Land Recent QUIC Changes.

Fix a flaky EndToEndTest where packet loss is possible with a large
buffer that may overrun.

Merge internal change: 66821654
https://codereview.chromium.org/280383003/


QUIC change to invoke SetNotPending for retransmissions in
MarkForRetransmission and a minor cleanup in UnackedPacketMap.

Merge internal change: 66729074
https://codereview.chromium.org/278823005/


Remove unused IsHandshake argument to QuicConnection::OnCanWrite.

Merge internal change: 66652078
https://codereview.chromium.org/286563003/


Revert of Fix a bug where if a crypto packet is spuriously retransmitted
and thenneutered, it remains in missing_packets until the end of the
connection.

Reverted chromium cl: https://codereview.chromium.org/270213002/

Rollback of Merge internal change: 66262890.

*** Reason for rollback ***

 This appears to have caused a major QUIC loadtest regression on
 internal server.

Additionally, it's wrong in some other subtle ways, so rolling it back
and trying a new approach.

*** Original change description ***

 Fix a bug where if a crypto packet is spuriously retransmitted and then
 neutered, it remains in missing_packets until the end of the
 connection.

***

Merge internal change: 66615668
https://codereview.chromium.org/287583002/


Revert of Minor QUIC cleanup to combine QuicUnackedPacketMap's
RemovePacket and NeuterPacket into RemoveOrNeuterPacket.

Rollback of merge internal change: 66306887
Rollback of Chromium's CL https://codereview.chromium.org/271443008/

*** Reason for rollback ***

 After further issues with the QUIC handshake packets, rolling back this
 change in favor of a different approach.

*** Original change description ***

 Minor QUIC cleanup to combine QuicUnackedPacketMap's RemovePacket and
 NeuterPacket into RemoveOrNeuterPacket.

 This prevents potential bugs where a packet can be set not pending, then
 neuter'd, which cuases the packet to never be removed from the unacked
 packet map.

***

Merge internal change: 66573334
https://codereview.chromium.org/280753002/


Merging end_to_end_test.cc with the internal source tree.

+ Fixed comments (MB for megabytes and 256KB per sec instead of 1Mbit).
+ Made LargePostWithPacketLossAndBlockedSocket and
  LargePostNoPacketLossWithDelayAndReordering to be in the same
  order as the internal source tree.

Merge internal change: 66556306
https://codereview.chromium.org/283683002/


Allows QUIC connections to remain established even though the peer's
port has changed. Should make things a bit better for clients
experiencing NAT rebindign while talking QUIC to Bandaid.

QUIC connection port migration. Protected behind
FLAGS_quic_allow_port_migration

Merge internal change: 66523001
https://codereview.chromium.org/279453004/


Remove unnecessary methods from QuicUnackedPacketMap and a minor
QuicSentPacketManagerTest fix as a result.

Merge internal change: 66427453
https://codereview.chromium.org/287443003/

R=rch@chromium.org

Review URL: https://codereview.chromium.org/283693002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270046 0039d316-1c4b-4281-b951-d872f2087c98

net/quic/congestion_control/tcp_loss_algorithm_test.cc

@@ -171,7 +171,7 @@ TEST_F(TcpLossAlgorithmTest, DontEarlyRetransmitNeuteredPacket) {
   // Early retransmit when the final packet gets acked and the first is nacked.
   unacked_packets_.SetNotPending(2);
   unacked_packets_.NackPacket(1, 1);
-  unacked_packets_.NeuterIfPendingOrRemovePacket(1);
+  unacked_packets_.NeuterPacket(1);
   VerifyLosses(2, NULL, 0);
   EXPECT_EQ(QuicTime::Zero(), loss_algorithm_.GetLossTimeout());
 }

net/quic/quic_connection.cc

@@ -204,6 +204,7 @@ QuicConnection::QuicConnection(QuicConnectionId connection_id,
       random_generator_(helper->GetRandomGenerator()),
       connection_id_(connection_id),
       peer_address_(address),
+      migrating_peer_port_(0),
       last_packet_revived_(false),
       last_size_(0),
       last_decrypted_packet_level_(ENCRYPTION_NONE),
@@ -234,7 +235,10 @@ QuicConnection::QuicConnection(QuicConnectionId connection_id,
       version_negotiation_state_(START_NEGOTIATION),
       is_server_(is_server),
       connected_(true),
-      address_migrating_(false),
+      peer_ip_changed_(false),
+      peer_port_changed_(false),
+      self_ip_changed_(false),
+      self_port_changed_(false),
       max_flow_control_receive_window_bytes_(
           max_flow_control_receive_window_bytes) {
   if (max_flow_control_receive_window_bytes_ < kDefaultFlowControlSendWindow) {
@@ -1105,18 +1109,7 @@ void QuicConnection::ProcessUdpPacket(const IPEndPoint& self_address,
   last_packet_revived_ = false;
   last_size_ = packet.length();
 
-  address_migrating_ = false;
-
-  if (peer_address_.address().empty()) {
-    peer_address_ = peer_address;
-  }
-  if (self_address_.address().empty()) {
-    self_address_ = self_address;
-  }
-
-  if (!(peer_address == peer_address_ && self_address == self_address_)) {
-    address_migrating_ = true;
-  }
+  CheckForAddressMigration(self_address, peer_address);
 
   stats_.bytes_received += packet.length();
   ++stats_.packets_received;
@@ -1141,20 +1134,45 @@ void QuicConnection::ProcessUdpPacket(const IPEndPoint& self_address,
   SetPingAlarm();
 }
 
+void QuicConnection::CheckForAddressMigration(
+    const IPEndPoint& self_address, const IPEndPoint& peer_address) {
+  peer_ip_changed_ = false;
+  peer_port_changed_ = false;
+  self_ip_changed_ = false;
+  self_port_changed_ = false;
+
+  if (peer_address_.address().empty()) {
+    peer_address_ = peer_address;
+  }
+  if (self_address_.address().empty()) {
+    self_address_ = self_address;
+  }
+
+  if (!peer_address.address().empty() && !peer_address_.address().empty()) {
+    peer_ip_changed_ = (peer_address.address() != peer_address_.address());
+    peer_port_changed_ = (peer_address.port() != peer_address_.port());
+
+    // Store in case we want to migrate connection in ProcessValidatedPacket.
+    migrating_peer_port_ = peer_address.port();
+  }
+
+  if (!self_address.address().empty() && !self_address_.address().empty()) {
+    self_ip_changed_ = (self_address.address() != self_address_.address());
+    self_port_changed_ = (self_address.port() != self_address_.port());
+  }
+}
+
 void QuicConnection::OnCanWrite() {
   DCHECK(!writer_->IsWriteBlocked());
 
