summaryrefslogtreecommitdiffstats
path: root/net/quic/quic_sent_packet_manager.cc
blob: b8a702463f9810826808fee556989bc4f3f08e5c (plain)
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
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
// Copyright 2013 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#include "net/quic/quic_sent_packet_manager.h"

#include <algorithm>

#include "base/logging.h"
#include "base/stl_util.h"
#include "net/quic/congestion_control/pacing_sender.h"
#include "net/quic/crypto/crypto_protocol.h"
#include "net/quic/quic_ack_notifier_manager.h"
#include "net/quic/quic_connection_stats.h"
#include "net/quic/quic_flags.h"
#include "net/quic/quic_utils_chromium.h"

using std::make_pair;
using std::max;
using std::min;

namespace net {
namespace {
static const int kDefaultRetransmissionTimeMs = 500;
// TCP RFC calls for 1 second RTO however Linux differs from this default and
// define the minimum RTO to 200ms, we will use the same until we have data to
// support a higher or lower value.
static const int kMinRetransmissionTimeMs = 200;
static const int kMaxRetransmissionTimeMs = 60000;
static const size_t kMaxRetransmissions = 10;

// Only exponentially back off the handshake timer 5 times due to a timeout.
static const size_t kMaxHandshakeRetransmissionBackoffs = 5;
static const size_t kMinHandshakeTimeoutMs = 10;

// Sends up to two tail loss probes before firing an RTO,
// per draft RFC draft-dukkipati-tcpm-tcp-loss-probe.
static const size_t kDefaultMaxTailLossProbes = 2;
static const int64 kMinTailLossProbeTimeoutMs = 10;

// Number of samples before we force a new recent min rtt to be captured.
static const size_t kNumMinRttSamplesAfterQuiescence = 2;

bool HasCryptoHandshake(const TransmissionInfo& transmission_info) {
  if (transmission_info.retransmittable_frames == NULL) {
    return false;
  }
  return transmission_info.retransmittable_frames->HasCryptoHandshake() ==
      IS_HANDSHAKE;
}

}  // namespace

#define ENDPOINT (is_server_ ? "Server: " : " Client: ")

QuicSentPacketManager::QuicSentPacketManager(bool is_server,
                                             const QuicClock* clock,
                                             QuicConnectionStats* stats,
                                             CongestionFeedbackType type,
                                             LossDetectionType loss_type)
    : unacked_packets_(),
      is_server_(is_server),
      clock_(clock),
      stats_(stats),
      debug_delegate_(NULL),
      send_algorithm_(
          SendAlgorithmInterface::Create(clock, &rtt_stats_, type, stats)),
      loss_algorithm_(LossDetectionInterface::Create(loss_type)),
      largest_observed_(0),
      first_rto_transmission_(0),
      consecutive_rto_count_(0),
      consecutive_tlp_count_(0),
      consecutive_crypto_retransmission_count_(0),
      pending_tlp_transmission_(false),
      max_tail_loss_probes_(kDefaultMaxTailLossProbes),
      using_pacing_(false) {
}

QuicSentPacketManager::~QuicSentPacketManager() {
}

void QuicSentPacketManager::SetFromConfig(const QuicConfig& config) {
  if (config.HasReceivedInitialRoundTripTimeUs() &&
      config.ReceivedInitialRoundTripTimeUs() > 0) {
    rtt_stats_.set_initial_rtt_us(min(kMaxInitialRoundTripTimeUs,
                                      config.ReceivedInitialRoundTripTimeUs()));
  }
  // TODO(ianswett): BBR is currently a server only feature.
  if (config.HasReceivedConnectionOptions() &&
      ContainsQuicTag(config.ReceivedConnectionOptions(), kTBBR)) {
    send_algorithm_.reset(
        SendAlgorithmInterface::Create(clock_, &rtt_stats_, kTCPBBR, stats_));
  }
  if (config.congestion_feedback() == kPACE ||
      (config.HasReceivedConnectionOptions() &&
       ContainsQuicTag(config.ReceivedConnectionOptions(), kPACE))) {
    MaybeEnablePacing();
  }
  // TODO(ianswett): Remove the "HasReceivedLossDetection" branch once
  // the ConnectionOptions code is live everywhere.
  if ((config.HasReceivedLossDetection() &&
       config.ReceivedLossDetection() == kTIME) ||
      (config.HasReceivedConnectionOptions() &&
       ContainsQuicTag(config.ReceivedConnectionOptions(), kTIME))) {
    loss_algorithm_.reset(LossDetectionInterface::Create(kTime));
  }
  send_algorithm_->SetFromConfig(config, is_server_);
}

