blob: 6346ae008165b2158163eed8e90f987e01da3c7c [file] [log] [blame]
// Copyright (c) 2012 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/third_party/quiche/src/quic/core/quic_connection.h"
#include <string.h>
#include <sys/types.h>
#include <algorithm>
#include <iterator>
#include <limits>
#include <memory>
#include <set>
#include <utility>
#include "base/macros.h"
#include "net/third_party/quiche/src/quic/core/crypto/crypto_protocol.h"
#include "net/third_party/quiche/src/quic/core/crypto/quic_decrypter.h"
#include "net/third_party/quiche/src/quic/core/crypto/quic_encrypter.h"
#include "net/third_party/quiche/src/quic/core/proto/cached_network_parameters.proto.h"
#include "net/third_party/quiche/src/quic/core/quic_bandwidth.h"
#include "net/third_party/quiche/src/quic/core/quic_config.h"
#include "net/third_party/quiche/src/quic/core/quic_packet_generator.h"
#include "net/third_party/quiche/src/quic/core/quic_pending_retransmission.h"
#include "net/third_party/quiche/src/quic/core/quic_types.h"
#include "net/third_party/quiche/src/quic/core/quic_utils.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_bug_tracker.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_error_code_wrappers.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_exported_stats.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_flag_utils.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_flags.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_logging.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_map_util.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_str_cat.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_string.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_string_utils.h"
#include "net/third_party/quiche/src/quic/platform/api/quic_text_utils.h"
namespace quic {
class QuicDecrypter;
class QuicEncrypter;
namespace {
// Maximum number of consecutive sent nonretransmittable packets.
const QuicPacketCount kMaxConsecutiveNonRetransmittablePackets = 19;
// Maximum number of retransmittable packets received before sending an ack.
const QuicPacketCount kDefaultRetransmittablePacketsBeforeAck = 2;
// Minimum number of packets received before ack decimation is enabled.
// This intends to avoid the beginning of slow start, when CWNDs may be
// rapidly increasing.
const QuicPacketCount kMinReceivedBeforeAckDecimation = 100;
// Wait for up to 10 retransmittable packets before sending an ack.
const QuicPacketCount kMaxRetransmittablePacketsBeforeAck = 10;
// One quarter RTT delay when doing ack decimation.
const float kAckDecimationDelay = 0.25;
// One eighth RTT delay when doing ack decimation.
const float kShortAckDecimationDelay = 0.125;
// The minimum release time into future in ms.
const int kMinReleaseTimeIntoFutureMs = 1;
bool Near(QuicPacketNumber a, QuicPacketNumber b) {
QuicPacketCount delta = (a > b) ? a - b : b - a;
return delta <= kMaxPacketGap;
}
// An alarm that is scheduled to send an ack if a timeout occurs.
class AckAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit AckAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
AckAlarmDelegate(const AckAlarmDelegate&) = delete;
AckAlarmDelegate& operator=(const AckAlarmDelegate&) = delete;
void OnAlarm() override {
DCHECK(connection_->ack_frame_updated());
QuicConnection::ScopedPacketFlusher flusher(connection_,
QuicConnection::SEND_ACK);
}
private:
QuicConnection* connection_;
};
// This alarm will be scheduled any time a data-bearing packet is sent out.
// When the alarm goes off, the connection checks to see if the oldest packets
// have been acked, and retransmit them if they have not.
class RetransmissionAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit RetransmissionAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
RetransmissionAlarmDelegate(const RetransmissionAlarmDelegate&) = delete;
RetransmissionAlarmDelegate& operator=(const RetransmissionAlarmDelegate&) =
delete;
void OnAlarm() override { connection_->OnRetransmissionTimeout(); }
private:
QuicConnection* connection_;
};
// An alarm that is scheduled when the SentPacketManager requires a delay
// before sending packets and fires when the packet may be sent.
class SendAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit SendAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
SendAlarmDelegate(const SendAlarmDelegate&) = delete;
SendAlarmDelegate& operator=(const SendAlarmDelegate&) = delete;
void OnAlarm() override { connection_->WriteAndBundleAcksIfNotBlocked(); }
private:
QuicConnection* connection_;
};
class PathDegradingAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit PathDegradingAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
PathDegradingAlarmDelegate(const PathDegradingAlarmDelegate&) = delete;
PathDegradingAlarmDelegate& operator=(const PathDegradingAlarmDelegate&) =
delete;
void OnAlarm() override { connection_->OnPathDegradingTimeout(); }
private:
QuicConnection* connection_;
};
class TimeoutAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit TimeoutAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
TimeoutAlarmDelegate(const TimeoutAlarmDelegate&) = delete;
TimeoutAlarmDelegate& operator=(const TimeoutAlarmDelegate&) = delete;
void OnAlarm() override { connection_->CheckForTimeout(); }
private:
QuicConnection* connection_;
};
class PingAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit PingAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
PingAlarmDelegate(const PingAlarmDelegate&) = delete;
PingAlarmDelegate& operator=(const PingAlarmDelegate&) = delete;
void OnAlarm() override { connection_->OnPingTimeout(); }
private:
QuicConnection* connection_;
};
class MtuDiscoveryAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit MtuDiscoveryAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
MtuDiscoveryAlarmDelegate(const MtuDiscoveryAlarmDelegate&) = delete;
MtuDiscoveryAlarmDelegate& operator=(const MtuDiscoveryAlarmDelegate&) =
delete;
void OnAlarm() override { connection_->DiscoverMtu(); }
private:
QuicConnection* connection_;
};
class RetransmittableOnWireAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit RetransmittableOnWireAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
RetransmittableOnWireAlarmDelegate(
const RetransmittableOnWireAlarmDelegate&) = delete;
RetransmittableOnWireAlarmDelegate& operator=(
const RetransmittableOnWireAlarmDelegate&) = delete;
void OnAlarm() override { connection_->OnPingTimeout(); }
private:
QuicConnection* connection_;
};
class ProcessUndecryptablePacketsAlarmDelegate : public QuicAlarm::Delegate {
public:
explicit ProcessUndecryptablePacketsAlarmDelegate(QuicConnection* connection)
: connection_(connection) {}
ProcessUndecryptablePacketsAlarmDelegate(
const ProcessUndecryptablePacketsAlarmDelegate&) = delete;
ProcessUndecryptablePacketsAlarmDelegate& operator=(
const ProcessUndecryptablePacketsAlarmDelegate&) = delete;
void OnAlarm() override { connection_->MaybeProcessUndecryptablePackets(); }
private:
QuicConnection* connection_;
};
} // namespace
#define ENDPOINT \
(perspective_ == Perspective::IS_SERVER ? "Server: " : "Client: ")
QuicConnection::QuicConnection(
QuicConnectionId connection_id,
QuicSocketAddress initial_peer_address,
QuicConnectionHelperInterface* helper,
QuicAlarmFactory* alarm_factory,
QuicPacketWriter* writer,
bool owns_writer,
Perspective perspective,
const ParsedQuicVersionVector& supported_versions)
: framer_(supported_versions,
helper->GetClock()->ApproximateNow(),
perspective),
current_packet_content_(NO_FRAMES_RECEIVED),
is_current_packet_connectivity_probing_(false),
current_effective_peer_migration_type_(NO_CHANGE),
helper_(helper),
alarm_factory_(alarm_factory),
per_packet_options_(nullptr),
writer_(writer),
owns_writer_(owns_writer),
encryption_level_(ENCRYPTION_NONE),
clock_(helper->GetClock()),
random_generator_(helper->GetRandomGenerator()),
connection_id_(connection_id),
peer_address_(initial_peer_address),
direct_peer_address_(initial_peer_address),
active_effective_peer_migration_type_(NO_CHANGE),
last_packet_decrypted_(false),
last_size_(0),
current_packet_data_(nullptr),
last_decrypted_packet_level_(ENCRYPTION_NONE),
should_last_packet_instigate_acks_(false),
was_last_packet_missing_(false),
max_undecryptable_packets_(0),
max_tracked_packets_(kMaxTrackedPackets),
pending_version_negotiation_packet_(false),
send_ietf_version_negotiation_packet_(false),
save_crypto_packets_as_termination_packets_(false),
idle_timeout_connection_close_behavior_(
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET),
close_connection_after_five_rtos_(false),
received_packet_manager_(&stats_),
ack_queued_(false),
num_retransmittable_packets_received_since_last_ack_sent_(0),
num_packets_received_since_last_ack_sent_(0),
stop_waiting_count_(0),
ack_mode_(GetQuicReloadableFlag(quic_enable_ack_decimation)
? ACK_DECIMATION
: TCP_ACKING),
ack_decimation_delay_(kAckDecimationDelay),
unlimited_ack_decimation_(false),
fast_ack_after_quiescence_(false),
pending_retransmission_alarm_(false),
defer_send_in_response_to_packets_(false),
ping_timeout_(QuicTime::Delta::FromSeconds(kPingTimeoutSecs)),
retransmittable_on_wire_timeout_(QuicTime::Delta::Infinite()),
arena_(),
ack_alarm_(alarm_factory_->CreateAlarm(arena_.New<AckAlarmDelegate>(this),
&arena_)),
retransmission_alarm_(alarm_factory_->CreateAlarm(
arena_.New<RetransmissionAlarmDelegate>(this),
&arena_)),
send_alarm_(
alarm_factory_->CreateAlarm(arena_.New<SendAlarmDelegate>(this),
&arena_)),
timeout_alarm_(
alarm_factory_->CreateAlarm(arena_.New<TimeoutAlarmDelegate>(this),
&arena_)),
ping_alarm_(
alarm_factory_->CreateAlarm(arena_.New<PingAlarmDelegate>(this),
&arena_)),
mtu_discovery_alarm_(alarm_factory_->CreateAlarm(
arena_.New<MtuDiscoveryAlarmDelegate>(this),
&arena_)),
path_degrading_alarm_(alarm_factory_->CreateAlarm(
arena_.New<PathDegradingAlarmDelegate>(this),
&arena_)),
process_undecryptable_packets_alarm_(alarm_factory_->CreateAlarm(
arena_.New<ProcessUndecryptablePacketsAlarmDelegate>(this),
&arena_)),
visitor_(nullptr),
debug_visitor_(nullptr),
packet_generator_(connection_id_, &framer_, random_generator_, this),
idle_network_timeout_(QuicTime::Delta::Infinite()),
handshake_timeout_(QuicTime::Delta::Infinite()),
time_of_first_packet_sent_after_receiving_(
GetQuicReloadableFlag(
quic_fix_time_of_first_packet_sent_after_receiving)
? QuicTime::Zero()
: clock_->ApproximateNow()),
time_of_last_received_packet_(clock_->ApproximateNow()),
time_of_previous_received_packet_(QuicTime::Zero()),
sent_packet_manager_(
perspective,
clock_,
&stats_,
GetQuicReloadableFlag(quic_default_to_bbr) ? kBBR : kCubicBytes,
kNack),
version_negotiation_state_(START_NEGOTIATION),
perspective_(perspective),
connected_(true),
can_truncate_connection_ids_(perspective == Perspective::IS_SERVER),
mtu_discovery_target_(0),
mtu_probe_count_(0),
packets_between_mtu_probes_(kPacketsBetweenMtuProbesBase),
next_mtu_probe_at_(kPacketsBetweenMtuProbesBase),
largest_received_packet_size_(0),
write_error_occurred_(false),
no_stop_waiting_frames_(transport_version() > QUIC_VERSION_43),
consecutive_num_packets_with_no_retransmittable_frames_(0),
max_consecutive_num_packets_with_no_retransmittable_frames_(
kMaxConsecutiveNonRetransmittablePackets),
min_received_before_ack_decimation_(kMinReceivedBeforeAckDecimation),
ack_frequency_before_ack_decimation_(
kDefaultRetransmittablePacketsBeforeAck),
fill_up_link_during_probing_(false),
probing_retransmission_pending_(false),
stateless_reset_token_received_(false),
received_stateless_reset_token_(0),
last_control_frame_id_(kInvalidControlFrameId),
is_path_degrading_(false),
processing_ack_frame_(false),
supports_release_time_(false),
release_time_into_future_(QuicTime::Delta::Zero()),
no_version_negotiation_(supported_versions.size() == 1) {
if (ack_mode_ == ACK_DECIMATION) {
QUIC_RELOADABLE_FLAG_COUNT(quic_enable_ack_decimation);
}
if (perspective_ == Perspective::IS_SERVER &&
supported_versions.size() == 1) {
QUIC_RESTART_FLAG_COUNT(quic_no_server_conn_ver_negotiation2);
}
QUIC_DLOG(INFO) << ENDPOINT
<< "Created connection with connection_id: " << connection_id
<< " and version: "
<< QuicVersionToString(transport_version());
QUIC_BUG_IF(!QuicUtils::IsConnectionIdValidForVersion(connection_id,
transport_version()))
<< "QuicConnection: attempted to use connection ID " << connection_id
<< " which is invalid with version "
<< QuicVersionToString(transport_version());
framer_.set_visitor(this);
stats_.connection_creation_time = clock_->ApproximateNow();
// TODO(ianswett): Supply the NetworkChangeVisitor as a constructor argument
// and make it required non-null, because it's always used.
sent_packet_manager_.SetNetworkChangeVisitor(this);
if (GetQuicRestartFlag(quic_offload_pacing_to_usps2)) {
sent_packet_manager_.SetPacingAlarmGranularity(QuicTime::Delta::Zero());
release_time_into_future_ =
QuicTime::Delta::FromMilliseconds(kMinReleaseTimeIntoFutureMs);
}
// Allow the packet writer to potentially reduce the packet size to a value
// even smaller than kDefaultMaxPacketSize.