   WriteQueuedPackets();
   WritePendingRetransmissions();
 
-  IsHandshake pending_handshake = visitor_->HasPendingHandshake() ?
-      IS_HANDSHAKE : NOT_HANDSHAKE;
   // 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(NOT_RETRANSMISSION, HAS_RETRANSMITTABLE_DATA,
-                pending_handshake)) {
+  if (!CanWrite(NOT_RETRANSMISSION, HAS_RETRANSMITTABLE_DATA)) {
     return;
   }
 
@@ -1166,11 +1184,8 @@ void QuicConnection::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.
-  pending_handshake = visitor_->HasPendingHandshake() ?
-      IS_HANDSHAKE : NOT_HANDSHAKE;
   if (visitor_->HasPendingWrites() && !resume_writes_alarm_->IsSet() &&
-      CanWrite(NOT_RETRANSMISSION, HAS_RETRANSMITTABLE_DATA,
-               pending_handshake)) {
+      CanWrite(NOT_RETRANSMISSION, 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.
@@ -1185,12 +1200,23 @@ void QuicConnection::WriteIfNotBlocked() {
 }
 
 bool QuicConnection::ProcessValidatedPacket() {
-  if (address_migrating_) {
+  if ((!FLAGS_quic_allow_port_migration && peer_port_changed_) ||
+      peer_ip_changed_ || self_ip_changed_ || self_port_changed_) {
     SendConnectionCloseWithDetails(
         QUIC_ERROR_MIGRATING_ADDRESS,
-        "Address migration is not yet a supported feature");
+        "IP or port migration is not yet a supported feature");
     return false;
   }
+
+  // Port migration is supported, do it now if port has changed.
+  if (FLAGS_quic_allow_port_migration &&
+      peer_port_changed_) {
+    DVLOG(1) << ENDPOINT << "Peer's port changed from "
+             << peer_address_.port() << " to " << migrating_peer_port_
+             << ", migrating connection.";
+    peer_address_ = IPEndPoint(peer_address_.address(), migrating_peer_port_);
+  }
+
   time_of_last_received_packet_ = clock_->Now();
   DVLOG(1) << ENDPOINT << "time of last received packet: "
            << time_of_last_received_packet_.ToDebuggingValue();
@@ -1230,8 +1256,7 @@ void QuicConnection::WritePendingRetransmissions() {
     const QuicSentPacketManager::PendingRetransmission pending =
         sent_packet_manager_.NextPendingRetransmission();
     if (GetPacketType(&pending.retransmittable_frames) == NORMAL &&
-        !CanWrite(pending.transmission_type, HAS_RETRANSMITTABLE_DATA,
-                  pending.retransmittable_frames.HasCryptoHandshake())) {
+        !CanWrite(pending.transmission_type, HAS_RETRANSMITTABLE_DATA)) {
       break;
     }
 
@@ -1269,8 +1294,8 @@ void QuicConnection::RetransmitUnackedPackets(
   WriteIfNotBlocked();
 }
 
-void QuicConnection::DiscardUnencryptedPackets() {
-  sent_packet_manager_.DiscardUnencryptedPackets();
+void QuicConnection::NeuterUnencryptedPackets() {
+  sent_packet_manager_.NeuterUnencryptedPackets();
   // This may have changed the retransmission timer, so re-arm it.
   retransmission_alarm_->Cancel();
   QuicTime retransmission_time = sent_packet_manager_.GetRetransmissionTime();
@@ -1289,12 +1314,11 @@ bool QuicConnection::ShouldGeneratePacket(
     return true;
   }
 
-  return CanWrite(transmission_type, retransmittable, handshake);
+  return CanWrite(transmission_type, retransmittable);
 }
 
 bool QuicConnection::CanWrite(TransmissionType transmission_type,
-                              HasRetransmittableData retransmittable,
-                              IsHandshake handshake) {
+                              HasRetransmittableData retransmittable) {
   if (writer_->IsWriteBlocked()) {
     visitor_->OnWriteBlocked();
     return false;
@@ -1342,8 +1366,7 @@ bool QuicConnection::WritePacket(QueuedPacket packet) {
   // TODO(ianswett): The congestion control should have been consulted before
   // serializing the packet, so this could be turned into a LOG_IF(DFATAL).
   if (packet.type == NORMAL && !CanWrite(packet.transmission_type,
-                                         packet.retransmittable,
-                                         packet.handshake)) {
+                                         packet.retransmittable)) {
     return false;
   }
 

net/quic/quic_connection.h

@@ -417,9 +417,9 @@ class NET_EXPORT_PRIVATE QuicConnection
   // initially encrypted packets when the initial encrypter changes.
   void RetransmitUnackedPackets(RetransmissionType retransmission_type);
 
-  // Calls |sent_packet_manager_|'s DiscardUnencryptedPackets. Used when the
+  // Calls |sent_packet_manager_|'s NeuterUnencryptedPackets. Used when the
   // connection becomes forward secure and hasn't received acks for all packets.
-  void DiscardUnencryptedPackets();
+  void NeuterUnencryptedPackets();
 
   // Changes the encrypter used for level |level| to |encrypter|. The function
   // takes ownership of |encrypter|.
@@ -458,8 +458,7 @@ class NET_EXPORT_PRIVATE QuicConnection
   }
 
   bool CanWrite(TransmissionType transmission_type,
-                HasRetransmittableData retransmittable,
-                IsHandshake handshake);
+                HasRetransmittableData retransmittable);
 
   uint32 max_flow_control_receive_window_bytes() const {
     return max_flow_control_receive_window_bytes_;
@@ -611,6 +610,11 @@ class NET_EXPORT_PRIVATE QuicConnection
   // Sets the ping alarm to the appropriate value, if any.
   void SetPingAlarm();
 
+  // On arrival of a new packet, checks to see if the socket addresses have
+  // changed since the last packet we saw on this connection.
+  void CheckForAddressMigration(const IPEndPoint& self_address,
+                                const IPEndPoint& peer_address);
+
   QuicFramer framer_;
   QuicConnectionHelperInterface* helper_;  // Not owned.
   QuicPacketWriter* writer_;  // Not owned.
@@ -623,6 +627,8 @@ class NET_EXPORT_PRIVATE QuicConnection
   // client.
   IPEndPoint self_address_;
   IPEndPoint peer_address_;
+  // Used to store latest peer port to possibly migrate to later.
+  int migrating_peer_port_;
 
   bool last_packet_revived_;  // True if the last packet was revived from FEC.
   size_t last_size_;  // Size of the last received packet.
@@ -734,9 +740,21 @@ class NET_EXPORT_PRIVATE QuicConnection
   // close.
   bool connected_;
 