// TODO(ianswett): Combine this method with OnPacketSent once packets are always
// sent in order and the connection tracks RetransmittableFrames for longer.
void QuicSentPacketManager::OnSerializedPacket(
    const SerializedPacket& serialized_packet) {
  if (serialized_packet.retransmittable_frames) {
    ack_notifier_manager_.OnSerializedPacket(serialized_packet);
  }

  unacked_packets_.AddPacket(serialized_packet);
}

void QuicSentPacketManager::OnRetransmittedPacket(
    QuicPacketSequenceNumber old_sequence_number,
    QuicPacketSequenceNumber new_sequence_number) {
  TransmissionType transmission_type;
  PendingRetransmissionMap::iterator it =
      pending_retransmissions_.find(old_sequence_number);
  if (it != pending_retransmissions_.end()) {
    transmission_type = it->second;
    pending_retransmissions_.erase(it);
  } else {
    DLOG(DFATAL) << "Expected sequence number to be in "
        "pending_retransmissions_.  sequence_number: " << old_sequence_number;
    transmission_type = NOT_RETRANSMISSION;
  }

  // A notifier may be waiting to hear about ACKs for the original sequence
  // number. Inform them that the sequence number has changed.
  ack_notifier_manager_.UpdateSequenceNumber(old_sequence_number,
                                             new_sequence_number);

  unacked_packets_.OnRetransmittedPacket(old_sequence_number,
                                         new_sequence_number,
                                         transmission_type);
}

void QuicSentPacketManager::OnIncomingAck(
    const ReceivedPacketInfo& received_info,
    QuicTime ack_receive_time) {
  QuicByteCount bytes_in_flight = unacked_packets_.bytes_in_flight();

  // We rely on delta_time_largest_observed to compute an RTT estimate, so
  // we only update rtt when the largest observed gets acked.
  bool largest_observed_acked = MaybeUpdateRTT(received_info, ack_receive_time);
  if (largest_observed_ < received_info.largest_observed) {
    largest_observed_ = received_info.largest_observed;
    unacked_packets_.IncreaseLargestObserved(largest_observed_);
  }
  HandleAckForSentPackets(received_info);
  InvokeLossDetection(ack_receive_time);
  MaybeInvokeCongestionEvent(largest_observed_acked, bytes_in_flight);

  // If we have received a truncated ack, then we need to clear out some
  // previous transmissions to allow the peer to actually ACK new packets.
  if (received_info.is_truncated) {
    unacked_packets_.ClearPreviousRetransmissions(
        received_info.missing_packets.size() / 2);
  }

  // Anytime we are making forward progress and have a new RTT estimate, reset
  // the backoff counters.
  if (largest_observed_acked) {
    // Reset all retransmit counters any time a new packet is acked.
    consecutive_rto_count_ = 0;
    consecutive_tlp_count_ = 0;
    consecutive_crypto_retransmission_count_ = 0;
  }
}

void QuicSentPacketManager::MaybeInvokeCongestionEvent(
    bool rtt_updated, QuicByteCount bytes_in_flight) {
  if (rtt_updated || !packets_acked_.empty() ||
      !packets_lost_.empty()) {
    send_algorithm_->OnCongestionEvent(
        rtt_updated, bytes_in_flight, packets_acked_, packets_lost_);
    packets_acked_.clear();
    packets_lost_.clear();
  }
}

void QuicSentPacketManager::HandleAckForSentPackets(
    const ReceivedPacketInfo& received_info) {
  // Go through the packets we have not received an ack for and see if this
  // incoming_ack shows they've been seen by the peer.
  QuicTime::Delta delta_largest_observed =
      received_info.delta_time_largest_observed;
  QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
  while (it != unacked_packets_.end()) {
    QuicPacketSequenceNumber sequence_number = it->first;
    if (sequence_number > received_info.largest_observed) {
      // These packets are still in flight.
      break;
    }