SetMaxPacketLength(perspective_ == Perspective::IS_SERVER
? kDefaultServerMaxPacketSize
: kDefaultMaxPacketSize);
received_packet_manager_.set_max_ack_ranges(255);
MaybeEnableSessionDecidesWhatToWrite();
DCHECK(!GetQuicRestartFlag(quic_no_server_conn_ver_negotiation2) ||
perspective_ == Perspective::IS_CLIENT ||
supported_versions.size() == 1);
}
QuicConnection::~QuicConnection() {
if (owns_writer_) {
delete writer_;
}
ClearQueuedPackets();
}
void QuicConnection::ClearQueuedPackets() {
for (auto it = queued_packets_.begin(); it != queued_packets_.end(); ++it) {
// Delete the buffer before calling ClearSerializedPacket, which sets
// encrypted_buffer to nullptr.
delete[] it->encrypted_buffer;
ClearSerializedPacket(&(*it));
}
queued_packets_.clear();
}
void QuicConnection::SetFromConfig(const QuicConfig& config) {
if (config.negotiated()) {
// Handshake complete, set handshake timeout to Infinite.
SetNetworkTimeouts(QuicTime::Delta::Infinite(),
config.IdleNetworkTimeout());
if (config.SilentClose()) {
idle_timeout_connection_close_behavior_ =
ConnectionCloseBehavior::SILENT_CLOSE;
}
} else {
SetNetworkTimeouts(config.max_time_before_crypto_handshake(),
config.max_idle_time_before_crypto_handshake());
}
sent_packet_manager_.SetFromConfig(config);
if (config.HasReceivedBytesForConnectionId() &&
can_truncate_connection_ids_) {
packet_generator_.SetConnectionIdLength(
config.ReceivedBytesForConnectionId());
}
max_undecryptable_packets_ = config.max_undecryptable_packets();
if (config.HasClientSentConnectionOption(kMTUH, perspective_)) {
SetMtuDiscoveryTarget(kMtuDiscoveryTargetPacketSizeHigh);
}
if (config.HasClientSentConnectionOption(kMTUL, perspective_)) {
SetMtuDiscoveryTarget(kMtuDiscoveryTargetPacketSizeLow);
}
if (debug_visitor_ != nullptr) {
debug_visitor_->OnSetFromConfig(config);
}
if (GetQuicReloadableFlag(quic_enable_ack_decimation) &&
config.HasClientSentConnectionOption(kACD0, perspective_)) {
ack_mode_ = TCP_ACKING;
}
if (config.HasClientSentConnectionOption(kACKD, perspective_)) {
ack_mode_ = ACK_DECIMATION;
}
if ((!GetQuicReloadableFlag(quic_enable_ack_decimation) ||
GetQuicReloadableFlag(quic_keep_ack_decimation_reordering)) &&
config.HasClientSentConnectionOption(kAKD2, perspective_)) {
if (GetQuicReloadableFlag(quic_keep_ack_decimation_reordering)) {
QUIC_RELOADABLE_FLAG_COUNT_N(quic_keep_ack_decimation_reordering, 1, 2);
}
ack_mode_ = ACK_DECIMATION_WITH_REORDERING;
}
if (config.HasClientSentConnectionOption(kAKD3, perspective_)) {
ack_mode_ = ACK_DECIMATION;
ack_decimation_delay_ = kShortAckDecimationDelay;
}
if ((!GetQuicReloadableFlag(quic_enable_ack_decimation) ||
GetQuicReloadableFlag(quic_keep_ack_decimation_reordering)) &&
config.HasClientSentConnectionOption(kAKD4, perspective_)) {
if (GetQuicReloadableFlag(quic_keep_ack_decimation_reordering)) {
QUIC_RELOADABLE_FLAG_COUNT_N(quic_keep_ack_decimation_reordering, 2, 2);
}
ack_mode_ = ACK_DECIMATION_WITH_REORDERING;
ack_decimation_delay_ = kShortAckDecimationDelay;
}
if (config.HasClientSentConnectionOption(kAKDU, perspective_)) {
unlimited_ack_decimation_ = true;
}
if (config.HasClientSentConnectionOption(kACKQ, perspective_)) {
fast_ack_after_quiescence_ = true;
}
if (config.HasClientSentConnectionOption(k5RTO, perspective_)) {
close_connection_after_five_rtos_ = true;
}
if (config.HasClientSentConnectionOption(kNSTP, perspective_)) {
no_stop_waiting_frames_ = true;
}
if (config.HasReceivedStatelessResetToken()) {
stateless_reset_token_received_ = true;
received_stateless_reset_token_ = config.ReceivedStatelessResetToken();
}
if (GetQuicReloadableFlag(quic_send_timestamps) &&
config.HasClientSentConnectionOption(kSTMP, perspective_)) {
QUIC_RELOADABLE_FLAG_COUNT(quic_send_timestamps);
framer_.set_process_timestamps(true);
received_packet_manager_.set_save_timestamps(true);
}
supports_release_time_ =
writer_ != nullptr && writer_->SupportsReleaseTime() &&
!config.HasClientSentConnectionOption(kNPCO, perspective_);
if (supports_release_time_) {
UpdateReleaseTimeIntoFuture();
}
}
void QuicConnection::OnSendConnectionState(
const CachedNetworkParameters& cached_network_params) {
if (debug_visitor_ != nullptr) {
debug_visitor_->OnSendConnectionState(cached_network_params);
}
}
void QuicConnection::OnReceiveConnectionState(
const CachedNetworkParameters& cached_network_params) {
if (debug_visitor_ != nullptr) {
debug_visitor_->OnReceiveConnectionState(cached_network_params);
}
}
void QuicConnection::ResumeConnectionState(
const CachedNetworkParameters& cached_network_params,
bool max_bandwidth_resumption) {
sent_packet_manager_.ResumeConnectionState(cached_network_params,
max_bandwidth_resumption);
}
void QuicConnection::SetMaxPacingRate(QuicBandwidth max_pacing_rate) {
sent_packet_manager_.SetMaxPacingRate(max_pacing_rate);
}
void QuicConnection::AdjustNetworkParameters(QuicBandwidth bandwidth,
QuicTime::Delta rtt) {
sent_packet_manager_.AdjustNetworkParameters(bandwidth, rtt);
}
QuicBandwidth QuicConnection::MaxPacingRate() const {
return sent_packet_manager_.MaxPacingRate();
}
void QuicConnection::SetNumOpenStreams(size_t num_streams) {
sent_packet_manager_.SetNumOpenStreams(num_streams);
}
bool QuicConnection::SelectMutualVersion(
const ParsedQuicVersionVector& available_versions) {
// Try to find the highest mutual version by iterating over supported
// versions, starting with the highest, and breaking out of the loop once we
// find a matching version in the provided available_versions vector.
const ParsedQuicVersionVector& supported_versions =
framer_.supported_versions();
for (size_t i = 0; i < supported_versions.size(); ++i) {
const ParsedQuicVersion& version = supported_versions[i];
if (QuicContainsValue(available_versions, version)) {
framer_.set_version(version);
return true;
}
}
return false;
}
void QuicConnection::OnError(QuicFramer* framer) {
// Packets that we can not or have not decrypted are dropped.
// TODO(rch): add stats to measure this.
if (!connected_ || last_packet_decrypted_ == false) {
return;
}
CloseConnection(framer->error(), framer->detailed_error(),
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
}
void QuicConnection::OnPacket() {
last_packet_decrypted_ = false;
}
void QuicConnection::OnPublicResetPacket(const QuicPublicResetPacket& packet) {
// Check that any public reset packet with a different connection ID that was
// routed to this QuicConnection has been redirected before control reaches
// here. (Check for a bug regression.)
DCHECK_EQ(connection_id_, packet.connection_id);
DCHECK_EQ(perspective_, Perspective::IS_CLIENT);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPublicResetPacket(packet);
}
QuicString error_details = "Received public reset.";
if (perspective_ == Perspective::IS_CLIENT && !packet.endpoint_id.empty()) {
QuicStrAppend(&error_details, " From ", packet.endpoint_id, ".");
}
QUIC_DLOG(INFO) << ENDPOINT << error_details;
QUIC_CODE_COUNT(quic_tear_down_local_connection_on_public_reset);
TearDownLocalConnectionState(QUIC_PUBLIC_RESET, error_details,
ConnectionCloseSource::FROM_PEER);
}
bool QuicConnection::OnProtocolVersionMismatch(
ParsedQuicVersion received_version,
PacketHeaderFormat form) {
QUIC_DLOG(INFO) << ENDPOINT << "Received packet with mismatched version "
<< ParsedQuicVersionToString(received_version);
if (perspective_ == Perspective::IS_CLIENT) {
const QuicString error_details = "Protocol version mismatch.";
QUIC_BUG << ENDPOINT << error_details;
TearDownLocalConnectionState(QUIC_INTERNAL_ERROR, error_details,
ConnectionCloseSource::FROM_SELF);
return false;
}
if (no_version_negotiation_) {
// Drop old packets that were sent by the client before the version was
// negotiated.
return false;
}
DCHECK_NE(version(), received_version);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnProtocolVersionMismatch(received_version);
}
switch (version_negotiation_state_) {
case START_NEGOTIATION:
if (!framer_.IsSupportedVersion(received_version)) {
SendVersionNegotiationPacket(form != GOOGLE_QUIC_PACKET);
version_negotiation_state_ = NEGOTIATION_IN_PROGRESS;
return false;
}
break;
case NEGOTIATION_IN_PROGRESS:
if (!framer_.IsSupportedVersion(received_version)) {
SendVersionNegotiationPacket(form != GOOGLE_QUIC_PACKET);
return false;
}
break;
case NEGOTIATED_VERSION:
// Might be old packets that were sent by the client before the version
// was negotiated. Drop these.
return false;
default:
DCHECK(false);
}
// Store the new version.
framer_.set_version(received_version);
framer_.InferPacketHeaderTypeFromVersion();
version_negotiation_state_ = NEGOTIATED_VERSION;
visitor_->OnSuccessfulVersionNegotiation(received_version);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnSuccessfulVersionNegotiation(received_version);
}
QUIC_DLOG(INFO) << ENDPOINT << "version negotiated "
<< ParsedQuicVersionToString(received_version);
MaybeEnableSessionDecidesWhatToWrite();
no_stop_waiting_frames_ =
received_version.transport_version > QUIC_VERSION_43;
// TODO(satyamshekhar): Store the packet number of this packet and close the
// connection if we ever received a packet with incorrect version and whose
// packet number is greater.
return true;
}
// Handles version negotiation for client connection.
void QuicConnection::OnVersionNegotiationPacket(
const QuicVersionNegotiationPacket& packet) {
// Check that any public reset packet with a different connection ID that was
// routed to this QuicConnection has been redirected before control reaches
// here. (Check for a bug regression.)