-  // Set to true if the udp packet headers have a new self or peer address.
-  // This is checked later on validating a data or version negotiation packet.
-  bool address_migrating_;
+  // Set to true if the UDP packet headers have a new IP address for the peer.
+  // If true, do not perform connection migration.
+  bool peer_ip_changed_;
+
+  // Set to true if the UDP packet headers have a new port for the peer.
+  // If true, and the IP has not changed, then we can migrate the connection.
+  bool peer_port_changed_;
+
+  // Set to true if the UDP packet headers are addressed to a different IP.
+  // We do not support connection migration when the self IP changed.
+  bool self_ip_changed_;
+
+  // Set to true if the UDP packet headers are addressed to a different port.
+  // If true, and the IP has not changed, then we can migrate the connection.
+  bool self_port_changed_;
 
   // If non-empty this contains the set of versions received in a
   // version negotiation packet.

net/quic/quic_flags.cc

@@ -33,5 +33,9 @@ bool FLAGS_enable_quic_stream_flow_control_2 = true;
 bool FLAGS_enable_quic_connection_flow_control = true;
 
 bool FLAGS_quic_allow_oversized_packets_for_test = false;
+
 // When true, the use time based loss detection instead of nack.
 bool FLAGS_quic_use_time_loss_detection = false;
+
+// If true, allow port migration of established QUIC connections.
+bool FLAGS_quic_allow_port_migration = true;

net/quic/quic_flags.h

@@ -13,5 +13,6 @@ NET_EXPORT_PRIVATE extern bool FLAGS_enable_quic_stream_flow_control_2;
 NET_EXPORT_PRIVATE extern bool FLAGS_enable_quic_connection_flow_control;
 NET_EXPORT_PRIVATE extern bool FLAGS_quic_allow_oversized_packets_for_test;
 NET_EXPORT_PRIVATE extern bool FLAGS_quic_use_time_loss_detection;
+NET_EXPORT_PRIVATE extern bool FLAGS_quic_allow_port_migration;
 
 #endif  // NET_QUIC_QUIC_FLAGS_H_

net/quic/quic_reliable_client_stream.cc

@@ -94,8 +94,7 @@ void QuicReliableClientStream::OnError(int error) {
 
 bool QuicReliableClientStream::CanWrite(const CompletionCallback& callback) {
   bool can_write =  session()->connection()->CanWrite(
-      NOT_RETRANSMISSION, HAS_RETRANSMITTABLE_DATA,
-      id() == kCryptoStreamId ? IS_HANDSHAKE : NOT_HANDSHAKE);
+      NOT_RETRANSMISSION, HAS_RETRANSMITTABLE_DATA);
   if (!can_write) {
     session()->MarkWriteBlocked(id(), EffectivePriority());
     DCHECK(callback_.is_null());

net/quic/quic_sent_packet_manager.cc

@@ -244,26 +244,27 @@ void QuicSentPacketManager::RetransmitUnackedPackets(
     // numbers with no connection to the previous ones.
     if (frames != NULL && (retransmission_type == ALL_PACKETS ||
                            frames->encryption_level() == ENCRYPTION_INITIAL)) {
-      unacked_packets_.SetNotPending(unacked_it->first);
       MarkForRetransmission(unacked_it->first, ALL_UNACKED_RETRANSMISSION);
     }
     ++unacked_it;
   }
 }
 
-void QuicSentPacketManager::DiscardUnencryptedPackets() {
+void QuicSentPacketManager::NeuterUnencryptedPackets() {
   QuicUnackedPacketMap::const_iterator unacked_it = unacked_packets_.begin();
   while (unacked_it != unacked_packets_.end()) {
     const RetransmittableFrames* frames =
         unacked_it->second.retransmittable_frames;
     if (frames != NULL && frames->encryption_level() == ENCRYPTION_NONE) {
-      // Once you're forward secure, no unencrypted packets will be sent.
-      // Additionally, it's likely the peer will be forward secure, and no acks
-      // for these packets will be received, so mark the packet as handled.
+      // Since once you're forward secure, no unencrypted packets will be sent,
+      // crypto or otherwise. Unencrypted packets are neutered and abandoned, to
+      // ensure they are not retransmitted or considered lost from a congestion
+      // control perspective.
       pending_retransmissions_.erase(unacked_it->first);
-      unacked_it = MarkPacketHandled(unacked_it->first,
-                                     QuicTime::Delta::Zero());
-      continue;
+      // TODO(ianswett): This may cause packets to linger forever in the
+      // UnackedPacketMap.
+      unacked_packets_.NeuterPacket(unacked_it->first);
+      unacked_packets_.SetNotPending(unacked_it->first);
     }
     ++unacked_it;
   }
@@ -275,6 +276,9 @@ void QuicSentPacketManager::MarkForRetransmission(
   const TransmissionInfo& transmission_info =
       unacked_packets_.GetTransmissionInfo(sequence_number);
   LOG_IF(DFATAL, transmission_info.retransmittable_frames == NULL);
+  if (transmission_type != TLP_RETRANSMISSION) {
+    unacked_packets_.SetNotPending(sequence_number);
+  }
   // TODO(ianswett): Currently the RTO can fire while there are pending NACK
   // retransmissions for the same data, which is not ideal.
   if (ContainsKey(pending_retransmissions_, sequence_number)) {
@@ -337,7 +341,11 @@ void QuicSentPacketManager::MarkPacketRevived(
         sequence_number, delta_largest_observed);
   }
 
-  unacked_packets_.NeuterIfPendingOrRemovePacket(sequence_number);
+  if (!transmission_info.pending) {
+    unacked_packets_.RemovePacket(sequence_number);
+  } else {
+    unacked_packets_.NeuterPacket(sequence_number);
+  }
 }
 