    if (IsAwaitingPacket(received_info, sequence_number)) {
      // Consider it multiple nacks when there is a gap between the missing
      // packet and the largest observed, since the purpose of a nack
      // threshold is to tolerate re-ordering.  This handles both StretchAcks
      // and Forward Acks.
      // The nack count only increases when the largest observed increases.
      size_t min_nacks = received_info.largest_observed - sequence_number;
      // Truncated acks can nack the largest observed, so use a min of 1.
      if (min_nacks == 0) {
        min_nacks = 1;
      }
      unacked_packets_.NackPacket(sequence_number, min_nacks);
      ++it;
      continue;
    }
    // Packet was acked, so remove it from our unacked packet list.
    DVLOG(1) << ENDPOINT << "Got an ack for packet " << sequence_number;
    // If data is associated with the most recent transmission of this
    // packet, then inform the caller.
    if (it->second.in_flight) {
      packets_acked_[sequence_number] = it->second;
    }
    it = MarkPacketHandled(it, delta_largest_observed);
  }

  // Discard any retransmittable frames associated with revived packets.
  for (SequenceNumberSet::const_iterator revived_it =
           received_info.revived_packets.begin();
       revived_it != received_info.revived_packets.end(); ++revived_it) {
    MarkPacketRevived(*revived_it, delta_largest_observed);
  }
}

bool QuicSentPacketManager::HasRetransmittableFrames(
    QuicPacketSequenceNumber sequence_number) const {
  return unacked_packets_.HasRetransmittableFrames(sequence_number);
}

void QuicSentPacketManager::RetransmitUnackedPackets(
    RetransmissionType retransmission_type) {
  QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
  while (it != unacked_packets_.end()) {
    const RetransmittableFrames* frames = it->second.retransmittable_frames;
    // TODO(ianswett): Consider adding a new retransmission type which removes
    // all these old packets from unacked and retransmits them as new sequence
    // numbers with no connection to the previous ones.
    if (frames != NULL && (retransmission_type == ALL_PACKETS ||
                           frames->encryption_level() == ENCRYPTION_INITIAL)) {
      MarkForRetransmission(it->first, ALL_UNACKED_RETRANSMISSION);
    }
    ++it;
  }
}

void QuicSentPacketManager::NeuterUnencryptedPackets() {
  QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
  while (it != unacked_packets_.end()) {
    const RetransmittableFrames* frames = it->second.retransmittable_frames;
    QuicPacketSequenceNumber sequence_number = it->first;
    ++it;
    if (frames != NULL && frames->encryption_level() == ENCRYPTION_NONE) {
      // Once you're forward secure, no unencrypted packets will be sent, crypto
      // or otherwise. Unencrypted packets are neutered and abandoned, to ensure
      // they are not retransmitted or considered lost from a congestion control
      // perspective.
      pending_retransmissions_.erase(sequence_number);
      unacked_packets_.RemoveFromInFlight(sequence_number);
      // RemoveRetransmittibility is safe because only the newest sequence
      // number can have frames.
      unacked_packets_.RemoveRetransmittability(sequence_number);
    }
  }
}

void QuicSentPacketManager::MarkForRetransmission(
    QuicPacketSequenceNumber sequence_number,
    TransmissionType transmission_type) {
  const TransmissionInfo& transmission_info =
      unacked_packets_.GetTransmissionInfo(sequence_number);
  LOG_IF(DFATAL, transmission_info.retransmittable_frames == NULL);
  if (transmission_type != TLP_RETRANSMISSION) {
    unacked_packets_.RemoveFromInFlight(sequence_number);
  }
  // TODO(ianswett): Currently the RTO can fire while there are pending NACK
  // retransmissions for the same data, which is not ideal.
  if (ContainsKey(pending_retransmissions_, sequence_number)) {
    return;
  }

