeout: adaptive once RTT is known, else `config.ack_wait`.
    rto: Duration,

    outgoing: Vec<OutgoingReliable>,
    stats: DataOutputStats,
}

impl ReliableDataOutputChannel {
    /// Creates a new output channel. `cipher` is the initial RC4 key state; pass
    /// `Some(..)` to enable RC4 encryption of the proxied application data, or `None`
    /// to pass it through unencrypted.
    pub fn new(config: OutputConfig, cipher: Option<Rc4KeyState>, _now: Instant) -> Self {
        let initial_rto = config.ack_wait;
        Self {
            config,
            cipher,
            dispatch_queue: VecDeque::new(),
            total_sequence: 0,
            max_client_sequence: 0,
            current_dispatch_index: 0,
            srtt: None,
            rttvar: Duration::ZERO,
            rto: initial_rto,
            outgoing: Vec::new(),
            stats: DataOutputStats::default(),
        }
    }

    /// Returns the gathered output statistics.
    pub fn stats(&self) -> &DataOutputStats {
        &self.stats
    }

    /// Drains the outgoing reliable data packets accumulated so far.
    pub fn take_outgoing(&mut self) -> Vec<OutgoingReliable> {
        std::mem::take(&mut self.outgoing)
    }

    /// Returns the number of reliable data packets currently awaiting acknowledgement.
    pub fn queued_len(&self) -> usize {
        self.dispatch_queue.len()
    }

    /// Sets the maximum length of the data portion (sequence + data) of a single
    /// packet. Should not be called after data has been enqueued.
    pub fn set_max_data_length(&mut self, max_data_length: usize) {
        self.config.max_data_length = max_data_length;
    }

    fn max_chunk(&self) -> usize {
        self.config.max_data_length - SEQUENCE_SIZE
    }

    /// Enqueues application data to be sent on the reliable channel. The data is
    /// fragmented as required to fit within the configured maximum packet length.
    pub fn enqueue_data(&mut self, data: &[u8]) {
        if data.is_empty() {
            return;
        }

        let mut remaining: Bytes = match &mut self.cipher {
            Some(_) => self.encrypt(data),
            None => Bytes::copy_from_slice(data),
        };

        let is_fragment = remaining.len() > self.max_chunk();
        self.stash_fragment(&mut remaining, true, is_fragment);
        while !remaining.is_empty() {
            self.stash_fragment(&mut remaining, false, true);
        }
    }

    /// Runs a tick of the output channel, moving due packets into the outgoing
    /// buffer. If the oldest in-flight packet has gone unacknowledged for longer than
    /// the current (adaptive) retransmit timeout, dispatch restarts from the front of
    /// the window (go-back-N) and the RTO is backed off exponentially (Karn).
    pub fn run_tick(&mut self, now: Instant) {
        // Retransmission timeout is keyed off the OLDEST in-flight packet's own send time
        // (not a single global timer), so a long quiet period before the first ack can't
        // make every tick resend the whole window.
        let timed_out = match self.dispatch_queue.front() {
            Some((_, front)) if front.sent => front
                .sent_at
                .is_some_and(|sent_at| now.duration_since(sent_at) > self.rto),
            _ => false,
        };
        if timed_out {
            self.current_dispatch_index = 0;
            // Karn's exponential backoff; a fresh (unambiguous) RTT sample resets this.
            self.rto = (self.rto * 2).min(RTO_MAX);
        }

        let max_index = self
            .dispatch_queue
            .len()
            .min(self.config.max_queued_outgoing);

        while self.current_dispatch_index < max_index {
            let (_, packet) = &mut self.dispatch_queue[self.current_dispatch_index];
            let op_code = if packet.is_fragment {
                OpCode::ReliableDataFragment
            } else {
                OpCode::ReliableData
            };

            self.stats.total_sent += 1;
            if packet.sent {
                self.stats.total_resent += 1;
                // A retransmitted packet's RTT is ambiguous; exclude it from sampling.
                packet.resent = true;
            }
            packet.sent = true;
            packet.sent_at = Some(now);

            let payload = packet.data.clone();
            self.outgoing.push(OutgoingReliable { op_code, payload });
            self.current_dispatch_index += 1;
        }
    }