 QuicUnackedPacketMap::const_iterator QuicSentPacketManager::MarkPacketHandled(
@@ -349,7 +357,7 @@ QuicUnackedPacketMap::const_iterator QuicSentPacketManager::MarkPacketHandled(
   }
   const TransmissionInfo& transmission_info =
       unacked_packets_.GetTransmissionInfo(sequence_number);
-  // If this packet is pending, remove it and inform the send algorithm.
+  // If this packet is pending, remove it.
   if (transmission_info.pending) {
     unacked_packets_.SetNotPending(sequence_number);
   }
@@ -358,6 +366,13 @@ QuicUnackedPacketMap::const_iterator QuicSentPacketManager::MarkPacketHandled(
   SequenceNumberSet::reverse_iterator all_transmissions_it =
       all_transmissions.rbegin();
   QuicPacketSequenceNumber newest_transmission = *all_transmissions_it;
+  // Remove the most recent packet, if it is pending retransmission.
+  pending_retransmissions_.erase(newest_transmission);
+
+  // Two cases for MarkPacketHandled:
+  // 1) Handle the most recent or a crypto packet, so remove all transmissions.
+  // 2) Handle old transmission, keep all other pending transmissions,
+  //    but disassociate them from one another.
   if (newest_transmission != sequence_number) {
     stats_->bytes_spuriously_retransmitted += transmission_info.bytes_sent;
     ++stats_->packets_spuriously_retransmitted;
@@ -368,19 +383,25 @@ QuicUnackedPacketMap::const_iterator QuicSentPacketManager::MarkPacketHandled(
   ack_notifier_manager_.OnPacketAcked(newest_transmission,
                                       delta_largest_observed);
 
+  // TODO(ianswett): Instead of handling all crypto packets in a special way,
+  // only handle NULL encrypted packets in a special way.
   bool has_crypto_handshake = HasCryptoHandshake(
       unacked_packets_.GetTransmissionInfo(newest_transmission));
   while (all_transmissions_it != all_transmissions.rend()) {
     QuicPacketSequenceNumber previous_transmission = *all_transmissions_it;
-    // If this packet was marked for retransmission, don't bother retransmitting
-    // it anymore.
-    pending_retransmissions_.erase(previous_transmission);
+    const TransmissionInfo& transmission_info =
+        unacked_packets_.GetTransmissionInfo(previous_transmission);
+    DCHECK(!ContainsKey(pending_retransmissions_, previous_transmission));
     if (has_crypto_handshake) {
       // If it's a crypto handshake packet, discard it and all retransmissions,
       // since they won't be acked now that one has been processed.
       unacked_packets_.SetNotPending(previous_transmission);
     }
-    unacked_packets_.NeuterIfPendingOrRemovePacket(previous_transmission);
+    if (!transmission_info.pending) {
+      unacked_packets_.RemovePacket(previous_transmission);
+    } else {
+      unacked_packets_.NeuterPacket(previous_transmission);
+    }
     ++all_transmissions_it;
   }
 
@@ -490,8 +511,6 @@ void QuicSentPacketManager::RetransmitCryptoPackets() {
     }
     packet_retransmitted = true;
     MarkForRetransmission(sequence_number, HANDSHAKE_RETRANSMISSION);
-    // Abandon all the crypto retransmissions now so they're not lost later.
-    unacked_packets_.SetNotPending(sequence_number);
   }
   DCHECK(packet_retransmitted) << "No crypto packets found to retransmit.";
 }
@@ -529,10 +548,11 @@ void QuicSentPacketManager::RetransmitAllPackets() {
   bool packets_retransmitted = false;
   for (QuicUnackedPacketMap::const_iterator it = unacked_packets_.begin();
        it != unacked_packets_.end(); ++it) {
-    unacked_packets_.SetNotPending(it->first);
     if (it->second.retransmittable_frames != NULL) {
       packets_retransmitted = true;
       MarkForRetransmission(it->first, RTO_RETRANSMISSION);
+    } else {
+      unacked_packets_.SetNotPending(it->first);
     }
   }
 
@@ -582,16 +602,17 @@ void QuicSentPacketManager::InvokeLossDetection(QuicTime time) {
     // until it's known whether the FEC packet arrived.
     ++stats_->packets_lost;
     packets_lost_[sequence_number] = transmission_info;
-    unacked_packets_.SetNotPending(sequence_number);
+    DVLOG(1) << ENDPOINT << "Lost packet " << sequence_number;
 
     if (transmission_info.retransmittable_frames != NULL) {
       MarkForRetransmission(sequence_number, LOSS_RETRANSMISSION);
     } else {
       // Since we will not retransmit this, we need to remove it from
       // unacked_packets_.   This is either the current transmission of
-      // a packet whose previous transmission has been acked, or it
-      // is a packet that has been TLP retransmitted.
-      unacked_packets_.NeuterIfPendingOrRemovePacket(sequence_number);
+      // a packet whose previous transmission has been acked, a packet that has
+      // been TLP retransmitted, or an FEC packet.
+      unacked_packets_.SetNotPending(sequence_number);
+      unacked_packets_.RemovePacket(sequence_number);
     }
   }
 }
@@ -663,7 +684,7 @@ const QuicTime QuicSentPacketManager::GetRetransmissionTime() const {
       // Base the updated timer on the send time of the last packet.
       const QuicTime sent_time = unacked_packets_.GetLastPacketSentTime();
       const QuicTime tlp_time = sent_time.Add(GetTailLossProbeDelay());
-      // Ensure the tlp timer never gets set to a time in the past.
+      // Ensure the TLP timer never gets set to a time in the past.
       return QuicTime::Max(clock_->ApproximateNow(), tlp_time);
     }
     case RTO_MODE: {

net/quic/quic_sent_packet_manager.h

@@ -96,7 +96,7 @@ class NET_EXPORT_PRIVATE QuicSentPacketManager {
 
   // Removes the retransmittable frames from all unencrypted packets to ensure
   // they don't get retransmitted.
-  void DiscardUnencryptedPackets();
+  void NeuterUnencryptedPackets();
 
   // Returns true if the unacked packet |sequence_number| has retransmittable
   // frames.  This will only return false if the packet has been acked, if a

net/quic/quic_sent_packet_manager_test.cc

@@ -69,22 +69,13 @@ class QuicSentPacketManagerTest : public ::testing::TestWithParam<bool> {
 
   void VerifyRetransmittablePackets(QuicPacketSequenceNumber* packets,
                                     size_t num_packets) {
-    SequenceNumberSet unacked =
-        QuicSentPacketManagerPeer::GetUnackedPackets(&manager_);
-    for (size_t i = 0; i < num_packets; ++i) {
-      EXPECT_TRUE(ContainsKey(unacked, packets[i])) << packets[i];
-    }
-    size_t num_retransmittable = 0;
-    for (SequenceNumberSet::const_iterator it = unacked.begin();
-         it != unacked.end(); ++it) {
-      if (manager_.HasRetransmittableFrames(*it)) {
-        ++num_retransmittable;
-      }
-    }
     EXPECT_EQ(num_packets,
               QuicSentPacketManagerPeer::GetNumRetransmittablePackets(
                   &manager_));
-    EXPECT_EQ(num_packets, num_retransmittable);
+    for (size_t i = 0; i < num_packets; ++i) {
+      EXPECT_TRUE(manager_.HasRetransmittableFrames(packets[i]))
+          << " packets[" << i << "]:" << packets[i];
+    }
   }
 
   void ExpectAck(QuicPacketSequenceNumber largest_observed) {
@@ -125,15 +116,18 @@ class QuicSentPacketManagerTest : public ::testing::TestWithParam<bool> {
                                   Pointwise(KeyEq(), lost_vector)));
   }
 