  pending_retransmissions_[sequence_number] = transmission_type;
}

void QuicSentPacketManager::RecordSpuriousRetransmissions(
    const SequenceNumberSet& all_transmissions,
    QuicPacketSequenceNumber acked_sequence_number) {
  if (acked_sequence_number < first_rto_transmission_) {
    // Cancel all pending RTO transmissions and restore their in flight status.
    // Replace SRTT with latest_rtt and increase the variance to prevent
    // a spurious RTO from happening again.
    rtt_stats_.ExpireSmoothedMetrics();
    for (PendingRetransmissionMap::const_iterator it =
             pending_retransmissions_.begin();
         it != pending_retransmissions_.end(); ++it) {
      DCHECK_EQ(it->second, RTO_RETRANSMISSION);
      unacked_packets_.RestoreInFlight(it->first);
    }
    pending_retransmissions_.clear();
    send_algorithm_->RevertRetransmissionTimeout();
    first_rto_transmission_ = 0;
    ++stats_->spurious_rto_count;
  }
  for (SequenceNumberSet::const_iterator
           it = all_transmissions.upper_bound(acked_sequence_number),
           end = all_transmissions.end();
       it != end;
       ++it) {
    const TransmissionInfo& retransmit_info =
        unacked_packets_.GetTransmissionInfo(*it);

    stats_->bytes_spuriously_retransmitted += retransmit_info.bytes_sent;
    ++stats_->packets_spuriously_retransmitted;
    if (debug_delegate_ != NULL) {
      debug_delegate_->OnSpuriousPacketRetransmition(
          retransmit_info.transmission_type,
          retransmit_info.bytes_sent);
    }
  }
}

bool QuicSentPacketManager::HasPendingRetransmissions() const {
  return !pending_retransmissions_.empty();
}

QuicSentPacketManager::PendingRetransmission
    QuicSentPacketManager::NextPendingRetransmission() {
  DCHECK(!pending_retransmissions_.empty());
  QuicPacketSequenceNumber sequence_number =
      pending_retransmissions_.begin()->first;
  TransmissionType transmission_type = pending_retransmissions_.begin()->second;
  if (unacked_packets_.HasPendingCryptoPackets()) {
    // Ensure crypto packets are retransmitted before other packets.
    PendingRetransmissionMap::const_iterator it =
        pending_retransmissions_.begin();
    do {
      if (HasCryptoHandshake(unacked_packets_.GetTransmissionInfo(it->first))) {
        sequence_number = it->first;
        transmission_type = it->second;
        break;
      }
      ++it;
    } while (it != pending_retransmissions_.end());
  }
  DCHECK(unacked_packets_.IsUnacked(sequence_number)) << sequence_number;
  const TransmissionInfo& transmission_info =
      unacked_packets_.GetTransmissionInfo(sequence_number);
  DCHECK(transmission_info.retransmittable_frames);

  return PendingRetransmission(sequence_number,
                               transmission_type,
                               *transmission_info.retransmittable_frames,
                               transmission_info.sequence_number_length);
}

void QuicSentPacketManager::MarkPacketRevived(
    QuicPacketSequenceNumber sequence_number,
    QuicTime::Delta delta_largest_observed) {
  if (!unacked_packets_.IsUnacked(sequence_number)) {
    return;
  }

  const TransmissionInfo& transmission_info =
      unacked_packets_.GetTransmissionInfo(sequence_number);
  QuicPacketSequenceNumber newest_transmission =
      *transmission_info.all_transmissions->rbegin();
  // This packet has been revived at the receiver. If we were going to
  // retransmit it, do not retransmit it anymore.
  pending_retransmissions_.erase(newest_transmission);

  // The AckNotifierManager needs to be notified for revived packets,
  // since it indicates the packet arrived from the appliction's perspective.
  if (transmission_info.retransmittable_frames) {
    ack_notifier_manager_.OnPacketAcked(
        newest_transmission, delta_largest_observed);
  }

  unacked_packets_.RemoveRetransmittability(sequence_number);
}

QuicUnackedPacketMap::const_iterator QuicSentPacketManager::MarkPacketHandled(
    QuicUnackedPacketMap::const_iterator it,
    QuicTime::Delta delta_largest_observed) {
  LOG_IF(DFATAL, it == unacked_packets_.end())
      << "MarkPacketHandled must be passed a valid iterator entry.";
  const QuicPacketSequenceNumber sequence_number = it->first;
  const TransmissionInfo& transmission_info = it->second;

  QuicPacketSequenceNumber newest_transmission =
      *transmission_info.all_transmissions->rbegin();
  // Remove the most recent packet, if it is pending retransmission.
  pending_retransmissions_.erase(newest_transmission);

  // Notify observers about the ACKed packet.
  {
    // The AckNotifierManager needs to be notified about the most recent
    // transmission, since that's the one only one it tracks.
    ack_notifier_manager_.OnPacketAcked(newest_transmission,
                                        delta_largest_observed);
    if (newest_transmission != sequence_number) {
      RecordSpuriousRetransmissions(*transmission_info.all_transmissions,
                                    sequence_number);
    }
  }

  // Two cases for MarkPacketHandled:
  // 1) Handle the most recent or a crypto packet, so remove all transmissions.
  // 2) Handle old transmission, keep all other pending transmissions,
  //    but disassociate them from one another.