    /// Folds a fresh, unambiguous round-trip sample into the smoothed RTT/variance
    /// estimates and recomputes the adaptive retransmit timeout (Jacobson/Karels estimator).
    fn update_rto(&mut self, sample: Duration) {
        match self.srtt {
            None => {
                self.srtt = Some(sample);
                self.rttvar = sample / 2;
            }
            Some(srtt) => {
                let diff = srtt.abs_diff(sample);
                // RTTVAR = 3/4 * RTTVAR + 1/4 * |SRTT - sample|
                self.rttvar = (self.rttvar * 3 + diff) / 4;
                // SRTT = 7/8 * SRTT + 1/8 * sample
                self.srtt = Some((srtt * 7 + sample) / 8);
            }
        }
        let srtt = self.srtt.unwrap_or(sample);
        let rto = srtt + std::cmp::max(RTO_GRANULARITY, self.rttvar * RTT_K);
        self.rto = rto.clamp(RTO_MIN, RTO_MAX);
    }

    /// Notifies the channel that the remote has acknowledged a single sequence.
    pub fn notify_of_acknowledge(&mut self, sequence: u16, now: Instant) {
        let seq = self.true_incoming(sequence);
        self.stats.incoming_acknowledge_count += 1;

        if let Some(pos) = self.dispatch_queue.iter().position(|(s, _)| *s == seq) {
            let (_, pkt) = &self.dispatch_queue[pos];
            let sample = (pkt.sent && !pkt.resent)
                .then(|| pkt.sent_at.map(|sent_at| now.duration_since(sent_at)))
                .flatten();
            self.dispatch_queue.remove(pos);
            self.current_dispatch_index = self.current_dispatch_index.saturating_sub(1);
            self.stats.actual_acknowledge_count += 1;
            if let Some(sample) = sample {
                self.update_rto(sample);
            }
        }

        if seq > self.max_client_sequence {
            self.max_client_sequence = seq;
        }
    }

    /// Notifies the channel that the remote has acknowledged all sequences up to and
    /// including the given one.
    pub fn notify_of_acknowledge_all(&mut self, sequence: u16, now: Instant) {
        let seq = self.true_incoming(sequence);
        self.stats.incoming_acknowledge_count += 1;

        let mut sample: Option<Duration> = None;
        loop {
            let (pop, this_sample) = match self.dispatch_queue.front() {
                Some((s, pkt)) if *s <= seq => {
                    let smp = (pkt.sent && !pkt.resent)
                        .then(|| pkt.sent_at.map(|sent_at| now.duration_since(sent_at)))
                        .flatten();
                    (true, smp)
                }
                _ => (false, None),
            };
            if !pop {
                break;
            }
            // Keep the freshest (most recently sent) unambiguous sample in this batch.
            if this_sample.is_some() {
                sample = this_sample;
            }
            self.dispatch_queue.pop_front();
            self.current_dispatch_index = self.current_dispatch_index.saturating_sub(1);
            self.stats.actual_acknowledge_count += 1;
        }

        if let Some(sample) = sample {
            self.update_rto(sample);
        }

        if seq > self.max_client_sequence {
            self.max_client_sequence = seq;
        }
    }

    fn stash_fragment(&mut self, data: &mut Bytes, is_master: bool, is_fragment: bool) {
        let mut amount = data.len().min(self.max_chunk());

        let mut buf = BytesMut::with_capacity(SEQUENCE_SIZE + FRAGMENT_LENGTH_SIZE + amount);
        buf.put_u16(self.total_sequence as u16);

        if is_master && is_fragment {
            buf.put_u32(data.len() as u32);
            amount -= FRAGMENT_LENGTH_SIZE;
        }

        buf.extend_from_slice(&data[..amount]);

        self.dispatch_queue.push_back((
            self.total_sequence,
            StashedOutputPacket {
                is_fragment,
                data: buf.freeze(),
                sent: false,
                sent_at: None,
                resent: false,
            },
        ));

        self.total_sequence += 1;
        *data = data.slice(amount..);
    }

    /// Encrypts `data` with the channel's RC4 cipher. A leading zero byte is
    /// prepended when the ciphertext itself begins with a zero, mirroring the input
    /// channel's padding-strip logic.
    fn encrypt(&mut self, data: &[u8]) -> Bytes {
        let cipher = self
            .cipher
            .as_mut()
            .expect("encrypt called without a cipher");