+  // Retransmits a packet as though it was a TLP retransmission, because TLP
+  // leaves the |old_sequence_number| pending.
+  // TODO(ianswett): Test with transmission types besides TLP.
   void RetransmitPacket(QuicPacketSequenceNumber old_sequence_number,
                         QuicPacketSequenceNumber new_sequence_number) {
     QuicSentPacketManagerPeer::MarkForRetransmission(
-        &manager_, old_sequence_number, LOSS_RETRANSMISSION);
+        &manager_, old_sequence_number, TLP_RETRANSMISSION);
     EXPECT_TRUE(manager_.HasPendingRetransmissions());
     QuicSentPacketManager::PendingRetransmission next_retransmission =
         manager_.NextPendingRetransmission();
     EXPECT_EQ(old_sequence_number, next_retransmission.sequence_number);
-    EXPECT_EQ(LOSS_RETRANSMISSION,
+    EXPECT_EQ(TLP_RETRANSMISSION,
               next_retransmission.transmission_type);
     manager_.OnRetransmittedPacket(old_sequence_number, new_sequence_number);
     EXPECT_TRUE(QuicSentPacketManagerPeer::IsRetransmission(
@@ -197,7 +191,6 @@ class QuicSentPacketManagerTest : public ::testing::TestWithParam<bool> {
     SerializedPacket packet(CreateDataPacket(sequence_number));
     packet.retransmittable_frames->AddStreamFrame(
         new QuicStreamFrame(1, false, 0, IOVector()));
-    packet.retransmittable_frames->set_encryption_level(ENCRYPTION_NONE);
     manager_.OnSerializedPacket(packet);
     manager_.OnPacketSent(sequence_number, clock_.ApproximateNow(),
                           packet.packet->length(), NOT_RETRANSMISSION,
@@ -297,7 +290,7 @@ TEST_F(QuicSentPacketManagerTest, RetransmitThenAck) {
 TEST_F(QuicSentPacketManagerTest, RetransmitThenAckBeforeSend) {
   SendDataPacket(1);
   QuicSentPacketManagerPeer::MarkForRetransmission(
-      &manager_, 1, LOSS_RETRANSMISSION);
+      &manager_, 1, TLP_RETRANSMISSION);
   EXPECT_TRUE(manager_.HasPendingRetransmissions());
 
   // Ack 1.
@@ -557,9 +550,9 @@ TEST_F(QuicSentPacketManagerTest, TruncatedAck) {
   manager_.OnIncomingAck(received_info, clock_.Now());
 
   // High water mark will be raised.
-  QuicPacketSequenceNumber unacked[] = { 2, 3, 4 };
+  QuicPacketSequenceNumber unacked[] = { 2, 3, 4, 5 };
   VerifyUnackedPackets(unacked, arraysize(unacked));
-  QuicPacketSequenceNumber retransmittable[] = { 4 };
+  QuicPacketSequenceNumber retransmittable[] = { 5 };
   VerifyRetransmittablePackets(retransmittable, arraysize(retransmittable));
 }
 
@@ -988,25 +981,6 @@ TEST_F(QuicSentPacketManagerTest,
   EXPECT_FALSE(QuicSentPacketManagerPeer::HasPendingPackets(&manager_));
 }
 
-TEST_F(QuicSentPacketManagerTest,
-       CryptoHandshakeRetransmissionThenAbandonAll) {
-  // Send 1 crypto packet.
-  SendCryptoPacket(1);
-  EXPECT_TRUE(QuicSentPacketManagerPeer::HasUnackedCryptoPackets(&manager_));
-
-  // Retransmit the crypto packet as 2.
-  manager_.OnRetransmissionTimeout();
-  RetransmitNextPacket(2);
-
-  // Now discard all unacked unencrypted packets, which occurs when the
-  // connection goes forward secure.
-  manager_.DiscardUnencryptedPackets();
-  VerifyUnackedPackets(NULL, 0);
-  EXPECT_FALSE(manager_.HasPendingRetransmissions());
-  EXPECT_FALSE(QuicSentPacketManagerPeer::HasUnackedCryptoPackets(&manager_));
-  EXPECT_FALSE(QuicSentPacketManagerPeer::HasPendingPackets(&manager_));
-}
-
 TEST_F(QuicSentPacketManagerTest, TailLossProbeTimeoutUnsentDataPacket) {
   QuicSentPacketManagerPeer::SetMaxTailLossProbes(&manager_, 2);
   // Serialize two data packets and send the latter.

net/quic/quic_session.cc

@@ -422,7 +422,7 @@ void QuicSession::OnCryptoHandshakeEvent(CryptoHandshakeEvent event) {
           << "Handshake confirmed without parameter negotiation.";
       // Discard originally encrypted packets, since they can't be decrypted by
       // the peer.
-      connection_->DiscardUnencryptedPackets();
+      connection_->NeuterUnencryptedPackets();
       connection_->SetOverallConnectionTimeout(QuicTime::Delta::Infinite());
       max_open_streams_ = config_.max_streams_per_connection();
       break;

net/quic/quic_unacked_packet_map.cc

@@ -91,7 +91,7 @@ void QuicUnackedPacketMap::ClearPreviousRetransmissions(size_t num_to_clear) {
     }
 
     ++it;
-    NeuterIfPendingOrRemovePacket(sequence_number);
+    RemovePacket(sequence_number);
     --num_to_clear;
   }
 }
@@ -119,7 +119,7 @@ void QuicUnackedPacketMap::NackPacket(QuicPacketSequenceNumber sequence_number,
   it->second.nack_count = max(min_nacks, it->second.nack_count);
 }
 