DCHECK_EQ(connection_id_, packet.connection_id);
if (perspective_ == Perspective::IS_SERVER) {
const QuicString error_details =
"Server receieved version negotiation packet.";
QUIC_BUG << error_details;
QUIC_CODE_COUNT(quic_tear_down_local_connection_on_version_negotiation);
TearDownLocalConnectionState(QUIC_INTERNAL_ERROR, error_details,
ConnectionCloseSource::FROM_SELF);
return;
}
if (debug_visitor_ != nullptr) {
debug_visitor_->OnVersionNegotiationPacket(packet);
}
if (version_negotiation_state_ != START_NEGOTIATION) {
// Possibly a duplicate version negotiation packet.
return;
}
if (QuicContainsValue(packet.versions, version())) {
const QuicString error_details =
"Server already supports client's version and should have accepted the "
"connection.";
QUIC_DLOG(WARNING) << error_details;
TearDownLocalConnectionState(QUIC_INVALID_VERSION_NEGOTIATION_PACKET,
error_details,
ConnectionCloseSource::FROM_SELF);
return;
}
server_supported_versions_ = packet.versions;
if (GetQuicReloadableFlag(quic_no_client_conn_ver_negotiation)) {
QUIC_RELOADABLE_FLAG_COUNT(quic_no_client_conn_ver_negotiation);
CloseConnection(
QUIC_INVALID_VERSION,
QuicStrCat(
"Client may support one of the versions in the server's list, but "
"it's going to close the connection anyway. Supported versions: {",
ParsedQuicVersionVectorToString(framer_.supported_versions()),
"}, peer supported versions: {",
ParsedQuicVersionVectorToString(packet.versions), "}"),
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return;
}
if (!SelectMutualVersion(packet.versions)) {
CloseConnection(
QUIC_INVALID_VERSION,
QuicStrCat(
"No common version found. Supported versions: {",
ParsedQuicVersionVectorToString(framer_.supported_versions()),
"}, peer supported versions: {",
ParsedQuicVersionVectorToString(packet.versions), "}"),
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return;
}
QUIC_DLOG(INFO) << ENDPOINT << "Negotiated version: "
<< QuicVersionToString(transport_version());
no_stop_waiting_frames_ = transport_version() > QUIC_VERSION_43;
version_negotiation_state_ = NEGOTIATION_IN_PROGRESS;
RetransmitUnackedPackets(ALL_UNACKED_RETRANSMISSION);
}
bool QuicConnection::OnUnauthenticatedPublicHeader(
const QuicPacketHeader& header) {
if (header.destination_connection_id == connection_id_) {
return true;
}
++stats_.packets_dropped;
QUIC_DLOG(INFO) << ENDPOINT
<< "Ignoring packet from unexpected ConnectionId: "
<< header.destination_connection_id << " instead of "
<< connection_id_;
if (debug_visitor_ != nullptr) {
debug_visitor_->OnIncorrectConnectionId(header.destination_connection_id);
}
// If this is a server, the dispatcher routes each packet to the
// QuicConnection responsible for the packet's connection ID. So if control
// arrives here and this is a server, the dispatcher must be malfunctioning.
DCHECK_NE(Perspective::IS_SERVER, perspective_);
return false;
}
bool QuicConnection::OnUnauthenticatedHeader(const QuicPacketHeader& header) {
if (debug_visitor_ != nullptr) {
debug_visitor_->OnUnauthenticatedHeader(header);
}
// Check that any public reset packet with a different connection ID that was
// routed to this QuicConnection has been redirected before control reaches
// here.
DCHECK_EQ(connection_id_, header.destination_connection_id);
if (!packet_generator_.IsPendingPacketEmpty()) {
// Incoming packets may change a queued ACK frame.
const QuicString error_details =
"Pending frames must be serialized before incoming packets are "
"processed.";
QUIC_BUG << error_details << ", received header: " << header;
CloseConnection(QUIC_INTERNAL_ERROR, error_details,
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
// If this packet has already been seen, or the sender has told us that it
// will not be retransmitted, then stop processing the packet.
if (!received_packet_manager_.IsAwaitingPacket(header.packet_number)) {
if (framer_.IsIetfStatelessResetPacket(header)) {
QuicIetfStatelessResetPacket packet(
header, header.possible_stateless_reset_token);
OnAuthenticatedIetfStatelessResetPacket(packet);
return false;
}
QUIC_DLOG(INFO) << ENDPOINT << "Packet " << header.packet_number
<< " no longer being waited for. Discarding.";
if (debug_visitor_ != nullptr) {
debug_visitor_->OnDuplicatePacket(header.packet_number);
}
++stats_.packets_dropped;
return false;
}
if (version_negotiation_state_ != NEGOTIATED_VERSION &&
perspective_ == Perspective::IS_SERVER) {
if (!header.version_flag) {
// Packets should have the version flag till version negotiation is
// done.
QuicString error_details =
QuicStrCat(ENDPOINT, "Packet ", header.packet_number.ToUint64(),
" without version flag before version negotiated.");
QUIC_DLOG(WARNING) << error_details;
CloseConnection(QUIC_INVALID_VERSION, error_details,
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
} else {
DCHECK_EQ(header.version, version());
version_negotiation_state_ = NEGOTIATED_VERSION;
framer_.InferPacketHeaderTypeFromVersion();
visitor_->OnSuccessfulVersionNegotiation(version());
if (debug_visitor_ != nullptr) {
debug_visitor_->OnSuccessfulVersionNegotiation(version());
}
}
DCHECK_EQ(NEGOTIATED_VERSION, version_negotiation_state_);
}
return true;
}
void QuicConnection::OnDecryptedPacket(EncryptionLevel level) {
last_decrypted_packet_level_ = level;
last_packet_decrypted_ = true;
// Once the server receives a forward secure packet, the handshake is
// confirmed.
if (level == ENCRYPTION_FORWARD_SECURE &&
perspective_ == Perspective::IS_SERVER) {
sent_packet_manager_.SetHandshakeConfirmed();
if (sent_packet_manager_.unacked_packets().use_uber_loss_algorithm()) {
// This may have changed the retransmission timer, so re-arm it.
SetRetransmissionAlarm();
}
}
}
QuicSocketAddress QuicConnection::GetEffectivePeerAddressFromCurrentPacket()
const {
// By default, the connection is not proxied, and the effective peer address
// is the packet's source address, i.e. the direct peer address.
return last_packet_source_address_;
}
bool QuicConnection::OnPacketHeader(const QuicPacketHeader& header) {
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPacketHeader(header);
}
// Will be decremented below if we fall through to return true.
++stats_.packets_dropped;
if (!ProcessValidatedPacket(header)) {
return false;
}
// Initialize the current packet content state.
current_packet_content_ = NO_FRAMES_RECEIVED;
is_current_packet_connectivity_probing_ = false;
current_effective_peer_migration_type_ = NO_CHANGE;
if (perspective_ == Perspective::IS_CLIENT) {
if (!received_packet_manager_.GetLargestObserved().IsInitialized() ||
header.packet_number > received_packet_manager_.GetLargestObserved()) {
// Update peer_address_ and effective_peer_address_ immediately for
// client connections.
direct_peer_address_ = last_packet_source_address_;
effective_peer_address_ = GetEffectivePeerAddressFromCurrentPacket();
}
} else {
// At server, remember the address change type of effective_peer_address
// in current_effective_peer_migration_type_. But this variable alone
// doesn't necessarily starts a migration. A migration will be started
// later, once the current packet is confirmed to meet the following
// conditions:
// 1) current_effective_peer_migration_type_ is not NO_CHANGE.
// 2) The current packet is not a connectivity probing.
// 3) The current packet is not reordered, i.e. its packet number is the
// largest of this connection so far.
// Once the above conditions are confirmed, a new migration will start
// even if there is an active migration underway.
current_effective_peer_migration_type_ =
QuicUtils::DetermineAddressChangeType(
effective_peer_address_,
GetEffectivePeerAddressFromCurrentPacket());
QUIC_DLOG_IF(INFO, current_effective_peer_migration_type_ != NO_CHANGE)
<< ENDPOINT << "Effective peer's ip:port changed from "
<< effective_peer_address_.ToString() << " to "
<< GetEffectivePeerAddressFromCurrentPacket().ToString()
<< ", active_effective_peer_migration_type is "
<< active_effective_peer_migration_type_;
}
--stats_.packets_dropped;
QUIC_DVLOG(1) << ENDPOINT << "Received packet header: " << header;
last_header_ = header;
// An ack will be sent if a missing retransmittable packet was received;
was_last_packet_missing_ =
received_packet_manager_.IsMissing(last_header_.packet_number);
// Record packet receipt to populate ack info before processing stream
// frames, since the processing may result in sending a bundled ack.
received_packet_manager_.RecordPacketReceived(last_header_,
time_of_last_received_packet_);
DCHECK(connected_);
return true;
}
bool QuicConnection::OnStreamFrame(const QuicStreamFrame& frame) {
DCHECK(connected_);
// Since a stream frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnStreamFrame(frame);
}
if (frame.stream_id != QuicUtils::GetCryptoStreamId(transport_version()) &&
last_decrypted_packet_level_ == ENCRYPTION_NONE) {
if (MaybeConsiderAsMemoryCorruption(frame)) {
CloseConnection(QUIC_MAYBE_CORRUPTED_MEMORY,
"Received crypto frame on non crypto stream.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
QUIC_PEER_BUG << ENDPOINT
<< "Received an unencrypted data frame: closing connection"
<< " packet_number:" << last_header_.packet_number
<< " stream_id:" << frame.stream_id << " received_packets:"
<< received_packet_manager_.ack_frame();
CloseConnection(QUIC_UNENCRYPTED_STREAM_DATA,
"Unencrypted stream data seen.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
visitor_->OnStreamFrame(frame);
stats_.stream_bytes_received += frame.data_length;
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnCryptoFrame(const QuicCryptoFrame& frame) {
DCHECK(connected_);
// Since a CRYPTO frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
visitor_->OnCryptoFrame(frame);
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnAckFrameStart(QuicPacketNumber largest_acked,
QuicTime::Delta ack_delay_time) {
DCHECK(connected_);
if (processing_ack_frame_) {
CloseConnection(QUIC_INVALID_ACK_DATA,
"Received a new ack while processing an ack frame.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
// Since an ack frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
QUIC_DVLOG(1) << ENDPOINT
<< "OnAckFrameStart, largest_acked: " << largest_acked;
if (largest_seen_packet_with_ack_.IsInitialized() &&
last_header_.packet_number <= largest_seen_packet_with_ack_) {
QUIC_DLOG(INFO) << ENDPOINT << "Received an old ack frame: ignoring";
return true;
}
if (!sent_packet_manager_.GetLargestSentPacket().IsInitialized() ||
largest_acked > sent_packet_manager_.GetLargestSentPacket()) {
QUIC_DLOG(WARNING) << ENDPOINT
<< "Peer's observed unsent packet:" << largest_acked
<< " vs " << sent_packet_manager_.GetLargestSentPacket();
// We got an ack for data we have not sent.
CloseConnection(QUIC_INVALID_ACK_DATA, "Largest observed too high.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
if (!sent_packet_manager_.GetLargestObserved().IsInitialized() ||
largest_acked > sent_packet_manager_.GetLargestObserved()) {
visitor_->OnForwardProgressConfirmed();
} else if (largest_acked < sent_packet_manager_.GetLargestObserved()) {
QUIC_LOG(INFO) << ENDPOINT << "Peer's largest_observed packet decreased:"
<< largest_acked << " vs "
<< sent_packet_manager_.GetLargestObserved()
<< " packet_number:" << last_header_.packet_number
<< " largest seen with ack:" << largest_seen_packet_with_ack_
<< " connection_id: " << connection_id_;
// A new ack has a diminished largest_observed value.
// If this was an old packet, we wouldn't even have checked.