  // If it's a crypto handshake packet, discard it and all retransmissions,
  // since they won't be acked now that one has been processed.
  // TODO(ianswett): Instead of handling all crypto packets in a special way,
  // only handle NULL encrypted packets in a special way.
  if (HasCryptoHandshake(
          unacked_packets_.GetTransmissionInfo(newest_transmission))) {
    unacked_packets_.RemoveFromInFlight(newest_transmission);
  }
  unacked_packets_.RemoveFromInFlight(sequence_number);
  unacked_packets_.RemoveRetransmittability(sequence_number);

  QuicUnackedPacketMap::const_iterator next_unacked = unacked_packets_.begin();
  while (next_unacked != unacked_packets_.end() &&
         next_unacked->first <= sequence_number) {
    ++next_unacked;
  }
  return next_unacked;
}

bool QuicSentPacketManager::IsUnacked(
    QuicPacketSequenceNumber sequence_number) const {
  return unacked_packets_.IsUnacked(sequence_number);
}

bool QuicSentPacketManager::HasUnackedPackets() const {
  return unacked_packets_.HasUnackedPackets();
}

QuicPacketSequenceNumber
QuicSentPacketManager::GetLeastUnackedSentPacket() const {
  return unacked_packets_.GetLeastUnackedSentPacket();
}

bool QuicSentPacketManager::OnPacketSent(
    QuicPacketSequenceNumber sequence_number,
    QuicTime sent_time,
    QuicByteCount bytes,
    TransmissionType transmission_type,
    HasRetransmittableData has_retransmittable_data) {
  DCHECK_LT(0u, sequence_number);
  LOG_IF(DFATAL, bytes == 0) << "Cannot send empty packets.";
  pending_tlp_transmission_ = false;
  // In rare circumstances, the packet could be serialized, sent, and then acked
  // before OnPacketSent is called.
  if (!unacked_packets_.IsUnacked(sequence_number)) {
    return false;
  }

  if (unacked_packets_.bytes_in_flight() == 0) {
    // TODO(ianswett): Consider being less aggressive to force a new
    // recent_min_rtt, likely by not discarding a relatively new sample.
    DVLOG(1) << "Sampling a new recent min rtt within 2 samples. currently:"
             << rtt_stats_.recent_min_rtt().ToMilliseconds() << "ms";
    rtt_stats_.SampleNewRecentMinRtt(kNumMinRttSamplesAfterQuiescence);
  }

  // Only track packets as in flight that the send algorithm wants us to track.
  const bool in_flight =
      send_algorithm_->OnPacketSent(sent_time,
                                    unacked_packets_.bytes_in_flight(),
                                    sequence_number,
                                    bytes,
                                    has_retransmittable_data);
  unacked_packets_.SetSent(sequence_number, sent_time, bytes, in_flight);

  // Reset the retransmission timer anytime a pending packet is sent.
  return in_flight;
}

void QuicSentPacketManager::OnRetransmissionTimeout() {
  DCHECK(unacked_packets_.HasInFlightPackets());
  DCHECK(!pending_tlp_transmission_);
  // Handshake retransmission, timer based loss detection, TLP, and RTO are
  // implemented with a single alarm. The handshake alarm is set when the
  // handshake has not completed, the loss alarm is set when the loss detection
  // algorithm says to, and the TLP and  RTO alarms are set after that.
  // The TLP alarm is always set to run for under an RTO.
  switch (GetRetransmissionMode()) {
    case HANDSHAKE_MODE:
      ++stats_->crypto_retransmit_count;
      RetransmitCryptoPackets();
      return;
    case LOSS_MODE: {
      ++stats_->loss_timeout_count;
      QuicByteCount bytes_in_flight = unacked_packets_.bytes_in_flight();
      InvokeLossDetection(clock_->Now());
      MaybeInvokeCongestionEvent(false, bytes_in_flight);
      return;
    }
    case TLP_MODE:
      // If no tail loss probe can be sent, because there are no retransmittable
      // packets, execute a conventional RTO to abandon old packets.
      ++stats_->tlp_count;
      ++consecutive_tlp_count_;
      pending_tlp_transmission_ = true;
      // TLPs prefer sending new data instead of retransmitting data, so
      // give the connection a chance to write before completing the TLP.
      return;
    case RTO_MODE:
      ++stats_->rto_count;
      RetransmitAllPackets();
      return;
  }
}