-void QuicUnackedPacketMap::NeuterIfPendingOrRemovePacket(
+void QuicUnackedPacketMap::RemovePacket(
     QuicPacketSequenceNumber sequence_number) {
   UnackedPacketMap::iterator it = unacked_packets_.find(sequence_number);
   if (it == unacked_packets_.end()) {
@@ -127,28 +127,41 @@ void QuicUnackedPacketMap::NeuterIfPendingOrRemovePacket(
     return;
   }
   TransmissionInfo* transmission_info = &it->second;
-  if (transmission_info->retransmittable_frames != NULL) {
-    if (transmission_info->retransmittable_frames->HasCryptoHandshake()
-            == IS_HANDSHAKE) {
-      --pending_crypto_packet_count_;
+  DCHECK(!transmission_info->pending);
+  MaybeRemoveRetransmittableFrames(transmission_info);
+  transmission_info->all_transmissions->erase(sequence_number);
+  if (transmission_info->all_transmissions->empty()) {
+    delete transmission_info->all_transmissions;
   }
-    delete transmission_info->retransmittable_frames;
-    transmission_info->retransmittable_frames = NULL;
+  unacked_packets_.erase(it);
+}
+
+void QuicUnackedPacketMap::NeuterPacket(
+    QuicPacketSequenceNumber sequence_number) {
+  UnackedPacketMap::iterator it = unacked_packets_.find(sequence_number);
+  if (it == unacked_packets_.end()) {
+    LOG(DFATAL) << "packet is not unacked: " << sequence_number;
+    return;
   }
-  if (transmission_info->pending) {
-    // Neuter it so it can't be retransmitted.
+  TransmissionInfo* transmission_info = &it->second;
+  // TODO(ianswett): Ensure packets are pending before neutering them.
+  MaybeRemoveRetransmittableFrames(transmission_info);
   if (transmission_info->all_transmissions->size() > 1) {
     transmission_info->all_transmissions->erase(sequence_number);
     transmission_info->all_transmissions = new SequenceNumberSet();
     transmission_info->all_transmissions->insert(sequence_number);
   }
-  } else {
-    // Remove it.
-    transmission_info->all_transmissions->erase(sequence_number);
-    if (transmission_info->all_transmissions->empty()) {
-      delete transmission_info->all_transmissions;
+}
+
+void QuicUnackedPacketMap::MaybeRemoveRetransmittableFrames(
+    TransmissionInfo* transmission_info) {
+  if (transmission_info->retransmittable_frames != NULL) {
+    if (transmission_info->retransmittable_frames->HasCryptoHandshake()
+            == IS_HANDSHAKE) {
+      --pending_crypto_packet_count_;
     }
-    unacked_packets_.erase(it);
+    delete transmission_info->retransmittable_frames;
+    transmission_info->retransmittable_frames = NULL;
   }
 }
 
@@ -165,13 +178,6 @@ bool QuicUnackedPacketMap::IsUnacked(
   return ContainsKey(unacked_packets_, sequence_number);
 }
 
-bool QuicUnackedPacketMap::IsPending(
-    QuicPacketSequenceNumber sequence_number) const {
-  const TransmissionInfo* transmission_info =
-      FindOrNull(unacked_packets_, sequence_number);
-  return transmission_info != NULL && transmission_info->pending;
-}
-
 void QuicUnackedPacketMap::SetNotPending(
     QuicPacketSequenceNumber sequence_number) {
   UnackedPacketMap::iterator it = unacked_packets_.find(sequence_number);
@@ -286,15 +292,6 @@ QuicUnackedPacketMap::GetLeastUnackedSentPacket() const {
   return unacked_packets_.begin()->first;
 }
 
-SequenceNumberSet QuicUnackedPacketMap::GetUnackedPackets() const {
-  SequenceNumberSet unacked_packets;
-  for (UnackedPacketMap::const_iterator it = unacked_packets_.begin();
-       it != unacked_packets_.end(); ++it) {
-    unacked_packets.insert(it->first);
-  }
-  return unacked_packets;
-}
-
 void QuicUnackedPacketMap::SetSent(QuicPacketSequenceNumber sequence_number,
                                    QuicTime sent_time,
                                    QuicByteCount bytes_sent,

net/quic/quic_unacked_packet_map.h

@@ -31,9 +31,6 @@ class NET_EXPORT_PRIVATE QuicUnackedPacketMap {
   // Returns true if the packet |sequence_number| is unacked.
   bool IsUnacked(QuicPacketSequenceNumber sequence_number) const;
 
-  // Returns true if the packet |sequence_number| is pending.
-  bool IsPending(QuicPacketSequenceNumber sequence_number) const;
-
   // Sets the nack count to the max of the current nack count and |min_nacks|.
   void NackPacket(QuicPacketSequenceNumber sequence_number,
                   size_t min_nacks);
@@ -72,10 +69,6 @@ class NET_EXPORT_PRIVATE QuicUnackedPacketMap {
   // been acked by the peer.  If there are no unacked packets, returns 0.
   QuicPacketSequenceNumber GetLeastUnackedSentPacket() const;
 
-  // Returns the set of sequence numbers of all unacked packets.
-  // Test only.
-  SequenceNumberSet GetUnackedPackets() const;
-
   // Sets a packet as sent with the sent time |sent_time|.  Marks the packet
   // as pending and tracks the |bytes_sent| if |set_pending| is true.
   // Packets marked as pending are expected to be marked as missing when they
@@ -120,15 +113,21 @@ class NET_EXPORT_PRIVATE QuicUnackedPacketMap {
   // Returns true if there are any pending crypto packets.
   bool HasPendingCryptoPackets() const;
 
-  // Deletes the retransmittable frames associated with the packet and removes
-  // it from unacked packets if it's not pending.
+  // Removes entries from the unacked packet map, and deletes
+  // the retransmittable frames associated with the packet.
   // Does not remove any previous or subsequent transmissions of this packet.
-  void NeuterIfPendingOrRemovePacket(QuicPacketSequenceNumber sequence_number);
+  void RemovePacket(QuicPacketSequenceNumber sequence_number);
+
+  // Neuters the specified packet.  Deletes any retransmittable
+  // frames, and sets all_transmissions to only include itself.
+  void NeuterPacket(QuicPacketSequenceNumber sequence_number);
 
   // Returns true if the packet has been marked as sent by SetSent.
   static bool IsSentAndNotPending(const TransmissionInfo& transmission_info);
 
  private:
+  void MaybeRemoveRetransmittableFrames(TransmissionInfo* transmission_info);
+
   QuicPacketSequenceNumber largest_sent_packet_;
 
   // Newly serialized retransmittable and fec packets are added to this map,

net/quic/test_tools/quic_sent_packet_manager_peer.cc

@@ -112,12 +112,6 @@ size_t QuicSentPacketManagerPeer::GetNumRetransmittablePackets(
   return sent_packet_manager->unacked_packets_.GetNumRetransmittablePackets();
 }
 