CloseConnection(QUIC_INVALID_ACK_DATA, "Largest observed too low.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
processing_ack_frame_ = true;
sent_packet_manager_.OnAckFrameStart(largest_acked, ack_delay_time,
time_of_last_received_packet_);
return true;
}
bool QuicConnection::OnAckRange(QuicPacketNumber start, QuicPacketNumber end) {
DCHECK(connected_);
QUIC_DVLOG(1) << ENDPOINT << "OnAckRange: [" << start << ", " << end << ")";
if (largest_seen_packet_with_ack_.IsInitialized() &&
last_header_.packet_number <= largest_seen_packet_with_ack_) {
QUIC_DLOG(INFO) << ENDPOINT << "Received an old ack frame: ignoring";
return true;
}
sent_packet_manager_.OnAckRange(start, end);
return true;
}
bool QuicConnection::OnAckTimestamp(QuicPacketNumber packet_number,
QuicTime timestamp) {
DCHECK(connected_);
QUIC_DVLOG(1) << ENDPOINT << "OnAckTimestamp: [" << packet_number << ", "
<< timestamp.ToDebuggingValue() << ")";
if (largest_seen_packet_with_ack_.IsInitialized() &&
last_header_.packet_number <= largest_seen_packet_with_ack_) {
QUIC_DLOG(INFO) << ENDPOINT << "Received an old ack frame: ignoring";
return true;
}
sent_packet_manager_.OnAckTimestamp(packet_number, timestamp);
return true;
}
bool QuicConnection::OnAckFrameEnd(QuicPacketNumber start) {
DCHECK(connected_);
QUIC_DVLOG(1) << ENDPOINT << "OnAckFrameEnd, start: " << start;
if (largest_seen_packet_with_ack_.IsInitialized() &&
last_header_.packet_number <= largest_seen_packet_with_ack_) {
QUIC_DLOG(INFO) << ENDPOINT << "Received an old ack frame: ignoring";
return true;
}
bool acked_new_packet =
sent_packet_manager_.OnAckFrameEnd(time_of_last_received_packet_);
// Cancel the send alarm because new packets likely have been acked, which
// may change the congestion window and/or pacing rate. Canceling the alarm
// causes CanWrite to recalculate the next send time.
if (send_alarm_->IsSet()) {
send_alarm_->Cancel();
}
if (supports_release_time_) {
// Update pace time into future because smoothed RTT is likely updated.
UpdateReleaseTimeIntoFuture();
}
largest_seen_packet_with_ack_ = last_header_.packet_number;
// If the incoming ack's packets set expresses missing packets: peer is still
// waiting for a packet lower than a packet that we are no longer planning to
// send.
// If the incoming ack's packets set expresses received packets: peer is still
// acking packets which we never care about.
// Send an ack to raise the high water mark.
PostProcessAfterAckFrame(GetLeastUnacked() > start, acked_new_packet);
processing_ack_frame_ = false;
return connected_;
}
bool QuicConnection::OnStopWaitingFrame(const QuicStopWaitingFrame& frame) {
DCHECK(connected_);
// Since a stop waiting frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (no_stop_waiting_frames_) {
return true;
}
if (largest_seen_packet_with_stop_waiting_.IsInitialized() &&
last_header_.packet_number <= largest_seen_packet_with_stop_waiting_) {
QUIC_DLOG(INFO) << ENDPOINT
<< "Received an old stop waiting frame: ignoring";
return true;
}
const char* error = ValidateStopWaitingFrame(frame);
if (error != nullptr) {
CloseConnection(QUIC_INVALID_STOP_WAITING_DATA, error,
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
if (debug_visitor_ != nullptr) {
debug_visitor_->OnStopWaitingFrame(frame);
}
largest_seen_packet_with_stop_waiting_ = last_header_.packet_number;
received_packet_manager_.DontWaitForPacketsBefore(frame.least_unacked);
return connected_;
}
bool QuicConnection::OnPaddingFrame(const QuicPaddingFrame& frame) {
DCHECK(connected_);
UpdatePacketContent(SECOND_FRAME_IS_PADDING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPaddingFrame(frame);
}
return true;
}
bool QuicConnection::OnPingFrame(const QuicPingFrame& frame) {
DCHECK(connected_);
UpdatePacketContent(FIRST_FRAME_IS_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPingFrame(frame);
}
should_last_packet_instigate_acks_ = true;
return true;
}
const char* QuicConnection::ValidateAckFrame(const QuicAckFrame& incoming_ack) {
if (LargestAcked(incoming_ack) > packet_generator_.packet_number()) {
QUIC_DLOG(WARNING) << ENDPOINT << "Peer's observed unsent packet:"
<< LargestAcked(incoming_ack) << " vs "
<< packet_generator_.packet_number();
// We got an error for data we have not sent. Error out.
return "Largest observed too high.";
}
if (LargestAcked(incoming_ack) < sent_packet_manager_.GetLargestObserved()) {
QUIC_LOG(INFO) << ENDPOINT << "Peer's largest_observed packet decreased:"
<< LargestAcked(incoming_ack) << " vs "
<< sent_packet_manager_.GetLargestObserved()
<< " packet_number:" << last_header_.packet_number
<< " largest seen with ack:" << largest_seen_packet_with_ack_
<< " connection_id: " << connection_id_;
// A new ack has a diminished largest_observed value. Error out.
// If this was an old packet, we wouldn't even have checked.
return "Largest observed too low.";
}
if (!incoming_ack.packets.Empty() &&
incoming_ack.packets.Max() != LargestAcked(incoming_ack)) {
QUIC_BUG << ENDPOINT
<< "Peer last received packet: " << incoming_ack.packets.Max()
<< " which is not equal to largest observed: "
<< incoming_ack.largest_acked;
return "Last received packet not equal to largest observed.";
}
return nullptr;
}
const char* QuicConnection::ValidateStopWaitingFrame(
const QuicStopWaitingFrame& stop_waiting) {
if (received_packet_manager_.peer_least_packet_awaiting_ack()
.IsInitialized() &&
stop_waiting.least_unacked <
received_packet_manager_.peer_least_packet_awaiting_ack()) {
QUIC_DLOG(ERROR)
<< ENDPOINT
<< "Peer's sent low least_unacked: " << stop_waiting.least_unacked
<< " vs " << received_packet_manager_.peer_least_packet_awaiting_ack();
// We never process old ack frames, so this number should only increase.
return "Least unacked too small.";
}
if (stop_waiting.least_unacked > last_header_.packet_number) {
QUIC_DLOG(ERROR) << ENDPOINT
<< "Peer sent least_unacked:" << stop_waiting.least_unacked
<< " greater than the enclosing packet number:"
<< last_header_.packet_number;
return "Least unacked too large.";
}
return nullptr;
}
bool QuicConnection::OnRstStreamFrame(const QuicRstStreamFrame& frame) {
DCHECK(connected_);
// Since a reset stream frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnRstStreamFrame(frame);
}
QUIC_DLOG(INFO) << ENDPOINT
<< "RST_STREAM_FRAME received for stream: " << frame.stream_id
<< " with error: "
<< QuicRstStreamErrorCodeToString(frame.error_code);
visitor_->OnRstStream(frame);
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnApplicationCloseFrame(
const QuicApplicationCloseFrame& frame) {
// TODO(fkastenholz): Need to figure out what the right thing is to do with
// this when we get one. Most likely, the correct action is to mimic the
// OnConnectionCloseFrame actions, with possibly an indication to the
// application of the ApplicationClose information.
return true;
}
bool QuicConnection::OnStopSendingFrame(const QuicStopSendingFrame& frame) {
DCHECK(connected_);
// Since a reset stream frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnStopSendingFrame(frame);
}
QUIC_DLOG(INFO) << ENDPOINT << "STOP_SENDING frame received for stream: "
<< frame.stream_id
<< " with error: " << frame.application_error_code;
visitor_->OnStopSendingFrame(frame);
return connected_;
}
bool QuicConnection::OnPathChallengeFrame(const QuicPathChallengeFrame& frame) {
// Save the path challenge's payload, for later use in generating the
// response.
received_path_challenge_payloads_.push_back(frame.data_buffer);
// For VERSION 99 we define a "Padded PATH CHALLENGE" to be the same thing
// as a PADDED PING -- it will start a connectivity check and prevent
// connection migration. Insofar as the connectivity check and connection
// migration are concerned, logically the PATH CHALLENGE is the same as the
// PING, so as a stopgap, tell the FSM that determines whether we have a
// Padded PING or not that we received a PING.
UpdatePacketContent(FIRST_FRAME_IS_PING);
should_last_packet_instigate_acks_ = true;
return true;
}
bool QuicConnection::OnPathResponseFrame(const QuicPathResponseFrame& frame) {
should_last_packet_instigate_acks_ = true;
if (!transmitted_connectivity_probe_payload_ ||
*transmitted_connectivity_probe_payload_ != frame.data_buffer) {
// Is not for the probe we sent, ignore it.
return true;
}
// Have received the matching PATH RESPONSE, saved payload no longer valid.
transmitted_connectivity_probe_payload_ = nullptr;
UpdatePacketContent(FIRST_FRAME_IS_PING);
return true;
}
bool QuicConnection::OnConnectionCloseFrame(
const QuicConnectionCloseFrame& frame) {
DCHECK(connected_);
// Since a connection close frame was received, this is not a connectivity
// probe. A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnConnectionCloseFrame(frame);
}
QUIC_DLOG(INFO) << ENDPOINT << "Received ConnectionClose for connection: "
<< connection_id()
<< ", with error: " << QuicErrorCodeToString(frame.error_code)
<< " (" << frame.error_details << ")";
if (frame.error_code == QUIC_BAD_MULTIPATH_FLAG) {
QUIC_LOG_FIRST_N(ERROR, 10) << "Unexpected QUIC_BAD_MULTIPATH_FLAG error."
<< " last_received_header: " << last_header_
<< " encryption_level: " << encryption_level_;
}
TearDownLocalConnectionState(frame.error_code, frame.error_details,
ConnectionCloseSource::FROM_PEER);
return connected_;
}
bool QuicConnection::OnMaxStreamIdFrame(const QuicMaxStreamIdFrame& frame) {
return visitor_->OnMaxStreamIdFrame(frame);
}
bool QuicConnection::OnStreamIdBlockedFrame(
const QuicStreamIdBlockedFrame& frame) {
return visitor_->OnStreamIdBlockedFrame(frame);
}
bool QuicConnection::OnGoAwayFrame(const QuicGoAwayFrame& frame) {
DCHECK(connected_);
// Since a go away frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnGoAwayFrame(frame);
}
QUIC_DLOG(INFO) << ENDPOINT << "GOAWAY_FRAME received with last good stream: "
<< frame.last_good_stream_id
<< " and error: " << QuicErrorCodeToString(frame.error_code)
<< " and reason: " << frame.reason_phrase;
visitor_->OnGoAway(frame);
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnWindowUpdateFrame(const QuicWindowUpdateFrame& frame) {
DCHECK(connected_);
// Since a window update frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnWindowUpdateFrame(frame, time_of_last_received_packet_);
}
QUIC_DLOG(INFO) << ENDPOINT << "WINDOW_UPDATE_FRAME received for stream: "
<< frame.stream_id
<< " with byte offset: " << frame.byte_offset;
visitor_->OnWindowUpdateFrame(frame);
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnNewConnectionIdFrame(
const QuicNewConnectionIdFrame& frame) {
return true;
}
bool QuicConnection::OnRetireConnectionIdFrame(
const QuicRetireConnectionIdFrame& frame) {
return true;
}
bool QuicConnection::OnNewTokenFrame(const QuicNewTokenFrame& frame) {
return true;
}
bool QuicConnection::OnMessageFrame(const QuicMessageFrame& frame) {
DCHECK(connected_);
// Since a message frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnMessageFrame(frame);
}
visitor_->OnMessageReceived(
QuicStringPiece(frame.data, frame.message_length));
should_last_packet_instigate_acks_ = true;
return connected_;
}
bool QuicConnection::OnBlockedFrame(const QuicBlockedFrame& frame) {
DCHECK(connected_);
// Since a blocked frame was received, this is not a connectivity probe.
// A probe only contains a PING and full padding.