void QuicSentPacketManager::RetransmitCryptoPackets() {
  DCHECK_EQ(HANDSHAKE_MODE, GetRetransmissionMode());
  // TODO(ianswett): Typical TCP implementations only retransmit 5 times.
  consecutive_crypto_retransmission_count_ =
      min(kMaxHandshakeRetransmissionBackoffs,
          consecutive_crypto_retransmission_count_ + 1);
  bool packet_retransmitted = false;
  for (QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
       it != unacked_packets_.end(); ++it) {
    QuicPacketSequenceNumber sequence_number = it->first;
    const RetransmittableFrames* frames = it->second.retransmittable_frames;
    // Only retransmit frames which are in flight, and therefore have been sent.
    if (!it->second.in_flight || frames == NULL ||
        frames->HasCryptoHandshake() != IS_HANDSHAKE) {
      continue;
    }
    packet_retransmitted = true;
    MarkForRetransmission(sequence_number, HANDSHAKE_RETRANSMISSION);
  }
  DCHECK(packet_retransmitted) << "No crypto packets found to retransmit.";
}

bool QuicSentPacketManager::MaybeRetransmitTailLossProbe() {
  if (!pending_tlp_transmission_) {
    return false;
  }
  for (QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
       it != unacked_packets_.end(); ++it) {
    QuicPacketSequenceNumber sequence_number = it->first;
    const RetransmittableFrames* frames = it->second.retransmittable_frames;
    // Only retransmit frames which are in flight, and therefore have been sent.
    if (!it->second.in_flight || frames == NULL) {
      continue;
    }
    DCHECK_NE(IS_HANDSHAKE, frames->HasCryptoHandshake());
    MarkForRetransmission(sequence_number, TLP_RETRANSMISSION);
    return true;
  }
  DLOG(FATAL)
    << "No retransmittable packets, so RetransmitOldestPacket failed.";
  return false;
}

void QuicSentPacketManager::RetransmitAllPackets() {
  DVLOG(1) << "RetransmitAllPackets() called with "
           << unacked_packets_.GetNumUnackedPackets() << " unacked packets.";
  // Request retransmission of all retransmittable packets when the RTO
  // fires, and let the congestion manager decide how many to send
  // immediately and the remaining packets will be queued.
  // Abandon any non-retransmittable packets that are sufficiently old.
  bool packets_retransmitted = false;
  QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
  while (it != unacked_packets_.end()) {
    const RetransmittableFrames* frames = it->second.retransmittable_frames;
    QuicPacketSequenceNumber sequence_number = it->first;
    ++it;
    if (frames != NULL) {
      packets_retransmitted = true;
      MarkForRetransmission(sequence_number, RTO_RETRANSMISSION);
    } else {
      unacked_packets_.RemoveFromInFlight(sequence_number);
    }
  }

  send_algorithm_->OnRetransmissionTimeout(packets_retransmitted);
  if (packets_retransmitted) {
    if (consecutive_rto_count_ == 0) {
      first_rto_transmission_ = unacked_packets_.largest_sent_packet() + 1;
    }
    ++consecutive_rto_count_;
  }
}

QuicSentPacketManager::RetransmissionTimeoutMode
    QuicSentPacketManager::GetRetransmissionMode() const {
  DCHECK(unacked_packets_.HasInFlightPackets());
  if (unacked_packets_.HasPendingCryptoPackets()) {
    return HANDSHAKE_MODE;
  }
  if (loss_algorithm_->GetLossTimeout() != QuicTime::Zero()) {
    return LOSS_MODE;
  }
  if (consecutive_tlp_count_ < max_tail_loss_probes_) {
    if (unacked_packets_.HasUnackedRetransmittableFrames()) {
      return TLP_MODE;
    }
  }
  return RTO_MODE;
}