-// static
-SequenceNumberSet QuicSentPacketManagerPeer::GetUnackedPackets(
-    const QuicSentPacketManager* sent_packet_manager) {
-  return sent_packet_manager->unacked_packets_.GetUnackedPackets();
-}
-
 // static
 QuicByteCount QuicSentPacketManagerPeer::GetBytesInFlight(
     const QuicSentPacketManager* sent_packet_manager) {

net/quic/test_tools/quic_sent_packet_manager_peer.h

@@ -60,9 +60,6 @@ class QuicSentPacketManagerPeer {
   static size_t GetNumRetransmittablePackets(
       const QuicSentPacketManager* sent_packet_manager);
 
-  static SequenceNumberSet GetUnackedPackets(
-      const QuicSentPacketManager* sent_packet_manager);
-
   static QuicByteCount GetBytesInFlight(
       const QuicSentPacketManager* sent_packet_manager);
 

net/tools/quic/end_to_end_test.cc

@@ -28,6 +28,7 @@
 #include "net/quic/test_tools/quic_test_utils.h"
 #include "net/quic/test_tools/reliable_quic_stream_peer.h"
 #include "net/test/gtest_util.h"
+#include "net/tools/epoll_server/epoll_server.h"
 #include "net/tools/quic/quic_epoll_connection_helper.h"
 #include "net/tools/quic/quic_in_memory_cache.h"
 #include "net/tools/quic/quic_packet_writer_wrapper.h"
@@ -46,6 +47,7 @@
 
 using base::StringPiece;
 using base::WaitableEvent;
+using net::EpollServer;
 using net::test::GenerateBody;
 using net::test::QuicConnectionPeer;
 using net::test::QuicSessionPeer;
@@ -481,7 +483,7 @@ TEST_P(EndToEndTest, DISABLED_LargePostNoPacketLoss) {
 
   client_->client()->WaitForCryptoHandshakeConfirmed();
 
-  // 1 Mb body.
+  // 1 MB body.
   string body;
   GenerateBody(&body, 1024 * 1024);
 
@@ -499,7 +501,7 @@ TEST_P(EndToEndTest, LargePostNoPacketLoss1sRTT) {
 
   client_->client()->WaitForCryptoHandshakeConfirmed();
 
-  // 1 Mb body.
+  // 100 KB body.
   string body;
   GenerateBody(&body, 100 * 1024);
 
@@ -521,7 +523,7 @@ TEST_P(EndToEndTest, LargePostWithPacketLoss) {
   client_->client()->WaitForCryptoHandshakeConfirmed();
   SetPacketLossPercentage(30);
 
-  // 10 Kb body.
+  // 10 KB body.
   string body;
   GenerateBody(&body, 1024 * 10);
 
@@ -530,42 +532,42 @@ TEST_P(EndToEndTest, LargePostWithPacketLoss) {
   request.AddBody(body, true);
 
   EXPECT_EQ(kFooResponseBody, client_->SendCustomSynchronousRequest(request));
+  VerifyCleanConnection(true);
 }
 
-TEST_P(EndToEndTest, LargePostNoPacketLossWithDelayAndReordering) {
+TEST_P(EndToEndTest, LargePostWithPacketLossAndBlockedSocket) {
+  // Connect with lower fake packet loss than we'd like to test.  Until
+  // b/10126687 is fixed, losing handshake packets is pretty brutal.
+  SetPacketLossPercentage(5);
   ASSERT_TRUE(Initialize());
 
+  // Wait for the server SHLO before upping the packet loss.
   client_->client()->WaitForCryptoHandshakeConfirmed();
-  // Both of these must be called when the writer is not actively used.
-  SetPacketSendDelay(QuicTime::Delta::FromMilliseconds(2));
-  SetReorderPercentage(30);
+  SetPacketLossPercentage(10);
+  client_writer_->set_fake_blocked_socket_percentage(10);
 
-  // 1 Mb body.
+  // 10 KB body.
   string body;
-  GenerateBody(&body, 1024 * 1024);
+  GenerateBody(&body, 1024 * 10);
 
   HTTPMessage request(HttpConstants::HTTP_1_1,
                       HttpConstants::POST, "/foo");
   request.AddBody(body, true);
 
   EXPECT_EQ(kFooResponseBody, client_->SendCustomSynchronousRequest(request));
-  VerifyCleanConnection(true);
 }
 
-TEST_P(EndToEndTest, LargePostWithPacketLossAndBlockedSocket) {
-  // Connect with lower fake packet loss than we'd like to test.  Until
-  // b/10126687 is fixed, losing handshake packets is pretty brutal.
-  SetPacketLossPercentage(5);
+TEST_P(EndToEndTest, LargePostNoPacketLossWithDelayAndReordering) {
   ASSERT_TRUE(Initialize());
 
-  // Wait for the server SHLO before upping the packet loss.
   client_->client()->WaitForCryptoHandshakeConfirmed();
-  SetPacketLossPercentage(10);
-  client_writer_->set_fake_blocked_socket_percentage(10);
+  // Both of these must be called when the writer is not actively used.
+  SetPacketSendDelay(QuicTime::Delta::FromMilliseconds(2));
+  SetReorderPercentage(30);
 
-  // 10 Kb body.
+  // 1 MB body.
   string body;
-  GenerateBody(&body, 1024 * 10);
+  GenerateBody(&body, 1024 * 1024);
 
   HTTPMessage request(HttpConstants::HTTP_1_1,
                       HttpConstants::POST, "/foo");
@@ -655,19 +657,17 @@ TEST_P(EndToEndTest, LargePostFEC) {
   VerifyCleanConnection(true);
 }
 
-// TODO(rtenneti): DISABLED_LargePostLargeBuffer seems to be flaky.
-// http://crbug.com/370087.
-TEST_P(EndToEndTest, DISABLED_LargePostLargeBuffer) {
+TEST_P(EndToEndTest, LargePostSmallBandwidthLargeBuffer) {
   ASSERT_TRUE(Initialize());
   SetPacketSendDelay(QuicTime::Delta::FromMicroseconds(1));
-  // 1Mbit per second with a 128k buffer from server to client.  Wireless
+  // 256KB per second with a 256k buffer from server to client.  Wireless
   // clients commonly have larger buffers, but our max CWND is 200.
   server_writer_->set_max_bandwidth_and_buffer_size(
-      QuicBandwidth::FromBytesPerSecond(256 * 1024), 128 * 1024);
+      QuicBandwidth::FromBytesPerSecond(256 * 1024), 256 * 1024);
 
   client_->client()->WaitForCryptoHandshakeConfirmed();
 
-  // 1 Mb body.
+  // 1 MB body.
   string body;
   GenerateBody(&body, 1024 * 1024);
 
@@ -676,6 +676,8 @@ TEST_P(EndToEndTest, DISABLED_LargePostLargeBuffer) {
   request.AddBody(body, true);
 