        let mut buf = BytesMut::with_capacity(data.len() + 1);
        buf.put_u8(0);
        buf.extend_from_slice(data);
        cipher.transform_in_place(&mut buf[1..]);

        let frozen = buf.freeze();
        if frozen[1] == 0 {
            frozen
        } else {
            frozen.slice(1..)
        }
    }

    fn true_incoming(&self, packet_sequence: u16) -> i64 {
        true_incoming_sequence(
            packet_sequence,
            self.max_client_sequence,
            self.config.max_queued_outgoing as i64,
        )
    }
}

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

    const MAX_DATA_LENGTH: usize = 506; // 512 (udp) - 2 (op) - 2 (seq) - 2 (crc)
    const FRAGMENT_WINDOW_SIZE: usize = 8;

    struct Clock {
        now: Instant,
    }

    impl Clock {
        fn new() -> Self {
            Self {
                now: Instant::now(),
            }
        }
        fn advance(&mut self, by: Duration) -> Instant {
            self.now += by;
            self.now
        }
    }

    fn new_channel(clock: &Clock) -> ReliableDataOutputChannel {
        let config = OutputConfig {
            max_data_length: MAX_DATA_LENGTH + SEQUENCE_SIZE,
            max_queued_outgoing: FRAGMENT_WINDOW_SIZE,
            ack_wait: Duration::from_millis(500),
        };
        ReliableDataOutputChannel::new(config, None, clock.now)
    }

    /// A deterministic pseudo-random byte buffer.
    fn generate_packet(size: usize) -> Vec<u8> {
        let mut state: u32 = 0x1234_5678 ^ size as u32;
        (0..size)
            .map(|_| {
                state = state.wrapping_mul(1_664_525).wrapping_add(1_013_904_223);
                (state >> 24) as u8
            })
            .collect()
    }

    /// Asserts that the data carried by `packets` (stripping the sequence and, for
    /// the first packet if `expect_master_fragment`, the length prefix) concatenates
    /// to exactly `buffer`.
    fn assert_packets_equal_buffer(
        packets: &[OutgoingReliable],
        buffer: &[u8],
        mut expect_master_fragment: bool,
    ) {
        let mut position = 0;
        for packet in packets {
            let data_offset = SEQUENCE_SIZE
                + if expect_master_fragment {
                    FRAGMENT_LENGTH_SIZE
                } else {
                    0
                };
            expect_master_fragment = false;

            let data = &packet.payload[data_offset..];
            assert!(
                position + data.len() <= buffer.len(),
                "received more data than expected"
            );
            assert_eq!(&buffer[position..position + data.len()], data);
            position += data.len();
        }
        assert_eq!(position, buffer.len(), "did not receive the whole buffer");
    }

    #[test]
    fn repeats_data_on_ack_failure() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let fragment_count = 4;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);

        ch.enqueue_data(&packet);
        ch.run_tick(clock.advance(Duration::from_millis(1)));
        assert_packets_equal_buffer(&ch.take_outgoing(), &packet, true);

        // Don't acknowledge; after the ack wait elapses the data is resent in full.
        ch.run_tick(clock.advance(Duration::from_millis(600)));
        assert_packets_equal_buffer(&ch.take_outgoing(), &packet, true);
    }

    #[test]
    fn repeats_data_from_arbitrary_position_on_ack_delay() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let fragment_count = 4;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);

        ch.enqueue_data(&packet);
        ch.run_tick(clock.advance(Duration::from_millis(1)));
        assert_packets_equal_buffer(&ch.take_outgoing(), &packet, true);

        ch.notify_of_acknowledge_all(1, clock.advance(Duration::from_millis(1)));

        ch.run_tick(clock.advance(Duration::from_millis(600)));
        // The master fragment (MAX-4) and the next fragment (MAX) were acknowledged.
        let expected_consumed = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH;
        assert_packets_equal_buffer(&ch.take_outgoing(), &packet[expected_consumed..], false);
    }

    #[test]
    fn repeats_full_window_from_arbitrary_position_on_ack_delay() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let fragment_count = FRAGMENT_WINDOW_SIZE * 2;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);

        ch.enqueue_data(&packet);
        ch.run_tick(clock.advance(Duration::from_millis(1)));