void QuicSentPacketManager::OnIncomingQuicCongestionFeedbackFrame(
    const QuicCongestionFeedbackFrame& frame,
    const QuicTime& feedback_receive_time) {
  send_algorithm_->OnIncomingQuicCongestionFeedbackFrame(
      frame, feedback_receive_time);
}

void QuicSentPacketManager::InvokeLossDetection(QuicTime time) {
  SequenceNumberSet lost_packets =
      loss_algorithm_->DetectLostPackets(unacked_packets_,
                                         time,
                                         largest_observed_,
                                         rtt_stats_);
  for (SequenceNumberSet::const_iterator it = lost_packets.begin();
       it != lost_packets.end(); ++it) {
    QuicPacketSequenceNumber sequence_number = *it;
    const TransmissionInfo& transmission_info =
        unacked_packets_.GetTransmissionInfo(sequence_number);
    // TODO(ianswett): If it's expected the FEC packet may repair the loss, it
    // should be recorded as a loss to the send algorithm, but not retransmitted
    // until it's known whether the FEC packet arrived.
    ++stats_->packets_lost;
    packets_lost_[sequence_number] = transmission_info;
    DVLOG(1) << ENDPOINT << "Lost packet " << sequence_number;

    if (transmission_info.retransmittable_frames != NULL) {
      MarkForRetransmission(sequence_number, LOSS_RETRANSMISSION);
    } else {
      // Since we will not retransmit this, we need to remove it from
      // unacked_packets_.   This is either the current transmission of
      // a packet whose previous transmission has been acked, a packet that has
      // been TLP retransmitted, or an FEC packet.
      unacked_packets_.RemoveFromInFlight(sequence_number);
    }
  }
}

bool QuicSentPacketManager::MaybeUpdateRTT(
    const ReceivedPacketInfo& received_info,
    const QuicTime& ack_receive_time) {
  if (!unacked_packets_.IsUnacked(received_info.largest_observed)) {
    return false;
  }
  // We calculate the RTT based on the highest ACKed sequence number, the lower
  // sequence numbers will include the ACK aggregation delay.
  const TransmissionInfo& transmission_info =
      unacked_packets_.GetTransmissionInfo(received_info.largest_observed);
  // Don't update the RTT if it hasn't been sent.
  if (transmission_info.sent_time == QuicTime::Zero()) {
    return false;
  }

  QuicTime::Delta send_delta =
      ack_receive_time.Subtract(transmission_info.sent_time);
  rtt_stats_.UpdateRtt(
      send_delta, received_info.delta_time_largest_observed, ack_receive_time);
  return true;
}

QuicTime::Delta QuicSentPacketManager::TimeUntilSend(
    QuicTime now,
    HasRetransmittableData retransmittable) {
  // The TLP logic is entirely contained within QuicSentPacketManager, so the
  // send algorithm does not need to be consulted.
  if (pending_tlp_transmission_) {
    return QuicTime::Delta::Zero();
  }
  return send_algorithm_->TimeUntilSend(
      now, unacked_packets_.bytes_in_flight(), retransmittable);
}

// Ensures that the Delayed Ack timer is always set to a value lesser
// than the retransmission timer's minimum value (MinRTO). We want the
// delayed ack to get back to the QUIC peer before the sender's
// retransmission timer triggers.  Since we do not know the
// reverse-path one-way delay, we assume equal delays for forward and
// reverse paths, and ensure that the timer is set to less than half
// of the MinRTO.
// There may be a value in making this delay adaptive with the help of
// the sender and a signaling mechanism -- if the sender uses a
// different MinRTO, we may get spurious retransmissions. May not have
// any benefits, but if the delayed ack becomes a significant source
// of (likely, tail) latency, then consider such a mechanism.
const QuicTime::Delta QuicSentPacketManager::DelayedAckTime() const {
  return QuicTime::Delta::FromMilliseconds(kMinRetransmissionTimeMs/2);
}