   EXPECT_EQ(kFooResponseBody, client_->SendCustomSynchronousRequest(request));
+  // This connection will not drop packets, because the buffer size is larger
+  // than the default receive window.
   VerifyCleanConnection(false);
 }
 
@@ -790,7 +792,7 @@ TEST_P(EndToEndTest, DISABLED_LimitCongestionWindowAndRTT) {
   // Now use the negotiated limits with packet loss.
   SetPacketLossPercentage(30);
 
-  // 10 Kb body.
+  // 10 KB body.
   string body;
   GenerateBody(&body, 1024 * 10);
 
@@ -953,7 +955,10 @@ class WrongAddressWriter : public QuicPacketWriterWrapper {
   IPEndPoint self_address_;
 };
 
-TEST_P(EndToEndTest, ConnectionMigration) {
+TEST_P(EndToEndTest, ConnectionMigrationClientIPChanged) {
+  // Tests that the client's IP can not change during an established QUIC
+  // connection. If it changes, the connection is closed by the server as we do
+  // not yet support IP migration.
   ASSERT_TRUE(Initialize());
 
   EXPECT_EQ(kFooResponseBody, client_->SendSynchronousRequest("/foo"));
@@ -971,6 +976,59 @@ TEST_P(EndToEndTest, ConnectionMigration) {
   EXPECT_EQ(QUIC_ERROR_MIGRATING_ADDRESS, client_->connection_error());
 }
 
+TEST_P(EndToEndTest, ConnectionMigrationClientPortChanged) {
+  // Tests that the client's port can change during an established QUIC
+  // connection, and that doing so does not result in the connection being
+  // closed by the server.
+  FLAGS_quic_allow_port_migration = true;
+
+  ASSERT_TRUE(Initialize());
+
+  EXPECT_EQ(kFooResponseBody, client_->SendSynchronousRequest("/foo"));
+  EXPECT_EQ(200u, client_->response_headers()->parsed_response_code());
+
+  // Store the client address which was used to send the first request.
+  IPEndPoint old_address = client_->client()->client_address();
+
+  // Stop listening on the old FD.
+  EpollServer* eps = client_->client()->epoll_server();
+  int old_fd = client_->client()->fd();
+  eps->UnregisterFD(old_fd);
+  close(old_fd);
+
+  // Create a new socket, which will result in a new ephemeral port.
+  QuicClientPeer::CreateUDPSocket(client_->client());
+
+  // The packet writer needs to be updated to use the new FD.
+  client_->client()->CreateQuicPacketWriter();
+
+  // Change the internal state of the client and connection to use the new port,
+  // this is done because in a real NAT rebinding the client wouldn't see any
+  // port change, and so expects no change to incoming port.
+  // This is kind of ugly, but needed as we are simply swapping out the client
+  // FD rather than any more complex NAT rebinding simulation.
+  int new_port = client_->client()->client_address().port();
+  QuicClientPeer::SetClientPort(client_->client(), new_port);
+  QuicConnectionPeer::SetSelfAddress(
+      client_->client()->session()->connection(),
+      IPEndPoint(
+          client_->client()->session()->connection()->self_address().address(),
+          new_port));
+
+  // Register the new FD for epoll events.
+  int new_fd = client_->client()->fd();
+  eps->RegisterFD(new_fd, client_->client(), EPOLLIN | EPOLLOUT | EPOLLET);
+
+  // Send a second request, using the new FD.
+  EXPECT_EQ(kBarResponseBody, client_->SendSynchronousRequest("/bar"));
+  EXPECT_EQ(200u, client_->response_headers()->parsed_response_code());
+
+  // Verify that the client's ephemeral port is different.
+  IPEndPoint new_address = client_->client()->client_address();
+  EXPECT_EQ(old_address.address(), new_address.address());
+  EXPECT_NE(old_address.port(), new_address.port());
+}
+
 TEST_P(EndToEndTest, DifferentFlowControlWindows) {
   // Client and server can set different initial flow control receive windows.
   // These are sent in CHLO/SHLO. Tests that these values are exchanged properly

net/tools/quic/quic_client.cc

@@ -83,6 +83,16 @@ bool QuicClient::Initialize() {
   epoll_server_.set_timeout_in_us(50 * 1000);
   crypto_config_.SetDefaults();
 
+  if (!CreateUDPSocket()) {
+    return false;
+  }
+
+  epoll_server_.RegisterFD(fd_, this, kEpollFlags);
+  initialized_ = true;
+  return true;
+}
+
+bool QuicClient::CreateUDPSocket() {
   int address_family = server_address_.GetSockAddrFamily();
   fd_ = socket(address_family, SOCK_DGRAM | SOCK_NONBLOCK, IPPROTO_UDP);
   if (fd_ < 0) {
@@ -153,8 +163,6 @@ bool QuicClient::Initialize() {
     LOG(ERROR) << "Unable to get self address.  Error: " << strerror(errno);
   }
 
-  epoll_server_.RegisterFD(fd_, this, kEpollFlags);
-  initialized_ = true;
   return true;
 }
 

net/tools/quic/quic_client.h

@@ -181,6 +181,10 @@ class QuicClient : public EpollCallbackInterface,
  private:
   friend class net::tools::test::QuicClientPeer;
 
+  // Used during initialization: creates the UDP socket FD, sets socket options,
+  // and binds the socket to our address.
+  bool CreateUDPSocket();
+
   // Read a UDP packet and hand it to the framer.
   bool ReadAndProcessPacket();
 

net/tools/quic/test_tools/quic_client_peer.cc

@@ -15,6 +15,16 @@ QuicCryptoClientConfig* QuicClientPeer::GetCryptoConfig(QuicClient* client) {
   return &client->crypto_config_;
 }
 
+// static
+bool QuicClientPeer::CreateUDPSocket(QuicClient* client) {
+  return client->CreateUDPSocket();
+}
+
+// static
+void QuicClientPeer::SetClientPort(QuicClient* client, int port) {
+  client->client_address_ = IPEndPoint(client->client_address_.address(), port);
+}
+
 }  // namespace test
 }  // namespace tools
 }  // namespace net

net/tools/quic/test_tools/quic_client_peer.h

@@ -20,6 +20,8 @@ namespace test {
 class QuicClientPeer {
  public:
   static QuicCryptoClientConfig* GetCryptoConfig(QuicClient* client);
+  static bool CreateUDPSocket(QuicClient* client);
+  static void SetClientPort(QuicClient* client, int port);
 
  private:
   DISALLOW_COPY_AND_ASSIGN(QuicClientPeer);