        // Only a full window of packets is sent initially.
        let expected_receive_length =
            MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (FRAGMENT_WINDOW_SIZE - 1);
        assert_packets_equal_buffer(
            &ch.take_outgoing(),
            &packet[..expected_receive_length],
            true,
        );

        ch.notify_of_acknowledge_all(
            (FRAGMENT_WINDOW_SIZE - 2) as u16,
            clock.advance(Duration::from_millis(1)),
        );
        ch.run_tick(clock.advance(Duration::from_millis(600)));

        let expected_consumed = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (FRAGMENT_WINDOW_SIZE - 2);
        let expected_repeat_length = MAX_DATA_LENGTH * FRAGMENT_WINDOW_SIZE;
        assert_packets_equal_buffer(
            &ch.take_outgoing(),
            &packet[expected_consumed..expected_consumed + expected_repeat_length],
            false,
        );
    }

    #[test]
    fn single_small_packet_is_not_fragmented() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let data = generate_packet(32);
        ch.enqueue_data(&data);
        ch.run_tick(clock.advance(Duration::from_millis(1)));

        let outgoing = ch.take_outgoing();
        assert_eq!(outgoing.len(), 1);
        assert_eq!(outgoing[0].op_code, OpCode::ReliableData);
        // No length prefix: payload is [seq u16][data].
        assert_eq!(&outgoing[0].payload[SEQUENCE_SIZE..], &data[..]);
    }

    #[test]
    fn single_ack_removes_specific_packet() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * 3;
        let packet = generate_packet(packet_length);
        ch.enqueue_data(&packet);
        assert_eq!(ch.queued_len(), 4);

        ch.run_tick(clock.advance(Duration::from_millis(1)));
        let _ = ch.take_outgoing();

        ch.notify_of_acknowledge(2, clock.advance(Duration::from_millis(1)));
        assert_eq!(ch.queued_len(), 3);
        assert_eq!(ch.stats().actual_acknowledge_count, 1);
    }

    /// Across consecutive ticks WITHOUT acknowledgement, the number of unacknowledged
    /// packets in flight must never exceed `max_queued_outgoing`. (Regression: the window
    /// ceiling was computed relative to the already-advanced dispatch index, so each tick
    /// admitted another full window -> unbounded in-flight growth -> client RCVBUF overflow.)
    #[test]
    fn window_does_not_grow_across_ticks_without_ack() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        // Enqueue far more than one window's worth of fragments.
        let fragment_count = FRAGMENT_WINDOW_SIZE * 4;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);
        ch.enqueue_data(&packet);

        // Tick 1: a full window goes out.
        ch.run_tick(clock.advance(Duration::from_millis(1)));
        let mut in_flight = ch.take_outgoing().len();
        assert_eq!(
            in_flight, FRAGMENT_WINDOW_SIZE,
            "first tick should send exactly one window"
        );

        // Several more ticks, no ack, well within ack_wait: nothing new may be sent
        // because the window is still full of unacknowledged packets.
        for _ in 0..5 {
            ch.run_tick(clock.advance(Duration::from_millis(10)));
            in_flight += ch.take_outgoing().len();
            assert!(
                in_flight <= FRAGMENT_WINDOW_SIZE,
                "in-flight unacked packets ({in_flight}) exceeded the window ({FRAGMENT_WINDOW_SIZE})",
            );
        }
    }

    /// Once a real round-trip time has been measured, the retransmit timeout must adapt
    /// upward so a quiet gap LONGER than the initial `ack_wait` no longer triggers a
    /// spurious resend. A fixed RTO (== ack_wait) would resend here; the adaptive RTO
    /// (SRTT + 4*RTTVAR after a ~500ms sample => ~1.5s) must not.
    #[test]
    fn adaptive_rto_suppresses_resend_after_learning_high_rtt() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock); // ack_wait = 500ms

        // Send one window.
        let fragment_count = FRAGMENT_WINDOW_SIZE + 4;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);
        ch.enqueue_data(&packet);
        ch.run_tick(clock.advance(Duration::from_millis(1)));
        let _ = ch.take_outgoing();