const QuicTime QuicSentPacketManager::GetRetransmissionTime() const {
  // Don't set the timer if there are no packets in flight or we've already
  // queued a tlp transmission and it hasn't been sent yet.
  if (!unacked_packets_.HasInFlightPackets() || pending_tlp_transmission_) {
    return QuicTime::Zero();
  }
  switch (GetRetransmissionMode()) {
    case HANDSHAKE_MODE:
      return clock_->ApproximateNow().Add(GetCryptoRetransmissionDelay());
    case LOSS_MODE:
      return loss_algorithm_->GetLossTimeout();
    case TLP_MODE: {
      // TODO(ianswett): When CWND is available, it would be preferable to
      // set the timer based on the earliest retransmittable packet.
      // Base the updated timer on the send time of the last packet.
      const QuicTime sent_time = unacked_packets_.GetLastPacketSentTime();
      const QuicTime tlp_time = sent_time.Add(GetTailLossProbeDelay());
      // Ensure the TLP timer never gets set to a time in the past.
      return QuicTime::Max(clock_->ApproximateNow(), tlp_time);
    }
    case RTO_MODE: {
      // The RTO is based on the first outstanding packet.
      const QuicTime sent_time =
          unacked_packets_.GetFirstInFlightPacketSentTime();
      QuicTime rto_time = sent_time.Add(GetRetransmissionDelay());
      // Wait for TLP packets to be acked before an RTO fires.
      QuicTime tlp_time =
          unacked_packets_.GetLastPacketSentTime().Add(GetTailLossProbeDelay());
      return QuicTime::Max(tlp_time, rto_time);
    }
  }
  DCHECK(false);
  return QuicTime::Zero();
}

const QuicTime::Delta QuicSentPacketManager::GetCryptoRetransmissionDelay()
    const {
  // This is equivalent to the TailLossProbeDelay, but slightly more aggressive
  // because crypto handshake messages don't incur a delayed ack time.
  int64 delay_ms = max<int64>(kMinHandshakeTimeoutMs,
                              1.5 * rtt_stats_.SmoothedRtt().ToMilliseconds());
  return QuicTime::Delta::FromMilliseconds(
      delay_ms << consecutive_crypto_retransmission_count_);
}

const QuicTime::Delta QuicSentPacketManager::GetTailLossProbeDelay() const {
  QuicTime::Delta srtt = rtt_stats_.SmoothedRtt();
  if (!unacked_packets_.HasMultipleInFlightPackets()) {
    return QuicTime::Delta::Max(
        srtt.Multiply(1.5).Add(DelayedAckTime()), srtt.Multiply(2));
  }
  return QuicTime::Delta::FromMilliseconds(
      max(kMinTailLossProbeTimeoutMs,
          static_cast<int64>(2 * srtt.ToMilliseconds())));
}

const QuicTime::Delta QuicSentPacketManager::GetRetransmissionDelay() const {
  QuicTime::Delta retransmission_delay = send_algorithm_->RetransmissionDelay();
  // TODO(rch): This code should move to |send_algorithm_|.
  if (retransmission_delay.IsZero()) {
    // We are in the initial state, use default timeout values.
    retransmission_delay =
        QuicTime::Delta::FromMilliseconds(kDefaultRetransmissionTimeMs);
  } else if (retransmission_delay.ToMilliseconds() < kMinRetransmissionTimeMs) {
    retransmission_delay =
        QuicTime::Delta::FromMilliseconds(kMinRetransmissionTimeMs);
  }

  // Calculate exponential back off.
  retransmission_delay = retransmission_delay.Multiply(
      1 << min<size_t>(consecutive_rto_count_, kMaxRetransmissions));

  if (retransmission_delay.ToMilliseconds() > kMaxRetransmissionTimeMs) {
    return QuicTime::Delta::FromMilliseconds(kMaxRetransmissionTimeMs);
  }
  return retransmission_delay;
}

const RttStats* QuicSentPacketManager::GetRttStats() const {
  return &rtt_stats_;
}

QuicBandwidth QuicSentPacketManager::BandwidthEstimate() const {
  return send_algorithm_->BandwidthEstimate();
}

QuicByteCount QuicSentPacketManager::GetCongestionWindow() const {
  return send_algorithm_->GetCongestionWindow();
}

void QuicSentPacketManager::MaybeEnablePacing() {
  if (!FLAGS_enable_quic_pacing) {
    return;
  }

  if (using_pacing_) {
    return;
  }

  // Set up a pacing sender with a 5 millisecond alarm granularity.
  using_pacing_ = true;
  send_algorithm_.reset(
      new PacingSender(send_algorithm_.release(),
                       QuicTime::Delta::FromMilliseconds(5)));
}

}  // namespace net