UpdatePacketContent(NOT_PADDED_PING);
if (debug_visitor_ != nullptr) {
debug_visitor_->OnBlockedFrame(frame);
}
QUIC_DLOG(INFO) << ENDPOINT
<< "BLOCKED_FRAME received for stream: " << frame.stream_id;
visitor_->OnBlockedFrame(frame);
stats_.blocked_frames_received++;
should_last_packet_instigate_acks_ = true;
return connected_;
}
void QuicConnection::OnPacketComplete() {
// Don't do anything if this packet closed the connection.
if (!connected_) {
ClearLastFrames();
return;
}
if (IsCurrentPacketConnectivityProbing()) {
++stats_.num_connectivity_probing_received;
}
QUIC_DVLOG(1) << ENDPOINT << "Got packet " << last_header_.packet_number
<< " for " << last_header_.destination_connection_id;
QUIC_DLOG_IF(INFO, current_packet_content_ == SECOND_FRAME_IS_PADDING)
<< ENDPOINT << "Received a padded PING packet. is_probing: "
<< IsCurrentPacketConnectivityProbing();
if (perspective_ == Perspective::IS_CLIENT) {
QUIC_DVLOG(1) << ENDPOINT
<< "Received a speculative connectivity probing packet for "
<< last_header_.destination_connection_id
<< " from ip:port: " << last_packet_source_address_.ToString()
<< " to ip:port: "
<< last_packet_destination_address_.ToString();
// TODO(zhongyi): change the method name.
visitor_->OnConnectivityProbeReceived(last_packet_destination_address_,
last_packet_source_address_);
} else if (IsCurrentPacketConnectivityProbing()) {
// This node is not a client (is a server) AND the received packet was
// connectivity-probing, send an appropriate response.
QUIC_DVLOG(1) << ENDPOINT << "Received a connectivity probing packet for "
<< last_header_.destination_connection_id
<< " from ip:port: " << last_packet_source_address_.ToString()
<< " to ip:port: "
<< last_packet_destination_address_.ToString();
visitor_->OnConnectivityProbeReceived(last_packet_destination_address_,
last_packet_source_address_);
} else {
// This node is not a client (is a server) AND the received packet was
// NOT connectivity-probing. If the packet had PATH CHALLENGES, send
// appropriate RESPONSE. Then deal with possible peer migration.
if (transport_version() == QUIC_VERSION_99 &&
!received_path_challenge_payloads_.empty()) {
// If a PATH CHALLENGE was in a "Padded PING (or PATH CHALLENGE)"
// then it is taken care of above. This handles the case where a PATH
// CHALLENGE appeared someplace else (eg, the peer randomly added a PATH
// CHALLENGE frame to some other packet.
// There was at least one PATH CHALLENGE in the received packet,
// Generate the required PATH RESPONSE.
SendGenericPathProbePacket(nullptr, last_packet_source_address_,
/* is_response= */ true);
}
if (last_header_.packet_number ==
received_packet_manager_.GetLargestObserved()) {
direct_peer_address_ = last_packet_source_address_;
if (current_effective_peer_migration_type_ != NO_CHANGE) {
StartEffectivePeerMigration(current_effective_peer_migration_type_);
}
}
}
current_effective_peer_migration_type_ = NO_CHANGE;
// An ack will be sent if a missing retransmittable packet was received;
const bool was_missing =
should_last_packet_instigate_acks_ && was_last_packet_missing_;
// It's possible the ack frame was sent along with response data, so it
// no longer needs to be sent.
if (ack_frame_updated()) {
MaybeQueueAck(was_missing);
}
ClearLastFrames();
CloseIfTooManyOutstandingSentPackets();
}
bool QuicConnection::IsValidStatelessResetToken(QuicUint128 token) const {
return stateless_reset_token_received_ &&
token == received_stateless_reset_token_;
}
void QuicConnection::OnAuthenticatedIetfStatelessResetPacket(
const QuicIetfStatelessResetPacket& packet) {
// TODO(fayang): Add OnAuthenticatedIetfStatelessResetPacket to
// debug_visitor_.
const QuicString error_details = "Received stateless reset.";
QUIC_CODE_COUNT(quic_tear_down_local_connection_on_stateless_reset);
TearDownLocalConnectionState(QUIC_PUBLIC_RESET, error_details,
ConnectionCloseSource::FROM_PEER);
}
void QuicConnection::MaybeQueueAck(bool was_missing) {
++num_packets_received_since_last_ack_sent_;
// Determine whether the newly received packet was missing before recording
// the received packet.
if (was_missing) {
// Only ack immediately if an ACK frame was sent with a larger
// largest acked than the newly received packet number.
const QuicPacketNumber largest_sent_largest_acked =
sent_packet_manager_.unacked_packets().largest_sent_largest_acked();
if (largest_sent_largest_acked.IsInitialized() &&
last_header_.packet_number < largest_sent_largest_acked) {
ack_queued_ = true;
}
}
if (should_last_packet_instigate_acks_ && !ack_queued_) {
++num_retransmittable_packets_received_since_last_ack_sent_;
if (ack_mode_ != TCP_ACKING &&
last_header_.packet_number >=
received_packet_manager_.PeerFirstSendingPacketNumber() +
min_received_before_ack_decimation_) {
// Ack up to 10 packets at once unless ack decimation is unlimited.
if (!unlimited_ack_decimation_ &&
num_retransmittable_packets_received_since_last_ack_sent_ >=
kMaxRetransmittablePacketsBeforeAck) {
ack_queued_ = true;
} else if (ShouldSetAckAlarm()) {
// Wait for the minimum of the ack decimation delay or the delayed ack
// time before sending an ack.
QuicTime::Delta ack_delay =
std::min(sent_packet_manager_.delayed_ack_time(),
sent_packet_manager_.GetRttStats()->min_rtt() *
ack_decimation_delay_);
const QuicTime approximate_now = clock_->ApproximateNow();
if (fast_ack_after_quiescence_ &&
(approximate_now - time_of_previous_received_packet_) >
sent_packet_manager_.GetRttStats()->SmoothedOrInitialRtt()) {
// Ack the first packet out of queiscence faster, because QUIC does
// not pace the first few packets and commonly these may be handshake
// or TLP packets, which we'd like to acknowledge quickly.
ack_delay = QuicTime::Delta::FromMilliseconds(1);
}
ack_alarm_->Set(approximate_now + ack_delay);
}
} else {
// Ack with a timer or every 2 packets by default.
if (num_retransmittable_packets_received_since_last_ack_sent_ >=
ack_frequency_before_ack_decimation_) {
ack_queued_ = true;
} else if (ShouldSetAckAlarm()) {
const QuicTime approximate_now = clock_->ApproximateNow();
if (fast_ack_after_quiescence_ &&
(approximate_now - time_of_previous_received_packet_) >
sent_packet_manager_.GetRttStats()->SmoothedOrInitialRtt()) {
// Ack the first packet out of queiscence faster, because QUIC does
// not pace the first few packets and commonly these may be handshake
// or TLP packets, which we'd like to acknowledge quickly.
ack_alarm_->Set(approximate_now +
QuicTime::Delta::FromMilliseconds(1));
} else {
ack_alarm_->Set(approximate_now +
sent_packet_manager_.delayed_ack_time());
}
}
}
// If there are new missing packets to report, send an ack immediately.
if (received_packet_manager_.HasNewMissingPackets()) {
if (ack_mode_ == ACK_DECIMATION_WITH_REORDERING) {
DCHECK(!GetQuicReloadableFlag(quic_enable_ack_decimation) ||
GetQuicReloadableFlag(quic_keep_ack_decimation_reordering));
// Wait the minimum of an eighth min_rtt and the existing ack time.
QuicTime ack_time =
clock_->ApproximateNow() +
0.125 * sent_packet_manager_.GetRttStats()->min_rtt();
if (ShouldSetAckAlarm() || ack_alarm_->deadline() > ack_time) {
ack_alarm_->Update(ack_time, QuicTime::Delta::Zero());
}
} else {
ack_queued_ = true;
}
}
if (fast_ack_after_quiescence_) {
time_of_previous_received_packet_ = time_of_last_received_packet_;
}
}
if (ack_queued_) {
ack_alarm_->Cancel();
}
}
void QuicConnection::ClearLastFrames() {
should_last_packet_instigate_acks_ = false;
}
void QuicConnection::CloseIfTooManyOutstandingSentPackets() {
// This occurs if we don't discard old packets we've seen fast enough. It's
// possible largest observed is less than leaset unacked.
if (sent_packet_manager_.GetLargestObserved().IsInitialized() &&
sent_packet_manager_.GetLargestObserved() >
sent_packet_manager_.GetLeastUnacked() + max_tracked_packets_) {
CloseConnection(
QUIC_TOO_MANY_OUTSTANDING_SENT_PACKETS,
QuicStrCat("More than ", max_tracked_packets_,
" outstanding, least_unacked: ",
sent_packet_manager_.GetLeastUnacked().ToUint64()),
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
}
}
const QuicFrame QuicConnection::GetUpdatedAckFrame() {
return received_packet_manager_.GetUpdatedAckFrame(clock_->ApproximateNow());
}
void QuicConnection::PopulateStopWaitingFrame(
QuicStopWaitingFrame* stop_waiting) {
stop_waiting->least_unacked = GetLeastUnacked();
}
QuicPacketNumber QuicConnection::GetLeastUnacked() const {
return sent_packet_manager_.GetLeastUnacked();
}
bool QuicConnection::HandleWriteBlocked() {
if (!writer_->IsWriteBlocked()) {
return false;
}
visitor_->OnWriteBlocked();
return true;
}
void QuicConnection::MaybeSendInResponseToPacket() {
if (!connected_) {
return;
}
// If the writer is blocked, don't attempt to send packets now or in the send
// alarm. When the writer unblocks, OnCanWrite() will be called for this
// connection to send.
if (HandleWriteBlocked()) {
return;
}
// Now that we have received an ack, we might be able to send packets which
// are queued locally, or drain streams which are blocked.
if (defer_send_in_response_to_packets_) {
send_alarm_->Update(clock_->ApproximateNow(), QuicTime::Delta::Zero());
} else {
WriteAndBundleAcksIfNotBlocked();
}
}
void QuicConnection::SendVersionNegotiationPacket(bool ietf_quic) {
pending_version_negotiation_packet_ = true;
send_ietf_version_negotiation_packet_ = ietf_quic;
if (HandleWriteBlocked()) {
return;
}
QUIC_DLOG(INFO) << ENDPOINT << "Sending version negotiation packet: {"
<< ParsedQuicVersionVectorToString(
framer_.supported_versions())
<< "}, ietf_quic: " << ietf_quic;
std::unique_ptr<QuicEncryptedPacket> version_packet(
packet_generator_.SerializeVersionNegotiationPacket(
ietf_quic, framer_.supported_versions()));
WriteResult result = writer_->WritePacket(
version_packet->data(), version_packet->length(), self_address().host(),
peer_address(), per_packet_options_);
if (IsWriteError(result.status)) {
OnWriteError(result.error_code);
return;
}
if (IsWriteBlockedStatus(result.status)) {
visitor_->OnWriteBlocked();
if (result.status == WRITE_STATUS_BLOCKED_DATA_BUFFERED) {
pending_version_negotiation_packet_ = false;
}
return;
}
pending_version_negotiation_packet_ = false;
}
size_t QuicConnection::SendCryptoData(EncryptionLevel level,
size_t write_length,
QuicStreamOffset offset) {
if (write_length == 0) {
QUIC_BUG << "Attempt to send empty crypto frame";
return 0;
}
ScopedPacketFlusher flusher(this, SEND_ACK_IF_PENDING);
return packet_generator_.ConsumeCryptoData(level, write_length, offset);
}
QuicConsumedData QuicConnection::SendStreamData(QuicStreamId id,
size_t write_length,
QuicStreamOffset offset,
StreamSendingState state) {
if (state == NO_FIN && write_length == 0) {
QUIC_BUG << "Attempt to send empty stream frame";
return QuicConsumedData(0, false);
}
// Opportunistically bundle an ack with every outgoing packet.
// Particularly, we want to bundle with handshake packets since we don't know
// which decrypter will be used on an ack packet following a handshake
// packet (a handshake packet from client to server could result in a REJ or a
// SHLO from the server, leading to two different decrypters at the server.)
ScopedPacketFlusher flusher(this, SEND_ACK_IF_PENDING);
return packet_generator_.ConsumeData(id, write_length, offset, state);
}
bool QuicConnection::SendControlFrame(const QuicFrame& frame) {
if (!CanWrite(HAS_RETRANSMITTABLE_DATA) && frame.type != PING_FRAME) {
QUIC_DVLOG(1) << ENDPOINT << "Failed to send control frame: " << frame;
// Do not check congestion window for ping.
return false;
}
ScopedPacketFlusher flusher(this, SEND_ACK_IF_PENDING);
packet_generator_.AddControlFrame(frame);
if (frame.type == PING_FRAME) {
// Flush PING frame immediately.
packet_generator_.FlushAllQueuedFrames();
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPingSent();
}
}
if (frame.type == BLOCKED_FRAME) {
stats_.blocked_frames_sent++;
}
return true;
}
void QuicConnection::OnStreamReset(QuicStreamId id,
QuicRstStreamErrorCode error) {
if (error == QUIC_STREAM_NO_ERROR) {
// All data for streams which are reset with QUIC_STREAM_NO_ERROR must
// be received by the peer.
return;
}
// Flush stream frames of reset stream.
if (packet_generator_.HasPendingStreamFramesOfStream(id)) {
ScopedPacketFlusher flusher(this, SEND_ACK_IF_PENDING);
packet_generator_.FlushAllQueuedFrames();
}
sent_packet_manager_.CancelRetransmissionsForStream(id);
// Remove all queued packets which only contain data for the reset stream.