        // Acknowledge the whole first window after a ~500ms round trip: this feeds the
        // RTO estimator a 500ms sample, raising the RTO well above the 500ms ack_wait.
        ch.notify_of_acknowledge_all(
            (FRAGMENT_WINDOW_SIZE - 1) as u16,
            clock.advance(Duration::from_millis(500)),
        );

        // The newly admitted window is now in flight. Advance 600ms (> the old fixed
        // ack_wait) with no further ack. With a fixed RTO this resends the window; with
        // the adaptive RTO (~1.5s) it must NOT.
        ch.run_tick(clock.advance(Duration::from_millis(1)));
        let _ = ch.take_outgoing();
        ch.run_tick(clock.advance(Duration::from_millis(600)));
        let resent = ch.take_outgoing();

        assert!(
            resent.is_empty(),
            "adaptive RTO must not resend within the learned RTT, but resent {} packets",
            resent.len()
        );
        assert_eq!(
            ch.stats().total_resent,
            0,
            "no packet should have been retransmitted after the RTO adapted to the RTT"
        );
    }

    /// End-to-end drain at RTT ~= the initial `ack_wait`: the channel must deliver every
    /// packet while keeping in-flight within ~1 window and the on-wire datagram count close
    /// to the unique count (no resend storm). Models a delayed pipe with cumulative acks.
    #[test]
    fn adaptive_rto_bounds_inflight_at_high_rtt() {
        let mut clock = Clock::new();
        let mut ch = new_channel(&clock);

        let one_way = Duration::from_millis(250); // RTT ~= ack_wait (500ms)
        let tick = Duration::from_millis(5);

        let fragment_count = 30;
        let packet_length = MAX_DATA_LENGTH - 4 + MAX_DATA_LENGTH * (fragment_count - 1);
        let packet = generate_packet(packet_length);
        ch.enqueue_data(&packet);
        let unique = ch.queued_len();

        let mut to_client: Vec<(Instant, u16)> = Vec::new();
        let mut to_server: Vec<(Instant, u16)> = Vec::new();
        let mut received = vec![false; unique];

        let mut total_on_wire = 0usize;
        let mut highest_sent: i64 = -1;
        let mut last_ack: i64 = -1;
        let mut max_in_flight: i64 = 0;

        for _ in 0..800 {
            let now = clock.advance(tick);

            // Deliver acks that have arrived back at the server.
            to_server.retain(|&(at, ack)| {
                if at <= now {
                    ch.notify_of_acknowledge_all(ack, now);
                    last_ack = last_ack.max(ack as i64);
                    false
                } else {
                    true
                }
            });

            // Deliver datagrams that have arrived at the client; ack the highest
            // contiguous sequence seen so far.
            let mut delivered_any = false;
            to_client.retain(|&(at, seq)| {
                if at <= now {
                    received[seq as usize] = true;
                    delivered_any = true;
                    false
                } else {
                    true
                }
            });
            if delivered_any {
                let mut hw: i64 = -1;
                for (seq, got) in received.iter().enumerate() {
                    if *got {
                        hw = seq as i64;
                    } else {
                        break;
                    }
                }
                if hw >= 0 {
                    to_server.push((now + one_way, hw as u16));
                }
            }

            ch.run_tick(now);
            for out in ch.take_outgoing() {
                let seq = u16::from_be_bytes([out.payload[0], out.payload[1]]);
                total_on_wire += 1;
                highest_sent = highest_sent.max(seq as i64);
                to_client.push((now + one_way, seq));
            }

            max_in_flight = max_in_flight.max(highest_sent - last_ack);

            if last_ack >= 0 && last_ack as usize + 1 == unique {
                break;
            }
        }

        assert!(
            last_ack >= 0 && last_ack as usize + 1 == unique,
            "channel did not drain all {unique} packets (acked through {last_ack})"
        );
        assert!(
            max_in_flight <= FRAGMENT_WINDOW_SIZE as i64 + 2,
            "in-flight ({max_in_flight}) far exceeded the window ({FRAGMENT_WINDOW_SIZE}) -> resend storm",
        );
        assert!(
            total_on_wire <= unique + unique / 4,
            "sent {total_on_wire} datagrams for {unique} unique packets (>1.25x = resend storm)",
        );
    }
}