// TODO(fayang): consider removing this because it should be rarely executed.
auto packet_iterator = queued_packets_.begin();
while (packet_iterator != queued_packets_.end()) {
QuicFrames* retransmittable_frames =
&packet_iterator->retransmittable_frames;
if (retransmittable_frames->empty()) {
++packet_iterator;
continue;
}
// NOTE THAT RemoveFramesForStream removes only STREAM frames
// for the specified stream.
RemoveFramesForStream(retransmittable_frames, id);
if (!retransmittable_frames->empty()) {
++packet_iterator;
continue;
}
delete[] packet_iterator->encrypted_buffer;
ClearSerializedPacket(&(*packet_iterator));
packet_iterator = queued_packets_.erase(packet_iterator);
}
// TODO(ianswett): Consider checking for 3 RTOs when the last stream is
// cancelled as well.
}
const QuicConnectionStats& QuicConnection::GetStats() {
const RttStats* rtt_stats = sent_packet_manager_.GetRttStats();
// Update rtt and estimated bandwidth.
QuicTime::Delta min_rtt = rtt_stats->min_rtt();
if (min_rtt.IsZero()) {
// If min RTT has not been set, use initial RTT instead.
min_rtt = rtt_stats->initial_rtt();
}
stats_.min_rtt_us = min_rtt.ToMicroseconds();
QuicTime::Delta srtt = rtt_stats->SmoothedOrInitialRtt();
stats_.srtt_us = srtt.ToMicroseconds();
stats_.estimated_bandwidth = sent_packet_manager_.BandwidthEstimate();
stats_.max_packet_size = packet_generator_.GetCurrentMaxPacketLength();
stats_.max_received_packet_size = largest_received_packet_size_;
return stats_;
}
void QuicConnection::OnCoalescedPacket(const QuicEncryptedPacket& packet) {
QueueCoalescedPacket(packet);
}
void QuicConnection::ProcessUdpPacket(const QuicSocketAddress& self_address,
const QuicSocketAddress& peer_address,
const QuicReceivedPacket& packet) {
if (!connected_) {
return;
}
QUIC_BUG_IF(current_packet_data_ != nullptr)
<< "ProcessUdpPacket must not be called while processing a packet.";
if (debug_visitor_ != nullptr) {
debug_visitor_->OnPacketReceived(self_address, peer_address, packet);
}
last_size_ = packet.length();
current_packet_data_ = packet.data();
last_packet_destination_address_ = self_address;
last_packet_source_address_ = peer_address;
if (!self_address_.IsInitialized()) {
self_address_ = last_packet_destination_address_;
}
if (!direct_peer_address_.IsInitialized()) {
direct_peer_address_ = last_packet_source_address_;
}
if (!effective_peer_address_.IsInitialized()) {
const QuicSocketAddress effective_peer_addr =
GetEffectivePeerAddressFromCurrentPacket();
// effective_peer_address_ must be initialized at the beginning of the
// first packet processed(here). If effective_peer_addr is uninitialized,
// just set effective_peer_address_ to the direct peer address.
effective_peer_address_ = effective_peer_addr.IsInitialized()
? effective_peer_addr
: direct_peer_address_;
}
stats_.bytes_received += packet.length();
++stats_.packets_received;
// Ensure the time coming from the packet reader is within 2 minutes of now.
if (std::abs((packet.receipt_time() - clock_->ApproximateNow()).ToSeconds()) >
2 * 60) {
QUIC_BUG << "Packet receipt time:"
<< packet.receipt_time().ToDebuggingValue()
<< " too far from current time:"
<< clock_->ApproximateNow().ToDebuggingValue();
}
time_of_last_received_packet_ = packet.receipt_time();
QUIC_DVLOG(1) << ENDPOINT << "time of last received packet: "
<< time_of_last_received_packet_.ToDebuggingValue();
ScopedPacketFlusher flusher(this, NO_ACK);
if (!framer_.ProcessPacket(packet)) {
// If we are unable to decrypt this packet, it might be
// because the CHLO or SHLO packet was lost.
if (framer_.error() == QUIC_DECRYPTION_FAILURE) {
if (encryption_level_ != ENCRYPTION_FORWARD_SECURE &&
undecryptable_packets_.size() < max_undecryptable_packets_) {
QueueUndecryptablePacket(packet);
} else if (debug_visitor_ != nullptr) {
debug_visitor_->OnUndecryptablePacket();
}
}
QUIC_DVLOG(1) << ENDPOINT
<< "Unable to process packet. Last packet processed: "
<< last_header_.packet_number;
current_packet_data_ = nullptr;
is_current_packet_connectivity_probing_ = false;
MaybeProcessCoalescedPackets();
return;
}
++stats_.packets_processed;
QUIC_DLOG_IF(INFO, active_effective_peer_migration_type_ != NO_CHANGE)
<< "sent_packet_manager_.GetLargestObserved() = "
<< sent_packet_manager_.GetLargestObserved()
<< ", highest_packet_sent_before_effective_peer_migration_ = "
<< highest_packet_sent_before_effective_peer_migration_;
if (active_effective_peer_migration_type_ != NO_CHANGE &&
sent_packet_manager_.GetLargestObserved().IsInitialized() &&
(!highest_packet_sent_before_effective_peer_migration_.IsInitialized() ||
sent_packet_manager_.GetLargestObserved() >
highest_packet_sent_before_effective_peer_migration_)) {
if (perspective_ == Perspective::IS_SERVER) {
OnEffectivePeerMigrationValidated();
}
}
MaybeProcessCoalescedPackets();
MaybeProcessUndecryptablePackets();
MaybeSendInResponseToPacket();
SetPingAlarm();
current_packet_data_ = nullptr;
is_current_packet_connectivity_probing_ = false;
}
void QuicConnection::OnBlockedWriterCanWrite() {
if (GetQuicRestartFlag(quic_check_blocked_writer_for_blockage)) {
QUIC_RESTART_FLAG_COUNT_N(quic_check_blocked_writer_for_blockage, 3, 6);
writer_->SetWritable();
}
OnCanWrite();
}
void QuicConnection::OnCanWrite() {
DCHECK(!writer_->IsWriteBlocked());
// Add a flusher to ensure the connection is marked app-limited.
ScopedPacketFlusher flusher(this, NO_ACK);
WriteQueuedPackets();
if (!session_decides_what_to_write()) {
WritePendingRetransmissions();
}
WriteNewData();
}
void QuicConnection::WriteNewData() {
// Sending queued packets may have caused the socket to become write blocked,
// or the congestion manager to prohibit sending. If we've sent everything
// we had queued and we're still not blocked, let the visitor know it can
// write more.
if (!CanWrite(HAS_RETRANSMITTABLE_DATA)) {
return;
}
{
ScopedPacketFlusher flusher(this, SEND_ACK_IF_QUEUED);
visitor_->OnCanWrite();
}
// After the visitor writes, it may have caused the socket to become write
// blocked or the congestion manager to prohibit sending, so check again.
if (visitor_->WillingAndAbleToWrite() && !send_alarm_->IsSet() &&
CanWrite(HAS_RETRANSMITTABLE_DATA)) {
// We're not write blocked, but some stream didn't write out all of its
// bytes. Register for 'immediate' resumption so we'll keep writing after
// other connections and events have had a chance to use the thread.
send_alarm_->Set(clock_->ApproximateNow());
}
}
void QuicConnection::WriteIfNotBlocked() {
if (!HandleWriteBlocked()) {
OnCanWrite();
}
}
void QuicConnection::WriteAndBundleAcksIfNotBlocked() {
if (!HandleWriteBlocked()) {
ScopedPacketFlusher flusher(this, SEND_ACK_IF_QUEUED);
WriteIfNotBlocked();
}
}
bool QuicConnection::ProcessValidatedPacket(const QuicPacketHeader& header) {
if (perspective_ == Perspective::IS_SERVER && self_address_.IsInitialized() &&
last_packet_destination_address_.IsInitialized() &&
self_address_ != last_packet_destination_address_) {
// Allow change between pure IPv4 and equivalent mapped IPv4 address.
if (self_address_.port() != last_packet_destination_address_.port() ||
self_address_.host().Normalized() !=
last_packet_destination_address_.host().Normalized()) {
if (!visitor_->AllowSelfAddressChange()) {
CloseConnection(
QUIC_ERROR_MIGRATING_ADDRESS,
"Self address migration is not supported at the server.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
}
self_address_ = last_packet_destination_address_;
}
if (GetQuicRestartFlag(quic_enable_accept_random_ipn)) {
QUIC_RESTART_FLAG_COUNT_N(quic_enable_accept_random_ipn, 2, 2);
// Configured to accept any packet number in range 1...0x7fffffff
// as initial packet number.
if (last_header_.packet_number.IsInitialized()) {
// The last packet's number is not 0. Ensure that this packet
// is reasonably close to where it should be.
if (!Near(header.packet_number, last_header_.packet_number)) {
QUIC_DLOG(INFO) << ENDPOINT << "Packet " << header.packet_number
<< " out of bounds. Discarding";
CloseConnection(QUIC_INVALID_PACKET_HEADER,
"Packet number out of bounds.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
} else {
// The "last packet's number" is 0, meaning that this packet is the first
// one received. Ensure it is in range 1..MaxRandomInitialPacketNumber(),
// inclusive.
if ((header.packet_number > MaxRandomInitialPacketNumber())) {
// packet number is bad.
QUIC_DLOG(INFO) << ENDPOINT << "Initial packet " << header.packet_number
<< " out of bounds. Discarding";
CloseConnection(QUIC_INVALID_PACKET_HEADER,
"Initial packet number out of bounds.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
}
} else { // if (GetQuicRestartFlag(quic_enable_accept_random_ipn))
// Count those that would have been accepted if FLAGS..random_ipn
// were true -- to detect/diagnose potential issues prior to
// enabling the flag.
if (header.packet_number >
received_packet_manager_.PeerFirstSendingPacketNumber() &&
header.packet_number <= MaxRandomInitialPacketNumber()) {
QUIC_CODE_COUNT_N(had_possibly_random_ipn, 2, 2);
}
bool out_of_bound =
last_header_.packet_number.IsInitialized()
? !Near(header.packet_number, last_header_.packet_number)
: header.packet_number >=
(received_packet_manager_.PeerFirstSendingPacketNumber() +
kMaxPacketGap);
if (out_of_bound) {
QUIC_DLOG(INFO) << ENDPOINT << "Packet " << header.packet_number
<< " out of bounds. Discarding";
QuicStringPiece packet_data = GetCurrentPacket();
const size_t kMaxPacketLengthInErrorDetails = 64;
CloseConnection(
QUIC_INVALID_PACKET_HEADER,
QuicStrCat("Packet number out of bounds. ",
last_header_.packet_number.IsInitialized()
? QuicStrCat("last_pkn=",
last_header_.packet_number.ToUint64())
: "first received packet",
", current_pkn=", header.packet_number.ToUint64(),
", current_pkt_len=", packet_data.length(),
", current_hdr=",
QuicTextUtils::HexEncode(
packet_data.length() > kMaxPacketLengthInErrorDetails
? QuicStringPiece(packet_data.data(),
kMaxPacketLengthInErrorDetails)
: packet_data)),
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return false;
}
}
if (version_negotiation_state_ != NEGOTIATED_VERSION) {
if (perspective_ == Perspective::IS_CLIENT) {
DCHECK(!header.version_flag || header.form != GOOGLE_QUIC_PACKET);
if (framer_.transport_version() <= QUIC_VERSION_43) {
// If the client gets a packet without the version flag from the server
// it should stop sending version since the version negotiation is done.
// IETF QUIC stops sending version once encryption level switches to
// forward secure.
packet_generator_.StopSendingVersion();
}
version_negotiation_state_ = NEGOTIATED_VERSION;
visitor_->OnSuccessfulVersionNegotiation(version());
if (debug_visitor_ != nullptr) {
debug_visitor_->OnSuccessfulVersionNegotiation(version());
}
}
}
if (last_size_ > largest_received_packet_size_) {
largest_received_packet_size_ = last_size_;
}
if (perspective_ == Perspective::IS_SERVER &&
encryption_level_ == ENCRYPTION_NONE &&
last_size_ > packet_generator_.GetCurrentMaxPacketLength()) {
SetMaxPacketLength(last_size_);
}
return true;
}
void QuicConnection::WriteQueuedPackets() {
DCHECK(!writer_->IsWriteBlocked());
if (pending_version_negotiation_packet_) {
SendVersionNegotiationPacket(send_ietf_version_negotiation_packet_);
}
while (!queued_packets_.empty()) {
// WritePacket() can potentially clear all queued packets, so we need to
// save the first queued packet to a local variable before calling it.
SerializedPacket packet(std::move(queued_packets_.front()));
queued_packets_.pop_front();
const bool write_result = WritePacket(&packet);
if (connected_ && !write_result) {
// Write failed but connection is open, re-insert |packet| into the
// front of the queue, it will be retried later.
queued_packets_.emplace_front(std::move(packet));
break;
}
delete[] packet.encrypted_buffer;
ClearSerializedPacket(&packet);
if (!connected_) {
DCHECK(queued_packets_.empty()) << "Queued packets should have been "
"cleared while closing connection";
break;
}
// Continue to send the next packet in queue.
}
}
void QuicConnection::WritePendingRetransmissions() {
DCHECK(!session_decides_what_to_write());
// Keep writing as long as there's a pending retransmission which can be
// written.
while (sent_packet_manager_.HasPendingRetransmissions() &&
CanWrite(HAS_RETRANSMITTABLE_DATA)) {
const QuicPendingRetransmission pending =
sent_packet_manager_.NextPendingRetransmission();
// Re-packetize the frames with a new packet number for retransmission.
// Retransmitted packets use the same packet number length as the
// original.
// Flush the packet generator before making a new packet.
// TODO(ianswett): Implement ReserializeAllFrames as a separate path that
// does not require the creator to be flushed.
// TODO(fayang): FlushAllQueuedFrames should only be called once, and should
// be moved outside of the loop. Also, CanWrite is not checked after the
// generator is flushed.
{
ScopedPacketFlusher flusher(this, NO_ACK);
packet_generator_.FlushAllQueuedFrames();
}
DCHECK(!packet_generator_.HasQueuedFrames());
char buffer[kMaxPacketSize];
packet_generator_.ReserializeAllFrames(pending, buffer, kMaxPacketSize);
}
}
void QuicConnection::SendProbingRetransmissions() {
while (sent_packet_manager_.GetSendAlgorithm()->ShouldSendProbingPacket() &&
CanWrite(HAS_RETRANSMITTABLE_DATA)) {
const bool can_retransmit =
sent_packet_manager_.MaybeRetransmitOldestPacket(
PROBING_RETRANSMISSION);
if (!can_retransmit) {
QUIC_DVLOG(1)
<< "Cannot send probing retransmissions: nothing to retransmit.";
break;
}
if (!session_decides_what_to_write()) {
DCHECK(sent_packet_manager_.HasPendingRetransmissions());
WritePendingRetransmissions();
}
}
}
void QuicConnection::RetransmitUnackedPackets(
TransmissionType retransmission_type) {
sent_packet_manager_.RetransmitUnackedPackets(retransmission_type);
WriteIfNotBlocked();
}
void QuicConnection::NeuterUnencryptedPackets() {
sent_packet_manager_.NeuterUnencryptedPackets();
// This may have changed the retransmission timer, so re-arm it.
SetRetransmissionAlarm();
}
bool QuicConnection::ShouldGeneratePacket(
HasRetransmittableData retransmittable,
IsHandshake handshake) {
// We should serialize handshake packets immediately to ensure that they
// end up sent at the right encryption level.
if (handshake == IS_HANDSHAKE) {
return true;
}
return CanWrite(retransmittable);
}
bool QuicConnection::CanWrite(HasRetransmittableData retransmittable) {
if (!connected_) {
return false;
}
if (session_decides_what_to_write() &&
sent_packet_manager_.pending_timer_transmission_count() > 0) {
// Force sending the retransmissions for HANDSHAKE, TLP, RTO, PROBING cases.
return true;
}
if (HandleWriteBlocked()) {
return false;
}
// Allow acks to be sent immediately.
if (retransmittable == NO_RETRANSMITTABLE_DATA) {
return true;
}
// If the send alarm is set, wait for it to fire.
if (send_alarm_->IsSet()) {
return false;
}
QuicTime now = clock_->Now();
QuicTime::Delta delay = sent_packet_manager_.TimeUntilSend(now);
if (delay.IsInfinite()) {
send_alarm_->Cancel();
return false;
}
// Scheduler requires a delay.
if (!delay.IsZero()) {
if (delay <= release_time_into_future_) {
// Required delay is within pace time into future, send now.
return true;
}
// Cannot send packet now because delay is too far in the future.
send_alarm_->Update(now + delay, QuicTime::Delta::FromMilliseconds(1));
QUIC_DVLOG(1) << ENDPOINT << "Delaying sending " << delay.ToMilliseconds()
<< "ms";
return false;
}
return true;
}
bool QuicConnection::WritePacket(SerializedPacket* packet) {
if (ShouldDiscardPacket(*packet)) {
++stats_.packets_discarded;
return true;
}
if (sent_packet_manager_.GetLargestSentPacket().IsInitialized() &&
packet->packet_number < sent_packet_manager_.GetLargestSentPacket()) {
QUIC_BUG << "Attempt to write packet:" << packet->packet_number
<< " after:" << sent_packet_manager_.GetLargestSentPacket();
CloseConnection(QUIC_INTERNAL_ERROR, "Packet written out of order.",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return true;
}
// Termination packets are encrypted and saved, so don't exit early.
const bool is_termination_packet = IsTerminationPacket(*packet);
if (HandleWriteBlocked() && !is_termination_packet) {
return false;
}
QuicPacketNumber packet_number = packet->packet_number;
QuicPacketLength encrypted_length = packet->encrypted_length;
// Termination packets are eventually owned by TimeWaitListManager.
// Others are deleted at the end of this call.
if (is_termination_packet) {
if (termination_packets_ == nullptr) {
termination_packets_.reset(
new std::vector<std::unique_ptr<QuicEncryptedPacket>>);
}
// Copy the buffer so it's owned in the future.
char* buffer_copy = CopyBuffer(*packet);
termination_packets_->emplace_back(
new QuicEncryptedPacket(buffer_copy, encrypted_length, true));
// This assures we won't try to write *forced* packets when blocked.
// Return true to stop processing.
if (HandleWriteBlocked()) {
return true;
}
}
DCHECK_LE(encrypted_length, kMaxPacketSize);
DCHECK_LE(encrypted_length, packet_generator_.GetCurrentMaxPacketLength());
QUIC_DVLOG(1) << ENDPOINT << "Sending packet " << packet_number << " : "
<< (IsRetransmittable(*packet) == HAS_RETRANSMITTABLE_DATA
? "data bearing "
: " ack only ")
<< ", encryption level: "
<< QuicUtils::EncryptionLevelToString(packet->encryption_level)
<< ", encrypted length:" << encrypted_length;
QUIC_DVLOG(2) << ENDPOINT << "packet(" << packet_number << "): " << std::endl
<< QuicTextUtils::HexDump(QuicStringPiece(
packet->encrypted_buffer, encrypted_length));
// Measure the RTT from before the write begins to avoid underestimating the
// min_rtt_, especially in cases where the thread blocks or gets swapped out
// during the WritePacket below.
QuicTime packet_send_time = clock_->Now();
if (supports_release_time_ && per_packet_options_ != nullptr) {
QuicTime next_release_time = sent_packet_manager_.GetNextReleaseTime();
QuicTime::Delta release_time_delay = QuicTime::Delta::Zero();
QuicTime now = packet_send_time;
if (next_release_time > now) {
release_time_delay = next_release_time - now;
// Set packet_send_time to the future to make the RTT estimation accurate.
packet_send_time = next_release_time;
}
per_packet_options_->release_time_delay = release_time_delay;
}
WriteResult result = writer_->WritePacket(
packet->encrypted_buffer, encrypted_length, self_address().host(),
peer_address(), per_packet_options_);
QUIC_HISTOGRAM_ENUM(
"QuicConnection.WritePacketStatus", result.status,
WRITE_STATUS_NUM_VALUES,
"Status code returned by writer_->WritePacket() in QuicConnection.");
if (IsWriteBlockedStatus(result.status)) {
// Ensure the writer is still write blocked, otherwise QUIC may continue
// trying to write when it will not be able to.
DCHECK(writer_->IsWriteBlocked());
visitor_->OnWriteBlocked();
// If the socket buffers the data, then the packet should not
// be queued and sent again, which would result in an unnecessary
// duplicate packet being sent. The helper must call OnCanWrite
// when the write completes, and OnWriteError if an error occurs.
if (result.status != WRITE_STATUS_BLOCKED_DATA_BUFFERED) {
return false;
}
}
// In some cases, an MTU probe can cause EMSGSIZE. This indicates that the
// MTU discovery is permanently unsuccessful.
if (IsMsgTooBig(result) && packet->retransmittable_frames.empty() &&
packet->encrypted_length > long_term_mtu_) {
mtu_discovery_target_ = 0;
mtu_discovery_alarm_->Cancel();
// The write failed, but the writer is not blocked, so return true.
return true;
}
if (IsWriteError(result.status)) {
OnWriteError(result.error_code);
QUIC_LOG_FIRST_N(ERROR, 10)
<< ENDPOINT << "failed writing " << encrypted_length
<< " bytes from host " << self_address().host().ToString()
<< " to address " << peer_address().ToString() << " with error code "
<< result.error_code;
return false;
}
if (debug_visitor_ != nullptr) {
// Pass the write result to the visitor.
debug_visitor_->OnPacketSent(*packet, packet->original_packet_number,
packet->transmission_type, packet_send_time);
}
if (IsRetransmittable(*packet) == HAS_RETRANSMITTABLE_DATA) {
if (!is_path_degrading_ && !path_degrading_alarm_->IsSet()) {
// This is the first retransmittable packet on the working path.
// Start the path degrading alarm to detect new path degrading.
SetPathDegradingAlarm();
}
if (GetQuicReloadableFlag(
quic_fix_time_of_first_packet_sent_after_receiving)) {
// Update |time_of_first_packet_sent_after_receiving_| if this is the
// first packet sent after the last packet was received. If it were
// updated on every sent packet, then sending into a black hole might
// never timeout.
if (time_of_first_packet_sent_after_receiving_ <
time_of_last_received_packet_) {
QUIC_RELOADABLE_FLAG_COUNT(
quic_fix_time_of_first_packet_sent_after_receiving);
time_of_first_packet_sent_after_receiving_ = packet_send_time;
}
} else {
// Only adjust the last sent time (for the purpose of tracking the idle
// timeout) if this is the first retransmittable packet sent after a
// packet is received. If it were updated on every sent packet, then
// sending into a black hole might never timeout.
if (time_of_first_packet_sent_after_receiving_ <=
time_of_last_received_packet_) {
time_of_first_packet_sent_after_receiving_ = packet_send_time;
}
}
}
MaybeSetMtuAlarm(packet_number);
QUIC_DVLOG(1) << ENDPOINT << "time we began writing last sent packet: "
<< packet_send_time.ToDebuggingValue();
bool reset_retransmission_alarm = sent_packet_manager_.OnPacketSent(
packet, packet->original_packet_number, packet_send_time,
packet->transmission_type, IsRetransmittable(*packet));
if (reset_retransmission_alarm || !retransmission_alarm_->IsSet()) {
SetRetransmissionAlarm();
}
SetPingAlarm();
// The packet number length must be updated after OnPacketSent, because it
// may change the packet number length in packet.
packet_generator_.UpdatePacketNumberLength(
sent_packet_manager_.GetLeastUnacked(),
sent_packet_manager_.EstimateMaxPacketsInFlight(max_packet_length()));
stats_.bytes_sent += result.bytes_written;
++stats_.packets_sent;
if (packet->transmission_type != NOT_RETRANSMISSION) {
stats_.bytes_retransmitted += result.bytes_written;
++stats_.packets_retransmitted;
}
return true;
}
void QuicConnection::FlushPackets() {
if (GetQuicRestartFlag(quic_check_blocked_writer_for_blockage) &&
!connected_) {
QUIC_RESTART_FLAG_COUNT_N(quic_check_blocked_writer_for_blockage, 5, 6);
return;
}
if (!writer_->IsBatchMode()) {
return;
}
if (HandleWriteBlocked()) {
QUIC_DLOG(INFO) << ENDPOINT << "FlushPackets called while blocked.";
return;
}
WriteResult result = writer_->Flush();
if (HandleWriteBlocked()) {
DCHECK_EQ(WRITE_STATUS_BLOCKED, result.status)
<< "Unexpected flush result:" << result;
QUIC_DLOG(INFO) << ENDPOINT << "Write blocked in FlushPackets.";
return;
}
if (IsWriteError(result.status)) {
OnWriteError(result.error_code);
}
}
bool QuicConnection::IsMsgTooBig(const WriteResult& result) {
return (result.status == WRITE_STATUS_MSG_TOO_BIG) ||
(IsWriteError(result.status) && result.error_code == QUIC_EMSGSIZE);
}
bool QuicConnection::ShouldDiscardPacket(const SerializedPacket& packet) {
if (!connected_) {
QUIC_DLOG(INFO) << ENDPOINT
<< "Not sending packet as connection is disconnected.";
return true;
}
QuicPacketNumber packet_number = packet.packet_number;
if (encryption_level_ == ENCRYPTION_FORWARD_SECURE &&
packet.encryption_level == ENCRYPTION_NONE) {
// Drop packets that are NULL encrypted since the peer won't accept them
// anymore.
QUIC_DLOG(INFO) << ENDPOINT
<< "Dropping NULL encrypted packet: " << packet_number
<< " since the connection is forward secure.";
return true;
}
return false;
}
void QuicConnection::OnWriteError(int error_code) {
if (write_error_occurred_) {
// A write error already occurred. The connection is being closed.
return;
}
write_error_occurred_ = true;
const QuicString error_details = QuicStrCat(
"Write failed with error: ", error_code, " (", strerror(error_code), ")");
QUIC_LOG_FIRST_N(ERROR, 2) << ENDPOINT << error_details;
switch (error_code) {
case QUIC_EMSGSIZE:
CloseConnection(
QUIC_PACKET_WRITE_ERROR, error_details,
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET_WITH_NO_ACK);
break;
default:
// We can't send an error as the socket is presumably borked.
if (transport_version() > QUIC_VERSION_43) {
QUIC_CODE_COUNT(quic_tear_down_local_connection_on_write_error_ietf);
} else {
QUIC_CODE_COUNT(
quic_tear_down_local_connection_on_write_error_non_ietf);
}
TearDownLocalConnectionState(QUIC_PACKET_WRITE_ERROR, error_details,
ConnectionCloseSource::FROM_SELF);
}
}
char* QuicConnection::GetPacketBuffer() {
return writer_->GetNextWriteLocation(self_address().host(), peer_address());
}
void QuicConnection::OnSerializedPacket(SerializedPacket* serialized_packet) {
if (serialized_packet->encrypted_buffer == nullptr) {
// We failed to serialize the packet, so close the connection.
// TearDownLocalConnectionState does not send close packet, so no infinite
// loop here.
// TODO(ianswett): This is actually an internal error, not an
// encryption failure.
if (transport_version() > QUIC_VERSION_43) {
QUIC_CODE_COUNT(
quic_tear_down_local_connection_on_serialized_packet_ietf);
} else {
QUIC_CODE_COUNT(
quic_tear_down_local_connection_on_serialized_packet_non_ietf);
}
TearDownLocalConnectionState(
QUIC_ENCRYPTION_FAILURE,
"Serialized packet does not have an encrypted buffer.",
ConnectionCloseSource::FROM_SELF);
return;
}
if (serialized_packet->retransmittable_frames.empty() &&
!serialized_packet->original_packet_number.IsInitialized()) {
// Increment consecutive_num_packets_with_no_retransmittable_frames_ if
// this packet is a new transmission with no retransmittable frames.
++consecutive_num_packets_with_no_retransmittable_frames_;
} else {
consecutive_num_packets_with_no_retransmittable_frames_ = 0;
}
SendOrQueuePacket(serialized_packet);
}
void QuicConnection::OnUnrecoverableError(QuicErrorCode error,
const QuicString& error_details,
ConnectionCloseSource source) {
// The packet creator or generator encountered an unrecoverable error: tear
// down local connection state immediately.
if (transport_version() > QUIC_VERSION_43) {
QUIC_CODE_COUNT(
quic_tear_down_local_connection_on_unrecoverable_error_ietf);
} else {
QUIC_CODE_COUNT(
quic_tear_down_local_connection_on_unrecoverable_error_non_ietf);
}
TearDownLocalConnectionState(error, error_details, source);
}
void QuicConnection::OnCongestionChange() {
visitor_->OnCongestionWindowChange(clock_->ApproximateNow());
// Uses the connection's smoothed RTT. If zero, uses initial_rtt.
QuicTime::Delta rtt = sent_packet_manager_.GetRttStats()->smoothed_rtt();
if (rtt.IsZero()) {
rtt = sent_packet_manager_.GetRttStats()->initial_rtt();
}
if (debug_visitor_ != nullptr) {
debug_visitor_->OnRttChanged(rtt);
}
}
void QuicConnection::OnPathMtuIncreased(QuicPacketLength packet_size) {
if (packet_size > max_packet_length()) {
SetMaxPacketLength(packet_size);
}
}
void QuicConnection::OnHandshakeComplete() {
sent_packet_manager_.SetHandshakeConfirmed();
if (sent_packet_manager_.unacked_packets().use_uber_loss_algorithm()) {
// This may have changed the retransmission timer, so re-arm it.
SetRetransmissionAlarm();
}
// The client should immediately ack the SHLO to confirm the handshake is
// complete with the server.
if (perspective_ == Perspective::IS_CLIENT && !ack_queued_ &&
ack_frame_updated()) {
ack_alarm_->Update(clock_->ApproximateNow(), QuicTime::Delta::Zero());
}
}
void QuicConnection::SendOrQueuePacket(SerializedPacket* packet) {
// The caller of this function is responsible for checking CanWrite().
if (packet->encrypted_buffer == nullptr) {
QUIC_BUG << "packet.encrypted_buffer == nullptr in to SendOrQueuePacket";
return;
}
// If there are already queued packets, queue this one immediately to ensure
// it's written in sequence number order.
if (!queued_packets_.empty() || !WritePacket(packet)) {
// Take ownership of the underlying encrypted packet.
packet->encrypted_buffer = CopyBuffer(*packet);
queued_packets_.push_back(*packet);
packet->retransmittable_frames.clear();
}
ClearSerializedPacket(packet);
}
void QuicConnection::OnPingTimeout() {
if (!retransmission_alarm_->IsSet()) {
visitor_->SendPing();
}
}
void QuicConnection::SendAck() {
ack_alarm_->Cancel();
ack_queued_ = false;
stop_waiting_count_ = 0;
num_retransmittable_packets_received_since_last_ack_sent_ = 0;
num_packets_received_since_last_ack_sent_ = 0;
packet_generator_.SetShouldSendAck(!no_stop_waiting_frames_);
if (consecutive_num_packets_with_no_retransmittable_frames_ <
max_consecutive_num_packets_with_no_retransmittable_frames_) {
return;
}
consecutive_num_packets_with_no_retransmittable_frames_ = 0;
if (packet_generator_.HasRetransmittableFrames() ||
visitor_->WillingAndAbleToWrite()) {
// There are pending retransmittable frames.
return;
}
visitor_->OnAckNeedsRetransmittableFrame();
}
void QuicConnection::OnPathDegradingTimeout() {
is_path_degrading_ = true;
visitor_->OnPathDegrading();
}
void QuicConnection::OnRetransmissionTimeout() {
DCHECK(!sent_packet_manager_.unacked_packets().empty());
if (close_connection_after_five_rtos_ &&
sent_packet_manager_.GetConsecutiveRtoCount() >= 4) {
// Close on the 5th consecutive RTO, so after 4 previous RTOs have occurred.
CloseConnection(QUIC_TOO_MANY_RTOS, "5 consecutive retransmission timeouts",
ConnectionCloseBehavior::SEND_CONNECTION_CLOSE_PACKET);
return;
}
sent_packet_manager_.OnRetransmissionTimeout();
WriteIfNotBlocked();
// A write failure can result in the connection being closed, don't attempt to
// write further packets, or to set alarms.
if (!connected_) {
return;
}
// In the TLP case, the SentPacketManager gives the connection the opportunity
// to send new data before retransmitting.
if (sent_packet_manager_.MaybeRetransmitTailLossProbe()) {
// Send the pending retransmission now that it's been queued.
WriteIfNotBlocked();
}
// Ensure the retransmission alarm is always set if there are unacked packets
// and nothing waiting to be sent.
// This happens if the loss algorithm invokes a timer based loss, but the
// packet doesn't need to be retransmitted.
if (!HasQueuedData() && !retransmission_alarm_->IsSet()) {
SetRetransmissionAlarm();
}
}
void QuicConnection::SetEncrypter(EncryptionLevel level,
std::unique_ptr<QuicEncrypter> encrypter) {
packet_generator_.SetEncrypter(level, std::move(encrypter));
}
void QuicConnection::SetDiversificationNonce(
const DiversificationNonce& nonce) {
DCHECK_EQ(Perspective::IS_SERVER, perspective_);
packet_generator_.SetDiversificationNonce(nonce);
}
void QuicConnection::SetDefaultEncryptionLevel(EncryptionLevel level) {
if (level != encryption_level_ && packet_generator_.HasQueuedFrames()) {
// Flush all queued frames when encryption level changes.
ScopedPacketFlusher flusher(this, NO_ACK);
packet_generator_.FlushAllQueuedFrames();
}
encryption_level_ = level;
packet_generator_.set_encryption_level(level);
}
void QuicConnection::SetDecrypter(EncryptionLevel level,
std::unique_ptr<QuicDecrypter> decrypter) {
framer_.SetDecrypter(level, std::move(decrypter));
if (!undecryptable_packets_.empty() &&
!process_undecryptable_packets_alarm_->IsSet()) {
process_undecryptable_packets_alarm_->Set(clock_->ApproximateNow());
}
}
void QuicConnection::SetAlternativeDecrypter(
EncryptionLevel level,
std::unique_ptr<QuicDecrypter> decrypter,
bool latch_once_used) {
framer_.SetAlternativeDecrypter(level, std::move(decrypter), latch_once_used);
if (!undecryptable_packets_.empty() &&
!process_undecryptable_packets_alarm_->IsSet()) {
process_undecryptable_packets_alarm_->Set(clock_->ApproximateNow());
}
}
const QuicDecrypter* QuicConnection::decrypter() const {
return framer_.decrypter();
}
const QuicDecrypter* QuicConnection::alternative_decrypter() const {
return framer_.alternative_decrypter();
}
void QuicConnection::QueueUndecryptablePacket(
const QuicEncryptedPacket& packet) {
QUIC_DVLOG(1) << ENDPOINT << "Queueing undecryptable packet.";
undecryptable_packets_.push_back(packet.Clone());
}
void QuicConnection::MaybeProcessUndecryptablePackets() {
process_undecryptable_packets_alarm_->Cancel();
if (undecryptable_packets_.empty() || encryption_level_ == ENCRYPTION_NONE) {
return;
}
while (connected_ && !undecryptable_packets_.empty()) {
// Making sure there is no pending frames when processing next undecrypted
// packet because the queued ack frame may change.
packet_generator_.FlushAllQueuedFrames();
if (!connected_) {
return;
}
QUIC_DVLOG(1) << ENDPOINT << "Attempting to process undecryptable packet";
QuicEncryptedPacket* packet = undecryptable_packets_.front().get();
if (!framer_.ProcessPacket(*packet) &&
framer_.error() == QUIC_DECRYPTION_FAILURE) {